Trends in Drug Research III: Proceedings of the 13th Noordwijkerhout-Caminero Symposium, the Netherlands, 6-11 May 2001 (Pharmacochemistry Library)

P H A R M A C O C H E M l S T R Y L I B R A R Y - V O L U M E 32

TRENDS IN DRUG RESEARCH III Proceedings of the 13th Noordwijkerhout-Camerino Symposium

PHARMACOCHEMISTRY LIBRARY, edited by H.Timmerman Other titles in this series Volume 24 Perspectives in Receptor Research. Proceedings of the 10'hNoordwijkerhout-Camerino Symposium, Camerino (Italy), 10-14 September 1995 edited by" D. Giardinb, A. Piergentili and M. Pigini Volume 25 Approaches to Design and Synthesis of Antiparasitic Drugs edited by Nitya Anand Volume 26 Stable Isotopes in Pharmaceutical Research edited by'Thomas R. Browne Volume 27 Serotonin Receptors and their Ligands edited by B.Olivier et al. Volume 28 Proceedings XIVth International Symposium on Medicinal Chemistry edited by E Awouters Volume 29 Trends in Drug Research I1. Proceedings of the 11th Noordwijkerhout-Camerino Symposium, Noordwijkerhout (The Netherlands), 11-15 May,1997 edited by H. van der Goot Volume 30 The Histamine H3Receptor. ATarget for New Drugs edited by• R. Leurs and H.Timmerman Volume 31 Receptor Chemistry towards the Third Millennium. Proceedings of the 12'h Noordwijkerhout-Camerino Symposium, Camerino (Italy), 5-9 September 1999 edited by U. Gulini et al.

PHARMACOCHEMISTRY

LIBRARY

E d i t o r : H. T i m m e r m a n

V o l u m e 32

TRENDS IN DRUG RESEARCH III Proceedings of the ] 3'" Noordwijkerhout-Caminero Symposium, The Netherlands, 6-11 May 2001

Edited by: Henk v a n der G o o t Department of Pharmacochemistry, Free University Amsterdam, The Netherlands

2002 ELSEVIER Amsterdam - London - NewYork-

O x f o r d - Paris - S h a n n o n - T o k y o

ELSEVIER SCIENCE B.V. Sara Burgerhartstraat 25 RO. Box 211, 1000 AE Amsterdam,The Netherlands ©2002 Elsevier Science B.V. All rights reserved. This work is protected under copyright by Elsevier Science, and the following terms and conditions apply to its use: Photocopying Single photocopies of single chapters may be made for personal use as allowed by national copyright laws. Permission of the Publisher and payment of a fee is required for all other photocopying, including multiple or systematic copying, copying for advertising or promotional purposes, resale, and all forms of document delivery. Special rates are available for educational institutions that wish to make photocopies for non-profit educational classroom use. Permissions may be sought directly from Elsevier Science Global Rights Department, PO Box 800, Oxford OX5 1DX, UK; phone: (+44) 1865 843830, fax: (+44) 1865 853333, e-mail: [email protected]. You may also contact Global Rights directly through Elsevier's home page (http://www.elsevier.com), by selecting 'Obtaining Permissions'. In the USA, users may clear permissions and make payments through the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA; phone: (+1) (978) 7508400, fax: (+1) (978) 7504744, and in the UK through the Copyright Licensing Agency Rapid Clearance Service (CLARCS), 90Tottenham Court Road, London WlP 0LP, UK; phone: (+44) 207 631 5555; fax: (+44) 207 631 5500. Other countries may have a local reprographic rights agency for payments. Derivative Works Tables of contents may be reproduced for internal circulation, but permission of Elsevier Science is required for external resale or distribution of such material. Permission of the Publisher is required for all other derivative works, including compilations and translations. Electronic Storage or Usage Permission of the Publisher is required to store or use electronically any material contained in this work, including any chapter or part of a chapter. Except as outlined above, no part of this work may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without prior written permission of the Publisher. Address permissions requests to: Elsevier Science Global Rights Department, at the mail, fax and e-mail addresses noted above. Notice No responsibility is assumed by the Publisher for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions or ideas contained in the material herein. Because of rapid advances in the medical sciences, in particular, independent verification of diagnoses and drug dosages should be made. First edition 2002 Library of Congress Cataloging in Publication Data A catalog record from the Library of Congress has been applied for. ISBN:

0-444-50760-4

The paper used in this publication meets the requirements of ANSI/NISO Z39.48-1992 (Permanence of Paper). Printed in The Netherlands.

PHARMACOCHEMISTRY LIBRARY ADVISORY BOARD T. Fujita E. Muts©hler N.J. de Souza EJ. Zeelen

Department of Agricultural Chemistry, Kyoto University, Kyoto, Japan Department of Pharmacology, University of Frankfurt, Frankfurt, Germany Research Centre, Wockhardt Centre, Bombay, India Heesch,TheNetherlands

This Page Intentionally Left Blank

vii

CONTENTS Preface

.....................................................................................................

ix

A. Hogner, J.S. Kastrup, J. Greenwood, S.B. Vogensen, E.H. M¢ller, T.B. Stensb¢l, J. Egebjerg and P. Krogsgaard-Larsen Towards rational design of AMPA receptor ligands: an integrated medicinal, computational, biostructural and molecular pharmacological approach ..................... 1

C. C.A. van Boeckel From heparin to synthetic antithrombotics. The pentasaccharide story and follow-up .................................................................................................... 13

A. Bruggink A new future for synthesis? ............................................................................ 21

M. T. Reetz Directed evolution of enantioselective enzymes as catalysts in the production of chiral pharmaceuticals .................................................................................. 27

R. Schoevaart and A.P.G. Kieboom At the interface of organic synthesis and biosynthesis ......................................... 39

R.E. Hubbart What can structure tell us about function in the estrogen receptors? ...................... 53

M. Kouwijzer and J. Mestres Molecular docking and dynamics simulations in the ligand binding domain of steroid hormone receptors .............................................................................. 57

S. Kliewer Peroxisome proliferator-activated receptors and reverse endocrinology .................... 67

A.J.R. Heck and C.S. Maier Biomolecular mass spectrometry related to drug research ..................................... 81

P. Angeli and G. Gaviraghi Chemical and biological diversity in drug discovery ........................................... 95

J. Gomeza, 11/1. Yamada, A. Duttaroy, W. Zhang, R. Makita, T. Miyakawa, J. Crawley, L. Zhang, H. Shannon, F.P. Bymaster, C.C. Felder, C. Deng and J. Wess Muscarine acetylcholine receptor knockout mice: phenotypical analysis and clinical implications ....................................................................................

97

P.F. Zaratin, A. Quattrini, S. Previtali, G. Hervieu and M.A. Scheideler Changes in expression of the orphan G-protein coupled receptor GPR7 in human painful peripheral neuropathies .................................................................... 115

o o °

VIII

V.J. Gillet and P. Willett Computational methods for the analysis of molecular diversity ............................ 125

D. Langley Enhancing drug discovery by acquisition of chemical diversity ............................ 135

P. Seneci Chemical diversity as a driving force to design and put in practice synthetic strategies leading to combinatorial libraries for lead discovery and lead optimization ....................................................................................... , ..... 147

H. Just, E. Stefan, C. Czupalla, B. Niirnberg, Chr. Nanoff and M. Freissmuth Beyond G proteins: the role of accessory proteins in G protein-coupled receptor signalling ......................................................................................... . ....... 161

Ph.G. Strange Mechanisms of action of antipsychotic drugs: the role of inverse agonism at the D2 dopamine receptor ................................................................................. 175

K.E.O. /lkerman, J. Niisman, T. Holmqvist and J.P. Kukkonen Agonist channeling of o~2-adrenoceptor function ............................................... 181

D. Golemi, L. Maveyraud, J. Haddad, W. Lee, A. Ishiwata, K. Miyashita, L. Mourey, S. Vakulenko, L. Kotra, J. Samama, and S. Mobashery Antibacterials as wonder drugs and how their effectiveness is being compromised .... 193

E.P. Greenberg Pseudomonas aeruginosa quorum sensing: a target for antipathogenic drug d i s c o v e r y . ................................................................................................. 207

I. Chopra, L. Hesse and A. O'Neill Discovery and development of new anti-bacterial drugs ...................................... 213

B. B. Zhan g Discovery of small molecule insulin mimetics as potential novel antidiabetic agents .................................................................. , ..... , ............................. 227

L.M. Furness Expression databases for pharmaceutical lead optimisation ................................. 237

R. G. Pertwee New developments in the pharmacology of cannabinoids ................................... 249 Author index ............................................................................................. 259 Subject index ............................................................................................ 261

IX

PREFACE Trends in Drug Research followers or setters This volume of Pharmacochemistry Library comprises the text of invited lecturers presented at the Noordwijkerhout-Camerino Symposium Trends in Drug Research, held in Noordwijkerhout, The Netherlands, from 6-11 May 2001. During the 13^^ symposium in the series the question was asked whether medicinal chemists are following trends or perhaps trendsetters. The answer was clear: trendsetters. Through the years of the series - the first one dates back to 1974 - topics of the programme have been developing into almost routine aspects of medicinal chemistry; QSAR, modelling, receptor models. The 13th symposium fitted perfectly well in this tradition. On the programme were sessions on chemical and biological diversity, on new paradigms in drug action, on new insights in receptor mechanisms. A session which got much attention - and which brought new insights - was on green chemistry, the interface between organic synthesis and biosynthesis. A special symposium was devoted to the growing problem of resistant micro-organisms and the possibilities to identify new - and better - antibiotics. In a final session on very recent developments the new finding of small molecules with insulin sensitizing properties received much attention. Would an insulin-mimetic, a small molecule, be possible? The organizers of the Noordwijkerhout-Camerino Symposia express their sincere thanks to those who supported the 2001 symposium financially: Astra Zeneca, Byk Nederland, DSM, Glaxo Wellcome, Janssen Research Foundation, Lundbeck A/S, E. Merck, Organon Research, Pfizer (Parke Davis), UCB, Yamanouchi Europe. H.Timmerman, Chairman Organizing Committee

This Page Intentionally Left Blank

H. van der Goot (Editor) Trends in Drug Research III © 2002 Elsevier Science B.V. All rights reserved

Towards Rational Design of AMPA Receptor Ligands: An Integrated Medicinal, Computational, Biostructural and Molecular Pharmacological Approach Anders Hogner, Jette S. Kastrup, Jeremy Greenwood, Stine B. Vogensen, Eva H. M0ller, Tine B. Stensbol, Jan Egebjerg^ and Povl Krogsgaard-Larsen*

Department of Medicinal Chemistry, The Royal Danish School of Pharmacy, 2 Universitetsparken, DK-2100 Copenhagen, Denmark and ^Department of Molecular Genetics, H. Lundbeck A/S, 9 Ottiliavej, DK-2500 Valby-Copenhagen, Denmark

1. Medicinal Chemistry, a Science Undergoing Rapid Transformations

The field of medicinal chemistry is in a state of swift development and is at present undergoing major restructuring. The molecular biological revolution and the progressing mapping of the human genome have created a new biochemical and biostructural "world order". These developments have provided new challenges and opportunities for drug research in general and for drug design in particular. The major objectives of the medicinal chemists are transformation of pathobiochemical and - physiological data into a "chemical language" with the aim of designing molecules interacting specifically with the derailed or degenerating processes in the diseased organism. Potential therapeutic targets are being disclosed with increasing frequency, and this exponential growth will continue during the next decades. In this situation, there is a need for rapid and effective target validation and for accelerated lead discovery procedures. Consequently, most industrial medicinal chemistry laboratories have built up new technologies in order to meet these demands. Key words in this regard are construction of compound libraries, high or ultrahigh throughput screening, accelerated ADME and toxicity tests, and automatized cellular assay systems. In parallel with this development, biostructure-based drug design and intelligent molecular mimicry or bioisosterism are areas of growing importance in the medicinal chemistry "playing field". Structural biology is becoming an increasingly important part of molecular biology and biochemistry, and, furthermore, organic chemists are increasingly directing their attention towards synthetic aspects of biomolecules and biologically active compounds biosynthesized by plants and animals. Thus the borderland between biology, biochemistry, and chemistry is rapidly broadening and is becoming the most fruitful working field for innovative and intuitive drug design scientists.

2. Industrial Drug Discovery - Academic Drug Design

Where are the academic medicinal chemistry departments in this area of drug research, which is now undergoing profound changes, and which is moving towards an increasing degree of integration of scientific disciplines? Furthermore, how should medicinal chemistry teaching programmes be organized and taught in this highly dynamic research area? These burning questions need to be effectively addressed, and if the responsible academics fail to meet these challenges, academic medicinal chemistry will degenerate into traditional organic synthesis, from where it originates, or into trivial service functions in relation to industrial drug design and development programmes. The equipment for automatized combinatorial chemistry and for high throughput screening procedures, now in operation in most industrial medicinal chemistry departments, is expensive, and purchase of such technical facilities is normally far beyond the financial capacity of academic departments. Furthermore, in terms of operation, these automatized procedures are predominantly technical, and although students should understand the prospects and limitations of such technologies, these aspects can only be limited parts of student courses in medicinal chemistry. The scientific challenges of the conversion of solution synthetic chemistry procedures into solid-phase synthetic methodologies are mainly of basic chemical nature, and the development of cell-based assay systems is predominantly a biochemical pharmacological task. In order to attract the attention of intelligent students, the creative and fascinating nature of drug design must be the imderlying theme of basic and advanced student courses in medicinal chemistry. In relation to industrial screening programmes and "hit-finding" procedures, students should be taught that the conversions of "hits" into lead structures and further into drug candidates require advanced synthetic chemistry supported by computational chemistry. Furthermore, these medicinal chemistry approaches should be integrated with molecular pharmacology studies using cloned target receptors, ion channels, or enzymes, expressed in appropriate model systems. It is beyond doubt that a steadily increasing number of biomolecules will be subjected to X-ray crystallographic structural analysis. The number of enzymes with established three-dimensional structure is now increasing exponentially [1], and this growth will continue during the next decades. Even oligomeric membrane-boimd receptors can now be crystallized and subjected to X-ray crystallographic analysis [2], but such analyses of mono- or oligomeric receptors are still hampered by major experimental difficulties. In recent years, however, biostructural scientists have succeeded in crystallizing recombinant versions of the binding domains of a G protein-coupled receptor [3] as well as a ligand-gated ion charmel [4]. Structural analyses of these binding domains cocrystallized with agonist and antagonist ligands have already provided insight into the structural basis of receptor-ligand interactions and of receptor activation and blockade.

These breakthroughs in biostructural chemistry have opened up new avenues in drug design. Structural information derived from X-ray analyses of enzyme-inhibitor conglomerates has been and continues to be very valuable for the design of new types of inhibitors. Similar pieces of information derived from studies of receptor binding domains co-crystallized with different types of competitive or noncompetitive ligands undoubtedly will be of key importance in receptor ligand design projects. These approaches which are in the nature of drug design on a rational basis will become important parts of student teaching programmes in medicinal chemistry. In academic research and teaching, biologically active natural products probably will play a progressively important role as lead structures. Not only do such compounds often possess novel structural characteristics, but they alsofrequentlyexhibit unique biological mechanisms of action, although naturally occurring "toxins" typically show nonselective pharmacological effects. By systematic structural modification, including molecular mimicry approaches, it may be possible to "tame" such "toxins" and convert them into leads with specific actions on biofiinctions of key importance in diseases. Biologically active natural products undoubtedly will be continue to be important starting points for academic drug design projects, and such approaches will continue to be exciting case stories in student medicinal chemistry courses.

Figure 1. Leading academic medicinal chemistry departments or centres capable of establishing innovative collaborative projects with major industrial drug discovery units will optimally have the above "four-leaf clover" integrated composition of expertises.

In conclusion, there are growing indications that industrial and academic medicinal chemistry approaches will develop differently. Industrial medicinal chemistry projects will be in the nature of drug discovery with fast and effective hit-to-lead-to-clinical candidate development as key words. Innovative academic approaches are likely to focus on long-term development of rational approaches to drug design based on biostructural analyses and molecular mimicry. In both cases computational chemistry will be a key discipline, and molecular pharmacology certainly will be an essential and fully integrated discipline in medicinal chemistry (Figure 1). Hopefully, these two major lines of medicinal chemistry research will develop in a complementary fashion, which will open up the prospects of establishing fruitful collaborative projects between industrial and academic drug design scientists. A prerequisite for the build up of productive and innovative collaborative projects along these lines is the recognition of the collaborators as equal partners. Furthermore, mutual scientific appreciation will form the basis for the extremely important participation of industrial medicinal chemists in teaching courses and training programmes in drug design. In the following we describe the development of a long-term academic medicinal chemistry project on glutamate receptor ligands from a classical drug design project based on re-design of a naturally occurring amino acid "toxin", ibotenic acid, into an integrated rational approach involving medicinal chemistry. X-ray crystallographic protein structural analysis, computational chemistry, and molecular pharmacology. 3. AMPA Receptor Ligands: Therapeutic Prospects The central excitatory neurotransmitter effects of (5)-glutamic acid [(5)-Glu] are mediated by three heterogeneous classes of ionotropic receptors named 7V-methyl-Daspartic acid (NMDA), 2-amino-3-(3-hydroxy-5-methyl-4-isoxazolyl)propionic acid (AMPA), and kainic acid (KA) receptors [5,6] (Figure 2), and a number of subtypes of metabotropic receptors [6,7]. These or perhaps distinct subtypes of these receptors have been associated with certain neurologic and psychiatric diseases and are potential therapeutic targets in such diseases [5-7]. In recent years, much interest has been directed towards the role of AMPA receptors in the mechanisms associated with cognitive functions [8], and enhancement of AMPA receptor functions has been shown to facilitate learning and memory [6,9]. Although AMPA receptor agonists may not be used therapeutically due to potential neurotoxicity, these observations have focused interest on the molecular mechanisms of receptor activation and, thus, on the structural basis of AMPA receptor-agonist interactions.

NR1 NR2A NR2B NR2C NR2D NR3A GluR1 GluR4 GluR2 GluR3

{s

NMDA

AMPA

GluR7 -1 GluR6 GluR5 KA1 KA2 51 52

20

40

60 Percent identity

80

KA

-*

] ORPHAN

100

Figure 2. Multiplicity and phylogenetic comparison of the ionotropic glutamate receptor subunits.

The AMPA Receptor Recognition Site - Hypothesis of the Presence of a Lipophilic Binding Pocket Extensive structure-activity studies (SARs) on analogues of the classical AMPA receptor agonist, AMPA [10], in which a variety of alkyl [11], aryl [12], or heteroaryl groups [ISIS] have been substituted for the methyl group, have shed some light on the structural requirements for activation of AMPA receptors by this class of ligands. A series of AMPA analogues, in which the size of the alkyl substituents in the 5-position of the 3-isoxazolol ring has been systematically increased, have been synthesized and pharmacologically characterized. Whereas the ethyl analogue of AMPA, Et-AMPA, is slightly more potent than AMPA [16], analogues containing larger alkyl groups are much weaker or inactive, and the isopentyl analogue, Pe-AMPA, does not interact detectably with AMPA receptors [11] (Figure 3). Interestingly, all of the active members of this series of compounds show full agonist effects. There are no indications of partial agonist effects of any of these analogues.

(S)-2-Me-Tet-AMPA

P

1-Me-Tet-AMPA

Tet-AMPA

P

(S)-Thio-ATPA

Figure 3. Structures of a number of analogues of AMPA. A large number of 3-isoxazolol amino acids containing a heterocyclic unit in the 5position of the ring have been synthesized and characterized pharmacologically. SAR studies disclosed that only heterocyclic substituents containing heteroatom(s) in the 2position(s) of the ring showed potent AMPA agonist effects [13,14], as exemplified by (5)-2-Me-Tet-AMPA (Figure 3), which is the most potent AMPA agonist so far tested [13,17]. On the other hand, (/?^-2-Me-Tet-AMPA and the isomeric compound, 1-MeTet-AMPA, and also Tet-AMPA possessing an imsubstituted tetrazole ring, were essentially inactive [13]. Whereas the inactivity of Tet-AMPA probably reflects the fact that the tetrazole ring carries a negative charge at physiological pH, the methyl substituent in 1-Me-Tet-AMPA appears to sterically hinder binding to the AMPA receptor. These S ARs were interpreted in terms of the presence of a lipophilic pocket of limited volume at the binding site of the AMPA receptors [16,18]. Since the demethyl analogue of AMPA is markedly weaker than AMPA itself [11], the occupancy of this proposed pocket seems to be important for binding to and activation of the receptors. Although it still is unclear, why the presence of heteroatom(s) in the 2-position(s) of the heteroaromatic substituents of the bicyclic AMPA analogues is important for AMPA agonist activity, it is tentatively concluded that the presence of a heteroatom, and thus absence of a hydrogen atom, in this position facilitate hydrogen bond formation in the proposed lipophilic pocket.

5. From AMPA Receptor Agonists to Subtype-Selective Kainic Acid Receptor Agonists Substitution of a tert-butyl group for the methyl group of (5)-AMPA to give (5)-ATPA (Figure 3) has profound pharmacological consequences. Thus, whereas (5)-AMPA is a potent and highly selective agonist at AMPA receptors, (5)-ATPA is a very potent and selective agonist at kainic acid (KA) preferring receptors of the GluR5 subtype, showing only weak agonist effects at AMPA receptor subtypes [19]. Hence, the bulky tert-hutyl group of (5)-ATPA does not seem to be easily accommodated by the proposed lipophilic cavity of the AMPA receptors. On the other hand, the unique pharmacology of ATPA [20], and in particular its (5)-enantiomer [19], suggests that the spherical structure of the tert-hutyl group almost perfectly fits into a cavity at the binding site of the GluRS receptor. The structurally related 3-isothiazolol amino acid, (5)-Thio-ATPA (Figure 3), shows very similar pharmacological effects but is markedly more potent at GluRS receptors than (5)-ATPA [21]. This observation suggests that the increased lipophilicity of the "bottom part" of the molecule of (5)-Thio-ATPA contributes to its remarkably high agonist potency at GluR5. (iS)-ATPO is a structural hybrid between (5)-ATPA and the classical competitive NMDA antagonist 2-amino-7-phosphonoheptanoic acid (AP7) [22]. Quite surprisingly, (S)ATPO turned out to be a competitive AMPA receptor antagonist showing similar potency at subtypes of AMPA receptors but substantially weaker antagonist effects at the GluR5 subtype of KA receptors (Figure 2) [23]. These observations emphasize that the structural parameters of importance for the interaction of agonists and competitive antagonists with the binding domain of the GluR2 subtype of AMPA receptors are different [24].

6. X-Ray Crystallographic Studies of a Recombinant AMPA Receptor Binding Domain Co-Crystallized with (5)-Glutamic Acid or AMPA Receptor Agonists Until recently, no structural data of (5)-Glu receptors have been available. A breakthrough did, however, occur in 1998 with the publication of the first crystal structure of the ligand-binding domain of the AMPA receptor GluR2 subtype in complex with kainic acid [4]. This achievement was based on the finding that the transmembrane region separating the two parts of the receptor protein forming the ligand binding site could be replaced by a peptide linker. These fused protein units, known as S1S2, quite remarkably retain binding affinities similar to those of the wild-type membrane-bound receptor [4,25]. The development of large-scale expression and purification methodologies for the GluR2-SlS2 constructs have paved the way for a number of highresolution structures of GluR2-SlS2 in complex with AMPA agonists and an AMPA

antagonist and in the apo state of GluR2 [24]. This series of structures provides the basis for detailed structure-function analyses. In collaboration with the group of E. Gouaux, a number of AMPA agonists and antagonists, developed at this Department, have now been co-crystallized with the GluR2-SlS2 construct. The structure of S1S2 in complex with (S)-2-Me-Tet-AMPA is illustrated in Figure 4. (S)-2-Me-Tet-AMPA is bound in the cleft between the two domains. Domain 1 is composed of segment SI and the C-terminal end of segment S2. The C-terminal end of segment SI ends in domain 2, which primarily is composed of segment 2. An analysis of the ligand-protein complex indicates that the methyltetrazole unit almost perfectly fits into a binding pocket.

N-Terminus

C-Tenninus

Linker: G-T

Figure 4. Ribbon representation of the GluR2-SlS2 structure in complex with (5)-2-MeTet-AMPA. The SI and S2 segments are in black and grey, respectively. One of the challenges, now evolving, is to interpret the structural data with the goal of designing new ligands on a rational basis. So far, the binding mode of (S)-Glu bioisosteres, such as (5)-AMPA (Figure 3), has been unclear. From the published crystal structure complexes of (S)-Glu and (S)-AMPA it is evident that the a-carboxylate and the a-ammonium group in both complexes bind to SIS2 in a consistent manner [24].

These groups form strong interactions via specific hydrogen bonds and ion-pair interactions with amino acid residues from both domain 1 and 2 (Figure 5).

Figure 5. The binding site of GluR2-SlS2 in complex with (*S)-Glu. (5)-Glu is in dark grey, and the backbones of domain 1 and of domain 2 are in black and light grey, respectively. Oxygen and nitrogen atoms are displayed as black spheres and carbon atoms as white spheres. Dashed lines indicate potential hydrogen bonds within 3.2A and water molecules are shown as dark grey spheres. Pdb id-code IFTJ (Armstrong and Gouaux [24]).

A superposition of (5)-Glu and (5)-AMPA structures shows that the fundamental interactions between S1S2 and the amino acid group are conserved (Figure 6). The difference between these two agonists lies in the different positioning of the distal carboxylate group of (iS)-Glu and the 3-isoxazolol anion of (5)-AMPA. It turns out that (5)-AMPA does not bind as a true structural bioisostere of (5)-Glu. Instead, a water molecule in the (iS)-AMPA structure occupies a similar site to one of the oxygens of the distal carboxylate group of (iS)-Glu. This piece of information emphasizes the role of tightly bound water molecules within the binding site. The binding of the 3-isoxazolol

10 anion of (iS)-AMPA is further stabilized by strong hydrogen bonds both directly to the protein but also through water-mediated hydrogen bonds. The methyl group of (5)AMPA partially fills a hydrophobic pocket in domain 1. By comparing the interactions of the 5-position substituents of (iS)-Me-Tet-AMPA and (5)-AMP A with the binding site, it becomes evident that the ethyl group of Et-AMPA (Figure 3) fits better into the hydrophobic cavity than the methyl group of (5)-AMPA. This observation may explain why Et-AMPA is slightly more potent than AMP A as an AMP A receptor agonist [16].

Figure 6. The binding site after superposition of the structures of GluR2-SlS2 in complex with (5)-Glu and (S)-AMPA. (5)-Glu and (5)-AMPA are in dark grey and white, respectively. Otherwise as in Figure 5. Pdb id-codes IFTJ and IFTM (Armstrong and Gouaux [24]).

The X-ray crystallographic data derived from structural analyses of complexes between the recombinant GluR2-SlS2 binding domain and different AMPA receptor ligands are now being exploited in terms of rational design of new types of AMPA receptor ligands.

11

7. References [I] D. Leung, G. Abbenante and D.P. Fairlie, J.Med.Chem. 43 (2000) 305. [2] K. Brejc, W J. van Dijk, R.V. Klaassen, M. Schuurmans, J. van der Oost, A.B. Smit and T.K. Sixma, Nature 411 (2001) 269. [3] N. Kunishima, Y. Shimada, Y. Tsuji, T. Sato, M. Yamamoto, T. Kumasaka, S. Nakanishi, H. Jingami and K. Morikawa, Nature 407 (2000) 971. [4] N. Armstrong, Y. Sun, G.-Q. Chen and E. Gouaux, Nature 395 (1998) 913. [5] D.T. Monaghan and R.J. Wenthold, Eds., The lonotropic Glutamate Receptors, Humana Press, Totowa, New Jersey, 1997. [6] H. Brauner-Osbome, J. Egebjerg, E.0. Nielsen, U. Madsen and P. KrogsgaardLarsen, J.Med.Chem. 43 (2000) 2609. [7] P.J. Conn and J. Patel, Eds., The Metabotropic Glutamate Receptors, Humana Press, Totowa, New Jersey, 1994. [8] H.K. Lee, M. Barbarosie, K. Kameyama, M.F. Bear and R.L. Huganir, Nature 405 (2000) 955. [9] U. Staubli, G. Rogers and G. Lynch, Proc.Natl.Acad.Sci. U.S.A. 91 (1994) 777. [10] P. Krogsgaard-Larsen, T. Honore, J.J. Hansen, D.R. Curtis and D. Lodge, Nature 284(1980)64. [II] F.A. Sl0k, B. Ebert, Y. Lang, P. Krogsgaard-Larsen, S.M. Lenz and U. Madsen, Eur.J.Med.Chem. 32 (1997) 329. [12] B. Ebert, S.M. Lenz, L. Brehm, P. Bregnedal, J.J. Hansen, K. Frederiksen, K.P. Bogeso and P. Krogsgaard-Larsen, J.Med.Chem. 37 (1994) 878. [13] B. Bang-Andersen, S.M. Lenz, N. Skjaerbaek, K.K. Soby, H.O. Hansen, B. Ebert, K.P. Bogeso and P. Krogsgaard-Larsen, J.Med.Chem. 40 (1997) 2831. [14] E. Falch, L. Brehm, L Mikkelsen, T.N. Johansen, N. Skjaerbaek, B. Nielsen, T.B. Stensbol, B. Ebert and P. Krogsgaard-Larsen, J.Med.Chem. 41 (1998) 2513. [15] B. Bang-Andersen, H. Ahmadian, S.M. Lenz, T.B. Stensbol, U. Madsen, K.P. Bogeso and P. Krogsgaard-Larsen, J.Med.Chem. 43 (2000) 4910.

12 [16] U. Madsen, B. Frolund, T.M. Lund, B. Ebert and P. Krogsgaard-Larsen, Eur.J.Med.Chem. 28 (1993) 791. [17] S.B. Vogensen, H.S. Jensen, T.B. Stensb0l, K. Frydenvang, B. Bang-Andersen, T.N. Johansen, J. Egebjerg and P. Krogsgaard-Larsen, Chirality 12 (2000) 705. [18] L. Brehm, F.S. Jorgensen, J.J. Hansen and P. Krogsgaard-Larsen, Drug News Perspect. 1 (1998) 138. [19] T.B. Stensbol, L. Borre, T.N. Johansen, J. Egebjerg, U. Madsen, B. Ebert and P. Krogsgaard-Larsen, Eur. J. Pharmacol. 380 (1999) 153. [20] V.R.J. Clarke, B.A. Ballyk, K.H. Hoo, A. Mandelzys, A. Pellizzari, C.P. Bath, J. Thomas, E.F. Sharpe, C.H. Davies, P.L. Omstein, D.D. Schoepp, R.K. Kamboj, G.L. CoUingridge, D. Lodge and D. Bleakman, Nature 389 (1997) 599. [21] T.B. Stensbol, H.S. Jensen, B. Nielsen, T.N. Johansen, J. Egebjerg, K. Frydenvang and P. Krogsgaard-Larsen, Eur.J.Pharmacol. 411 (2001) 245. [22] G.L. CoUingridge and J.C. Watkins, Eds., The NMDA Receptor, Oxford University Press, Oxford, 1994. [23] E.H. MoUer, J. Egebjerg, L. Brehm, T.B. Stensbol, T.N. Johansen, U. Madsen and P. Krogsgaard-Larsen, Chirality 11 (1999) 752. [24] N. Armstrong and E. Gouaux, Neuron 28 (2000) 165. [25] A. Kuusinen, M. Arvola and K. Keinanen, EMBO J. 14 (1995) 6327.

H. van der Goot (Editor) Trends in Drag Research III © 2002 Elsevier Science B.V. All rights reserved

13

From Heparin to Synthetic Antithrombotics The pentasaccharide story and follow-up

C.A.A. van Boeckel N. V. Organon, Lead Discovery Unit, P.O. Box 20, 5340 BH Oss, The Netherlands

Since 1936, heparin has been used in clinics for the prevention and treatment of thrombosis. Its main antithrombotic activity is explained by its ability to potentiate the activity of the serine protease inhibitor antithrombin III (AT-III), which inactivates a number of serine proteases- such as thrombin and factor Xa- in the coagulation cascaded By the end of the 1970's heparin fragments (obtained by chemical or enzymatic degradation) had been isolated by affinity chromatography on immobilised AT-III and the high affinity fractions had been analysed. From these studies it was deduced^ in 1981 that a unique pentasaccharide (PS) fragment, that occurs in about one-third of the heparin polysaccharide chains, constitutes the minimal binding domain for AT-III. The pentasaccharide fragment (also known as the DEFGH part of heparin) was synthesised^' "* a couple of years later to confirm the earlier proposal.

