List of Contributors P.D. Adelson, Department of Neurosurgery, University of Pittsburgh, Pittsburgh, PA 900951769, USA F. Andermann, Montreal Neurological Hospital and Institute, McGill University, 3801 University St., Montreal, PQ H3A 2T5, Canada D.L. Arnold, Epilepsy Clinic and Brain Imaging Center, Montreal Neurological Institute and Hospital, 3801 University Street, Montreal, PQ H3A 2B4, Canada B.K. August, Department of Neurology, University of Wisconsin, 1300 University Avenue, Madison, WI 53706, USA J.K. Austin, Indiana University School of Nursing, 1111 Middle Drive, NU 492, Indianapolis, IN 46202-5107, USA R. Baldwin, Epilepsy Research Laboratory, VA Greater Los Angeles Healthcare System, Department of Neurology and Brain Research Institute, UCLA School of Medicine, Los Angeles, CA 90095, USA T.Z. Baram, Departments of Pediatrics and Anatomy/Neurobiology and Neurology, University of California at Irvine, Irvine, CA 92697-4475, USA D.E Barboriak, Department of Radiology (Neuroradiology), Duke University Medical Center, Durham, NC 27710, USA C. Barlow, The Salk Institute for Biological Studies, The Laboratory of Genetics, 10010 North Torrey Pines Road, La Jolla, CA 92037, USA N.G. Bazan, Neuroscience Center of Excellence and Department of Ophthalmology, LSU State University Health Sciences Center, 2020 Gravier Street, New Orleans, LA 70112, USA A.J. Becker, Department of Neuropathology, University of Bonn Medical Center, SigmundFreud-Str. 25, 53105 Bonn, Germany B. Bell, Department of Neurology, University of Wisconsin, 600 North Highland Avenue, Madison, WI 53792, USA Y. Ben-Ari, Institut National de la Sant6 de la Recherche Medicale, Unit 29, Institut de Neurobiologie de la M6diterran6e, Marseille, France R.A. Bender, Departments of Pediatrics and Anatomy/Neurobiology and Neurology, University of California at Irvine, CA 92697-4475, USA J. Bengzon, Department of Neurosurgery, University Hospital, S-221 85 Lund, Sweden A. Bernasconi, Epilepsy Clinic and Brain Imaging Center, Montreal Neurological Institute and Hospital, 3801 University Street, Montreal, PQ H3A 2B4, Canada I. Bltimcke, Department of Neuropathology, University of Bonn Medical Center, SigmundFreud-Str. 25, 53105 Bonn, Germany K.J. Buchheim, Johannes Mtiller Institut ffir Physiologie, Universit~itsklinikum Charit6, Humboldt Universit~it Berlin, Tucholskystrasse 2, D 10117 Berlin, Germany L.D. Cahan, Neurosurgery, Southern California Permanente Medical Group, Los Angeles, CA 15261, USA F. Cendes, Epilepsy Clinic and Brain Imaging Center, Montreal Neurological Institute and Hospital, 3801 University Street, Montreal, PQ H3A 2B4, Canada
vi H.R. Cock, Department of Clinical and Experimental Epilepsy, Institute of Neurology, Queen Square, London WC1N 3BG, UK A.J. Cole, MGH Epilepsy Service, VBK 830, Massachusetts General Hospital, 55 Fruit Street, Boston, MA 02114, USA J.A. Del Rio, The Salk Institute for Biological Studies, The Laboratory of Genetics, 10010 North Torrey Pines Road, La Jolla, CA 92037, USA C.B. Dodrill, Harborview Medical Center, Epilepsy Center (Box 359745), 325 9th Avenue, Seattle, WA 98104-2499, USA RE. Dudek, Department of Anatomy and Neurobiology, Colorado State University, Fort Collins, CO 80523, USA J.S. Duncan, National Society for Epilepsy, Chalfont St. Peter, Buckinghamshire SL9 0LR, UK D.W. Dunn, Indiana University School of Nursing, 1111 Middle Drive, NU 492, Indianapolis, IN 46202-5107, USA A. Ebner, Epilepsy Center Bethel, Epilepsy Surgery Program, Maraweg 21, D-33617 Bielefeld, Germany M. Eghbal-Ahmadi, Departments of Pediatrics and Anatomy/Neurobiology and Neurology, University of California at Irvine, CA 92697-4475, USA C.T. Ekdahl, Section of Restorative Neurology, Wallenberg Neuroscience Center, BMC A-11, S-221 84 Lund, Sweden J. Engel, Jr., Departments of Neurology and Neurobiology, and the Brain Research Institute, UCLA School of Medicine, 710 Westwood Plaza, Los Angeles, CA 90095-1769, USA D.J. Ferraro, Department of Anatomy and Neurobiology, Colorado State University, Fort Collins, CO 80523, USA S. Gabriel, Johannes MUller Institut ftir Physiologie, Universit~itsklinikum Charitr, Humboldt Universitfit Berlin, Tucholskystrasse 2, D 10117 Berlin, Germany W.D. Galliard, Clinical Epilepsy Section, National Institutes of Health, Building 10, Room 5N-250, Bethesda, MD 20892, USA O.H.J. Gr6hn, NMR Research Group, A.I. Virtanen Institute for Molecular Sciences, University of Kuopio, P.O. Box 1627, FIN-70 211 Kuopio, Finland R.W. Guillery, Department of Anatomy, University of Wisconsin, 1300 University Avenue, Madison, WI 53706, USA W.A. Hauser, Sergievsky Center, Columbia University, College of Physicians and Surgeons, 630 W 168th St, New York, NY 10032, USA U. Heinemann, Johannes Mtiller Institut ftir Physiologie, Universit~tsklinikum Charitr, Humboldt Universit~it Berlin, Tucholskystrasse 2, D 10117 Berlin, Germany J.L. Hellier, Department of Neurology, University of Colorado Health Science Center, Denver, CO 80262, USA C. Helmstaedter, University Clinic of Epileptology, University of Bonn, Sigmund Freud Strasse 25, D-53105 Bonn, Germany B.P. Hermann, Department of Neurology, University of Wisconsin, 600 North Highland Avenue, Madison, WI 53792, USA G.L. Holmes, Clinical Neurophysiology Laboratory, Hunnewell 2, Children's Hospital, 300 Longwood Avenue, Boston, MA 02115, USA H. Jokeit, Swiss Epilepsy Center, Bleulerstrasse 60, CH-8008 Zurich, Switzerland R. K~ilvi~iinen, Department of Neurology, Kuopio University Hospital, P.O. Box 1777, FIN-70 211 Kuopio, Finland
vii O. Kann, Johannes Mtiller Institut fur Physiologie, Universit/itsklinikum Charit6, Humboldt Universit~it Berlin, Tucholskystrasse 2, D 10117 Berlin, Germany H. Katsumori, Epilepsy Research Laboratory, VA Greater Los Angeles Healthcare System, Department of Neurology and Brain Research Institute, UCLA School of Medicine, Los Angeles, CA 90095, USA R. Kauppinen, NMR Research Group, A.I. Virtanen Institute for Molecular Sciences, University of Kuopio, P.O. Box 1627, FIN-70 211 Kuopio, Finland R. Khazipov, Institut National de la Sant6 de la Recherche Medicale, Unit 29, Institut de Neurobiologie de la M6diterran6e, Marseille, France S. Koh, Epilepsy Research Laboratory, Massachusetts General Hospital and Department of Neurology, Harvard Medical School, Boston, MA 02114, USA R. Kotloski, Department of Neurology, H6/570, 600 Highland Avenue, University of Wisconsin, Madison, WI 53792, USA R. Kovacs, Johannes Mtiller Institut ftir Physiologie, Universitatsklinikum Charit6, Humboldt Universit~it Berlin, Tucholskystrasse 2, D 10117 Berlin, Germany S. Lauersdorf, Department of Neurology, H6/570, 600 Highland Avenue, University of Wisconsin, Madison, WI 53792, USA J.R. Lee, California Medical Review Inc., San Francisco, CA 94104, USA J.P. Leite, Department of Neurology, Ribeirao Preto School of Medicine, University of S~o Paulo, Ribeirao Preto, Brazil SA D.V. Lewis, Department of Pediatrics (Neurology), Duke University Medical Center, Durham, NC 27710, USA L.M. Li, Epilepsy Clinic and Brain Imaging Center, Montreal Neurological Institute and Hospital, 3801 University Street, Montreal, PQ H3A 2B4, Canada O. Lindvall, Section of Restorative Neurology, Wallenberg Neuroscience Center, BMC A-11, S-221 84 Lund, Sweden H. Liu, Epilepsy Research Laboratory, VA Greater Los Angeles Healthcare System, Department of Neurology and Brain Research Institute, UCLA School of Medicine, Los Angeles, CA 90095, USA D.H. Lowenstein, Harvard Medical School and Department of Neurology, Beth IsraelDeaconess Medical Center, 333 Brookline Avenue, Boston, MA 02215, USA K. Lukasiuk, Epilepsy Research Laboratory, A.I. Virtanen Institute, University of Kuopio, Neulaniementie 2, P.O. Box 1627, FIN-70 211 Kuopio, Finland M. Lynch, Department of Neurology, H6/570, 600 Highland Avenue, University of Wisconsin, Madison, WI 53792, USA J.R. MacFall, Department of Radiology (Neuroradiology), Duke University Medical Center, Durham, NC 27710, USA G.W. Mathern, Division of Neurosurgery, Reed Neurological Research Center, 710 Westwood, Plaza, Room 2123, Los Angeles, CA 90095-1769, USA A.M. Mazarati, Epilepsy Research Laboratory, VA Greater Los Angeles Healthcare System, Department of Neurology and Brain Research Institute, UCLA School of Medicine, Los Angeles, CA 90095, USA B.S. Meldrum, GKT School of Biomedical Sciences, Henriette Raphael House, Guy's Campus, London SE1 1UL, UK R. Miettinen, Department of Neurology and Neurosciences, University of Kuopio, P.O. Box 1777, FIN-70 211 Kuopio, Finland T.V. Mitchell, Department of Radiology (Neurology), Duke University Medical Center, Durham, NC 27710, USA
viii E Mohapel, Section of Restorative Neurology, Wallenberg Neuroscience Center, BMC A-11, S-221 84 Lurid, Sweden S.L. Mosh6, Departments of Neurology and Neuroscience, Albert Einstein College of Medicine, Einstein/Montefiore Epilepsy Management Center, 1410 Pelham Parkway South, Bronx, NY 10461, USA J. Nairism~igi, NMR Research Group, A.I. Virtanen Institute for Molecular Sciences, University of Kuopio, EO. Box 1627, FIN-70 211 Kuopio, Finland D. Naylor, Epilepsy Research Laboratory, VA Greater Los Angeles Healthcare System, Department of Neurology and Brain Research Institute, UCLA School of Medicine, Los Angeles, CA 90095, USA J. Niquet, Epilepsy Research Laboratory, VA Greater Los Angeles Healthcare System, Department of Neurology and Brain Research Institute, UCLA School of Medicine, Los Angeles, CA 90095, USA J. Nissinen, Epilepsy Research Laboratory, A.I. Virtanen Institute, University of Kuopio, Neulaniementie 2, EO. Box 1627, FIN-70 211 Kuopio, Finland J.M. Parent, University of Michigan Medical Center, Neuroscience Laboratory Building, 1103 East Huron Street, Ann Arbor, MI 48104-1687, USA E. Pauli, Department of Neurology, Epilepsy-Center University, Erlangen-Niimberg, Schwabachanlage 6, 91054 Erlangen, Germany A. Pitkanen, Epilepsy Research Laboratory, A.I. Virtanen Institute for Molecular Sciences, University of Kuopio, Neulaniementie 2, P.O. Box 1627, FIN-70 211 Kuopio, Finland J.M. Provenzale, Department of Radiology (Neuroradiology), Duke University Medical Center, Durham, NC 27710, USA E.B. Rodriguez de Turco, Neuroscience Center of Excellence and Department of Ophthalmology, LSU State University Health Sciences Center, New Orleans, LA 70112, USA T. Salmenper~i, Department of Neurology, Kuopio University Hospital, P.O. Box 1777, FIN-70 211 Kuopio, Finland R. Sankar, Epilepsy Research Laboratory, VA Greater Los Angeles Healthcare System, Department of Neurology and Brain Research Institute, UCLA School of Medicine, Los Angeles, CA 90095, USA N. Santilli, Elan Pharmaceuticals, 800 Gateway Blvd, South San Fransisco, CA 94080, USA P. Schauwecker, Department of Cell and Neurobiology, University of Southern California, Keck School of Medicine, BMT 401, 1333 San Pablo Street, Los Angeles, CA 900899112, USA S. Schuchmann, Johannes MUller Institut fiir Physiologie, Universit~tsklinikum Charit6, Humboldt Universit~it Berlin, Tucholskystrasse 2, D 10117 Berlin, Germany M. Seidenberg, Department of Psychology, Chicago Medical School, North Chicago, IL 60064-3095, USA S. Shinnar, Comprehensive Epilepsy Management Center, Montefiore Medical Center, Albert Einstein College of Medicine, 111 E 210th Street, Bronx, NY 10467, USA Y. Shirasaka, Epilepsy Research Laboratory, VA Greater Los Angeles Healthcare System, Department of Neurology and Brain Research Institute, UCLA School of Medicine, Los Angeles, CA 90095, USA S. Shorvon, Institute of Neurology, Queen Square, London WC1N 3BG, UK C.E. Stafstrom, Department of Neurology, H6-528, University of Wisconsin, 600 Highland Avenue, Madison, WI 53792, USA
ix K.J. Staley, Department of Neurology, University of Colorado Health Science Center, Denver, CO 80262, USA H. Stefan, Department of Neurology, Epilepsy-Center University, Erlangen-Ntirnberg, Schwabachanlage 6, 91054 Erlangen, Germany L. Suchomelova, Epilepsy Research Laboratory, VA Greater Los Angeles Healthcare System, Department of Neurology and Brain Research Institute, UCLA School of Medicine, Los Angeles, CA 90095, USA T.E Sutula, Departments of Neurology and Anatomy, H6/570, University of Wisconsin, Madison, WI 53792, USA J.W. Swann, The Cain Foundation Laboratories, Department of Pediatrics, Baylor College of Medicine, 6621 Fannin St., MC3-6365, Houston, TX 77030, USA E. Tasch, Epilepsy Clinic and Brain Imaging Center, Montreal Neurological Institute and Hospital, 3801 University Street, Montreal, PQ H3A 2B4, Canada W.H. Theodore, Clinical Epilepsy Section, National Institutes of Health, Building 10, Room 5N-250, Bethesda, MD 20892, USA K.W. Thompson, Epilepsy Research Laboratory, VA Greater Los Angeles Healthcare System, Department of Neurology and Brain Research Institute, UCLA School of Medicine, Los Angeles, CA 90095, USA D.M. Treiman, Barrow Neurological Institute, 350 West Thomas Road, Phoenix, AZ 85013, USA B. Tu, Neuroscience Center of Excellence and Department of Ophthalmology, LSU State University Health Sciences Center, New Orleans, LA 70112, USA K.E. VanLandingham, Department of Medicine (Neurology), Duke University Medical Center, Durham, NC 27710, USA L. Veli~ek, Departments of Pediatrics, Neurology and Neuroscience, AECOM, K314, 1410 Pelham Parkway South, Bronx, NY 10461, USA C.G. Wasterlain, Department of Neurology, VA Medical Center (127), 11301 Wilshire Boulevard, West Los Angeles, CA 90073, USA M.J. West, Department of Neurobiology/Anatomy, University of Arhus, 8000 Arhus C, Denmark O.D. Wiestler, Department of Neuropathology, University of Bonn Medical Center, Sigmund-Freud-Str. 25, 53105 Bonn, Germany EA. Williams, Department of Anatomy and Neurobiology, Colorado State University, Fort Collins, CO 80523, USA Y. Zheng, Epilepsy Research Laboratory, Massachusetts General Hospital and Department of Neurology, Harvard Medical School, Boston, MA 02114, USA
xi
Preface People with severe epilepsy may experience multiple seizures each day and hundreds to thousands of seizures during their lifetimes. Even for people with severe epilepsy, the time represented by the seizures, which typically last only seconds to minutes, is only a very small fraction of their total existence. The cumulative adverse impact of the brief seizures, however, is usually substantial and, in some cases, may be quite debilitating. How do brief episodes of abnormal brain electrical activity and the accompanying involuntary behaviors produce prolonged disruption of normal functions that persists beyond the seizures, and why in some cases does this disruption appear to be progressive and cumulative? Seizures can be regarded as symptoms of an underlying brain disorder, and progression of neurological disability may be caused by progression of the primary disorder. In some cases, the causative lesion may be static, but progression still occurs, which raises questions about the potentially damaging effects of the repeated seizures. For many patients, recurring seizures become increasingly frequent and are associated with progressive disability that may include memory disturbances, cognitive impairments, and diminished quality of life. These observations and concerns, which are unfortunately familiar for too many patients, families, and physicians, raise the question "Do seizures damage the brain?" The association of epilepsy with brain damage is well established, and includes the observation that as many as 70% of cases of drug refractory epilepsy are accompanied by obvious hippocampal sclerosis, a lesion characterized by neuronal loss and gliosis involving the CA3, CA1 subregions of the hippocampus, and the hilus of the dentate gyrus. The association of epilepsy with hippocampal sclerosis is of interest because the hippocampus has been implicated in memory formation, and many patients with chronic epilepsy have significant memory dysfunction. In addition to memory disturbances, the cognitive dysfunction experienced by patients with long-standing epilepsy may involve other domains, and may be observed as a cumulative consequence of poorly controlled seizures even when seizure control is eventually achieved. What are the cumulative effects of recurring brief seizures? A firm answer to this question has been surprisingly elusive for a variety of reasons. Clearly there is a subset of patients who appear to tolerate seizures with relatively limited long-term consequences, and not all patients are destined to progress to intractability with frequent seizures and disability. This variability and individual susceptibility has made it difficult to make statements that fairly apply to the full range of people with epileptic disorders, whose disorders span a broad spectrum from mild with excellent control and few limitations, to severe with multiple daily seizures and pronounced disability that affects employment, educational performance, and personal life. Another factor that has made it difficult to assess the effects of recurring brief seizures is the broad range of underlying pathologies, which also vary in respect to severity. As noted, progression of a primary condition may obscure the potentially progressive adverse effects of the superimposed symptomatic seizures. When the primary condition is static, but has produced severe cognitive disability, such as cerebral palsy or syndromes of mental
xii retardation, the incremental adverse effects of seizures may be difficult to detect on the background of significant primary impairments. Finally, in an effort to make the best of unfavorable circumstances, many people afflicted by epilepsy have been understandably reluctant to draw attention to adverse consequences, and to add to the burden of stigma and exclusion from many aspects of normal life. So what are the consequences, if any, of the repeated brief seizures that are the defining feature of epilepsy? Is there a spectrum of severity of seizure-induced damage? Should damage be regarded only as loss of neurons or atrophy of hippocampus and other brain structures? Is damage any change in cells (or properties of cells) that is sufficient to permanently alter functional properties of neural circuits, as detected by physiological assessment? In the developing nervous system, is damage any change in cells, properties of cells, or neural circuits that cumulatively modifies or alters developmental outcome? Is damage any change in cells (or properties of cells) that cumulatively results in behaviorally significant alterations? Few would dispute that inadequately treated status epilepticus produces brain damage detectable by imaging evidence of reduced brain volume and permanent cognitive dysfunction. The potential subtlety of incremental effects at the cellular level may make detection of changes induced by brief seizures quite difficult, but these seizures may still lead to cumulative structural and functional defects. This volume seeks to explore the spectrum of severe to more subtle damage that may be a consequence of seizures. The contributing authors have addressed these questions and related issues using a variety of methods in experimental models and in patients with epilepsy. The picture that has emerged in response to the question "Do seizures damage the brain?" is that cumulative damage from repeated seizures should not be regarded only in terms of loss of neurons or brain atrophy, but should be broadened to include irreversible or permanent dysfunction that is a consequence of recurring seizures. The dysfunction may or may not be associated with clear structural alterations, and may be caused by complex molecular, cellular, synaptic, and systems level dysfunction that results in permanent cognitive and behavioral impairments. T. Sutula A. Pitk~inen (Editors)
xiii
Acknowledgements The contributions to this volume were initially presented and discussed at a workshop held in Rovaniemi, Finland from June 27-July 1, 2001, which was supported by grants from Elan Pharmaceuticals and the American Epilepsy Society. The editors gratefully acknowledge not only the support, but also the encouragement from Elan and the American Epilepsy Society to pursue the difficult and sometimes unsettling questions that are the subject of this volume. We especially thank the participating authors and their 'behind the scenes' supporters, whose sacrifice of time and dedication have significantly contributed to this effort to address important questions for people with epilepsy. We also thank the many members of the Department of Neurology at the University of Wisconsin and the A.I. Virtanen Institute for Molecular Sciences at the University of Kuopio who organized the workshop, including Daryn Belden, Carol Chijimatsu, and Greg Zalesak (Madison) and Samuli Kemppainen, Sanna Viitanen (Kuopio).
T. Sutula and A. Pitkanen (Eds.) Progress in Brain Research, Vol. 135 © 2002 Published by Elsevier Science B.V.
CHAPTER 1
Concept of activity-induced cell death in epilepsy: historical and contemporary perspectives Brian S. M e l d r u m * GKT Department of Biomedical Sciences, Kings College, London, UK
Abstract: Selective neuronal loss following status epilepticus was first described just under 100 years ago. The acute pathology following status epilepticus was shown to be 'ischemic cell change' and was assumed to arise through hypoxia/ischemia. Less than 30 years ago it was proposed, from experiments in primates, that the selective neuronal loss in hippocampus and cortex resulted from the abnormal electrical discharges. Selectively vulnerable neurons show swollen, calcium-loaded mitochondria in the soma and focally in dendrites. Burst firing with a massive Ca2+ entry needs to be sustained for 30-120 min to produce necrotic cell death. Lesser stress may produce apoptosis or immediate early gene expression with enhanced expression of many enzymes and receptor subunits. Changes in enzyme, transporter, ion-channel or receptor function or in network properties may lead to altered vulnerability to the effects of seizures. This type of modification and the cumulative effect of oxidative damage to proteins and lipids may explain the long-term consequences of repetitive brief seizures.
Introduction
The concept that it is the abnormal discharges associated with epileptic activity that cause selectively vulnerable neurons to die is so widely accepted today that it appears to be virtually self-evident. I suspect that most young researchers in epilepsy would be amazed to learn that the concept did not exist 30 years ago. Thus I am grateful for this opportunity to give a simple historical account of the development of the concept. A fuller account is provided by Shorvon (1994). The first account of a selective pattern of brain damage (unilateral hippocampal sclerosis, diffuse cortical atrophy, lobular cerebellar atrophy) found in institutionalized patients with intractable seizures
*Correspondence to: B.S. Meldrum, GKT School of Biomedical Sciences, Henriette Raphael House, Guy's Campus, London SE1 1UL, UK. Tel.: +44-207-848-6420; E-mail: brian.meldrum @kcl.ac.uk
was given by Bouchet and Cazauvieilh in 1825. In 1880, Sommer described the cellular pathology in the hippocampus very precisely and concluded that this focal pathology was the cause ('aetiologisches Moment') of generalized seizures (Sommer, 1880). Pfleger also in 1880 reviewed a large series of post mortem studies and remarked that some patients dying during or a few days after status epilepticus had what appeared (on gross inspection) to be acute vascular lesions in the amygdala or hippocampus. He suggested that disturbances in local cerebral circulation might be associated with prolonged seizures (Pfleger, 1880). Selective neuronal loss
The first clear account of selective neuronal loss following status epilepticus appeared in the American Journal of Insanity in three articles by Clark and Prout (1903-1904). They monitored blood pressure, respiration and temperature throughout status epilepticus, emphasizing the sinister significance of
hyperthermia. They described a sequence of cellular changes observed with Nissl staining in the brains of seven patients dying during the course of status epilepticus involving pyramidal neurons in lamina II and III in the cortex. There was initially chromatolysis (loss of Nissl staining) sometimes associated with protoplasmic vacuolation or shrinkage of the cytoplasm. Later there were glial cell changes that could be evidence of neuronophagia. Classical neuropathology developed in Germany in the early 1900s. Alzheimer (1907) was the first of about a dozen noted pathologists to write in German on the neuropathology of epilepsy and status epilepticus (quoted in Scholz, 1951, 1959; Peiffer, 1963). By far the most influential paper concerning status epilepticus was that of Spielmeyer (1927). He described the predominant early finding in selectively vulnerable neurons as 'ischemic cell change' ('Ischaemische Zellerkrankung'), a type of necrotic cell death observed in Nissl stained preparations that was characteristic of anoxic/ischemic damage. The selective pattern of damage in the hippocampus, neocortex and cerebellum was also similar to that seen following cerebral ischemia, as after cardiac arrest, severe arterial hypotension or strangulation. It was then thought that vascular spasm ('angiospasmen') played an important part in the initiation of seizures, thus it was concluded that selective patterns of brain damage observed after status epilepticus were the consequence of cerebral hypoxia/ischemia occurring as a result of vasospasm, or possibly hypoxia and cerebrovascular problems during or directly after seizures or associated with later cerebral edema (Spielmeyer, 1930). The concept of selective vulnerability in neurodegenerative disorders was explored by Vogt and Vogt (1922) who spoke of 'pathoclisis' evoking a mechanism whereby differences in fine structure or biochemistry of particular neurons could account for different patterns of selective vulnerability encountered in different degenerative disorders.
man, 1985). The first such study was from Gibbs et al. (1934) who used thermocouples in the internal jugular vein. Similarly Penfield and colleagues (Penfield et al., 1939; Penfield and Jasper, 1954) recorded focal seizure discharges with corticography in the open skull and using thermocouples and direct observation noted that there was vasodilation in association with seizure onset. Indeed, the hyperemia was such that the venous blood became red, suggesting that blood flow increased relatively more than oxygen consumption.
Incisural sclerosis Earle et al. (1953) used suction to remove mesial temporal structures as a treatment for drug-resistant complex partial seizures and described the frequent occurrence (100 of 157 cases) of sclerotic lesions in the medial tip of the temporal lobe (uncus). They called this pathology 'incisural sclerosis' to indicate that it arose from herniation of the medial border of the anterior temporal lobe over the tentorium causing ischemic lesions (through compression of branches of the anterior choroidal and posterior cerebral arteries). The postulation that this event occurred during moulding of the head in the normal birth process was supported by experimental studies on fetal brains. The ischemic lesion was seen as the focal pathology responsible for the epileptic focus. This view was very influential for 20-30 years. It was rejected by Falconer (1968) on the grounds that: (1) postmortem studies in neonates did not reveal this pattern of pathology (Veith, 1960); (2) the lack of correlation of the pathology with a history of difficult birth (as acknowledged by Penfield) is problematic; and (3) the extent of the pathology in the anterior temporal lobe and elsewhere in the brain was not compatible with compression of the anterior choroidal and posterior cerebral arteries. Nevertheless the clinical concept remained current in the 1980s (Turner and Wyler, 1981).
Changes in cerebral blood flow and metabolism in seizures
The Meyer hypothesis
This concept of the ischemic nature of the damage was widely accepted, but came into conflict with a progressive body of evidence that cerebral blood flow increased during seizures (see review by Chap-
The introduction of the en bloc resection of the anterior temporal lobe by Murray Falconer provided material suitable for detailed neuropathological study. The initial analysis of this material by Cavanagh and
Meyer (1956) is a landmark. It showed that laminar necrosis was widespread in the temporal cortex. It also noted the striking clinical correlation between mesial temporal sclerosis and a history of an early episode of status epilepticus or prolonged febrile convulsions (11 out of 17 patients with Ammon's horn sclerosis, compared with 0 out of 9 without Ammon's horn sclerosis) that preceded the clinical onset of complex partial seizures by 1 or more years. This clinical correlation whereby approximately two-thirds of drug-refractory complex partial seizure patients with mesial temporal sclerosis give a history of an early (6 months to 5 years) episode of status epilepticus or complicated febrile convulsions (more than 1 in 24 h, or lasting more than 30 min, or with focal features) has been confirmed in a large number of neurosurgical reports, e.g. Falconer, 1974; Sagar and Oxbury, 1987; Kim et al., 1990; Cendes et al., 1993; Mathem et al., 1995; Mathem et al., 2002, this volume). At the 1954 Marseille Colloquium Gastaut (1956) described Meyer as having proposed that the initial event causes the mesial temporal sclerosis and this lesion, after an appropriate maturational period, is the focal site of onset of the complex partial seizures. This causal sequence has been described as the Meyer hypothesis (Meldrum, 1997).
Primate experimental studies and activity-dependent cell death Although several authors had induced generalized or focal seizures in animals chemically or electrically and looked for pathological changes, none had successfully addressed the issue of relating the pathology to local and generalized physiological changes. The first studies permitting conclusions regarding these issues were those of Brierley and Meldrum in the early 1970s. We monitored a variety of physiological parameters in adolescent baboons during prolonged seizures induced by intravenous bicuculline and studied ischemic cell change in selectively vulnerable neurons in the acutely perfused brain (Meldrum and Brierley, 1973; Meldrum and Horton, 1973). We studied the relationship between physiological changes and acute pathology both in unmodified seizures and in paralyzed, ventilated animals in which cerebral hypoxia was minimized
(Meldrum et al., 1973). We concluded that, whereas arterial hypotension and hyperthermia could contribute to cerebellar damage, hippocampal and cortical damage was largely the result of the local seizure activity itself, with the critical duration of seizure activity for inducing ischemic cell change lying between 82 and 120 min. We also used allylglycine to induce multiple brief seizures (6-63 seizures in 2-11 h) and saw similar selective patterns of ischemic cell change in the hippocampus, but also saw, with 7-21 day survival, classical appearances of selective neuronal loss, phagocytosis and gliosis in CA1 and CA3 subzones (Meldrum and Brierley, 1972; Meldmm et al., 1974). The concept that the burst discharges were the primary (and sufficient) cause of the selective neuronal damage remained controversial, but was confirmed a decade later when Sloviter (1983) showed that perforant path stimulation could lead to damage in the hilus and CA3 (see also Olney et al., 1983; Sloviter et al., 1996).
Subsequent rodent experiments with generalized seizures With the collaboration of Astrid Chapman in Bo Siesj6's laboratory I subsequently established a similar model of status epilepticus induced by bicuculline in paralysed ventilated rats, that allowed confirmation of the selective pathology occurring in the hippocampus but also permitted detailed study of cerebral blood flow and metabolism during seizures (Meldrum and Nilsson, 1976; Borgstrrm et al., 1976; Chapman et al., 1977). These studies showed massive increases in oxygen and glucose consumption that were, however, fully compensated by the increases in blood flow. It was possible to manipulate the critical physiological parameters, confirming that hyperthermia could exacerbate cerebellar damage but perhaps surprisingly showing that arterial hypotension and mild hypoxia might be protective, possibly because they reduced the intensity of the seizure discharge (Blennow et al., 1978; Nevander et al., 1985).
Experiments with limbic seizures We and others showed that limbic seizures induced by the focal injection of kainate or NMDA (N-
methyl-D-aspartate) into the amygdala or hippocampus in rodents and primates could produce various patterns of hippocampal and amygdala damage but with a strong effect on CA3 pyramidal neurons (BenAri et al., 1980; Menini et al., 1980). NMDA receptor antagonists protect against the remote damage seen after focal injections of kainate into amygdala or hippocampus, and most of the damage seen after seizures induced by systemic kainate (Fariello et al., 1989; Clifford et al., 1990; Lerner-Natoli et al., 1991; Jarrard and Meldrum, 1993). Only a limited amount of the CA3 damage seems to be a direct consequence of kainate receptor activation; most of the damage is due to burst activity and NMDA receptor activation related to seizure spread.
Calcium and mitochondrial swelling In a series of talks in 1980, I outlined for the first time the hypothesis that mitochondrial poisoning by calcium overload was the critical link between the burst discharges and ischemic cell change (see Meldrum, 1981, 1983). This hypothesis derived from the observations of Schanne et al. (1979) in liver and Wrogemann and Pena (1976) in muscle. In collaboration with Griffiths and Evans we used the oxalate pyroantimonate procedure to visualize free calcium in EM images, during the course of status epilepticus induced in rats by bicuculline, allylglycine or kainic acid (Evans et al., 1983, 1984; Griffiths et al., 1983). In the hippocampus, we observed massive calcium loading of mitochondria focally in dendritic fields of CA1 and CA3 after 60-120 min of seizure activity, irrespective of the seizure-inducing agent. The focal dendritic swellings appeared to relate to excitatory inputs. We subsequently showed that the grossly swollen, calcium-loaded mitochondria were visible after 30 min of seizure-like discharge, but that these changes were reversible over a 30-60-min period if the seizure was terminated by diazepam (Evans et al., 1984; Griffiths et al., 1984). In the hippocampal dendritic fields, the mitochondrial changes are highly focal with a large number of mitochondria remaining morphologically normal. This finding does not appear to be sufficiently considered in ex vivo studies of mitochondrial function following seizure activity (see Cock, 2002, this volume).
Necrotic cell death and apoptotic cell death Ischemic cell change with its initial stage of mitochondrial overload with calcium is the classic form of necrotic cell death. Apoptotic cell death is the preferred form of neuronal suicide, but requires energy in the form of ATP and protein synthesis. It can be triggered either via cell surface death receptors or via the mitochondrial release of either cytochrome c (activating caspase 9, and subsequently the executioner caspases) or the apoptosis-inducing factor (Joza et al., 2001). It is clear that severe and prolonged seizure discharges trigger both forms of cell death, sometimes in the same cell population but with different time courses or in different cell populations. This is seen morphologically in dentate granule cells with pilocarpine seizures (Covolan et al., 2000) or repetitive perforant path stimulation (Sloviter et al., 1996), and with immunocytochemistry in CA3 neurons after intra-amygdaloid kainate (Henshall et al., 1999, 2000). It is clear that dentate granule cells more readily show apoptosis, and in the limbic seizure model with intra-amygdaloid kainate they may show a minimal level of apoptosis when CA3 neurons are showing severe necrosis. They also may show increased apoptosis after a single kindling seizure (Bengzon et al., 1997).
Cumulative effects and'the two hit concept The conclusion from a large series of experiments relating to activity-dependent cell death is that the threshold duration of sustained seizure activity to produce acute necrotic cell death with an appropriate selective pattern is around or above 30 min. Single brief seizures do not cause necrotic cell death. As shown by Bengzon et al. (2002, this volume) they may double the natural rate of apoptosis in certain cell types, such as dentate granule cells, but they are not documented as inducing selective patterns of cell loss. Nevertheless, sequences of brief seizures in various circumstances do appear to be capable of inducing cell loss. There is a wide range of possible mechanisms for such an effect. A single brief seizure is associated with intracellular changes in ionic concentration well outside the physiological ranges that have multiple consequences with complex time courses (see Holmes
et al., 2002, this volume). Immediate early gene expression is seen 0.5-4 h following the seizure and mRNA and protein for a variety of enzymes, receptors and ion channel subunits can be shown to be altered subsequently, including for example COX-2 (see Bazan et al., 2002, this volume). These changes will alter the vulnerability of an individual neuron to subsequent 'excitotoxic' stresses and will modify network responses to future abnormal inputs. There is of course a wide variety of experimental evidence that the consequences of brief episodes of cerebral ischemia or of epileptic activity or both combined are different if they occur with a brief separation (in the range 0.5-24 h) compared with longer intervals. Of course, some functional consequences of a single seizure may be long-term and therefore cumulative. These include long-term changes in gene expression, oxidative damage to membrane lipids and proteins or damage to nuclear or mitochondrial DNA. The changes in gene expression may result in altered subunit composition for excitatory or inhibitory receptors with major effects on synaptic function. GABAA receptor subunits are hanged after pilocarpine seizures (Rice et al., 1996) or after absence seizures (Banerjee et al., 1998). Altered expression of GABA receptor subunits may alter epileptogenesis (Poulter et al., 1999). There may also be changes in ion channel function as described by Chen et al. (2001) and Bender et al. (2001) after hyperthermic seizures (where the function of hyperpolarization activated, cyclicnucleotide-gated cation channels is altered) and by Becker et al. (2002, this volume) in the pilocarpine status model (where spontaneous limbic seizures are associated with enhanced expression of the cL-1H subunit of the T-type voltage-sensitive calcium channel). Genetic influences
There has long been evidence that different strains of mice have different seizure susceptibilities. This applies to almost all seizure models, including electroshock, chemically induced seizures and reflex epilepsies. These models offer the possibility of detecting single locus differences (see Frankel et al., 2001 for electroconvulsive shock in inbred strains, Ferraro et al., 1997 for kainic acid seizures, Fer-
raro et al., 1999 for pentylenetetrazol seizures and Skradski et al., 1998 for sound-induced seizures). The altemative approach of using microchip arrays to identify differentially expressed genes comparing brain regions in different mouse strains and the effects of seizures in those strains is described by Sandberg et al. (2000) and Del Rio and Barlow (2002, this volume). This approach shows that the induction of immediate early genes 60 min after pentylenetetrazol seizures is closely similar in 129SvEv and C57BL/6 mice and confirms the early induction of COX2. Genetic factors also influence the consequences of seizures. This has been elegantly demonstrated by Schauwecker (2002, this volume). Kainic acidinduced cell death in the mouse hippocampus is strongly modulated by genetic factors (Schauwecker and Steward, 1997), with some standard mouse strains C57BL/6, 129/SvJ and BALB/c) being relatively resistant to hilar and pyramidal cell loss (compared to 129/SvEMS and DBA/2J) in spite of similar seizure severity. As Dr. Schauwecker shows, strain crosses suggest that one or more genes with a dominant effect are responsible for the resistance to kainate-induced cell death. The complexity of the genetic factors influencing the development of Ammon's horn sclerosis following prolonged febrile convulsions in man can be broken down into several headings. The tendency to show convulsions during the course of a febrile illness is under strong genetic control (Hauser et al., 1985). The extent to which this gives rise to selective neuronal death is probably under a separate genetic Control. The nature and severity of the inflammatory response is probably independently genetically controlled as is suggested by the recent study showing specific polymorphisms in interleukin receptor antagonist genes in patients with temporal lobe epilepsy (Kanemoto et al., 2000). There may be additional genetic influences that influence the subsequent process of epileptogenesis. This has been clearly demonstrated in the rodent amygdala kindling model (Mclntyre et al., 1999) where fast and slow kindling strains have been identified. Responsiveness to anti-epileptic drug therapy may also be under genetic control as is suggested by studies on phenytoin resistance in kindled rats (Ebert and L6scher, 1999).
TABLE 1
References
Some consequences of burst firing 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12.
Increase in [Ca2+]i Activation of phospholipase A2 and phospholipase C Immediate early gene expression Altered kinase activity, altered phosphorylation of enzymes, receptors, ion channels Altered ion channel function Altered subunit expression of excitatory and inhibitory receptors Altered synaptic morphology, remodelled dendritic spines Enhanced neurogenesis in dentate Sprouting, altered connectivity Oxidative damage to proteins, lipids and DNA Apoptosis (caspase activation) Necrosis (mitochondrial Ca 2+ poisoning)
Conclusions The burst firing associated with epileptic discharges when sustained for 30-120 min leads to selective neuronal necrosis in hippocampus, amygdala and cortex that is a consequence of calcium overloading of mitochondria. Some consequences of less sustained burst firing are listed in Table 1. At the threshold, there will be effects mediated via actin on dendritic spine shapes that are indistinguishable from physiological effects of AMPA and NMDA receptor activation (Fischer et al., 2000). Activation of phospholipase 2 initiates the arachidonic acid cascade with many complex consequences for gene expression, free radical production and apoptosis as discussed by Bazan et al. (2002, this volume). Functional changes occur through altered phosphorylation of many enzymes and receptors and through altered expression of receptor subunits. These and various morphological changes will lead to significant changes in network function and excitability. These types of change provide one mechanism whereby single seizures can have cumulative effects; the other mechanism is by the progressive accumulation of oxidative damage to proteins, lipids and DNA. Understanding these mechanisms allows the design and testing of a wide variety of neuroprotective strategies, as discussed in Meldrum (2002, this volume).
Alzheimer, A. (1907) Die Gruppierung der Epilepsie. Alg. Z Psychiatric, 64: 418-448. Banerjee, P.K., Tillakaratne, N.J., Brailowsky, S., Olsen, R.W., Tobin, AJ. and Snead, O.C. (1998) Alterations in GABAA receptor alphal and alpha4 subunit mRNA levels in thalamic relay nuclei following absence-like seizures in rats. Exp. Neurol., 154: 213-223. Bazan, N.G., Tu, B. and Rodriguez de Turco, E.B. (2002) What synaptic lipid signaling tells us about seizure-induced damage and epileptogenesis. In: T. Sutula and A. Pitk~inen (Eds.), Do Seizures Damage the Brain. Progress in Brain Research, Vol. 135. Elsevier, Amsterdam, pp. 175-185. Becker, A.J., Wiestler, O.D. and Bltimcke, I. (2002) Functional genomics in experimental and human temporal lobe epilepsy: powerful new tools to identify molecular disease mechanisms of hippocampal damage. In: T. Sutula and A. Pitk'anen (Eds.), Do Seizures Damage the Brain. Progress in Brain Research, Vol. 135. Elsevier, Amsterdam, pp. 161-173. Ben-Aft, Y., Tremblay, E., Ottersen, O.P. and Meldrum, B.S. (1980) The role of epileptic activity in hippocampal and 'remote' cerebral lesions induced by kainic acid. Brain Res., 191: 79-97. Bender, R.A., Brewster, A., Santoro, B., Ludwig, A., Hofmann, F., Biel, M. and Baram, T.Z. (2001) Differential and agedependent expression of hyperpolarization activated, Cyclic nucleotide-gated cation channel iso-forms 1-4 suggest evolving roles in the developing rat hippocampus. Neuroscience (in press). Bengzon, J., Kokala, Z., Elmer, E., Nanobashvili, A., Kokaia, M. and Lindvall, O. (1997) Apoptosis and proliferation of dentate gyms neurons after single and intermittent limbic seizures. Proc. Natl. Acad. Sci. USA, 94: 10432-10437. Bengzon, J., Mohapel, P., Ekdahl, C. and Lindvall, O. (2002) Neuronal apoptosis after brief and prolonged seizures. In: T. Sutula and A. Pitk~inen (Eds.), Do Seizures Damage the Brain. Progress in Brain Research, Vol. 135. Elsevier, Amsterdam, pp. 111-119. Blennow, G., Brierley, J.B., Meldrum, B.S. and Siesj6, B.K. (1978) Epileptic brain damage. The role of systemic factors that modify cerebral energy metabolism. Brain, 101: 687-700. Borgstr6m, L., Chapman, A.G. and Siesj6, B.K. (1976) Glucose consumption in the cerebral cortex of rat during bicucullineinduced status epilepticus. J. Neurochem., 27: 971-973. Bouchet, C. and Cazauvieilh, M. (Bouchet and Cazauvieilh) De l'6pilepsie consider6e dans ses rapports avec l'ali6nation mentale. Arch. G(n. Med., 9: 510-542. Cavanagh, J.B. and Meyer, A. (1956) Aetiological aspects of Ammon's horn sclerosis associated with temporal lobe epilepsy. Br. Med. J., 2: 1403-1407. Cendes, E, Andermann, E, Dubeau, E, Gloor, P., Evans, A. and Jones-Gotman, M. et al. (1993) Early childhood prolonged febrile convulsions, atrophy and sclerosis of mesial structures, and temporal lobe epilepsy: An MRI volumetric study. Neurology, 43: 1083-1087. Chapman, A.G. (1985) Cerebral energy metabolism and seizures.
In: T.A. Pedley and B.S. Meldrnm, Recent Advances in Epilepsy, Vol. 2. Churchill Livingstone, Edinburgh, pp. 1963. Chapman, A.G., Meldrum, B.S. and Siesj6, B.K. (1977) Cerebral metabolic changes during prolonged epileptic seizures in rats. J. Neurochem., 28: 1025-1035. Chen, K., Aradi, I., Thon, N., Eghbal-Ahmadi, M., Baram, T.Z. and Soltesz, I. (2001) Persistently modified H-channels after complex febrile seizures convert the seizure-induced enhancement of inhibition to hyperexcitability. Nat. Med., 7: 1-7. Clark, L.E, Prout, T.R (1903-1904) Status epilepticus: A clinical and pathological study in epilepsy. Am. J. Insanity 60: 291306, 645-675; 61: 81-108. Clifford, D.B., Olney, J.W., Benz, A.M., Fuller, T.A. and Zorumski, C.E (1990) Ketamine, phencyclidine, and MK-801 protect against kalnic acid-induced seizure-related brain damage. Epilepsia, 31: 382-390. Cock, H.R. (2002) The role of mitochondria and oxidative stress in neuronal damage after brief and prolonged seizures. In: T. Sutula and A. Pitk~inen (Eds.), Do Seizures Damage the Brain. Progress in Brain Research, Vol. 135. Elsevier, Amsterdam, pp. 187-196. Covolan, L., Smith, R.L. and Mello, L.E. (2000) Ultrastructural identification of dentate granule cell death from pilocarpine induced-seizures. Epilepsy Res., 41: 9-21. Del Rio, J.A. and Barlow, C. (2002) Genomics and neurological phenotypes: applications for seizure-induced damage. In: T. Sutula and A. Pitk~inen (Eds.), Do Seizures Damage the Brain. Progress in Brain Research, Vol. 135. Elsevier, Amsterdam, pp. 149-160. Earle, K.M., Baldwin, M. and Penfield, W. (1953) Incisural sclerosis and temporal lobe seizures produced by hippocampal herniation at birth. Arch. Neurol. Psychiatry, 69: 27-42. Ebert, U. and Ltischer, W. (1999) Characterization of phenytoinresistant kindled rats, a new model of drag resistant partial epilepsy: influence of genetic factors. Epilepsy Res., 33: 217226. Evans, M., Griffiths, T. and Meldrum, B.S. (1983) Early changes in the rat hippocampus following seizures induced by bicuculline or L-allylglycine: A light and electron microscope study. Neuropathol. Appl. Neurobiol., 9: 39-52. Evans, M.C., Griffiths, T. and Meldrum, B.S. (1984) Kainic acid seizures and the reversibility of calcium loading in vulnerable neurons in the hippocampus. Neuropathol. Appl. Neurobiol., 10: 285-302. Falconer, M.A. (1968) The significance of mesial temporal sclerosis (Ammon's horn sclerosis) in epilepsy. Guy's Hosp. Rep., 117: 1-12. Falconer, M.A. (1974) Mesial temporal (Ammon's horn) sclerosis as a common cause of epilepsy. Aetiology, treatment and prevention.. Lancet, 2: 767-770. Fariello, R.G., Golden, G.T., Smith, G.G. and Reyes, P.E (1989) Potentiation of kainic acid epileptogenicity and sparing from neuronal damage by an NMDA receptor antagonist. Epilepsy Res., 3: 206-213. Ferraro, T.N., Golden, G.T., Smith, G.G., Schork, N.J., St Jean, P., Ballas, C., Cho, H. and Berrettini, W.H. (1997) Mapping
murine loci for seizure response to kainic acid. Mammalian Genome, 8: 200-208. Ferraro, T.N., Golden, G.T., Smith, G.G., St Jean, P., Schork, N.J., Mulholland, N., Ballas, C., Schill, J., Buono, R.J. and Berrettini, W.H. (1999) Mapping loci for pentylenetetrazolinduced seizure susceptibility in mice. J. Neurosci., 19, 67336739. Fischer, M., Kaech, S., Wagner, U., Brinkhaus, H. and Matus, A. (2000) Glutamate receptors regulate actin-based plasticity in dendritic spines. Nat. Neurosci., 3: 887-894. Frankel, W.N., Taylor, L., Beyer, B., Tempel, B.L. and White, H.S. (2001) Electroconvulsive thresholds of inbred mouse strains. Genomics, 74: 306-312. Gastaut, H. (1956) Colloque sur les probl~mes d'anatomie normale et pathologique poses par les drcharges 6pileptiques Marseille. Acta NeuroL Belg., 56:29 (discussion). Gibbs, E, Lennox, W.G. and Gibbs, E.L. (1934) Cerebral blood flow preceding and accompanying epileptic seizures in man. Arch. Neurol. Psychiatry, 32: 257-272. Griffiths, T., Evans, M.C. and Meldrum, B.S. (1983) Intracellular calcium accumulation in rat hippocampus during seizures induced by bicuculline or L-allylglycine. Neuroscience, 10: 385-395. Griffiths, T., Evans, M.C. and Meldrum, B.S. (1984) Status epilepticus: The reversibility of calcium loading and acute neuronal pathological changes in the rat hippocampus. Neuroscience, 12: 557-567. Hauser, W.A., Annegers, J.F., Anderson, V.E. and Kurland, L.T. (1985) The risk of seizure disorders among relatives of children with febrile convulsions. Neurology, 35: 1268-1273. Henshall, D.C., Sinclair, J. and Simon, R.P. (1999) Relationship between seizure-induced transcription of the DNA damageinducible gene GADD45, DNA fragmentation, and neuronal death in focally evoked limbic epilepsy. J. Neurochem., 73: 1573-1583. Henshall, D.C., Sinclair, J. and Simon, R.P. (2000) Spatiotemporal profile of DNA fragmentation and its relationship to patterns of epileptiform activity following focally evoked limbic seizures. Brain Res., 858: 290-302. Holmes, G.L., Khazipov, R. and Ben-Ari, Y. (2002) Seizureinduced damage in the developing human: relevance of experimental models. In: T. Sutula and A. Pitkiinen (Eds.), Do Seizures Damage the Brain. Progress in Brain Research, Vol. 135. Elsevier, Amsterdam, pp. 321-334. Jarrard, L.E. and Meldrum, B.S. (1993) Selective excitotoxic pathology in the rat hippocampus. NeuropathoL AplJl. Neurobiol., 19: 381-389. Joza, N., Susin, S.A., Daugas, E., Stanford, W.L., Cho, S.K. and Li, C.Y.J. et al. (2001) Essential role of the mitochondrial apoptosis-inducing factor in programmed cell death. Nature, 410: 549-554. Kanemoto, K., Kawasaki, J., Miyamoto, T., Obayashi, H. and Nishimura, M. (2000) Interleukin (IL)-ll3, IL-lct, and ILl receptor antagonist gene polymorphisms in patients with temporal lobe epilepsy. Ann. NeuroL, 47: 571-574. Kim, J.H., Guimaraes, P.O., Shen, M.Y., Masukawa, L.M. and Spencer, D.D. (1990) Hippocampal neuronal density in tempo-
10
ral lobe epilepsy with and without gliomas. Acta NeuropathoL, 80: 41-45. Lerner-Natoli, M., Rondouin, G., Relaidi, M., Baldy-Moulinier, M. and Kamenka, J.M. (1991) N-(2-thienyl)cyclohexyl 1piperidine (TCP) does not block kainic acid-induced status epilepticus but reduces secondary hippocampal damage. Neurosci. Lett., 122: 174-178. McIntyre, D., Kelly, M.E. and Dufresne, C. (1999) Fast and slow arnygdala kindling rat strains: comparison of amygdala, hippocampal, piriform and perirhinal cortex kindling. Epilepsy Res., 35: 197-209. Mathern, G.W., Pretorius, J.K. and Babb, T.L. (1995) Influence of the type of initial precipating injury and at what age it occurs on course and outcome in patients with temporal lobe seizures. J. Neurosurg., 82: 220-227. Mathern, G.W., Adelson, ED., Cahan, L.D. and Leite, J.E (2002) Hippocampal neuron damage in human epilepsy: Meyer's hypothesis revisited. In: T. Sutula and A. Pitk~inen (Eds.), Do Seizures Damage the Brain. Progress in Brain Research, Vol. 135. Elsevier, Amsterdam, pp. 237-251. Meldrum, B.S. (1981) Metabolic effects of prolonged epileptic seizures and the causation of epileptic brain damage. In: EC. Rose (Ed.), Metabolic Disorders of the Nervous System. Pitman, London, pp. 175-187. Meldrum, B.S. (1983) Metabolic factors during prolonged seizures and their relation to nerve cell death. Adv. Neurol., 34: 261-275. Meldrum, B.S. (1997) The First Alfred Meyer memorial lecture. Epileptic brain damage: a consequence and a cause of seizures. Neuropathol. Appl. Neurobiol., 23: 185-202. Meldrum, B.S. (2002) Implications for neuroprotective treatments, ln: T. Sutula and A. Pitk~inen (Eds.), Do Seizures Damage the Brain. Progress in Brain Research, Vol. 135. Elsevier, Amsterdam, pp. 487-495. Meldrum, B.S. and Brierley, J.B. (1972) Neuronal loss and gliosis in the hippocampus following repetitive epileptic seizures induced in adolescent baboons by allylglycine. Brain Res., 48: 361-365. Meldrum, B.S. and Brierley, J.B. (1973) Prolonged epileptic seizures in primates: ischaemic cell change and its relation to ictal physiological events. Arch. Neurol., 28: 10-17. Meldrum, B.S. and Horton, R.W. (1973) Physiology of status epilepticus in primates. Arch. Neurol., 28: 1-9. Meldrum, B.S. and Nilsson, B. (1976) Cerebral blood flow and metabolic rate early and late in prolonged epileptic seizures induced in rats by bicuculline. Brain, 99: 523-542. Meldrum, B.S., Vigoroux, R.A. and Brierley, J.B. (1973) Systemic factors and epileptic brain damage. Prolonged seizures in paralysed artificially ventilated baboons. Arch. Neurol., 29: 82-87. Meldrum, B.S., Horton, R.W. and Brierley, J.B. (1974) Epileptic brain damage in adolescent baboons following seizures induced by allylglycine. Brain, 97: 417-428. Menini, C., Meldrum, B.S., Riche, D., Silva-Comte, C. and Stutzmann, J.M. (1980) Sustained limbic seizures induced by intraamygdaloid kainic acid in the baboon: symptomatology
and neuropathological consequences. Ann. Neurol., 8: 501509. Nevander, G., Ingvar, M., Auer, R.N. and Siesjo, B.K. (1985) Status epilepticus in well-oxygenated rats causes neuronal necrosis. Ann. Neurol, 18: 281-290. Olney, J.W., De Gubareff, T. and Sloviter, R.S. (1983) 'Epileptic' brain damage in rats induced by sustained electrical stimulation of the perforant path. II. Ultrastructural analysis of acute hippocampal pathology. Brain Res. Bull., 10:699-712. Peiffer, J. (1963) Morphologische aspekte der epilepsien. Pathogenetische, Pathogisch-Anatomische und Klinische Probleme der Epilepsien, Springer Verlag, Berlin, pp. 1-185. Penfield, W. and Jasper, H. (1954) Epilepsy and the Functional Anatomy of the Human Brain. Churchill, London, pp. 1-896. Penfield, W., von Sfintha, K. and Cipriani, A. (1939) Cerebral blood flow during induced epileptiform seizures in animals and man../. Neurophysiol., 2: 257-267. Pfleger, L. (1880) Beobachtungen tiber Schrumpfung und Sklerose des Ammonshorns bei Epilepsie. Alg. Z. Psychiatrie, 36: 359-365. Poulter, M.O., Brown, L.A., Tynan, S., Willick, G., William, R. and McIntyre, D.C. (1999) Differential expression of alphal, alpha2, alpha3 and alpha5 GABAA receptor subunits in seizure-prone and seizure-resistant rat models of temporal lobe epilepsy. J. Neurosci., 19: 4654-4661. Rice, A., Rafiq, A., Shapiro, S.M., Jakoi, E.R., Coulter, D.A. and De Lorenzo, R.J. (1996) Long-lasting reduction of inhibitory function and gamma-aminobutyric acid type A receptor subunit mRNA expression in a model of temporal lobe epilepsy. Proc. Natl. Acad. Sci. USA, 93: 9665-9669. Sagar, H.J. and Oxbury, J.M. (1987) Hippocampal neuron loss in temporal lobe epilepsy: Correlation with early childhood convulsions. Ann. Neurol., 22: 334-340. Sandberg, R., Yasuda, R., Pankratz, D.G., Carter, T.A., Del Rio, J., Wodicka, L., Mayford, M., Lockhart, D.J. and Barlow, C. (2000) Regional and strain-specific gene expression mapping in the adult mouse brain. Proc. Natl. Acad. Sci. USA, 97: 11038-11043. Schanne, EA.X., Kane, A.B., Young, E.E. and Farber, J.L. (1979) Calcium dependence of toxic cell death: a final common pathway. Science, 206: 700-702. Schauwecker, EE. (2002) Complications associated with genetic background effects in models of experimental epilepsy. In: T. Sutula and A. Pitk~inen (Eds.), Do Seizures Damage the Brain. Progress in Brain Research, Vol. 135. Elsevier, Amsterdam, pp. 139-148. Schauwecker, EE. and Steward, O. (1997) Genetic determinants of susceptibility to excitotoxic cell death: Implications for gene targeting approaches. Proc. Natl. Acad. Sci. USA, 94: 4103-4108. Scholz, W. (1951) Die Krampfschiidigungen des Gehirns. Springer, Berlin. Scholz, W. (1959) The contribution of patho-anatomical research to the problem of epilepsy. Epilepsia, 1: 36-55. Shorvon, S. (1994) Status Epilepticus. Cambridge University Press, Cambridge. Skradski, S.L., White, H.S. and Ptacek, L.J. (1998) Genetic
11
mapping of a locus (mass 1) causing audiogenic seizures in mice. Genomics, 49: 188-192. Sloviter, R.S. (1983) 'Epileptic' brain damage in rats induced by sustained electrical stimulation of the perforant path, I. Acute electrophysiological and light microscopic studies. Brain Res., 10: 675-697. Sloviter, R.S., Dean, E., Sollas, A.L. and Goodman, J.H. (1996) Apoptosis and necrosis induced in different hippocampal neuron populations by repetitive perforant path stimulation in the rat. J. Comp. Neurol., 366: 516-533. Sommer, W. (1880) Erkrankung des Ammonshornes als aetiologisches Moment der Epilepsie. Arch. Psychiat. Nervenkrank., 10: 631-675. Spielmeyer, W. (1927) Die Pathogenese des epileptischen
Krampfes. Z. Ges. NeuroL Psychiatrie, 109: 501-520. Spielmeyer, W. (1930) Anatomic substratum of the convulsive state. Arch. NeuroL Psychiatry, 23: 869-875. Turner, D.A. and Wyler, A.R. (1981) Temporal lobectomy for epilepsy: mesial temporal herniation as an operative and prognostic finding. Epilepsia, 22: 623-629. Veith, G. (1960) On the pathogenesis of perinatal brain lesions. Geburtsh. Frauenheilk., 20: 905-909. Vogt, C. and Vogt, O. (1922) Erkrankungen der Groszhirnrinde im Lichte der topistik Pathoklise, und Pathoarchitektonik. J. PsychoL Neurol. (Leipzig), 28: 1-171. Wrogemann, K. and Pena, S.D.J. (1976) Mitochondrial calcium overload: a general mechanism for cell-necrosis in muscle diseases. Lancet, 1: 672-674.
T. Sutula and A. Pitkanen (Eds.) Progress in Brain Research, Vol. 135 © 2002 Elsevier Science B.V. All rights reserved
CHAPTER 2
Are seizures harmful: what can we learn from animal models? Andrew J. Cole *, Sookyong Koh and Yi Zheng Epilepsy Research Laboratory, Massachusetts General Hospital and Department of Neurology, Harvard Medical School, Boston, MA 02114, USA
Abstract: Epilepsy is a brain disease that requires distributed neuronal networks for its expression. Several characteristics of epilepsy, including its natural history, the latency between an initial insult and the first manifestation of seizures, the complex interaction of seizures with development as a function of developmental stage, the modulating effect of systemic physiological responses, and the fact that seizures are ultimately defined by a combination of electrical and behavioral criteria all suggest that epilepsy should ideally be studied in an intact whole animal preparation. Such preparations offer the ability to study acute and chronic changes in brain structure and function after single or repeated seizures. Animal models have major limitations, however, including strain specificity, difficulty in isolating potentially confounding variables, a relative lack of accessible higher cortical functions, such as language and abstract processing, and shorter lifespans that may be insufficient to allow the complete expression of seizure-related injury. Information we have learned from animal studies includes a broad understanding of the chemical, molecular and anatomic consequences of seizures, including their temporal and spatial relationships to each other, and information on the consequences of seizures as a function of development. Recent studies have cast light on potential mechanisms of resistance to seizure-induced injury in the developing brain. In the future, we can anticipate that animal models will continue to be useful, especially when whole-animal preparations are used to generate material for detailed in vitro examination.
Introduction Epilepsy is a brain disease, and epileptic seizures result from abnormal paroxysmal activity o f populations o f neurons. The careful reader will note that by definition epilepsy and epileptic seizures cannot be conceptualized as disorders o f single neurons per se. It is perhaps less clear whether the disease state and its primary s y m p t o m can be successfully recapitulated in an isolated array of neurons, either a plate o f cultured cells or an ex vivo preparation,
* Correspondence to: A.J. Cole, MGH Epilepsy Service, VBK 830, Massachusetts General Hospital, Fruit Street, Boston, MA 02114, USA. Tel.: -t-1-617-726-3311; Fax: +1-617-726-9250; E-maih cole.andrew @mgh.harvard.edu
such as a h i p p o c a m p a l or cortical slice. Furthermore, epilepsy, a disorder manifest by recurrent unprovoked seizures, is by definition a chronic disease, or at least a disease that manifests itself over a period o f time, not simply at an instant. Reductionist models typically are acute, in the case of ex vivo preparations, or developmentally disturbed, in the case o f cultured cells. Moreover, epilepsy occurs in the context of ongoing physiological processes such as perfusion, oxygenation, glucose metabolism, acidbase regulation, thermoregulation, endocrine modulation and the like, some of which m a y have critical interactions with the disease state itself. W h i l e the confounding influences of normal physiological responses m a y make the interpretation of whole animal experiments difficult, their absence m a y represent an important and under-recognized limitation of reductionist approaches. For these reasons, it seems that
14 TABLE 1 The spectrum of animal models Electrical stimulation
Maximal electroconvulsiveseizures (MECS) Perforant-path stimulation (PPS) Kindling
Chemoconvulsants
Systemic
Kainate Pilocarpine Picrotoxin Bicuculline Penicillin Kainate Pilocarpine Picrotoxin Bicuculline Tetanus toxin Pertussis toxin Alumina cream Penicillin
Intracerebral
Topical Physical models
Hyperthermia Freeze lesions Photic stimulation (Papiopapio) Auditory stimulation (Swiss DBA2 mice)
Genetic models
Spontaneous
Mutant
Genetically epilepsy-prone rat strain (GEPRS) Strasbourg rats (Absence) Epileptic beagles Stargazer Lurcher Totterer Mocha
Transgenic Knockout Spontaneous seizure models
Post-kindling Post-kainate Post-pilocarpine
the most effective way to model epilepsy should be to utilize whole animal preparations. And indeed a considerable body of epilepsy research work has utilized animal models, leading to many fundamental insights. In this chapter, we will review some of the important uses of animal models, discuss their limitations, and consider the role of animal models going forward. In particular, we will focus on the issue of whether a single or initial seizure is harmful, and how it might relate to the development of epilepsy.
The spectrum of animal models A wide variety of animal models of epilepsy and epileptic seizures exist. The major models are listed
in Table 1. The tremendous diversity of available animal models offers both opportunity and challenge. Many of the models are acute, and many are, in fact, models of status epilepticus which may have important differences from isolated seizures. Genetic models often have phenotypes that are complex, with seizures as only one manifestation of a more pervasive structural, functional or developmental problem. Strain differences in responses to epileptogenic or convulsant stimuli make comparisons between animals from different laboratories difficult, but may offer opportunity to identify pro- or anti-epileptic genes using differential screening approaches (Schauwecker and Steward, 1997; Sandberg et al., 2000; Schauwecker, 2000). Chemo-
15 TABLE 2
TABLE 3
Advantages and limitationsof animal models
Situations in which whole-animalmodels are required
Advantages
•
•
Allowassessment of seizure causes and consequences in an intact preparation • Survivaltime can be adjusted to examine temporal evolution of post-ictal changes • Developmentalstage can be selected, and events at one point in developmentcan be studied with respect to effects manifested at a later stage • Repeated events can studied (e.g. kindling) • Effects of physiologicalmilieu are integrated
• • • • • •
Limitations • •
Manyoffer snapshot picture of ictal/post-ictal events Manyresult in status epilepticus, which may not adequately model typical epilepsy • Straindifferences in seizures and responses highlight the difficulties in generalizingfindings in a specific model • Effects of physiologicalmilieu are integrated
• • •
To examine transductionof an input function into an output function without requiring knowledge of the mechanism To examine systems level physiology where distant or unknown connectionsmay have a role To study anatomic patterns of responses To study the relationship of anatomic findings,e.g. injury, plasticity to specific molecular markers To study developmentallyregulated anatomic, biochemical and functional events To study the effects of chronic or recurrent seizures To study the effects of specific genetic manipulationson phenotype To examine the influenceof physiologicalmilieu on seizures and their consequences To generate biological material for examinationafter seizures To confirm findingsfrom reductionist systems in the scaled-up whole animal situation
Insights from animal models convulsants may have systemic effects that are either completely independent from seizures, or even more problematic, result in seizures only as a secondary consequence to injury or functional disturbance. Finally, some of the important correlates of seizures and epilepsy, such as memory loss, behavioral changes, and secondary psychiatric disturbance are at best difficult to measure in epileptic animals. By contrast, all animal models share the property of having anatomically intact central nervous systems with functional connections and measurable efferent responses. A n i m a l s can be developmentally monitored, and prolonged survival is possible to allow examination of delayed effects of specific stimuli or treatment. Most recently, the ability to manipulate the genetic endowment of model animals has opened the door to allowing examination of complex interactions between genes and behavior, between nature and nurture. The advantages and disadvantages of animal models are listed in Table 2. In the remainder of this chapter, we will identify specific situations where animal models should be particularly useful, and offer examples of insights that have come from animal models that might not have been available elsewhere. Several situations in which animal models have an obvious utility can be defined and are listed in Table 3.
Seizures trigger a cascade of biochemical, anatomic and functional changes in the central nervous system It has long been recognized that critical aspects of brain development are activity-dependent. For example, the pioneering experiments of Hubel and Weisel established the critical role of visual input in the post-natal organization of the visual system, including the establishment of cortical columns, pruning of redundant connections, and establishment of the critical property of surround inhibition (Hubel and Wiesel, 1970; Hubel et al., 1977). Later studies have demonstrated that cortical plasticity, both functional and anatomic, occurs in response to altered afferent activity, e.g. amputation or fusion of digits in primates (Merzenich et al., 1984; Allard et al., 1991). Perhaps surprisingly, our appreciation that brief seizures trigger long-term changes in CNS properties is relatively recent. Fig. 1A summarizes, in a schematic form, some of the events that occur after seizures. A critical point is that each of these observations came from animal studies. As an example, consider the observation that brief seizures induced by pentylenetetrazole or maximal electroconvulsive treatment results in rapid and transient expression of a large class of immediate early genes, many encoding transcription factors. While cell culture studies of
16
Time Course of Biochemical, Anatomic and Functional Changes After Seizures (a)
S u s c e iflbilit to recur 'ent se z u r e s
(week to m, ,nths) O - - - - - ' 4 ) N e u "ogene ds (da~ s- wee ts)
Be] laviot~ J Def'N its (61 tours- montl ~) A
dm
leuro~ al Cell Loss hour
Glial Activa ion (6 hol lrs - 5 lays)
0=== ='=0
A v
0"
E
arly G me Ac tivafio (.J: m t n - 6 horn s)
; a + + I m Infl ax
0"
rnse~, min~t .~s)
I
10-3
P r o t i inexp] ession e.g. S o n ~ tostafi ~, NPI (2 ho ]rs - 3 l a y s )
K i n a ~e Acti ration (5 mi nutes I o 2A h,,urs)
. -
-4
- 14 q lays)
10-2 lO-I
1
10
10 ~
1~
1 0 I~
1 hour
1 min
I
10 4
1 day
10 9
1010
1 week
Time (seconds)
Time Course of Biochemical, Anatomic and Functional Changes After Seizures S u s c e )tibilit tto recur *nt se zures (week ; to m~ ~uths)
(b)
O - - - - - - 4 i Neu rogem iis
( d a s - we, ~s)
Be tavior IDefie Is (6 ours, montt s) A v
v
0-----4
~eurot al Cel Loss 6 horn s - 141 ays)
Glial~ ictival Ion (6 ho, irs - 5 ~lays) FCOR
2
Som~ t,ostati n, NP~ (2 ho Jrs, 3 Ja3~) ~atil)|l ltlJles
ECT
o 24 h ~ c s l
lkatio 6 hm/ s}
}lR ~{
(a++
It" I 10 a
10 -2
10 "~
1
10
10 1 min
10 3
|,m hfft p~.
{ n~, t- - . i , l q ~ s! i
I0 a
1 hour
I day
I
tO~ I week
10 9
1010
Time (seconds) Fig. 1. (A) Diagrammatic representation of some of the biochemical, anatomic and functional changes seen after seizures in a variety of electrical and chemoconvulsant animal models. (B) Diagrammatic representation of changes seen in ECT models (indicated in green). Note that many of the lasting changes have not been reported after single or repeated ECT.
17
Parallel Model ]
(a)
Seizures [ Sprouting
(b)
I Injury
Gliosis
Neurogenesis
Functional Change
Series Model Seizures Iv wy G] psis Spro ating Neure enesis Functional Change
Fig. 2. Theoreticalrepresentationof the relationship between various documentedchanges occurring after seizures. (A) Parallel model indicates that specific consequences of seizures may occur independently of each other. (B) Series model represents the alternative hypothesisthat later latencychanges depend on earlier events for their expression.
activity-dependent gene expression could have suggested this phenomenon, only whole animal studies allowed us to appreciate the extraordinary anatomic specificity of this response (Morgan et al., 1987; Saffen et al., 1988; Cole et al., 1989). Similarly, observation of a variety of events including kinase activation (Murray et al., 1998, 1999; Anderson et al., 2000), neuropeptide regulation (Gall et al., 1990; Baraban et al., 1993; Vezzani et al., 1999; Madsen et al., 2000), cell loss (Margerison and Corsellis, 1966; Corsellis and Meldrum, 1976; Schwob et al., 1980; Gloor, 1991; Sloviter, 1994), mossy fiber sprouting (Sutula et al., 1988; Cavazos et al., 1991; Wuarin and Dudek, 1996; Patrylo et al., 1999), enhanced neurogenesis (Parent and Lowenstein, 1994), altered receptor expression, chronic behavioral deficits, and altered susceptibility to recurrent seizures are all the result of animal studies. With the exception of the latter phenomenon (see next section), it remains un-
clear whether any or all of these effects of a single seizure contribute to the development of epilepsy. Moreover, it is unclear whether whatever contribution they may have is organized in series or in parallel (Fig. 2). Obviously this issue has critical therapeutic implications.
Early life seizures increase susceptibility to later life seizures and neuronal injury An important observation from animal studies is that seizures early in life result in a long-term susceptibility to recurrent seizures with resultant neuronal injury and behavioral deficits later in life. This finding has been established in multiple laboratories using a variety of animal models including repeated kainate administration (Koh et al., 1999), early life hyperthermic seizures (Dube et al., 2000), and earlylife fluroythyl-induced status (Holmes et al., 1998;
18 Schmid et al., 1999). Each of these models shares the property that anatomic injury is difficult or impossible to detect after the initial insult, suggesting that the resulting susceptibility is the consequence of a functional change in network properties. These studies raise several important questions: (1) What is the transduction process that results in enhanced seizure-susceptibility later in life? (2) Why are juvenile animals resistant to seizureinduced injury? (3) Is there a critical point in development before or after which seizures no longer have this long-term effect?
Early-life seizures alter synaptic connectivity in developing brain One hypothesis, unproven, is that early life seizures may stabilize immature synaptic connections normally destined for removal, thereby resulting in an intrinsically hyperexcitable brain in adulthood. Evidence for this concept comes from Grigonis and Murphy (1994) who showed that topical application of penicillin to immature rabbit visual cortex resulted in persistence of the immature pattern of callosal projections without the pruning that occurs in normal development. By contrast, Swann and colleagues have found that early seizures resulted in a reduction in dendritic spine density, suggesting an alternative substrate for lasting functional change (Jiang et al., 1998). Similar findings have been reported in human surgical tissue (Multani et al., 1994) and in chronic animal seizure models (Willmore et al., 1980) as well.
Resistance to neuronal injury in the juvenile brain Many studies have established that in the rodent, seizures prior to P21 result in little, if any, detectable injury. In the adult brain, we have observed that treatment with nerve growth factor attenuates hippocampal injury after kainate-induced seizures (Weiss et al., 1995). Ambient levels of NGF are maximal during development, peaking at P14-15.
Because we have no data on this point, this question will not be discussed.
We therefore hypothesized that high ambient levels of NGF may prevent seizure-induced neuronal injury in the immature brain. To test this hypothesis, we selectively lesioned cholinergic neurons bearing the low-affinity neurotrophin receptor, p75ntfr using a selective immunotoxin, 192-IgG-saporin. In preliminary experiments, rats treated with 192-IgG-saporin on P7 that showed complete loss of basal forebrain cholinergic neurons and marked depletion of acetylcholinesterase stained terminals in hippocampus also demonstrated severe selective cell loss in CA3 after subsequent treatment with kainate on P16. By contrast, animals treated with saline on P7 and animals in which the saporin treatment was unsuccessful in depleting cholinergic neurons had no kainate-induced hippocampal injury (Fig. 3). While these experiments are consistent with our hypothesis, b e c a u s e p75ntfr binds several neurotrophins including BDNF, NT-3 and NT-4/5 it remains unclear whether high levels of NGF are critical to neuronal survival in this experimental paradigm. Moreover, in light of the fact that seizures induce NGF expression, and the findings of Grigonis and Murphy described above, we must consider the possibility that enhanced neuronal survival may have negative consequences with respect to later seizure susceptibility.
Are seizures neuroprotective ? The preceding discussion emphasizes the commonly held belief that seizures are harmful. Seizure-induced phenomena, such as neuronal loss, synaptic remodeling and aberrant neuronal proliferation, especially in the context of decreased performance on behavioral testing and enhanced susceptibility to recurrent seizures would seem to support this notion. Long clinical experience and recent experimental data, however, have challenged this axiom. For example, for many years psychiatrists have been treating patients with refractory depression with electroconvulsive seizures, often with gratifying results. There is little to suggest that ECT, as typically applied causes significant anatomic injury or behavioral dysfunction, although patients who receive hundreds of treatments may provide anecdotal exceptions. Suggestions that ECT causes progressive brain atrophy and hippocampal changes have not been supported by careful clinical studies (Sheline et al., 1999; Ende
19
DNAF
CV
Saline - Kainate
Saporin - Kainate
Fig. 3. Neuronal injury after kainate-induced seizures in P15 rats. DNAF indicates DNA fragmentation, a marker of neuronal injury. CV indicates Cresyl violet Nissl staining showing neuronal integrity. Saline-Kainate indicates animals pretreated with intraventricular saline on P7, prior to kainate on P15. Saporin-Kainate indicates animal treated with 192-IgG-Saporinintraventricularlyon P7 prior to kainate on PI5. Note injury and cell loss in CA3 after saporin treatment. et al., 2000) which have concluded that depression itself and its pharmacological treatment are confounding variables with a more powerful effect on structure and function. By contrast, animal studies during development have pointed out injurious effects of ECT (Wasterlain and Plum, 1973; Jorgensen et al., 1980) and in experimental systems electroconvulsive seizures have been shown to induce abnormal gene expression (Morgan et al., 1987; Saffen et al., 1988; Cole et al., 1990, 1997), kinase activation (Baraban et al., 1993), protein synthesis (Cole et al., 1990; Gall et al., 1991; Bhat et al., 1993), and neurotransmitter receptor expression (Bergstrom and Kellar, 1979; Kellar et al., 1981; Lerer, 1984; Green et al., 1986) (See Fig. 1B). These observations suggest that many of the early events occurring after seizures, while perhaps necessary, are not sufficient to mediate later emergence of neuronal injury, synaptic reorganization, and network dysfunction. They also emphasize the possibility that many of the events occurring af-
ter seizures are arranged in parallel, rather than in series (see Fig. 2A,B). Whether seizures are harmful, neutral or beneficial may depend on seizure type, e.g. brief electrically induced generalized attacks occurring in controlled clinical circumstances versus spontaneous seizures of variable duration occurring in an uncontrolled environment, or host characteristics, such as the presence of underlying neurological (as opposed to psychiatric) dysfunction or disease. In any case, these observations emphasize the need for cautious and unbiased interpretation of the observations gathered from experimental systems. Recent experimental data have also challenged the notion that seizures are harmful. For example, Greenberg and colleagues (Sasahira et al., 1995) found that repeated bicuculline seizures separated by 1, 3, 5 or 7 days conferred a time-dependent protective effect against hippocampal injury induced by subsequent seizures in the CA3c sector of the hippocampus. They coined the term 'epileptic tolerance'
20 to describe this phenomenon, and suggested that enhanced expression of heat-shock proteins caused by the initial seizure might be responsible for subsequent protection against recurrent seizure-induced injury. Their study suggested, however, that the protection was brief and of uncertain clinical relevance. Mclntyre and colleagues (Kelly and Mclntyre, 1994) have shown that kindling stimulation protects against kainate seizure-induced injury for up to 28 days, and Penner and colleagues have confirmed this result in a rapid kindling paradigm (Penner et al., 2001). Similarly, Gale and colleagues have reported in abstract form that repeated electroconvulsive seizures protect against kainate-seizure induced injury in experimental animals. ECT has been shown to cause declining seizure severity with repeated administration (Cole et al., 1990), perhaps the behavioral analogue of the relative refractory period described in synaptic physiology. It will therefore be important to review Gale's work critically to determine the duration of protection conferred.
Transgenic and knockout experiments The ability to determine the genetic endowment of experimental animals has offered an important alternative to classical pharmacology for hypothesis testing. Rather than relying on specific agonists or antagonists, one can now directly block either the synthesis or activity of specific molecules and then examine the resultant phenotype. These studies may be confounded by at least two practical issues. First, animals born with altered genomes may utilize compensatory mechanisms to overcome the induced deficits, or they may develop abnormally in a manner that renders hypothesis testing impossible or irrelevant. The extreme example, of course, is the embryonic-lethal transgene or knockout. More subtle is the situation where alternative isoforms of specific gene products serve to partially or completely compensate for the genetic alteration. To some extent, the development of inducible expression systems for transgenes and conditional knockouts may partially overcome these limitations. Second, strain differences in genetic background may, in fact, have more influence on phenotype than the targeted genetic alteration itself. A graphic example of this phenomenon came from studies of p53 knockout
animals. Several groups of investigators found that p53 knockout animals were protected against kainate seizure-induced neuronal injury (Morrison et al., 1996). Another group, however, using p53 knockout animals from a different source, were unable to reproduce that result (Schauwecker and Steward, 1997). It subsequently became apparent that the difference in the two studies was the result of strain differences in the genetic backgrounds of mice used to develop the knockout lines. Animal models in the new millennium
As we move into the new millennium, a new trend in animal models is emerging that promises to offer powerful insights into the cause and effect of seizures. These models share the property that in vivo and in vitro techniques are combined to allow experiments that could not be conducted in either environment exclusively. While these strategies are perhaps best thought of as evolutionary, rather than revolutionary, they deserve special mention none the less. Three experimental paradigms provide illustrations of these approaches.
Use of biological material from genetically manipulated animals for in vitro study Perhaps one of the most valuable uses of genetic manipulation results from the ability to harvest biological material from manipulated animals for in vitro studies. For example, recent studies have demonstrated that conditional knockout of the neuronal MAP kinase kinase (MEK) gene in hippocampus alters the characteristics of long-term potentiation as studied in hippocampal slices (Atkins et al., 1998; Selcher et al., 1999; Schafe et al., 2000) (Kelleher, personal communication). Similar approaches are now routinely undertaken to examine the effect of genetic manipulation on in vitro physiology using slice preparations and patch clamp techniques, and on neuronal viability and biochemical responses using primary neuronal cell culture techniques.
Receptor alterations in epileptic animals Another example of the combination of in vitro and in vivo techniques comes from the work of
21 Coulter and colleagues. These investigators have developed spontaneously epileptic animals using the pilocarpine model. After documenting recurrent seizures, they have prepared hippocampal slices and documented physiological abnormalities, especially in the properties of G A B A receptors (Gibbs et al., 1997). They have then gone on to use RT-PCR techniques on single cells from these slices to document changes in the specific G A B A receptor subunits expressed in epileptic animals as compared to controls (Brooks-Kayal et al., 1998). Interestingly, an extension of this study to examine G A B A receptor subunit expression in animals after the initial seizure, but before the development of spontaneous seizures could provide insight into the fundamental question before this meeting, whether a single seizure is harmful. Microarray analysis o f epileptic animals
A third example of the use of in vivo and in vitro techniques in combination illustrates the idea that this approach is really an evolutionary extension of older analytical methods where in situ techniques were used to characterize molecular responses to seizures. Microarray analysis of cDNAs generated from epileptic material offers an open-ended method for documenting altered gene expression that requires little de facto knowledge of the targets to be examined. In its earliest incarnations, 2-D get electrophoresis of proteins and differential screening of cDNA libraries led to the identification of previously unknown molecular responses to abnormal activity. For example, the finding that Homer (Brakeman et al., 1997), GRIP (Dong et al., 1997) and ARC (Lyford et al., 1995) were regulated after seizures came from a differential screening strategy. Homer and GRIP are PDZ-domain containing proteins that interact with metabotropic- and AMPA-type glutamate receptors, respectively. ARC, by contrast, is expressed mainly in dendrites where it may mediate synaptic plasticity. Each of these proteins has the potential to mediate changes in synaptic efficiency that are long-lasting. More recently, this approach has been extended by the use of microarray analysis in which thousands of known and unknown transcripts can be quantitatively assessed in response to seizures and compared to controls to look for both up- and down-regulation of activity-dependent tran-
scripts (Sandberg et al., 2000). Dingledine has used this approach to examine differences between early and late transcriptional responses, whereas Lowenstein and colleagues have concentrated on examining transcripts that are regulated both during development and after seizures. Applications of this technology, which starts with the whole animal and quickly moves to the in vitro environment will be limited only by the arrays of targets available and the imagination of investigators.
References Allard, T., Clark, S.A., Jenkins, W.M. and Merzenich, M.M. (1991) Reorganization of somatosensory area 3b representations in adult owl monkeys after digital syndactyly. J. Neurophysiol., 66(3): 1048-1058. Anderson, A.E., Adams, J.E, Varga, A.W., Selcher, J.C., Feng, J., Trzaskos, J.M. and Sweatt, J.D. (2000) Essential role for ERK MAPK activation and K+ channel phosphorylation in the kainate model of epilepsy. Soc. Neurosci. Abstr., 26:1836. Atkins, C.M., Selcher, J.C., Petraitis, J.J., Trzaskos, J.M. and Sweatt, J.D. (1998) The MAPK cascade is required for mammalian associative learning. Nat. Neurosci., 1(7): 602-609. Baraban, J.M., Fiore, R., Sanghera, J.S., Paddon, H.B. and Pelech, S. (1993) Identification of p42 mitogen-activated protein kinase as a tyrosine kinase substrate activated by maximal electroconvulsive shock in hippocampus. J. Neurochem., 60: 330-336. Bergstrom, D.A. and Kellar, K.J. (1979) Effect of electroconvulsive shock on monoaminergic receptor binding sites in rat brain. Nature, 278: 464-466. Bhat, R.V., Tansk, F.A., Baraban, J.M., Mains, R.E. and Eipper, B.A. (1993) Rapid increases in peptide processing enzyme expression in hippocampal neurons. J. Neurochem., 61(4): 1315-1322. Brakeman, ER., Lanahan, A.A., O'Brien, R., Roche, K., Barnes, C.A., Huganir, R.L. and Worley, EE (1997) Homer: a protein that selectively binds metabotropic glutamate receptors. Nature, 386(6622): 284-288. Brooks-Kayal, A.R., Shumate, M.D., Jin, H., Rikhter, T.Y. and Coulter, D.A. (1998) Selective changes in single cell GABA(A) receptor subunit expression and function in temporal lobe epilepsy. Nat. Med., 4(10): 1166-1172. Cavazos, J.E., Golarai, G. and Sutula, T.E (1991) Mossy fiber synaptic reorganization induced by kindling: time course of development, progression, and permanence. J. Neurosci., 11(9): 2795-2803. Cole, A.J., Saffen, D.W., Baraban, J.M. and Worley, EE (1989) Rapid increase of an immediate early gene messenger RNA in hippocampal neurons by synaptic NMDA receptor activation. Nature, 340: 474-476. Cole, A.J., Abu Shakra, S., Saffen, D.W., Baraban, J.M. and Worley, P.E (1990) Rapid rise in transcription factor mRNAs
22
in rat brain after electroshock-induced seizures. J. Neurochem., 55: 1920-1927. Cole, A.J., Koh, S., Liu, Z. and Holmes, G.L. (1997) DNA fragmentation in developing brain after kainate-induced seizures. Epilepsia, 38(Suppl. 8): 40. Corsellis, J.A.N. and Meldrum, B.S. (1976) Epilepsy. In: W. Blackwood and J.A.N. Corsellis (Eds.), Greenfield's Neuropathology, 3 edn. Edward Arnold, London, pp. 771-795. Dong, H., O'Brien, R.J., Fung, E.T., Lanahan, A.A., Worley, P.F. and Huganir, R.L. (1997) GRIP: a synaptic PDZ domaincontaining protein that interacts with AMPA receptors. Nature, 386(6622): 279-284. Dube, C., Chen, K., Eghbal-Ahmadi, M., Brunson, K., Soltesz, I. and Baram, T.Z. (2000) Prolonged febrile seizures in the immature rat model enhance hippocampal excitability long term. Ann. Neurol., 47(3): 336-344. Ende, G., Braus, D.E, Walter, S., Weber-Fahr, W. and Henn, EA. (2000) The hippocampus in patients treated with electroconvulsive therapy: a proton magnetic resonance spectroscopic imaging study. Arch. Gen. Psychiatry, 57( 10): 937-943. Gall, C., Lauterborn, J., Isackson, E and White, J. (1990) Seizures, neuropeptide regulation, and mRNA expression in the hippocampus. Prog. Brain Res., 83: 371-390. Gall, C., Lauterbom, J., Bundman, M., Murray, K. and Isackson, EJ. (1991) Seizures and the regulation of neurotrophic factor and neuropeptide gene expression in brain [Review]. Epilepsy Res. Suppl., 4: 225-245. Gibbs, J.W., Shumate, M.D. and Coulter, D.A. (1997) Differential epilepsy-associated alterations in postsynaptic GABA(A) receptor function in dentate granule and CA1 neurons. J. Neurophysiol., 77(4): 1924-1938. Gloor, E (1991) Mesial temporal sclerosis: historical background and an overview from a modem perspective. In: H. Luders (Ed.), Epilepsy Surgery. Raven Press, New York, pp. 689-703. Green, A.R., Heal, D.J. and Goodwin, G.M. (1986) The effects of electroconvulsive therapy and antidepressant drugs on monoamine receptors in rodent brain-similarities and differences. Ciba Found. Symp., 123: 246-267. Grigonis, A.M. and Murphy, E.H. (1994) The effects of epileptic cortical activity on the development of callosal projections. Dev. Brain Res., 77(2): 251-255. Holmes, G.L., Gairsa, J.L., Chevassus-Au-Louis, N. and Ben Ari, Y. (1998) Consequences of neonatal seizures in the rat: morphological and behavioral effects. Ann. Neurol., 44(6): 845-857. Hubel, D.H. and Wiesel, T.N. (1970) The period of susceptibility to the physiological effects of unilateral eye closure in kittens. J. Physiol., 206: 419-436. Hubel, D.H., Wiesel, T.N. and LeVay, S. (1977) Plasticity of ocular dominance columns in monkey striate cortex. Philos. Trans. R. Soc. Lond. B Biol. Sci., 278(961): 377-409. Jiang, M., Lee, C.L., Smith, K.L. and Swann, J.W. (1998) Spine loss and other persistent alterations of hippocampal pyramidal cell dendrites in a model of early-onset epilepsy. J. Neurosci., 18(20): 8356-8368. Jorgensen, O.S., Dwyer, B. and Wasterlain, C.G. (1980) Synaptic
proteins after electroconvulsive seizures in immature rats. J. Neurochem., 35(5): 1235-1237. Kellar, K.J., Cascio, C.S., Butler, J.A. and Kurtze, R.N. (1981) Differential effects of electroconvulsive shock and antidepressant drugs on serotonin-2 receptors in rat brain. Eur. J. Pharmacol., 69: 515-518. Kelly, M.E. and McIntyre, D.C. (1994) Hippocampal kindling protects several structures from the neuronal damage resulting from kainic acid-induced status epilepticus. Brain Res., 634(2): 245-256. Kob, S., Storey, T.W., Santos, T.C., Mian, A.Y. and Cole, A.J. (1999) Early-life seizures in rats increase susceptibility to seizure-induced brain injury in adulthood [In Process Citation]. Neurology, 53(5): 915-921. Lerer, B. (1984) Electroconvulsive shock and neurotransmitter receptors: implications for mechanism of action and adverse effects of electroconvulsive therapy. Biol. Psychiatry, 19(3): 361-383. Lyford, G.L., Yamagata, K., Kaufmann, W.E., Barnes, C.A., Sanders, L.K., Copeland, N.G., Gilbert, D.J., Jenkins, N.A., Lanahan, A.A. and Worley, EE (1995) Arc, a growth factor and activity-regulated gene, encodes a novel cytoskeletonassociated protein that is enriched in neuronal dendrites. Neuron, 14(2): 433-445. Madsen, T.M., Greisen, M.H., Nielsen, S.M., Bolwig, T.G. and Mikkelsen, J.D. (2000) Electroconvulsive stimuli enhance both neuropeptide Y receptor Y1 and Y2 messenger RNA expression and levels of binding in the rat hippocampus. Neuroscience, 98(1): 33-39. Margerison, J.H. and Corsellis, J.A.N. (1966) Epilepsy and the temporal lobes: a clinical encephalographic and neuropathological study of the brain in epilepsy, with particular reference to the temporal lobes. Brain, 89: 499-530. Merzenich, M.M., Nelson, R.J., Stryker, M.E, Cynader, M.S., Schoppmann, A. and Zook, J.M. (1984) Somatosensory cortical map changes following digit amputation in adult monkeys. J. Comp. Neurol., 224(4): 591-605. Morgan, J.I., Cohen, D.R., Hempstead, J.L. and Curran, T. (1987) Mapping patterns of c-los expression in the central nervous system after seizure. Science, 237(4811): 192-197. Morrison, R.S., Wenzel, H.J., Kinoshita, Y., Robbins, C.A., Donehower, L.A. and Schwartzkroin, EA. (1996) Loss of the p53 tumor suppressor gene protects neurons from kainateinduced cell death. J. Neurosci., 16(4): 1337-1345. Multani, R, Myers, R.H., Blume, H.W., Schomer, D.L. and Sotrel, A. (1994) Neocortical dendritic pathology in human partial epilepsy: a quantitative Golgi study. Epilepsia, 35(4): 728-736. Murray, B., Alessandrini, A., Cole, A.J., Yee, A.G. and Furshpan, E.J. (1998) Inhibition of the p44/42 MAP kinase pathway protects hippocampal neurons in a cell-culture model of seizure activity. Proc. Natl. Acad. Sci. USA, 95(20): 11975-11980. Murray, B., Beer, T. and Cole, A.J. (1999) Pre- and postsynaptic activation of p44/42 MAP kinase in hippocampus after kainate-induced seizures. Epilepsia, 40:2 I. Parent, J.M. and Lowenstein, D.H. (1994) Treatment of refrac-
23
tory generalized status epilepticus with continuous infusion of midazolam. Neurology, 44(10): 1837-1840. Patrylo, P.R., Schweitzer, J.S. and Dudek, EE. (1999) Abnormal responses to perforant path stimulation in the dentate gyms of slices from rats with kainate-induced epilepsy and mossy fiber reorganization. Epilepsy Res., 36(1): 31-42. Penner, M.R., Pinaud, R. and Robertson, H.A. (2001) Rapid kindling of the hippocampus protects against neural damage resulting from status epilepticus. NeuroReport, 12(3): 453457. Saffen, D.W., Cole, A.J., Worley, P.E, Christy, B.A., Ryder, K. and Baraban, J.M. (1988) Convulsant-induced increase in transcription factor messenger RNAs in rat brain. Proc. Natl. Acad. Sci. USA, 85: 7795-7799. Sandberg, R., Yasuda, R., Pankratz, D.G., Carter, T.A., Del Rio, J.A., Wodicka, L., Mayford, M., Lockhart, D.J. and Barlow, C. (2000) Regional and strain-specific gene expression mapping in the adult mouse brain. Proc. Natl. Acad. Sci. USA, 97(20): 11038-11043. Sasahira, M., Lowry, T., Simon, R.P. and Greenberg, D.A. (1995) Epileptic tolerance: Prior seizures protect against seizureinduced neuronal injury. Neurosci. Lett., 185: 95-98. Schafe, G.E., Atkins, C.M., Swank, M.W., Bauer, E.P., Sweatt, J.D. and LeDoux, J.E. (2000) Activation of ERK/MAP kinase in the amygdala is required for memory consolidation of pavlovian fear conditioning. J. Neurosci., 20(21): 8177-8187. Schauwecker, P.E. (2000) Seizure-induced neuronal death is associated with induction of c-Jun N-terminal kinase and is dependent on genetic background. Brain Res., 884(1-2): 116128. Schauwecker, P.E. and Steward, O. (1997) Genetic determinants of susceptibility to excitotoxic cell death: implications for gene targeting approaches. Proc. Natl. Acad. Sci. USA, 94(8): 4103-4108. Schmid, R., Tandon, P., Stafstrom, C.E. and Holmes, G.L. (1999) Effects of neonatal seizures on subsequent seizure-induced brain injury. Neurology, 53(8): 1754-1761.
Schwob, J.E., Fuller, T., Price, J.L. and Olney, J.W. (1980) Widespread patterns of neuronal damage following systemic or intracerebral injections of kainic acid: a histological study. Neuroscience, 5: 991=1014. Selcher, J.C., Atkins, C.M., Trzaskos, J.M., Paylor, R. and Sweatt, J.D. (1999) A necessity for MAP kinase activation in mammalian spatial learning. Learn. Memory, 6(5): 478490. Sheline, Y.I., Sanghavi, M., Mintun, M.A. and Gado, M.H. (1999) Depression duration but not age predicts hippocampal volume loss in medically healthy women with recurrent major depression. J. Neurosci., 19(12): 5034-5043. Sloviter, R.S. (1994) On the relationship between neuropathology and pathophysiology in the epileptic hippocampus of humans and experimental animals [Review]. Hippocampus, 4(3): 250253. Sutula, T.E, He, X.X., Cavazos, J. and Scott, G. (1988) Synaptic reorganization in the hippocampus induced by abnormal functional activity. Science, 239(4844): 1147-1150. Vezzani, A., Sperk, G. and Colmers, W.E (1999) Neuropeptide Y: emerging evidence for a functional role in seizure modulation. Trends Neurosci., 22(1): 25-30. Wasterlain, C.G. and Plum, E (1973) Vulnerability of developing rat brain to electroconvulsive seizures. Arch. Neurol., 29(1): 38-45. Weiss, S., Cataltepe, O. and Cole, A.J. (1995) NGF reduced DNA fragmentation in CA1 neurons but not elsewhere after systemic kainate-induced seizures in the rat. Soc. Neurosci. Abstr., 21: 985. Willmore, L.J., Ballinger Jr., W.E., Boggs, W., Sypert, G.W. and Rubin, J.J. (1980) Dendritic alterations in rat isocortex within an iron-induced chronic epileptic focus. Neurosurgery, 7(2): 142-146. Wuarin, J.-E and Dudek, EE. (1996) Electrographic seizures and new recurrent excitatory circuits in the dentate gyms of hippocampal slices from kainate-treated epileptic rats. J. Neurosci., 16(14): 4438-4448.
T. Sutula and A. Pitk~nen (Eds.)
Progress in Brain Research, Vol. 135 © 2002 Elsevier Science B.V. All rights reserved
CHAPTER 3
Doubt and certainty in counting R.W. Guillery 1,, and B.K. August 2,** 1Department of Anatomy and 2 Department of Neurology, School of Medicine, University of Wisconsin, Madison, W1 53706, USA
Abstract: Some of the methods used for counting objects in histological sections are discussed. The method best suited for any particular counting program depends on many variables, which include the level of accuracy required, the type of preparation available for study, the size of the objects to be counted, the thickness of the sections that can be used, the equipment available and the amount of labor that can reasonably be invested. For light and electron microscopy, profile counts are simple and quick for objects that are small relative to section thickness and whose dimensions are readily defined. The 'physical disector' is particularly useful where objects to be counted are large relative to sections thickness, or where their dimensions are unknown or highly variable. For light microscopy, the optical disector is often easier to use. However, it makes more assumptions than the physical disector; some of these can introduce serious bias in the counts, and they are explored. Electron microscopy raises some special problems that relate to the depth of focus, the relatively very thin sections, and the tendency for thin structures that do not span the full thickness of a section to be lost or unrecognizable in some section planes. The importance of recognizing the assumptions that underlie any method of counting and its interpretation is stressed.
However, if one remains aware that these methods yield approximations, one or another of them will prove useful. The choice of method depends on the actual conditions and is, moreover, largely a matter of taste. E.R. Weibel (1969) writing on stereological principles. Introduction The importance of reliable counting methods in neuroscience has increased over the past decade, and has received significant discussion in a literature that is surprisingly extensive. (Recent references include:
*Correspondence to: R.W. Guillery, Department of Anatomy, School of Medicine, University of Wisconsin, 1300 University Avenue, Madison, WI 53706, USA. Tel.: +1-608-263-4763; Fax: +1-608-262-7036; E-mail: rguiller @facstaff.wisc.edu ** Responsible for the electron microscopic studies.
Gundersen et al., 1988; Pakkenberg and Gundersen, 1995; Coggeshall and Lekan, 1996; Guillery and Herrup, 1997; West, 1999; Geuna, 2000; Benes and Lange, 2001). Counts of nerve cells, glial cells, vesicles, or synapses in sections prepared for light or electron microscopy are being undertaken more commonly than in the past. They can help to define the progress of disease, allow comparisons between experimental and normal conditions, or show changes that characterize development. If the counting is based on reliable methods, it can provide crucial information. If not, it can be seriously misleading. However, no method of counting can be certified as free of error. The investigator who needs to count is best placed if the need for counts is established before the tissue is prepared for sectioning. Matching tissue preparation to the objects to be counted and to the degree of accuracy required by the counts should be a first step wherever possible. Often, however, counts need to be done on tissue that has already been prepared, perhaps many years ago, or that needs to serve more
26 than one experimental purpose. For such material one needs to devise the best counting method available, establish possible sources of error, and decide whether the likely size of the error of the final counts will justify the necessary labor and lead to sufficiently trustworthy conclusions. Although neuroscience has been greatly advanced by studies that included counts from sections, it should be recognized that there are situations where, on the basis of the material available and the likely size of the error of a count, it may be prudent not to count. An erroneous count may be more of a hindrance to advancing knowledge than no count at all; authors, reviewers and editors should strive to make the description of methods relevant to a counting study as complete and translucent as possible, so that readers can be helped to evaluate the reliability of a count. Where two comparable quantitative studies produce conflicting results, it is often difficult to understand the basis of the difference. For example, Haug (1986) reports that the number of nerve cells reported in human cerebral cortex has varied from 0.6 x 109 to 16 x 109, and Pakkenberg and Gundersen (1995) report 20-25 x 109 (see also footnote 2). Although different methods of counting have been clearly presented, and sources of errors as well as possible corrections of errors have been discussed (e.g. Floderus, 1944; Abercrombie, 1946; Konigsmark, 1970; Pakkenberg and Gundersen, 1989; Clarke, 1992, 1993; Coggeshall and Lekan, 1996; Guillery and Herrup, 1997; Gundersen et al., 1999; West, 1999; Benes and Lange, 2001), serious differences concerning methods can be found in the literature, and discrepancies in the results obtained are still common. There is an increasingly widely held, but overly simple faith that only one method of counting, using the 'disector', often described as 'un-biased' or 'assumption-free' (see: Coggeshall and Lekan, 1996; Mayhew and Gundersen, 1996; West, 1999), can produce valid results. Although this is a powerful and useful method, the view that only counts that use the disector should be accepted for use in publications and grant-supported research is too sweeping. So also is the claim that these methods are free of assumptions of the type that can, and often do, lead other methods into systematic errors. For almost any counting method that has been used, there have been many publications in which the method of count-
ing is either quite evidently flawed, or is presented without the detail that is needed to evaluate the significance of the final count. In this brief review, we stress that there is no one 'correct' way to undertake a count, and we indicate that there are innumerable, often unforeseen errors that one can make, no matter which method of counting is used. Any student interested in the subject can find examples of more or less egregious errors in the literature (in fact, encouraging students to find such errors is a good way to introduce them to the problems of counting). Since almost all of us who have made counts have committed errors of omission or commission, we cite only few examples of the sorts of things that can go wrong. No matter what method of counting is used, the range of possible sources of errors is great, far beyond the scope of this review. Planning a count
The first step in evaluating a counting method is to decide what sort of information is being sought. This is a first step for the investigator who has to provide a clear statement; it is also an essential first step for those who give advice on counting and for those who have to evaluate studies that include counts. For some studies one simply needs a ballpark figure to indicate how many cells or axons or synapses characterize a particular structure or brain. More commonly, one is looking to make comparisons, between nerve cells and glia, nerve ceils and synapses, between structure A and structure B, or between different experimental conditions, disease states, or developmental stages or species. There is, as we shall see, an important difference between wanting to know the number of objects within a particular nucleus or cell layer, and wanting to know the number of objects per unit volume of a brain part. Perhaps an important opening statement for the methods section of a quantitative study should be one that clearly describes the nature of the counts to be made, and defines the margins of error that are considered acceptable. This not only alerts the reader to the relevant variables of the study, but will also protect the author from being attacked for not achieving a level of accuracy greater than that actually needed in the study. A second step should perhaps be recognition of possible errors, their likely
27 size, and a summary of the steps that have been taken to reduce or eliminate such errors. Possibly the most important source of error is the bias introduced by the observer who is making a comparison between two populations. Unless the observer is blind to the conditions of the study, there will be a possibility of a serious bias in the results. Labeling a counting method as 'unbiased' or 'assumption-free' may at times have led investigators into a false sense of security about the unbiased nature of their results. Whenever a count reports a comparison, the methods section should indicate whether the observer was blind to the conditions being compared, and if the methods section does not include such a statement, then one is well advised to treat the results as unreliable. It is tempting to think that some defined formula could be devised for the method to be used in any quantitative study, but since each tissue, each preparatory method, and each type of count raises its own particular array of problems, the development of such a 'formulaic' approach is to be discouraged. There is already too great a tendency for investigators to name their method, cite a general account of the method, and leave out specific and important details that, as we shall show, can make a significant difference to the reliability of a count. In general, any study that does not look at the possibility of errors in the counts deserves particularly close scrutiny, especially if it claims to be free of 'assumptions'.
Defining the 'acceptable' error Preliminary observations can often give a clue as to the size of the error that is acceptable. For a doctoral dissertation one of the authors (R.W.G.) counted the cells in the mamillary nuclei and found about 80,000 cells in the medial mamillary nucleus of the cat, and only about 3000 cells in the lateral mamillary nucleus. The difference tells something significant about the functions of the two parts of the system, and the general conclusion about a large difference between the two nuclei could have been firmly established even by counts that had a large error. However, for comparisons between species, and for ratios of cells relative to axons, methods that were sensitive to smaller differences were needed. Comparably, when electron microscopic images of the visual relay in
the thalamus (the lateral geniculate nucleus) first became available, one of the striking features was that a relatively small proportion of the total synaptic junctions were established by retinal axons. The predominance of non-retinal afferents was demonstrable by counts (Guillery, 1969). These cannot be regarded as having provided an accurate quantitative survey, but they did stress what was at the time a surprising conclusion about the preponderance of non-retinal afferents. More recent counts have confirmed this general conclusion, but have provided rather more accurate figures (Eri~ir et al., 1998). Counts of ratios between (e.g.) Purkinje cells and granule cells in the cerebellum can similarly show that the latter greatly outnumber the former, and a large error would not alter the conclusion. However, if a theoretician needed more precise ratios, or if one wanted to compare the ratios between species, then the smaller size of the acceptable error must be taken into consideration when the counts are planned. The smaller the acceptable error, the more important is the method, and the greater the necessary investment in the study. It has been pointed out that errors differ in their nature and can differ in the effects that they produce. For many purposes it is important to distinguish between errors that are random, and thus affect the variance of the mean obtained, and errors that are systematic (or 'biased'), which shift the mean up or down (West, 1999). The distinction is important when one is planning a count, but it should be noted that distinguishing the effects of each type of error in any particular published count can be extremely difficult, and often impossible. The first type of error, provided that an appropriate method of sampling has been used, produces a set of counts that are randomly scattered around the actual value, and the extent of this scatter, the variance, can be estimated, and recorded in terms of standard errors. Up to a point, the more counts that one has, the smaller this error is likely to be, although the labor required for accurate counts often leads to relatively small sample sizes. It should be stressed that the estimate of the variance will depend on the distribution of the population of objects being counted, on the nature of the population of organisms being studied, and on the sampling procedure used. In some tissues the problem of variation from
28 one section to another or from one part of a section to another is relatively small. Gundersen and Osterby (1981) have made the point that this can be of 'negligible importance' relative to the 'biological variation' between individuals. However, the nervous system is a highly structured (i.e. non-random) tissue, where objects that are to be counted are likely to be heterogeneously distributed. Here a limited random sampling procedure of the sort that is often used can give a false estimate of the size and of the variability of the population (see Benes and Lange, 2001). Consider a population of cells that is highly concentrated in small clumps or layers in a tissue. If 90% of the cells occupy only about 5% of the volume of the tissue, then there is significant probability that a sample of 20 counts will not include a region of the high cell concentration. That is, these 20 counts could provide not only a significant undercount, but would also provide an entirely false picture of the variability, and so provide a seriously misleading clue for evaluating the data. Investigators should have a clear view of how the population of objects that is to be counted is distributed, should ensure that their sampling method provides a representative sample of all parts of this population, and should also indicate how the sampling procedure used relates to the known distribution of the objects being counted. The problem of variance applies not only to the objects being counted but also to the population of individuals from whom the tissue for counting has been obtained. A point that may be particularly important for the subject of this conference concerns the measure of variance that one can expect to find in any population of individuals likely to be used. Whereas one can expect a highly inbred strain of mouse to produce a relatively low variance for many counts (Williams et al., 1996), especially when age and sex are taken into account 1, a human population is likely to show very much more variation 2. This will not only produce high standard errors for the relatively small samples that often have to be used, Note, however,that where two investigatorsuse different inbred strains the chances of obtaining discordant results are increased. 2 Pakkenberg and Gundersen (1997) show a range from <15 x 10 9 to >30 x 10 9 for numbers of neocortical neurons in human brains.
but it may also mean that subjects used for counts in one laboratory may differ in real terms (general health, diet, environment etc.) from subjects used in another. The same is likely to hold for animals obtained from the wild or for animals that, though bred for laboratory research, have not been inbred, such as the cats or monkeys most commonly used in CNS research. Again, one should be on guard for small sample sizes, which can provide ballpark figures, but which may be seriously misleading not only about the population mean, but also about the variance. Heavy investments made to obtain highly accurate individual counts may prove counter productive where this leads to records from relatively few individuals in a highly variable population. Experimental studies of ageing that involve more than a few years face a particularly difficult problem because housing conditions for the experimental subjects change as governments adopt new guidelines, and as the training received by animal caretakers changes. For human studies the occurrence of major wars, famines, or changing diets may be more relevant than age in itself. The second major type of error is a systematic error that produces undercounts or overcounts. The most common example occurs where objects of unknown size are counted in sections. Since many of the objects are cut and appear in more than one section, an uncorrected count will provide an overcount, and the larger the objects, the greater the overcount. Where two conditions are being compared, it is possible to record a significant difference in cell number where, in reality, there is a difference in cell sizes, or cell shapes, with no difference in the numbers. That is, there is a systematic error (or bias) because there has been a confusion of parameters. Larger objects produce spuriously higher counts. In this example, the confusion is between cell size (or shape) and cell number; in following sections we illustrate other confusions of parameters that can also lead to comparable systematic errors, or biases, and that are less widely recognized. In practice, many of the problems and errors that one encounters in light microscopy and in electron microscopy are different, and for this reason, we consider the two separately in what follows.