Structure - Activity Org31540/SR90107 (Arixtra®)

lessential sulphates / carboxylates contributing sulphates

Figure 1 The slightly modified synthetic pentasaccharide fragment 1 ( ORG 31540/SR90107 in Figure 1) was found to elicit a very selective antithrombotic mode of action, in that it only accelerates the AT-III mediated inhibition of coagulation factor Xa but not that of thrombin^ (see Figure 2). The results of four Phase III clinical trials show that the pentasaccharide 1 provides a superior benefit over a low molecular weight heparin in

14 preventing deep-vein thrombosis (DVT) in major orthopedic surgery patients, with an overall relative risk reduction of 50% and a similar safety profile. In August 2001 Organon and Sanofi-Synthelabo have received an "approvable letter" from the U.S. Food and Drug Administration (FDA) for the registration of 1 as a new antithrombotic drug called Arixtra®

Pointa^ffiettaricte shows selective Anti Xa Activity

+ Arg +

+ +

+ Lys

Arg

factor Xa

ATin

Figure 2 The specificity of the interaction of the sulphated pentasaccharide with the protein was confirmed when heparin pentasaccharide analogues were synthesised and tested^ for inhibition of blood coagulation factor Xa. First it was established which of the charged groups play an important role in the activation of AT-III. It was found that some groups are strictly required for the activation of AT-III while other groups contribute significantly during the AT-III activation (see Figure 1). Taking into account these structure-activity relationships and by contemplating molecular modelling data we postulated a simplified AT-III/PS interaction model. On the basis of this model we introduced^ an extra sulphate group at position 3 of unit H of the naturally occurring fragment to give analogue 2 (see Figure 3). This extra-sulphated analogue displays higher affinity towards AT-III and an enhanced AT-III mediated anti-Xa activity (1250 U/mg for 2 vs. 700 U/mg for 1).

15

Tirsf ATIII/PSInl^r^ttonMQd^

OSO,

OH

V V ^ ^

AT n i BINDING SITE 1

Figure 3 Subsequently, attention was turned to a new simplified series^ in which all hydroxyl groups are methylated and in which all the N-sulphate groups are replaced by Osulphate groups . Quite to our surprise these modifications did not affect the biological activity of the PS. It should be stressed that the synthesis of these methylated analogues is much easier than that of heparin-like fragments. In this series we prepared several analogues methylated at the 2-0 and 3-0 positions of both uronic acid moieties. At first sight it was expected that such analogues would loose at least half of their biological activity as was observed for the "natural" counterparts lacking the 2-0-sulphate group of iduronic acid. However, quite unexpectedly, one of these methylated analogues, (i.e. compound 3; SanOrg 34006; see Figure 4) turned out to be highly potent^, displaying 1600 anti-Xa U/mg.

San Org 34006

aXa 700 U/mg 1600 U/mg

Ko(ATin)

tVj rat

600 nM 20 nM

0.7 hr H.Ohr

Figure 4

16 The potent compound 3 not only binds much stronger to AT-III (Kd=20nM), relative to the PS (compound 1, Kd = 700 nM), but also its elimination half-life is about fifteen fold longer. SanOrg 34006 is clinically investigated for the prevention of thrombosis using a once a week dose regimen. For many PS analogues it was found^ that the elimination half-life is proportional to the affinity of AT-III. The next challenge was to extend the concept of AT-III mediated inhibition of factor Xa by pentasaccharides towards synthetically feasible derivatives displaying both antifactor Xa and anti-thrombin activity. It is known that for AT-III mediated inhibition of thrombin a heparin fragment comprising at least 16 saccharide units is required to facilitate the binding of AT-III and thrombin to the same polysaccharide chain (the so called "bridge" or "template" mechanism).

Doftl^n of Synttette CoiftiyQattisft heparin v

\

ABD

*»»+

+

y thrombin -f

+

AT III synthetic conjugates

TBD

1 v_

ABD

t

neutra^spaceT

TBD

D-unit

Figure 5 Our model^^ of the ternary complex (see Figure 5 for a schematic representation) revealed that heparin analogues may be obtained when a thrombin binding oligosaccharide is tethered to the non-reducing terminus of the AT-III binding pentasaccharide with a neutral spacer of about 50 atoms in length. To this end glycoconjugates (e.g. compound 4 in Figure 6) were synthesised which comprise a PS as AT-III binding domain (ABD), a linear spacer and a persulphated oligosaccharide as thrombin binding domain (TBD).

17

ABD

spacer

OSOs'

TBD -OSCb'

ABD 0CH3

°f^'');^">r°'t^v^'-

TBD

anti Xa = 740 U/mg antilla= 140 U/mg

Figure 6 Compound 4 was one of the first conjugates that has been synthesised and which indeed displayed good to strong AT-III mediated anti-thrombin activity (4 = 140 U/mg; heparin =160 U/mg) besides the expected anti-factor Xa activity (740 U/mg). The potency, the anti-factor Xa/anti-thrombin ratio and half-life in circulation of this new type of heparin like molecules can be adjusted^ ^'^^ in a rational way by varying the AT-III affinity of the PS (ABD), the TBD (charge density) and the spacer (length and rigidity).

18 Design of Dual Inhibitor

ATIII-mediated anti-Xa: 885 U mg-^ Direct anti-IIa activity:

ECggsCas ^.M

Figure 7 Furthermore, a different class of conjugates was designed and prepared ^^ in which a pentasaccharide is covalently linked to a direct thrombin inhibitor (e.g. NAPAP) displaying a dual mode of action (AT-III activation and direct thrombin inhibition). Such dual inhibitors (e.g compound 5 in Figure 8) have also the advantage that the PS component (bringing about AT-III mediated anti-Xa activity) and the direct thrombin inhibitor have the same half-life in circulation. In addition such dual inhibitors are expected to neutralise clot-bound thrombin more efficiently than the heparin /AT-III complex, which because of its size is hampered to penetrate the blood clot. Several of the conjugates/ dual-inhibitors are now in (pre)-clinical development.

Nmtn Dmal Antiilhiiroinfitotiic OSO 3"

0

OSO 3-

^ OSO 3"

OSO 3-

OSO 3-

^ OSO 3-

OSO 3-

H2N

NAPAP-PS Coniugate

Figure 8

OSO 3-

19 In conclusion, the discovery of the AT-III binding pentasaccharide domain in heparin opened an avenue to various new synthetic antithrombotics showing tailor-made profiles both with respect to anti-factor Xa and anti-thrombin activity (either via AT-III or a dual mode of action) and duration of action.

REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13.

Heparin (Eds.rD.A. Lane, U. Lindahl), Edward Arnold, London (1989). J. Choay et al., N.Y. Acad, Sci. 370, 644 (1981). P. Sinay et al., Carbohydr.Res. 132, C5 (1984). C.A.A. van Boeckel et al., J.Carbohydr.Chem. 4, 293 (1985). J.M. Herbert et al.. Cardiovascular Drug Reviews 15, 1 (1997). C.A.A. van Boeckel, M. Petitou, Angew.Chem.Int.Ed.Engl. 32, 1671 (1993). C.A.A. van Boeckel, J.E.M. Basten, H. Lucas, S.F. van Aelst, Angew. Chem. Int. Ed. Engl. 27, 1177(1988). G.Jaurand, J. Basten, I. Lederman, C.A.A. van Boeckel, M. Petitou, Bioorg & Med. Chem. Lett. 2, 897 (1992). P. Westerduin et al., Bioorg. & Med. Chem. 2, 1267 (1994). P.D.J. Grootenhuis et al.. Nature Struct. Biol., 2 736 (1995). J.E.M. Basten, CM. Dreef-Tromp, B. de Wijs and C.A.A. van Boeckel, Bioorg. & Med. Chem. Lett. 8, 1201 (1998). CM. Dreef-Tromp et al., Bioorg. & Med. Chem. Lett. 8, 2081 (1998). R.C Buijsman et al. Bioorg. & Med. Chem. Lett. 9, 2013(1999).

Part of this work was done in collaboration with Sanofi Recherche.

This Page Intentionally Left Blank

H. van der Goot (Editor) Trends in Drug Research III © 2002 Elsevier Science B.V. All rights reserved

21

A new Future for Synthesis? Alle Bruggink a) Department of Organic Chemistry, University of Nijmegen, Toernooiveld 1, 6525 ED Nijmegen b) DSM Research, P.O. Box 18, 6160 MD Geleen, The Netherlands

1. Summary New scientific challenges with great implications for synthetic chemistry are developing rapidly, in particular in Life Sciences, Performance Materials and Nanotechnology. Without a drastic and radical change in approach synthetic chemistry can not be expected to deliver the required contributions, whereas the timely availability of the desired molecules, employing sustainable processes in their manufacturing, has to be at the basis of these new developments.

2. Introduction At the start of the chemical and pharmaceutical industry stoichiometric synthetic chemistry was sufficient to obtain the required molecules. With the increasing scale and volume, in particular in the petrochemicals industry, chemocatalysis was required to reach efficient and economical processes. The increasing complexity and functionality of the desired molecules, in particular in the pharmaceutical industry, could be met through the introduction of biocatalysis allowing mild reaction conditions and subtle processes. However, biocatalysis is often combined with traditional, stoichiometric chemistry to reach the desired synthesis goals. Translation of chemocatalysis from petrochemicals to a broader application in chemical and pharmaceutical industry is still rather remote from maturity and, more importantly, will not be sufficient to meet the present challenges. Moreover, all our synthetic methodology is still characterized by a "step-by-step" approach; i.e. chemical bonds are manipulated one by one, requiring several protecting groups for the remaining functionalities and various activation mechanisms for the desired transformations. Prolonged development times, increased waste streams and laborious recycle loops are important drawbacks in these processes, hampering a sustainable chemical industry for the future.

3. New Challenges With the introduction of combinatorial chemistry, high-throughput-screenings and robotics the pharmaceutical industry has greatly expanded the possibilities for generating lead compounds. The molecular complexity of these products is increasing

22 as well, stimulated by the growing impact of molecular biology and the results of genomics and proteomics research programs. The (bio-)chemical industry is highly challenged in quickly synthesizing the required molecules for further pharmaceutical development. Several small scale contract research companies and kg-shops have emerged in the past few years to meet these needs. Due to the pressure on "time to market" their synthetic methodology is very often based on the versatile diversity of traditional stoichiometric chemistry. This poses a threat to future development of sustainable, catalytic processes for industrial application as the initial syntheses are often also the start of product and process registration in drug master files. In fact, improvements in processes for existing drugs through the introduction of (bio-)catalysis might be found difficult to translate to application in processes for new drugs. Thus, there is a great challenge in applying catalytic methods right from the start in the synthesis of lead compounds. Moreover, in order to meet the demands for short development times catalyst screening and development has to be done in highthroughput-systems and process steps should be highly integrated without isolation or purification of intermediates. However, this dream of "one-time-right", i.e. direct conversion of simple starting materials in a cascade of (catalytic) reactions without activating and protecting agents into (complex) end products, is still very far away in today's organic synthesis. In fact, upcoming demands from important industrial segments such as Life Sciences, Performance Materials and Nanotechnology, are bringing us a molecular complexity that no longer can be met with the presently available tools for synthesis.

4. Learningfi-omNature In nature synthesis in microorganisms occurs in a chain of reactions each catalyzed by combined and simultaneous enzyme actions. In this cascade of reactions, concentrations of starting materials, intermediates and end products are kept very low allowing maximum selectivity and no side reactions or byproducts being formed. Starting materials are brought in in a fed-batch mode using controlled membrane transport proteins. End products are continuously removed according to the in-situ-productremoval (ISPR) principle, employing again controlled transport systems. This allows, in principle, high throughputs and efficiency. Required protecting and activating agents are kept in-situ through recyclable energy and redox carriers (ATP, NADP, etc.). The desire to utilize these features of biosynthesis in nature for a wider scope of molecules might be a historic one, but now time has come to allow a meaningful and responsible scientific program towards these ends. The first steps have been made, i.e. metabolic engineering for protein synthesis can now be done. For all other molecules the molecular biology has just started to be developed. From an industrial point of view biosynthesis is mainly limited to biocatalysis (employing the 1 enzyme/1 step approach) and fermentation processes, which are mainly empirically based (see also fig. 1). These two extremes in utilizing enzymes for syntheses should be brought together.

23 Nr. of catalyst Many

More

Full Fermentations Cascade Catalysis (in concert) Bio-transPrecursor formations Fermentations Bio-Redox

Present Catalysis

Cascade Catalysis (1 by 1)

Sto'i'chiom. Chemistry

One-pot Reactions More

Many

Nr. of bonds manipulated simultaneously

Figure 1 Synthetic Methodology in Industry Fundamental understanding of these processes is developing rapidly through metabolic flux analyses, metabolic pathway engineering and related developments in molecular biology (genomics, biomolecular informatics). In particular the intricate molecular recognition mechanisms in biosystems are slowly being revealed. Application of this knowledge in improved biocatalysis is underway. Utilization of this know-how for conversions of relative simple but often unreactive molecules is in its infancy. Studies on methane-mono-oxygenase are an outstanding example. In the end all kind of inter-atomic bonds will be manageable in an efficient and economic way employing cascades of bio- or bio-inspired-catalysts. The diversity of chemistry and the complexity of biology are brought together in a new fiiture for synthesis (see also Fig. 2).

5. Towards the Best of Both Worlds. Molecular Biology & Biosynthesis Fundamental knowledge about biological systems has increased enormously in the last 25 years. At molecular level the actions of enzymes and metabolic pathways can now be identified. To translate this know-how into new biosyntheses and biocatalysts, additional insight is needed in the operation of complete micro-organisms, individual cells, cell compartments, enzyme interactions and metabolic fluxes. Challenges are in particular the role of compartments in cells and further study of the mechanisms of the several process couplings. The control of enzyme levels and activation between genome and proteome and the associated molecular recognition systems need to be understood in detail. In order to widen the scope of fermentation processes interactions between

24 primary and secondary metabolisms should be elucidated. Transport-mechanisms into, within and out of a cell or cell compartment should be known as well for natural products as for (new or known) non-natural molecules. In particular transport-proteins and mechanisms involved in product removal should be understood in much greater detail. New or revised systems might be needed to allow high throughputs. Genetically engineered biosystems should be made available to improve biosyntheses of both enzymes for biocatalyst development and metabolic end products (natural and nonnatural). In particular fermentation of non-natural molecules will need full deployment of all available tools in molecular biology. Further exploitation of directed evolution methods will greatly enhance this (including developments in genomics, biomolecular informatics etc.) Bio-transformations Although the present approach of using one enzyme for a particular conversion can still be widened in scope the challenge should be at employing combinations of enzymes. Cascades of enzymatic reactions are already emerging but can be researched much wider. Biocatalysts consisting of a number of enzymes acting in concert and interdependent such as in redox reactions should be made available. More simple and cheaper co-factor regenerating systems should be developed and build into a biocatalyst. The present focus on single enzyme biocatalysts should further be shifted to enzyme combinations, cell compartments or even complete cells as catalyst whereby several enzymes are indeed employed. Nowadays whole cell systems are already used as biocatalyst but mainly employing only one of the enzymes present. Often these enzymes cannot be isolated or are deactivated upon isolation. Further challenges are new techniques for formulation of enzymes and enzyme systems into stable, robust and efficient biocatalysts.

Q.

E o o O) O

Joint Future

Biosyntheses Direct Fermentation Precursor Fermentation Bio-oxidation BiotrMigformations fwcatalysis Ctiemocatalysis Stoichiometric synthesis

Chemistry (after J.IVI. Lehn) Figure 2 SynthesisfromChemical and Biological Perspective

Diversity

25

Single biocatalytic conversions and bio-transformations using enzyme combinations should become a synthetic continuum with precursor fermentations and direct fermentations. Insight in interaction of enzymes with their environment (i.e. membranes) will hereby be needed. Methods to tune enzyme kinetics will be required. Artificial co-factor regenerating systems will have to be developed as well as new ISPR methods. Bio-inspired Organic Synthesis Many organic syntheses are already inspired by nature. However, the complexity of biology has forced organic chemistry to very inaccurate translations of enzyme systems into man-made catalysts (i.e. catalysts with a simple molecular structure and mol. weight below 500 vs. enzymes with highly intricate structures and mol. weights up to 500.000). The present bio-inspired trend in organic synthesis towards macromolecular systems is meeting the advancement in molecular biology at the same level. This should lead to joint design of new multifunctional (bio-)catalyst systems, which can either be used in a modified metabolic path for a fermentation process or as an efficient catalyst in a series of organic syntheses. Bio-mimetic catalyst systems will be made available designed on growing knowledge of metabolic pathways and detailed insight in bio-recognition phenomena. Biomolecular informatics will provide guidelines for the design of new and robust catalysts. When combined with mechanistic know-how of chemocatalysis 'de novo' enzyme design comes within the realm of current chemistry. Combination with directed evolution methods would be another way to new catalyst design. High selectivity for a specific target molecule can be reached. Better bio-, chemo- and hybride-catalyst formulations for a wide range of (bio-) syntheses will be the result utilizing enzymes, metals and a range of dedicated ligands. Increased knowledge of active site structures, effect of protein modifications, functional insight in enzyme systems, cell compartments and complete cells will be at the basis of these new developments together with the mechanistic insights from chemocatalysis. Specific challenges in synthetic organic chemistry are: _ direct functionalization of aromatic compounds; for instance replacing Friedel Crafts type chemistry by direct arene alkylation/acylation using olefins; _ cross coupling reactions, which play a prominent role in current synthetic repertoire, based on olefins; _ (bio-)catalytic reductions and oxidations; _ synthetic conversions without protective groups; _ catalytic methodology in heterocyclic chemistry. Numerous bioactive products in particular pharmaceuticals and agrochemicals are based on multifunctional heterocyclic compounds. Hardly any of the current catalytic methods can be employed for heterocyclic substrates due to rapid catalyst poisoning. Another major challenge is the reduction of the number of steps in common multi-step synthesis. The combination of mutually depending bio- and chemocatalysis is only one

26 of the possibilities to develop synthetic methods not depending on exhaustive protective group manipulation. Eventually multi-step, once-through processes will evolve from exceptions today to common methodology tomorrow. Combining combinatorial approaches with understand, design and build methods will be an additional challenge in reaching these new synthesis methodologies.

Acknowledgement l.This essay is the introduction of a new national Dutch research program aiming at integration of organic synthesis and biosynthesis in a joint effort of academia and the Dutch life sciences industries. The program will be organized by the Council for Chemical Sciences of NWO, the national Dutch organization for science and technology. 2.This essay has been used as a guideline for the panel discussion on "Green Chemistry" at the 13* Noordwijkerhout-Camerino Symposium on "Trends in Drug Research" on May 6-11,2001 in The Netherlands.

Further reading 1. P.S. Zurer, "Annulation Strategies (cascade reactions)", Chem. Eng. News, 79 (2001)27-30. 2. "New Voices in Chemistry", Chem. Eng. News, 79 (2001) 51-291. 3. A.I. Scott, "Towards a Total, Genetically-Engineered Synthesis of Vitamin-B12", Synlett. (1994), 871-883 4 . R. Schoevaart, F. van Rantwijk and R.A. Sheldon, "Class I fructose-1,6bisphosphate aldolases as catalysts for asymmetric aldol reactions". Tetrahedron Asymmetry, 10 (1999), 705-711. 5. R.A. Sheldon and H. van Bekkum, "Future Outlook", in "Fine Chemicals through Heterogeneous Catalysis", pg. 589-592, Wiley-VCH, 2001. 6. B. Zwanenburg (ed.), "Enzymes in Action, Green Solutions for Chemical Problems", Kluwer Academic Publishers, 2000. 7. A. Bruggink (ed.), "Synthesis of P-lactam Antibiotics, Chemistry, Biocatalysis and Process Integration", Kluwer Academic Publishers, 2001. 8. J.J. Heijnen, "Microorganisms as Micro-Chemical Factories for Sustainable Precision Production of Chemicals" (Conference Report), pg. 69-70, Gratama Workshop 2000, Osaka, Japan . 9. H.C.Kolb, M.G.Finn and K.B.Sharpless, "Click Chemistry: Diverse Chemical Functions from a Few Good Reactions", Angew. Chem. Intern. Edit., 40 (2001),2004-2021.

H, van der Goot (Editor) Trends in Drug Research III © 2002 Elsevier Science B.V. All rights reserved

27

Directed Evolution of Enantioselective Enzymes as Catalysts in the Production of Chiral Pharmaceuticals Manfred T. Reetz Max-Planck-Institut fur Kohlenforschung Kaiser-Wilhelm-Platz 1 45470 Mtilheim an der Ruhr, Germany

1

Introduction

The development of methods for the stereoselective synthesis of chiral organic compounds is of enormous academic and industrial interest [1-3]. Indeed, a major portion of research in organic chemistry during the last 30 years has been devoted to asymmetric synthesis. Such activities are certainly driven in part by the need to prepare chiral therapeutic drugs in both enantiomerically pure forms. In fact, the so-called "chiral market" of enantiomerically pure or enriched organic compounds is expanding rapidly, total sales of chiral pharmaceuticals alone exceeding 100 billion $ US in 2000 [le]. Not all, but certainly many of these compounds are prepared in the laboratories of organic chemists. Presently, classical antipode separation is used most often in industry [Id]. However, this requires stoichiometric amounts of an appropriate optically active reagent as well as large amounts of organic solvents. Due to ecological and economic reasons, asymmetric catalysis can be expected to be more efficient, provided that active and highly enantioselective catalysts can be found. Catalytic asymmetric transformations can be carried out either in the form of kinetic resolution of racemates or in reactions involving prochiral substrates. Two options are available, namely transition metal catalysis [2] or biocatalysis [3]. Success in the area of asymmetric transition metal catalysis entails a difficult goal [2], namely efficient ligand tuning (Fig. la), which in turn requires experience, intuition, knowledge of the reaction mechanism and the ability to apply molecular modeling as well as time-consuming trial and error. Numerous successful examples have been reported in the ongoing academic literature, and a few of these have been commercialized [1].

a)

I >D.

!'9^"^ ^. tuning b)

C l^W ^— site specific ^ ^ | J ^ mutagenesis near active site

VD'

Fig. 1 a) Schematic representation of ligand tuning in the design and synthesis of a chiral transition metal (M) catalyst, C2-symmetry arbitrarily being shown; the arrow symbolizes points of potential structural variation and D denote donor atoms, b) Schematic representation of "de novo design" of an enantioselective enzyme, the arrow symbolizing the exchange of amino acids on the basis of site specific mutagenesis.

28

In the general area of biocatalysis, enzymes [3] and catalytic antibodies constitute [4] the most important options. Since enzymes are inherently the more active catalysts, they have been used most often. Indeed, a number of industrial processes for the enantioselective synthesis of intermediates needed in the production of chiral pharmaceuticals are based on the application of enzymes. A prominent example is the lipase-catalyzed kinetic resolution of an epoxy-ester used in the production of the antihypertensive therapeutic Dilthiazem® [5]. There seems to be a trend in industry to use enzymes more often than in the past [6]. However, these catalysts suffer from the disadvantage that for a given synthetic transformation of interest. A—>B, enantioselectivity may well be poor. In principle, it should be possible to apply sitedirected mutagenesis [7] in order to increase enantioselectivity to an acceptable level (Fig. lb), similar to ligand tuning in transition metal catalysis (Fig. la). However, this has not turned out to be a straightforward process. We have recently proposed a radically different approach to the development of enantioselective catalysts which does not rely on any knowledge regarding the structure of the enzyme nor on any speculations concerning enzyme mechanism [8]. The combination of proper molecular biological methods for random mutagenesis and gene expression coupled with high-throughput screening systems for the rapid identification of enantioselective variants of the natural (wild-tpye) enzyme forms the basis of the concept [8-10]. The idea is to start with a wild-type enzyme showing an unacceptably low enantiomeric excess (ee) or selectivity factor (E) value for a given transformation of interest. A—>B, to create a library of mutant genes, to identify the most enantioselective enzyme-variant following expression, and to repeat the process as often as necessary using in each case an improved mutant gene for the next round of mutagenesis. Due to the fact that the inferior mutant genes and enzyme variants are discarded, an evolutionary "pressure" in the overall process builds up (Fig. 2) [8-10].

gene (DNA)

I

wild-type enzyme

random mutagenesis

library of mutated genes repeat

I expression

\

' 9 w

W"^'^-

library of enzyme-variants

I

screening (or selection) for enantioselectivity

positive mutants Fig. 2 Directed evolution of an enantioselective enzyme [9]

29 During the last 15 years molecular biologists have developed new and practical techniques for random mutagenesis. For example, Leung, Chen and Goedell described the technique of error prone polymerase chain reaction (epPCR) in which the conditions of the classical PCR are varied empirically (e.g., the MgC^ concentration) so as to attain the desired mutation rate [Ua]. Later the method was improved by Cad well and Joyce [lib]. This procedure of inducing point mutations was followed in 1994 by Stemmer's method of DNA shuffling [12] and in 1998 and 1999 by Arnold's staggered extension process [13] and random priming recombination method [14], respectively, which are all recombinative processes resulting in a high diversity of mutant genes. Since then these and other methods such as saturation mutagenesis (in which the substitution or insertion of codons is performed leading to all possible 20 amino acids at any predetermined position in the gene) or cassette mutagenesis (using DNAfragments made up of nucleotides encoding one to several hundred amino acids in a defined region of the enzyme) have been applied in the quest to obtain structurally altered enzymes with improved stability and activity [11-18]. However, enantioselectivity is a difficult parameter, and at the outset of our efforts it was not clear whether the tools of directed evolution can be applied successfully in the area. The major challenges in putting the concept described in Figure 2 into practice involve the development of efficient strategies for exploring protein sequence space with respect to enantioselectivity and the establishment of high-throughput screening or selection methods for assaying enantioselective enzymes. In this chapter we summarize the current status of this new and exciting branch of asymmetric catalysis [9]. 2

High-Throughput Screening Systems for Enantioselectivity

The determination of ee values is traditionally performed by gas chromatography or HPLC using chiral phases, but only a few dozen samples can be analyzed per day. When we initiated research in 1995 concerning the evolution of enantioselective enzymes (see below) [8], high-throughput assays for enantioselectivity were unknown. Several methods have been developed since then, including systems based on UVA^is [8, 19-21], IR-thermography [22], MS [23], and capillary array electrophoresis [24]. Since a complete review concerning the scope and limitations of these and other methods has appeared recently [10], only a few highlights as well as new developments are mentioned here. A very practical assay is a method based on electrospray ionization mass spectrometry (ESI-MS) [23]. The (/?)- and (5)- enantiomers of a given chiral product have identical mass spectra and, in the absence of chromatographic separation, cannot be distinguished. However, if one of the enantiomers is deuterium-labeled, the parent peaks appear separately in the mass spectrum of the mixture, and integration then provides the ee value. Accordingly, deuterium-labeled substrates in the form of pseudoenantiomers or pseudo-prochiral compounds are used to test a potentially enantioselective (bio)catalyst. The method is restricted to the kinetic resolution of chiral compounds and to reactions of prochiral compounds having enantiotopic groups. The system has been automated, allowing for about 1000 ee determinations per day [23]. The assay has been applied in several cases [10,23]. A new example involves the hydrolytic kinetic resolution of the epoxide 1 catalyzed by an epoxide hydrolase (Fig. 3) [25]. The chiral diol 2 is of considerable interest in the pharmaceutical industry. As in other kinetic resolutions, the reacton is allowed to reach the ideal value of 50 %.

30

Instead of employing a genuine racemate (J?)-l/(5)-l as in a normal lab-scale experiment, a 1 : 1 mixture of pseudo-enantiomers (/?)-l/(5)-(D5)-l needs to be used because at any point of the reaction the ratio of (/?)-! : (5)-(D5)-l and (/?)-2 : (S)'(Ds)-2 can be determined simply by integrating the appropriate ESI-MS peaks. This delivers the enantiomeric excess (ee) as well as the selectivity factor (£), provided the appropriate "time window" is used. If a given library of enzyme-variants contains very active as well as less active (or even inactive) enzymes, a rough pre-screening test is advisable. Several MS measurements as the reaction progresses provides time resolution if necessary.

3/^'

^O

0

^\

H2O ^. epoxide hydrolase

(fl)-i HO

OH

^

(S)-(D5)-1

o

HO

OH

D

D^

D

\ D

(S)-(D5)-2

Fig. 3 Kinetic resolution of pseudo-enantiomers (/?)-l/(5)-(D5)-l [25] If reactions involve desynmietrization of meso-type substrates, kinetic resolution is not involved, which means that the transformation can be run to 100 % conversion [10,23], and the "time window" is of no concern. The use of the ESI-MS-based assay in such cases requires the synthesis and application of pseudo-meso substrates, i.e., mesosubstrates that contain deuterium-labeled enantiotopic groups. In sunmiary, although the MS-based assay is restricted to the two symmetry classes as delineated here, it is highly efficient. Moreover, new MS instruments utilizing an 8-channel multiplex electrospray source allow for even higher throughput, which means that about 8000 ee determinations can be performed in one day. An alternative and also practical MSbased assay requires derivatization using chiral reagents [26]. Other options include high-throughput e^-assays based on capillary array electrophoresis [24], color tests [8,19-21,27], circular dichroism [28] and DNA microarrays [29]. 3

First Example of Directed Evolution of an Enantioselective Enzyme

We initially studied the kinetic resolution of the lipase-catalyzed hydrolysis of the chiral ester 3 in which a maximum of 50 % conversion is aimed for [8]. Lipases are enzymes

31 that catalyze the hydrolysis of esters [5], the reverse reaction in organic solvents also being possible. The particular enzyme used in our case was the bacterial lipase from Pseudomonas aeruginosa, which showed an ee-value of only 5 % in favor of the (S)acid 4 at 50 % conversion.

\^

/

p. P.aeruginosa aeruginosa

rac-3 (R = n-CsH^ 7)

11

^..

(S)-4

.

r"r

OH

(^-4

Fig. 4 Lipase-catalyzed hydrolytic kinetic resolution

The first step in directed evolution is the consideration of the mutation rate, which has to do with the problem of exploring protein sequence space. The lipase from Pseudomonas aeruginosa has 285 amino acids. Complete randomization would result in 20^^^ different enzyme-variants, which is more than the mass of the universe, even if only one molecule of each enzyme were to be produced [8,9]. The other extreme entails the minimum amount of structural change, namely the substitution of a single amino acid per molecule of enzyme by one of the other 19 naturally occurring amino acids. In this case, on the basis of the algorithm A^= 19M-285!/[(285-Af)! -M!], where M = number of amino acid substitutions per enzyme molecule, the library of variants would theoretically have 5415 members [8, 9]. However, when using epPCR as the random mutagenesis method, a library of 5000 - 6000 members is not expected to contain all theoretically possible permutations. This is because the genetic code is degenerate. If two amino acids are exchanged per enzyme molecule (M = 2), then the number of enzyme-variants increases dramatically (about 14 million!). In the case of M = 3, it is more than 52 billion. We therefore initially chose a low mutation rate so as to induce an average of only one amino acid exchange per enzyme molecule [8]. Thus, in the case of the kinetic resolution of the ester 3, epPCR was adjusted to cause about 1 - 2 base substitutions per 1000 base pairs of the gene, resulting in an average of one amino acid exchange. Typically, 2000 - 3000 enzyme-variants per generation were screened. Following expression in E. colilP, aeruginosa, a screening system based on the UVA^is absorption of the liberated p-nitrophenolate at 410 nm was employed. As a consequence of the first round of mutagenesis and screening, a variant displaying an ^^-value of 31 % was identified (£ = 2.1). The corresponding mutant gene was then subjected once more to mutagenesis, and the process was repeated several times. The results after four generations of mutants led to an ee-value of 81 %, the selectivity factor being £ = 11.3 (Fig. 5).

32

81%^.»« (E=11.3)

1

2 3 mutant generations

Fig. 5 Increasing the enantioselectivity of the lipase-catalyzed hydrolysis of the model ester 3 These remarkable results constitute proof of principle. Nevertheless, a selectivity factor of £ = 11.3 cannot be viewed as industrially viable. Thus, a fifth round of mutagenesis was performed, and indeed the usual library of about 2000 mutants contained slightly improved variants. In spite of this advancement it became clear to us that we needed to develop methods which allow for even more efficient ways to explore protein sequence space with respect to enantioselectivity [9, 30]. Accordingly, DNA-analyses leading to amino acid sequence determinations of the variants were carried out as a first step. For example, the best mutants of the first four generations turned out to have the following amino acid substitutions (Fig. 6). Variant 01E4(E= 2.1):

Seri

Variant 08H3(E = 4.4):

Glyi49 Leui55

Variant 13D10{E= 9.4): Ser^g Seri55 Val47 Variant 04H3(E= 11.3):

Glyi49

Giyi49 Leui55 Gly47 Glyi49 Leui55 Gly47 LeU259

Fig. 6 Data of amino acid exchanges in the best lipase-variants of the first four generations [9, 30]

33 We then drew the following conclusions [9, 30]: 1. The process of random mutagenesis/screening identifies sensitive positions ("hot spots") in the enzyme which are responsible for improved enantioselectivity. 2. Such positions are likely to be correct, but the particular amino acid identified may not be optimal. 3. Saturation mutagenesis at the "hot spots" can be expected to generate improved mutants. Rather than continuing with epPCR in further cycles of random mutagenesis, we decided to utilize appropriate combinations of various types of mutagenesis. Saturation mutagenesis is a molecular biological method with which mutations at a given position of an enzyme can be introduced, a small library of only 300 - 400 variants being necessary to ensure that all of the remaining 19 amino acids have been introduced. Upon applying this strategy at one of the hot spots (e.g., at position 155), it was discovered that phenylalanine (F) is the best amino acid . Saturation mutagenesis using the best gene in the third generation led to the identification of a variant which showed a selectivity factor of ^ = 20, phenylalanine again "showing up" as the best amino acid at position 155. Thereafter, epPCR was applied again, which resulted in E= 25! Clearly, the combination of mutagenesis methods, namely epPCR and saturation mutagenesis, constitutes an efficient method to explore protein sequence space with respect to enantioselectivity. Thus, a small family of enzymes was created, all showing £-values of 20 - 25 and £^-values of 88 - 91 % for the model reaction [9, 30]. Following these developments, recombinant methods such as DNA-shuffling in combination with methods based on point mutations, were applied. Specifically, Combinatorial Multiple Cassette Mutagenesis using two genes obtained at high mutation rate and an oligo-cassette at two "hot spots" led to a selectivity factor of E>51 (^^>95%) [31]. Moreover, it was possible to invert the sense of enantioselectivity [9, 32]. The present results are summarized in Figure 7. (S)-selectivJty E>51 (ee> 95 %)

4o4 304

E=20-25

2o4

cassette! mutagenesis pointl mutations

1o4

l4

^ ^ = 3 4 pQjptj mutations and DNAi shuffling

wild"lype

1o4

• point| mutations and DNAi shuffling

20^ (/lO-selectlvity

E=20

©

Fig. 7 Optional (S)- or (/?)-selectivity in the lipase-catalyzed hydrolysis of ester 3 [9,30,31]

34

4

Extending the Flow of Genetic Information from DNA to Transition Metal Catalysts

Following our initial reports concerning the directed evolution of enantioselective lipase-variants, further examples of the underlying principle have been reported [33]. Thus, the general concept schematized in Figure 2 may well turn out to be general. Moreover, directed evolution of enantioselective enzymes provides a unique opportunity to study structure/selectivity relations, provided the crystal structures of the newly evolved enzyme-variants become available. Parallel to these efforts we have started to develop a concept which goes far beyond directed evolution of improved enzymes. Accordingly, we are extending the flow of genetic information from the gene (DNA) to transition metal catalysts (Fig. 8) [34]. riMA transcription^ _ . , . translation^ _ chemical DNA l i — • RNA ^^^j^^-——-^

^ i-«,„^^I f^^M Enzyme I H ^ L ^ M

Fig. 8 The flow of genetic information from DNA to transition metal catalysts (L = ligand; M = transition metal) A wild-type enzyme is chosen having a cavity large enough for potential substrate binding. It should contain a "reactive" amino acid residue (e.g., cystein) at the cavity so that chemical modification with introduction of ligands (phosphines, nitrogen-moieties, etc.) capable of binding transition metals such as palladium, rhodium, nickel, etc. Such catalysts have enzyme-like structures and therefore appropriate binding properties, yet the actual transformations are not "enzymatic" in the traditional sense because transition metal catalyzed reactions such as hydrogenation, hydroformylation, oxidation, allylic substitution, cycloadditions, etc. are the focus of interest. We have shown that such implantation of catalytically active transition metal sites is possible [34]. The next step is to perform mutagenesis on the wild-type gene as described above in the directed evolution of enzymes, to express the enzyme-variants and finally to introduce ligands/metals at defined positions for thousands of variants using robotics before testing traditional transition metal catalyzed reactions en masse employing the ^e-screening systems previously developed [10]. Not only enantioselectivity, but also activity as well as chemo- and regioselectivity can then be optimized.

5

Conclusion

Directed evolution of enantioselective enzymes is emerging as a new and fascinating area of research. Two major problems have already been dealt with successfully, namely the development of strategies for efficiently exploring protein sequence space with respect to enantioselectivity, and the establishment of high-throughput eescreening systems. Nevertheless, more efforts are necessary before generality can be claimed, including the study of a wide variety of enzymes and substrates. It can be predicted that directed evolution will be used successfully to create novel enzymevariants which are highly enantioselective and active as well as stable enough to allow for a variety of applications in organic and/or pharmaceutical chemistry. This applies all the more to the idea of conceptually fusing molecular biology with transition metal chemistry in order to evolve completely new types of catalysts.

35 References [1]

[2]

[3]

[4]

[5] [6]

[7]

a) A.N. Collins, G.N. Sheldrake, J. Crosby, (Eds.), Chirality in Industry: The Commercial Manufacture and Applications of Optically Active Compounds, Wiley, Chichester, 1992; b) A.N. Collins, G.N. Sheldrake, J. Crosby, (Eds.), Chirality in Industry II: Developments in the Commercial Manufacture and Applications of Optically Active Compounds, Wiley, Chichester, 1997; c) R.A. Sheldon, Chirotechnology: Industrial Synthesis of Optically Active Compounds, Dekker, New York, 1993; d) S.C. Stinson, Chem, Eng. News 1999, 77(41), 101120; e) S.C. Stinson, Chem. Eng. News 2000, 75(43), 55-78. a) E.N. Jacobsen, A. Pfaltz, H. Yamamoto, (Eds. j , Comprehensive Asymmetric Catalysis, Vol. Mil, Springer, Berlin, 1999; b) H. Brunner, W. Zettlmeier, Handbook of Enantioselective Catalysis with Transition Metal Compounds, Vol. I-II, VCH, Weinheim, 1993; c) R. Noyori, Asymmetric Catalysis in Organic Synthesis, Wiley, New York, 1994; d) I. Ojima, (Ed.), Catalytic Asymmetric Synthesis, VCH, Weinheim, 1993; e) D.J. Berrisford, C. Bolm, K.B. Sharpless, Angew. Chem. 1995, 107, 1159-1171; Angew. Chem. Int. Ed. Engl. 1995, 34, 1059-1070. a) H.G. Davies, R.H. Green, D.R. Kelly, S.M. Roberts, Biotransformations in Preparative Organic Chemistry: The Use of Isolated Enzymes and Whole Cell Systems in Synthesis, Academic Press, London, 1989; b) C.H. Wong, G.M. Whitesides, Enzymes in Synthetic Organic Chemistry (Tetrahedron Organic Chemistry Series, Vol. 12), Pergamon, Oxford, 1994; c) K. Drauz, H. Waldmann (Eds.), Enzyme Catalysis in Organic Synthesis: A Comprehensive Handbook, Vol. MI, VCH, Weinheim, 1995; d) K. Faber, Biotransformations in Organic Chemistry, 3rd ed.. Springer, Berlin, 1997; e) G. Carrea, S. Riva, Angew. Chem. 2000, 112, 2312-2341; Angew. Chem. Int. Ed. 2000, 39, 22262254; f) K.M. Koeller, C.-H. Wong, Nature (London) 2001,409, 232-240. a) P.G. Schultz, R.A. Lerner, Science (Washington, D.C.) 1995, 269, 18351842: b) J.-L. Reymond, Top. Curr. Chem. 1999, 200, 59-93; D. Hilvert, Top. Stereochem. 1999,22, 85-135. R.D. Schmid, R. Verger, Angew. Chem. 1998, 110, 1694-1720; Angew. Chem. Int. Ed. 199S, 37, 1608A633. a) M. McCoy, Chem. Eng. News 1999, 77(1), 10-14; b) R.S. Rogers, Chem. Eng. News 1999, 77(29), 87-91; c) B. Schulze, M.G. Wubbolts, Curr. Opin. Biotechnol. 1999, 10, 609-615; d) A. Schmid, J.S. Dordick, B. Hauer, A. Kiener, M. Wubbolts, B. Witholt, Nature (London) 2001,409, 258-268. a) M.M. Altamirano, J.M. Blackburn, C. Aguayo, A.R. Fersht, Nature (London) 2000, 403, 617-622; b) P.N. Bryan, Biotechnol. Adv. 1987, 5, 221-234; c) J.A. Gerlt, Chem. Rev. 1987, 87, 1079-1105; d) J.R. Knowles, Science (Washington, D.C.) 1987, 236, 1252-1258; e) S.J. Benkovic, C.A. Fierke, A.M. Naylor, Science (Washington, D.C.) 1988, 239, 1105-1110; f) J.A. Wells, D.A. Estell, Trends Biochem. Sci. 1988,13, 291-297; g) Y. Hirose, K. Kariya, Y. Nakanishi, Y. Kurono, K. Achiwa, Tetrahedron Lett. 1995,36, 1063-1066; h) Z. Shao, F.H. Arnold, Curr. Opin. Struct. Biol. 1996, 6, 513-518; i) C. Heiss, M. Laivenieks, J.G. Zeikus, R.S. Phillips, / Am. Chem. Soc. 2001,123, 345-346; j) S. Patkar, J. Vind, E. Kelstrup, M.W. Christensen, A. Svendsen, K. Borch, O. Kirk, Chem. Phys. Lipids 1998, 93, 95-101; k) M. Holmquist, P. Berglund, Org. Lett. 1999,

36

[8] [9] [10]

[11] [12] [13] [ 14] [15] [16] [17] [18] [19] [20]

[21]

[22] [23] [24] [25] [26] [27] [28]

[29] [30] [31]

/, 763-765; 1) P. Kast, D. Hilvert, Cum Opin. Struct Biol 1997, 7,470-479; m) J.P. Schanstra, A. Ridder, J. Kingma, D.B. Janssen, Protein Eng. 1997, 10, 5361; n) M.H.J. Cordes, N.P. Walsh, C.J. McKnight, R.T. Sauer, Science (Washington, D.C) 1999,284, 325-327. M.T. Reetz, A. Zonta, K. Schimossek, K. Liebeton, K.-E. Jaeger, Angew, Chem. 1997,109,2961-2963; Angew. Chem, Int. Ed. Engl. 1997,36,2830-2832. M.T. Reetz, K.-E. Jaeger, Chem.^-Eur. J. 2000,6,407-412. Review of high-throughput e^-screening systems in the directed evolution of enzymes and in asymmetric transition metal catalysis: M.T. Reetz, Angew. Chem. 2001,113, 292-320; Angew. Chem. Int. Ed. 2001,40,284-310. a) D.W. Leung, E. Chen, D.V. Goeddel, Technique (Philadelphia) 1989, 1, 1115; b) R.C. Cadwell, G.F. Joyce, PCR Methods Appl. 1994,3, S136-S140. a) W.P.C. Stemmer, Nature (London) 1994, 370, 389-391; b) P.A. Patten, R.J. Howard, W.P.C. Stemmer, Curr. Opin. Biotechnol. 1997, 8, llA-m^. H. Zhao, L. Giver, Z. Shao, J.A. Affholter, F.H. Arnold, Nat. Biotechnol. 1998, 76,258-261. Z. Shao, H. Zhao, L. Giver, F. Arnold, Nucleic Acids Res. 1998,26, 681-683. a) F.H. Arnold, Ace. Chem. Res. 1998,31, 125-131; b) J. Affholter, F.H. Arnold, CHEMTECH1999,29, 34-39. M. Kanunann, J. Laufs, J. Schell, B. Gronenborn, Nucleic Acids Res. 1989, 17, 5405. A. Crameri, S.-A. Raillard, E. Bermudez, W.P.C. Stemmer, Nature (London) 1998, i97, 288-291. M.T. Reetz, K.-E. Jaeger, Top. Curr. Chem. 1999,200, 31-57. L.E. Janes, R.J. Kazlauskas, J. Org. Chem. 1997,62,4560-4561. a) L.E. Janes, A.C. Lowendahl, R. J. Kazlauskas, Chem.—Eur. J. 1998,4, 23242331; b) F. Moris-Varas, A. Shah, J. Aikens, N.P. Nadkarni, J.D. Rozzell, D.C. Demirjian, Bioorg. Med. Chem. 1999, 7,2183-2188. a) F. Zocher, M.M. Enzelberger, U.T. Bomscheuer, B. Hauer, R.D. Schmid, Anal. Chim. Acta 1999, 391, 345-351; b) E. Henke, U.T. Bomscheuer, Biol. Chem. 1999,380, 1029-1033. M.T. Reetz, M.H. Becker, K.M. Kuhling, A. Holzwarth, Angew. Chem. 1998, 110,2792-2795; Angew. Chem. Int. Ed. 1998,37, 2647-2650. M.T. Reetz, M.H. Becker, H.-W. Klein, D. Stockigt, Angew. Chem. 1999, 111, 1S72A875; Angew. Chem. Int. Ed. 1999,38,1758-1761. M.T. Reetz, K.M. KUhling, A. Deege, H. Hinrichs, D. Belder, Angew. Chem. 2000, 772,4049-4052; Angew. Chem. Int. Ed. 2000,39, 3891-3893. M.T. Reetz, A. Eipper, C. Torre, Y. Genzel, R. Furstoss, unpublished results. J. Guo, J. Wu, G. Siuzdak, M.G. Finn, Angew. Chem. 1999, 777, 1868-1871; Angew. Chem. Int. Ed. 1999,38,1755-1758. D.C. Demirjian, P.C. Shah, F. Moris-Varas, Top. Curr. Chem. 1999,200, 1-29. a) K. Ding, A. Ishii, K. Mikami, Angew. Chem. 1999, 777, 519-523; Angew. Chem. Int. Ed. 1999, JS, 497-501; b) K. Mikami, R. Angelaud, K. Ding, A. Ishii, A. Tanaka, N. Sawada, K. Kudo, M. Senda, Chem. Eur. J. 2001, 7, 730737; c) M.T. Reetz, K.M. Kuhling, H. Hinrichs, A. Deege, Chirality, 2000, 72, 479-482. G.A. Korbel, G. Lalic, M.D. Shair, /. Am. Chem. Soc. 2001, 72i, 361-362. K. Liebeton, A. Zonta, K. Schimossek, M. Nardini, D. Lang, B.W. Dijkstra, M.T. Reetz, K.-E. Jaeger, Chem. Biol. 2000, 7,709-718. M.T. Reetz, S. Wilensek, D. Zha, K.-E. Jaeger, submitted.

37 [32] [33]

[34]

K.-E. Jaeger, T. Eggert, A. Eipper, M.T. Reetz, Appl. Microbiol. Biotechnol 2001, in press. a) U.T. Borascheuer, J. Altenbuchner, H.H. Meyer, Biotechnol Bioeng. 1998, 58, 554-559; b) O. May, P.T. Nguyen, F.H. Arnold, Nat Biotechnol 2000, 18, 317-320; c) S. Pong, T.D. Machajewski, C.C. Mak, C.H. Wong, Chem, Biol 2000, 7, 873-883. M.T. Reetz, P. Maiwald, P. Wedemann, unpublished.

This Page Intentionally Left Blank

H. van der Goot (Editor) Trends in Drug Research III © 2002 Elsevier Science B.V. All rights reserved

39

AT THE INTERFACE OF ORGANIC SYNTHESIS AND BIOSYNTHESIS ROB SCHOEVAART^ and TOM KIEBOOM^'^ ^Industrial Fermentative Chemistry, Leiden Institute of Chemistry, Leiden University, Leiden, Netherlands; ^DSM Food Specialties R&D, Delft, Netherlands

Organic synthesis^ i.e. chemistry by mankind, uses traditionally often a step-by-step approach to convert a starting material A into a final product D, in which intermediate products B and C have to be isolated and purified for each next conversion step: Conversion steps

Recovery steps

Such multi-step organic syntheses are also still quite common in to-day's fine chemical industry, i.e.: • often carried out in a non-catalytic way by using relatively large amounts of reagents, resulting in the production of multi-kg of waste per kg of final product; • requiring separation and purification steps after each conversion step in order to be able to do the next conversion, resulting in the production of heat waste due to the consumption of the appropriate amount of energy; • using extra energy to overcome eventual thermodynamic hurdles, i.e. to reach the final product D in case of higher energetic intermediates C and/or D.

40

Biosynthesisy i.e, chemistry by nature, in the cells of living organisms, goes through a multi-step cascade approach to convert a starting material A into the final product D without separation of intermediates B and C: Conversion steps

Such multi-step combined syntheses are quite common in every-day's life, i.e.: • carried out in a full-catalytic way by using enzymes with relatively limited amounts of reagents (cofactors) and so producing much less waste; • without intermediate recovery steps as possible by the mutual compatibility and high selectivity of the enzymatic conversions; • saving energy to overcome eventual thermodynamic hurdles, i.e. to reach the final product D in case of higher energetic intermediates B and/or C. For the next generation of organic synthesis^ it is the challenge to: • combine the power of the chemo-catalytic, enzymatic and microbial conversions; • search for multi-step conversions without recovery steps like nature does, i,e, to go for one-pot multi-catalytic procedures; • fine-tune reaction conditions and catalytic systems in order to allow for the right concerted cooperation without any intermediate isolation and/or purification steps. This to drastically diminish energy and waste, i.e. costs, of most of the present-day multi-step conversion processes of the fine chemical industry. For the next generation of biosynthesis., i.e. microbial conversions, it is the challenge to: • improve the productivity in terms of carbon source efficiency of micro-organisms, i.e. towards a much higher product^iomass ratio; • broaden the scope of products, i.e. from natural towards modified products often required for drugs;

41

•

re-engineer metabolic pathways as most efficient way to reach these two goals; This to free fermentation processes from excessive use of renewables that end up in undesired, low-value, biomass as by-product as well as to minimize additional chemical modification steps to obtain the final product. The above mentioned challenges for both organic synthesis and biosynthesis are by no way kind of wishful thinking. This will be demonstrated by the evolution during the past 30 years of the industrial process route for the synthesis of the antibiotic Cephalexin within DSM^ Here, the integration of chemistry, enzymatic conversion and fermentation forms the basis of an efficient production method from both an economic and environmental point of view (Figure 1). Figure 1. Integration of bio- and organic synthesis in modem industrial Cephalexin making.

D-Phenylglycine

CHEMISTRY ENZYME

7-ADCA

Fe^MENTATION

The 1970's Cephalexin synthesis consisted of one fermentation (step 3) together with eight organic synthesis steps^: 1. Benzaldehyde -^ D-Phenylglycine 2. [1] "> D-Phenylglycine chloride 3. Sugar + Phenylacetic acid -> Penicillin-G 4. [3] ^ Sulfoxide of Penicillin-G 5. [4] -> Trimethylsilyl ester of Penicillin-G sulfoxide

42

6. [5] "> Ring enlargement towards full protected 7-ADCA moiety 7. [6] ^ Hydrolysis to 7-ADCA 8. [2] + [7] ^ Cephalexin This multi-step sequence used high levels of energy (e.g. low temperature conversions 5-7, many recovery and solvent recycle steps), of reagents and organic solvents (steps 2 and 4-8) and of phenylacetic acid (as no recycling was possible after the chemical hydrolysis step 7). The 2000's Cephalexin synthesis consists of one organic synthesis (step 1), together with one fermentation (step 2) and two enzymatic steps: 1. Benzaldehyde -> D-Phenylglycine amide 2. Sugar + Adipic acid ^ Adipic amide of 7-ADCA 3. [2] -> 7-ADCA + Adipic acid (the latter recycled for use in step 2) 4. [1] + [3]-^ Cephalexin This sequence avoids low temperature conversions, major part of reagents and organic solvents, and consumption of adipic acid by: • enzymatic coupling of D-phenylglycine amide with 7-ADCA in water (avoiding the synthesis of acid chloride and use of reagents and organic solvents); • metabolic pathway engineered micro-organism (GMO) that is able to ferment sugar directly into the desired 7-ADCA moiety (thus avoiding 3 chemical conversion steps) ^; • aqueous enzymatic hydrolysis of the adipyl side chain that easily allows recycling of the adipic acid to be used again in the fermentation step (Figure 2). The list of reagents (often used in more than stoichiometric amounts) and solvents (that could not be recycled completely) that are skipped by the novel cephalexin synthesis are^: Peracetic acid Trimethylsilyl chloride and Bis(trimethylsilyl)urea Pyridine.HBr and Trimethylamine.HCl Phosphorpentachloride Dichloromethane Butanol Phenylacetic acid Acid chloride HCl salt of D-phenylglycine

43

SUGAR

Figure 2. Fermentative 7-ADC A production by GMO including enzymatic recycling of adipic acid\ Apart from such a sequential cooperation of chemical, enzymatic and microbial conversions on an industrial multi-tons scale, a number of combined multi-step one-pot catalytic conversions have been described on both lab- and industrial scale, using either a combination of enzymes or a combination of an enzyme and a chemo-catalyst (Table 1). Table 1. Combined Catalytic One-pot Conversions Conversion D-Glucose -> D-Fructose -> D-Mannitol Glycerol - » - > ^ ^ D-Heptulose

Benzaldehyde -> (S)-Mandelonitrile acetate j Acetoacetate ester + D-Glucose -» (R)-3-hydroxybutanoate ester + D-Gluconate

Reaction steps & Conditions Hydrogenation Isomerisation 70°C, 50 atm H2 Phosphorylation Oxidation Aldol reaction pH 4 •> 7.5 -> 4 Cyanohydrin form. Racemisation Acylation Oxidation Reduction Cofactor-regeneration Two-phase system Ambient conditions

ChemoCatalyst Cu/Si02

BioCatalyst Glucoseisomerase Phytase Oxidase+Catalase Aldolase

HOexchanger

Ref 3

4

Lipase 5

Glucose dehydrogenase Aldehyde dehydrogenase

6

44 D/L-Hydantoin -> D-Amino acid Alkyipyruvate + Formiate -> L-amino acid + CO2 a-Keto acids -> D-Amino acids

(R/S)-epoxide -> (S)-diols

Starch -> D-Fructose

Sucrose -> D-Fructooligosacchardes Unsaturated triacylglycerols -> Hydroperoxy fatty acids ^^C-formiate + glycine ^ 3-^^C-L-Serine D-glucose 6phosphate -> D-Gluconic acid 6phosphate (R/S)-Phenylethyiamine + EtOAc -> (S)-N-Acetylphenylethylamine 1 Cephalosporin C + Methyl tetrazolylacetate + 2-Mercapto-5methylthiadiazole -> Cefazolin 4-Methylcylcohexanone + Formiate -> (S)-4-Methylcaprolacton + C02

Racemisation Hydrolysis pH8.5,50°C Reductive amination Cofactor regeneration pH8,25°C Aminotransfer Redox reaction Racemisation

HO-

Enzymatic followed by acidic hydrolysis T30"^10°C pH 7.5^1 Hydrolysis Isomerisation Double-immobilized enzyme system pH=6,70°C Fructosyl transfer Oxidation pH=6,50°C Hydrolysis Peroxidation pH=9,26"C,02, octane/water Hydroxymethylation Cofactor regeneration Oxidation Ambient conditions

H"

Acylation Racemisation Ethyl acetate

Pd/C

Oxidation Amide hydrolysis Amine acylation Thio-ether formation pH=8">6.5 T25->4^65"C Oxidation Cofactor regeneration pH8,30^C

HO-

Hydantoinase 7

Amino acid + Formiate dehydrogenases Racemase Dehydrogenases D-Aminoacid aminotransferase Epoxide hydrolase

Glucoamylase Glucose isomerase

Fructosyltransferase Glucose oxidase Lipase Lipoxygenase

8

9

10

11

12

13

Serinehydroxymethyltransferase Dihydrofolate reductase Glucose 6phosphate dehydrogenase Lipase

14

15

Aminoacid oxidase Glutaryl acylase Penicillin G acylase Cyclohexanone mono-oxygenase Formate dehydrogenase

16

17

45 (R/S)-lphenylethanol + 4Chlorophenyl acetate -> (R)-l-Phenylethyl acetate+ 4Chlorophenol (R/S)-Allylic alcohols + AcOR-> (R)-Allylic esters (R/S)-2-Phenyl-3Acetoxycyclohexene -> (R)-2-Phenyl-3hydoxycyclohexene N-Ac-D-glucosamine + a-D-Glucose 1phosphate + Phospho-enolpyruvate -> N-Acetyllactosamine

Esterification Racemisation 20-70 °C, t-BuOH, cyclohexane

Sugars -> Complex Carbohydrates and Glycoconjugates

Glycosidic bond formation Epimerisation Phosphorylation Glycosyl transfer

Racemisation Esterification Organic solvent 100% cv/ee concept Ester hydrolysis Racemisation Disaccharide coupling Phosphorylation Epimerisation Ambient conditions

Ru catalyst

Lipase

Rh2(OAc)4 18

Ru catalyst

Lipase 19

PdCb

Lipase 20

Galactose transferase Phosphokinase UDP-Glucose pyrophosphorylase UDP-Galactose 4epimerase Transferases Phosphorylases Epimerases

21

22

Two major features are apparent from the data of Table 1: • By far, multi-step one-pot conversions have been reported in the field of carbohydrates using combinations of enzymatic 4,11,12,21,22

•

conversions ' ^ ^' ^ ^^^ ^ '^^; Combinations of metal- and bio-catalytic conversions are not yet that ^^^^^^3,15,18,19,20.

•

common ' ' ' ' ; After some early examples"^'^^ in the 1980's, there has been more than a decade of 'silence', followed by a clearly increasing interest during the past few years for combined catalytic conversions.

A very first example of the combined action of an enzyme and a metal catalyst is the direct one-pot conversion of glucose into mannitol, which is trice as expensive as glucitol:

46

Glucose

I^" Fructose

V Glucitol

Mannitol

Glucose isomerase on silica Copper on silica Hydrogen pressure of 70 atm Water, pH=7-8, at 70 ^C

Here, the isomerase enzyme converts glucose into a -1:1 glucose-fructose mixture and takes care that this mixture remains in equilibrium, while at the same time the copper catalyst hydrogenates preferentially fructose from this equilibrium into mannitol. This combi approach looks, at first sight, quite simple but in practice a number of fine-tuning measures had to be taken to achieve a balanced cooperation of the two simultaneous catalytic conversion steps, e.g.: • Immobilization of the enzyme onto silica to prevent poisoning of the copper metal by protein sulphur moieties; • Protection of the enzyme by a copper ion complexing agent (EDTA) to avoid inhibition by traces of copper ions from the copper catalyst; • Right compromise of hydrogen pressure and temperature to fulfil stability and activity requirements for both catalyst systems; • Slightly basic pH to avoid that mutarotation of glucose, i.e. the interconversion of a- to p-glucopyranose forms, becomes rate limiting as the enzyme only converts the a-form. What's really happening can be seen from the quite complicated kinetic and molecular picture (Figure 3) including three 'different' types of kinetics, expressed in TON's (sec"*), i.e.: • Michaelis-Menten (enzymatic isomerisation; only two of the six sugar forms are substrates for the enzyme; KM-values for glucose, fructose, glucitol and mannitol are 0.13, 0.04, 0.4 and > 1 M, respectively); • Langmuir-Hinshelwood (heterogeneous hydrogenation on copper; adsorption constants b vary from 3-10 M"*; only --25% of the copper surface covered with fructose that, however, reacts much faster than glucose adsorbed on the copper catalyst); • Homogeneous catalysis (acid/base catalysed mutarotation, i.e. interconversion of the different glucose and fructose forms; the rate for glucose is 50 times slower than that for fructose).

47

M U T A R 0 T A T I 0 N

OH

H - C - H HO--CH HO-CH HC - O H HC - O H H , C - OH 6 5 V.

mannitol

Figure 3. Facsimile^^ of molecular and kinetic picture of one-pot glucose to mannitol conversion: enzymatic isomerisation, mutarotation and copper catalysed hydrogenation.

48

The above mentioned principle of an equilibrium of two compounds of which one is selectively converted, together with the required fine-tuning of simultaneous catalytic conversion steps in one pot, is of great importance for the so-called 100% ee-100% yield synthesis of enantiomeric pure compounds from racemic starting materials (c/Table 1, r5,7,10,15,18,19,20y

(S)-A

(R)-A

(R)-B

Some recent examples on lab-scale have been reported^^'^^"^^ for the concomitant action of transition metal catalysts with lipases for the racemisation and esterification, respectively, of racemic 1-arylalcohols into high ee esters in high yield, e,g, ^^: CH3

v^

v^

K^

Rh2(OAc)4 Lipase Racemisation Esterification 70 ^C, cyclohexane There is no doubt that these kind synthesis of enantiomeric pure compounds with 100% e.e. and in 100% yield from relatively inexpensive racemates will find its way into the fine chemical industry in the near future (as proven, already, by the industrial hydantoinase process for amino acids, in which spontaneous racemisation occurs^)

Another elegant multi-step one-pot approach recently developed is ihQ four-enzyme catalysedfour-step one-pot conversion of glycerol into a heptose sugar derivative, in which a pH switch method is applied to temporary turn off phytase enzyme during the second and third steps of the concerted synthesis'*:

49

4 steps, one-pot

,

pyrophospha* Phytase

PO4 PO4

Phytase OPO3

OH FruA butanal

OPO3

I OH

Catalase

The four consecutive enzymatic conversion steps in one and the same reactor without any separation of intermediates consist of: • Phosphorylation: Glycerol is phosphorylated with pyrophosphate by phytase at pH 4.0 at 37 ^C. Racemic glycerol-3-phosphate is obtained in 100% yield (based on pyrophosphate) in 95% glycerol after 24h. • Oxidation: By raising the pH (to 7.5) phytase activity is "switched off, hydrolysis is prevented. Oxidation of L-glycerol-3-phosphate to DHAP by GPO at 55% glycerol (v/v) is quantitative. Catalase is added to suppress the build-up of hydrogen peroxide. The D-isomer is converted back to glycerol and phosphate in the last step. • Aldol reaction: More than twenty aldehydes are known to be substrates for the aldolases from S. carnosus and S. aureus. Stereoselectivity of the aldolases must be looked at for each acceptor substrate, since isomers are formed in different proportions. The oxidation and aldol reaction can be carried out simultaneously. • Dephosphorylation: Lowering the pH back to 4 "switches on" phytase's activity, hydrolysis of the aldol adduct is initiated. Combined with the broad substrate specificity of DHAP aldolases it constitutes a simple procedure for the synthesis of a wide variety of carbohydrates from readily available glycerol and pyrophosphate'*.

50

In conclusion, the concept and various examples given show that we may foresee a renaissance in synthesis methodology by integration of biO' and organic synthesis for fine chemicals by various approaches, from one-pot multi-step (bio)catalytic procedures towards metabolic engineered microbial transformations and combinations thereof^'*. In this respect, future clean synthesis methods should be inspired by the achievements in the field of modem detergent formulations that have up to six different enzymes in them^^, i.e. an advanced multi-catalytic onepot conversion of dirty laundry -> ^ - > - > - > -> clean laundry + dirt with the washing machine as in-house catalytic reactor that simultaneously separates the product (clean laundry) from waste (dirt). Finally, investigations of such multi-step synthesis methods without isolation of intermediate products require appropriate in situ analytical methods to know what's really happening during the consecutive conversions. Quite a powerful window to this information is the use of selectively isotope {e.g. ^^C, ^^N, ^^O^ enriched starting materials in combination with NMR^^. In this way, a sequence of conversions can be well characterized, even in complicated matrices of catalysts, reagents and mixed solvent systems, e.g. the investigation of galactose oxidase mediated cross-linking phenomena of D-galactose protein mixtures^"^.

References 1. DSM Magazine 147 (1998) 18. 2. J. Verweij and E. de Vroom, Reel. Trav. Chim. Pays-Bas 112 (1993) 66. 3. M. Makkee, A.P.G. Kieboom, H. van Bekkum, and J.A. Roels, J. Chem. Soe., Chem. Commun. (1980) 930; M. Makkee, A.P.G. Kieboom, and H. van Bekkum, Carbohydr. Res. 138 (1985) 237. 4. R. Schoevaart, F. van Rantwijk, and R.A. Sheldon, J. Chem. Soc. Chem. Comm. (1999) 2465 and Tetrahedron: Asymmetry 10 (1999) 705. 5. J. Oda, J. Am. Chem. Soc. 113 (1991) 9360. 6. S. Shimizu; M. Kataoka, M. Katoh, T. Miyoshi, and H. Yamada, Appl. Environ. Microbiol. 56 (1990) 2303. 7. P. Rasor and W. Tischer, in "Advances in Industrial Biocatalysis", Bio-Europe (1998) 50.

51 8.

9. 10.

11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23.

24.

25. 26.

27.