29
Light microscopical studies Profile c o u n t s
T h e m o s t c o m m o n l y d i s c u s s e d s y s t e m a t i c error in counts that use light m i c r o s c o p i c a l study o f tissue slices is the one i n t r o d u c e d above, p r o d u c e d by double c o u n t i n g o f objects in sections, w h e r e s o m e o f the objects are cut so that they a p p e a r in two or m o r e sections (Fig. 1). I f all o f the profiles in the s a m p l e d parts o f the sections are c o u n t e d (profile counts), then the severity o f this bias, and the extent to w h i c h it can be r e c o g n i z e d and c o r r e c t e d d e p e n d s on h o w m u c h the o b s e r v e r k n o w s about the shape and the size o f the objects and the thickness o f the sections. T h e less one k n o w s about the shape and the
Fig. 1. Schema to represent the use of the Abercrombie correction for a profile count. (A) Twenty-five cells (or nuclei or nucleoli or mitochondria etc.) are distributed through a tissue. One of a series of sections is represented by the tissue between the two solid lines. This includes some of the cells, cuts some and excludes others. (B) h, the mean dimension of the 25 cells in the z-axis is equal to H/25. Estimates of h and T are based on measurements that represent dimensions in the z-plane. If it can be shown that on average the dimensions of the cells are equivalent in all directions, then h can be obtained from sections cut in any plane, but if this cannot be shown, then h should be measured in the z-plane itself, either directly, using measurement of movements of the microscope stage along the z-axis (see section on measurements along the z-axis), or by cutting sections perpendicular to the original section plane, using either equivalent tissues to those used for counting or, where the method allows this, re-cutting the same sections perpendicular to the original section plane. (C) When the section is viewed through the microscope from above, a profile count will, on average, represent cells having a unique point (top, bottom or some other, theoretical, unique central point) in a thickness of the original tissue equivalent to T + h. That is, the profile count over-estimates the cell number by a factor of T + h / T . (D) If, for the measurement of h and T sections are re-cut perpendicular to the original section plane, then, if there is further shrinkage during this second process of cutting (which may require re-embedding), then, provided that the shrinkage acts equally on all parts of the section, for calculating the Abercrombie correction factor, no further correction is required, since the ratio T / ( T + h) = s T / ( s T + sh), where s represents the degree of shrinkage. However, the point about equivalent shrinkage of all parts of the tissue should be treated as an assumption (see text). It is useful to compare measurements made on the original sections (by calibrated vertical movements of the microscope stage) with measurements made directly along the same axis on the re-cut sections.
size o f the objects b e i n g counted, the stronger the a r g u m e n t for using an alternative, such as the 'disector' m e t h o d (Pakkenberg and G u n d e r s e n , 1989, 1995; C o g g e s h a l l and Lekan, 1996; West, 1999), w h i c h is i n d e p e n d e n t o f o b j e c t shape and size. It has b e e n argued that the d i s e c t o r t e c h n i q u e should always be used, and that profile counts are intrinsic a l l y unreliable, and thus should be avoided. S i n c e profile counts are relatively s i m p l e to do and can be carried out fairly rapidly on m a n y different types o f section, it is worth l o o k i n g at the d e g r e e to w h i c h they are unreliable, and defining c o n d i t i o n s u n d e r w h i c h their use should be e n c o u r a g e d . A point that is o f p r i m a r y i m p o r t a n c e in e v a l u a t i n g counts is the tension that a l w a y s exists b e t w e e n , on the one hand, using a labor-intensive, h i g h l y accurate m e t h o d such as the p h y s i c a l disector, w h i c h often leads to small s a m p l e sizes, or, on the o t h e r hand, using a m o r e
2
3
4
5
I • •
6
7
9
12
16
•
19
b
25
3
7
4
13
j h =H/25
9
10
12
•
• •
11
• • •
t
22
24
•
B I
21
2
•
, "
17
zu 23
C
I0
13
]4
t6
17
1
"I" /
• /
t5
18 i
i
1
lh/S •
6
7
8
13
D
• •
12 11
~"
.TIS
30 rapid method that allows for larger samples. The more that is known about the sources of variability discussed above, the easier it will be to decide on the most appropriate approach. Fig. 1 shows that the size of the error in a profile count depends on the thickness of the sections and on the mean dimension of the counted objects in a plane perpendicular to the section (the z-plane). That is, the smaller the objects being counted and the thicker the section, the smaller the error. A suitable correction for this error was published by Abercrombie 55 years ago and can be expressed as N = Nl ( T / T + h), where N is the actual number of objects that should be recorded, N1 is the recorded count, T is the thickness of the sections, and h is the mean height of the objects in the z-plane (Abercrombie, 1946). As h becomes larger relative to T, the error becomes larger, and is quite unacceptably large if h is greater than T. Clarke (1992) has suggested that profile counts combined with the Abercrombie corrections should not be used where T exceeds h by a ratio of 1.5, which is good general rule, although one is wise to work well on the safe side of this limit. Where h is small relative to T, the error will be correspondingly small. That is, given
that one has reliable information about h and T and h / T is small, profile counts with the Abercrombie correction, or with an appropriate variant of that correction (e.g. Floderus, 1944; Konigsmark, 1970) can produce reliable results. Arguments against the use of such counts are often based on situations where the objects to be counted are large relative to section thickness or of unknown shape (Coggeshall and Lekan, 1996; West, 1999). Where the objects are small and of well-defined shape, those arguments do not apply. Furthermore, since the method is simple and relatively straight forward, and since it can be used on material where adjacent sections are not stained by the same method (which is important if the physical disector method described below is to be used) it is a mistake to discourage its proper use. An example, where a profile count can be useful, is provided by nucleolar counts (which can also be used for cell counts in tissues where cells are known to have one and not more than one nucleolus). In sections that are 10-30 ~tm in thickness and where the nucleoli are roughly spherical and 1-3 Ixm in diameter, as they are in many neural structures, pro-
file counts provide a simple and rapid method. For the dimensions given above, an uncorrected profile count will produce an error of about 3% for a population of small nucleoli in 30-1xm sections and of about 23% for large nucleoli in 10-txm sections, so that for small nucleoli and 30-1xm sections, even with no correction, the counts can produce useful results for many purposes. With larger nucleoli or thinner sections, the correction becomes important, but once the correction is made, provided that it is based on reliable information about the dimensions h and T, one can expect the estimate to be relatively close to the actual number in that sample. It should be noted that there have been several versions of the Abercrombie correction. Some (e.g. Floderus, 1944; Konigsmark, 1970) make allowance for small pieces of the counted structures that are lost in the cutting (so-called lost caps), or that cannot be recognized because they are too thin at the surface of a section. Others (e.g. Coupland, 1968; Hendry, 1976; Hedreen, 1998) address the problem posed by objects whose size represents a significant fraction of section thickness, causing measurements of h to be biased in favor of smaller objects that are completely included in the sections. These corrections are refinements that can provide a somewhat closer approach to the 'real' number but, as can be seen from the above example of nucleolar counts, such corrections are a case of gilding the lily for small objects in relatively thick sections. Where the ratio of object size to section thickness is relatively high, these additional corrections may help, but the correction for the lost caps can then also be of slightly dubious value, since often the assessment of just what allowance to make for lost or unrecognizable 'caps' can be arbitrary. Nuclear counts in thin sections may give a rather high h / T ratio, but can still be used in relatively thick sections provided there is good evidence about the value of h. Ideally h and T should be measured in some of the sections of the series that is used for the counts, and this can be done after some of the sections have been cut perpendicular to the original section plane (see Marengo, 1944, and Fig. 1). That is the measurements should be made in the z-plane. However, alternate sections, sections from the other hemisphere or possibly even from other, closely matched brains can also be used; alternatively, the measurements can be made on the sections used
31 for the counts by measuring stage movements of the microscope in the z-direction, beating in mind, however, the problems involved in making such measurements, which are considered in a separate section below. It may be of some interest for those who have done or plan to do profile counts to note that the Abercrombie correction is wrongly presented in two recent articles that argue strongly against profile counts in general and in support of the disector methods in particular (Coggeshall and Lekan, 1996; West, 1999). The correction factors given there would not lead one to recognize the importance of the ratio of h to T for profile counts, and could lead to serious error if used to correct a profile count. For anyone planning a profile count, the best strategy is to compute the correction factor on the basis of a figure such as Fig. 1C, which shows clearly that a profile count is actually counting objects representative of numbers in a volume equivalent to a section having a thickness of h + T. Alternatively reference can be made to Abercrombie's original paper, which also includes some interesting further thoughts on counting and measuring the objects to be counted. One of the suggestions made in that study is that the correction can be avoided if one cuts adjacent sections at different thickness (TI and T2, where TI > T2). The difference between the profile counts (Nl minus N2) made in the two sets of sections will then give the correct number for the number of objects in a (notional) section having thickness T1 minus T2. Where celloidin or frozen sections are cut on a sliding microtome, this method can be quite practical, provided that actual section thicknesses are measured. 3 Shrinkage of the tissue, either before sectioning, or on the slide after sectioning, can be considerable. The issue may become important if one is comparing two tissues that do not undergo the same amount of shrinkage (Uylings et al., 1986; Haug, 1987), and if one is recording cell densities rather than total
3 For paraffin sections, two knives mounted precisely parallel to each other, one above the other and one slightly closer to the block face than the other can also providethe needed two series, but collecting two paraffin ribbons concurrently takes quite some skill, even if one managesto arrange for the knives to be properly mounted, which is, of course, critical.
numbers. This would lead to a different confusion of parameters, where a greater amount of tissue shrinkage becomes interpreted as a higher cell density. The issue can also be important if one measures volume before shrinkage and cell packing densities after shrinkage. The use of the 'disector'
This method of counting has been fully described in several publications (Gundersen et al., 1988, 1999; Coggeshall and Lekan, 1996; West, 1999, 2002, this volume) and was first brought to the attention of contemporary investigators by a paper published almost two decades ago (Sterio, 1984). Benes and Lange (2001) have pointed out that an earlier use of the same principle was described in 1895 at the University of Wisconsin for counting the glomeruli in the kidney of a cat (Miller and Carlton, 1895). The method is based on a comparison of two adjacent sections, recording the number of profiles of objects seen in one section, the 'sampling section', and then only counting those objects that do not also appear on the adjacent 'look-up' section. This is a simple method for avoiding double counts, and its particular strength is that it is completely independent of the size and shape of the objects being counted. That is, with this method, the two parameters, size and number, cannot be confused. If the two sections are physically separate sections (the 'physical disector') then the process is difficult because the two sections need to be precisely matched so that objects on one section can be matched against objects on the next section. Furthermore, the method often cannot be used on archival material where complete series may not be available, either because different stains were used for adjacent sections, or sections were not mounted as complete series. For these reasons the 'optical disector' has considerable appeal, since it uses two or more optical sections within a single histological section to achieve the same result. In principle, this method can be used rather like the physical disector, recording number of profiles at one plane of focus, and then only counting those that are not still present at another plane. Within a reasonably thick section, several pairs of optical sections can be used, with the look-up section of one pair serving as a sample
32 section for another (West, 1999)4. In essence, it is necessary to ensure that within any volume of tissue in the single physical section, only one point on any object to be counted is recorded; a simple choice is to record the top of any object that first comes into focus as the section is raised towards the objective (West, 1999; Geuna, 2000). This then resolves the method into a simple procedure of recording the number of such 'tops' within an optically defined thickness within the thicker histological section. The volume used for such counts should avoid the edges of the sections, where there may be unevenness and where there is likely to be a 'lost caps' problem. The optical disector has justifiably gained considerably in popularity in recent years. Its use is to be encouraged since it is relatively simple and so can produce larger sample sizes than the physical disector. Furthermore, it can serve where the objects to be counted are too large or too irregular to be counted by profile counts. However, it is important that two issues be clearly appreciated by those using this method. One is the importance of the basic rules that govern the production of optical images with a light microscope, and a second is the problems of tissue shrinkage associated with the production of histological sections. Failure to recognize the importance of these can introduce serious bias into a count. The most important point about the optical disector concerns measurements in the z-axis. Whereas the physical disector avoids many (but not all) of the problems related to measurements along this axis, these problems can play a major role in the production of systematic errors when the optical disector is used.
Measurements in the z-axis We have seen that measurements of section thickness are critical for profile counts. They can sometimes be ignored where a physical disector is used to sample
4 If this method is used then it is important to ensure that the distance between the two optical sections is smaller than the smallest object to be counted, since, of course, such small objects would be missed as they fell between two chosen optical sections. The problem does not arise if one simply counts tops as the section is raised from one focal plane to the other.
from a known proportion of sections through a well defined volume of tissue (Gundersen et al., 1988). That is, if the sections are all of essentially the same thickness, and if each section then represents a known fraction of the total tissue dimension in the z-plane, then a count in a sample of each section will represent a proportion of the total tissue volume that can be defined without recording section thickness. However, often, for comparisons of densities of objects in a tissue, which are likely to be important where boundaries are difficult to define (see below), section thicknesses need to be defined. For counts that use the optical disector, it is important to define the thickness of the optical section from which any one count is obtained, and this is crucial and can readily lead to undercounts or overcounts if the measure is wrongly determined. In a recent review of quantitative methods Geuna (2000), discussing the optical disector, argues that the top of an object: "is a point (and thus adimensional), it has no size, shape or orientation (therefore no assumptions are needed for these parameters) and can be sampled in only one disector volume (i.e. it cannot be split into two disectors)". This represents a serious misunderstanding of optical sections and one that, to judge from published accounts, is shared by several other users of the optical disector. Although the top of the object is a dimensionless point, its microscopic image is not. The image has three dimensions, and these dimensions depend upon the optical conditions of the study. The depth of the field when using widefield optics 5 and actual measurements made along the z-axis depend on the numerical aperture of the lens, and on the optical properties of the specimen. With good optical conditions (see below) and a good oil immersion lens having a high numerical aperture (1.35 or 1.4), one can expect the accuracy to be of the order of 0.5 txm, or even less6. This assumes that
The term refers to light microscopic methods that are distinct from confocal microscopy and other methods using a restricted beam of light. 6 Under ideal conditions, it can be as small as 0.3 txm (Williams and Rakic, 1988), but conditions, as argued below, are rarely ideal. It is best to check empirically. Lange and Edstrrm (1954) discussed problems of measurements in z-axis in detail and recorded a 10% error for measurements of 5 ltm along the
33 a suitably thin cover glass has been used and that the refractive index of the mounting medium and the tissue is very close to that of the glass and of the immersion oil. However, these ideal conditions are rarely, if ever, met, and often the details provided are not sufficient for judging the extent to which they have been met. Mounting media that have been used in the preparation of histological sections vary greatly in their refractive index. Lillie (1965) shows a range from 1.413 to 1.8225 7, but it is rare, in publications that use measurements along the z-axis, to find any information about refractive index, either of the mounting medium, or, where this is relevant, of the embedding compound used. Concerns about the refractive index of the mounting medium are particularly important where sections have not been cleared and are viewed in an aqueous medium. Each change of refractive index encountered by the light on its way through the specimen, the mounting medium and the cover glass to the objective lens will produce a shift from the ideal condition and a consequent change o f the distance along the z-axis over which the object appears to be in focus. The effect on the apparent size of an object in the z-axis produced by changes in refractive index can be surprisingly large (amounting to about 65% for objects viewed without immersion oil; see Glaser, 1982; West and Slomianka, 1998). The refractive index of the tissue itself could play a significant role in the formation of the image, but is usually not considered since one assumes that dehydration and clearing replace water and lipids from z-axis. West et al. (1996) using a high numerical aperture, oil immersion lens give an estimate of 'depth of focus' as 1-2 Ixm, which is probably the sort of figure one can generally expect. Uylings et al. (1986), using x63 and xl00 lenses (NA not given) report inter-individual variation with a range of 3 txm for measuring a section that was 5 Itm in thickness. We have used a Nikon 'microcator' (Gundersen et al., 1988) to record vertical movements of the microscope stage and have measured the thickness of number 1 cover glasses that had ink marks on both surfaces, and were mounted in Eukitt mounting medium (RI = 1.51: Calibrated Instruments Inc., Hawthorne, NY 10502). With a ×40 lens (NA = 1.0) we recorded measurements ranging from 146 to 159 txm (mean = 153.0, SE = 0.51, n = 40) for one cover glass. With a x 100 lens (NA 1.25) the corresponding values for the same cover glass were 147-158 tim (mean = 152.7, SE = 0.54, n = 20). 7 Immersion oil has a refractive index close to 1.51.
the tissue so that it takes on the refractive index of the mounting medium. However, if an unstained section of the tissue can be imaged by phase contrast or interference contrast optics (either of these providing a suitable test) then the tissue itself is acting to refract the light passing through it, and this is likely to be the case for any tissue of interest to neuroscientists. That is, some parts of the section, most probably proteins that have not been replaced by the dehydration and clearing procedures, are producing variations in the refractive index of the cleared tissue. The question for measures along the z-axis, is whether the overall effect of these variations in the refractive index affect the measurements. Preliminary observations of lines drawn on a microscope slide and observed through a 60-1xm tissue section, suggest that the sections do not seriously affect measures along the z-axis, except in so far as the resolution of the image is poorer when viewed through the section. That is, the refractive properties of the CNS tissues we have used (cerebral cortex, cerebellar cortex) produced some slight loss in the accuracy of the measures, but not a systematic shift in the mean of the measures that were recorded. Where measures along the z-axis are critical for a particular set of observations, it would be useful to have commercially available slides that are marked with fine lines. These could serve two purposes if appropriately designed. If the lines were spaced a known distance apart along the z-axis ( 5 - 1 0 Ixm apart between layers of glass, or of plastic having the same refractive index as the glass) they could serve to calibrate the equipment used for measurements along this axis. Furthermore, sample sections could be mounted over such lines so that any changes in resolution and focal depth produced by the presence of the section could be recorded and reported as a part of the study 8. A simple method of calibration that may be more practical is to use c o m m e r c i a l l y available microspheres, and check that mean and standard errors for measures in the x - and y-axes match the measures in the z-axis. A minor, but interesting, variable relevant to measurements in the z-axis is represented by the optics s However, measurements of fine lines present a best case scenario. As noted below, the nature of the object being observed is relevant to the accuracy with which its position and dimensions in the z-axis can be defined.
34 of the observer's eye (see e.g. Haug, 1956; Uylings et al., 1986). They are best eliminated by using an image on a screen or a photographic image. Ideally, perhaps, if one is not using a screen or a photograph, one should always use very old observers whose lenses have hardened (it should be regarded as unethical to paralyze the accommodative mechanisms of young observers). Mechanical problems are also relevant to measurements in the z-axis. The accuracy of the fine markings on the focal adjustments of microscopes varies greatly, and some instruments may give rather inaccurate measures. Devices for recording the actual vertical movements of the microscope stage ('microcators': Gundersen et al., 1988) have been described and these provide more reliable measures of stage movements, although reports that use such devices would be well served by brief mention of their calibration before use. 9 Accurately defining the focal plane in which the top of an object lies, depends not only on the optics, but also on the nature of the object. Theoretical considerations generally apply to very small objects having high contrast, i.e., a point source. In practice, we do not have many point sources in our sections, and generally that is not what we count. For very small, well-defined objects that have a high contrast, finding exactly where the top is in focus is relatively straight forward, whereas a larger opaque object, a pale one, or a slightly granular one may prove much more difficult. If one is studying a Nisslstained section, then defining the top of the cell can be extremely difficult, since one is assessing where the very small Nissl granules, scattered immediately below the surface of the cell, first come into focus. Defining the top of the nucleus may prove easier, because the nuclear membrane is generally (but not invariably) more clearly defined, and since it has a higher curvature, the 'top' is nearer to Geuna's idealized point. If the control and the experimen-
9 We are uncertain about the degree to which the variance of the measures reported in footnote 5 was due to the mechanism used, to the optics of the preparation, to the observer's visual system, or to slight variations in the cover glass thickness. Although the means and standard errors we have recorded are relatively consistent, the range of these observations is somewhat disconcerting.
tal samples differ in the staining properties or the dispersal of the structures (such as the Nissl granules near the cell surface) that define the 'top' of the counted objects, then this could well influence judgments about best focus. In summary, measurements along the z-axis are subject to systematic errors that depend on the optics of the preparation and the mechanics of the recording equipment. Some errors, like the changes in refractive index, can lead to serious under counts. Others, like the uncertainties of defining the precise focal plane of an object, will increase the size of the random errors that characterize a result. Where relatively small distances along the z-axis are being recorded (e.g. 10 txm or less between the optical slices), the margin of error introduced by the optical and mechanical problems may well represent a significant percentage of the total z-dimension. There are some good reasons for having optical disectors closely spaced, but the likely errors summarized above suggest that it may be better to have a fairly wide spacing. Where comparisons are being made that depend on judgment calls as to whether an object is or is not in focus, it is absolutely essential that the observer be blind to the conditions under investigation, and that this be clearly stated in the description of the methods. The z-axis also presents difficulties for quantitative methods because of the shrinkage that occurs along this axis x0. Any tissue when it is dehydrated or hydrated during histological processing is likely to change in volume, and different methods of preparation introduce different amounts of change. The amount of the volume change in any part of a section depends on the nature of the tissue, on the nature of the treatment, and on forces that may be acting on the tissue to counteract those produced by the shrinkage itself. When a tissue block or a freefloating section is processed, one generally assumes that all parts of the tissue shrink equally, although anyone who has handled individual sections of cerebral or cerebellar cortex after they have been exposed to significant shrinkage, will know that the gray and
10Linear shrinkages to 70-80% are commonly found. Since this has to be cubed to calculate the volume shrinkage one is dealing with very significant volume changes.
35 the white matter do not shrink equally. In spite of this, measurements along the x- and the y-axes, before and after processing are generally considered to give a reasonable estimate of shrinkage, where this measure is needed. When frozen sections or vibratome sections are used, if these are fixed to the slide, then, provided that the sections adhere firmly to the slide, there can be no further shrinkage of the attached section surface in the x- and y-axes, and much of the shrinkage must occur along the z-axis. Consequently, the amount of this shrinkage can be relatively large. Furthermore, since one surface of the section is firmly adherent to the slide but the other is free, there is every reason to expect that the amount of the shrinkage will not be uniform along the z-axis. We know of no observations that have investigated this point, and would expect it to vary depending on the tissue, the method of adhesion, and the type of treatment. Where the optical disector is used, therefore, two caveats should be considered. One is that two optical slices taken one above the other from the same area of a single histological section and measured as equal in their dimensions along the z-axis may not represent equal thicknesses of the original unshrunk section, and the second, a consequence of the first, is that the precise position along the z-axis, of a single optical slice, may make a difference to a count if the tissue has not undergone even shrinkage along this axis. Some of the issues addressed above apply to all sections, no matter how they are viewed, but the problem of defining the focal plane is different when the confocal microscope is used. Here the refractive indices of tissue, mounting medium and specimen, and the optical properties of the lens are still important, as is tissue shrinkage, but one can expect to have a narrower extent of the z-plane in focus than with widefield methods (see footnote 4). Given that setting up optimal conditions for confocal microscopy is not simple and that many investigators rely on technical support that a reviewer or referee cannot evaluate, some calibration or empirical evidence regarding the depth of focus of a confocal image should be provided where images that are relatively closely spaced are used.
Other problems Although the problem of double counting cut objects with profile counts has received the most attention, there are many other sources of error, several of them being a confusion of parameters comparable to the confusion of cell size and cell number. Where the number of objects within a cell group (e.g. a nuclear group or a lamina) needs to be determined, defining the outline of the particular cell group accurately is critical. Many such outlines are very hard to define in terms of objective criteria, and may depend upon features that seem obvious to an experienced microscopist, but that are not easily rendered into a clear written summary. There are several important outcomes: one is that two observers working independently in different laboratories and at different times may well be using different criteria to identify boundaries. Another is that the features that serve to identify boundaries may well be affected by the variable under study, the disease state, the age, or the experimental condition. Architectonic studies often rely on subtle changes from one region to another in cell or background staining, in cell size, or cell distribution, and if these are features that differ in the two conditions being compared, then an objective quantitative comparison may prove very difficult to establish. Here the staining properties of cells or background represent parameters that influence judgments about borders, and so can lead to erroneous conclusions about size of a cell group or the number of the objects that they contain. Where comparisons between two disease states or experimental conditions are being studied, the criteria used for establishing borders should be clearly defined and the methods section should indicate that the observers who drew the borders were blind to the disease state or experimental condition. The issue is raised not because we have a suggestion for entirely avoiding the problems, but because there is a tendency in some published accounts to treat a so-called 'unbiased' method as guarantee for accuracy; it is not, and especially in comparative counts, it is important to identify likely sources of error no matter what particular counting method is being used. Sometimes one can change strategy and count objects that show a particular property rather than cells that are included in a nuclear group or lamina.
36
37 For instance, one may choose to count cells that are retrogradely labeled following a particular delivery of axonally transported marker, or to count cells having particular staining properties (GABA or G A D reactive, ACh containing, or immunoreactive to one of any number of epitopes). The obvious caveat here is to make sure that the distinction between a change in actual cell number and a change in the affinity of the particular marker or stain (or the concentrations used) is clearly recognized. Again, there is a possibility of confusing two parameters, here, staining properties and cell numbers. For example, the difference between saying that there is a loss of cells in the nucleus of Meynert, and saying that there is a change in the staining properties of the cells in the nucleus of Meynert is important unless one is willing to assert boldly and clearly (as some are) that a cell does not belong to the nucleus of Meynert unless it has particular staining properties, and then one leaves changes in total number of cells in that region of the brain deliberately unresolved. The important point is to be clear about what the results may mean in terms of cell number or cell staining. A further point about many methods that stain particular components of a cell is that they are often not all-or-none methods. With these methods, the strength of the reaction and, therefore, the identifiability of a positively stained cell, that is, the final decision about the threshold between a stained and an unstained cell, depends on the vagaries of the method and the judgement of the observer. For example, immunohistochemical methods often stain the surface of a section well, but stain the inner parts of a section weakly or not at all. If the border between these two regions is sharp, then one can choose to study the well-stained part and ignore the rest. However, the change in staining intensity is more likely to be along a gradient, and decisions about the parts of the section that are suitable for
counts will generally then be rather arbitrary. The extent to which the reaction product can be seen in the depth of a section may depend on the amount of the epitope available, on properties of the matrix that surrounds the positive cells, or on the concentrations of antibody used, and these, rather than the number of cells reported, may be the variable that differs between two populations being studied. This provides yet another example of a confusion of parameters.
Electron microscopical studies Counting objects in electron microscopic materials presents a number of problems not encountered in light microscopy and is, on the whole, more difficult. If one starts with tissue preparation, then problems of tissue shrinkage are generally comparable to problems encountered with light microscopical methods, with one important exception. Cutting thin sections can readily lead to a compaction of the tissue perpendicular to the knife edge. That is, measurements along this axis of a section may not be equivalent to measurements along the corresponding part of the block face or to measurements taken parallel to the knife edge. The point is readily checked, and it seems reasonable to assume that where compacting has occurred it is even across the whole block face. Measurements of section thickness, where they are needed, can be based on interference colors, or (better) on views of folds in the section, where the thickness of the section can be measured directly (Small, 1968). It is generally assumed that section thickness is even throughout any one section, since unevenesses are readily spotted on the basis of interference colors. Since sections are extremely thin (down to about 50 nm) variations in section thickness can be a significant fraction of the thickness itself, and the evenness of the sections should be reported. Occasionally, comparisons need to be made be-
Fig. 2. Six micrographs taken at six different tilt angles from the same region of the hippocampus of a rat. The axis of tilt is indicated by the short black line at the upper left of figure A and the angle of tilt is shown in the upper right of each micrograph. The junction labeled I is a symmetrical junction with a visible cleft in A at +47 °. At the other angles some of the vesicles that lie close to the presynaptic membrane of this junction are more clearly shown, but otherwise there is little to identify this as a synaptic junction in B-E Junction 2 is clearly shown in D, E and F, from 0° to -27 °, but at the other angles, particularly in A, at +47 °, it is poorly defined as a synaptic junction. Junction 3 is recognizable at all angles in this figure because the axis of tilt was roughly perpendicular to the plane of the junction.
38 tween measurements or counts in adjacent thin and semithin sections made by electron and light microscopy, respectively. The points discussed above about the refractive index of the embedding resin may then become important. If this differs from the refractive index of glass and immersion oil, then some corrections for measurements along the z-axis will be required for the light microscopical observations. Depth of focus in an electron micrograph exceeds section thickness, so that all parts of the image are in focus at the same time. This limits the ways in which electron micrographs can be used. Profile counts based on electron micrographs are generally not advisable, since most structures to be counted are large in relation to section thickness, which is usually less than 100 nm. However, useful sections can be slightly thicker than this and, since synaptic vesicles are roughly 2 0 - 4 0 nm in diameter, they are an exception. Not only are their dimensions small relative to section thickness, but also one can reasonably treat them as spherical or ovoid objects that are generally not oriented relative to section plane, or that can be measured in more than one section plane if there is reason to suspect some orientation of 'flattened' vesicles. The extent to which a section through a part of a vesicle may be invisible in an electron micrograph will depend on the size and the nature of the vesicle, but arriving at a determination of this particular 'lost caps' problem may prove elusive. In practice, counts of vesicles, unless the vesicles are relatively large and sparsely packed, could not be based on serial reconstructions or use the physical disector, since the small, closely packed vesicles, will all be in focus in any one micrograph, so that matching from one section to the next is not feasible for most synapses. Probably profile counts with appropriate corrections are the best that one can expect to do, and where two conditions are being compared on the basis of vesicle packing or number, a good record of vesicle sizes should also be provided. Counts of larger objects, such as mitochondria,
synaptic terminals or synaptic junctions 11, are best carried out with the physical disector method. Profile counts would need to include unacceptably large corrections based on the size of the objects and the thickness of the sections. In principle, object size can be estimated (see Anker and Cragg, 1974; Colonnier and Beaulieu, 1985), but such estimates represent approximations and, more importantly, are labor-intensive so that there is a danger that the final number will be based on relatively small samples and thus represent rather rough estimates (which, in some instances, may, of course, be sufficient for the purposes of the study). Two problems concerning counts of synaptic junctions deserve some attention. One is the question of defining exactly what is to be classified as a synaptic junction. If one insists that the synaptic cleft must be visible, the pre- and postsynaptic thickenings clearly distinguishable, and a small crowd of vesicles associated with the presynaptic thickening, then one will count relatively few synapses. Most investigators accept obliquely cut synapses, where the membrane thickenings are recognizable, but not necessarily distinct (see Figs. 2 and 3), but when this is done, it is not easy to define the point at which an apparent thickening ceases to be classifiable as a synapse. Similarly, the rigidity with which the juxtasynaptic presence of more than one vesicle is used as a criterion can vary. This is fine, provided that exactly the same criteria are used throughout any one study, but becomes more difficult where two independent studies need to be compared. Unless the criteria are rather strictly outlined in the methods section, it is unlikely that two laboratories will be using exactly the same criteria. Furthermore, where two conditions are being compared by one observer, it is essential that a 'blind' observer make the counts. J1The distinction between synaptic junctions and synaptic terminals is important since one terminal will often establish several junctional specializations.
Fig. 3. Four micrographs taken at four different tilt angles from the same region of a section from the hippocampus of a rat. Conventions as in Fig. 2. The junction labeled 1 shows a clear synaptic cleft in A at -30 °, but is only poorly defined in D at +50 °. In contrast, junction 2 has a well-defined cleft at all angles, except in A at -30 °. Junction 3, which is a symmetricaljunction, shows a well defined cleft at -30 °, but only the presynaptic dense projections are recognizable at +30 ° and +50 °. The region labeled 4 probably represents a junction roughly en face in A and B, with some sign of a thickening identifiable in D.
39
40
Fig. 4. When a synaptic junction is viewed en face it is difficult to recognize it as a junction, particularly if not all of it is included in a single section. The three structures indicated by arrows in this figure may represent such en face views, partially (3) or completely (1 and 2) included in the section. However, interpreting them as synaptic junctions is necessarily tentative, and for the purpose of synaptic counts they would almost certainly not be counted. The second problem, which is closely related to the first, is presented in Figs. 2 - 4 , which are electron microscopic images of sections that have been tilted through a total of 9 0 - 1 0 0 °. Zero degrees in the figures represents the usual position of the grid perpendicular to the beam. It is evident that the appearance of a synaptic junction changes with a change in the angle at which it is viewed. Most tellingly, the appearance can be changed from one that quite obviously represents a synaptic junction to one that might not readily be accepted as a synaptic junction, being only a blur or smudge that represents a synaptic thickening viewed en face or partially en face. It is important to notice that although the quality of the image deteriorates with increasing tilt (because the beam is passing through more tissue), some synaptic junctions are recognizable at maximum tilt, but are absent or dubious in the untilted section. In any single electron micrograph one can identify structures such as those illustrated in Fig. 4 that can be interpreted as en face images of synaptic junctions, some probably representing most or all of a junction, whereas others represent a part of a cut, en face junction. Unless these junctions are long and curve into the adjacent section, they will not be recognizable as synaptic junctions even where serial sections are available. With thicker sections, junctions viewed en face disappear completely (Scott and Guillery, 1974), so that information about section thickness may be important for an interpretation of
synaptic counts that are based on electron microscopic images. Even so, however, we know of no way in which one can estimate the proportion of synaptic junctions that are unrecognizable in any one section. Since the maximum tilt possible is generally about 90 ° , circa 45 ° in either direction, and with greater tilt the image becomes unusable, one cannot examine even a small piece of tissue at all possible tilt angles. Furthermore, for our studies, we only had one tilt axis (indicated in the figure) available, and could not rotate the stage in order to study a single junction with several different tilt axes. The conclusion that a significant proportion of synapses cannot be seen in electron micrographs is probably not surprising to many electron microscopists. The problem for synaptic counts is two-fold. One is that we have no estimate of what that proportion is. The other, intuitively obvious, and illustrated in Figs. 2 and 3, is that the angle over which any one synapse remains recognizable will vary. Asymmetrical junctions with a thick postsynaptic thickening can take more tilt than symmetrical junctions before they are lost to view, and curved junctions, depending on how the curvature relates to axis of tilt, will often remain identifiable over greater angles than junctions that are fiat. Small junctions are more likely to be lost than large junctions, particularly if serial sections are used, because few large junctions are completely flat. The conclusions that arise from these considerations are rather disconcerting, because they suggest
41 that in any study of synaptic junctions, no matter whether they are profile counts or use the physical disector, only a certain proportion of all synapses can be identified, and, so far, we know of no methods that would define what that proportion is. However, this still allows useful estimates of synaptic numbers in a tissue, by providing orders of magnitude and ratios, and these are often all that is needed. Whether one can use such counts for making comparisons between two ages or two clinical or experimental conditions, will depend on the extent to which one can establish that variables affecting the visibility of synaptic junctions are unchanged; that is, that there is not a change in the curvature of some or all of the junctions, and that the junctions do not change in terms of the relative density of the synaptic thickenings. In view of the problems considered above, staining methods that may be selective for synaptic junctions are likely to prove useful for more accurate counts. A method earlier proposed by Bloom and Aghajanian (1968), using ethanolic phosphotungstic acid, reveals synaptic thickenings quite strikingly in electron micrographs, and may still be a useful tool, but it is likely that other methods employing post-embedding immunohistochemical staining of specific components of the juxta-synaptic specializations could prove more useful. The use of the physical disector on such preparations may provide a means of recording numbers of synaptic junctions in a relatively accurate way. Of course, given a method for revealing all synaptic junctions in an electron microscopic section, it should then be possible to estimate the proportion of synaptic junctions that are not seen in conventional electron micrographs, although one should expect this proportion to vary somewhat, from one part of the brain to another, from one synaptic type to another, and also to depend on section thickness.
Conclusions The major conclusion for investigators planning counts is that there is no one best method. The method has to be suited to the nature of the material and to the nature of the problem. It may also need to be suited to the equipment available. There is often significant tension between highly accurate, often
labor-intensive counts, which are likely to produce small samples, on the one hand, and, on the other hand, more rapid, possibly less accurate counts, that can, however, produce relatively large samples. The more that is known about the variability of the populations, and this includes the populations of the objects in terms of their sizes, shapes and distributions, as well as the populations of organisms, people, cats, mice etc. being studied, the clearer will be the choice as to methods and necessary sample sizes. No counting method is entirely free of assumptions in all of its applications. Given the danger that the term 'assumption-free', when applied to a quantitative method, will lead to a false sense of security about the reliability of the results and their interpretation, it may be best for the term to he avoided altogether and replaced by a thoughtful and complete account of the tissues that have been used, and the methods employed to avoid a biased result.
Acknowledgements We thank Dr. J.B. Pawley for helpful discussions of some of the optical problems, Dr. A.O.W. Stretton for helpful comments on an earlier draft, and Dr. A. Messing for the loan of a microcator. Supported by grants EY 11494 and EY 12936.
References Abercrombie, M. (1946) Estimation of nuclear population from microtome sections. Anat. Rec., 94: 239-247. Anker, R.L. and Cragg, B.G. (1974) Estimation of the number of synapses in a volume of nervous tissue from counts in thin sections by electron microscopy.J. NeurocytoL, 3: 725-735. Benes, EM. and Lange, N. (2001) Two dimensional versus three dimensional cell counting: a practical perspective. Trends Neurosci., 24:11-17. Bloom, EE. and Aghajanian, G.K. (1968) Fine structural and cytochemical analysis of the staining of synaptic junctions with phosphotungstic acid. J. Ultrastruct. Res., 22: 361-375. Clarke, P.G.H. (1992) How inaccurate is the Abercrombie correction factor for cell counts?. Trends Neurosci., 15:211-212. Clarke, P.G.H. (1993) An unbiased correction factor for cell counts in histological sections. J. Neurosci. Methods, 49:133140. Coggeshall, R.E. and Lekan, H.A. (1996) Methods for determining numbers of cells and synapses: a case for more uniform standards of review. J. Comp. Neurol., 364: 3-16. Colonnier, M. and Beaulieu, C. (1985) An empirical assessment of stereological formulae applied to the counting of synaptic
42
discs in the cerebral cortex. J. Comp. Neurol., 231: 175-179. Coupland, R.E. (1968) Determining sizes and distribution of sizes of spherical bodies such as chromaffin granules in tissue sections. Nature, 217: 384-388. Eri~ir, A., Van Horn, S.C., Bickford, M.E. and Sherman, S.M. (1998) Distribution of synapses in the lateral geniculate nucleus of the cat. Differences between laminae A and A1 and between relay cells and interneurons. J. Comp. Neurol., 390: 247-255. Floderus, S. (1944) Untersuchungen fiber den Bau der menschlichen Hypophyse mit besonderer BeriJcksichtung der quantitativen mikromorphologischen Verh~iltnisse. Acta Pathol. Microbiol. Scand. Suppl., 53: 1-266. Geuna, S. (2000) Appreciating the difference between designbased and model-based sampling strategies in quantitative morphology of the nervous system. J. Comp. Neurol., 427: 333-339. Glaser, E.M. (1982) Snell's law: the bane of computer microscopists. J. Neurosci. Methods, 5: 201-202. Guillery, R.W. (1969) A quantitative study of synaptic interconnections in the laminae of the dorsal lateral geniculate nucleus of the cat. Z. Zellforsch., 96: 1-38. Guillery, R.W. and Herrup, K. (1997) Quantification without pontification: choosing a method for counting objects in sectioned tissues. J. Comp. NeuroL, 386: 2-7. Gundersen, H.J.G. and Osterby, R. (1981) Optimizing sampling efficiency of stereological studies in biology: or 'do more less well!'. J. Microsc., 121: 65-73. Gundersen, H.J.G., Bagger, P., Bendtsen, T.E, Evans, S.M., Korbo, L., Marcussen, N., Moiler, A., Nielsen, K., Nyengaard, J.R. and Pakkenberg, B. (1988) The new stereological tools: disector, fractionator, nucleator and point sampled intercepts and their use in pathological research and diagnosis. APM1S, 96: 857-881. Gundersen, H.J.G., Jensen, E.B., Kieu, K. and Nielsen, J. (1999) The efficiency of systematic sampling in stereology - - reconsidered. J. Microscop., 193:199-211. Haug, H. (1956) Remarks on the determination and significance of the gray cell coefficient. J. Comp. Neurol., 104: 473-492. Haug, H. (1986) History of neuromorphometry. J. Neurosci. Methods, 18: 1-17. Haug, H. (1987) Brain sizes, surfaces and neuronal sizes of the cortex cerebri: a stereological investigation of man and his variability and a comparison with some mammals. Am. J. Anat., 180: 126-142. Hedreen, J.C. (1998) What was wrong with the Abercrombie and empirical cell counting methods? A review. Anat. Rec., 250: 373-380. Hendry, I.A. (1976) A method to correct adequately for the change in neuronal size when estimating neuronal numbers after nerve growth factor treatment. J. Neurocytol., 5: 337-349. Konigsmark, B.W. (1970) Methods for the counting of neurons. In: W.J.H. Nauta and S.O.E. Ebbesson (Eds.), Contemporary Research Methods in Neuroanatomy. Springer, Heidelberg, pp. 315-338. Lange, P.W. and Edstr6m, A. (1954) Determination of thickness of microscopic objects. Lab. Invest., 3:116-131.
Lillie, R.D. (1965) Histopathologic Technic and Practical Histochemistry. 3rd edn., McGraw Hill, New York, NY. Marengo, N.P. (1944) Paraffin section thickness - - a direct method of measurement. Stain Technol., 19: 1-10. Mayhew, T.M. and Gundersen, H.J.G. (1996) "If you assume you can make an ass out of u and me": a decade of the disector for stereological counting of particles in 3D space. J. Anat., 188: 1-15. Miller, W.S. and Carlton, E.P. (1895) The relation of the cortex of the cat's kidney to the volume of the kidney and an estimation of the number of glomeruli. Trans. Wisc. Acad. Sci., 10: 525-538. Pakkenberg, B. and Gundersen, H.J.G. (1989) New stereological method for obtaining unbiased and efficient estimates of total cell number in human brain areas. Exemplified by the mediodorsal nucleus in schizophrenics. APMIS, 97: 677-681. Pakkenberg, B. and Gundersen, H.J.G. (1995) Solutions to old problems in the quantitation of the central nervous system. J. Neurol. Sci., 129(Suppl 95): 65-67. Pakkenberg, B. and Gundersen, H.J.G. (1997) Neocortical neuron number in humans. Effect of sex and ages. J. Comp. NeuroL, 384: 312-320. Small, J.V. (1968) Measurements of section thickness. Proc. 4th Eur. Cong. Electron Microsc., 1: 609. Sterio, D.C. (1984) The unbiased estimation of number and sizes of arbitrary particles using the disector. J. Microscop., 134: 127-136. Scott, G.L. and Guillery, R.W. (1974) Studies with the high voltage electron microscope of normal, degenerating and Golgi impregnated neural processes. J. Neurocytol., 3: 567-590. Uylings, H.B., van Eden, C.G. and Hofman, M.A. (1986) Morphometry of size/volume variables and comparison of their bivariate relations in the nervous system under different conditions. J. Neurosci. Methods, 18: 19-37. Weibel, E.R. (1969) Stereological principles for morphometry in electron microscopic cytology. Int. Rev. Cytol., 26: 235-302. West, M.J. (1999) Stereological methods for estimating the total number of neurons and synapses: issues of precision and bias. Trends Neurosci., 22:51-61. West, M.J. (2002) Design based stereological methods for counting neurons. In: T. Sutula and A. Pitk~inen (Eds.), Do Seizures Damage the Brain. Progress in Brain Research, Vol. 135. Elsevier, Amsterdam, pp. 43-51. West, M.J. and Slomianka, L. (1998) Total numbers of neurons in the layers of the human entorhinal cortex. Corrigendum. Hippocampus, 8: 426. West, M.J., Ostergaard, K., Andreassen, O.A. and Finsen, B. (1996) Estimation of the number of somatostatin neurons in the striatum: an in situ hybridization study using the optical fractionator method. J. Comp. Neurol., 370:11-22. Williams, R.W. and Rakic, P. (1988) Three dimensional counting: An accurate and direct method to estimate numbers of cells in sectioned material. J. Comp. Neurol., 278: 344-352. Williams, R.W., Strom, R.C., Rice, D.S. and Goldowitz, D. (1996) Genetic and environmental control of variation in retinal ganglion cell number in mice. J. Neurosci., 16: 7193-7205.
T. Sutula and A. Pitkanen (Eds.) Progress in Brain Research, Vol. 135 © 2002 Published by Elsevier Science B.V.
CHAPTER 4
Design-based stereological methods for counting neurons M a r k J. W e s t * Department of Neurobiology, University of Aarhus, 8000 Arhus C, Denmark
Abstract: Recently developed stereological methods for counting neurons have a number of advantages over previously available stereological methods. These methods are most aptly referred to as 'design-based' because, in contrast to their predecessors, the probes and the sampling schemes that define the newer methods are 'designed', that is, defined a priori, in such a manner that one need not take into consideration the size, shape, orientation, and distribution of the objects to be counted. The elimination of the need for information about the geometry of the objects to be counted results in more robust data regarding estimates of total neuron number and neuronal loss because potential sources of systematic errors in the calculations are eliminated. In this article I will describe the salient features of the newer, design-based, methods and why they represent improvements over previously available methods.
Introduction A recent search of the electronic database maintained by the National Library of Medicine indicates that of the approximately 9000 articles dealing with the morphology of epilepsy, only 16 deal with morphometry or quantitative anatomy and only 3 papers in the last three years involve counting neurons. In view of the fact that neurons are the fundamental functional units of the nervous system and that neuron loss can be viewed as a robust measure of the cumulative damage to a region of the central nervous system, the question arises as to why this is so. It is the author's opinion that this state of affairs is the consequence of the confusion that has been generated by the manner and style of the first articles that described the new stereology and the somewhat uninformed debate that is presently going on in the literature between the advocates of the classic
* Correspondence to: M.J. West, Department of Neurobiology/Anatomy, University of Aarhus, 8000 Arhus C, Denmark. Tel.: +45-8942-3011; Fax: +45-8942-3060; E-maih
[email protected]
assumption-based counting techniques and those of newer design-based counting techniques. Here I will attempt to eliminate some of this confusion by describing the differences between the two approaches and explaining some of the unique features of the newer methods that lead to better data.
Assumption-based stereology Previously available stereological methods for counting, in large part, were based on modeled relationships between the number of objects embedded in a structure and the number of times two dimensional, sectional probes intercept objects of known size, shape, and orientation. At a theoretical level, model-based approaches are valid and have made valuable contributions to the field of stochastic geometry, the mother of stereology. In practice, however, these model-based approaches are more aptly referred to as 'assumption-based' methods because one often assumed, rather than determined or estimated, the size, shape, orientation, and distribution of the modeled objects. This was primarily because it was difficult and time consuming to actually determine the degree to which the 'model' parameters
44
Fig. 1. Low-power light micrographs of a silver-stained horizontal section through the brain of a mouse. The micrographs are taken from different parts of the CA3 pyramidal cell layer on the same section. Note that the profiles of the nuclei of the pyramidal cells appear to have different sizes and shapes, which reflect the different shape and orientation of the nuclei in different parts of the layer. Intraand inter-sectional differences in the size, shape, and orientation of the nuclei within the same structure make it difficult to determine or estimate the height, H, of objects and thereby make estimates with Eq. 1 that do not have a systematic error.
were accurate representations of the true values of the parameters in real structures. For example, it was often assumed that neurons or neuronal nuclei (a frequently used counting unit) were spheres. In many parts of the nervous system, particularly cortical structures, this is the exception rather than the rule (Fig. 1). The principle that underlies the most fundamental model-based approach for estimating the number of objects in a unit volume of tissue, Nv, is based on the relationship between the number of object profiles per unit area of the sections, QA, the mean height of the objects measured orthogonal to the sectioning plane, H, and the thickness of the sections, h (CruzOrive, 1997): Nv = Q A / ( H q-h)
(1)
H + h is the average number of times that a profile of an object can be identified in a section series. This is the basis of the methods described by Abercrombie (1946) and Konigsmark (1970). Note that this method requires either (1) a determination of H (the mean object height) from serial constructions, if H > h, or (2) a determination of H along the focal axis of a thick section, if H < h. In practice, an assumption about H has most often been used (hence the name 'assumption-based' methods) because of the difficulties encountered when
attempting to make determinations of this parameter (Cruz-Orive and Weibel, 1990). In Fig. 2, the mean height, H , of the objects orthogonal to the plane of sectioning, is approximately four section thicknesses (actually 3.94) and there are 326 sectional profiles. According to the formula presented above, Nv = Q A / ( H + h) = 326/5 (which is approximately 66). In this case the volume of the region containing the objects of interest, V(REF), would be the same as the volume sampled, i.e. the sum of the areas of all of the sections multiplied by the thickness of the sections, and the estimate of the total number of neurons is estN = Nv • V ( R E F ) , that is, 66 = 6 6 . 1 . Note, that with this approach, the accuracy of the determination is dependent upon the accuracy of the geometrical description of the objects, more specifically H. If an assumption about the H of the object is not accurate, the resulting determination of N will systematically deviate from the true number and, in a statistical sense, be biased. An error in measurement or false assumption about object H, comparable to one section thickness will in this case result in a 20% bias, that is estN will be 326/4 or 82 rather than 66. If one is determined to avoid making assumptions and actually make determinations of H , this should be done in each individual, in order to avoid assumptions.
45
A.
B. []
oR,
Borrn
m
,,,[], oPn
,nn, B
S=mqnB,,,n, B
96 SECTIONS C.
LOOKUP
D.
Fig. 2. Assumption-based and disector counting. (A) Representation of a green structure that contains 66 objects of different size, shape and orientation and that are unevenly distributed throughout the structure. The structure has been serially sectioned into 96 sections and is viewed orthogonal to the plane of sectioning. The number of objects can be determined by (1) counting the total number of sectional profiles that appear in the sections, 326, and dividing this number by the mean number of sectional profiles per object, 5, gives approximately 66 objects (indirect or assumption-based counting), or (2) counting the first profiles of the objects as they are encountered, that is the leading edges shown in yellow, as one proceeds sequentially through the series (design-based or disector counting). (B) A histogram in which the number of leading edges, small yellow squares, is plotted as a function of the position in the series where the first profile of an object appears when proceeding from left to right. Note that the distribution of objects is not even along the sectioning axis. (C) A disector composed of two sections, the red and blue sections shown in (A) after being rotated 90 degrees. The objects that have sectional profiles within these sections are shown in their entirety. Using the leading edge counting rule, three objects are counted in the red section. (D) An expanded view of the disector seen in (C) which shows the spatial positions of the sectional profiles. The red section is referred to as the 'sample section' and the blue as the 'lookup section'. There are sectional profiles of objects (yellow) in the sample section that do not have sectional profiles in the 'lookup' section. (With permission of Elsevier, from West, 1999.)
46
Design-based stereology The new stereological methods are not dependent upon the need for assumptions about the size, shape, and orientation o f the objects being counted. This is because of the use of a 3-D counting probe, the disector (Sterio, 1984), rather than a 2-D probe (i.e. the section) used in the previously available methods. Unlike the assumption-based approach described above, in which the numerical density, Nv, is derived from a model relationship between the number of object profiles counted on 2-D probes (i.e. sections), disector counting involves the direct counting of objects in a known volume of tissue. In its simplest form, a disector is c o m p o s e d of two sections: a ' s a m p l e section' and a 'lookup section' (Fig. 2C). The volume being probed by the disector is the product o f the area o f the ' s a m p l e section' and the distance between the two sections. The two requirements for proper use o f the disector probe are: (1) any object placed within the region o f interest must be able to be identified on at least one o f the sections that pass through the region, and (2) sectional profiles of the same object must be able to be identified. The disector counting rule is then: an object is considered to be in a disector probe when a sectional
A°
profile o f the object is apparent in the second section, the ' s a m p l e section', and not in the first, the 'lookup section', as one proceeds through the section series. In essence, what one is doing is directly counting the number o f leading edges 'tops' present in the volume defined by the disector. Regardless of the direction o f sectioning and the size, shape, and orientation of the objects, there will be only one leading edge for each object. In order to identify sectional profiles that belong to the same object, as in the case of branching objects, it may, however, be necessary to have access to additional sections that are between and adjacent to the disector pair. Although this method o f counting has been discovered and rediscovered over the centuries (Bendtsen and Nyengaard, 1989), a relatively recent development has made disector counting feasible in histological tissue in which the numbers of objects reaches thousands and millions. This is the unbiased areal counting frame o f Gundersen (1977) (Fig. 3), which enables one to perform unbiased sub-sampling of sections that have large numbers of sectional profiles of objects of interest. Accordingly, one samples, at random, an area of the test section with an unbiased areal counting frame. The profiles o f the objects that lie partially or entirely within the frame and do not intercept the forbidden line (i.e. a hyper-plane
B°
Fig. 3. Subsampling sections. (A) A physical disector consisting of two separate sections. The small blue square in the sample section (red) represents an unbiased 2-D counting frame that can be used to sample a limited area of the section. Profiles of objects that are either entirely within the frame or partially within the frame, but do not touch the green 'forbidden' line are sampled. (Not shown is the infinite extension of the forbidden line in both directions.) When disector counting rules are used, only one object is counted (yellow profile in upper left of frame. The volume of the disector is defined by the area of the counting frame and the thickness of the sample section. (B) A diagrammatic representation of an optical disector. In this case, the counting grid is superimposed on an image of a thin focal plane that is moved a known distance through a thick section. An object is counted if its leading edge comes into focus within the counting frame, as the latter is moved through the section. The volume of the disector is defined by the area of the counting frame and the extent of the movement of the frame through the thick section. In this example, only one object is counted. (With permission of Elsevier, from West, 1999.)
47 that divides the sampling field), are defined as the objects that are to be 'tested'. One then applies disector counting rules to the profiles sampled by the frame and the corresponding part of the 'lookup section'. If the sectional profile of an object, 'sampled' by the areal counting frame placed on the 'sample section', does not have a profile in the 'lookup section', it is defined as an object that should be counted in the volume defined by the disector. In this case, the latter is the product of the area of the counting frame and the distance between the corresponding surfaces of the two sections. A short time after the first descriptions of how disector counting rules could be applied to unbiased sub-samples of large sections, it became apparent that the sections used to define a disector need not be physically separate sections (Gundersen, 1986; Gundersen et al., 1988b). By adjusting the optics of the light microscope so that the depth of focus was minimized (i.e. opening the diaphragm of the sub-stage condenser lens), it was possible to apply disector counting rules to optical sections positioned within thick sections. It was also possible to increase the volume of the sample by increasing the number of consecutive optical sections, so that a virtual 'stack' of optical sections then defined the probe. This probe was subsequently referred to as an optical disector (West and Gundersen, 1990) and the original disector referred to as a physical disector in order to distinguish between the two. The volume of an optical disector probe is then, the product of the area of the unbiased areal counting frame and the distance between the corresponding surfaces of the upper and lower optical sections in the stack. Optical disector counting is performed by superimposing an unbiased areal counting frame on an image of an optical section and 'moving' the counting frame a known distance, through the thickness of the section, with the focus control of the microscope (Fig. 4). An object is considered to be in an optical disector if its top comes into focus within the unbiased counting frame, as one focuses through the section. Objects in focus at the uppermost focal level of the optical disector are not counted because this represents the 'look up section' of the first disector pair in the stack. They are counted at the bottom level, since this is the 'sample section' of the last disector in the stack. The optical disector probe has a number of advan-
tages over the physical disector probe when used at the light microscopic level. First and foremost, one's ability to find the corresponding parts of the sections that have to be compared, when applying the counting rules, is greatly simplified in that one only has to focus up and down to find the corresponding areas of the 'sample' and 'lookup' sections. This is a major problem when using physically separate sections that contain large numbers of profiles of the objects of interest. When using the optical disector probe it is also considerably easier to look at other 'sections' when attempting to determine whether or not profiles of objects at one level belong to the same object. Unfortunately, the optical disector concept cannot be used at the electron microscopic (EM) level (e.g. to count synapses). This is because the depth of focus of an electron image is very large (in the order of meters) and cannot be positioned or moved as a section within of an EM section.
Disector counting is not enough In order to obtain a truly unbiased estimate of total object number, it is not enough to just count the objects with disector probes, it is also important that the sections used in the analysis and the positions on those sections to be sampled with disectors be chosen in a statistically unbiased manner. In order to make this point clear, one should recall that there are two, basic, 'design-based' methods for making unbiased estimates of total object number, N, using optical disector probes (West, 1993). One is the 'two step' method which involves (a) making estimates of the numerical density of objects, Nv, from multiple samples made with optical disector probes, and (b) making estimates of the volume of the tissue in which they are found, V(REF), which can be readily and efficiently obtained by point counting (Gundersen et al., 1988a). According to the first method, Nv. V(REF)= N. This method was unfortunately referred to earlier in the literature as the optical disector. To avoid confusion with the optical disector probe, it is now recommended that it be referred to as either the 'two step' method or the ' N V • V(REF)' method. The other of the two methods is the 'optical fractionator' method (West et al., 1991), with which one counts, also with optical disector probes, the number of objects, ~ Q - , in a known
48
49 fraction, f , of the volume of the structure of interest (Fig. 5). In this case, ~ Q - x ( I / f ) = N. The proper implementation of optical disector probes with both of these methods involves unbiased sampling at two additional levels of the sampling scheme. To make an unbiased estimate of total cell number (i.e. an estimate obtained with a method that on average gives the true number) with either method, there must be a random selection of (1) the sections used in the analysis and (2) the positions on those sections that are sampled with the optical disectors (Fig. 5). If this is not done at both levels, there are constraints with regard to the conclusions that can be drawn from the resulting estimate. For example if one uses a 'standardized' section (Hyman et al., 1998) or a set of sections taken from one end of the region of interest to make counts, the estimate can only be considered to be representative of the entire region when, and only when (1) the Nv estimated in that section is the same as the ratio of the total number of neurons to the reference volume, i.e. NCroT~/V(REF),and (2) the reference volumes, V~REF), are the same in all individuals. The same will be the case if one only samples on one edge or side of the sectional profiles of the region of interest. Without a priori knowledge about NCroTI/V(REV),this approach would also fall into the category of 'assumptionbased' methods because the validity of the resulting data is dependent upon the validity of the assumption about Nv stated above and the assumption that the reference volumes are the same in all individuals in the study. In the biological world, assumptions of this type are generally weak and must be addressed openly in the discussion of any data of this type.
The potential biases inherent in the assumptionbased sampling scheme described above can be eliminated by designing the selection and sampling of sections in such a manner that one does not have to make assumptions about the distribution of the objects of interest. As already alluded to above, the assumption about Nv being the same as the ratio N(TOT)/V(REF)can be eliminated if the sections and the positions within the sections sampled by disectors are randomly sampled (Miles and Davy, 1976). That is, one uses a method of selecting (1) the sections from all of the sections that pass through the region of interest, and (2) the positions within the sections so that all parts of the region of interest (i.e. along all three spatial axes) have equal probabilities of being sampled. The random selection and sampling procedures can be either independent random or systematic random. Systematic random is preferred because in general it is more efficient and more readily applied to histological preparations because they are cut along one axis (Cruz-Orive, 1993; Gundersen et al., 1999). Fig. 5 depicts the application of such a sampling scheme. An example of a scheme for the unbiased sampling of structures that can only be identified at the electron microscopic level (e.g. synapses) can be found in Geinisman et al. (1996).
Summary Data from design-based stereological methods for estimating total neuron number are free from potential biases related to inaccuracies in the geometrical description of the objects being counted. This is
Fig. 4. An optical disector. A stack of optical sections through the granule cell layer, of the dentate gyrus of the human hippocampus, used to make an estimate of the numerical density Nv of granule cells with the optical disector technique. An unbiased counting frame of known area (0.02 m m x 0.02 mm) is superimposed on an optical section obtained with a high numerical aperture oil immersion objective. Each optical section (A-L) is separated by 0.002 ram. Starting with the first lookup section, (A), the nuclei sampled by the frame are counted as one proceeds to focus through a known distance of the section thickness. (In this example nuclei, rather than cell bodies are counted because it is easier and because there is only one nucleus per granule cell.) The profiles of nuclei within the frame or in contact with the thin lines of the frame are considered to be inside the counting frame. Those touching the thick forbidden line are defined as being outside the frame. The two nuclei in focus at the top level of the optical disector, level (A) (black arrows are not counted because they do not come into focus and one proceeds to focus through the section. That is, the top of the disector is also a forbidden line. This point is emphasized by omitting the counting frame from this level. In this optical disector, four nuclei are counted (white arrow heads). Other nuclei that come into focus within the field, but are not sampled by the optical disector, are shown in black. Note that the bottom level of the optical disector is (K). Profiles at this level are sampled, unlike those at level (A). Level (L) has only been included to resolve ambiguities that may arise with branched objects (seldom the case with convex structures such as nuclei) and is not used for counting. (With permission of Wiley-Liss, from West and Gundersen, 1990.)
50
/ a(frame)
l//Ill
L-. ',,l
kkXl
a I It
I I ./I
/ I //,~
I / / ~1
/11
'
I"£N
'\
\ t
,q
P P f~l,~l
~,
P ~ II 1// r ~Y'' '
'
IIIII)kk
7
~d~
1 / /
I'Ll
I I
t \
I t / l ' l
11 \ \
lilly
/
-
a(step)
tll I111
111 I I I I
C
b
Fig. 5. A diagrammatic representation of the optical fractionator sampling scheme for estimating total number of neurons expressing somato-statin mRNA in the striatum of the rat. (a) A systematic random sample of 10-13 sections that span the entire length of the striatum are selected for analysis. The sections are selected at equal intervals, i.e. every nth section after a random start within the first interval, to ensure that all parts of the striatum have equal probabilities of being in the sample. The selected sections therefore constitute a known fraction of the sections in the series, the section sampling fraction (ssf). (b) The labeled neurons are counted under a known fraction of the section area, the area sampling fraction (asf). This fraction corresponds to the ratio of the area of the disector counting frame, a(frame) (shown here as small black rectangles), to the area associated with each step movement of the slide, a(step) (shown here as large white rectangles); asf = a(step)/a(frame). (c) The neurons are counted in optical disectors positioned in the central part of the section thickness. The height of the optical disector, h, constitutes a known fraction of the section thickness, t. The ratio h/t is the thickness sampling fraction (tsf). The area of the counting frame, a(frame), is shaded. After systematically sampling at all levels, one has directly counted the number of neurons Y~ Q - in a known fraction of the region of interest without having to make assumptions about the size, shape and orientation of the objects. The sum of the number of neurons in the disectors, Y~.Q - , times the product of the inverse of the fractions, constitutes an unbiased estimate of the total number of labeled neurons in the striatum, estN = ~ Q - • l/ssf. 1/asf. t/h Note that the volume of the structure and the numerical density are never estimated. (With permission of Wiley-Liss, from West et al., 1996.) b e c a u s e 3 - D p r o b e s a r e u s e d to d i r e c t l y c o u n t t h e n u m b e r o f o b j e c t ' t o p s ' in a k n o w n v o l u m e o f t i s s u e . The tops are zero-dimensional and therefore their
n u m b e r is n o t a f f e c t e d b y t h e g e o m e t r y o f t h e o b jects. Equally important, the probes are distributed in a r a n d o m
manner
throughout
the three dimen-
sions of the region of interest, so that all neurons in the region of interest have an equal probability of being sampled, and assumptions about the distribution of the objects (which also have the potential for leading to biases) need not be made. An appreciation of the differences between the recently developed 'design-based' methods and previously available 'assumption-based' methods can help the investigator of structural dynamics to understand why the newer 'designed-based' stereological methods provide more robust and useful data. The practical aspects of the application of these methods have been described by West (1993). References Abercrombie, M. (1946) Estimation of nuclear population from microtome sections. Anat. Rec., 94: 239-247. Bendtsen, T.E and Nyengaard, J.R. (1989) Unbiased estimation of particle number using sections - - an historical perspective with special reference to the stereology of glomeruli. J. Microsc., 153: 93-102. Cruz-Orive, L.M. (1993) Systematic sampling in stereology. Bull. Int. Stat. Inst., 55: 451-468. Cruz-Orive, L.M. (1997) Stereology of single objects. J. Microsc., 186: 93-107. Cruz-Orive, L.M. and Weibel, E.R. (1990) Recent stereological methods for cell biology: a brief survey. Am, J. Physiol., 258 (Lung Cell. Mol. Physiol., 2): LI48-L156. Geinisman, Y., Gundersen, H.J.G. and West, M.J. (1996) Unbiased stereological estimation of the total number of synapses in a brain region. J. Neurocytol., 25: 805-819. Gundersen, H.J.G. (1977) Notes on the estimation of the numerical density of arbitrary particles: the edge effect. J. Microsc., 1 I1: 219-233. Gundersen, H.J.G. (1986) Stereology of arbitrary particles. A review of unbiased number and size estimators and the presentation of some new ones, in memory of William R. Thompsen. J. Microsc., 143: 3-45.
Gundersen, H.J.G., et al. (1988a) Some new, simple, and efficient stereological methods and their use in pathological research and diagnosis. APMIS, 96: 379-394. Gundersen, H.J.G., et al. (1988b) The new stereological tools: disector, fractionator, nucleator, and point sampled intercepts and their use in pathological research and diagnosis. APMIS, 96: 857-881. Gundersen, H.J.G., Jensen, E.B., Kieu, K. and Nielsen, J. (1999) The efficiency of systematic sampling in stereology - reconsidered. J. Microsc., 193: 199-211. Hyman, B.T., Gomez-Isla, T. and Irizarry, M.C. (1998) Stereology: a practical primer for neuropathology. J. NeuropathoL Exp. Neurol., 57: 305-310. Konigsmark, B.W. (1970) Methods for the counting of neurons. In: W.H.J. Nauta and S.O.E. Ebbesson (Eds.), Contemporary Research Methods in Neuroanatomy. Springer, New York, NY, pp. 315-340. Miles, R.E. and Davy, EJ. (1976) Precise and general conditions for the validity of a comprehensive set of stereological fundamental formulae. J. Microsc., 107:211-226. Sterio, D.C. (1984) The unbiased estimation of number and size of arbitrary particles using the disector. J. Microsc., 134: 127136. West, M.J. (1993) New stereological methods for counting neurons. Neurobiol. Aging, 14: 275-285. West, M.J. (1999) Stereological methods for estimating the total number of neurons and synapses: issues of precision and bias. Trends" Neurosci., 22:51-61. West, M.J. and Gundersen, H.J.G. (1990) Unbiased stereological estimation of the number of neurons in the human hippocampus. J. Comp. Neurol., 296: 1-22. West, M.J., Slomianka, L. and Gundersen, H.J.G. (1991) Unbiased stereological estimation of the total number of neurons in the subdivisions of the rat hippocampus using the optical fractionator. Anat. Rec., 231: 482-497. West, M.J., Ostergaard, K., Andreassen, O. and Finsen, B. (1996) Estimation of the number of somatostatin neurons in the striatum: an in situ study using the optical fractionator. J. Comp. Neurol., 370:11-22.
T. Sutula and A. Pitk~inen (Eds.) Progress in Brain Research, Vol. 135 © 2002 Elsevier Science B.V. All rights reserved
CHAPTER 5
The course of cellular alterations associated with the development of spontaneous seizures after status epilepticus F. Edward Dudek 1,,, Jennifer L. Hellier 2, Philip A. Williams Kevin J. Staley 2
1 Damien
J. Ferraro 1 and
1 Department of Anatomy and Neurobiology, Colorado State University, Fort Collins, CO 80523, USA 2 Department of Neurology, University of Colorado Health Science Center, Denver, CO 80262, USA
Abstract: Chronic epilepsy, as a consequence of status epilepticus, has been studied in animal models in order to analyze the cellular mechanisms responsible for the subsequent occurrence of spontaneous seizures. Status epilepticus, induced by either kainic acid or pilocarpine or by prolonged electrical stimulation, causes a characteristic pattern of neuronal death in the hippocampus, which is followed - - after an apparent latent period - - by the development of chronic, recurrent, spontaneous seizures. The question most relevant to this conference is the degree to which the subsequent chronic seizures contribute further to epileptogenesis and brain damage. This article addresses the temporal and anatomical parameters that must be understood in order to address this question. (1) How does one evaluate experimentally whether the chronic epileptic seizures that follow status epilepticus contribute to epileptogenesis and lead to brain damage? To answer this question, we must first know the time course of the development of the chronic epileptic seizures, and whether the interval between subsequent individual chronic seizures is a relevant factor. (2) What anatomical parameters are most relevant to the progression of epilepsy? For instance, how does loss of inhibitory intemeurons potentially influence seizure generation and the progressive development of epileptogenesis? Does axon sprouting and formation of new synaptic connections represent a form of seizure-induced brain damage? These specific issues bear directly on the general question of whether seizures damage the brain during the chronic epilepsy that follows status epilepticus.
Introduction Considerable research at the molecular and cellular levels has been conducted on the mechanisms of epileptogenesis, but most of this work has focused on hypothetical changes thought to be important for the development of spontaneous seizures (e.g., Dudek et al., 1994; McNamara, 1994, 1999; Dudek and Spitz,
* Correspondence to: EE. Dudek, Department of Anatomy and Neurobiology, Colorado State University, Fort Collins, CO 80523, USA. Tel.: +1-970-491-2942; Fax: +1-970-491-2623; E-mail:
[email protected]
1997). Experimental status epilepticus (SE) is known to induce death o f susceptible neurons, and to lead to the development of chronic epilepsy (e.g., BenAft, 1985 and Nadler, 1991 for kainic acid-induced SE). However, relatively little is known about how the subsequent seizures associated with the chronic epileptic state, which follows SE after an apparent latent period, further affect the brain. The present article will focus primarily on recent work from our group concerning progressive changes in the dentate gyrus after kainate-induced SE. We will consider the SE-induced loss of hilar neurons, and the issue of whether the chronic seizures contribute to further neurodegeneration in the hippocampus. Data on
54 the time course of mossy fiber sprouting and the formation of local excitatory circuits will be considered in relation to the evolution of epilepsy in the kainate-treated rat. The dentate gyrus, aside from its importance for hippocampal function, is considered a model for studying changes that could occur in the other parts of the temporal lobe. Neuronal loss in the hilus and this form of synaptic reorganization in the dentate gyrus have been well documented in human tissue from patients undergoing surgical treatment for intractable epilepsy (e.g., de Lanerolle et al., 1989; Sutula et al., 1989; Houser et al., 1990; Babb et al., 1991). Several technical and conceptual issues relevant to whether chronic seizures contribute to epileptogenesis and damage the brain, at least after experimental SE, will be discussed. Induction of status epilepticus (SE) Several experimental treatments have been used to induce SE, and they can generally be divided into two types: injection of chemotoxins and electrical stimulation. Kainic acid (i.e., kainate; Ben-Ari, 1985; Nadler, 1991) and pilocarpine (Turski et al., 1983) are the two primary chemotoxins that have been used to induce SE. The electrical stimulation protocols involve prolonged, repetitive stimulation of the perforant path, hippocampus, or other limbic structures (e.g., Lothman et al., 1990). In both types of models, repetitive seizures are induced for hours, and they likely cause widespread brain damage. Although these models (and their variants) probably have some significant differences, they have many similarities. In the hippocampus, SE causes loss of neurons in the hilus and in the CA3 and CA1 areas, but other regions of the hippocampus and brain are also clearly affected. Many physiological (e.g., receptors) and structural (e.g., dendrites and axons) alterations, which probably occur with different time courses, follow the SE-induced death of susceptible neurons. Seizures are thought to occur after a latent period of days, weeks or months (e.g., see Hellier et al., 1998, 1999, for kainate-treated rat). The time course of chronic seizures after SE Several studies have reported that chronic SEinduced motor seizures begin to occur after a latent
period, although only a few studies have provided quantitative data (see Stafstrom et al., 1992 and Hellier et al., 1998). We have analyzed this question with direct observation (approximately 6 h/week) and 24 h video monitoring of motor seizures (Racine, 1972). With 6 h/week monitoring, the latent period appeared to be an average of 2-3 months, and there was considerable variability across animals (Hellier et al., 1998). This approach has the limitation that one is monitoring the animals only about 5% of the time. As expected, a different but complementary picture was obtained with 24 h video monitoring (Hellier et al., 1999). In this study, 81% of 26 kainate-treated rats did not have any motor seizures in the first week after kainate treatment (Fig. 1). With this treatment protocol, virtually all kainate-treated rats develop epilepsy (Hellier et al., 1998), and most of these particular rats were later observed to have had one or more motor seizures with 6 h/week monitoring. Of the 19% (i.e., five rats) of the kainate-treated rats that had one or more seizures during the first week after treatment, two of the five rats had one or two motor seizures that occurred only 1 or 2 days after the SE. Two rats had no seizures until day 7, and then had either one or two seizures. One rat, however, had several motor seizures over a 3-day period, 5 - 7 days after the treatment. Our interpretation is that the two rats that had motor seizures within the first 2 days after SE, but no motor seizures for the next several days, had not become epileptic; instead, they had seizures that were an extension of the SE. Based on the criterion that epilepsy is defined as the condition of having had two or more unprovoked seizures, only two of the twenty-six animals became epileptic within the first week after treatment, and their motor seizures began on days 5 and 7 after kainate-induced SE. Thus, the latent period for motor seizures in most kainate-treated rats, using this protocol, is at least 1 week. These data are similar to the results of Stafstrom et al. (1992). Another issue relevant to the concept of progressive changes in epileptogenesis relates to the question of whether seizure frequency increases with time after SE. It is generally believed that many people with temporal lobe epilepsy, although not all of them, have a progressive worsening of their condition. We, therefore, asked whether the frequency of motor seizures increases with time after SE (Hellier
55 100
A
8O "o (1)
60
40
o
"E (13
//
20
Q.
0
No seizures
> 1 seizure
u) ..~ o
B
'-9 O
E
.--
0 "J
z
6 Days after kainate treatment
o
.Q E Z
oJ
i
, 0
B
----
I
I
I
1
2
3
4
5
6
7
Days alter kainate treatment Fig. 1. Analysis of motor seizures during the first week after kainate-induced status epilepticus. (A) With continuous 24 h video monitoring during the first week after kainate treatment, 81% of the rats had no spontaneous motor seizures, but the other 19% were observed to have > 1 motor seizure. (B and C) Number of spontaneous motor seizures per day as a function of time after kainate treatment. Of the five rats that had motor seizures during the first week after treatment, two experienced seizures within the first 27 h, but had no subsequent seizures during the remainder of the first week. The other three rats had their first motor seizure at 5-7 days after kainate treatment. (Reproduced from Hellier et al., 1999 with permission.)
56
~ 0.8
0@
o
A
0.8
t-" (/)
"O
1~20
~0.60.6
~
N~12
1~35
0.2
1
(D
Plus sign represents animals used for analysis
~ 0.2D"
N~471~47
23 N
0.4-
(,9
N-~47
0.0
P
.~1',4
N~47
0.4 ~
N~36
09
23 N
C
B
O
I
I
I
I
I
I
2
3
4
5
6
7
~
~4z 0.0
23 N
Months after kainate treatment
0
(/3
I
I
I
I
I
I
I
1
2
3
4
5
6
7
Months after onset of seizures
to 0.6
0.8
C 0.5 ~0.6
D
t--
C
~
"~ o.4 N
0.4
-N
0.3
N~23
t
8'0.2
N
~
in Figure 3D.
C 03
O.2
,,$.-
0.0 0
I
I
I
I
25
50
75
100
Duration of epileptic period (%)
,~ o.o ._N 0 03
6O
tAsterisk represents significant difference (P<0.05) from the first month after onset of seizures.
I
I
I
I
I
I
I
1
2
3
4
5
6
7
8
Months after onset of seizures
Fig. 2. Seizure frequency as a function of time after kainate treatment. These data are based on a population of 47 kainate-treated rats that had at least one seizure every 30 days after their initial seizure, and that were euthanized >4 months after kainate treatment. (A) Graph of seizure frequency for all 47 rats. At later periods, fewer kainate-treated rats were available for the analysis, because they were either used for experiments or died. (B) In this graph, seizure frequency is assessed as a function of time after the onset of seizures to compensate for the different latent periods. (C) Seizure frequency as a percent of the duration of the total epileptic period. The epileptic period was defined as the number of weeks between the onset of chronic seizures and death of the animal. These data compensate for the effect of preferentially using animals with a high seizure frequency for electrophysiological experiments and for the observation that those animals that had a high seizure frequency were more susceptible to death. (D) Analysis of seizure frequency as a function of time for seven rats that were analyzed for 4-8 months. The seizure frequency at 4-8 months was significantly higher compared to 1 month (note asterisks). (Reproduced with permission from Hellier et al., 1998.)
et al., 1998). Fig. 2 shows the results of an analysis of motor seizure frequency as a function of time after SE. The data indicate that initially the seizures occur at a low rate, and then increase in frequency
over the next several months. On average, seizure frequency reaches an apparent maximum at approximately one seizure every 2 h (0.5 seizures/h). These data, and those of Stafstrom et al. (1992), support the
57 hypothesis that the epileptic state worsens for most animals. Furthermore, the motor seizures can occur nearly every hour, at least under some conditions in some animals, although this requires several months to develop. The data in Fig. 1, and described above, emphasize the difficulty assessing the actual latent period. Our experience suggests that analyses based on seizure frequency are more practical than assessments of the latent period, and are potentially a more accurate approach to assessing the progressive nature of the epileptogenesis in SE-based models. SE causes neuronal death presumably because repetitive seizures occur at a high rate, and several seizures precede each subsequent seizure by a relatively short time period (i.e., minutes). Does the amount of time between seizures (i.e., the interseizure interval) during chronic epilepsy contribute to the likelihood that a seizure will cause neuronal damage? The initial chronic motor seizures that occur shortly after SE are usually quite infrequent in experimental animals, but after several weeks and months, motor seizures can be observed every hour or two. When chronic seizures occur nearly every hour, are they more likely to be deleterious? With intervals as short as an hour, it seems likely that brain recovery systems are inadequate, so that an additive effect of successive chronic seizures may be present in a manner reminiscent to what occurs during SE. A continuum between infrequent seizures, clusters of seizures over several hours with short inter-seizure intervals, and actual SE could hypothetically exist. This possibility emphasizes the need for quantitative analysis of seizure frequency as a variable when considering this question, since the frequency of spontaneous seizures during chronic epilepsy may occasionally come relatively close to the seizure frequency that occurs during SE (i.e., one per hour versus several per hour). Because seizure frequency increases for at least several months after kainate-induced SE, this may be a highly vulnerable time period in the post-SE models of temporal lobe epilepsy.
Anatomical changes after SE: neuronal death The issue of how one measures neuronal death is a fundamental and long-standing problem for the field of epilepsy research. We have chosen to analyze
this issue using neuron counting with stereological methods (e.g., West and Gundersen, 1990; West et al., 1991). In our initial analysis of kainate-treated rats (Buckmaster and Dudek, 1997), we assessed neuronal loss in the hilus of the dentate gyms along the septal-temporal axis using cresyl violet and immunocytochemically stained tissue. We found that about 52% of the hilar neurons in rats with kainateinduced epilepsy (i.e., several months after treatment) were lost compared to saline-treated controls, and that neuronal loss occurred predominantly at the temporal end of the hippocampus. In a subsequent study (Hellier et al., 1999), we analyzed the number of neurons with similar although not identical methods at 1 week after kainate treatment (Fig. 3). The estimated loss of hilar neurons 1 week after kainate treatment was about 35% relative to the controis, also with a preferential loss at the temporal end of the hippocampus. If chronic seizures did kill neurons, one would expect more neuronal loss in the chronically epileptic rats than in rats evaluated 1 week after SE. The apparent loss of hilar neurons was greater (i.e., 52% compared to 35%) in the rats with kainate-induced epilepsy compared to rats 1 week after treatment (i.e., before most rats have chronic motor seizures). For the group studied many weeks after SE, chronic seizures had been occurring for a considerable period of time, and presumably at a high frequency for many of them. Although we did find fewer neurons in the hilus of the rats with kainate-induced epilepsy compared to kainateinjected rats 1 week after treatment, this is a post hoc comparison and an actual quantitative analysis would be inappropriate. First, different investigators conducted the studies, and differences in neuronal counting could have been a source of variance. Second, the animals were prepared at different times, and thus were not appropriately matched. Therefore, although similar stereological analyses were used in these two studies (i.e., Buckmaster and Dudek, 1997 versus Hellier et al., 1999), and the animal model was virtually identical, it is not possible to make firm conclusions from a direct comparison of these two data sets. Furthermore, even if a difference in the two experimental groups were found, it would not be possible to conclude that the chronic seizures caused the loss of neurons. In the long term, it would also seem necessary to use an experimental design that
58 U) t-
2 cO
E
"m c-
400
septal middle temporal Portion of the septotemporal axis
versus after SE and chronic seizures). Furthermore, the neurons that are lost after SE would be expected to be those most susceptible to chronic seizures. These two general issues represent major potential problems in regard to using SE-based models of experimental epilepsy to determine whether chronic seizures damage the brain. Experiments that aim to mark damaged neurons destined to die (e.g., TUNNEL stain) necessarily assume that these neurons will die, and when applied to the SE-based models, still have the limitations and caveats summarized above.
B
Anatomical changes after SE: mossy fiber sprouting
A
-~--~a[in~e, / ~
300 200 100
t-6}
0
3.0 o
2.5
E E
2.0
O 0~
-,-saline -o-kainate
iv. 1.5 m 1.0 ¢-
0.5
0.0
0
O--------------~
septal middle temporal Portion of the septotemporal axis
Fig. 3. Distribution of hilar neurons and Timm stain in the inner molecular layer as a function of position along the septaltemporal axis for saline- and kainate-treated rats. (A) The salinetreated rats had more neurons at the temporal end (100% septotemporal distance), compared to the septal end (0%). All three regions of the hippocampus had fewer neurons in kainate-treated rats. (B) The difference in Timm score for kainate- and salinetreated rats was small but significant along the full extent of the hippocampus. (Reproduced with permission from Hellier et al.,
1999.)
first blocks the chronic seizures after SE for an appropriate period, and then assesses neuronal damage relative to a group with unblocked chronic seizures. Additional technical and conceptual problems exist in this type of experiment, and these issues should be considered in future work. The number and proportion of neurons damaged or killed after SE can be large and variable. In this type of experiment, this variability will be an important source of error in a quantitative comparison of the two experimental groups (i.e., after SE but before chronic seizures
Several laboratories have reported that Timm stain in the inner molecular layer increases with time after SE and during kindling (e.g., Mathern et al., 1992, 1993; Cavazos et al., 1994). We have analyzed both the histological and electrophysiological changes that occur in the dentate gyrus as a function of time after SE (Wuarin and Dudek, 2001). A semi-quantitative analysis of mossy fiber sprouting (i.e., scores ranging from 0 to 3) was based on the intensity and area of Timm stain in the inner molecular layer (see Tauck and Nadler, 1985). We found that mossy fiber sprouting clearly began to occur within the first week or two after kainate treatment. It had progressed substantially by 2-4 weeks, and had become extremely robust 10-51 weeks after treatment (Fig. 4), when virtually all of the animals were displaying spontaneous motor seizures. Using whole-cell recording from dentate granule cells in hippocampal slices from kainatetreated rats, we found that the frequency and amplitude of spontaneous EPSCs increased in parallel with the Timm stain in the inner molecular layer. Furthermore, with flash photolysis of caged glutamate to stimulate relatively small regions of the dentate granule-cell layer, we found that hippocampal slices from kainate-treated rats with mossy fiber sprouting showed EPSCs to focal stimulation of granule cells (Fig. 5; see also Molnar and Nadler, 1999). The number of granule cells that received excitatory input from nearby granule cells increased progressively as a function of time after kainateinduced SE (Fig. 6). Only 1 of 52 controls showed
59
KA-treated
Controls 100
n=441 7 rats
n = 519 8 rats
n = 395 7 rats
n = 678
n = 783
n = 832
1K r=f¢
1~ r~f~
14 r~fe
10
~mm staining rading scale rauck and Nadler, • Neurosci., 5, 1985)
80 i
09
i 60.
i = 3
t-
0 = 2 ~ = 1
09 09 a.
40.
C--] = 0
!
20.
i
! !
0"
1-2
2-4
i
!
I
17-72
1-2
2-4
10-51
Time after treatment (weeks) Fig. 4. Analysis of Timm staining score as a function of time after kainate treatment. Hippocampal sections from kainate-treated rats used in electrophysiological experiments (see below) were analyzed with the Timm stain after saline or kainate treatment. The Timm score was significantly greater in kainate-treated rats and showed a progressive increase as a function of time after kainate treatment• All assessments were done with blind procedures, and the number of sections and rats is shown above each bar. (Reproduced with permission from Wuarin and Dudek, 2001 .)
a similar response (Fig. 6). These data indicate that excitatory connections from granule cells to other granule cells increase in density as a function of time after SE. Previous studies from our group and others (e.g., Cronin et al., 1992; Wuarin and Dudek, 1996, 2001; Patrylo and Dudek, 1998; Hardison et al., 2000; Lynch and Sutula, 2000) showed that epileptiform bursts could be evoked in the isolated dentate gyms after robust mossy fiber sprouting had occurred, months after kainate- or pilocarpine-induced SE, if inhibition was depressed and/or extracellular potassium was elevated slightly. Similar observations have been made in slices of human hippocampus with mossy fiber sprouting from patients with intractable epilepsy (Franck et al., 1995). Activation of a small population of granule cells (using focal photolysis of caged glutamate) evoked network bursts when slices were treated with bicuculline and high extracellular potassium, but only in preparations with
robust sprouting several months after kainate treatment (Fig. 7). Burst discharges were not evoked at earlier time periods with less Timm stain in the inner molecular layer. These data indicate that a progressive increase in local excitatory circuits occurs in the dentate gyrus after SE. As the density of axon sprouting increases with more time after SE, the dentate gyms progressively becomes more capable of generating epileptiform bursts when granule cells are activated after inhibition is depressed and potassium is elevated. These data and others indicate that SE and the consequent neuronal death are followed by a progressive increase in seizure frequency after an apparent latent period. In the dentate gyms, which serves as an experimental model, an increase in neuronal death in the hilus may occur as a function of the chronic seizures, but data from our group are not definitive on this point. A profound in-
60
A
3
B
lO
C
15
Fig. 5. Photoactivationof caged glutamate in the dentate granule-cell layer evoked repetitive EPSCs in granule cells from kainate-treated rats with mossy fiber sprouting. Whole-cell recording at resting membrane potential from a granule cell in the outer blade from a rat 39 weeks after kainate treatment. Repetitive photostimulations(0.05 Hz) were given at one site in the granule-cell layer, 600 txm from the recorded cell. The numbers in each trace indicate the stimulation number, and the traces are continuous in A-C. The top trace in each panel shows baseline activity.The arrows indicate the stimulus artifact from the flash. The bottom trace in C is an expansion of the time period marked by the dashed lines. (Reproduced with permissionfrom Wuarin and Dudek, 2001.)
crease in mossy fiber axons in the inner molecular layer is, however, observed during the months after SE. Several laboratories have provided evidence that this increase in mossy fiber sprouting is also associated with the development of local excitatory circuits (e.g., Cronin et al., 1992; Wuarin and Dudek, 1996, 2001; Molnar and Nadler, 1999; Lynch and Sutula, 2000), and possibly also an increase in lo-
cal inhibitory circuits (Cronin et al., 1992; Sloviter, 1992; Buhl et al., 1996). Ultrastructural observations support these electrophysiological data (Kotti et al., 1997; Zhang and Houser, 1999; Wenzel et al., 2000). Local inhibitory circuits are known to mask the effects of local excitatory circuits (Miles and Wong, 1987; Christian and Dudek, 1988), and when inhibition is intact, the responses of the granule cells
61 100
¢q ID O tt~ t_ O O~
90
•
KA-treatedrats
80
[]
Saline-injected
controls 66% (n = 32)
70 60 50 40
¢Q~
30
Q}
13-
20
10% (n = 29)
10 0
0%
0% (n = 18) 1-2
(n = 17)
2-4
10-51 (kainate) 17-72 (saline)
Time after treatment (weeks) Fig. 6. Plot of the percentage of granule cells responding to photoactivation of caged glutamate with an increase in EPSCs as a function of time after treatment for saline- and kainate-injected rats. For each group, the number of tested cells is indicated above the bar. (Reproduced with permission from Wuarin and Dudek, 200l.)
are either similar to those in the normal dentate gyrus or slightly hyperexcitable (e.g., see Patrylo et al., 1999). Thus, the granule-cell network establishes new forms of connectivity after the loss of approximately half of the neurons in the hilus, but inhibitory circuits tend to mask the new excitatory interconnections. Although we cannot determine from our data whether neuronal loss occurred over time, the progressive nature of the axon sprouting and synaptic reorganization is well established. Anatomical changes after SE: loss of inhibitory interneurons When considering neuronal damage, it is obvious that one has to give particular consideration to the loss of inhibitory interneurons versus principal neurons. Loss of GABAergic interneurons themselves is likely to cause hyperexcitability and seizures in many cortical areas. Several studies have been con-
ducted on dentate granule cells using hippocampal slices in vitro, which have had the excitatory inputs from the entorhinal cortex removed. In these studies, blockade of GABAA-receptor-mediated inhibition and/or elevation of extracellular potassium in control animals has minimal effect or may lead to mild hyperexcitability in response to perforant-path stimulation (Wuarin and Dudek, 1996, 2001; Patrylo et al., 1999; Hardison et al., 2000; Lynch and Sutula, 2000). These experimental conditions, however, do not lead to network bursts from hilar stimulation or focal activation of granule cells with caged glutamate in slices from control animals. On the other hand, as summarized above, after robust mossy fiber sprouting has occurred following experimental SE, pronounced burst discharges have been evoked when GABAA-receptor-mediated inhibition was depressed and/or the concentration of extracellular potassium was raised. Therefore, loss of inhibitory interneurons could have a much more profound effect on the func-
62
t
t 50 mV
500 ms
~.__ ~___. ~_.., ..
t
® Fig. 7. Photoactivation of caged glutamate in the granule-cell layer in the presence of 30 IxM bicuculline and 6 mM extracellular potassium. Photoactivation of caged glutamate in the granule-cell layer evoked epileptiform bursts of action potentials at numerous locations throughout the dentate gyrus when the flash was localized to the granule-cell layer (but not the hilus). The patch pipette indicates the position of the recorded cell in the outer blade of the dentate gyms. Each number shows the location of the photostimulation in the diagram and the correspondingbursts of action potentials. The recording was conducted in current-clampconfigurationat resting potential (-71 mV), and the rat had been treated 33 weeks previously with kainate. The arrowheads indicate the stimulus artifact from the flash. (Reproduced with permission from Wuarin and Dudek, 2001.)
tion of the granule-cell network when new recurrent excitatory circuits are present than when they are not. Several studies have provided evidence for neurodegeneration, axon sprouting and formation of new recurrent excitatory circuits in the CA1 area of the hippocampus after treatments that induce SE (Nadler et al., 1980; Meier and Dudek, 1996; Perez et al., 1996; Esclapez et al., 1999; Smith and Dudek, 2001). Thus, the formation of new recurrent excitatory cir-
cuits could be a widespread p h e n o m e n o n that occurs in many cortical areas under a variety of epileptogenic conditions. Frequent seizures (i.e., during SE or even during chronic epilepsy) could damage or kill some interneurons, and thus not only increase overall excitability, but also u n m a s k new recurrent excitatory circuits. These two mechanisms would be expected to have a combined effect to increase seizure susceptibility, and would lead to a progres-
63
sive worsening of the epilepsy. Gorter et al. (2001) found that rats with a progressive, post-SE increase in the frequency of spontaneous seizures had both extensive loss of parvalbumin- and somatostatinimmunoreactive neurons and also robust T i m m stain in the inner molecular layer compared to rats without a progressive increase in the frequency of spontaneous chronic seizures, which had less of these histopathological markers. The c o m b i n e d loss of inhibitory interneurons and the formation of new recurrent excitatory circuits would potentially have a positive-feedback effect to promote epileptogenesis.
Conclusions Several technical and conceptual issues are relevant to evaluating whether the chronic seizures that develop after SE damage the brain. Because SE itself causes significant and variable neuronal death, an evaluation of subsequent injury attributable only to the chronic seizures is problematic. Neuronal counts, even with an appropriate experimental design, will be a difficult but important approach for assessing whether the post-SE chronic seizures damage the brain. If neuronal injury does result from the chronic seizures, presumably it will become more of a problem as seizure frequency increases weeks and months after SE. Although it has not been clear in our studies whether there is a gradual seizuredependent decline (i.e., independent of prior SE) in the number of neurons, progressive and profound synaptic reorganization occurs in animals and humans for months after SE. This synaptic reorganization signifies, at least partially, the formation of new recurrent excitatory circuits, which are masked in some areas (e.g., the dentate gyms) by strong inhibitory circuits. Loss of inhibitory interneurons, either during or after the injury, would tend to unmask abnormal recurrent excitatory circuits and other potential epileptogenic mechanisms (e.g., upregulated N M D A receptors). Progressive loss of inhibitory interneurons would be expected to augment the electrophysiological effects of new excitatory local circuits. These abnormal circuits would presumably contribute to cognitive deficits that can occur in animal models (Stafstrom et al., 1993) and some people with epilepsy. These issues and concerns are
likely to be important to an analysis of whether seizures lead to brain damage.
References Babb, T.L., Kupfer, W.R., Pretorius, J.K., Crandall, RH. and Levesque, M.E (1991) Synaptic reorganization by mossy fibers in human epileptic fascia dentate. Neuroscience, 42: 351-363. Ben-Aft, Y. (1985) Limbic seizures and brain damage produced by kainic acid: Mechanisms and relevance to human temporal lobe epilepsy. Neuroscience, 14: 375-403. Buckmaster, ES. and Dudek, EE. (1997) Neuron loss, granule cell reorganization, and functional changes in the dentate gyrus of epileptic kainate-treated rats. J. Comp. Neurol., 385: 385404. Buhl, E.H., Otis, T.S. and Mody, I. (1996) Zinc-induced collapse of augmented inhibition by GABA in a temporal lobe epilepsy model. Science, 271: 369-373. Cavazos, J.E., Das, I. and Sutula, T.E (1994) Neuronal loss induced in limbic pathways by kindling: evidence for induction of hippocampal sclerosis by repeated brief seizures. J. Neurosci., 14: 3106-3121. Christian, E.E and Dudek, EE. (1988) Characteristics of local excitatory circuits studied with glutamate microapplication in the CA3 area of rat hippocampal slices. J. Neurophysiol., 59: 90-109. Cronin, J., Obenaus, A., Houser, C.R. and Dudek, EE. (1992) Electrophysiology of dentate granule cells after kainateinduced synaptic reorganization of the mossy fibers. Brain Res., 573: 305-310. De Lanerolle, N.C., Kim, J.H., Robbins, R.J. and Spencer, D.D. (1989) Hippocampal interneuron loss and plasticity in human temporal lobe epilepsy. Brain Res., 495: 387-395. Dudek, EE. and Spitz, M. (1997) Hypothetical mechanisms for the cellular and neurophysiologic basis of secondary epileptogenesis: proposed role of synaptic reorganization. J. Clin. Neurophys., 14: 90-101. Dudek, EE., Obenaus, A., Sehweitzer, J.S. and Wuarin, J.-E (1994) Functional significance of hippocampal plasticity in epileptic brain: electrophysiological changes of the dentate granule cells associated with mossy fiber sprouting. Hippocampus, 4: 259-265. Esclapez, M., Hirsch, J.C., Ben-Ari, Y. and Bernard, C. (1999) Newly formed excitatory pathways provide a substrate for hyperexcitability in experimental temporal lobe epilepsy. J. Comp. Neurol., 408: 449-460. Franck, J.E., Pokorny, J., Kunkel, D.D. and Schwartzkroin, EA. (1995) Physiologic and morphologic characteristics of granule cell circuitry in human epileptic hippocampus. Epilepsia, 36: 543-558. Gorter, J.A., van Vliet, E.A., Aronica, E. and Lopes da Silva, F.H. (2001) Progression of spontaneous seizures after status epilepticus is associated with mossy fibre sprouting and extensive bilateral loss of hilar parvalbumin and somatostatinimmunoreactive neurons. Eur. J. Neurosci., 13: 657-669.
64
Hardison, J.L., Okazaki, M.M. and Nadler, J.V. (2000) Modest increase in extracellular potassium unmasks effect of recurrent mossy fiber growth. J. Neurophysiol., 84: 2380-2389. Hellier, J.L., Patrylo, P.R., Buckmaster, P.S. and Dudek, EE. (1998) Recurrent spontaneous motor seizures after repeated low-dose systemic treatment with kainate: assessment of a rat model of temporal lobe epilepsy. Epilepsy Res., 31: 73-84. Hellier, J.L., Patrylo, P.R., Dou, P., Nett, M., Rose, G.M. and Dudek, EE. (1999) Assessment of inhibition and epileptiform activity in the septal dentate gyrus of freely behaving rats during the first week after kainate treatment. J NeuroscL, 19: 10053-10064. Houser, C.R., Miyashiro, J.E., Swartz, B.E., Walsh, G.O., Rich, J.R. and Delgado-Escueta, A.V. (1990) Altered patterns of dynorphin immunoreactivity suggest mossy fiber reorganization in human hippocampal epilepsy. J. Neurosci., 10: 267282. Kotti, T., Riekkinen Sr., P.J. and Miettinen, R. (1997) Characterization of target cells for aberrant mossy fiber collaterals in the dentate gyrus of epileptic rat. Exp. NeuroL, 146: 323-330. Lothman, E.W., Bertram, E.H., Kapur, J. and Stringer, J.L. (1990) Recurrent spontaneous hippocampal seizures in the rat as a chronic sequela to limbic status epilepticus. Epilepsy Res., 6: 110-118. Lynch, M. and Sutula, T. (2000) Recurrent excitatory connectivity in the dentate gyrus of kindled and kainic acid-treated rats. J. Neurophysiol., 83: 693-704. Mathern, G.W., Kupfer, W.R., Pretorius, J.K., Babb, T.L. and Levesque, M.E (1992) Onset and patterns of hippocampal sprouting in the rat kainate seizure model: evidence for progressive cell loss and neo-innervation in regio inferior and superior. Dendron, 1: 69-84. Mathern, G.W., Cifuentes, E, Leite, J.P., Pretorius, J.K. and Babb, T.L. (1993) Hippocampal EEG excitability and chronic spontaneous seizures are associated with aberrant synaptic reorganization in the rat intrahippocampal kainate model. Elec. Clin. Neurophysiol., 87: 326-339. McNamara, J.O. (1994) Cellular and molecular basis of epilepsy. J. Neurosci., 14: 3413-3425. McNamara, J.O. (1999) Emerging insights into the genesis of epilepsy. Nature, 24: 15-22. Meier, C.L. and Dudek, EE. (1996) Spontaneous and stimulation-induced synchronized bursts after discharges in the isolated CA1 of kainate-treated rats. J. Neurophysiol., 76: 2231-2239. Miles, R. and Wong, R.K. (1987) Inhibitory control of local excitatory circuits in the guinea-pig hippocampus. J. Physiol., 388: 611-629. Molnar, P. and Nadler, J.V. (1999) Mossy fiber-granule cell synapses in the normal and epileptic rat dentate gyrus studied with minimal laser stimulation. J. NeurophysioL, 82: 18831894. Nadler, J.V. (1991) Kainic acid as a tool for the study of temporal lobe epilepsy. Life Sci., 29:2031-2042. Nadler, J.V., Perry, B.W. and Cotman, C.W. (1980) Selective reinnervation of hippocampal area CA1 and the fascia den-
tata after destruction of CA3-CA4 afferents with kainic acid.
Brain Res., 182: 1-9. Perez, Y., Morin, E, Beaulieu, C. and Lacaille, J.-C. (1996) Axonal sprouting of CA1 pyramidal cells in hyperexcitable hippocampal slices of kainate-treated rats. Eur. J. Neurosci., 8: 736-748. Patrylo, P.R., Schweitzer, J.S. and Dudek, EE. (1999) Abnormal responses to perforant path stimulation in the dentate gyrus of slices from rats with kainate-induced epilepsy and mossy fiber reorganization. Epilepsy Res., 36:3631-3642. Patrylo, P.R. and Dudek, F.E. (1998) Physiological unmasking of new glutamatergic pathways in the dentate gyrus of hippocampal slices from kainate-induced epileptic rats. J. Neurophysiol., 79: 418-429. Racine, R.J. (1972) Modification of seizure activity by electrical stimulation, II. Motor seizures. Electroencephalogr. Clin. Neurophysiol., 32: 281-294. Sloviter, R.S. (1992) Possible functional consequences of synaptic reorganization in the dentate gyrus of kainate-treated rats. Neurosci. Lett., 137: 91-96. Smith, B.N. and Dudek, EE. (2001) Short- and long-term changes in CA1 network excitability after kainate treatment in rats. J. Neurophysiol., 85: 1-9. Stafstrom, C.E., Thompson, J.L and Holmes, G.L. (1992) Kainic acid seizures in the developing brain: status epilepticus and spontaneous recurrent seizures. Dev. Brain Res., 65: 227-236. Stafstrom, C.E., Chronopoulos, A., Thurber, S., Thompson, J.L. and Holmes, G.L. (1993) Age-dependent cognitive and behavioral deficits after kainic acid seizures. Epilepsia, 34: 420432. Sutula, T., Cascino, G., Cavazos, J., Parada, I. and Ramirez, L. (1989) Mossy fiber synaptic reorganization in the epileptic human temporal lobe. Ann. Neurol., 26: 321-330. Tauck, D.L. and Nadler, J.V. (1985) Evidence of functional mossy fiber sprouting in hippocampal formation of kainic acid-treated rats. J. Neurosci., 5: 1016-1022. Turski, W.A., Cavalheiro, E.A., Schwarz, M., Czuczwar, S.J., Kleinrok, Z. and Turski, L. (1983) Limbic seizures produced by pilocarpine in rats: behavioural, electroencephalographic and neuropathological study. Behav. Brain Res., 9: 315-335. Wenzel, H.J., Woolley, C.S., Robbins, C.A. and Schwartzkroin, P.A. (2000) Kainic acid-induced mossy fiber sprouting and synapse formation in the dentate gyrus of rats. Hippocampus, 10: 244-260. West, M.J. and Gundersen, H.J. (1990) Unbiased stereological estimation of the number of neurons in the human hippocampus. J. Comp. NeuroL, 296: 1-22. West, M.J., Slomianka, L. and Gundersen, H.J.G. (1991) Unbiased stereological estimation of the total number of neurons in the subdivisions of the rat hippocampus using the optical fractionator. Anat. Rec., 231: 482-497. Wuarin, J.-E and Dudek, EE. (1996) Electrographic seizures and new recurrent excitatory circuits in the dentate gyrus of hippocampal slices from kainate-treated epileptic rats. J. Neurosci., 16: 4438-4448. Wuarin, J.-P. and Dudek, F.E. (2001) Excitatory synaptic input
65
to granule cells increases with time after kainate treatment. J.
NeurophysioL, 85: 1067-1077. Zhang, N. and Houser, C.R. (1999) Ultrastructural localization
of dynorphin in the dentate gyrus in human temporal lobe epilepsy: a study of reorganized mossy fiber synapses. J. Comp. Neurol., 405: 472-490.
T. Sutulaand A. Pitk~inen(Eds.) Progress in Brain Research, Vol. 135
© 2002 ElsevierScienceB.V.All rightsreserved CHAPTER 6
Progression of neuronal damage after status epilepticus and during spontaneous seizures in a rat model of temporal lobe epilepsy Asia Pitk~inen 1,3,,, Jari Nissinen 1, Jaak Nairism~igi 2, Katarzyna Lukasiuk 1, Olli H.J. Gr6hn 2, Riitta Miettinen 3,4 and Risto Kauppinen 2 I Epilepsy Research Laboratory and e NMR Research Group, A.L Virtanen Institute for Molecular Sciences, University of Kuopio, PO. Box 1627, F1N-70 211 Kuopio, Finland 3Department of Neurology, Kuopio University Hospital, P.O. Box 1777, FIN-70 211 Kuopio, Finland 4 Department of Neurology and Neurosciences, University of Kuopio, P.O. Box 1627, FIN- 70 211 Kuopio, Finland
Abstract:
The present study was designed to address the question of whether recurrent spontaneous seizures cause progressive neuronal damage in the brain. Epileptogenesis was triggered by status epilepticus (SE) induced by electrically stimulating the amygdala in rat. Spontaneous seizures were continuously monitored by video-EEG for up to 6 months. The progression of damage in individual rats was assessed with serial magnetic resonance imaging (MRI) by quantifying the markers of neuronal damage (T2, T10, and Dav) in the amygdala and hippocampus. The data indicate that SE induces structural alterations in the amygdala and the septal hippocampus that progressively increased for approximately 3 weeks after SE. T2, T~p, and Dav did not normalize during the 50 days of follow-up after SE, suggesting ongoing neuronal death due to spontaneous seizures. Consistent with these observations, Fluoro-Jade B-stained preparations revealed damaged neurons in the hippocampus of spontaneously seizing animals that were sacrificed up to 62 days after SE. The presence of Fluoro-Jade B-positive neurons did not, however, correlate with the number of spontaneous seizures, but rather with the time interval from SE to perfusion. Further, there were no Fluoro-Jade B-positive neurons in frequently seizing rats that were perfused for histology 6 months after SE. Also, the number of lifetime seizures did not correlate with the severity of neuronal loss in the hilus of the dentate gyms assessed by stereologic cell counting. The methodology used in the present experiments did not demonstrate a clear association between the number or occurrence of spontaneous seizures and the severity of hilar cell death. The ongoing hippocampal damage in these epileptic animals detected even 2 month after SE was associated with epileptogenic insult, that is, SE rather than spontaneous seizures.
Introduction Analysis of histologic preparations stained using Nissl, silver, terminal deoxytransferase nick end* Correspondence to: A. Pitk~inen, Epilepsy Research Laboratory, A.I. Virtanen Institute, University of Kuopio, P.O. Box 1627 (street address: Neulaniementie 2), FIN-70 211 Kuopio, Finland. Tel.: +358-17-16-3296; Fax: -+-358-1716-3025; E-mail:
[email protected]
labeling (TUNEL), or immunohistochemical techniques indicates that even a few electrically induced kindled seizures can cause a loss of subpopulations of neurons in the amygdala (Callahan et al., 1991; Pretel et al., 1997; Tuunanen et al., 1997; Tuunanen and Pitk~inen, 2000) and hippocampus (Cavazos et al., 1994; Bengzon et al., 1997; Pretel et al., 1997; Dalby et al., 1998; Zhang et al., 1998). In the hippocampus, the severity of neuronal damage correlates with the number of afterdischarges, that is,
68 evoked seizures in a kindling model (Cavazos et al., 1994), and its distribution depends on the stimulation site (Kotloski et al., 2002). Finally, the severity of damage correlates with the impairment in cognitive performance (Sutula et al., 1995). Several lines of evidence suggest that brief seizures can also cause damage in the human epileptic brain. The neurochemical basis for this damage is explained by the study of During and Spencer (1993) who demonstrated a potentially neurotoxic 8-fold increase in the extracellular levels of glutamate in the hippocampus after secondarily generalized spontaneous seizures. Several studies indicate an increase in the cerebrospinal fluid (CSF) level of neuronspecific enolase (y-enolase), a marker of irreversible neuronal damage, after brief seizures (Jacobi and Reiber, 1988; Rabinowicz et al., 1994, 1996; Greffe et al., 1996; Tumani et al., 1999). Further evidence comes from a classic histologic study of MourizenDam (1982) who reported that hippocampal neuronal loss correlates with the number of generalized seizures and the duration of epilepsy. More recent histopathologic analysis of patients undergoing hippocampal resection due to drug-refractory temporal lobe epilepsy (TLE) favors the idea that both the initial insult as well as recurrent seizures contribute to the damage (Mathern et al., 1995). The study by Henshall et al. (2000a) suggests that programmed cell death contributes to ongoing neuronal damage in the brain of patients with epilepsy. In addition to neuronal damage, an 11-fold increase in microglia in the CA1 subfield and a 3-fold increase in the CA3 subfield of the hippocampus in patients with TLE undergoing surgery due to drug-refractory seizures support the idea of continuing neuronal injury due to ongoing seizure activity (Beach et al., 1995). Whether recurrent seizures cause progressive neuronal loss has also recently been addressed in crosssectional as well as follow-up magnetic resonance imaging (MRI) studies. Our cross-sectional MRI volumetry studies of patients with TLE indicated that only 5% of patients with newly diagnosed epilepsy (< 1 year of TLE) had a unilateral hippocampal volume reduction of at least 2 standard deviations (SDs) of the control mean (Salmenper~i et al., 2001). In chronic patients who experienced more than two seizures per year for more than 20 years, however, there was a volume reduction of at least 2 SDs in the
hippocampus in 50% of patients (Salmenper~i et al., 2001). The severity of the hippocampal damage correlated with the lifetime number of seizures (K~ilvi~iinen et al., 1998; Salmenper~i et al., 2001). Consistent with our observations, Van Paesschen et al. (1997) concluded that the degree of hippocampal volume loss might be associated with the number of generalized seizures. The cross-sectional MR spectroscopy study by Tasch et al. (1999) demonstrated that the Nacetyl aspartate (NAA)/creatine (Cr) ratio correlated negatively with the duration of epilepsy, and they concluded that generalized seizures might cause progressive neuronal dysfunction or loss. The follow-up study by Van Paesschen et al. (1998) demonstrated that two patients exhibited a hippocampal volume loss over a 1-year follow-up. O'Brien et al. (1999) described a case study in which a 28-year-old man with refractory TLE had progressive hippocampal atrophy of 12% over a period of 4 years. Another case study by Jackson et al. (1999) demonstrated a 47% hippocampal volume decrease over 8 months in a 23year-old male with three tonic-clonic seizures. More recently, there have been several other case studies demonstrating progressive hippocampal damage in patients with epilepsy (see Sutula and Pitk~inen, 2001). Taken together, data accumulating from histologic analyses of brains from kindled rats support the idea that brief seizures can cause neuronal loss. Imaging studies indicate that spontaneous brief seizures lasting less than 2 min can also cause progressive damage in humans, at least in some individuals. In kindling, however, the seizures are induced electrically and the data from humans are based on volumetric measurement, which is an indirect indicator of cellular death. Also, no direct evidence is available on spontaneous seizure models in animals. Therefore, it has remained under dispute whether spontaneous seizures can cause neuronal loss. In retrospective human studies it has also been difficult to differentiate the contribution of the underlying epileptogenic insult versus recurrent seizures to overall damage. The present study was designed to address the question of whether recurrent spontaneous seizures cause progressive neuronal damage in the brain using MRI and histologic techniques. We used an experimental model of TLE in which epileptogenesis is triggered by electrically induced status epilepticus
69 (SE). In this model, a subpopulation of rats develops epilepsy with frequent spontaneous seizures (mean seizure frequency up to 31 seizures per day; Nissinen et al., 2000). Spontaneous seizures were monitored using a continuous video-EEG recording system. According to our hypothesis, the lifetime number of spontaneous seizures should correlate with the severity of neuronal loss. Further, rats with frequent seizures should have signs of ongoing brain damage.
after more than 50 HAFDs (n = 8). Rats were monitored with video-EEG (every other day, 24 h/day) for 60 days, and thereafter, perfused for histology. In addition, there were 14 electrode-implanted unstimulated controls. In 'the chronic group', self-sustained SE was induced in 16 rats. The rats were monitored with videoEEG for 6 months (24 h/day, every other day). In addition, there were 8 electrode-implanted controls.
Materials and methods
Implantation of electrodes
Animals
The animals were anesthetized with intraperitoneal injection of sodium pentobarbital (60 mg/kg) and chloral hydrate (100 mg/kg) and placed in a stereotaxic frame. A pair of stimulation electrodes (0.5 m m vertical tip separation) was implanted into the lateral nucleus of the left amygdala (3.6 m m posterior, 5.0 m m lateral, and 6.5 m m ventral to bregma). One stainless-steel screw was inserted as a cortical electrode into the skull above the right frontal cortex (3 mm anterior to bregma, 2 m m lateral to midline). Two stainless-steel screws were inserted as indifferent and ground electrodes into the skull bilaterally over the cerebellum. The electrodes were fixed with dental acrylate. The rats were allowed to recover from the surgical operation for 14 days until electrical stimulation was started.
Male Sprague-Dawley rats (275-325 g) were used in the present study. After implantation of electrodes, rats were housed in individual cages at a temperature of 19-21°C, with humidity maintained at 50 to 60% and lights on from 0700 to 1900. Standard food pellets and water were freely available. All animal procedures were conducted in accordance with the guidelines set by the European Community Council Directives 86/609/EEC. To mimic the process of human TLE, three rat groups were included in the study: 'newly diagnosed', 'recently diagnosed', and 'chronic' rats with epilepsy. In 'the newly diagnosed group', selfsustained SE was induced by electrically stimulating the lateral nucleus of the amygdala in 16 adult male Sprague-Dawley rats. The animals were allowed to recover from SE spontaneously. Nonstimulated electrode-implanted rats (n = 16) served as controls. Rats were monitored with video-EEG continuously 24 h/day via cortical electrode until they developed a second spontaneous seizure and 11 days thereafter, before being perfused for histology. In 'the recently diagnosed group', self-sustained SE was induced in 30 rats. To investigate the effect of the duration of SE on neuronal damage, SE was stopped with diazepam (15 m g / k g and 10 mg/kg 8 h later, intraperitoneally) after the rats experienced 5 or fewer high-amplitude and high-frequency discharges ( H A F D s 1; n = 7), after 6 to 50 HAFDs (n = 15), or
I A conspicuous feature of EEG activity during SE is the occurrence of high-amplitude and high-frequency discharges (HAFDs), which are typically associated with behavioral
Induction of self-sustained SE Afterdischarge threshold was assessed by stimulating the amygdala with a 1-s train of 60 Hz, 50 to 400 IxA (peak to peak), 1-ms bipolar square-wave pulses. Only those rats in which afterdischarges could be induced at a current level of 400 txA or lower were included in the study. To induce SE, the amygdala was stimulated with 2 trains/s of 1-ms, 60-Hz bipolar pulses (each train lasting 100 ms) at 400 lxA current (peak to peak). After 20 rain of continuous stimulation, the stimulation was interrupted, and the behavioral and electrographic seizure activity of the animal was observed
seizures. HAFDs are defined as a high-amplitude (>2 x baseline) and high-frequency (>8 Hz) discharge in the amygdala or in the cortex (or both) that lasts for at least 5 s.
70 for 60 s. If the behavior of the animal revealed the presence of epileptic activity (head nodding/or limb clonus), the observation period was extended up to 10 min. If the animal did not meet the criterion of clonic SE (continuous epileptiform spiking and recurrent clonic seizures), stimulation was resumed and the behavior of the animal was checked again after 5 min. Once the criterion of SE was achieved, no further stimulation was given.
Electrophysiologic characterization of SE and spontaneous seizures To electrophysiologically characterize the duration and severity of SE, seizure activity was recorded by a digital video-EEG system as described previously in detail (Nissinen et al., 2000). The severity of SE was assessed by counting the number of HAFD during SE. The duration of SE was defined as an interval between the first and last HAFD. The occurrence of spontaneous seizures was determined based on the analysis of EEG data. If an electrographic seizure was observed, the severity of the behavioral seizure was scored based on the video. Criterion for an epileptic seizure was a highfrequency (>5 Hz), high-amplitude (>2 x baseline) discharge either in the amygdala or in the cortex (or both) that lasted for at least 5 s.
Characterization of behavioral seizures During the stimulation and follow-up periods, the behavior of the rats was recorded using a video camera that was time-locked with the digital EEG. Behavioral motor seizure activity was classified according to a slightly modified Racine scale (Racine, 1972). Score 0: electrographic seizure without any detectable motor manifestation. Score 1: mouth and face clonus, head nodding. Score 2: clonic jerks of one forelimb. Score 3: bilateral forelimb clonus. Score 4: forelimb clonus and rearing. Score 5: forelimb clonus with rearing and falling.
Histologic analysis of brain tissue Fixation Rats were deeply anesthetized and perfused intra-
cardially according to the fixation protocol for Timm staining (Sloviter, 1982). The brains were postfixed in the fixative for 4 h and then cryoprotected in a solution containing 20% glycerol in 0.02 M potassium phosphate buffer, pH 7.4 for 24 h. The brains were then blocked, frozen in dry ice, and stored at - 7 0 ° C until cut. They were sectioned in the coronal plane in a one-in-five series at a thickness of 30 Ixm (newly diagnosed, recently diagnosed groups) or 50 txm (chronic group) with a sliding microtome. The sections were stored in a cryoprotectant tissuecollecting solution (TCS; 30% ethylene glycol, 25% glycerol in 0.05 M sodium phosphate buffer) at - 2 0 ° C until processing. Assessment of neuronal damage in thionin staining One series of sections was stained with thionin to characterize the cytoarchitectonic boundaries of various brain regions and to locate the electrode tips. These sections were also used to assess the severity of neuronal damage by counting cells. The hippocampus was partitioned into different regions according to the nomenclature described by Amaral and Witter (1995).
Stereologic estimation of total neuronal numbers in the hilus. Stereologic analyses were conducted blindly with respect to the treatment status of the animal. The optical fractionator method was implemented using Stereo Investigator software in a Neuro Lucida morphometry system (MicroBrightField, Germany) with guidelines described by West et al. (1991). A color video camera (Hitachi HVC20, Japan), interfaced with an Olympus BX50 microscope, was used to view sections on a highresolution monitor, and neuroanatomic borders of the hilus were digitized under low-power magnification. Subsequent cell counting was confined within these borders. The sections were inspected according to a systematic random sampling scheme such that counts were derived from a known and representative fraction of the hilus. Specifically, the motorized stage of the microscope system was under computer control, and the hilar fields in every histologic section were surveyed at evenly spaced x-y intervals of 180 x 180 txm for the hilus. For each x-y step, cell counts were derived from a known fraction of
71 the total area using an unbiased counting frame that was 36 x 36 txm. Counting was performed throughout the section, avoiding the neurons that were in focus at the surface of the section. Neuronal nuclei were counted only as they first came into focus within each optical dissector. Glia, identified by size and cytologic characteristics, were excluded from the counts. Finally, the total neuron number was estimated by multiplying the sum of the neurons counted by the reciprocal of the fraction of the hippocampus that was sampled (i.e., a multiple of the fraction of the histologic sections examined, the fraction of the x-y step interval covered by the counting frame, and the fraction of the total section thickness examined). Assessment of neuronal density in the septal and temporal hilus. Because the severity of hilar cell damage varies along the septotemporal axis of the hippocampus, the density of hilar cells was estimated from the septal and temporal ends of the hippocampus separately. At the septal end, the neuronal density was assessed from three sections that were systematically sampled at 450-txm (newly diagnosed and recently diagnosed animals) or 200-~tm intervals (chronic animals) starting at the level at which the suprapyramidal and infrapyramidal blades of the granule cell layer form a continuous band of cells. At the temporal end, the neuronal density was assessed from two adjacent sections (150 txm apart) at the level where the granule cell layer forms an oval shape. Using Stereo Investigator software in the Neuro Lucida morphometry system described above, hilar fields in every histologic section were surveyed at evenly spaced x-y intervals (70 x 70 ~tm septally, 150 × 150 Ixm temporally). For each x-y step, cell counts were derived from a known fraction of the total area using a counting frame that was 25 × 25 Ixm septally and 30 x 30 Ixm temporally. Because the area of the hilus in the epilepsy groups did not differ from that in controls, neuronal density was calculated by dividing the neuronal number by the area. The mean density was used for the statistical analysis. Analyses were conducted blindly with respect to the treatment status of the animal. To compare the severity of damage between the groups, the severity of damage was transformed to damage-%, that is (density of neurons in experimental animal/mean density of neurons in the control group) x 100%.
Assessment of neuronal damage in Fluoro-Jade B preparations An adjacent series of sections (1 in 10) was stained with Fluoro-Jade B according to the protocol described by Schmued et al. (1997). The density of Fluoro-Jade B-positive neurons was semiquantitatively assessed as follows (Fig. 3A). Score 0: no damage. Score 1: lesions involving less than 20% of neurons in the region of interest. Score 2: lesions involving 21 to 50% of neurons. Score 3: lesions involving 51 to 100% of neurons. The damage score was based on the analysis of all sections in which the region of interest was present. To visualize the distribution of Fluoro-Jade Bpositive cells, fluorescent neurons were plotted using a fluorescent microscope, Leitz DMRD (Leitz, Wetzlar, Germany) equipped with a computer-aided digitizing system (Minnesota Datametrics, St. Paul, MN). To determine the location of plotted fluorescent neurons, cytoarchitectonic borders of the region of interest were drawn using a camera lucida from adjacent thionin-stained sections with a stereo microscope and a drawing tube. Thereafter, cytoarchitectonic outlines were superimposed on computergenerated plots using Canvas 3.5 (Deneba, Miami, FL) software on a Macintosh computer. MRI follow-up of rats during epileptogenesis and epilepsy To follow the progression of structural alterations in individual rats, a separate group of rats (n = 10) was prepared for MRI analysis. In these animals, a bipolar stainless-steel stimulating electrode was implanted into the left amygdala. To detect spontaneous seizures, three platinum-iridium wires were implanted into the skull as described in the section 'Implantation of electrodes'. To improve the quality of MRI, the stimulating electrode was removed 2 days after the induction of SE. To monitor the spontaneous seizures, video-EEG was recorded via a cortical platinum electrode for 4 days before each imaging session. Three rats with electrode implantations without stimulation served as controls. MRI of the amygdala and the septal hippocampus was performed using a scanner operating at 4.7 T (a bird-cage-type volume-coil, coronal 1 mm slice
72 - 2.8 mm from bregma, matrix 128 • 256, FOV 35 mm) at 2, 9, 21 and 50 days after SE. Diffusion (Dav, TR = 1.5 s, TE -----55 ms, b-values 0, 470, 856 s/mm 2) and T2 (TE = 20 to 60 ms, TR = 1.5 s, 4 averages) images were sequentially acquired. Tip was quantified at the same time points (variable length (10-90 ms) adiabatic spin-lock pulses followed by a fast spin echo imaging sequence, TR = 2.5 s, echo spacing 10 ms, 16 echoes/excitation, 4 averages). Regions analyzed included the amygdala (stimulation site, primary focus for spontaneous seizures) and the septal hippocampus (remote area connected polysynaptically with the primary stimulation site).
Results
crease in the number of hilar neurons contralaterally in epileptic animals compared to controls (44,600 4- 6021 vs 37,369 4- 9337). Neuron counts did not differ between the ipsilateral and contralateral sides. The severity of hilar damage varies along the septotemporal axis of the hippocampus, being more severe temporally than septally (Nissinen et al., 2000). During the course of these studies it has become apparent that even substantial damage involving a small segment of the septotemporal extent of the hilus might lead to a relatively small decline in the total neuronal numbers. Therefore, we assessed the association between hilar cell loss and lifetime seizure number in the septal and temporal ends separately. In the ipsilateral septal hilus, the density of neurons in newly diagnosed, recently diagnosed, and chronic animals was reduced by 23 4- 17%, 25 4- 14%, and 12 + 9%, respectively, compared to that in controls (all, p < 0.001; Fig. 2A). In the contralateral septal hilus, the density of neurons in the newly diagnosed, recently diagnosed, and chronic animals was reduced by 20q- 17% (p < 0.001), 15-t- 13% (p < 0.001), and 144- 12% (p < 0.01), respectively, compared to that in controls (p > 0.05, p < 0.05, p > 0.05 compared to ipsilateral side). In the ipsilateral temporal hilus, the neuronal density in newly diagnosed rats was reduced by 38 + 16% and in recently diagnosed rats by 30-4-17% compared to that in controls (both, p < 0.001). In the contralateral temporal hilus, the neuronal density in newly diagnosed animals was 36 4- 24% and in recently diagnosed rats 31 5:23% lower than in controls (both p < 0.001; Fig. 3A). The severity of damage did not differ between the ipsilateral and contralateral sides.
Severity of neuronal damage in the hilus at different stages of the epileptic process
Association of lifetime seizure number with neuronal numbers and densities
According to our hypothesis, the neuronal numbers would be lowest in rats with chronic epilepsy. Therefore, we first estimated the total neuronal numbers in the hilus of the chronic group by stereologic cell counting. The number of neurons in the ipsilateral hilus did not differ between the controls (n = 8) and epileptic (n = 13) animals (44,650 q- 3923 vs 40,3544-9992; Fig. 1A). There was also no de-
If recurrent spontaneous seizures cause further damage to the hilar cells, we hypothesized that there would be a correlation between the severity of hilar damage and lifetime seizure number. In chronically epileptic animals, the total number of lifetime seizures did not correlate with the total number of neurons in the ipsilateral (Fig. 1B) or contralateral hilus (both, p > 0.05). When the association be-
Photomicrography Photomicrographs were taken with a Leica DM RB microscope equipped with a CoolSnap digital camera system (RS Photometrics, Sweden). Statistical analysis Data were analyzed using SPSS for Windows (version 9.0) or Statview 4.5 for Macintosh. MRI data were analyzed using Student's t-test. The severity of neuronal damage in different temporal lobe structures in the different treatment groups was compared using the Mann-Whitney U-test. The severity of neuronal damage between the stimulated and contralateral sides was compared using the Wilcoxon signed rank test. Correlations between hilar cell damage and seizure frequency were analyzed using Spearman's rank correlation. A p-value of less than 0.05 was considered significant.
73
B
60000
50000
A6OOOO
40000 30000
50000 ..................... i ..................... I ........................
20000
Chronic epilepsy
10000 O9
40000 .............................................................................. • |
R.S.
0
2000
Z
O II rr" 30000 ................................................................................ LU
4000
6000
LIFETIME Sz NUMBER
Z
20000 ................................................................................ 1 0 0 0 0 ........ 44 650 ].......................................
C
60000 50000 40000
................................................. . .............
.....
30000 [ 40~ 0
'
Control
] '
20000 10000
Chronic epilepsy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
n.s.
Chronic
Epilepsy
0
50
100
150
200
DURATION OF EPILEPSY (d) Fig. 1. (A) Total neuronal number in the ipsilateral hilus was not decreased in chronically epileptic animals compared to that in controls (6 months after SE). (B) The number of hilar neurons did not correlate with the lifetime seizure number in chronically epileptic animals. Note that some animals with a very high seizure number have normal hilar cell counts (arrows). (C) The number of hilar neurons did not correlate with the duration of epilepsy in chronically epileptic animals. Note that some animals with the longest duration of epilepsy have normal hilar cell counts (arrows). Abbreviations: d = days; n.s. = statistically nonsignificant; Sz = seizure.
tween lifetime seizure number and damage to hilar cells at the septal and temporal ends was assessed by combining the data from the different animal groups, the lifetime seizure number did not correlate with neuronal density in the hilus ipsilaterally, but did correlate contralaterally (r = 0.335, p < 0.05, n = 37) at the septal end (data available from newly diagnosed, recently diagnosed, and chronic animals) (Fig. 2B). At the temporal end (data available from newly diagnosed and recently diagnosed animals), however, there was a correlation both ipsilaterally (r = 0.447, p < 0.05, n = 24; Fig. 3B) and contralaterally (r = 0.605, p < 0.01, n = 24). It should be noted, however, that some animals with a very low lifetime seizure number had damage as severe as the animals with a large number of seizures (Fig. 3B).
Association of seizure frequency with neuronal numbers and densities Exposure of neurons to the toxic effects of glutamate presumably depends on the seizure frequency, and therefore, the association of seizure frequency with neuronal loss was assessed. There was no correlation between seizure frequency and total neuronal numbers in the hilus. Ipsilaterally seizure frequency correlated with the severity of damage in the temporal hilus (r = 0.498, p < 0.05, n = 23) and contralaterally, both septally (r = 0.365, p < 0.05, n = 36) and temporally (r = 0.612, p < 0.01, n = 23). Analysis of data from individual animals, however, indicates that some rats with a very low seizure frequency had substantial ( > 3 0 % ) hilar cell damage. In contrast, some animals with a mean daily seizure frequency of more than 15 had the same magnitude of damage as rats with very low seizure frequency.
74
B 100
/x. ._J
100
o~ 8O
......................l ....................: ...................z ........................
~ z 0rr
..............................
~
2o
z
0
I
•
.0.
............. i ................... i .......................
60 40
.iiit!i.ii.i...i_iiii..i.ii.i.. .....,
............................................................................ 1J ~n.s. " L~'""t""l;~
500 1500 2500 3500 4500 5500 6500
W _J
o~ 60 z O rr w z
I
..................................................................
LIFETIME Sz NUMBER
0
40
o~ 100 h o t -28%
..............o , ,
.............• .....................t ~
. . . . . . . . . . . . .
8o
20
6o
co z 40 '
'
'
Chronic Recent New
~
20
D w z
0
,;:;: ::; ; ;: :i; t ,anmaSns 0
50
100
150
200
DURATION OF EPILEPSY (d) Fig. 2. (A) Density of neurons in the ipsilateral septal hilus (expressed as a percentage of neurons remaining compared to controls; 100%) was decreased both in the newly diagnosed, recently diagnosed, and chronically epileptic animals (all, p < 0.001). (B) The density of neurons in the ipsilateral septal hilus did not correlate with the lifetime seizure number when data from all animal groups were combined. Note that some animals with a very high seizure number have normal hilar cell counts (arrows). (C) The density of neurons in the ipsilateral septal hilus did not correlate with the duration of epilepsy when data from all animal groups were combined. Note that some animals with the longest duration of epilepsy have normal hilar cell counts (arrows). Abbreviations: d = days; n.s. = statistically nonsignificant; Sz = seizure.
Association of duration of epilepsy with neuronal numbers and densities
Association of duration of SE with neuronal numbers and density
Longer duration of epilepsy was assumed to associate with more severe damage caused by both the progression of degenerative processes initiated by SE and recurrent seizures. In the chronic group, there was no correlation between the duration of epilepsy and the neuronal numbers in the hilus (Fig. 1C). Also, the duration of epilepsy did not correlate with neuronal density in the ipsilateral or contralateral septal or temporal hilus (ipsilateral septal in Fig. 2C and ipsilateral temporal in Fig. 3C).
SE is a brain damaging insult, the duration of which is associated with the severity of neuronal damage (Meldrum and Brierley, 1973). We retrospectively assessed the association of the duration of SE with cell counts. Interestingly, the duration of SE did not correlate with the decrease in the total neuronal numbers in the hilus (chronic group). Also, there was no association between the duration of SE or the number of HAFDs and the density of neurons in the septal or temporal hilus (all animals; Fig. 4).
75
~°100
A 100
80 t-
I.I_ UJ
- 60 co z O
~
60
o
4o 1 All animals 2o .............................................................................
UJ
z
| ................................................................t • |
I p=0.033
0
40
90
140
LIFETIME SZ NUMBER
................................
C
g 4o
o~
IJ.I
~_ 100
z
20
._1
0
Recent
New
~z 0 D uJ z
80
60 .... tt- ................................... • ' ' ° , ............................ 40 ...............°.~. .................................................. All animals 20 i n.s. 0 ......................................................................... 0 20 40 60 DURATION OF EPILEPSY (d)
Fig. 3. (A) Density of hilar neurons (expressed as a percentage of neurons remaining compared to controls; 100%) was decreased at the temporal end of the dentate gyrus both in the newly diagnosed and recently diagnosed (both p < 0.001) epileptic animals. (B) The density of neurons in the ipsilateral temporal hilus correlated with the lifetime seizure number when data from newly and recently diagnosed animal groups were combined. Note, however, that many of the animals with only very few seizures (arrows) had hilar cell loss as severe as the rats with frequent seizures. (C) The density of neurons in the ipsilateral temporal hilus did not correlate with the duration of epilepsy when data from newly and recently diagnosed animal groups were combined (Spearman's rank correlation). Note that many of the animals with a very short duration of epilepsy (arrows) had as severe damage as the rats with the longest duration of epilepsy. Abbreviations: d = days; n.s. = statistically nonsignificant; Sz = seizure.
MRI of brain during SE-induced epileptogenesis and epilepsy Histologic analysis provides insight into only one time point, which compromises the analysis of the temporal course of damage in individual animals. Therefore, we used modem imaging techniques to follow the progression of damage after SE (Fig. 5). At 2 days, T2 was prolonged in the ipsilateral amygdala (by 162%, p < 0.01) and in the hippocampus (127%) compared to electrode-implanted controls (is 100%). At 9, 23 and 50 days, T2 remained elevated but tended to decrease in the ipsilateral amygdala (to 117% at 50 days) as well as in the ipsilateral hippocampus (to 114%). Changes in T10 paralleled
the changes in T2 even though the abnormalities were more pronounced and Tl0 remained elevated even at day 50 (Fig. 5). At 2 days, D,v in the ipsilateral amygdala did not differ from controls, whereas it was increased in the ipsilateral hippocampus (127%). At later time points, Day was elevated in both the amygdala and the hippocampus. MRI parameters during the first 23 days did not predict the severity of epilepsy at later time points.
Assessment of Fluoro-Jade B staining at different stages of epilepsy Fluoro-Jade B stains irreversibly damaged neurons (electron microscopic observations by Miettinen and
76 SEPTAL HILUS
100
....~,~...._•...•
.................................................
. ,......:'....
u4, 80
~ •
o ............... •
,ip..~
100
..............
- ......................................
. . . . . . . . . . . .
m 80 ~
60
O or 40 D 20 Iii z 0
0 rr
40
i iiiiiiiiiiiiiiiiiiiiiiiiiiiii'iiiiiiiiiiiiiiiiiiiiii "nsan,n a,s 0
500 1000 DURATION OF SE (min)
. . . . . . . . . . . . . . . . . . . . . . .
O.._l.Z.l.._.o__i -
......... ~ . O . 0
............... a •
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6o
~z
............,
__I~I
.................. ~
_ ~ O
...................................
..............................
.O. ................................
............................ • ......................................
[ p=0.037
LU 20 Z 0
1500
•
0
50
100 HAFDs
150
200
TEMPORAL HILUS o~ ii UJ
G0
Z
O
fr D L.U Z
C 100 ..o----..--
.......................................................................................................
80 ~
............................. i-".......................................................................
60 r ~
................go .......................................................................... _ _ 40 I-- ........................................o ........................I All animals ...................................................................... [n.s. 0-0 500 1000 1500 DURATION OF SE (min)
I ~ 1 0 0 l.............l ........................l .tu
80
l
..............................................................................
........................... i i - - - ~ ..............o ............................................................
COZ 60 ",," 0 • D rr 40 iiiiiiiiiiii iiii iiii iiii i°iiiill iiii iiiiiiiii t All animals UJ z n.s. 0 0 50 100 150 200 HAFDs
Fig. 4. Association of the severity of SE with neuronal damage in the ipsilateral septal (A and B) and temporal (C and D) hilus. (A) The severity of cell loss (expressed as a percentage of neurons remaining compared to controls; 100%) did not correlate with the duration of SE (data pooled from all three rat groups). Note that some of the rats with the longest SE (arrows) had normal hilar cell densities. (B) The severity of cell loss correlated with the number of high-amplitude and high-frequency discharges during SE. The association, however, suggested that animals with a higher number of seizures during SE had a higher density of hilar cells. At the temporal end, there was no association between hilar cell loss and duration of status epilepticus (C) or the number of high-amplitude and high-frequency discharges during SE (D). Abbreviations: HAFD = high-amplitude and -frequency discharge; rain = minutes; n.s. = statistically nonsignificant; SE = status epilepticus.
Pitk~inen, unpublished). In the present study, FluoroJade B-positive neurons were assessed in the newly diagnosed and chronic groups (Fig. 6). None of the newly diagnosed rats had positive neurons in the hilus, whereas 10/14 had Fluoro-Jade B-positive neurons in the CA3, 9/14 in the CA1, 14/14 in the amygdala, 3/14 in the piriform cortex, and 11/14 in the thalamus. Also, one of the two rats that was stimulated but did not develop epilepsy had a few Fluoro-Jade B-positive neurons in the amygdala and thalamus, and the other rat had a few in just the amygdala. In the newly diagnosed group, the density of Fluoro-Jade B-positive neurons corre-
lated negatively with the time from SE to perfusion in the CA3 (r = -0.488, p < 0.05), the perirhinal cortex (r = - 0 . 2 8 4 , p < 0.05), and the thalamus (r = -0.286, p < 0.05). There were no Fluoro-Jade B-positive cells in any of the animals in the chronic group. When the data from newly diagnosed and chronic groups were pooled, there was no correlation between the lifetime seizure number and hippocampal total damage score (Fig. 7A). Time to perfusion, however, correlated with the total hippocampal damage score (p < 0.001; Fig. 7B). Approximately 60 days after SE, there were no positive neurons in any
77
HIPPOCAMPUS
AMYGDALA
B
A
180 160 140 120
•-.L
180 160 140 120
seizures
........................................................................
...................................................................................................................................
...... 3.x .................................
~..I--.-.~...**.-.ii....=.......
.
i i!!!!!!
80
80 60
Lseizures
........................................................................
I.
...............................................................................................• -
l_._ Day
40 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
20 I
0
I
10
I
t
I
40 20
1- I - Tip
............................................................................
0
60
T 2
1
I
I
20 30 40 FOLLOW-UP (d)
I
I
50
.................................................................................................. iiiiiii ii iiiiiiiiii ii iiiiiiiiii ii iiiiiiii ii "O'T2~ T1 °Dav
0
I
60
I
0
I
10
I
I
I
I
I
I
20 30 40 FOLLOW-UP (d)
I
I
50
I
60
Fig. 5. Long-term magnetic resonance imaging follow-up of rats after SE. (A) In the ipsilateral amygdala, T2 and Tip were increased 2 days after SE (p < 0.01), returned to normal levels during the next 2 weeks, and thereafter tended to increase again. The diffusion coefficient, Day, became progressively more abnormal during the first 3 weeks. Importantly, all parameters were above the control level even 50 days after SE. (B) Unlike in the amygdala, there were no rapid increases in the T2 and Tip values in the ipsilateral septal hippocampus after SE. Instead, all parameters (T2, Tip, Day) increased progressively during the first 3 weeks after SE and remained above the control level even at 50 days. Values are expressed as a percentage of that in controls animals (control value is defined as 100%). Abbreviations: d = days; Day = 1/3 of the trace of the diffusion tensor; TLp = longitudinal relaxation time in the rotating frame; T2 = transverse relaxation time. of the rats despite a high frequency of daily seizures ( > 1 0 seizures/day) in some chronically epileptic animals (Fig. 7B).
Discussion Methodologic considerations When comparing data obtained from animal studies to those obtained from human studies, a critical question arises. What does the animal model model? In the present experiments, epileptogenesis was induced by electrically stimulating the amygdala in adult male rats. This results in the development of self-sustained SE and the appearance of spontaneous seizures of temporal lobe origin after a latency period of approximately 1 month. In some animals seizures might occur tens of times per day (Nissinen et al., 2000). We propose that our model models TLE in humans that has an adult onset with acute nonsymptomatic SE as an etiology. It is important to realize, however, that the lack of data on the development of brain damage in spontaneous seizure
models triggered by other etiologies, such as head trauma or stroke, compromises the extrapolation of data obtained in SE models to sequelae of other etiologies. In the present study, SE was induced by electrical stimulation of the amygdala, which is also the site of origin for most of the spontaneous seizures occurring later in life (Nissinen et al., 2000). The amygdala provides substantial inputs to the CA1 and CA3 subfields of the temporal half of the hippocampus (Pikkarainen et al., 1999). Projections to the septal hippocampus are meager, and there are no monosynaptic projections to the hilus (Pikkarainen et al., 1999). With these caveats in mind, we assumed that the polysynaptically connected hippocampus would provide a good candidate region to study seizureinduced neuronal death that would not be contaminated by monosynaptic neuronal death associated with the SE-triggering stimulation itself. In the present study, neuronal loss was assessed by estimating the total neuronal numbers in the hilus and measuring neuronal density in the septal and temporal hilus. These approaches have a limited sen-
78
~--;
~..~
~:x8 ~t~
~
o
..~ ~ ~ ~ "
~-~ ~ o~-~
.~'~ t.~
~ ~ =
~ - ~
~.~
0
"0
,~
02
O~
c~
0
."
"~
-= ~
~.~ ~
'
:~
"
II
~
~:
~
~ -~
a.~ o ~ I e . ~
o
~ . ~
•~ 5
• .0
oo~
~
~
o'd ,
~
0
.~ ~-~ ~
. ~ ~ ~ ~ o ~
79
A uJ nO u~ m
8 7 6 5 4 3
B ......... • ........................
Newly
..............a .......................
chronic
..............~ t .......................
n.s.
diagnosed
no o
animals
1 0
54
.........• .............................
Newly
,,~, ~ ' ~ ; ua " n o s e
.............w
chronic
animals
.............. a .................................................. ~ . . . ~ ¢ . . . . ~ ' . ................... ~ .............Q l
0
......
...................................... -Q....Q---.-Q .................... Q ...........
2000
4000
6000
LIFETIME SEIZURE NUMBER
"?, ,,
.......................
..............a .........................
.........Q
2 "
8
and
'~ a n d
p<0.001
...............................................................................................................
1 0 0
50
100
150
200
TIME TO PERFUSION (d)
Fig. 7. (A) Number of Fluoro-Jade B-positive cells in the hippocampus was not associated with the lifetime seizure number (the correlation was also not significant if calculated separately for partial or secondarily generalized seizures in the newly diagnosed group; data not shown). Note that several animals with a very high lifetime seizure number (and high daily seizure frequency; arrows) had no Fluoro-Jade B-positive cells in the hippocampus. (B) Number of Fluoro-Jade B-positive cells correlated inversely with the time to perfusion. Note that there were no positive cells in animals that were sacrificed 180 days after SE (13 rats, arrow) even though some of the animals had several spontaneous seizures per day. Abbreviations: d = day; F-J B = Fluoro-Jade B.
sitivity to detect minor loss of a subpopulation of vulnerable neurons (Tuunanen et al., 1997; Tuunanen and Pitk~inen, 2000). Further, there is substantial interanimal variability in the severity of damage caused by SE. Seizure and SE-induced neurogenesis is another factor confounding the assessment of neuronal damage by retrospective cell counting (Parent et al., 1997). In fact, the reconstitutive effect of neurogenesis on hilar cell counts might be the most remarkable in frequently seizing animals (Scott et al., 1998, 2000; Madsen et al., 2000; Nagawa et al., 2000). Another tool used to assess ongoing neuronal damage was Fluoro-Jade B staining, which has been reported to label irreversibly damaged neurons (Schmued et al., 1997). This assumption is based on the demonstration of Fluoro-Jade B-positive cells in sections adjacent to those showing acid-fuchsine (Schmued et al., 1997) or silver-positive (Kubov~i et al., 2001) neurons in the same brain areas. There are, however, no previous studies that assessed the ultrastructure of Fluoro-Jade B-positive cells. We recently investigated Fluoro-Jade B-positive neurons in rats that experienced kainate-induced SE using electron microscopy. The preliminary data show that all Fluoro-Jade B neurons in the hippocampus or the thalamus have an ultrastructure of irreversible damage (Pitk~inen and Miettinen, unpublished). Therefore, we conclude that all Fluoro-Jade B-positive
cells in the present study are dying neurons, not just cells undergoing temporary dysfunction.
Time course of progression of amygdaloid and hippocampal damage after SE The MRI follow-up of rats after SE provided evidence that structural alterations continue to progressively increase for approximately 3 weeks after SE. At the stimulation site in the amygdala, the T2 and T] 0 elevations were most prominent 2 days after SE, declining towards normal values within the next week, and again becoming more abnormal during the third week after SE. In the polysynaptically connected septal hippocampus, there was no such peak elevation at 2 days. Rather, there was a more linear increase in MRI parameters analyzed during the first 3 weeks following SE. Interestingly, both at the stimulation site and in the septal hippocampus, MRI parameters tended to remain elevated even 50 days after SE. A critical question remains as to whether the abnormalities in MRI at later follow-up points when the animal is seizing spontaneously indicate ongoing neuronal loss. The present data indicate a dynamic pattern of both T2 and T10 in the ipsilateral amygdala after electrical stimulation and consequent SE. Substantial recovery of the relaxation times in the amygdala is reconciled as reinstated water home-
80 ostasis following seizures (see Ebisu et al., 1996). Prolonged T2 is a well-documented indicator of irreversible brain damage, and thus, the time-dependent increase in T2, particularly when accompanied by an increased T10, is considered a sign of irreversible neural damage. The T10 contrast serves as a sensitive MRI indicator of irreversible neuronal damage during reperfusion following acute cerebral ischemia in the rat (Grrhn et al., 1999). The present data add a new dimension to this picture, demonstrating that T~ 0 may recover following prolonged seizures. Finally, diffusion is a well-established MRI index of acute ischemia (Moseley et al., 1990), epileptic seizure (Zhong et al., 1993), and brain infarction (Welch et al., 1986, 1995). Acute ischemia and seizure cause reduced diffusion, whereas infarcted tissue has elevated water diffusivity. The present data indicate elevated diffusivity in the amygdala after 4 weeks or so; diffusion increased much earlier in the hippocampus, suggesting a differential time course of neural cell damage in the brain structures. Taken together, the MRI data obtained suggest that during the first 3 weeks after SE, there is a progressive increase in structural damage in the amygdala and the hippocampus. At later time points, milder abnormalities in T2, T10, and Dav are associated with ongoing neuronal cell death, which is consistent with the presence of Fluoro-Jade B-positive neurons in the amygdala and hippocampus 50 days after SE.
Contribution of spontaneous seizures to hilar cell damage
shorter duration of spontaneous seizures compared to that of induced seizures in the kindling model explains the difference in the severity of seizureinduced damage. We had a few chronically seizing animals with secondarily generalized seizures but even in these animals we could not find any FluoroJade B-positive ceils in the hilus or brain areas at 6 months after SE. Also, when the correlation analysis was performed separately between the severity of damage assessed by cell counting versus partial or secondarily generalized seizures, no association was found. The timing of sacrifice could have affected our ability to detect neuronal damage with Fluoro-Jade B, particularly in rarely seizing animals. We had, however, several animals with frequent seizures each day. Therefore, we consider it unlikely that we missed damaged neurons because of the time interval between the seizure and sacrifice. A more extensive analysis of neuronal cell counts in brain areas other than the hilus, including regions monosynaptically linked with the primary seizure focus, might have provided a more thorough view to seizure-induced neuronal damage. The lack of Fluoro-Jade B-positive neurons in regions receiving monosynaptic inputs from the lateral nucleus of the amygdala (e.g., other amygdaloid nuclei, layer III of the entorhinal cortex, perirhinal cortex; see Pitk~inen, 2000) in chronically epileptic animals with several seizures per day argues against ongoing damage caused by seizures in these regions.
Contribution of SE to hilar cell damage There was no correlation between the number of lifetime seizures and the severity of neuronal loss in the hilus of the dentate gyrus. The longest follow-up of spontaneously seizing rats was 6 months, during which four animals had more that 2700 spontaneous seizures (Nissinen et al., 2000). Therefore, it is unlikely that infrequency of spontaneous seizures compromised our ability to detect seizure-induced damage. As shown previously, most of the behavioral seizures in frequently seizing animals are partial and are shorter than the secondarily generalized seizures (44 vs 61 s, Nissinen et al., 2000). Further, the proportion of partial seizures of all seizures increases during the course of the disease (Nissinen et al., 2000). It remains to be studied whether the
Previous studies demonstrated that hilar cells are among the most vulnerable cell types to become damaged by SE (Fujikawa, 1996). This raises a question as to whether there are any hilar cells left after SE to be exposed to the damaging effects of spontaneous seizures later in life in our model. As we show in the present experiments, only 23% (3/13) of the chronic animals had a 20% reduction in the total neuronal numbers in the ipsilateral hilus. When the neuronal density was estimated in the ipsilateral septal and temporal hilus separately, 46% (17/37) of rats had >20% damage in the septal and 79% (19/24) in the temporal hilus. Therefore, most of our epileptic animals had a substantial num-
82
account in spontaneous seizure models currently in use and remains a challenge for researchers attempting to model the human condition. References Amaral, D.G. and Witter, M.E (1995) Hippocampal formation. In: G. Paxinos (Ed.), The Rat Nervous System. Academic Press, San Diego, CA, pp. 443-493. Amaral, D.G. and Insausti, R. (1990) The human hippocampal formation. In: G. Paxinos (Ed.), The Human Nervous System. Academic Press, San Diego, CA, pp. 711-755. Beach, T.G., Woodhurst, W.B,, MacDonald, D.B. and Jones, M.W. (1995) Reactive microglia in hippocampal sclerosis associated with human temporal lobe epilepsy. Neurosci. Lett., 191 : 27-30. Bengzon, J., Kokaia, Z., Elm@, E., Nanobashvili, A., Kokaia, M. and Lindvall, O. (1997) Apoptosis and proliferation of denatate gyrus neurons after single and intermittent limbic seizures. Proc. Natl. Acad. Sci. USA, 94: 10432-10437. Callahan, P.M., Paris, J.M., Cunningham, K.A. and ShinnickGallagher, E (1991) Decrease of GABA-immunoreactive neurons in the amygdala after electrical kindling in the rat. Brain Res., 555: 335-339. Cavazos, J.E., Das, I. and Sutula, T.E (1994) Neuronal loss induced in limbic pathways by kindling: evidence for induction of hippocampal sclerosis by repeated brief seizures. J. Neurosci., 14: 3106-3121. Dalby, N.O., West, M. and Finsen, B. (1998) Hilar somatostatinmRNA containing neurons are preserved after perforant path kindling in the rat. Neurosci. Lett., 255: 45-48. During, M.J. and Spencer, D.D. (1993) Extracellular hippocampal glutamate and spontaneous seizure in the conscious human brain. Lancet, 341: 1607-1610. Ebisu, T., Rooney, W.D., Graham, S.H., Mancuso, A., Weiner, M.W. and Maudsley, A.A. (1996) MR spectroscopic imaging and diffusion-weighted MRI for early detection of kainateinduced status epilepticus in the rat. Magn. Reson. Med., 36(6): 821-828. Fujikawa, D.G. (1996) The temporal evolution of neuronal damage from pilocarpine-induced status epilepticus. Brain Res., 725(l): 11-22. Fujikawa, D.G., Shinmei, S.S. and Cai, B. (1999) Lithiumpilocarpine-induced seizures produces necrotic neurons with internucleosomal DNA fragmentation in adult rats. Eur. J. Neurosci., 1 l: 1605-1614. Fujikawa, D.G., Shinmei, S.S. and Cai, B. (2000) Kainic acidinduced seizures produce necrotic, not apoptotic, neurons with internucleosomal cleavage: implications for programmed cell death mechanisms. Neuroscience, 98( 1): 41-53. Greffe, J., Lemoine, E, Lacroix, C., Brunon, A.-M., Terra, J.-L., Dalery, J., Bernier, E., Soares-Boucaud, I., Rochet, T., Leduc, T.A., Balvay, G. and Mathieu, E (1996) Increased serum levels of neuron-specific enolase in epileptic patients and after electroconvulsive therapy - - a preliminary report. Clin. Chim. Acta, 244: 199-208.
Gr6hn, O.H.J., Lukkarinen, J.A., Silvennoinen, M.J., Pitkanen, A., van Zijl, P.C.M. and Kauppinen, R. (1999) Assignment of reversible and irreversible ischaemic cerebral damage in a rat using quantitative Tlo, T2 and the trace of the diffusion tensor magnetic resonance imaging. Magn. Reson. Med., 42: 268-276. Henshall, D.C., Clark, R.S., Adelson, ED., Chen, M., Watkins, S.C. and Simon, R.E (2000a) Alterations in bcl-2 and caspase gene family protein expression in human temporal lobe epilepsy. Neurology, 55(2): 250-257. Henshall, D.C., Chen, J. and Simon, R.E (2000b) Involvement of caspase-3-1ike protease in the mechanism of cell death following focally evoked limbic seizures. J. Neurochem., 74(3): 1215-1223. Jackson, G.D., Chambers, B.R. and Berkovic, S.E (1999) Hippocampal sclerosis: development in adult life. Dev. Neurosci., 21(3-5): 207-214. Jacobi, C. and Reiber, H. (1988) Clinical relevance of increased neuron-specific enolase concentration in cerebrospinal fluid. Clin. Chim. Acta, 177: 49-54. K~ilvi~iinen, R., Salmenper~i, T., Partanen, K., Vainio, E, Riekkinen, E and Pitkanen, A. (1998) Recurrent seizures may cause hippocampal damage in temporal lobe epilepsy. Neurology, 50: 1377-1382. Kotloski, R., Lynch, M. and Sutula, T.E (2002) Repeated brief seizures induce progressive hippocampal neuron loss and memory deficits. Prog. Brain Res., in press. Kubov~, H., Druga, R., Lukasiuk, K., Suchomelov~i, L., Haugvicoyly, R. and Pitk~inen, A. (2001) Status epilepticus causes necrotic damage in the mediodorsal nucleus of the thalamus in immature rats. J. Neurosci., 21(10): 3593-3599. Lukasiuk, K. and Pitk~inen, A. (1998) Distribution of early neuronal damage after status epilepticus in chronic model of TLE induced by amygdala stimulation. Epilepsia, 39(Suppl. 6): 9. Madsen, T.M., Treschow, A., Bengzon, J., Bolwig, T.G., Lindvail, O. and Tingstr6m, A. (2000) Increased neurogenesis in a model of electroconvulsive therapy. Biol. Psychiatry, 47: 1043-1049. Mathern, G.W., Pretorius, J.K. and Babb, T.L. (1995) Influence of the type of initial precipitating injury and at what age it occurs on course and outcome in patients with temporal lobe seizures. J. Neurosurg., 82(2): 220-227. Meldrum, B.S. and Brierley, J.B. (1973) Prolonged epileptic seizures in primates. Ischemic cell change and its relation to ictal physiological events. Arch. Neurol., 28: 10-17. Moseley, M.E., Cohen, Y., Mintorovitch, J., Chileuitt, L., Shimizu, H., Kucharczyk, J., Wendland, M.F. and Weinstein, ER. (1990) Early detection of regional cerebral ischemia in cats: comparison of diffusion- and T2-weighted MRI and spectroscopy. Magn. Reson. Med., 14(2): 330-346. Mourizen-Dam, A. (1982) Hippocampal neuron loss in epilepsy and after experimental seizures. Acta Neurol. Scand., 66: 601642. Nagawa, E., Aimi, Y., Yasuhara, O., Tooyama, I., Shimada, M., McGeer, EL. and Kimura, H. (2000) Enhancement of progenitor cell division in the dentate gyrus triggered by initial
81 ber of hilar cells remaining after SE, particularly septally. Another possibility is that the time-course of hilar cell death after SE is faster than that of hippocampal pyramidal cells. Supporting this idea, there were no Fluoro-Jade B-positive neurons in the hilus of animals that had a substantial number of hilar cells remaining in thionin preparations, whereas there were numerous Fluoro-Jade B-positive cells in other hippocampal subfields in the same section. When we assessed the neuronal numbers or densities retrospectively several months after SE, there was no clear association between the duration or severity (number of HAFDs) of SE and the severity of neuronal damage. We previously demonstrated that duration of SE is associated with the presence of damage. In the present model, SE has to continue for approximately 40 min to induce damage (Lukasiuk and Pitkanen, 1998). If SE is allowed to continue uncontrolled for several hours, the association between the duration of SE and the severity of damage, however, disappears. Previous studies using silver and TUNEL techniques as well as markers of programmed cell death, demonstrated that SE-induced neuronal loss might continue for several days (Fujikawa et al., 1999, 2000; Henshall et al., 2000b; Tuunanen and Pitkanen, 2000). Fluoro-Jade B staining can be detected in the amygdala, hippocampus, endopiriform nucleus, perirhinal cortex, and thalamus approximately 60 days after SE. The number of Fluoro-Jade B-positive cells did not, however, correlate with the number of spontaneous seizures, but correlated with the time interval from SE to perfusion. Further, there were no Fluoro-Jade B-positive neurons present in frequently seizing rats that had been perfused for histology 6 months after SE. These observations raise the question as to whether the damage even at about 2 months after SE was more related to SE rather than to the occurrence of spontaneous seizures. Therefore, the methodology used in the present experiments did not demonstrate a clear association between the number or occurrence of spontaneous seizures and the severity of hilar cell death. The hippocampal damage in our epileptic animals seemed to associate with progressive neuronal loss triggered by the underlying epileptogenic insult, that is, SE rather than with the spontaneous seizures.
Conclusions The present data did not provide evidence to support the idea that spontaneous seizures cause neuronal damage. Rather, the loss of neurons in the present model was associated with the epileptogenic insult, that is, SE. One explanation is that more sensitive methods are needed to detect the few neurons damaged by brief spontaneous seizures. As shown by Bengzon et al. (1997), a single afterdischarge in a kindling model might cause positive TUNEL labeling of only one or two cells per 20-p~m-thick section. Consequently, the total number of neurons killed by a single kindled seizure is estimated to be only a few hundred. In humans, the total number of neurons in the granule cell layer, hilus, and CA subfields of the hippocampus has been estimated to be 35.7 million (West and Gundersen, 1990). We recently demonstrated that in patients with cryptogenic TLE with no known underlying brain damaging etiology, the number of spontaneous seizures needed to cause a 50% decrease in the total volume of the hippocampus is approximately 6500 (K~ilviainen et al., 1998). If the hippocampal volume loss in MRI is assumed to derive to the same magnitude from the cellular and noncellular compartments, it is estimated that each spontaneous seizure would damage approximately 2746 neurons (17.85 million/6500). Considering that the total rostrocaudal length of the hippocampus is 4 cm (Amaral and Insausti, 1990) giving 800 50-1xm-thick sections, there would be approximately three damaged neurons (2746/800) visible in each section after seizures. These numbers correlate well with the data from a kindling model (Bengzon et al., 1997). Further studies using sensitive markers of different cell death pathways are needed to assess the presence of damage induced by spontaneous seizures. In humans, progressive seizure-induced neuronal damage is observed in a subpopulation of patients (Van Paesschen et al., 1998; see Sutula and Pitk~inen, 2001). Analysis of the clinical history of these individuals indicates that factors like genotype, gender, head trauma, environment, or medication might have contributed to the sensitivity of an individual to seizure-induced neuronal damage (Sutula and Pitk~nen, 2001). A possibility of multifactorial risk of seizure-induced neuronal damage is not taken into
83
limbic seizures in rat models of epilepsy. Epilepsia, 41(1): 10-18. Nissinen, J., Halonen, T., Koivisto, E. and Pitkanen, A. (2000) A new model of chronic temporal lobe epilepsy induced by electrical stimulation of the amygdala in rat. Epilepsy Res., 38: 177-205. O'Brien, T.J., So, E.L., Meyer, F.B., Parisi, J.E. and Jack, C. (1999) Progressive hippocampal atrophy in chronic intractable temporal lobe epilepsy. Ann. Neurol., 45: 526-529. Parent, J.M., Yu, T.W., Leibowitz, R.T., Geschwind, D.H., Sloviter, R.S. and Lowenstein, D.H. (1997) Dentate granule cell neurogenesis is increased by seizures and contributes to aberrant network reorganization in the adult rat hippocampus. J. Neurosci., 17(10): 3727-3738. Pikkarainen, M., ROnkk6, S., Savander, V., Insausti, R. and Pitk~inen, A. (1999) Projections from the lateral, basal and accessory basal nuclei of the amygdala to the hippocampal formation in rat. J. Comp. Neurol., 403: 229-260. Pitk~inen, A. (2000) Connectivity of the rat amygdaloid complex. In: J.E Aggleton (Ed.), The Amygdala: a Functional Analysis. Oxford University Press, Oxford, pp. 31-115. Pretel, S., Applegate, C.D. and Piekut, D. (1997) Apoptotic and necrotic cell death following kindling induced seizures. Acta Histochem., 99: 71-79. Rabinowicz, A.L., Correale, J.D., Couldwell, W.T. and DeGiorgio, C.M. (1994) CSF neuron-specific enolase after methohexital activation during electrocorticography. Neurology, 44: 1167-1169. Rabinowicz, A.L., Correale, J., Boutros, R.B., Couldwell, W.T., Henderson, C.W. and DeGiorgio, C.M. (1996) Neuron-specific enolase is increased after single seizures during inpatient video/EEG monitoring. Epilepsia, 37(2): 122-125. Racine, R.J. (1972) Modulation of seizure activity by electrical stimulation, II. Motor seizures. Electroencephalogr. Clin. Neurophysiol., 32: 281-294. Salmenperfi, T., K~ilvi~iinen, R., Partanen, K. and Pitkanen, A. (2001) Hippocampal and amygdaloid damage in partial epilepsy: A cross-sectional MRI study of 241 patients. Epilepsy Res., 46: 69-82. Schmued, L.C., Albertson, C. and Slikker Jr., W. (1997) FluoroJade: a novel fluorochrome for the sensitive and reliable histochemical localization of neuronal degeneration. Brain Res., 751(1): 37-46. Scott, B.W., Wang, S., Burnham iU, W.M., De Boni, U. and Wojtowicz, J.M. (1998) Kindling-induced neurogenesis in the dentate gyms of the rat. Neurosei. Lett., 248(2): 73-76. Scott, B.W., Wojtowicz, J.M. and Burnham, W.M. (2000) Neurogenesis in the dentate gyms of the rat following elctroconvulsive shock seizures. Exp. NeuroL, 165: 231-236. Sloviter, R.S. (1982) A simplified Timm stain procedure compatible with formaldehyde fixation and routine paraffin embedding of rat brain. Brain Res. Bull., 8: 771-774. Sutula, T., Lauersdorf, S., Lynch, M., Jurgella, C. and Woodard,
A. (1995) Deficits in radial arm maze performance in kindled rats: evidence for long-lasting memory dysfunction induced by repeated brief seizures. J. Neurosci., 15(12): 8295-8301. Sutula, T.P. and Pitk~inen, A. (2001) More evidence for seizureinduced neuron loss: Is hippocampal sclerosis a cause and an effect of epilepsy?. Neurology, 57: 169-170. Tasch, E., Cendes, F., Li, L.M., Dubeau, F., Andermann, E and Arnold, D.L. (1999) Neuroimaging evidence of progressive neuronal loss and dysfunction in temporal lobe epilepsy. Ann. Neurol., 45: 568-576. Tumani, H., Otto, M., Gefeller, O., Wiltfang, J., Herrendorf, G., Mogge, S. and Steinhoff, B.J. (1999) Kinetics of serum neuron-specific enolase and prolactin in patients after single epileptic seizures. Epilepsia, 40(6): 713-718. Tuunanen, J. and Pitk~inen, A. (2000) Do seizures cause neuronal damage in rat amygdala kindling?. Epilepsy Res., 39: 171176. Tuunanen, J., Halonen, T. and Pitkanen, A. (1997) Decrease in somatostatin-immunoreactive neurons in the rat amygdaloid complex in a kindling model of temporal lobe epilepsy. Epilepsy Res., 26: 315-327. Van Paesschen, W., Revesz, T., Duncan, J.S., King, M.D. and Connelly, A. (1997) Quantitative neuropathology and quantitative magnetic resonance imaging of the hippocampus in temporal lobe epilepsy. Ann. Neurol., 42(5): 756-766. Van Paesschen, W., Duncan, J.S., Stevens, J.M. and Connelly, A. (1998) Longitudinal quantitative hippocampal magnetic resonance imaging study of adults with newly diagnosed partial seizures: one-year follow-up results. Epilepsia, 39(6): 633639. Welch, K.M., Levine, S.R. and Ewing, J.R. (1986) Viewing stroke pathophysiology: an analysis of contemporary methods. Stroke, 17(6): 1071-1077. Welch, K.M., Windham, J., Knight, R.A., Nagesh, V., Hugg, J.W., Jacobs, M., Peck, D., Booker, P., Dereski, M.O. and Levine, S.R. (1995) A model to predict the histopathology of human stroke using diffusion and T2-weighted magnetic resonance imaging. Stroke, 26(11): 1983-1989. West, M.J. and Gundersen, H.J.G. (1990) Unbiased stereological estimation of the number of neurons in the human hippocampus. J. Comp. NeuroL, 296: 1-22. West, M.J., Slomianka, L. and Gundersen, H.J.G. (1991) Unbiased stereological estimation of the total number of neurons in the subdivisions of the rat hippocampus using optical fractionator. Anat. Ree., 231 : 482-497. Zhang, L.-X., Smith, M.A., Li, X.L., Weiss, S.R.B. and Post, R.M. (1998) Apoptosis of hippocampal neurons after amygdala kindled seizures. Brain Res. MoL Brain Res., 55: 198208. Zhong, J., Petroff, O.A., Pilchard, J.W. and Gore, J.C. (1993) Changes in water diffusion and relaxation properties of rat cerebrum during status epilepticus. Magn. Reson. Med., 30(2): 241-246.
T. Sutulaand A. Pitk~inen(Eds.) Progress in Brain Research, Vol. 135 © 2002 ElsevierScienceB.V.All rightsreserved CHAPTER 7
Does convulsive status epilepticus (SE) result in cerebral damage or affect the course of epilepsy the epidemiological and clinical evidence? Simon Shorvon * Institute of Neurology, University College London, Queen Square, London WCIN 3BG, UK
Introduction The fact that convulsive status epilepticus (SE) can result in brain damage in experimental models has been irrefutably demonstrated, since the pioneering work of Meldrum and colleagues in the adolescent baboon (Meldrum and Brierley, 1973; Meldrum and Horton, 1973; Meldrum et al., 1973, 1974; Meldrum, 1997). Numerous animal models, in which seizures produced either by chemical means or by electrical stimulation have repeatedly shown a distinctive pattern of cerebral damage (see Fountain and Lothman, 1995; Coulter and DeLorenzo, 1999). Considerable progress has been made in clarifying mechanisms, as indeed the work presented in this volume itself testifies. In spite of this, evidence of brain damage in humans has been more difficult to define or quantify. In the past, indeed, several authorities have doubted that damage is prominent or clinically relevant. This inadequate state of affairs reflects badly upon clinical research. In this paper, I will try to summarise, in short form, the lines of evidence which do exist and the reasons for lack of clarity on this subject. I will concentrate on tonic clonic SE and not on the
* Correspondence to: S. Shorvon, Institute of Neurology, University College London, Queen Square, London WC1N 3BG, UK. Tel.: +44-20-7829-8758; E-mail:
[email protected]
TABLE 1 Adverse outcomesof human SE Death Focal neurological deficits Intellectual deficit Epilepsy
other types of SE (e.g. non-convulsive SE, epilepsia partialis continua, SE in childhood epilepsy syndromes, etc.). The prognosis and outcome of these conditions is covered elsewhere (Shorvon, 1994). The experimental animal evidence is much stronger and provides fascinating insights into the extent and mechanisms of cerebral damage following SE; this evidence is reviewed elsewhere in this volume and will not be covered here. In Table 1 are the categories of adverse outcome in SE which have been the subject of study. I will consider each in turn. In Table 2 are the confounding TABLE 2 Factors which influence outcome and which render clinical studies difficult Definitions of SE (the duration of episode) Types of SE The underlying pathology Patient factors (e.g. age, drug treatment) Ceiling effects Insensitivity of clinical measurementtools
86 factors that have made this a subject difficult to study from the clinical perspective and which complicate the assessment of results (and these are difficulties which can be largely avoided in experimental investigations in animal models).
Factors which complicate the assessment of the findings from clinical studies
Definitions of SE There is considerable (and to an extent futile) debate about what constitutes SE, a debate which resolves mainly around the length of time continuous or repeated seizure activity should persist before SE is considered to be present. Studies with differing definitions will not be comparable. The traditional literature, influenced heavily by Gastaut and colleagues (Gastaut et al., 1967), defined SE as present if seizures persist for 60 rain or more. Subsequent work proposed time periods of 20 or 30 min, and in recent times much shorter periods have been suggested (5 min being the current record). The radical shortening to 5 min is based on studies (usually small) of video-EEG convulsive seizures suggesting that if a seizure continues for 5 rain or more, it will become self-sustaining (Lowenstein et al., 1999). Here is not the place to debate the merits of these suggestions. Even if this proposition is correct, and epidemiological evidence from the Richmond study (DeLorenzo et al., 1999) as well as everyday clinical experience, suggests that it is not always the case, the fact remains that the propensity to cerebral damage, and prognosis and outcome, is heavily dependent on the duration of the seizure activity. Animal experimentation has yielded findings which are consistently clear about the importance of duration of seizures as a major factor influencing the degree of cerebral damage. Almost all models have shown that there is a threshold below which histological damage is not seen, and above which the longer the seizure, the more likely is consequent cerebral damage. This was first demonstrated by Meldrum who showed that cortical damage is unlikely in the photosensitive baboon with seizures of less than 90 rain duration (see Meldrum, 1997). Findings similar in principle (but varying in detail) have been made by many subsequent authors. The same is probably
true in human SE, although few studies have actually explored this. One of the earliest of these was that of Barois et al. (1985) who reported 90 cases of SE of under 24 h duration of whom only 4% died, and 29 cases of SE lasting more than 24 h, of whom 24% died and 24% were left with residual deficits or in a persisting vegetative state. Towne et al. (1994), in a retrospective survey of 253 adult cases of SE, showed a 9.8 fold (odds ratio) increase in mortality amongst those with SE lasting more than 1 h when compared to those with less long-lasting SE. There has been a significant reduction in mortality and morbidity of childhood SE since the recognition that prolonged seizures are potentially dangerous, and a much more urgent approach to treatment has been taken. This follows the landmark work of Aicardi and colleagues (see below; Aicardi and Chevrie, 1970). Interestingly, this improvement in morbidity was interpreted in an editorial 20 years later suggesting that SE was not as dangerous as once taught (Freeman, 1989). This fallacy was well countered by Aicardi and others in the correspondence that followed (Aicardi and Chevrie, 1989), who emphasised that it was the urgent treatment of acute seizures (and therefore the reduction in seizure duration) which was the reason for the improvement in morbidity, and that delayed treatment of SE was as dangerous as ever. A final methodological point worth making is that, in human studies, the point of onset of SE is often poorly observed and thus estimates of 'duration' are surely subject to wild inaccuracy. Quite how this problem is overcome in epidemiological studies is not clear, but this simple fact clearly complicates all human studies purporting to measure outcome related to seizure-duration.
Types of SE It has long been thought that convulsive SE carries far greater risk of brain damage than non-convulsive SE. Certainly, the mortality rates are much higher in convulsive SE (see below). There is however unresolved debate concerning the extent of morbidity of non-convulsive SE. This is a non-trivial subject, for the urgency of treatment hinges on this point. The evidence from animal experimentation is however not reassuring. In Meldrum's original work,
87 it was clear that convulsive SE was more damaging than non-convulsive SE, although both caused damage (Meldrum, 1997). Furthermore the damage caused by convulsive SE could not be wholly attributed to systemic physiological disturbances (e.g. hyperthermia, hypoxia, hypoglycaemia, low blood pressure, etc.) although these worsened damage. The damage is largely due to electrographic activity, and the clear demonstration that hippocampal damage consistently follows prolonged non-convulsive limbic SE in rodents and other species have also highlighted the dangers in humans (and this is discussed further below and extensively elsewhere in this volume). Many of the older studies of SE, however, did not differentiate between seizure types, and this renders comparison difficult.
The underlying pathology Status epilepticus is much more likely to occur in patients with symptomatic epilepsy than in idiopathic epilepsy. The underlying pathology has often a propensity to cause cerebral damage itself, and the disentanglement of this effect from that of the SE is a problem which has plagued all clinical studies of outcome. This is particularly true of de novo SE in patients who have no previous history of epilepsy (40-70% of all cases) and in whom the SE is due to an acute brain insult. The common causes are trauma, encephalitis, stroke, tumour etc. Morbidity in such cases is often more likely to be due to the underlying pathology than the SE, although both may play a part. An interesting observation on this point was made in the study of Goulon et al. (1985). They noted that the prognosis of the underlying condition worsened if SE occurred; thus, the mortality of bacterial meningitis admissions to the ITUs was 33% of 87 cases without SE, compared to 82% of the 11 cases complicated by SE. Similarly, the recent Richmond epidemiological project examined 37 deaths, and most were in patients with acute symptomatic epilepsy (see below). The mortality rate of stroke in the same hospital is less than 4%, but when the stroke is complicated by SE, the death rate is 35-50% (DeLorenzo et al., 1995). Of course even in these studies, it is likely that the pathologies (e.g. stroke, meningitis) which resulted in SE were more profound than those that did not. Causality in
such uncontrolled situations is really not possible to prove.
Patient factors (e.g. age, therapy) Children have been recognised for many years to be more prone to SE than adults, yet the propensity for cerebral damage may be less (Fountain and Lothman, 1995). This implies that the mechanisms of epileptogenesis are not the same as the mechanisms of seizure-induced cerebral damage, but this is a little explored point. The relative resistance of immature brain to seizure-induced damage has been repeatedly shown in animal work, although there are very few clinical data. It is also possible that some of the drugs used to treat SE (for instance the barbiturates, general anaesthesia) have neuro-protectant potential, although again clinical evidence is slight.
Ceiling effects It seems very likely that if cerebral damage is induced by seizures (or SE) it is subject to a ceiling effect. Routine clinical experience shows that whilst cortical gliosis and a degree of hippocampal atrophy are common in chronic epilepsy, it is seldom very severe, and does not strongly correlate with number of seizures or SE episodes. This is discussed further elsewhere in this volume. Even in patients with very chronic and severe epilepsy, the degree of atrophy can be mild. It is possible that the majority of damage is caused by the initial episodes and that subsequent episodes have less propensity to result in damage. This might explain the apparent lack of obvious evidence of clinical deterioration in non-convulsive complex partial SE which is the SE type, par excellence, which is regularly repeated (Cockerell et al., 1994).
Insensitivity of clinical measurement tools A great advantage of animal experimentation is the ability directly to study tissue in a controlled fashion by histological methods. In human SE, reliance has to be put on methods which are fundamentally insensitive to the sorts of small changes induced by seizures. Serial MRI and psychometric testing, for instance, are better suited to detecting large structural
88 change. Even massive changes in gene expression, or in receptor regulation, or in synaptic reorganisation - - the sorts of changes which are likely to be seizure-induced and result in cerebral damage - are not easily detected by these methods. Strenuous efforts, however, have produced some results, and these are discussed elsewhere in this volume.
Mortality of SE The fact that SE can result in death was recognised from the very first documentation of epilepsy (Sakikku), and has been frequently emphasised ever since (see Shorvon, 1994). Even in the pre-treatment era, however, the condition was by no means always fatal and only between 5 and 50% of patients with convulsive SE died in the acute attack in the series before the early part of the twentieth century (and the introduction of effective treatment).
Mortality in adult SE In the first important study of outcome in adult SE (Oxbury and Whitty, 1971), 6 (11%) of the 54 patients with known cerebral pathology died in SE, and a further 5 (9%) died within the next 6 months. Only 1 of the 32 cases without known pathology died. Aminoff and Simon (1980) reported death in 16 (16%) of 98 adult patients admitted to hospital with SE, but in only 2 was death directly attributed to SE, and in the rest it was due to the underlying cause or to medical or therapeutic complications. In a study of 282 consecutive admissions in SE to two intensive care units (Goulon et al., 1985), 100 (35%) died, but in only 2 was the death attributable to SE. The recent epidemiological study from Richmond, Virginia, of 137 adult and 29 paediatric cases of SE showed an overall mortality of 22%. The mortality of the elderly patients was highest (38%) compared with that in young adult age groups (14%). The great majority of these deaths were due to the underlying causes, although 26% of the adult deaths occurred in either idiopathic (3%) or remote symptomatic epilepsy (22%) (DeLorenzo et al., 1995). Towne et al. (1994) published univariate and multivariate analyses of mortality amongst 253 adults with SE identified retrospectively from the hospital computerised database of discharge details (a source
which has inherent inaccuracies). Those with SE of less than one hour duration had lower mortality rates than those with SE duration of over one hour. Anoxic aetiologies and advancing age (which are interrelated factors) were associated with higher rates.
Mortality in childhood SE Aicardi and Chevrie (1970) reported an 11% mortality rate amongst 239 infants and children, with death both during the acute stage of convulsive SE and in the months later. Most of the cases (85%) were under the age of five years, and the authors reported that the duration of convulsions was critical and that prolonged convulsions were particularly devastating in young babies. In later studies, lower mortality rates were reported (e.g. 3.6%) which reflects the success of more urgent therapy. Phillips and Shanahan (1989), in a study deliberately designed to see whether prognosis had improved in the two decades since the Aicardi study, reported a 6% mortality rate in 218 episodes of childhood SE admitted to a paediatric intensive care unit over a 5year period. In 1 l of the 13 deaths, there were acute cerebral insults, and there was only 1 death amongst the 99 episodes of idiopathic SE. In the most recent study (from Richmond), the mortality amongst children was only 2.5% (1 case) and none in idiopathic or remote symptomatic cases (DeLorenzo et al., 1995).
Mortality in non-convulsive SE There have been few studies of mortality in nonconvulsive SE. It is possible that some cases were included in the earlier studies cited above, although usually seizure type was not documented and in these early studies only convulsive SE was included in definitions. One study reported death in three of ten cases of complex partial SE admitted to intensive care, and in two of these cases the cause of the acute SE was not known. However, the great majority of patients with non-convulsive SE are not admitted to intensive care facilities, and the extent of mortality in the generality of case is unknown - - although routine clinical experience suggests that this common event is very rarely fatal.
89
Clinical studies of neurological morbidity resulting from SE Adults In the study of Oxbury and Whitty (1971), five of 86 cases of SE were 'undoubtedly deteriorated' after the episode, and in two (2%) no cause other than the epilepsy itself could be found for the neurological impairment. Rowan and Scott (1970) recorded a 26% rate 'neurological sequelae' but did not provide any further details. Six of the 90 cases reported by Aminoff and Simon (1980) had intellectual impairment following the SE. The main difficulty in assessing this problem is that of differentiating the effects of seizures from that of the underlying cause. Nevertheless, it is common clinical experience that memory and personality change is common after a prolonged bout of convulsive SE, although these deficits often improve over months. There is one prospective psychometric study of patients before and after SE (albeit without SE being the major focus of the study; Dodrill and Wilenski, 1990). In this investigation, 143 adult epileptic patients were tested 5 years apart. SE occurred in the intervening period in 9 cases, and there was a bigger (but not significantly so) deterioration in the WAIS amongst these 9 cases when compared to the other 144. However, the patients who experienced SE had markedly lower scores than the controls even at the first evaluation (Dodrill and Wilenski, 1990). Because of small numbers and other confounding factors, the statistical value of this study is limited. Dodrill and Wilenski also reviewed 14 other studies and case reports and concluded that SE had only a slight adverse effect on cognitive abilities amongst survivors, and that in many individuals there were no adverse effects. My own reading of these limited data, however, is not so reassuring. The correct tests have seldom been done, and there are very few prospective data. There are, furthermore, no detailed serial data in the aftermath of SE. In adults, at least, it can be probably concluded that serious morbidity is relatively uncommon (although everyday clinical practice shows that effects do occur), yet there can be no doubt from individual cases that permanent cognitive sequelae can result from a severe episode of convulsive SE. Acute intellectual disturbances
often improve over the months following SE, and the timing of testing is important. Conversely, serial imaging evidence suggests that consecutive atrophy may progress in the months after an episode of SE. Children In children, the risk of morbidity from SE may differ from that in adults. The younger the child, the more likely are both motor or cognitive deficits. In the series of studies of Aicardi and Chevrie (Aicardi and Chevrie, 1970) of the 239 cases of SE between the ages of 1 month and 15 years, 47 (20%) subsequently developed motor deficits and 55 (23%) mental impairment which could be directly attributable to the SE (and not the underlying aetiology). The motor problems were hemiplegia in 28 (12%) of the type conforming to the HH syndrome, and diplegia, extrapyramidal and cerebellar signs. The motor and mental sequelae often co-existed and overall 82 (34%) children were affected. In these studies, 118 children were seen in the acute phase of SE and followed prospectively. Of these, 47 (40%) were left with neurological and 51 (43%) mental sequelae (a total of 53 (45%) of cases). In a similar study from Japan, deficits occurred in 40 (51%) of 79 children, of whom 25 (32%) were of the HH type, and 37% mental deficits. A 28% morbidity was found amongst 52 children by Yager et al. (1988), and a 9.1% morbidity amongst 186 survivors from SE by Maytal et al. (1989). The experimental data in rodent models suggest that juvenile animals are less liable to damage following SE than adult animals (Coulter and DeLorenzo, 1999), but this issue is, in clinical practice, unresolved. The HH and HHE syndromes (initials referring to the permanent hemiplegia or hemiparesis) and chronic epilepsy (in about three quarters), which can follow a prolonged asymmetrical or unilateral febrile convulsion in a child under 4 years of age (usually under 2 years) (Aicardi and Chevrie, 1969). Many of these children have some neurological dysfunction before the SE, and the SE may be of greater severity than usual reflecting this. There is severe venous congestion and thrombosis, and massive cerebral oedema pathologically, although angiography is usually normal. The damage is probably due to a mixture of vascular and excitotoxic mechanisms. This
90 used to be a common sequel to febrile SE (Gastaut et al., 1967; Aicardi and Chevrie, 1969, 1970, 1983), but with the more rapid and effective early therapy, HH and HHE are now rare occurrences in developed countries, although they are still frequent and preventable in the developing world. This improvement is very likely to be the consequence of more rapid and urgent control of convulsive and febrile SE.
Histological and MRI studies of hippocampal and cortical damage following SE Hippocampal atrophy and gliosis have been long recognised to be an association with human SE, both chronically and in the acute phase. There are definitive pathological studies in the older literature on this point. These findings are very similar to those in experimental animal models, and these are discussed elsewhere in this volume. Corsellis and Bruton (1983), for instance, found almost complete loss of neurones in the Sommer sector of the hippocampus in 20 patients dying during or soon after an episode of SE. This acute lesion was recognised as the precursor of Ammon's Horn Sclerosis. They also noted damage in the cerebral cortex, the cerebellum and thalamus. Indeed, widespread gliosis (Chaslin's gliosis) and neuronal loss in the cortex was initially regarded as more significant than that in the hippocampus. Acute neuronal necrosis was found especially in the middle cortical layers, stretching over wide areas of the cortical mantle in some cases, and only patchily distributed in others (Corsellis and Bruton, 1983). In survivors of the acute lesion, gradual atrophy of the cerebellum was noted, in both the Purkinje cells and the granular layer, and also widespread shrinkage, gliosis and neuronal loss in neocortex and basal ganglions. In 55 patients with severe chronic epilepsy, Margerison and Corsellis (1966) found significant damage in the hippocampus in 65%, in the cerebellum in 45%, in the amygdala in 27%, in the thalamus in 25% and in the cortex in 22%. It was concluded that hippocampal sclerosis was due to ischaemic brain damage in early seizures or SE. The relative vulnerability of the prosubiculum, CA1 and CA3 regions to damage in human SE has been confirmed by quantitative measures of neuronal densities (DeGiorgio et al., 1992; Chapter 21 in this volume). Although all the studies reported
above were of convulsive SE, there is also a single paper describing three patients without pre-existing epilepsy who died 11-27 days after the onset of non-convulsive SE lasting 1-3 days (Fujikawa et al., 1991). In all three cases there were changes similar to those outlined above, with neuronal loss in CA1, CA3 and hilar cells, and also in amygdala, thalamus, cerebellum and cerebral cortex. There are also a small number of serial MRI studies which show evidence of progressive atrophy following SE. These are reviewed elsewhere in this volume and will be discussed only briefly here. In the study of Wieshmann et al. (1997) for instance, a patient with an episode of SE which lasted 2 weeks was studied. She was scanned during the episode, 2 months later and then 56 months later. Within 2 months of the SE episode, bilateral hippocampal atrophy was demonstrated and this had progressed further at the time of the scan 56 months later. Limited psychometric analysis over this period (using the Warrington recognition test for words and faces) also showed evidence of cognitive decline at 2 months which had also progressed further at 58 months. Meierkord et al. (1997) and Chee and Lo (1997) also reported three cases scanned during and after episodes of prolonged SE and showed atrophy and signal change. VanLandingham et al. (1998) showed changes in hippocampal volume after prolonged or focal febrile SE. These confirm the findings from the pre-MRI days of air-encephalographic studies which showed ventricular enlargement following SE (Aicardi and Baraton, 1971), and from CT studies (Labate et al., 1991).
Epilepsy resulting from SE A fundamental question which has been very inadequately researched is whether human SE increases the propensity for further epilepsy. The animal experimentation is clear, but there are at present no modem direct clinical studies of this phenomenon and data are sparse. The situation is complicated by the fact that no study has successfully disentangled the effects of the SE from that of the underlying aetiology. Four areas can be cited which are relevant to this aspect. First is of course the analogy with animal experimentation. In many animal models, experimentally
92 the NCPP (Nelson and Ellenberg, 1981), permanent hemiplegia occurred after the febrile seizure in only 0.4%, and no child developed the H H E syndrome. The risk of subsequent epilepsy after a febrile convulsion has varied from 2 to 11%. The NCPP found a four fold increase in risk of epilepsy in children who had a febrile convulsion compared to those who had not (although the overall risk was small). In the CHES cohort (16004 neonatal survivors, 98.5% of all children born in the U K in one w e e k in April, 1970), 14676 were studied at 10 years and of these who were neurologically normal before their first febrile seizure (382 cases), only 2.4% (9 cases) developed epilepsy. 32 patients were recorded to have had a prolonged febrile convulsion and 2 (6%) developed subsequent epilepsy (Verity, 1998). None o f the other studies differentiated febrile SE from shorter febrile seizures, and the risks of epilepsy are no doubt greater after prolonged febrile seizures. It thus seems clear that epilepsy can develop after febrile SE, although the majority of children who experience febrile SE will not develop subsequent epilepsy or cerebral damage. The subject o f the relationship between febrile seizures and epilepsy is discussed elsewhere in this volume. Future studies of the potential of SE to cause epilepsy are needed. These are important clinically to assess the place of neuroprotection. F r o m the clinical perspective, studies o f epilepsy after SE are complicated by the confounding effects of the aetiology, the type of SE, and ceiling effects. Investigations of de novo SE in patients without previous epilepsy are useful only if the cause of the SE is known itself not to result independently in subsequent epilepsy (and this excludes cases due, for instance, to cerebral infection, trauma, stroke or tumour). Ideally, a case control methodology should be utilised, with SE cases matched to other epilepsy patients without a history of SE. The results should be stratified by age, and by the duration of the SE. These are formidable design problems and because of the difficulties these pose, there are no published studies to date which have satisfactorily explored this question. References
Aicardi, J. and Baraton, J. (1971) A pneumoencephalographic demonstration of brain atrophy following status epilepticus.
Dev. Med. Child Neurol., 13: 660-667. Aicardi, J. and Chevrie, J.-J. (1969) Acute hemiplegia in infancy and childhood. Dev. Med. Child Neurol., 11: 162-173. Aicardi, J. and Chevrie, J.-J. (1970) Convulsive status epilepticus in infants and children. Epilepsia, 11: 187-197. Aicardi, J. and Chevrie, J.-J. (1989) Status epilepticus. Pediatrics, 84: 939-940. Aicardi, J. and Chevrie, J.-J. (1983) Consequences of status epilepticus in infants and children. In: A.V. Delgado Escueta, C.G. Wasterlain, D.M. Treiman and R.J. Porter (Eds.), Status Epilepticus. Mechanisms of Brain Damage and Treatment.
Advances in Neurology 34, Raven Press, New York, NY, pp. 115-128. Aminoff, M. and Simon, R.P. (1980) Status epilepticus: causes, clinical features and consequences in 98 patients. Am. J. Med., 69: 657-666. Annergers, J.F., Hauser, W.A., Shirts, S.B. and Kurtland, L.T. (1987) Factors prognostic of unprovoked seizures after febrile convultions. New England J. Med., 316: 493~-98. Barois, A., Estournet, B., Baron, S. and Levy-Alcover, M. (1985) Prognostic h long terme des 6tats de mal convulsifs prolongrs. A propos de vingt-neuf observations d'6tats de mal convulsifs de plus du vingt-quatre heures. Ann. Pediatr., 32: 621-626. Berg, A.T., Shinnar, S., Hauser, W.A., Alemany, M., Shaprio, E.D., Salomon, M.E. and Crain, E.F. (1992) A prospective study of recurrent febrile studies. New England J. Med., 327: 1122-1127. Chee, M.W.L. and Lo, N.K. (1997) Asymmetric hippocampal atrophy and extra-hippocampal epilepsy following refractory status epilepticus in an adult. J. Neurol. Sci., 147: 203-204. Corsellis, J.A.N. and Bruton, C.J. (1983) Neuropathology of status epilepticus in humans. In: A.V. Delgado Escueta, C.G. Wasterlain, D.M. Treiman and R.J. Porter (Eds.), Status Epilepticus. Mechanisms of Brain Damage and Treatment.
Advances in Neurology 34, Raven Press, New York, NY, pp. 129-139. Cockerell, O.C., Walker, C.M., Sander, J.W.A.S. and Shorvon, S.D. (1994) Complex partial status epilepticus: a recurrent problem. J. Neurol., Psychiatry Neurosurg., 57: 835-837. Coulter, D.A. and DeLorenzo, R.J. (1999) Basic mechanisms of status epilepticus. In: A.V. Delgado, W.A. Wilson, R.W. Olsen and R.J. Porter (Eds.), Jaspers Basic Mechanisms of the Epilepsies. Lippencott Williams and Wilkins, Philadelphia, PA, pp. 725-733. DeGiorgio, C.M., Tomiyasu, U., Cott, P.S. and Treiman, D.M. (1992) Hippocampal pyramidal cell loss in human status epilepticus. Epilepsia, 33: 23-37. DeLorenzo, R.J., Garnett, L.K., Towne, A.R., Waterhouse, E.J., Boggs, J.G., Morton, L., Afzal Choudhry, M., Barnes, T. and Ko, D. (1999) Comparison of status epilepticus with prolonged seizure episodes lasting from 10 to 29 minutes. Epilepsia, 40: 164-169. DeLorenzo, R.J., Pellock, J.M., Towne, A.R. and Boggs, J.G. (1995) Epidemiology of status epilepticus. Z Clin. Neurophysiol., 12: 316-325. Dodrill, C.B. and Wilenski, A.J. (1990) Intellectual impairment
91 TABLE 3 The course of epilepsy after an episode of convulsive status in 5 patients with pre-existing epilepsy Patient Sex/age Seizure type prior to SE
Aetiology of epilepsy
Approx Course in subsequent 12 months duration of status
1.
F/32
CPS, SGTCS, > 1/month
Post-tranmatic
36 h
2. 3.
M/34 M/27
CPS,SGTCS, > 1/month Hippocampalsclerosis CPS, SPS, SGTCS, 1/3 month Hippocampal sclerosis
4 5.
M/38 F/22
CPS, SPS, SGTCS SGTCS
24 h 24 h
Post-operative dysplasia 12 h Not known 6h
7 months seizure free, then epilepsy returned at previous frequency No change Szs unchanged in form but more frequent (> 1/month) No change New seizure type developed (motor partial Szs), frequency unchanged
Course of epilepsy in 5 patients with chronic epilepsy, followed for at least 12 months after an episode of convulsive status. The duration of status is given approximately (to nearest 6 h). Seizure frequencies are categorised into: < 1/month; l-3/month; <3/month). Patients 1 and 4 had at least one documented previous episode of status). Key: M = male; F = female; CPS = complex partial seizures; SPS = simple partial seizures; SGTCS = secondarily generalised tonic clonic seizures; Szs = seizures. induced SE results in a continuing propensity to seizures. This is attributed usually to the statusinduced hippocampal damage, and this resembles human hippocampal sclerosis. A good (but by no means the only) example is that of the self-sustaining limbic status epilepticus model of Lothman (see Fountain and Lothman, 1995). The induced SE usually lasts several hours and the animals gradually return to normal. The episode results in chronic cell loss in CA1 and in the middle layers of entorhinal cortex. About one month later, the animals begin to have spontaneous seizures, of a complex partial type, and this chronic epilepsy' persists for at least 12 months (the longest reported period yet of follow up; Fountain and Lothman, 1995). Epilepsy is a sequel of chemically and electrically induced SE in a wide range of models, and of course also of kindling (Coulter and DeLorenzo, 1999). Second, SE is sometime the first manifestation of epilepsy - - and it could be hypothesised that, at least in some cases, the status-induced damage is the cause of the subsequent epilepsy. In various series, between 40 and 70% of all cases of SE occur in people without a previous history of epilepsy, and in 12% of all cases of epilepsy the seizures started with SE. Third, an episode of SE can also change seizure type. In the 239 children reported by Aicardi and Chevrie (1970) for instance, the following seizure types developed de novo: infantile spasms in 10 (7%) cases, further SE in 15 (10%), Lennox Gastaut
Syndrome in 16 (11%), and partial epilepsy in 35 (15%). Tonic clonic seizures occurred after the SE in five fewer patients than before. I have reviewed the subsequent course of epilepsy in five patients with chronic epilepsy who sustained an episode of SE from m y own practice (Table 3). In two, the SE made no difference to the seizure pattern. In one, a new seizure type emerged (partial motor seizures), in one there was a period of freedom from seizures for 7 months before the epilepsy returned at approximately its previous frequency, and in one the epilepsy was significantly worsened after the SE. Such observations of course are uncontrolled and furthermore, treatment changes were made in all except one case, yet they do provide anecdotal evidence that SE can alter, in various ways, the course of epilepsy. Epilepsy is more likely if the episode of SE resulted in overt cerebral damage. About three quarters of those children with post-status hemiplegia, for instance, will also develop seizures (Gastaut et al., 1967; Aicardi and Chevrie, 1969, 1970, 1983). Finally, there is the evidence that epilepsy can result from febrile SE. Although the case series from hospital practice provide rather ominous evidence of the potential for febrile seizures to cause SE, the epidemiological studies are more reassuring. There were no deaths from febrile seizures or SE recorded amongst the 2740 children in the three largest prospective studies (Nelson and Ellenberg, 1981; Annergers et al., 1987; Berg et al., 1992). In
93
as an outcome of status epilepticus. Neurology, 40(Suppl. 2): 23-27. Freeman, J.M. (1989) Status epilepticus: it's not what we've thought or taught. Pediatrics, 83: 444-445. Fountain, N.B. and Lothman, E.W. (1995) Pathophysiology of status epilepticus. J. Clin. Neurophysiol., 12: 326-342. Fujiwara, T., Watanabe, M., Matsuda, K., Senbongi, M., Yagi, K. and Seino, M. (1991) Complex partial status epilepticus provoked by ingestion of alcohol. Epilepsia, 32: 650-656. Gastaut, H., Roger, J. and Lob, H. (1967) Les dtats de mal gpileptiques. Masson, Paris. Goulon, M., Levy-Alcover, M.A. and Nouailhat, F. (1985) Etat de mal 6pileptique de l'adulte, 6tude 6pid6miologique et clinique en r6animation. Rev. EEG Neurophysiol., 14: 277-285. Labate, C., Magaudda, A., Fava, C., Meduri, M. and Di Perri, R. (1991) Hemispheric brain atrophy following unilateral status epilepticus. Boll. Lega Ital. Contro L'Epilepsia, 74: 103-104. Lowenstein, D.H., Bleck, T. and MacDonald, R. (1999) It's time to revise the definition of status epilepticus. Epilepsia, 40: 120-122.
Margerison, J.H. and Corsellis, J.A.N. (1966) Epilepsy and the temporal lobes: a clinical, electroencephalographic and neuropathological study of the brain with particular reference to the temporal lobes. Brain, 89: 499-530. Maytal, J., Shinnar, S., Moshe, S.L. and Alvarez, L.A. (1989) Low morbidity and mortality of status epilepticus in children. Pediatrics, 83: 323-331. Meierkord, H., Wieshmann, U., Niehaus, L. and Lehmann, R. (1997) Structural consequences of status epilepticus demonstrated with serial magnetic resonance imaging. Acta Neurol. Scand., 96: 127-132. Meldrum, B.S. (1997) First Alfred Meyer Memorial Lecture. Epileptic brain damage: a consequence and a cause of seizures. Neuropathol. Appl. Neurobiol., 23: 185-202. Meldrum, B.S. and Brierley, J.B. (1973) Prolonged epileptic seizures in primates: ischaemic cell change and its relation to ictal physiological events. Arch. Neurol., 28: 10-17.
Meldrum, B.S. and Horton, R.W. (1973) Physiology of status epilepticus in primates. Arch. Neurol., 28: 1-9. Meldrum, B.S., Vigouroux, R.A. and Brierley, J.B. (1973) Systemic factors and epileptic brain damage. Arch. Neurol., 29: 82-87. Meldrum, B.S., Horton, R.W. and Brierley, J.B. (1974) Epileptic brain damage in adolescent baboons following seizures induced by alloglycine. Brain, 99: 523-542. Nelson, K.B. and Ellenberg, J.H. (Eds.) (1981) Febrile Seizures. Raven Press, New York. Phillips, S.A. and Shanahan, R.J. (1989) Etiology and mortality of status epilepticus in children: a recent update. Arch. Neurol., 46: 74-76. Oxbury, J.M. and Whitty, C.W.M. (1971) Causes and consequences of status epileptieus in adults: a study of 86 cases. Brain, 94: 733-744. Rowan, A.J. and Scott, D.F. (1970) Major status epilepticus: a series of 42 patients. Acta Neurol. Scand., 46: 573-584. Shorvon, S.D. (1994) Status Epilepticus: Its Clinical Features and Treatment in Children and Adults. Cambridge University Press, Cambridge. Towne, A.R., Pellock, J.M., Ko, D. and DeLorenzo, R.J. (1994) Determinants of mortality in status epilepticus. Epilepsia, 35: 22-34. VanLandingham, K.E., Heinz, E.R., Cavazos, J.E. and Lewis, D.V. (1998) MRI evidence of hippocampal injury after prolonged febrile convulsions. Ann. Neurol., 43: 413-426. Verity, C.M. (1998) Do seizures damage the brain? The epidemiological evidence. Arch. Dis. Child., 78: 78-84. Yager, J.Y., Cheang, M. and Seshia, S.S. (1988) Status epilepticus in children. Can. J. Neurol. Sci., 15: 402-405. Wieshmann, U.C., Woermann, F.G., Lemieux, L., Free, S.L., Bartlett, P.A., Smith, S.J.M., Duncan, J.S., Stevens, J.M. and Shorvon, S.D. (1997) Development of hippocampal atrophy: a serial magnetic resonance imaging study in a patient who developed epilepsy after generalised status epilepticus. Epilepsia, 38: 1238-1241.
T. Sutula and A. Pitk~en (Eds.)
Progressin BrainResearch,Vol. 135 © 2002 Elsevier Science B.V. All rights reserved
CHAPTER 8
Repeated brief seizures induce progressive hippocampal neuron loss and memory deficits Robert Kotloski 1, Michael Lynch 1, Suzanne Lauersdorf I and Thomas Sutula 1,2,* 1Department of Neurology and2Department of Anatomy, University of Wisconsin, Madison, W153792, USA
Abstract: The long-term effects of repeated brief seizures on spatial memory and hippocampal neuronal populations were assessed in kindled rats. Rats that experienced a range of 3 afterdischarges to 134 secondary generalized tonic-clonic (Class V) seizures evoked by stimulation of the olfactory bulb were evaluated in a radial arm maze task that is a measure of spatial memory and is disrupted by hippocampal damage. After completion of the memory task and a minimum of "-~3 months after the last evoked seizure, stereological methods were used to assess neuronal populations at septal and temporal locations of the hippocampus and dentate gyms. Repeated brief seizures induced a long-lasting deficit in spatial memory performance that was detected after a cumulative total of "-~6partial and 30 secondary generalized seizures. The memory deficit progressively increased as a function of the number of seizures, and was not observed in age-matched, electrodeimplanted, unstimulated, but otherwise similarly handled paired controls. Neuronal loss was detected in the temporal hilus of the dentate gyms, CA1, and CA3 of the hippocampus after 69 or more secondary generalized tonic-clonic seizures, and was associated with the progressive memory dysfunction. Repeated brief seizures induced progressive, permanent functional and structural abnormalities in the hippocampus, which included spatial memory deficits accompanied by gradually evolving neuronal loss in a pattern resembling human hippocampal sclerosis. These experimental results support the view that hippocampal sclerosis and associated memory dysfunction are induced by repeated seizures, and imply that seizure control could prevent adverse long-term consequences of seizures on hippocampal dependent functions.
Introduction
The question of whether seizures cause brain damage has provoked controversy for more than a century. While there has been continuing debate about whether brief seizures induce neural damage, both experimental and clinical studies have firmly established that intense or continuous seizures during status epilepticus induce widespread neural damage prominently involving the hippocampus. The damage is a direct consequence of the seizures and is
* Correspondence to: T.E Sutula, Department of Neurology H6/570, University of Wisconsin, Madison, WI 53792, USA. Tel.: +1-608-263-5448; Fax: -t-1-608-2630412; E-mail:
[email protected]
not merely the result of hypoxia or metabolic disturbances (Meldrum et al., 1973). Prompt, effective treatment of status epilepticus in rodent models not only suppresses the acute seizures, but also reduces or prevents long-term consequences, such as seizure-induced damage in specifically vulnerable neurons in the hippocampal dentate gyms, associated mossy fiber sprouting, behavioral abnormalities, and increased susceptibility to evoked kindied seizures (Ylinen et al., 1991; Sutula et al., 1992). These beneficial long-term effects of treatment in status epilepticus have important clinical implications for resolving the controversy about neural damage induced by recurring brief seizures. If neural damage is also a consequence of the brief recurring seizures that are the defining feature of epilepsy, prompt effective treatment that achieves complete
96 control and suppression of sporadic seizures would be expected to forestall cognitive impairment and development of intractable epilepsy. The most commonly encountered pattern of damage in human epilepsy is neuronal loss and gliosis in the hippocampus prominently involving CA1, CA3, and the hilus of the dentate gyms, which has been referred to as hippocampal sclerosis, Ammon's Horn sclerosis, or mesial temporal sclerosis (Gloor, 1991). Autopsy studies and pathological examination of the surgically resected human temporal lobe have demonstrated that the damage may variably involve other regions of the hippocampus (Mouritzen Dam, 1980; Margerison and Corsellis, 1966), including CA2 and the granule cell layer of the dentate gyms, which are regarded as relatively resistant to hypoxic and excitotoxic damage (Sloviter, 1989). Volumetric MRI methods have established a correlation between atrophy of the hippocampus and the histological lesion of hippocampal sclerosis (Cascino et al., 1991), and have confirmed that the damage may involve not only the hippocampus, but other limbic areas such as the entorhinal cortex, lateral temporal cortex, and regions beyond the hippocampal formation (DeCarli et al., 1998; Lee et al., 1998; Bernasconi et al., 1999). MRI studies have also helped to define the relationship between seizure onset, duration of epilepsy, and the development of hippocampal damage. Hippocampal volume loss may be a consequence of an initial precipitating injury such as prolonged febrile seizures or febrile status epilepticus in children (Sagar and Oxbury, 1987; Vanlandingham et al., 1998), and in adults the extent of hippocampal volume loss is correlated with duration of epilepsy (DeCarli et al., 1998; Lee et al., 1998; Theodore et al., 1999) and in some studies, with estimates of cumulative seizure frequency (Kalviainen et al., 1998; Salmenpera et al., 1998; see also Theodore and Gaillard, 2002, this volume; Holmes et al., 2002, this volume). While volumetric MRI studies have demonstrated that hippocampal atrophy or sclerosis is progressive in patients with intractable epilepsy, the roles of initial precipitating injury and/or repeated seizures in the induction of hippocampal sclerosis remain controversial. The question of whether hippocampal damage is a consequence of an initial injury or might also be caused by cumulative seizure-induce brain damage
can be directly addressed in experimental animal models of chronic epilepsy. Although it had generally been believed that brief sporadic seizures do not induce neuronal damage, recent work from multiple laboratories has provided increasingly strong evidence that repeated brief seizures induce neuronal death. The possibility that repeated seizures might induce neuronal damage followed the observation that kindling induced progressive sprouting of the mossy fiber pathway in the dentate gyms, which suggested that the sprouting might be a reactive consequence to progressive seizure-induced deafferentation or neuronal loss (Sutula et al., 1988; Cavazos et al., 1991). Stereological methods initially demonstrated that there was a reduction in neuronal density in the hilus of the dentate gyrus which increased as a function of the number of seizures, and was consistent with neuronal loss (Cavazos and Sutula, 1990). Subsequent studies reported that brief repeated seizures evoked by kindling progressively reduced neuronal density in a variety of hippocampal and extra-hippocampal areas. The reduction in density was initially most apparent in the hilus of the dentate gyms, but also gradually developed in CA1, CA3, and the entorhinal cortex, and thus cumulatively resembled the pattern of the neuronal loss in hippocampal sclerosis (Cavazos et al., 1994). In this study, there was non-significant trend to a seizureinduced increase in overall hippocampal volume, and there has been controversy about whether the seizure-induced reduction in neuronal density in the hilus was caused by an increase in the volume or by neuronal loss (Bertram and Lothman, 1993; Adams et al., 1997). Dramatic increases in volume of the dentate gyrus have been reported in a chronic mouse model of epilepsy induced by kainic acid (Bouilleret et al., 2000). More recent studies using so-called unbiased stereological methods have confirmed the observation of seizure-induced neuronal loss in the hilus after seizures (Dalby et al., 1998), and multiple laboratories have reported that repeated brief seizures evoked by kindling induce apoptotic neuronal death in both hippocampal and extrahippocampal regions (Bengzon et al., 1997; Pretel et al., 1997; Zhang et al., 1998), which supports the interpretation that seizures induce neuronal death and cumulatively result in a pattern of neuronal loss resembling hip-
97 pocampal sclerosis as well as more widespread damage which is often encountered in human temporal lobe epilepsy. As a further complication for assessment of seizure-induced neuronal loss, it is now apparent that seizures evoked by pilocarpine and kindling also induce neurogenesis in the rat dentate gyms (Parent et al., 1997, 1998) which could restore neuronal number and possibly contribute to the seizure-induced increase in volume of this region reported in rodent models. To further address the controversial questions about seizure-induced neuronal loss, stereological and morphometric analysis was performed in a group of rats that experienced seizures evoked by kindling stimulation of the olfactory bulb. This group of kindled rats was previously examined in a radial arm maze task and demonstrated progressive seizureinduced memory dysfunction as a function of the number of evoked seizures (Sutula et al., 1995). Stereological analysis in these kindled rats with characterized seizure-induced memory deficits provided an opportunity to determine if a relationship existed between the number of seizures, development of spatial memory dysfunction, and alterations in specific hippocampal neuronal populations. The results revealed an increase in volume of the dentate gyms that may partially account for reduced neuronal density in the hilus, but as changes in volume were not observed in CA1 and CA3, the seizure-induced reductions in these regions were most likely caused by neuronal loss. Evidence for neuronal loss in CA1 and CA3 was observed only after ~ 3 0 evoked Class V seizures, which corresponded to the onset of performance deficits in the radial arm maze that requires integrity of the hippocampus. Methods
Surgical and kindling procedures Adult male Sprague-Dawley rats (250-350 g) were anesthetized with a combination of ketamine 80 m g / k g i.m. and xylazine 10 mg/kg i.m., and were stereotaxically implanted with an insulated stainless steel bipolar electrode for stimulation and recording. The electrode was implanted in the olfactory bulb (9 mm anterior, 1.2 m m lateral, 1.8 mm ventral to bregma). After a recovery period of 2 weeks,
age-matched pairs of electrode implanted rats were randomly assigned to a group that received kindling stimulation, or to a paired control group that was similarly handled, but did not receive electrical stimulation. The unrestrained awake animals in the kindling group received twice daily kindling stimulation (5 days per week) with a 1-s train of 62 Hz biphasic constant current 1.0 ms square-wave pulses, according to standard procedures (Cavazos et al., 1994). The evoked behavioral seizures were classified according to standard criteria (Sutula and Steward, 1986). Rats received stimulation until 3 afterdischarges (ADs), or 3, 30, 69-75, or >83 Class V seizures were evoked. A group of paired agematched control rats were handled and placed in the recording cage twice daily according to the same protocol, but did not receive stimulation. Each control rat was handled twice daily until its paired kindled rat experienced the assigned number of Class V seizures.
Behavioral testing A complete description of the behavioral assessment of these rats has been published previously (Sutula et al., 1995). After completion of the last kindling stimulation, the kindled rats were treated with a 1month rest interval. During this interval, the kindled rats were placed into the recording cage and handled twice daily in the same manner as the control rats, but did not receive electrical stimulation. Control rats continued to undergo twice daily handling and placement in the recording cage. After completion of the 1-month rest interval, behavioral testing in an 8-arm radial maze was performed 5 days per week, according to previously described procedures. For the duration of the study, the same 4 arms of the 8-arm radial maze were baited with a raisin, which was hidden from view in a recess at the end of the arm. Each rat was placed in the center area of the maze, and was observed for the sequence of arm entries and consumption of the raisins. Criterion performance was defined as consumption of the raisin in all 4 baited arms during no more than 5 entries. The sequence of entries into baited and unbaited arms of the radial maze was recorded during daily behavioral trials. According to terminology in previ-
98 ous studies, entry into an unbaited arm was scored as a reference error. Reentry into a baited arm was scored as a working error. Reentry into an unbaited arm was scored as a reference and a working error. Completion of the behavioral task was defined as achievement of criterion performance on each of 5 consecutive daily trials. After achievement of criterion performance on 5 consecutive days, the rats continued daily behavioral trials for an additional week, and were then evaluated for recall or memory of the correct baited arms by an additional testing procedure. In this additional testing procedure, all arms of the radial maze were unbaited. Each rat was then placed in the center of the maze, and the ratio of entries into previously baited and unbaited arms, and time spent in previously baited arms was recorded during a 10-min period of observation.
illustrated in Figs. 1-3 of Cavazos et al. (1994), were identified by the following criteria: (1) In a horizontal section about 800 txm ventral from the most dorsal hippocampus, neuronal density was measured in the hilus of the dentate gyms, CA3a, CA3c, CA2, CAla, and CAlc. At this septotemporal level, the polymorphic neurons in the hilus of the dentate gyms are enclosed in an oval ring formed by granule cells in the stratum granulosum. (2) In a horizontal section about 2800 Ixm ventral from the most dorsal hippocampus that included the motor nuclei of cranial nerves Ili and IV, neuronal density was measured in the hilus of the dentate gyms, CA3a, CA3c, CA2, CAla, and CAlc. The investigator performing the counts was unaware of the identity of the sections, and the order of examination of the sections was randomized.
Histological procedures
Counting methods
After completion of behavioral testing, each rat was deeply anesthetized and perfused transcardially with an aqueous solution of 10% (v/v) formalin in 0.9% (w/v) NaC1. The brains were removed, postfixed for at least 10 days in same solution, dehydrated for 35 days in graded concentrations of alcohols, and were then embedded in graded concentrations of low viscosity nitrocellulose (celloidin) for 50 days, which provides superior preservation of the neuronal architecture. Horizontal 20-1xm sections were cut using a sliding microtome from the surface of neocortex throughout the most ventral hippocampus. All sections were retained and stored in 70% ethanol. Every fifth section was stained with cresyl violet, but additional sections were also stained to insure that equivalent areas were available for quantitative assessment in each specimen.
Counts of nuclei were obtained ipsilateral and contralateral to the stimulating electrode at the standardized locations by an optical disector method. All nuclei (objects) that came into focus within the area of an eyepiece reticule (25 m m 2) while moving the microscope headstage through the thickness of the section were manually counted. In each reticule field, nuclei overlapping the lower and left edges of the reticule were counted, but nuclei that overlapped the upper and right edges were not counted. In each location, the number of neurons and their relative position were recorded using a camera lucida, and were summed to obtain the nuclear count per reticule field through the thickness of the section. In previous experiments, the nuclear counting procedure in this study had an inter-observer variability of less than 10%; the intra-observer variability was 3% (Cavazos and Sutula, 1990; Cavazos et al., 1994). Previous studies have revealed that these methods are quite sensitive for assessment of neuronal loss, which typically exceeds 15-20% before a lesion is reliably detected by visual inspection (Konigsmark, 1969; Cavazos and Sutula, 1990; Cavazos et al., 1994). Mean neuronal density (Ni) for each counted region was calculated from the average nuclear count per reticule field (ni) obtained from the standardized septal and temporal sections according to the Floderus formula: Ni = n i x t / ( t + d - 2b), where
Stereological analysis Locations for neuron counting Neuronal counts were obtained in the dentate gyms and hippocampus in horizontal sections at two standarized locations along the septotemporal (dorsoventral) axis of the hippocampal formation, as in previous studies (Cavazos and Sutula, 1990; Cavazos et al., 1994). The location of the regions, which are
99
Ni is the corrected neuronal density, ni is the average nuclear count per reticule field, t is the section thickness, d is the average nuclear diameter at each of the counted regions, and b is a constant which represents the limit of optical resolution for each objective (Konigsmark, 1969; Cavazos and Sutula, 1990; Cavazos et al., 1994). For counts obtained with a 20× objective (n.a. 0.75), b = 0.36 I~m. The thickness (t) of each section was also determined by measuring the distance between upper and lower focal planes of the section with the stage micrometer using the 100 × oil immersion objective. The average nuclear diameter (d) for neurons in the hilus of the dentate gyms and hippocampal pyramidal neurons ranged from "--6.5 to 11 Ixm. The ratio of section thickness (t)/nuclear diameter (d) was always above 1.5 (range: 1.81-3.07), which supports the stereological assumptions of this method (Clarke, 1992). Volume measurements As systematic changes in tissue volume could alter the neuronal density, estimates of the volume of the hippocampal region (hippocampus proper, i.e., CA1, CA2, CA3, and dentate gyms were obtained from ipsilateral and contralateral hemispheres using a stereological approximation method. The volume of the combined hippocampus proper and dentate gyms in the temporal region (Vtot~) was measured through six consecutive sections, including the section chosen for counting, using an image analysis system. (Vtotal) was measured using the formula Vtotal : ~ n = l - 6 (An -q-An+l)t/2 where An is the area of the combined hippocampus proper and dentate gyms in the n-th section, An+l is the combined area of the next section, and t is the thickness of the consecutive sections. For calculation of the volume of the dentate gyms (VoG), An was measured by outlining the circumference of the dentate gyms from crest to crest along the hippocampal fissure, and connecting the tips with straight lines that excluded pyramidal neurons in CA3c. The hippocampal volume (VH) was calculated by subtraction according to the following formula: VH = Vtotal- VDG. For each rat, the volume corrected mean neuronal densities (Nvcor) of the hilus of the dentate gyms and hippocampal subfields of the septal and temporal locations were calculated according to the formula
Nvcor = Ni x Vx/Vc, where Vx is the estimated volume (Vto~, VH, or VoG) of the given rat, and Vc is the corresponding estimated mean volume of control rats. Statistical methods All measures were expressed as mean :t: SEM. The differences in nuclear counts per reticule field (ni), mean neuronal densities (NiL and volume-corrected neuronal densities (Nvcor) as a function of the number of evoked seizures were statistically analyzed using an ANOVA and were considered significant when P < 0.05. Individual comparisons were evaluated for significance using the Bonferroni correction. Results
Effects of brief repeated seizures on spatial memory performance Spatial memory was assessed by radial arm maze testing in 32 pairs of kindled rats and age-matched control rat, which have been reported in detail previously (Sutula et al., 1995). Kindled rats experienced a range of three evoked ADs to 134 Class V seizures. The age-matched control rat of each pair was implanted with an olfactory bulb electrode and was handled similarly, but received no stimulation. Kindled rats studied at 1 month after the last of 30134 evoked Class V seizures acquired competence in radial arm maze performance at a rate that was indistinguishable from controls, but demonstrated a deficit in the ability to repeat the task on consecutive days. At 1 month after the last evoked seizure, these performance deficits are unlikely to be affected by acute consequences of recent seizures. The spatial memory deficit was detected in rats with greater than 30 evoked Class V seizures, but was not observed in kindled rats that experienced three ADs or three Class V seizures. The severity of the deficit, as assessed by the number of reference errors in the radial arm maze, increased as a function of the number of evoked seizures (Sutula et al., 1995; see also Fig. 3A).
100 TABLE 1
Septal hippocampus Control
3ADs
3 Class V
30 Class V
69-75 Class V
>83 Class V
ANOVA
32 20.44-0.5
6 214- 1.2
6 18.84- 1
4 16.3-t- 1
8 17.55:0.9
8 16.3-t-0.8
F = 9.84,
290144-1115 28657 4- 1049
299924-3204 29659 4- 2053
226514-4237 24016 -I- 3791
201764-3586 22431 4- 2813
265804-2355 28254 4- 2658
233624-1435 26045 4-1115
32 9.24-0.2 1917724-4850 189941 4-4679
6 8.54-0.7 1743864- 15943 1701564- 11627
6 8.24-0.3 173414 4- 10832 1687804- 12165
4 9.14-0.3 1939704-5104 1991944-10481
8 9.34-0.5 1914414- 10718 192838 4- 12935
8 9.14-0.5 185449 4- 10993 184731 4-15178
32 194-0.4 2764864-7170 2737794-6677
6 21.84-0.9 3109104-14617 3056284- 12719
6 20.14-0.8 2892274-12651 2868134- 16032
4 18.74-1.4 2768734-24785 2817044-18366
8 18.84-0.9 2666054-9252 271343-4-13898
8 18.94-0.9 2708874-15773 2647124- 15133
32 10.14-0.2 2106514-6480 2087554-5803
6 10.54-0.7 2212264- 15589 2172964-17598
6 9.34-0.5 192011 4- 14098 1889684-13904
4 114-0.6 2383534- 11171 2432704-12228
8 10.74-0.7 2284864-11937 2261154-15601
8 11.44-0.7 2372734-16302 2341524-20469
32 25.3 4- 0.6 567022 4-16816 5596274-13672
6 25.2 4- 0.8 559413 4-20576 5505704-21184
6 24.1 4-1.3 568237 4-37920 5643334-37919
4 24.2 4-1.1 5591024- 31839 5737914-32925
8 25.3 4-1.0 552675 4-15886 5687264-28412
8 30.1 4-1.3 568728 4-18478 5588494-36905
32 224-0.6 5010644-17716 5066254-15319
6 21.94- 1.0 5083064-27552 5013564-27279
6 23.84-1.9 5951194-81945 5791504-62356
4 25.94-1.2 6262164-33420 6415154-40157
8 23.24- 1.3 5244584-34426 5384434-31166
8 23.64-1.0 5414824-28548 5347624-40319
Hilus n
Counts
P < 0.0001 Ni
Nv~,~
F = 1.28, P = 0.2816
CA3a n
Counts IV/
Nvcor CA3c n
Counts N~
Nvoot CA2 n
Counts
Ni
Nvcor CAla n
Counts Ni
Nvoo~ CAlc n
Counts Ni
NV~T
n, number of rats;
Ni,
neuronal density;
Nvoo,, volume
corrected neuronal density.
Effects of brief repeated seizures on hippocampal
neuronal density Spatial memory performance in the radial ann maze may be disrupted by hippocampal damage, so it was of interest to determine if the radial arm maze performance deficits, which increased as a function of the number of evoked seizures, were accompanied by hippocampal neuron loss. Neuronal densities in the hilus of the dentate gyms and the CA3, CA2, and CA1 subfields were measured at septal and temporal levels of the hippocampal formation at 7 days after
completing the maze task in the 32 pairs of kindled and age-matched control rats. There were decreases in hippocampal neuronal counts and volume-corrected neuronal densities (Nvcor) that developed as a function of the number of evoked seizures (see Tables 1 and 2, and Fig. 1), which confirmed previous reports of seizure-induced neuronal loss in other groups of kindled rats (Cavazos and Sutula, 1990; Cavazos et al., 1994). Decreases in NVcor in kindled rats were observed primarily in the temporal regions of the hippocampus, and included the hilus of the dentate gyms, CA3c, CAla,
101
TABLE 2 Temporal hippocampus Control
3ADs
3 Class V
30 Class V
69-75 Class V
>83 Class V
n Counts
32 37.74-0.7
6 36.64-1.5
6 39.1.1.5
4 38.3-1-1.7
8 33.34-1.0
8 31.9,1,1.3
Ni Nv cot
505384-1423 498984- 1 2 6 7
50466±1601 49574-4-1 5 2 0
49917-1-1941 5 4 0 8 3"1"2890
49866.1.754 53849"1"3 0 7 5
43384±1135 43537 ± 2772
426054-2578 4 7 7 9 54- 2435
n Counts
32 14.3+0.3
6 12.9.1.0.5
6 14.4,1,0.6
4 13.54-0.6
8 12.94-0.5
8 13.64-0.6
Ni
182988,1,4516 172740+7297 180102.1.4147 170475-4-7059
1797134-8159 177141,1,7374
171357+4161 1754544-8655
1634724-7365 1534234- 13310
1761664-7932 173121+7902
n Counts
32 16.64-0.3
6 15.74-0.8
6 17.1.0.7
4 15.34-0.9
8 15.6.1.0.5
8 14.94-0.6
Ni Nvcor
2328724-5782 229018-1-5044
228144-4-10525 231114,1,7961 2246154-10197 226723,1,8766
2117444-14748 216833-t-15869
2171404-8918 201944±15621
210943+9087 2074524-9322
32 12.54-0.3 170532.1.3786 168304,1,3815
6 12.5.1.0.4 177990-4-5541 175407-t-6527
6 13.8+0.8 182069±12669 178145~z8814
4 11.84-0.8 159715±11239 1631804-13134
8 12.74-0.8 170670±12971 162891,1,16649
8 12.5±0.4 1723414-3217 1701244-7947
n Counts
32 18.44-0.4
6 20.6,1, 1.2
6 19.84-1.2
4 18.3,1,0.7
8 15.64-1.0
8 17.5±0.8
Ni Nvcor
282203 4-6963 281290±6151
3250634-22476 321216± 19712
290200±22315 283751± 18777
2736864-17519 278240± 12326
246387,1,22464 222561±23423
267287± 12049 2666334- 17771
n Counts
32 30.14-0.5
6 30.3 + 1.1
6 34.1 ±2.1
4 30.4± 1.2
8 29.44- 1.4
8 27.7± 1.2
Ni
501065,1,10578 497861± 10274
5228644-25112 514945±23052
538912-t-34614 532884±32063
491926±11960 503206q-23873
475056zk24866 4539604-43607
459254±17399 449099± 18912
ANOVA
Hilus F = 12.6, P < 0.0001 F = 4.05, P = 0.02
CA3a
Nvcor
F = 2.95, P = 0.0559 F = 2.36, P = 0.099
CA3c F = 3.13, P = 0.0473 F=3.48, P = 0.034
CA2 n Counts
Ni Nvcor CAIa
F = 6.78, P = 0.0016 F = 5.75, P = 0.0042
CAIc
Nvcor
F = 3.84, P = 0.024 F = 4.51, P = 0.013
Abbreviations as in Table 1. and C A l c , where the seizure-induced decreases were r e s p e c t i v e l y , 9 2 % , 9 0 % , 87%, a n d 9 1 % o f c o n t r o l s .
Effects of brief repeated seizures on volume of the hippocampus and dentate gyrus R e d u c t i o n in n e u r o n a l d e n s i t y m a y b e c a u s e d b y n e u r o n a l loss or an i n c r e a s e in v o l u m e . T h e r e w a s
n o c h a n g e in the total v o l u m e o f t h e h i p p o c a m p u s i n c l u d i n g the C A 1 , C A 2 , C A 3 subfields a n d the d e n tate gyrus, ( F = 0.302, P = 0.74), but t h e r e w a s an i n c r e a s e in the v o l u m e o f the d e n t a t e g y r u s in kind l e d rats c o m p a r e d to c o n t r o l s ( F = 3.19, P < 0.045, Fig. 2). T h e v o l u m e i n c r e a s e in the d e n t a t e g y r u s w a s -'~7% in t h e k i n d l e d rats w h i c h e x p e r i e n c e d > 6 9 C l a s s V g e n e r a l i z e d t o n i c - c l o n i c s e i z u r e s (t = 2.35,
102
~0
hilus (temporal)
~
CA3¢ (temporal)
1=*
o ~
N
+~
-r
-o m
conlrol
comr~ =~o,-sea v u- +s4av Number of seizures
:; AOt -:;¢ CI V O - 1:;4C~ V
Number of seizures
E "
~
CAIa (temporal)
"
~
(onbol
$ ADZ-$$O V ~ - !114(:1V
E "+]
CA1 ¢ (temporal)
~"'1
¢onfrol
Number of seizures
$ ADI -lie CI I/ G) - t$4CI V
¢ordlrol
Nmnber of seizures
$ AO! - $ $ CI II ¢9 - 1 5 4 0 V
Number of seizures
Fig. 1. Repeated brief seizures evoked by kindling of the olfactory bulb in rats decreased volume corrected neuronal density in temporal regions of the hippocampus. Volume corrected neuronal density in the hilus of the dentate gyms, CA3c, CAla, and CAlc decreased after 69 secondary generalized tonic-clonic (Class V) seizures. As neuronal loss was observed in the hilus of the dentate gyms, CA3c, CAla, and CAlc, but not in CA2, the distribution of neuronal loss resembled the pattern of hippocampal sclerosis. The results demonstrated that hippocampal sclerosis was induced by brief repeated seizures. Asterisks indicate statistical significance: hilus, P = 0.02; CA3c, P =0.034; CAla, P =0.0042; CAlc, P =0.013.
onolJ
Total volume
1=01/ Dentate gyrus volume
i,,,1
I
+,,
++
I~
¢on~'ol
3ADz-$OCIV ~ - 1 ~ 4 C I V
Number of seizures
"
le
~+
Hinnhr.=zmt~lJ¢ u~hJm~
|,,, +,,
¢onkol
I~AOz-$0CIV | ) - I ~ I 4 C I V
Number of seizures
IS
¢on~'ol
3ADI-$eCIV
I;[~-154CIV
Number of seizures
Fig. 2. Alterations in volume of subregions of the hippocampal formation induced by repeated brief seizures. There was an increase in the volume of the dentate gyrns that achieved statistical significance after 69 Class V seizures (asterisk indicates P = 0.045, ANOVA; P < 0.05, Bonferoni correction). There was a trend toward reduction in volume of the hippocampus proper that did not achieve statistical significance.
P < 0.05, Bonferroni correction). This increase in volume in the dentate gyrus was accompanied by a non-significant trend to decrease in volume of the hippocampus proper, i.e., CA1, CA2, CA3, (~3%, F = 0.716, P = 0.49).
On the basis of this morphometric analysis of volume, the reduction in Nvco, in the hippocampus proper is most likely caused by neuronal loss. The increase in volume of the dentate gyrus after repeated seizures may contribute to the seizure-induced de-
103
B
A
¢o ¢J O
7O
120
~@ so
lOO
•~ 5o
so
O)
6O
e-
40,
Is ~ 30 ~ 2e .o
o , O• ~
,
.|i :t "t
. - 140 1
o
110q
o i
; ;e
• ,
Number
,;o
,;a ,;o ,;o 2;o
o
control
3ADs- 3 0 C I V
69- 134CIV
Number of seizures
of afterdischarges
Fig. 3. (A) Seizure-induceddeficits in radial arm maze performance as a function of the number of afterdischarges or evoked seizures. There was a strong correlationbetween the number of reference errors and the cumulativenumber of seizures (r = 0.69, P = 0.00001). Reference errors were defined as entry into unbaited arms of the radial arm maze (see Methods for additional details). Dark circles indicate rats from groups with no detectable cumulative neuronal loss. Open circles are rats from groups with significantneuronal loss. Rats with greater than 90 evoked afterdischarges had neuronal loss and also made more reference errors during radial arm maze testing. (B) Reference errors in electrode-implanted,age-matched control rats, in rats that experienced 30 or fewer Class V seizures, and rats with greater than 69 Class V seizures that had neuronal loss. There was a significantincrease in reference errors in rats with neuronal loss that developed after 69 Class V seizures (P < 0.0001, Mann- Whitney Rank Sum Test). crease in neuronal density in this region, but as there was a significant decrease in neuronal density of the dentate gyrus after correction for the increase in volume (Nvco,), neuronal loss probably also occurred in the dentate gyrus at the temporal level.
in the subgroup of kindled rats that experienced more than 69 Class V seizures and demonstrated significant neuronal loss (Fig. 3A,B).
Relationship of seizure-induced spatial memory deficits and neuronal loss
Stereological and morphometric analysis in kindled rats with characterized deficits in spatial memory function demonstrated that repeated brief seizures induced subfield specific hippocampal neuronal loss in a pattern resembling hippocampal sclerosis. The results not only confirmed previous studies, but also provided additional new evidence that relatively subtle seizure-induced neuronal loss in specific subpopulations of hippocampal neurons is associated with long-term, probably permanent deficits in spatial memory.
As reported previously, the number of reference errors, defined as entry into unbaited arms of the radial maze, increased as a function of the number of evoked ADs in kindled rats (r = 0.69, P = 0.00001, Fig. 3A). The maze performance of rats that experienced 3 ADs or 3 Class V seizures did not differ from their paired controls, but a difference in ability to repeat criterion performance on consecutive days was observed in kindled rats after ~ 4 0 evoked ADs, or ~ 3 0 or more Class V seizures. There were no significant differences in reference errors in rats that experienced 30 or fewer Class V seizures compared to controls, but more reference errors were observed
Discussion
Technical issues in the assessment of neuronal loss There are substantial theoretical and technical questions that must be addressed in analysis of neuron
104 number in the central nervous system. The so-called unbiased methods offer some potential advantages, particularly when variations in size and shape of the objects of interest may influence accuracy of counts, or when sampling bias in a structure with cellular heterogeneity might result in inaccurate estimates of cell number in the overall structure (West, 1999; see also West, 2002, this volume). While the so-called unbiased methods have a number of potential advantages, there has been a range of viewpoints about their application and relative merits compared to older conventional counting methods (Guillery and Hermp, 1997; see also Guillery and August, 2002, this volume). Several methodological aspects of this study were informative regarding the sources of error in estimating neuronal density using counting techniques, and also provided some insight into the sources of variability among studies applying these techniques for assessment seizure-induced alterations in the hippocampus. Inspection of Tables 1 and 2 revealed that the major source of variability arose in the calculation of volume in the region of interest. In comparison to our previous studies, the measurements for the calculation of estimated volume employed greater sampling which would be expected to improve accuracy, and confirmed previously published reports of an increase in the volume of the dentate gyms in some rodent models of epilepsy (Adams et al., 1997; Bouilleret et al., 2000). These results suggested that the seizure-induced reduction in neuronal density in this region may not be caused by neuronal loss, but the significant reduction in neuronal density after volume correction (Nvcor) in this study nevertheless implied that neuronal loss also occurred in the dentate gyms. Importantly, there was no increase in volume of the hippocampus proper, but rather a non-significant trend to a decrease in volume, which supports the interpretation that neuronal loss occurred in the hippocampus. With the significant increase in volume of the dentate gyrus accompanied by a trend to smaller volume of the hippocampus proper, there was no overall change in the combined volume of the dentate gyrus and hippocampus, as reported previously (Cavazos et al., 1994). The unbiased methods, which offer some methodological advantages, have also provided evidence supporting seizure-induced neuronal loss (Dalby et al., 1998).
Conventional stereological methods employed in this and previous studies of kindled rats and unbiased methods thus support the viewpoint that repeated brief seizures induce hippocampal neuronal loss.
Spatial features of hippocampal neuronal loss induced by brief repeated seizures: relationship to hippocampal sclerosis and site of seizure initiation This analysis and previous studies of the effects of kindled seizures on hippocampal neuronal populations using stereological counting techniques (Cavazos et al., 1994) have demonstrated that repeated brief seizures evoked by kindling induced a consistent pattern of neuronal loss resembling human hippocampal sclerosis. Kindled rats that experienced seizures evoked exclusively by stimulation of the olfactory bulb demonstrated neuronal loss in the temporal regions of CA3c, CAla, CAlc, and the hilus of the dentate gyms, with sparing of CA2, a subfield which is known to be resistant to hypoxia and is also relatively preserved in human hippocampal sclerosis (Mouritzen Dam, 1980; Babb et al., 1984; Sloviter, 1989; Gloor, 1991). This distinctive subfield specific pattern was also observed in kindled rats that received perforant path or amygdala stimulation (Cavazos et al., 1994). The cumulative damage in hippocampal subfields in kindled rats that received olfactory bulb stimulation was apparent in the temporal hippocampus, which contrasts with the more widespread neuronal loss involving both septal and temporal regions in kindled rats that experienced seizures evoked by perforant path or amygdala stimulation (Cavazos et al., 1994). These observations suggest that the distribution of seizure-induced neuronal loss is dependent on the site of seizure initiation and pathways that are activated by the epileptogenic process. Compared to our previous studies using the same counting techniques, the amount of damage in rats experiencing repeated seizures evoked by olfactory bulb stimulation was less than in rats kindled by stimulation of the perforant path. The differences in seizureinduced damage in the studies may be caused by variability in the technical aspects of the analysis, but the use of the same techniques in both studies, the consistency of the pattern of loss in specific hippocampal subfields in both studies, and the differ-
105 ences in the septal and temporal distribution of the neuron loss suggest that the seizure-induced damage was pathway-specific and cannot be explained merely by variability or technical factors. These observations are of interest in view of the reported differences in damage in human temporal lobe epilepsy between tumor associated cases and idiopathic cases (Kim et al., 1990). While this study assessed neuronal densities only in the hippocampus, previous analysis in groups of rats experiencing seizures evoked by perforant path and amygdala stimulation demonstrated neuronal loss in extrahippocampal regions including the entorhinal cortex and endopyriform nucleus (Cavazos et al., 1994), which is consistent with imaging and pathological observations in human epileptic temporal lobe indicating that damage may involve extra-hippocampal and other limbic areas, such as the entorhinal cortex, lateral temporal cortex, and regions beyond the hippoeampal formation (DeCarli et al., 1998; Kalviainen et al., 1998; Lee et al., 1998; Bernasconi et al., 1999).