S. Rissom, U. Schwarz-Linek, M. Vogel, V.I. Tishkov, and U. Kragl, Tetrahedron: Asymmetry 8 (1997) 2523; A.S. Bommarius, K. Drautz, W. Hummel, M.R. Kula, and C. Wandrey, Biocatalysis 10 (1994) 37. A. Galkin, L. Kulakova, H. Yamamoto, K. Tanizawa, H. Tanaka, N. Esaki, and K.J. Soda, J. Ferment. Bioeng. 83 (1997) 299. R.V.A. Orru, S.F. Mayer, W. Kroutil, K. Faber, Tetrahedron 54 (1998) 859; S. Pedragosa-Moreau, C. Morisseau, J. Baratti, J. Zylber, A. Archelas, and R. Furstoss, Tetrahedron 53 (1997) 9707. Y. Ge, Y. Wang, H. Zhou, S. Wang, Y. Tong, W.J. Li, Biotechnol. 67 (1999) 33. G.S. Wang, Y.T. Liu, Taiwan Tangye Yanjiuso Yanjiu Huibao 151 (1996) 55. M. Gargouri, M.D. Legoy, Enzyme Microb. Technol. 21 (1997) 7. H. Maede, K. Takata, A. Toyoda, T. Niitsu, M. Iwakura, and K. Shibata, J. Ferment. Bioeng. 83 (1997) 113. M.T. Reetz and K. Schimossek, Chimia 50 (1996) 668. R. Femandez-Lafuente, J.M. Guisan, M. Pregnolato, and M. Terreni, Tetrahedron Lett. 38 (1997) 4693 and J. Org. Chem. 62 (1997) 9099. K. Seelbach and U. Kragl, Enzyme Microb. Technol. 20 (1997) 389. A.L.E. Larsson, B.A. Persson, and J.E. Backvall, Angew. Chem. Int. Ed. Engl. 36 (1997)1211; J.V. Allen and J.M.J. Williams, Tetrahedron Lett. 37 (1996) 1859. D. Lee, E.A. Huh, M.-J. Kim, H.M. Jung, J.H. Koh, and J. Park, Org. Lett. 2 (2000) 2377. P.M. Dinh, J.A. Howarth, A.R. Hudnott, J.M.J. Williams, and W. Harris, Tetrahedron Lett. 37 (1996)7623. C.-H. Wong, S.L. Haynie, and G.M. Whitesides, J. Org. Chem. 47 (1982) 5418. K.M. Koeller and C.-H. Wong, Chem. Rev. 100 (2000) 4465 and references cited herein. A.P.G. Kieboom, M. Makkee, and H. van Bekkum, unpublished scheme and data (1985) derived from both ref 1 and from: M. Makkee, "Combined action of enzyme and metal catalyst, applied to the preparation of D-mannitol", PhD Thesis, Delft University of Technology, Delft, Netherlands (1984). J.J. Heijnen, C.A.G. Haasnoot, A. Bruggink, R.A.L. Bovenberg, E.W. Meijer, B.L. Feringa, and A. Driessen, NWO Programme Proposal "Integration of Biosynthesis and Organic Synthesis, A New Future for Synthesis", 9 October 2000, The Hague. M. Mccoy, Chem. Eng. News, 19 Februari 2001, p. 23. J. Lugtenburg and H.J.M. de Groot, Photosynth. Res. 55 (1998) 241 and in: Stable Isotopes in Pharmaceutical Research, Pharmacochemistry Library 26(1997) and references cited therein. R. Schoevaart and T. Kieboom, Carbohydr. Res, submitted.

This Page Intentionally Left Blank

H. van der Goot (Editor) Trends in Drug Research III © 2002 Elsevier Science B.V. All rights reserved

53

What can Structure tell us about Function in the Estrogen Receptors? Roderick E. Hubbard, Structural Biology Laboratory, Department of Chemistry, University of York, Heslington, York, YOlO 5DD. [email protected] The estrogen receptor (ER) is a ligand-inducible transcription factor that controls expression of a number of genes in a wide variety of tissues. Binding of the natural hormone, estradiol, triggers dimerisation and nuclear location of the receptor v^here it binds to a response element (ERE). It then recruits a succession of large multi-protein complexes to modify chromatin and modulate the transcriptional machinery. The classical ER is ER-a, found in female reproductive organs but also important for maintaining bone and cardiovascular function. A number of important drug therapies depend on achieving agonism or antagonism of the receptor in these different tissues. Antagonists of the receptor are effective in treating some forms of breast cancer, whereas agonists are required to maintain bone function in post-menopausal women and thus protect against osteoporosis. About five years ago, a new isoform of the receptor (ER-P) was discovered which is distributed widely in both males and females. The differential role of these receptors is currently being investigated. We have determined the structure of the ligand binding domain (LBD) of both the a and P forms of ER complexed to different ligands, including partial and full agonists and selective and full antagonists. This has provided insights at the molecular level into some key aspects of the pharmacology and function of this molecule. Our results are described in detail in the papers listed in the bibliography. The sections below provide a brief summary of the key points. The distinctive ER pharmacophore Estradiol (E2), the main female sex hormone, contains two hydroxyl groups separated by a relatively flat and hydrophobic steroid core. A large number of estrogenic compounds have been characterised which have the same general features -. the distinctive ER pharmacophore. The overall structure of the LBD is a triple sandwich of alpha helices that generate a totally enclosed, mainly hydrophobic ligand binding pocket. In all structures determined to date, the A ring phenolic hydroxyl is found in the same position, hydrogen bonding to a water molecule and an arginine and glutamic acid residue. The D ring hydroxyl (and equivalents in other ligands) forms a hydrogen bond to a histidine amino acid. There are no other polar amino acids in the ligand binding pocket, which is tightly defined around the A ring, but there is additional space around the D ring.

54 The structure of ER LBD bound to a variety of ligands has demonstrated how this additional space is exploited to accommodate ligands of different shapes. The structures also confirm that the agonist conformation of the receptor has a key helix, helix 12, lying across the putative entrance to the binding cavity. Selective antagonism by molecules such as raloxifene There are a class of ligand molecules, known as Selective Estrogen Response Modulators (SERM) - such as tamoxifen and raloxifene (RAL) - which show different agonist and antagonist properties in different tissues. The structure of ER LBD complexed to RAL shows, as expected, that the core of the ligand is bound in essentially the same orientation as E2, with the small difference in shape accommodated by a movement of the histidine residue. The large pendant side chain exits from the binding cavity, preventing helix 12 from occupying its agonist position. Instead, helix 12 is found in an alternative binding site, identified as the major coactivator binding site on the protein. The agonism exhibited by these SERM ligands in some tissues could be due to residual coactivator binding sites elsewhere on the full length receptor, which are used in different cells with different coactivators expressed. Full antagonism A more extreme antagonist is the ICI compound, which consists of the estradiol core with a very long, hydrophobic side chain. This ligand acts as a full antagonist, abolishing estrogen receptor activity and in addition, there is some evidence that it increases receptor turnover. The structure of ER LBD complexed to ICI, shows that the ligand occupies the binding pocket, but has to flip from the standard E2 conformation to allow the side chain to exit by the same route used by RAL. The side chain then occupies the coactivator binding cleft - where helix 12 is found in the RAL structure. This means that helix 12 does not have an available alternative binding site and the helix is not seen in our electron density maps. It is possible that this large conformational change also disrupts the other activation function on the protein, found in the large N terminal domain which is not a part of the LBD structure. Partial agonism in ER-^ The plant phyto-oestrogen genistein is one of a number of compounds that act as full agonists for ER-a but are only partial agonists in ER-P. In addition, there is some evidence that ER-p is more readily antagonised than ER-a by a series of compounds. The structure of ER-P complexed with genistein shows that helix 12 adopts an unusual conformation, halfway between the agonist and antagonist positions seen for E2 and RAL respectively complexed to ER-a Taken together, these results suggest that helix 12 is more easily displaced in ER-a than in ER-p.

55 Co-activator recruitment The alternate binding location for helix 12 seen in the ER-a RAL structure masks key amino acids (in particular Lys 362) which are implicated in co-activator recruitment. In addition, the buried surface of helix 12 has similar features to the LXLL motif identified as the nuclear receptor binding region on coactivators (the so-called NR-box). Structures of agonist bound LBD, complexed to representative NR-box peptides, have confirmed that the NR-box peptides adopt a helical conformation and a binding site very similar to that adopted by helix 12 in the ER RAL structure. Concluding Remarks The published structures of the ER LBD a and (3 forms, in complex with a variety of ligands, provide a structural rationale for the measured agonist and antagonist properties of the ligands, and have identified the principle co-activator binding site on the receptor. However, there are many additional features of receptor function that remain to be explained. This will require structure determinations of larger constructs of the receptor (to include the DNA binding domain and the N terminal A/B domain), possibly complexed to larger fragments of co-activators. Acknowledgements This work was funded by Karo Bio Inc and the infrastructure of the Structural Biology Laboratory at York is supported by the BBSRC. The main contributors to the work were Ashley C.W. Pike, Andrzej M. Brzozowski, Julia Walton of York, together with Tomas Bonn\ Jan-Ake Gustafsson^ and Mats CarlquistV ^Karo Bio AB, NOVUM, S14157 Huddinge, Sweden. ^Departments of Medical Nutrition and Biosciences, Karolinska Institute, NOVUM, S-14186 Huddinge, Sweden.

Bibliography Brzozowski, A.M., Pike, A.C.W., Dauter, Z., Hubbard, R.E., Bonn, T., Engstrom, O., Ohman, L., Greene, G.L., Gustafsson, J.A. and Carlquist, M. (1997), "Molecular basis of agonism and antagonism in the oestrogen receptor". Nature, 389, 753-758 Pike, A.C.W., Brzozowski, A.M., Hubbard, R.E., Bonn, T., Thorsell, A.G., Engstrom, O., Ljunggren, J., Gustafsson, J.K. and Carlquist, M. (1999) "Structure of the ligandbinding domain of oestrogen receptor beta in the presence of a partial agonist and a full antagonist" Embo J, 18 4608-4618 Hubbard, R.E., Pike, A.C.W., Brzozowski, A.M., Walton, J., Bonn, T., Gustafsson, J.A. and Carlquist, M., (2000), "Structural insight into the mechanisms of agonism and antagonism in oestrogen receptor isoforms", Eur. J. Cancer, 36, S17-S18 Pike, A. C. W., Brzozowski, A. M., Walton, J., Hubbard, R. E., Bonn, T., Gustafsson, JA. and Carlquist, A. M. (2000), "Structural aspects of agonism and antagonism in the oestrogen receptor", Bioch. Soc. Trans. 28, 396-400

56 Pike, A. C. W., Brzozowski, A. M. and Hubbard, R. E. (2000)" A structural biologist's view of the oestrogen receptor", J of Steroid Bioch and Mol. Biol 74, 261-268 Pike, A. C. W., Brzozowski, A, M., Walton, J., Hubbard, R. E., Thorsell, A.-G., Li, YL, Gustafsson, J.-A. and Carlquist, M. (2001) "Structural insights into the mode of action of a pure antiestrogen". Structure, 9, 145-153

H. van der Goot (Editor) Trends in Drug Research III © 2002 Elsevier Science B.V. All rights reserved

57

Molecular Docking and Dynamics Simulations in the Ligand Binding Domain of Steroid Hormone Receptors Milou Kouwijzer-f- and Jordi Mestres§ i Molecular Design & Informatics, N.V. Organon, 5340 BH Oss, The Netherlands § Computational Medicinal Chemistry, Organon Laboratories Ltd., Newhouse, Lanarkshire MLl 5SH, Scotland

Introduction During the process of protein-ligand recognition, the structures of both the protein and the ligand undergo conformational changes as a result of a mutual induced fit aiming at optimising their interaction. Although ligand flexibility is now efficiently accounted for in docking calculations [1-8], proper treatment of protein flexibility still remains a critical issue for docking methods due to the complexity of sampling and scoring and the amount of computer power required [9-17]. The inability to account for protein flexibility can sometimes have a significant effect on the results of docking calculations. On one hand, it is known from crystal structures of proteins cocrystallized with different ligands that the backbone and/or the side chains of the residues in the binding pocket adapt to the ligand bound. This situation is illustrated in Figure la for two crystal structures of the estrogen receptor a (ERa). On the other hand, there is an intrinsic ambiguity in the orientation of heteroatoms for some residues (particularly threonine, histidine, glutamine and asparagine) due to the fact that differences in the electron density of carbon, nitrogen, and oxygen atoms are often difficult to appreciate. This situation is illustrated in Figure lb for the two independent molecules in the crystal structure of the progesterone receptor. It is thus clear that the result of a docking calculation depends on such details in the protein structure, and that scoring functions cannot correct for the inflexibility of the protein. One way of taking into account the flexibility of a protein is using molecular dynamics (MD) simulations. Because of the computer time requirements, it is not applicable for large numbers of ligands but it can be used for binding mode assessment activities. The aim of the present work is to evaluate the performance of MD simulations when applied to the ligand binding domain of steroid hormone receptors and compare the results with those obtained from flexible ligand docking methods that make use of a single rigid crystal structure.

58

Figure 1. Side chain adjustments and ambiguities in the binding pockets of steroid hormone receptors, a (left): superposition of ERa cocrystallized with estradiol [23] (light gray) or DBS [24] (dark gray), b (right): superposition of two PR structures [26] (molecule A in light gray and B in dark gray). The choice of a single rigid structure for docking purposes will clearly affect the results.

Methodology For the MD simulations, only the ligand and the residues that have at least one atom within 8 A of the ligand were allowed to move (i.e. -800 out of -4000 atoms). From a 100 ps run at 400 K (after a heating phase of 10 ps) coordinate sets were saved every picosecond. After the MD simulation, these coordinate sets were energy minimized (still taking into account the constraints) and the average interaction energy from these 100 frames was calculated. The orientation having the lowest average interaction energy was taken as the most favorable orientation. All calculations were performed with QUANTA/CHARMm [18] and charges were taken from templates. A 15 A cutoff radius was used, with a shift and a switch function between 11 and 14 A for the Coulomb and van der Waals energy, respectively. For the minimizations and MD runs, a distance-dependent dielectric constant 8 of 4r was used, whereas for the calculation of the interaction energy, 8 was set to r [19]. Docking calculations were performed with DOCK 4.0 [4]. Sphere centers were generated with SPHGEN [20]. For computational reasons, initial sphere sets were reduced to a number of around 30, ensuring appropriate coverage of the binding pocket. Energy (AMBER [21]) and chemical scoring grids of 0.3 A were generated using a united atom model and an 8 of 4r with a 10 A cut-off radius. Mol2 files of the ligands were generated with SYBYL [22] using Gasteiger charges. Virtual screening calculations were performed using uniform sampling with a maximum of 50 orientations and 5 seeds, whereas higher sampling values of 500 orientations and 25 seeds were set for binding mode

59 calculations. The minimal anchor size was set to four atoms and the maximum number of steric clashes between protein and ligand was set to three. Energy/chemical score minimization of maximally 100 iterations to a convergence of 0.1 kcal/mol was performed to each docked ligand. A database of 1000 diverse compounds selected from the Available Chemicals Directory (ACD) was used for virtual screening purposes. For validation purposes, MD simulations in the ligand-binding domains of steroid hormone receptors with their co-crystallized ligands were performed, namely, ERa__lere (estrogen receptor a, structure from pdb code lere) with estradiol [23], ERa_3erd (structure from pdb code 3erd) with diethylstilbestrol (DES) [24], ERp with genistein [25], and the progesterone receptor (PR) with progesterone [26]. In addition, MD simulations of ERa structures with the ligand that was co-crystallized in another structure were also performed, namely, ERa^lere with DES, and ERa_3erd with estradiol. The former calculation was repeated for 1000 ps at 300 K to validate the high temperature used in the simulations. The aim of these calculations was mainly to assess the effect and the extent of the accommodation of the binding pocket to a ligand other than the native co-crystallized ligand. Finally, as an application example, a prediction of the orientation of genistein in the ERa binding pocket based on MD simulations is presented (the crystal structure of the ERa-genistein complex has not yet been determined experimentally). Due to the high computer time requirements for the MD simulations, sampling of ligand orientations in the binding pocket was limited to the most probable orientations. The four starting orientations for the ligands containing a steroidal core are represented below, orientation 1 being the one observed in the crystal structures containing estradiol [23] and progesterone [26]. Analogous orientations were used for the non-steroidal ligands, DES and genistein.

1

2

3

4

Final structural details: in the ERo/p receptors, the proton of the (uncharged) histidine residue close to the steroidal D ring was placed on Ne2 and, in the PR structure (molecule A), the labels of atoms O E I and NE2 in Gln-725 were exchanged, in accordance with the hydrogen bonding scheme published [26] and the second molecule in the asymmetric unit.

60 Molecular docking In order to assess to which extent the use of a single rigid crystal structure affects the final results of a docking calculation, the structures of ERa cocrystallized with estradiol [23] and DES [24] were used in a virtual screening exercise. As shown above in Figure 1, the binding pocket of the two structures differs mainly in the conformation of four residues, namely, Met343, Thr347, Met421, and His524. While DES pushes away Met343 and His524 with respect to the position they adopt with estradiol, estradiol induces a significant conformational change to Met421 with respect to the position it adopts with DES. More interestingly, while in lere the hydroxy 1 of Thr347 points towards the ligand cavity, in 3erd it points away from the ligand cavity. All these conformational changes will define essentially a different accessible space and interaction pattem in the binding pocket and thus result potentially in a different ranking for the molecules in a database. The correlation between the energy and chemical score ranks obtained by docking 1000 diverse molecules from the ACD in the two rigid ERa structures (lere and 3erd) is presented in Figure 2. As can be visually observed, a better correlation between molecule ranks is found when using the chemical score (r^=0.82) than when using the energy score (r^=0.67). In this respect, the chemical score seems to be less dependent than the energy score on the conformational changes of the residues defining the binding pocket. This is a particularly interesting statement that would require though further investigation to other protein binding pockets to assess its generality. In conclusion, it has been shown that the use of different single rigid structures does have a significant effect in the final outcome of docking calculations and that efficient treatment of protein flexibility during the course of a docking calculation would be highly desired. BOO

'."^.s-S&l^'

••••••.• • • •••

700 600

••"••.Js^^A-

111 500 Ul •- 400

k'^-^"'0

100

200

300

'••• 300

400 3ER0

SOO

200

600

700

800

»30

100 0<

pr\. }

100

•': ••••-% • 200

300

•. 400

500

600

700

800

90

3ERD

Figure 2. Correlation between the energy score ranks (left) and the chemical score ranks (right) obtained by docking 1000 diverse molecules from the ACD in the ligand binding domains defined by the ERa crystal structures with PDB entries lere and 3erd.

61 Molecular dynamics simulations Simulations of the ERa ligand-binding domain as defined by the PDB entry 3erd crystal structure with estradiol, DES, and genistein were recently reported in the literature [27]. In the work presented here, MD simulations of the ERa, ERP, and PR ligand-binding domains were performed and the results are discussed in the remainder of the paper. The first validation test is to check whether the experimentally observed binding mode is found as the most favorable orientation from MD simulations of a protein crystal structure with its cocrystallized ligand. The results of this exercise are presented in Figure 3. It can be seen that in all cases orientation 1 (see above for orientation numbering), which corresponds to the orientation observed in the crystal structure, is indeed the most favorable. This shows that MD simulations are capable of discriminating the correct from alternative orientations of the ligand in the binding pocket. Only for DES in ERa__3erd, differences between alternative solutions are small, in agreement with the highly symmetrical nature of this ligand. For the other cases, orientation 1 is clearly preferred over alternative orientations. The next step is to check whether the experimentally observed binding mode is reproduced from MD simulations when a ligand is docked in the binding pocket of a non-native protein crystal structure. Figure 4 shows the results for the binding of estradiol and DES in their counterpart crystal structures. Again, the experimentally observed orientation is found to be the most favorable, although the amino acid side chains in the binding pocket were initially not optimally adapted to the ligands. This proves that the conformations of the residues in the binding pocket can indeed accommodate for other ligands than the one with which the crystal structure was determined. When the MD simulation for the ERa_lere complex with DES was run for 1000 ps at 300 K the results did not significantly differ from those obtained from that for 100 ps at 400 K.

62 Estradiol in ERa (1ere)

Progesterone in PR (1a28) 1

1

I

o

I LU V

.1

•45

-45

cd

1

•55

i

-55

-65

65

-75

7C

i

1

1

-

t 1

•

1

1 2 3 4 Genistein in ERp {1ql<m)

1 2 3 4 Diethylstilbestrol in ERa (3erd)

O

E cd

LU V

1 2 3 4 orientation number

-75

1 2 3 4 orientation number

Figure 3. Average interaction energies obtained from MD simulations of different ligands in four orientations in the ligand-binding domain of steroid hormone receptors. A comparison of the results obtained from MD and DOCK calculations for the docking of estradiol and DES in their native and non-native protein structures (lere and 3erd) is presented in Table 1. RMSD values are calculated between the most favorable orientation found with MD or DOCK and the experimentally determined ligand orientation. As can be observed, results for the two native protein-ligand structures (estradiol in lere and DES in 3erd) are reasonably accurate from both methods (RMSD<0.60 A), MD giving slightly lower RMSD values than DOCK (0.05 A for estradiol and 0.28 A for DES). However, MD results for the two non-native protein-ligand structures (DES in lere and estradiol in 3erd) are significantly better than DOCK results (0.74 A for estradiol and 0.32 A for DES). The poorer results obtained by DOCK in the non-native cases are indicative of docking calculations using rigid protein structures.

63 Estradiol in ERa (3erd) 1

Diethylstilbestrol in ERa (lere)

1

-45

-45 h L

o

E Iri o c

UJ"

-55 -65

V 7 C

-

T

-if 1

1

-55 -65

1

-75

1 2 3 4 orientation number

1 2 3 4 orientation number

Figure 4. Average interaction energies from MD simulations of estradiol in ERa_3erd and DES in ERa_lere in the same four orientations as in Figure 3.

Table I. RMSD (A) for the non-hydrogen atoms between the optimal ligand orientation obtained from MD or DOCK and the experimentally determined ligand orientation. Plain and italic numbers correspond to native and non-native protein-ligand structures, respectively. ERa_.lere

ERa_3erd

MD

DOCK

MD

DOCK

Estradiol

0.52

0.57

0.44

0.76

DES

0.34

1.08

0.32

0.60

Simulations on the preferred orientation of genistein in the ligand-binding domain of ERa were also performed. In the crystal structure of genistein bound to ERP [25], helix 12 adopts the antagonistic conformation. In the present MD simulations, helix 12 in ERa was retained in the agonistic conformation (as in lere and 3erd). The basis for this assumption is twofold: first, genistein is a partial agonist in ERp but, given its rather small size, it is very well possible that helix 12 in ERp with genistein bound can adopt both agonistic and antagonistic conformations [25] and, second, although the affinity of genistein for ERa is lower than for ERp, it appears to be a full agonist (even a slight superagonist) in ERa [28].

64 The results of the MD simulations of genistein in four orientations in ERa are shown in Figure 5. As can be observed, the results are consistent with orientation 4 being the preferred orientation, irrespective of the ERa crystal structure used originally. However, interestingly enough, the predicted binding mode of genistein in ERa (orientation 4) differs from the experimentally observed binding mode in ERP (orientation 1), as illustrated in Figure 6. Since the crystal structure of the ERa-genistein complex is not available, the scope of this prediction cannot be established yet. However, since the results from both ERa_lere and ERa_3erd agree and since the same methodology applied to genistein in ERp gives indeed the correct result, we are confident that the present prediction may have a high degree of reliability.

Genistein in ERa (1ere) 1

1

1

Genistein in ERa (3erd) 1

1

-45

-45

Id -55

-55

lij" -65

-65

1

1

1

1

1

"o

g V

-75

1

1

1

1

1 2 3 4 oritentation number

-75

I

I

1 2 3 4 orientation number

Figure 5. Average interaction energies from MD simulations of genistein in ERa.

65

Figure 6. Left: experimentally determined orientation of genistein in ERp [25], including hydrogen bonding partners (Glu-305, Arg-346, a water molecule, and His-475). Right: predicted binding orientation in ERa (including hydrogen bonding partners Glu-353, Arg-394 and His-524; water was not included)

Conclusions One of the remaining problems with docking ligands in receptors is the flexibility of the protein. Standard docking programs do not take protein flexibility into account in an efficient manner yet. The dynamical calculations described here do take protein flexibility properly into account, as shown by the correct reproduction of the experimentally determined ligand binding modes in the crystal structures of steroid hormone receptors. This is also the case when a ligand is simulated in the binding pocket of a crystal structure that was cocrystallized with another ligand. Of course, these calculations take much more computer time than standard docking programs (one orientation takes about 12 hours on a SGI octane with a RIOOOO processor). The methodology is therefore not to be used for large numbers of ligands. However, it can be highly valuable in structure-based drug design problems, when one is interested in the binding mode of a small series of structurally diverse ligands for which the binding mode is not known. Finally, the relevance of the prediction that the binding mode of genistein in ERa is different than in ERP will have to await experimental confirmation.

References [1] M. Rarey, B. Kramer, T. Lengauer, G. Klebe, 7. MoL Biol. 261 (1996) 470. [2] W. Welch, J. Ruppert, A.N. Jain, Chem. Biol 3 (1996) 449. [3] G. Jones, P. Willett, R.C. Glen, A.R. Leach, R. Taylor, J, MoL Biol, 267 (1997) 727. [4] T. Ewing, I.D. Kuntz. /. Comput. Chem. 18 (1997) 1175.

66 [5] G.M. Morris, D.S. Goodsell, R.S. Halliday, R. Huey, W.E. Hart, R.K. Belew, A.J. Olson, J. Comput. Chem. 19 (1998) 1639. [6] S. Makino, I.D. Kuntz. J. Comput. Chem. 19 (1998) 1834. [7] C.A. Baxter, C.W. Murray, D.E. Clark, D.R. Westhead, M.D. Eldridge, Proteins 33 (1998) 367. [8] J. Wang, P.A. KoUman, I.D. Kuntz, Proteins 36 (1999) 1. [9] A.R. Leach, J. Mot. Biol. 235 (1994) 345. [10] R.M.A. Knegtel, I.D. Kuntz, CM. Oshiro, J. Mol. Biol. 266 (1997) 424. [11] J. Apostolakis, A. Pliickthun, A. Caflisch, J. Comput. Chem. 19 (1998) 21. [12] B. Sandak, H.J. Wolfson, R. Nussinov, Proteins 32 (1998) 159. [13] V. Schnecke, C.A. Swanson, E.D. Getzoff, J.A. Trainer, L.A. Kuhn, Proteins 33 (199S) 14. [14] M. Zacharias, H. Sklenar, J. Comput. Chem. 20 (1999) 287. [15] J.Y. Trosset, H.A. Scheraga, J. Comput. Chem. 20 (1999) 412. [16] M. Mangoni, D. Roccatano, A. Di Nola, Proteins 35 (1999) 153. [17] A.C. Anderson, R.H. O'Neil, T.S. Surti, R.M. Stroud, Chem. Biol. 8 (2001) 445. [18] CHARMm version 25.2, QUANTA 98, Molecular Simulations Inc., San Diego, USA. [19] P.D.J. Grootenhuis, P.J.M. van Galen, Acta Cryst. D 51 (1995) 560. [20] I.D. Kuntz, J.M. Blaney, S.J. Oatley, R. Langridge, T.E. Ferrin, J. Mol. Biol. 161 (1982) 269. [21] S.J. Weiner, P.A. KoUman, D.T. Nguyen, D.A. Case, J. Comput. Chem. 7 (1986) 230. [22] SYBYL version 6.7, Tripos Inc., St. Louis, USA. [23] A.M. Brzozowski, A.C.W. Pike, Z. Dauter, R.E. Hubbard, T. Bonn, O. Engstrom, L. Ohman, G.L. Greene, J.-A. Gustafsson, M. Carlquist, Nature 389(1997)753. [24] A.K. Shiau, D. Barstad, P.M. Loria, L. Cheng, P.J. Kushner, D.A. Agard, G.L. Greene, Cell 95 (1998) 927. [25] A.C. Pike, A.M. Brzozowski, R.E. Hubbard, T. Bonn, A.G. Thorsell, O. Engstrom, J. Ljunggren, J.-A. Gustafsson, M. Carlquist, EMBO J. 18 (1999)4608. [26] S.P. Williams, P.B. Sigler, Nature 393 (1998) 392. [27] B.C. Oostenbrink, J.W. Pitera, M.M.H. van Lipzig, J.H.N. Meerman, W.F. van Gunsteren, / Med. Chem. 43 (2000) 4594. [28] Barkhem T, Carlsson B, Nilsson Y, Enmark E, Gustafsson J & Nilsson S. Mol. Pharmacol. 54(1998) 105-112.

H. van der Goot (Editor) Trends in Drug Research III © 2002 Elsevier Science B.V. All rights reserved

Peroxisome Proliferator-Activated Endocrinology

67

Receptors

and

Reverse

Steven Kliewer GlaxoSmithKline Five Moore Drive Research Triangle Park, NC 27709

1. Orphan Nuclear Receptors and Reverse Endocrinology Members of the nuclear receptor family, which includes the steroid, retinoid, and thyroid hormone receptors, function as Ugand-activated transcription factors and have critical roles in nearly every aspect of vertebrate development and adult physiology (1). Family members share several structural features including a highly conserved DNA binding domain (DBD) with two C4 zinc fingers. The DBD targets the receptor to short stretches of DNA, termed hormone response elements in the regulatory regions of target genes. The carboxy-terminal region of the nuclear receptors includes the ligand binding domain (LBD). The binding of hormone to the LED results in conformational changes that effectively flip the receptor "on," allowing it to interact with coactivator proteins and components of the transcriptional machinery and activate the expression of target genes. In addition to the receptors with known hormones, the nuclear receptor family includes many members that lack identified hormones. The identification of these so-called "orphan receptors" suggested that there might be additional hormones yet to be discovered and opened the era of "reverse endocrinology" (2, 3). Historically, novel endocrine signaling pathways were discovered based upon their effects on physiological and/or developmental processes. The isolated hormone was then used to identify its partner receptor. In reverse endocrinology, this path is inverted. The orphan receptor is used to screen for novel small molecule ligands, either natural or synthetic, that modulate its transcriptional activity. The ligand, in turn, is used as a chemical tool to dissect the role of the receptor in physiology and pathophysiology. Over the past several years, reverse endocrinology has been used to link a number of orphan receptors to ligands and biological activities. Among the orphan nuclear receptors for which ligands have been identified and used to unravel their biology are the peroxisome proliferator-activated receptors (PPAR) as fatty acid and eicosanoid receptors; the liver X receptors (LXR) as oxysterol receptors; the famesoid X receptor (FXR) as a bile acid receptor; and, the pregnane X receptor (PXR) as a xenobiotic and bile acid receptor.

68 2. PPARs In 1990, Issemaim and Green reported the cloning and initial characterization of a novel murine orphan nuclear receptor that was activated by a group of compounds that cause dramatic increases in the size and number of peroxisomes — organelles that are involved in the P-oxidation of long-chain chain fatty acids — in rodent liver (4). They named this receptor peroxisome proliferator-activated receptor a (PPARa). Among the compounds that activate PPARa are the fibrates — drugs, which are used clinically to lower triglycerides and raise high density lipoprotein cholesterol (HDL-c), the so-called "good" cholesterol," in the treatment and prevention of coronary artery disease. It is now well established that fibrates exert their therapeutic effects through activation of PPARa. In addition to PPARa, two closely related orphan nuclear receptors encoded by distinct genes were subsequently cloned and named PPARy and PPAR5, though neither of these receptors is activated efficiently by the classical peroxisome proliferators (5). Each of the three PPAR subtypes is expressed in a distinct tissuespecific pattern, suggesting that they subserve distinct biological roles. PPARa is highly expressed in liver, heart, kidney, skeletal muscle, and brown adipose, tissues which are metabolically very active. PPARy is most highly expressed in adipose tissue, colon, and macrophages. In contrast to PPARa and PPARy, which are abundantly expressed in just a few tissues, PPARS is expressed in virtually all tissues at comparable levels. Like several other members of the nuclear receptor superfamily, the three PPARs bind to DNA as obligate heterodimers with the 9-cis retinoic acid receptors (RXRs). The PPAR/RXR heterodimers bind to two half sites of the consensus sequence AGGTCA organized as a direct repeat with a single nucleotide spacer, a so-called DR-1 motif Peroxisome proliferator response elements (PPREs) have been identified in the transcriptional regulatory regions of several genes involved in carbohydrate and lipid metabolism (5). Shortly after their cloning, it was noted that the PPARs are unusual in their ligand activation properties in two important respects. First, they are relatively promiscuous in their ligand activation compared to most other nuclear receptors such as the steroid hormone and retinoid receptors. The PPARs are activated by structurally-diverse amphipathic acids, including both natural and synthetic compounds. Second, the PPAR/RXR heterodimers are "permissive" for activation by 9-cis retinoic acid and synthetic RXR ligands through the RXR subunit of the heterodimer. Although the PPARa/RXR complex was the first permissive heterodimer to be characterized, we now know that heterodimers formed between RXR and several other nuclear receptors, including all three PPAR subtypes are permissive for activation by RXR ligands (2). Since 1993, our nuclear receptor group at GlaxoSmithKline has focused much of its effort on the identification of PPAR subtype-selective agonists and antagonists, and the use of these chemical tools to uncover the physiological roles of the PPARs and the therapeutic utility of PPAR ligands. This chapter will focus on our work relating to the PPARy and PPARS subtypes.