Evidence that repeated brief seizures induce progressive hippocampal neuronal loss Neuronal loss was detected in this study only after >69 Class V seizures evoked by olfactory bulb stimulation. In contrast, neuronal loss was initially detected after 30 repeated seizures evoked by perforant path or amygdala stimulation, and progressively increased after additional seizures (Cavazos and Sutula, 1990; Cavazos et al., 1994). An important question is whether this neuronal loss, which occurs in a pattern that appears to be pathway-specific, begins only after a series of repeated seizures, or is a consequence of each seizure. Studies using the TUNEL method, which detects of DNA fragmentation and is a marker for apoptosis, have demonstrated that even a single or a few brief seizures induce apoptotic death in neurons of the dentate gyrus and hippocampus (Bengzon et al., 1997; Pretel et al., 1997; Zhang et al., 1998). These observations directly support the observation of cumulative seizure-induced neuronal loss in the dentate gyms and hippocampus as detected by stereological methods. Furthermore, multiple studies in experimental models suggest that cumulative, gradually progressive neuronal loss is
likely to be a consequence of the repeated brief seizures (for a summary, see Table 3). While the direct observation of seizure-induced apoptosis after single or a few seizures suggests that each seizure produces neuronal damage, the stereological methods employed in this and previous studies detected cumulative neuronal loss only after multiple seizures. This may reflect the limited sensitivity of the counting methods, which are clearly subject to substantial variability due to a range of potential errors, particularly calculation of volume changes, which may influence neuronal density. Seizure-induced neurogenesis may also be a factor that contributes to the cumulative effect of seizures on hippocampal neuronal populations (Bengzon et al., 1997; Parent et al., 1997; Parent et al., 1998; see also Parent and Lowenstein, 2002, this volume). With direct evidence of apoptotic neuronal death after a few seizures and the evidence of evolving loss in multiple stereological studies, it seems most likely that individual seizures progressively induce neuronal loss that cumulatively results in a pattern of damage resembling hippocampal sclerosis. The distribution of this loss may depend on the pathways activated by the site of seizure initiation, and raise that possibility that specific functional abnormalities may result from cumulative seizure-induced neuronal loss. It is of interest that repeated seizures appear to have different effects on hippocampal neurons and circuits after a previous episode of status epilepticus. After neuronal damage, induced by status epilepticus, spontaneous recurrent seizures have not consistently induced additional neuronal loss (see Dudek et al., 2002, this volume and Pitk~inen et al., 2002, this volume). The reasons for differences between these studies of repeated brief seizures and recurrent spontaneous seizures after status epilepticus are uncertain, but several possibilities deserve consideration. First, the secondary generalized tonic-clonic seizures evoked by kindling are typically several minutes in duration, which is longer than the recurring seizures that follow some models of status epilepticus (see Pitk~inen et al., 2002, this volume). Second, the initial prolonged seizures may have damaged the most vulnerable neuronal subpopulations, leaving surviving neurons that are relatively resistant to additional seizure-induced loss. Third, vulnerabil-
106 TABLE 3 Experimental studies assessing neuronal loss after repeated brief seizures Model
Site of seizure induction
Numberof seizures
Method of assessment
Sampled locations and results (bold indicates loss or damage)
Reference
Kindling
perforant path
3-30 secondary generalized tonic--clonic seizures
stereological
neuronal loss in hilus of the dentate gyrus
Cavazos and Sutula (1990)
Kindling (rapid)
amygdala
~1500 seizures
stereological
decreased neuronal density and increased volume of dentate gyrus
Bertram and Lothman (1993)
Kindling
perforant path amygdala olfactory bulb
3-150 secondary generalized tonic-clonic seizures
stereological
neuronal loss in hilus of the dentate gyrus, CA1, CA3, entorllinal cortex
Cavazos et al. (1994)
Kindling
amygdala
3 secondary generalized tonic-clonic seizures
grid counting
decreased neuronal density and increased volume of dentate gyrus
Adams et al. (1997)
Kindling (rapid)
hippocampus
1 afterdischarge and 2-3 secondary generalized tonic-clonic seizures
TUNEL staining for apoptosis
increase in apoptotie cells in the hilus of dentate gyrus after 1 AD (~214%) and 2-3 generalized seizures (~415%)
Bengzon et al. (1997)
Kindling
entorhinal cortex
5-85 secondary generalized tonic-clonic seizures
TUNEL staining for apoptosis; silver degeneration staining
apoptosis in hilus, CA1, granule cells, subiculum, neocortex
Pretel et ai. (1997)
Kindling
amygdala
1-20 partial and secondary generalized tonic--clonicseizures
TUNEL staining, Bax/Bcl ratio
30-80% increase in apoptotic cells after 1 or 20 seizures
Zhang et al. (1998)
Kindling
perforant path
5 secondary generalized tonic--clonicseizures
unbiased stereological
neuronal loss in the hilus of the dentate gyrus
Dalby et al. (1998)
Kindling
amygdala
5 secondary generalized tonic--clonicseizures
unbiased stereological
Tuunanen and no loss in amygdala or hilus of the dentate gyrus Pitk~inen(2000)
Kindling
olfactory bulb
3-134 secondary generalized tonic-clonic seizures
stereological
neuronal loss in temporal hilus, CA1, CA3
Kotloski et al. (2002)
Spontaneous seizures after status epilepticus
amygdala
as many as 6000 spontaneous seizures
stereological, fluoro-Jade B immunochemistry, MRI
no effect of repeated spontaneous seizures
Pitk~inenet al. (2002)
ity to additional damage may be altered by previous seizures, as suggested by the apparent protective effects of kindled seizures against damage induced by status epilepticus in kindled rats (Kelley and Mclntyre, 1994). These experimental studies suggest that the effects of seizures may vary as a function of the previous activity and seizures in neural circuitry.
after status epilepticus
Relationship between repeated brief seizures, neuronal loss, and memory dysfunction The possible cognitive effects of seizure-induced neuronal loss are of potentially major clinical significance. In rats that experienced repeated brief seizures evoked by stimulation of the olfactory bulb,
107 impairment of spatial memory became apparent after ~30 secondary generalized (Class V) seizures, and increased in severity as a function of the cumulative number of seizures. The maze performance of kindied rats that experienced 30 or fewer evoked Class V seizures (or <40-50 ADs in Fig. 3), as assessed by reference errors or ability to repeat the criterion task, did not differ from controls without seizures. In contrast, kindled rats with >69 Class V seizures (or ~80-90 ADs in Fig. 3) that had significant cumulative seizure-induced neuronal loss also demonstrated spatial memory deficits. The results thus provide evidence of an association between the development of seizure-induced memory dysfunction and loss of hippocampal neurons. Although the relationship between hippocampal damage and memory dysfunction has been recognized for many years on the basis of human observations including the patient H.M. (Scoville and Milner, 1957; Corkin et al., 1997) and numerous rodent and primate studies (Squire and Zola, 1997), this study provides quantitative evidence of an association between the induction of relatively subtle, subfield specific seizure-induced damage in the rodent hippocampus and spatial memory dysfunction. These observations confirm the importance of the hippocampus in the acquisition of short-term memories, and have potential clinical significance in regard to the consequences of repeated or uncontrolled seizures.
Implications for human epilepsy The observations of this combined behavioral and anatomical study have potentially important implications for human epilepsy. Human imaging studies have directly demonstrated that hippocampal atrophy may follow an initial precipitating injury such a prolonged or complex febrile seizures (Vanlandingham et al., 1998; see also Lewis et al., 2002, this volume), but the possibility that repeated seizures may be contributing to progressive neuronal loss has also been suggested by multiple imaging studies demonstrating that the atrophy increases with the duration of the epilepsy (Davies et al., 1996; Lee et al., 1998; Tasch et al., 1999; Theodore et al., 1999; Fuerst et al., 2001), and in some cases occurs without an initial precipitating injury (Briellmann et al., 2001). Human MRI observations are consistent with the possibility
that hippocampal sclerosis, in addition to being the sequelae of an initial injury, may also be an acquired lesion that is at least partially caused by poorly controlled seizures (Mathem et al., 1996; Tasch et al., 1999). This experimental study strongly supports the view that brief repeated seizures cumulatively produce hippocampal damage which is accompanied by memory dysfunction, and thus suggests that hippocampal sclerosis could in part be caused by poorly treated epilepsy. This possibility, particularly with respect to the associated seizure-induced memory dysfunction, may be important in regard to cases of medically intractable or poorly controlled epilepsy, which is accompanied by memory dysfunction (Jokeit and Ebner, 1999; see the following in this volume: Austin and Dunn, 2002; Hermann et al., 2002; Helmstaedter, 2002; Jokeit and Ebner, 2002; Sutula and Pitk~inen, 2002; Treiman, 2002). In considering the implications of these experimental observations for human epilepsy, the following additional points deserve emphasis. First, the majority of patients with temporal lobe epilepsy experience recurring partial complex rather than secondary generalized seizures, and the overwhelming majority of seizures evoked by kindling are partial with secondary generalization. As these seizures are likely to produce more significant metabolic stress, it would be valuable to determine if repeated partial seizures also induce neuronal loss. This study using kindled rats could not discriminate whether recurring partial seizures (Class I-IV) might also be inducing neuronal loss, but this question could perhaps be addressed using markers for apoptosis. Second, while the results using markers for apoptosis and stereological methods both support the viewpoint that individual seizures are likely to induce neuronal loss, memory dysfunction was observed only after many seizures, which might seem to suggest that there is a threshold for seizure-induced cognitive dysfunction. In regard to this apparent threshold for induction of memory dysfunction (>30 secondary generalized seizures), hippocampal circuits may have a capacity to maintain normal function after some number of seizures or other adverse injurious events, but emergent properties of these complex circuits, such as memory, may become affected only after a significant proportion of the components of the circuits are damaged.
108 Stereological analysis and m e t h o d s for detection o f apoptosis strongly support the v i e w p o i n t that n e u ronal death is n o t o n l y a c o n s e q u e n c e o f status epilepticus, but also is i n d u c e d b y repeated brief seizures. T h e association of s e i z u r e - i n d u c e d hipp o c a m p a l n e u r o n a l loss with spatial m e m o r y i m p a i r m e n t i m p l i e s that the c o n s e q u e n c e s o f n e u r o n a l loss from repeated brief seizures are n o t b e n i g n . In e x p e r i m e n t a l m o d e l s o f status epilepticus, p r o m p t t r e a t m e n t that suppresses seizures reduces or prevents l o n g - t e r m c o n s e q u e n c e s of the seizures, such as d a m a g e in specifically v u l n e r a b l e n e u r o n s in the dentate gyms, associated m o s s y fiber sprouting, behavioral abnormalities, and i n c r e a s e d susceptibility to additional seizures (Sutula et al., 1992; Y l i n e n et al., 1991). These observations suggest that p r o m p t effective t r e a t m e n t o f repeated brief seizures m a y prevent n e u r o n a l loss a n d r e o r g a n i z a t i o n o f hipp o c a m p a l circuitry that contribute to m e m o r y dysf u n c t i o n and p r o m o t e s susceptibility to additional seizures. T h e adverse c o n s e q u e n c e s o f repeated brief seizures strongly support the i m p o r t a n c e of u r g e n t t r e a t m e n t a n d a c h i e v e m e n t o f c o m p l e t e seizure control as therapeutic goals for this c o m m o n disorder.
References Adams, B., Sazgar, M., Osehobo, P., Vanderzee, C., Diamond, J., Fahnestock, M. and Racine, R.J. (1997) Nerve growth factor accelerates seizure development, enhances mossy fiber sprouting, and attenuates seizure-induced decreases in neuronal density in the kindling model of epilepsy. J. Neurosci., 17(14): 5288-5296. Austin, J.K. and Dunn, D.W. (2002) The neurodevelopmental impact of childhood onset temporal lobe epilepsy on brain structure and function and the risk of progressive cognitive effects. In: T. Sutula and A. Pitk~inen (Eds.), Do Seizures Damage the Brain. Progress in Brain Research, Vol. 135. Elsevier, Amsterdam, pp. 419-427. Babb, T., Pretorius, W., Davenport, C., Lieb, J. and Crandall, P. (1984) Temporal lobe volumetric cell densities in temporal lobe epilepsy. Epilepsia, 25: 721-732. Bengzon, J., Kokaia, Z., Elmer, E., Nanobashvili, A., Kokaia, M. and Lindvall, O. (1997) Apoptosis and proliferation of dentate gyms neurons after single and intermittent limbic seizures. Proc. Natl. Acad. Sci. USA, 94(19): 10432-10437. Bernasconi, N., Bernasconi, A., Andermann, E, Dubeau, E, Feindel, W. and Reutens, D.C. (1999) Entorhinal cortex in temporal lobe epilepsy - - a quantitative MRI study. Neurology, 52: 1870. Bertram, E. and Lothman, E. (1993) Morphometric effects of
intermittent kindled seizures and limbic status epilepticus in the dentate gyrus of the rat. Brain Res., 603(1): 25-31. Bouilleret, V., Loup, E, Kiener, T., Marescaux, C. and Fritschy, J. (2000) Early loss of interneurons and delayed subunit specific changes in GABAa-receptor expression in a mouse model of mesial temporal lobe epilepsy. Hippocampus, 10(3): 305-324. Briellmann, R., Newton, M., Wellard, M. and Jackson, G. (2001) Hippocampal sclerosis following brief generalized seizures in adulthood. Neurology, 57: 318-320. Cascino, G.D., Jack Jr., C.R., Parisi, J.E., Sharbrough, EW., Hirschorn, K.A., Meyer, EB., Marsh, W.R. and PC, O.B. (1991) Magnetic resonance imaging-based volume studies in temporal lobe epilepsy: pathological correlations. Ann. NeuroL, 30(1): 31-36. Cavazos, J.E. and Sutula, T. (1990) Progressive neuronal loss induced by kindling: a possible mechanism for mossy fiber synaptic reorganization and hippocampal sclerosis. Brain Res., 527: 1-6. Cavazos, J.E., Golarai, G. and Sutula, T. (1991) Mossy fiber synaptic reorganization induced by kindling: time course of development, progression, and permanence. J. Neurosci., 11: 2795-2803. Cavazos, J.E., Das, I. and Sutula, T. (1994) Neuronal loss induced in limbic pathways by kindling: evidence for induction of hippocampal sclerosis by repeated brief seizures. J. Neurosci., 14(5): 3106-3121. Clarke, P.G. (1992) How inaccurate is the Abercrombie correction factor for cell counts?. Trends Neurosci., 15:211-212. Corkin, S., Amaral, D.G., Gonzalez, R.G., Johnson, K.A. and Hyman, B.T. (1997) H.M.'s medial temporal lobe lesion: findings from magnetic resonance imaging. J. Neurosci., 17(10): 3964-3979. Dalby, N.O., West, M. and Finsen, B. (1998) Hilar somatostatinmRNA containing neurons are preserved after perforant path kindling in the rat. Neurosci. Lett., 255(1): 45-48. Davies, K.G., Hennann, B.P., Dohan Jr., F.C., Foley, K.T., Bush, A.J. and Wyler, A.R. (1996) Relationship of hippocampal sclerosis to duration and age of onset of epilepsy, and childhood febrile seizures in temporal lobectomy patients. Epilepsy Res., 24(2): 119-126. DeCarli, C., Hatta, J., Fazilat, S., Fazilat, S., Gaillard, W.D. and Theodore, W.H. (1998) Extratemporal atrophy in patients with complex partial seizures of left temporal origin. Ann. Neurol., 43(1): 41-45. Dudek, F.E., Hellier, J.L., Williams, P.A., Ferraro, D.J. and Staley, K.J. (2002) The course of cellular alterations associated with the development of spontaneous seizures after status epilepticus. In: T. Sutula and A. Pitkiinen (Eds.), Do Seizures Damage the Brain. Progress in Brain Research, Vol. 135. Elsevier, Amsterdam, pp. 53-65. Fuerst, D., Shah, J., Kupsky, W., Johnson, R., Shah, A., HaymanAbello, B., Ergh, T., Poore, Q., Canada, A. and Watson, C. (2001) Is hippocampal sclerosis a progressive disorder? A volumetric MRI, pathological and neuropsychological study. Neurology, 57: 184-188. Gloor, P. (1991) Mesial temporal sclerosis: historical background and an overview from a modern prospective. In: H. Luders
109
(Ed.), Epilepsy Surgery. Raven Press, New York, NY, pp. 689-703. Guillery, R.W. and August, B.K. (2002) Doubt and uncertainty in counting. In: T. Sutula and A. Pitkanen (Eds.), Do Seizures Damage the Brain. Progress in Brain Research, Vol. 135. Elsevier, Amsterdam, pp. 25-42. Guillery, R.W. and Herrup, K. (1997) Quantification without pontification: choosing a method for counting objects in sectioned tissues. J. Comp. Neurol., 386(1): 2-7. Helmstaedter, C. (2002) Effects of chronic epilepsy on declarative memory systems. In: T. Sutula and A. Pitk~inen (Eds.), Do Seizures Damage the Brain. Progress in Brain Research, Vol. 135. Elsevier, Amsterdam, pp. 439-453. Hermann, B.P., Seidendberg, M. and Bell, B. (2002) The neurodevelopmental impact of childhood onset temporal lobe epilepsy on brain structure and function and the risk of progressive cognitive effects. In: T. Sutula and A. Pitk~inen (Eds.), Do Seizures Damage the Brain. Progress in Brain Research, Vol. 135. Elsevier, Amsterdam, pp. 429-438. Holmes, G.L., Khazipov, R. and Ben-Aft, Y. (2002) Seizureinduced damage in the developing human: relevance of experimental models. In: T. Sutula and A. Pitk~inen (Eds.), Do Seizures Damage the Brain. Progress in Brain Research, Vol. 135. Elsevier, Amsterdam, pp. 321-334. Jokeit, H. and Ebner, A. (1999) Long term effects of refractory temporal lobe epilepsy on cognitive abilities: a cross sectional study. J. Neurol. Neurosurg. Psychiatry, 67(1): 44-50. Jokeit, H. and Ebner, A. (2002) Effects of chronic epilepsy on intellectual functions. In: T. Sutula and A. Pitkanen (Eds.), Do Seizures Damage the Brain. Progress in Brain Research, Vol. 135. Elsevier, Amsterdam, pp. 455-463. Kalviainen, R., Salmenpera, T., Partanen, K., Vainio, P., Riekkinen, P. and Pitkanen, A. (1998) Recurrent seizures may cause hippocampal damage in temporal lobe epilepsy. Neurology, 50(5): 1377-1382. Kelley, M.E. and Mclntyre, D.C. (1994) Hippocampal kindling protects several structures from the neuronal damage resulting from kainic acid induced status epilepticus. Brain Res., 634(2): 245-256. Kim, J., Guimaraes, P.O., Shen, M.Y., Masukawa, L.M. and Spencer, D. (1990) Hippocampal neuronal density in temporal lobe epilepsy with and without gliomas. Acta Neuropathol., 80(1): 41-45. Konigsmark, B. (1969) Methods for the counting of neurons. In: W. Nauta and S. Ebbesson (Eds.), Contemporary Research Methods in Neuroanatomy. Springer Verlag, Berlin, pp. 315340. Kotloski, R., Lynch, M., Lauersdorf, S. and Sutula, T. (2002) Repeated brief seizures induce progressive hippocampal neuron loss and memory deficits. In: T. Sutula and A. Pitk~inen (Eds.), Do Seizures Damage the Brain. Progress in Brain Research, Vol. 135. Elsevier, Amsterdam, pp. 95-110. Lee, J.W., Andermann, F., Dubeau, F., Bernasconi, A., MacDonald, D., Evans, A. and Reutens, D.C. (1998) Morphometric analysis of the temporal lobe in temporal lobe epilepsy. Epilepsia, 39(7): 727-736. Lewis, D.V., Barboriak, D., MacFall, J.R., Provenzale, J.M.,
Mitchell, T.V. and van Landingham, K.E. (2002) Do prolonged febrile seizures produce medial temporal sclerosis? Hypotheses, MRI evidence and unanswered questions. In: T. Sutula and A. Pitkanen (Eds.), Do Seizures Damage the Brain. Progress in Brain Research, Vol. 135. Elsevier, Amsterdam, pp. 263-278. Margerison, J.H. and Corsellis, J.A. (1966) Epilepsy and the temporal lobes. A clinical, electroencephalographic and neuropathological study of the brain in epilepsy, with particular reference to the temporal lobes. Brain, 89(3): 499-530. Mathern, G.W., Babb, T., Leite, J., Pretorius, J., Kuhlman, P., Spradlin, S. and Mendoza, D. (1996) The pathogenic and progressive features of chronic human hippocampal epilepsy. Epilepsy Res., 26: 151-161. Meldrum, B., Vigouroux, R. and Brierley, J. (1973) Systemic factors and epileptic brain damage: prolonged seizures in paralyzed artificially ventilated baboons, Arch. Neurol. Psychiatry, 29: 82-87. Mouritzen Dam, A. (1980) Epilepsy and neuron loss in the hippocampus. Epilepsia, 21: 617-629. Parent, J.M. and Lowenstein, D.H. (2002) Seizure-induced neurogenesis: are more new neurons good for an adult brain? In: T. Sutula and A. Pitk~inen (Eds.), Do Seizures Damage the Brain. Progress in Brain Research, Vol. 135. Elsevier, Amsterdam, pp. 121-132. Parent, J.M., Yu, T.W., Leibowitz, R.T., Geschwind, D.H., Sloviter, R.S. and Lowenstein, D.H. (1997) Dentate granule cell neurogenesis is increased by seizures and contributes to aberrant network reorganization in the adult rat hippocampus. J. Neurosci., 17(10): 3727-3738. Parent, J.M., Janumpalli, S., McNamara, J.O. and Lowenstein, D.H. (1998) Increased dentate granule cell neurogenesis following amygdala kindling in the adult rat. Neurosci. Lett., 247(1): 9-12. Pitk~nen, A., Nissinen, J., Nairism/igi, J., Lukasiuk, K., Grrhn, O.H.J., Miettinen, R. and Kauppinen, R. (2002) Progression of neuronal damage after status epilepticus and during spontaneous seizures in a rat model of temporal lobe epilepsy. In: T. Sutula and A. Pitk~inen (Eds.), Do Seizures Damage the Brain. Progress in Brain Research, Vol. 135. Elsevier, Amsterdam, pp. 67-83. Pretel, S., Applegate, C.D. and Piekut, D. (1997) Apoptotic and necrotic cell death following kindling induced seizures. Acta Histochem., 99(1): 71-79. Sagar, H. and Oxbury, J. (1987) Hippocampal neuronal loss in temporal lobe epilepsy: correlation with early childhood convulsions. Ann. Neurol., 22: 334-340. Salmenpera, T., Kalviainen, R., Partanen, K. and Pitkanen, A. (1998) Hippocampal damage caused by seizures in temporal lobe epilepsy. Lancet, 351 (9095): 35. Scoville, W. and Milner, B. (1957) Loss of recent memory after bilateral hippocampal lesions. J. Neurol. Neurosurg. Psychiatry, 20:11-21. Sloviter, R. (1989) Calcium-binding protein (calbindin-D28) and parvalbumin immunocytochemistry: localization in the rat hippocampus with specific reference to the selective vulnerability
110
of hippocampal neurons to seizure activity. J. Comp. Neurol., 280: 183-196. Squire, L.R. and Zola, S.M. (1997) Amnesia, memory and brain systems. PhiL Trans. R. Soc. Lond., 352(1362): 1663-1673. Sutula, T. and Pitk~inen, A. (2002) Summary: Neuropsychological consequences of human epilepsy. In: T. Sutula and A. Pitk~inen (Eds.), Do Seizures Damage the Brain. Progress in Brain Research, Vol. 135. Elsevier, Amsterdam, pp. 465-467. Sutula, T. and Steward, O. (1986) Quantitative analysis of synaptic potentiation during kinding of the perforant path. J. NeurophysioL, 56: 732-745. Sutula, T., He, X.X., Cavazos, J. and Scott, G. (1988) Synaptic reorganization in the hippocampus induced by abnormal functional activity. Science, 239:1147-1150. Sutula, T., Cavazos, J.E. and Golarai, G. (1992) Alteration of long-lasting structural and functional effects of kainic acid in the hippocampus by brief treatment with phenobarbital. J. Neurosci., 12(11): 4173-4187. Sutula, T., Lauersdorf, S., Lynch, M., Jurgella, C. and Woodard, A. (1995) Deficits in radial arm maze performance in kindled rats: evidence for long-lasting memory dysfunction induced by repeated brief seizures. J. Neurosci., 15(12): 8295-8301. Tasch, E., Cendes, E, Li, L.M., Dubeau, E, Andermann, E and Arnold, D.L. (1999) Neuroimaging evidence of progressive neuronal loss and dysfunction in temporal lobe epilepsy. Ann. Neurol., 45(5): 568-576. Theodore, W.H. and Gaillard, W.D. (2002) Neuroimaging and the progression of epilepsy. In: T. Sutula and A. Pitkfinen (Eds.), Do Seizures Damage the Brain. Progress in Brain
Research, Vol. 135. Elsevier, Amsterdam, pp. 305-313. Theodore, W.H., Bhatia, S., Hatta, J., Fazilat, S., DeCarli, C., Bookheimer, S.Y. and Gaillard, W.D. (1999) Hippocampal atrophy, epilepsy duration, and febrile seizures in patients with partial seizures. Neurology, 52(1): 132-136. Treiman, D.M. (2002) Will brain damage after status epilepticus be history in 2010? In: T. Sutula and A. Pitk~inen (Eds.), Do Seizures Damage the Brain. Progress in Brain Research, Vol. 135. Elsevier, Amsterdam, pp. 471-478. Tuunanen and Pitk~inen, A. (2000) Do seizures cause neuronal damage in rat amygdala kindling? Epilepsy Res., 39:171-176. Vanlandingham, K.E., Heinz, E.R., Cavazos, J.E. and Lewis, D.V. (1998) Magnetic resonance imaging evidence of hippocampal injury after prolonged focal febrile convulsions. Ann. Neurol., 43(4): 413-426. West, M.J. (1999) Stereological methods for estimating the total number of neurons and synapses: issues of precision and bias. Trends Neurosci., 22(2): 51-56. West, M.J. (2002) Design based stereological methods for counting neurons. In: T. Sutula and A. Pitk~inen (Eds.), Do Se&ures Damage the Brain. Progress in Brain Research, Vol. 135. Elsevier, Amsterdam, pp. 43-51. Ylinen, A., Miettinen, R., Pitkanen, A., Guyla, A., Freund, T. and Riekkinen, E (1991) Enhanced GABAergic inhibition preserves hippocampal structure and function in a model of epilepsy. Proc. Natl. Acad. Sci. USA, 88: 7650-7653. Zhang, L.X., Smith, M., Li, X., Weiss, S. and Post, R.M. (1998) Apoptosis of hippocampal neurons after amygdala kindled seizures. Mol. Brain Res., 55(2): 198-208.
T. Sutula and A. Pitk~inen (Eds.) Progress in Brain Research, Vol. 135 O 2002 Elsevier Science B.V. All rights reserved
CHAPTER 9
Neuronal apoptosis after brief and prolonged seizures Johan Bengzon 1,,, Paul Mohapel
2, Christine T. Ekdahl 2 and Olle Lindvall 2
I Department of Neurosurgery, University Hospital, S-221 85 Lund, Sweden 2 Section of Restorative Neurolog), Wallenberg Neuroscience Center, BMC A-I1, S-221 84 Lund, Sweden
Abstract: Evidence has accumulated that apoptotic cell death contributes to brain damage following experimental seizures.
A substantial number of degenerating neurons within limbic regions display morphological features of apoptosis following prolonged seizures evoked by systemic or local injections of kainic acid, systemic injections of pilocarpine and sustained stimulation of the perforant path. Although longer periods of seizures consistently result in brain damage, it has previously not been clear whether brief single or intermittent seizures lead to cell death. However, recent results indicate that also single seizures lead to apoptotic neuronal death. A brief, non-convulsive seizure evoked by kindling stimulation was found to produce apoptotic neurons bilaterally in the rat dentate gyrus. The mechanism triggering and mediating apoptotic degeneration is at present being studied. Alterations in the expression and activity of cell-death regulatory proteins such as members of the Bcl-2 family and the cysteinyl aspartate-specific proteinase (caspase) family occur in regions vulnerable to cell degeneration, suggesting an involvement of these factors in mediating apoptosis following seizures. Findings of decreased apoptotic cell death following administration of caspase inhibitors prior to and following experimentally induced status epilepticus, further suggest a role for caspases in seizure-evoked neuronal degeneration. Intermediate forms of cell death with both necrotic and apoptotic features have been found after seizures and investigation into the detailed mechanisms of the different forms of cell degeneration is needed before attempts to specific prevention can be made.
Introduction
The mechanism of epileptic cell damage has for long been attributed to excitotoxicity-induced necrosis. However, in recent years, evidence has accumulated that neurons also die by apoptosis following seizures. The term apoptosis was introduced by Kerr et al. (1972). The original definition of apoptosis was based on morphological criteria including cell shrinkage, condensation, blebbing of the cytoplasmand nuclear membranes, and budding off of cellular fragments. However, apoptosis is now often used synonymously with programmed cell death. Programmed cell death constitutes a cell-autonomous, * Correspondence to: J. Bengzon, Department of Neurosurgery, University Hospital, S-221 85 Lund, Sweden. Tel.: +46-46-171580; Fax: +46-46-2220560; E-mail: johan.bengzon @neurokir.lu.se
controlled, and active lysis of single cells requiring de novo protein synthesis (Kerr et al., 1972; Wyllie et al., 1980; Clarke, 1990). During the development of the nervous system, neurons that die through programmed cell death display the morphological criteria of apoptosis. The most widely used techniques for the study of apoptosis are in situ terminal deoxynucleotidyltransferase-mediated dUTP nick-end labeling (TUNEL) of fragmented DNA, sometimes in combination with a DNA stain, in tissue sections and agarose gel electrophoresis of fragmented DNA from homogenized tissue samples. Necrosis, in contrast, deletes groups or clusters of cells and is characterized by a random breakdown of cellular constituents due to energy failure. Swelling and lysis of the cell is often followed by an inflammatory reaction and injury to the surrounding tissue. Importantly, modes of cell degeneration intermediate between apoptosis and necrosis have
112
TABLE 1 Summary of the literature characterizing the patterns of apoptosis following a variety of animal models of epilepsy and status epilepticus Method of seizure induction
Reference
Method of apoptotic detection
Time after insult
TUNEL
5h
hippocampal GCL and SGZ
TUNEL and propidium
0.5-4 h
hippocampal GCL, SGZ, and hilus
TUNEL and biotin
20-72 h
TUNEL
24 h, 2 weeks
Bax and Bcl-2 in situ hybr. TUNEL
4-24 h 18 h
none detected
TUNEL
2h
Sloviter et al., 1996
EM and toluidine blue
12-24 h
hippocampal GCL, SGZ, and hilus; white matter hippocampal GCL and SGZ
Thompson et al., 1998 Pollard et al., 1994a,b
TUNEL
2 h, 24 h 18 h, 24 h
Venero et al., 1999
Silver stain, TUNEL, EM, DNA electrophoresis TUNEL
Henshall et al., 2000a
PANT, TUNEL, Caspase-3 immuno,
4-96 h
Filipkowski et al., 1994
DNA electrophoresis
18 h, 72 h
Sakhi et al., 1994
p53 in situ hybr., TUNEL
4 h, 8 h, 16 h
Kindling, single Bengzon et al., 1997 stimulation Kindling, Bengzon et al., 1997 multiple stimulations Pretel et al., 1997
Zhang et al., 1998
Nakagawa et al., 2000 Umeoka et al., 2000 Perforant path stimulation
Kainic acid, local application
Kainic acid, systemic application
Gillardon et al., 1995 TUNEL, Bcl-2 and Bax Gillardon et al., 1995 CPP-32 in situ hybr., TUNEL Liu et al., 1996 cyclin D1 immuno. Sakhi et al., 1996 p53 immuno.
1-10 days
after 8 h of stim. in rat pups intraamygdaloid injections intra-septal injections intraamygdaloid injections
Tuunanen et al., 1999 DNA electrophoresis, TUNEL, Bcl-2 and Bax
8 h, 16 h
Faherty et al., 1999
30 h to 4 days
hippocampal GCL and hilus (only at 24 h) hippocampal CA3 and CA4; amygdala hippocampal CA1, CA3, and CA4; amygdala; septum; thalamus hippocampal CA3 and CA4
hippocampus; entorhinal and sensory cortices mice
hippocampal CA1 and CA3; amygdala; piriform cortex; thalamus hippocampus; neocortex hippocampal CA3 and CA4 hippocampal pyramidal layers hippocampal CA1 and CA3; piriform cortex; thalamus
8-16 h 4 h, 30 h 2-12 h
Regions expressing apoptotic damage
qualitative hippocampal GCL, hilus, assessment only interna/molecular layer, CA1, and subiculum; deep cortical layers diffusely throughout hippocampus hippocampal GCL
24 h, 48 h 6 h, 24 h
Simonian et al., 1996 DNA electrophoresis, TUNEL
TUNEL, Caspase-3 immuno.
Comments
cerebellar granular cell cultures examined only amygdala damage in KA FVB/N sensitive mice
variable between different amygdaloid nuclei hippocampal CA1, CA3, and CA4
113
TABLE l (continued) Method of seizure induction
Pilocarpine
Reference
Method of apoptotic detection
Time after insult
Comments
Regions expressing apoptotic damage
Fujikawa et al., 2000
TUNEL, EM, DNA electrophoresis, H and E stain
24 h, 72 h
necrotic cells can exhibit 'apoptotic' markers
dorsal and ventral hippocampus; amygdala; entorhinal, piriform, and frontal cortices
Kondratyev and Gale, 2000
Caspase-3 immuno., 24 h Hoechst and Texas Red: DNA electrophoresis TUNEL, EM, DNA 24 h electrophoresis, ethidium bromide
Sankar et al., 1998
Fujikawa et al., 1999
TUNEL, EM, DNA electrophoresis, H and E stain
24 h, 72 h
Roux et al., 1999
TUNEL and p75 immuno.
1 day, 3 days
Covolan et al., 2000b EM
2.5-48 h
Ekdahl et al., 2001
2 days, 8 days
TUNEL and Hoechst
amygdala; rhinal cortex; hippocampus immature and mature rats
necrotic cells can exhibit 'apoptotic' markers
cells show both apoptotic and necrotic features
hippocampal CA1 and SGZ; progressively more damage in hippocampal CA3 and amygdala with increasing age ventral hippocampal CA1, CA2, and CA3; piriform and entorhinal cortices; thalamus hippocampal CA1 and GCL; entorhinal, perirhinal and piriform cortices; amygdala; thalamus hippocampal GCL and SGZ
hippocampal GCL and SGZ, hilus, CA1, and CA3
Abbreviations: CA1 = CA1 pyramidal layer; CA3 = CA3 pyramidal layer; CA4 = CA4 pyramidal layer within the dentate gyms; EM = electron microscopy; GCL = granular cell layer of dentate gyms; KA = kainate ac!d; PANT = DNA polymerase 1-mediated biotin-dATP nick translation; SGZ = subgranular zone of dentate gyms; TUNEL = terminal deoxynucleotidyltransferase-mediated dUTP nick-end labeling.
been observed after seizures (Covolan et al., 2000b; Fujikawa, 2000; Fujikawa et al., 2000). The following section summarizes evidence of apoptotic cell death following brief and prolonged seizures induced by kindling, perforant path stimulation, pilocarpine, and kainate and reviews the literature on the mechanisms of seizure-induced apoptotic neuronal degeneration.
Apoptosis after brief seizures Status epilepticus lasting for more than 30 min has been well documented to cause neuronal damage in both experimental animals and in humans (Hauser, 1983; Sperk, 1994) (Table 1). Although longer periods of seizures consistently produce degeneration of neurons, brief, single or intermittent seizures
have previously not been thought to lead to cell death. However, work from our own laboratory provides evidence for cell death following brief single seizures in rats (Bengzon et al., 1997). Five hours after a single hippocampal kindling stimulation, producing focal epileptiform activity lasting for about 80 s, a marked increase (214 4-32% of control) of TUNEL-positive pycnotic nuclei was observed bilaterally within the dentate gyrus. Most labeled nuclei were located in the subgranular zone of the dorsal and ventral blade of the dentate granule cell layer. A few TUNEL-positive nuclei were also seen within the granule cell layer and in the dentate hilus. Degenerating cells displayed morphological characteristics of apoptosis. In addition to being intensely labeled using the TUNEL technique, propidium iodide DNA staining showed that degenerating nuclei
114
Fig. 1. (A) Nuclei labeled for fragmented DNA (arrows) along the hilar border of the granule cell layer 2 h after 40 hippocampal kindling stimulations, (B and C) High-powerphotomicrographsof labeled nuclei with morphologicalfeatures characteristic of apoptosis, such as condensation and lobulation (arrow), following kindling stimulation. Note numerous apoptotic bodies (arrowheads). (D) Confocal scanning laser image showing nuclear TUNEL (green) and cytoplasmicNeuN immunolabeling(red) demonstrating the neuronal identity of a degenerating cell 2 h after 40 hippocampalkindling stimulations. (Bars: A = 80 ~m; B and C = 15 Ixm;D = 10 ~tm.) Illustrations from Bengzon et al. (1997). were consistently shrunken, irregularly shaped, and lobulated compared to normal nuclei. After 40 recurring stimulations with 5-min intervals, so-called rapid kindling, markedly higher numbers of apoptotic cells were observed within the dentate gyrus. The kindling-induced increase in TUNEL-positive cells was blocked by the protein synthesis inhibitor cycloheximide. Some TUNEL-positive degenerated cells were double-labeled with the neuron-specific antigen NeuN, indicating a neuronal phenotype of these apoptotic cells (Fig. 1). Degeneration of dentate granule neurons following brief kindled seizures was confirmed in the study by Zhang et al. (1998). The number of cells displaying fragmented DNA as assessed by the TUNEL technique increased by 30% compared to control after a single epileptic discharge produced by amygdala kindling stimulation. The authors suggested that the degenerating cells mainly consisted of hippocampal interneurons and that the functional effect might be a disinhibition within the hippocampus. In support of these findings, Pretel et al. (1997) demonstrated that some of the kindling induced TUNEL-labeled
cells in the hippocampus double stained with somatostatin, a marker for interneurons. However, further experiments are needed to clarify the identity of the degenerating cells. The cell loss induced by each seizure in all of the above studies is mild which may explain why conventional degeneration staining protocols in the past have failed to detect this change. Corresponding studies in humans aimed at detecting subtle cell loss following seizures are technically impossible to carry out. However, postmortem analysis of humans with chronic idiopathic epilepsy suggests that neuronal damage is cumulative and related to the frequency of seizures and the duration of the disease (Dam, 1980). This could imply that a mild cell loss, as seen after every seizure in the kindling model of epilepsy, over several years of repeated and frequent seizures could lead to substantial pathological changes.
Apoptosis following prolonged seizures Widespread apoptosis has been found in a variety of animal models of status epilepticus. The first
115 studies demonstrating apoptotic cell death following seizures were performed by Pollard et al. (1994a,b). After experimental status epilepticus induced by intra-amygdaloid injection of kainate in adult rats, silver-impregnated damaged neurons were observed in the amygdala and pyramidal neurons of the hippocampal CA3 region a few hours after injection. In both areas the degeneration had apoptotic features, including nuclear chromatin condensation and marginalization and positive nuclear labeling with the TUNEL-staining method. Combined TUNEL labeling and silver staining showed that the DNA fragmentation occurred in dying neurons. Subsequent studies using both kainic acid systemically (Tuunahen et al., 1999) and locally (Venero et al., 1999) have extended these initial findings and confirmed neuronal degeneration with apoptotic features within the hippocampal pyramidal regions CA1 and CA3/4 and in the amygdala. Following pilocarpine-induced status epilepticus, Roux et al. (1999) observed numerous TUNEL-positive cells throughout the piriform cortex and entorhinal cortex in addition to the hippocampus. Also the immature brain shows vulnerability to status-epilepticus-induced apoptosis. Sankar et al. (1998) noted neurons displaying features of apoptotic death in the CA1 region of the 2-weeks-old pups, and in the subgranular zone of the dentate gyrus in the 3-weeks-old animals after lithiumpilocarpine-induced status epilepticus. Thompson et al. (1998) observed that intermittent perforant path stimulation in rat pups produced apoptotic hippocampal cell loss. After 16 h of stimulation, TUNEL labeling performed 2 h after the end of stimulation showed an intense band of positively labeled eosinophilic cells with condensed profiles bilaterally in the dentate granule cell layer. Cell-death mechanisms
Programmed cell death in response to various insults can be triggered through cell membrane receptor activation, mitochondrial injury, or by direct damage to the DNA (Graham and Chen, 2001). Membrane-receptor-activated apoptosis has been documented following seizures. Activation of the kainate glutamate receptor 6 (GluR6) leads to signaling via the c-Jun N-terminal kinase (Jnk) family
(Savinainen et al., 2001). One particular member of the Jnk family, Jnk3, may be required for stressinduced neuronal apoptosis. Yang et al. (1997) reported that disruption of the gene encoding Jnk3 caused mice to be resistant to excitotoxicity caused by kainic acid. They showed that a disruption of Jnk3 caused a reduction in seizure activity and prevented hippocampal CA1 and CA3 neuronal apoptosis. Further evidence implicating the Jnk family in seizure-induced apoptosis comes from the finding that systemic kainic-acid-evoked seizures leads to activation of Jnkl and phosphorylation of c-Jun (Mielke et al., 1999). Furthermore, the magnitude and period of induction of Jnk-1 protein following kainic acid seizures were associated with impending cell death, while increased phosphorylation of c-Jun protein was associated with resistance to cell death (Schauwecker, 2000). Other receptor candidates involved in triggering cell degeneration following seizures might include the p75 neurotrophin receptor (p75NTR) (Roux et al., 1999). Numerous TUNEL-positive cells throughout the post-seizure hippocampus, piriform cortex, and entorhinal cortex after pilocarpine-induced seizures were found to be double labeled for the p75NTR, suggesting that seizure-induced neuronal loss within the CNS might occur through apoptotic signaling cascades involving p75NTR. However, further work is needed to clarify the involvement of this receptor in the induction of apoptosis following seizures. Mitochondrial stress produced by prolonged depolarization, oxidative stress, and opening of the mitochondrial permeability transition pore are powerful triggers for programmed cell death (Reed, 1998). The Bcl-2 family genes are key determinants in regulating mitochondrial permeability. The function of Bcl-2 is to block the mitochondrial releases of cytochrome c in response to mitochondrial stressor stimuli. In addition to the anti-apoptotic Bcl-2 gene, this family consists of more than 20 genes including the anti-apoptotic genes Bcl-xL, Bcl-w, and the pro-apoptotic gene Bax. Following amygdala kindling stimulations, the ratio of Bax/Bcl-2 expression was found to increase in the hippocampus (Zhang et al., 1998). Also the levels of Bcl-w protein increase within the hippocampus following seizures (Henshall et al., 2001). A herpes simplex virus-1 vector used to deliver the Bcl-2 gene in order to overex-
116 press it within vulnerable parts of the hippocampus granule cells can counteract the neuronal degeneration and decline in hippocampal function seen after kainic-acid-induced status epilepticus (McLaughlin et al., 2000). It has been demonstrated that the tumor suppressor protein p53, which stimulates production and/or mitochondrial translocation of Bax, can reduce damage to CA1 and CA3 hippocampal neurons following kainic-acid-induced seizures (Culmsee et al., 2001). Cytochrome c release from mitochondria triggers cell death through cysteinyl aspartate-specific proteinase (caspase) activation. The caspases comprise a family of 14 proteases that, in their active proteolysed state, function as initiators and effectors of programmed cell death. Clear evidence now points to a role for caspase-dependent neuronal degeneration following seizures. Faherty et al. (1999) examined the levels of activated caspase-3 following kainic-acid-induced seizures in two mice strains either sensitive or resistant to kainic-acid-induced neuronal degeneration. Catalytically active caspase3 was detected 30 h following kainic acid treatment in the sensitive strain before the appearance of pyknosis and TUNEL labeling. This expression of activated caspase-3 continued up to 4 days following injection. Caspase-3 immunoreactivity was never detected in the resistant strain and there was no evidence of pyknosis or TUNEL staining. In support, other groups (Henshall et al., 2000a; Kondratyev and Gale, 2000) have demonstrated increases in active caspase-3 fragments following kainic-acid-induced seizures. Together these data suggest that activation of caspase-3 is a necessary component of kainicacid-induced TUNEL cell death. Further evidence for the involvement of caspase in neuronal death following seizures was the finding that the caspase-3 inhibitor z-DEVD-fmk, which was injected into the lateral ventricle prior to and following the epileptic insult, substantially attenuated apoptotic cell death both in hippocampus and rhinal cortex (Kondratyev and Gale, 2000). Subsequently, in recent work in our laboratory, we gave multiple intracerebroventftcular infusions of caspase inhibitors during and following pilocarpine-induced status epilepticus and reduced the number of TUNEL/Hoechst-positive cells (Ekdahl et al., 2001). This treatment also increased the number of newly formed cells in the subgran-
ular zone of the dentate gyrus at 1 week after the epileptic insult (see below). Viswanath et al. (2000) created transgenic mice that neuronally expressed the baculoviral caspase inhibitor p35 and found an attenuation in both caspase activity and neurodegeneration in response to kainic acid in vitro and in vivo. In particular, kainic acid administration to p35 mice resulted in decreased caspase activity and TUNEL staining in pyramidal CA1 hippocampal neurons. Necrosis versus apoptosis
Considerable controversy exists conceming the degree of true apoptotic damage with kainic-acid- and pilocarpine-induced seizures. The nature of dentate granule cell damage in epilepsy has been reported as either apoptotic, necrotic or both. Adding to the problem of identifying the mode of degeneration are shortcomings in the techniques used. There is evidence that in situ TUNEL labeling may give falsepositive labeling of necrotic cells and the detection of DNA laddering on agarose gels may not be exclusive to apoptotic cells (e.g. Charriaut-Marlangue and Ben-Aft, 1995; Fujikawa, 2000; Fujikawa et al., 2000). Covolan et al. (2000a) analyzed dentate gyrus granule cell morphology with electron microscopy after pilocarpine and reported a variety of cell death morphologies ranging from apoptosis to necrosis. Some cells displayed coalescence of chromatin against nuclear membranes in the absence of obvious apoptotic cytoplasmic budding or typical membranebound apoptotic bodies. The authors concluded that pilocarpine-induced status epilepticus promotes a degenerative process in the dentate granule cell layer with both apoptotic and necrotic features. Intermediate forms of degeneration of hippocampal neurons were seen also in a series of studies by Fujikawa et al. (1999, 2000) after kainic-acid- and pilocarpine-evoked seizures. Degenerating neurons displayed morphologies characteristic of necrosis; however, with electron microscopy these neurons appeared dark and shrunken with pyknotic nuclei containing small and dispersed TUNEL-negative chromatin clumps. Furthermore, many of these 'necrotic' looking cells exhibited internucleosomal DNA cleavage (DNA 'laddering'). In other less severe animal models of epilepsy, such as kindling or perforant path stimulation, clear apoptosis is more discern-
118
to controls. Evidence of caspase-1 and caspase-3 activity was determined by detecting increases in their respective cleavage byproducts. Bcl-2, Bax, and caspase-3 immunoreactivity were increased predominantly in cells with neuronal morphology, whereas Bcl-xL immunoreactivity was increased in cells with glia morphology.
Conclusions and implications Over the past few years it has become clear that neuronal apoptotic degeneration occurs both as an early and late consequence of seizures. It is not unusual that apoptotic degeneration is detected following prolonged seizures; but what may be more surprising is that death of hippocampal neurons is detected following brief and focal non-convulsive seizures. The discovery that part of the damage to the brain following seizures is the result of programmed cell death, suggests alternative new treatment strategies, such as caspase inhibitors, to prevent cell degeneration. However, intermediate forms of degeneration between necrosis and apoptosis clearly exist. Further investigations into the precise molecular mechanisms of these different forms of degeneration and their relationship to regenerative processes, such as reactive neurogenesis, are needed before the full therapeutic potential can be harnessed.
Abbreviations caspase: TUNEL:
cysteinyl aspartate-specific proteinase terminal deoxynucleotidyltransferasemediated dUTP nick-end labeling
Acknowledgements Supported by grants from the Swedish Medical Research Council, the Elsa and Thorsten Segerfalk Foundation, the Kock Foundation and the Royal Physiographic Society.
References Bengzon, J., Kokaia, L., Elmer, E., Nanobashvili, A., Kokaia, M. and Lindvall, O. (1997) Apoptosis and proliferation of dentate gyrus neurons after single and intermittent limbic seizures. Proc. Natl. Acad. Sci. USA, 94: 10432-10437.
Bonfoco, E., Krainc, D., Ankarcrona, M., Nicotera, P. and Lipton, S.A. (1995) Apoptosis and necrosis: two distinct events induced, respectively, by mild and intense insults with Nmethyl-D-aspartate or nitric oxide/superoxide in cortical cell cultures. Proc. Natl. Acad. Sci. USA, 92: 7162-7166. Charriaut-Marlangue, C. and Ben-Ari, Y. (1995) A cautionary note on the use of the TUNEL stain to determine apoptosis. Neuroreport, 7: 61-64. Clarke, P.G. (1990) Developmental cell death: morphological diversity and multiple mechanisms. Anat. Embryol., 181: 195213. Covolan, L., Ribeiro, L.T., Longo, B.M. and Mello, L.E. (2000a) Cell damage and neurogenesis in the dentate granule cell layer of adult rats after pilocarpine or kainate-induced status epilepticus. Hippocampus, 10: 169-180. Covolan, L., Smith, R.L. and Mello, L.E.A.M. (2000b) Ultrastructural identification of dentate granule cell death from pilocarpine-induced seizures. Epilepsy Res., 41: 9-21. Culmsee, C., Zhu, X., Yu, Q.S., Chan, S.L., Camandola, S., Guo, L., Greig, N.H. and Mattson, M.R (2001) A synthetic inhibitor of p53 protects neurons against death induced by ischemic and excitotoxic insults, and amyloid beta-peptide. J. Neurochem., 77: 220-228. Dam, A.M. (1980) Epilepsy and neuron loss in the hippocampus. Epilepsia, 21: 617-629. Ekdahl, C.T., Mohapel, P., Elmdr, E. and Lindvall, O. (2001) Caspase inhibitors increase short-term survival of progenitor cell progeny in the adult rat dentate gyrus following status epilepticus. Fur. J. Neurosci., 14: 937-945. Faherty, C.L., Xanthoudakis, S. and Smeyne, R.J. (1999) Caspase-3-dependent neuronal death in the hippocampus following kainic acid treatment. Brain Res. Mol. Brain Res., 70: 159-163. Filipkowski, R.K., Hetman, M., Kaminska, B. and Kaczmarek, L. (1994) DNA fragmentation in the rat brain after intraperitoneal administration of kainate. Neuroreport, 5: 1538-1540. Fujikawa, D.G. (2000) Confusion between neuronal apoptosis and activation of programmed cell death mechanisms in acute necrotic insults. Trends Neurosci., 23:410-411. Fujikawa, D.G., Shinmei, S.S. and Cai, B. (1999) Lithiumpilocarpine-induced status epilepticus produces necrotic neurons with internucleosomal DNA fragmentation in adult rats. Eur. J. Neurosci., 11: 1605-1614. Fujikawa, D.G., Shinmei, S.S. and Cai, B. (2000) Kainic acidinduced seizures produce necrotic, not apoptotic, neurons with internucleosomal DNA cleavage: implications for programmed cell death mechanisms. Neuroscience, 98: 41-53. Gage, F.H. (2000) Mammalian neural stem cells. Science, 287: 1433-1438. Gillardon, E., Wickert, H. and Zimmermann, M. (1995) Upregulation of bax and down-regulation of bcl-2 is associated with kainate-induced apoptosis in mouse brain. Neurosci. Lett., 192: 85-88. Graham, S.H. and Chen, J. (2001) Programmed cell death in cerebral ischemia. J. Cereb. Blood Flow Metab., 21: 99-109. Hauser, W.A. (1983) Status epilepticus: frequency, etiology, and neurological sequelae. Ad~: Neurol., 34: 3-14.
117 able (Sloviter et al., 1996; Bengzon et al., 1997). Perforant path stimulation for 8 h induces acute degeneration of dentate granule cells, whereas 24 h additionally injures hilar and hippocampal pyramidal neurons (Sloviter et al., 1996). Light and electron microscopic analyses revealed that the degenerating hilar and pyramidal neurons exhibited morphological features of necrosis. In contrast, acutely degenerating granule neurons exhibited morphological features of apoptosis, including coalescence of nuclear chromatin into multiple nuclear bodies, compaction of the cytoplasm, cell shrinkage, and budding-off of apoptotic bodies. The authors concluded that the nature of the neuronal death induced by excessive excitation could be determined postsynaptically by the manner in which different target cells react to an excitatory insult. Alternatively, it could be that the nature of the degeneration is determined by the severity of the insult, as it has been proposed for ischemia-induced neurodegeneration (Graham and Chen, 2001). Excitotoxin exposure at a high concentration for a prolonged period of time has been shown to produce necrosis of neurons in vitro, whereas brief exposure to the toxin at lower concentrations results in apoptosis in the same neuronal population (Bonfoco et al., 1995). Factors such as the variation in the duration and intensity of seizure activity, metabolic disturbances and energy failure during and after seizures, and specific cell death triggering factors may all contribute to determining the eventual pathway of degeneration that a particular cell commits to.
Relationship of apoptosis to dentate gyrus neurogenesis The dentate gyrus in rats contains precursor cells that continue to produce new neurons throughout adulthood. Proliferation of neuroblasts takes place in the subgranular zone and the newly formed cells then migrate into the granule cell layer and develop phenotypic characteristics of granule neurons. The newly formed cells seem to be well integrated and extend projections to the hippocampal CA3 region, but the functional capacity of these neurons remains unknown. Dentate granule neurogenesis can be influenced by various stimuli, including aging, learning, exercise, adrenal steroids, growth factors
and various insults, see Gage (2000). Neurogenesis is also increased following seizures. Brief, kindied seizures give rise to moderate increases in the number of newly formed cells and more prolonged or repetitive seizures result in larger increases in neurogenesis (Bengzon et al., 1997; Parent et al., 1997). In a recent study from our own laboratory we investigated the relationship between seizureinduced apoptosis and neurogenesis in the dentate gyms of adult rats by blocking caspases. Multiple intraventricular infusions of caspase inhibitors just prior to and up to 1 week following pilocarpineevoked status epilepticus reduced the number of TUNEL/Hoechst-positive cells and increased the number of bromodeoxyuridine-stained proliferated cells in the dentate subgranular zone at 1 week following the insult (Ekdahl et al., 2001). Our findings suggest that caspases modulate seizure-induced neurogenesis in the dentate gyms, probably by regulating apoptosis of newly born neurons, and that this action can be suppressed by caspase inhibitors. Along the same lines, administration of the protein synthesis inhibitor cycloheximide was found to increase the number of proliferated cells in the hippocampus following systemic pilocarpine administration (Covolan et al., 2000a). The authors speculated that such increased mitotic rates might be associated with an anti-apoptotic-mediated protection of a vulnerable precursor cell population that would otherwise degenerate after pilocarpine-induced status epilepticus. Our work supports this notion and points to the caspases as this anti-apoptotic mechanism in the protection of these precursor cells. Taken together, our own work and the findings of Covolan et al. (2000a) point to a dynamic balance between apoptosis and the generation of new cells in the dentate gyms following seizures.
Apoptosis in humans To address the role of cell death regulatory genes in the neuropathology of human epilepsy, Henshall et al. (2000b) investigated the expression of Bcl-2, Bcl-xL, Bax, caspase-1, and caspase-3 proteins in temporal cortex samples from patients who had undergone temporal lobectomy surgery for intractable epilepsy. Levels of Bcl-2 and Bcl-xL proteins were significantly increased in epileptic brains compared
ll9
Henshall, D.C., Chen, J. and Simon, R.E (2000a) Involvement of caspase-3-1ike protease in the mechanism of cell death following focally evoked limbic seizures. J. Neurochem., 74: 1215-1223. Henshall, D.C., Clark, R.S., Adelson, ED., Chen, M., Watkins, S.C. and Simon, R.E (2000b) Alterations in bcl-2 and caspase gene family protein expression in human temporal lobe epilepsy. Neurology, 55: 250-257. Henshall, D.C., Skradski, S.L., Lan, J., Ren, T. and Simon, R.P. (2001) Increased'Bcl-w expression following focally evoked limbic seizures in the rat. Neurosci. Lett., 305: 153-156. Kerr, J.F., Wyllie, A.H. and Currie, A.R. (1972) Apoptosis: a basic biological phenomenon with wide-ranging implications in tissue kinetics. Br. J. Cancer, 26: 239-257. Kondratyev, A. and Gale, K. (2000) Intracerebral injection of caspase-3 inhibitor prevents neuronal apoptosis after kainic acid-evoked status epilepticus. Brain Res. Mol. Brain Res., 75: 216-224. Liu, W., Bi, X., Tocco, G., Baudry, M. and Schreiber, S.S. (1996) Increased expression of cyclin D1 in the adult rat brain following kainate acid treatment. Neuroreport, 7: 2785-2789. McLaughlin, L., Roozendaal, B., Dumas, T., Gupta, A., Ajilore, O., Hsieh, J., Ho, D., Lawrence, M., McGaugh, J.L. and Sapolsky, R. (2000) Sparing of neuronal function postseizure with gene therapy. Proc. Natl. Acad. Sci. USA, 97: 1280412809. Mielke, K., Brecht, S., Dorst, A. and Herdegen, T. (1999) Activity and expression of JNK1, p38 and ERK kinases, c-Jun N-terminal phosphorylation, and c-jun promoter binding in the adult rat brain following kainate-induced seizures. Neuroscience, 91: 471-483. Nakagawa, E., Aimi, Y., Yasuhara, O., Tooyama, L., Shimada, M., McGeer, EL. and Kimura, H. (2000) Enhancement of progenitor cell division in the dentate gyrus triggered by initial limbic seizures in the rat models of epilepsy. Epilepsia, 41: 10-18. Parent, J.M., Yu, T.W., Leibowitz, R.L., Geschwind, D.H., Sloviter, R.S, and Lowenstein, D.H. (1997) Dentate granule cell neurogenesis is increased by seizures and contributes to aberrant network reorganization in the adult rat hippocampus. J. Neurosci., 17: 3727-3738. Pollard, H., Cantagrel, S., Cbarriant-Marlangue, C., Morean, J. and Ben Ari, Y. (1994a) Apoptosis associated DNA fragmentation in epileptic brain damage. Neuroreport, 5: 1053-1055. Pollard, H., Charriaut-Marlangue, C., Cantagrel, S., Represa, A., Robain, O., Moreau, J. and Ben-Aft, Y. (1994b) Kainateinduced apoptotic cell death in hippocampal neurons. Neuroscience, 63: 7-18. Pretel, S., Applegate, C.D. and Piekut, D. (1997) Apoptotic and necrotic cell death following kindling induced seizures. Acta Histochem., 99: 71-79. Reed, J.C. (1998) Bcl-2 family proteins. Oncogene, 17: 32253236. Roux, EE, Colicos, M.A., Barker, P.A. and Kennedy, T.E. (1999) p75 neurotrophin receptor expression is induced in apoptotic neurons after seizure. J. Neurosci., 19: 6887-6896. Sakhi, S., Bruce, A., Sun, N., Tocco, G., Baudry, M. and
Schreiber, S.S. (1994) p53 induction is associated with neuronal damage in the central nervous system. Proc. Natl. Acad Sci. USA, 91: 7525-7529. Sakhi, S., Sun, N., Wing, L.L., Mehta, E and Schreiber, S.S. (1996) Nuclear accumulation of p53 protein following kainic acid-induced seizures. Neuroreport, 7: 493-496. Sankar, R., Shin, D.H., Liu, H., Mazarati, A., Pereira de Vasconcelos, A. and Wasterlain, C.G. (1998) Patterns of status epilepticus-induced neuronal injury during development and long-term consequences. J. Neurosci., 18: 8382-8393. Savinainen, A., Garcia, E.P., Dorow, D., Marshall, J. and Liu, Y.F. (2001) Kainate receptor activation induces mixed lineage kinase-mediated cellular signaling cascades via post-synaptic density protein 95. J. Biol. Chem., 276: 11382-11386. Schanwecker, EE. (2000) Seizure-induced neuronal death is associated with induction of c-Jun N-terminal kinase and is dependent on genetic background. Brain Res., 884:116-128. Simonian, N.A., Getz, R.L., Leveque, J.C., Konradi, C. and Coyle, J.T. (1996) Kainic acid induces apoptosis in neurons. Neuroscience, 75: 1047-1055. Sloviter, R.S., Dean, E., Sollas, A.L. and Goodman, J.H. (1996) Apoptosis and necrosis induced in different hippocampal neuron populations by repetitive perforant path stimulation in the rat. J. Comp. Neurol., 366: 516-533. Sperk, G. (1994) Kainic acid seizures in the rat. Prog. Neurobiol., 42: 1-32. Thompson, K., Holm, A.M., Schousboe, A., Popper, P., Micevych, P. and Wasterlain, C. (1998) Hippocampal stimulation produces neuronal death in the immature brain. Neuroscience, 82: 337-348. Tuunanen, L., Lukasiuk, K., Halonen, T. and Pitkanen, A. (1999) Status epilepticus-induced neuronal damage in the rat amygdaloid complex: distribution, time-course and mechanisms. Neuroscience, 94: 473-495. Umeoka, S., Miyamoto, O., Janjua, N.A., Nagao, S. and Itano, T. (2000) Appearance and alteration of TUNEL positive cells through epileptogenesis in amygdaloid kindled rat. Epilepsy Res., 42: 97-103. Venero, J.L., Revuelta, M., Machado, A. and Cano, J. (1999) Delayed apoptotic pyramidal cell death in CA4 and CA1 hippocampal subfields after a single intraseptal injection of kainate. Neuroscience, 94:1071-1081. Viswanath, V., Wu, Z., Fonck, C., Wei, Q., Boonplueang, R. and Andersen, J.K. (2000) Transgenic mice neuronally expressing baculoviral p35 are resistant to diverse types of induced apoptosis, including seizure-associated neurodegeneration. Proc. Natl. Acad. Sci. USA, 97: 2270-2275. Wyllie, A.H., Kerr, J.F. and Currie, A.R. (1980) Cell death: the significance of apoptosis. Int. Rev. Cytol., 68: 251-306. Yang, D.D., Kuan, C.Y., Whitmarsh, A.L., Rincon, M., Zheng, T.S., Davis, R.J., Rakic, P. and Flavell, R.A. (1997) Absence of excitotoxicity-induced apoptosis in the hippocampus of mice lacking the Jnk3 gene. Nature, 389: 865-870. Zhang, L.X., Smith, M.A., Li, X.L., Weiss, S.R. and Post, R.M. (1998) Apoptosis of hippocampal neurons after amygdala kindled seizures. Brain Res. Mol. Brain Res., 55: 198-208.
T. Sutula and A. PitkRnen (Eds.) Progress in Brain Research, Vol. 135 © 2002 Elsevier Science B.V. All rights reserved
CHAFFER 10
Seizure-induced neurogenesis: are more new neurons good for an adult brain? Jack M. Parent 1,* and Daniel H. Lowenstein 2 Department of Neurology, University of Michigan Medical Center, Ann Arbor, M1 48104-1687, USA 2 Harvard Medical School and Department of Neurology, Beth Israel-Deaconess Medical Center, 333 Brookline Avenue, Boston, MA 02215, USA
Abstract: The idea that neural stem cells may play a role in the pathophysiology or potential treatment of specific epilepsy syndromes is relatively new. This notion relates directly to advances in the field of stem cell biology over the past decade, which have confirmed prior theories that both neural stem cells and neurogenesis, the birth of new neurons, persist in specific regions of the adult mammalian brain. The physiological role of persistent neurogenesis is not known, although recent work implicates this process in specific learning and memory tasks. Knowledge of the normal neurogenic pathways in the mature brain has led to recent studies of neurogenesis in rodent models of acute seizures or epileptogenesis. Most of these studies have examined neurogenesis in the adult rodent dentate gyms, and current evidence indicates that single brief or prolonged seizures, as well as repeated kindled seizures, increase dentate granule cell (DGC) neurogenesis. The models studied to date include pilocarpine and kainic acid models of temporal lobe epilepsy, limbic kindling, and intermittent perforant path stimulation. Recent work also suggests that pilocarpine-induced status epilepticus increases rostral forebrain subventricular zone (SVZ) neurogenesis and caudal SVZ gliogenesis. Several lines of evidence implicate newly generated neurons in structural and functional network abnormalities in the epileptic hippocampal formation of adult rodents. These abnormalities include aberrant mossy fiber reorganization, persistence of immature DGC structure (e.g. basal dendrites), and the abnormal migration of newborn neurons to ectopic sites in the dentate gyms. Taken together, these findings suggest a pro-epileptogenic role of seizure- or injury-induced neurogenesis in the epileptic hippocampal formation. However, the induction of forebrain SVZ neurogenesis and directed migration to injury after seizures and other brain insults underscores the potential therapeutic use of neural stem cells as a source for neuronal replacement after injury.
Neurogenesis in the adult mammalian brain Neurogenesis in the mammalian CNS is confined largely to the embryonic period, after which neuronal precursor cells undergo terminal division and new neuronal tissue ceases to be generated. Despite the recognition nearly 90 years ago that mitotically
* Correspondence to: J.M. Parent, University of Michigan Medical Center, Neuroscience Laboratory Building, 1103 E. Huron St., Ann Arbor, MI 48104-1687, USA. Tel.: +1-734-936-1988; Fax: -t-1-734-763-7686; E-mail:
[email protected]
active cells persist in the adult rodent brain (Allen, 1912), the widely held belief that the adult CNS lacks any regenerative potential delayed acceptance of the notion that new neurons could be generated in the adult mammalian brain. Based on tritiated thymidine mitotic labeling studies, however, investigators first proposed over three decades ago that neurons continue to be produced in the adult rodent rostral SVZ-olfactory bulb pathway and hippocampal dentate gyrus (Altman and Das, 1965; Hinds, 1968; Altman, 1969). These findings were confirmed by electron microscopy a decade later (Kaplan and Hinds, 1977), and have been identified in every adult mammalian species examined to date, including hu-
122
Fig. 1. Schematic parasagittal view of the adult rodent brain showing regions of persistent neurogenesis. Dark circles denote neuronal precursor cells, and white circlesrepresentdifferentiatingneurons. See text for details. DG = dentate gyrus; SVZ = subventricularzone; RMS = rostral migratorystream; OB = olfactorybulb.
man (for the dentate gyrus) and non-human primates (Eriksson et al., 1998; Gould et al., 1998, 1999; Kornack and Rakic, 1999). Adult mammalian neurogenesis has been studied most extensively in the rodent dentate gyrus. In the adult rat, neuronal precursor cells proliferate in clusters in the dentate subgranular zone, located at the border of the granule cell layer and hilus (Fig. 1) (Kaplan and Hinds, 1977; Cameron et al., 1993; Kuhn et al., 1996). Their progeny disperse and migrate into the DGC layer where they differentiate into mature granule neurons (Cameron et al., 1993; Kuhn et al., 1996). A smaller number of progeny probably also differentiate into radial glia-like cells in the granule cell layer (Cameron et al., 1993). However, it is not known whether individual precursor cells are multipotentent or lineage restricted in terms of the daughter cells they can generate. Although the majority of DGCs in the rat are produced near the end of the first postnatal week, new DGCs continue to be generated at a lower rate throughout adulthood and into senescence (Kuhn et al., 1996). Newly differentiating granule neurons in the mature hippocampal formation express a number of immature neuronal markers putatively involved in cell migration and axon outgrowth, including collapsin response mediator protein-4 (CRMP-4), the polysialylated form of neural cell adhesion molecule (PSANCAM), and doublecortin (Seki and Arai, 1993; Parent et al., 1997, 1999; Nacher et al., 2001). Newly
generated DGCs in the adult also appear to integrate normally into existing hippocampal networks. Combined retrograde tracer and mitotic labeling studies in adult rodent have shown that mossy fibers of newly born DGCs project to appropriate targets in hippocampal area CA3 (Stanfield and Trice, 1988; Markakis and Gage, 1999). Neuronal precursors also persist and continue to proliferate in the adult rodent forebrain SVZ (Altman, 1969; Kaplan and Hinds, 1977; Lois and Alvarez-Buylla, 1994; Lois et al., 1996; Thomas et al., 1996). However, unlike in the dentate gyrus, SVZ neuronal progenitors migrate long distances to their final destinations in the olfactory bulb (Fig. 1) (Lois and Alvarez-Buylla, 1994; Lois et al., 1996). The immature neurons migrate from the rostral SVZ to the olfactory bulb using a relatively unique form of tangential, chain migration (Lois and Alvarez-Buylla, 1994; Lois et al., 1996; Doetsch and Alvarez-Buylla, 1996; Wichterle et al., 1997) in a restricted forebrain pathway known as the rostral migratory stream (RMS) (Altman, 1969; Kishi, 1987). As in the dentate gyrus, the immature neuronal progeny in the SVZ and RMS of adult rodents can be identified by their expression of characteristic markers such as PSA-NCAM, neuron-specific beta tubulin, doublecortin, and CRMP-4 (Bonfanti and Theodosis, 1994; Doetsch and Alvarez-Buylla, 1996; Thomas et al., 1996; Perreto et al., 1999; Magavi et al., 2000; Nacher et al., 2000). Once the neuroblasts reach the
123 subependymal region of their olfactory bulb target, they disperse radially and differentiate into granule and periglomernlar neurons (Luskin, 1993; Lois and Alvarez-Buylla, 1994; Lois et al., 1996; Thomas et al., 1996). Several relatively recent advances have accelerated our understanding of persistent neurogenesis in the adult mammalian brain. First, the ability to identify proliferating precursors and track their migration and differentiation in vivo has markedly improved with the advent of bromodeoxyuridine (BrdU) and retroviral labeling techniques. Second, a number of groups reported in the early 1990s that a population of cells in the adult rodent forebrain SVZ and dentate gyms possess neural stem cell-like properties, i.e. they can self-renew and generate neurons, astrocytes and oligodendrocytes in vitro (Reynolds and Weiss, 1992; Richards et al., 1992; Lois and Alvarez-Buylla, 1993; Palmer et al., 1997). Proliferation in vitro requires growth factors, such as epidermal growth factor (EGF), basic fibroblast growth factor (bFGF), or brain-derived neurotrophic factor (BDNF) (Reynolds and Weiss, 1992; Lois and Alvarez-Buylla, 1993; Kirschenbaum and Goldman, 1995; Gritti et al., 1996). Both quiescent and constitutively proliferating populations of rostral SVZ precursor cells appear to exist in vivo (Morshead and van der Kooy, 1992; Morshead et al., 1998; Doetsch et al., 1999), although the identity of the putative neural stem cell in the adult rodent forebrain SVZ remains controversial (Johansson et al., 1999; Doetsch et al., 1999). Remarkably, evidence from a recent study of neurogenesis in the adult primate brain indicates that the capacity for neuronal renewal may not be limited solely to the phylogenetically older olfactory bulb and dentate gyms brain regions suggested by rodent neurogenesis studies. Gould and colleagues have shown that the forebrain SVZ of the adult monkey is a proliferative region that generates precursor cells capable of migrating through the mature white matter to differentiate into neurons in multiple cortical regions, including neocortical association areas (Gould et al., 1999). This finding, if confirmed, has important implications for the therapeutic use of endogenous neural precursors as a source for neuronal replacement after injury.
Modulation of adult neurogenesis The physiological role of neurogenesis in the mature brain is unknown. Recent evidence raises the possibility that DGC neurogenesis in the adult rat is necessary for certain forms of hippocampal-dependent learning and memory (Shots et al., 2001). Similarly, studies of altered neurogenesis in the adult rodent SVZ-olfactory bulb pathway implicate this system in specific types of olfactory learning (Gheusi et al., 2000). These data fit well with the current understanding of the role of neurogenesis in adult songbird learning (Scharff et al., 2000). In addition to its physiological function, the molecular mechanisms that regulate adult neurogenesis remain poorly understood. The rate of DGC neurogenesis during early postnatal and adult life appears to be influenced, at least in part, by factors such as aging, environmental stimulation, exercise, glucocorticoid hormone levels and glutamatergic input to the DGC layer (reviewed in Gage et al., 1998). Other neurochemical systems implicated in modulating adult DGC neurogenesis include serotonin, dopamine, and opioids (Dawirs et al., 1998; Brezun and Daszuta, 1999; Eisch et al., 2000). Growth or neurotrophic factors have also been shown to influence cell proliferation and/or neurogenesis in adult rodent germinative zones. Increased DGC neurogenesis has been reported after administration of bFGF or insulin-like growth factor1 (IGF-1) (Wagner et al., 1999; Aberg et al., 2000), while bFGF, EGF and BDNF appear to alter neurogenesis and/or gliogenesis in the adult rodent rostral SVZ-olfactory bulb pathway (Craig et al., 1996; Kuhn et al., 1997; Zigova et al., 1998; Wagner et al., 1999). As mentioned above, the presence of ongoing neurogenesis in the mature brain raises the possibility that endogenous precursor cells could be used therapeutically for repair of neuronal loss associated with brain injuries or degenerative disorders (Lowenstein and Parent, 1999). However, the response of endogenous neural stem or precursor cells to cerebral injury and their potential involvement in neurological disease pathophysiology have received relatively little attention. A number of recent investigations suggest that various forms of injury accelerate neural (i.e. both neuronal and glial) precursor proliferation in the adult rodent dentate gyms and rostral fore-
124 brain SVZ. In addition to studies of seizure-induced injury described below, increased DGC neurogenesis has been found after mechanical, excitotoxic, or ischemic lesions of adult rodent brain (Gould and Tanapat, 1997; Liu et al., 1998; Takagi et al., 1999). Several investigations also describe injury-induced increases in adult rodent forebrain SVZ precursor cell proliferation. The types of injury include aspiration or transection lesions of the forebrain (Willis et al., 1976; Szele and Chesselet, 1996; Weinstein et al., 1996), inflammatory or chemical demyelination (Calz~ et al,, 1998; Nait-Oumesmar et al., 1999), and percussion trauma (Holmin et al., 1997). Some of this work suggests that SVZ precursors give rise to astrocytes and oligodendrocytes after brain injury (Holmin et al., 1997; Nait-Oumesmar et al., 1999). However, a recent investigation of endogenous forebrain SVZ precursors in an adult rat Parkinson's disease model has shown increased proliferation, directed migration and neuronal differentiation of SVZ progenitor cells induced by combined 6-hydroxydopamine lesions and transforming growth factor-ct infusion (Fallon et al., 2000).
Seizure-induced adult neurogenesis and gliogenesis Several recent studies have shown that DGC neurogenesis in adult rats increases in various rodent models of limbic epileptogenesis or acute seizures (Bengzon et al., 1997; Parent et al., 1997, 1998; Gray and Sundstrom, 1998; Scott et al., 1998). In the kainate and pilocarpine models of temporal lobe epilepsy, chemoconvulsant-induced status epilepticus (SE) increases cell proliferation by approximately 5- to 10fold in the adult rat dentate gyrus after a latent period of at least several days (Parent et al., 1997; Gray and Sundstrom, 1998). Most of the newly born cells differentiate into DGCs and disperse into the granule cell layer. A similar, albeit less dramatic, effect on DGC neurogenesis also occurs in electrical kindling models of epileptogenesis, including amygdala (Scott et al., 1998; Parent et al., 1998), hippocampal (Bengzon et al., 1997) and perforant path (Nakagawa et al., 2000) kindling. Acute seizures in adult rats induced by intermittent perforant path stimulation or brief hippocampal stimulation also accelerate DGC neurogenesis (Bengzon et al., 1997; Parent et
al., 1997). Remarkably, even single afterdischarges produced by hippocampal stimulation are capable of increasing the number of newly differentiated DGCs two weeks later (Bengzon et al., 1997). Although these findings raise the possibility that electrical activation directly stimulates mitotic activity, apoptotic cell death in the DGC layer also occurs after even brief, discrete seizure-like discharges (Sloviter et al., 1996; Bengzon et al., 1997). Therefore, seizures may act to increas e neurogenesis indirectly through injury leading to cell turnover in the dentate gyrns. This idea is supported by findings of a relationship between cell death and subsequent cell birth in a number of postnatal neurogenic systems, including the higher vocal center of adult songbirds (Scharff et al., 2000) and the rodent olfactory bulb and dentate gyrus (Gould and McEwen, 1993; Biebl et al., 2000). The molecular link between seizure-induced injury and increased proliferation and/or survival of newly generated DGCs is not known. Among the candidate molecules upregulated by seizure activity are growth factors (Riva et al., 1992; Humpel et al., 1993; Gall et al., 1994; Young and Dragunow, 1995; Opanashuk et al., 1999), neurotrophins (Ernfors et al., 1991; Isackson et al., 1991; Dugich-Djordjevic et al., 1992) and extracellular matrix molecules (Ferhat et al., 1996). Specific neurotransmitters or neuromodulatory systems mentioned above that normally influence DGC neurogenesis may also be altered by seizure activity. To begin to address potential mechanisms of seizure-induced DGC neurogenesis, we recently asked whether constitutively proliferating precursors are activated by seizures or if, instead, quiescent progenitors are recruited to proliferate in the dentate subgranular zone. We labeled the constitutively proliferating cells by systemic BrdU administration one day prior to inducing SE with pilocarpine, and then identified the labeled cells immunohistochemically between 2 and 14 days later (Parent et al., 1999). Interestingly, after a latent period of 4-7 days we found that SE accelerated the proliferation of cells normally dividing prior to any injury. Although these data do not exclude the involvement of quiescent precursors, it suggests that the mechanisms involved in seizure-induced neurogenesis at least in part relate to accelerated cell division of mitotically active neural precursors.
125 More recently, we have identified a second constitutively proliferating precursor population that expands after pilocarpine-induced SE. These precursors arise from the caudal SVZ at the level of the dorsal hippocampus. When they are labeled with BrdU 1-2 days before SE, the number of BrdUimmunoreactive cells 10-14 days later is markedly increased in pilocarpine-treated adult rats compared to controls. We confirmed this increase in immature cells by immunostaining for the differentiating precursor markers PSA-NCAM, doublecortin and CRMP-4. These studies showed expansion of precursors with migratory cell morphology in the caudal SVZ, infracallosal region and areas CA1 and CA3 of the hippocampus after seizure-induced injury. To confirm migration, retroviral reporters were stereotaxically injected into the caudal SVZ prior to seizures, and 2-3 weeks later labeled cells were found in these same regions. Surprisingly, all of these cells showed a glial morphology (oligodendrocytic or astrocytic) and failed to co-express neuronal markers. In controls, retroviral reporter-labeled cells appeared only in the caudal SVZ and corpus callosum after caudal SVZ injections. These findings suggest that pilocarpine-induced SE accelerates the proliferation of glial-lineage restricted precursors in the caudal SVZ. Moreover, it is likely that cues arising from the injured hippocampus redirect the migration of newly generated glioblasts to sites of damage after SE. Understanding how newly generated glia influence network function in the hippocampus may provide insight into epileptogenic mechanisms. Furthermore, these precursors may be useful as vehicles to deliver specific gene products to sites of injury for future antiepileptogenic treatment strategies. The effect of seizures on the other persistent germinative zone in the adult, the rostral forebrain SVZ, has been relatively unexplored. Based on previous findings of injury-induced cell proliferation in the rostral SVZ and seizure-induced DGC neurogenesis in adult rodents, we asked whether prolonged seizures also increase neurogenesis in the adult rat rostral SVZ. Using the pilocarpine model of limbic epileptogenesis, we recently found that 2 h of SE markedly upregulates cell proliferation, as measured by BrdU labeling and immunostaining for an endogenous cell cycle marker, in the adult rat forebrain SVZ and RMS. Immunostaining for markers of
immature neurons revealed that this change in proliferative activity resulted in increased neurogenesis in these same brain regions. The majority of neurons newly generated after seizures migrate through the normal RMS pathway to their appropriate targets in the olfactory bulb. However, their migration to the olfactory bulb is accelerated after SE. We labeled proliferating cells by systemic BrdU administration or focal injection of retroviral reporters into the rostral SVZ, and found that they reached the olfactory bulb much more rapidly after pilocarpine treatment as compared to saline-treated controls. Moreover, a significant proportion of the neuroblasts arising from the SVZ after SE appeared to exit the RMS prematurely and migrate into injured forebrain regions. The results of these experiments are summarized in Fig. 2.
Potential effects of altered neurogenesis in epilepsy models We are only at the earliest stages of understanding the effects of seizure-induced neurogenesis in the adult mammalian brain. Again, most of the work addressing this question has been directed at altered neurogenesis in the epileptic adult rodent dentate gyms. Although the occurrence of increased neuronal birth after seizures suggests the potential for compensatory effects in the setting of injury, our initial hypothesis was that newly generated DGCs were responsible for aberrant mossy fiber reorganization in the epileptic hippocampal formation. In the pilocarpine model of TLE, aberrant mossy fiber reorganization typically begins during the second week after SE and peaks after approximately two months (Cavalheiro et al., 1991; Mello et al., 1993). Thus, the birth and subsequent differentiation of increased numbers of developing DGCs induced by prolonged seizure activity parallels the time course of mossy fiber remodeling in this model. Importantly, seizureinduced mossy fiber synaptic reorganization in the adult rat closely resembles pathological findings in human TLE (Tauck and Nadler, 1985; Cronin and Dudek, 1988; Sutula et al., 1989; Houser et al., 1990; reviewed in Parent and Lowenstein, 1997). To test our hypothesis, we first used BrdU labeling and immunostaining for axonal markers to determine whether newly born DGCs contribute to
126
IIIIIIIII v
Normal . , ., _ , . . . . .s v z LV
~oo t~
RMS
,,,
,,,,,
RAizurA
Fig. 2. Model of seizure-induced rostral SVZ neurogenesis in the adult rat. The top panel shows normal adult SVZ-olfactory bulb neurogenesis. Pilocarpine-induced status epilepticus (bottom panel) induces the following: (1) the SVZ and RMS expand with increased neuronal precursors; (2) neuroblast migration to the olfactory bulb is accelerated; (3) some neuroblasts exit the RMS prematurely, migrate ectopically into the forebrain, and differentiate into neurons. Cx = cortex; LV = lateral ventricle; SVZ = subventricular zone; RMS = rostral migratory stream; OB ----olfactory bulb.
seizure-induced mossy fiber remodeling in adult rats (Parent et al., 1997). We found that developing axons from newly born DGCs participated in aberrant mossy fiber reorganization in both area CA3 and the dentate inner molecular layer. To further test our hypothesis, we inhibited DGC neurogenesis with whole brain X-irradiation and then examined whether pilocarpine-induced SE still caused mossy fiber remodeling (Parent et al., 1999). Despite a nearly complete reduction of newborn DGCs after irradiation, robust aberrant supragranular mossy fiber
remodeling appeared four weeks after pilocarpine treatment. Taken together, these data are consistent with the idea that both newly born and mature DGCs respond to seizure-induced injury with the aberrant outgrowth or sprouting of axons to atypical sites (Fig. 3). The molecular signals responsible for mossy fiber remodeling after seizures remain unknown. A second abnormality related to seizure-induced DGC neurogenesis concerns the ectopic location of newborn granule neurons in the epileptic hippocampal formation. In human TLE, the granule cell layer
128 ever, unlike mature granule neurons in the DGC layer, the ectopic hilar granule cells exhibited abnormal burst firing in synchrony with CA3 pyramidal cells. In addition, many putatively newborn DGC located in the hilus and hilar aspect of the DGC layer after seizures exhibit a much higher percentage of persistent basal dendrites than is normally seen in DGCs of adult rodents (Spigelman et al., 1998; Buckmaster and Dudek, 1999; Ribak et al., 2000). Similar changes also occur in the dentate gyms of the p35 (neuronal-specific activator of cyclin-dependent kinase 5)-deficient mouse, a model of cerebral dysgenesis and epilepsy (Wenzel et al., 2001), suggesting that such changes are developmental abnormalities. Importantly, Ribak and colleagues have found morphological evidence of substantially increased net excitatory synapses on the basal dendrites of hilar ectopic DGCs (Ribak et al., 2000). These investigators have suggested that this synaptic reorganization may be a mechanism for seizure generation. Beyond the data supporting a pro-epileptogenic role of seizure-induced DGC neurogenesis, the findings of ectopic hilar DGCs have implications for brain repair potential after injury. We have found newborn hilar and molecular layer granule-like cells after SE even in animals with significant injury to the superior blade of the DGC layer (J.M. Parent and D.H. Lowenstein, unpublished data). Furthermore, endogenous DGC precursors still appear to differentiate into DGCs (by the criteria of morphology and antigen expression) even when they migrate to atypical locations. These findings suggest that brain repair strategies will need to overcome at least two obstacles to achieve neuronal replacement after injury using endogenous or transplanted neural precursor cells. First, competing cues may exist at sites of injury that could direct migration of neural progenitors to inappropriate locations (e.g. the dentate hilus instead of injured DGC layer). Second, the local environment may not maintain the cues necessary to direct differentiation of neural progenitors into the appropriate cell types. In addition to epileptogenesis, seizure-induced alterations in neurogenesis may have implications for the memory dysfunction associated with chronic epilepsy. Although unproven, existing evidence supports a relationship between adult DGC neurogenesis
and certain forms of hippocampal learning and memory in the rodent (Shors et al., 2001). Learning and memory dysfunction has been demonstrated following limbic kindling of adult rats (Sutula et al., 1995), and alterations of DGC networks may participate in such disturbances given the presumed involvement of the hippocampal formation in mammalian learning and memory (Jarrard, 1993). The potential role of altered DGC neurogenesis in cognitive impairments associated with brain irradiation to treat humans with brain tumors has also been raised recently (Parent et al., 1999; Tada et al., 2000). Perhaps seizures alter the network integration of newly generated DGCs (via ectopic location or persistent basal dendrites), and this leads to dysfunction of circuits necessary for memory acquisition or retrieval. How potential seizure-induced changes in forebrain SVZ neurogenesis, which is putatively involved in olfactory learning in rodents (Gheusi et al., 2000), may relate to interictal abnormalities in human epilepsy is even more speculative. The recent finding of seizure-induced neurogenesis in the adult mammalian forebrain SVZ also raises some other interesting questions. For example, how does seizure activity or seizure-induced injury stimulate SVZ neurogenesis and alter the normal pattern of neuroblast migration? Do differentiating neurons that migrate ectopically into the forebrain survive and integrate into existing networks? If so, do they serve to reestablish normal connections after injury, or do they participate in the formation of abnormal networks that may predispose to seizure generation? These issues are important in light of the evidence that endogenous neuronal precursors in the adult rat SVZ and dentate gyrus respond similarly to seizures and other forms of brain injury. Understanding the molecular regulation of adult mammalian neural precursors is a necessary step toward achieving the capacity to modify or direct the proliferation, migration and differentiation of neural stem cells in the setting of acute brain injury or neurodegeneration. The ability to manipulate this response may then offer strategies for brain repair or antiepileptogenic therapies. References ~berg, M.A.I., ,~berg, N.D., Hedb~icker, H., Oscarsson, J. and Eriksson, ES. (2000) Peripheral infusion of IGF-1 selectively
127
G,.,,.j%j ¢
Hilu
to CA3
Fig. 3. Model of seizure-induced dentate granule cell neurogenesis in the adult rat. Proliferation of neural precursors (round gray cell under the GCL) in the subgranular zone accelerates after pilocarpine-induced status. Some precursors differentiate into DGCs in the granule cell layer (GCL), and both newly generated and mature DGCs contribute to aberrant mossy fiber reorganization in the inner molecular layer (above the GCL). Other precursors may form glial ceils or undergo seizure-induced chain migration into the hilus to produce hilar ectopic DGCs. is often relatively preserved but may show abnormalities that include dispersion of the layer and the presence of ectopic granule-like neurons in the hilus and inner molecular layer (Houser, 1990). DGC dispersion also occurs in the adult rat pilocarpine model of TLE (Mello et al., 1992). In our initial studies of seizure-induced neurogenesis (Parent et al., 1997), we found that newly differentiating neurons with granule cell morphology appeared in unusual locations, i.e. the dentate hilus and inner molecular layer. These cells resemble the 'ectopic' granule-like neurons identified in surgical specimens from humans with temporal lobe epilepsy (Houser, 1990). We also observed a progressive increase in large BrdU-immunolabeled nuclei in the hilus with increasing time after SE, as well as chains of migrating
neuroblasts extending from the inner DGC layer to the hilus (Fig. 3). Although the precise origin of these cells is unknown, their appearance following pilocarpine treatment, but not in controls, suggests that they migrate aberrantly from the dentate SGZ to the hilus after prolonged seizures. Several groups have subsequently confirmed and extended the finding of newly generated ectopic hilar DGCs after kainic acid- or pilocarpine-induced SE (Scharfman et al., 2000; Dashtipour et al., 2001). Importantly, Scharfman and colleagues studied the electrophysiological features of the hilar ectopic granule-like neurons using intracellular recordings in hippocampal slices from epileptic adult rats. They found that these ectopic cells maintained many of the electrophysiological characteristics of DGCs. How-
129
induces neurogenesis in the adult rat hippocampus. J. Neurosci., 20: 2896-2903. Allen, E. (1912) The cessation of mitosis in the central nervous system of the albino rat. J. Comp. Neurol., 22: 547-568. Altman, J. (1969) Autoradiographic and histological studies of postnatal neurogenesis, IV. Cell proliferation and migration in the anterior forebrain, with special reference to persisting neurogenesis in the olfactory bulb. J. Comp. Neurol., 137: 433-458. Altman, J. and Das, G.D. (1965) Autoradiographic and histological evidence of postnatal hippocampal neurogenesis in rats. J. Comp. Neurol., 124: 319-335. Bengzon, J., Kokaia, Z., Elmer, E., Nanobashvili, A., Kokaia, M. and Lindvall, O, (1997) Apoptosis and proliferation of dentate gyrus neurons after single and intermittent limbic seizures. Proc. Natl. Acad. Sci. USA, 94: 10432-10437. Biebl, M., Cooper, C.M., Winkler, J. and Kuhn, H.G. (2000) Analysis of neurogenesis and programmed cell death reveals a self-renewing capacity in the adult rat brain. Neurosci Lett., 291: 17-20. Bonfanti, L. and Theodosis, D.T. (1994) Expression of polysialylated neural cell adhesion molecule by proliferating cells in the subependymal layer of the adult rat, in its rostral extension and in the olfactory bulb. Neuroscience, 62: 291-305. Brezun, J.M. and Daszuta, A. (1999) Depletion in serotonin decreases neurogenesis in the dentate gyrus and the subventricular zone of adult rats. Neuroscience, 89: 999-1002. Buckmaster, P.S. and Dudek, EE. (1999) In vivo intracellular analysis of granule cell axon reorganization in epileptic rats. J. Neurophysiol., 81: 712-721. Calz~t, L., et al. (1998) Proliferation and phenotype regulation in the subventricular zone during experimental allergic encephalomyelitis: in vivo evidence of a role for nerve growth factor. Proc. Natl. Acad. Sci. USA, 95: 3209-3214. Cameron, H.A., Woolley, C.S., McEwen, B.S. and Gould, E. (1993) Differentiation of newly born neurons and glia in the dentate gyrus of the adult rat. Neuroscience, 56: 337-344. Cavalheiro, E.A., Leite, J.R, Bortolotto, Z.A., Turski, W.A., Ikonomidou, C. and Turski, L. (1991) Long-term effects of pilocarpine in rats: structural damage of the brain triggers kindling and spontaneous recurrent seizures. Epilepsia, 32: 778-782. Craig, C.G., Tropepe, V., Morshead, C.M., Reynolds, B.A., Weiss, S. and van der Kooy, D. (1996) In vivo growth factor expansion of endogenous subependymal neural precursor cell populations in the adult mouse brain. J. Neurosci., 16: 26492658. Cronin, J. and Dudek, EE. (1988) Chronic seizures and collateral sprouting of dentate mossy fibers after kainic acid treatment in rats. Brain Res., 474: 181-184. Dashtipour, K., Tran, EH., Okazaki, M.M., Nadler, J.V. and Ribak, C.E. (2001) Ultrastructural features and synaptic connections of hilar ectopic granule cells in the rat dentate gyms are different from those of granule cells in the granule cell layer. Brain Res., 890: 261-271. Dawirs, R.R., Hildebrandt, K. and Teuchert-Noodt, G. (1998) Adult treatment with haloperidol increases dentate granule cell
proliferation in the gerbil hippocampus. J. Neural Transm., 105: 317-327. Doetsch, F. and Alvarez-Buylla, A. (1996) Network of tangential pathways for neuronal migration in adult mammalian brain. Proc. Natl. Acad. Sci. USA, 93: 14895-14900. Doetsch, F., Caillt, I., Lim, D.A., Garcfa-Verdugo, J.M, and Alvarez-Buylla, A. (1999) Subventricular zone astrocytes are neural stem cells in the adult mammalian brain. Cell, 97: 703716. Dugich-Djordjevic, M.M., et ak (1992) BDNF mRNA expression in the developing rat brain following kainic acid-induced seizure activity. Neuron, 8:1127-1138. Eisch, A.J., Barrot, M., Schad, C.A., Self, D.W. and Nestler, EJ. (2000) Opiates inhibit neurogenesis in the adult rat hippocampus. Proc. Natl. Acad. Sci. USA, 97: 7579-7584. Eriksson, P.S., Perfilieva, E., Bjork-Eriksson, T., Alborn, A., Nordborg, C., Peterson, D.A. and Gage, EH. (1998) Neurogenesis in the adult human hippocampus. Nat. Med., 4: 13131317. Ernfors, P., Bengzon, J., Kokaia, Z., Persson, H. and Lindvall, O. (1991) Increased levels of messenger RNAs for neurotrophic factors in the brain during epileptogenesis. Neuron, 7: 165176. Fallon, J., et al. (2000) In vivo induction of massive proliferation, directed migration, and differentiation of neural cells in the adult mammalian brain. Proc. Natl. Acad. Sci. USA, 97: 14686-14691. Ferhat, L., Chevassus-Au-Louis, N., Khrestchatisky, M., BenAri, Y. and Represa, A. (1996) Seizures induce tenascin-C mRNA expression in neurons. J. NeurocytoL, 25: 535-546. Gage, F.H., Kemperman, G., Palmer, T.D., Peterson, D.A. and Ray, J. (1998) Multipotent progenitor cells in the adult dentate gyrus. J. Neurobiol., 36: 249-266. Gall, C.M., Berschauer, R. and Isackson, P.J. (1994) Seizures increase basic fibroblast growth factor mRNA in adult rat forebrain neurons and glia. Mol. Brain Res., 21: 190-205. Gheusi, G., Cremer, H., McLean, H., Chazal, G., Vincent, J. and Lledo, P. (2000) Importance of newly generated neurons in the adult olfactory bulb for odor discrimination. Proc. Natl. Acad. Sci. USA, 97: 1823-1828. Gould, E. and McEwen, B.S. (1993) Neuronal birth and death. Curr. Opin. Neurobiol., 3: 676-682. Gould, E. and Tanapat, P. (1997) Lesion-induced proliferation of neuronal progenitors in the dentate gyrus of the adult rat. Neuroseience, 80: 427-436. Gould, E., Tanapat, P., McEwen, B.S., Fltigge, G. and Fuchs, E. (1998) Proliferation of granule cell precursors in the dentate gyms of adult monkeys is diminished by stress. Proc. Natl. Acad. Sci. USA, 95: 3168-3171. Gould, E., Reeves, AJ., Graziano, M.S.A. and Gross, C.G. (1999) Neurogenesis in the neocortex of adult primates. Science, 286: 548-552. Gray, W.P. and Sundstrom, L.E. (1998) Kainic acid increases the proliferation of granule cell progenitors in the dentate gyrus of the adult rat. Brain Res., 790: 52-59. Gritti, A., et al. (1996) Multipotential stem cells from the adult
130
mouse brain proliferate and self-renew in response to basic fibroblast growth factor. J. Neurosci., 16:1091-1100. Hinds, J.W. (1968) Autoradiographic study of histogenesis in the mouse olfactory bulb. J. Comp. Neurol., 134: 287-304. Holmin, S., Almqvist, P., Lendahl, U. and Mathiesen, T. (1997) Adult nestin-expressing subependymal cells differentiate to astrocytes in response to brain injury. Eur. J. Neurosci., 9: 65-75. Houser, C.R. (1990) Granule cell dispersion in the dentate gyms of humans with temporal lobe epilepsy. Brain Res., 535: 195204. Houser, C.R., Miyashiro, J.E., Swartz, B.E., Walsh, G.O., Rich, J.R. and Delgado-Escueta, A.V. (1990) Altered pattens of dynorphin immunoreactivity suggest mossy fiber reorganization in human hippocampal epilepsy. J. Neurosci., 10: 267282. Humpel, C., Lippoldt, A., Chadi, G., Ganten, D., Olson, L. and Fuxe, K. (1993) Fast and widespread increase of basic fibroblast growth factor messenger RNA and in the forebrain after kainate-induced seizures. Neuroscience, 57: 913-922. Isackson, P.J., Huntsman, M.M., Murray, K.D. and Gall, C. (1991) BDNF mRNA expression is increased in adult rat forebrain after limbic seizures: temporal pattens of induction distinct from NGF. Neuron, 6: 937-948. Jarrard, L. (1993) On the role of the hippocampus in learning and memory in the rat. Behav. Neural Biol., 60: 9-26. Johansson, C.B., Momma, S., Clarke, D.L., Risling, M., Lendahl, U. and Fris6n, J. (1999) Identification of a neural stem cell in the adult mammalian central nervous system. Cell, 96: 25-34. Kaplan, M.S. and Hinds, J.W. (1977) Neurogenesis in adult rat: electron microscopic analysis of light radioautographs. Science, 197: 1092-1094. Kirschenbaum, B. and Goldman, S. (1995) BDNF promotes the survival of neurons arising from the adult rat forebrain subventricular zone. Proc. Natl. Acad. Sci. USA, 92: 210-214. Kishi, K. (1987) Golgi studies on the development of granule cells of the rat olfactory bulb with reference to migration in the subependymal layer. J. Comp. Neurol., 258:112-124. Kornack, D.R. and Rakic, P. (1999) Continuation of neurogenesis in the hippocampus of the adult macaque monkey. Proc. NatL Acad. Sei. USA, 96: 5768-5773. Kuhn, H.G., Dickinson-Anson, H. and Gage, EH. (1996) Neurogenesis in the dentate gyms of the adult rat: age-related decrease of neuronal progenitor proliferation. J. Neurosci., 16: 2027-2033. Kuhn, H.G., Winkler, J., Kempermann, G., Thal, L.J. and Gage, EH. (1997) Epidermal growth factor and fibroblast growth factor-2 have different effects on neural progenitors in the adult rat brain. J. Neurosci., 17: 5820-5829. Liu, J., Solway, K., Messing, R.O. and Sharp, ER. (1998) Increased neurogenesis in the dentate gyms after transient global ischemia in gerbils. J. Neurosci., 18: 7768-7778. Lois, C. and Alvarez-Buylla, A. (1993) Proliferating subventricular zone cells in the adult mammalian forebrain can differentiate into neurons and glia. Proc. Natl. Acad. Sci. USA, 90: 2074-2077. Lois, C. and Alvarez-Buylla, A. (1994) Long-distance neuronal
migration in the adult mammalian brain. Science, 264: 11451148. Lois, C., Garcfa-Verdugo, J.M. and Alvarez-Buylla, A. (1996) Chain migration of neuronal precursors. Science, 271: 978981. Lowenstein, D,H. and Parent, J.M. (1999) Brain, heal thyself. Science, 283:1126-1127. Luskin, M. (1993) Restricted proliferation and migration of postnatally generated neurons derived from the forebrain subventricular zone. Neuron, 11: 173-189. Magavi, S.S., Leavitt, B.R. and Macklis, J.D. (2000) Induction of neurogenesis in the neocortex of adult mice. Nature, 405: 951-955. Markakis, E.A. and Gage, EH. (1999) Adult-generated neurons in the dentate gyms send axonal projections to field CA3 and are surrounded by synaptic vesicles. J. Comp. Neurol., 406: 449-460. Mello, L.E., Cavalheiro, E.A., Tan, A.M., Kupfer, W.R., Pretoflus, J.K., Babb, T.L. and Finch, D.M. (1993) Circuit mechanisms of seizures in the pilocarpine model of chronic epilepsy: cell loss and mossy fiber sprouting. Epilepsia, 34: 985-995. Mello, L.M., Cavalheiro, E., Tan, A., Pretorius, J.K., Babb, T. and Finch, D. (1992) Granule cell dispersion in relation to mossy fiber sprouting, hippocampal cell loss, silent period and seizure frequency in the pilocarpine model of epilepsy. In: J. Engel Jr., C. Wasterlain, E.A. Cavalheiro, U. Heinemann and G. Avanzini (Eds.), Molecular Neurobiology of Epilepsy. Elsevier, Amsterdam, pp. 51-60. Morshead, C.M. and van der Kooy, D. (1992) Postmitotic death is the fate of constitutively proliferating cells in the subependymal layer of the adult mouse brain. J. Neurosci., 12: 249-256. Morshead, C.M., Craig, C.G. and van der Kooy, D. (1998) In vivo clonal analyses reveal the properties of endogenous neural stem cell proliferation in the adult mammalian forebrain. Development, 125: 2251-2261. Nacher, J., Crespo, C, and McEwen, B.S. (2001) Doublecortin expression in the adult rat telencephalon. Eur. J. Neurosci., 14: 629-644. Nacher, J., Rosell, D.R. and McEwen, B.S. (2000) Widespread expression of collapsin response-mediated protein 4 in the telencephalon and other areas of the adult rat central nervous system. J. Comp. Neurol., 424: 628-639. Nait-Oumesmar, B., et al. (1999) Progenitor cells of the adult mouse subventricular zone proliferate, migrate and differentiate into oligodendrocytes after demyelination. Eur. J. Neurosci., 11: 4357-4366. Nakagawa, E., Aimi, Y., Yasuhara, O., Tooyama, 1., Shimada, M,, McGeer, EL. and Kimura, H. (2000) Enhancement of progenitor cell division in the dentate gyms triggered by initial limbic seizures in rat models of epilepsy. Epilepsia, 41: 10-18. Opanashuk, L.A., Mark, R.J., Porter, J., Datum, D., Mattson, M.E and Seroogy, K.B. (1999) Heparin-binding epidermal growth factor-like growth factor in hippocampus: modulation of expression by seizures and anti-excitotoxic action. J. Neurosci., 19: 133-146. Palmer, T.D., Takahashi, J. and Gage, EH. (1997) The adult rat
131
hippocampus contains primordial neural stem cells. Mol. Cell. Neurosci., 8: 389-404. Parent, J.M. and Lowenstein, D.H. (1997) Mossy fiber reorganization in the epileptic hippocampus. Curr. Opin. Neurol., 10: 103-109. Parent, J.M., Yu, T.W., Leibowitz, R.T., Geschwind, D.H., Sloviter, R.S. and Lowenstein, D.H. (1997) Dentate granule cell neurogenesis is increased by seizures and contributes to aberrant network reorganization in the adult rat hippocampus. J. Neurosci., 17: 3727-3738. Parent, J.M., Janumpalli, S., McNamara, J.O. and Lowenstein, D.H. (1998) Increased dentate granule cell neurogenesis following amygdala kindling in the adult rat. Neurosci Lett., 247: 9-12. Parent, J.M., Tada, E., Fike, J.R. and Lowenstein, D.H. (1999) Inhibition of dentate granule cell neurogenesis with brain irradiation does not prevent seizure-induced mossy fiber synaptic reorganization in the rat. J. Neurosci., 19: 4508-4519. Perreto, P., Merighi, A., Fasolo, A. and Bonfanti, L. (1999) The subependymal layer in rodents: a site of structural plasticity and cell migration in the adult mammalian brain. Brain Res. Bull., 49: 221-243. Reynolds, B.A. and Weiss, S. (1992) Generation of neurons and astrocytes from isolated ceils of the mammalian central nervous system. Science, 255: 1707-1710. Ribak, C.E., Tran, P.H., Spigelman, I., Okazaki, M.M. and Nadler, J.V. (2000) Status epilepticus-induced hilar basal dendrites on rodent granule cells contribute to recurrent excitatory circuitry. J. Comp. Neurol., 428: 240-253. Richards, L.J., Kilpatrick, T.J. and Bartlett, P.F. (1992) De novo generation of neuronal cells from the adult mouse brain. Proc. Natl. Acad. Sci. USA, 89: 8591-8595. Riva, M.A., Gale, K. and Mocchetti, I. (1992) Basic fibroblast growth factor mRNA increases in specific brain regions following convulsive seizures. Mol. Brain Res., 15:311-318. Scharff, C., Kirn, J.R., Grossman, M., Macklis, J.D. and Nottebohm, E (2000) Targeted neuronal death affects neuronal replacement and vocal behavior in adult songbirds. Neuron, 25: 481-492. Scharfman, H.E., Goodman, J.H. and Sollas, A.L. (2000) Granule-like neurons at the hilar/CA3 border after status epilepticus and their synchrony with area CA3 pyramidal cells: functional implications of seizure-induced neurogenesis. J. Neurosci., 20: 6144-6158. Scott, B.W., Wang, S., Burnham, W.M., De Boni, U. and Wojtowicz, J.M. (1998) Kindling-induced neurogenesis in the dentate gyms of the rat. Neurosci. Lett., 248: 73-76. Seki, T. and Arai, Y. (1993) Highly polysialylated neural cell adhesion molecule (NCAM-H) is expressed by newly generated granule cells in the dentate gyms of the adult rat. J. Neurosci., 13: 2351-2358. Shors, T.J., Miesegaes, G., Beylin, A., Zhao, M., Rydel, T. and Gould, E. (2001) Neurogenesis in the adult is involved in the formation of trace memories. Nature, 410: 372-376. Sloviter, R.S., Dean, E., Sollas, A.L. and Goodman, J.H. (1996) Apoptosis and necrosis induced in different hippocampal neuron populations by repetitive perforant path stimulation in the rat. J. Comp. Neurol., 366: 516-533.
Spigelman, I., Yan, X.X., Obenaus, A., Lee, E.Y., Wasterlain, C.G. and Ribak, C.E. (1998) Dentate granule cells form novel basal dendrites in a rat model of temporal lobe epilepsy. Neuroscience, 86: 109-120. Stanfield, B.B. and Trice, J.E. (1988) Evidence that granule cells generated in the dentate gyms of adult rats extend axonal projections. Exp. Brain Res., 72: 399-406. Sutula, T., Cascino, G., Cavazos, J., Parada, I. and Ramirez, L. (1989) Mossy fiber synaptic reorganization in the epileptic human temporal lobe. Ann. Neurol., 26: 321-330. Sutula, T., Lauersdorf, S., Lynch, M., Jurgella, C. and Woodard, A, (1995) Deficits in radial ann maze performance in kindled rats: evidence for long-lasting memory dysfunction induced by repeated brief seizures. J. Neurosci., 15: 8295-8301. Szele, F.G. and Chesselet, M. (1996) Cortical lesions induce an increase in cell number and PSA-NCAM expression in the subventricular zone of adult rats. J. Comp. Neurol., 368: 439-454. Tada, E., Parent, J.M., Lowenstein, D.H. and Fike, J.R. (2000) X-irradiation causes a prolonged reduction in cell proliferation in the dentate gyms of adult rats. Neuroscience, 99: 33-41. Takagi, Y., Nozaki, K., Takahashi, J., Yodoi, J., Ishikawa, M. and Hashimoto, N. (1999) Proliferation of neuronal precursor cells in the dentate gyms is accelerated after transient forebrain ischemia in mice. Brain Res., 831: 283-287. Tauck, D. and Nadler, J. (1985) Evidence of functional mossy fiber sprouting in hippocampal formation of kainic acid-treated rats. J. Neurosci., 5: 1016-1022. Thomas, L.B., Gates, M.A. and Steindler, D.A. (1996) Young neurons from the adult subependymal zone proliferate and migrate along an astrocyte, extracellular matrix-rich pathway. Glia, 17: 1-14. Wagner, J.P., Black, I.B. and DiCicco-Bloom, E. (1999) Stimulation of neonatal and adult brain neurogenesis by subcutaneous injection of basic fibroblast growth factor. J. Neurosci., 19: 6006-6016. Weinstein, D.E., Burrola, P. and Kilpatrick, T.J. (1996) Increased proliferation of precursor cells in the adult rat brain after targeted lesioning. Brain Res., 743:11-16. Wenzel, H.J., Robbins, C.A., Tsai, L. and Schwartzkroin, P.A. (2001) Abnormal morphological and functional organization of the hippocampus in a p35 mutant model of cortical dysplasia associated with spontaneous seizures. J. Neurosci., 21: 983-998. Wichterle, H., Garcfa-Verdugo, J.M. and Alvarez-Buylla, A. (1997) Direct evidence for homotypic, glia-independent neuronal migration. Neuron, 18: 779-791. Willis, P., Berry, M. and Riches, A.C. (1976) Effects of trauma on cell production in the subependymal layer of the rat neocortex. Neuropathol. Appl. Neurobiol., 2: 377-388. Young, D. and Dragunow, M. (1995) Neuronal injury following electrically induced status epilepticus with and without adenosine receptor antagonism. Exp. Neurol., 133: 125-137. Zigova, T., Pencea, V., Wiegand, S.J. and Luskin, M.B. (1998) Intraventricular administration of BDNF increases the number of newly generated neurons in the adult olfactory bulb. Mol. Cell. Neurosci., 11 : 234-245.
T. Sutula and A. Pitk~nen (Eds.) Progress in Brain Research, Vol. 135 O 2002 Published by Elsevier Science B.V.
CHAPTER 11
Summary: Seizure-induced damage in experimental models Thomas Sutula x,2,, and Asla Pitk~inen 3,4 1 Department of Neurology and 2 Department of Anatomy, University of Wisconsin, Madison, WI 53792, USA 3 Epilepsy Research Laboratory, A.L Virtanen Institute for Molecular Sciences and 4 Department of Neurology, Kuopio University Hospital, Kuopio, Finland
The association of seizures with brain damage was initially recognized in the early 19th Century with the identification of hippocampal sclerosis in autopsy studies of patients with essentially untreated epilepsy. From this neuropathological observation linking epilepsy to a specific form of hippocampal damage, subsequent experimental efforts have indisputably demonstrated that prolonged seizure activity, as in status epilepticus, is sufficient to produce brain damage even when systemic metabolic conditions were optimized (Chapter 1). There is also longstanding awareness from clinical studies that inadequately treated status epilepticus is often followed by adverse neurological sequelae (Chapter 7). But what about the effects of repeated brief seizures in people with established epilepsy? Do these seizures, which typically are limited to several minutes or less, cause neuronal damage or injury? Section I has addressed this issue at conceptual, experimental, and translational levels based on currently available data. Experimental efforts to determine when and how seizures induce damage have proceeded in what seems to be an ever-expanding range of models of seizures with different clinical manifestations and
* Correspondence to: T. Sutula, Department of Neurology H6/570, University of Wisconsin, Madison, WI 53792, USA. Tel.: +1-608-263-5448; Fax: +1-608-263-0412; E-mail: sutula@ neurology.wisc.edu
seizure durations. While reductionistic approaches have been relatively successful for understanding specific mechanisms contributing to damage caused by severe seizures (see Section II), another form of a reductionistic approach, that is, determining the minimal seizure event that produces damage, has produced as many questions as answers. Some reasons for this complexity are suggested in Chapter 2. For example, in experimental models and in human epilepsy, the effects of seizures cannot be conceptualized merely in terms of how a seizure affects a single neuron. Indeed, neural circuits, as complex systems with many levels of organization, may become dysfunctional (or damaged) by subtle, incremental effects involving a relatively small number of components. Emphasis on the effects of seizures at the level of the whole animal, while difficult both clinically and experimentally, may reveal dysfunction that seems disproportionate to the abnormality detected in a single pathway or by a single measure, for example, the extent of neuron loss. This is a significant obstacle for experimental approaches. Yet it is precisely the system level dysfunctions, for example, increasing frequency of seizures and cumulative seizure-related memory dysfunction, that are troubling permanent features of poorly controlled epilepsy. Among the substantial challenges for assessing the possible damaging effects of status epilepticus and brief seizures are the technical difficulties presented in detecting neuronal loss, which are dis-
134 cussed with critical perspective in Chapters 3 and 4. The epilepsy research literature is filled with a bewildering range of results supporting and denying the occurrence of seizure-induced neuronal loss in a variety of experimental models. Divergent or contradictory interpretations in such studies are often based on technical methods with substantial limitations in sensitivity or variability. Examples of the challenges of stereological analysis are evident in studies of status epilepticus (Chapters 5 and 6) and in models of brief seizures (Chapter 8 and 9). In Chapters 5 and 6, it is reasonably proposed that if spontaneous seizures after status epilepticus add to the extent of neuronal loss, more neuronal loss should be detected at longer intervals after the initial status. While this was observed in Chapter 5, differing quantitative methods at the early and later time points after kainic acid induced status epilepticus in this model precluded the interpretation that the additional seizures contributed to further neuronal loss. In a model of recurring spontaneous seizures following status epilepticus induced by amygdala stimulation (Chapter 6), stereological analysis, histological methods for labeling damaged neurons (Fluoro-Jade B), and MRI methods detected initial status-induced neuronal loss, but there was no evidence that recurring spontaneous seizures produced additional damage. Progressive cell loss within the 2 months after status epilepticus was associated with the initial insult, that is, status epilepticus itself. In contrast, cumulative neuronal loss was detected after repeated brief seizures evoked by kindling in a pattern that resembled hippocampal sclerosis, which was associated with long-term seizure-induced memory dysfunction (Chapter 8). Congruent with these findings and multiple previous stereological studies (Chapter 8), apoptosis has been detected in the dentate gyrus after single evoked seizures (Chapter 9), an observation also confirmed in the dentate gyrns and in other hippocampal and cortical regions (see Pretel et al., 1997; Zhang et al., 1998, and Table 3 in Chapter 8). While application of quantitative methods for detection of seizure-induced neuronal loss is not simple, and rigorous application of these methods may be challenging, it seems doubtful that the apparently divergent outcomes in these studies (Chapters 6, 8 and 9) in regard to damaging effects of brief seizures can be discounted simply as artifacts of the particu-
lar analysis methods employed in a given study. Because severe brain damage precedes the occurrence of spontaneous seizures in status epilepticus models (Chapters 5 and 6), the most seizure-sensitive neuronal populations may already be partially or completely lost. In kindling, however, the neuronal circuitry is intact at the time of the first stimulations. Both models ultimately are associated with memory dysfunction, which strongly indicates that repeated seizures, whether the result of status epilepticus or cumulative brief seizures, have long-term deleterious effects on neuronal circuitry. Similar long-term adverse effects are now also apparent in a variety of experimental models of seizures in developing animals, even when overt damage or neuronal loss is not apparent (Chapters 28 and 32). What are the possible explanations for the apparent discrepancy between the effects of repeated seizures after status and in animals whose neural circuitry is normal when seizures are first induced? Certainly different sensitivity of the assessment methods must always be considered in these studies, and in the design of future experiments. A second possibility is that model-related differences in seizure type and duration may be contributing. The recurrent seizures that evolve in status models are often partial or quite brief in duration (44-60 s, see Chapter 6). In contrast, repeated secondary generalized seizures evoked in kindled animals are typically of 1-2 min or more in duration in advanced stages of kindling. This difference is potentially significant from the point of view of neuronal metabolism (Section II), and is also of interest given evidence in human epilepsy that hippocampal volume reduction (Chapters 22, 24, 25, 26) and cognitive impairment (Chapter 35) are related to the number of lifetime generalized tonicclonic seizures. A third possibility is that seizureinduced neurogenesis (Chapter 10) contributes to the outcome and the differences among studies in the different models, which deserves further experimental consideration. One is left with the impression that a definitive experiment on the minimally damaging seizure has not yet been performed, and that the possible differences in factors such as seizure duration and neurogenesis in different models have not received adequate attention. A further challenge is the possibility, if not the likelihood, that the effects of seizures in a 'complex
135 system' such as neural circuitry may be influenced conditionally by so many variables that the m i n i m a l seizure that produces damage may vary depending on the dynamic conditions of the neural circuits and systemic factors. Whatever the uncertainty about the m i n i m a l seizure that produces damage, studies in a variety of experimental models have irrefutably demonstrated that repeated seizures, whether prolonged or brief, have extensive and profound long-term effects in neural circuitry, and provide no reassurance that these effects can be casually regarded as benign. This impression is unavoidable in the studies of adult animals in this volume, and is also supported by experiments assessing the long-term effects of seizures in developing animals, where numerous adverse longterm consequences are now being recognized in the absence of overt morphological damage (Chapters 28, 32 and 33).
In attempting to assess the translational significance of these experimental studies for human epilepsy, Simon Shorvon critically comments in Chapter 7 on the widely held view that "evidence in humans has been more difficult to define or quantify. In the past, indeed, several authorities have doubted that damage is prominent or clinically relevant." The reader is encouraged to critically assess the latter viewpoint in light of emerging clinical studies in humans presented in Sections III-VI. References Pretel, S., Applegate, C.D. and Piekut, D. (1997) Apoptotic and necrotic cell death following kindling induced seizures. Acta Histochem., 99(1): 71-79. Zhang, L.X., Smith, M., Li, X., Weiss, S. and Post, R.M. (1998) Apoptosis of hippocampal neurons after amygdala kindled seizures. Mol. Brain Res., 55(2): 198-208.
T. Sutula and A. Pitk~inen (Eds.) Progress in Brain Research, Vol. 135 © 2002 Elsevier Science B.V. All rights reserved
CHAPTER 12
Complications associated with genetic background effects in models of experimental epilepsy P. Elyse S c h a u w e c k e r * Department of Cell and Neurobiology, University of Southern California, Keck School of Medicine, Los Angeles, CA 90089-9112, USA
Abstract: To elucidate the genetic influences contributing to susceptibility to seizure disorders, researchers have long used selected lines and inbred strains of rodents. In recent years, the use of genetically altered mice as models of complex human disease has revolutionized biomedical research into the genetics of disease pathogenesis and potential therapeutic interventions. In particular, the study of transgenic and gene-deleted (knockout) mice can provide important insights into the in vivo function and interaction of specific gene products. While a variety of inbred mouse mutations have been used to directly evaluate the genetic basis of seizure disorders, data obtained from such genetically altered mice must be interpreted carefully. An increasing number of scientific articles have reported that the phenotype of a given single gene mutation in mice can be modulated by the genetic background of the inbred strain in which the mutation is maintained. This effect is attributable to so-called modifier genes, which act in combination with the causative gene. In this review, the author points out the importance of considering the genetic background of the strain used to create these animal models, the potential problems with interpretation of phenotype, and solutions to selecting an appropriate mouse model of experimental epilepsy. Despite these potential limitations, knockout mice provide a powerful tool for understanding the genetic and neurobiological mechanisms contributing to experimental epilepsy.
Introduction
While recent progress in mapping human genes for epilepsy indicates that genetic factors contribute to the etiology of seizure disorders (reviewed in Anderman, 1982; Elmslie and Gardiner, 1995; reviewed in Bate and Gardiner, 1999; Serratosa, 1999; Steinlein, 1999; Gardiner and Lehesjoki, 2000), it is becoming clear that the genetic susceptibility to these disorders is complex. Additionally, it has been suggested that a cascade of biological events involving increased
* Correspondence to: P.E. Schauwecker, Department of Cell and Neurobiology, University of Southern California, Keck School of Medicine, BMT 401, 1333 San Pablo Street, Los Angeles, CA 90089-9112, USA. Tel.: +1-323442-2116; Fax: + 1-323-442-3466; E-mail: schauwec @hsc.usc.edu
excitatory amino acid release, glutamate receptor activation, influx of calcium into neurons, and subsequent neuronal degeneration and cell death, may underlie the development and progression of epilepsy. However, while reports of neuronal damage or loss have been noted in patients with epilepsy for many years (Mauritzen Dam, 1982; Babb et al., 1984; De Lanerolle et al., 1989), it has been difficult to determine in human subjects whether neuronal loss is a direct result of status epilepticus or results from other systemic factors, such as preexisting brain pathology. Differences in the phenotypic expression of epilepsy disorders can occur between individuals or within a single pedigree, suggesting that more than one gene may predispose an individual to epilepsy and likely in different etiologic combinations (reviewed in Bate and Gardiner, 1999; Gardiner and Lehesjoki, 2000). Regardless, the search for genes influencing traits and disorders with a complex genetic background,
140 such as epilepsy, has become a realistic task based on recent advances in molecular technology using newly developed animal models. As a result, animal modeling of seizure disorders has been instrumental in elucidating the pathophysiological mechanisms underlying epilepsy and in designing more effective therapies for seizure-induced damage. Importantly, both spontaneous and genetically engineered animal models of inherited epilepsy have been shown to mimic closely the biological phenomena associated with the seizure conditions, such as neurologic abnormalities, predisposition to hyperexcitability, and behavioral traits associated with increased seizure activity (Frankel et al., 1994; reviewed in McNamara and Puranam, 1998; reviewed in Prasad et al., 1999). While the abnormalities in these animal models may be the result of a specific single gene effect or may be the result of multiple genetic interactions, genetic studies using these animal models offer promising alternative approaches to localize and identify genes involved with seizure susceptibility and/or seizure-induced cell death. Similar to the human situation, phenotypic variability is also seen in mice when spontaneous or induced mutations have been placed on different strain backgrounds. Thus, identification of the modifying genes in mice that confer resistance to seizure-induced cell death could be homologous to some of the multiple loci involved in resistance to epilepsy in human. In this review, we will discuss the use of normal inbred mice as a model system to investigate intrastrain (background) variation in susceptibility to seizureinduced cell death, the effects of genetic background strain on the phenotypes of mice lacking specific genes, and the means by which to overcome potential genetic background effects in experimental models of epilepsy. Strain-related differences in seizure-induced cell death
Over the last ten years, substantial evidence has indicated that laboratory mouse strains differ markedly with regard to behavioral studies of complex learning (Diana et al., 1994; Andrews et al., 1995; reviewed in Crawley et al., 1997; Logue et al., 1997; Gerlai, 1998), tumor susceptibility (Malkinson and Beer, 1983; Jacoby et al., 1994; reviewed in Lee, 1998),
TABLE 1 Inbred mouse strains displaying differential vulnerability to kainic acid-induced neurotoxicity Resistant mouse strains
Susceptible mouse strains
BALB/c C3H C57BL/6 ICR 129/SvJ SJL
DBA/2J FVB/N 129/SvEMS
and sensitivity to toxins (Seale et al., 1984; Zocchi et al., 1998; reviewed in Crabbe et al,, 1999). A number of studies have determined that genetic differences can also affect seizure susceptibility in mice. Ferraro et al. (1995, 1997, 1998, 1999) have shown that strain-dependent differences in seizure susceptibility to both electroconvulsive and chemically induced seizures are conferred in a polygenic manner, suggesting that sensitivity to seizures is a multifactorial trait. While it is well known that different strains of rat or mice vary in their response to chemically induced seizures, (Sanberg et al., 1979; Ferraro et al., 1995, 1997; Golden et al., 1995), our studies have also demonstrated that genetic factors in mice influence the degree of sensitivity or resistance to kainic acid-induced cell death (Schauwecker and Steward, 1997). Studies in our laboratory have found that commonly used inbred strains of mice demonstrate dramatic differences in susceptibility to kainic acid-induced excitotoxic cell death. One remarkable finding is that certain inbred strains of mice exhibit seizures following kainic acid (KA) administration, but exhibit little, if any, excitotoxic cell death (Schauwecker and Steward, 1997; Schauwecker, 2000; Schauwecker et al., 2000). Other strains of mice exhibit similar seizures and a pattern of excitotoxic cell death similar to what has been described in rats (Table 1). Resistant strains (C57BL/6, BALB/c, 129/SvJ, C3H, ICR, and SJL) show no degeneration or cell damage until doses at or exceeding the LD90 are administered. Even at these high doses, degeneration and cell death is only evident in a small sector of the hippocampus (area CA3b) in a minority of the mice (Schauwecker and Steward, 1997). Those
141
C57BI.J6
FVBIN
Nissl stain
Fink-Heimer silver stain
Gallyas silver stain Fig. l. Neuronal cell loss and degeneration 7 days following systemic kainate administration in a representative 'cell death-resistant' and representative 'cell death-susceptible' strain of mice. Note that in the 'susceptible' strain, extensive cell loss is evident within the dentate hilar region, as well as in area CA3. Less extensive damage is observed in area CA1. In contrast, cell loss is not evident in the 'resistant' strain. Numerous cells positive for silver impregnation are observed through the dentate hilus and in area CA3 in the 'susceptible' strain, while no degenerative debris is present in the 'resistant' strain.
mice vulnerable to K A - i n d u c e d cell death ( F V B / N , 129/SvEMS, and D B A / 2 J ) show loss of hippocampal pyramidal neurons and neurons in the hilus of the dentate gyms, similar to what has been previously described in rats (Fig. 1; Nadler and Cuthbertson, 1980; Nadler et al., 1980; Sperk et al., 1983; BenAri, 1985). The virtual invulnerability cannot be explained by decreased alterations in the extent of seizure activity because representative 'cell death-resistant' and representative 'cell death-susceptible' mice exhibit the same classes of behavioral seizures, a qualitatively
similar level of seizure intensity, and a similar duration of severe seizures. Moreover, the pattern and extent of neuronal activity is comparable as revealed by 2-deoxyglucose autoradiography (Schauwecker and Steward, 1997). Thus, differences in cell death do not appear to be the result o f differences in b l o o d - b r a i n barrier permeability or differences in the pattern of neuronal activity during the seizures. In addition, it is unlikely that strain differences in the response to K A - i n d u c e d cell death result from differences in the pharmacokinetics of K A as previous studies have shown no difference in the uptake of 3H-KA into the
142 murine central nervous system (Ferraro et al., 1995) in strains that have been identified as 'resistant' or 'susceptible'. Rather, inter-strain differences in excitatory amino acid receptor function (Lipartiti et al., 1993; Magnusson and Cotman, 1993; Kelly et al., 1998), and/or the presence of genes that either mediate neuroprotection or predispose to excitotoxic damage may govern the level of injury. In order to examine the contribution of background genetic factors that may be linked to the differential response of inbred strains to KA-induced cell death, we examined the susceptibility of F1 hybrid mice (129/SvJ × 129/SvEMS) to KA-induced cell death. We chose the 129/Sv strain as it was considered a relatively 'new' strain and since we had observed clear-cut phenotypic differences within the different lines of the 129 strain (e.g. 129/SvJ = resistant; 129/SvEMS -= susceptible), we assumed that any genetic differences would be easier to differentiate within the same inbred strain versus between different inbred strains. By crossing strains that are resistant or susceptible to KA-induced cell death, valuable information about the dominant or recessive nature of resistance can be determined. Nearly all of the mice generated from this cross (99%) were resistant to excitotoxic cell death, suggesting that resistance to excitotoxic cell death is a dominant trait. However, since previous studies have suggested that the 129/SvJ strains may be contaminated with genomic material from another inbred strain (C57BL/6), and thus may not be considered an 'inbred' strain (Simpson et al., 1997; Threadgill et al., 1997), we chose to examine the genetic basis of resistance in intercrosses between other inbred strains. As a thorough genetic analysis of resistance to excitotoxic cell death requires the identification of any elements that may positively or negatively control cell death, we chose to use strains that display large phenotypic (susceptibility or resistance to excitotoxic cell death) and genotypic differences. Matings, performed between the inbred C57BL/6 (resistant) and FVB/N (susceptible) strains, produced 73 FI mice in which we determined the susceptibility of heterozygous animals to KA-induced cell death. Nearly all of the F1 mice were resistant to cell death following KA administration (89%) confirming our previous findings with the 129/SvJ strain. These results suggest
that resistance to KA-induced cell death involves one or more genes with a dominant effect. The finding that many commonly used inbred mouse strains and several Fls are resistant to KAinduced excitotoxicity has important implications for studies using these strains to examine neuronal death. It may be that these strains, and any others with similar resistance, will show very different responses in any situation in which excitotoxic cell death plays a role (for example, after brain or spinal cord injury, transient ischemia, or epileptogenesis). Furthermore, the relative resistance of many mouse strains to seizure-induced cell death also has implications for those studies using gene targeting approaches.
Modeling experimental epilepsy using genetically manipulated mice To elucidate the molecular mechanisms responsible for the production of seizure-induced cell death, previous studies have utilized genetic approaches, including: (1) creation of transgenic mice in which an exogenous gene is introduced into the mouse genome; and (2) gene targeting in which an endogenous gene is removed. The goal of experiments using a genetic manipulation approach is to enable the assignment of functions to single genes and determine the role of a particular gene within a specified system. While it is clear that transgenic and knockout strategies offer the opportunity to study the effects of specific genes thought to be candidates for modulating seizure and cell death susceptibility (Dawson et al., 1996; Watanabe et al., 1996; Morrison et al., 1996; Tsirka et al., 1997; Yang et al., 1997; Jiang et al., 2000; Mazarati et al., 2000), several caveats exist with regard to the design of these experiments. The genetic background of mice may influence the transgenic or knockout phenotype as unlinked genes contained in the strain background (strain-specific modifiers) can have a dramatic effect on the expected phenotype. Modifier genes can greatly affect the manifestation of a mutant phenotype. Secondly, with regard to gene targeting studies, if the genetic background in which targeted gene deletions are constructed is mixed (i.e. comprised of two different strains), phenotypic differences observed may either be the result of the null mutation or genetic link-
143 age from background (modifier) genes (reviewed in Gerlai, 1996; Choi, 1997; Schauwecker and Steward, 1997; Frankel, 1998). The effects of these confounds on interpretation of experimental epilepsy models will be discussed below.
Phenotypic changes resulting from strain-specific modifier genes A major confound in assessment of a knockout phenotype is the issue of genetic background. Genetic background can be defined as the collection of all genes present within an organism that can influence a particular trait or many traits. Thus, many mutations are likely to have different consequences in different genetic backgrounds. It has been well established that even in monogenic diseases, caused by mutations in a single gene, marked variations in the symptoms of patients with the same disease can exist. Variations in the phenotype of a disease are thought to be the result of modifying effects of other genes. Studies performed over the last few years have clearly illustrated that phenotypes caused by specific genetic modifications are strongly influenced by genes unlinked to the target locus. For example, whereas deletion of the p53 tumor suppressor gene causes a dramatic increase in the frequency of tumor formation in those mice compared with wild-type mice, the types of tumor formed, their numbers per animal, and age of tumor onset vary in different genetic backgrounds (Van Meyel et al., 1998; Kuperwasser et al., 2000). In addition, while mutations at different loci can produce a similar phenotype, allelic variation at the same locus can produce different phenotypes (Prasad et al., 1999). Previous studies have shown that wild-type genes can modify the progression of phenotypic traits in either transgenic or knockout mice in a strain-dependent manner (reviewed in Gingrich and Hen, 2000). Furthermore, silencing of a targeted gene using traditional gene knockout approaches can also result in the organism attempting to compensate for the alteration. These compensatory changes can then result in unexpected phenotypic changes (reviewed in Crabbe et al., 1999; and Clark, 2000). The effects of strain on transgenic and knockout mice have been reported extensively (Threadgill et al., 1995; Schauwecker and Steward, 1997; Kelly
et al., 1998; Paradee et al., 1999). These and other studies have demonstrated that a large number of phenotypes observed in transgenic and gene-targeted animals can be influenced by genetic background including ethanol tolerance, locomotor activity, behavior, and seizure susceptibility (Diana et al., 1994; Ferraro et al., 1995, 1998; Crabbe, 1996; Crawley et al., 1997; Logue et al., 1997; Gerlai, 1998; Kelly et al., 1998; Zocchi et al., 1998; Crabbe et al., 1999). As an example of the latter, the susceptibility to pentylenetetrazole (PTZ) seizures in mice differs when the knockout loci are present on C57BL/6, 129/Sv or FVB/N backgrounds (Ferraro et al., 1998). While a number of recent gene targeting studies have reported effects of null mutations on processes that are triggered by or related to glutamate-mediated excitotoxic cell death, the background genotype of the null mutant mice, and its potential confounding effects have been virtually ignored. As a result, it remains unclear whether the phenotypical changes observed in the mutant animals result from the targeted mutation or in fact result from the effects of other genes whose alleles are also different between the mutant and control mice. One example of such background gene effects comes from a study in p53 tumor suppressor knockout mice. It has been reported that deletion of the p53 gene in animals of a mixed genetic background (129/Sv x C57BL/6) conferred protection against KA-induced degeneration (Morrison et al., 1996). However, while these results suggested that p53 may play a critical role in enhancing neuronal viability following KA-induced seizures, our results documenting strain differences in susceptibility to excitotoxic cell death suggest that considerable genetic variability between strains hosting the mutation may profoundly influence the resultant phenotype. In order to examine this issue further, we obtained p53 - / - mice that had been constructed on three different genetic backgrounds from Jackson Laboratories (Bar Harbor, ME). On a C57BL/6 background, p 5 3 - / - mice were resistant to KA-induced cell death. However, on either a 129/SvEMS or FVB/N background, p 5 3 - / - mice were susceptible to KA-induced cell death (Fig. 2). In essence, the p53-null mutation did not influence KA-induced neuronal death even on those genetic backgrounds susceptible to excitotoxic
144
120
t.L
"6
96
9"//.
,.4
72
N
o
48
"6
/'i
lb..
m .o
24
E
Z
0 Hilus
CA3
CA1
DG
Hippocampal region C57BL/6
~
129/SvEMS ~
FVB/N
Fig. 2. Susceptibility to kainate-induced neuronal loss in p53-/- mice is dependent on the genetic background. Quantification of hippocampal cell loss following systemic kainate administration in p53 /- mice on three different genetic backgrounds (C57BL/6, 129/SvEMS, or FVB/N). Note that cell loss is observed throughout area CA3, area CAI, and the dentate hilus when the p53-/- is generated in a 129/SvEMS or FVB/N genetic background. No cell loss is observed when the p53-/- is generated in a C57BL/6 background. Data represent the mean -t- SEM of 5-6 mice for each backgroundstrain. *P < 0.05. cell death. Thus, the difference in susceptibility to KA-induced cell death was attributable to a difference in the progenitor strains used to establish the F2 background upon which the knockout existed. These results have important implications for previous studies assessing the potential neuroprotective effects of particular genes on seizure-induced cell death. As the majority of these studies have been performed on undefined genetic backgrounds, and to a greater extent on mixed backgrounds, the phenotypical differences observed between mutant and wild-type mice may be due either to the introduced null mutation or to the background genes linked to the targeted locus.
Use of hybrid strains for gene targeting Another potential confound for gene targeting experiments is the use of mixed genetic backgrounds, such that loci from one strain may co-segregate with the mutated gene when crossed to a different strain. Unfortunately, the large majority of knockout mutants are often hybrids of two mouse strains (typically, C57BL/6 × 129), While most studies compare homozygous mutant, heterozygous mutant,
and wild-type littermates of an F2 population to determine phenotypical changes resulting from the null mutation, it is important to note that the genetic background composition of these hybrid mice varies among littermates because of gene segregation from the hybrid (F1) parents. In particular, the alleles of genes that surround the targeted locus in a mixed background can be derived from one strain in null mutant mice and of another strain in wild-type mice. Thus, the mutation may actually be viewed as a marker for background genes that are linked to the locus of interest. Many recent studies have examined genes involved in susceptibility to excitotoxin-induced cell death using gene targeting techniques in hybrid mice. For example, null mutations of nitric oxide synthase (Ayata et al., 1997), c-fos (Watanabe et al., 1996), tissue plasminogen activator (Tsirka et al., 1997), poly (ADP-ribose) polymerase (Eliasson et al., 1997), prion protein (Coiling et al., 1997), presenilin-1 (Gut et al., 1999), and glutathione peroxidase (Jiang et al., 2000) in 129Sv x C57BL/6 hybrids have been reported to influence excitotoxic cell death. Although the mechanism of elicitation of excitotoxicity differs from the paradigm used in our
145 studies, our data suggests that the use of the hybrids with C57BL/6 genes might influence the resultant phenotype. Solutions to selecting or creating an appropriate mouse model of experimental epilepsy In order to identify genes that may modulate susceptibility to seizure activity or seizure-induced damage, it is important to keep in mind the complexity underlying genetic studies. Thus, while gene targeting techniques offer the opportunity to study the effect of specific genes that may be candidates for modulating seizure activity or the extent of seizure-induced cell death, careful attention must be paid to the interpretation of observed phenotypes when using gene targeting approaches. This review has revealed that genetic background can confound experiments designed to assess the effect of neuroprotective measures against seizure onset or seizure-induced cell death. Based on the accumulated literature on the genetic dissection of complex traits, following are some suggestions as to how to cope best with the genetic background problem. While it would be most ideal to maintain a mutation on a pure inbred genetic background, our results demonstrating strain-dependent differences in susceptibility to seizure-induced cell death suggest that the correct selection of the most suitable background parental strain requires a thorough understanding of which background genes might influence the resultant phenotype. Thus, the ideal situation would be to generate mutant mice with a pure genetic background that display a meaningful phenotype. By assessing mice that differ only in the presence or absence of a targeted locus, the power of identifying a 'candidate' gene responsible for conferring susceptibility or resistance can be substantially increased. If the use of an inbred strain is not possible, then congenic mice should be generated. Congenic strains are created by transferring a short segment of the chromosome around a marker gene from one strain into an inbred genetic background by repeated backcrossing and selection (Weft et al., 1997). This approach assumes that the resultant strain pair retains a phenotypic difference and then subsequent crosses are made so that the trait locus is the only one segregating. Thus, similar to inbred mouse strains,
congenic strains are homozygous at the majority of loci, eliminating the variability that may confound the resultant phenotype. Secondly, when targeted gene deletions are constructed on a mixed background, the genetic background composition of the appropriate control mice also varies among littermates. Thus, these control mice can only provide an approximate genetic match to mixed backgrounds. While one potential solution involves backcrossing the hybrid mice to a strain of the desired genetic background for 4-5 generations, a small section of one of the genomes will always flank the targeted gene, and like the null mutation, become homozygous following matings between siblings to generate null mice. To check for the possibility of flanking gene effects, one could obtain littermates which are different from the targeted locus, but homozygous for the flanking region or use a polymorphic marker for the targeted genomic region that can permit rapid exclusion of flanking allele effects. However, the easiest solution to this dilemma may be to avoid using null mutants created on a mixed background, in accordance with recommendations from geneticists at the Banbury Conference on Genetic Background in Mice (1997). The optimal approach to determine whether the observed phenotypic effect is the result of a mutated gene is through the use of rescue strategies, in which the functional gene is introduced and studied for its ability to reverse the observed phenotypic effect(s). A number of recent studies have utilized transgenic or lineage-specific rescue of disrupted genes and shown that mutant phenotypes could be rescued (reviewed in Ishida et al., 1998; Lipp and Wolfer, 1998; Schomberg et al., 1999; Aoyama et al., 2000). These studies have not only aided in the development of more sophisticated methods to create genomic alterations, but can assist in defining phenotypes more accurately. Thus future studies will need to focus on dissecting pleiotropic effects of removal of a specific gene by means of rescue experiments, or through the use of other approaches including chimera and mosaic studies, creation of dominant negative mutants, or conditional (tissue-specific) gene knockout techniques.
146
Conclusions The availability of numerous inbred strains represents a useful tool for the genetic dissection of the basis of differential susceptibility to excitatory amino acid-induced cell death. As is clear from the present discussion of genetic susceptibility to KA-induced cell death, endogenous genes present in certain strains of inbred mice can determine resistance. Although the gene(s) conferring resistance to excitotoxin-induced cell death have not been positionally cloned, analysis of phenotypic effects can provide additional information regarding the mechanism of seizure-induced cell death and assist in understanding the extent and nature of the genetic complexity that underlies seizure disorders.
References Anderman, E. (1982) Multifactorial inheritance of generalized and focal epilepsy. In: V.E. Anderson, W.A. Hauser, J.K. Penry and C.E Sing (Eds.), Genetic Basis of the Epilepsies. Raven Press, New York, pp. 355-374. Andrews, J.S., Jansen, J.H.M., Linders, S., Princen, A. and Broekkamp, C.L.E. (1995) Performance of four different rat strains in the autoshaping, two-object discrimination, and swim maze tests of learning and memory. Physiol. Behav., 57: 785-790. Aoyama, S., Kase, H. and Borrelli, E. (2000) Rescue of locomotor impairment in dopamine D2 receptor-deficient mice by an adenosine A2A receptor antagonist. J. Neurosci., 20: 58485852. Ayata, C., Ayata, G., Hara, H., Matthews, R.T., Beal, M.E, Ferrante, R.J., Endres, M., Kim, A., Christie, R.H., Waeber, C., Huang, EL., Hyman, B.T. and Moskowitz, M.A. (1997) Mechanisms of reduced striatal NMDA excitotoxicity in type I nitric oxide synthase knock-out mice. J. Neurosci., 17: 89068917. Babb, T.L., Brown, W.J., Pretorius, J., Davenport, C., Lieb, J.R and Crandall, RH. (1984) Temporal lobe volumetric cell densities in temporal lobe epilepsy. Epilepsia, 25: 729-740. Banbury Conference on Genetic Background in Mice (1997) Neuron, 19: 755-759. Bate, L. and Gardiner, M. (1999) Genetics of inherited epilepsies. Epileptic Disord., 1: 7-19. Ben-Ari, Y. (1985) Limbic seizure and brain damage produced by kainic acid: mechanisms and relevance to human temporal lobe epilepsy. Neuroscience, 14: 375-403. Choi, D.W. (1997) Background genes: out of sight, but not out of brain . . . . Trends Neurosci., 20: 499-500. Clark, A.G. (2000) Limits to prediction of phenotype from knowledge of genotypes. In: M.T. Clegg, M. Hecht and R.J. Maclntrye (Eds.), Limits to Knowledge in Evolutionary Genetics. Kluwer Academic/Plenum, New York, pp. 205-224.
Coiling, S.B., Khana, M., Collinge, J. and Jefferys, J.G.R. (1997) Mossy fibre reorganization in the hippocampus of prion protein null mice. Brain Res., 755: 28-35. Crabbe, J.C. (1996) A genetic animal model of alcohol withdrawal. Alcohol Clin. Exp. Res., 20: 96A-100A. Crabbe, J.C., Phillips, T.J., Buck, K.J., Cunningham, C.L. and Belknap, J.K. (1999) Identifying genes for alcohol and drug sensitivity: recent progress and future directions. Trends Neurosci., 22: 173-179. Crawley, J.N., Belknap, J.K., Collins, A., Crabbe, J.C., Frankel, W., Henderson, N., Hitzemann, R.J., Maxon, S.C., Miner, L.L., Silva, A.J., Wehner, J.M., Wynshaw-Boris, A. and Paylor, R. (1997) Behavioral phenotypes of inbred mouse strains: implications and recommendations for molecular studies. Psychopharmacology, 132: 107-124. Dawson, V.L., Kizushi, V.M., Huang, EL., Snyder, S.H. and Dawson, T.M. (1996) Resistance to neurotoxicity in cortical cultures from neuronal nitric oxide synthase-deficient mice. J. Neumsci., 16: 2479-2487. De Lanerolle, N.C., Kim, J.H., Robbins, R.J. and Spencer, D.D. (1989) Hippocampal interneuron loss and plasticity in human temporal lobe epilepsy. Brain Res., 495: 387-395. Diana, G., Domenici, M.R., Loizzo, A., Scott de Carolis, A. and Saratella, S. (1994) Age and strain differences in rat place learning and hippocampal dentate gyrus frequencypotentiation. Neurosci. Lett., 171: 113-116. Eliasson, M.J., Sampei, K., Mandir, A.S., Hum, P.D., Traystman, R.J., Bao, J., Pieper, A., Wang, Z.Q., Dawson, T.M., Snyder, S.H. and Dawson, V.L. (1997) Poly (ADP-ribose) polymerase gene disruption renders mice resistant to cerebral ischemia. Nat. Med., 3: 1089-1095. Elmslie, E and Gardiner, M. (1995) Genetics of the epilepsies. Curr. Opin. NeuroL, 8: 126-129. Ferraro, T.N., Golden, G.T., Smith, G.G. and Berrettini, W.H. (1995) Differential susceptibility to seizures induced by systemic kainic acid treatment in mature DBA/2J and C57BL/6J mice. Epilepsia, 36:301-307. Ferraro, T.N., Golden, G.T., Smith, G.G., Schork, N.J., St. Jean, P., Ballas, C., Choi, H. and Berrettini, W.H. (1997) Mapping murine loci for seizure response to kainic acid. Mamm. Genome, 8: 200-208. Ferraro, T.N., Golden, G.T., Snyder, R., Laibinis, M., Smith, G.G., Buono, R.J. and Berrettini, W.H. (1998) Genetic influences on electrical seizure threshold. Brain Res., 813: 207210. Ferraro, T.N., Golden, G.T., Smith, G.G., St. Jean, P., Schork, N.J., Mulholland, N., Ballas, C., Schill, J., Buono, R.J. and Berrettini, W.H. (1999) Mapping loci for pentylenetetrazolinduced seizure susceptibility in mice. J. Neurosci., 19: 67336739. Frankel, W.N. (1998) Mouse strain backgrounds: more than black and white. Neuron, 20: 183. Frankel, W.N., Taylor, B.A., Noebels, J.L. and Lutz, C.M. (1994) Genetic epilepsy model derived from common inbred mouse strains. Genetics, 138: 481-489. Gardiner, M. and Lehesjoki, A.E. (2000) Genetics of the epilepsies. Curr. Opin. NeuroL, 13: 157-164.
147
Gerlai, R. (1996) Molecular genetic analysis of mammalian behavior and brain processes: caveats and perspectives. Sem. Neurosci., 8: 153-161. Gerlai, R. (1998) Contextual learning and cue association in fear conditioning in mice: a strain comparison and a lesion study. Behav. Brain Res., 95: 191-203. Gingrich, J.A. and Hen, R. (2000) The broken mouse: the role of development, plasticity and environment in the interpretation of phenotypic changes in knockout mice. Curr. Opin. Neurobiol., 10: 146-152. Golden, G.T., Smith, G.G., Ferraro, T.N. and Reyes, P.E (1995) Rat strain and age differences in kainic acid induced seizures. Epil. Res., 20: 151-159. Guo, Q., Fu, W., Sopher, B.L., Miller, M.W., Ware, C.B., Martin, G.M. and Mattson, M.P. (1999) Increased vulnerability of hippocampal neurons to excitotoxic necrosis in presenilin-1 mutant knock-in mice. Nat. Med., 5: 101-106. lshida, J., Sugiyama, E, Tanimoto, K., Taniguchi, K., Syouji, M., Takimoto, E., Horiguchi, H., Murakami, K., Yagami, K. and Fukamizu, A. (1998) Rescue of angiotensinogen-knockout mice. Biochem. Biophys. Res. Commun., 252: 610-616. Jacoby, R.F., Hohman, C., Marshall, D.J., Frick, T.J., Schlack, S., Broda, M., Smutko, J. and Elliott, R.W. (1994) Genetic analysis of colon cancer susceptibility in mice. Genomics, 15: 381-387. Jiang, D., Akopian, G., Ho, Y.-S., Walsh, J.P. and Andersen, J.K. (2000) Chronic brain oxidation in a glutathione peroxidase knockout mouse model results in increased resistance to induced epileptic seizures. Exp. Neurol., 164: 257-268. Kelly, A., Conroy, S. and Lynch, M.A. (1998) Evidence that nerve growth factor plays a role in long-term potentiation in the rat dentate gyrus. Neuropharmacology, 37: 561-570. Kuperwasser, C., Hurlbut, G.D., Kittrell, F.S., Dickinson, E.S., Laucirica, R., Medina, D., Naber, S.P. and Jerry, D.J. (2000) Development of spontaneous mammary tumors in BALB/c p53 heterozygous mice. A model for Li-Fraumeni syndrome. Am. J. Pathol., 157: 2151-2159. Lee, G.-H. (1998) Genetic dissection of murine susceptibilities to liver and lung tumors based on the two-stage concept of carcinogenesis. Path. Int., 48: 925-933. Lipartiti, M., Fadda, E., Savoini, G., Siliprandi, R., Sautter, J., Arban, R. and Maney, H. (1993) In rats, the metabotropic glutamate receptor-triggered hippocampal neuronal damage is strain-dependent. Life Sci., 52: PL85-PL90. Lipp, H.P. and Wolfer, D.P. (1998) Genetically modified mice and cognition. Curt Opin. NeurobioL, 8: 272-280. Logue, S.F., Owen, E.H., Rasmussen, D.L. and Wehner, J.M. (1997) Assessment of locomotor activity, acoustic and tactile startle, and prepulse inhibition of startle in inbred mouse strains and F/ hybrids: implications of genetic background for single gene and quantitative trait loci analyses. Neuroscience, 80: 1075-1086. Magnusson, K.R. and Cotman, C.W. (1993) Age-related changes in excitatory amino acid receptors in two mouse strains. Neurobiol. Aging, 14: 197-206. Malkinson, A.M. and Beer, D.S. (1983) Major effect on suscepti-
bility to urethan-induced pulmonary adenoma by a single gene in BALB/cBy mice. J. Natl. Cancer Inst., 70: 931-936. Mauritzen Dam, A. (1982) Hippocampal neuron loss in epilepsy and after experimental seizures. Acta Neurol. Scand., 66: 589606. Mazarati, A.M., Hohmann, J.G., Bacon, A,, Liu, H., Sankar, R., Steiner, R.A., Wynick, D. and Wasterlain, C.G. (2000) Modulation of hippocampal excitability and seizures by galanin. J. Neurosci., 20: 6276-6281. McNamara, J.O. and Puranam, R.S. (1998) Epilepsy genetics: an abundance of riches for biologists. Curr. Biol., 8: 168-170. Morrison, R.S., Wenzel, H.J., Kinoshita, Y., Robbins, C.A., Donehower, L.A. and Schwartzkroin, P.A. (1996) Loss of the p53 tumor suppressor gene protects neurons from kainateinduced cell death. Z Neurosci., 16: 1337-1345. Nadler, LV. and Cuthbertson, G.J. (1980) Kainic acid neurotoxicity toward hippocampal formation: dependence on specific excitatory pathways. Brain Res., 195: 47-56. Nadler, J.V., Perry, B.W., Gentry, C. and Cotman, C.W. (1980) Loss and reacquisition of hippocampal synapses after selective destruction of CA3-CA4 afferents with kainic acid. Brain Res., 191: 387-403. Paradee, W., Melikian, H.E., Rasmussen, D.L., Kenneson, A., Corm, P.J. and Warren, S.T. (1999) Fragile X mouse: strain effects of knockout phenotype and evidence suggesting deficient amygdala function. Neuroscience, 94: 185-192. Prasad, A.N., Prasad, C. and Stafstrom, C.E. (1999) Recent advances in the genetics of epilepsy: insights from human and animal studies. Epilepsia, 40: 1329-1352. Sanberg, P.R., Pisa, M. and McGeer, E.G. (1979) Strain differences and kainic acid neurotoxicity. Brain Res., 166: 431435. Schauwecker, P.E. (2000) Seizure-induced neuronal death is associated with induction of c-Jun N-terminal kinase and is dependent on genetic background. Brain Res., 884: 116-128. Schauwecker, P.E. and Steward, O. (1997) Genetic determinants of susceptibility to excitotoxic cell death: implications for gene targeting approaches. Proc. Natl. Acad. Sci. USA, 94: 4103-4108. Schauwecker, P.E., Ramirez, J.J. and Steward, O. (2000) Genetic dissection of the signals that induce synaptic reorganization. Exp. Neurol., 161: 139-152. Schomberg, D.W., Couse, J.F., Mukherjee, A., Lubahn, D.B., Sar, M., Mayo, K.E. and Korach, K.S. (1999) Targeted disruption of the estrogen receptor-alpha gene in female mice: characterization of ovarian responses and phenotype in the adult. Endocrinology, 140: 2733-2744. Seale, T.W., Johnson, P., Carney, J.M. and Rennert, O.M. (1984) Interstrain variation in acute toxic response to caffeine among inbred mice. Pharmacol. Biochem. Behav., 20: 567-573. Serratosa, J.M. (1999) Idiopathic epilepsies with a complex mode of inheritance. Epilepsia, 40(Suppl. 3): 12-16. Simpson, E.M., Linder, C.C., Sargent, E.E., Davisson, M.T., Mobraaten, L.E. and Sharp, J.J. (1997) Genetic variation among 129 substrains and its importance for targeted mutagenesis in mice. Nat. Genet., 16: 19-27. Sperk, G., Lassmann, H., Baran, H., Kish, S.J., Seitelberger, F.
148 and Horuykiewicz, O. (1983) Kainic acid induced seizures: neurochemical and histopathological changes. Neuroscience, 10: 1301-1315. Steinlein, O.K. (1999) New insights into the molecular and genetic mechanisms underlying idiopathic epilepsies. Clin. Genet., 54: 169-175. Threadgill, D.W., Dlugosz, A.A., Hansen, L.A., Tennenbaum, T., Lichti, U., Yee, D., LaMantia, C., Mourton, T., Hen'up, K. and Harris, R.C. (1995) Targeted disruption of mouse EGF receptor: effect of genetic background on mutant phenotype. Science, 269: 230-234. Threadgill, D.W., Yee, D., Matin, A., Nadeau~ J.H. and Magnuson, T. (1997) Genealogy of the 129 inbred strains: 129/SvJ is a contaminated inbred strain. Mamm. Genome, 8: 390-393. Tsirka, S.E., Rogove, A.D., Bugge, T.H., Degen, J.L. and Strickland, S. (1997) An extracellular proteolytic cascade promotes neuronal degeneration in the mouse hippocampus. J. Neurosci., 17: 543-552. Van Meyel, D.J., Sanchez-Sweatman, O.H., Kerkvliet, N., Stitt, L., Ramsay, D.A., Khokha, R., Chambers, A.F. and Cairncross,
J.G. (1998) Genetic background influences timing, morphology and dissemination of lymphomas in p53-deficient mice. Int. J. Oncol,, 13: 917-922. Watanabe, Y., Johnson, R.S., Butler, L.S., Binder, D.K., Spiegelman, B.M., Papaioannou, V.E. and McNamara, J.O. (1996) Null mutation of c-fos impairs structural and functional plasticities in the kindling model of epilepsy. J. Neurosci., 16: 3827-3836. Weil, M.M., Brown, B.W. and Serachitopol, D.M. (1997) Genotype selection to rapidly breed congenic strains. Genetics, 146: 1061-1069. Yang, D.D., Kuan, C.-Y., Whitmarsh, A.J., Rincon, M., Zheng, T.S., Davis, R.J., Rakic, P. and Flavell, R.A. (1997) Absence of excitotoxicity-induced apoptosis in the hippocampus of mice lacking the Jnk3 gene. Nature, 389: 865-870. Zocchi, A., Orsini, C., Cabib, S. and Puglisi-Allegra, S. (1998) Parallel strain-dependent effect of amphetamine on locomotor activity and dopamine release in the nucleus accumbens: an in vivo study in mice. Neuroseience, 82:521-528.
T. Sutula and A. Pitk~inen (Eds.) Progress in Brain Research, Vol. 135 © 2002 Elsevier Science B,V. All rights reserved
CHAPTER 13
Genomics and neurological phenotypes" applications for seizure-induced damage Jo A. Del Rio and Carrolee Barlow * The Salk Institute for Biological Studies, The Laboratory of Genetics, 10010 North Torrey Pines Road, La Jolla, CA 92037, USA
Abstract: It is sometimes assumed that because the brain is such a complex organ, experimental genomics methods are not directly applicable to neurobiological studies. In fact, it is because the brain and brain process are complex that it is even more important to apply methods that allow large numbers of genes to be monitored across a significant number of experiments. How can we begin to understand the mechanisms underlying various brain functions, and how can we understand what can and does go wrong in disease? How can such tasks be accomplished without being overly costly and time- and labor-intensive? We and others have put DNA microarray technology to work to address a variety of biological problems, and in particular to study the brain and various brain functions. This review provides an overview of how we use DNA microarray technology to identify the genes that are responsible for specific neurological responses, seizure-induced responses, and the unique structures and functions of different brain regions.
Introduction
New experimental methods have made it possible to gain a global view of molecular and cellular events at the level of transcription. Among the most useful and versatile tools developed for molecular and cellular studies are high-density D N A arrays that allow complex mixtures of RNA and DNA to be interrogated in a highly parallel fashion (Lipshutz et al., 1995, 1999; Lockhart et al., 1996; Marshall and Hodgson, 1998; Lockhart, 1999; Lockhart and Barlow, 2001). DNA arrays can be employed for many different purposes, and they have been put to greatest use to measure gene expression levels (messenger RNA abundance) for tens of thousands of genes simultaneously. The goal of these methods is to understand the underlying workings of the cell, and
* Correspondence to: C. Barlow, The Salk Institute for Biological Studies, The Laboratory of Genetics, 10010 North Torrey Pines Road, La Jolla, CA 92037, USA. E-mail: barlow@ salk.edu
how all the various components work together. By combining these technologies with existing genetic and neurobiological experimentation, it is possible to broaden the scope of our questions to discover the genetic determinants that underlie several readily tractable biological phenotypes. We can ask for example, what genes are important for mediating resistance and susceptibility to genetic and environmental insults? W h y are specific brain regions and cell populations more susceptible to damage? In this review, we provide an overview of the use of genomic technologies. We discuss how to apply array-based methods to the study of cells and complex tissue, and describe some special considerations for applying these methods to the study of the brain. Global gene expression e x p e r i m e n t s - - an overview
The collection of genes that are expressed or transcribed from genomic DNA (often referred to as the transcriptome) is a major determinant of cellular phenotype and function. Differences in gene
150 expression are both responsible for morphological and phenotypic differences a's well as indicative of cellular responses to environmental stimuli and perturbations. Unlike the genome, the transcriptome is highly dynamic and changes in response to perturbations. Changes in multi-gene patterns of expression can provide clues about regulatory mechanisms and broader cellular functions.
density arrays of relatively short, specifically chosen, in situ synthesized DNA oligonucleotides on glass (often called DNA microarrays, oligonucleotide arrays, Affymetrix GeneChip arrays, or simply 'chips') and their use for parallel gene expression (mRNA abundance) measurements (Fodor et al., 1991; Lipshutz et al., 1995; Lockhart et al., 1996).
DNA arrays
Oligonucleotide arrays (Affymetrix GeneChip arrays)
Nucleic acid arrays have been used successfully to measure transcript abundance in many different experiments (for a review see Lockhart and Winzeler, 2000). The arrays are passive devices that work by hybridization of labeled RNA or DNA samples to DNA molecules attached at specific locations on a surface. The DNA probes on the surface effectively 'count' the number of molecules of each type by binding to molecules that contain their complementary sequence. DNA arrays are generally produced in one of two basic ways: by deposition of nucleic acids (PCR products, plasmids, or oligonucleotides) onto a glass slide, or by in situ synthesis of oligonucleotides using photolithography (Lipshutz et al., 1999) or by the spatially specific application of reactants. Regardless of how they are made, DNA arrays are simply large collections of oligonucleotides or cDNAs (generally 500-1000 bp of double stranded DNA) at distinct positions on glass, and most of their uses amount to the counting of different molecules. In both cases, surface-bound probes are usually chosen from sequences located nearer the 3'-end of the gene (near the poly-A tail in eukaryotic mRNA), and different probes can be used for different exons to enable the detection of variant splice forms. The monitored genes can be of known or unknown function; all that is needed to design probes for an array is at least a couple hundred bases of sequence information to design either PCR primers to make cDNAs, or from which to choose appropriate complementary oligonucleotides. There are a number of possible variations on the basic experimental approach, but the key elements of parallel hybridization to localized, surface-bound nucleic acid probes and subsequent detection and quantification of bound molecules are ubiquitous. In this review, we will emphasize the use of high-
The Affymetrix GeneChip arrays are photolithographically synthesized arrays where ~ 107 copies of each selected oligonucleotide (usually 25 nucleotides in length) are synthesized base by base in a highly parallel, combinatorial synthesis strategy to make hundreds of thousands of different oligonucleotide probes in distinct regions on a flat glass surface. Typically for these arrays, multiple 16-20 probes (a probe refers to the 25-mer oligonucleotide corresponding to a unique sequence for the gene of interest) per gene are placed on the array (Fig. 1). The oligonucleotide probes for gene expression measurements are designed directly from gene, gene fragment or EST sequence information. Therefore, for each RNA, multiple, different, and specific oligonucleotide probes are chosen that are complementary in sequence to each monitored mRNA. The advantage of having multiple, different oligonucleotide probes (non-overlapping if possible, but minimally overlapping if necessary) is that they serve as independent detectors for the same gene. The use of redundant, but different probes for each mRNA also increases both the qualitative and quantitative accuracy of the results because consistent patterns of hybridization (or hybridization differences) across probes of different sequences for the same gene can be recognized. The average behavior across the entire probe set can be used for quantitation rather than relying on the intensity of only a single detector or 'spot'. Several unique features are incorporated into these types of arrays. The high density and uniformity of probe cells on oligonucleotide arrays permit the building of redundancy as well as a variety of controls into the design. For example, as mentioned above, in order to increase confidence and improve the accuracy of mRNA quantitation, multiple independent probes (typically 25-mers) are used to detect
151
mRNA referencesequence
,
,
,//t'
........
/
~~\Spaced DNA probepairs Reference sequence "" TGTGATGGTGGGAATGGGTCAGAA~ACTCCTATGTGGGTGACGAGGCC"' Z TTACCCAGTCTT}C~3TGAGGATACACCCAC PerfectMatch~lOo l [_TTACCCAGTCTT~JCTGAG GATACACCCAC MismatchOligo Perfect match orobecells
\ Mismatch probeceils Fig. 1. Gene expression probe set layout for oligonucleotide arrays. Multiple oligonucleotide probe sequences near the 3'-end and complementary to the mRNA of interest are chosen. In a position physically adjacent (below in the schematic) to each perfect match (PM) probe, is a probe that has a single base difference in the middle (the mismatch, or MM probe). The MM probes serve as specificity Controls, and allow the discrimination between signals that are due to the specific RNA of interest and those that may be due to cross-hybridization. The PM probes are chosen based on a measure of sequence uniqueness, expected absence of secondary structure, and a set of sequence-based selection rules. The probes are not necessarily chosen from equally spaced regions of the transcript, and they may have some degree of sequence overlap. For quantitation, the PM minus MM signal intensity differences are used because subtracting the MM signals helps reduce contributions due to background and cross-hybridization. The patterns of hybridization (i.e., the consistency of PM signals that are larger than MM signals, as expected from specific hybridization of the RNA for which the probes were designed) are used to make a qualitative assessment of 'Present' or 'Absent' for each probe set (more precisely, 'detectable' and 'not detectable'). The average of the PM-MM values (referred to as the 'average difference'), after discarding outliers, is used to make a quantitative assessment of RNA abundance. In some newer designs, the multiple PM/MM pairs for each gene are not located next to each other, but are scattered on the array to minimize effects of spatial variation.
each expressed s e q u e n c e (see Fig. 1). In addition, each probe is d e s i g n e d to be a 'perfect m a t c h ' (PM) to a specific region of the expressed target sequence. In addition, each P M probe is partnered with a single m i s m a t c h ( M M ) to the target sequence. The M M probe is synthesized i m m e d i a t e l y adjacent to and b e l o w its c o r r e s p o n d i n g P M probe, f o r m i n g a series of P M / M M probe pairs that m a k e up a probe set (Fig. 1) for each gene or EST. T h e n u m b e r of probe pairs in each probe set is typically 1 4 - 2 0 , d e p e n d i n g on the type of array. The M M probes in each set provide a m e a s u r e for non-specific hybridization, a way to d e t e r m i n e if the h y b r i d i z a t i o n signal is truly due to the i n t e n d e d m e s s e n g e r R N A , and is also a conve-
n i e n t way to subtract out the potentially c o n f o u n d i n g contributions of cross-hybridization and non-specific b a c k g r o u n d signals. Therefore, the M M probes serve as internal controls for h y b r i d i z a t i o n specificity, a n d e n a b l e the effective subtraction of local b a c k g r o u n d a n d cross-hybridization signals. In addition to the sets of probes for each gene or E S T (referred to as the probe set), there are a host of control probes that are used for grid a l i g n m e n t , spatial n o r m a l i z a t i o n , array identification, and for assessments of R N A , array and data quality, and overall detection sensitivity and specificity.
152
Procedural overview
section can be readily minimized (Sandberg et al., 2000; Lockhart and Barlow, 2001). It is important that animals be handled in a systematic and consistent manner prior to obtaining tissue. All animals are singly housed for 7 days prior to sacrifice. All euthanasia is performed using cervical dislocation. We also go so far as to perform all dissections at specified hours of the day. Dissections are carried out on petri dishes filled with wet ice. Samples are dissected and immediately frozen in dry ice and stored at - 8 0 ° C until RNA is extracted. To prepare total RNA from the frozen tissue, TRlzol (GibcoBRL) is added at approximately 1 ml per 100 mg tissue and then homogenized (Polytron, Kinematica) at maximum speed for 2 min. RNA is resuspended in RNase-free water at a concentration of at least 1 mg/ml. The quality of the total RNA is checked on an agarose gel (to check the distribution of RNA lengths) and by an absorption measurement of the RNA in TE and H20.
The basic steps, prior to hybridization, for performing an array-based expression measurement are similar to those necessary for any mRNA measurement (e.g., northerns, RT-PCR), and involve the handling of animals, tissues, cells and RNA. The exact procedures after total R N A extraction tend to be more array specific (e.g., cDNA arrays versus oligonucleotide GeneChip arrays available from Affymetrix), but the protocols all employ basic molecular biological techniques and reagents. For the Affymetrix arrays, the cellular mRNA is usually amplified (by a factor of 50-200) using a linear in vitro transcription (IVT) reaction. The IVT reaction is run in the presence of labeled ribonucleotides to produce labeled, complementary RNA (cRNA). The single-stranded cRNA is randomly fragmented to an average size of 30-50 bases prior to hybridization to minimize the possible effects of RNA secondary structure and to enhance hybridization specificity. Following hybridization (typically overnight at a temperature of 40-50°C), the sample is recovered from the hybridization cartridge and saved for future use (samples can be rehybridized multiple times) (Lockhart and Barlow, 2001). The arrays are washed to remove weakly bound molecules and to reduce background signals and are then 'read' using a specially designed laser confocal scanner that scans the entire array in only 5-10 rain at a spatial resolution of 3 Ixm. This scan produces the raw data file (the 'image') that is then quantitatively analyzed and interpreted, as discussed below. The raw data file (the '.dat' file) for a 1.28 x 1.28 cm array read at a resolution of 3 Ixm per pixel is approximately 44 MB in size, and the subsequent processed files (the '.cel' and the '.chp' files) are both about 10 MB in size. That means that the basic data from a single experiment requires at least 64 MB of storage. Using a single scanner, data can be collected as often as every 10-15 rain.
In most current implementations of array-based approaches, the RNA or DNA to be hybridized must be labeled prior to the hybridization reaction so that surface-bound molecules can be fluorescently detected and quantitated. Either RNA or DNA can be hybridized to arrays, and different methods can be used to prepare labeled material. We routinely start with 5-10 ~tg of total RNA for each experiment. This generally yields between 60 and 100 txg of labeled cRNA. Approximately 30 Ixg of this cRNA is used per hybridization. We have found that the biotin-phycoerythrin labeling yields approximately 10 times as much signal per bound molecule than when using straight incorporation of fluorescein. With oligonucleotide arrays, each sample is labeled identically and hybridized independently to different arrays. Signal intensities can then be compared directly between any individual array experiments.
Tissue dissection and RNA preparation
Array data analysis
Several methods exist for obtaining high-quality RNA from brain tissue. Based on our experience in studies of the mouse brain, we have found that animal to animal variation, and variation due to dis-
Following a quantitative fluorescence scan of a typical, photolithographically synthesized oligonucleotide array, a grid is aligned to the image using the known dimensions of the array and the corner and
Sample preparation
153 edge controls (laid out in specific patterns on every array) as markers. The individual pixels (typically 50-60 per 24 × 24 Ixm synthesis feature - - newer designs available from Affymetrix use 20 × 20-1xm features) within each region are averaged (or most commonly, the 75th percentile pixel intensity value is used) after systematically eliminating those at the border and discarding outliers. The details of noise calculations for each array image, threshold settings, and the logic of the voting scheme were established based on extensive quantitative spiking and reconstruction experiments, and on an assessment of an acceptable false-positive rate (the false-positive rate for 'present' calls in any single measurement on an array is generally less than 1% of all genes monitored when using standard conditions and default analysis parameters). The default analysis algorithms examine the data in a number of ways to generate the final qualitative and quantitative results. After background subtraction, normalization or scaling of the data is generally carried out to equalize the overall signal intensities across different arrays used in a given set of experiments. Next, a number of different metrics are determined for every probe set. Along with the qualitative assessment of the pattern to make a call of present or absent, there is a quantitative assessment to estimate the RNA concentration or abundance. The determination of quantitative RNA abundance is calculated from the average of the pairwise PM minus MM differences, referred to as the 'average difference', (the quantity shown to be proportional to RNA concentration) across the set of probes for each RNA. For example, each probe pair in a set is checked for specific hybridization performance (PM vs. MM) to insure reliable detection of the sampled regions of the transcript above the noise of the assay. The various metrics are integrated into a call of 'P' for present, 'N for undetected, and 'M' for marginal for the transcript or EST cluster represented by the probe set. Because of the use of multiple, independent probes for each gene or EST, it is possible to use a consistent overall pattern of hybridization to the PM and MM probes to determine if the signal is due to the designated transcript. In effect, the set of probes act as a jury, with each member given a vote in order to make a qualitative assessment. The members of the jury must agree to a reasonable extent (more like a civil trial in which unanimity is not
required) in order to make a call of present. In this way, no single probe in a set has an undue influence, and this makes the approach much more impervious to the occasional outlier or strong and unpredictable cross-hybridization event. When assessing the differences between two different RNA samples (hybridized independently to two different arrays), similar logic and criteria are used, except the primary determinants in this case are the changes in the individual P M - M M values across the probe set. Prior to comparing any two or more measurements, all signal intensities on an array are multiplied by a factor (a linear 'scaling factor' in the simplest case) that makes the mean P M - M M value for any array measurement equal to a preset value. This simple global scaling process is designed to correct for any inter-array differences, or small differences in sample concentration, labeling efficiency or fluorescence detection, and it appears to work rather well when experiments are performed in a consistent fashion (e.g., identical hybridization and washing conditions, and the same amount of labeled material hybridized). In the case of a pairwise comparison of array results, the patterns of change (with consistent 'voting') and the magnitude of the changes are used to make both qualitative calls of 'Increase' or 'Decrease', and quantitative assessments of the absolute size (differences in signal, related to changes in the number of copies per cell) and the relative size (ratio or 'fold change') of any differences. These methods for qualitative and quantitative assessments of mRNA abundance and differential expression are codified in the standard, commercially available Affymetrix GeneChip analysis software. Using this method, messenger RNAs present at one to a few copies (relative abundance of 1 : 300,000) to thousands of copies per mammalian cell can be detected (Lockhart et al., 1996; Wodicka et al., 1997), and changes as subtle as a factor of 1.1 to 2.0, can be reliably detected (although changes of at least a factor of 1.5 are more routinely trustworthy) if data quality is high and replicate experiments are performed. The software integrates the comparisons for every probe set into a call of 'I' for increase, 'D' for decrease, and 'NC' for no change (along with marginal calls when patterns of change are more ambiguous). The expression algorithms also produce quantitative
154 results that reflect transcript abundance and relative changes between compared samples. The quantitative metric, termed average difference, is literally the average of the P M - M M differences across the probes in the set. The average difference for a probe set is calculated as a 'trimmed' mean (e.g. after outlier rejection) of the intensity differences (PM-MM) for each probe pair in the set, and it has the advantage of being automatically background subtracted. The average difference is useful as a measure of expression level because it has a nearly linear relationship with the transcript abundance over a wide dynamic range of more than three orders of magnitude. In addition, an estimate of the relative change, or fold change, of expression levels is calculated based on the ratio of the average difference values between any two experiments (after setting a minimum possible denominator based on the size of the noise to avoid dividing by zero or values that are not significantly above the noise). Because of the richness of the data and the builtin redundancy (at the level of having both multiple pixels per feature and multiple features per gene or EST), there are of course a number of alternative ways in which data of this type could be assessed. These issues are being explored by many groups, with the attainable goal of a significant increase in the information content per array and data quality without an increase in the difficulty, expense or time required for an experiment.
Important considerations To obtain results with the highest confidence, it is necessary to perform experiments in a consistent and careful fashion, and to perform quality control checks at several points during the experiments. It is very important to handle animals, tissue and cells appropriately and to handle total RNA in ways that minimize degradation. To insure that samples are of suitable quality before hybridizing them to arrays, the following procedures are employed: (1) total RNA is run on a gel to check the size distribution relative to rRNA bands and by spectrophotometer to ensure an OD 260/280 ratio of greater than or equal to 2.0. (2) labeled, purified and unfragmented cRNA is run on a gel to check for the correct size
distribution relative to quality standards, and the amount of labeled product is quantitated using a measurement of the absorbance at 260 nm (based on a full absorption spectrum from 220 to 340 nm); (3) following fragmentation, the labeled cRNA is run on a low molecular weight gel to check for a suitable distribution of fragment lengths (typically between 30 and 50 bases). Following hybridization of a sample to an array, collection of an image, and basic image analysis, data 'triage' is performed to make sure that the array data are of sufficient quality for further analysis and comparison with other data sets. For example, the background, noise, overall signal strength, the ratio of the 3'- and Y-signals for actin and GAPDH mRNA (a measure of RNA length and quality - degraded RNA will result in high 3'/5' ratios because only the region of the mRNA near the 3' poly-A tail will be amplified and labeled), and the percentage of genes scored as 'present' should be similar between chips. Typically, we expect to see % Present values that are within 5% of each other, background and Q values (Q is a measure of the minimum background noise across the array image) within a factor of two of each other, scaling factors (SF) within a factor of two, and 3'/5' ratios for both actin and GAPDH of less than 2.0. When experiments meet these standards for sample and data quality, the false-positive rate can be expected to be acceptably low (Lockhart and Barlow, 2001). The extent of change in expression level for any gene is commonly given as the 'fold change'. For example, if the expression level went from 5 to 10 copies per cell, this would be a two-fold change, 5-15 copies per cell, a three-fold change, and so on. Usually we care most about the relative size of a change rather than exactly how many copies of mRNA per cell are found for a given gene, and we generally would not interpret a change from 5 to 10 copies per cell any differently than we would a change from 20 to 40 copies. Another reason the ratio or fold change is used is that with spotted cDNA arrays, the readout is a ratio of two intensities at each 'spot' after a competitive hybridization of two samples labeled with different fluorophores that emit at different wavelengths (i.e., only the ratio is interpreted). But a problem arises no matter which
155 type of array is being used if in one of the two cases, the mRNA is absent or the level is extremely low. For example, if the abundance of a transcript really goes from zero to 10 copies per cell, the fold change is infinite, and the difference between the signals is a more appropriate measure than the ratio and for oligonucleotide arrays the signal difference has been shown to be quantitatively related to the change in mRNA abundance. In cases such as this, the ratio is also likely to be rather variable because it is difficult to know where to set 'zero' and even if the transcript abundance is not strictly zero, the signal is not large relative to the background noise, and cannot be quantified with any confidence. In both of these cases, it is typical to have a minimum allowable value (often set by a measure of the background or the noise in the signals) for the denominator to avoid dividing by zero or an unreasonably small and overly noisy value. When at least one of the two values is too small, an approximate fold change can be given, but it must be remembered that it is an approximation that is completely dependent on the specifics of how the minimum denominator value was set, and that it is likely to be an underestimate of the true value.
The importance of well-controlled, replicate measurements For high-throughput, parallel measurements, data quality is of critical importance if one is attempting to identify with high confidence specific genes that are differentially expressed. The reason is that when monitoring, for example, 10,000 genes, even a low false-positive rate of 1% results in 100 incorrect difference calls, comparable to the number of true changes observed in many types of experiments (a false-positive here is defined as an assignment of a gene as 'differentially expressed' when in fact the mRNA abundance is not significantly changed). We find that when experiments are performed with sufficient care, the source of most of these false-positives (which are in large part the result of setting the lowest possible thresholds in the interest of sensitivity) is random noise, small variations in sample preparation and other experimental steps, and the occasional array-specific physical defect. Because these various factors lead to largely random variations, observations made consistently in independent replicates
can yield a false-positive rate closer to 0.01% (i.e., 1% of 1% ), or only one false call of 'different' ('increased' or 'decreased') for every 10,000 genes monitored (Carter et al., 2001; Lockhart and Barlow, 2001).
Analysis of replicate data To obtain a low false-positive rate, it is important to use multiple criteria for assessing differences. One key to obtaining a low false-positive rate is good, consistent experimental technique while controlling as much as possible all sources of experimental variation (e.g., mouse handling, dissection protocols, tissue handling, RNA extractions, amplification and labeling reactions, hybridization and washing conditions and array usage). For example, in experiments done as independent duplicates using cell lines or different isogenic mice, we typically require that: (1) the probe set score as 'increased' or 'decreased' in 2/2 comparisons, and (2) the fold change be at least 1.8 fold in 2/2 comparison, and (3) the gene scores as clearly 'present' in at least one of the four (2 × 2) data sets, and (4) that the difference in the signal be at least 50 in 2/2 comparisons (in arbitrary units after scaling the overall intensity to a mean of 200 which corresponds to an RNA abundance in a mammalian cell of about 3-5 copies per cell - - so a signal change of 50 corresponds to a change in mRNA abundance of roughly 1-2 copies per cell). These specific thresholds are somewhat arbitrary, but we have found that requiring all of the qualitative and quantitative criteria be met together makes it so each of the individual criteria can be fairly permissive while the overall requirements are quite strict. For example, requiring only a signal change of 50 alone, or a quantitative fold change of at least 1.8 without the other requirements would lead to an increase in the false-positive rate by more than a factor of 10. Again, it is important to be cautious about interpreting a negative result because it is possible for some genes to miss passing the stringent set of criteria for being differentially expressed. It is always possible to reanalyze the data using more permissive criteria to identify additional genes that may have changed, but these should be interpreted with greater caution than those that meet the stricter criteria. Also, it is straightforward to query the data
156 to examine the behavior of any specific gene or any chosen set of genes in which one has a particular interest, apart from whether or not their behavior meets the global selection criteria.
Verification and follow-up of array-based observations Although the array-based expression measurements can be made highly quantitative and reproducible, specific genes that are found to be differentially expressed on arrays should be viewed as highprobability candidates. Based on our experience and that of others, we cannot stress strongly enough the importance of great experimental care, wellcharacterized and rigorous analysis, and the need for appropriate follow-up and verification. When verifying candidates and designing experiments, in almost all cases, experiments should be performed at least in duplicate, with replicates performed as independently as possible (e.g., different mice or independent dissections of a region, independent sample preparations, and independent hybridizations to physically different arrays). It is not sufficient to merely remake samples from the same extracted RNA from the same mouse or tissue sample, or to simply rehybridize samples to additional arrays, as has been done. If genetically identical, inbred mice are not used, then it is necessary to perform additional experiments or to pool mice to effectively average out differences due to genetic inhomogeneity (independently pooled samples should be used as replicates). The same considerations apply when using any other animal or human tissue. We routinely use northern blotting and quantitative RT-PCR for selected sets of genes to verify results and to validate experimental and analytical methods. In these follow-up experiments, it is important to use independently prepared samples and not simply the same RNA that was used for the array experiments. Independent verification is even more critical if untested or less stringent analysis criteria are used, or if extremely subtle expression differences are to be interpreted. In addition, Western blots can be used to measure corresponding protein levels, and immunohistochemistry and in situ hybridization can be very useful to measure cell or region specificity of proteins and mRNAs to both
confirm and extend the results obtained using mRNA measurements on arrays. Finally, global expression measurements should be considered a starting point for the understanding of a biological problem, and as a valuable tool for obtaining information concerning a large number of genes. These methods should be used in the context of other types of measurements, knowledge and information, and it should be understood that findings will often need to be followed up with further experiments of various, more conventional types.
Gene expression profiling in neurobiology Obviously, the brain is a complex and inhomogeneous organ containing a large number of different regions and cell types. This does not mean, however, that the brain is too complex to be studied using these new tools. Instead, what is clear is that extra care must be taken, experiments need to be designed with the unique features of the brain in mind, and that array-based measurements need to be applied in combination with other methods. We and others have used these techniques to study the brain. The next section provides an overview of specific experiments, with an emphasis on appropriate experimental procedures and the potential use of the technology for understanding brain function (Ginsberg et al., 2000; Lee et al., 2000; Mimics et al., 2000; Sandberg et al., 2000; Carter et al., 2001; Lockhart and Barlow, 2001).
Transcriptional response to seizure The molecular response to seizure has been extensively studied. Many of these studies have been designed to test the transcriptional or signal transduction response of a particular gene or small set of genes at various timepoints after a seizure. We used a genomic approach to identify the global changes in gene expression that occur in response to seizure (Sandberg et al., 2000). In this study, C57BL/6J (B6) and 129/SvEvTac (129) male mice at 8 weeks of age were treated with pentylenetetrazol (PTZ) to induce seizure. Two animals of each strain which showed a similar response to seizure were studied along with control animals to determine the tran-
157 scriptional response to seizure in the hippocampus and cerebellum one-hour after seizure induction. Importantly, the experiment successfully detected the induction of several known immediate-early genes including members of the fos and jun family, serum and glucocorticoid-regulated kinase (sgk), growth factor inducible immediate early gene (3CH134), cox-2, and the transcription factors KROX20 and zif/268 verifying that changes are detectable and that they recapitulate data generated in more traditional types of studies. Several questions could be addressed using the data. For example, what are the differences and similarities between the transcriptional response of the hippocampus and the cerebellum? In this analysis, the two similar brain regions from the 129 samples at baseline were c o m p a r e d to the two similar regions from the 129 samples after seizure, and two B6 samples at baseline were compared to two B6 samples after seizure using pair-wise comparisons. The number of genes that were changed in the four pair-wise comparisons for the 129 hippocampus and for the B6 hippocampus were determined. A similar analysis was performed on the cerebellar samples. The number of genes differentially expressed after seizure were determined by summing the genes that were changed in 129 and in B6 using the criteria of a 1.8-fold change or greater, an absolute difference change (ADC) of >50, and a call of I, MI, D or MD in three of the four comparison files for each strain analyzed independently. As shown in Fig. 2, 84 genes were induced and 19 repressed in the hippocampus whereas in the cerebellum, 85 genes were induced and 32 genes repressed. Of those, 35 genes were regulated co-ordinately in both the hippocampus and the cerebellum. Another key feature of the study was the finding of a differential response to seizure between the two strains. In this analysis of the data, all comparisons were included and the standard criteria had to be met in at least six o f the eight files, thereby ensuring that the gene was consistently changed in both strains. The results of this analysis are shown in Fig. 2 as the numbers in parentheses. As shown, the number of genes that changed in both strains was significantly less (a total of 84 induced in one of the two strains versus 54 in both strains and a total of 51 repressed in one of the two strains versus 12 repressed in
Induced
Repressed Fig. 2. Seizure-induced differential gene expression in the cerebellum and hippocampus of C57BL/6 and 129/SvEv TAC mice. Gene expression in the hippocampus and cerebellum from C57BL/6 and 129/SvEvTac mice l h after seizure was compared to gene expression from C57BL/6 and 129/SvEvTac mice at baseline. Genes that had a 1.8 or greater fold change increase with an ADC in signal intensity of 50 or more were considered induced (upper panel), whereas genes that had a 1.8 or greater fold change decrease with an ADC in signal intensity of 50 or more were considered repressed (lower panel). The number of genes meeting these criteria in either the hippocampus or the cerebellum is indicated in the outer portion of the circles, whereas the number of genes meeting these criteria that are expressed in both cerebellum and hippocampus are indicated in the overlapping portion of the circles. The numbers in parentheses are from a more stringent test in which the genes meeting these criteria were required to have behaved consistently in both mouse strains (see text for details of the analysis method used).
both strains). This difference was largely due to a significant increase in the number of genes induced in the B6 hippocampus (49 in C 5 7 B L / 6 c o m p a r e d to 12 in 129SvEv, P < 0.001). Interestingly, however, the transcriptional response of several known
158 immediate-early genes, including members of thefos and jun family, serum and glucocorticoid-regulated kinase (sgk), growth factor inducible immediate early gene (3CH134), cox-2, and the transcription factors KROX20 and zif/268 showed a similar level of postseizure induction in the two strains (Sandberg et al., 2000). Therefore, the immediate-early response to seizure and the response to seizure in the cerebellum were similar between the two strains whereas the overall transcriptional response in the hippocampus was blunted in 129 compared to that in B6. The complete list of genes as well as the data used to generate the lists in each of the categories described above are available at our web site at
http :/ / www.salk.edu/ docs /labs /barlow /brainstrain /. Brain region specific gene expression measurements An important use of region-specific expression studies is to identify uniquely expressed genes and their promoters, which can be used to drive expression of a transgene in specific cell types or tissues in animal models. The paucity of site-specific tools in the mouse makes this an important use of the expression results. In addition, determining which genes are responsible for the unique structures and functions of specific brain regions will also prove informative. Perhaps most importantly, for the case of seizure induced damage, it will be interesting to identify genes which are responsible for one sub-region to be uniquely resistant or sensitive to seizure induced damage. We have performed gene expression analysis on multiple brain regions to begin to address some of these questions (Sandberg et al., 2000; Lockhart and Barlow, 2001). Of the 13,069 probe sets analyzed, 7,169 (55%) gave a hybridization signal consistent with a call of 'present' in at least one brain region. This indicates that at least 55% of the genes covered on the murine arrays are detected in one or more areas of the adult male mouse brain. We next compared the expression profiles of cortex, cerebellum and midbrain within the same strain and found that, on average, a relatively small number of genes (70/13,069 or 0.54%) were expressed in a pattern suggesting they were highly enriched in or restricted to a specific brain region. For example, 23 genes
were expressed in the cerebellum that were not detected in other regions, and another 28 were not expressed in cerebellum but were present in other brain regions indicating that the cerebellum appears to be the most unique region of those tested. Importantly, genes such as PCP-2, a known cerebellar-specific gene, were identified as being specifically expressed in the cerebellum, providing further validation of the approach. In contrast to the cerebellum, the structures of the medial temporal lobe (hippocampus, amygdala and entorhinal cortex) showed extremely similar expression profiles. Only eight genes were unique to one of the three regions. Of the seven genes present in hippocampus but not amygdala or entorhinal cortex, six were also expressed outside of the medial temporal lobe. There was only one gene uniquely expressed in the amygdala, and none in the entorhinal cortex. This suggests that forebrain structures, despite some functional differences, are highly similar at the molecular level. Finally, the midbrain was interesting in that, although there were ten genes uniquely expressed, no genes were exclusively 'absent'. In contrast to the very small number of differences between brain regions, 13.6% (1,780/13,069) of the monitored genes were found to be uniquely expressed between brain and fibroblasts, even though the two very different types of cells express a similar overall number of genes. This indicates, as might be expected, that various brain regions are considerably more similar to each other than to fibroblasts. However, in contrast to the allor-none analysis described above, many more genes showed differential levels of expression at 1.8-fold or higher between the various brain regions (Lockhart and Barlow, 2001). One important point is that these studies compared large brain regions rather than sub-regions or specific cell types. It may be that differences in gene expression between various brain regions are much more pronounced in certain cell types, and that the high similarity in the expression patterns from different regions is due to averaging over all the cell types in the tissue. More recently we have begun to perform analysis on smaller sub-regions of the brain (Fig. 3). Shown in Fig. 3 is an analysis of the subregions of the hippocampus including the dentate gyrus, CA1 and CA3 regions. Note that the dentate gyrus is the most unique of the three regions. Impor-
159
CA1 CA3
43 (23)
CA3
CA1 & CA3
Dentate Gyms
88 (39)
58 (26)
67 (7)
Fig. 3. Differential gene expression in hippocampal sub-regions. The three major subregions of the hippocampus, CAI, CA3 and dentate gyms, were compared to each other to determine the number of genes that differed between these regions. The number of genes meeting the criteria as described in the text in 2/2 comparisons between each region is shown. The numbers in parentheses represent genes meeting these criteria in both mouse strains.
tantly, of the 88 genes that were unique between CA 1 and the dentate gyms, 32 genes are n o t present in CA3 suggesting they are unique to CA1. This example suggests that by studying brain regions that are uniquely sensitive to specific insults it may be possible to identify genes that are uniquely expressed to determine why a particular region is sensitive or resistant to neurotoxic insults. As it becomes possible to use this technology for nuclei or even small cell populations in the CNS, higher resolution, regionspecific and cell-type specific information will be gained.
Summary These highly parallel gene expression approaches allow one to look globally at the interactions of genes and modifiers and their effects, and will greatly enhance our ability to identify the genes that contribute to important phenotypes, and to define the role of developmental alterations, mutations, and compensatory mechanisms in causing or modifying particular behaviors. The studies described in this review demonstrate the feasibility and utility of expression profiling in the brain. The expression results serve as a framework to begin to understand, for example, the factors responsible for the variation in behavioral phenotypes, drug sensitivity and neurotoxic-induced cell death. There is no doubt that the combination of gene targeting technology, robust behavioral analysis, genetics, biochemistry and global gene expression measurements will provide new avenues for
studying the brain and further our ability to understand the interplay between genes that give rise to unique brain functions and complex behaviors.
Abbreviations 129 A ADC B6 D I IVT M MD MI MM NC P PM PTZ Q
129/SvEvTac absent or undetected average difference change C57BL/6 decrease increase in vitro transcription marginal marginal decrease marginal increase mismatch no change present or detected perfect match pentylenetetrazol noise
Acknowledgements We would like to thank Cindy Doane for help with manuscript preparation, David J. Lockhart for continued advice, Daniel J. Lockhart for the development of the gene expression filtering tools and members of the Barlow laboratory for comments. This work was supported by grants from the Esther A. and Joseph Klingenstein Fund and the Frederick B. Rentschler Developmental Chair to C.B, and by the Joe W. and Dorothy Brown Foundation.
References Carter, T.A., Del Rio, J.A., Greenhall, J.A., Latronica, M.L., Lockhart, D.J. and Barlow, C. (2001) Chipping away at complex behavior: Transcriptome/phenotype correlations in the mouse brain. Physiol. Behav., 73: 849-857. Fodor, S.P.A., Read, J.L., Pirrung, M.C., Stryer, L., Lu, A.T. and Solas, D. (1991) Light-directed, spatially addressable parallel chemical synthesis. Science, 251: 767-773. Ginsberg, S.D., Hereby, S.E., Lee, V.M., Eberwine, J.H. and Trojanowski, J.Q. (2000) Expression profile of transcripts in Alzheimer's disease tangle-bearing CA1 neurons. Ann. Neurol., 48: 77-87. Lee, C.K., Weindruch, R. and Prolla, T.A. (2000) Gene-
160 expression profile of the aging brain in mice. Nat. Genet., 25: 294-297. Lipshutz, R.J., Morris, D., Chee, M., Hubbell, E., Kozal, M.J., Shah, N., Shen, N., Yang, R. and Fodor, S.P. (1995) Using oligonucleotide probe arrays to access genetic diversity. Biotechniques, 19: 442-447. Lipshutz, R.J., Fodor, S.P., Gingeras, T.R. and Lockhart, D.J. (1999) High density synthetic oligonucleotide arrays. Nat. Genet., 21: 20-24. Lockhart, D.J. (1999) The chipping forecast. Nat. Genet. SuppL, 21: 3-50. Lockhart, D.J. and Barlow, C. (2001) Expressing what's on your mind: DNA arrays and the brain . Nat. Rev. Neurosci., 2: 63-68. Lockhart, D.J. and Winzeler, E.A. (2000) Genomics, gene expression and DNA arrays. Nature, 405: 827-836. Lockhart, D.J., Dong, H., Byrne, M.C., Follettie, K.T., Gallo, M.V., Chee, M.S., Mittmann, M., Wang, C., Kobayashi, M.,
Horton, H. and Brown, E.L. (1996) Expression monitoring by hybridization to high-density oligonucleotide arrays. Nat. Biotechnol., 14: 1675-1680. Marshall, A. and Hodgson, J. (1998) DNA chips: an array of possibilities. Nat. Biotechnol., 16: 27-31. Mimics, K., Middleton, A.M., Lewis, D.A. and Levitt, P. (2000) Molecular characterization of schizophrenia viewed by microarray analysis of gene expression in prefrontal cortex. Neuron, 28: 53-67. Sandberg, R., Yasuda, R., Pankratz, D.G.° Carter, T.A., Del Rio, J.A., Wodicka, L., Mayford, M., Lockhart, D.J. and Badow, C. (2000) From the cover: regional and strain-specific gene expression mapping in the adult mouse brain. Proc. Natl. Acad. Sci. USA, 97:11038-11043. Wodicka, L., Dong, H., Mittmann, M., Ho, M.-H. and Lockhart, D.J. (1997) Genome-wide expression monitoring in Saccharomyces cerevisiae. Nat. Biotechnol., 15: 1359-1367.
T. Sutula and A. Pitkanen (Eds.) Progress in Brain Research, Vol. 135 © 2002 Elsevier Science B.V. All rights reserved
CHAPTER 14
Functional genomics in experimental and human temporal lobe epilepsy: powerful new tools to identify molecular disease mechanisms of hippocampal damage Albert J. Becker *, Otmar D. Wiestler and lngmar Bltimcke Department of Neuropathology, University of Bonn Medical Center, Bonn, Germany
Abstract: The human genome project is a milestone for molecular genetic studies on complex, sporadic disorders in the human central nervous system (CNS). Functional analysis and tissue-/cell-specific expression profiles will be of particular importance anticipating the magnitude of expressed genes in the brain and their dynamic epigenetic modifications. The recent progress in microarray technologies allows expression studies for a large number of genes. In combination with laser-microdissection and quantitative reverse transcription-polymerase chain reaction technologies, such large-scale expression analyses can be successfully addressed in well-defined tissue specimens or cellular subpopulations. Complex, sporadic diseases, such as temporal lobe epilepsy (TLE), are challenging for functional genomics. Issues of particular importance in this field include molecular mechanisms of neurodevelopmental abnormalities, neuronal plasticity and hyperexcitability as well as neuronal cell damage in affected CNS areas. The availability of anatomically well-preserved surgical specimens, i.e. hippocampus obtained from epilepsy patients with Ammon's horn sclerosis or focal lesions not affecting the hippocampus proper as well as comparisons with experimental TLE models may help to elucidate specific molecular-pathological mechanisms during epileptogenesis and in chronic conditions of the disease.
Introduction
Temporal lobe epilepsy Epilepsy is a common neurological disorder characterized by recurrent spontaneous seizures that affects about 2 - 3 % of the population worldwide. A substantial fraction of epileptic patients does not respond to antiepileptic drug therapy. In most of these patients, seizures originate in the mesial temporal lobe (Elger and Schramm, 1993). Several lines of evidence sug-
* Correspondence to: A.J. Becker, Department of Neuropathology, University of Bonn Medical Center, Sigmund-Freud Str. 25, 53105 Bonn, Germany. Tel.: +49228-287-9108; Fax: +49-228-287-4331; E-mail: albert_becker @uni-bonn.de
gest the hippocampal formation to be critically involved in temporal lobe epilepsy (TLE): Recordings from intracerebrally implanted electrodes demonstrate that the first electrographic abnormalities in temporal lobe seizures often appear within this structure (Van Roost et al., 1998). Surgical removal of the amygdala and hippocampal formation considerably diminishes or abolishes seizures in most pharmacoresistant TLE patients (Zentner et al., 1995). TLE pathogenesis involves a variety of developmental, metabolic and/or hypoxic alterations, while it lacks significant genetic inheritance (Jackson et al., 1998). A central question addresses the intriguing issue whether the alterations observed in the chronic epileptic state resemble an end stage of the disease after long-term additive pathophysiological events, maintenance of an initial etiologic episode or a combination of both. Data from human and experimental
162 TLE suggest that recurrent seizures, but not necessarily status epilepticus, progressively affect the hippocampal formation (Cavazos et al., 1994; Kalviainen et al., 1998; Salmenpera et al., 1998). While numerous molecular genetic alterations of cellular injury have been identified in hippocampal neurons following status epilepticus and within epileptogenesis (Lynch et al., 1996; Coulter and DeLorenzo, 1999), molecular pathways associated with neuronal damage and recurrent brief seizure episodes, i.e. the chronic state of human TLE, are less characterized. With this review we will address the question whether pathogenetic casades similar to those induced by status epilepticus are active in the chronic state of TLE and how functional genomics can gain our understanding of region- and cell-specific epileptogenesis. How can such delicate experiments be successfully applied in human tissue, in particular since proper controls are, for obvious reasons, not available? Neuropathological evaluation of surgical specimens is an important strategy to address this obstacle. The majority of resected mesial temporal lobe structures can be classified in two groups, i.e. Ammon's horn sclerosis versus focal lesions not affecting the hippocampus proper. The comparative analysis between both groups of patients, i.e. with respect to specific cell types and/or anatomical regions, may help to identify pathogenetic mechanisms specifically associated with each epileptogenic lesion. Ammon's horn sclerosis
Approximately 60% of TLE patients present with severe unilateral atrophy of either the fight or left hippocampus (fight/left = 1.08/1, n = 293, data obtained from the archives of the Department of Neuropathology, University of Bonn Medical Center). Histopathologically, the hippocampal formation shows segmental neuronal loss in CA1 and CA4, whereas CA2 and dentate gyms granule cells appear more resistant (Bliamcke et al., 1999a). Dense fibrillary astrogliosis and sclerosis of the tissue are observed in all segments with prominent neuronal cell loss. This macroscopic aspect has been first described in 1880 and classified as Ammon's horn sclerosis (AHS) (Sommer, 1880; Margerison and
Corsellis, 1966). Neuronal cell loss is also observed in hippocampal segments others than CA1 and CA4 (Kim et al., 1990; BliJmcke et al., 1996b). In the dentate gyms, specific cytoarchitectural abnormalities have been described in AHS, which may reflect seizure associated postnatal neurogenesis and persistence of Cajal-Retzius-like interneurons (Bltimcke et al., 1996a, 1999b, 2001; Nakagawa et al., 2000). Along with hippocampal cell loss, the entorhinal cortex and amygdala complex is affected in most patients (Pitkanen et al., 1998; Yilmazer-Hanke et al., 2000). Lesion-associated TLE
A second group, representing approximately 3040% of TLE patients exhibit focal lesions within the temporal lobe, which usually do not involve the hippocampus proper. This group covers lowgrade glio-neuronal neoplasms, i.e. gangliogliomas and dysembryoplastic neuroepithelial tumors (DNT), low-grade astrocytomas and oligodendrogliomas as well as glio-neuronal malformations, i.e. focal cortical dysplasia (Wolf and Wiestler, 1993; Bltimcke et al., 1999a). These lesions share predominant localization within the temporal lobe and frequent association with chronic, intractable seizures combined with benign biological behavior and rare recurrence after surgical removal. Gangliogliomas and DNTs are composed of a neoplastic glial and dysplastic neuronal cell population (Bltimcke et al., 1999a). Some of these tumors were found together with a malformative lesion pointing towards a maldevelopmental origin (Wolf and Wiestler, 1993; Wolf et al., 1994; Blfimcke et al., 1999a). Recently, a novel mutation in the TSC2 gene was selectively detected within the glial component of a ganglioglioma suggesting that the glioma portion derives from clonal evolution (Becker et al., 2001). In contrast to the characteristic pattern of neuronal cell loss in AHS, no significant neuropathological alterations are observed in the hippocampal formation of lesionassociated epilepsy (Bltimcke et al., 2000). A small subgroup of patients presents with dual pathology, i.e. AHS in addition to focal lesions (Bliamcke et al., 1999a).
163 The use of animal models to study functional genomics of TLE
A major obstacle for the systematic analysis of surgical specimens obtained from patients with pharmacoresistant TLE is the lack of non-epileptic, agematched controls. Comparative analysis between two groups of TLE patients, i.e. AHS versus lesionassociated TLE can be applied with respect to different clinico-pathological features, i.e. extent of structural changes and duration/severity of seizures. Rarely, biopsy samples from tumor patients without epileptic seizures can be obtained as truly nonepileptic controls. Autopsy specimens suffer from a variable post mortem delay and are therefore not appropriate for delicate molecular biological studies, such as mRNA expression analysis. Another strategy for the evaluation of epilepsyassociated changes follows the comparison between human epilepsy tissue and experimental animal models. In particular, alterations observed in both human and different experimental models appear more likely to be of pathogenic relevance (Bltimcke et al., 2000). Since surgical specimens are usually obtained at a late stage of the disease, experimental data based on human samples may not allow to distinguish between primary pathogenetic lesions and secondary changes. Animal models provide the possibility to analyze epileptogenesis and epilepsyassociated structural and molecular changes. As an example, studies on seizure-induced neuronal apoptosis are almost exclusively restricted to animal models, because terminal deoxynucleotidyl transferase dUTP nick-end labeling (TUNEL-staining) can be employed only for a limited time interval after the excitotoxic event (Tuunanen et al., 1999; Venero et al., 1999). It is extremely challenging to collect groups of human surgical specimens within appropriate post seizure intervals. Commonly used animal models for focal limbic epilepsies are kainate-, pilocarpine- and kindlinginduced chronic seizures (Mello et al., 1993; Ben-Ari and Cossart, 2000; Sutula, 2001). Induction of status epilepticus by intracerebral or intraperitoneal injection of epileptogenic compounds induces delayed segmental neuronal cell loss in the hippocampal formation. The segmental pattern of cell loss partially resembles that in human AHS with less severe neu-
ronal loss of CA1 in animals. On the other hand, subconvulsive electrical kindling of the amygdala or tractus perforans produces sustained hippocampal seizure activity, which usually results in less significant histopathological alterations compared to application of epileptogenic compounds (Clusmann et al., 1992; Bertram and Lothman, 1993). The degree of histopathological changes also depends on the severity and frequency of seizures, in particular following status epilepticus (Bertram and Lothman, 1993; Cavazos et al., 1994; Ebert and L/Sscher, 1995). In summary, different experimental paradigms have been established with neuropathological changes similar to TLE patients, i.e. pilocarpine or kainate injections modeling AHS, whereas kindling associated epileptogenesis resembles the epileptogenic hippocampus in patients with focal lesions. With increasing availability of transgenic mice carrying targeted mutations in epilepsy-related candidate genes, such models will play a significant role to study epileptogenesis and epilepsy-associated structural and molecular alterations. However, it is pivotal to note that mouse strains bear different susceptibilities for kainic acid-induced excitotoxic neurodegeneration. With respect to studies on functional genomics, mouse strain specific mRNA expression levels as well as developmentally regulated and regionally different gene expression profiles have to be considered (Schauwecker and Steward, 1997; Wen et al., 1998; Cantallops and Routtenberg, 2000; Sandberg et al., 2000). Functional genomics of TLE specimens
Microarray technology offers the opportunity to analyze human gene expression profiles on a genome wide level (Brown and Botstein, 1999; Lipshutz et al., 1999). In recent years, different expression array technologies have been developed (Table 1) including cDNA nylon arrays (Wellmann et al., 2000), large-scale oligonucleotide (Lipshutz et al., 1999) and glass microscope slide DNA arrays (Brown and Botstein, 1999) (Fig. 1). Besides such large-scale 'chip' approaches, designed arrays and real-time reverse transcription-polymerase chain reaction (RTPCR) quantification represent useful tools to substantiate hypothesis based expression studies (Bartosiewicz et al., 2000; Miyajima et al., 2001).
164
• ,,
L
',;
,
~" ~ ' ~ ",~'c u.~
.~
~
.'~
0 •
.
%'it
q,'
' . %% •
8
•
•~
[--
o
~n
"
"".~'~,, •, ~
o6
.
0
/~ ",
~
~
~ . ~
!
m
O
]
o
~ ~.~ ~.~
E
. m
< Z
~ ~.~ ~ -
e= . i i
es
C~
~
~
i
tU -i
z
~-
i
Q.
E
~
E
~
~ ~ z
~
o
m
OH
e~
E
mm
0
"~.
N
0
e~ e~
3::
0
0
~
m
~
m
165 TABLE 1 Differentexpressionanalysis tools are outlined Type
Approach
Scale
Label type
Methodicalfeature
System
Publication
Atlas-Array filters GeneChip microarrays cDNA microarrays PIQOR Real-time RT-PCR In-situ hybridization
inductive inductive inductive inductive deductive deductive
low density high density high density low density low density singlegene
radioactive fluorescent fluorescent fluorescent fluorescent fluorescent, radioactive
cDNA,nylon oligonucleotide cDNA cDNA relativequantitation oligonucleotides, RT-PCRproducts
Clontech Affymetrix P. Brown Memorec Applera
Wellmann et al. (2000) Lipshutz et al. (1999) Brown and Botstein (1999)
Major strategies for chip analysis include comparisons between identified cell populations, brain regions or groups of neuropathologically characterized individuals (Sandberg et al., 2000). The careful selection of appropriate controls or matched pairs of samples plays a pivotal role for expression profiling experiments. Especially when starting from hippocampal biopsy specimens of pharmacoresistant TLE patients, there are a number of problems inherent in such an approach. Besides significant expression differences between human subjects due to individual genetic background, temporal lobe epilepsy patients may exhibit considerable heterogeneity with respect to drug treatment, progression of the disease and frequency, type and intensity of seizures. These problems should be addressed by a careful matching of patient subjects with respect to clinical criteria and by increasing numbers of studied individuals per group, for which real-time RT-PCR may provide a particularly economical method. In complex, sporadic brain disorders, such as temporal lobe epilepsy a variety of molecular pathways and genes are involved and differentially regulated during the long medical history of the disease (Table 2) (BRimcke et al., 1999a). The bioinformatic analysis of large scale or even transcriptome expression (i.e. the level of each mRNA detectable in the genome) is a major challenge (Mimics, 2001). Introduction of certain reference or housekeeping genes enables a systematic comparison between different sets of experiments, including complex estimations, such as cluster analysis (Eisen et al., 1998; Bassett et al., 1999). However, the identification of reliable housekeeping genes may considerably change with experimental paradigms as has been shown to be
Fink et al. (1998) Lie et al. (2000); Chen et al. (2001)
relevant for the pilocarpine epilepsy model (Waha et al., 1998; Chen et al., 2001). At the present time, a major limitation of microarray technology is the need of sufficient amounts of region/cell specific mRNA. For the majority of microarray systems, approximately 50 Izg of total RNA are recommended for reverse transcription (Duggan et al., 1999). Due to these amounts of required starting material, such strategies do not provide information about cell specific patterns of gene regulation. Furthermore, certain expression alterations may reflect changes in the composition of neuronal tissue, if pronounced degeneration of specific cell types or reactive cell infiltration is observed. Such problems have to be considered for expression studies in TLE since patients with AHS exhibit segmental neuronal cell loss, reactive astrogliosis as well as structural and molecular reorganization in the hippocampal formation (Bltimcke et al., 1999a). Since altered mRNA levels between hippocampi of patients with AHS and control individuals may then reflect simply altered tissue composition, expression array analyses from TLE tissue have to be supplemented by a detailed analysis of expression alterations at the cellular level. Several approaches have been taken to establish expression analysis with very low amounts of input mRNA including RT-PCR and antisense mRNA (aRNA) amplification techniques, aRNA amplification has been used to generate sufficient amounts of aRNA for array hybridization starting from individual cells (Eberwine et al., 1992; Phillips and Eberwine, 1996; Luo et al., 1999). An advantage of this approach is the opportunity to screen large numbers of genes starting from minute amounts of mRNA. This linear amplification technique may se-
166 TABLE 2 Pathogenic mechanisms potentially involved in TLE
1 2 3 4 5
TLE-associated pathomechanism
Candidate genes
Apoptosis Cytoarchitectural malformations Axonal reorganization Cellular hyperexcitability Gliosis
iNOS, PIN, JIP-1, JNK, c-JUN, caspases, heat shock, ubiquitin, bcl, bax Reelin, CDK5, p35, TSC1, TSC2 Extracellular matrix molecules and receptors, cadherin, neurotransmitter-receptors Voltage dependent Ca2+, Na +, K+-channels, neurotransmitter-receptors Connexins, extracellular matrix molecules and receptors, K+-inward rectifiers
lect for certain populations of mRNAs, a problem which may be overcome by a novel strategy combining linear amplification and a template switch effect for microarray probe preparation (Wang et al., 2000). Compared to these techniques, real-time RT-PCR allows to monitor reaction dynamics of the PCR amplification and relative efficiency of target as well as reference gene amplification for every cycle. It provides detailed information regarding the linear dynamic range of the reactions (Fink et al., 1998). However, RT-PCR will usually be restricted to a limited number of genes. RT-PCR combined with laser microdissection (Fink et al., 1998; Schtitze and Lahr, 1998; Lahr, 2000) of hippocampal subfields will provide a reproducible tool to confirm and/or localize differentially regulated genes of interest to respective hippocampal cell populations. In addition to the confirmation of differential gene expression levels with RT-PCR, experimental errors introduced by false annotation of spotted sequences have to be controlled (Knight, 2001). These obstacles underline the need for alternative strategies to analyze expression profiles and to confirm their regional and cellular origin. Large-scale expression studies offer the unique opportunity to identify novel pathways potentially involved in sporadic diseases, such as TLE and AHS. It is important to note that the choice of the expression-monitoring tool, i.e. real-time RT-PCR or microarrays, strongly influences the experimental design. Simultaneous, transcriptome wide expression analysis describes an inductive approach. Expression profiles are compared between a variety of physiological and/or pathophysiological states. The result is an indefinite number of differentially expressed genes, which is used to build up a hypothesis for further experiments. However, for certain genes, it might be difficult to transfer differential expression into functional consequences. Using a deductive or
'top down' approach, expression analysis of a limited number of genes requires a certain hypothesis to be verified or disproven (Bassett et al., 1999).
Molecular pathways of seizure-induced hippocampal damage Necrosis and apoptosis have been shown as two independent pathways of excitotoxic neuronal damage (Ankarcrona et al., 1995; Van Lookeren Campagne et al., 1995). While necrosis results from cellular swelling, bursting and lysis, apoptosis follows a programmed mode of active cellular degeneration (Nicotera et al., 1997). Hippocampal apoptosis has been described in several epilepsy models. Following pilocarpine and kainate-induced status epilepticus, neuronal apoptosis is pronounced in CA3 and CA1 neurons, whereas dentate gyrus granule cells are more resistant (Mello et al., 1993; Ben-Ari and Cossart, 2000). The molecular pathogenesis of neuronal apoptosis has been associated with a glutamate receptor-mediated pronounced intracellular Ca 2+ increase (Choi, 1987; Wahlestedt et al., 1993). Subsequently, excitotoxicity proceeds via stimulation of various intracellular signaling cascades including the c-Jun amino-terminal kinase (JNK) group of mitogen-activated protein kinases (Gupta et al., 1996; Martin et al., 1996; Kawasaki et al., 1997; Schwarzschild et al., 1997) and the formation of nitric oxide (NO) with caspase-mediated apoptosis (Bruno et al., 1993; Leist et al., 1997; Montecot et al., 1998). Mice with a targeted mutation of the JNK3 isoform selectively expressed in the nervous system show an increased resistance to kainic acid-induced cell loss (Yang et al., 1997). Kainate-induced expression of JNK-1 relates to increased apoptosis in hippocampal neurons, while serine-73 phosphorylation of c-Jun is associated with resistance to cell
167 death (Schauwecker, 2000). In the kainate model, an inverse correlation is observed between the hippocampal distribution of kainic acid receptors and the pattern of neuronal cell loss, i.e. low receptor density in the highly vulnerable segment CA1 and vice versa in the dentate gyrus (Sperk et al., 1983). A striking relationship occurs between expression of the endogenous protein inhibitor of neuronal nitric oxide synthase (PIN), a cytoplasmic inhibitor of the JNK signal transduction pathway designated JNK interacting protein-1 (JIP-1) and the gene for the apoptosis-executing protease caspase-3 to patterns of hippocampal vulnerability after kainate-induced seizures (Dickens et al., 1997; Jaffrey and Snyder, 1996; Becker et al., 1999). In the dentate gyrus, no delayed cell loss is observed although high kainate receptor densities are encountered in this area (Sperk et al., 1983). Here, PIN and JIP-1 mRNA signals increase significantly, whereas caspase-3 expression remains at basal levels. In CA1 with extensive neuronal cell loss and low kainate receptor density, weaker expression of JIP-1 and PIN vs. induction of caspase-3 are observed compared to the dentate gyrus (Sperk et al., 1983; Becker et al., 1999). This selective regulation may serve as example for the capacity of downstream apoptotic signaling cascades to interfere with excitotoxic apoptotic stimuli in different hippocampal subfields. Functional pathways involved in seizure-associated apoptosis include expression of the TP53 tumor suppressor (Sakhi et al., 1994; Liu et al., 1999) and the tissue plasminogen activator gene (Tsirka et al., 1995). Certain lines of evidence suggest that single intermittent seizures, resembling the chronic state of TLE, induce apoptosis. Severe neuronal cell loss is observed after repeated kindling seizures (Cavazos et al., 1994). Also, apoptosis occurs in the dentate gyrus following intermittent kindling stimulation in the ventral CA1 region (Bengzon et al., 1997). There is evidence that a limited number of brief repeated kindling seizures do not alter total amygdaloid or hilar neuronal cell numbers, but may induce degeneration of certain neuronal subpopulations (Pretel et al., 1997; Tuunanen et al., 1997; Tuunanen and Pitkiinen, 2000). However, novel data suggest that neurodegenerative pathways in the chronic TLE state may be similar to those which occur early during TLE pathogenesis. Expression of bcl-2, bcl-xL, bax, caspase-3 and
caspase-1 proteins shows alterations in resected temporal lobe structures from patients with long-term pharmacoresistant TLE (Henshall et al., 2000). The molecular signals predisposing hippocampal neurons to enhanced or reduced susceptibility for seizureinduced damage in the chronic TLE state have not yet been fully characterized. Potential candidates include a variety of ionotropic and metabotropic neurotransmitter receptors.
Neurodegeneration or neuroprotection: role of neurotransmitter receptors Several lines of evidence suggest that recurrent spontaneous seizures in human as well as experimental chronic TLE are caused by alterations in the balance between inhibitory and excitatory neurotransmitter systems (Meldrum et al., 1999; Ben-Aft and Cossart, 2000; Chapman, 2000; Kullmann et al., 2000). Changes in neurotransmitter receptor expression, subunit composition and their functional consequences may not only contribute to enhanced seizure susceptibility but also predispose or protect neuronal cells for/from cellular damage. This has been demonstrated for excitatory ionotropic and metabotropic glutamate receptors as well as for inhibitory ionotropic GABAA receptor pathways (Jacobs et al., 2000; Meldrum, 2000; Coulter, 2001). With respect to epilepsy-associated neuronal damage, neuroprotection as well as novel pharmacological treatment strategies, metabotropic glutamate receptors (mGluRs) have emerged as interesting target molecules. The mGluR family consists of at least 8 different subtypes (Nicoletti et al., 1996). Activation of class I mGluRs (i.e. mGluR1 and mGluR5) results in excitatory membrane depolarization followed by release of Ca 2+ from intracellular stores, which appears to be mediated by inositol phosphate hydrolysis. Class II (mGluR2 and mGluR3) and class III (mGluR4, mGluR6-8) mGluRs operate mainly via a G-protein-mediated inhibition of adenylate cyclase (Nicoletti et al., 1996). Immunohistochemical studies and in-situ hybridization revealed distinct preor postsynaptic localization of mGluR isoforms in rat (Baude et al., 1993; Shigemoto et al., 1997) and human hippocampus (Bliimcke et al., 1996c; Lie et al., 2000). Recent molecular, pharmacological and physiological data point to a role for specific mGluR
168 subtypes in the generation and propagation of epileptiform activity (Mayat et al., 1994; Attwell et al., 1995; Holmes et al., 1996; Aronica et al., 1997; Merlin et al., 1998). In particular, agonists of excitatory class I mGluRs exert significant convulsant properties, whereas class I antagonists can prevent excitotoxic neuronal damage (Mukhin et al., 1996; Strasser et al., 1998; O'Leary et al., 2000). Furthermore, kainate and kindling models revealed enhanced expression of class I mGluRs as well as increased phosphoinositide hydrolysis (Nicoletti et al., 1987; Akbar et al., 1996). Expression alterations of excitatory class I (mGluR1 and mGluR5) and inhibitory class III (mGluR4) metabotropic glutamate receptors were observed in chronic TLE. mRNA expression and protein distribution analysis of mGluR1 and mGluR5 revealed a striking induction of mGluRlc~ in the hippocampal dentate gyms with an almost identical regional distribution in kainic acid-treated and amygdala-kindled, chronic epileptic animals as well as in human TLE specimens (Bltimcke et al., 2000). This expression alteration may significantly predispose cells with enhanced mGluRlc~ expression to neuronal excitability. An opposite functional effect may be the result of regional and cellular induction of the mGluR4 subtype in chronic TLE. In contrast to control hippocampus obtained from nonepileptic controls, i.e. patients suffering from diffuse infiltrating malignant gliomas, most TLE specimens showed a significant increase of mGluR4 protein and mRNA expression within the dentate gyms and residual CA4 neurons (Lie et al., 2000). With respect to a neuroprotective potential of mGluR4 in various cell culture models, mGluR4 induction may constitute a cellular mechanism to antagonize excitatory hippocampal activity and critical intracellular Ca 2+ overload (Gasparini et al., 1999; Bruno et al., 2000). In the pilocarpine animal model, the acute status epilepticus is frequently followed by a silent period of weeks before chronic spontaneous limbic seizures occur (Coulter, 2000). In the hippocampus, sprouting of zinc containing mossy fibers can be observed during this adaptation phase (Cavazos et al., 1991; Mello et al., 1993). GABAA receptors of dentate gyms granule cells show enhanced sensitivity to blockade by zinc in chronic TLE. Expression profiling of single cells and functional analysis revealed major alterations in subunit composition of
the GABAA receptor (Brooks-Kayal et al., 1998). The enhanced sensitivity of dentate gyms granule cell GABAA receptors to blockade by zinc in chronic TLE may be due to decreased expression of the c~l subunit of the GABAA receptor. The combination of increased zinc sensitive GABAA receptors and sprouted zinc-containing mossy fiber terminals may result in a failure of inhibition and concomitant enhanced seizure propensity triggering chronic TLE and cellular damage (Brooks-Kayal et al., 1998). Enhanced potency of GABA in activating GABAA receptors as well as reduced c~2 and c~5 GABA subunit expression in CA1 and additional changes in subunit expression in other hippocampal neurons have also been described using a combined approach of expression arrays and electrophysiology (Rice et al., 1996; Gibbs et al., 1997; Becker et al., 1998; Coulter, 1999; Coulter and DeLorenzo, 1999).
Perspectives for functional genomics in human TLE The plethora of functional cascades involved in the pathogenesis of chronic TLE is a challenging feature of this complex, sporadic disease. With the opportunity to study large-scale gene expression using novel microarray technologies, we may discover novel pathways of TLE associated hyperexcitability, neuronal damage or functional/stmctural plasticity. Due to the complexity of human TLE tissue, TLE animal models and the expression array data, epilepsy researchers using array technology would benefit from intemet platforms for functional genomics providing access to brief annotations of specific genes as well as links to known biochemical pathways and interactions at the transcriptional level to verify and extend their observations. Currently, experimental conditions and potential problems of expression data are discussed intensively (Geschwind, 2001; Lockhart and Barlow, 2001) and first internet platforms for expression array data such as Gene Expression Omnibus (GEO) and ArrayExpress are established. Uniform expression data formats and integrated analysis tools will also be essential for a successful application of a functional genomics approach in human and experimental TLE.
169
Acknowledgements T h e authors thank S. N o r m a n n for e x c e l l e n t technical assistance. W e like to a c k n o w l e d g e the grateful support and contribution of our clinical c o l l e a g u e s Profs. E l g e r and S c h r a m m to the interdisciplinary e p i l e p s y p r o g r a m . Our w o r k is g e n e r o u s l y supported by D e u t s c h e F o r s c h u n g s g e m e i n s c h a f t ( S F B - T R 3 ) , B M B F ( G e n o m i c N e t w o r k s , S P l l ) and the B O N F O R p r o g r a m o f the U n i v e r s i t y of B o n n M e d i c a l Center.
References Akbar, M.T., Rattray, M., Powell, J.F. and Meldrum, B.S. (1996) Altered expression of group I metabotropic glutamate receptors in the hippocampus of amygdala-kindled rats. Mol. Brain Res., 43(1-2): 105-116. Ankarcrona, M., Dypbukt, J.M., Bonfoco, E., Zhivotovsky, B., Orrenius, S., Lipton, S.A. and Nicotera, R (1995) Glutamateinduced neuronal death: a succession of necrosis or apoptosis depending on mitochondftal function. Neuron, 15(4): 961973. Aronica, E.M., Gorter, J.A., Paupard, M.C., Grooms, S.Y., Bennett, M.V. and Zukin, R.S. (1997) Status epilepticus-induced alterations in metabotropic glutamate receptor expression in young and adult rats. J. Neurosci., 17(21): 8588-8595. Attwell, R, Kaura, S., Sigala, G., Bradford, H.F., Croucher, M.J., Jane, D.E. and Watkins, J.C. (1995) Blockade of both epileptogenesis and glutamate release by (lS,3S)-ACPD, a presynaptic glutamate receptor agonist. Brain Res., 698(1-2): 155-162. Bartosiewicz, M., Trounstine, M., Barker, D., Johnston, R. and Buckpitt, A. (2000) Development of a toxicological gene array and quantitative assessment of this technology. Arch. Biochem. Biophys., 376(1): 66-73. Bassett, D.E., Eisen, M.B. and Boguski, M.S. (1999) Gene expression informatics--it's all in your mine. Nat. Genet., 21: 51-55. Baude, A., Nusser, Z., Roberts, J.D., Mulvihill, E., Mcllhinney, R.A. and Somogyi, P. (1993) The metabotropic glutamate receptor (mGluR1 alpha) is concentrated at perisynaptic membrane of neuronal subpopulations as detected by immunogold reaction. Neuron, 11(4): 771-787. Becket, A.J., Rikhter, T.Y., Bl0mcke, I., Wiestler, O.D. and Coulter, D.A. (1998) Alterations of GABAA receptor subunit mRNA expression in the hippocampal CA3 region of pilocarpine-treated rats [Abstract]. Epilepsia, 39(Suppl. 6): 11-12. Becker, A.J., Gillardon, F., Bltimcke, I., Langendorfer, D., Beck, H. and Wiestler, O.D. (1999) Differential regulation of apoptosis-related genes in resistant and vulnerable subfields of the rat epileptic hippocampus. Mol. Brain Res., 67(1): 172176.
Becker, AJ., Ltibach, M., Klein, H., Normann, S., Ntithen, M.M., von Deimling, A., Mizuguchi, M., Elger, C.E., Schramm, J., Wiestler, O.D. and Bltimcke, I. (2001) Mutational analysis of TSCI and TSC2 genes in gangliogliomas. Neuropathol. Appl. NeurobioL, 27: 105-114. Ben-Aft, Y. and Cossart, R. (2000) Kainate, a double agent that generates seizures: two decades of progress. Trends Neurosci.. 23(11): 580-587. Bengzon, J., Kokaia, Z., Elmer, E., Nanobashvili, A., Kokaia, M. and Lindvall, O. (1997) Apoptosis and proliferation of dentate gyrus neurons after single and intermittent limbic seizures. Proc. Natl. Acad. Sci. USA, 94(19): 10432-10437. Bertram, E.H.D. and Lothman, E.W. (1993) Morphometric effects of intermittent kindled seizures and limbic status epilepticus in the dentate gyrus of the rat. Brain Res., 603(1): 2531. Bltimcke, I., Beck, H., Nitsch, R., Eickhoff, C., Scheffler, B., Celio, M.R., Schramm, J., Elger, C.E., Wolf, H.K. and Wiestler, O.D. (1996a) Preservation of calretinin-immunoreactive neurons in the hippocampus of epilepsy patients with Ammon's horn sclerosis. J. Neuropathol. Exp. Neurol., 55(3): 329-341. Bltimcke, I., Beck, H., Scheffler, B., Hof, P.R., Morrison, J.H., Wolf, H.K., Schramm, J., Elger, C.E. and Wiestler, O.D. (1996b) Altered distribution of the alpha-amino-3-hydroxy-5methyl-4-isoxazole propionate receptor subunit GluR2(4) and the N-methyl-D-aspartate receptor subunit NMDARI in the hippocampus of patients with temporal lobe epilepsy. Acta Neuropathol. (Berl. ), 92(6): 576-587. Bltimcke, I., Behle, K., Malitschek, B., Kuhn, R., Kn6pfel, T., Wolf, H.K. and Wiestler, O.D. (1996c) Immunohistochemical distribution of metabotropic glutamate receptor subtypes mGluRlb, mGluR2/3, mGluR4a and mGluR5 in human hippocampus. Brain Res., 736(1-2): 217-226. Bltimcke, I., Beck, H., Lie, A.A. and Wiestler, O.D. (1999a) Molecular neuropathology of human mesial temporal lobe epilepsy. Epilepsy Res., 36(2-3): 205-223. Bltimcke, I., Zuschratter, W., Schewe, J.C., Suter, B., Lie, A.A., Riederer, B.M., Meyer, B., Schramm, J., Elger, C.E. and Wiestler, O.D. (1999b) Cellular pathology of hilar neurons in Ammon's horn sclerosis. J. Comp. Neurol., 414: 437-453. Bltimcke, I., Becker, A.J., Klein, C., Scheiwe, C., Lie, A.A., Beck, H., Waha, A., Friedel, M., Kuhn, R., Emson, P., EIger, C. and Wiestler, O.D. (2000) Temporal lobe epilepsy associated up-regulation of metabotropic glutamate receptors: correlated changes in mGluR1 mRNA and protein expression in experimental animals and human patients. J. Neuropathol. Exp. Neurol., 59(I): 1-10. Bltimcke, I., Schewe, J.C., Normann, S., Brtistle, O., Schranma, J., Elger, C.E. and Wiestler, O.D. (2001) Increase of nestinimmunoreactive cells in the dentate gyrus of pediatric patients with early onset temporal lobe epilepsy. Hippocampus, 11(3): 311-321. Brooks-Kayal, A.R., Shumate, M.D., Jin, H., Lin, D.D., Rikhter, T.Y. and Coulter, D.A. (1998) Selective changes in single GABA(A) receptor subunit expression and function in temporal lobe epilepsy. Nat. Med., 4(10): 1166-1172.