69 3. PPARy Regulates Glucose Homeostasis Adult onset (Type 2) diabetes mellitus afflicts more than 100 million people worldwide, and its prevalence is soaring in developed countries with high fat diets and sedentary lifestyles (6). Type 2 diabetes is characterized by a resistance of peripheral tissues, including skeletal muscle, liver, and adipose, to the actions of insulin. The development of insulin resistance is an early event in the progression of Type 2 diabetes. In the early 1980's, scientists at the pharmaceutical company Takeda in Japan reported the first of a promising new class of antidiabetic drugs, which were subsequently termed glitazones (7). These drugs lowered glucose levels in rodent models of insulin resistance without increasing insulin secretion from the pancreas. The glitazones were originally derived from the fibrate clofibrate, which was noted to have weak antihyperglycemic activity in rodents. Optimization of the glucose-lowering activity of clofibrate through a combination of traditional medicinal chemistry and in vivo pharmacology yielded the prototypical glitazone, ciglitazone (7). In addition to its antihyperglycemic effects, ciglitazone also reduced levels of insulin and lipids, including both triglycerides and non-esterified fatty acids. Subsequent optimization of these glitazone activities in animal models led to more potent members of this chemical class, including troglitazone, pioglitazone and rosiglitazone (8, 9). Rosiglitazone and pioglitazone are currently approved for use for Type 2 diabetes under the trade names Avandia and Actos, respectively. Troglitazone (Rezulin), the first of the glitazones to be approved by the FDA, was removed from the market due to idiosyncratic hepatotoxicity, which in extreme cases caused liver failure. In 1995, we reported that PPARy was the receptor for glitazones (10). Two observations led us to test whether glitazones modulated PPARy activity. First, PPARy was shown to be highly expressed in adipocytes and to serve as a critical regulator of fat cell differentiation in vitro (11, 12). Second, the glitazones had been reported to dramatically enhance insulin-dependent differentiation of various preadipocyte and stem cell lines to fat cells (13, 14). Among the genes regulated by glitazones in fat cells was the adipocyte fatty acid binding protein aP2, a gene that is regulated directly by PPARy (11, 15). We initially demonstrated that glitazones were selective PPARy ligands in vitro (10). We subsequently showed that the potencies of different glitazones in activating PPARy in vitro correlated closely with their glucose-lowering activities in vivo (16). Taken together, these findings provided compelling evidence that PPARy was the molecular target for the antidiabetic actions of the glitazones. At first blush, it appeared counterintuitive that a key regulator of fat cell differentiation would be the molecular target for these antidiabetic drugs. However, work from a number of laboratories has unequivocally established that PPARy is the molecular target for the antidiabetic actions of the glitazones (5, 8, 9). The glitazones were developed in rodent models of insulin resistance in the absence of any knowledge about their molecular target. Moreover, the glitazones contain a chiral center at carbon 5 of the thiazolidinedione headgroup, but have been developed as a

70 mixture of isomers (racemates) since they undergo racemization in vivo. With the knowledge of their molecular target, we quickly established that only the (S)enantiomers of the glitazones bind with high affinity to PPARy (17). Thus, only 50% of the drug substance in the currently approved glitazones is likely to be active. We also found that one of the glitazones, troglitazone, has significant activity on the nuclear receptor PXR, a key transcriptional regulator of CYP3A expression in the liver and intestine (18). Consistent with this observation, troglitazone was reported to interact with drugs metabolized by CYP3A, such as oral contraceptives. During the past five years, our group has focused much of its effort on the identification of antidiabetic drugs optimized for their activity on PPARy. Recently, we reported a series of tyrosine-based PPARy agonists, including GI262570, GW1929, and GW7845, that fulfill these criteria (19-21). These compounds are single enantiomers synthesized from the amino acid L-tyrosine and are extremely potent PPARy agonists, binding to the receptor at low nanomolar concentrations. Moreover, these PPARy agonists do not activate PXR. Importantly, GW1929 was shown to lower glucose levels in ZDF rats at serum plasma concentrations > 100-fold lower than troglitazone, which mirrors the differences in the potencies of these two compounds at PPARy in vitro (22). One of these compounds, GI262570, has subnanomolar activity on human PPARy and submicromolar activity on PPARa. GI262570 represents the first PPAR agonist optimized against the human receptor to be progressed into the clinic, where it is currently in phase III clinical trials for the treatment of Type 2 diabetes. The phase II clinical data show that GI 262570 has potent glucose-lowering effects and also lowers triglycerides and raises HDL-c in diabetes patients (23, 24). Although it is clearly established that PPARy modulates insulin sensitivity, a paradox remains: How does activation of a nuclear receptor that is highly expressed in fat sensitize skeletal muscle and liver to the actions of insulin? Several plausible explanations have been proposed. First, PPARy could be activated in the muscle and liver and regulate gene expression and insulin sensitivity directly in these tissues. However, there is very little PPARy expressed in either muscle or liver. Alternatively, activation of PPARy in fat cells could modulate the levels of signaling molecules that affect glucose metabolism in muscle and liver. Indeed, PPARy activation is known to alter the levels of several potential signaling molecules including tumor necrosis factor a, leptin, and free fatty acids (9). In order to better understand the mechanism underlying the antidiabetic effects of PPARy agonists, we recently examined the effects of the potent, selective PPARy agonist GW1929 in Zucker diabetic fatty (ZDF) rats, a standard animal model of Type 2 diabetes. There were two aspects to this study. First, we examined the temporal relationship in the decreases in serum glucose and lipid parameters. Second, we identified genes that were regulated by GW1929 in major insulin-sensitive tissues, including adipose tissue, liver, and skeletal muscle. ZDF rats were treated for 24 hrs, 48 hrs, or 7 days with the potent, selective PPARy agonist GW1929 or vehicle alone. As expected, GW1929 treatment resulted in marked decreases in serum glucose, triglyceride, and FFA levels at the 7 day time point (25). Notably, FFA levels were

71 significantly reduced as early as 24 hrs post treatment and continued to decline to the 7 day time point. By contrast, glucose and triglyceride levels were only decreased by GW1929 treatment at the 7 day time point. These data demonstrate that drops in FFA levels precede those in glucose and triglyceride levels, suggesting that FFAs levels may be an important component in the antidiabetic actions of PPARy agonists. We next soUght to identify genes regulated by PPARy in major insulin-sensitive tissues. RNA was prepared from epididymal white adipose tissue (WAT), soleus skeletal muscle, and liver of ZDF rats treated for 7 days with either GW1929 or vehicle alone. Genes whose expression was either increased or decreased in response to GW1929 were identified by a comprehensive and unbiased mRNA profiling technology called GeneCalling (26). In this technique, cDNA is generated from the tissue of interest, fragmented using different combinations of restriction enzymes, and subjected to high resolution electrophoresis. Fragments representing differentially expressed genes are identified or "called" from a sequence database based on the precise length of the fragment and the restriction enzymes that were used to generate it. The identities of differentially expressed genes are confirmed by competitive PCR using gene-specific oligonucleotides (26). Based on the fraction of GeneCalling fragments that were changed by treatment, we estimate that approximately 10%, 2%, and 1% of expressed genes were regulated by GW1929 in WAT, liver, and skeletal muscle, respectively. Thus, many more genes are regulated in fat than in other tissues. The genes and metabolic pathways that were regulated by GW1929 in these three tissues have been described in detail by Way et al. (25) and are discussed briefly below. WAT. GW1929 treatment resulted in a coordinate increase in the expression of a large number of genes involved in glucose and fatty acid homeostasis in WAT. Among these were many genes whose products are required for lipogenesis, including the dihydrolipoamide acyltransferase subunit of pyruvate dehydrogenase (PDH), acetylCoA carboxylase, which catalyzes the rate limiting step in long-chain fatty acid biosynthesis, fatty acid synthase, proteins of the pyruvate/malate cycle (citrate synthase, pyruvate carboxylase, tricarboxylate transport protein, ATP-citrate lyase, cytosolic malate dehydrogenase, malic enzyme), and proteins involved in fatty acid desaturation (stearyl-CoA desaturase) and triglyceride biosynthesis [phosphoenolpyruvate carboxykinase (PEPCK), long-chain acyl-CoA synthetase, acyl-CoA synthetase 5, and glycerol-3-phosphate acyltransferase]. Glycogen synthase expression was also markedly increased by GW1929 treatment. The increased expression of these genes is consistent with GW1929 stimulating the storage of glucose in adipocytes as either triglyceride or glycogen. GW1929 treatment also stimulated the expression of genes involved in different aspects of fatty acid metabolism in WAT, including lipoprotein catabolism (lipoprotein lipase), fatty acid transport (CD36, fatty acid transport protein, heart fatty acid binding protein), and fatty acid oxidation (carnitine palmitoyltransferase I, short-chain and long-chain acyl-CoA dehydrogenases, monoglyceride lipase). Uncoupling protein 3 (UCP3) expression was also increased by PPARy agonist treatment. Taken together, these changes in gene expression suggest that increased uptake, storage, and oxidation of

72 fatty acids in WAT contribute to the marked hypolipidemic actions of GW1929. GW1929 treatment also increased expression of the transcription factor ADDl/SREBPlc, which cooperates with PPARy in promoting adipocyte differentiation in vitro (27). Moreover, ADDl/SREBPlc regulates a program of genes involved in lipogenesis in mature adipocytes and may play a broad role in mediating the actions of insulin on genes involved in lipid and carbohydrate metabolism (28). Skeletal muscle. Nearly all of the genes regulated by GW1929 in skeletal muscle were decreased in expression. Notably, expression of pyruvate dehydrogenase kinase 4 (PDK4) was markedly decreased by GW1929 treatment. Since PDK4 phosphorylates and inactivates the PDH complex, thereby inhibiting oxidative glucose metabolism, this finding suggests a molecular basis for increased glucose utilization in muscle of PPARy agonist-treated animals. In agreement with this observation, PDH activity was recently shown to be reduced in ZDF rats and to be restored by troglitazone treatment (29). GW1929 treatment also resulted in a marked decrease in UCP3 expression and the coordinate repression of ten genes involved in fatty acid transport and oxidation. In sum, the changes in gene expression are in line with PPARy agonists tilting the ratio of energy substrate used by the muscle from fatty acid to glucose. Liver. Roughly equal numbers of genes were up-regulated and down-regulated in liver by GW1929. Notably, GW1929 treatment resulted in decreases in the expression of PEPCK, pyruvate carboxylase and glucose 6-phosphatase, which are required for hepatic gluconeogenesis. These changes in gene expression suggest a molecular basis for the finding that PPARy agonists reduce hepatic glucose production in vivo. In contrast, GW1929 treatment increased expression of several genes involved in lipogenesis and increased hepatic expression of glucokinase, which catalyzes a key step in glucose metabolism. In sum, our ZDF rat studies provide strong evidence that modulation of systemic FFA levels contributes to the glucose-lowering activity of PPARy agonists. We show that decreases in circulating FFA levels precede the drops in both glucose and triglyceride levels. Moreover, we show that GW1929 has marked effects on genes that are involved in fatty acid metabolism in multiple insulin sensitive tissues. Overall, our data are consistent with PPARy agonists promoting a repartitioning of FFAs from muscle and liver to WAT, where they are either metabolized or stored as either triglycerides or glycogen. Such a repartitioning of FFAs would be predicted to enhance glucose utilization in the muscle through the glucose-fatty acid (Randle) cycle (30). In this regard, the marked decrease in PDK4 expression by GW1929 treatment is particularly noteworthy. Since PDK4 phosphorylates and inactivates the PDH complex, its down regulation by GW1929 provides a direct mechanism linking PPARy activation to increased glucose oxidation in muscle. In liver, decreases in FFA levels would be predicted to decrease glucose production (31). In line with this prediction, we observed decreased expression of several genes involved in gluconeogenesis. Thus, PPARy activation results in the coordinate regulation of genes that cause increases in glucose utilization in muscle and decreases in glucose production in the liver.

73

4. PPAR5 Regulates Reverse Cholesterol Transport Unlike PPARa and PPARy, no drugs have been identified that work through PPAR5. Thus, much less is known about the biology of PPAR5 than the other two PPAR subtypes. It seemed likely that PPAR5 is involved in lipid metabolism since fatty acids activate the receptor and fibroblasts transfected with a PPAR8 expression vector become responsive to fatty acids as measured by the induction of early genes in the adipocyte differentiation cascade (32). In support of this idea, the Gonzalez laboratory reported that mice lacking functional PPAR5 have reduced adipose tissue mass compared to control animals (33). Scientists at Merck demonstrated that the leukotriene antagonist L-165041 activated PPAR8 and raised serum cholesterol levels in db/db mice (34). Although this compound is not selective for murine PPAR8 over murine PPARy, the pharmacological effect was attributed to activation of PPAR8 since neither serum glucose nor triglycerides were lowered at the same dose. These data suggested that PPAR8, like the other two PPAR subtypes, may also have important roles in lipid metabolism. In order to understand the biological role of PPAR8 and its therapeutic potential, we discovered a potent subtype-selective ligand by using combinatorial chemistry and structure-based drug design. This compound, termed GW501516, binds to PPAR8 with low nanomolar affinity and is > 1000-fold selective for PPAR8 over PPARa, PPARy, and other nuclear receptors (35). We used GW501516 as a chemical tool to probe the function of PPAR8 in a number of different bioassays. Recently, several different nuclear receptors have been shown to stimulate expression of the ATP-binding cassette Al (ABCAl) protein, which regulates cholesterol and phospholipid transport from cells (36). Through a process called reverse cholesterol transport, ABCAl transfers excess cholesterol from peripheral cells, including macrophage-derived foam cells, to HDL-c for transport to the liver and removal from the body. Patients with Tangier disease have been identified with loss-of-function mutations in the ABCAl gene. These patients suffer from low HDL-c levels, high triglyceride levels, and an increased incidence of cardiovascular disease (37). We examined whether GW501516 would modulate ABCAl expression in differentiated THP-1 macrophages in vitro. Notably, treatment of these cells with GW501516 resulted in an - 3-fold increase in ABCAl expression (35) As expected, stimulation of ABCAl expression was accompanied by a corresponding increase in the rate at which the macrophages were able to efflux cholesterol. In addition to macrophages, GW50r516 also increased ABCAl expression in human fibroblast and intestinal cell lines. These data suggest that PPAR8 may play a critical role in regulating reverse cholesterol transport in a variety of different tissues. In order to assess whether its activity on ABCAl expression and cholesterol efflux in vitro would translate to a therapeutic benefit in vivo, GW501516 was administered to obese rhesus monkeys in a dose escalation study (35). This colony of primates displays

74

many of the features of the human metaboHc syndrome X including central obesity, dyslipidemia, insulin resistance, hyperinsulinemia, and hypertension. Importantly, the dyslipidemia in these animals is manifest as low HDL-c and high triglyceride levels, much like in the human metabolic X syndrome. Treatment with GW501516 had dramatic effects on a number of important serum lipid parameters. In agreement with its effects on cholesterol efflux in vitro, HDL-c increased in a dose-dependent manner in primates treated with GW501516, with levels raised -80% relative to baseline at the highest concentration of drug. ApoAI and ApoAII, the two major lipoprotein constituents of HDL-c, were raised 43% and 21%, respectively, at this dose, indicating that GW501516 treatment results in an increase in the number of HDL-c particles, not simply an increase in the size of pre-existing particles. GW501516 treatment also resulted in an --30% decrease in LDL-c levels. Subclass analysis revealed a marked reduction in the smaller, more atherogenic LDL-c particles. Finally, GW501516 treatment produced an -50% decrease in fasting triglyceride levels and a dosedependent decrease insulin levels. All of these data are consistent with PPAR5 counteracting the principal features of metabolic syndrome X, suggesting that PPAR5 agonists may have utility in the treatment of human disease. More broadly, these findings firmly establish PPAR5 as a key regulator of lipid homeostasis in primates.

5. Natural PPAR Ligands A comprehensive understanding of PPAR physiology will require the identification of the natural ligands for these nuclear receptors. It was first shown that PPARa is activated by micromolar concentrations of a surprisingly diverse collection of fatty acids that vary in both chain length and degree of saturation (38). A search for natural PPARa ligands in fractionated human serum identified palmitic acid, oleic acid, linoleic acid, and arachidonic acid as naturally-occurring activators of this orphan receptor (39). Subsequent work showed that fatty acids also activate PPARy and PPAR6. More recently, the availability of high-affinity, synthetic radioligands for all three of the PPAR subtypes has provided the opportunity to address whether fatty acids bind directly to the PPARs. Many of these structurally-diverse fatty acids interact directly with the PPARs at concentrations in the low micromolar range (40). Although these concentrations are higher than those typically required for ligands to bind to their cognate nuclear receptors, they are consistent with the levels of nonesterified fatty acids found in human serum. Interestingly, the three PPAR subtypes have different fatty acid binding profiles. PPARa is the most promiscuous subtype, interacting efficiently with both saturated and unsaturated fatty acids. By contrast, PPARy is the most selective subtype, interacting efficiently with only a subset of the polyunsaturated fatty acids, including eicosapentaenoic and arachidonic acid. PPAR5 binds to both unsaturated and saturated fatty acids but with slightly lower affinity than PPARa. Thus, the PPARs are capable of interacting with multiple fatty acids in vitro. Certain oxidized fatty acid metabolites have also been shown to fiinction as PPAR ligands in vitro. Two of these eicosanoids, IS-deoxy-A^^'^'^-prostaglandin J2 (15d-PGJ2)

75 and 8(S)-hydroxyeicosatetraenoic acid (8(S)-HETE), were found to be slightly more potent activators of the PPARs than their polyunsaturated fatty acids precursors (4143). Interestingly, 15d-PGJ2 and 8(S)-HETE were subtype-selective in their interactions with the PPARs: whereas 15d-PGJ2 was selective for PPARy, 8(S)-HETE interacted preferentially with PPARa. More recently, the lipoxygenase products 9hydroxyoctadecadienoic acid (HODE), 13-HODE, and 15-HETE, which are components of oxidized low density lipoprotein, were shown to bind and activate both PPARa and PPARy (44). Thus, the regulated conversion of polyunsaturated fatty acids to eicosanoids through either the cyclooxygenase or lipoxygenase pathways may provide a mechanism for the modulating the activities of one or more of the PPAR subtypes. Given their important role in energy balance, the idea that the PPARs might serve as fatty acid receptors is attractive. The discovery that the PPARs are capable of binding to a variety of fatty acids and their metabolites suggests that their activation in vivo may not determined through interactions with a single, high-affinity ligand, like the classical steroid and thyroid hormone receptors, but rather through interactions with a number of fatty acids and fatty acid metabolites. Thus, the PPARs may function as generalized sensors of fatty acid levels, coupling fluxes in the levels of these fatty acids to the transcriptional regulation of genes involved in lipid and glucose homeostasis.

6. Summary Over the past decade, reverse endocrinology has resulted in an explosion in our understanding of metabolism and its regulation by nuclear receptors. Nowhere has the impact of reverse endocrinology been greater than with the PPARs and fatty acid metaboHsm. Our understanding of PPARa and PPARy has been dramatically enhanced based on the association of these one-time orphan receptors with the fibrate and glitazone class of drug, respectively. With the knowledge of their molecular targets in hand, new generations of PPAR agonists with improved efficacy and side-effect profiles can be developed for the treatment of diabetes and dyslipidemia. Indeed, GI262570 represents the first PPAR compound optimized against the human receptor to be progressed into the clinic. In the case of PPARS, the availability of potent, selective ligands has revealed the fundamental role this receptor has in lipid homeostasis and provided evidence that PPARS agonists may prove effective in the treatment of dyslipidemia and perhaps other aspects of the metabolic syndrome X. Ultimately, our increasing knowledge of PPAR target genes and mechanism of action coupled with the power of combinatorial chemistry will permit the discovery of PPAR modulators — molecules that specifically activate subsets of PPAR target genes in appropriate tissues. Such drugs will undoubtedly represent powerful tools for the future treatment of human metabolic disease.

76 7. Acknowledgments I acknowledge the contributions of my many colleagues at GlaxoSmithKline to the work presented in this chapter. Portions of this chapter are from edited versions of Kliewer et al. (45).

8. References 1.

2. 3. 4. 5. 6. 7.

8. 9. 10.

11.

12.

13.

14.

Mangelsdorf DJ, Thummel C, Beato M, Herrlich P, Schutz G, Umesono K, et al. The nuclear receptor superfamily: the second decade. Cell 1995;83(6):8359. Mangelsdorf DJ, Evans RM. The RXR heterodimers and orphan receptors. Cell 1995;83(6):841-50. Kliewer SA, Lehmann JM, Willson TM. Orphan nuclear receptors: shifting endocrinology into reverse. Science 1999;284(5415):757-60. Issemann I, Green S. Activation of a member of the steroid hormone receptor superfamily by peroxisome proliferators. Nature 1990;347(6294):645-50. Willson TM, Brown PJ, Stembach DD, Henke BR. The PPARs: from orphan receptors to drug discovery. J Med Chem. 2000;43(4):527-50. Saltiel AR. New perspectives into the molecular pathogenesis and treatment of type 2 diabetes. Cell 2001; 104(4):517-29. Sohda T, Mizuno K, Tawada H, Sugiyama Y, Fujita T, Kawamatsu Y. Studies on antidiabetic agents. I. Synthesis of 5-[4-(2-methyl-2-phenylpropoxy)benzyl]tliiazolidine-2,4-dione (AL-321) and related compounds. Chem Pharm Bulletin 1982;30(10):3563-73. Saltiel AR, Olefsky JM. Thiazolidinediones in the treatment of insulin resistance and type II diabetes. Diabetes 1996;45(12): 1661-9. Spiegelman BM. PPAR-gamma: adipogenic regulator and thiazolidinedione receptor. Diabetes 1998;47(4):507-14. Lehmann JM, Moore LB, Smith-Oliver TA, Wilkison WO, Willson TM, Kliewer SA. An antidiabetic thiazolidinedione is a high affinity ligand for peroxisome proliferator-activated receptor gamma (PPAR gamma). J Biol Chem 1995;270(22): 12953-6. Tontonoz P, Hu E, Graves RA, Budavari Al, Spiegelman BM. mPPAR gamma 2: tissue-specific regulator of an adipocyte enhancer. Genes Dev 1994;8(10): 1224-34. Tontonoz P, Hu E, Spiegelman BM. Stimulation of adipogenesis in fibroblasts by PPAR gamma 2, a lipid-activated transcription factor. Cell 1994;79(7): 1147-56. Sparks RL, Strauss EE, Zygmunt Al, Phelan TE. Antidiabetic AD4743 enhances adipocyte differentiation of 3T3 T mesenchymal stem cells. J Cell. Physiol 1991;146:101-109. Hiragun A, Sato M, Mitsui H. Preadipocyte differentiation in vitro: identification of a highly active adipogenic agent. J Cell Physiol 1988;134:124-130.

77

15.

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

27.

Kletzein RF, Foellmi LA, Harris PKW, Wyse BM, Clarke SD. Adipocyte fatty acid-binding protein: regulation of gene expression in vivo and in vitro by an insulin-sensitizing agent. Molec Pharmacol 1992;42:558-562. Willson TM, Cobb JE, Cowan DJ, Wiethe RW, Correa ID, Prakash SR, et al. The structure-activity relationship between peroxisome proliferator-activated receptor gamma agonism and the antihyperglycemic activity of thiazolidinediones. J Med Chem 1996;39(3):665-8. Parks DJ, Tomkinson NC, Villeneuve MS, Blanchard SG, Willson TM. Differential activity of rosiglitazone enantiomers at PPAR gamma. Bioorg. Med Chem Letters 1998;8(24):3657-8. Jones SA, Moore LB, Shenk JL, Wisely GB, Hamilton GA, McKee DD, et al. The pregnane X receptor: a promiscuous xenobiotic receptor that has diverged during evolution. Mol Endocrinol 2000; 14(l):27-39. Henke BR, Blanchard SG, Brackeen MF, Brown KK, Cobb JE, Collins JL, et al. N-(2-Benzoylphenyl)-L-tyrosine PPARgamma agonists. 1. Discovery of a novel series of potent antihyperglycemic and antihyperlipidemic agents. J Med Chem 1998;41(25):5020-36. Cobb JE, Blanchard SG, Boswell EG, Brown KK, Charifson PS, Cooper JP, et al. N-(2-Benzoylphenyl)-L-tyrosine PPARgamma agonists. 3. Structureactivity relationship and optimization of the N-aryl substituent. J Med Chem 1998;41(25):5055-69. Collins JL, Blanchard SG, Boswell GE, Charifson PS, Cobb JE, Henke BR, et al. N-(2-Benzoylphenyl)-L-tyrosine PPARgamma agonists. 2. Structureactivity relationship and optimization of the phenyl alkyl ether moiety. J Med Chem 1998;41(25):5037-54. Brown PJ, Winegar DA, Plunket KD, Moore LB, Lewis MC, Wilson JG, et al. A ureido-thioisobutyric acid (GW9578) is a subtype-selective PPARalpha agonist with potent lipid-lowering activity. J Med Chem 1999;42(19):3785-8. Fiedorek FT, Wilson GG, Frith L, Patel J, Abou-Donia M. Monotherapy with GI262570, a tyrosine-based non-thiazolidinedione PPARy agonist, improves metabolic control in Type 2 diabetes mellitus patients. Diabetes 2000;49:157. Wilson GG, Abou-Donia M, Frith L, Patel J, Fiedorek FT. Monotherapy with GI262570, a tyrosine-based non-thiazolidinedione PPARg agonist, significantly reduces triglyceride and increases HDL-C concentrations in patients with Type 2 diabetes mellitus. Diabetes 2000;49:525. Way JM, Harrington WW, Brown KK, Gottschalk WK, Sundseth SS, Mansfield TA, et al. Comprehensive messenger ribonucleic acid profiling reveals that peroxisome proliferator-activated receptor gamma activation has coordinate effects on gene expression in multiple insulin-sensitive tissues. Endocrinology 2001; 142(3): 1269-77. Shimkets RA, Lowe DG, Tai JT, Sehl P, Jin H, Yang R, et al. Gene expression analysis by transcript profiling coupled to a gene database query. Nature Biotechnol. 1999;17(8):798-803. Kim JB, Spiegelman BM. ADDl/SREBPl promotes adipocyte differentiation and gene expression linked to fatty acid metabolism. Genes Dev 1996;10(9):1096-107.

78 28. 29.

30.

31. 32.

33.

34.

35.

36.

37. 38.

39.

40.

41.

Flier JS, Hollenberg AN. ADD-1 provides major new insight into the mechanism of insuUn action. Proc Natl Acad Sci USA 1999;96(25):14191-2. Sreenan S, Keck S, Fuller T, Cockbum B, Burant CF. Effects of troglitazone on substrate storage and utilization in insulin-resistant rats. Am J Physiol 1119;276(6Ptl). Randle PJ. Regulatory interactions between lipids and carbohydrates: the glucose fatty acid cycle after 35 years. Diabetes Metab Rev 1998;14(4):26383. Bergman RN, Ader M. Free fatty acids and pathogenesis of type 2 diabetes mellitus. Trends Endocrinol Metab 2000;11(9):351-6. Bastie C, Hoist D, Gaillard D, Jehl-Pietri C, Grimaldi PA. Expression of peroxisome proliferator-activated receptor PPARdelta promotes induction of PPARgamma and adipocyte differentiation in 3T3C2 fibroblasts. J Biol Chem 1999;274(31):21920-5. Peters JM, Lee SST, Li W, Ward JM, Gavrilova O, Everett C, et al. Growth, adipose, brain, and skin alterations resulting from targeted disruption of the mouse peroxisome proliferator-activated receptor beta (delta). Mol Cell Biol 2000;20:5119-5128. Leibowitz MD, Fievet C, Hennuyer N, Peinado-Onsurbe J, Duez H, Bergera J, et al. Activation of PPARdelta alters lipid metabolism in db/db mice. FEBS Lett2000;473(3):333-6. Oliver WR, Jr., Shenk JL, Snaith MR, Russell CS, Plunket KD, Bodkin NL, et al. A selective peroxisome proliferator-activated receptor delta agonist promotes reverse cholesterol transport. Proc Natl Acad Sci U S A 2001;98(9):5306-5311. Repa JJ, Turley SD, Lobaccaro JA, Medina J, Li L, Lustig K, et al. Regulation of absorption and ABC 1-mediated efflux of cholesterol by RXR heterodimers. Science 2000;289(5484): 1524-9. Schmitz G, Langmann T. Structure, fimction and regulation of the ABCl gene product. Curr Opin Lipidol 2001; 12(2): 129-40. Gottlicher M, Widmark E, Li Q, Gustafsson JA. Fatty acids activate a chimera of the clofibric acid-activated receptor and the glucocorticoid receptor. Proc Natl Acad Sci USA 1992;89(10):4653-7. Banner CD, Gottlicher M, Widmark E, Sjovall J, Rafter JJ, Gustafsson JA. A systematic analytical chemistry/cell assay approach to isolate activators of orphan nuclear receptors from biological extracts: characterization of peroxisome proliferator-activated receptor activators in plasma. J Lipid Res 1993;34(9):1583-91. Xu HE, Lambert MH, Montana VG, Parks DJ, Blanchard SG, Brown PJ, et al. Molecular recognition of fatty acids by peroxisome proliferator-activated receptors. Mol Cell 1999;3(3):397-403. KHewer SA, Lenhard JM, Willson TM, Patel I, Morris DC, Lehmann JM. A prostaglandin J2 metabolite binds peroxisome proliferator-activated receptor gamma and promotes adipocyte differentiation. Cell 1995;83(5):813-9.

79 42.

43.

44.

45.

Forman BM, Tontonoz P, Chen J, Bmn RP, Spiegelman BM, Evans RM. 15Deoxy-delta 12, 14-prostaglandin J2 is a ligand for the adipocyte determination factor PPAR gamma. Cell 1995;83(5):803-12. Yu K, Bayona W, Kallen CB, Harding HP, Ravera CP, McMahon G, et al. Differential activation of peroxisome proliferator-activated receptors by eicosanoids. J Biol Chem 1995;270(41):23975-83. Nagy L, Tontonoz P, Alvarez JG, Chen H, Evans RM. Oxidized LDL regulates macrophage gene expression through ligand activation of PPARgamma. Cell 1998;93(2):229-40. Kliewer SA, Xu HE, Lambert MH, Willson TM. Peroxisome proliferatoractivated receptors: from genes to physiology. Recent Prog Horm Res 2001; 56:239-63.

This Page Intentionally Left Blank

H. van der Goot (Editor) Trends in Drug Research III © 2002 Elsevier Science B.V. All rights reserved

81

Biomolecular Mass Spectrometry related to Drug Research Albert J. R. Heck and Claudia S. Maier Department of Biomolecular Mass Spectrometry, Bijvoet Center for Biomolecular Research and Utrecht Institute for Pharmaceutical Sciences, Utrecht University Sorbonnelaan 16, 3584 CA Utrecht, The Netherlands Tel: (+31)30-2536797, Fax: (+31) 30-2518219, E-mail: [email protected] or [email protected] WEB: http://v^rsvw.chem.uu.nl/bioms/

New ionization methods in mass spectrometry Mass spectrometry (MS) has been around as an analytical tool for more than 50 years. However, the application of mass spectrometry to biomolecules, including larger drug related molecules was until recently rather limited because the analyte molecules need to be ionized, le. brought into the vapor phase and charged. With the introduction of new ionization techniques in mass spectrometry, such as matrix-assisted laser desorption (MALDI) [1] and electrospray ionization (ESI) [2], the field has seen a renaissance, particulary in the area of biomedical applications. Mass spectrometry is of course still a technique, which does not measure much more than the molecular weight of compounds or fragments of those compounds, but the new ionization techniques have broadened the applicability of this method so much that nowadays almost any compound may be analyzed. One of the advantages of mass spectrometry over other physio-analytical techniques used in structural biology and drug research is its great sensitivity. Nowadays, for many compounds a few femtomoles of sample is sufficient to determine its molecular weight and often also sufficient to get structural information through collision induced dissociation experiments. The recent expansion of biomolecular mass spectrometry is to a large extent due to its widespread use in sequencing of proteins and peptides for applications in proteomics and functional genomics [3-6]. In parallel, however, methodologies have been developed to use mass spectrometry in studying biomolecular interactions [7-9] enzyme-inhibitors [10], antibiotic-cell-wall target [11], drug-DNA and drug-RNA [12, 13], protein-carbohydrate [14] and protein-protein interactions [15-17]. The present chapter in this book reviews some of the contributions made by the authors' group to this emerging field. Noncovalent interactions and their role in drug function Noncovalent interactions between biomolecules are important determinants in molecular recognition processes. At the most basic level, the formation of secondary and tertiary structure during protein folding depends on the interplay of van der Waals'

82 forces, hydrophobic effects, hydrogen bond interactions and electrostatic interactions. Many proteins are only active as oligo- or multimeric protein complexes, whose subunit interactions are directed by noncovalent forces. Additionally, the formation, stability and function of complex protein assembUes, such as the sophisticated protein machinery of ribosomes and viruses, are based on noncovalent forces that dictate the interaction of their building blocks. Furthermore, the basfe of cellular communication is based on transient specific interactions of different biomolecules that act as molecular switches to trigger signal propagation in transduction pathways. Noncovalent interactions also play a pivotal role in the molecular origin of some human diseases in which the formation of fibril-like structures through self-association of misfolded proteins is thought to initiate the pathway that leads to the disease. The investigation of biomolecular interactions is an important and challenging problem in structural biology. Formation and composition of multimeric complexes are highly dependant on environmental constraints and, thus, suitable conditions that allow the study of noncovalent complexes are difficult to establish in vitro. Mass spectrometry is emerging as a powerful technique to study and characterize noncovalent complexes. The necessary requirement that a macromolecular sample has to be ionized and transferred intact into the gas-phase for mass spectrometric detection has been a considerable barrier hampering the earlier development of biomolecular mass spectrometry. The introduction of matrix assisted laser desorption ionization (MALDI) and electrospray ionization (ESI) has overcome these earlier limitations and provided breakthroughs in the determination of the molecular mass and primary structure of peptides, proteins and other biopolymers. In particular, electrospray ionization mass spectrometry has opened up opportunities for applications of mass spectrometry in the fields of analytical biochemistry and structural biology mainly because of its capability for analyzing proteins from aqueous solution under nearly physiological conditions of concentration, pH and temperature. Additionally, the efficient ionization and relatively gentle phase transfer of biomolecules and biomolecular assemblies from solution to the gas-phase realized by nano-electrospray ionization has enabled the intact mass spectrometric detection of multi-component complexes with solution dissociation constants (Ka) down to the micromolar range. Nevertheless, a protein in vitro and/or in solution is different from the same protein in the gas phase. The generated ions are likely to exhibit extra charges with no solvent molecules in their environment. Therefore, the use of mass spectrometry to probe biomolecular interactions is not self-evident. Proteins have traditionally been analyzed by ESI-MS using aqueous solutions containing an organic modifier, e.g. 10-50 % (v/v) methanol or acetonitrile and organic acids, e.g. 0.1-5 % formic or acetic acid. Under these conditions, ESI-MS allows mass determinations of proteins with high mass accuracy. However, since these conditions often cause protein denaturation they are not suitable for the investigation of noncovalent complexes. It is therefore essential to establish solution conditions that favor the maintenance of the noncovalent complex of interest. Solution conditions that often allow the preservation of specific noncovalent complexes and are compatible with the electrospray ionization process are aqueous

83

buffer solution at near physiological pH containing low concentration of volatile buffer salts, e.g. ammonium acetate. Over the last ten years, many studies have focused on how mass spectrometry can contribute to the investigation of biomolecular interactions. These studies have revealed that, using carefully designed approaches, biologically relevant parameters concerning noncovalent complexes and noncovalent interactions may be investigated by mass spectrometry. In this report several illustrative examples are described, mainly from the author's laboratory, highlighting the potential as well as some of the drawbacks of mass spectrometry-based approaches to characterize noncovalent biomolecular complexes. Drug-receptor interactions One of the first systems studied by bio-affinity mass spectrometry was the interaction of vancomycin-group antibiotics interacting with their receptor [18]. These clinically important group of antibiotics, of which vancomycin (Figure 1) is the most applied one, act by binding to bacterial cell-wall precursors terminating in the sequence -Lys-D-AlaD-Ala. This interaction inhibits cross-linking of the growing cell wall, leading to bacterial cell death [19]. To investigate this interaction the binding constants of these antibiotics with a variety of peptide cell-wall precursor analogs (e.g., N,N'-diAc-Lys-DAla-D-Ala) have been determined using various, quite laborious, methods including IH NMR spectroscopy [20], UV, microcalorimetry [21], and capillary electrophoresis [22]. If we consider the biomolecular association of vancomycin with a model peptide as V+L -> V*L, then, when a solution containing a mixture of V and L is introduced into a mass spectrometer the resulting spectrum may show ion signals not only due to V and L but also to V»L. A critical question is whether the ion signal of V»L indicates a specific interaction, or is caused by non-specific association in solution or during the ionization process. Figure 1. Chemical structure of vancomycin.

Ideally, the relative abundance of these three ion signals directly reflects the relative abundance/concentration of the species V, L and V«L in solution, as these may then be used to derive solution-phase association constants. It is critical that the assumption must be made that the ionization efficiency of at least the complex and either the free ligand or receptor is identical. Figure 2 shows an illustrative example of such an ESI spectrum. In this case vancomycin was combined with an equimolar mixture of three bacterial cell-wall mimicking peptides N-Ac-Gly-D-Ala, N-Ac-D-Ala-D-Ala and N-AcD-Ala-D-Ala-D-Ala. The spectruni displays the dominant ions in the mass spectrum; the doubly protonated ions of vancomycin and its noncovalent complexes with each of the three peptides. Although, the concentrations of the three peptides were equal in solution the three ions representing the noncovalent complexes are different in intensity.

84

By integrating the ion signals equilibrium concentrations can be [V + N-Ac-D-Ala-D-Ala-D-Ala] determined which subsequently may be Vancomycin used to calculate association constants. For the vancomycin system good agreement has been observed between the ESI-MS data and literature values [11, 23-25]. The advantages of the MS method lies not only in the sensitivity and efficiency, but also in the fact that m/z 750 800 850 900 MS may be used to multiplex the 700 technique, i.e. measuring more than one Figure 2. ESI mass spectrum of a mixture of vancomycin with three cell-wall receptor mimicking peptides. association constant at the time. To illustrate this advantage in more detail Figure 3 shows the ESI mass spectrum of 3 chemically related but modified vancomycin-group antibiotics, i.e. dechlorovancomycin (the phenyl-Cl has been replaced by a H) , demethylvancomysin (the Nterminal methyl has been replaced by a H) and vancomycin itself. In an equimolar mixture of the three they show in the ESI mass spectrum three equally intense doubly protonated ions. However, when a cell-wall mimicking peptide is mixed into the solution the three subsequently formed noncovalent complexes display strikingly dissimilar intensities. The differences are rather small, and, therefore, difficult to measure by alternative methods. The differences in intensities of the noncovalent complexes reveal that demethylvancomycin (clinically used in China) has a higher affinity for the cell-wall mimicking peptide than vancomycin. The determined affinity constants are given in Figure 3 as well. demethyl( vancomycin |m|f

(vancomycin y

Association constant Kj M''

dechloro- ^ vancomycin

i650

700

730 000 [6, 12]

vancomycin N-demethylvancomycin

900 000 [12]

dechlorovancomycin

350 000 [12]

"(vancomycin + 12)"

< 500 [15]

a 750

800

850

900

950

1000

1050

Figure 3. ESI mass spectrum of a mixture of vancomycin-group antibiotics with a cell-wall receptor mimicking peptides.

In another example, we showed recently that several vancomycin-group antibiotics (vancomcyin, eremomycin and avoparcin) undergo spontaneous chemical modifications when kept at room temperature in aqueous solutions containing traces of

85 formaldehyde or acetaldehyde at neutral pH [26]. This chemical modification leads predominantly to a mass increase of 12 Da in the reaction with formaldehyde and 26 Da in the case of acetaldehyde. This modification could be identified to originate from the formation of a ring-closed 4-imidazolidinone moiety at the N-terminus of the glycopeptide antibiotics, i.e. near the receptor binding pocket of the glycopeptide antibiotics. Figure 4 shows the ESI mass spectrum of a mixture of vancomycin and acetaldehyde-modified vancomycin, combined in solution with the cell-wall mimicking peptide N,N'-Ac2-Lys-D-Ala-D-Ala. It is clear from this spectrum that the acetaldehyde-modified vancomycin does not bind to the cell-wall mimicking peptide N,N'-Ac2-Lys-D-Ala-D-Ala as no noncovalent complex could be observed (expected at a m/z value indicated by the arrow). As a control, in the same experiment, the noncovalent complex of vancomycin with the cell-wall peptide was observed, as expected (see Figure 4). [vancomycin + 26 Da + 2H] 2

[vancomycin + 2H] 2+

[vancomycin + ac^-L-Lys-D-Ala-D-Ala + 2 H] 2+

IM^^

ijlllirt

Wft.iii)H^

700 800 900 1000 Figure 4. ESI mass spectrum of a mixture of vancomycin and acetaldehyde-modified vancomycin, combined with a cell-wall receptor mimicking peptide. The non-observation of a noncovalent complex of the modified vancomycin reveals that the modification blocks the interaction with the cell-wall receptor.

By using bio-affinity mass spectrometry we were able to reveal quite a few new aspects of the relation between the chemical structures of vancomycin-group antibiotics and their affinity for their natural receptor mimic D-Ala-D-Ala. Besides the subtle but significant effects observed when substituting only one atom in the vancomycin molecule (e.g. Cl/H) and the complete blocking of the affinity for the receptor mimic D-Ala-D-Ala when the N-terminus reacts via a Maillard-like reaction with formaldehyde [26], we have also reported on the ability of vancomycin-group antibiotics to noncovalently dimerize [24] (which is thought to promote antibacterial efficacy) and on the thermal degradation of vancomycin and avoparcin, which leads partly to stereo-isomerization of the antibiotics [27]. However, it should be noted that MS based techniques do not seem to work for all biomolecular associations. Especially, when the noncovalent interactions are primarily hydrophobic or when the class of ligands is not very homogeneous false positives and false negatives may be observed

86 [28]. Nevertheless, when used carefully bio-affinity mass spectrometry appears to be a fast and useful complementary technique to screen biomolecular interactions. Protein-protein and protein-drug interactions In recent years, ESI-MS has emerged as a rapidly progressing technique to study also biomolecular interactions of larger proteins due to its unique ability to directly analyze proteins and their noncovalent complexes from solution [7, 9, 10, 29]. In the electrospray ionization process of proteins multiply charged ions are generated, transferred intact into the gas phase, then introduced into the mass spectrometer and analyzed according to their mass-to-charge (m/z) ratios. Since proteins have multiple protonation/deprotonation sites, a typical ESI mass spectrum of a protein M contains a series of ion peaks (M+nH)°^ or (M-nH)°", in which each ion peak represents a population of protein molecules carrying n charges. One example that illustrates the use of mass spectrometry-based approaches to characterize a noncovalent protein-protein complex is the study of the enzyme Glyoxalase I [30]. Glyoxalase I is involved in the ubiquitous glyoxalase system, the major pathway that leads to detoxification of alphaketoaldehydes, such as methylglyoxal [31]. At present, there is considerable interest in studying the enzymes involved in the glyoxalase system because they are considered as potential targets for the development of novel anticancer and anti-malaria drugs. It is hypothesized that the intervention of the glyoxalase detoxification route leads to increased levels of cytotoxic methylglyoxal which ultimately may lead to cell death. The nature of the electrospray solution may (a) have a dramatic effect on the composition of the associated mass spectrum. To illustrate this, in Figure 5 ESI mass spectra are shown of Escherichia coli Glyoxalase I acquired from 7+ ammonium acetate solutions acidified to pH 3 2133 (Fig. 5A) and pH 5 (Fig. 5B), respectively. The spectrum at pH 3 yielded a mass spectrum 6+ JUL showing intense ion peaks in the m/z-range JL 1000-2000. Most of these ion peaks originate (b) from multiply protonated protein monomer, with the [M+9H]^'' and [M+SH]^"" being the most abundant ions. 10+

3(K)2

Figure 5. ESI mass spectra of Glyoxalase I at pH 3 (a) and pH 5 (b). The top spectrum displays mainly monomer species, whereas the bottom spectrum displays more dimer species (the latter are indicated by the black circles).

8+ 1867

'" -133

87 Deconvolution of the mass spectrum yielded a molecular mass of 14919 Da, in good agreement with the theoretical mass based on the known amino acid sequence [31]. In contrast, when Glyoxalase I is analyzed in an aqueous ammonium acetate solution adjusted to pH 5 the ESI mass spectrum displays a dramatically different charge state distribution. The dominant species is now the protein dimer, with the [2M+llHy^^ being the most abundant ion signal. A second minor charge state distribution (below m/z 2000) originates from monomeric protein species. For Glyoxalase I the stability of the dimeric complex is highly dependent on the solution pH. In case of Glyoxalase I, the mass spectrometric-derived complex stability versus solution-pH curve resembles closely the pH activity profile of Glyoxalase I in solution. These new techniques, which directly probe biomolecular interactions, are also suitable for the study of drug-target interactions. In order to evaluate whether mass spectrometry may play a role in the screening of potential inhibitors of Glyoxalase we studied the interaction of a small library of inhibitors by mass spectrometry [30]. Glutathione is an essential co-factor in the functional process of Glyoxalase I, and therefore several Salkyl-glutathions were chosen as potential inhibitors. Figure 6 shows mass spectra of Glyoxalase I, to which the potential inhibitor and S- {2-[3-(hexyloxy)benzoyl] vinyl} glutathione had been added at increasing concentrations (a to c). Figure 6. ESI mass spectra of Glyoxalase I with increasing concentrations of S-{2-[3-(hexyloxy)benzoyl]vinyl}-glutathione.

The spectra show intact noncovalent complexes of the protein-dimer complexed with 0, 1 or 2 inhibitor-molecules (as marked by a 0, 1 and 2). The molecular mass of these intact noncovalent protein-substrate/inhibitor complexes is exceeding 30 kDa. However, even at these high values the mass differences induced by binding of the small drug molecules (mass < 500 Da) can still be resolved as seen in Figure 6. The titration curves obtained from such data did provide an indication of the affinity between the protein and the potential inhibitor. Therefore, such data may be used for initial drug-screening, with advantages over other methods such as the sensitivity and speed of the MS-based analysis, but also the fact that the stoichiometry of the interaction is directly visualized. In this particular example a clear correlation could be made between the inactivation of the enzyme by the screened library of inhibitors, and their bioaffinity.

Very large macromolecular protein complexes: Vanillyl alcohol oxidase Several recent reports have shown that by ESI and MALDI very large biomacromolecules (e.g. DNA, viruses, protein-complexes) may be ionized and, therefore, analyzed by mass spectrometry [16, 32]. However, extraction of useful information from such data is often hampered by the limited mass resolution achieved in the spectra partly caused by the heterogeneity of the sample. To measure the mass of a biomacromolecular complex a primary requisite is that the detected charge states of the protein complex need to be resolved and identified. When homogeneous, mass spectrometry can nowadays measure quite accurately protein complexes of masses exceeding 1 million Dalton. To illustrate this, in Figure 7 are shown mass spectra of very large protein complexes of the Penicillium simplicissimum flavoprotein vanillylalcohol oxidase (VAO) [33, 34]. The VAO monomer has an average mass of 63,691 Da, including the covalently linked FAD. The bottom spectrum in Figure 7 shows the protein dissolved in a denaturing solution containing water/acetonitrile/formic acid. This spectrum may be interpreted as showing the VAO monomer picking up many charges (= protons). The maximum in the charge distribution is found around the [M+37H]^^^ ion. These ions still have mass-to-charge (m/z) ratios below 2000, an accessible range for most mass spectrometers. The recently determined crystal structure of VAO has however revealed that the enzyme is at neutral pH predominantly an octamer. In the top spectrum of Figure 7 an overview is given of the nanoflow ESI spectrum of VAO when sprayed from an aqueous ammonium acetate solution (pH = 5) at a VAO monomer concentration of 10 jiM. The spectrum displays a few clusters of ion signals. As the different charge-state signals are well resolved the number of charges and thus also the mass of the species could easily be derived allowing the interpretation of the spectrum. Several different protein assembhes can be identified in this ESI spectrum, each displaying relatively narrow charge state distributions. The most abundant species in the spectrum at the top of Figure 7 was found to be the VAO octamer, centered around a m/z value of 10,000. The molecular mass of this species as determined from the spectrum was 508,540 ± 1 5 0 Da. Because of the limited number of charges these noncovalent assemblies carry, giving the ions high m/z values, in particular time-offlight analyzers with their principally unlimited mass range are suited for the study of these complexes by mass spectrometry. Other assemblies observed appeared to be the VAO 16-mer for which a molecular weight of 1,017,100 ± 600 Da was calculated and even the 24-mer for which a mass of 1,525,600 ± 1,000 Da was obtained. The 24-mer ions displayed a narrow charge state distribution ranging from approximately 87 to 93. These ion signals appear around m/z = 17000, indicating that this protein assembly has picked up on average one proton per 150 amino-acids. The width (at half height) of each of the latter ion signals was approximately 30 m/z units. This renders to a width in the calculated mass of the 24-mer protein assembly of approximately 2,000 Da on an overall mass exceeding 1,500,000 Da, i.e. approximately 0.13%. These results indicate that mass spectrometry has at the present time grown into a technique that is able to detect very large molecular machineries.

89

non-denaturing solution condition, 50 mlVI NH4AC, pH 5

-5T+

II

. Octamer I . 0.51IVIDa ^iiti II

,

lit

IVIonomer (+FAD) 63.5 kDa denaturing solution conditicbn, I.e. 25% ACN, 0.05% FA

10000

12000

14000

Figure 7. ESI mass spectra of the protein vanillyl alcohol oxidase taken under native conditions (top) and denaturing conditions (bottom). The top spectrum shows primarily the intact octamer oligomer and some dimer, the bottom spectrum only the monomer.

Protein-membrane interactions Membrane proteins are biologically very important as they are responsible for crucial functions in the cell, like signal transduction, hormone reception and transport of proteins, nutrients and ions across cell membranes. Many human diseases ranging from cystic fibrosis to cancer are the result of mutant membrane proteins and some of the most widely used pharmaceuticals are directed at membrane protein targets. The way in which membrane proteins are embedded in the lipid bilayer and interact with surrounding lipids is of fundamental significance for membrane protein structure and function. The precise (interfacial) positioning of membrane proteins is important for accessibility of sites near the interface of for instance receptor proteins or channel forming proteins. The structural analysis of membrane proteins by mass spectrometry is traditionally hampered by the fact that membrane spanning regions of these proteins are strongly hydrophobic and therefore they are insoluble in aqueous solutions. To overcome this problem detergents may be used, however, they are often not compatible with mass spectrometry either. Recently, we introduced a new method in which we analyzed membrane spanning peptides incorporated in lipid bilayers directly by nanoflow electrospray mass spectrometry [35]. First of all this allowed the analysis of these normally water insoluble peptides by electrospray mass spectrometry. Additionally, by combining this direct proteo-liposome technique with H/D exchange

90

we were able to investigate the detailed architecture of these peptides when incorporated in the bilayer. The H/D exchange process of amide hydrogens with solvent deuteriums can only take place when these hydrogens are exposed to the solvent. For a membraneincorporated peptide the rate of H/D exchange is influenced by the extent of permeation of the solvent to the site of exchange in different regions of the bilayer, as well as by the participation of the amide hydrogens in the hydrogen-bonding network that defines secondary and tertiary structure. In our initial studies we investigated the peptides WALP16 and WALP16(+10). The WALP16 peptide consists of an alternating Leu/Ala hydrophobic core sequence flanked by tryptophan residues, which are found in many membrane proteins near the membrane/water interfacial region and are assumed to interact in a specific way with the membrane interface. Moreover, this hydrophobic core forms a transmembrane a-helix in phosphatidylcholine bilayers and, therefore, it may resemble a consensus transmembrane a-helical segment of integral membrane proteins. The a-helical length of WALP16 (23 A) is similar to the hydrophobic thickness of a l,2-dimyristoyl-5w-glycero-3-phosphocholine (DMPC) bilayer, which implicates that in this lipid system WALP16 is completely embedded in the hydrophobic region of the bilayer. The WALP16(+10) peptide has alternating Gly/Ala extensions at both termini of WALP16. These tails were made to provide hydrophilicity to penetrate into the water phase surrounding the bilayer and to adopt no secondary structure because of the ahelix breaking Gly residues. Since these hydrogens are therefore in the aqueous phase and not involved in hydrogen bonding they are expected to exchange more rapidly than those in the hydrophobic core of the peptide. . The hypothesized membrane incorporation of • t t ^ t t t ©ttit@# both peptides is schematically depicted in U U UmU M uf} UmU U U ^^S^^^ ^- ^^ ^^^^l W A L P 1 6 and WALP16(+10) lliiiiBiiM HHil 1111111 ^^^^ ^^ ^^^ ^^ exchangeable hydrogens, i##9®#i d i i y i i i respectively. However, when dissolved in D2O it was shown that even after several hours ^ Figures. Schematic structures of WALP 16 and WALP16(+10) incorporated in a membrane.

TTTATT»I^ WALP16

J and

^TT A T T»I ^ / . iri\ I_ J WALP16(+10) had

1 only

exchanged approximately 12 and 21 hydrogens, respectively. These finding were already consistent with the schematic depiction of the membrane incorporation shown in Figure 8, assuming that the 10 hydrogens in the membrane spanning hydrophobic core had not been exchanged. To test this hypothesis the partly exchanged peptides were submitted to collision induced dissociation (CID) experiments. Such experiments provide fragmentation patterns of the peptides, fi:om which the extent of deuteration per amino acid could be derived. Figure 9 shows, in a bar diagram, the determined deuterium content per peptide fi:agment of WALP16(+10). An advantage is that the CID spectra provide complementary sequencing information starting fi:om the C-terminus and the Nterminus. The observed flat region between amino-acid 10 and 18 indicates that almost no deuteriums are incorporated in this (transmembrane) part of the peptide. The solid line indicates the theoretical deuterium content as a function of the peptide-length

91 assuming that the hydrophobic core does not exchange at all, and that all other exchangeable hydrogens exchange at alike rates. Similar results were obtained for the shorter WALP16 peptide. All these finding were consistent with the simple model shown schematically in Figure 8. In further experiments we have recently tested the detailed interplay of the transmembrane peptides with the membrane by varying the length of the hydrophobic core, by moving the membrane anchoring tryptophan and by breaking the a-helix by introducing a proline. All these experiments have taught us that our novel proteoliposome nano-ESI-MS technique may be an extremely powerful and sensitive tool to analyze the exact positioning of membrane peptides in a lipid bilayer and to reveal site-specific information on peptide-membrane interactions. We believe that this method could be equally well applied to study the positioning of larger transmembrane proteins in model or biological membranes.

11

N

11

13 15 17 19 21 23 25

A fragment ion

C

13 15 17 19 21 23 25

Y" fragment ion

N

Figure 9. Deuterium content per peptide fragment of WALP16(+10). The flat region between amino-acid 10 and 18 indicates that almost no deuteriums are incorporated in this (transmembrane) part of the peptide.

Conclusions With the aid of new ionization methods and new mass analyzers mass spectrometry has evolved into one of the major tools in biomedical analysis and structural biology. Particulary, electrospray ionization has opened up novel ways to investigate biomolecular interactions by mass spectrometry. Even very large noncovalent complexes can now be transferred into the gas-phase and analyzed by mass spectrometry. In favorable cases association constants may be determined as well. Moreover, using these techniques, preferably combined with other methods (such as for instance hydrogen/deuterium exchange and limited proteolysis), the investigation of protein folding and protein-membrane interactions is today achievable by mass spectrometry. These developments therefore provide many innovative applications in drug research as illustrated by a few examples in the current report. Acknowledgements The work described here has been the effort of many people in the author- group together with many (inter)national collaborators. In particular I would like to acknowledge the people in my group who performed most of the work described here;

92 Jeroen Demmers, Claudia Maier, Nora Tahallah, Pauline Bonnici, Ellen Stokvis, Cees Versluis and our collaborators Dudley Williams and Thomas Staroske (Cambridge University, UK), Peter Roepstorff and Thomas Jorgensen (Odense University, Denmark), Susan Clugston and John Honek (University of Waterloo, Canada), Martin Wills (Warwick University, UK), Robert van den Heuvel and Willem van Berkel (Wageningen, The Netherlands) and Antoinette Killian and Ee^an Breukink (Utrecht University). References (1)

(2)

(3) (4) (5)

(6)

(7)

(8)

(9) (10)

(11) (12)

Karas, M.; Hillenkamp, F. Laser desorption ionization of proteins with molecular masses exceeding 10,000 daltons Anal. Chem. 60 (1988) 22992301. Fenn, J. B.; Mann, M.; Meng, C. K.; Wong, S. F.; Whitehouse, C. M. Electrospray ionization for mass spectrometry of large biomolecules Science 246(1989)64-71. Roepstorff, P. Protein sequencing or genome sequencing. Where does mass spectrometry fit into the picture? J. Protein Chem. 17 (1998) 542-543. Andersen, J. S.; M., M. Functional genomics by mass spectrometry FEBS Lett. 480(2000)25-31. Washburn, M. P.; Wolters, D.; Yates, J. R. Large-scale analysis of the yeast proteome by multidimensional protein identification technology Nat. Biotechnol. 19 (2001) 242-247. Banks, R. E.; Dunn, M. J.; Hochstrasser, D. F.; Sanchez, J. C ; Blackstock, W. P.; Pappin, D. J.; Selby, P. J. Proteomics: new perspectives, new biomedical opportunities Lancet 356 (2000) 1749-1756. Przybylski, M.; Glocker, M. O. Electrospray mass spectrometry of biomacromolecular complexes with noncovalent interactions-new analytical perspectives for supramolecular chemistry and molecular recognition processes Angew. Chem. Int. Ed. Engl. 35 (1996) 806-826. Smith, R. D.; Bruce, J. E.; Wu, Q.; Lei, Q. P. New mass spectrometric methods for the study of noncovalent associations of biopolymers Chem. Soc. Rev. 26 (1997) 191-202. Loo, J. A. Studying noncovalent protein complexes by electrospray ionization mass spectrometry Mass Spectrom. Rev. 16 (1997) 1-23. Ganem, B.; Li, Y.-T.; Henion, J. D. Observation of noncovalent enzymesubstrate and enzyme-product complexes by ion-spray MS J. Am. Chem. Soc. 113(1991)7818-7819. Jorgensen, T. J. D.; Roepstorff, P.; Heck, A. J. R. Direct determination of binding constants Anal. Chem. 70 (1998) 4427-4432. Gale, D. C ; Goodlett, D. R.; Light-Wahl, K. J.; Smith, R. D. Observation of duplex DNA-drug noncovalent complexes by ESI-MS J. Am. Chem. Soc. 116 (1994) 6027-6028.

93 (13) (14)

(15)

(16)

(17) (18)

(19) (20) (21)

(22)

(23)

(24)

(25)

(26)

Sannes-Lowery, K. A.; Hu, P.; Mack, D. P.; Mei, H.-Y.; Loo, J. A. HIV-1 Tat Peptide binding to TAR RNA by ESI MS Anal. Chem. 69 (1997) 5130-5135. van Dongen, W. D.; Heck, A. J. R. Binding of selected carbohydrates to apoconcanavalin A studied by electrospray ionization mass spectrometry Analyst 125 (2000) 583-589. Light-Wahl, K. J.; Schwartz, B. L.; Smith, R. D. Observation of the Noncovalent Quaternary Associations of Proteins by Electrospray ionization Mass Spectrometry J. Am. Chem. Soc. 116 (1994) 5271-5278. Zhang, Z.; Krutchinsky, A.; Endicott, S.; Realini, C ; Rechsteiner, M.; Standing, K. G. Proteasome activator 1 IS REG or PA28: recombinant REG alpha/REG beta hetero-oligomers are heptamers. Biochem. 38 (1999) 56515658. Rostom, A, A.; Robinson, C. V. Disassembly of intact multiprotein complexes in the gas phase Protein Sci. 8 (1999) 1368-1370. Hamdan, M.; Curcuruto, O.; DiModugno, E. Investigation of complexes between some glycopeptide antibiotics and bacterial cell-wall analogues by electrospray- and capillary zone electrophoresis/electrospray-mass spectrometry Rapid Commun. Mass Spectrom. 9 (1995) 883-887. WiUiams, D. H. The Glycopeptide Story - How to Kill the Deadly 'Superbugs' Nat. Prod. Rep. 13 (1996) 469-477. Bama, J. C ; Williams, D. H. The structure and mode of action of glycopeptide antibiotics of the vancomycin group Annu. Rev. Microbiol. 38 (1984) 339-357. Rodriguez-Tebar, A.; Vazquez, D.; Perez Velazquez, J. L.; Laynez, J.; Wadso, I. Thermochemistry of the interaction between peptides and vancomycin or ristocetin J. Antibiot. 39 (1986) 1578-1583. Colton, I. J.; Carbeck, J. D.; J., R.; Whitesides, G. M. Affinity capillary electrophoresis: a physical-organic tool for studying interactions in biomolecular recognition Electrophoresis. 1998 Mar;19(3):367-82. Review 19 (1998)367-382. Jorgensen, T. J. D.; Staroske, T.; Roepstorff, P.; Williams, D. H.; Heck, A. J. R. Subtle differences in molecular recognition between modified glycopeptide antibiotics and bacterial receptor peptides characterised by electrospray ionization mass spectrometry J. Chem. Soc. Perkin Trans. I 9 (1999) 18591864. Staroske, T.; O'Brien, D. P.; Jorgensen, T. J. D.; Roepstorff, P.; Williams, D. H.; Heck, A. J. R. The formation of hetero-dimers by vancomycin antibiotics Chem. Eur. J. 6 (2000) 504-509. van der Kerk - van Hoof, A.; Heck, A. J. R. Interactions of a - and (J avoparcin with bacterial cell-wall receptor mimicking peptides studied by electrospray ionization mass spectrometry J. Antimicrob. Chemother. 44 (1999) 593-599. Heck, A. J. R.; Bonnici, P. J.; Breukink, E.; Morris, D.; Wills, M. Modification and inhibition of Vancomycin-group antibiotics by formaldehyde and acetaldehyde Chem. Eur. J. 7 (2001) 910-916.

94 (27)

(28)

(29)

(30)

(31)

(32)

(33)

(34)

(35)

Bonnici, P. J.; Damen, M.; Waterval, J. C. M.; Heck, A. J. R. Formation and Efficacy of Vancomycin Group Glycopeptide Antibiotic Stereoisomers Studied by Capillary Electrophoresis and Bioaffinity Mass Spectrometry Anal. Biochem. 290 (2001) 292-301. Robinson, C. V.; Chung, E. W.; Kragelund, B. B.; Knudsen, J.; Aplin, R. T.; Poulsen, F. M.; Dobson, C. M. Probing the nature of noncovalent interactions by mass spectrometry. A study of protein-Co A ligand binding and assembly J. Am. Chem. Soc. 118 (1996) 8646-8653. Smith, D. L.; Deng, Y.; Zhang, Z. Probing the noncovalent structure of proteins by amide hydrogen-exchange and mass spectrometry J. Mass Spectrom. 32 (1997) 135-146. Stokvis, E.; Clugston, S. L.; Honek, J. F.; Heck, A. J. R. Characterization of Glyoxalase I (E. Coli) - Inhibitor Interactions by electrospray time-of-flight mass spectrometry and enzyme kinetic analysis J. Prot. Chem. 19 (2000) 389397. Clugston, S. L.; Daub, E.; Kinach, R.; Miedema, D.; Barnard, J. F.; Honek, J. F. Isolation and sequencing of a gene coding for glyoxalase I activity from Salmonella typhimurium and comparison with other glyoxalase I sequences. Gene 186(1997) 103-111. Tito, M. A.; tars, K.; Valegrad, K.; Hajdu, J.; Robinson, C. V. Electrospray time-of-flight mass spectrometry of the intact MS2 Virus capsid J. Am. Chem. Soc. 122(2000)3550-3551. van Berkel, W. D.; van den Heuvel, R. H. H.; Versluis, C ; Heck, A. J. R. Detection of intact megadalton protein assemblies of vanillyl alcohol oxidase by mass spectrometry Prot. Science 9 (2000) 435-439. Tahallah, N.; Pinkse, M.; Maier, C. S.; Heck, A. J. R. The effect of the source pressure on the abundance of ions of noncovalent protein assemblies in an electrospray ionization orthogonal time-of-flight Rapid Commun. Mass Spectrom. 15 (2001) 596-601. Demmers, J. A. A.; Haverkamp, J.; Heck, A. J. R.; Koeppe, R. E.; Killian, J. A. Electrospray ionization mass spectrometry as a tool to analyze hydrogen/deuterium exchange kinetics of transmembrane peptides in lipid bilayers Proc. Natl. Acad. Sci. USA 97 (2000) 3189-3194.

H. van der Goot (Editor) Trends in Drug Research III © 2002 Elsevier Science B.V. All rights reserved

95

C H E M I C A L A N D B I O L O G I C A L D I V E R S I T Y IN D R U G DISCOVERY Introduction P. Angeli, G. Gaviraghi* Dipartimento di Scienze Chimiche, University di Camerino, Via S. Agostino 1, 62032 Camerino, Italy •Verona, Italy

In spite of dramatic technological advances in the biopharmaceutical field, R&D productivity has significantly declined in the last decade. The pipelines of major Pharma Drug Companies are almost empty and their growth is being assured only by megamergers and acquisition. This leads to a big concentration of Industry and to cost cutting, including downsizing of research activity. Biotechnology Companies possess a larger portfolio of opportunities, particularly in biotech products, such as proteins, vaccines and antibodies, but the high risk nature of these products predicts that only a small fraction of them will ultimately reach the market. However, the recent achievement of complete genome mapping, together with the development of new approaches, such as functional and structural genomics, proteomics and transgenics is revolutionizing the Discovery process once again. To decrease the attrition rate during the Research and Development process we need to have access to new and validated targets from genomics as well as to have wider chemical diversity. Combinatorial chemistry and High Through Put screening can generate novel ligands which in turn can be optimized to obtain more solid drug candidates. Once optimized, these candidates could be developed with a lower attrition rate than before. This session aims to address the issues of biological and chemical diversities in drug discovery and their complete mtegration to achieve a more efficient productive process. Dr. JUrgen Wess will show the usefulness of gene targeting techniques in elucidating the physiological roles of individual muscarinic receptors, a fundamental step to offer new perspectives for drug therapy. The study of GPR7, a human orphan G-protem coupled receptor expressed in the nervous system, with sequence similarity to both somatostatin and opioid receptors, will be the topic of Dr. Mark Scheideler's lecture. He will deal with changes in expression of this receptor in human painful peripheral neuropathies. Prof Peter Willett will introduce us to the world of computational methods available for quantification of molecular diversity and for design of structurally diverse combinatorial libraries, while Dr. David Langley, as past coordinator of the compound acquisition programme for Glaxo Wellcome Research will tell us how he used the acquisition of compounds from external sources to expand the compound collection.