170 Brown, EO. and Botstein, D. (1999) Exploring the new world of the genome with DNA microarrays. Nat. Genet., 21(1): 33-37. Bruno, V., Scapagnini, U. and Canonico, EL. (1993) Excitatory amino acids and neurotoxicity. Funct. Neurol., 8(4): 279-292. Bruno, V., Battaglia, G., Ksiazek, I., van der Putten, H., Catania, M.V., Giuffrida, R., Lukic, S., Leonhardt, T., Inderbitzin, W., Gasparini, E, Kuhn, R., Hampson, D.R., Nicoletti, E and Flor, EJ. (2000) Selective activation of mGlu4 metabotropic glutamate receptors is protective against excitotoxic neuronal death. J. Neurosci., 20(17): 6413-6420. Cantallops, I. and Routtenberg, A. (2000) Kainic acid induction of mossy fiber sprouting: dependence on mouse strain. Hippocampus, 10(3): 269-273. Cavazos, J.E., Golarai, G. and Sutula, T.E (1991) Mossy fiber synaptic reorganization induced by kindling: time course of development, progression, and permanence. J. Neurosci., 11(9): 2795-2803. Cavazos, J.E., Das, I. and Sutula, T.P. (1994) Neuronal loss induced in limbic pathways by kindling: evidence for induction of hippocampal sclerosis by repeated brief seizures. J. Neurosci., 14: 3106-3121. Chapman, A.G. (2000) Glutamate and epilepsy. J. Nutr., 130(4S Suppl.): 1043-1045. Chen, J., Sochivko, D., Beck, H., Marechal, D., Wiestler, O.D. and Becker, A.J. (2001) Activity-induced expression of common reference genes in individual CNS neurons. Lab. Invest., 81(6): 913-916. Choi, D.W. (1987) Ionic dependence of glutamate neurotoxicity. J. Neurosci., 7(2): 369-379. Clusmann, H., Stabel, J., Stephens, D.N. and Heinemann, U. (1992) Alterations in medial perforant path and mossy fiber induced field potentials in amygdala and beta-carboline (FG 7142) kindled rats. Neurosci. Lett., 146(1): 65-68. Coulter, D.A. (1999) Chronic epileptogenic cellular alterations in the limbic system after status epilepticus. Epilepsia, 40(Suppl 1): 23-33. Coulter, D.A. (2000) Mossy fiber zinc and temporal lobe epilepsy: pathological association with altered garnmaaminobutyric acid A receptors in dentate granule cells. Epilepsia, 41(Suppl 6): 96-99. Coulter, D.A. (2001) Epilepsy-associated plasticity in gammaaminobutyric acid receptor expression, function, and inhibitory synaptic properties. Int. Rev. Neurobiol., 45(6): 237-252. Coulter, D.A. and DeLorenzo, R.J. (1999) Basic mechanisms of status epilepticus. Adv. Neurol., 79(4): 725-733. Dickens, M., Rogers, J.S., Cavanagh, J., Raitano, A., Xia, Z., Halpern, J.R., Greenberg, M.E., Sawyers, C.L. and Davis, R.J. (1997) A cytoplasmic inhibitor of the JNK signal transduction pathway. Science, 277(5326): 693-696. Duggan, D.J., Bittner, M., Chen, Y., Meltzer, P. and Trent, J.M. (1999) Expression profiling using cDNA microarrays. Nat. Genet., 21(1): 10-14. Ebert, U. and L/~scher, W. (1995) Differences in mossy fibre sprouting during conventional and rapid amygdala kindling of the rat. Neurosci. Lett., 190(3): 199-202. Eberwine, J., Yeh, H., Miyashiro, K., Cao, Y., Nair, S., Finnell, R., Zettel, M. and Coleman, P. (1992) Analysis of gene ex-
pression in single live neurons. Proc. Natl. Acad. Sci. USA, 89(7): 3010-3014. Eisen, M.B., Spellman, ET., Brown, P.O. and Botstein, D. (1998) Cluster analysis and display of genome-wide expression patterns. Proc. Natl. Acad. Sci. USA, 95(25): 14863-14868. Elger, C.E. and Schramm, J. (1993) The surgical treatment of epilepsy. Radiologe, 33(4): 165-171. Fink, L., Seeger, W., Ermert, L., Hanze, J., Stahl, U., Grimminger, F., Kummer, W. and Bohle, R.M. (1998) Real-time quantitative RT-PCR after laser-assisted cell picking. Nat. Med., 4(11): 1329-1333. Gasparini, E, Bruno, V., Battaglia, G., Lukic, S., Leonhardt, T., Inderbitzin, W., Laurie, D., Sommer, B., Varney, M.A., Hess, S.D., Johnson, E.C., Kuhn, R., Urwyler, S., Saner, D., Portet, C., Schmutz, M., Nicoletti, E and Flor, P.J. (1999) (R,S)-4-phosphonophenylglycine, a potent and selective group III metabotropic glutamate receptor agonist, is anticonvulsive and neuroprotective in vivo. J. Pharmacol. Exp. Ther., 289(3): 1678-1687. Geschwind, D.H. (2001) Sharing gene expression data: an array of options. Nat. Rev. Neurosci., 2(6): 435-438. Gibbs, J.W.I., Shumate, M.D. and Coulter, D.A. (1997) Differential epilepsy-associated alterations in postsynaptic GABA(A) receptor function in dentate granule and CA1 neurons. J. Neurophysiol., 77(4): 1924-1938. Gupta, S., Barrett, T., Whitmarsh, A.J., Cavanagh, J., Sluss, H.K., Derijard, B. and Davis, R.J. (1996) Selective interaction of JNK protein kinase isoforms with transcription factors. EMBO J., 15(11): 2760-2770. Henshall, D.C., Clark, R.S., Adelson, P.D., Chen, M., Watkins, S.C. and Simon, R.E (2000) Alterations in bcl-2 and caspase gene family protein expression in human temporal lobe epilepsy. Neurology, 55(2): 250-257. Holmes, K.H., Keele, N.B. and Shinnick-Gallagher, P. (1996) Loss of mGluR-mediated hyperpolarizations and increase in mGluR depolarizations in basolateral amygdala neurons in kindling-induced epilepsy. J. Neurophysiol., 76(4): 28082812. Jackson, G.D., Mclntosh, A.M., Briellmann, R.S. and Berkovic, S.E (1998) Hippocampal sclerosis studied in identical twins. Neurology, 51(1): 78-84. Jacobs, K.M., Graber, K.D., Kharazia, V.N., Parada, 1. and Prince, D.A. (2000) Postlesional epilepsy: the ultimate brain plasticity. Epilepsia, 41(Suppl 6): 153-161. Jaffrey, S.R. and Snyder, S.H. (1996) PIN: an associated protein inhibitor of neuronal nitric oxide synthase. Science, 274(5288): 774-777. Kalviainen, R., Salmenpera, T., Partanen, K., Vainio, P., Riekkinen, E and Pitkanen, A. (1998) Recurrent seizures may cause hippocampal damage in temporal lobe epilepsy. Neurology, 50(5): 1377-1382. Kawasaki, H., Morooka, T., Shimohama, S., Kimura, J., Hirano, T., Gotoh, Y. and Nishida, E. (1997) Activation and involvement of p38 mitogen-activated protein kinase in glutamateinduced apoptosis in rat cerebellar granule cells. J. Biol. Chem., 272(30): 18518-18521. Kim, J.H., Guimareas, P.O., Shen, M.Y., Masukawa, L.M. and
171 Spencer, D.D. (1990) Hippocampal neuronal density in temporal lobe epilepsy with and without gliomas. Acta Neuropathol., 80: 41-45. Knight, J. (2001) When the chips are down. Nature, 410(6831): 860-861. Kullmann, D.M., Asztely, F. and Walker, M.C. (2000) The role of mammalian ionotropic receptors in synaptic plasticity: LTP, LTD and epilepsy. Cell. Mol. Life. Sci., 57(11): 1551-1561. Lahr, G. (2000) RT-PCR from archival single cells is a suitable method to analyze specific gene expression. Lab. Invest., 80(9): 1477-1479. Leist, M., Volbracht, C., Kuhnle, S., Fava, E., Ferrando-May, E. and Nicotera, P. (1997) Caspase-mediated apoptosis in neuronal excitotoxicity triggered by nitric oxide. Mol. Med., 3(11): 750-764. Lie, A.A., Becket, A., Behle, K., Beck, H., Malitschek, B., Conn, P.J., Kuhn, R., Nitsch, R., Plaschke, M., Schramm, J., Elger, C.E., Wiestler, O.D. and BRimcke, I. (2000) Upregulation of the metabotropic glutamate receptor mGluR4 in hippocampal neurons with reduced seizure vulnerability. Ann. Neurol., 47(1): 26-35. Lipshutz, R.J., Fodor, S.P., Gingeras, T.R. and Lockhart, D.J. (1999) High density synthetic oligonucleotide arrays. Nat. Genet., 21(1): 20-24. Liu, W., Rong, Y., Baudry, M. and Schreiber, S.S. (1999) Status epilepticus induces p53 sequence-specific DNA binding in mature rat brain. Mol. Brain Res., 63(2): 248-253. Lockhart, D.J. and Barlow, C. (2001) Expressing what's on your mind: DNA arrays and the brain. Nat. Rev. Neurosci., 2(1): 63-68. Luo, L., Salunga, R.C., Guo, H., Bitmer, A., Joy, K.C., Galindo, J.E., Xiao, H., Rogers, K.E., Wan, J.S., Jackson, M.R. and Erlander, M.G. (1999) Gene expression profiles of laser-captured adjacent neuronal subtypes. Nat. Med., 5(1): 117-122. Lynch, M.W., Rutecki, P.A. and Sutula, T.P. (1996) The effects of seizures on the brain. Curr. Opin. Neurol., 9(2): 97-102. Margerison, J.H. and Corsellis, J.A.N. (1966) Epilepsy and the temporal lobes: A clinical, electroencephalographic and neuropathological study of the brain in epilepsy, with particular reference to the temporal lobes. Brain, 89: 499-530. Martin, J.H., Mohit, A.A. and Miller, C.A. (1996) Developmental expression in the mouse nervous system of the p493F12 SAP kinase. Mol. Brain Res., 35(1-2): 47-57. Mayat, E., Lerner Natoli, M., Rondouin, G., Lebrun, E, Sassetti, I. and Reasens, M. (1994) Kainate-induced status epilepticus leads to a delayed increase in various specific glutamate metabotropic receptor responses in the hippocampus. Brain Res., 645(1-2): 186-200. Meldrum, B.S. (2000) Glutamate as a neurotransmitter in the brain: review of physiology and pathology. J. Nutr., 130(4S Suppl.): 1007-1015. Meldrum, B.S., Akbar, M.T. and Chapman, A.G. (1999) Glutamate receptors and transporters in genetic and acquired models of epilepsy. Epilepsy Res., 36(2-3): 189-204. Mello, L.E., Cavalheiro, E.A., Tan, A.M., Kupfer, W.R., Pretoflus, J.K., Babb, T.L. and Finch, D.M. (1993) Circuit mecha-
nisms of seizures in the pilocarpine model of chronic epilepsy: cell loss and mossy fiber sprouting. Epilepsia, 34(6): 985-995. Merlin, L.R., Bergold, P.J. and Wong, R.K. (1998) Requirement of protein synthesis for group I mGluR-mediated induction of epileptiforrn discharges. J. Neurophysiol., 80(2): 989-993. Mirnics, K. (2001) Microarrays in brain research: the good, the bad and the ugly. Nat. Rev. Neurosci., 2(6): 444-447. Miyajima, K., Tamiya, S., Oda, Y., Adachi, T., Konomoto, T., Toyoshiba, H., Masuda, K. and Tsuneyoshi, M. (2001) Relative quantitation of p53 and MDM2 gene expression in leiomyosarcoma; real-time semi-quantitative reverse transcription-polymerase chain reaction. Cancer Lett., 164(2): 177-188. Montecot, C., Rondi-Reig, L., Springhetti, V., Seylaz, J. and Pinard, E. (1998) Inhibition of neuronal (type 1) nitric oxide synthase prevents hyperaemia and hippocampal lesions resulting from kainate-induced seizures. Neuroseience, 84(3): 791800. Mukhin, A., Fan, L. and Faden, A.I. (1996) Activation of metabotropic glutamate receptor subtype mGluR1 contributes to post-traumatic neuronal injury. J. Neurosci., 16(19): 60126020. Nakagawa, E., Aimi, Y., Yasuhara, O., Tooyama, I., Shimada, M., McGeer, EL. and Kimura, H. (2000) Enhancement of progenitor cell division in the dentate gyrus triggered by initial limbic seizures in rat models of epilepsy. Epilepsia, 41(i): 10-18. Nicoletti, F., Wroblewski, J.T., Alho, H., Eva, C., Fadda, E. and Costa, E. (1987) Lesions of putative glutamatergic pathways potentiate the increase of inositol phospholipid hydrolysis elicited by excitatory amino acids. Brain Res., 436(1): 103-112. Nicoletti, E, Bruno, V., Copani, A., Casabona, G. and Knopfel, T. (1996) Metabotropic glutamate receptors: a new target for the therapy of neurodegenerative disorders?. Trends Neurosci., 19(7): 267-271. Nicotera, P., Ankarcrona, M., Bonfoco, E., Orrenius, S. and Lipton, S.A. (1997) Neuronal necrosis and apoptosis: two distinct events induced by exposure to glutamate or oxidative stress. Adv. Neurol., 72(4): 95-101. O'Leary, D.M., Movsesyan, V., Vicini, S. and Faden, A.I. (2000) Selective mGluR5 antagonists MPEP and SIB-1893 decrease NMDA or glutamate-mediated neuronal toxicity through actions that reflect NMDA receptor antagonism. Br. J. Pharmacol., 131(7): 1429-1437. Phillips, J. and Eberwine, J.H. (1996) Antisense RNA amplification: a linear amplification method for analyzing the mRNA population from single living cells. Methods, 10(3): 283-288. Pitkanen, A., Tuunanen, J., Kalviainen, R., Partanen, K. and Salmenpera, T. (1998) Amygdala damage in experimental and human temporal lobe epilepsy. Epilepsy Res., 32(1-2): 233253. Pretel, S., Applegate, C.D. and Piekut, D. (1997) Apoptotic and necrotic cell death following kindling induced seizures. Acta Histochem., 99(1): 71-79. Rice, A., Rafiq, A., Shapiro, S.M., Jakoi, E.R., Coulter, D.A. and Delorenzo, R.J. (1996) Long-lasting reduction of inhibitory
172
function and gamma- aminobutyric acid type A receptor subunit mRNA expression in a model of temporal lobe epilepsy. Proc. Natl. Acad. Sci. USA, 93(18): 9665-9669. Sakhi, S., Bruce, A., Sun, N., Tocco, G., Baudry, M. and Schreiber, S.S. (1994) p53 induction is associated with neuronal damage in the central nervous system. Proc. Natl. Acad. Sci. USA, 91(16): 7525-7529. Salmenpera, T., Kalviainen, R., Partanen, K. and Pitkanen, A. (1998) Hippocampal damage caused by seizures in temporal lobe epilepsy. Lancet, 351(9095): 35. Sandberg, R., Yasuda, R., Pankratz, D.G., Carter, T.A., Del Rio, J.A., Wodicka, L., Mayford, M., Lockhart, D.J. and Barlow, C. (2000) Regional and strain-specific gene expression mapping in the adult mouse brain. Proc. Natl. Acad. Sci. USA, 97(20): 11038-11043. Schauwecker, EE. (2000) Seizure-induced neuronal death is associated with induction of c-Jun N-terminal kinase and is dependent on genetic background. Brain Res., 884(1-2): 116128. Schauwecker, EE. and Steward, O. (1997) Genetic determinants of susceptibility to excitotoxic cell death: implications for gene targeting approaches. Proe. Natl. Acad. ScL USA, 94(8): 4103-4108. Sch~itze, K. and Lahr, G. (1998) Identification of expressed genes by laser-mediated manipulation of single cells. Nat. Biotechnol., 16(8): 737-772. Schwarzschild, M.A., Cole, R.L. and Hyman, S.E. (1997) Glutamate, but not dopamine, stimulates stress-activated protein kinase and AP-l-mediated transcription in striatal neurons. J. Neurosci., 17(10): 3455-3466. Shigemoto, R., Kinoshita, A., Wada, E., Nomura, S., Ohishi, H., Takada, M., Flor, EJ., Neki, A., Abe, T., Nakanishi, S. and Mizuno, N. (1997) Differential presynaptic localization of metabotropic glutamate receptor subtypes in the rat hippocampus. J. Neurosci., 17(19): 7503-7522. Sommer, W. (1880) Die Erkrankung des Ammonshorns als aetiologisches Moment der Epilepsie. Arch. Psychiat. Nervenkr., 308: 631-675. Sperk, G., Lassmann, H., Baran, H., Kish, S.J., Seitelberger, E and Hornykiewicz, O. (1983) Kainic acid induced seizures: neurochemical and histopathological changes. Neuroscienee, 10(4): 1301-1315. Strasser, U., Lobner, D., Behrens, M.M., Canzoniero, L.M. and Choi, D.W. (1998) Antagonists for group I mGluRs attenuate excitotoxic neuronal death in cortical cultures. Eur J. Neurosci., 10(9): 2848-2855. Sutula, T.P. (2001) Secondary epileptogenesis, kindling, and intractable epilepsy: a reappraisal from the perspective of neural plasticity. Int. Rev. Neurobiol., 45: 355-386. Tsirka, S.E., Gualandris, A., Amaral, D.G. and Strickland, S. (1995) Excitotoxin-induced neuronal degeneration and seizure are mediated by tissue plasminogen activator. Nature, 377(6547): 340-344. Tuunanen, J. and Pitk~inen, A. (2000) Do seizures cause neuronal damage in rat amygdala kindling?. Epilepsy Res., 39(2): 171176. Tuunanen, J., Halonen, T. and Pitk~inen, A. (1997) Decrease in
somatostatin-immunoreactive neurons in the rat amygdaloid complex in a kindling model of temporal lobe epilepsy. Epilepsy Res., 26(2): 315-327. Tuunanen, J., Lukasiuk, K., Halonen, T. and Pitkanen, A. (1999) Status epilepticus-induced neuronal damage in the rat amygdaloid complex: distribution, time-course and mechanisms. Neuroscience, 94(2): 473-495. Van Lookeren Campagne, M., Lucassen, P.J., Vermeulen, J.P. and Balazs, R. (1995) NMDA and kainate induce internucleosomal DNA cleavage associated with both apoptotic and necrotic cell death in the neonatal rat brain. Eur. J. Neurosci., 7(7): 16271640. Van Roost, D., Solymosi, L., Schramm, J., van Oosterwyck, B. and Elger, C.E. (1998) Depth electrode implantation in the length axis of the hippocampus for the presurgical evaluation of medial temporal lobe epilepsy: a computed tomographybased stereotactic insertion technique and its accuracy. Neurosurgery, 43(4): 819-826. Venero, J.L., Revuelta, M., Machado, A. and Cano, J. (1999) Delayed apoptotic pyramidal cell death in CA4 and CA1 hippocampal subfields after a single intraseptal injection of kainate. Neuroscience, 94(4): 1071-1081. Waha, A., Watzka, M., Koch, A., Pietsch, T., Przkora, R., Peters, N., Wiestler, O.D. and von Deimling, A. (1998) A rapid and sensitive protocol for competitive reverse transcriptase (cRT) PCR analysis of cellular genes. Brain Pathol., 8(1): 13-18. Wahlestedt, C., Golanov, E., Yamamoto, S., Yee, E, Ericson, H., Yoo, H., Inturrisi, C.E. and Reis, D.J. (1993) Antisense ofigodeoxynucleotides to NMDA-R1 receptor channel protect cortical neurons from excitotoxicity and reduce focal ischaemic infarctions. Nature, 363(6426): 260-263. Wang, E., Miller, L.D., Ohnmacht, G.A., Liu, E.T. and Marincola, F.M. (2000) High-fidelity mRNA amplification for gene profiling. Nat. Biotechnol., 18(4): 457-459. Wellmann, A., Thieblemont, C., Pittaluga, S., Sakai, A., Jaffe, E.S., Siebert, P. and Raffeld, M. (2000) Detection of differentially expressed genes in lymphomas using cDNA arrays: identification of clusterin as a new diagnostic marker for anaplastic large-cell lymphomas. Blood, 96(2): 398-404. Wen, X., Fuhrman, S., Michaels, G.S., Carr, D.B., Smith, S., Barker, J.L. and Somogyi, R. (1998) Large-scale temporal gene expression mapping of central nervous system development. Proc. Natl. Acad. Sci. USA, 95(1): 334-339. Wolf, H.K. and Wiestler, O.D. (1993) Surgical pathology of chronic epileptic seizure disorders. Brain Pathol., 3(4): 371380. Wolf, H.K., Zentner, J., Hufnagel, A., Campos, M.G., Schramm, J., Elger, C.E. and Wiestler, O.D. (1994) Morphological findings in temporal lobe epilepsy: experience with 216 consecutive surgical specimens. Verh. Dtsch. Ges. Pathol., 78: 438442. Yang, D.D., Kuan, C.Y., Whitmarsh, A.J., Rincon, M., Zheng, T.S., Davis, R.J., Rakic, P. and Flavell, R.A. (1997) Absence of excitotoxicity-induced apoptosis in the hippocampus of mice lacking the Jnk3 gene. Nature, 389(6653): 865-870. Yilmazer-Hanke, D.M., Wolf, H.K., Schramm, J., Elger, C.E., Wiestler, O.D. and Bliimcke, I. (2000) Subregional pathology
173
of the amygdala complex and entorhinal region in surgical specimens from patients with pharmacoresistant temporal lobe epilepsy. J. Neuropathol. Exp. Neurol., 59(10): 907-920. Zentner, J., Hufnagel, A., Wolf, H.K., Ostertun, B., Behrens, E,, Campos, M.G., Solymosi, L., Elger, C.E., Wiestler, O.D.
and Schramm, J. (1995) Surgical treatment of temporal lobe epilepsy: clinical, radiological, and histopathological findings in 178 patients. J. Neurol. Neurosurg. Psychiat~, 58(6): 666673.
T. Sutula and A. Pitk~inen (Eds.) Progress in Brain Research, Vol. 135 © 2002 Elsevier Science B.V. All rights reserved
CHAPTER 15
What synaptic lipid signaling tells us about seizure-induced damage and epileptogenesis Nicolas G. Bazan *, Bin Tu and Elena B. Rodriguez de Turco Neuroscience Center of Excellence and Department of Ophthalmology, Louisiana State University Health Sciences Center, New Orleans, LA 70112, USA
Abstract: Glutamate, the most abundant excitatory neurotransmitter in the mammalian CNS, plays a central role in many neuronal functions, such as long-term potentiation, which is necessary for learning and memory formation. The fast excitatory glutamate neurotransmission is mediated by ionotropic receptors that include AMPA/kainate and N-methyl-Daspartate (NMDA) receptors, while the slow glutamate responses are mediated through its interaction with metabotropic receptors (mGluRs) coupled to G-proteins. During seizures, massive release of glutamate underlies excitotoxic neuronal damage as it triggers an overflow of calcium in postsynaptic neurons mediated by NMDA-gated channels. The early upstream postsynaptic events involve the activation of phospholipases, with the release of membrane-derived signaling molecules, such as free arachidonic acid (AA), eicosanoids, and platelet-activating factor (PAF). These bioactive lipids modulate the early neuronal responses to stimulation as they affect the activities of ion channels, receptors, and enzymes; and when released into the extracellular space, they can contribute to the modulation of presynaptic neurotransmitter release/re-uptake, and/or affect other neighboring neuronal/glial cells. The downstream postsynaptic events target the nucleus, leading to activation of gene-expression cascades. Syntheses of new proteins are the basis for seizure-induced sustained physiological and/or pathological changes that occur hours, days, or months later, such as synaptic reorganization and repair, and apoptotic/necrotic neuronal death. The intricate mesh of signaling pathways converging to the nucleus, and connecting upstream to downstream synaptic events, are at present the focus of many research efforts. We describe in this chapter how seizure-induced glutamate release activates the hydrolysis of membrane AA-phospholipids via phospholipase A2 (PLA2), PLC, and PLD, thus releasing bioactive lipids that, in turn, modulate neurotransmission. We discuss mechanisms through which lipid messengers, such as AA and PAF, may turn into injury mediators participating in seizure-induced brain damage.
Introduction
Arachidonoyl phospholipids (AA-PLs) and docosahexaenoyl-PLs (DHA-PLs) are highly unsaturated lipid components of neuronal membranes that provide fluidity and the proper environment for active
* Correspondence to: N,G. Bazan, LSU Health Sciences Center, Neuroscience Center of Excellence, 2020 Gravier Street, Suite D, New Orleans, LA 70112, USA. Tel.: +1504-599-0832; Fax: +1-504-568-5801; E-mail:
[email protected]
integral protein functions (i.e., receptors, ion channels, enzymes) (Stubbs and Smith, 1984; Spector and Yorek, 1985; Yeagle, 1989). AA- (and DHA-) PLs play a highly dynamic role in cellular function as a reservoir of messengers for agonists such as neurotransmitters, growth factors, and cytokines that, by interacting with plasma-membrane receptors, modulate phospholipase activity, thus switching on intracellular signaling pathways by releasing membrane PL-derived second messengers (Fig. 1). Phospholipase A2 (PLA2) hydrolyzes AA-PLs, releasing AA; PLC generates AA-diacylglycerols (DAG) from polyphosphoinositides (PPI); and PLD preferentially
176 Membrane Substrates
N-AA-PE
I I
[
Phospholipases
Bioactive Products
I
PC
PPI
AA-PLs
AlkyI-AA-PC Oxidized Pl.s
I
PLD
.Anandamide (A.EA)
•PA ~ I
.DAG ~
l
.lsoprostanes .Neuroprostanes
.AA.4"~
..,o,,
I
;®
COX-2 Gene Expression
. ~ oAA-LysoPA4=.~ .2-AG •12-OH-AEA -PGE2-ethanolamide
V Cellular Responses
°HPTES •HETES •LTs
-Protaglandins -Prostacyclin .Thromboxanes
Neurogenesis - S y n a p t o g e n e s i s ~> A p o p t o s i s
Fig. 1. Cascade of bioactive metabolites triggered by activation of PLAz, PLC, and PLD that are involved in cellular responses to stimulation. Closed circlesindicatebioactivelipids including IP3. For details, see text. targets phosphatidylcholine (PC), releasing phosphatidic acid (PA). The excitatory neurotransmitter glutamate, which is involved in the induction of long-term potentiation (LTP, a synaptic model of learning and memory) (Bliss and Collingridge, 1993; Nakanishi, 1994), activates PLA2 and PLC, generating AA, PAR and DAG (Fig. 1), bioactive lipids that are implicated in neuronal plasticity. The same signaling pathways that contribute to synaptic plasticity, when they are overstimulated during seizures, ischemia, trauma, and neurodegeneration, lead to excitotoxic brain damage (Bazan et al., 1995; Bazan and Allan, 1998). The goals of this chapter are: (1) to give an overview of the different signaling pathways that, activated at the plasma membrane by glutamate, contribute to bioactive lipid generation; and (2) to show how different avenues and pathways of the cascade may underlie plasticity changes and/or neuronal injury in epilepsy.
Metabotropic glutamate receptors and PLC-mediated DAG signaling Glutamate neurotransmission is mediated by different types and subtypes of receptors located at the pre- and postsynapse and in glial cells. The glutamate ionotropic receptors (iGluRs), AMPA/kainate, and NMDA, mediate the fast excitatory neurotransmission, while the slow glutamate responses are mediated by metabotropic receptors (mGluRs) coupled to G-proteins (for review, see Nakanishi, 1994; Conn and Pin, 1997). The mGluRs are classified into three groups: group I mGluRs includes mGluR1 and mGluR5, which are linked to phosphatidylinositol 4,5-bis-phosphate (PIP2)-PLC pathway activation; and Groups II and III mGluRs are negatively coupled to adenylate cyclase. Their distribution varies among different neuronal populations, with NMDA receptors located postsynaptically, while mGluRs subtypes are at the pre- and postsynapse. Glial cell response to glutamate is mediated by AMPA and
177 mGluR5 receptors (Conn and Pin, 1997). All these widely distributed types of glutamate receptors contribute to and modulate glutamate neurotransmission. NMDA-gated calcium channels are central players in excitatory glutamate neurotransmission facilitated by post-synaptic group I mGluR. Activation of group I mGluRs coupled to PLC during seizures leads to a rapid accumulation of AA-DAG in the brain (Bazan et al., 1995). This is a short-lived signal, since the released AA-DAG activates PKC, which, in turn, contributes to feedback inhibition of the PLC pathway (Nishizuka, 1995). However, the mGluR-PLC pathway elicits potent and sustained consequences in other signaling pathways, since PKC activates PLD and PLA2. DAG kinase epsilon (DGKs) selectively phosphorylates AA-DAG to generate AA-phosphatidic acid (PA), thereby shutting off the DAG signal (Tang et al., 1996; Pettitt and Wakelam, 1999). Activation of group I mGluR and PKC¥ is involved in synaptic plasticity, such as in learning, memory, and LTP (Abeliovich et al., 1993a,b; Aiba et al., 1994a; Conquet et al., 1994; Nakanishi, 1994; Wilsch et al., 1998), and long-term depression (Aiba et al., 1994b). Moreover, alterations of this signaling pathway have been implicated in neurological and psychiatric diseases, such as epilepsy, Alzheimer's, and depression (Bazan et al., 1995; Pacheco and Jope, 1996; Conn and Pin, 1997; Bordi and Ugolini, 1999). The central role played by mGluR signaling in glutamate neurotransmission was revealed by studies in mice with targeted disruption of the DGK~ gene (Rodriguez de Turco et al., 2001). DGKe-/mice display higher resistance to seizures induced by electroconvulsive shock (ECS) and attenuation of LTP in the hippocampus. The genetic background of these DGK~-/- mice (generated from 129Ola-type ES introduced into blastocyst-stage embryos from C57BL/6 mice, followed by intercrossing of the heterozygous null-mutant mice with BL6) may contribute to the observed phenotypic changes (Gerlai, 1996). However, the magnitude of seizure susceptibility displayed by heterozygous ( + / - ) mice was intermediate between wild-type ( + / + ) and DGKs-/mice, indicating that background genes were not responsible for the DGKs-/- mouse response to ECS. Nevertheless, other adaptive changes may occur as a consequence of DGKe targeting, such as overex-
pression of other DGK enzymes to take over the functional role of DGK~. Interestingly, not only the PPI-PLC pathway was greatly affected by the mutation, but also the cPLA2-AA and the PLD-DAG pathways, reflecting the direct impact of the former in modulating multiple signaling pathways essential for synaptic activity and neuronal plasticity.
Activation of phospholipase A2 triggered by seizures Seizures trigger an early activation of synaptic PLA2, reflected in a rapid accumulation of free fatty acids (FFA) (Bazan, 1970; Bazan et al., 1993). However, the detailed events involved in the activation of neuronal and/or glial phospholipases under physiological and pathophysiological conditions are still not fully understood. Furthermore, PLA2 is a large family of enzymes classified into three types: cytosolic calcium-dependent PLA2 (cPLA2; type IV), calcium-independent (iPLA2; type VI) and low-molecular weight, secretory PLAzs (sPLAzs) (Balsinde et al., 1999). The cPLA2 displays high selectivity for AA-PLs, thus activating the AA cascade and the generation of eicosanoids (Clark et al., 1991). There is a general consensus that cPLA2 is the enzyme involved in signaling, and which is activated by agonists that either trigger increased calcium influx or that stimulate calcium mobilization from the intracellular stores pathway (Clark et al., 1991; Nicotera et al., 1992). Increased postsynaptic calcium permeation through channels gated by NMDA-glutamate receptor leads to activation of cPLA2 and AA release in primary cultures of striatal neurons and cerebellar granule cells (Dumuis et al., 1988), and in hippocampal slices (Pellerin and Wolfe, 1990), and is blocked by PLA2 inhibitors (Sanfeliu et al., 1990). Also, cPLA2 can be activated by glutamate interaction with group I mGluRs coupled to G-proteins and PLC, which promotes calcium mobilization from intracellular stores (Conn and Pin, 1997). The release of AA triggered by the mGluRs plays a central role in hippocampal LTP (Izumi et al., 2000). Activation of calcium-dependent cPLA2 also involves its phosphorylation by mitogen-activated protein kinase (MAPK) and its translocation from the cytosol to the nuclear membrane and endoplasmic
178 reticulum, where AA-PLs are hydrolyzed (Clark et al., 1991; Peters-Golden et al., 1996; Leslie, 1997; Hirabayashi et al., 1999). Its full activation depends upon the duration of cytosolic Ca 2+ elevation (Hirabayashi et al., 1999). The glutamate-NMDA pathway activates MAPK in hippocampal neurons (Kurino et al., 1995), and PAF, a potent second messenger generated by the glutamate-NMDA-PLA2 pathway, is used as the messenger (Mukherjee et al., 1999). Activation of PLA2 by glutamate opens the window for a cascade of second messengers, such as AA, eicosanoids, and PAF (as summarized in Fig. 1), which are directly involved in the modulation of excitotoxic neurotransmission, as is discussed in the following sections.
Arachidonic acid and eicosanoid signaling Free AA is a potent signaling molecule that is maintained under normal physiological conditions at very low levels and which, upon neuronal stimulation, is released from membrane AA-PLs by cPLA2 (Piomelli, 1993; Bazan et al., 1995). PLC and PLD signaling pathways also contribute to the free AA pool used for the synthesis of the biologically active prostaglandins, leukotrienes, thromboxanes, and hydroxyeicosatrienoic acids (Fig. 1). Multiple proteins are targeted by AA at the synapse, leading to a positive modulation of glutamatergic neurotransmission. Postsynaptically, AA increases NMDA open-channel probability (Miller et al., 1992) and presynaptically glutamate and AA interaction with mGluRs autoreceptors increases glutamate release (Corm and Pin, 1997). The levels of glutamate at the synapse are tightly controlled by glutamate transporters present in glia and neuronal cells (Vesce et al., 1999; Danbolt, 2001). AA, by blocking glutamate re-uptake, elevates its synaptic concentration, favoring excitotoxicity (Volterra et al., 1992; Vesce et al., 1999). Glial cells are actively involved in glutamate neurotransmission. They play a central role in the uptake of glutamate at the synapse, which is then metabolized to glutamine and delivered back to the neuronal cell, thus fueling glutamate synthesis (Vesce et al., 1999). But glial cells can also release glutamate into the synapse, when presynaptically released glutamate interacts with glial type I
mGluRs and activates the cPLA2-AA cascade (Conn and Pin, 1997). The number of sites in glutamatergic neurotransmission under modulation through the cPLA2 pathway is greatly enlarged when AA is metabolized to eicosanoids, which are potent modulators of synaptic transmission, glial function, and cerebrovasculature properties (Piomelli and Greengard, 1990; Shimizu and Wolfe, 1990; Piomelli, 1994). The cyclooxygenase (COX) enzymes catalyze the cyclooxygenation of AA to PGG2, which then undergoes hydroperoxidation to PGH2, the substrate for the synthesis of biologically active prostaglandins and thromboxanes (Vane et al., 1998). There are two forms of COX enzymes: COX-l, which is constitutively expressed in all tissues, and COX-2, which is expressed in response to inflammatory signals (Vane et al., 1998). The brain is one of the few organs where the COX2 enzyme is constitutively expressed, mainly in the cortex and hippocampus (Yamagata et al., 1993; Vane et al., 1998), and consistently localized in neuronal dendritic spines actively involved in neurotransmission (Kaufmann et al., 1996). The expression of the COX-2 gene in rat brain is dynamically regulated by the NMDA receptordependent synaptic activity that is implicated in both LTP and exeitotoxic neuronal damage (Yamagata et al., 1993). The most abundant prostaglandins generated in the brain through COX-I/COX-2 pathways, PGE2, PGF2~, and PGD, are modulators of synaptic activity as they exert paracrine functions through pre- and postsynaptic receptors as well as autocrine roles as intraneuronal second messengers (Piomelli, 1994). COX-2, but not COX-l, can signal to the nucleus as it is translocated to the nuclear membrane, generating prostanoids involved in gene expression (Morita et al., 1995; Vane et al., 1998). PGE2 interacts with a nuclear EP1 receptor, and its activation leads to calcium mobilization and gene transcription (Bhattacharya et al., 1998). Thus, the NMDACOX-2-PGE2-EP1 receptor-gene expression pathway may lead to long-lasting changes related to synaptic plasticity. The lipoxygenase (LOX)-mediated oxygenation of AA opens a new cascade of bioactive metabolites involved in the modulation of synaptic activity (Piomelli and Greengard, 1990; Shimizu and Wolfe, 1990). These include hydroperoxy-eicosatetraenoic
179 acids (HPETEs), which are further hydrolyzed to HETEs. The 5- and 12-LOX pathways, which are the most active in the brain and retina (for review see Birkle and Bazan, 1986; Piomelli, 1994), generate (in the case of 5-LOX) leukotrienes (LTB4), sulfidopeptide LTC4, LTD4, LTE4, and lipoxins (Piomelli and Greengard, 1990), and in the case of 12-LOX, 12S-HPETE, 12-HETES, and hepoxilins (HxA3 and HxB3; Piomelli and Greengard, 1990; Pace-Asciak, 1994). Interestingly, metabolites generated through the 12-LOX pathway (12-HPETE/12-HETE/Hx) inhibit, at the presynaptic level, neurotransmitter release (Piomelli and Greengard, 1990), including glutamate release from hippocampal mossy-fiber terminals (Freeman et al., 1991).
PAF signaling: intracellular and extracellular targets PAF (1-O-alkyl-2-acetyl-glycero-3-phosphocholine) is a potent lipid mediator that is actively involved in glutamatergic neurotransmission, both under physiological and pathophysiological conditions (Bazan and Allan, 1998; Prescott et al., 2000). Stimulation of NMDA receptors activates PAF synthesis and PAF is produced in brain in response to seizures and ischemia (Kumar et al., 1988; Nishida and Markey, 1996). Glutamate-NMDA-induced calcium increase and activation of cPLA2 in postsynaptic neurons activates PAF synthesis through 'the remodeling pathway' (Bazan and Rodriguez de Turco, 1995). Alkyl-AA-glycerophosphorylcholine is acted upon by cPLA2, releasing AA and lyso-PAF, which is then acetylated by lyso-PAF acetyltransferase, generating PAF (Fig. 1). PAF action is rapidly terminated by PAF-acetylhydrolase (PAF-AH) (Bazan, 1995). This enzyme can also hydrolyze certain species of oxidatively damaged phospholipids that are generated during brain oxidative stress, a component of seizures, ischemia-reperfusion, neurotrauma, and neurodegenerative diseases (Prescott et al., 2000). These PAF-like species possess PAF-like activity at the PAF receptor (Prescott et al., 2000). PAF actions are mediated by its interaction with extracellular receptors that are members of the seven membrane-spanning domain, G protein-coupled receptor superfamily (for review, see Prescott et al., 2000), and which are present in neuronal and glial
cells, with a very high expression in microglia (Mori et al., 1996). In addition to the PAF extracellular receptor, two other different types of PAF-binding sites have been found in brain synaptosomal membranes and in microsomal membranes. These extraand intracellular binding sites can be pharmacologically differentiated, since the former is inhibited by the synthetic hetrazepine BN 52021, and the latter by BN 50730, a terpenoid extracted from the leaf of the Ginkgo biloba tree (Marcheselli et al., 1988; Marcheselli and Bazan, 1990; Marcheselli and Bazan, 1994). Presynaptic receptor-mediated PAF actions potentiate glutamatergic transmission, and downstream of postsynaptic NMDA receptors, PAF is the messenger of glutamate actions that lead to gene expression and neuronal plasticity (Bazan, 1998). Postsynaptic neuronal activity activates PAF synthesis, which in turn acts as a retrograde messenger, stimulating presynaptic glutamate release (Clark et al., 1992), thus contributing to LTP (Wieraszko et al., 1993; Kato et al., 1994) and memory formation (Jerusalinsky et al., 1994: Izquierdo et al., 1995; Bazan, 1998). In fact, animals deficient in the PAF receptor displayed attenuated LTP (Chen et al., 2001). PAF-mediated effects through the BN 50730-sensitive intracellular receptor are linked with PAF-mediated effects on gene expression in the CNS (Bazan, 1998). Electroconvulsive shock and kainic acid-induced seizures activate early gene expression in the hippocampus, including expression of the inducible COX-2, and this induction is inhibited by BN 50730, but not by BN 52021 (Marcheselli and Bazan, 1994, 1996). Glutamate triggers the activation of the MAPK signaling pathway (Bading and Greenberg, 1991; Fiore et al., 1993; Kurino et al., 1995), and PAF is the mediator of glutamate-NMDA activation of JNK, p38, and ERK MAPK (Mukherjee et al., 1999). The MAPK cascade is also activated in the hippocampus during kainic acid-induced seizures (Kim et al., 1994) and transient cerebral ischemia (Hu and Wieloch, 1994). Because PAF-MAPK activation may stimulate cPLA2 (Clark et al., 1991) and also may be upstream of PAF-induced COX-2 expression, both effects may converge in potentiating the AA cascade and eicosanoid synthesis. Finally, PAF can be the trigger for the previously mentioned COX-2-PGE2-EP1 receptor pathway at
180 the nuclear level, leading to a late gene-transcription activation.
Can lipid mediators contribute to neuronal death? Glutamate signaling has dual properties: under physiological conditions it is involved in synaptic plasticity, learning, and memory, and in pathological conditions, when high and sustained levels of glutamate are present at the synapse, it triggers excitotoxic neuronal damage (Bazan et al., 1995). It is logical to argue that the same signaling pathways, including those regulated by phospholipases, may be the ones used by glutamatergic neurotransmission to reach the two opposite ends of the spectrum: plasticity changes and cell survival or cell death. One of the central pathways contributing to glutamate excitotoxicity is the calcium-mediated activation of cPLA2. Its activation during seizures and ischemia contributes to neuronal injury and degeneration (Bazan et al., 1993; Bonventre, 1996). In fact, cPLA2 knockout mice are more resistant to brain ischemic insult (Bonventre et al., 1997). One direct consequence of cPLA2 overstimulation is the generation of free AA and lyso-phospholipids and perturbations in membrane structure affecting the activity of receptors, ion channels, enzymes, and cytoskeletal proteins (Fig. 2). In addition, the viability of free AA is the rate-limiting step for the COX1/COX-2-eicosanoid pathways, with the release of free radicals and potential mediators of peroxidative membrane damage (Bazan et al., 1995). The cPLA2-mediated PAF synthesis and its transcriptional activation of COX-2 are activated during seizures and ischemia, and its expression precedes apoptotic neuronal death (Yamagata et al., 1993; Adams et al., 1996; Kaufmann et al., 1996; Marcheselli and Bazan, 1996; Miettinen et al., 1997; Tocco et al., 1997). COX-2 overexpression is involved in NMDA-induced neuronal cell death (Hewett et al., 2000). Several lines of experimental evidence support the central role of PAF-COX-2 in excitotoxic damage: (1) transgenic mice overexpressing COX-2 are more susceptible to neuronal excitotoxic damage (Kelley et al., 1999); (2) COX-2 inhibition prevents ischemia and NMDA-induced cell death (Nakayama et al., 1998; Hewett et al., 2000); (3)
NMDA neurotoxicity is reduced in COX-2-deficient mice (Iadecola et al., 2001); (4) the antagonist of the PAF intracellular binding site, BN50730, inhibits seizure-induced COX-2 expression (Marcheselli and Bazan, 1996) and greatly reduces seizure-induced hippocampal damage (Marcheselli and Bazan, unpublished observations); (5) PAF-AH attenuates NMDA-induced hippocampal neuronal apoptosis (Ogden et al., 1998); (6) seizure-induced COX-2 induction occurs in those areas with the highest neuronal damage, i.e., hippocampus > cortex (Marcheselli and Bazan, 1996); (7) the inflammatory cytokine 1L-l[3, whose expression is increased in the brain during seizures, contributes to neuronal damage by activating PAF-COX-2 signaling (Serou et al., 1999). All the above observations strongly support the involvement of cPLA2 activation in excitotoxic pathways mediated by PAF-COX-2 induction leading to neuronal death. We have recently reported that PAE by direct interaction with the mitocbondria, activates the opening of the transition pore and cytochrome c release (Parker et al., in press). This newly identified target of PAF action leads, through caspase activation, to apoptotic neuronal death (Fig. 2). Mitochondrial dysfunction and oxidative stress play a central role in neurodegenerative diseases (Beal, 1998) and may also, during seizures, severely compromise neuronal function. Thus, the reported NMDA-induced alteration of mitochondrial function and free-radical generation (Dugan et al., 1995) may be mediated by activation of the PLA2-PAF pathway.
Kindling: COX-2 and cPLA2 induction in the cortex and hippocampus Kindling is a widely used model of human temporallobe epilepsy that is characterized by hippocampal cell death and sclerosis, and also by plasticity changes, including sprouting of mossy fibers, synaptic reorganization, and cell proliferation in the dentate gyrus (Cavazos et al., 1994; Bengzon et al., 1997). The mechanisms that contribute to this aberrant synaptic plasticity and epileptogenesis are not defined but could be the ones that set in motion enhanced excitability contributing to seizure development and propagation. The PAF-COX-2 pathway that is activated during seizures (Marcheselli
181
tcPLAa
Lyso PLs i FreeAA ~ Peroxidation I Alterationsin: I • Receptors ~ , ~ _ _ _ _ ~ Free Radicals ] • Ion Channels Pimp F 'C O X ' ~ ids -j • CytoskeletalProteins
1 l
/'kProtein COX-2 mRNA ~Kinases _ ~ \ ~'~ Transcription ~\ Factors
Excitotoxicity
~k~Gi(~uPresy;Tetl: ase "~ ~ Opening
PermeabilityTransitionPore + Cytochromec Release Caspases
Ib
t
tPLC
~
DAG IP3 -+ t[Ca2*]il Actv iatoi nSUStani edpKcP / LA2
Dysregulationof signaling modulated by PKC
Fig. 2. Involvement of phospholipid-derived second messengers in neuronal excitotoxic damage.
and Bazan, 1996) may contribute to epileptogenesis (Bazan and Serou, 1999). To explore the correlation between this signaling pathway and the progression of excitability, the induction of cPLA2 and COX-2 during kindling was followed in the hippocampus and the cortex (Tu and Bazan, submitted). Rats were treated with subconvulsive electrical stimulation at 30-min intervals 12 times daily for 4 days (Fig. 3). Seizure behaviors gradually increase, reaching class 5 by the scale of Racine (1972) by days 3 and 4. One month later, maximal seizure response was obtained using the same low stimulation, indicating that neuronal circuitries involved in seizures were permanently modified by the kindling protocol. In both hippocampi, COX-2 messenger RNA increased by 13-fold after the first day of treatment, decreased in-between daily stimulations, and reached the highest increase by day 3 (20-fold), when animals reached class-5 seizures. Interestingly, in the
cortex, no changes in COX-2 mRNA were detected during the first 2 days post-seizure, but 2 h prior to day-3 stimulation it was increased 12-fold, and then it reached a 29-fold increase after seizure. The profile of cPLA2 induction was of much lower magnitude than that of COX-2, with a sustained increase from day 1 in the hippocampus and from day 3 in the cortex. The early increase of COX-2 in the hippocampus could contribute to neuronal damage in this highly susceptible area of the brain, while the profile of changes by day 3 in both the cortex and hippocampus suggest the involvement of COX-2 signaling in the maturation of epileptogenesis and in evoking epileptic activity in the motor cortex. The spreading of COX-2 and cPLA2 gene induction from the hippocampus to the cortex during kindling may result from permanently modified synaptic networks, as new connections are generated and/or old ones are modified by seizure activity.
182
+
+
+
+
30
-5
25
,. ,. ,. . ., ., . .
:.:.2":.>2
.....'.'.'.
2:::::::2:2
:::::::::::
:::..::
;:::::::::~,
:::::::::::
iiii!iiii
20 15 -
iii
~
!iiiiil
4
8
3m
RHI COX 2 _
LH~
"
2.~,
10
>, 5
'o
_
i
2" "
"~ f~ 0
l
I:i:!:!:i:i:
t~ ,,Q
1
.!iliiii!iii " LH
'
l
~
t
::::::::~:
t
Day 1
m
' I
Day 2
=
~
I::::::::::
I
cPLA2
~:~. . . . . . . ,,,,.%. ::::::::::~
I
0
I
Day 4
Day 3
o
4
20
i i i i !i !i'iiii!i i i;": iiiiii : i i i i i) ~i ~li!))i))i ~~LC}'!)i)))))) i: i))))))Re ))))COX.2
2.--.o
10 5 -
0
O
; i E i i ~ i:i:]}:~i:i : : : i : i " : i : i I :::::::::::
10
I
I
20
30
::::~:::~:
40
I
I::::::::::
I
50 60 70 Time(houm)
..:.':':::: LCJ ,
I
:::::::;:;:1
80
90
I
0
100
Fig. 3. Early induction of COX-2 and cPLA2 mRNA in the hippocampus and delayed changes in the cortex during kindling epileptogenesis.
Conclusion Do seizures damage the brain? And are lipid mediators involved in neuronal damage? Many times, structural damage and death of neurons may occur as a consequence of a disarray and imbalance of neuronal signaling pathways activated by excessive excitatory or deficient inhibitory stimulation. Other times, seizures do not result in cell death, but in functional damage as a consequence of aberrant signaling, with overstimulation of membrane PL hydrolysis, the source of messengers underlying long-lasting and progressive neuronal dysfunctions. Synaptic activity correlates with the degree of phospholipase activation and the release of the potent lipid mediator PAF, which mediates LTP, plasticity changes, and, on
the other side of the coin, when overproduced, controis multiple downstream neuronal events involved in apoptotic cell death.
Acknowledgements The authors' research described here was supported by NS23002 (NIH).
References Abeliovich, A., Chen, C., Goda, Y., Silva, A.J., Stevens, C.F. and Tonegawa, S. (1993a) Modified hippocampal long-term potentiation in PKC gamma-mutant mice. Cell, 75: 12531262. Abeliovich, A., Paylor, R., Chen, C., Kim, J.J., Wehner, J.M.M. and Tonegawa, S. (1993b) PKC gamma mutant mice exhibit
183
mild deficits in spatial and contextual learning. Cell, 75: 12631271. Adams, J., Colla~o-Moraes, Y. and de Belleroche, J. (1996) Cyclooxygenase-2 induction in cerebral cortex: an intracellular response to synaptic excitation. J. Neurochem., 66: 6-13. Aiba, A., Chen, C., Herrup, K., Rosenmund, C., Steven, C.E and Tonegawa, S. (1994a) Reduced hippocampal long-term potentiation and context-specific deficit in associative learning in mGluR1 mutant mice. Cell, 79: 365-375. Aiba, A., Kano, M., Chen, C., Stanton, M.E., Fox, G.D., Herrup, K., Zwingman, T.A. and Tonegawa, S. (1994b) Deficient cerebellar long-term depression and impaired motor learning in mGluR1 mutant mice. Cell, 79: 377-388. Bading, H. and Greenberg, M.E. (1991) Stimulation of protein tyrosine phosphorylation by NMDA receptor activation. Science, 253: 912-914. Balsinde, J., Balboa, M.A., Insel, P.A. and Dennis, E.A. (1999) Regulation and inhibition of phospholipase A2. Annu. Rev. Pharmacol. Toxicol., 39: 175-189. Bazan, N.G. (1970) Effects of ischemia and electroconvulsive shock on free fatty acid pool in the brain. Biochim. Biophys. Acta, 218: 1-10. Bazan, N.G. (1995) A signal terminator. Nature, 374: 501-502. Bazan, N.G. (1998) Bioactive lipids and gene expression in neuronal plasticity. Adv. Exp. Med. Biol., 446: 37-49. Bazan, N.G. and Allan, G. (1998) Platelet-activating factor and other bioactive lipids. In: M.D. Ginsberg and J. Bogousslavsky (Eds.), Cerebrovascular Disease, Pathophysiology, Diagnosis and Management. Blackwell Science, Malden, MA, pp. 532555. Bazan, N.G. and Rodriguez de Turco, E.B. (1995) Plateletactivating factor is a synapse messenger and a modulator of gene expression in the nervous system. Neurochem. Int., 26: 435-441. Bazan, N.G. and Serou, M.J. (1999) Second messengers, longterm potentiation, gene expression and epileptogenesis. In: A.V. Delgado-Escueta, W.A. Wilson, R.W. Olsen and R.J. Porter (Eds.), Jasper's Basic Mechanisms of the Epilepsies, 3rd ed. Advances in Neurology, Vol 79. Lippincott, Williams and Wilkins, Philadelphia, PA, pp. 659-664. Bazan, N.G., Allan, G. and Rodriguez de Turco, E.B. (1993) Role of phospholipase A2 and membrane-derived lipid second messengers in excitable membrane function and transcriptional activation of genes: implications in cerebral ischemia and neuronal excitability. Prog. Brain Res., 96: 247-257. Bazan, N.G., Rodriguez de Turco, E.B. and Allan, G. (1995) Mediators of injury in neurotrauma: intracellular signal transduction and gene expression. J. Neurotrauma, 12: 791-814. Beal, M.F. (1998) Mitochondfial dysfunction in neurodegenerative diseases. Biochim. Biophys. Acta, 1366:211-223. Bengzon, J., Kokaia, Z., Elmer, E., Nanobashvili, A., Kokaia, M. and Lindvall, O. (1997) Apoptosis and proliferation of dentate gyms neurons after single and intermittent limbic seizures. Proc. Natl. Acad. Sci. USA, 94: 10432-10437. Bhattacharya, M., Peri, K.G., Almazan, G., Ribeiro-da-Silva, A., Shichi, H., Durocher, Y., Abramovitz, M., Hou, X., Varma, D.R, and Chemtob, S. (1998) Nuclear localization
of prostaglandin E2 receptors. Proc. Natl. Acad. Sci. USA, 95: 15792-15797. Birkle, D.L. and Bazan, N.G. (1986) The arachidonic acid cascade and phospholipid and docosahexaenoic acid metabolism in the retina. In: N. Osborne and J. Chader (Eds.), Progress in Retinal Research, Vol. 5. Pergamon Press, London, pp. 309322. Bliss, T.V.E and Collingridge, G.L. (1993) A synaptic model of memory: long-term potentiation in the hippocampus. Nature, 361: 31-39. Bonventre, J.V. (1996) Roles of phospholipases A2 in brain cell and tissue injury associated with ischemia and excitotoxicity. J. Lipid Med. Cell Signal., 14: 15-23. Bonventre, J.V., Huang, Z., Taheri, M.R., O'Leary, E., Li, E., Moskowitz, M.A. and Sapirstein, A. (1997) Reduced fertility and postischaemic brain injury in mice deficient in cytosolic phospholipase A2. Nature, 390: 622-625. Bordi, E and Ugolini, A. (1999) Group I metabotropic glutamate receptors: implications for brain diseases. Prog. Neurobiol., 59: 55-79. Cavazos, J.E., Das, I. and Sutula, T.E (1994) Neuronal loss induced in limbic pathways by kindling: evidence for induction of hippocampal sclerosis by repeated brief seizures. J. Neurosci., 14: 3106-3121. Chen, C., Magee, J.C., Marcheselli, V., Hardy, M. and Bazan, N.G. (2001) Attenuated LTP in hippocampal dentate gyms of mice deficient in the PAF receptor. J. Neurophysiol., 85: 384-390. Clark, J.D., Lin, L.L., Kriz, R.W., Ramesha, C.S., Sultzman, J.A., Lin, A.Y., Minola, N. and Knopf, J.L. (1991) A novel arachidonic acid-selective cytosolic PLA2 contains a Ca 2+dependent translocation domain with homology to PKC and GAE Cell, 65: 1043-1051. Clark, G.D., Happel, L.T., Zorumski, C.E and Bazan, N.G. (1992) Enhancement of hippocampal excitatory synaptic transmission by platelet-activating factor. Neuron, 9:1211-1216. Conn, EJ. and Pin, J.E (1997) Pharmacology and functions of metabotropic glutamate receptors. Annu. Rev. Pharmacol. Toxicol., 37: 205-237. Conquet, F., Bashir, Z.I., Davies, C.H., Daniel, H., Ferraguti, F., Bordi, E, Franz-Bacon, K., Reggiani, A., Matarese, V. and Conde, E et al. (1994) Motor deficit and impairment of synaptic plasticity in mice lacking mGluR1. Nature, 372: 237-243. Danbolt, N.C. (2001) Glutamate uptake. Prog. Neurobiol., 65: 1-105. Dugan, L.L., Sensi, S.L., Canzoniero, L.M., Handran, S.D., Rothman, S.M., Lin, T.S., Goldberg, M.E and Choi, D.W. (1995) Mitochondrial production of reactive oxygen species in cortical neurons following exposure to N-methyl-D-aspartate. J. Neurosci., 15: 6377-6388. Dumuis, A., Sebben, M., Haynes, J.-E and Bockaert, J. (1988) NMDA receptors activate the arachidonic acid cascade system in striatal neurons. Nature, 336: 68-70. Fiore, R.S., Murphy, T.H., Sanghera, J.S., Pelech, S.L. and Baraban, J.M. (1993) Activation of p42 mitogen-activated protein
184 kinase by glutamate receptor stimulation in rat primary cortical cultures. J. Neurochem., 61: 1626-1633. Freeman, E.J., Damron, D.S., Tertian, D.M. and Dorrnan, R.V. (1991) 12-Lipoxygenase products attenuate the glutamate release and Ca2+ accumulation evoked by depolarization of hippocampal mossy fiber nerve endings. J. Neurochem., 56: 1079-1082. Gerlai, R. (1996) Gene-targefing studies of mammalian behavior: is it the mutation of the background genotype?. Trends Neurosci., 19: 177-181. Hewett, S.J., Uliasz, T.E, Vidwans, A.S. and Hewett, J.A. (2000) Cyclooxygenase-2 contributes to N-methyl-D-aspartatemediated neuronal cell death in primary cortical cell culture. J. Pharmacol. Exp. Ther., 293: 417--425. Hirabayashi, T., Kume, K., Hirose, K., Yokomizo, T., Iino, M., Itoh, H. and Shimizu, T. (1999) Critical duration of intracellular Ca2+ response required for continuous translocation and activation of cytosolic phospholipase A2. J. Biol. Chem., 274: 5163-5169. Hu, B.R. and Wieloch, T. (1994) Tyrosine phosphorylation and activation of mitogen-activated protein kinase in the rat brain following transient cerebral ischemia. J. Neurochem., 62: 1357-1367. Iadecola, C., Niwa, K., Nogawa, S., Zhao, X., Nagayama, M., Araki, E., Morham, S. and Ross, M.E. (2001) Reduced susceptibility to ischemic brain injury and N-methyl-D-aspartatemediated neurotoxicity in cyclooxygenase-2-deficient mice. Proc. Natl. Acad. Sci. USA, 98: 1294-1299. Izquierdo, I., Fin, C., Schmitz, P.K., Da Silva, R.C., Jerusalinsky, D., Quillfeldt, J.A., Ferreira, M.B.G., Medina, J.H. and Bazan, N.G. (1995) Memory enhancement by intrahippocampal, intraamygdala, or intraentorhinal infusion of platelet-activating factor measured in an inhibitory avoidance task. Proc. Natl. Acad. Sci. USA, 92: 5047-5051. Izumi, Y., Zarrin, A.R. and Zorurnski, C.F. (2000) Arachidonic acid rescues hippocampal long-term potentiation blocked by group I metabotropic glutamate receptor antagonists. Neuroscience, 100: 485-491. Jerusalinsky, D., Fin, C., Quillfeldt, J.A., Ferreira, M.B., Schmitz, P.K., Da Silva, R.C., Walz, R., Bazan, N.G., Medina, J.H. and Izquierdo, I. (1994) Effect of antagonists of plateletactivating factor receptors on memory of inhibitory avoidance in rats. Behav. Neural Biol., 62: 1-3. Kato, K., Clark, G.D., Bazan, N.G. and Zorumski, C.F. (1994) Platelet-activating factor as a potential retrograde messenger in CA1 hippocampal long-term potentiation. Nature, 367: 175179. Kaufmann, W.E., Worley, P.E, Pegg, J., Bremer, M. and Isakson, P. (1996) COX-2, a synaptically induced enzyme, is expressed by excitatory neurons at postsynaptic sites in rat cerebral cortex. Proc. Natl. Acad. Sci. USA, 93: 2317-2321. Kelley, K.A., Ho, L., Winger, D., Freire-Moar, J., Borelli, C.B., Aisen, P.S. and Pasinetti, G.M. (1999) Potentiation of excitotoxicity in transgenic mice overexpressing neuronal cyclooxygenase-2. Am. J. PathoL, 155: 995-1004. Kim, Y.S., Hong, K.S., Seong, Y.S., Park, J.B., Kuroda, S., Kishi, K., Kaibuchi, K. and Takai, Y. (1994) Phosphorylation
and activation of mitogen-activated protein kinase by kainic acid-induced seizure in rat hippocampus. Biochem. Biophys. Res. Commun., 202: 1163-1168. Kumar, R., Harvey, S.A.K., Kester, M., Hanahan, D.J. and Olson, M.S. (1988) Production and effects of platelet-activating factor in the rat brain. Biochim. Biophys. Acta, 963: 375-383. Kurino, M., Fukunaga, K., Ushio, Y. and Miyamoto, E. (1995) Activation of mitogen-activated protein kinase in cultured rat hippocampal neurons by stimulation of glutamate receptors. J. Neurochem., 65: 1282-1289. Leslie, C.C. (1997) Properties and regulation of cytosolic phospholipase A2. J. BioL Chem., 272: 16709-16712. Marcheselli, V.L. and Bazan, N.G. (1990) Quantitative analysis of fatty acids in phospholipids, diacylglycerols, free fatty acids, and other lipids. J. Nutr. Biochem., 1: 382-388. MarcheseUi, V.L. and Bazan, N.G. (1994) Platelet-activating factor is a messenger in the electroconvulsive shock-induced transcriptional activation of c-fos and zif-268 in hippocampus. J. Neurosci. Res., 37: 54-61. Marcheselli, V.L. and Bazan, N.G. (1996) Sustained induction of prostaglandin endoperoxide synthase-2 by seizures in hippocampus: inhibition by a platelet-activating factor antagonist. J. Biol. Chem., 271: 24794-24799. Marcheselli, V.L., Scott, B.L., Reddy, T.S. and Bazan, N.G. (1988) Quantitative analysis of acyl group composition of brain phospholipids and neutral lipids, and free fatty acids. In: A.A. Boulton, G.B. Baker and L.A. Horrocks (Eds.), Lipids and Related Compounds, Neuromethods. Humana Press, New Jersey, pp. 68-85. Miettinen, S., Fusco, ER., Yrj~nheikki, J., Kein~inen, R., Hirvonen, T., Roivainen, R., Narhi, M., H6kfelt, T. and Koistinaho, J. (1997) Spreading depression and focal brain ischemia induce cyclooxygenase-2 in cortical neurons through N-methylD-aspartic acid-receptors and phospholipase A2. Proc. Natl. Acad. Sci. USA, 94: 6500-6505. Miller, B., Sarantis, M., Traynelis, S.E and Attwell, D. (1992) Potentiation of NMDA receptor currents by arachidonic acid. Nature (Lond.), 355: 722-725. Moil, M., Aihara, M., Kume, K., Hamanoue, M., Kohsaka, S. and Shimizu, T. (1996) Predominant expression of plateletactivating factor receptor in the rat brain microglia. J. Neurosci., 16: 3590-3600. Morita, I., Schindler, M., Regier, M.K., Otto, J.C., Hori, T., DeWitt, D.L. and Smith, W.L. (1995) Different intracellular locations for prostaglandin endoperoxide H synthase-1 and -2. J. Biol. Chem., 270: 10902-10908. Mukherjee, P.K., Decoster, M., Campbell, EZ., Davis, R.J. and Bazan, N.G. (1999) Glutamate receptor signaling interplay modulates stress-sensitive mitogen-activated protein kinases and neuronal cell death. J. Biol. Chem., 274: 6493-6498. Nakanishi, S. (1994) Metabotropic glutamate receptors: synaptic transmission, modulation, and plasticity. Neuron, 13: 10311037. Nakayama, N., Uchimura, K., Zhu, R.L., Nagayama, T., Rose, M.E., Stetler, R.A., Isakson, P.C., Chert, J. and Graham, S.H. (1998) Cyclooxygenase-2 inhibition prevents delayed death of
185
CAI hippocampal neurons following global ischemia. Proc. Natl. Acad. Sci. USA, 95: 10954-10959. Nicotera, P., Bellomo, G. and Orrenius, S. (1992) Calciummediated mechanisms in chemically induced cell death. Annu. Rev. Pharmacol. Toxicol., 32: 449-470. Nishida, K. and Markey, S.P. (1996) Platelet-activating factor in brain regions after transient ischemia in gerbils. Stroke, 27: 514-519. Nishizuka, Y. (1995) Protein kinase C and lipid signaling for sustained cellular responses. FASEB J., 9: 489-496. Ogden, E, DeCoster, M.A. and Bazan, N.G. (1998) Recombinant plasma-type platelet-activating factor acetylhydrolase attenuates NMDA-induced hippocampal neuronal apoptosis. J. Neurosci. Res., 53: 677-684. Pace-Asciak, C.R. (1994) Hepoxilins: a review on their cellular actions. Biochim. Biophys. Acta, 1215: 1-8. Pacheco, M.A. and Jope, R.S. (1996) Phosphoinositide signaling in human brain. Prog. Neurobiol., 50: 255-273. Parker, M.A., Bazan, H.E.P., Marcheselli, V., Rodriguez de Turco, E.B. and Bazan, N.G. (2002) Platelet-activating factor induces permeability transition and cytochrome c release in isolated brain mitochondria, in press. Pellerin, L. and Wolfe, L.S. (1990) Release of arachidonic acid by NMDA-receptor activation in the rat hippocampus. Neurochem. Res., 16: 983. Peters-Golden, M., Song, K., Marshall, T. and Brock, T. (1996) Translocation of cytosolic phospholipase A2 to the nuclear envelope elicits topographically localized phospholipid hydrolysis. Biochem. J., 318: 797-803. Pettitt, T.R. and Wakelam, M.J.O. (1999) Diacylglycerol kinase e, but not g, selectively removes polyunsaturated diacylglycerol, inducing altered protein kinase C distribution in vivo. J. Biol. Chem., 274: 36181-36186. Piomelli, D. (1993) Arachidonic acid in cell signaling. Curr. Opin. Cell. Biol., 5(2): 274-280. Piomelli, D. (1994) Eicosanoids in synaptic transmission. Crit. Rev. Neurobiol., 8: 65-83. Piomelli, D. and Greengard, P. (1990) Lipoxygenase metabolites of arachidonic acid in neuronal transmembrane signaling. Trends Pharmacol. Sci., 11: 367-373. Prescott, S.M., Zimmerman, G.A., Stafforini, D.M. and McIntyre, T.M. (2000) Platelet-activating factor and related lipid mediators. Annu. Re~: Biochem., 69: 419-445. Racine, R. (1972) Modification of seizure activity by electrical stimulation. II. Motor seizure. Electroencephalogr. Clin. Neurophysiol., 32: 281-294. Rodriguez de Turco, E.B., Tan, W., Tophan, M.K., Sakane, E, Marcheselli, V.L., Chen, C., Taketomi, A., Prescott, S. and Bazan, N.G. (2001) Diacylglycerol kinase epsilon regulates seizures susceptibility and long-term potentiation through inositol lipid signaling. Proc. Natl. Acad. Sci. USA, 98: 47404745.
Sanfeliu, C., Hunt, A. and Patel, A.J. (1990) Exposure to Nmetbyl-D-aspartate increases release of arachidonic acid in primary cultures of rat hippocampal neurons and not in astrocytes. Brain Res., 526: 241-248. Serou, M., DeCoster, M.A. and Bazan, N.G. (1999) Interleukin-I beta activates expression of cyclooxygenase-2 and inducible nitric oxide synthase in primary hippocampal neuronal culture: platelet-activating factor as a preferential mediator of cyclooxygenase-2 expression. J. Neurosci. Res., 58: 593-598. Shimizu, T. and Wolfe, L.S. (1990) Arachidonic acid cascade and signal transduction. J. Neurochem., 55: 1-15. Spector, A.A. and Yorek, M.A. (1985) Membrane lipid composition and cellular function. J. Lipid Res., 26: 1015-1035. Stubbs, C.D. and Smith, A.D. (1984) The modification of mammalian membrane polyunsaturated fatty acid composition in relation to membrane fluidity and function. Biochim. Biophys. Acta, 779: 89-137. Tang, W., Bunting, M., Zimmerman, G.A., McIntyre, T.M. and Prescott, S.M. (1996) Molecular cloning of a novel human diacylglycerol kinase highly selective for arachidonate-containing substrates. J. Biol. Chem., 271: 10237-10241. Tocco, G., Freire-Moar, J., Schreiber, S.S., Sakhi, S.H., Aisen, P.S. and Pasinetti, G.M. (1997) Maturation regulation and regional induction of cyclooxygenase-2 in rat brain: implications for AIzheimer's disease. Exp. Neurol., 144: 339-349. Tu, B. and Bazan, N.G. (submitted) Hippocampal kindling epileptogenesis upregulates neuronal neocortical COX-2 expression. Vane, J.R., Bakhle, Y.S. and Botting, R.M. (1998) Cyclooxygenases 1 and 2. Annu. Rev. Pharmacol. Toxieol., 38: 97-120. Vesce, S., Bezzi, P. and Volterra, A. (1999) The highly integrated dialogue between neurons and astrocytes in brain function. Sci. Prog., 82: 251-270. Volterra, A., Trotti, D., Cassutti, P., Tromba, C., Galimberti, R., Lecchi, P. and Racgni, G. (1992) A role for the arachidonic acid cascade in fast synaptic modulation: ion channels and transmitter uptake systems as target proteins. Adv. Exp. Med. Biol., 3138: 147-158. Wieraszko, A., Li, G., Kornecki, E., Hogan, M.V. and Ehrlich, Y.H. (1993) Long-term potential in the hippocampus induced by platelet-activating factor. Neuron, 10: 553-557. Wilsch, V.W., Behnisch, T., Jager, T., Reymann, K.G. and Balschun, D. (1998) When are class I metabotropic glutamate receptors necessary for long-term potentiation?. J. Neurosci., 18: 6071-6080. Yamagata, K., Andreasson, K.I., Kaufmann, W.E., Barnes, C.A. and Worley, RE (1993) Expression of a mitogen-inducible cyclooxygenase in brain neurons: Regulation by synaptic activity and glucocorticoids. Neuron, 11: 371-386. Yeagle, EL. (1989) Lipid regulation of cell membrane structure and function. FASEB J., 3: 1833-1842.
T. Sutula and A. Pitk~inen (Eds.) Progress in Brain Research, Vol. 135 © 2002 Elsevier Science B.V. All fights reserved
CHAPTER 16
The role of mitochondria and oxidative stress in neuronal damage after brief and prolonged seizures H a n n a h R. C o c k * Department of Clinical and Experimental Epilepsy, Institute of Neurology, Queen Square, London WCIN 3BG, UK
Abstract: Studies in vitro and in other disease states where excitotoxicity is believed to be important have demonstrated that mitochondrial function is a critical determinant of cell death, reflecting key roles in intracellular calcium homeostasis, energy production and oxidative stress. Central to this is the process of mitochondrial permeability transition, for which there are numerous influencing factors, although many, if not all, may specifically act though effects on the redox state of the cell and oxidative stress. Mitochondrial function in relation to seizure-induced cell death has been little studied until recently, but there is now accumulating evidence that similar mechanisms operate, certainly in cell death, following prolonged seizures. To what extent these same mechanisms might contribute to non-fatal but pathologically significant functional cellular changes in epilepsy, and the significance of reported free radical production after brief seizures is as yet uncertain. However, with the wide range of established techniques available to study mitochondrial function and oxidative stress, and those currently under development, these questions are undoubtedly answerable in the near future. Increased understanding of the mechanisms involved in seizure-induced cellular damage is an essential basis for the development of rational neuroprotective strategies.
Introduction Seizure activity results in a large number of changes and cascades of events at a cellular level. Changes in gene expression, receptor composition, synaptic physiology and the activation of some late cell death pathways (e.g. caspase activation) will have been covered elsewhere. This chapter will focus on the potential role of mitochondria, including their capacity to produce free radicals, in seizure-associated neuronal damage. Following an introduction to normal mitochondrial functions, I will briefly discuss some of the methodological issues in this area. I will
* Correspondence to: H.R. Cock, Department of Clinical and Experimental Epilepsy, Institute of Neurology, Queen Square, London WC1N 3BG, UK. Tel.: -t-44-207-8373611, ext. 4256; Fax: +44-207-278-5616; E-mail:
[email protected]
then summarize the major conclusions from studies on neuronal death in vitro and in other disease states, where there has been extensive work on excitotoxic cell death, which underpins the more limited work to date in epilepsy. Finally, I will review the work that has been done in this field with respect to seizureassociated cell death, before concluding with my own perspective on the future in this area. Mitochondrial structure and function Mitochondria are ubiquitous intracellular organelles, whose primary function is the production of cellular energy in the form of adenosine triphospate (ATP) from food-derived fuels. Each mitochondrion (Fig. 1) consists of a double membranebound structure, with an internal matrix in which many metabolic systems involved in breaking down food fuels reside. These include the fatty acid [~oxidation enzymes, and those of the tricarboxylic
188
glugose
fatty acids
0 mtDI
MP"
Inner membrane// outer membrane Fig. 1. Schematicrepresentationof a mitochondrion.See text for details, mtDNA,mitochondrialDNA; MPT,mitochondrialpermeability transition; TCA, tricarboxylicacid cycle; I, II, III, IV, V, complexesof the mitochondrialrespiratorychain. acid (TCA/Kreb's) cycle to break down carbohydrates. Electrons from these systems are passed to the mitochondrial respiratory chain (MRC) situated on the inner mitochondrial membrane, which through a series of enzymatic processes (complexes I-IV), passes the electrons to the final acceptor, oxygen, which is reduced to water (Darley-Usmar et al., 1994). At complexes I, III and IV, electron transport is coupled to vectoral proton translocation, creating an electrochemical gradient across the inner mitochondrial membrane. This potential energy is then utilized by Complex V to generate ATE ATP is the basic unit of cellular energy and as such not only a pre-requisite for cell survival, but also essential for a wide range of cellular functions (McCormack and Denton, 1994), including several ionic homeostasis mechanisms (e.g. Na+-K + ATPase; Na+-Ca 2-ATPase), repair systems, and the ability of neurons to generate action potentials. In addition to their oxidative metabolism function, mitochondria play a key role in a variety of other processes believed to be important in cell death (Fig. 2). Mitochondria are crucial to intra-
cellular calcium homeostasis (Duchen, 2000), and possess several calcium transport systems (Nicholls, 1985). The concentration of free intracellular calcium is central to normal neuronal functioning, and in turn this has been shown to be critically dependent on functioning mitochondria, as well as secondarily on sodium/calcium exchange (White and Reynolds, 1995). Intramitochondrial calcium levels also have important regulatory functions, including direct influence on enzymes of the TCA cycle and consequent metabolic rate, which will be further discussed by Heinemann et al. (2002, this volume). Finally, the MRC has long been recognized as the major source of free radicals in the cell (Cadenas et al., 1977). Free radicals are highly reactive oxygen species, which, unopposed, can damage all cell structures, including lipids, proteins and DNA (Halliwell and Gutteridge, 1985). Some radicals are toxic via secondary reactions, one of the most important being the production of the highly damaging peroxynitrite from nitric oxide and the superoxide radical. Mitochondria possess a calcium-dependent nitric oxide synthase, thus have the potential for per-
189
Glutamate
_ ~
C a ++
""~
MPT ~ Caspase , ~ ~
ATPase
activation L ~
| ATP
/--/"~
~-,~.t
~l~r"v
Tca*÷~J"-"~"-'"-'~/
~ ............
\,, ~i
"/ /
// .........
Repa~ system Proteins
~,~
.
/" ....
Free ~[adicals
"
Antioxidants
Fig. 2. Mechanisms of mitochondrial involvementin excitotoxic cell death in a single neuron. Points of interaction are represented by arrows, see text for details. ONOO-, peroxynitrite; nNOS, neuronal nitric oxide synthase; NO, nitric oxide; MPT, mitochondrial permeability transition. oxynitrite production in addition to those directly produced by the MRC. Under physiological conditions numerous antioxidant systems exist, including superoxide dismutase, catalase, and glutathione, to 'mop' up any free radicals as soon as they are produced. Under pathological conditions, however, this balance can be disturbed, either due to an excess production of free radicals, or failure of the normal antioxidant systems, resulting in a state known as oxidative stress. As inhibition of the MRC results in excess free radical production, and free radicals themselves are direct inhibitors of the MRC (Zhang et al., 1990) this can result in a vicious cycle leading to considerable oxidative cell damage. Furthermore, the inherent biochemical and physiological characteristics of the brain, including high lipid content and
energy requirements make it particularly susceptible to oxidative damage. Reflecting their central role in cellular metabolism, mitochondria have been proposed to act as the 'stress sensor', and in extreme circumstances 'executioner' of the cell (Green and Reed, 1998). Pivotal to this role is a process known as mitochondrial permeability transition (MPT). This is traditionally represented as a pore straddling the two membranes which when opened releases large molecules including cytochrome c in to the cell cytoplasm (Martinou, 1999). This not only directly activates known cell death pathways (e.g. caspases), but also dissipates the mitochondrial membrane potential crucial to oxidative phosphorylation. To date, although it is generally agreed that MPT is protein mediated, no novel
190 inner membrane pore with appropriate characteristics has been identified, and MPT is almost certainly caused by a group of modified and assembled inner and outer membrane components, including the ADP/ATP translocator, cyclophillin D and possibly porin and hexokinase (Kowaltowski et al., 2001). Numerous factors are known to regulate MPT both from within the mitochondria (e.g. calcium, oxidative stress, membrane potential) and the cytosol (e.g. bcl-2 proteins, nitric oxide) (Kroemer and Reed, 2000); however, there is evidence to suggest that oxidative stress links all of these factors and is the critical determinant (Kowaltowski et al., 2001).
Methods for assessing mitochondrial function/oxidative stress Many of the methods used to study mitochondria and oxidative stress are well established, and have been extensively validated over many years. These include spectrophotometric analysis of MRC and matrix enzyme activities, and substrate-linked respiration studies using oxygen electrodes (polarography: Darley-Usmar et al., 1994). These can be applied to tissue homogenates, purified mitochondrial/enzyme preparations, or in the case of polarography to brain slice preparations. HPLC methods can quantify a variety of important metabolites (e.g. glutathione, N-acetyl aspartate (Heales et al., 1995)), and gene expression and protein synthesis within mitochondria can also be readily studied (Darley-Usmar et al., 1994). Various patterns of enzymatic inactivation in relation to particular insult type/mechanisms are also well recognized. For instance, complex I of the respiratory chain is particular susceptible to superoxide damage (Zhang et al., 1990), and the matrix enzyme aconitase is readily inactivated in the presence of peroxynitrite (Gardner et al., 1994). Methods for measuring ionic transport across mitochondrial membranes (e.g. calcium) and mitochondrial membrane potential are also available, most involving the use of fluorescent dyes (e.g. Schuchmann et al., 2000). Fluorescent dyes can also be used to detect free radical production in vitro and ex-vivo, but do have limitations with variable auto-oxidation, photoconversion and retention in cells, and possible direct effects themselves on mitochondrial function (Buck-
man et al., 2001). Free radical detection, particularly in vivo, is difficult due to their short half-life and low concentrations. Direct colorimetric and luminescent probes (e.g. nitro-blue tetrazoleum (NBT), cytochrome c reduction (Bauknight et al., 1992)) only measure free radicals that leave cells, making them of limited value. Most commonly, investigators instead look for evidence of free radical damage, and there are established methods for quantifying lipid (http: / / www.oxisresearch.com / products / assays), protein (Cini and Moretti, 1995) and DNA oxidation (Williams et al., 1998). This is a vast field, and I have been able only to touch on what is a formidable repertoire of available techniques available in addition to which new refinements to many are constantly developing.
Mitochondria, oxidative stress and cell death Mitochondria and excitotoxic cell death
As will have been discussed in previous chapters, and has been comprehensively reviewed elsewhere (Meldrum, 1993), cell death in epilepsy is believed to involve excitotoxicity, whereby excessive glutamate causes over stimulation of the post-synaptic NMDA receptors, with a resultant accumulation of intracellular calcium leading to ultimately to cell death. Excitotoxic cell death is also believed to be important in a wide range of other diseases including stroke, trauma and neurodegenerative conditions (Sattler and Tymianski, 2000), where it has been extensively studied. Conventional thinking in recent years has subclassifted cell death as either necrotic (acute) or apoptotic (delayed) on largely morphological grounds (Bonfoco et al., 1995). Apoptosis is an energy-requiring process, thus mitochondrial respiratory function is believed to be a critical determinant of the mode of excitotoxic cell death (Ankarcrona et al., 1995): essentially where MRC function is significantly impaired, for example due to oxidative damage, necrosis occurs; in contrast, where there is adequate MRC function to survive the initial insult, a delayed cell death by apoptosis occurs. The definition of apoptosis, however, is primarily a morphological one (Kerr, 1971), and other than implying a certain energy reserve within the cell at the time of commitment
191 to death, its presence alone does not tell us what mechanisms initiated the process. Furthermore it is increasingly accepted that neuronal apoptosis in the context of excitotoxicity differs from 'classical' apoptosis (originally described in a developmental context). The primary event in the latter has been considered to be independent of mitochondrial function, being genetically based and protein activated, and focused on the caspase-8 activation pathway (Ashkenazi and Dixit, 1998). In contrast, in excitotoxicity, the primary event probably centers on mitochondrial dysfunction, involving calcium influx and oxidative stress, with secondary MPT and caspase (9) activation. However, secondary changes in mitochondrial function including MPT can and undoubtedly do occur in classical apoptosis, and in the context of excitotoxicity and neurodegenerative diseases, features of both necrosis and apoptosis may identifiable in individual cells. Furthermore, in neural tissue 'apoptosis' can be variably triggered by a variety of 'necrotic' insults (Roy and Sapolsky, 1999) making classification somewhat confusing. The distinction has been further confounded by the relative non-specificity of some of the methods used (e.g. TUNEL histochemistry) to identify apoptotic changes (Charriaut-Marlangue and Ben-Ari, 1995). Thus some authors have argued that the distinction between apoptosis and necrosis in excitotoxic cell death is somewhat artificial, each representing one end of a continuum depending on the nature/severity of the causative insult and properties of the cell in question (Martin et al., 1998). From a pragmatic perspective, in the context of seizure-related cell death, many of the involved mechanisms are probably common to both 'excitotoxic apoptosis' and 'necrosis'. For this reason, in the rest of this chapter I will be referring to excitotoxic cell death as a whole, without drawing mechanistic conclusions from morphological data. I will focus on the role of mitochondria and oxidative stress, but not be discussing pathways specific to the development of apoptotic changes (e.g. caspases activation, genetic regulation of MPT) as these will have been covered elsewhere in this volume. Neither will I cover the undoubtedly important role of bioactive lipid pathways, although these can affect mitochondrial function, as this will be covered by Bazan et al. (2002, this volume).
Studies in vitro and in other disease status have highlighted a number of mechanisms contributing to excitotoxic cell death (Fig. 2). These include calcium influx into the cell (Hartley et al., 1993), oxidative damage via the production of free radicals (Beal, 1996; Lafoncazal et al., 1993), and specifically mitochondrial nitric oxide/peroxynitrite production (Almeida et al., 1998). Of note, from a mitochondrial perspective, it appears that ATP depletion, although it may lower the threshold for cell death, is not the crucial step. Eloquent studies have demonstrated that impairment of mitochondrial calcium sequestration is the key determinant (Stout et al., 1999). This in turn requires ATE but is also in part potential driven, involving the MPT pore mentioned previously (Schinder et al., 1996). Thus even in the presence of adequate ATP, if the mitochondrion is unable to sequester calcium cell death ensues. It is likely that all of these factors contribute to excitotoxic cell death, though some may be more important than others depending on the exact system/disease under study. Equally, there are numerous ways of counteracting each of these parts of the cascade (e.g. calcium channel blockers, antioxidants). The key in addressing this in epilepsy, however, is to ascertain which, if any, of these mechanisms are important in seizure-associated cell death.
Non-fatal cell damage Whilst the focus of this book has concentrated largely on cell death, it should be remembered that the same cellular mechanisms might also result in non-fatal cell damage and dysfunction. Free radicals, probably through membrane (lipid) oxidation may influence channel/receptor function (reviewed by Cock and Schapira, 1999). Oxidative protein modifications (Orrenius et al., 1992) have wide potential ramifications, and calcium is recognized as a key messenger in a range of cell functions (Duchen, 2000), including probable feedback inhibition of the NMDA receptor (Rosenmund et al., 1995), and regulation metabolic activity. Thus understanding of cell death mechanisms could have a wider application in preventing perhaps more widespread non-fatal functional damage, although to date this possibility has been little explored.
t92
Evidence for mitochondrial dysfunction/oxidative stress in epilepsy The first observation to make is that compared to the wealth of in vitro data on excitotoxicity, and the large number of studies in other disease states, both using animal models and human samples, there is a relative paucity of data about mitochondrial function and oxidative stress in epilepsy. In part, this probably reflects that human epilepsy, by definition with spontaneous seizures is not a condition easily modelled, and the study of human tissue is largely confined to surgical specimens or post-mortem material, which inevitably represents the end stage of whatever processes are occurring. What data there is largely relates to prolonged seizures (status epilepticus). I will review this first, before going on to the little data on brief seizures, and drawing some conclusions.
In vitro evidence of oxidative stress/mitochondrial dysfunction Frantseva et al. (1999) have reported studies using fluorescent dyes in hippocampal slice preparations. These dyes have limitations, as discussed previously, and of course epilepsy, with clinical manifestations by definition, cannot be replicated in a slice. However, hippocampal slices are widely used to study synaptic physiology and ionic changes, and the reported results are persuasive. Rhythmic synchronous activity was induced with topical bicuculline, and increased free radical production was apparent within 10-15 min, particularly in the CA3 region where neurons subsequently degenerated. Both the free radical production and cell death correlated with a persistent and progressive increase in intracellular calcium, beyond the duration of activity. Heinemann et al. (2002, this volume) have taken this a stage further using similar methods applied to organotypic slice cultures and a variety of seizure inducers. This has demonstrated a clear relationship between neuronal activity, intramitochondrial calcium handling, metabolic rate and mitochondrial depolarization, which would be anticipated to lead to cell death. It has been further demonstrated (Frantseva et al., 2000) that free radical scavengers (vitamin E and glutathione) significantly reduced the seizure-induced
neurodegeneration, suggesting that free radical overproduction is directly related to seizure-induced cell death.
In vivo studies of oxidative stress/mitochondrial function after prolonged seizures There have been a number of studies looking for evidence of oxidative damage following seizures, but most involve acute seizure provocation models and it is not possible to definitively separate the effect of the agent used (e.g. iron, bicuculline, kainate) from the seizures per se. This is particularly true for iron, which is well recognized as a potent stimulant of free radical production in its own right (Gutteridge, 1992). Intracortical iron, used as a model of post-traumatic epilepsy, certainly produces both free radicals (mainly hydroxyl radical) and seizures (e.g. Kucukkaya et al., 1998), but cause and effect have not been clearly established. Pretreatment with free radical scavengers (adenosine (Yokoi et al., 1995) and melatonin (Kabuto et al., 1998)) have been shown to reduce iron-induced hydroxyl radical production in vitro, and variably delay/prevent the occurrence of spike discharges in vivo in this model. However, both have adenosine and melatonin have actions in addition to their antioxidant capacities, and the conclusion that free radical scavenging suppresses the epileptogenesis following iron injection is perhaps premature. A number of groups have looked at oxidative stress in kainate-induced seizures. Bruce and Baudry (1995) demonstrated that in the early stages (8-16 h) after prolonged (5-6 h) seizure activity, there was increased lipid peroxidation and protein oxidation in regional brain homogenates, correlating with and preceding histological cell damage that was most apparent at 5 days post-status (predominantly affecting CA1 and CA3 in mature, but not young animals). There was also a later (5 days post-status) increase in antioxidant enzyme activities (glutathione peroxidase), postulated to reflect compensatory micro-glial activation and proliferation. Gulyaeva et al. (1999) reported similar findings with evidence of lipid peroxidation, and reduced glutathione levels 3 days after kainate-induced status. Melatonin pretreatment has also been studied in this model, and shown to partially ameliorate the development of acute seizures
193 (Giusti et al., 1996) and neuronal damage (Uz et al., 1996; TUNEL and Nissl staining at 72 h). This is all in keeping with the hypothesis that oxidative stress plays an early role in the cascade to cell death in kainate-induced excitotoxicity. The anticonvulsant and antioxidant effects of melatonin cannot be distinguished in these studies. However, further support for the hypothesis has been provided by studies involving pretreatment with a synthetic antioxidant (EUK-134), which had no effect on kainate-induced seizures, also partially ameliorated early evidence of oxidative stress (specific gene induction, and nitrotyrosine immunohistochemistry at 8-16 h post-status) and histological damage evident at 5 days (Rong et al., 1999). A handful of studies in other chemoconvulsant models have also been reported: (Bauknight et al., 1992) demonstrated increased free radicals (spectrophotometric analysis of the SOD inhibitable reduction of NBT) in the CSF overlying the seizure focus (bicuculline-induced) in cats; Rauca et al. (1999) studied free radical production after acutely provoked and kindled seizures following pentylenetetrazole administration to rats. The method used is reported to detect free hydroxyl radicals, trapped by systemically applied salicylate, resulting in a stable and quantifiable product. Increased free radical production was seen in the early minutes (1-15) after acute or kindled seizures, but had normalized by 60 rain. However despite the small number and limitations of some of these studies, the overall message is consistent - - supporting free radical production as a consequence of seizures. In contrast to the more detailed studies of MRC function and other enzyme systems known to be sensitive to oxidative stress, undertaken in other disease models, there has been little in this area in epilepsy. Kunz et al. (1999) have performed polarography and fluorescent spectroscopy on hippocampal slices prepared from kainate treated rats. The animals were sacrificed at least a month after initial kainate treatment, having exhibited initial status and subsequent spontaneous seizures. The main finding was an increase in oxidative metabolic rates in the seizure animals compared to controls, suggested to represent increased mitochondrial energy turnover, in turn possibly reflecting futile calcium cycling, which would fit with existing hypotheses. Markers of oxida-
tive stress, or individual enzyme activities were not, however, reported so no conclusions can be drawn in this respect. Our own work has been in the perforant path stimulation model of status epilepticus, which avoids the potentially confounding effects of extrinsic chemoconvulsants (Cock et al., 2002). We have observed significant decreases in the activities of aconitase and a-ketoglutarate dehydrogenase, two mitochondrial enzymes known to be critically sensitive to oxidative stress, particularly involving nitric oxide (peroxynitrite), as well as reduced levels of glutathione, in the brain homogenates of status animals compared to sham-operated controls. These changes were detectable up to 40 h after the 5 h of seizure activity, and preceded maximal histological neuronal damage (detected at 8 days post-status in this study), again supporting a role for free radical production in the pathogenesis of cell death after prolonged seizures.
In vivo studies of oxidative stress/mitochondrial dysfunction after brief seizures There has been less work looking for oxidative stress after brief seizures. Following electroconvulsive shock in rats, an acute decrease in regional brain antioxidant levels has been reported (Erakovic et al., 2000), persisting for at least 48 h after a single seizure, although exact seizure duration is not specified in the paper. The decrease was even more marked where there had been repeated brief seizures. This could be either as a result of, or contributing to, oxidative stress in this model, and direct evidence of free radical production was not sought. Arnaiz et al. (1998) used a chemiluminescent assay to look for lipid peroxidation after 3-mercaptopropionic acidinduced seizures in rats. Even in animals sacrificed during the seizures (3-6 min after onset), a 20-40% increase in lipid peroxidation was found in some vulnerable brain regions, although not in all. This had largely normalized 20 min later, which is somewhat surprising given the rate of lipid turnover, and total antioxidant capacity did not change at all throughout. An immediate increase in free radical production following individual seizures has also been demonstrated in PTZ-kindled animals (Rauca et al., 1999). Thus it does appear that even brief seizures can increase free radical production, though perhaps only
194 very transiently and the patho-physiological consequences of this have not been sufficiently elucidated to draw further conclusions.
Human studies of oxidative stress/mitochondrial function in epilepsy Studies of mitochondrial function using spectrophotometric enzyme analysis on regional brain homogenates, and polarography on hippocampal slices have been reported from 40 surgical resection specimens from patients with hippocampal sclerosis (Kudin et al., 1999). Control tissue is of course difficult to obtain for such studies, and the three 'controls' had pathologically normal hippocampi, but did also have epilepsy associated with other lesions. Complex I activity was found to be decreased in the CA3 region in the sclerosed specimens, and the authors speculate that this might be of significance in the pathogenesis of temporal lobe epilepsy. However, whilst clearly of interest, it is difficult to draw mechanistic conclusions from such work as the tissue clearly represents the end stage of a very long process, both in terms of neuronal damage and epileptogenesis. Further studies of this nature in appropriate disease models may help address this issue, and are underway in our own laboratory. Conclusions and areas for future attention
Overall, the fairly consistent data from a range of animal models, suggests that impaired mitochondrial calcium handling and significant free radical production occur following prolonged seizures, and furthermore that there is evidence of local oxidative cell damage preceding neuronal death in vulnerable brain regions. Many of the models studied have only complex partial seizures, without additional cardiorespiratory compromise, and the changes observed can probably be ascribed directly to the seizure activity, particularly where extrinsic chemoconvulsants have been avoided. Following status epilepticus, there appears to be a potential therapeutic window following status of at least a few hours, although all the neuroprotective studies to date in this area have involved antioxidant pre-treatment, so this is at present speculative. However, it is by no means clear which free radicals are of particular impor-
tance in epilepsy (e.g. hydroxl radicals, superoxide, peroxynitrite?), where they come from, or which cell components are especially vulnerable (e.g. lipid membranes, MRC or other enzymes, channels or receptors etc.). Different antioxidant neuroprotective strategies may apply depending on the particular mechanisms operate (e.g. glutathione mainly protects against complex I damage, vitamin E against complex IV), and 'random' neuroprotective attempts without a good understanding are probably inappropriate (Delanty and Dichter, 2000). Perhaps more important to establish is to what extent the various pathways activated act in parallel or in series. If the former applies, then multiple neuroprotective strategies would be needed to prevent cell death/damage, either via a single agent with multiple actions, or using multiple agents. Such an approach is clearly likely to have more widespread, perhaps damaging, implications on other cell functions and may account for why antioxidant/neuroprotective strategies in the clinical arena have been so disappointing to date (Delanty and Dichter, 2000). It is further possible that different mechanisms will prove important in different models and in different epilepsy syndromes, which will need to be considered in experimental design. However, despite these hurdles, there are a number of well-established techniques in this area, which have been sparsely applied to epilepsy models, so further laboratory-based work addressing these questions is clearly not only necessary, but possible. This should be followed by rationale appropriately designed clinical trials of neuroprotective strategies both in animal models and in man, based on an increased understanding of the mechanisms involved in epilepsy. The question of whether repeated brief seizures might be trigger the same mechanism(s) also needs addressing from the mitochondrial/oxidative stress perspective. The results of Rauca et al. (1999) suggest that any oxidative stress with brief seizures may be short lived. If this finding is replicated in other models, it would suggest a very short window of opportunity for antioxidant therapy in the acute situation, but there may be a cumulative effect with frequent seizures such that prophylaxis in patients with chronic epilepsy may be justified. A randomized double blind trial of vitamin E in children with epilepsy (Ogunmekan and Hwang, 1989) showed
195 that 10 o f the 12 in the t r e a t m e n t group had at least a 6 0 % i m p r o v e m e n t in seizure f r e q u e n c y o v e r a 3 - m o n t h p e r i o d c o m p a r e d to n o n e in the p l a c e b o group. A l t h o u g h this is a single small study, and it has perhaps surprisingly not b e e n replicated on a larger scale, this surely should fuel the n e e d for m o r e w o r k in this area, both in a n i m a l models, and u l t i m a t e l y in man.
Acknowledgements W i t h thanks to collaborators on m y o w n w o r k at the Institute o f N e u r o l o g y : S i m o n Shorvon, M a t t h e w Walker, Phillip Patsalos, X i n Tong ( D e p a r t m e n t o f C l i n i c a l and E x p e r i m e n t a l Epilepsy); J o h n Clark, Sim o n H e a l e s , Iain H a r g r e a v e s ( D e p a r t m e n t o f N e u r o chemistry); Maria Thom; Mike Groves (Department o f N e u r o p a t h o l o g y ) and to Tony S c h a p i r a at the Royal Free Hospital Department of Clinical Neurosciences.
References Almeida, A., Heales, S.R., Bolanos, J.E and Medina, J.M. (1998) Glutamate neurotoxicity is associated with nitric oxidemediated mitochondrial dysfunction and glutathione depletion. Brain Res., 790: 209-216. Ankarcrona, M., Dypbukt, J.M., Bonfoco, E., Zhivotovsky, B., Orrenius, S., Lipton, S.A. and Nicotera, P. (1995) Glutamateinduced neuronal death: a succession of necrosis or apoptosis depending on mitochondrial function. Neuron, 15: 961-973. Arnaiz, S.L., Travacio, M., Llesuy, S. and Arnaiz, G.R.D. (1998) Regional vulnerability to oxidative stress in a model of experimental epilepsy. Neuroehem. Res., 23: 1477-1483. Ashkenazi, A. and Dixit, V.M. (1998) Death receptors: signalling and modulation. Science, 281: 1305-1308. Bauknigbt, G.C., Wei, E.E and Kontos, H.A. (1992) Superoxide production in experimental seizures in cats. Stroke, 23: 15121514. Bazan, N.G., Tu, B. and Rodriguez de Turco, E.B. (2002) What synaptic lipid signaling tells us about seizure-induced damage and epileptogenesis. In: T. Sutula and A. Pitk~inen (Eds.), Do Seizures Damage the Brain. Progress in Brain Research, Vol. 135. Elsevier, Amsterdam, pp. 175-185. Beal, M.E (1996) Mitochondria, free radicals, and neurodegeneration. Curr. Opin. Neurobiol., 6: 661-666. Bonfoco, E., Krainc, D., Ankarcrona, M., Nicotera, P. and Lipton, S.A. (1995) Apoptosis and necrosis: two distinct events induced, respectively, by mild and intense insults with Nmethyl-D-aspartate or nitric oxide/superoxide in cortical cell cultures. Proc. Natl. Acad. Sci. USA, 92: 7162-7166. Bruce, A.J. and Bandry, M. (1995) Oxygen-free radicals in rat
limbic structures after kainate-induced seizures. Free Radic. Biol. Med., 18: 993-1002. Buckman, J.E, Hernandez, H., Kress, G.J., Votyakova, T.V., Pal, S. and Reynolds, I.J. (2001) MitoTracker labeling in primary neuronal and astrocytic cultures: influence of mitochondftal membrane potential and oxidants. J. Neurosci. Methods, 104: 165-176. Cadenas, E., Boveris, A., Ragan, C.I. and Stoppani, A.O.M. (1977) Production of superoxide radicals and hydrogen peroxide by NADH-ubiquinone reductase and ubiquinol-cytochrome c reductase from beef heart mitochondria. Arch. Biochem. Biophys., 180: 248-257. Charriaut-Marlangue, C. and Ben-Aft, Y. (1995) A cautionary note on the use of the TUNEL stain to determine apoptosis. NeuroReport, 7: 61-64. Cini, M. and Moretti, A. (1995) Studies on lipid peroxidation and protein oxidation in the ageing brain. Neurobiol. Ageing, 16: 53-57. Cock, H.R. and Schapira, A.H.V. (1999) Mitochondrial DNA mutations and mitochondrial dysfunction in epilepsy. Epilepsia, 40: 33-40. Cock, H.R., Tong, X., Hargreaves, I.P., Heales, S.J.R., Clark, J.B., Patsalos, P.N., Thorn, M., Groves, M., Schapira, A.H.V., Shorvon, S.D. and Walker, M.C. (2002) Mitochondrial dysfunction associated with neuronal death following status epilepticus in rat. Epilepsy Res., in press. Darley-Usmar, V.M., Ragan, C.I., Smith, P.R. and Wilson, M.T. (1994) The proteins of the mitochondrial inner membrane and their role in oxidative phosphorylation. In: V.M. Darley-Usmar and A.H.V. Schapira (Eds.), Mitochondria: DNA, Proteins and D&ease, Portland Press Research Monographs. Portland Press, London, pp. 1-27. Delanty, N. and Dichter, M.A. (2000) Antioxidant therapy in neurologic disease. Arch. NeuroL, 57: 1265-1270. Duchen, M.R. (2000) Mitochondria and calcium: from cell signalling to cell death. J. PhysioL, 529: 57-68. Erakovic, V., Zupan, G., Varljen, J., Radosevic, S. and Simonic, A. (2000) Electroconvulsive shock in rats: changes in superoxide dismutase and glutathione peroxidase activity. Mol. Brain Res., 76: 266-274. Frantseva, M.V., Velazquez, J.L.E, Hwang, EA. and Carlen, EL. (1999) Free radical production correlates with cell death in an in vitro model of epilepsy. Epilepsia, 40: 139-140. Frantseva, M.V., Velazquez, J.L.E, Hwang, P.A. and Carlen, EL. (2000) Free radical production correlates with cell death in an in vitro model of epilepsy. Eur. J. Neurosci., 12: 1431-1439. Gardner, ER., Nguyen, D.D. and White, C.W. (1994) Aconitase is a sensitive and critical target of oxygen poisoning in cultured mammalian cells and in rat lungs. Proc. Natl. Acad. Sci. USA, 91: 12248-12252. Giusti, E, Lipartiti, M, Franceschini, D., Schiavo, N., Floreani, M. and Manev, H. (1996) Neuroprotection by melatonin from kainate-induced excitotoxicity in rats. FASEB J., 10: 891-896. Green, D.R. and Reed, J.C. (1998) Mitochondria and apoptosis. Science, 281: 1309-1310. Gulyaeva, N.V., Stepanichev, M . Y . , Onufftev, M.V., Chernyavskaya, L.I., Moiseeva, Y.V., Lazareva, N.A. and Fliss,
196
H. (1999) Seizure severity in kainic acid-treated rats correlates with transcriptional activity, oxidative stress, and apoptosis in the hippocampus. J. Neurochem., 73: $9. Gutteridge, J.M.C. (1992) Iron and oxygen radicals in brain. Ann. Neurol., 32: S16-$21. Halliwell, B. and Gutteridge, J.M.C. (1985) Free Radicals in Biology and Medicine. Clarendon Press, Oxford. Hartley, D.M., Kurth, M.C., Bjerkness, L., Weiss, J.H. and Choi, D.W. (1993) Glutamate receptor induced 45Ca2+ accumulation in cortical cell culture correlates with subsequent neuronal degeneration. J. Neurosci., 13: 1993-2000. Heales, S.J., Davies, S.E.C., Bates, T. and Clark, J.B. (1995) Depletion of brain glutathione is accompanied by impaired mitochondrial function and decreased N-acetyl aspartate concentration. Neurochem. Res., 20: 31-38. Heinemann, U., Buchheim, K., Gabriel, S., Kann, O., Kovacs, R. and Schuchmann, S. (2002) Cell death and metabolic activity during epileptiform discharges and status epilepticus in the hippocampus. In: T. Sutula and A. Pitk~inen (Eds.), Do Seizures Damage the Brain. Progress in Brain Research, Vol. 135. Elsevier, Amsterdam, pp. 197-210. Kabuto, H., Yokoi, I. and Ogawa, N. (1998) Melatonin inhibits iron-induced epileptic discharges in rats by suppressing peroxidation. Epilepsia, 39: 237-243. Kerr, J.ER. (1971) Shrinkage necrosis: a distinct mode of cellular death. J. Pathol., 105: 13-20. Kowaltowski, A.J., Castilho, R.E and Vercesi, A.E. (2001) Mitochondrial permeability transition and oxidative stress. FEBS Lett., 495: 12-15. Kroemer, G. and Reed, J.C. (2000) Mitochondrial control of cell death. Nat. Med., 6: 513-519. Kucukkaya, B., Aker, R., Yuksel, M., Onat, E and Yalcin, A.S. (1998) Low dose MK-801 protects against iron-induced oxidative changes in a rat model of focal epilepsy. Brain Res., 788: 133-136. Kudin, A., Vielhaber, S., Beck, H., Elger, E.C. and Knnz, W.S. (1999) Mitochondrial respiratory chain complex I deficiency in human epileptic hippocampus. J. Neurochem., 73: $217. Kunz, W.S., Goussakov, I.V., Beck, H. and Elger, C.E. (1999) Altered mitochondrial oxidative phosphorylation in hippocampal slices of kainate-treated rats. Brain Res., 826: 236-242. Lafoncazal, M., Pietri, S., Culcasi, M. and Bockaert, J. (1993) NMDA-dependent superoxide production and neurotoxicity. Nature, 364: 535-537. Martin, L.J., AI Abdulla, N.A., Brambrink, A.M., Kirsch, J.R., Sieber, EE. and Portera-Cailliau, C. (1998) Neurodegeneration in excitotoxicity, global cerebral ischemia, and target deprivation: A perspective on the contributions of apoptosis and necrosis. Brain Res. Bull., 46:281-309. Martinou, J.C. (1999) Apoptosis: key to the mitochondrial gate. Nature, 399:411-412. McCormack, J.G. and Denton, R.M. (1994) Mammalian mitochondrial metabolism and its regulation. In: V.M. DarleyUsmar and A.H.V. Schapira (Eds.), Mitochondria: DNA, Proteins and Disease. Porton Press, London, pp. 81-112. Meldrum, B.S. (1993) Excitotoxicity and selective neuronal loss
in epilepsy. Brain Pathol., 3: 405-412. Nicholls, D.G. (1985) A role for the mitochondrion in the protection of cells against calcium overload?. Prog. Brain Res., 63: 97-106. Ogunmekan, A.O. and Hwang, P.A. (1989) A randomised double-blind, placebo-controlled, clinical trial of D-alphatocophorol acetate (vitamin E), as add on therapy for epilepsy in children. Epilepsia, 30: 84-89. Orrenius, S., Burkitt, M.J., Kass, G.E.N., Dypbukt, J.M. and Nicotera, P. (1992) Calcium ions and oxidative cell injury. Ann. Neurol., 32: $33-$42. Rauca, C., Zerbe, R. and Jantze, H. (1999) Formation of free hydroxyl radicals after pentylenetetrazol- induced seizure and kindling. Brain Res., 847: 347-351. Rong, Y.Q., Doctrow, S.R., Tocco, G. and Baudry, M. (1999) EUK-134, a synthetic superoxide dismutase and catalase mimetic, prevents oxidative stress and attenuates kainateinduced neuropathology. Proc. Natl. Acad. Sci. USA, 96: 9897-9902. Rosenmund, C., Feltz, A. and Westbrook, G.L. (1995) Calciumdependent inactivation of synaptic NMDA receptors in hippocampal neurons. J. Neurophysiol., 73: 427-430. Roy, M. and Sapolsky, R. (1999) Neuronal apoptosis in acute necrotic insults: why is this subject such a mess?. Trends Neurosci., 22: 419-422. Sattler, R. and Tymianski, M. (2000) Molecular mechanisms of calcium-dependent excitotoxicity. J. MoL Med., 78: 3-13. Schinder, A.E, Olson, E.C., Spitzer, N.C. and Montal, M. (1996) Mitochondrial dysfunction is a primary event in glutamate neurotoxicity. J. Neurosci., 16: 6125-6133. Schuchmann, S., Luckermann, M., Kulik, A., Heinemann, U. and Ballanyi, K. (2000) Ca(2+) - and metabolism-related changes of mitochondrial potential in voltage-clamped CA1 pyramidal neurons in situ. J. Neurophysiol., 83: 1710-1721. Stout, A.K., Raphael, H.M., Kanterewicz, B.I., Klann, E. and Reynolds, I.J. (1999) Glutamate-induced neuron death requires mitochondrial calcium uptake. Nat. Neurosci., 1: 366-373. Uz, T., Giusti, P., Franceschini, D., Kharlamov, A. and Manev, H. (1996) Protective effect of melatonin against hippocampal DNA damage induced by intraperitoneal administration of kainate to rats. Neuroscience, 73:631-636. White, R.J. and Reynolds, I.J. (1995) Mitochondria and Na+/Ca 2+ exchange buffer glutamate-induced calcium loads in cultured cortical neurons. J. Neurosci., 15: 1318-1328. Williams, M.D., Van Remmen, H., Conrad, C.C., Huang, T.T., Epstein, C.J. and Richardson, A. (1998) Increased oxidative damage is correlated to altered mitochondrial function in heterozygous manganese superoxide dismutase knockout mice. J. Biol. Chem., 273: 28510-28515. Yokoi, I., Toma, J., Liu, J.K., Kabuto, H. and Mori, A. (1995) Adenosines scavenged hydroxyl radicals and prevented posttraumatic epilepsy. Free Radic. Biol. Med., 19: 473-479. Zhang, Y., Marcillat, O., Giulivi, C., Emster, L. and Davies, J.A. (1990) The oxidative inactivation of mitochondrial electron transport chain components and ATPase. J. Biol. Chem., 265: 16330-16336.
T. Sutula and A. Pitk~inen (Eds.) Progress in Brain Research, Vol. 135 © 2002 Elsevier Science B.V. All rights reserved
CHAFFER 17
Cell death and metabolic activity during epileptiform discharges and status epilepticus in the hippocampus U. Heinemann *, K. Buchheim, S. Gabriel, O. Kann, R. Kovacs and S. Schuchmann Johannes Mailer Institute of Physiology, Charitd, Humboldt University Berlin, D-lOll7 Berlin, Germany
Abstract: Mechanisms of seizure-induced cell death were studied in organotypic hippocampal slice cultures. These develop after withdrawal of magnesium recurrent seizure-like events (SLE), which lead to intracellular and intramitochondrial calcium accumulation. The intramitochondrial Ca accumulation seems to be involved in causing increased production of NADH, measured as NAD(P)H autofluorescence. During SLEs, depolarization of mitochondria and increased production of free radicals is indicated by fluorescence measurements with appropriate dyes. During recurrent seizures, an increased failure to produce NADH is noted while at the same time free radical production seems to increase. This increase and the decline in NADH production could be involved in transition to late recurrent discharges, a phase in which status epilepticus becomes pharmacoresistant. It also coincides with increased cell death as determined with propidium iodide fluorescence. Interestingly, some of these changes can be prevented by application of a-tocopherol, a free radical scavenger, which also has neuroprotective effects under our experimental conditions. The results suggest that free radical-induced mitochondrial impairment is involved in seizure-induced cell death.
Introduction Interictal and ictal discharges indicate synchronized hyperactivity in large ensembles of neurons. These discharges are associated with significant changes of the extracellular ionic microenvironment (Lux et al., 1986). During a seizure, extracellular potassium concentration ([K+]o) can rise to 12 mM, while [Na+]o, [Ca2+]o and the size of the extracellular space decreases (Lux et al., 1986). Consequently intracellular ion concentrations change (Ballanyi et al., 1987; Gloveli et al., 1999) and transmembrane fluxes of C1- appear (Dietzel et al., 1982). Moreover, pH measurements reveal an initial transient alkalosis followed by an acidic shift (Gutschmidt et al.,
*Correspondence to: U. Heinemann, Johannes Mtiller Institute of Physiology, Charit6, Humboldt University Berlin, D-10117 Berlin, Germany. Tel.: +49-30-450528091; Fax: +49-30-4505-28962; E-mail: uwe.heinemann @charite.de
1999). After a seizure, transport processes have to be activated in order to restore ionic gradients. These processes depend on sufficient supply of ATE Biochemical evidence suggests that about 60% of cerebral ATP consumption is used for operation of the electrogenic Na,K-pump which transports three Na ions out of the cell in exchange for two K ions (Ames, 2000). The Na,K-ATPase is activated by intracellular Na accumulation, but some variants of the Na,K-ATPase, particularly those in glial cells, can also be activated by extracellular K accumulation (Grisar et al., 1979). Many other transport processes in nerve cells, such as uptake of glutamate, choline and GABA are dependent on the transmembrane Na gradient. Ca can, in addition to Na/Ca exchange, also be transported by the Ca,Mg-ATPase. The ATP content within a nerve cell is rather limited and other stores for energy production are also scarce. Neurons in the CNS can utilize GABA for ATP production through the GABA shunt and also metabolize lactate (Schousboe et al., 1997; Waagepetersen et al., 1999). Particularly consumption of GABA for ATP
198 synthesis may be a dangerous event, as this would lead to depletion of the GABA pool during recurring seizures. Indeed, transition of recurring seizures to drug resistant late status epilepticus (Dreier and Heinemann, 1991; Zhang et al., 1995) may depend on increased GABA consumption. It is widely held that the energy demands of a group of nerve cells are covered by local adaptation of blood flow (Mathiesen et al., 1998; Caesar et al., 1999), which indeed strongly increases (by up to a factor of seven) (Nilsson et al., 1976; Meldrum, 1983) in areas participating in seizure activity (Horton et al., 1980; Ingvar and Siesjo, 1983). Rises in [K+]o, decreases in Ca, acidosis, release of adenosine and generation of NO seem to be factors involved in this coupling process (Dirnagl et al., 1994; Dirnagl, 1997). Prolonged status epilepticus is a condition which can cause considerable cell loss (Meldrum and Chapman, 1993). This cell loss seems to include glial cells as well (Schmidt-Kastner and Ingvar, 1996). Four hypotheses were proposed to explain status epilepticus-induced cell death. The original idea that energy supply to the brain may be reduced due to systemic factors was rejected early on the basis of glucose consumption and blood flow measurements (Pinard et al., 1984). However, when status epilepticus lasts for a prolonged period, a decline in ATP content (Folbergrova et al., 1985) and a change in the redox potential (Wasterlain and Plum, 1973; Fujikawa et al., 1988) was found, suggesting that during recurring seizures, energy production, in spite of increased supply, may be hampered (Folbergrova et al., 1985). The idea that excitotoxic cell damage alone is responsible for seizure-induced cell death always faced the difficulty that glutamate-induced cell death normally spares glial cells which contribute to cell loss during status epilepticus. More recently, it was suggested that cell death could occur, when intracellular Ca is elevated, causing mitochondrial depolarization (Duchen, 1999). Depolarized mitochondria may exploit 02 incompletely, resulting in an increased production of radical oxygen species (ROS). As a result, mitochondrial function may be compromised, leading to reduced generation of NADH and subsequently reduced production of ATE On the other hand, increases in intracellular Ca concentration may lead to uptake of Ca in mitochondria and
to increased formation of NADH and FADH. Indeed, some enzymes in the tricarboxylic acid cycle are sensitive to Ca and thereby Ca may play an important role in adjusting the ATP production to a given state of neuronal activity (McCormack and Denton, 1993a,b; Hansford and Zorov, 1998). We decided to exploit imaging techniques to get an insight into possible damage cascades during glutamate exposure and during seizures. Methods
The experiments were done on three types of preparations. Studies on glutamate-induced cell damage were done in dissociated hippocampal cell cultures (Schuchmann et al., 1998; Schuchmann and Heinemann, 2000a) with some additional experiments in organotypic slice cultures. Both preparations were performed as previously described (Peacock et al., 1979; Stoppini et al., 1991). Subsequently, we turned to complex entorhinal cortex and hippocampal slices, where recurrent seizures are readily induced by lowering of extracellular Mg concentration or application of 4AP in the entorhinal cortex and neighboring structures, such as the subiculum and the temporal neocortex (Walther et al., 1986). This activity progresses after some time into late recurrent discharges (Dreier and Heinemann, 1991) which are resistant to the presently available anticonvulsant drugs (Zhang et al., 1995). All experiments with respect to ictal activity in this paper were done by removing extracellular Mg concentration. We recently exploited the advantages of organotypic slice cultures. These cultures develop a strong excitatory coupling which leads to facilitated seizure generation during application of low Mg or bicuculline in comparison to age-matched slices (Gutierrez et al., 1999). Dissociated and slice cultures offer the advantage that they can be readily bulk loaded with different dyes which permit imaging of cytosolic and mitochondrial Ca concentration changes, measurements of mitochondrial potentials, formation of ROS and of NAD(P)H. When excited with 360 nm light, NAD(P)H produces a bright autofluorescence. The recordings were done under an upright microscope equipped with a photomultiplier and a CCD camera and a monochromator suitable to generate light with
199 wavelength between 200 and 1000 nm. Most frequently, the photomultiplier was used to sample light emission from area CA3, the hilus and part of area CA1. In slices, we either injected single cells with a given dye or used the NAD(P)H autofluorescence in order to gain insight into mechanisms involved in cellular metabolism. For methodological details see Schuchmann et al. (1999, 2000, 2001) and Kovacs et al. (2001).