96 Modem drug discovery projects derive from chemical diversity and combinatorial libraries. Dr. Pierfausto Seneci will explain how a successful project completion strongly depends on the correct selection of chemical diversity in the first screening campaign and of the most promising hits, and how it also depends on the rational design of focused libraries for the optimization of the selected hit structures. And finally, examples showing the progression of optimized drug candidates to development are presented and discussed by Dr. Giovanni Gaviraghi who spent a long career in Glaxo Wellcome. He proposes the intriguing challenge to integrate medicinal chemistry, chemical diversity and biology.

H, van der Goot (Editor) Trends in Drug Research III © 2002 Elsevier Science B.V. All rights reserved

97

MUSCARINIC ACETYLCHOLINE RECEPTOR KNOCKOUT MICE: PHENOTYPICAL ANALYSIS AND CLINICAL IMPLICATIONS J. Gomeza\ M. Yamada^'^, A. Duttaroy\ W. Zhang\ R. Makita^, T. Miyakawa^, J. Crawley^ L. Zhang^ H. Shannon\ F. P. Bymaster^ C. C. Felde/, C. Deng^ and J. Wess^ ^Laboratory of Bioorganic Chemistry and ^Laboratory of Biochemistry and Metabolism, National Institute of Diabetes and Digestive and Kidney Diseases, Bethesda, Maryland 20892, U.S.A. ^Laboratory for Cell Culture Development, RIKEN Brain Science Institute, Wako-shi, Saitama 351-0198, Japan ^Section on Behavioral Pharmacology, Experimental Therapeutics Branch, National Institute of Mental Health, Bethesda, Maryland 20892, U.S.A. "^Lilly Research Laboratories, Eli Lilly and Company, Indianapolis, Indiana 46285, U.S.A.

INTRODUCTION Molecular cloning studies have revealed the existence of five molecularly distinct muscarinic acetylcholine (ACh) receptor (mAChR) subtypes (M1-M5) [1-3]. At a molecular level, the Mi, M3, and M5 receptors preferentially couple to G proteins of the Gq/Gii family, whereas the M2 and M4 receptors are primarily linked to G proteins of the Gi/Go class [1-3]. Studies with mAChR agonists and antagonists have shown that mAChRs are involved in the control of numerous fundamental physiological processes [1-10]. Central mAChRs are known to regulate a large number of vegetative, sensory, and motor functions [4]. Moreover, central muscarinic mechanisms play important roles in arousal, attention, rapid eye movement (REM) sleep, emotional responses, stress modulation, and higher cognitive processes such as memory and learning [4-10]. In addition, disturbances in the central mAChR system have been implicated in a variety of pathophysiological conditions including Alzheimer's and Parkinson's disease, depression, schizophrenia, and epilepsy [4-10]. In the body periphery, mAChRs mediate the well-known functions of ACh at parasympathetically innervated effector organs such as reduction in heart rate, contraction of smooth muscle, and stimulation of glandular secretion [2, 4]. A major problem that has complicated research in the mAChR field has been the lack of ligands, either agonists or antagonists, that can bind to individual mAChR subtypes with high selectivity [2, 11]. Due to this fact, in combination with the presence of multiple mAChRs in most tissues [12-14], it remains unclear in many cases which specific mAChR subtypes participate in the diverse muscarinic functions of ACh, particularly as far as the central muscarinic processes are concerned. To address this question, we and others recently developed mutant mouse lines in which specific mAChR genes had been inactivated by the use of gene targeting

98 techniques [15-19]. This chapter will review the major phenotypes displayed by mutant mice lacking M2 [16], M3 [19], or M4 [17] mAChRs.

ANALYSIS OF M2 AND M4 mAChR MUTANT MICE The M2 and M4 mAChR genes were inactivated via homologous recombination in mouse embryonic stem (ES) cells [16, 17]. Homozygous M2"^' (M2R'^') and M4"^' receptor (M4R"^") mutant mice were obtained with the expected Mendelian frequency, indicating that there was no increase in embryonic or postnatal mortality. Moreover, wild-type (WT) and M2 and M4 receptor mutant mice did not differ in overall health, were fertile, and bred normally. Immunoprecipitation studies using antisera selective for the individual mouse mAChRs showed that the M2 and M4 receptors were widely expressed throughout the brains of WT mice (though in varying densities) [16, 17]. These studies also confirmed the absence of M2 and M4 receptors in the M2R"^" and M4R"^" mutant mice, respectively. In addition, the immunoprecipitation studies demonstrated that the lack of M2 or M4 receptors did not lead to compensatory changes in the levels of M4 or M2 receptors, respectively [16, 17]. Role of peripheral M2 mAChRs In contrast to the M4 mAChR, the M2 receptor is widely expressed throughout the body periphery including heart and smooth muscle tissues [2-4, 20, 21]. Functional studies showed that carbachol, a non-selective muscarinic agonist, produced a dose-dependent bradycardia in spontaneously beating atrial preparations derived from WT and M4R"^' mice [22]. Strikingly, this activity was totally abolished in atria derived from M2R"^" mice. Consistent with previous pharmacologic studies, these data provide direct evidence that the bradycardic effects of ACh in the heart are mediated by the M2 receptor subtype. M2 receptors also represent the predominant mAChR subtype in smooth muscle tissues [20, 21]. Interestingly, Stengel et al. [22] showed that carbachol was slightly less potent (reduction of EC50 values by approximately 2-fold) in contracting isolated urinary bladder, stomach fundus, and trachea smooth muscle strips from M2R"^" mice as compared to the corresponding WT preparations. These findings support the concept that M2 receptors partially contribute to the efficiency of muscarinic agonist-induced smooth muscle contraction, an effect predicted to be mediated primarily by activation of M3 receptors [18,20,21]. Role of central M2 and M4 mAChRs Locomotor activity measurements The M1-M4 mAChRs are abundantly expressed in the striatum [23], a region known to play a key role in the extrapyramidal control or locomotor activity. To examine whether the lack of M2 or M4 receptors affected basal locomotor activity, M2 and M4 receptor mutant mice and their WT littermates were placed in an open field apparatus endowed with photocells, and horizontal activity was monitored over a 1 hr period by counting the number of photobeam disruptions. In these experiments, the M4R'^" mice showed a small but significant increase in basal locomotor activity (Fig. 1 A), which was observed

99 throughout the 1 hr observation period [17]. This effect was M4 receptor-specific since it was not observed with the M2R"'^" mice. In the striatum, the M4 mAChR as well as other mAChRs are coexpressed with different dopamine receptor subtypes [24-29]. To study potential interactions between striatal dopamine and M4 muscarinic receptors, we injected M4R"^' mice and their WT littermates with different dopamine receptor agonists and monitored the resulting changes in locomotor activity. In WT mice, s.c. injection of apomorphine, a nonsubtype selective dopamine receptor agonist, resulted in a modest increase in locomotor activity at intermediate doses (Fig. IB). Strikingly, the locomotor effects of apomorphine were greatly enhanced in M4R'^' mice (Fig. IB). To examine which dopamine receptor subtype was involved in this effect, we carried out analogous experiments with SKF 38393, an agonist selective for Dl-type dopamine receptors, and with quinpirole, an agonist selective for D2-type dopamine receptors. In WT mice, the

"

600

^1-

400

0 O (0 J^

200

n

100

u.

E CO a.

E 0

0 0

^

1000

• WT

800 600

t

400 H 200 0

Veh

Basal locomotor activity ^3500

V^ 0 0

Apomorphine (D1/D2 agonist)

1200

0

0 X} 0 •*->0

1600 1400

300

CO 0)

^^ >

•

500

V.

B

3000

0.3 1 3 10 (mg/kg s.c.)

1000^ SKF 38393 (D1 agonist)

800-]

Quinpirole (D2 agonist)

2500 2000 1500

• WT

600 H

A M4'-/-

400

• WT M4-/-

* A

1000

200

500 0

Veh

3 10 30 (mg/kg s.c.)

100

1—I—I

Veh

0.003

1

1—r

0.03 0.1 1 (mg/kg s.c.)

10

FIG. 1. Locomotor activity of M4R'^' mice and their WT littermates. (A) Basal locomotor activity. Horizontal locomotor activity was assessed in an 'open field' test by determining the number of photobeam breaks during a 1 hr observation period (n=115). (B-D) Locomotor effects induced by dopamine receptor agonists. (B) Apomorphine (non-selective), (C) SKF 38393 (selective for Dl-type receptors), and (D) quinpirole (selective for D2-type receptors). For additional experimental details, see ref. [17]. Data are given as means ± S.E.M. *, p<0.05 (compared to WT). Data were taken from ref. [17].

100 Dl receptor agonist produced a dose-dependent increase in locomotor activity (Fig. IC). This increase in locomotor stimulation was found to be greatly enhanced in M4R"^" mice (Fig. IC). In contrast, administration of the D2 receptor agonist resulted in reduced locomotor activity in WT mice, an effect that remained essentially unaffected in mice lacking the M4 receptor (Fig. ID). These data are consistent v^ith the concept that M4 receptor activation inhibits Dl receptor-stimulated locomotor activity in WT animals. In the striatum, Dl and D2 dopamine receptors are known to be expressed by different subpopulations of spiny projection neurons which comprise the majority of striatal neurons (reviewed in [30]). Receptor localization studies have shown that almost all Dl receptor-expressing striatal projection neurons also express M4 (as well as Mi) mAChRs [28, 29]. DI/M4 receptor-expressing striatal neurons project directly to the substantia nigra (striatonigral pathway), and Dl receptor-mediated activation of this pathway is known to facilitate locomotion [30]. Biochemically, Dl dopamine and M4 muscarinic receptors mediate opposite effects. Whereas Dl receptor stimulation leads to an increase of intracellular cAMP levels via activation of Gs, M4 receptor activation results in a Gidependent reduction in intracellular cAMP levels. It is therefore likely that this functional antagonism underlies the observed M4 receptor-dependent inhibition of Dl receptor-stimulated locomotor activity. From a therapeutic point of view, our data suggest that selective M4 receptor antagonists might be useful in the treatment of Parkinson's disease due to their ability to potentiate the beneficial locomotor effects of L-DOPA or other dopamine receptor agonists. Such agents are likely to cause fewer peripheral and central side effects than classical non-selective muscarinic antagonists [4]. Muscarinic agonist-induced tremor activity When administered to experimental animals, oxotremorine or other centrally acting muscarinic agonists cause akinesia and tremor [31, 32], two of the key symptoms of Parkinson's disease. These responses can also be elicited by direct intrastriatal injection of muscarinic agonists [33] and can be suppressed by pretreatment of animals with muscarinic antagonists or L-DOPA [34, 35], drugs that are widely used in the treatment of Parkinson's disease. Because of these findings, oxotremorine-induced tremor has served as an experimental model to identify new anti-Parkinson drugs. To identify the mAChR subtype mediating muscarinic agonist-induced tremor activity, we injected WT and M2 and M4 receptor mutant mice with increasing doses of oxotremorine, a non-selective muscarinic agonist. At a dose of 0.3 mg/kg (s.c), oxotremorine induced massive whole body tremor in all WT animals tested [16]. Tremor responses remained fully intact in mice lacking the M4 receptor subtype [17]. In contrast, oxotremorine-mediated tremor was totally abolished in both homozygous and heterozygous M2 receptor mutant mice [16]. The loss of tremor activity in the heterozygous M2 receptor mutant mice suggests that agonists must occupy more than 50% of M2 receptors in order to trigger tremor activity. These data convincingly demonstrate that muscarinic agonist-induced tremor is mediated by the M2 receptor subtype. The cellular and molecular mechanisms involved in this activity remain to be elucidated.

101 Muscarinic agonist-induced hypothermia Several lines of evidence suggest that hypothalamic mAChRs play a role in the regulation of body temperature [36]. Consistent with this notion, oxotremorine or other centrally active muscarinic agonists, when administered to mice or other experimental animals, trigger a pronounced hypothermia response. In WT mice, oxotremorine induced a dose-dependent decrease in core body temperature which amounted to about 8 °C at the highest oxotremorine dose used (0.3 mg/kg, s.c.) [16]. The lack of M4 receptors had little effect on oxotremorine-dependent reductions in body temperature [17]. In contrast, in heterozygous M2 receptor mutant mice, oxotremorine dose-response curves were shifted to the right by approximately 3-fold, and in homozygous M2 receptor mutant mice (M2R"^'), oxotremorine-induced hypothermia was impaired to an even greater extent [16]. These data indicate that muscarinic agonist-induced reductions in body temperature are mediated primarily (but not exclusively) by the M2 receptor subtype. Muscarinic agonist-induced analgesia It is well documented that centrally active muscarinic agonists can induce profound analgesic effects in experimental animals and man [37-40]. The antinociceptive effects of muscarinic agoni-sts are known to be mediated by both spinal and supraspinal mechanisms [37-40] and are similar in magnitude to those of morphine [38, 40]. In addition, several studies suggest that muscarinic agonists may be less prone to trigger tolerance and addiction, two major side effects which limit the usefulness of classical opioid analgesics [40, 41]. Identification of the mAChR subtype mediating muscarinic agonist-induced antinociceptive responses is therefore of considerable therapeutic interest. To address this issue, we assessed oxotremorine-induced analgesic effects in M2 and M4 receptor mutant mice and their WT littermates, using the tail-flick and hot plate analgesia tests. Whereas the tail-flick method assesses pain sensitivity primarily at the spinal level, the hot plate test measures pain responses and analgesia mediated predominantly by supraspinal mechanisms [42]. In WT mice, oxotremorine induced dose-dependent analgesic effects in both assays (Fig. 2). The analgesic potency of oxotremorine remained virtually unaffected by the lack of M4 receptors, as studied with M4R"^" mice (Fig. 2C, D). However, in both tests, oxotremorine-dependent antinociceptive responses were found to be markedly reduced in M2R"^" mice (Fig. 2A, B). These findings demonstrated that the M2 receptor subtype plays a predominant role in mediating mAChR-dependent analgesia. Since analgesic responses were reduced but not abolished in the M2R'^' mice, our data suggested that non-M2 mAChRs also contribute to muscarinic agonist-induced analgesia. Since M4 receptors couple to the same class of G protein as the M2 receptors, it seemed reasonable to assume that the agonist-dependent analgesic effects remaining in the M2R''^" mice are dependent on the presence of M4 receptors. We speculated that the contribution of M4 receptors to agonist-mediated antinociceptive responses may have remained undetected in the M4R"^' mice due to the presence of the predominant M2 receptor pathway. To examine the potential role of M4 receptors in muscarinic agonist-induced analgesic effects, we generated double KO mice that lack both M2 and M4 mAChRs. These animals were obtained by intermating M2R"^" and M^R'^'mutant mice (J. Gomeza and J. Wess, unpublished results). Homozygous M2/M4 receptor double KO mice (M2R"^'

102 /M4R"^") were obtained with the expected Mendeiian frequency, did not show any obvious morphological abnormalities, and did not differ from their WT littermates in overall health, fertility and longevity. Moreover, immunoprecipitation studies with receptor subtype-selective antisera indicated that the lack of M2 and M4 receptors did not lead to compensatory changes in the levels of the remaining three mAChR subtypes. Ml, M3, and M5 ((J. Gomeza and J. Wess, unpublished results). Strikingly, oxotremorine was completely devoid of analgesic activity when injected into the M2/M4 receptor double KO mice (A. Duttaroy and J. Wess, unpublished results). The absence of agonist-mediated analgesia was observed in both the tail-flick and the hot plot assays. Identical results were obtained with other centrally active muscarinic agonists (A. Duttaroy and J. Wess, unpublished results). These results clearly indicate that both M2 and M4 receptors mediate the analgesic effects of muscarinic agonists.

B

Tail-Flick

Hot Plate

%MPE 80-

M2

6040-

j J(^

^

/

200

•

t

-20-

Veh

0.01 0.03

0.1

0.3

%^^^>^

Y^

Veh

0.01

0.03

0.1

0.3

Veh

0.01

0.03 0.1

0.3

Oxotremorine (mg/kg s.c.) FIG. 2. Oxotremorine-induced analgesic responses in M2 and M4 receptor mutant mice and their WT littermates. (A, C) Tail-flick test. (B, D) Hot plate assay. Mice of the indicated genotypes were injected s. c. with increasing doses of oxotremorine or saline (Veh). Analgesia measurements were carried out as described in ref. [16]. Data are given as means ± S.E.M. (n=18-20 per dose and genotype) and are expressed as % maximum possible effect (MPE). Data were taken from refs. [16] and [17].

103 The tail-flick data suggested that spinal cord mAChRs play a key role in the suppression of pain responses. To examine the expression of M2 and M4 receptors in mouse spinal cord, we labeled spinal cord mAChRs with a saturating concentration (2 nM) of the non-selective muscarinic antagonist, [^H]N-methylscopolamine ([^H]NMS). Strikingly, the number of [^H]NMS binding sites was reduced by > 90% in preparations from M2/M4 receptor double KO mice and M2R"^' mice (A. Duttaroy and J. Wess, unpublished results). On the other hand, no significant reduction in [^H]NMS binding activity was seen in preparations from M4R"^" mice. These data indicate that the M2 receptor subtype is the predominant mAChR species expressed in mouse spinal cord, whereas M4 receptor expression appears to be too low to be detectable in radioligand binding assays. From a therapeutic point of view, the potential use of M2 receptor agonists as analgesic drugs appears problematic, primarily because of the expected cardiac side effects. On the other hand, the availability of agonists with a high degree of M4 receptor selectivity would allow the administration of relatively high agonist doses required to achieve maximum analgesia without causing unwanted side effects associated with the activation of other mAChR subtypes. Molecular characterization of central inhibitory muscarinic autoreceptors It is well known that acetylcholine (ACh), like most other neurotransmitters, can inhibit its own release via stimulation of so-called inhibitory autoreceptors present on cholinergic nerve endings [43, 44]. Physiologically, this mechanism may serve to prevent excessive neurotransmitter release and subsequent overstimulation of target cells. Classical pharmacologic studies suggest that multiple mAChRs can function as inhibitory autoreceptors in both peripheral and central tissues. However, in many cases, the identity of the specific receptor subtypes involved in this activity remains controversial, primarily due to the limited subtype selectivity of the ligands used in these studies. To address this issue in a more direct fashion, we initiated a series of in vitro ACh release studies, using brain tissues from different mAChR mutant mouse strains. Specifically, we studied hippocampal, cortical, and striatal preparations from mice which lacked either M2 or M4 receptors or which were deficient in both M2 and M4 receptors. For these studies, superfused hippocampal, cortical, and striatal slices were incubated with [^H]choline to label cellular ACh pools. Potassium-stimulated ["^HJACh release was then determined in the absence of drugs (SI phase) and in the presence of the muscarinic agonist, oxotremorine, or other drugs (S2 phase). The S2/S1 release ratio was then used as a parameter to quantitate drug effects on transmitter release. As already stated above, the M2 and M4 mAChRs (but not the Mi, M3, and M5 receptors) are efficiently coupled to G proteins of the GJGQ family [1-3]. Since mAChR-activated Gi/Go proteins mediate the inhibition of voltage-sensitive Ca^^ channels [2, 45] which are known to be intimately involved in the regulation of neurotransmitter release, we speculated that the M2 and/or M4 receptor subtypes represent the major inhibitory muscarinic autoreceptors. To test this concept in a direct and unambiguous fashion, we initially analyzed ACh release using brain tissues from M2/M4 receptor double KO mice. Incubation of hippocampal, cortical, and striatal slices prepared from WT mice with

104 increasing concentrations of oxotremorine led to concentration-dependent reductions (up to 80%) of potassium-stimulated [^H]ACh release (W. Zhang and J. Wess, unpublished results). Strikingly, in hippocampal, cortical, and striatal slices prepared from M2/M4 receptor double KO mice, oxotremorine completely lost its ability to mediate inhibition of stimulated [^H] ACh release (W. Zhang and J. Wess, unpublished results). This observation demonstrates in a very direct fashion that either M2 or M4 receptors (or a mixture of the two receptors) mediate autoinhibition of ACh release in these brain tissues. We are currently carrying out analogous studies with brain slices from M2 and M4 receptor single KO mice in order to assess the contribution of each of these receptors to autoinhibition of ACh release in different areas of the brain. Proper regulation of ACh release in the striatum is known to be critical for coordinated locomotor control [30, 46, 47]. Likewise, the maintenance of proper synaptic ACh levels in hippocampus and cerebral cortex is thought to be important for facilitating learning and memory [48-50]. Identification of the presynaptic mAChRs regulating ACh release in these brain regions is therefore of considerable therapeutic interest.

ANALYSIS OF M3 mAChR MUTANT MICE The M3 mAChR is known to be widely expressed in the CNS and in peripheral tissues [2, 3, 13, 51]. Peripheral M3 receptors are predicted to play a role in ACh-dependent smooth muscle contraction and glandular secretion [2, 4, 20, 21]. At present, little is known about the roles of the central M3 receptors. To learn more about the physiological importance of the M3 receptor subtype, we generated mice with a targeted disruption of the M3 mAChR gene [19]. Western blotting and immunoprecipitation studies confirmed the absence of M3 receptor protein in the homozygous M3 receptor mutant mice (M3R"^" mice). Moreover, immunoprecipitation studies using subtype-selective antisera demonstrated that the lack of M3 receptors had no significant effect on the expression levels of the remaining four mAChR subtypes [19]. Body weight measurements and pharmacologic and endocrinologic studies MsR'^" mice were obtained at the expected Mendelian frequency, showed no obvious behavioral abnormalities, and did not differ from their WT littermates in fertility and longevity. Immediately after birth, WT and M3R"''" mice had similar body weights (Fig. 3). However, starting at about 2-3 weeks after birth, M3R*^' mice showed a significant reduction in body weight (Fig. 3). This difference continued to increase during the following weeks and persisted throughout the life of the animals. Generally, adult male and female M3R'^' mice weighed about 25% less than their WT littermates (Fig. 3). Despite the observed differences in body weight, the body length of adult MsR'^" mice did not differ significantly form that of their WT littermates (Fig. 4A), indicating that the lack of M3 receptors does not interfere with proper linear growth. We observed, however, that the mass of peripheral fat deposits was considerably reduced in the M3R"^" mice. As shown in Fig. 4B, male M3R"" mice showed an about 50% reduction in gonadal fat pad mass, a parameter which generally correlates well with total body fat content [52].

105 Strikingly, the lack of M3 receptors also led to pronounced reductions (5-10-fold) in serum leptin and insulin levels (Fig. 4C, D). The low serum leptin and insulin levels found in the M3R"^" mice are probably primarily due to the reduction in total body fat.

B

25-1

Females

20 H 15H lO-j 51 - I — I — I — I

3 5 7 9 11 13 1517 19 21 Time (week)

3

5 7 9 Time (week)

11

FIG. 3. Lack of M3 mAChRs results in reduced body weight. Growth curves for male (A) and female (B) MsR'^" mice and their WT littermates fed with standard food pellets (n = 17 or 18 per group). Data are given as means ± S.E.M.

A

Bcxly length

B

Fat pads

i£3

10

^ ^ 2

^ 20

f 12 0)

1

8

E

i '

Leptin

3l

12

I

Insulin

E

n

M3WT(+/+)

•

M3K0(-/-)

250 ^

t

200

Glucose I

r^

JO 1501

i 100 E E 50 0} CO

0

FIG. 4. Body length, mass of peripheral fat pads, and serum leptin, insulin, and glucose levels in WT and MsR'^' mice. (A) Body length (nose-anus) measurements were carried out with 20-week old male mice (n=8). (B) Percentage of epididymal fat pad weight relative to total body weight (10-week old male mice; n=8). (C-D) Serum levels of leptin (C), insulin (D), and glucose (E). Blood was collected from freely fed male mice that were about 20 weeks old (n=6). Data are given as means ± S.E.M. *, p<0.05 (compared to WT). Data were taken from ref [19].

106 since both hormones usually circulate at levels that correlate well with body fat content [53-55]. However, it is possible that the absence of pancreatic M3 receptors which may play a role in facilitating insulin release [56] makes an additional contribution to the observed reduction in serum insulin levels. Despite reduced insulin levels, MsR'^' mice showed normal serum glucose levels (Fig. 4E) and did not become diabetic. Moreover, in glucose (administered p.o. or i.p.) tolerance tests, MSR"'^" mice were able to clear glucose from the blood at least as efficiently as WT mice [19]. The MsR'^'mice showed an increase in insulin sensitivity, as determined in an insulin tolerance test [19]. This phenomenon is typically seen in lean animals, explaining why the MsR"^' mice have normal blood glucose levels despite reduced insulin levels. Additional studies showed that the lack of M3 receptors had no significant effect on the serum levels of several other hormones involved in growth and anabolic or catabolic pathways such as the thyroid hormones, insulin-like growth factor 1, and corticosterone [19]. Moreover, the lack of M3 receptors did not significantly affect locomotor activity patterns, metabolic rate, and gastrointestinal motor activity in vivo, as determined in a charcoal transit test [19]. MsR'^' mice also showed normal learning, memory, sensory, and motor abilities, as assessed in a battery of behavioral and neurological tests [57]. Pharmacologic studies suggest that M3 receptors play a role in mediating AChdependent stimulation of salivary gland secretion [2, 4]. To test this concept, we injected WT and M3R'^' mice with three different doses of the muscarinic agonist, pilocarpine, and quantitated saliva output over a 30 min period by the use of a standard filter paper method [58]. Interestingly, a significant reduction in saliva production, by about 50%, was seen only with the intermediate pilocarpine dose (5 mg/kg, s.c). On the other hand, no significant differences in salivary flow were seen with 1 and 15 mg/kg of pilocarpine. These data indicate that both M3 and non-Ms mAChRs contribute to cholinergic stimulation of salivary flow. Food intake studies To examine whether the reduced body weight of the MsR'^' mice was associated with decreased food intake, we next carried out systematic food consumption measurements. The results of the food intake studies are summarized in Fig. 5 in which food intake is expressed either as daily food intake (in g) per mouse or as daily food intake per (body weight in g)^^^, the so-called 'metabolic weight' [59]. Interestingly, these experiments showed that the MsR"^" mice consumed significantly less food than their WT littermates. This difference in food intake was observed independent of the type of food that the mice were offered, either standard dry pellet food (Fig. 5A) or a wet mash diet (Fig. 5B). The wet mash food intake data suggest that it is unlikely that impaired salivary flow is responsible for the hypophagia displayed by the MsR'^" mice. Since the hypothalamus plays a preeminent role in the regulation of food intake, we next carried out radioligand binding studies to examine M3 receptor expression in this brain region. When hypothalamic tissues from WT and MsR"^" mice were labeled with a saturating concentration (2 nM) of the non-selective muscarinic antagonist, [^H]quinuclidinyl benzilate ([^H]QNB), the number of hypothalamic [^H]QNB binding sites was reduced by about one half in the MsK^' mice. This observation suggested that

107

B • 0.2 '^^ .g>

li Jilliri: i

Q

Dry pellet food

§5

-^

1

r2.0

(D 0>

0.3 i ? - n <1>

11^

20 j

£ o CO "

Q

'

D

M3WT(+/+) M3KO(-/-)

5 o

Wet paste food

FIG. 5. Lack of M3 mAChRs leads to reduced food intake. (A) Daily consumption of standard dry food pellets by 12-week old male mice. (B) Daily consumption of wet mash food by 20-week old male mice. Food intake was measured daily over a 5-day observation period. Data are expressed as food intake per mouse or as food intake per (body weight in g) [59]. Data are given as means ± S.E.M. (n=10 per group). p<0.05 (compared to WT). Data were taken from ref. [19].

§

D

M3WT(+/+)

•

M3KO(7-)

i5 0)

c Q

AGRP

POMC

MCH

Orexin

FIG. 6. Expression of hypothalamic neuropeptide mRNAs in WT and MsR"^' mice. Total RNA was isolated from hypothalami of 20-week old, freely fed male mice and subjected to semi-quantitative RT-PCR analysis using peptide-specific primers as described [19]. Neuropeptide expression levels of M3R'^' mice are expressed relative to those of WT mice. Data are given as means ± S.E.M. (n=5 per group). *, p<0.05 (compared to WT). Data were taken from ref. [19].

about 50% of hypothalamic mAChRs (corresponding to about 1 pmol/mg membrane protein) represent M3 receptors. Roles and expression levels of hypothalamic feeding factors Insulin and particularly leptin are considered the two key reporter molecules signaling the peripheral energy status to the CNS [53-55]. Insulin and leptin serum levels are sensed by specific receptors in the arcuate nucleus of the hypothalamus, stimulation of which gives rise to pathways that either stimulate or inhibit eating and energy expenditure. Physiologically, reduced serum leptin levels lead to the activation of one class and the inhibition of another class of neurons in the arcuate nucleus [53-55]. The

108 neurons that are activated contain appetite-stimulating (orexigenic) neuropeptides such as neuropeptide Y and agouti-related protein (AGRP). The neurons that are inhibited by low serum leptin levels contain appetite-suppressing (anorexigenic) neuropeptides such as POMC which gives rise to the active peptide, a-MSH. These first-order leptinsensitive neurons of the arcuate nucleus project to and regulate the activity of secondary hypothalamic feeding centers including the lateral hypothalamus [55, 60-62]. The second-order neurons in the lateral hypothalamus primarily contain orexigenic neuropeptides such as melanin-concentrating hormone (MCH) [63, 64] or the orexins [65], release of which eventually triggers an increase in appetite and food intake. Although the MsR'^' mice showed rather low serum leptin and insulin levels, which normally trigger an increase in food intake [53-55], the lack of M3 receptors was associated with a reduction in food intake (Fig. 5). To shed light on this issue, we employed semi-quantitative RT-PCR analysis to measure the expression levels of several key hypothalamic feeding factors (AGRP, POMC, MCH, and orexins). As shown in Fig. 6, MsR'^' mice showed increased expression levels of the orexigenic peptide, AGRP, and reduced levels of the anorexigenic peptide, POMC, both of which are expressed in first-order hypothalamic neurons. This pattern is typically seen in fasted mice or under conditions of leptin deficiency and serves to stimulate food intake by reducing the activity of hypothalamic melanocortin receptors [53-55]. Interestingly, the loss of M3 receptors led to a significant decrease in the expression of MCH (Fig. 6), an orexigenic peptide synthesized virtually exclusively in second-order neurons of the lateral hypothalamus [63]. This was a surprising observation since increased AGRP levels and reduced POMC levels normally trigger an increase in MCH expression [5355]. We therefore speculated that hypothalamic M3 receptors may be required for the transfer of information from the leptin-sensitive first-order hypothalamic neurons to the second-order MCH neurons. Consistent with this hypothesis, in situ hybridization/ immunohistochemistry double labeling studies demonstrated that M3 receptors are expressed by the majority of MCH neurons of the lateral hypothalamus [19]. Intracerebroventricular administration of orexigenic peptides To further test the hypothesis that the lack of M3 receptors interferes with signaling between first- and second-order hypothalamic feeding centers, we also determined food intake in WT and M3R"^' mice following intracerebroventricular (i.c.v.) infusion of three orexigenic neuropeptides, AGRP, MCH, and orexin-A. In WT mice, as expected, i.c.v. infusion of each of the three peptides resulted in a significant stimulation of food intake (Fig. 7). Strikingly, AGRP failed to stimulate food intake in the M3R"^" mice (Fig. 7). On the other hand, M3R"^" mice responded normally to MCH, indicating that MCHdependent downstream signaling pathways are fully functional in the M3 receptor mutant mice. The i.c.v. infusion data therefore support the hypothesis that M3 receptors are required for the proper responsiveness of second-order hypothalamic feeding centers to orexigenic peptides released from first-order neurons of the arcuate nucleus. In summary, the data discussed above are consistent with the model illustrated in Fig. 8. We showed that M3 receptors are expressed by second-order MCH neurons of the lateral hypothalamus [19]. Previous studies have shown that MCH neurons receive abundant cholinergic innervation from lower brain regions and that muscarinic agonists can stimulate MCH expression in hypothalamic slice preparations [66]. These data.