Cell death determinations In order to determine cell death in organotypic and dissociated cultures we used propidium iodide staining (Kov~ics et al., 1999). This dye is normally excluded from healthy cells which can be marked by acridine orange, for example. When the plasma membrane is damaged the propidium iodide enters nerve cells and forms a bright fluorescence after binding to RNA and DNA. The intensity of this staining was used to determine the degree of cell loss after 2 h of status epilepticus or exposure to glutamate. Results
Glutamate-induced fluorescence signals in hippocampal dissociated cell cultures and organotypic slice cultures Glutamate dose-dependently induced an increase in cytosolic Ca concentration from a baseline concentration of about 80 nM as determined by ratiometric Fura-2 measurements. Application of 100 ttM of glutamate induced an increase in cytosolic Ca concentration in the order of 400 nM in cultures older then 2 weeks (Schuchmann et al., 1998). This compared to a 300-1000 nM increase in intracellular free Ca concentration in slices during seizure-like events (Gloveli et al., 1999). Application of 100 IxM glutamate for 1 h led to a reduction of viable cells by roughly 60% within 6 h of glutamate exposure and by roughly 80% after 24 h (Schuchmann and Heinemann, 2000a). Application of cyclosporin A but not of tocopherol could protect against this cell death. This suggested an involvement of mitochondria and perhaps development of transition pores in the glutamate-induced cell death.
We therefore decided to obtain more information on the effects of glutamate on mitochondrial potential. For this we employed the fluorescent dye rhodamine-123 which is positively charged and accumulated, therefore, within mitochondria where the fluorescence is quenched. When mitochondria become depolarized, part of the rhodamine-123 leaves the mitochondria resulting in a rhodamine-123 fluorescence increase (see e.g. Schuchmann et al., 1998, 2000). Application of 100 IxM glutamate-induced a pronounced mitochondrial depolarization which was absent when glutamate was applied in the presence of lowered extracellular Ca concentration. The rhodamine-123 fluorescence increase amounted to about 10% in cultures older than 2 weeks. The fluorescence increase was, moreover, dose- and agedependent as well as being dependent on application time. The depolarization of mitochondria might interfere with their capability to generate NADH. We therefore determined the NAD(P)H autofluorescence in cultured hippocampal cells and found that following an initial decrease in NAD(P)H autofluorescence there was a subsequent increase in NAD(P)H fluorescence, suggesting that in spite of mitochondrial depolarizations, the cells were able to generate NAD(P)H. This increase lasted for some 200 s before it returned to baseline (Schuchmann et al., 1998). It was about 3%. in cultures older than 2 weeks. In the presence of depolarized mitochondria, utilization of 02 is less complete and the formation of free radicals is facilitated. We therefore determined whether glutamate-induced Ca load leads to an increased formation of free radical oxygen species. Unfortunately dyes which are used to measure ROS production are not very specific. We therefore compared the oxidation of three dyes which become fluorescent upon oxidation. These were dihydroethidine (HEt), 2'-7'-dichloro dihydroftuorescein (DCF) and dihydrorhodamine (DHR). Upon exposure to 100 IzM glutamate, all three dyes became rapidly oxidized and thereby fluorescent (Fig. 1). Control measurements with biochemical methods indicated there was, indeed, an increased production of ROS species (Schuchmann and Heinemann, 2000b). The increased production of ROS will eventually lead to increased consumption of glutathione. We therefore also studied the effect of glutamate
200
A =~
100 pM glutamate
I
B o~"
100 pM glutamate
n, -1- 0 J 121
C o~"
100 pM glutamate
._~ 0
OJ
o
I I 100 s
Fig. 1. Measurements of ROS production induced by glutamate in cultured hippocampal neurons using different ROS indicators. Application of 100 IxM glutamate for 100 s induced an increase of the fluorescence signal of ethidium, the oxidized form of hydroethidium (HEt, A), rhodamine-123, the oxidized form of dihydrorhodamine (DHR, B) and dichlorofluorescein (DCF, C). All signals were expressed as changes in baseline signal in %.
exposure on the glutathione content in cultured hippocampal neurons. For this, we employed the dye monochlorobimane (MBCL) which predominantly reacts with glutathione - - but only in its reduced form - - to emit a bright fluorescence (Stabel-Burow et al., 1997; Schuchmann and Heinemann, 2000a; Reichelt et al., 1997; Huster et al., 2000). Glutamate applied with 100 txM for 1 h leads to a fall in GSH content by about 5%, which slowly recovers to baseline within 6 h. In the presence of glutamate receptor antagonists (NBQX and 2APV), glutamate instead caused an increase in MBCL fluorescence due to a glutamate-dependent increased synthesis of glutathione which could be further augmented by
cystine or cysteine (Schuchmann and Heinemann, 2000a). The findings suggested that exposure of neurons to elevated glutamate levels induces a Ca-dependent depolarization of mitochondria leading to an increased formation of ROS. The glutamate-induced cell death could be prevented in part by upregulation of glutathione or by application of cyclosporin A, an inhibitor of transition pores in the mitochondria. If such pores are formed, release of cytochromes is expected which might be involved in induction of apoptosis (Bernardi, 1996; Zamzami et al., 1996).
201
Properties of status epilepticus in combined entorhinal cortex hippocampal slices We noted earlier (Walther et al., 1986) that the lowering of Mg can induce different patterns of epileptiform activity in combined slices of the hippocampus and neighboring structures, such as the ento- and perirhinal cortex. Lowering of extracellular Mg concentration in these preparations induces recurrent seizure-like events characterized by slow negative shifts superimposed by tonic- and clonic-like discharges. The SLEs are accompanied by similar ionic changes, as in vivo, and are blocked by anticonvulsant drugs (Zhang et al., 1995; Dreier et al., 1998). These events recur regularly, but after some 20-40 repetitions they change their appearance (Dreier and Heinemann, 1991). The late recurrent discharges are shorter in duration and recur with a relatively high frequency. Studies on the pharmacological sensitivity to clinically employed anticonvulsants has revealed that these late recurrent discharges no longer respond to clinically employed anticonvulsants and thus seem to model the late pharmacoresistant status epilepticus which presents with considerable problems in clinical care. In slices, it was shown that this drugresistant status epilepticus can readily be reversed to treatable status epilepticus, when GABA is supplemented (Pfeiffer et al., 1996). This is in contrast to high levels of midazolam or phenobarbital, which are without effect. The finding that GABA and muscimol can stop the late recurrent discharges in this model of drug-resistant discharges then points to a loss of GABA during recurrent seizures, presumably due to consumption by neurons and glia in the GABA shunt of the tricarboxylic acid cycle. To test this hypothesis further, we studied the effects of anticonvulsants on 4AP-induced seizurelike events. These are similar in appearance to low Mg-induced SLEs, but recur, in our hands, in a somewhat lower frequency (Brtickner and Heinemann, 2000; Buchheim et al., 2000; Schuchmann et al., 1999). They differ from those induced by low Mg in that one type of interictal discharge can persist (albeit reduced in amplitude) when the seizurelike events are blocked by CNQX combined with 2APV, antagonists of ionotropic glutamate receptors (Briickner et al., 2000). These interictal discharges are further reduced in amplitude when bicuculline
is applied (Perreault and Avoli, 1991). This suggests an involvement of GABA in the generation of these events. However, the used concentrations also lead to blockade of glycinergic currents (Shirasaki et al., 1991) and to blockade of Ca-dependent small conductance K channels (Khawaled et al., 1999; Strobaek et al., 2000). The interictal discharges of this type frequently occur just at the onset of a SLE (Lucke et al., 1995). However, thorough counting reveals that this varies from slice to slice and may also change during the course of recurring SLEs. Also in the 4AP model transition to late recurrent discharges is frequently observed. This process can be accelerated when bicuculline is employed together with 4AP (Brtickner et al., 1999). Under that condition, SLEs change almost immediately to late recurrent discharges. The same is observed with the combined application of low Mg and bicuculline (Pfeiffer et al., 1996). Tests on the pharmacological sensitivity of these late discharges revealed that they are also insensitive to clinically employed anticonvulsants, even in the toxic concentration range (Zhang et al., 1995).
Seizure-induced changes in NAD(P )H autofluorescence in entorhinal cortex slices The seizure-like events in parahippocampal structures, such as the subiculum the entorhinal cortex, the perirhinal cortex and neighboring temporal neocortex were characterized by 30-90-s-long negative potential shifts superimposed by initial tonic-like and then clonic-like field potential transients. These events were followed by interictal discharges. Fura-2 measurements revealed rises in [Ca2+]i by 200-900 txM, depending on cell type (Gloveli et al., 1999). Interestingly, the rises in [Ca2+]i were particularly large in layer III neurons, a cell group which is particularly vulnerable during status epilepticus (Du et al., 1993; Du et al., 1995). The SLEs were usually initiated in medial entorhinal cortex from where they spread to neighboring areas (Buchheim et al., 2000). In adult tissue from normal rats, invasion of SLEs into the hippocampus were usually not observed. This was different in slices from juvenile animals (Weissinger et al., 2000) and from adult animals which had previously experienced a pilocarpine status epilepticus or which were kindled by recur-
202 ring stimulation of the amygdala (Behr et al., 1998; Wozny et al., 2000). With time, the appearance of the SLEs changed and after roughly 20-40 SLEs, the activity changed abruptly into late recurrent discharges. In order to test for the hypothesis that metabolism is altered during recurring seizure-like events and may be compromised during transition to late drug-resistant discharges, we measured the changes in NAD(P)H autofluorescence during recurring seizure-like events. We found that each seizure-like event was accompanied by an initial decrease in NAD(P)H autofluorescence followed by a long-lasting increase. With recurring numbers of SLEs, the amplitude of the rises in NAD(P)H autofluorescence declined, while the initial decreases in these signals remained constant. At the time when seizure-like events were replaced by late recurrent discharges, the NAD(P)H overshoots had disappeared (Schuchmann et al., 1999). These findings suggested that recurrent seizures can damage mitochondrial functions and that this process may be involved in causing cell death. Unfortunately, detailed studies with respect to this damage cascade cannot be readily performed in slices. This is due to the fact that slices underwent a period of hypoxia during preparation, that they have damaged axons and dendrites and that due to these alterations also microglial cells become activated. Moreover, the oxygen tension in the slice is variable depending on distance to the cut surface of the slice. Staining of slices with fluorescent probes is also not readily performed as exposure to dyes in stagnant chambers may alter viability of slices further. We therefore took advantage of organotypic slice cultures which also develop seizure-like events when exposed to low Mg concentration (Gutierrez et al., 1999).
Recurring seizure-like events in organotypic hippocampal cultures Unlike hippocampal slices from rats aged 2-3 weeks, where lowering of extracellular Mg concentration induces SLEs only in area CA1 and the subiculum, slice cultures of similar developmental age generate SLEs, which rapidly synchronize throughout the preparation. These events involve the DG, the hilus, area CA3 and CA1 (Gutierrez et al.,
1999; Kov~ics et al., 1999). This is due to the development of aberrant connectivity in the slice culture presumably due to deafferentation and deefferentation in the isolation procedure. In such cultures, recurrent axon collaterals can be demonstrated for mossy fibers and mutual connections between CA1 and CA3 and CA1 and the DG also exist (Gutierrez and Heinemann, 1999). This is actually comparable to the synaptic organization in slices from rats with pilocarpine- or kainate-status epilepticusinduced hippocampal sclerosis where similar aberrant connectivities were also demonstrated (Esclapez et al., 1999; Lehmann et al., 2000, 2001; Smith and Dudek, 2001). The SLEs induced by exposure to Mg free ACSF are rather similar to the SLEs induced in the entorhinal cortex. They recur with an average frequency of one SLE every 15 min. By applying short-stimulus trains to the mossy fibers, such events can also be electrically triggered (Fig. 2). They consist of an initial bursting discharge followed by a tonic- and clonic-like discharge period and a postictal depression (Kovacs et al., 2001) Before and after a SLE, interictal discharges appeared. The ionic changes accompanying such SLEs are quite comparable to those which we observed in intact animals. The [Ca2+]o drops by, on average, 0.6 mM and [K+]o rises to about 9 mM. After the 15th to 20th SLE, recurrent late discharges develop. We preloaded, in the incubator, slice cultures with different dyes and used photomultiplier and imaging techniques to follow the intracellular events during recurring SLEs. Staining with Ca green, an indicator which signals cytosolic Ca concentration changes, revealed that each seizure-like event was characterized by typical intracellular Ca fluctuations. The Ca concentration rose rapidly during the IBP, declined just before the tonic discharge period during which the Ca climbed to a plateau level. During the CLADE the Ca declined slowly although each clonic discharge was accompanied by a transient increase in Ca concentration. These kinetics were very similar from SLE to SLE. However, the amplitudes of Ca green signals declined rapidly by about 30% from the first to 3rd seizure-like event and then remained constant (Kovacs et al., 2001) (Fig. 3A). When we stained the cultures with Rhod-2, a Casensing fluorescent probe, which, due to its positive charge, accumulates within mitochondria, we were
203
TLP
CLADP
IBP
2 mV
0.45 mM
Fig. 2. Typical seizure-likeevents in an organotypichippocampal slice culture, prepared at around P7 and studied about 1 week later. IBP, initial bursting discharge; TLP, tonic-likedischarge phase; CLADP,clonic-likeafter discharge period; fp, field potential recording. The Ca signal was linearizedby us using the Nernst equation. able to monitor changes in intramitochondrial Ca concentration (Fig. 3B). During SLEs, the Rhod-2 fluorescence signals indicated a rapid rise of intramitochondrial Ca during the initial burst discharge followed by a secondary rise during the tonic discharge phase. By contrast to Ca green and Fluo signals single afterdischarges were not reflected in the Rhod-2 signals. Moreover, the decay of the Rhod-2 signals during and following the clonic afterdischarge period was much slower than that of the Ca green signals. These findings suggest that the Rhod-2 fluorescence came from a different compartment than that of the Ca green signals and reliably reflected intramitochondrial rises in [Ca 2+] (see Fig. 3B). During the course of recurring SLEs, the intramitochondrial Ca concentration signals also declined in amplitude. The amplitudes decreased by roughly 60% from the first to the 15th SLE. This is much more than indicated by the cytosolic Ca signals and may point to a loss of mitochondrial function. We also used rhodamine-123 to follow changes in mitochondrial membrane potential. It turned out that each SLE was associated with a mitochondrial depolarization, which during late recurrent discharges reflected in a steadily increased mitochondrial potential. To determine whether production of ROS signals is increased during single seizure-like events, we measured the changes in HEt fluorescence (Fig. 4). We found that each SLE was accompanied by an in-
crease in ethidium fluorescence. However, while Ca signals declined in amplitude these signals increased in amplitude from seizure to seizure. In a further step, we analyzed the NAD(P)H fluorescence signals (Fig. 5). As was the case in slices, these signals also declined in amplitude during recurring seizures and rises in NAD(P)H autofluorescence were abolished shortly before or at the time of transition into late recurrent discharges. These findings indicated that during status epilepticus, Ca enters not only the cytoplasm, but also the mitochondria where they probably stimulated Casensitive enzymes in the tricarboxylic acid cycle, resulting in increased production of NADH. However, as the amplitudes of these signals declined with time, we hypothesized that production of free radicals might have affected the mitochondrial function. In order to test this hypothesis, we pretreated our slice cultures with ct-tocopherol, which is a widely used free radical scavenger acting mostly at lipid membranes. In the presence of ct-tocopherol, the rises in HEt fluorescence were initially reduced while the decline in NAD(P)H autofluorescence signals no longer occurred (Figs. 4 and 5) and transition to late recurrent discharges was protracted. This corroborated the idea that the HEt fluorescence increase was indeed due to increases in ROS production and that ROS-induced impairment of mitochondrial function might be involved in a reduced capability of nerve cells and glia to adapt their cellu-
204
A
CaGreen
avf0
2%
fp ~
~
I 1 mV
ICa"l.
O.4Sm 50 s
SLE, No 1
B
SLE, No 15
Rhod-2
~ 1
Af/fo
2%
[Ca Is
I 0.45mM S
SLE, No 1
SLE, No 15
Fig. 3. Fluorescence signals of Ca green and Rhod-2 during the first (SLE, No. 1) and 15th SLE (SLE, No. 15). Ca green indicates changes in cytosolic Ca concentration and Rhod-2 predominantly changes in intramitochondrial Ca concentration. Simultaneously recorded changes in field potentials (fp) and extracellularCa concentrationare also displayed.
lar metabolism to the energy demands imposed onto cells by increased activity. It was therefore of interest to test whether tocopherol also protected against seizure-induced cell loss.
Propidium iodide staining following 2 h of status epilepticus We have previously shown that slice cultures, when stained with propidium iodide after 2 h of status
205
A Ethidium
untreated
~'~
AV,o ~
fp [Ca2*]e
~
1
k"~"~'J
i 2 o/° 1mY
0.2 mM
SLE, No 15
SLE, No 1
B Ethidium
(~-tocopherol
Af/fo
12%
f0
I 1mY
[Ca2*]e
0.2 mM
SLE, No 1
SLE, No 15
Fig. 4. Effects of recurrent seizures on increases in ethidium fluorescence after staining with hydroethidium(HEt). (A) Note increase in fluorescence signal from the first to 15th seizure-likeevent, while fp and changes in calcium concentrationremain largely unaltered. (B) In the presence of a-tocopherol, the increases in HEt fluorescence are reduced while fp and Ca concentrationchanges are not largely altered.
epilepticus, developed an intense increase in PI fluorescence. This affected all principal cell layers in the organotypic slice culture, namely the granule cell layer, the CA3 region and the CA1 region. In the presence of c~-tocopherol, this fluorescence increase
was much reduced suggesting that seizure associated production of free radicals were indeed involved in status epilepticus-induced cell loss (Fig. 6).
206
A
NAD(P)H untreated
Af/fo __~~
fp
[Ca2+]e
~
SLE, Nol
12mV 0.2 mM
SLE, No15
B
NAD(P)H ~-tocopherol
Af/fo J
~
fp 4~ll~
-~~12mv
[Ca2+]e SLE, No 1
12%
--~Js
0.2 mM
SLE, No 15
Fig. 5. Changes in NAD(P)H autofluorescence during the first and 15th SLE in untreated slice cultures (A) and in slice cultures pretreated with a-tocopherol (B). Note reduced decline in NAD(P)H autofluorescence from the first to the 15th SLE in B.
Discussion and conclusions Our findings suggest that during status epilepticus, Ca has a role in adapting N A D H synthesis in mitochondria and thereby ATP synthesis to the needs for ion homeostasis which require activation of the
Na,K-ATPase. This depends on oxygen and glucose. As during seizures mitochondria depolarize, there is the risk of increased production of free radicals. Due to the increase in blood flow during seizures, oxygen supply may be augmented and consequently the risk of ROS generation could be further enhanced. This
207
nontreated
o~-tocopherol
CA1
CA3 Fig. 6. Exampleof propidium iodide staining in an untreated and c~-tocopherol-treatedslice culture after 2 h of status epilepticus. Note neuroprotectiveeffect of the free radical scavenger.
seems to result in damage of mitochondrial function as indicated by the reduced capability to take up Ca, the decrease in NAD(P)H autofluorescence and the increase in HEt-fluorescence. The fact that some of these alterations as well as cell death can be prevented by application of a free radical scavenger raises the interesting possibility that cell loss during status epilepticus involves generation of free radicals. We have previously shown that the intracellular levels of glutathione influence sensitivity to free
radical-induced cell death (Schuchmann and Heinemann, 2000a). The synthesis of GSH depends on uptake of glycine, glutamate and cystine in glial cells while neurons require glycine and glutamylcysteine for GSH synthesis. The cystine uptake into glia and the cysteine uptake into neurons depend on exchange transport against glutamate. As during seizures extracellular glutamate is elevated, the efficacy of glutamate-cystine antiport and the supply of neurons with cysteine may be hampered while, at the same time, due to increased ROS production, GSH may become oxidized. It was shown in oligodendrocyte cultures as well as in hippocampal dissociated cell cultures (Schuchmann and Heinemann, 2000a) that supply of cysteine can enhance intracellular glutathione levels. In oligodendrocyte cultures, the clinically well known N-acetylcysteine was also effective in cell protection. This might imply that N-acetylcysteine and c~-tocopherol could exert some neuroprotection during status epilepticus. We are on the way to test these interesting hypotheses. The increases in intracellular Ca concentration correspond to those observed during application of about 100 IxM glutamate in dissociated cultures. In addition, this treatment induces cell death which likely involves activation of apoptosis. One signal commonly considered to stimulate apoptosis is mitochondrial release of cytochrome c. We do not yet know whether this also occurs during status epilepticus, but it will be interesting to see whether SE-induced cell death can also be prevented by inhibitors of mitochondrial transition pore formation.
Acknowledgements This research was supported by the BMBF, the SFB 507 C3 and the Graduate College 238: Damage processes in the central nervous system: studies with imaging techniques. We are grateful to H. Siegmund and H.-J. Gabriel for technical assistance.
References Ames III, A. (2000) CNS energy metabolism as related to function. Brain Res. Rev., 34: 42-68. Ballanyi, K., Grafe, P. and ten Bruggencate, G. (1987) Ion activities and potassium uptake mechanisms of glial cells in guinea-pig olfactorycortex slices. J. Physiol., 382: 159-174. Behr, J., Lyson, K.J. and Mody, I. (1998) Enhanced propagation
208
of epileptiform activity through the kindled dentate gyrus. J. Neurophysiol., 79: 1726-1732. Bemardi, P. (1996) The permeability transition pore. Control points of cyclosporin A-sensitive mitochondrial channel involved in cell death. Biochim. Biophys. Acta, 1275: 5-9. Brtickner, C. and Heinemann, U. (2000) Effects of standard anticonvulsant drugs on different patterns of epileptiform discharges induced by 4-aminopyridine in combined entorhinal cortex-hippocampal slices. Brain Res., 859: 15-20. Brtickner, C., Stenkamp, K., Meierkord, H. and Heinemann, U. (1999) Epileptiform discharges induced by combined application of bicuculline and 4-aminopyridine are resistant to standard anticonvulsants in slices of rats. Neurosci. Lett., 268: 163-165. Brtickner, C., Stenkamp, K., Meierkord, H. and Heinemann, U. (2000) Effects of bicuculline and different glutamate receptor antagonists on 4-aminopyridine-induced epileptiform discharges in rat hippocampal entorhinal cortex slices. Neurosci. Res. Commun., 26: 41-49. Buchheim, K., Schuchmann, S., Siegmund, H., Weissinger, F., Heinemann, U. and Meierkord, H. (2000) Comparison of intrinsic optical signals associated with low Mg 2+- and 4aminopyridine induced seizure-like events reveals characteristic features in adult rat limbic system. Epilepsia, 41: 635641. Caesar, K., Akgoren, N., Mathiesen, C. and Lauritzen, M. (1999) Modification of activity-dependent increases in cerebellar blood flow by extracellular potassium in anaesthetized rats. J. Physiol., 520(1): 281-292. Dietzel, I., Heinemann, U., Hofmeier, G. and Lux, H.D. (1982) Stimulus-induced changes in extracellular Na + and Cl- concentration in relation to changes in the size of the extracellular space. Exp. Brain Res., 46: 73-84. Dirnagl, U. (1997) Metabolic aspects of neurovascular coupling. Adv. Exp. Med. BioL, 413: 155-159. Dirnagl, U., Niwa, K., Lindauer, U. and Villringer, A. (1994) Coupling of cerebral blood flow to neuronal activation: role of adenosine and nitric oxide. Am. J. Physiol., 267: H296-H301. Dreier, J.P. and Heinemann, U. (1991) Regional and time dependent variations of low magnesium induced epileptiform activity in rat temporal cortex. Exp. Brain Res., 87: 581-596. Dreier, J.P., Zhang, C.L. and Heinemann, U. (1998) Phenytoin, phenobarbital, and midazolam fail to stop status epilepticuslike activity induced by low magnesium in rat entorhinal slices, but can prevent its development. Acta Neurol. Scand., 98: 154-160. Du, E, Whetsell Jr., W.O., Abou-Khalil, B., Blumenkopf, B., Lothman, E.W. and Schwarcz, R. (1993) Preferential neuronal loss in layer III of the entorhinal cortex in patients with temporal lobe epilepsy. Epilepsy Res., 16: 223-233. Du, E, Eid, T., Lothman, E.W., Kohler, C. and Schwarcz, R. (1995) Preferential neuronal loss in layer III of the medial entorhinal cortex in rat models of temporal lobe epilepsy. J. Neurosci., 15: 6301-6313. Duchen, M.R. (1999) Contributions of mitochondria to animal physiology: from homeostatic sensor to calcium signalling and cell death. J. Physiol., 516(Pt 1): 1-17.
Esclapez, M., Hirsch, J.C., Ben Ari, Y. and Bernard, C. (1999) Newly formed excitatory pathways provide a substrate for hyperexcitability in experimental temporal lobe epilepsy. J. Comp. Neurol., 408: 449-460. Folbergrova, J., Ingvar, M., Nevander, G. and Siesjo, B.K. (1985) Cerebral metabolic changes during and following fluorothylinduced seizures in ventilated rats. J. Neurochem., 44: 14191426. Fujikawa, D.G., Vannucci, R.C., Dwyer, B.E. and Wasterlain, C.G. (1988) Generalized seizures deplete brain energy reserves in normoxemic newborn monkeys. Brain Res., 454: 51-59. Gloveli, T., Egorov, A.V., Schmitz, D., Heinemann, U. and Mtiller, W. (1999) Carbachol-induced changes in excitability and [Ca2+]i signalling in projection cells of medial entorhinal cortex layers II and IlL Eur. J. Neurosci., 11: 3626-3636. Grisar, T., Frere, J.-M. and Franck, G. (1979) Effect of K+-ions on kinetic properties of the (Na+,K+)-ATPase (EC 3.6.1.3) of bulk isolated glial cells, perikarya and synaptosomes from rabbit brain cortex. Brain Res., 165: 87-103. Gutierrez, R. and Heinemann, U. (1999) Synaptic reorganization in explanted cultures of rat hippocampus. Brain Res., 815: 304-316. Gutierrez, R., Armand, V., Schuchmann, S. and Heinemann, U. (1999) Epileptiform activity induced by low Mg 2+ in cultured rat hippocampal slices. Brain Res., 815: 294-303. Gutschmidt, K.U., Stenkamp, K., Buchheim, K., Heinemann, U. and Meierkord, H. (1999) Anticonvulsant actions of furosemide in vitro. Neuroscience, 91: 1471-1481. Hansford, R.G. and Zorov, D. (1998) Role of mitochondrial calcium transport in the control of substrate oxidation. Mol. Cell. Biochem., 184: 359-369. Horton, R.W., Meldrum, B.S., Pedley, T.A. and McWilliam, J.R. (1980) Regional cerebral blood flow in the rat during prolonged seizure activity. Brain Res., 192: 399-412. Huster, D., Reichenbach, A. and Reichelt, W. (2000) The glutathione content of retinal Miiller (glial) cells: effect of pathological conditions. Neurochem. Int., 36: 461-469. Ingvar, M. and Siesjo, B.K. (1983) Local blood flow and glucose consumption in the rat brain during sustained bicucullineinduced seizures. Acta Neurol. Scand., 68: 129-144. Khawaled, R., Bruening-Whright, A., Adelman, J.P. and Maylie, J. (1999) Bicuculline block of small-conductance calciumactivated potassium channels. Pflugers Arch., 438: 314-328. Kov~cs, R., Gutierrez, R., Kivi, A., Schuchmann, S., Gabriel, S. and Heinemann, U. (1999) Acute cell damage after low Mg 2+induced epileptiform activity in organotypic hippocampal slice cultures. NeuroReport, 10: 207-213. Kovacs, R., Schuchmann, S., Gabriel, S., Kardos, J. and Heinemann, U. (2001) Ca 2+ signalling and changes of mitochondrial function during low-Mg2+-induced epileptiform activity in organotypic hippocampal slice cultures. Eur. J. Neurosci., 13: 1311-1319. Lehmann, T.N., Gabriel, S., Kovacs, R., Eilers, A., Kivi, A., Schulze, K., Lanksch, W.R., Meencke, H.J. and Heinemann, U. (2000) Alterations of neuronal connectivity in area CA1 of hippocampal slices from temporal lobe epilepsy patients and
209
from pilocarpine-treated epileptic rats. Epilepsia, 41(Suppl. 6): S190-S194. Lehmann, T.N., Gabriel, S., Eilers, A., Njunting, M., Kov~cs, R., Schulze, K. and Heinemann, U. (2001) Fluorescent tracer in pilocarpine-treated rats shows widespread aberrant hippocampal neuronal connectivity. Eur. J. Neurosci., 14: 83-95. Lucke, A., Nagao, T., Kohling, R. and Avoli, M. (1995) Synchronous potentials and elevations in [K+]o in the adult rat entorhinal cortex maintained in vitro. Neurosci. Lett., 185: 155-158. Lux, H.D., Heinemann, U. and Dietzel, I. (1986) Ionic changes and alterations in the size of the extracellular space during epileptic activity. In: A.V. Delgado-Escueta, A.A. Ward, D.M. Woodbury and R.J. Porter (Eds.), Advances in Neurology, Vol. 44, Basic Mechanisms of the Epilepsies: Molecular and Cellular Approaches. Raven Press, New York, pp. 619-639. Mathiesen, C., Caesar, K., Akgoren, N. and Lauritzen, M. (1998) Modification of activity-dependent increases of cerebral blood flow by excitatory synaptic activity and spikes in rat cerebellar cortex. Z Physiol., 512(Pt 2): 555-566. McCormack, J.G. and Denton, R.M. (1993a) Mitochondrial Ca 2+ transport and the role of intramitochondrial Ca 2+ in the regulation of energy metabolism. Dev. Neurosci., 15: 165-173. McCormack, J.G. and Denton, R.M. (1993b) The role of intramitochondrial Ca 2+ in the regulation of oxidative phosphorylation in mammalian tissues. Biochem. Soc. Trans., 21(3): 793-799. Meldrum, B.S. (1983) Metabolic factors during prolonged seizures and their relation to nerve cell death. In: A.V. Delgado-Escueta, C.G. Wasterlain, D.M. Treiman and R.J. Porter (Eds.), Advances in Neurology. Raven Press, New York, pp. 261-275. Meldrum, B.S. and Chapman, A.G. (1993) Epilepsy and epileptic brain damage. Brain Pathol., 3: 355-356. Nilsson, B., Meldrum, B., Norberg, K. and Siesj6, B.K. (1976) Correlation of changes in blood flow and acid-base changes in the brain during induced epileptic seizures. In: E. Betz (Ed.), Ionic Actions on Vascular Smooth Muscle. Springer, Berlin, pp. 105-116. Peacock, J.H., Rush, D.E and Mathers, L.H. (1979) Morphology of dissociated hippocampal cultures from fetal mice. Brain Res., 169: 231-346. Perreault, E and Avoli, M. (1991) Physiology and pharmacology of epileptiform activity induced by 4-aminopyridine in rat hippocampal slices. J. Neurophysiol., 65: 771-785. Pfeiffer, M., Dragubn, A., Meierkord, H. and Heinemann, U. (1996) Effects of 7-aminobutyric acid (GABA) agonists and GABA uptake inhibitors on pharmacosensitive and pharmacoresistant epileptiform activity in vitro. Br. J. Pharmacol., 119: 569-577. Pinard, E., Tremblay, E., Ben-Ari, Y. and Seylaz, J. (1984) Blood flow compensates oxygen demand in the vulnerable CA3 region of the hippocampus during kainate-induced seizures. Neuroscience, 13: 1039-1099. Reichelt, W., Stabel-Burow, J., Pannicke, T., Weichert, H. and Heinemann, U. (1997) The glutathione level of retinal Mtiller
glial cells is dependent on the high-affinity sodium-dependent uptake of glutamate. Neuroscience, 77: 1213-1224. Schmidt-Kastner, R. and Ingvar, M. (1996) Laminar damage of neurons and astrocytes in neocortex and hippocampus of rat after long-lasting status epilepticus induced by pilocarpine. Epilepsy Res. Suppl., 12: 309-316. Schousboe, A., Westergaard, N., Waagepetersen, H.S., Larsson, O.M., Bakken, I3. and Sonnewald, U. (1997) Trafficking between glia and neurons of TCA cycle intermediates and related metabolites. Glia, 21: 99-105. Schuchmann, S. and Heinemann, U. (2000a) Diminished glutathione levels causes spontaneous and mitochondria-mediated cell death in neurons from trisomy 16 mice: a model of Down's syndrome. J. Neurochem., 74: 1205-1214. Schucbmann, S. and Heinemann, U. (2000b) Increased mitochondrial superoxide generation in neurons from trisomy 16 mice: a model of Down's syndrome. Free Radic. Biol. Med., 28: 235-250. Schuchmann, S., Buchheim, K., Meierkord, H. and Heinemann, U. (1999) A relative energy failure is associated with lowMg 2+ but not with 4-aminopyridine induced seizures-like events in entorhinal cortex. J. Neurophysiol., 81: 399403. Schuchmann, S., Kovfics, R., Kann, O., Heinemann, U. and Buchheim, K. (2001) Monitoring NAD(P)H autofluorescence to assess mitochondrial metabolic functions in rat hippocampal-entorhinal cortex slices. Brain Res. Protoc., 7: 267-276. Schuchmann, S., L~ckermann, M., Kulik, A., Heinemann, U. and Ballanyi, K. (2000) Ca 2+- and metabolism-related changes of mitochondrial potential in voltage-clamped CA1 pyramidal neurons in situ. J. Neurophysiol., 83: 1710-1721. Schuchmann, S., Mtiller, W. and Heinemann, U. (1998) Altered CaZ+-signaling and mitochondrial deficiencies in hippocampal neurons of trisomy 16 mice: a model of Down's syndrome. J. Neurosci., 18: 7216-7231. Shirasaki, T., Klee, M.R., Nakaye, T. and Akaike, N. (1991) Differential blockade of bicuculline and strychnine on GABAand glycine-induced responses in dissociated rat hippocampal pyramidal cells. Brain Res., 561: 77-83. Smith, B.N. and Dudek, F.E. (2001) Short- and long-term changes in CA1 network excitability after kainate treatment in rats. J. Neurophysiol., 85: 1-9. Stabel-Burow, J., Kleu, A., Schuchmann, S. and Heinemann, U. (1997) Glutathione levels and nerve cell loss in hippocampal cultures from trisomy 16 mouse - - a model of Down syndrome. Brain Res., 765: 313-318. Stoppini, L., Buchs, P.-A. and Muller, D. (1991) A simple method for organotypic cultures of nervous tissue. J. Neurosci. Methods, 37: 173-182. Strobaek, D., Jorgensen, T.D., Christophersen, E, Ahring, EK. and Olesen, S.E (2000) Pharmacological characterization of small-conductance Ca(Z+)-activated K(+)channels stably expressed in HEK 293 cells. Br. J. Pharmacol., 129: 991-999. Waagepetersen, H.S., Sonnewald, U. and Schousboe, A. (1999) The GABA paradox: multiple roles as metabolite, neurotransmitter, and neurodifferentiative agent. J. Neurochem., 73: 1335-1342.
210
Walther, H., Lambert, J.D.C., Jones, R.S.G., Heinemann, U. and Hamon, B. (1986) Epileptiform activity in combined slices of the hippocampus, subiculum and entorhinal cortex during perfusion with low magnesium medium. Neurosci. Lett., 69: 156-161. Wasterlain, C.G. and Plum, E (1973) Vulnerability of developing rat brain to electroconvulsive seizures. Arch. Neurol., 29: 3845. Weissinger, E, Buchheim, K., Siegmund, H., Heinemann, U. and Meierkord, H. (2000) Optical imaging reveals characteristic seizure onsets, spread patterns and propagation velocities in hippocampal-entorhinal cortex slices of juvenile rats. Neurobiol. Dis., 7: 286-298.
Wozny, C., Heinemann, U., Gabriel, S. and Behr, J. (2000) The entorhinal cortex entrains epileptiform activity in area CA1 in a model of temporal lobe epilepsy. Eur. J. Neurosci., 12 (Suppl. 11): 102.02 (Abstract). Zamzami, N., Susin, S.A., Marchetti, E, Hirsch, T., G6mezMonterrey, I., Castedo, M. and Kroemer, G. (1996) Mitochondrial control of nuclear apoptosis. J. Exp. Med., 182: 15331544. Zhang, C.L., Dreier, J.P. and Heinemann, U. (1995) Paroxysmal epileptiform discharges in temporal lobe slices after prolonged exposure to low magnesium are resistant to clinically used anticonvulsants. Epilepsy Res., 20:105-111.
T. Sutula and A. Pitk~inen (Eels.) Progress in Brain Research, Vol. 135 © 2002 Published by Elsevier Science B.V.
CHAPTER 18
Summary: Mechanisms of seizure-induced damage Thomas Sutula 1,2,, and Asla Pitk~inen 3,4 I Department of Neurology and 2 Department of Anatomy, University of Wisconsin, Madison, WI 53792, USA 3 Epilepsy Research Laboratory, A.I. Virtanen Institute for Molecular Sciences, and 4 Department of Neurology, Kuopio University Hospital, Kuopio, Finland
Following the pioneering observation of Meldrum and coworkers demonstrating that prolonged seizures during status epilepticus cause neuronal damage, there have been significant experimental insights into the mechanisms by which neurons experiencing synchronous activity may undergo injury and death. These experimental efforts have revealed that neurons may die by at least two distinctive mechanisms, namely, excitotoxic necrosis and apoptosis, although some have pointed out that there may be overlap of these modes of cell death (e.g., Chapter 29). Epilepsy research has not only greatly benefited from, but indeed has played a significant role in advancing fundamental knowledge about mechanisms of neuronal death. Understanding of the ionic fluxes, second messenger systems, and intrinsic death pathways that lead to neuronal loss after severe seizures have proceeded with some success, as indicated by the historical and contemporary perspectives developed in Chapters 1, 15, 16, 17. These chapters provide a detailed review of some of the metabolic pathways activated by seizures, beginning with excessive release of the excitatory neurotransmitter glutamate, and proceeding with second messenger systems including Ca 2+, G-proteins, membrane derived signaling molecules, and eventually involvement of mitochondrial systems, oxidative damage, * Correspondence to: T. Sutula, Department of Neurology H6/570, University of Wisconsin, Madison, WI 53792, USA. Tel.: -t-1-608-263-5448; Fax: +1-608-263-0412; E-mail:
[email protected]
activation of caspases and death proteins, and neuronal death (Chapters 15, 16, 17). The experiments described in these chapters represent the successful use of a reductionistic approach, which has provided increasing insight into the molecular and cellular features of seizure-induced processes that lead to neuronal death, and may eventually permit therapeutic interventions and neuroprotection (Chapters 1 and 44). The reviews of second messenger systems and mitochondrial pathways activated by seizures (Chapter 15, 16) offer a glimpse of the potentially rich opportunity for development of drugs that specifically act on downstream pathways that are sequentially activated after repeated seizures, whether prolonged or brief, which contribute to neuronal dysfunction or death, and ultimately to gene expression that plays a role in adverse consequences of seizures for neural circuitry. The influence of genetics on the consequences of seizures is not limited to seizure-induced gene expression that contributes to long-term effects of seizures on neural circuits (Chapter 12, 13). There is a powerful role of genetic background in determining the acute effects of seizures, as demonstrated by the pronounced differences in susceptibility to acute seizure-induced damage among different mouse strains (Chapter 12). The effects of genetic background and strain differences are often overlooked in interpretation of experiments addressing the effects of seizures. These background genetic influences are sufficient to drastically modify the effects of genes on seizure-induced damage, such
212 as p53 (Chapter 12). The implications of these observations are significant not only for interpretation of experimental studies, but also are potentially of major importance for understanding individual differences in susceptibility to seizure-induced damage from both status epilepticus and repeated seizures in humans. While there has been major interest and emphasis on 'epilepsy genes' that underlie genetic syndromes in human families, these syndromes represent rare or relatively uncommon causes of epilepsy. In the more common idiopathic, cryptogenic, and symptomatic epilepsy syndromes, comprehensive descriptions of the mouse, rat, and human genomes are likely to provide insights into genetic influences on metabolic and signal transduction pathways activated by
seizures, and on gene-dependent individual differences in susceptibility to seizure-induced damage. The application of contemporary genomics-based methods for study of epilepsy are presented in Chapters 13 and 14. These approaches are in their infancy, but offer promise for unraveling the genetic contributions to epileptogenesis and to consequences of seizures in both animals and humans. If genetic background has a powerful effect on acute and long-term susceptibility to damage in animals, is it not likely that these influences are also a potent factor in humans? The reader is encouraged to keep this perspective in mind as the effects of seizures in humans are presented using epidemiological, pathological, imaging, and neuropsychological approaches in Sections III-VI.
T. Sutula and A. Pitk~inen (Eds.) Progress in Brain Research, Vol. 135 © 2002 Published by Elsevier Science B.V.
CHAPTER 19
Do seizures beget seizures? W. Allen Hauser 1,, and Ju R. Lee 2 Sergievsky Center, College of Physicians and Surgeons, Columbia University, 630 W 168 Street, New York, NY 10032, USA 2 California Medical Review Inc., San Francisco, CA 94104, USA
Abstract: There have been suggestions that seizures in some way modify brain function and that each seizure increases the risk for further seizures. Reports thus far on this phenomenon have been flawed because of inappropriate study design. We have evaluated the risk for seizure recurrence following a first unprovoked seizure in a cohort identified at their first unprovoked seizure. Individuals with low risk for a seizure recurrence demonstrate a significant increase in risk for seizure recurrence with increasing numbers of seizures. This is the first time that a progressive increase in risk for seizures with increasing number of seizures has been demonstrated in humans. Since the majority of these cases will ultimately go into remission and discontinue antiseizure medication, there must be competing forces that increase seizure risk and promote seizure suppression. We need appropriate animal models to better understand both processes and their interactions.
Introduction There is substantial data from animals suggesting that seizures cannot only damage the brain but can also result in self-propagation. Most of the animal models deal with seizures rather than epilepsy. The concept of epilepsy as a progressive illness is not new. Gowers in 1881 suggested that each seizure in some way increases the risk for further seizures (Gowers, 1881). Reynolds and colleagues suggested that the risk for persistent epilepsy increased with the number of seizures prior to initiation of treatment for epilepsy (Reynolds, 1995) although this is an indirect assessment of the hypothesis. There are more recent data suggesting that it is seizure density rather than seizure frequency that predicts persistence of epilepsy. None of these studies directly address the question of seizure worsening with increasing numbers of seizures. We have used a patient population followed from their first unprovoked seizure to ad-
* Correspondence to: W.A. Hauser, 630 W 168th St., New York, NY 10032, USA. Tel.: +1-212-305-2447; Fax: +1212-305-2518; E-mail: wahl @spyral.net
dress the question of increasing risk for seizure recurrence with increasing numbers of seizures.
Methods The methods involved in this study are described elsewhere but are briefly reviewed here (Hauser et al., 1998). Patients
A surveillance system was established to identify patients who were evaluated because of newly identified seizures. Once their informed consent was obtained, the patients were screened to determine their eligibility for a series of studies, in which case it was scheduled after discharge. The subjects included in this study were restricted to the patients who at the time of the initial evaluation had had a definite unprovoked seizure, documented by an eyewitness; had no evidence in the history of a previous unprovoked seizure; were identified and signed informed-consent forms within 24 h of the initial seizure; and completed the baseline interview within 30 days of the initial seizure. Altogether, 271 patients were seen
216 and recruited on the day of the index seizure. For 67 of these patients, the intake interview was not completed within 30 days. The remaining 204 patients were included in the analysis.
Classification of seizures All seizures were categorized as partial or generalized on the basis of the description of the onset of the seizure by an eyewitness, according to criteria recommended by the International League against Epilepsy (Commission on Classification and Terminology of the International League Against Epilepsy, 1981; Commission on Epidemiology and Prognosis, International League Against Epilepsy, 1993). Seizures were categorized according to the witness's description, without regard to the findings on EEG or the neurologic examination. Each patient's seizure was also categorized as idiopathic/cryptogenic (seizures occurring in the absence of a documented insult that was thought to increase substantially the risk of unprovoked seizures) or as remote symptomatic (seizures in persons with a history of insult to the central nervous system that was known to increase substantially the risk for subsequent epilepsy). In those with a first unprovoked seizure, information was collected for each subsequent seizure, thus potentially allowing reclassification of the seizure according to type or cause. Persons with status epilepticus (seizures continuing for 30 min or more without interruption) or clusters of seizures (two or more) in the same 24-h period were considered to have had a single seizure. This was an observational study, and we made no attempt to influence the practice of the treating physicians once patients had been identified. Thus, there was neither standardization of treatment (if any) nor systematic monitoring of the adequacy of therapy in patients for whom antiseizure medication was recommended.
Follow-up Subjects were contacted by telephone at 6-month intervals for 2 years from the date of the first seizure and annually thereafter. Data collected included the date, duration, and clinical characteristics of any subsequent seizures, potential precipitating events,
concurrent and previous use of antiseizure medication, and the details of neurologic insults, if any, since the previous follow-up contact. The medical records of those who reported additional seizure activities were reviewed to document the occurrence of seizures and to confirm reported medication use. The medical records generally confirmed the recurrence of seizures, but seldom provided specific information on the type, frequency, or specific dates of seizures. Thus, the information used in our analysis came primarily from the interviews with patients. The medical records of those who did not report additional seizures were also reviewed periodically for other reasons, and no additional patients with seizures were identified. Follow-up was terminated for any of the following reasons: death, the occurrence of an event associated with an increased risk of unprovoked seizures (e.g., head injury with loss of consciousness), or a decision by the patient to terminate participation in the study. Data were obtained on each seizure through the fourth episode. The institutional review board of the University of Minnesota approved all protocols.
Statistical analysis To determine recurrence risk for a second unprovoked seizure, subjects who had experienced a first unprovoked seizure were entered into the analysis on the date of the first unprovoked seizure and were followed until the date of the second unprovoked seizure or the last date of follow-up. A similar strategy was used to evaluate risk for a third/fourth unprovoked seizure following a second/third. Seizure characteristics and etiology was classified based upon most recent data. The cumulative risks of recurrence for subgroup seizure were determined by Kaplan-Meier methods, with an 'event' defined as an unprovoked seizure recurrence (Kaplan and Meier, 1958). The computed risks, therefore, represent the risk of recurrence conditional on survival. 'Events' were classified into subgroups to evaluate the possibility of evidence for increasing risk for seizure recurrence with increasing numbers of seizures. Based on previous work, we have identified groups at differential risk for seizure recurrence (Hauser et al., 1990, 1998). To evaluate the possibility of 'Progression' or 'kindling' effect following a
217
percent recurrence 80 r 60
40 20
0
20
40
60
80
100
120
monthsfrom index seizum I-x-no risk factor 1 to 2 -*-with risk factor 1to 2 --2 to3 --3 to 41 Fig. 1. Estimated recurrence risk from index seizure. first seizure. Six groups were formed based on etiology, family history of epilepsy, EEG G S W feature, number of recurrences and sample size of subgroup. They were: first idiopathic seizure event with no G S W (generalized spike and wave) in EEG and no family history of epilepsy (group 1, the reference group, n = 122); first idiopathic seizure events with presence of G S W in EEG or a family history of epilepsy (group 2, n = 23); first remote symptomatic seizures (group 3, n = 59); second idiopathic symptomatic seizures (group 4, n = 37); third idiopathic seizures (group 5, n = 20); second and third remote symptomatic seizures (group 6, n = 47). The proportional hazards model was used to estimate rate ratios, defined as the ratio of the rate of seizure recurrence in the group of patients with a given factor to the rate of seizure recurrence to the reference group (Cox, 1972). All statistical analyses were performed using the SAS software (SAS). Findings were considered significant when the bounds of the 95% confidence interval did not include unity. All P-values are two tailed. Results
Overall recurrence risk A second seizure occurred in 63 of the 204 (31%) patients identified at the time of the first seizure. A
third seizure occurred in 41 of the 63 (65%) who had a second seizure. A fourth seizure occurred in 63% of those with three seizures. Analysis of the Kaplan Meier curves failed to demonstrate any difference in the estimated percentages in recurrence in those with two and with three seizures.
Recurrence risk within etiologic subgroups The absolute recurrence risk was significantly greater in those with remote symptomatic epilepsy when compared with those with idiopathic/cryptogenic epilepsy, and the Kaplan-Meier curves were different when these two groups were compared (Fig. 1). Because of this, we separately analyzed those with and without a presumed etiology to evaluate the impact of number of seizures on risk of recurrence.
Effect of number of seizures on recurrence risk Estimated recurrence among people with a first unprovoked seizure
People with no identified antecedent. We evaluated the risk for further seizures among people with a first seizure and no clear antecedent. The recurrence risk by 5 years after the initial event was 25% for those with neither EEG abnormalities nor a first-degree relative with epilepsy, 39% for those with either or
218 TABLE 1 Seizure recurrence in each assigned risk group Risk group
Number
1-Year risk
3-Year risk
5-Year risk
1st idiopathic Family H x - and G S W Family Hx + / G S W + 1st remote symptomatic 2nd idiopathic 3rd idiopathic 2nd and 3rd remote symptomatic
122 23 59 37 20 47
13.3% 39.1% 30.8% 49.5% 59.2% 65.0%
20.7% 39.1% 44.5% 52.4% 75.5% 82.2%
24.8% 39.1% 47.9% 64.3% 75.5% 82.2%
both of those risk factors (Table 1). The risk for a third seizure was 64% at 5 years; the risk for a fourth seizure was 76%.
People with an identified antecedent. Among people with an identified antecedent, the risk for a second seizure was 48%. The risk was 82% for a third and a fourth seizure. There was no discernable increment between the third and fourth seizure. Risk for seizure recurrence based upon number of seizures
Unprovoked seizures of unknown cause. We also determined the risk for subsequent seizures using as a referent, the recurrence following a first seizure in the group for low risk of recurrence. We find an incremental increase in risk ratio with increasing numbers of seizures (Table 2). This was true if analysis was restricted to those with no risk factors or to those with a family history and or an abnormal EEG.
TABLE 2 Recurrence rate ratio a by seizure category Risk group
Number
Rate ratio
95% Confidence
2.3 2.2 3.7 5.8 6.4
1.1-4.8 1.3-3.9 2. I--6.5 2.9-11.6 3.8-10.6
1 st idiopathic
Family H x + / G S W + 23 1st remote symptomatic 59 2nd idiopathic 37 3rd idiopathic 20 2nd and 3rd remote 47 symptomatic
a The reference group consisted of patients with 1st idiopathic seizure, EEG negative of GSW and no family history of epilepsy (n = 122).
Unprovoked seizures with an identified etiology. Unlike those with epilepsy of unknown cause, we saw no progression in risk after the second seizure (Table 2). Discussion In his 1881 text, Gowers wrote "When one attack has occurred, whether in apparent consequence of an immediate excitant or not, others usually follow without any immediate traceable cause. The effect of a convulsion on the nerve centers is such as to render the occurrence of another more easy, to intensify, the predisposition that already exists. This every fit may be said to be, in part, the result of those that have preceded it, the cause of those which follow it." The first component of this hypothesis seems answered. Acute symptomatic seizures do not in conventional terms beget (unprovoked) seizures. In studies of brain injury, early seizures seem a surrogate for severity of head injury (Annegers et al., 1998). Similarly, acute symptomatic seizures in stroke are associated with size and location of the lesion (Labowitz). These same factors are associated with an increased risk for epilepsy following stroke. Acute symptomatic seizures associated with systemic metabolic disturbance do not increase the risk for subsequent unprovoked seizures (Hesdorffer et al., 1998). There is a considerable literature on the question of seizures begetting seizures in humans but most consist of review articles with indirect evidence used to support varying opinions on the outcome. Only the article by Elwes et al. (1988) attempted to address this question directly. This study seems fatally flawed from a methodological standpoint as reviewed (Berg
219
and Shinnar, 1997). The other studies in humans use indirect evidence to answer this question by assessing remission in people with epilepsy. This is the first study to address the question of an escalating risk for unprovoked seizures in humans. The analyses would seem to support an increasing risk for further seizures with increasing numbers of seizures. In this small sample, the phenomenon can be identified only in those with epilepsy of unknown cause. If there is accelerating pathology associated with seizures, it would appear to occur to appear only in those with more subtle pathology underlying the epileptogenic tendency. This study does not address the question of treatment effect. Treatment was recommended for the majority of people in the current study, but seldom was optimal dosing used, and most did not take medications as recommended. These data suggest that a complex process of competing forces is involved in the persistence of epileptogenesis. A tendency for increasing ease of seizure occurrence is counterbalanced by a drive to inhibit seizures. From our epidemiologic studies it would appear that for most people with recurrent unprovoked seizures, the factors favoring normalization are dominant. This seems true regardless of therapeutic interventions (Feksi et al., 1991). To understand the dynamics of the process of epileptogenesis, we need animal models of epilepsy rather than acute seizures, and need to be able to distinguish the effects of the mechanisms used to induce seizures from the effects of the seizures per se. We also need to understand the mechanisms associated with resolution of the tendency to have seizures.
References Annegers, J.E, Hauser, W.A., Coan, S. and Rocca, W.A. (1998) A population-based study of seizures after traumatic brain injuries. New Engl. J. Med., 338: 20-24. Berg, A.T. and Shinnar, S. (1997) Do seizures beget seizures? An assessment of the clinical evidence in humans. J. Clin. Neurophysiol., 14: 102-110. Commission on Classification and Terminology of the International League Against Epilepsy (1981) Proposal for revised clinical and electroencephalographic classification of epileptic seizures. Epilepsia, 22: 489-501. Commission on Epidemiology and Prognosis, International League Against Epilepsy (1993) Guidelines for epidemiologic studies on epilepsy. Epilepsia, 34: 592-596. Cox, D.R. (1972) Regression models and life-tables. J. R. Stat. Soc. (B), 34: 187-220. Elwes, R.D., Johnson, A.L. and Reynolds, E.H. (1988) The course of untreated epilepsy. Br. Med. J., 297: 948-950. Feksi, A., Kaamugisha, J., Gatiti, S., Sander, J. and Shorvon, S. (1991) Comprehensive primary health care antiepileptic drug treatment programme in rural and semi-urban Kenya. Lancet, 337: 406-409. Gowers, W.R. (1881) Epilepsy and other chronic convulsive diseases. London, Churchill. Hauser, W.A., Rich, S.S., Annegers, J.E and Anderson, V.E. (1990) Seizure recurrence after a 1st unprovoked seizure: an extended follow-up. Neurology, 40:1163-1170. Hauser, W.A., Rich, S.S., Lee, J.R., Annegers, J.E and Anderson, V.E. (1998) Risk of recurrent seizures after two unprovoked seizures. New Engl. J. Med., 338: 429-434. Hesdorffer, D.C., Logroscino, G., Cascino, G., Annegers, J.E and Hauser, W.A. (1998) Risks of unprovoked seizure following acute symptomatic seizure: effects of status epilepticus. Ann. Neurol, 44: 908-912. Kaplan, E. and Meier, E (1958) Nonparametric estimation from incomplete observations. J. Am. Stat. Assoc., 53: 457-481. Reynolds, E.H. (1995) Do anticonvulsants alter the natural course of epilepsy? Treatment should be started as early as possible. Br. Med. J., 310: 176-177. SAS Software for HP-UX, Release 6.12. SAS Institute Inc., Cary, NC.
T. Sutulaand A. Pitk~inen(Eds.) Progress in Brain Research, Vol. 135 © 2002 ElsevierScienceB.V. All rightsreserved
CHAPTER 20
Do occasional brief seizures cause detectable clinical consequences? Shlomo Shinnar 1,, and W. Allen Hauser 2 l Comprehensive Epilepsy Management Center, Montefiore Medical Center, Albert Einstein College of Medicine, Bronx, NY 10467, USA 2 Sergievsky Center, College of Physicians and Surgeons, Columbia University, New York, NY 10032, USA
Abstract: Seizures, particularly when prolonged or frequent, have been associated with a variety of adverse outcomes. However, epidemiological data provide little evidence for adverse effects of isolated brief seizures per se. Even the animal data is mostly for prolonged or frequent seizures. Febrile seizures lasting < 10 min have not been associated with adverse seizures or cognitive outcomes. Treating either febrile seizures or other acute symptomatic seizures does not reduce the risk of subsequent epilepsy. In subjects with a first unprovoked seizure, seizure duration does not influence recurrence risk. Furthermore, treatment after a first unprovoked seizure reduces recurrence risk, but does not alter long-term prognosis. In epidemiological studies of newly diagnosed epilepsy, the number of seizures prior to therapy does not influence prognosis. There are a variety of specific epilepsy syndromes associated with poor cognitive outcomes and with progressive loss of function. However, the poor outcomes in these syndromes do not appear to be the result of seizures per se but rather to the specific syndrome and to the frequent interictal spike activity seen in these patients. Antiepileptic drugs, while effective in reducing seizure recurrence are also associated with a variety of potential adverse effects. On a risk-benefit basis, the available epidemiologic data do not justify starting treatment after the first seizure to attempt to influence long-term prognosis.
Introduction
Gowers (1881) wrote that "The tendency of the disease is toward self-perpetuation; each attack facilitates the occurrence of another by increasing the instability o f the nerve elements." This became d o g m a for many years and provided the justification for treatment after the first seizure and for the view that what made epilepsy intractable was not the underlying disorder, but rather that each seizure promoted the occurrence of the next seizure and made the epileptic focus more severe (Shorvon and Reynolds, 1982; Reynolds et al., 1983; Elwes et al., 1984;
Reynolds, 1995). In other words, that seizures beget seizures. More recent epidemiological data have cast considerable doubt on this view (Sander, 1993; Chadwick, 1995; Camfield et al., 1996; Shinnar and Berg, 1996; Berg and Shinnar, 1997; Sillanpaa et al., 1998a). There is no question that even a brief seizure, if it occurs in the wrong place at the wrong time (e.g. while swimming or driving) can have disastrous consequences. However, the evidence that one or several brief seizures per se have significant long-term functional consequences is less clear. This paper will review the clinical evidence that occasional brief seizures can affect long-term outcomes. Febrile seizures
* Correspondence to: S. Shinnar, Comprehensive Epilepsy Management Center, Montefiore Medical Center, 111 E 210th St., Bronx, NY 10467, USA. Tel.: +1-718-9204378; Fax: +1-718-655-8070; E-mail:
[email protected]
Febrile seizures are a form of acute symptomatic seizures (National Institutes of Health, 1980; Commission on E p i d e m i o l o g y and Prognosis, 1993).
222 They occur in 2 - 5 % of children and are the most common form of childhood seizures. In the past, it was believed that most febrile seizures represented a form of epilepsy and that the prognosis was not favorable. Febrile seizures were believed to cause brain damage as well as subsequent epilepsy (Taylor and Ounsted, 1971; Wallace, 1980). However, data from large epidemiological studies over the past 25 years have shown that they are largely benign and are not associated with adverse outcomes (Ellenberg and Nelson, 1978; Nelson and Ellenberg, 1978; Annegers et al., 1979b, 1987; Maytal and Shinnar, 1990; Verity and Golding, 1991; Verity et al., 1998; Shinnar et al., 2001a). Consideration of febrile seizures is relevant as for many years they were thought to cause epilepsy. Whether or not febrile seizures cause mesial temporal sclerosis remains one of the more controversial topics in epileptology (Shinnar, 1998). A small proportion of children with febrile seizures do develop subsequent epilepsy, but the risk is increased primarily in those with complex febrile seizures, especially when prolonged, those who are neurologically abnormal and those with a family history of epilepsy (Nelson and Ellenberg, 1978; Annegers et al., 1987; Verity and Golding, 1991; Berg and Shinnar, 1996a; Hesdorffer and Hauser, 2002). Cognitive outcomes are also favorable (Ellenberg and Nelson, 1978; Verity et al., 1998; Chang et al., 2000, 2001; Hirtz, 2002). Children with simple febrile seizures perform as well as controls in measures of intelligence and academic performance and behavior in large studies that utilized either sibling (Ellenberg and Nelson, 1978) or population based (Verity et al., 1998; Chang et al., 2000) controls. A recent population based study from Taiwan reported that children with simple febrile seizures perform at least as well as population based controls on memory tasks (Chang et al., 2001). Recent animal data has suggest that prolonged febrile seizures may lead to long-lasting changes in hippocampal circuits. In a rat model of prolonged febrile seizures, cyto-skeletal changes in neurons were evident within 24 h and persisted for several weeks without leading to cell loss (Toth et al., 1998). However, altered functional properties of these injured neurons were evident (Chen et al., 1999; Dube et al., 2000; Dube, 2002). Thus, it appears that exposure of hippocampal neurons to prolonged febrile
seizures in early childhood may lead to transient injury and more sustained dysfunction of these neurons. However, even in this model that has produced convincing data for functional changes, required a seizure duration of 20 min or more. Seizures lasting 10 min or less were not associated with any anatomic or functional changes. Recent studies of prolonged febrile seizures that have imaged children within 72 h of the seizure episode have demonstrated acute changes on MRI in some cases which were followed by later chronic changes in a few children (VanLandingham et al., 1998; Mitchell and Lewis, 2002). These studies provide the most convincing evidence to date in humans of seizure induced injury. However, these were very prolonged febrile seizures with a mean duration of over 90 min. Of note is that only children with seizures that were both focal and prolonged developed either acute changes or mesial temporal sclerosis (MTS). Also to date, only one of the children who developed MTS has gone on to have temporal lobe epilepsy. Long-term follow-up of such cohorts utilizing both clinical and imaging data may provide more answers as to how often this phenomenon actually occurs. While of great interest, these studies are not directly relevant to the issues of whether brief seizures cause damage. If brief seizures promote the occurrence of subsequent seizures, then preventing febrile seizures should reduce the risk of subsequent epilepsy. There have been three well designed randomized clinical trials for preventing febrile seizure recurrence (Wolf et al., 1977; Knudsen and Vestermark, 1985; Rosman et al., 1993a) which have also assessed the risk of subsequent epilepsy (Wolf and Forsythe, 1989; Rosman et al., 1993b; Knudsen et al., 1996). In two of these studies, follow-ups of 10 or more years were available (Wolf and Forsythe, 1989; Knudsen et al., 1996). Despite the fact that the treatment arm was effective in reducing the risk of recurrent febrile seizures in all three studies, there was no difference in the rate of developing epilepsy in the treatment arm compare to those untreated in any of the three studies. In general, there is no evidence from prospective randomized trials that treating febrile seizures or any other form of acute symptomatic seizures prevents subsequent epilepsy (Shinnar and Berg, 1996; Berg and Shinnar, 1997;
223 Knudsen, 2002). While these trials included only a small number of children with very prolonged febrile seizures, the implications in terms of brief seizures are clear and consistent with those seen in other settings such as other forms of acute symptomatic seizures and first unprovoked seizures.
First unprovoked seizure
Other acute symptomatic seizures
Observational epidemiological studies
The occurrence of acute symptomatic seizures other than febrile seizures is associated with an increased risk of subsequent epilepsy but the data do not suggest a causal relationship between brief acute symptomatic seizures and subsequent epilepsy. The data, in both children and adults, for acute symptomatic seizures such as post-traumatic and post-craniotomy are similar to the data from randomized treatment trials for febrile seizures. The occurrence of seizures in the immediate post-traumatic period following head trauma is associated with an increased risk of subsequent epilepsy (Annegers et al., 1980). Randomized clinical trials have demonstrated that antiepileptic drugs (AEDs) are effective in preventing acute posttraumatic seizures (Temkin et al., 1990). However, prevention of acute post-traumatic seizures using AEDs does not seem to alter the risk of developing post-traumatic epilepsy a year or two later (Temkin et al., 1990). After controlling for other factors, acute post traumatic seizures do not alter prognosis (Hauser et al., 1984; Annegers et al., 1998). We have previously argued (Hauser et al., 1984; Shinnar and Berg, 1996; Annegers et al., 1998) that these seemingly contradictory findings are best explained by regarding the occurrence of seizures in the acute post-injury period as a marker for the severity of the brain injury and its epileptogenic potential, not the cause itself of later epilepsy. One can mask or suppress this marker with AEDs, but doing so does not alter the severity of the injury or its later consequences, specifically the development of epilepsy. If the seizures per se resulted in damage, then preventing them should alter the subsequent clinical course. Similar data have been reported from studies of seizures in the acute post-craniotomy period where pretreatment with AEDs will prevent the occurrence of seizures in the immediate postoperative period but does not influence the subsequent risk of developing epilepsy (Foyet al., 1992).
Following a first unprovoked seizure the best estimates of the risk of seizure recurrence within 2 years are approximately 40% (95% confidence interval 37-43%) (Berg and Shinnar, 1991). These data come from a meta-analysis of prospective studies in children and adults done up to 1990. Subsequent studies have produced similar recurrence risks (Van Donselaar et al., 1992; First Seizure Trial Group, 1993; Shinnar et al., 1996, 2000; Hauser et al., 1998; Stroink et al., 1998). Factors consistently associated with an increased risk of recurrence include a remote symptomatic etiology, an abnormal electroencephalogram and the occurrence of the first seizure in sleep (Hauser et al., 1982, 1990; Camfield et al., 1985; Annegers et al., 1986; Hopkins et al., 1988; Shinnar et al., 1990, 1996, 2000; Berg and Shinnar, 1991; Van Donselaar et al., 1992; First Seizure Trial Group, 1993; Hauser et al., 1998; Stroink et al., 1998; Hirtz et al., 2000). These factors appear to be relevant in both children and adults. The duration of the first seizure is not associated with a differential risk of recurrence in patients with cryptogenic or idiopathic seizures (Shinnar et al., 1996, 2001b). The data regarding seizure duration is of particular relevance to this discussion. If seizures 'begat' seizures, then the risk of recurrence should be higher following prolonged seizures than following brief seizures. This is not the case. However, in children with a first unprovoked seizure who do experience a recurrence, the duration of the second seizure is correlated with the duration of the first seizure. Thus, in the 137 children with recurrent seizures whose first seizure lasted <10 min, the second seizure lasted >10 min in 11 (8%), >20 min in 5 (4%) and >30 min in only 2 (1%). On the other hand, in the 25 children with recurrent seizures whose first seizure lasted >30 rain, the second seizure lasted >10 min in 11 (44%), _>20 min in 9 (36%) and >30 min in 6 (24%) (P < 0.001) (Shinnar et al., 2001b). The
There is a substantial amount of data from prospective studies of children and adults who present with a first unprovoked seizure. These include observational studies as well as randomized therapeutic trials
224 same results have been reported for febrile seizures. Having a prolonged febrile seizure does not increase the risk of another febrile seizure (Nelson and E1lenberg, 1978; Annegers et al., 1990; Berg et al., 1990, 1992, 1997; Offringa et al., 1992, 1994; Berg and Shinnar, 1996b). However, if another febrile seizure does occur, it is more likely to be prolonged (Berg and Shinnar, 1996b). Twin studies have also shown that if one twin has status epilepticus, the other twin is at higher risk not just for seizures but for status epilepticus (Corey et al., 1998). In a population-based study of patients with childhoodonset epilepsy followed for over 30 years, if status epilepticus did occur, the first episode occurred early in the course of the disorder (Sillanpaa et al., 1998b). Furthermore, the occurrence of status epilepticus in otherwise normal children does not significantly alter long-term prognosis or remission following either a first unprovoked seizure (Shinnar et al., 1995) or newly diagnosed childhood onset epilepsy (Sillanpaa et al., 1998a,b; Berg et al., 1999, 2001). Thus the epidemiologic data argue for a subgroup of children with a predisposition for prolonged seizures rather than the duration of the seizures altering prognosis. Randomized treatment trials following a first unprovoked seizure If seizures indeed do beget seizures then early treatment may be crucial to preventing the evolution of a chronic process. There are four randomized clinical trials that have examined the effect of treatment after a first seizure in both children and adults (Camfield et al., 1989; Chandra, 1992; First Seizure Trial Group, 1993; Gilad et al., 1996). All four studies reported that recurrence risk in the treatment arm was reduced by 50% or more. The magnitude of the reduction ranged from 51% in the Italian multicenter study of 417 children and adults (First Seizure Trial Group, 1993) which is the largest study with the longest follow-up, to 97% reduction in risk in the study by Chandra which included 91 adults. The Italian study followed the subjects long term to address the crucial question of whether delayed treatment affects long-term prognosis. While treatment following the first seizure reduced recurrence risk (First Seizure Trial Group, 1993), it did not alter the probability of attaining 1- or 2-year remission
in long-term follow-up (Musicco et al., 1997). This is the only study addressing the important issue of prevention of future epilepsy, and does not support a long-term benefit of treatment after a single seizure. Number of seizures and outcome
What is the clinical evidence that occasional brief seizures adversely affect outcome? In a well known series of papers, Reynolds and colleagues (Shorvon and Reynolds, 1982; Reynolds et al., 1983; Elwes et al., 1984) reported that in patients with newly diagnosed epilepsy, the probability of attaining 1year remission following initiation of AED therapy was inversely proportional to the number of seizures prior to the initiation of therapy. These data are frequently cited as clinical evidence that even brief seizures are associated with a worse prognosis and that treatment therefore needs to be initiated after the first seizure in order to prevent the development of intractable epilepsy (Reynolds, 1995). However, there are several flaws in this argument. This was not a randomized study. The reason there are patients in this study who were not treated until they had 10 or more seizures, is because they did not present to medical attention until that time. Therefore, the subjects with higher numbers of seizures prior to treatment were those with complex partial seizures who did not get diagnosed until later whereas those with tonic-clonic seizures presented to medical attention after only a few seizures. Careful scrutiny of the data from this British study reveals that outcome was only correlated with the number of complex partial seizures prior to initiating AEDs. There was no difference in the number of generalized tonicclonic seizures in those who did and did not respond to therapy (Shorvon and Reynolds, 1982). Complex partial seizures are known to be associated with a poorer prognosis and a lower probability of attaining remission on medications than tonic-clonic seizures (Annegers et al., 1979a). Therefore, these data do not really provide solid evidence for an adverse effect of brief seizures on long-term prognosis. Several other studies have found that the number of seizures prior to treatment does not alter longterm prognosis. A collaborative multi-center study from Italy reported no difference in outcomes between patients where AED therapy was started after
225 two to five seizures and those where therapy was not initiated until six or more seizures had occurred (Collaborative Group for the Study of Epilepsy, 1992). In a population-based study of childhood onset seizures in Nova Scotia, Canada, having up to 10 seizures prior to initiation of AED therapy did not alter the likelihood of attaining remission (Camfield et al., 1996). While children with more than 10 seizures had a lower chance of entering remission, this group was heavily weighted toward those with complex partial seizures (59% if more than 10 seizures vs 16% if 10 or fewer seizures). Note that none of these studies randomized to delayed versus immediate therapy. In all of them, therefore, the group with a higher number of seizures prior to initiation of therapy was skewed to those with complex partial seizures. In the studies from developed countries, essentially all patients who presented to medical attention with two or more seizures were treated. The differences between groups with multiple seizures before treatment and those with only a few may therefore be due to the differences in the underlying epilepsy syndromes. Ideally, to determine if delayed treatment has an impact on prognosis, one would design a randomized study of early versus late treatment. However, ethical issues prevent these studies being carried out except in the case of a first seizure or of specific benign childhood epilepsy syndromes, such as benign rolandic epilepsy. In these benign syndromes, however, randomized treatment trials may provide useful information about the efficacy of the drug in preventing seizures but are unlikely to provide useful data regarding long-term outcomes due to the generally favorable clinical course of these syndromes. The situation in developing countries, where AEDs may not be readily available, is different. There, even patients with tonic-clonic seizures may be untreated due to lack of resources. This unfortunate situation allows one to examine whether patients whose treatment was delayed respond as well as newly diagnosed patients (Feksi et al., 1991; Sander, 1993). In a randomized trial comparing two antiepileptic drugs in drug-naive patients in Kenya, the response to treatment was quite comparable to what is seen in more economically developed countries in patients with newly diagnosed epilepsy
(Feksi et al., 1991). Half of the patients in the Kenyan study had epilepsy of more than 5-years' duration, and about a third had a history of more than 100 generalized tonic-clonic seizures. Within the trial, patients with longer duration of epilepsy and those with a history of more than 100 seizures responded equally well to medication as did those with epilepsy of briefer duration and fewer seizures. Such comparability of outcomes would not be expected if the occurrence of seizures did, in fact, exacerbate the underlying disease or process. The epidemiologic data from developing countries has been comprehensively reviewed by Sander (1993). Further evidence against an adverse impact of occasional brief seizures on long-term prognosis comes from longitudinal epidemiological studies of the prognosis of childhood onset seizures. Sillanpaa et al. (1998a) followed a population based cohort of childhood onset epilepsy in Turku Finland for over 30 years. On multivariable analysis of predictors of remission or lack of remission, the number of seizures prior to treatment was not significant. Rather, it was whether the patient responded to treatment within the 3 months or not. Berg et al. (2001) have examined early predictors of intractability in a community based cohort of 613 children with newly diagnosed epilepsy in Connecticut. The number of seizures prior to diagnosis was not significant. What was very significant was the seizure frequency. Children with a higher seizure frequency were more likely to develop medically refractory epilepsy within a few years of diagnosis. Both these studies indicate that one can indeed identify subjects who may not do well early, but that it is the underlying biology of the particular epilepsy syndrome that is important rather than a specific number of seizures. The results from the British National General Practice Study of Epilepsy are also consistent with this finding (Cockerell et al., 1997; MacDonald et al., 2000). In this analysis, the probability of remission was related to the number of seizures in the 6-month period after the first identified seizure. While the authors talk about the number of seizures as the predictive variable in the study, it is clear that in fact they were looking at seizure frequency rather than the absolute number of seizures. Having more than 10 seizures prior to diagnosis was not associated
226 with a differential probability of remission. However, having > 10 seizures in the 6 months following diagnosis, presumably on AED therapy though this is not explicitly stated, was associated with a substantially reduced chance of remission. The number of seizures in 6 months is a measure of seizure frequency rather than of absolute number of seizures. Thus these results are very consistent with those from epidemiological studies of childhood onset seizures discussed above (Sillanpaa et al., 1998a; Berg et al., 2001). Seizure frequency at the time of initial referral was also reported to be associated with long-term prognosis in a recent study from Japan (Ohtsuka et al., 2001). Time course of seizures
In a retrospective analysis of newly diagnosed patients with epilepsy, Elwes et al. (1988) reported that untreated epilepsy follows a progressive course with decreasing intervals between seizures and argued that this demonstrates the Gowers hypothesis that seizures beget seizures. However, other studies have not found this progression. In the Dutch prospective study of newly diagnosed childhood epilepsy, Van Donselaar et al. (1997) found that the interseizure interval was variable showing a decelerating course in some and an accelerating course in others with no consistent pattern being found. They concluded that their data did not demonstrate a progressive course and that fear of disease progression should not be used as an argument for early therapy. In a prospective study of children followed from the time of their first seizure, Shinnar et al. (2000) examined the risk of subsequent seizures after each seizure. Once a second seizure has occurred, the risk of subsequent seizures was approximately 70%, an observation that has also been made in adults (Hauser et al., 1998). However, the subsequent risk of seizures remained approximately the same over the first five or six seizures. In other words, the probability of a third seizure once a second seizure has occurred was the same as the probability of a fourth seizure after a third seizure which was the same as the probability of a fifth seizure once a fourth seizure has occurred. These findings argue against a progressive course as a result of a few seizures early in the course of the disorder.
In the current volume, Hauser and Lee (Chapter 19) report an increasing risk for seizure recurrence with increasing numbers of seizures in a select group of people with a first seizure. Paradoxically, this finding is limited to those with no risk factors for recurrence - the group with lowest initial recurrence risk and a group with high probability of remission (Shafer et al., 1988). Thus the long-term impact of this phenomenon which was not found in the analysis of Shinnar et al. (2000) of a similar data set remains unclear. Seizures or epileptiform EEG?
Epileptiform activity on the electroencephalogram (EEG) is the hallmark interictal finding in patients with epilepsy. Interictal spikes are by definition not a seizure but an interictal signature of the seizure focus. They do represent abnormal brain activity. Paradoxically, the evidence for damage for interictal spikes is far more convincing than the evidence for actual clinical seizures. The issues of kindling and of secondary epileptogenesis are discussed elsewhere in this volume and are beyond the scope of this discussion. It should be noted that, while of theoretical concern, their occurrence in humans remains controversial (Goldensohn, 1984; Engel and Shewmon, 1991). However, there are several epilepsy syndromes where it is the interictal spikes rather than the seizures per se that are thought to be responsible for the damage. The best known of these are continuous spike wave in sleep also known as electrographic status epilepticus in sleep (Tassinari et al., 1992), the Landau-Kleffner syndrome (Beaumanoir, 1992) and infantile spasms (Jeavons and Livet, 1992).
Epilepsy with electrical status epilepticus during sleep (ESES) also known as epilepsy with continuous spikes and waves during slow sleep (CSWS) Children with this form of childhood onset epilepsy usually have a variety of seizure types both generalized and partial (Tassinari et al., 1992). The seizures are not usually medically refractory. The hallmark EEG of this syndrome is that during slow wave sleep, 80% or more of the EEG tracing consists of spike and wave activity. Cognitive and behavioral deterioration are common in this syndrome with the specific
227 area of deterioration often associate with the location of the spike focus. It may involve language, memory, executive functions, affect and behavior. The deterioration occurs even if the clinical seizures are well controlled with AED therapy. Seizures usually remit in adolescence but the cognitive and behavioral problems often persist. Treatment is therefore aimed not just at seizure control but at elimination of the abnormal electrical activity and therefore involves agents, such as steroids, which are thought to suppress not just seizures but also the underlying EEG focus.
Acquired epileptic aphasia (Landau-Kleffner syndrome) The Landau-Kleffner syndrome (LKS) is a childhood disorder consisting of an acquired aphasia and epileptiform discharges involving the temporal or parietal regions of the brain (Beaumanoir, 1992). Onset is usually between age 4 and 10 years. The majority of children (70-80%) have clinical seizures, usually readily controlled with AED therapy, but the occurrence of clinical seizures is not part of the diagnostic criteria. The onset of the aphasia may be acute or insidious and may precede or follow the onset of clinical seizures by several months. LKS is a predominantly receptive aphasia with verbal auditory agnosia being the typical language disorder though many children become mute. Many children learn to communicate using sign language. The EEG criteria for the diagnosis of LKS are less well defined than those for ESES (Beaumanoir, 1992). The most common pattern is of spikes or spike-and-wave activity, usually in the temporal or parietal-occipital area accentuated by sleep. There are no clear frequency criteria for the spikes and all night EEG recordings may be needed to detect them. While LKS is a clinically distinct entity, there is some debate as to how much it overlaps with ESES, but this is beyond the scope of this review. The seizures in LKS, if they occur, readily respond to conventional AED therapy. However, the language deterioration occurs despite good seizure control and is thought to be related to the interictal epileptiform activity involving the temporal lobes, though the mechanism for this is not established. Seizures essentially always remit in this syndrome.
This is why it was considered a relatively benign form of childhood epilepsy for many years. However, while language function may improve, deficits in language, often quite severe, persist even after the disappearance of the spike focus (Bishop, 1985; Marescaux et al., 1990; Dugas et al., 1991). As is the case with ESES, treatment is aimed not so much at suppressing seizure which are usually infrequent and respond to any of the standard AEDs, but at suppression of the spike focus and even reversal of he underlying encephalopathic process. Drugs such as valproate and benzodiazepines which suppress generalized spikes have been used (Marescaux et al., 1990). High-dose steroids are commonly used for cases where language function fails to improve with conventional AED therapy though controlled studies as to their efficacy are not available (Marescaux et al., 1990; Lerman et al., 1991). A few centers have even used subpial transection as a surgical means of abolishing the interictal spike activity (Morrell et al., 1995; Rintahaka et al., 1995).
Infantile spasms (West syndrome) Infantile spasms or West syndrome is an age-specific epileptic disorder of infancy with peak onset between 4 and 6 months of age. The usual features are onset of infantile spasms associated with a hypsarrhythmic EEG followed by developmental regression and loss of previously acquired milestones (Jeavons and Livet, 1992). The etiologies are varied. Of particular interest to our discussion are the 10-20% of cases who are of cryptogenic etiology (i.e. were previously developmentally normal and have normal imaging studies). While the spasms themselves are an agedependent phenomenon that eventually remits with or without treatment, these children are often left with severe cognitive impairment. Children with infantile spasms, even if initially developmentally normal, often regress and lose milestones following the onset of the disorder which is characterized as an epileptic encephalopathy. The typical EEG pattern is hypsarrhythmia which consists of high-voltage, multifocal independent spikes and waves superimposed on a very disorganized background. The presumption is that it is this wildly abnormal EEG pattern rather than the brief seizures
228 that is responsible for the regression. Treatment for cryptogenic cases is usually high-dose adrenocorticotrophic hormone (ACTH) (Baram et al., 1996) and therapy is aimed not just at suppressing seizures, but at reversing the hypsarrhythmic EEG (Baram et al., 1996; Riikonen, 2000). While many children do poorly despite treatment, a proportion of cryptogenic cases will normalize their EEGs and have normal developmental outcomes as well as no further seizures. Favorable outcomes are more common in population-based cohorts. In a population-based incident series from Iceland, all five children with cryptogenic infantile spasms were seizure free and in a normal school setting at follow-up (Ludvigsson et al., 1994). Note that treatment is aimed at reversing the EEG abnormality not just suppressing seizures and that developmental improvement is associated with improvement of the EEG. ACTH is not a conventional AED and is a drug that may well be considered to be not just antiseizure but antiepileptogenic in the true sense of the word as it does seem to alter the underlying process. What all these three seemingly very different syndromes share in common is the presence of cognitive and behavioral deterioration in the presence of an epileptiform process. However, one cannot leap to the conclusion that seizures per se are the cause of this decline. In cases of LKS, clinical seizures are not always even present and in the case of ESES they are usually readily controlled with AEDs. In neither case is seizure control associated with either preventing or reversing the decline in function which is thought to be due to the frequent interictal activity. Infantile spasms is a very different syndrome where the seizures may be difficult to control and which is often associated with subsequent medically refractory epilepsy. Discussing it in the same context serves to highlight the concept that it is not the brief myoclonic seizures per se that cause the damage, but the epileptiform encephalopathy that underlies it. Even in cases where the spasms remit and no further seizures occur, the child may be left with devastating cognitive and behavioral sequelae. Infantile spasms are also a disorder where therapy is explicitly aimed at not just controlling seizures but at reversing the underlying EEG abnormalities. ACTH is used precisely because of its demonstrated ability to reverse the hypsarrhythmic EEG pattern thought to be re-
sponsible for the clinical deterioration (Baram et al., 1996; Riikonen, 2000). The evidence reviewed suggests that it is not the seizure per se that cause deterioration, but the underlying epileptic process. The implication of this for those who argue for early treatment of seizures to prevent sequelae is that in these syndromes it is insufficient to treat seizures, but one must eliminate the seizure focus. Most of our conventional AEDs including specifically sodium channel blockers such as carbamazepine and phenytoin are excellent at suppressing seizures, but do not significantly affect the interictal spike focus and may even exacerbate the EEG. Elimination of the spike focus would be an argument for early surgical rather than medical intervention with early being defined as after only a few seizures and regardless of whether clinical seizures are occurring (Moshe and Shinnar, 1993; Shinnar and Berg, 1996). This would be difficult to justify on a risk-benefit basis without considerably more supportive data than are currently available.
Epilepsysyndrome It increasingly appears that, in humans, prognosis both for seizure remission or development of medically refractory epilepsy and for development of cognitive and behavioral sequelae is primarily a function of the specific epilepsy syndrome. Seizures per se, while clearly relevant, may not be the primary determinant. We have already discussed several syndromes (LKS, ESES, infantile spasms) associated with adverse cognitive and behavioral outcomes. Several other syndromes merit specific discussion in this context.
Benign Rolandic epilepsy Benign Rolandic epilepsy (BRE), also called benign childhood epilepsy with centrotemporal spikes, is characterized by daytime partial seizures and nocturnal tonic-clonic seizures of presumed focal onset. Age of onset is between ages 3 and 12 years (Lerman, 1992). The EEG pattern is that of centratemporal spikes with a characteristic dipole. The spikes are more frequent in sleep. Seizure frequency is usually low and up to half the people with the EEG trait never have clinical seizures. Seizures are
229 usually easily controlled with appropriate AEDs, and the condition virtually always remits by midadolescence regardless of whether AED therapy is used or not. Many children with centro-temporal spikes never experience clinical seizures. Because the seizures are usually infrequent, brief, and often occur at night, there is growing debate over the need to treat children with this self-limited form of epilepsy (Freeman et al., 1987; Ambrosetto and Tassinari, 1990; O'Dell and Shinnar, 2001).
Childhood absence epilepsy and juvenile myoclonic epilepsy Two primary generalized idiopathic epilepsy syndromes, childhood absence and juvenile myoclonic epilepsy provide additional evidence that it is usually not seizures per se that alter prognosis (Loiseau, 1992; Wolf, 1992). The seizures in both syndromes are usually readily controlled with appropriate AED therapy. The prognosis is quite different. The majority of children with childhood absence epilepsy enter remission during adolescence, and the generalized spike and wave EEG abnormalities typical of this syndrome also disappear. While treatment is beneficial in that it improves the child's ability to attend to a task in school and learn, there is no evidence that treatment improves the likelihood of ultimately achieving remission off medication. In contrast, juvenile myoclonic epilepsy rarely remits even with AED therapy, and the EEG features persists at least into the fourth or fifth decade though they may be masked by AED therapy. Long-term therapy is needed to maintain control of seizures. In terms of cognitive outcomes, however, juvenile myoclonic epilepsy has a better prognosis with favorable cognitive outcomes. A recent study of children with childhood absence epilepsy found that these children had impaired academic performance compared with a control population of children with juvenile rheumatoid arthritis which is a far more disabling disorder. While those with uncontrolled seizures did worse, even those whose seizures were fully controlled did worse than the controls (Wirrell et al., 1997). Children with idiopathic epilepsy also performed worse than population based normal controls on a variety of measures of psychosocial function (Sillanpaa et al., 1998a).
Mesial temporal lobe epilepsy The syndrome of mesial temporal lobe epilepsy (MTLE) has gained increased recognition as an epilepsy syndrome that can be progressive. Patients with medically refractory MTLE frequently have impairment of memory (Sass et al., 1992; Bell and Davies, 1998; Fuerst et al., 2001). In addition, recent studies have suggested that the degree of hippocampal atrophy may be correlated with the duration of the seizure disorder (Fuerst et al., 2001; Sutula and Pitkanen, 2001). As this epilepsy syndrome responds well to resective surgery, a cogent argument can be made for early surgical intervention in this epilepsy syndrome for cases that are medically refractory (Engel, 2001; Sutula and Pitkanen, 2001; Wiebe et al., 2001). As the majority of seizures in patients with MTLE are brief complex partial seizures, the above data has also been cited as evidence that even brief seizures can cause damage (Fuerst et al., 2001; Sutula and Pitkanen, 2001). There are several problems with this line of reasoning. Firstly the evidence comes primarily from patients with medically refractory MTLE. These patients have typically had hundreds if not thousands of seizures by the time they are being evaluated for surgery. While there is no question that some of these patients have had a progressive course, it is difficult to understand how one can extrapolate the data from these patients to patients with occasional brief seizures. Other studies have suggested that there is a wide spectrum of clinical manifestations of MTS and that it is not always associated with medically refractory MTLE (Kim et al., 1999; Kobayashi et al., 2001). In addition, while the interictal EEGs from surface recordings are often relatively silent, intracranial recordings with depth electrodes frequently demonstrate a very active spike focus with very frequent discharges that are not being seen at the surface due to the anatomy of the hippocampus which creates a closed field. Is it possible that in analogy with ESES and Landau-Kleffner syndrome, at least some of the progressive symptomatology, including impairment of memory, may be due to the very frequent spike activity that is taking place in the hippocampus?
230
Summary The specific epilepsy syndrome plays a key role in determining long-term prognosis. There are clearly syndromes, such as benign rolandic epilepsy, whose prognosis is favorable regardless of therapy. Other syndromes, such as LKS and ESES, have a favorable prognosis in terms of seizures, but an unfavorable one in terms of cognitive outcomes. Infantile spasms and the Lennox-Gastaut syndrome are associated with a poor prognosis for both remission of seizures and cognitive and behavioral outcomes. MTLE is a syndrome with variable outcomes. In cases of medically refractory MTLE, the data suggest that the syndrome may be progressive and early surgical intervention may be justified once medical intractability is established. However, the data do not suggest an adverse effect on outcome as a consequence of an occasional brief seizure. Rationale for early intervention
From a clinical perspective, the issue needs to be framed from a risk-benefit perspective. Few people believe that seizures are good for you. Even brief seizures have the potential for adverse outcomes, especially if they occur in the wrong place at the wrong time (O'Dell and Shinnar, 2001). As elegantly outlined throughout this volume, there are also potential physiologic changes that occur following seizures. Some of these changes may be persistent though here the evidence is strongest for prolonged or frequent seizures rather than occasional brief seizures. Thus from a theoretical perspective, it is appropriate to be concerned about the possible adverse effects of even brief seizures. However, AEDs, both new and old, are also not without adverse effects which have been reviewed extensively elsewhere (Committee on Drugs, 1995; O'Dell and Shinnar, 2001). These include idiosyncratic and dose-related effects as well as the potential for teratogenicity. Subtle effects on cognition and behavior are common even in individuals who do not demonstrate overt cognitive symptomatology (Vining et al., 1987; Committee on Drugs, 1995; O'Dell and Shinnar, 2001). In all the discussion of adverse effects of seizures, the reverse side of the coin is rarely considered. For example, in Hauser and Lee (2002, this volume) an
increasing risk for seizure recurrence with increasing numbers of seizures is reported in a select group of people with a first seizure. However, this finding is limited to those with no risk factors for recurrence - - the group with lowest initial recurrence risk and a group with high probability of remission (Shafer et al., 1988). What are the factors that influence remission in this group of patients which includes adults as well as some children? The brain must have some mechanisms that will suppress the seizure focus as, in this primarily adult population, developmental factors are unlikely to play a major role. Could the seizures themselves have a role to play? Is it possible that, at least in some cases, seizures may be in some way a response to an underlying pathologic state and reflect a response of homeostatic mechanisms to modify a pathologic substrate? They may in some settings be responsible for establishing the milieu for the remission that characterizes the majority of cases of epilepsy regardless of etiology or age of onset. The majority of the seizure models are models of acute symptomatic seizures rather than of chronic epilepsy. While it is difficult to seriously argue a beneficial effect for seizures, the epidemiological data certainly do not demonstrate adverse effects from occasional brief seizures per se. For the clinician, the discussion does not focus on whether even brief seizures may be harmful, but on a risk-benefit analysis. In adults, there is now consensus that after the second seizure, treatment is usually indicated. This is not based on evidence that waiting longer will worsen outcome but rather on epidemiological data that once a second seizure has occurred the risk of a third seizure is approximately 70% and that therefore, treatment is indicated on a risk-benefit basis (Hauser et al., 1998; O'Dell and Shinnar, 2001). Even in this setting special circumstances, such as a woman who wishes to have children and does not need to drive, may change the risk-benefit analysis towards no treatment in selected cases. In children, the recurrence risks are similar, but there is much more controversy on the need to treat following a second seizure (Freeman et al., 1987; Duchowny, 2000; Shinnar et al., 2000; O'Dell and Shinnar, 2001). This is because on a risk-benefit basis, the potential consequences of seizures are somewhat less than in adults, the risk of adverse effects from AEDs somewhat greater
231 and the probability that they have a self-limited agedependent childhood seizure disorder must also be factored in (Shinnar et al., 2000; O'Dell and Shinnar, 2001). If even brief seizures produce damage, one would logically initiate treatment after the first seizure. To justify this, one would have to demonstrate that AED therapy after the first seizure improves prognosis compared with waiting until at least two have occurred. While there is basic science data suggesting that even brief seizures have potential adverse consequences, the clinical data from randomized and epidemiological studies have not demonstrated any adverse effects on long-term outcomes from delaying therapy until at least two seizures have occurred. This needs to remain the fundamental justification for putting patients on drugs all of which have potential toxicities. The evidence to date while intriguing and justifying further research does not provide a sufficient rationale justify treating individual with a first unprovoked seizure with AEDs in order to alter the long-term prognosis. It clearly does justify treating most individuals with recurrent seizures in order to prevent further seizures. The data on potential adverse effects from interictal spike activity is also of concern and raises similar risk-benefit issues. In general, our conventional AEDs do not suppress interictal spike activity with the exception of the generalized spike and wave activity seen in the primary generalized epilepsy syndromes. Therefore, if one desires to suppress them, one needs to either use nonconventional medications, such as steroids, or to advocate for very early surgical intervention to remove the epileptic focus even in patients whose clinical seizures are controlled with AEDs. Indeed this rationale has been used to justify subpial transection of the posterior temporal lobe in some cases of children with language regression and LKS who had epileptic loci as the presumed cause of the regression (Morrell et al., 1995; Rintahaka et al., 1995). The same argument, namely that one would need a surgical rather than medical intervention, can be made with regard to preventing kindling or secondary epileptogenesis (Moshe and Shinnar, 1993; Shinnar and Berg, 1996). The clinical data at this time do not justify the use of these procedures in this setting outside of a controlled clinical trial.
Conclusions The epidemiological data do not provide clear evidence for adverse effects from occasional brief seizures. Evidence for a progressive adverse effects of seizures comes primarily from specific epilepsy syndromes, often with markedly abnormal interictal patterns. Treatment with conventional AEDs appears to be effective in reducing seizure recurrence but does not alter long-term prognosis. Chronic AED therapy is also associated with a variety of adverse effects. Animal models elegantly presented in this volume report data that are a cause for concern and justify further study. However, on a risk-benefit basis, the clinical data do not justify early intervention after a single seizure to attempt to alter the long-term outcome.
Acknowledgements Supported in part by Grant R01 NS26151 (to S.S.) from the National Institute of Neurological Disorders and Stroke, NIH, Bethesda, MD.
References Ambrosetto, G. and Tassinari, C.A. (1990) Antiepileptic drug treatment of benign childhood epilepsy with rolandic spikes: is it necessary?. Epilepsia, 31: 802-805. Annegers, J.F., Hauser, W.A. and Elveback, L.R. (1979a) Remission of seizures and relapse in patients with epilepsy. Epilepsia, 20: 729-737. Annegers, J.E, Hauser, W.A., Elveback, L.R. and Kurland, L.T. (1979b) The risk of epilepsy following febrile convulsions. Neurology, 29: 297-303. Annegers, J.F., Grabow, J.D., Groover, R.V., Laws Jr., E.R., Elveback, L.R. and Kurland, L.T. (1980) Seizures after head trauma: a population study. Neurology, 30: 683-689. Annegers, J.F., Shirts, S.B., Hauser, W.A. and Kurland, L.T. (1986) Risk of recurrence after an initial unprovoked seizure. Epilepsia, 27: 43-50. Annegers, J.F., Hauser, W.A., Shirts, S.B. and Kurland, L.T. (1987) Factors prognostic of unprovoked seizures after febrile convulsions. New Engl. J. Med., 316: 493-498. Annegers, J.F., Blakely, S.A., Hauser, W.A. and Kurland, L.T. (1990) Recurrence of febrile convulsions in a population-based cohort. Epilepsy Res., 66: 1009-1012. Annegers, J.F., Hauser, W.A., Coan, S. and Rocca, W.A. (1998) A population-based study of seizures after traumatic brain injuries. New Engl. J. Med., 338: 20-24. Baram, T.Z., Mitchell, W.G., Touruay, A., Snead, O.C., Hanson, R.A. and Horton, E.J. (1996) High-dose corticotropin
232
(ACTH) versus prednisone for infantile spasms: a prospective, randomized, blinded study. Pediatrics, 97: 375-379. Beaumanoir, A. (1992) The Landau-Kleffner syndrome. In: J. Roger, M. Bureau, C. Dravet, EE. Dreifuss et al. (Eds.), Epileptic Syndromes in Infancy, Childhood and Adolescence, 2nd edn. John Libbey, London, pp. 231-243. Bell, B. and Davies, K. (1998) Hippocampal sclerosis and memory: Recent neuropsychological findings. Neuropsychol. Rev., 8: 25-41. Berg, A.T. and Shinnar, S. (1991) The risk of seizure recurrence following a first unprovoked seizure: a quantitative review. Neurology, 41: 965-972. Berg, A.T. and Shinnar, S. (1996a) Unprovoked seizures in children with febrile seizures: short term outcomes. Neurology, 47: 562-568. Berg, A.T. and Shinnar, S. (1996b) Complex febrile seizures. Epilepsia, 37: 126-133. Berg, A.T. and Shinnar, S. (1997) Do seizures beget seizures? An assessment of the clinical evidence in humans. J. Clin. Neurophysiol., 14: 102-110. Berg, A.T., Shinnar, S., Hauser, W.A. and Leventhal, J.M. (1990) Predictors of recurrent febrile seizures: a meta-analytic review. J. Pediatr., 116: 329-337. Berg, A.T., Shinnar, S., Hauser, W.A., Alemany, M., Shapiro, E.D., Salomon, M.E. and Crain, E.E (1992) A prospective study of recurrent febrile seizures. New Engl. J. Med., 327: 1122-1127. Berg, A.T., Shinnar, S., Darefsky, A.S., Holford, T.R., Shapiro, E.D., Salomon, M.E., Crain, E.E and Hauser, W.A. (1997) Predictors of recurrent febrile seizures. Arch. Pediatr. Adolesc. Med., 151: 371-378. Berg, A.T., Shinnar, S., Levy, S.R. and Testa, EM. (1999) Status epilepticus in children with newly diagnosed epilepsy. Ann. Neurol., 45: 618-623. Berg, A.T., Shinnar, S., Levy, S.R., Testa, EM., Smith-Rapaport, S. and Beckerman, B. (2001) Early development of intractable epilepsy in children: a prospective study. Neurology, 56: 14451452. Bishop, D.V.M. (1985) Age of onset and outcome in 'acquired aphasia with convulsive disorder' (Landau-Kleffner syndrome). Dev. Med. Child Neurol., 27: 705-712. Camfield, ER., Camfield, C.S., Dooley, J.M., Tibbles, J.A.R., Fung, T. and Garner, B. (1985) Epilepsy after a first unprovoked seizure in childhood. Neurology, 35: 1657-1660. Camfield, E, Camfield, C., Dooley, J., Smith, E. and Garner, B. (1989) A randomized study of carbamazepine versus no medication following a first unprovoked seizure in childhood. Neurology, 39:851-852. Camfield, C., Camfield, E, Gordon, K. and Dooley, J. (1996) Does the number of seizures before treatment influence ease of control or remission of childhood epilepsy? Not if the number is l0 or less. Neurology, 464: 41-44. Chadwick, D. (1995) Do anticonvulsants alter the natural course of epilepsy? Case for early treatment is not established. Br. Med. J., 310: 177-178. Chandra, B. (1992) First seizure in adults: to treat or not to treat. Clin. Neurol. Neurosurg., 94(Suppl.): $61-$63.
Chang, Y.-C., Gut, N.-W., Huang, C.-C., Wang, S.-T. and Tsai, J.-J. (2000) Neurocognitive attention and behavior outcome of school-age children with a history of febrile convulsions: a population study. Epilepsia, 41: 412-420. Chang, Y.C., Gut, N.W., Wang, S.T., Huang, C.C. and Tsai, J.J. (2001) Working memory of school-aged children with a history of febrile convulsions: a population study. Neurology, in press. Chen, K., Baram, T.Z. and Soltesz, I. (1999) Febrile seizures in the developing brain result in persistent modification of neuronal excitability in limbic circuits. Nat. Med., 5: 888-894. Cockerell, O.C., Johnson, A.L., Sander, J.W. and Sborvon, S.D. (1997) Prognosis of epilepsy: a review and further analysis of the first nine years of the British National General Practice Study of Epilepsy, a prospective population-based study. Epilepsia, 38: 31-46. Collaborative Group for the Study of Epilepsy (1992) Prognosis of epilepsy in newly referred patients: a multicenter prospective study of the effects of monotherapy on the long-term course of epilepsy. Epilepsia, 33:45-51. Committee on Drugs (1995) American Academy of Pediatrics. Behavioral and cognitive effects of anticonvulsant therapy. Pediatrics, 96: 538-540. Commission on Epidemiology and Prognosis (1993) International League Against Epilepsy. Guidelines for epidemiologic studies on epilepsy. Epilepsia, 34: 592-596. Corey, L.A., Pellock, J.M., Boggs, J.G., Miller, L.L. and DeLorenzo, R.J. (1998) Evidence for a genetic predisposition for status epilepticus. Neurology, 50: 558-560. Dube, C. (2002) Do prolonged 'febrile' seizures in an immature rat model cause epilepsy? In: T.Z. Baram and S. Shinnar (Eds.), Febrile Seizures. Academic Press, San Diego, CA, pp. 215-229. Dube, C., Chen, K., Eghbal-Ahmadi, M., Brunson, K., Soltesz, I. and Baram, T.Z. (2000) Prolonged febrile seizures in the immature rat model enhance hippocampal excitability long term. Ann. Neurol., 47: 336-344. Duchowny, M. (2000) Seizure recurrence in childhood epilepsy: "the future ain't what it used to be". Ann. NeuroL, 48: 137139. Dugas, M., Gerard, C.L., Franc, S. et al. (1991) Natural history, course and prognosis of the Landau and Kleffner syndrome. In: I.P. Martins, A. Castro-Caldas, H.R. Van Dongen and A. van Hout (Eds.), Acquired Aphasia in Children: Acquisition and Breakdown of Language in the Developing Brain. Kluwer Academic, Dordrecht, pp. 263-277. Ellenberg, J.H. and Nelson, K.B. (1978) Febrile seizures and later intellectual performance. Arch. Neurol., 35: 17-21. Elwes, R.D., Johnson, A.L., Shorvon, S.D. and Reynolds, E.H. (1984) The prognosis for seizure control in newly diagnosed epilepsy. New Engl. J. Med., 311 : 944-947. Elwes, R.D., Johnson, A.L. and Reynolds, E.H. (1988) The course of untreated epilepsy. Br. Med. J., 297: 948-950. Engel Jr., J. (2001) Finally, a randomized, controlled trial of epilepsy surgery. New Engl. J. Med., 345: 365-367. Engel, J. and Shewmon, D.A. (1991) The impact of the kindling phenomenon on clinical epileptology. In: E Morrell (Ed.),
233
Kindling and Synaptic Plasticity: The Legacy of Graham Goddard. Birkhauser, Boston, MA, pp. 195-210. Feksi, A.T., Kaamugisha, J., Sander, J.W., Gatiti, S. and Shorvon, S.D. (1991) Comprehensive primary health care antiepileptic drug treatment in rural and semi-urban Kenya. Lancet, 337: 406-407. First Seizure Trial Group (1993) Randomized clinical trial on the efficacy of antiepileptic drugs in reducing the risk of relapse after a first unprovoked tonic-clonic seizure. Neurology, 43: 478-483. Foy, P.M., Chadwick, D.W., Rajgopalan, N., Johnson, A.L. and Shaw, M.D. (1992) Do prophylactic anticonvulsant drugs alter the pattern of seizures after craniotomy?. J. Neurol. Neurosurg. Psychiatry, 55: 753-757. Freeman, J.M., Tibbles, J., Camfield, C. and Camfield, P. (1987) Benign epilepsy of childhood: a speculation and its ramifications. Pediatrics, 79: 864-868. Fuerst, D., Shah, J., Kupsky, W.J., Johnson, R., Shah, A., Hayman-Abello, B., Ergh, T., Poore, Q., Canady, A. and Watson, C. (2001) Volumetric MRI, pathological, and neuropsychological progression in hippocampal sclerosis. Neurology, 57: 184-188. Gilad, R., Lampl, Y., Gabbay, U., Eshel, Y. and Sarova-Pinhas, I. (1996) Early treatment of a single generalized tonic-clonic seizure of prevent recurrence. Arch. Neurol., 53:1149-1152. Goldensohn, E.S. (1984)The relevance of secondary epileptogenesis to the treatment of epilepsy: kindling and the mirror focus. Epilepsia, 25(Suppl. 2): S156-S168. Gowers, W.R. (1881) Epilepsy and Other Chronic Convulsive Disorders. Churchill, London. Hauser, W.A. and Lee, J.R. (2002) Do seizures beget seizures? In: T. Sutula and A. Pitk~nen (Eds.), Do Seizures Damage the Brain. Progress in Brain Research, Vol. 135. Elsevier, Amsterdam, pp. 215-219. Hauser, W.A., Anderson, V.E., Loewenson, R.B. and McRoberts, S.M. (1982) Seizure recurrence after a first unprovoked seizure. New Engl. J. Med., 307: 522-528. Hauser, W.A., Tabaddor, K., Factor, P. and Feiner, C. (1984) Seizures and head injury in an urban community. Neurology, 34: 746-758. Hauser, W.A., Rich, S.S., Annegers, J.E and Anderson, V.E. (1990) Seizure recurrence after a 1st unprovoked seizure: an extended followup. Neurology, 40:1163-1170. Hauser, W.A., Rich, S.S., Lee, J.R., Annegers, J.F. and Anderson, V.E. (1998) Risk of recurrent seizures after two unprovoked seizures. New Engl. J. Med., 338: 429-434. Hesdorffer, D.C. and Hauser, W.A. (2002) Febrile seizures and the risk for epilepsy. In: T.Z. Baram and S. Shinnar (Eds.), Febrile Seizures. Academic Press, San Diego, CA, pp. 63-76. Hirtz, D. (2002) Cognitive outcome of febrile seizures. In: T.Z. Baram and S. Shinnar (Eds.), Febrile Seizures. Academic Press, San Diego, CA, pp. 53-61. Hirtz, D., Ashwal, S., Berg, A., Bettis, D., Camfield, C., Camfield, P., Crumrine, P., Elterman, R., Schneider, S. and Shinnar, S. (2000) Practice Parameter: Evaluating a first nonfebrile seizure in children: report of the Quality Standards Subcommittee of the American Academy of Neurology, the Child
Neurology Society and the American Epilepsy Society. Neurology, 55: 616-623. Hopkins, A., Garman, A. and Clarke, C. (1988) The first seizure in adult life: value of clinical features electroencephalography and computerized tomographic scanning in prediction of seizure recurrence. Lancet, 1: 721-726. Jeavons, P.M. and Livet, M.O. (1992) West syndrome: infantile spasms. In: J. Roger, M. Bureau, C. Dravet et al. (Eds.), Epileptic Syndromes in Infancy, Childhood and Adolescence, 2nd edn. John Libbey, London, pp. 53-65. Kim, W.J., Park, S.C. and Lee, S.J. et al. (1999) The prognosis for control of seizures with medications in patients with MRI evidence for mesial temporal sclerosis. Epilepsia, 40: 290293. Knudsen, EU. (2002) Treatment of febrile seizures: practical management approaches to simple and complex febrile seizures. In: T.Z. Baram and S. Shinnar (Eds.), Febrile Seizures. Academic Press, San Diego, CA, pp. 273-304. Knudsen, EU. and Vestermark, S. (1985) Recurrence risk after first febrile seizure and effect of short-term diazepam prophylaxis. Arch. Dis. Child., 60: 1045-1049. Knudsen, EU., Paerregaard, A., Andersen, R. and Andresen, J. (1996) Long term outcome of prophylaxis for febrile convulsions. Arch. Dis. Child., 74: 13-18. Kobayashi, E., Lopes-Cendes, I., Guerreiro, C.A.M., Sousa, S.C., Guerreiro, M.M. and Cendes, E (2001) Seizure outcome and hippocampal atrophy in familial mesial temporal lobe epilepsy. Neurology, 56: 166-172. Lerman, P. (1992) Benign partial epilepsy with centro-temporal spikes. In: J. Roger, M. Bureau, C. Dravet et al. (Eds.), Epileptic Syndromes in Infancy, Childhood and Adolescence. 2nd edn. John Libbey, London, pp. 189-200. Lerman, P., Lerman-Sagie, T. and Kivity, S. (1991) Effect of early steroid therapy for Landau-Kleffner syndrome. Dev. Med. Child Neurol., 33: 257-266. Loiseau, P. (1992) Childhood absence epilepsy. In: J. Roger, M. Bureau, C. Dravet et al. (Eds.), Epileptic Syndromes in Infancy, Childhood and Adolescence, 2nd edn. John Libbey, London, pp. 135-150. Ludvigsson, P., Olafsson, E., Siguroardottir, S. and Hauser, W.A. (1994) Epidemiologic features of infantile spasms in Iceland. Epilepsia, 35: 802-805. MacDonald, B.K., Johnson, A.L., Goodridge, D.M., Cockrell, O.C., Sander, J.W.A.S. and Shorvon, S.D. (2000) Factors predicting prognosis of epilepsy after presentation with seizures. Ann. Neurol., 48: 833-841. Marescaux, C., Hirsch, E. and Finck, S. et al. (1990) LandauKleffner syndrome. A pharmacologic study of five cases. Epilepsia, 31: 768-777. Maytal, J. and Shinnar, S. (1990) Febrile status epilepticus. Pediatrics, 86: 611-616. Mitchell, T.V. and Lewis, D.V. (2002) Do prolonged febrile seizures injure the hippocampus? Human MRI studies. In: T.Z. Baram and S. Shinnar (Eds.), Febrile Seizures. Academic Press, San Diego, CA, pp. 103-126. Morrell, E, Whisler, W.W. and Smith, M.C. et al. (1995)
234
Landau-Kleffner syndrome: Treatment with subpial intracortical transection. Brain, 118: 1529-1546. Moshe, S.L. and Shinnar, S. (1993) Early Intervention. In: J. Engel Jr. (Ed.), Surgical Treatment of the Epilepsies, 2nd edn. Raven Press, New York, NY, pp. 123-132. Musicco, M., Beghi, E., Solari, A. and Viani, F. (1997) Treatment of first tonic-clonic seizure does not improve the prognosis of epilepsy. Neurology, 49: 991-998. National Institutes of Health (1980) Febrile Seizures: Consensus Development Conference Summary. Vol. 3, no. 2. National Institutes of Health, Bethesda, MD. Nelson, K.B. and Ellenberg, J.H. (1978) Prognosis in children with febrile seizures. Pediatrics, 61: 720-727. O'DeI1, C. and Shinnar, S. (2001) Initiation and discontinuation of antiepileptic drugs. Neurol. Clin., 19: 289-311. Offringa, M., Derksen-Lubsen, G., Bossuyt, P.M.M. and Lubsen, J. (1992) Seizure recurrence after a first febrile seizure: a multivariate approach. Dev. Med. Child Neurol., 34: 15-24. Offringa, M., Bossuyt, P.M.M. and Lubsen, J. et al. (1994) Risk factors for seizure recurrence in children with febrile seizures: a pooled analysis of individual patient data from five studies. J. Pediatr., 124: 574-584. Ohtsuka, Y., Yoshinaga, H., Kobayashi, K., Murakami, N., Yamatogi, Y. Oka, E. and Tsuda, T. (2001) Predictors and underlying causes of medically intractable localization-related epilepsy in childhood. Pediatr. Neurol., 24:209-213. Reynolds, E.H. (1995) Do anticonvulsants alter the natural course of epilepsy? Treatment should be started as early as possible. Br. Med. J,, 310: 176-177. Reynolds, E.H., Elwes, R.D.C. and Shorvon, S. (1983) Why does epilepsy become intractable? Prevention of chronic epilepsy. Lancet, 2(356): 952-954. Riikonen, R.S. (2000) Steroids or vigabatrin in the treatment of infantile spasms?. Pediatr. Neurol., 23: 403-408. Rintahaka, P.J., Chugani, H.T. and Sankar, R. (1995) LandauKleffner syndrome with continuous spikes and waves during slow-wave sleep. J. Child Neurol., 10: 127-133. Rosman, N.P., Colton, T. and Labazzo, J. et al. (1993a) A controlled trial of diazepam administered during febrile illnesses to prevent recurrence of febrile seizures. New Engl. J. Med., 329: 79-84. Rosman, N.P., Labazzo, J.L. and Colton, T. (1993b) Factors predisposing to afebrile seizures after febrile convulsions and preventive treatment. Ann. Neurol., 34: 452. Sander, J.W.A.S. (1993) Some aspects of prognosis in the epilepsies: a review. Epilepsia, 34: 1007-1016. Sass, K., Sass, A. and Westerveld, M. et al. (1992) Specificity in the correlation of verbal memory and hippocampal neuron loss: dissociation of memory, language, and verbal intellectual ability. J. Clin. Exp. Neuropsychol., 14: 662-672. Shafer, S., Hauser, W.A., Annegers, J.E and Klass, D.W. (1988) EEG and other early predictors of later epilepsy remission: a community study. Epilepsia, 29: 590-600. Shinnar, S. (1998) Prolonged febrile seizures and mesial temporal sclerosis. Ann. Neurol., 43:411-412. Shinnar, S. and Berg, A.T. (1996) Does antiepileptic drug therapy
prevent the development of 'chronic' epilepsy. Epilepsia, 37: 701-708. Shinnar, S., Berg, A.T., Moshe, S.L., Petix, M., Maytal, J., Kang, H., Goldensohn, E.S. and Hauser, W.A. (1990) Risk of seizure recurrence following a first unprovoked seizure in childhood: a prospective study. Pediatrics, 85: 1076-1085. Shinnar, S., Berg, A.T. and Moshe, S.L. (1995) The effect of status epilepticus on the long term outcome of a cohort of children prospectively followed from the time of their first idiopathic unprovoked seizure. Dev. Med. Child Neurol., 37(Suppl. 72): 116. Shinnar, S., Berg, A.T., Moshe, S.L., O'Dell, C., Newstein, D., Kang, H., Goldensohn, H. and Hauser, W.A. (1996) The risk of seizure recurrence following a first unprovoked afebrile seizure in childhood: an extended follow-up. Pediatrics, 98: 216-225. Shinnar, S., Berg, A.T,, O'Dell, C., Newstein, D., Moshe, S.L. and Hauser, W.A. (2000) Predictors of multiple seizures in a cohort of children prospectively followed from the time of their first unprovoked seizure. Ann. Neurol., 48: 140-147. Shinnar, S., Pellock, J.M., Berg, A.T., O'Dell, C., Driscoll, S.M., Maytal, J., Moshe, S.L. and DeLorenzo, R.J. (2001a) Short-term outcomes of children with febrile status epilepticus. Epilepsia, 42: 47-53. Shinnar, S., Berg, A.T., Moshe, S.L. and Shinnar, R. (2001b) How long do new-onset seizures in children last?. Ann. Neurol., 49: 659-664. Shorvon, S.D. and Reynolds, E.H. (1982) Early prognosis of epilepsy. Br. Med. J., 285: 1699-1701. Sillanpaa, M., Jalava, M., Kaleva, O. and Shinnar, S. (1998a) Long-term prognosis of seizures with onset in childhood. New Engl. J. Med., 338: 1715-1722. Sillanpaa, M., Jalava, M. and Shinnar, S. (1998b) Status epilepticus in a population based cohort with childhood onset epilepsy in Finland. Epilepsia, 39(Suppl. 6): 219-220. Stroink, H., Brouwer, O.F., Arts, W.E, Geerts, A.T., Peters, A.C. and van Donselaar, C.A. (1998) The first unprovoked seizure in childhood: a hospital based study of the accuracy of the diagnosis, rate of recurrence, and long term outcome after recurrence. Dutch study of epilepsy in childhood. J. NeuroL Neurosurg. Psychiatry, 64: 595-600. Sutula, T.P. and Pitkanen, A. (2001) More evidence for seizureinduced neuron loss: Is hippocampal sclerosis both cause and effect of epilepsy?. Neurology, 57(2): 169-170. Tassinari, C.A., Bureau, M., Dravet, C., Dallla Bernardina, B. and Roger, J. (1992) Epilepsy with continuous spikes and waves during slow wave sleep - - otherwise described as ESES (epilepsy with electrical status epilepticus in slow wave sleep. In: J. Roger, M. Bureau, C. Dravet et al. (Eds.), Epileptic Syndromes in lnfano', Childhood and Adolescence, 2nd edn. John Libbey, London, pp. 245-256. Taylor, D.C. and Ounsted, C. (1971) Biological mechanisms influencing the outcome of seizures in response to fever. Epilepsia, 12: 33-45. Temkin, N.R., Dikmen, S.S., Wilensky, A.J., Keihm, J., Chabal, S. and Winn, H.R. (1990) A randomized, double-blind study
235
of phenytoin for the prevention of post-traumatic seizures. New Engl. J. Med., 323: 497-502. Toth, Z., Yah, X.X., Haftoglou, S., Ribak, C.E. and Baram, T.Z. (1998) Seizure-induced neuronal injury: vulnerability to febrile seizures in immature rat model. J. Neurosci., 18: 42854294. Van Donselaar, C.E., Schimsheimer, R.J., Geerts, A.T. and Declerck, A.C. (1992) Value of the electroencephalogram in adult patients with untreated idiopathic first seizures. Arch. Neurol., 49: 231-237. Van Donselaar, C.A., Brouwer, O.F., Geerts, A.T., Arts, W.F., Stroink, H. and Peters, A.C. (1997) Clinical course of untreated tonic-clonic seizures in childhood: prospective, hospital based study. Br. Med. J., 314: 401-404. VanLandingham, K.E., Heinz, E.R., Cavazos, J.E. and Lewis, D.V. (1998) MRI evidence of hippocampal injury after prolonged, focal febrile convulsions. Ann. Neurol., 43: 413-426. Verity, C.M. and Golding, J. (1991) Risk of epilepsy after febrile convulsions: a national cohort study. Br. Med. J., 303: 13731376. Verity, C.M., Greenwood, R. and Golding, J. (1998) Long-term intellectual and behavioral outcomes of children with febrile convulsions. New Engl. J. Med., 338: 1723-1728. Vining, E.EG., Mellits, E.D., Dorsen, M.M., Cataldo, M.E, Quaskey, S.A., Spielberg, S.E and Freeman, J.M. (1987)
Psychologic and behavioral effects of antiepileptic drugs in children: A double-blind comparison between phenobarbital and valproic acid. Pediatrics, 80: 165-174. Wallace, S.J. (1980) They don't do very well. Pediatrics, 65: 678-679. Wiebe, S., Blume, W.T., Girvin, J.P. and Eliasziw, M. (2001) Effectiveness and Efficiency of Surgery for Temporal Lobe Epilepsy Study Group. A randomized, controlled trial of surgery for temporal-lobe epilepsy. New Engl. J. Med., 345: 311-318. Wirrell, E.C., Camfield, C.S., Camfield, P.R., Dooley, J.M., Gordon, K.E. and Smith, B. (1997) Long-term psychosocial outcome in typical absence epilepsy. Sometimes a wolf in sheeps' clothing. Arch. Pediatr. Adolesc. Med., 151: 152-158. Wolf, P. (1992) Juvenile myoclonic epilepsy. In: J. Roger, M. Bureau, C. Dravet et al. (Eds.), Epileptic Syndromes in Infancy, Childhood and Adolescence, 2nd edn. John Libbey, London, pp. 313-328. Wolf, S.M. and Forsythe, A. (1989) Epilepsy and mental retardation following febrile seizures in childhood. Acta Paediatr. Scand., 78: 291-295. Wolf, S.M., Carr, A. and Davis, D.C. et al. (1977) The value of phenobarbital in the child who has had a single febrile seizure: a controlled prospective study. Pediatrics, 59: 378-385.
T. Sutulaand A. Pitk~inen(Eds.) Progress in Brain Research, Vol. 135 © 2002 ElsevierScienceB.V. All rightsreserved CHAPTER 21
Hippocampal neuron damage in human epilepsy: Meyer's hypothesis revisited Gary W. Mathern 1,,, p. David Adelson 2, Leslie D. Cahan 3 and Joao R Leite 4 l Division of Neurosurgery, The Mental Retardation Research Center, and The Brain Research Institute, University of California, Los Angeles, CA 90095-1769, USA 2 Department of Neurosurgery, University of Pittsburgh, Pittsburgh, PA 15260, USA 3 Neurosurgery, Southern California Permanente Medical Group, Los Angeles, CA, USA 4 Department of Neurology, Ribeirto Preto School of Medicine, University of Sat Paulo, Ribeir~o Preto, SA, Brazil
Abstract:
Whether hippocampal neuron loss and/or hippocampal sclerosis is the 'cause' or 'consequence' of seizures has been a fundamental question in human epilepsy studies for over a century. To address this question, this study examined hippocampal specimens from temporal lobe epilepsy patients (TLE; n = 572) and those with extra-temporal seizures and pathologies (n : 73) for qualitative signs of hippocampal sclerosis and quantitative neuron loss using cell counting techniques. Patients were additionally classified based on pathological substrate, and history of an initial precipitating injury (IPI). Results showed that: (1) Hippocampal sclerosis was strongly linked with an IPI in both TLE and extratemporal seizure patients. (2) In TLE cases, IPIs showed an early age preference and often involved seizures, but IPIs were not age dependent and older IPI cases showed sclerosis that was indistinguishable from younger IPI patients. (3) In TLE patients, longer seizure durations were associated with decreased neuronal densities in all hippocampal subfields. The decrease was independent of the neuron loss linked with IPIs, it occurred in all pathological groups, it occurred over 30 years or more, and was not a consequence of aging. (4) Intractable seizures in the young human hippocampus were not associated with neuronal damage, but were linked with decreased postnatal granule cell development and aberrant axon sprouting. These results support the concept that hippocampal sclerosis is likely an acquired pathology, and most of the neuronal loss occurs with the IPI. In addition, there is progressive hippocampal damage from intractable TLE regardless of pathology. Hence, hippocampal neuron loss can be the 'consequence' of repeated limbic seizures over 30 years or more, but is unlikely to 'cause' hippocampal sclerosis unless there is also an IPI.
Introduction Whether hippocampal neuron loss is the 'cause' or 'effect' of seizures has been an important c l i n i c a l pathological question for over 100 years, especially in the syndrome of temporal lobe epilepsy (TLE) associated with hippocampal sclerosis. Elucidating the * Correspondence to: G.W. Mathern, Division of Neurosurgery, Reed Neurological Research Center, 710 Westwood, Plaza, Room 2123, Los Angeles, CA, 90095-1769, USA. Tel.: + 1-310-206-8777; Fax: + 1-310-206-8461 ; E-mail:
[email protected]
answer is important not only in understanding the pathogenic mechanisms of epilepsy, but also in determining the best therapeutic goals for patients. For example, if hippocampal sclerosis generates TLE and is a result of a pre-epilepsy brain injury, then preventing or augmenting the initial pathological process m a y avert chronic epilepsy. Alternately, if hippocampal neuron loss is the result of repeated seizures, then stopping all seizures at any age becomes an important treatment goal. Unfortunately, the medical literature is inconclusive and rife with controversy regarding this question since the earliest descriptions of hippocampal pathology and its
238 relationship with epilepsy (for review see Babb and Brown, 1987; Gloor, 1991; Mathern et al., 1997a). For example, early autopsy studies from epilepsy patients established that severe hippocampal damage in a pattern termed 'Ammon's horn' or 'hippocampal' sclerosis was strongly associated with complex partial TLE, while minor hippocampal neuronal loss in regions such as the end folium were associated with other seizure types (Meynert, 1867; Pfleger, 1880; Sommer, 1880; Stauder, 1936; Margerison and Corsellis, 1966). It was unclear from these postmortem studies, however, whether severe damage (i.e. hippocampal sclerosis) was the 'cause' or 'consequence' of TLE. With the advent of pre-mortem surgical resection for TLE, it became possible to carry out more detailed pathological studies and compare the surgical findings with pre-operative clinical data. M. Falconer and colleagues in London were one of the first groups to perform these studies, and within a few years Meyer et al. (1954) proposed that hippocampal sclerosis was the result of early childhood brain injuries such as febrile convulsions (see Falconer, 1974; Meldrum, 1997). This concept has been challenged by epidemiological studies showing that the risk of TLE after febrile convulsions is very low (Nelson and Ellenberg, 1976; Annegers et al., 1979; Verity et al., 1993; Camfield et al., 1994). In the early 1990s, our group re-addressed Meyer's hypothesis in clinical-pathological studies of surgical TLE and non-TLE patients with and without hippocampal sclerosis (Mathern et al., 1995a,b, 1996a). By expanding Meyer's concept of brain insult to include any significant medical event prior to habitual seizure onset we found that events (termed initial precipitating injuries; IPI) were strongly related to finding hippocampal sclerosis at surgery. Furthermore, IPIs did not always occur at a young age and did not require a febrile seizure to be associated with hippocampal sclerosis. We also found that chronic TLE was associated with progressive hippocampal neuronal damage, but it was limited to CA1 and prosubiculum subfields and required more than 1520 years of intractable seizures in order to detect. In other words, we found that hippocampal neuron loss can be both 'cause' and 'consequence' of limbic epilepsy, but that severe neuron loss in the sclerosis pattern was more strongly linked with IPIs than with
long-term seizures. In the current study, we have up-dated our clinical data sets to include all TLE patients regardless of pathology (the previous studies concentrated on sclerosis patients), and expanded the patient number to include cases operated during the 1990s. Our purpose was to determine when during a patient's life hippocampal neuron loss occurred, whether hippocampal sclerosis was the likely result of an IPI, and if repeated seizures independently produce hippocampal injury and/or sclerosis. Methods Clinical material
Two patient groups, surgically treated to control seizures, were studied. The first were patients with intractable complex partial TLE, and the second were patients with generalized and/or partial seizures from extra-temporal lesions in which the hippocampus was removed as part of the planned resection (extra-temporal). The patients in both groups were evaluated and treated at four collaborating medical institutions. Most of the TLE patients were from the University of California, Los Angeles, (UCLA; n = 524) operated between 1961 and 2000. The UCLA TLE cohort has been previously reviewed in several publications (Babb et al., 1984a,b, 1987). During the 1990s additional TLE cases were obtained from the Ribeirao Preto School of Medicine (n = 38); the University of Pittsburgh (n = 15); and Southern California Permanente Medical Center (n = 15). Most extra-temporal cases were also treated at UCLA as part of the pediatric epilepsy surgery program (n = 65) beginning in 1992 and the remainder from the Ribeirao Preto School of Medicine (n = 4) and the University of Pittsburgh (n = 4). The clinical evaluation and treatment protocols for both patient groups have been previously published (Engel, 1993; Mathern et al., 1996b, 1999). Informed consent was obtained for medical treatment and the use of any clinical-pathological data for research purposes. For comparison purposes, autopsy hippocampal tissue was obtained in individuals without neurological disease collected at UCLA and Ribeir~o Preto School of Medicine (n = 105).
239
Clinical data collection and patient classification
Cryptogenic
TLE and extra-temporal surgical cases were classified into pathological sub-groups based on review of pathology specimens and neuroimaging studies (MRI and PET). Prior to the MRI era this classification was based at UCLA on complete histopathological review of the en bloc surgical specimen (Crandall and Mathern, 2000), and more recently by comparing the pathology and neuroimaging reports. The TLE cases were sub-classified into: hippocampal sclerosis; lesion only; dual pathology; and cryptogenic.
Patients with intractable TLE but who did not have hippocampal sclerosis or a mass lesion were grouped into the cryptogenic category. These patients are less likely to be seizure free after temporal lobe resection compared with the previous TLE groups with a known pathological substrate (Babb and Brown, 1987; Mathern et al., 1995a). The extra-temporal surgical cases (n = 73) were sub-classified based on their pathological substrate into those with cortical dysplasia (CD; n = 41) or those without dysplastic pathologies (non-CD) such as cortical atrophy from stroke, infection (n = 21), or Rasmussen's encephalitis (n = 10) (Mathern et ai., 1994, 1996b). For the purposes of this study, the pathological findings in the CD and non-CD were similar and combined into a single extra-temporal seizure group for comparison with the TLE group. From the medical record, clinical data were collected to discern if there was an IPI, the IPI type (seizure versus non-seizure), IPI age, the age at habitual seizure onset, and age at surgery as previously described (Mathern et al., 1995a). An IPI was defined as any medical event or incident of probable cerebral injury that was associated with unconsciousness for more than 30 min or alteration in cognition for more than 4 h, and this information was collected retrospectively from the various pre-surgery patient interviews. The age of habitual seizure onset was defined as the age when the patient's typical seizure for which they were referred for surgical treatment began. From these data, the latent period was calculated as the time, in years, from the IPI to habitual seizure onset, and the duration of seizures as the interval between habitual seizure onset and age at surgery. These data were collected in a uniform standardized format without knowledge of the hippocampal pathology.
Hippocampal sclerosis Patients with damage to the hippocampus as determined by MRI/PET and by pathology shows a distinctive qualitative pattern of neuron loss known as hippocampal sclerosis, and no other significant substrate was noted in the temporal lobe specimen. By definition, sclerosis consisted of severe pyramidal neuron loss and gliosis through out the hippocampus that was especially severe in CA1 and prosubiculum (Sommer's sector) and CA4 (end folium) compared with the stratum granulosum and CA2 stratum pyramidal (Bratz, 1899; Mathern et al., 1997a). By comparison, subicular neurons were not destroyed, and there was a transition between the damaged hippocampus and subiculum. Lesion only These patients had extra-hippocampal macroscopic mass lesions, such as a low-grade glioma, cortical dysplasia, hamartoma, etc. and the hippocampus did not show significant cell loss. The hippocampus may or may not show some minor cell loss at pathological examination, but the amount of damage was not severe and not in the hippocampal sclerosis pattern.
Hippocampal neuron densities Dual pathology Surgical cases in which a mass lesion and hippocampal sclerosis were found together were classified in this category.
In addition to qualitative histopathological review, hippocampal sections were stained with cresylecht violet for cell counts (10 txm sections). Counts were performed at 400× using grid morphometric techniques with the corrections of Abercrombie (1946), and the anatomical subfields were based on the clas-
240 sification of Lorente de No (1934) as previously published (Mathern et al., 1995a). The subfields were the granule cells of the fascia dentate, CA4, CA3, CA2, CA1 stratum pyramidal, prosubicular and subicular neurons. For this study, the subfield density measures were averaged into a single variable to represent overall hippocampal neuronal density. Cell density measurements are estimates of packing density and are not a calculation of total neurons per hippocampus. This is because the total volume of the hippocampus cannot be reliably determined in surgical specimens, and our method is an accepted quantitative technique in surgical material.
Data analysis Data were entered into a database on a personal computer and analyzed using a statistical program (Statview 5, SAS Institute Inc., Cary, NC). Statistical tests included ANOVA, ANCOVA, X 2 a n d regression analysis comparing the pathology groups with the other clinical variables. Results were plotted using the same software, and tests were considered statistically different at a minimum confidence level of P < 0.01. Results
Analysis of the TLE clinical-pathological group show that hippocampal sclerosis was strongly associated with an IPI compared with the other pathology categories. IPIs showed a younger age preference, but were not age dependent. Furthermore, prolonged TLE seizure histories correlated with decreased hippocampal neuronal densities in all pathology categories, but the progressive neuronal loss from limbic seizures probably does not generate hippocampal sclerosis. Patients with extra-temporal partial and generalized seizures and pathologies do not show hippocampal sclerosis unless there was an IPI history, even with multiple seizures per day in early human life. Furthermore, the extra-temporal cohort supports the concept that seizures during early human life negatively impact postnatal hippocampal granule cell neurogenesis and mossy fiber axon development. The findings that support these conclusions from the two human clinical-pathological data sets are discussed below.
TLE dataset Of the 572 TLE surgical cases, 54.1% were classified as hippocampal sclerosis, 21.7% as lesion only, 16.5% as dual pathology, and 7.6% as cryptogenic. The mean ages at surgery (years 4- SEM) were not statistically different between hippocampal sclerosis (30.3 4-0.6), lesion only (28.3 ± 1.1), dual pathology (27.5 4- 1.3), and cryptogenic pathologies (29.2 4-1.6; P = 0.126). However, the age at habitual seizure onset for hippocampal sclerosis (12.3 4- 0.5), lesion only (16.04- 1.0), dual pathology (13.8 4- 1.0) and cryptogenic cases (13.74- 1.3) were different (P = 0.0043), with sclerosis patients less than lesion only (P = 0.0003). Further analysis of the TLE data set disclosed. IPIs were associated with hippocampal sclerosis and seizures In all TLE patients, 59.2% had an IPI history prior to the onset of their habitual limbic seizures. The percentages of patients with IPI histories were different between the surgical pathology groups (Table 1). Most of the hippocampal sclerosis patients (87%) had an IPI history compared to the other pathology subgroups (g2; P < 0.0001). Of note, within the UCLA TLE surgical series (excluding the non-UCLA cases) the percentage of hippocampal sclerosis patients with IPI histories has not significantly changed between 1960 and 2000. This supports the notion that IPIs are important in the pathogenesis of hippocampal sclerosis, and this association has been stable over time. The IPI type (seizure versus non-seizure) was also different between TLE pathology groups (Table 2). Most of the hippocampal sclerosis patients had IPIs that involved seizures (70.8%) compared with the other pathology categories (X2; P < 0.0001). Typical non-seizure IPIs included head trauma, cerebral hypoxia, systemic infection with coma, near drownTABLE 1 Initial precipitating injury by pathologyin temporal lobe epilepsy IPI Yes No
Hippocampal Lesion sclerosis only
pathology
87.0% 13.0%
54.4% 45.6%
13.9% 86.1%
Dual
Cryptogenic 23.3% 76.7%
241 during early human development were harmful to the hippocampus. In other words, they speculated that the immature hippocampus was vulnerable to seizure-induced injury that produced sclerosis. Careful inspection of our TLE data set would be consistent with this notion, but an alternative hypothesis is also supported when our extra-temporal group is considered. Fig. 1 shows IPI ages in TLE surgical patients. IPIs occurred by age 4 years in 78.7% of cases with the greatest peak (23.3%) between 6 and 12 months. Notice, however, that Fig. 1 shows a very long 'tail' with IPIs occurring up to age 23 years. Review of the surgical specimens from older IPI patients (i.e. over age 10 years) c o m p a r e d with those younger than age 5 years showed no significant difference in the qualitative hippocampal sclerosis pattern. Put another way, older IPIs produce hippocampal sclerosis at surgery that is indistinguishable from younger IPIs. W h a t is different is that IPIs associated
TABLE 2 Seizure versus no-seizure IPI by temporal lobe epilepsy pathology IPI type
Hippocampal Lesion sclerosis only
Seizures 70.8% No-seizures 29.2%
Dual Cryptogenic pathology
17.6% 49.0% 82.4% 51.0%
30.0% 70.0%
ing, etc. In TLE patients, therefore, IPIs involving seizures were strongly associated with hippocampal sclerosis. This is similar to the findings of Meyer, Falconer, and others more than 40 years ago (Meyer et al., 1954; Falconer, 1974). IPI age and seizures: a reflection of brain maturity IPIs involving seizures at a young age prompted M e y e r et al. (1954) to hypothesize that seizures
IPI a n d t h e a g e it o c c u r s in t e m p o r a l I
25
I
l
l
l
l
l
l
l
l
l
l
l
l
l
l
lobe epilepsy l
l
I
l
l
22.5 20 17.5 t-
O ,t 13_
15 12.5 10 7.5
2.5 0 0
5
10
15
20
25
IPI AGE Fig. 1. Frequency histogram showing the age at initial precipitating injury (IPI) in TLE (n = 352). Bar intervals equal 6 months of age. Note that: (1) The plot does n o t show a normal distribution; (2) the greatest frequency is between 6 and 12 months of age (23.3%); and (3) while most of the IPI ages are less than 5 years, there are a number of cases beyond that age up to 23 years. The mean (4- SEM) IP1 age was 3.74-0.3 years and the median was 1.5 years.
242
Relationship Between Latent Period and IPI Age In Human Temporal Lobe Epilepsy 40
'
....
30351101 •
1,.,.
'
•
....
, ....
, . . , , , , , , , i , , , , , ....
i ....
•
kO 25"J=,
• •
(1) 13._ .,,._, 20 E ,,i,.,,i
t'~ ._1
15
°Oo°
,0
• . . . .
.
8 loll
•
o8 ° O, -~ 0
5
10
15
20
25
30
35
40
IPI AGE Fig. 2. Scattergram displaying the latent period duration in years (y-axis) and IPI age (x-axis). IPls less than 7 years show variable but often longer latent periods compared with IPIs over age 10 years. Linearregression analysis indicates a negative correlation (r = -0.162; P = 0.0024). with a seizure occurred at a younger age (years 4SEM; 2.28-4-0.27) compared with non-seizure IPIs (5.89 + 0.67; P < 0.0001). This raises the question about the role of seizure-related IPIs in the pathogenesis of hippocampal sclerosis. If sclerosis was not restricted to younger IPI patients with seizures, but also occurred in older non-seizure IPI patients, then another possible interpretation is that seizure IPIs could be a surrogate marker of cerebral injury at a young age. Experimental studies support the concept that seizures are easier to provoke in the developing compared with the mature animal (Holmes and Thompson, 1988; Moshe, 1993; Leite et al., 1996; Holmes, 1997; Lado et al., 2000). Hence, the human TLE data could be interpreted to indicate that IPIs at a younger age are associated with seizures because the immature brain responds to cerebral injury with a seizure more frequently than IPIs in mature brains.
This interpretation is supported by our findings from young surgical patients with frequent extra-temporal seizures that rarely show hippocampal sclerosis (see below). IPI age influences the latent period In the TLE group, the mean duration of the latent period (years -4- SEM) was 8.8 ± 0.04 with a median of 7.0 years, and 67.8% of patients had latent periods of 10 years or less (range few months to 38 years). Furthermore, the IPI age influenced the latent period as shown in Fig. 2. IPIs occurring at age 10 years or less were associated with variable but often very long latent periods while IPIs over age 10 years were usually less than 10 years in duration. This distribution was statistically significant with a negative linear correlation (P = 0.0024). Therefore,
243
Decreased 25000
CA4
. . . .
'
Neuron
. . . .
Densities '
. . . .
With '
Longer
. . . .
TLE i
,
Seizure ,
•
. . . .
Duration ,
,
22500
20000
• ~:
17500
E
R= -0.394 •
• 15ooo-
,~
•
•
12500
-
10000
-
d
qPO
•
•
•
•
_
o 7500 -
5000 -
2500 -
0 0
10
20
30
40
50
60
Seizure Duration In Y e a r s Fig. 3. Scattergram showing CA4 neuron densities (y-axis) and total seizure duration (x-axis, in years) in TLE patients regardless of pathology. The negativelinear regression is significant (r = -0.394; P < 0.0001). Notice very long seizure durations extending for more than 40 years were necessary to discern the negativecorrelation. the IPI age can impact the clinical TLE history by influencing the latent period duration with younger IPI ages associated with the longest latent periods. Progressive hippocampal neuronal loss with longer TLE seizure histories TLE patients also showed that longer seizure durations were associated with decreased neuron densities in all hippocampal subfields. Our prior studies had disclosed progressive loss in CA1 and prosnbiculum subfields only. The progressive neuronal injury with longer limbic seizure durations occurred in all TLE pathology groups, it was independent of the neuronal loss associated with IPI histories, and probably does not lead to hippocampal sclerosis by itself. A typical example for CA4 is shown in Fig. 3 for all TLE pathology groups, and the negative linear regression was statistically significant (P < 0.0001).
Re-analysis using analysis of covariance (ANCOVA) controlling for pathology groups or age at surgery found that the negative linear regressions were still statistically significant (P < 0.0001). This indicates that decreased CA4 neuron counts with longer seizure histories occurred in all TLE pathology types, and was independent of the underlying substrate generating limbic seizures and any affect of aging. The progressive cell loss from long limbic seizure durations was also independent of the cell damage associated with IPI histories, as shown in Table 3 for averaged hippocampal neuron densities for all subfields (ANCOVA). IPIs and seizure duration were statistically significant (P < 0.002) without an interaction (P = 0.32). Notice in Fig. 3 that the time required to demonstrate the negative correlation was over several decades of seizures, and with an r 2 of 0.155 the regression was not very predictive of cell loss in any given specimen. Hence, cell loss from repeated limbic seizures is not likely to produce
244 TABLE 3 IPI and seizure duration both influence hippocampal cell loss in temporal lobe epilepsy patients Hippocampalcell density IPI Seizure duration Interaction
P = 0.001 P = 0.002 P = 0.32
ANCOVA P-values; surgical TLE cases.
the significant damage associated with hippocampal sclerosis (i.e. greater than 50% cell loss) unless there are more than 30 years of limbic seizures. Given that the average (4- SD) duration of habitual seizures for all TLE patients was 15.9 + 9 . 7 years, most of the neuron loss associated with hippocampal sclerosis was more likely from the IPI and not uncontrolled limbic seizures. Extra-temporal data set The patients with extra-temporal substrates and seizures were not as large as the TLE data set, and consisted of younger patients. However, this patient group addressed if seizures during early human life injured the postnatal developing hippocampus and/or produced hippocampal sclerosis. The extratemporal group consisted of 73 surgical patients with a mean age at surgery of 5.86 4-0.83 years (youngest 6 weeks of age), and a mean age of seizure onset of 19.1 4- 3.7 months (41% by age 2 months). These patients generally had many clinical behavioral seizures per day, and electrographically often had continuous ictal-like discharges. For purposes of this paper, the hippocampal findings in the cortical dysplasia (CD) and non-CD patients were the same, and the two groups combined. The relevant findings are given below: Hippocampal sclerosis is rare and is associated with an IPI Hippocampal damage in a sclerosis pattern was noted in three (4.1%) patients with extra-temporal substrates and early seizure histories, and all cases had an IPI history. The age at surgery for the three hippocampal sclerosis cases were 12 months, 8
years, and 14.75 years. The first case started seizing shortly after birth from a large multi-lobar cortical dysplasia, and at 28 days of age the child was placed into a drug induced coma in an attempt to stop or decrease prolonged status-like clinical and electrographic events. During the coma the child became hypotensive requiring emergency surgery to resect ischemic bowel. The seizures continued despite the treatment and hemispherectomy at age 12 months disclosed hippocampal sclerosis (see Mathem et al., 1997a for illustration). The second case involved a child who had prolonged status epilepticus at age 2 years, and thereafter had severe diffuse left cerebral brain damage by MRI with hemiparesis, hemianopsia, etc. (Fig. 4; left panel). Hemispheric resection at 8 years of age disclosed hippocampal sclerosis with aberrant supragranular mossy fiber sprouting (Fig. 4; right panels). The third case was a young adult with a known perinatal stroke involving the middle and posterior cerebral arteries (i.e. mesial temporal structures were involved in the initial ischemic event), seizures began at age 1 year, and surgery at nearly 15 years showed hippocampal sclerosis. These cases illustrate that hippocampal sclerosis was rare in the extra-temporal seizure group unless there was an IPI, and that IPI often involved hypoxia/ischemia. Seizures during early life adversely affected postnatal granule cell development The other hippocampal specimens from the extratemporal cohort did not consistently show significant pyramidal neuron loss (Mathern et al., 1996b). However, there were decreased granule cell densities and aberrant mossy fiber sprouting, and it was more severe in the youngest patients. Furthermore, in the extra-temporal cases we have recently studied signs of postnatal granule cell develop by staining for highly polysialylated neural cell adhesion molecule (PSA-NCAM) that marks newly formed and migrating neurons (Fig. 5) (Mathern et al., 1994). This shows that postnatal PSA-NCAM expression was decreased in children with frequent extra-temporal seizures supporting the concept that epilepsy in the young human brain negatively impacts postnatal granule cell formation and migration. In other words, our clinical-pathological analysis of children with extra-temporal epilepsy shows that fre-
245
Fig. 4. Example of hippocampal sclerosis in a patient with extra-temporal lobe epilepsy. This child had a prolonged episode of status epilepticus at age 1 year with residual hemiparesis from the resulting left hemispheric damage as shown in the MRI (left panel). Habitual seizures began within a few months consisting of partial motor events leading to secondary generalizations and further episodes of status. Left hemispherectomy was performed at age 8.4 years and examination of the hippocampus showed severe cell loss in the hippocampal sclerosis pattern (CV panel) with aberrant supragranular mossy fiber sprouting (Timm's panel).
quent seizures are not associated with hippocampal sclerosis, but are linked with signs of reduced postnatal granule cell development. In the extra-temporal group, linear regression analyses showed no evidence that seizures produce progressive hippocampal neuron loss or sclerosis. It is important to emphasize, however, that the issue of progressive neuronal loss with long seizure histories is still unclear in the extra-temporal cohort. Most of the extra-temporal cases were operated under age 6 years, and the data set consisted of 73 cases. Our experience with the TLE data set did not find statistically significant linear correlations until we had over 150 patients with seizure durations of 15 years or more. Hence, it is still possible that seizures will be associated with progressive hippocampal neuronal loss in the extra-temporal group, but if so it will probably be with very long seizure durations as noted in TLE patients.
Discussion Our clinical-pathological studies of TLE and extratemporal epilepsy patients indicate that hippocampal neuronal damage depends on the epilepsy syndrome, IPI history, and duration of intractable seizures. Specifically: (1) Hippocampal sclerosis was strongly associated with IPIs in TLE and extra-temporal patients. This supports the concept that hippocampal sclerosis is most likely an acquired pathology from an IPI involving cerebral injury (Figs. 6 and 7). The IPI associated with sclerosis may or may not involve a seizure. Furthermore, based on our analysis of children with extra-temporal epilepsy, it is likely that IPI-induced hippocampal damage involves more than one clinical mechanism such as hypoxia/ischemia plus seizures (Mathern et al., 1998b; Katzir et al., 2OOO).
246
Fig. 5. PSA-NCAM immunoreactivity (IR) was decreased in postnatal human hippocampi with seizures. The left panel shows a low power view of the fascia dentata from a 2-month-old child with severe hemispheric cortical dysplasia whose clinical seizures were first noted shortly after birth. Notice focal IR in the infragranular zone and individual IR fibers coursing through the hilus to CA3 stratum lucidum. The amount of IR in the epilepsy case is decreased compared with a 2-month-old autopsy hippocampus from a child without seizures. In the autopsy case, notice the diffuse IR in the infragranular region. At higher power, the IR labels cell bodies, and this is consistent with migrating newly formed granule cells. These findings are consistent with the notion that postnatal granule cell development is decreased as a consequence of epilepsy in young children.
(2) IPIs generally occur at a young age and very often involve seizures, but hippocampal damage from IPIs was not age or seizure dependent. Furthermore, repeated seizures at a young age were not associated with severe hippocampal damage in our patients with extra-temporal substrates. These findings re-enforce the concept that IPIs seem to be an important element to finding hippocampal sclerosis at surgery, but likewise challenge the notion that the immature hippocampus is vulnerable to seizurerelated injury. An alternate hypothesis, and one that the authors favor, is that seizures during younger IPIs probably reflect the propensity of the immature brain to seize with cerebral injury. Therefore, seizure-related IPIs may be surrogate markers of cerebral injury, and may not directly produce hippocampal damage unless accompanied by additional pathological processes at the time of the IPI.
(3) In TLE patients, prolonged seizure durations were associated with decreased hippocampal neuronal densities. This finding occurred in all hippocampal subfields and pathological sub-groups (Fig. 7). This was the most direct evidence that repeated non-status limbic seizures may damage the hippocampus. It is important to emphasize, however, that the time course to detect damage was very long (i.e. more than 30 years), does not preferentially affect Sommer's sector or the end-folium, and only partly accounts for the overall neuronal loss noted in hippocampal sclerosis. In other words, repeated TLE seizures most likely add to the damage in hippocampal sclerosis patients, and only mildly affect other TLE patient groups with mass lesions and tumors (Fig. 7). Likewise, our data showed linear correlations and cannot discern if seizures or factors linked with chronic seizures such as secondary
247
Modification of Meyer's Hypothesis: Pathogenesis of Ammon's Horn (Hippocampal) Sclerosis Brain Insul~ One/More Than One?
Genetic Factors
Initial Precipitating
Injury Latent Period
A g e Preference Not A g e Dependent
Short-Term Anatomic Changes
Long-Term Anatomic Changes
Early Hippocampal Sclerosis
Seizures Increase and Plateau
Cell Loss with Axon/Synaptic Reorganization Early Brief Seizures
Final Hippocampal Sclerosis Secondary Cell Loss and Other A n a t o m i c Changes Fig. 6. Graphic illustration depicting a modification of Meyer's original hypothesis concerning Ammon's horn sclerosis pathogenesis. From our analysis, sclerosis is most likely an acquired pathology as a consequence of an IPI. The IPI is generally at a young age, but is not age dependent. Whether the hippocampal damage is from multiple brain insults and/or genetic factors plus a brain injury is suggested from our retrospective analysis of human data. During the latent period, there are probably additional anatomical changes to the hippocampus including synaptic reorganization of excitatory and inhibitory axon systems that probably promote and/or generate spontaneous limbic seizures. Once limbic seizures become established, there are long-term anatomical changes including additional neuronal loss. However, the time course of the long-term changes is over 30 or more years.
h y p o x i a / i s c h e m i a , hormonal changes, anti-epileptic drugs, etc. were the critical factors responsible for the progressive neuronal injury. Therefore, while it is likely that limbic seizures over several decades 'damage' the brain, it is unlikely that repeated seizures over many years 'cause' hippocampal sclerosis based on our c l i n i c a l - p a t h o l o g i c a l analysis. The only possible exception to this concept might be the TLE
cases with dual pathology (lesion plus sclerosis). In this subgroup, 49% had IPIs and many of the macroscopic lesions were located within the limbic system (amygdala, parahippocampus, etc.). In the dual pathology group, this raises the question whether severe hippocampal damage can be the consequence of seizure-induced damage from lesions with direct connections to the hippocampus in the
248
Time Line of Hippocampal Damage In TLE Normal
IPI
Surgery
~t
0
©
TLE Sz Onset
20Latent Period
40-
HS Threshold k..
0
60-
80* - v-
Hipp. Sclerosis Tumor/Cryptogenic
Seizure Duration 20 to 40 Years
100 Fig. 7. Graphic depiction of a proposed hypothesis concerning hippocampal cell loss in patients with TLE. Time is the x-axis and greater cell loss is shown on the y-axis. It is hypothesized that patients with hippocampal sclerosis (solid line) incur most of their cell loss with the IPI. During the latent period, there may be additional anatomical changes that may include cell death, and once TLE seizures begin, there is additional cell loss over many years. Patients with mass lesions or with cryptogenic TLE without hippocampal sclerosis (dashed line) also show long-term progressive hippocampal cell loss, but in general do not reach the threshold of cell loss or the pattern of damage consistent with hippocampal sclerosis unless their seizures continue for more than 30+ years.
one-half of patients without IPIs. This might be a similar mechanism of injury to findings from rat studies showing that kindling might generate hippocampal neuronal loss if the stimulating electrode was in the perforant path (Cavazos and Sutula, 1990; Cavazos et al., 1991; Sutula et al., 1992; Sutula, 2001; but see Mathern et al., 1997b). Further experimental and human studies will be necessary to validate this concept. (4) Analysis to date shows that repeated brief seizures that propagate into the young developing human hippocampus were not linked with neuronal damage, but were associated with improper postnatal fascia dentata development. Granule cell neuronal densities were reduced in children with severe repeated extra-temporal epilepsy, aberrant supragranular mossy fiber sprouting was frequently observed, and PSA-NCAM expression was reduced for newly born granule cells (Fig. 4). These findings support the idea that while seizures during early postnatal development may not 'destroy' existing cells, they
probably adversely affect postnatal fascia dentate neurogenesis, axon formation and other processes that may have a negative impact on brain development and maturation. Based on these findings, we propose that Meyer's original concept of hippocampal sclerosis pathogenesis be modified and updated (Figs. 6 and 7). We concur with Meyer's original notion that most of the hippocampal neuron loss is likely from some initial brain injury, but would expand the hypothesis in support of the idea that the injury is not age dependent nor does it require a prolonged febrile seizure. Instead, IPI-induced hippocampal damage seems to show an age preference and likely results from more than one pathogenic mechanism at the time of injury. The injury factors may include genetic susceptibility to hippocampal injury, and/or more than one excitotoxic event occurring during the IPI. We would also modify Meyer's concept to include the notion of progressive pathological changes after the IPI to explain the latent phase. Specifically,
249 we suggest that anatomical changes occur in the hippocampus and other limbic circuits after IPI-induced neuronal damage that produce the necessary conditions to promote or generate spontaneous seizures. These pathological changes m a y include aberrant excitatory and inhibitory axon sprouting and changes in post-synaptic receptor subunit composition (Kapur and Coulter, 1995; Mathern et al., 1997c, 1998a, 1999; Brooks-Kayal et al., 1998, 1999; Babb, 1999; Coulter, 2001). Furthermore, limbic seizures over many years were associated with additional longterm anatomical changes, such as cell loss. The neuronal loss from chronic seizures is probably a secondary injury that adds to the substrate that is hippocampal sclerosis. Therefore, the generation of hippocampal sclerosis and repeated seizure-induce brain injury likely reflect acute and chronic progressive anatomical and physiological changes, and our pathogenic concepts should be modified to consider this disease as an entity that evolves with time. This concept has recently been supported by neuroimaging and neurocognitive studies (Theodore et al., 1999, 2001; see relevant chapters in this book). Our understanding o f the mechanisms of seizurerelated hippocampal neuronal damage in humans are far from complete, and additional studies will be necessary to confirm many of the concepts proposed in this paper. At present, however, we can say that hippocampal neuron loss can be the 'consequence' of limbic seizures in TLE patients, but that sclerosis is probably related to IPIs and 'causes' TLE. Just as importantly, there is still no substitute for studying the human surgical material using c l i n i c a l pathological principals in order to define and test concepts of pathogenesis in patients with epilepsy.
Acknowledgements This work was supported at the U C L A epilepsy laboratory by NIH grants P01 NS02808 and RO1 NS38992. The clinical work at Ribeir~o Preto was supported by F A P E S P (Proc. #99/11729-2) and CNPq. Thanks to the numerous investigators and epilepsy fellows at U C L A who have contributed to the collection of clinical and research pathological data for over 40 years including Thomas L. Babb, James K. Pretorius, Jann Brown, Harry V. Vinters, Jerome Engel Jr., and Paul H. Crandall.
References Abercrombie, M. (1946) Estimation of nuclear population from microtome sections. Anat. Rec., 94: 239-247. Annegers, J.E, Hauser, W.A., Elveback, L.R. and Kurland, L.T. (1979) The risk of epilepsy following febrile convulsions. Neurology, 29: 297-303. Babb, T.L. (t999) Synaptic reorganizations in human and rat hippocampal epilepsy. Adv. Neurol., 79: 763-779. Babb, T.L. and Brown, W.J. (1987) Pathological findings in epilepsy. In: J. Engel Jr. (Ed.), Surgical Treatment of the Epilepsies. Raven Press, New York, pp. 511-540. Babb, T.L., Brown, W.J., Pretorius, J., Davenport, C., Lieb, J.E and Crandall, P.H. (1984a) Temporal lobe volumetric cell densities in temporal lobe epilepsy. Epilepsia, 25: 729-740. Babb, T.L., Lieb, J.P., Brown, W.J., Pretorius, J. and Crandall, EH. (1984b) Distribution of pyramidal cell density and hyperexcitability in the epileptic human hippocampal formation. Epilepsia, 25: 721-728. Bratz, E. (1899) Ammonshornbefunde bei epileptikern. Arch. Psychiatr. Nervenkr, 32: 820-835. Brooks-Kayal, A.R., Shumate, M.D., Jin, H., Rikhter, T.Y. and Coulter, D.A. (1998) Selective changes in single cell GABA(A) receptor subunit expression and function in temporal lobe epilepsy. Nat. Med., 4: 1166-1172. Brooks-Kayal, A.R., Shumate, M.D., Jin, H., Lin, D.D., Rikhter, T.Y., Holloway, K.L. and Coulter, D.A. (1999) Human neuronal gamma-aminobutyric acid(A) receptors: coordinated subunit mRNA expression and functional correlates in individual dentate granule cells. J. Neurosci., 19: 8312-8318. Camfield, E, Camfield, C., Gordon, K. and Dooley, J. (1994) What types of epilepsy are preceded by febrile seizures? A population-based study of children. Dev. Med. Child Neurol., 36:887 892. Cavazos, J.E. and Sutula, T.E (1990) Progressive neuronal loss induced by kindling: a possible mechanism for mossy fiber synaptic reorganization and hippocampal sclerosis. Brain Res., 527: 1-6. Cavazos, J.E., Golarai, G. and Sutula, T.E (1991) Mossy fiber synaptic reorganization induced by kindling: time course of development, progression, and permanence. J, Neurosci., 11: 2795-2803. Coulter, D.A. (2001) Epilepsy-associated plasticity in gammaaminobutyric acid receptor expression, function, and inhibitory synaptic properties. Int. Rev. Neurobiol., 45: 237-252. Crandall, EH. and Mathern, G.W. (2000) Surgery for lesional temporal lobe epilepsy. In: H.O. Luders and Y.G. Comair (Eds.), Epilepsy Surgery, 2rid edn. Lippincott Williams and Wilkins, Philadelphia, PA, pp. 653-665. Engel Jr., J. (1993) Update on surgical treatment of the epilepsies. Summary of the Second International Palm Desert Conference on the Surgical Treatment of the Epilepsies (1992). Neurology, 43: 1612-1617. Falconer, M.A. (1974) Mesial temporal (Ammon's horn) sclerosis as a common cause of epilepsy. Aetiology, treatment, and prevention. Lancet, 2: 767-770. Gloor, P. (1991) Mesial temporal sclerosis: Historical back-
250
ground and an overview from a modern perspective. In: H.O. Ltiders (Ed.), Epilepsy Surgery. Raven Press, New York, pp. 689-703. Holmes, G.L. (1997) Epilepsy in the developing brain: lessons from the laboratory and clinic. Epilepsia, 38: 12-30. Holmes, G.L. and Thompson, J.L. (1988) Effects of kainic acid on seizure susceptibility in the developing brain. Brain Res., 467: 51-59. Kapur, J. and Coulter, D.A. (1995) Experimental status epilepticus alters gamma-aminobutyric acid type A receptor function in CA1 pyramidal neurons. Ann. Neurol., 38: 893-900. Katzir, H., Mendoza, D. and Mathern, G.W. (2000) Effect of theophylline and trimethobenzamide when given during kainate-induced status epilepticus: an improved histopathologic rat model of human hippocampal sclerosis. Epilepsia, 41: 1390-1399. Lado, F.A., Sankar, R., Lowenstein, D. and Moshe, S.L. (2000) Age-dependent consequences of seizures: relationship to seizure frequency, brain damage, and circuitry reorganization. Ment. Retard. Dev. Disabil. Res. Rev., 6: 242-252. Leite, J.P., Babb, T.L., Pretorius, J.K., Kuhlman, P.A., Yeoman, K.M. and Mathern, G.W. (1996) Neuron loss, mossy fiber sprouting, and interictal spikes after intrahippocampal kainate in developing rats. Epilepsy Res., 26: 219-231. Lorente de No, R. (1934) Studies on the structure of the cerebral cortex. II. Continuation of the study of the ammonic system. J. PsychoL Neurol., 45:113-177. Margerison, J.H. and Corsellis, J.A. (1966) Epilepsy and the temporal lobes. A clinical, electroencephalographic and neuropathological study of the brain in epilepsy, with particular reference to the temporal lobes. Brain, 89: 499-530. Mathern, G.W., Leite, J.P., Pretorius, J.K., Quinn, B., Peacock, W.J. and Babb, T.L. (1994) Children with severe epilepsy: evidence of hippocampal neuron losses and aberrant mossy fiber sprouting during postnatal granule cell migration and differentiation. Dev. Brain Res., 78: 70-80. Mathern, G.W., Babb, T.L., Vickrey, B.G., Melendez, M. and Pretorius, J.K. (1995a) The clinical-pathogenic mechanisms of hippocampal neuron loss and surgical outcomes in temporal lobe epilepsy. Brain, 118: 105-118. Mathern, G.W., Pretoflus, J.K. and Babb, T.L. (1995b) Influence of the type of initial precipitating injury and at what age it occurs on course and outcome in patients with temporal lobe seizures. J. Neurosurg., 82: 220-227. Mathern, G.W., Babb, T.L., Leite, J.P., Pretorius, K., Yeoman, K.M. and Kuhlman, P.A. (1996a) The pathogenic and progressive features of chronic human hippocampal epilepsy. Epilepsy Res., 26: 151-161. Mathern, G.W., Babb, T.L., Mischel, P.S., Vinters, H.V., Pretoflus, J.K., Leite, J.P. and Peacock, W.J. (1996b) Childhood generalized and mesial temporal epilepsies demonstrate different amounts and patterns of hippocampal neuron loss and mossy fibre synaptic reorganization. Brain, 119: 965-987. Mathern, G.W., Babb, T.L. and Armstrong, D.L. (1997a) Hippocampal sclerosis. In: J. Engel Jr. and T.A. Pedley (Eds.), Epilepsy: A Comprehensive Textbook, Vol. 1. LippincottRaven, Philadelphia, PA, pp. 133-155.
Mathern, G.W., Bertram III, E.H., Babb, T.L., Pretorius, J.K., Kuhlman, P.A., Spradlin, S. and Mendoza, D. (1997b) In contrast to kindled seizures, the frequency of spontaneous epilepsy in the limbic status model correlates with greater aberrant fascia dentata excitatory and inhibitory axon sprouting, and increased staining for N-methyl-D-aspartate, AMPA and GABA(A) receptors. Neuroscience, 77: 1003-1019. Mathern, G.W., Pretorius, J.K., Kornblum, H.I., Mendoza, D., Lozada, A., Leite, J.P., Chimelli, L.M., Fried, I., Sakamoto, A.C., Assirati, J.A., Levesque, M.E, Adelson, P.D. and Peacock, W.J. (1997c) Human hippocampal AMPA and NMDA mRNA levels in temporal lobe epilepsy patients. Brain, 120: 1937-1959. Mathern, G.W., Pretorius, J.K., Mendoza, D., Lozada, A. and Kornblum, H.I. (1998a) Hippocampal AMPA and NMDA mRNA levels correlate with aberrant fascia dentata mossy fiber sprouting in the pilocarpine model of spontaneous limbic epilepsy. J. Neurosci. Res., 54: 734-753. Mathern, G.W., Price, G., Rosales, C., Pretoflus, J.K., Lozada, A. and Mendoza, D. (1998b) Anoxia during kainate status epilepficus shortens behavioral convulsions but generates hippocampal neuron loss and supragranular mossy fiber sprouting. Epilepsy Res., 30: 133-151. Mathern, G.W., Pretorius, J.K., Mendoza, D., Leite, J.P., Chimelli, L., Born, D.E., Fried, I., Assirati, J.A., Ojemann, G.A., Adelson, P.D., Cahan, L.D. and Kornblum, H.I. (1999) Hippocampal N-methyl-D-aspartate receptor subunit mRNA levels in temporal lobe epilepsy patients. Ann. Neurol., 46: 343-358. Meldrum, B.S. (1997) First Alfred Meyer Memorial Lecture. Epileptic brain damage: a consequence and a cause of seizures. Neuropathol. Appl. Neurobiol., 23: 185-201; discussion 201202. Meyer, A., Falconer, M.A. and Beck, E. (1954) Pathological findings in temporal lobe epilepsy. J. Neurol. Neurosurg. Psychiatry, 17: 276-285. Meynert, T. (1867) Studien uber das pathologisch-anatomissche material der Wiener Irren-Anstalt. Vieteljahrssch., 3: 381-402. Moshe, S.L. (1993) Seizures in the developing brain. Neurology, 43: $3-$7. Nelson, K.B. and Ellenberg, J.H. (1976) Predictors of epilepsy in children who have experienced febrile seizures. New Engl. J. Med., 295: 1029-1033. Pfleger, L. (1880) Beobachtungen uber schrumpfung und skierose des ammonshorns bei epilepsie. Allg. Z. Psychiatr., 36: 359-365. Sommer, W. (1880) Erkrankung des ammonshorns als aetiologisches moment der epilepsie. Arch. Psychiatr. Nervenkr., 10: 631-675. Stander, K.H. (1936) Epilepsie und schlafenlappen. Arch. Psychiatr. Nervenkr., 104: 181-211. Sutula, T.P. (2001) Secondary epileptogenesis, kindling, and intractable epilepsy: a reappraisal from the perspective of neural plasticity. Int. Rev. Neurobiol., 45: 355-386. Sutula, T., Cavazos, J. and Golarai, G. (1992) Alteration of long-lasting structural and functional effects of kainic acid
251 in the hippocampus by brief treatment with phenobarbital. J. Neurosci., 12: 4173-4187. Theodore, W.H., Bhatia, S., Hatta, J., Fazilat, S., DeCarli, C., Bookheimer, S.Y. and Gaillard, W.D. (1999) Hippocampal atrophy, epilepsy duration, and febrile seizures in patients with partial seizures. Neurology, 52:132-136.
Theodore, W.H., Gaillard, W.D., De Carli, C., Bhatia, S. and Hatta, J. (2001) Hippocampal volume and glucose metabolism in temporal lobe epileptic foci. Epilepsia, 42: 130-132. Verity, C.M., Ross, E.M. and Golding, J. (1993) Outcome of childhood status epilepticus and lengthy febrile convulsions: findings of national cohort study. Br. Med. J., 307: 225-228.
T. Sutula and A. Pitk~inen (Eds.) Progress in Brain Research, Vol. 135 © 2002 Elsevier Science B.V. All rights reserved
CHAPTER 22
MRI studies. Do seizures damage the brain? John S. Duncan 1,2,* 1 University College London, London, UK 2 National Society for Epilepsy, Chalfont St. Peter, Buckinghamshire SL90LR, UK
Abstract: Methods to assess the development of cerebral damage need to be quantitative, reliable, reproducible and safe. They must be acceptable to patients and to a healthy control group, for repeated use and the acquisition and analytical methods must be stable over years. Longitudinal studies are necessary to determine whether secondary cerebral damage occurs as a consequence to the epilepsies. The principal aim of longitudinal studies is to detect physical evidence of brain damage when it occurs. Patient groups will be heterogeneous in this regard and analysis will need to be not only of changes in group means, but also of the number of patients who show significant changes in imaging parameters, that exceed the limits of test-retest reliability. MRI is attractive as a tool to evaluate the presence and development of cerebral damage in patients with epilepsy. MRI is readily available and non-invasive, making it acceptable to patients and controls. MRI volumetry is reliable and reproducible, but the sensitivity of the method to detect subtle abnormalities has not yet been established. Longitudinal studies are ongoing in patients with newly diagnosed and chronic epilepsy, with an inter-scan interval of 3.5 years, using complementary voxel-based and region-based methods that can detect changes in hippocampal and cerebellar volumes of 3% and neocortical volume changes of 1.6%. MR spectroscopy may be more sensitive for detecting abnormalities, but the test-retest reliability is less good. Other MRI tools, such as diffusion tensor imaging, may be useful methods for evaluating secondary cerebral damage acutely and chronically.
Introduction A superficially straightforward central question in clinical epileptology is "do seizures cause brain damage?" On further inspection, this leads to a range of questions and appreciation of the complexity of the situation. The epilepsies are a very heterogeneous range of conditions of which overt seizures are but one manifestation. Apparently similar seizures may result in cerebral damage in the context of one form of epilepsy but not in another. Subclinical seizures and interictal epileptiform activity might also result in cerebral damage. May some medications increase or decrease the risk of secondary cerebral * Correspondence to: J.S. Duncan, National Society for Epilepsy, Chalfont St. Peter, Buckinghamshire SL9 0LR, UK. Tel.: -I-44-14-9460-1341; Fax: +44-14-9487-6294; E-mail: j.duncan @ion.ucl.ac.uk
damage? Furthermore, by virtue of genetic predisposition, may some individuals be more at risk than others? It is necessary to be able to identify which patients are at risk of secondary cerebral damage. If therapies can be designed to prevent or ameliorate these effects, methods to determine whether the therapies are effective will be crucial. The assessment of physical status and cognitive function will measure the consequences of cerebral damage. Serial EEG studies have not been shown to be a sensitive indicator in this regard. The need is for more sensitive in vivo assessment methods. The requirements are that: The techniques have to be quantitative, reliable, reproducible and safe. They must be acceptable to patients and to a healthy control group, for repeated use over a period of years. Furthermore, the acquisition and analytical methods must be stable over years. If large numbers
254 of patients are to be evaluated it is beneficial if the methods can be reliably applied in multiple sites.
Study design Longitudinal studies need to be large enough to have adequate power to detect changes of clinical significance, and also need to be of sufficient duration to identify differences between healthy controls and patients and between active treatment and placebotreated groups of patients. The objectives of longitudinal imaging studies are: First, to detect physical evidence of brain damage when it occurs. It must be recognized that patient groups will be heterogeneous in this regard and that analysis will need to be not only of changes in group means, but also of the number of patients who show significant changes in imaging parameters, that exceed the limits of test-retest reliability. Second, in an interventional study, to show that treatment is associated with absence of damage, which occurs in the non-treated parallel group. Study populations will need to be stratified according to age, epilepsy and seizure types, and for partial seizures, the localization of seizure onset. The optimal duration of longitudinal studies needs careful consideration. A longer study increases the chance of cerebral changes and differences becoming evident, but is more costly and there is likely to be increased difficulty with subject dropout and with variation of the imaging acquisition and analysis equipment and protocols.
The range of potential imaging tools in evaluating cerebral damage in epilepsy In vivo imaging studies have the potential to identify and to quantify secondary cerebral damage as a result of epilepsy, before there is any clinical accompaniment, and to act as a surrogate endpoint for intervention and preventative strategies. MRI has a number of advantages that make it attractive as a tool to evaluate the presence and development of cerebral damage in patients with epilepsy. MRI testing is more readily available and comparatively less expensive than either positron emission tomography (PET) or single photon emis-
sion computed tomography (SPECT). Additionally, its non-invasive nature, and absence of ionizing radiation make it more acceptable to patients and controls participating in long-term studies.
Single photon emission computed tomography Single photon emission computed tomography is widely available, with tracers that are sensitive to cerebral blood flow, and there are specific tracers for the central benzodiazepine receptor. When combined with co-registration techniques, SPECT allows the mapping of the area of brain involved in the generation of seizures (O'Brien et al., 1999a). The limitations of temporal resolution result in the possibility of imaging secondary spread rather than the site of seizure onset. SPECT is more widely available than PET and is relatively inexpensive. However, the technique is only semi-quantifiable, with the use of an internal reference region and radiation exposure is a concern as it is with PET, and the test-retest coefficient of reliability is 15% at best (Varrone et al., 2000).
Positron emission tomography There are a variety of PET ligands available. These ligands allow the measurement of glucose metabolism, central benzodiazepine receptors, various opioid receptor subtypes, and dopamine receptors. PET produces data that may be analyzed quantitatively. Arterial cannulation is often necessary for accurate quantification of cerebral PET data. 18F-Fluorodeoxyglucose PET may give rise to parametric images of regional cerebral glucose utilization that is sensitive but non-specific to a range of cerebral pathologies. Hypometabolism is a sensitive but nonspecific marker of cerebral dysfunction. Regional hypometabolism occurs in about 90% of patients with medial temporal lobe epilepsy (TLE) (Gaillard et al., 1995). Focal or diffuse regional hypometabolism occurs in about 70% of patients with neocortical epilepsy (Engel et al., 1995). Hypometabolism is not only caused by neuronal loss. There is an additional metabolic disturbance. This is in contrast to MRI volume loss and neuronal loss that are very well correlated. The degree of medial temporal lobe hypometabolism correlated well with post-operative
255 outcome (Radtke et al., 1993). Limitations of PET for longitudinal studies include concerns with radiation exposure, lack of availability, high cost, the semiinvasive nature of the test, and a test-retest reliability of only 10% at best (e.g. Vilkman et al., 2000). Magnetic resonance spectroscopy
Magnetic resonance spectroscopy (MRS) is considered in Bernasconi et al. (2002, Chapter 25, this volume). This shares the advantages of MRI of being non-invasive and well-tolerated. The use of MRS allows the evaluation of both the integrity and function of the neurones by measuring N-acetylaspartate (NAA), a normal byproduct of neuronal cellular metabolism. NAA is a marker of neuronal cell dysfunction, not just volume loss. Other metabolites that can be measured with this technique include choline, creatine, lactate, GABA, glutamate, and glutamine. Abnormalities of metabolite profiles may be found in temporal lobes with normal MR/(Knowlton et al., 1997; Connelly et al., 1998) and bilateral abnormalities have been noted in up to 50% of patients with apparently unilateral structural abnormality (Ende et al., 1997), indicating that MRS may be more sensitive for detecting pathology. MRS may be more helpful in lateralizing the epileptic temporal lobe in patients with bilateral hippocampal atrophy than volumetric studies. The ability of MRS to detect abnormalities (83%) is similar to the ability of structural M R / t o detect hippocampal volume loss (Cendes et al., 1995). When these two methods are combined, the ability to detect abnormalities increases to 93%. (Cendes et al., 1995). The sensitivity to abnormality may be greater than MRI, but changes are non-specific and the testretest coefficient of reliability for the measurement of NAA using MRS is approximately 15-20%, and the reliability is less good for the other metabolites (Woermann et al., 1999a). Volumetric MRI
Volumetric 3D Tl-weighted MRI produces scans of good anatomical definition, with voxels typically of 0.9 mm 3. The technique of hippocampal volumetry has been established for a decade (Jack et al., 1990; Cook et al., 1992), using an interactive process. This
has been used to demonstrate the spectrum of severity of hippocampal sclerosis. Hippocampal atrophy is identified with hippocampal volume measures and this correlates well with hippocampal neurone loss, particularly in the CA1 sub-region (Van Paesschen et al., 1997a,b). Amygdala volumes may be similarly determined (Cendes et al., 1993a; Van Elst et al., 2000). Grey and white matter may be segmented, with operator dependent (Sisodiya et al., 1995) and automated procedures (Lemieux et al., 2000) so that measures may be derived of cerebral hemisphere grey matter and subcortical volumes. Furthermore, the distributions of grey and white matter may be compared between groups of subjects, and between a single subject and a group using regional measures (Sisodiya et al., 1995) and voxel-based morphometry (Richardson et al., 1997; Woermann et al., 1999b). The test-retest reliability of MRI-measured hippocampal volumes, cerebral volumes is 3% (Lemieux et al., 2000). The severity of hippocampal atrophy on the side of the language-dominant hemisphere is an important determinant of impairment of verbal memory following hippocampal resection. The more severe the atrophy pre-operatively, the less likely it is that there will be a significant decline of verbal memory after surgery (Trenerry et al., 1993). T2 relaxometry allows a quantitative determination of the T2-signal changes. An approximation of the T2-relaxation time may be obtained by a variety of methods, for example using 16-echo times (Jackson et al., 1993) or 2-echoes (Duncan et al., 1996). The latter had the advantage of compete brain coverage in 5-mm-thick slices. Partial volume effect with CSF is a confound that needs to be avoided with T2 relaxometry, in view of the long T2 of CSE Reliable T2-measures may be obtained in the hippocampus (Jackson et al., 1993; Duncan et al., 1996) and the amygdala (Van Paesschen et al., 1996) as boundaries with CSF may be avoided by careful region definition. Increases in hippocampal T2 relaxation time correlate with the glial/neurone ratio, particularly in the CA1 subregion (Van Paesschen et al., 1997a,b). The T2-relaxation time may be measured along the length of the hippocampus, giving a profile of abnormality (Woermann et al., 1998). Abnormalities are also seen contralaterally in about 30% of cases with clear cut HS (Jackson et al., 1993).
256
Diffusion tensor imaging Diffusion tensor imaging (DTI) holds promise for detecting areas of neuronal damage. DTI is an MR method for identifying the motion of water in the brain, that can be quantified by both voxel-based and region-based methods. The two main parameters determined by DTI are diffusivity and fractional anisotropy (Eriksson et al., 2001). Increased diffusivity is likely to correlate with neurone loss and gliosis, but formal correlative studies have not yet been performed. Fractional anisotropy reflects the asymmetry of the motion of the fluid. Motion of fluid within the brain is normally restricted to movement in the same axis as the axon or myelin sheath. When there is damage to neurones or myelin sheaths, the fractional anisotropy decreases because the fluid can move freely in various axes (Eriksson et al., 2001; Rugg-Gunn et al., 2001). The methods may identify abnormalities in patients with epilepsy that are not evident in vivo using conventional MRI (Rugg-Gunn et al., 2001). It is possible that serial DTI would be a sensitive indicator of developing cerebral damage.
Functional MRI (fMRI) At present, fMRI may lateralize language function (Detre et al., 1998). Attempts to reliably localize the parts of the brain that are involved in language and memory function are being made (Binder et al., 2000; Dupont et al., 2000). There are important caveats. First, the involvement of a part of brain in a task does not mean that that area is crucial for the task. Second, if an area is not activated with a particular fMRI paradigm it does not mean that that area is not involved in the task. Third, the extent and height of the activation in a task, found using fMRI, may bear no relation to the competence with which that task is performed.
MR studies of disease progression in other neurological conditions Alzheimer's disease The association of cerebral atrophy with Alzheimer's disease (AD) and other dementias is well established and there is reasonable correlation between
the severity of atrophy and cognitive impairment. The progression of AD may also be followed in vivo with serial quantitative MRI (Fox et al., 1996, 2000). This particular study found that the mean (SD) rate of brain atrophy for patients with AD was 2.37% (1.11%) per year, while in the control group it was 0.41% (0.47%) per year. From these figures, in order to have 90% power to detect a drug effect equivalent to a 20% reduction in the rate of atrophy, 207 patients would be needed in each treatment arm in a 1-year placebo-controlled trial with a 10% patient dropout rate, and if 10% of scan pairs were unusable. Methodologies need to be precise, to minimize the noise of the measurements. In a study of hippocampal volumes over 3 years in 27 patients with AD, the range of hippocampal volume loss was - 2 . 3 to -15.6%, compared to - 2 . 2 to - 5 . 8 % in control subjects (Laakso et al., 2000). The observed changes in individual subjects were small, and within the accuracy range of the measurements.
Multiple sclerosis In multiple sclerosis, cerebral atrophy reflects parenchymal destruction and in some studies, correlates with disability (Fisher et al., 2000). A reduction of cerebral volume in those treated with interferon~31b and those receiving placebo of 2.9 and 3.9%, respectively, was noted in a recent multicentre trial (Molyneux et al., 2000), but the finding that there was no correlation between disability and change in cerebral volume raises the query of the significance of atrophy. Other clinical-MRI correlations were also not as clear as might have been imagined, such as the absence of correlation between suppression of T2-1esion load with absence of progression of disability (Ebers, 2000).
MRI studies in epilepsy Most attention has focussed on the hippocampus because: • The techniques of hippocampal measurements have been established for many years. • There is a clear association between hippocampal sclerosis with temporal lobe epilepsy. • Temporal lobe epilepsy is associated with cognitive impairment, particularly memory.
257 • Temporal lobe epilepsy is one of the more common homogenous forms of epilepsy, so adequate numbers of patients are available for studies. Cross-sectional studies of the association between the severity of hippocampal sclerosis (HS), and seizure frequency and duration have produced conflicting results (Cendes et al., 1993b; Van Paesschen et al., 1997a; K~ilvi~iinen et al., 1998; Theodore et al., 1999; Salmenper~i et al., 2001). In a crosssectional analysis of a community-based cohort, the mean hippocampal volume (HV), corrected for intracranial volume (ICV), in patients with chronic localization-related epilepsy was 6% less than in those with newly diagnosed partial seizures and 10% less than the control group (Everitt et al., 1998). In a separate population, there was an 18% and 14% reduction of the left and right HV, respectively, on the side ipsilateral to the seizure focus in patients with chronic drug-resistant epilepsy (K~ilvi~iinen et al., 1998; Salmenper~i et al., 2001). In 82 patients with refractory TLE, hippocampal volumes were inversely related to duration of epilepsy, ipsilateral to the epileptic focus, but not contralaterally (Tasch et al., 1999). Complex partial seizure frequency was not related, but patients with frequent secondarily generalize seizures had smaller ipsilateral hippocampi. The conclusions that can be drawn on disease progression from cross-sectional studies, however, are very limited, because: (1) small changes in structure over time are often masked by large biological variability across subjects; and (2) cross-sectional studies are unable to give direct information on the causal relationship between seizures and structural brain damage. The question of whether chronic epilepsy results in smaller hippocampi, or whether a reduction in hippocampal volume determines intractability can only be addressed by longitudinal studies. There have been case reports of progressive hippocampal sclerosis in patients suffering from recurrent partial and secondarily generalized seizures (O'Brien et al., 1999b), and status epilepticus (e.g. Nohria et al., 1994; Wieshmann et al., 1997; Van Landingham et al., 1998). Van Paesschen et al. reported significant reductions of hippocampal volume in 8% of patients with partial seizures scanned 1 year apart (Van Paesschen et al., 1998). The changes were considered to be either the result of frequent seizures
or the resolution of oedema after initial seizures. It is likely that recurrent seizures, especially secondarily generalized seizures, can induce secondary hippocampal changes in some patients, but this is not universal. Only 50-75% of hippocampal resections for intractable temporal lobe epilepsy show neuronal loss in the dentate gyrus and hippocampus proper (Margerison and Corsellis, 1966; Honovar and Meldrum, 1991). This leads to the question - - what factors make an individual susceptible to secondary hippocampal and extra-temporal atrophy? In order to address these questions, in 1995, a prospective, community-based longitudinal followup study of patients with newly diagnosed seizures and chronic active epilepsy was established at the Chalfont Centre for Epilepsy, funded by the Wellcome Trust. Ninety patients with newly diagnosed seizures, 154 with chronic active epilepsy (defined as epilepsy for more than 4 years and a seizure in the last year) and 80 control subjects had baseline MRI scans between June 1995 and May 1997, and follow up scans 3.5 years later. As epilepsy might result in damage not only to the hippocampus, but also to the cerebral neocortex and the cerebellum, it was decided to make a quantitative assessment of all of these structures. This investigation required the establishment and implementation of optimal techniques for quantitative MRI and for serial measures using region-based hippocampal volumetry, hippocampal T2-relaxation times, automated measures of cerebral hemisphere volumes, cerebral grey matter, CSF and intracranial volumes and semiautomated cerebellar volumes. Prior to volumetry, a series of automatic processing steps were carried out on the Tl-weighted volume datasets. After an initial automatic brain segmentation of the baseline and repeat scan using a 2D version of our segmentation software Exbrain (Lemieux et al., 2000) non-uniformity correction was performed, using the automatic method, N3 (Sled et al., 1998; http://www.bic.mni.mcgill.ca/ brainweb/). Automatic brain segmentation of the non-uniformity corrected baseline scan was then performed using the 3D version of Exbrain, resulting in an accurate delineation of the brain (Lemieux et al., 2000) and CSF (Lemieux, 2001). In the segmented scans, all voxels outside the brain are set to zero intensity. The repeat scan was then co-
258 registered and intensity matched to the segmented baseline scan using our software MRreg (Lemieux et al., 1998), (Lemieux and Barker, 1998). In MRreg, a 9-parameter rigid body transformation (three rotation, three translation and three scaling), was used to register images with an accuracy of <0.06 mm in each linear dimension and correct for variations in voxel dimensions. The matched repeat scan was then resampled using sinc-based interpolation, with a kernel radius of 5 voxels. A final automatic segmentation of the brain and CSF in the matched repeat scan was then performed using Exbrain. These methods reliably detect individual hippocampal and cerebellar volume changes greater than 3.1 and 3%, respectively (Lemieux et al., 2000). Preliminary results from the first 53 subjects who have been rescanned (24 with chronic active epilepsy, 9 with newly diagnosed seizures and 20 healthy controls) has shown significant reduction of hippocampal volume in four individuals (three with chronic epilepsy and one control), and significant reductions of cerebellar volume, total brain volume and grey matter volume in two (chronic epilepsy), three and one subject, respectively (Liu et al., 2001). One patient with newly diagnosed seizures, who ceased a previously excessive alcohol intake after the baseline scan, showed a significant increase in hippocampal volumes of 6.6 and 8.8%, and similar increases in cerebellar and total brain volumes (Liu et al., 2000). Rescanning of the whole cohort will be complete in September 2001, and the data then analyzed and the identity of risk factors and associations with cerebral damage will be sought. Complementary voxel-based analysis of T1weighted volumetric images comprises co-registration and subtraction of the follow-up scan from the baseline image. The difference images display those voxels that change in signal intensity, i.e. grey matter to CSF, white matter to CSF. The raw difference images are then filtered to remove the noise of the serial imaging and co-registration process by ignoring those voxels which showed signal change in any one of 40 pairs of scans obtained in a prior group of control subjects (Lemieux et al., 1998). In subsequent analyses, to give anatomic localization to the findings, both the segmented neocortical images and the voxel-based subtraction images
are divided up into a series of anatomical subregions using an anatomical MRI-template that may be co-registered to individual subjects' MRI scans (Hammers et al., 2000).
Other possible MR tools for longitudinal studies A potential criticism of serial volumetric studies of cerebral structure is that although the results are reliable and reproducible, they may be insensitive to the initial development of secondary cerebral damage. It is a possibility that acute post-ictal imaging of cerebral perfusion and diffusion may identify abnormalities that are associated with the development of secondary cerebral damage. Neuronal loss results in increased diffusivity, and ictal diffusion-weighted imaging has shown reduced diffusivity at an epileptic focus, most likely due to cell swelling. The development of neuroprotective strategies to limit neuronal damage resulting from seizures is an important area of current interest. Non-invasive methods to assess secondary neuronal damage and which may be used as a measure of treatment efficacy could provide a surrogate marker for the acute evaluation of neuroprotective agents.
Conclusion Quantitative MRI techniques have the advantage of being flexible, reliable, reproducible and well tolerated by patients and controls. The need for consistency and great attention to detail will limit their application to dedicated centres. Quantitative analysis of Tl-weighted volumetric MRI can reliably detect changes in volume of cerebral structure of 3%. Ongoing longitudinal studies will determine whether changes of this magnitude in hippocampi, cerebral neocortex and cerebellum are common or exceptional over a 3.5-year interval in patients with newly diagnosed seizures and chronic epilepsy. These measures could find application as surrogate endpoints in the evaluation of neuroprotectant agents, but their sensitivity in relation to the development of functional impairment is not yet clear. Other quantitative MRI methods, such as the mean diffusivity of water as assessed using DTI and acute post-ictal imaging studies, may prove to be useful and sensitive measures of cerebral damage.
259
Acknowledgements I a m v e r y grateful to m y c o l l e a g u e s in the E p i l e p s y I m a g i n g G r o u p and for the support o f the N a t i o n a l S o c i e t y for Epilepsy, W e l l c o m e Trust, M e d i c a l Research C o u n c i l and A c t i o n R e s e a r c h .
References Binder, J.R., Frost, J.A., Hammeke, T.A., Bellgowan, ES., Springer, J.A., Kaufman, J.N. and Possing, E.T. (2000) Human temporal lobe activation by speech and nonspeech sounds. Cereb. Cortex, 10: 512-528. Bernasconi, A., Tasch, E., Cendes, E, Li, L.M. and Arnold, D.L. (2002) Proton magnetic resonance spectroscopic imaging suggests progressive neuronal damage in human temporal lobe epilepsy. In: T. Sutula and A. Pitk~nen (Eds.), Do Seizures Damage the Brain. Progress in Brain Research, Vol. 135. Elsevier, Amsterdam, pp. 297-304. Cendes, E, Andermann, F., Gloor, P., Evans, A., Jones-Gotman, M., Watson, C., Melanson, D., Olivier, A., Peters, T. and Lopes-Cendes, I. et al. (1993a) MRI volumetric measurement of amygdala and hippocampus in temporal lobe epilepsy. Neurology, 43: 719-725. Cendes, E, Andermann, F., Gloor, P., Lopes-Cendes, I., Andermann, E., Melanson, D., Jones-Gotman, M., Robitaille, Y., Evans, A. and Peters, T. (1993b) Atrophy of mesial structures in patients with temporal lobe epilepsy: cause or consequence of repeated seizures?. Ann. NeuroL, 34: 795-801. Cendes, F., Andermann, E, Dubeau, E and Arnold, D.L. (1995) Proton magnetic resonance spectroscopic images and MRI volumetric studies for lateralization of temporal lobe epilepsy. Magn. Reson. Imaging, 13:1187-1191. Connelly, A., Van Paesschen, W., Porter, D.A., Johnson, C.L., Duncan, J.S. and Gadian, D.G. (1998) Proton magnetic resonance spectroscopy in MRI-negative temporal lobe epilepsy. Neurology, 51: 61-66. Cook, M.J., Fish, D.R., Shorvon, S.D., Straughan, K, and Stevens, J.M. (1992) Hippocampal volumetric and morphometric studies in frontal and temporal lobe epilepsy. Brain, 115: 1001-1015. Detre, J.A., Maccotta, L., King, D., Alsop, D.C., Glosser, G., D'Esposito, M., Zarahn, E., Aguirre, G.K. and French, J.A. (1998) Functional MRI lateralization of memory in temporal lobe epilepsy. Neurology, 50: 926-932. Duncan, J.S., Bartlett, P. and Barker, G.J. (1996) Technique for measuring hippocampal T2 relaxation time. Am. J. NeuroradioI., 17: 1805-1810. Dupont, S., Van de Moortele, P.E, Samson, S., Hasboun, D., Poline, J.B., Adam, C., Lehericy, S., Le Bihan, D., Samson, Y. and Baulac, M. (2000) Episodic memory in left temporal lobe epilepsy: a functional MRI study. Brain, 123: 1722-1732. Ebers, G. (2000) MRI: measure of efficacy. Brain, 123: 21872188. Ende, G.R., Laxer, K.D., Knowlton, R.C., Matson, G.B., Schuff, N., Fein, G. and Weiner, M.W. (1997) Temporal lobe epilepsy:
bilateral hippocampal metabolite changes revealed at proton MR spectroscopic imaging. Radiology, 202: 809-817. Engel, J., Henry, T.R. and Swartz, B.E. (1995) Positron emission tomography in frontal lobe epilepsy. In: H.H. Jasper, S. Riggio and P.S. Goldman-Rakic (Eds.), Epilepsy and the Functional Anatomy of the Frontal Lobe. Raven Press, New York, pp. 223-238. Eriksson, S.H., Symms, M.R., Rugg-Gunn, EJ., Barker, G.J. and Duncan, J.S. (2001) Diffusion tensor imaging in patients with epilepsy and malformations of cortical development. Brain, 124: 617-626. Everitt, A.D., Birnie, K.D., Stevens, J.M., Sander, J.W., Duncan, J.S. and Shorvon, S.D. (1998) The NSE MRI study: structural brain abnormalities in adult epilepsy patients and healthy controls. Epilepsia, 39(Suppl. 6): 140. Fisher, E., Rudick, R.A., Cutter, G., Baier, M., Miller, D., Weinstock-Guttman, B., Mass, M.K., Dougherty, D.S. and Simonian, N.A. (2000) Relationship between brain atrophy and disability: an 8-year follow-up study of multiple sclerosis patients. Mult. Scler., 6: 373-377. Fox, N.C., Freeborough, P.A. and Rossor, M.N. (1996) Visualisation and quantification of rates of atrophy in Alzheimer's disease. Lancet, 348(9020): 94-97. Fox, N.C., Cousens, S., Scahill, R., Harvey, R.J. and Rossor, M.N. (2000) Using serial registered brain magnetic resonance imaging to measure disease progression in Alzheimer disease: power calculations and estimates of sample size to detect treatment effects. Arch. Neurol., 57: 339-344. Gaillard, W.D., Bhatia, S., Bookheimer, S.Y., Fazilat, S., Sato, S. and Theodore, W.H. (1995) FDG-PET and volumetric MRI in the evaluation of patients with partial epilepsy. Neurology, 45: 123-126. Hammers, A., Koepp, M.J., Free, S.L., Labbe, C., Brooks, D.J., Cunningham, V.J. and Duncan, J.S. (2000) Implementation and application of a new brain template for multiple volumes of interest. Neurolmage, 11 : $61. Honovar, M. and Meldrum, B.S. (1991) In: D.I. Graham and P.L. Arnold (Eds.), Greenfield's Neuropathology. Arnold, London, pp. 931-971. Jack Jr., C.R., Sharbrough, F.W., Twomey, C.K., Cascino, G.D., Hirschorn, K.A., Marsh, W.R., Zinsmeister, A.R. and Scheithauer, B. (1990) Temporal lobe seizures: lateralization with MR volume measurements of the hippocampal formation. Radiology, 175: 423-429. Jackson, G.D., Connelly, A., Duncan, J.S., Grunewald, R.A. and Gadian, D.G. (1993) Detection of hippocampal pathology in intractable partial epilepsy: increased sensitivity with quantitative magnetic resonance T2 relaxometry. Neurology, 43: 1793-1799. K~lvi~iinen, R., Salmenper~i, T., Partanen, K., Vainio, P., Riekkinen, P. and Pitkanen, A. (1998) Recurrent seizures may cause hippocampal damage in temporal lobe epilepsy. Neurology, 50: 1377-1382. Knowlton, R.C., Laxer, K.D., Ende, G., Hawkins, R.A., Wong, S.T., Matson, G.B., Rowley, H.A., Fein, G. and Weiner, M.W. (1997) Presurgical multimodality neuroimaging in electroen-
260 cephalographic lateralized temporal lobe epilepsy. Ann. NeuroL, 42: 829-837. Laakso, M.P., Lehtovirta, M., Partanen, K., Riekkinen, P.J. and Soininen, H. (2000) Hippocampus in Alzheimer's disease: a 3-year follow-up MRI study. Biol. Psychiatry, 47: 557-561. Lemieux, L. (2001) Automatic, segmentation of grey and white matter and cerebrospinal fluid in serial Tl-weighted volume MRI data. Proceedings of SPIE Medical Imaging Conference, Vol. 4322, in press. Lemieux, L. and Barker, G.J. (1998) Measurement of small interscan fluctuations in voxel dimensions in magnetic resonance images using registration. Med. Phys., 25: 1049-1054. Lemieux, L., Wieshmann, U.C., Moran, N.E, Fish, D.R. and Shorvon, S.D. (1998) The detection and significance of subtle changes in mixed-signal brain lesions by serial MRI scan matching and spatial normalization. Med. Image AnaL, 2: 227-242. Lemieux, L., Liu, R.S.N. and Duncan, J.S. (2000) Hippocampal and cerebellar volumetry in serially acquired MRI volume scans. Magn. Reson. Imaging, 18: 1027-1033. Liu, R.S.N., Lemieux, L., Shorvon, S.D., Sisodiya, S.M. and Duncan, J.S. (2000) Association between brain size with abstinence from alcohol. Lancet, 355: 1969-1970. Liu, R.S.N., Lemieux, L., Bell, G.S., Bartlett, P.A., Sander, J.W.A.S., Sisodiya, S.M., Shorvon, S.D. and Duncan, J.S. (2001) A longitudinal quantitative MRI study of communitybased patients with chronic epilepsy and newly diagnosed seizures: methodology and preliminary findings. Neurolmage, 14: 231-243. Margerison, J.H. and Corsellis, J.A. (1966) Epilepsy and the temporal lobes. A clinical, electroencephalographic and neuropathological study of the brain in epilepsy, with particular reference to the temporal lobes. Brain, 89: 499-530. Molyneux, P.D., Kappos, L., Polman, C., Pozzilli, C., Barkhof, E, Filippi, M., Yousry, T., Hahn, D., Wagner, K., Ghazi, M., Beckmann, K., Dahlke, E, Losseff, N., Barker, G.J., Thompson, A.J. and Miller, D.H. (2000) The effect of interferon beta-lb treatment on MRI measures of cerebral atrophy in secondary progressive multiple sclerosis. European Study Group on Interferon beta-lb in secondary progressive multiple sclerosis. Brain, 123: 2256-2263. Nohria, V., Lee, N., Tien, R.D., Heinz, E.R., Smith, J.S., DeLong, G.R., Skeen, M.B., Resnick, T.J., Crain, B. and Lewis, D.V. (1994) Magnetic resonance imaging evidence of hippocampal sclerosis in progression: a case report. Epilepsia, 35: 1332-1336. O'Brien, T.J., So, E.L., Mullan, B.E, Hauser, M.E, Brinkmann, B.H., Jack Jr., C.R., Cascino, G.D., Meyer, EB. and Sharbrough, EW. (1999a) Subtraction SPECT co-registered to MRI improves postictal SPECT localization of seizure foci. Neurology, 52: 137-146. O'Brien, T.J., So, E.L., Meyer, EB., Parisi, J.E. and Jack, C.R. (1999b) Progressive hippocampal atrophy in chronic intractable temporal lobe epilepsy. Ann. Neurol., 45: 526-529. Radtke, R.A., Hanson, M.W., Hoffman, J.M., Crain, B.J., Walczak, T.S., Lewis, D.V., Beam, C., Coleman, R.E. and Friedman, A.H. (1993) Temporal lobe hypometaholism on PET:
predictor of seizure control after temporal lobectomy. Neurology, 43: 1088-1092. Richardson, M.E, Friston, K.J., Sisodiya, S.M., Koepp, M.J., Ashburner, J., Free, S.L., Brooks, D.J. and Duncan, J.S. (1997) Benzodiazepine receptors in malformations of cortical development: a voxel-based comparison of structural and functional data: in cortical grey matter. Brain, 120: 1961-1974. Rugg-Gunn, EJ., Eriksson, S.H., Symms, M.R., Barker, G.J. and Duncan, J.S. (2001) Diffusion tensor imaging of cryptogenic and acquired partial epilepsies. Brain, 124: 627-636. SalmenperK T., K~ilvi~iinen, R., Partanen, K. and Pitk~inen, A. (2001) Hippocampal and amygdaloid damage in partial epilepsy: a cross-sectional MRI study of 241 patients. Epilepsy Res., 46: 69-82. Sisodiya, S.M., Free, S.L., Stevens, J.M., Fish, D.R. and Shorvon, S.D. (1995) Widespread cerebral structural changes in patients with cortical dysgenesis and epilepsy. Brain, 118: 1039-1050. Sled, J.G., Zijdenbos, A.P. and Evans, A.C. (1998) A non parametric method for automatic correction of intensity non uniformity in MRI data. IEEE Trans. Med. Imaging, 17: 8797. Tasch, E., Cendes, E, Li, L.M., Dubeau, E, Andermann, E and Arnold, D.L. (1999) Neuroimaging evidence of progressive neuronal loss and dysfunction in temporal lobe epilepsy. Ann. NeuroL, 45: 568-576. Theodore, W.H., Bhatia, S., Hatta, J., Fazilat, S., DeCarli, C., Bookheimer, S.Y. and Galliard, W.D. (1999) Hippocampal atrophy, epilepsy duration, and febrile seizures in patients with partial seizures. Neurology, 52: 132-136. Trenerry, M.R., Jack Jr., C.R., Ivnik, R.J., Sharbrough, EW., Cascino, G.D., Hirschorn, K.A., Marsh, W.R., Kelly, EJ. and Meyer, EB. (1993) MRI hippocampal volumes and memory function before and after temporal lobectomy. Neurology, 43: 1800-1805. Van Elst, L.T., Woermann, EG., Lemieux, L., Thompson, P.J. and Trimble, M.R. (2000) Affective aggression in patients with temporal lobe epilepsy: a quantitative MRI study of the amygdala. Brain, 123: 234-243. Van Landingham, K.E., Heinz, E.R., Cavazos, J.E. and Lewis, D.V. (1998) Magnetic resonance imaging evidence of hippocampal injury after prolonged febrile convulsions. Ann. Neurol., 43: 413-426. Van Paesschen, W., Connelly, A. and Duncan, J.S. (1996) The amygdala and intractable temporal lobe epilepsy: a quantitative magnetic resonance imaging study. Neurology, 47: 10211031. Van Paesschen, W., Connelly, A., Jackson, G.D., King, M. and Duncan, J.S. (1997a) The spectrum of hippocampal sclerosis. A quantitative MRI study. Ann. Neurol., 41 : 41-51. Van Paesschen, W., Revesz, T., Duncan, J.S., King, M.D. and Connelly, A. (1997b) Quantitative neuropathology and quantitative MRI of hippocampus in temporal lobe epilepsy. Ann. Neurol., 42: 756-766. Van Paesschen, W., Duncan, J.S., Stevens, J.M. and Connelly, A. (1998) Longitudinal quantitative hippocampal magnetic reso-
261
nance imaging study of adults with newly diagnosed partial seizures: one-year follow-up results. Epilepsia, 39: 633-639. Varrone, A., Fujita, M., Verhoeff, N,P., Zoghbi, S.S., Baldwin, R.M., Rajeevan, N., Charney, D.S., Seibyl, J.P. and Innis, R.B. (2000) Test-retest reproducibility of extrastriatal dopamine D2 receptor imaging with [123I]epidepride SPECT in humans. J. Nucl. Med., 41: 1343-1351. Vilkman, H., Kajander, J., Nagren, K., Oikonen, V., Syvalahti, E. and Hietala, J. (2000) Measurement of extrastriatal D2-1ike receptor binding with [11C]FLB 457 - - a test-retest analysis. Eur. J. Nucl. Med., 27: 1666-1673. Wieshmann, U.C., Woermann, EG., Lemieux, L., Free, S.L., Bartlett, P.A., Smith, S.J., Duncan, J.S., Stevens, J.M. and Shorvon, S.D. (1997) Development of hippocampal atrophy: a serial magnetic resonance imaging study in a patient who developed epilepsy after generalized status epilepticus. Epilep-
sia, 38: 1238-1241. Woermann, EG., Barker, G.J., Birnie, K.D., Meencke, H.J. and Duncan, J.S. (1998) Regional changes in hippocampal T2relaxation time - - a quantitative magnetic resonance imaging study of hippocampal sclerosis. J. Neurol. Neurosurg. Psychiat., 65: 656-664. Woermann, F.G., Maclean, M.A., Bartlett, P.A., Parker, G., Barker, G.J. and Duncan, J.S. (1999a) Short echo time single voxel MRS of hippocampal sclerosis and temporal lobe epilepsy. Ann. Neurol., 45: 369-375. Woermann, F.G., Free, S.L., Koepp, M.J., Ashburner, J. and Duncan, J.S. (1999b) Voxel-by-voxel comparison of automatically segmented cerebral grey matter - - a rater-independent comparison of structural MRI in patients with epilepsy. Neurolrnage, 10: 373-384.
T. Sutula and A. Pitk~inen (Eds.) Progress in Brain Research, Vol. 135 © 2002 Published by Elsevier Science B.V.
CHAPTER 23
Do prolonged febrile seizures produce medial temporal sclerosis? Hypotheses, MRI evidence and unanswered questions Darrell V. Lewis 1,,, Daniel E Barboriak 2, James R. MacFall 2, James M. Provenzale 2, Teresa V. Mitchell 2 and Kevan E. VanLandingham 3 1 Department of Pediatrics (Neurology), 2 Department of Radiology and 3 Department of Medicine (Neurology), Duke University Medical Center, Durham, NC 27710, USA
Abstract: Whether or not severe febrile seizures in infancy cause hippocampal injury and subsequent medial temporal sclerosis is an often debated question in epilepsy. Recent magnetic resonance imaging (MRI) of infants suffering from febrile seizures has provided preliminary evidence that abnormally increased T2 signal intensity can be seen in the hippocampi of infants following prolonged and focal febrile seizures. Follow-up MRIs in a few of these infants have confirmed that medial temporal sclerosis can develop following these acute MRI signal changes. In this article, we review the hypotheses and MRI evidence relating to hippocampal injury during prolonged febrile seizures and the later development of medial temporal sclerosis.
Introduction
Medial temporal sclerosis (MTS) is the most frequent pathological substrate of temporal lobe epilepsy (TLE) and is characterized by neuronal loss and gliosis in the hippocampal formation. In various series, from 30% (Cendes et al., 1993a; Mathern et al., 1995a; Harvey et al., 1997) to as high as 70% (Davidson and Falconer, 1975; Williamson et al., 1993) of patients with TLE due to MTS have histories of prolonged febrile convulsions in early childhood. Therefore, it has been hypothesized that prolonged febrile seizures may acutely damage the temporal lobe leading to MTS (Cavanagh and Meyer, 1956; Falconer et al., 1964; Ounsted et al., 1966). *Correspondence to: D.V. Lewis, Department of Pediatrics (Neurology), Duke University Medical Center, Durham, NC 27710, USA. Tel.: +1-919-684-3219, ext. 2; Fax: +1 919 681-8943; E-mail:
[email protected]
This chapter will examine the evidence for the hypothesis that prolonged febrile seizures produce hippocampal injury and MTS. After first defining febrile seizure types and describing the link between febrile seizures and epilepsy, we will examine several hypotheses regarding the relationship between prolonged or focal febrile seizures, hippocampal injury and later temporal lobe epilepsy and the MRI evidence that bears on these hypotheses. Finally, questions regarding mechanisms and consequences of hippocampal injury will be discussed. Febrile convulsions
Febrile convulsions are defined as seizures occurring in childhood after age 1 month, associated with a febrile illness in the absence of an infection of the CNS or other acute cause and with no history of previous afebrile seizures (Commission on Epidemiology and Prognosis, International League against
264 Epilepsy, 1993). Simple febrile convulsions are less than 15 rain in duration and non-focal and probably do not produce any brain injury (Shinnar, 1999). It is probable that any association of simple febrile convulsions with later epilepsy is due to a genetically determined predilection for both the febrile seizures and epilepsy (Shinnar, 1999). This discussion will focus on febrile seizures that are prolonged and/or focal. Any febrile seizure that is focal, greater than 15 min duration or occurs more than once in 24 h is termed a complex febrile seizure. A febrile seizure or a series of febrile seizures lasting longer than 30 min without recovery in between constitutes febrile status epilepticus or FSE (Maytal and Shinnar, 1990). In this discussion all the febrile seizures we are concerned with will be complex and most will be in the category of FSE. Two to 5% of children will have one or more febrile convulsions by the time they reach 5 years, and 30% of these will experience at least one complex febrile seizure. Approximately 5% of febrile convulsions are sufficiently prolonged to be classified as FSE (Shinnar, 1999). Given these figures and a population at risk of approximately 19 million children under 5 years of age, one can estimate 25,00060,000 children affected by complex febrile seizures and 4,000-10,000 children affected by FSE per year in the United States alone.
The association of prolonged and focal febrile seizures and subsequent epilepsy The cumulative later incidence of partial complex seizure disorders in infants who have had complex febrile seizures correlates with the number and severity of the complex features characterizing the febrile seizures (i.e. focality, prolonged duration and multiple events in 24 h). Because the average latency between febrile seizures and onset of later epilepsy may be from 8 (French et al., 1993) to 11 (Mathern et al., 1995b) years, long follow-up is essential to determine cumulative risk of developing epilepsy. In infants who have had febrile seizures with a single complex feature, population studies have shown that the cumulative risk of later epilepsy with 7 years follow-up (Nelson and Ellenberg, 1976) was 4%, whereas with 25 years follow-up (Annegers et al., 1987) it was 8%. Annegers et al. (1987) found dura-
tion >30 min, focality and repetitive seizures within 24 h all clearly increased the risk of unprovoked complex partial seizures (CPSs) by age 25 years. Focal features produced the highest relative risk in both studies (Nelson and Ellenberg, 1976; Annegers et al., 1987). Annegers et al. (1987) found the probability of later CPSs due to each of these risk factors to be additive and 50% of children with all three risk factors had CPSs by age 25 years. However, Berg and Shinnar (1996b) did not find the risks due to each complex feature to be additive. Verity and Golding (1991) using a 10-year follow-up, found after focal febrile seizures, 5 of 17 (29%) infants developed CPSs and concluded that risk of epilepsy was highest in those with focal febrile seizures. In the same cohort, of 19 infants with FSE, 3 (16%) had later afebrile CPSs (Verity et al., 1993). In a follow-up study of 48 children with complex febrile seizures, Sapir et al. (2000) found 13 or 27% with epilepsy (mean follow-up period 43 months), and although insufficient numbers prevented determination of the relative significance of each complex feature, those with focal seizures showed a trend for a greater risk of epilepsy. Sixty-one percent of the infants who developed epilepsy had a partial seizure disorder. In the above studies, partial seizures and complex partial seizures were not classified as to the lobe of origin. In a study directed specifically at TLE, Maher and McLachlan (1995), studied families with apparent genetic predisposition to febrile seizures and found TLE, diagnosed by seizure history and electroencephalography, in 8 of 59 family members with febrile seizures and in only 1 of 213 members without febrile seizures. Average febrile seizure duration in those who developed TLE was 100 min and in those without TLE was 9 min. In summary, there is a high association of focal and prolonged febrile seizures and later partial seizure disorders. It is unclear if focality and duration are independent risk factors. The strong association of focal prolonged febrile seizures and complex partial seizure disorders suggests that both the prolonged febrile seizure and the later epilepsy arise from the same brain region, which was either abnormal from the start or suffered an acute epileptogenic insult during the febrile seizure.
265
Hypotheses proposed to explain the association of prolonged febrile seizures, MTS and TLE These data support an association between complex febrile seizures and later CPSs, or more accurately between FSE and CPSs given the recurring theme of febrile seizures more than 30 min in duration. Several hypotheses have suggested different causal relationships. Perhaps the oldest hypothesis is that FSE occurs in a previously normal brain producing acute hippocampal injury sufficient to produce MTS causing later TLE (Ounsted et al., 1966; Davidson and Falconer, 1975). This scenario requires no preexisting brain abnormality and posits that the acute injury during the FSE is sufficient to produce classical MTS causing TLE. However, Ounsted et al. (1966) suggested also that the families of infants with prolonged seizures have a genetic tendency for simple febrile seizures, but for some reason the probands with TLE had very prolonged events. However, it is not clear why a 'normal brain', or even a brain predisposed to simple febrile seizures, would respond to a febrile illness with focal and prolonged seizure activity rather than with a brief generalized febrile seizure. Alternatively, it has been suggested that MTS predates and causes the complex febrile seizures (Annegers et al., 1987; Cendes et al., 1993b; Davies et al., 1996; Bower et al., 2000). In this hypothesis, the complex febrile seizures would be the first clinical manifestation of the MTS which at a later age would also cause TLE. Implicit in this hypothesis would be a prior insult producing the MTS, such as prenatal or perinatal injury (Earle et al., 1953), meningitis, encephalitis, or head trauma. Since approximately 20% of children with FSE have pre-existing neurological abnormalities (Shinnar et al., 2001), it is likely that previous neurological insults have occurred in many of these children which would be in favor of a hypothesis positing a pre-existing lesion. With additional clinical and basic research, it is becoming apparent that the pathogenesis of MTS might be a multifactorial and multistage process. Thus, it may often require the synergistic effects of acquired insults and genetic predisposition together to result in MTS of sufficient epileptogenicity to lead to the clinical expression of TLE. The sequence and types of insults and the modulating genetic
traits culminating in MTS and TLE are multiple and therefore there are probably several different pathways to the final common pathology. It has been well documented that many patients with TLE and MTS have a variety of initial precipitating injuries, whereas some have no history of precipitating events (Mathern et al., 1995b). We propose that prolonged febrile seizures could initiate or augment the causal pathogenic sequence of MTS evolution in several different ways. First, pre-existing hippocampal or temporal lobe abnormalities could lower the seizure threshold of the limbic system so that a febrile illness could trigger limbic seizure activity clinically expressed as a prolonged focal febrile seizure. The pre-existing hippocampal abnormalities could be prenatal, such as localized dysgenesis or acquired such as infection, hypoxic-ischemic injury or trauma. The prolonged febrile seizure itself then produces additional seizure-induced injury to the hippocampus. This seizure-induced injury, if severe, could produce acute edema with subsequent loss of volume, gliosis and neuronal death characteristic of MTS (VanLandingham et al., 1998). Second, prolonged febrile seizures might produce more subtle hippocampal injury not detectable by conventional imaging techniques. Animal studies have shown that seizures induced by hyperthermia in infant rats may produce subtle changes resulting in hyperexcitability of hippocampal circuitry without cell death or morphological features of MTS (Dube et al., 2000). If this can occur in human infants as well during febrile seizures, later evolution to MTS could be triggered by chronic repetitive subclinical hippocampal seizure activity or by additional insults triggering further seizure activity, with resultant excitotoxic injury ultimately generating fully developed MTS. Finally, evidence is accumulating that common childhood viral illnesses that can trigger febrile seizures may also be accompanied by viral invasion of the CNS. For example, human herpes virus 6, an agent that causes roseola, is commonly found as a primary infection in infants with febrile seizures (Hall et al., 1994) and may also be found in the cerebrospinal fluid in that setting (Yoshikawa and Asano, 2000). More research will be needed to determine if limited focal inflammation due to CNS viral invasion
266 has any role in hippocampal injury during febrile seizures. Superimposed on all of these factors are genetic influences that may modulate seizure duration (Corey et al., 1998), influence local neuronal migration (Fernandez et al., 1998) and modulate glial and neuronal responses to excessive excitation or inflammation (Kanemoto et al., 2000). Experimental animal models of hyperthermiainduced seizures have already provided examples of some of the above mechanisms. Germano et al. (1996) produced cortical and hippocampal dysgenesis in rats by exposure in utero to an alkylating agent. The rat pups with dysgenesis were more susceptible to hypertherrnia-induced seizures and hyperthermiainduced neuronal dropout than the controls. In addition, the rats with dysgenesis were more susceptible to hippocampal kindled seizures and showed evidence of neuronal injury in the kindled hippocampi that was absent from the hippocampi of control kindled rats (Germano et al., 1998). Dube et al. (2000) have shown that 20-min-long hyperthermic seizures in infant rats can permanently lower limbic seizure thresholds without gross morphological change or neuronal death. However, these rats did not develop spontaneous limbic seizures suggesting that in this model additional limbic injury would be needed to develop epilepsy. It seems that hyperthermia-induced seizures in infant rats do not typically produce gross cell death even when prolonged. Sarkisian et al. (1999) used hyperthermia and continuous hippocampal electrical stimulation for 45 min and did not see gross cell death. These animal studies argue that other factors in synergism with the seizure activity may be involved in those instances where severe febrile seizures appear to be associated with gross hippocampal injury. In the following discussion, we will focus on the MRI studies of children that provide some insight into the role of prolonged febrile seizures in the pathogenesis of MTS. Although, there are also studies of adult TLE patients that bear on this question, these will not be discussed in this chapter.
MRI studies of the pathogenesis of MTS Based on MRI studies of children with TLE, complex febrile seizures and afebrile status epilepticus,
some tentative statements can be made that provide clues about the role of complex febrile seizures in the pathogenesis of MTS.
MRI evidence of acute hippocampal injury following complex febrile seizures correlates with duration and focality of seizures In agreement with long-term follow-up studies of children with febrile seizures, the imaging data available show that the duration and focality of febrile seizures correlate with hippocampal abnormality on postictal imaging. Szabo et al. (1999) measured hippocampal volumes in controls and in five infants with relatively brief complex febrile seizures no more than 20 min in duration and found no statistically significant hippocampal volume abnormalities. Nohria et al. (1994) performed MRI scans on an infant after one 45-50-min-long left-body seizure, again 6 weeks later after a second identical seizure and then 13 months later. The scans documented progressive loss of fight hippocampal volume reflected in the ratio of fight to left hippocampal volumes of 0.94, 0.87 and 0.72 on the first, second and third scans, respectively. Right MTS was clear by the last scan. VanLandingham et al. (1998) described 27 infants between 8 and 24 months of age who were imaged following complex febrile seizures and found definite MRI abnormalities in six of the 15 infants with focal or lateralized prolonged febrile seizures, and in none of the 12 infants with generalized prolonged febrile seizures. In two of the six infants with lateralized prolonged febrile seizures and abnormal MRIs, the MRIs showed pre-existing hippocampal atrophy consistent with the history of perinatal insults in these infants. However, the remaining four infants had severe acute changes in the affected hippocampi and these infants had suffered significantly (P < 0.05) longer (mean 99 min) seizures than both the remaining infants with lateralized seizures (mean 41 min) and those with generalized seizures (mean 46 min). The cohort of infants followed for complex febrile seizures at our institution has increased since the original study (VanLandingham et al., 1998) was published. We have recently re-evaluated a subgroup of these infants limited to those subjects who were:
267 (1) recruited prospectively, i.e. at the time of presentation; and (2) imaged within 72 h of the initial seizure. All imaging was performed using a 1.5T GE Signa unit and included conventional clinical protocols plus a fast-spin echo T2-weighted sequence (TR/TE/NEX, 4000/100/4) with contiguous 3-mm-thick slices through the hippocampi with the plane of the slices perpendicular to the long axis of the left hippocampus. Hippocampal images were visually reviewed by a neuroradiologist blinded to the clinical history who graded abnormalities using the following three measures: (1) anatomical extent of the abnormality (body, head or total extent of hippocampus); (2) intensity of T2 signal (T2Score) on a numerical scale from 0 (normal) to + 4 (markedly increased); and (3) the hippocampal volume (VolScore) from - 3 (markedly decreased) through 0 (normal) to +3 (markedly increased). The hippocampal abnormalities were then classified into four categories: (1) normal = T2Score and VolScore both are zero; (2) mildly abnormal = T2Score is zero and VolScore not zero; (3) moderately abnormal = T2Score is greater than zero and less than three; (4) severely abnormal = T2Score greater than or equal to three. Quantitative hippocampal volumes (HVs) were also measured by a blinded observer. Control HVs were obtained from MRIs done on children who had no developmental delay, seizures or brain structural abnormalities. Interobserver reliability of measures as well as intraobserver reliablity were verified using repeated measures in 10 subjects and ranged from 3 to 4%. Twenty-four of the 30 subjects presented with FSE, 11 generalized and 13 lateralized. Six of the 30 presented with complex febrile seizures, two generalized and four lateralized. Twenty-six of 30 had normal development by history and four were delayed upon presentation. There were 18 normal scans, 4 mildly abnormal, 4 moderately abnormal and 4 severely abnormal. Therefore, in total, 12 (40%) of the 30 subjects were judged to have hippocampal abnormalities by visual inspection. All abnormalities were unilateral and located in the hemisphere of seizure origin based on clinical localizing signs. Seizure duration in the four subjects with severely abnormal hippocampi averaged 103 min compared to 49 min for the other 26 subjects (P < 0.05). Abnormalities were more fre-
quent in lateralized FSE (8 of 13 subjects) than in generalized FSE (2 of 11 subjects) (P < 0.05). Two children had extrahippocampal abnormalities on the initial MRIs consisting of periventricular leukomalacia and these infants had histories of perinatal insults. Clinical parameters were correlated with the severity scores of the initial hippocampal MRI abnormalities (T2Scores and VolScores). Total seizure duration (r = 0.42, P = 0.02) and age at seizure (r = 0.53, P = 0.002) correlated positively with T2Scores using Spearman rank order correlation coefficients. There was a suggestive but nonsignificant inverse correlation between rectal temperature at presentation and T2Score. There was no correlation between T2Score and gender or number of seizures. VolScores, which are more difficult to determine visually, did not correlate as well with clinical parameters. Only temperature correlated inversely (r = -0.57, P = 0.008) with VolScore. No significant correlations between age or seizure duration and VolScore were found. We tentatively propose, therefore, that acute hippocampal injury manifest as T2 hyperintensity in postictal scans correlates with prolonged seizures and focal seizures. This is in agreement with retrospective and prospective cohort studies suggesting CPSs may follow febrile seizures that are prolonged and focal. Together, these data argue that prolonged and focal seizures can sometimes cause acute hippocampal injury leading to MTS and TLE. The definition of 'acute hippocampal injury' in these patients is admittedly tenuous. The initial MRI scans clearly demonstrate that hippocampal T2 signal can be increased shortly following a prolonged febrile seizure. In addition, hippocampal volume can be increased slightly, but this is less consistent. Several months later, follow-up scans definitely show the volume of an affected hippocampus may decrease markedly (see below) and the T2 signal may remain abnormal. The ensuing rapid loss of volume argues that some acute injury had occurred. However, one cannot determine whether or not hippocampal volume and signal were normal prior to the prolonged febrile seizure. It remains possible that previously sclerotic hippocampi could briefly swell after a seizure, masking pre-existing atrophy. Nevertheless, the most parsimonious interpretation of
268
Fig. 1. Fast spin echo oblique coronal sections through the head of the hippocampi of a subject with severe initial hippocampalT2 signal abnormality. (A) MRI done 2 days after the patient had a 72-rain-long complex febrile convulsion with focal left-sidedjerking. Note the increased size and signal in the right hippocampal head (under arrow, on the left side of the MRI). (B) Follow-up MRI done 9 months later showing the subsequent decrease in size of the right hippocampus (arrow) but remaining increased signal. the MRI data is that hippocampal edema and subsequent volume loss can occur after a prolonged febrile seizure suggesting an acute insult did occur, whether or not it was superimposed on a previously normal substrate.
Initial MRI abnormalities seen after severe febrile seizures can be followed by rapid development of MTS In the last decade, there have been scattered MRI reports of acute increases of T2 signal or volume in hippocampus shortly following status epilepticus (DeCarolis et al., 1991; Nohria et al., 1994; Tien and Felsberg, 1995; Chan et al., 1996; Stafstrom et al., 1996). However, resultant MTS was documented in only a few cases (Nohria et al., 1994; Tien and Felsberg, 1995; Stafstrom et al., 1996) whereas in others the MRI changes were reversible (Lee et al., 1992; Cox et al., 1995; Chan et al., 1996). In these reports hippocampal volume measurements were not uniformly done, follow-up intervals were variable, and often pre-existing epilepsy complicated the interpretation of the T2 signal changes as acute versus chronic. Therefore, it was unclear from these reports how often and at what rate MTS developed after an acute insult, such as a prolonged febrile seizure. VanLandingham et al. (1998), using hippocampal volumetry, performed follow-up MRIs in two of the four infants with acute hippocampal abnormalities on their initial images and found in both that marked hippocampal atrophy with persistent increased T2 had developed by 8-10 months after the prolonged febrile seizures. The patient of Nohria et al. (1994) developed MTS over 13 months and two patients of
Perez et al. (2000) developed bilateral MTS over 8 months and 4 months following afebrile and febrile prolonged focal seizures, respectively. Our reanalysis of 30 prospectively identified infants included 8 who have had follow-up MRIs done from 6 to 30 months after the initial scans. Four of these subjects had normal or mildly abnormal hippocampi in the initial MRIs and in the follow-up scans, their hippocampi showed normal and symmetrical rates of growth. Four other subjects had severely abnormal hippocampi in the initial MRIs, and in the follow-up scans 3 of these 4 showed asymmetric and abnormal growth rates with atrophy of the severely injured hippocampi and persistent increased signal compatible with MTS. Fig. 1 illustrates the MRI slices from the anterior hippocampi of one of the infants with severe initial hippocampal abnormality. The initial study was performed at 48 h (Fig. 1A) and the follow-up at 9 months (Fig. 1B) after the initial prolonged seizure. In Fig. IA, increased T2 intensity and swelling of the head of the fight hippocampus can be seen (on the left in the MRI section). In Fig. 1B, at 9 months after presentation, the T2 hyperintensity persists, but the size is now clearly less than the contralateral side. Fig. 2 illustrates serial hippocampal volume measurements on this infant. The solid center line is a best fit to a growth curve for 32 normal control fight hippocampal volumes and the dashed lines represent calculated 95% confidence limits for the normal control volumes. On the initial scan, the volume of the fight hippocampus was greater than that of the left hippocampus. By the time of the first follow-up, the fight hippocampus had lost volume and was now much smaller than the left. On subsequent follow-ups, both hip-
269 5500 5000 ¢~11= 4500 E
4000
3000
:F 250o
f
................
2000 15001_.,'1,",," 1000 500
,
- - - Right Hippocampus - o - Left Hippocampus n
0 1
I
I
I
I
I
I
I
I
I
2
3
4
5
6
7
8
9
10
Age Years Fig. 2. Sequential total hippocampal volumes of Subject 8 from 32 months, the time of the complex febrile convulsion, to 79 months taken from four sequential MRIs. Although the right hippocampus (RH) was initially larger than the left (LH), the relative volumes are reversed by the time of the first follow-up MRI. Note that there was subsequent growth of both hippocampi on follow-up. There was also a slight initial decline in the volume of the left hippocampus suggesting that although the predominant insult was on the right, there may have been some injury to the left as well. The solid line is a growth curve regression equation fitted to the hippocampi of 31 control infants with the dotted lines giving the 95% confidence intervals for the normal volumes.
pocampi showed growth at slightly higher rates than controls, but compared to the left hippocampus, the right remained clearly smaller and continued to have increased T2 signal compatible with MTS. Similar growth curve deviations have been seen in the follow-up scans of two of the other three infants with severe initial hippocampal abnormality, whereas the fourth infant showed normal symmetrical hippocampal growth on follow-up. These cases suggested that the severity of the initial hippocampal T2 signal abnormality in the MRI scans of infants with severe febrile seizures can predict the development of MTS on follow-up scans. To examine this relationship, we compared the magnitude of initial T2Scores with changes in hippocampal size on follow-up. The change in HVR (hippocampal volume ratio = right HV/left HV) between the initial scan and the first follow-up was calculated as: (Initial HVR) - (Follow-up HVR) and serves as a measure of the asymmetry in growth of the hippocampi after the FSE. The severity and lateralization of the initial T2 abnormality was expressed as the T2 asymmetry
= (Right T2Score) - (Left T2Score). Even with our small sample (n = 8), the correlation between the T2 asymmetry and the change in HVR was significant (r ----0.798, P = 0.017), indicating that hippocampi with high T2Scores were likely to become atrophic. Thus, marked and diffuse changes in hippocampal T2 signal following severe febrile seizures or FSE may be reliable predictor of subsequent evolution to MTS and the typical MRI appearance of MTS in these infants can develop within several months to a year following the insult. These examples probably give a lower limit to the time interval required and represent unusually severe insults. Apparently unilateral initial injury can lead to bilateral volume loss Close inspection of Fig. 2 shows that both hippocampi showed a decrease in volume on the first follow-up MRI in this infant. This pattern has been seen in other patients in our series who have developed MTS (unpublished data) and is consistent
270 with the observations that TLE patients frequently have bilateral hippocampal abnormalities (Margerison and Corsellis, 1966; Tasch et al., 1999) and this may be particularly true in patients with histories of complex febrile seizures (Barr et al., 1997). Thus even though the MRI abnormalities were initially thought to be unilateral in the patients reported by VanLandingham et al. (1998), the occurrence of bilateral asymmetric volume loss on follo