109 0.8-1

1

cvi 0.6

-^ c

04 H

•§ o

0.2 i

D M3WT(+/+) •

M3K0(-/-)

X

1

LJL

0.0

Vehicle AGRP

MCH

Ji_ Orexin-A

(83-132)

FIG. 7. Food intake of WT and MaR'^'mice following i.c.v. administration of orexigenic neuropeptides. AGRP83-132 (0.9 nmol), MCH (0.2 nmol), or orexin-A (0.2 nmol) were administered to 20-week old male mice via i.c.v. infusion as described [19]. Consumption of standard pellet food was measured for 2 hrs after injections. Data are given as means ± S.E.M. (n=5 or 6 per group). *, p<0.05 (compared to WT). Data were taken from ref. [19].

Higher brain centers

Food intake'

Lateral hypothalamic area Second-order neurons 3rd ventri(cle

Cholinergic input

t

AGRP *v PgJ^^l + %M:i\ J

1

LDTn. PPTn.

Arcuate nucleus First-order neurons

X

Leptin (Insulin)

FIG. 8. Diagram depicting the expression of M3 mAChRs by MCH neurons in the lateral hypothalamus. Previous studies have shown that MCH neurons receive synaptic input from leptin-sensitive AGRP and POMC neurons of the arcuate nucleus [60-62], a primary hypothalamic feeding center, and from cholinergic neurons of the laterodorsal tegmental (LDT) and pedunculopontine tegmental nuclei (PPT) [66]. Our data suggest that M3 receptors are required for the proper responsiveness of MCH neurons to cholinergic stimulation and input from first-order hypothalamic neurons. See text for details (modified according to ref. [19]).

no together with the studies described in this chapter, suggest that M3 receptors are required for maintaining proper hypothalamic MCH expression and proper responsiveness of MCH neurons to input from first-order hypothalamic neurons. It is likely that loss of this activity is a major factor responsible for the reduced food intake displayed by the MsR"^' mice. In agreement with this proposal, mice that lacking the MCH peptide show a phenotype that is very similar to that of the MsR"^' mice [67]. Pharmacologic manipulation of this novel hypothalamic cholinergic pathway may offer new perspectives in the treatment of obesity.

CONCLUDING REMARKS Mutant mouse strains deficient in specific mAChR subtypes represent invaluable novel tools to dissect the physiological and pathophysiological roles of the individual mAChRs. It is likely that the comprehensive pharmacologic, physiological, and behavioral analysis of these mutant animals will reveal new targets for drug therapy. Intermating of mAChR single KO mice will allow the generation of mutant mouse strains lacking two or more mAChR subtypes. Analysis of these animals may reveal new phenotypes that may not manifest themselves in the single KO mice due to functional redundancy. Moreover, recent advances in gene targeting technology should allow the ablation of specific mAChR subtypes in an inducible, tissue-specific fashion.

REFERENCES [1] E.C. Hulme, N.J.M. Birdsall and N.J. Buckley, Annu. Rev. Pharmacol. Toxicol. 30 (1990)633. [2] M.P. Caulfield, Pharmacol. Ther. 58 (1993) 319. [3] J. Wess, Crit. Rev. Neurobiol. 10 (1996) 69. [4] J. Wess, T. Buhl, G. Lambrecht and E. Mutschler, in Comprehensive Medicinal Chemistry, Vol. 3, Ed. J.C. Emmett, pp. 423-491, Pergamon, Oxford, 1990. [5] R.R. Levine and N.J.M. Birdsall, Eds., Subtypes of Muscarinic Receptors. Proceedings of the Seventh International Symposium on Subtypes of Muscarinic Receptors. Life Sci. 60: 963-1211, 1997. [6] R.R. Levine, N.J.M. Birdsall and N.N. Nathanson, Eds., Subtypes of Muscarinic Receptors. Proceedings of the Eighth International Symposium on Subtypes of Muscarinic Receptors. Life Sci. 64: 355-596, 1999. [7 ] R.R. Levine, N.J.M. Birdsall and N.N. Nathanson, Eds., Subtypes of Muscarinic Receptors. Proceedings of the Ninth International Symposium on Subtypes of Muscarinic Receptors. Life Sci. 68: 2449-2642,2001. [8] A. Fisher, Ed., Muscarinic Agonists and the Treatment of Alzheimer's Disease, Neuroscience Intelligence Unit, R.G. Landes, Austin, 1996. [9] D.S. Janowsky, D.H. Overstreet and J.I. Numberger, Jr., Am. J. Med. Genet. 54 (1994) 335. [10] P.P. Bymaster, H.E. Shannon, K. Rasmussen, N.W. DeLapp, J.S. Ward, D.O. Calligaro, C.H. Mitch, C. Whitesitt, T.S. Ludvigsen, M. Sheardown, M. Swedberg,

Ill T. Rasmussen, P.H. Olesen, L. Jeppesen, P. Sauerberg, A. Fink-Jensen, Life Sci. 64 (1999) 527. [11] M.P. Caulfield and NJ.M. Birdsall, Pharmacol. Rev. 50 (1998) 279. [12] A.I. Levey, C.A. Kitt, W.F. Simonds, D.L. Price and M.R. Brann, Identification and localization of muscarinic acetylcholine receptor proteins in brain with subtype-specific antibodies. J. Neurosci. 11 (1991) 3218. [13] A.L Levey, Life Sci. 52 (1993)441. [14] M.T. Vilaro, G. Mengod and J.M. Palacios, Prog. Brain Res. 98 (1993) 95. [15] S.E. Hamilton, M.D. Loose, M. Qi, A.I. Levey, B. Hille, G.S. McKnight, R.L. Idzerda and N.M. Nathanson, Proc. Natl. Acad. Sci. U.S.A. 94 (1997) 13311. [16] J. Gomeza, H. Shannon, K. Kostenis, C. Felder, L. Zhang, J. Brodkin, A. Grinberg, H. Sheng and J. Wess, Proc. Natl. Acad. Sci. U.S.A. 96 (1999) 1692. [17] J. Gomeza., L. Zhang., E. Kostenis, C. Felder, F. Bymaster, J. Brodkin, H. Shannon, B. Xia, C. Deng and J. Wess, Proc. Natl. Acad. Sci. U.S.A. 96 (1999) 10483. [18] M. Matsui, D. Motomura, H. Karasawa, T. Fujikawa, J. Jiang, Y. Komiya, S. Takahashi and M.M. Taketo, Proc. Natl. Acad. Sci. U.S.A. 97 (2000) 9579. [19] M. Yamada, T. Miyakawa, A. Duttaroy, A. Yamanaka, T. Moriguchi, R. Makita, M. Ogawa, C.J. Chou, B, Xia, J.N. Crawley, C.C. Felder, C. Deng and J. Wess, Nature 410 (2001) 207. [20] R.M. Eglen, S.S. Hegde and N. Watson, Pharmacol. Rev. 48 (1996) 531. [21] F.J. Ehlert, R.S. Ostrom and G.W. Sawyer, Life Sci. 61 (1997) 1729. [22] P.W. Stengel, J. Gomeza, J. Wess and M.L. Cohen, J. Pharmacol. Exp. Ther. 292 (2000) 877. [23] S.M. Hersch, C.-A. Gutekunst, H.D. Rees, C.J. Heilman and A.I. Levey, J. Neurosci. 14(1994)3351. [24] C.R Gerfen, T.M. Engber, L.C. Mahan, Z. Susel, T.N. Chase, F.J. Monsma, Jr., and D.R. Sibley, Science 250 (1990) 1429. [25] D.M. Weiner, A.I. Levey and M.R. Brann. Proc. Natl. Acad. Sci. USA 87 (1990) 7050. [26] C. Le Moine, E. Normand, A.F. Guitteny, B. Fouque, R. Teoule and B. Bloch, Proc. Natl. Acad. Sci. U.S.A. 87 (1990) 230. [27] C. Le Moine, E. Normand and B. Bloch, Proc. Natl. Acad. Sci. USA 88 (1991) 4205. [28] V. Bernard, E. Normand and B. Bloch, J. Neurosci. 12 (1992) 3591. [29] E. Ince, B.J. Ciliax and A.I. Levey, Synapse 27 (1997) 357. [30] G. Di Chiara, M. Morelli and S. Consolo, Trends Neurosci. 17 1994) 228. [31] R. George, W.L. Haslett and D.J. Jenden, Life Sci. 1 (1962) 361. [32] C. Sanchez and E. Meier, Pharmacol. Toxicol. 72 (1993) 262. [33] J.D. Connor, G.V. Rossi and W.W. Baker, Int. J. Neuropharmacol. 5 (1966) 207. [34] R.M. Quock and T.S. Lucas, Eur. J. Pharmacol. 95 (1983) 193. [35] A.D. Korczyn and Y. Eshel, Neuropharmacology 18 (1979) 601. [36] R.D. Myers, in Handbook of the Hypothalamus, Vol. 3 (Part A), Eds. P.J. Morgane and J. Panksepp, pp. 83-210, Marcel Dekker, New York, 1980. [37] T.L. Yaksh, R. Dirksen and G.J. Harty, Eur. J. Pharmacol. 117 (1985) 81. [38] P. Hartvig, P.G. Gillberg, T. Gordh, Jr., and C. Post, Trends Pharmacol. Sci. 10 (Suppl.)(1989)75. [39] E.T. Iwamoto and L. Marion, J. Pharmacol. Exp. Ther. 266 (1993) 329.

112 [40] M.D. Swedberg, MJ. Sheardown, P. Sauerberg, P.H. Olesen, P.D. Suzdak, K.T. Hansen, P.P. Bymaster, J.S. Ward, C.H. Mitch, D.O. Calligaro, N.W. Delapp and H.E. Shannon, J. Pharmacol. Exp. Ther. 281 (1997) 876. [41] M. Widman, S. Tucker, D.A. Erase and W.L. Dewey, Life Sci. 36 (1985) 2007. [42] M. Konig, A.M. Zimmer, H. Steiner, P.V. Holmes, J.N. Crawley, M.J. Brownstein and A. Zimmer, Nature 383 (1996) 535. [43] H. Kilbinger, Trends Pharmacol. Sci. 7 (1984) 103. [44] K. Starke, M. Gothert and H. Kilbinger, Pharmacol. Rev. 69 (1989) 864. [45] M.S. Shapiro, M.D. Loose, S.E. Hamilton, N.M. Nathanson, J. Gomeza, and J. Wess and B. Hille, Proc. Natl. Acad Sci. U.S.A. 96 (1999) 10899. [46] O. Hornykiewicz, in Movement Disorders, Eds. D.C. Marsden and S. Fahn, pp 4158, Boston, Butterworth, 1981. [47] S. Fahn, R. Burke and Y. Stern, Prog. Brain Res. 84 (1990) 389. [48] J.T. Coyle, D.L. Price and M.R. DeLong, Science 219 (1983) 1184. [49] D.C. Mash, D.D. Flynn and L.T. Potter, Science 228 (1985) 1115. [50] R. Quirion, I. Aubert, P.A. Lapchak, R.P. Schaum, S. Teolis, S. Gauthier and D.M. Araujo, Trends Pharmacol. Sci. (Suppl.) (1989) 80. [51] A.I. Levey, S.M. Edmunds, C.J. Heilman, T.J. Desmond and K.A. Frey, Neuroscience 63 (1994) 207. [52] P. Rogers and G.P. Webb, Br. J. Nutr. 43 (1980) 83. [53] J. Flier and E. Maratos-Flier, Cell 92 (1998) 437. [54] J.K. Elmquist, C.F. Elias and C.B. Saper, Neuron 22 (1999) 221. [55] M.W. Schwartz, S.C. Woods, D. Porte, Jr., R.J. Seeley and D.G. Baskin, Nature 404(2000)661. [56] A.C. Boschero, M. Szpak-Glasman, E.M. Cameiro, S. Bordin, I. Paul, E. Rojas and I. Atwater, Am. J. Physiol. 268 (1995) E336. [57] J.N. Crawley and R. Paylor, Horm. Behav. 31 (1997) 197. [58] M.W. Parkes and J.C. Parks, Br J. Pharmacol. 46 (1972) 315. [59] M. Kleiber, in The Fire of Life (2nd edn.), pp. 179-222, R.E. Krieger Publishing Company, Huntington, New York, 1975. [60] C.F. Elias, C.B. Saper, E. Maratos-Flier, N.A. Tritos, C. Lee, J. Kelly, J.B. Tatro, G.E. Hoffman, M.M. Ollmann, G.S. Barsh, T. Sakurai, M. Yanagisawa and J.K. Elmquist, J. Comp. Neurol. 402 (1998) 442. [61] C. Broberger, L. De Lecea, J.G. Sutcliffe and T. Hokfelt, J. Comp. Neurol. 402 (1998)460. [62] C.F. Elias, C. Aschkenasi, C. Lee, J. Kelly, R.S. Ahima, C. Bjorbaek, J.S. Flier, C.B. Saper and J.K. Elmquist, Neuron 23 (1999) 775. [63] J.C. Bittencourt, F. Presse, C. Arias, C. Peto, J. Vaughan, J.L. Nahon, W. Vale and P.E. Sawchenko, J. Comp. Neurol. 319 (1992) 218. [64] D. Qu, D.S. Ludwig, S. Gammeltoft, M. Piper, M.A. Pelleymounter, M.J. CuUen, W.F. Mathes, R. Przypek, R. Kanarek and E. Maratos-Flier, Nature 380 (1996) 243. [65] T. Sakurai, A. Amemiya, M. Ishii, I. Matsuzaki, R.M. Chemelli, H.Tanaka, S.C. Williams, J.A. Richardson, G.P. Kozlowski, S. Wilson, J.R Arch, R.E. Buckingham, A.C. Haynes, S.A. Carr, R.S. Annan, D.E. McNulty, W.S. Liu, J.A. Terrett, N.A. Elshourbagy, D.J. Bergsma and M. Yanagisawa, Cell 92 (1998) 573. [66] L. Bayer, P.Y. Risold, B. Griffoond and D. Fellmann, Neuroscience 91 (1999) 1087.

113 [67] M. Shimada, N.A. Tritos, B.B. Lowell, J.S, Flier and E. Maratos-Flier, Nature 396 (1998)670.

Acknowledgements We thank E. Kostenis, J. Brodkin, A. Yamanaka, T. Moriguchi, R. Makita, M. Ogawa, A. S. Basile, C. J. Chou, M. L. Reitman, A. Grinberg, H. Sheng, P. W. Stengel, M. L. Cohen, H. Kodama, H. Kanki, and T. Sakurai for their contributions to the work described in this chapter. We are also grateful to J. Gan, B. Xia, Y. Cui, S. C. Peters, and C. Li for providing expert technical assistance.

This Page Intentionally Left Blank

H. van der Goot (Editor) Trends in Drug Research III © 2002 Elsevier Science B.V. All rights reserved

115

Changes in expression of the orphan G-protein coupled receptor GPR7 in human painful peripheral neuropathies

P. F Zaratin\ A. Quattrini^ S. Previtali^, G. Hervieu^and M. A. Scheideler^ ^Neurobiology Research, GlaxoSmithKline Pharmaceuticals, via Zambeletti, Baranzate di Bollate, Milan, Italy; ^Department of Neuroscience, San Raffaele Hospital I.R.C.C.S., Milan, Italy; ^Neurophysiology & Imaging Research, GlaxoSmithKline Pharmaceuticals, New Frontiers Science Park, Third Avenue, Harlow, Essex, United Kingdom, CM19 SAW.

Introduction Genomics has generated thousands of potential targets. Amoung these targets are the orphan 7 transmembrane receptors (GPCRs). GPCRs are already well established as a class of fruitful drug targets. Of the top 200 best selling prescription drugs, more than 20% interact with GPCRs (Wilson et al., 1998). Now the challenge for the pharmaceutical industry and for basic research is to link these receptors to an appropriate function. The overall strategy for characterizing orphan GPCRs is referred to as reverse pharmacology (Wilson et al., 1998). The reverse approach (Figure 1) starts with an orphan receptor of unknown function that is used as a 'hook' to fish out its ligand using relevant biological extracts. The ligand is then used to explore the biological and pathophysiological role of the receptor. In parallel, the high-throughput screening of available chemical banks is initiated on the receptor to develop agonists and antagonists. Finally, a medicinal chemistry approach is initiated to identify drug candidates. There have been a number of recent examples of the successful application of this reverse pharmacology approach to orphan GPCRs; this has lead to the identification of both endogenous receptor ligands and synthetic agonists or antagonists (Wilson et al., 1998).

116

Reverse Pharmacology Bioinformatic/molecular genetics Full lengh sequence • Orphan GPCR Ligand fishing Compound

y

Ligands

Link orphan GPCRs to diseases

Functional studies y Biological role/Pathophysiology w

Medicinal Chemistry

Drug Candidate

Figure 1. Reverse pharmacology strategy to chracterize orphan GPCRs In parallel to the identification of natural and synthetic ligands, a further challenge is to link orphan GPCRs to human diseases, as a means to determine the therapeutic or diagnostic value of the receptor. Within this approach we have designed a strategy to validate new targets expressed in the peripheral nervous system after injury, as candidate pain and nerve repair targets involved in the pathogenesis or the progression of human painful inflammatory peripheral neuropathies. Human Painful Peripheral Neuropathies The peripheral neuropathies that are accompanied by neuropathic pain sensations are usually thought to include cases of trauma (e.g. causalgia and phantom pain), postherpetic neuralgia and painful diabetic neuropathy. The occurrence of abnormal pain sensations in conjunction with many other peripheral neuropathies is less frequent. Nevertheless, neuropathic pain

117 sensations are known to occur in a very wide variety of conditions and in particular in inflammatory peripheral neuropathies. For example, neuropathic pain is reported in multiple sclerosis, Guillain-Barre syndromes and other inflammatory demyelinating polyneuropathies and in conditions that damage the nerve's blood vessels (i.e. epi- and endoneurial vasculitis). Most human inflammatory peripheral neuropathies result in a final common pathway of structural and functional changes in their axons. As a consequence of peripheral injury an inflammatory reaction occurs. Hematogenous macrophages cross the blood nerve barrier. Once within the nerve, macrophages promote inflammation by releasing pro-inflammatory cytokines such as Interleukin-1 and Tumor Necrosis Factor-alpha (TNF alpha).

Figure 2. Working hypothesis The inflammatory mediator can trigger gene expression changes in Schwann cells that distrupt myelination and induce phenotypic changes in surviving motor and sensory neurons. Changes in the axonal phenotype can contribute to augmented pain signaling which can be associated with human inflammatory peripheral neuropathies. Genes expressed in Schwann cells after injury represent candidate pain and nerve repair targets, and may have important utilities in the diagnosis and treatment of human inflammatory peripheral neuropathies (Figure 2).

118

A target validation strategy to link orphan GPCRs to Human Painful Peripheral Neuropathies The starting point of the above approach (Figure 3) is the information available on the receptor sequence. Based on homology with known GPCRs, the orphan GPCRs have been classified in families (Dennis et al., 2000). Similarity to other GPCRs is used to suggest physiological or pathological functions. Using information available from human and rat receptor sequences we have used this information to generate probes with which to study mRNA expression in tissue sample from patients with peripheral neuropathies. The expression of genes in diseased tissues and/or cells samples is in important step to the discovery and validation of therapeutic targets. The up or downregulation of gene activity can either be the cause or a result of the disease. In the first case, targeting disease-causing gene products is desirable to achieve disease modification. In the second case, interfering with genes that are expressed as a consequence of disease can lead to an allevation of symptoms. To quantify GPCR mRNA levels in human peripheral neuropathies we used sural nerves biopsies from patients with no evidence of peripheral neuropathies and from patients with different kinds of peripheral neuropathies. We screened orphan GPCRs against these tissues using realtime quantitative PCR^^. This technique uses forward and reverse specific primers and a double-labeled fluorescent probe designed against GPCR sequences. During the extension phase of PCR, the probe is cleaved by 5' nuclease activity of TaqGold polymerase resulting in an increase in fluorescent emission at 518 nm (Bustin, 2000) that is proportional to mRNA levels. Real-time quantitative PCR"^*^ offers advantages over existing techniques because it allows the measurement of PCR products during the log phase of PCR product accumulation when thefidelityof PCR is higher, and is a high-throughput technique performed in 96 well plates. Real-time quantitative PCR^"^ offers advantages over existing techniques because it allows the measurement of PCR products during the log phase of PCR product accumulation when the fidelity of PCR is higher, and is a highthroughput technique performed in 96 well plates. As a next step antibodies are generated to study the localization of the receptors in human peripheral nerves and in periheral nerves from preclinical animal models of disease. To complete the target validation strategy transgenic and/or KO mice are generated to established that the receptors are involved in the progression of the disease.

119

Target Validation Strategy Human and rat receptor sequences

mRNA and protein expression in human peripheral neuropathy

Protein expression in preclinical animal models of neuropathy

Generation of Transgenic/KO mice

Figure 3. Target validation strategy to link orphan GPCRs to human peripheral neuropathies

The orphan G-protein coupled receptor GPR7 Among the orphan GPCRs families are opioid receptor-like and somatostatinreceptor like families. A role of G-protein coupled receptors, including the somatostatin and opioid receptors, in nerve repair processes and pain has been reported (Darlinson and Richter, 1999, p. 81). The orphan GPR7 is a member of these family. The GPR7 originally described by O'Dowd et al. (1995), is a human orphan receptor sharing 45% indentity with ^i-, 6- and K- opioid receptors and 50% identity with somatostatin 3 (SST3) receptor. GPR7

120 receptor was shown to bind to several opioid drugs such as bremazocine, levorphanol and p-funaltrexamine, but not to the |i-, 6- and K- opioid receptor subtype selective agonists or somatostatin. This points the likely presence, in vivo, of endogenous selective ligand, as yet unidentified. The orthologs of GPR7 were also cloned in other species, including mouse and rat (Lee et al., 1999). Northern blot analysis using human mRNA and in situ hybridization studies demonstrated that GPR7 is expressed in human cerebellum, frontal cortex, hypothalamus and pituitary (O'Dowd et al., 1995). Overexpression of GPR7 in Schwann cells from patients with inflammatory peripheral neuropathies We showed a marked increased of GPR7 mRNA levels in human sural nerves biopsiy samples from patients with an inflammatory neuropathy, such as hronic Inflammatory Demyelinating Neuropathies (CIDP) and Vasculitic (V) Neuropathy, when compared samples from patients with no evidence of neuropathy and these with hereditary neuropathies, such as Charcot-Mariel, Charcot-Marie2 (CMTl, CMT2) and Hereditary Neuropathy with Liability to Pressure Palsies (HNPP). GPR7 mRNA levels were elevated when an inflammatory infiltration was present also in peripheral neuropathies (PN) of unknown etiology (Figure 4). To study the localization of GPR7 protein in human sural nerves we used antibodies designed against GPR7 and the axonal marker neurofilament-heavy (NF-H). Using these antibodies we performed double immunofluorescence on cross sections of human sural nerves from patients with no evidence of peripheral neuropathies and with vasculitic neuropathy or CIDP (neuritis). GPR7 immunoreactivity was not associated with NF-H immunoreactivity, but myelin forming units demonstrated a sharp concentric ring of outer GPR7 staining surrounding NF-H positive axons, indicating that GPR7 is Schwann cell associated. GPR7 staining was increased in human sural nerves from patients with vasculitic neuropathy and CIDP, confirming data obtained with mRNA expression studies.

121

i TAMRA

8 Plateau

(0

2 o Basbline PCR Cycles Ct

/I

.!550 'Z=7

ZI^

N CMT1CMT2 FN HNPP FN CIDF FAN V Epineurial and endoneurial perivascular inflammatory infiltration: Absent • • Present

Figure 4. GPR7 mRNA upregulation in human sural nerve from patients with inflammatory peripheral neuropathies. GPR7 mRNA levels in human sural nerve from patients with no evidence of peripheral neuropathies were ascribed a value of 1, bargraphs represent mean of relative GPR7 mRNA levels for each kind of neuropathies studied for a total of 30 patients.

122

Overexpression of GPR7 in Schwann cells from animal models of inflammatory peripheral neuropathies Rat experimental autoimmune neuritis (EAN) represents an animal model of autoimmune inflammatory neuropathies, including CIDP (Gold et al., 2000). EAN is induced in susceptible animals by immunization of peripheral nervous system antigens, such as PO and P2. A pathological hallmark of the disease, that is completely manifested 15 days after immunization, is the focal infiltration of peripheral nerves by lymphocytes and macrophages and segmental demyelination. We showed immunohistochemical localization of GPR7 in Schwann cells from control rat sciatic nerve. Immunostaining for GPR7 was increased in EAN Schwann cells 15 days post-imunization. Partial nerve ligation of the rat sciatic nerve (Seltzer et al, 1990) is an animal model of inflammatory and neuropathic pain (Lindenlaub and Sommer, 2000), where a mixture of intact and injured fibers is created. Partial ligation of the rat sciatic nerve produces responses characteristic of thermal hyperalgesia (increased pain to a thermal noxious stimulus) and mechanical allodynia (pain with mechanical stimuli that are not normally noxious) symptomsthat resemble those found in patients and that are completely manifested 14 days after surgery. We showed that GPR7 was expressed in Schwann cells from control rat sciatic nerve. Immunostaining for GPR7 was increased in rat sciatic nerve 14 days after ligation when hyperalgesia and allodynia were present. We confirmed expression of GPR7 in myelin forming Schwann cells using cultured rat Schwann cells treated with forskolin (a cellular model of myelination, as described by Bermingham JR et al., 2001). Discussion These studies indicate that GPR7 is expressed by Schwann cells and the expression of GPR7 in both human and rat nerves is increased in conditions of painful inflammatory neuropathy. The expression of GPR7 into Schwann cells may represent the sum of two competing influences in inflammatory neuropathy: the generation of new axon signals after injury and stimulatory cytokines or other factors secreted by macrophages that may induce the expression of GPR7 in Schwann cells. Tumor necrosis factor alpha (TNFalpha) plays a key role in promoting changes in gene expression after tissue damage due to trauma or disease (Shubavey and Mayer, 1999). TNFalpha is produced by activated macrophages and T cells in response to

123

inflammation (Bemelmans et al., 1996) and Schwann cellls in response to peripheral nerve injury (Wagner and Mayer, 1996). Endoneurial TNF-alpha causes demyelination, axonal degeneration and hyperaigesic pain states (Ledeen and Chakraborty, 1998; Redford et al., 1995; Wagner and Mayer, 1996). We further showed that TNFalpha was able to induce GPR7 mRNA expression, in vitro, in SH-SY5Y neuroblastoma cell line (Scott et al., 2000). Schwann cells are the major glial cells of the peripheral nervous system, where their prime function is to myelinate nerve fibres and promote the generation of rapid nerve impulses. Schwann cells also have a significant role in providing trophic support for spinal motoneurones and dorsal root ganglion neurones. Schwann cell-intrinsic defects have been shown to cause neuropathies associated with pain. For example, the recently described periaxin-null mouse manifests pathological alterations typical of hereditary neuropathies of man, but shows behavioural changes including allodynia (pain elicited by a stimulus that is usually associated with normal sensation) and hyperalgesia (Gillespie, et al., 2000). Po mutations have also been associated with pain syndromes in human neuropathies. For example, CMT2 due to the mutation MyelinProteinZero (thrl24met), is a neuropathy associated with lancinating pm\ and effects on small fibers in human nerve (De Jonghe, et al., 1999; Chapon,etal., 1999). Conclusion Molecules such as GPR7 whose expression in Schwann cells varies under pathological conditions may play a role in the pathogenesis of human neuropathies by disrupting myelination and altering axonal function. Moreover these results enlarge our view of the role that peripheral glia can play in nociceptive signaling in peripheral tissues and may have important implication in the management or diagnosis of painful inflammatory neuropathy.

124 References [1] Bermingham JR, Shumas S, Whisenhunt T, Rosenfeld MG, Scherer SS. Modification of representational difference analysis applied to the isolation of forskolin-regulated genes from Schwann cells. J Neurosci Res. 2001; 63(6):516-24. [2] Bustin S. A., 2000. Absolute quantification of mRNA using real-time reverse transcription polymerase chain reaction assays J. Mol. Endocrinol. 25,169-163. [3] Chapon F., P. Latour, P. Diraison, S. Schaeffer, and A. Vandenberghe, 1999. Axonal phenotype of Charcot-Marie-Tooth disease associated with a mutation in the myelin protein zero gene. J Neurol Neurosurg Psychiatry. 66:779-782. [4] Darlison M. G. and Richter D. ,1999. Multiple genes for neuropeptides and their receptors: co-evolution and physiology TINS. 22, 81-87. [5] Dennis K. Lee, Susan R. George, Brian F. O'Dowd, Susan R. George, Susan R. George, Brian F. O'Dowd, Jilly F. Evans and Kevin R. Lynch 2000. Orphan G protein-coupled receptors in the CNS Current Opinion in Pharmacology 1, 31-39 [6] Gillespie, C. S., D. L. Sherman, S. M. Fleetwood-Walker, D. F. Cottrell, S. Tait, E. M. Garry, V. C. Wallace, J. Ure, I. R. Griffiths, A. Smith, and P. J. Brophy, 2000. Peripheral demyelination and neuropathic pain behavior in periaxin-deficient mice. Neuron. 26:523-531. [7] Gold R., Hans-Peter hratung and Klaus V. Toyka, 2000. Animal models for autoimmune demyelinating disorders of the nervous system Mol. Med. Today 6, 8891. [8] O'Dowd B.F.; Scheideler M.A.; Nguyen T.; Cheng R.; Rasmussen J.S.; Marchese A.; Zastawny R.; Heng H.H.Q.; Tsui L.-C; Shi X.; Asa S.; Puy L.; George S.R., 1995. The cloning and chromosomal mapping of two novel human opioidsomatostatin-like receptor genes, GPR7 And GPR8, expressed in discrete areas of the brain. Genomics, 28, 84-91. [9] Seltzer Z, Dubner R, Shir Y, 1990. A novel behavioural model of neuropathic pain disorders produced in rats by partial sciatic nerve injury. Pain 43,205-218. [10] Thies Lindenlaub and Claudia Sommer, 2000. Partial sciatic nerve transection as a model of neuropathic pain: a qualitative and quantitative neuropathological study. Pain 89, 96-106.

H. van der Goot (Editor) Trends in Drug Research HI © 2002 Eisevier Science B.V. All rights reserved

125

Computational Methods For The Analysis Of Molecular Diversity Valerie J. Gillet and Peter Willett Krebs Institute for Biomolecular Research and Department of Information Studies, University of Sheffield, Western Bank, Sheffield SIO 2TN, UK. Email: {v.gillet, p.willett}©Sheffield.ac.uk 1. INTRODUCTION Developments in combinatorial chemistry and high-throughput screening over the last few years mean that it is now possible to generate and test far greater numbers of compounds in lead discovery programmes than was previously possible. However, molecules that are structurally similar are likely to exhibit comparable activity profiles, and considerations of cost-effectiveness hence dictate that the compounds considered in such programmes should be structurally diverse, so as to maximise the amounts of structure-activity information that can be obtained without redundant experimentation. The need for structural diversity has resulted in the development of computer-based methods for designing libraries that ensure coverage of the largest possible expanse of chemical space in the search for new leads, an area of study that is normally referred to as molecular diversity analysis [1-3]. Diversity analysis draws heavily on methods for the calculation of molecular similarity [4, 5]. Such calculations involve measuring the degree of structural similarity (or dissimilarity) between two molecules by a comparison of the sets of descriptors that characterise those molecules. There is thus much current interest both in the descriptors that are employed to characterise molecules, and in the coefficients that are employed to quantify the degree of resemblance between sets of descriptors. At the same time, new similarity-based methods have been devised for selecting compounds (or, more specifically, libraries of compounds) so as to maximise their structural diversity and for eliminating compounds that are unlikely to exhibit the characteristics of useful leads. These approaches are summarised below, and then exemplified by our program SELECT for combinatorial library design. 2. CALCULATION OF MOLECULAR SIMILARITY Perhaps the principal aim of a molecular diversity analysis is to identify a small, structurally diverse set of compounds, these generally being a subset of those available in an actual database (such as a corporate structure file) or in a virtual library (such as the full set of compounds obtainable by enumeration of the products of a combinatorial synthesis scheme). Diversity is normally taken to mean dissimilar (or disparate, heterogeneous, etc.) and the identification of diverse sets of compounds hence requires ways of calculating just how similar or dissimilar they are. There are many different ways of performing such calculations: here we restrict attention to those measures that can be calculated sufficiently rapidly for use in a molecular diversity context, where libraries can be very large (indeed, virtual libraries may contain hundreds of thousands, or even millions, of compounds). The two most important components of any similarity measure are the structural descriptors that are used to characterise the molecules, and the

126 similarity coefficient that is used to quantify the degree of similarity between pairs of molecules. Many different types of descriptor have been suggested for the calculation of structural similarity and several of these have been adopted for applications in diversity analysis, with most studies thus far involving fragment substructures or physicochemical properties [6]. The presence of fragment substructures (normally 2D substructures containing patterns of connected atoms and bonds) within a molecule are encoded by setting bits in a bit-vector (or fingerprint). Such representations have been used successfully in very many diversity studies [7-9], as have 3D substructural descriptors based upon potential pharmacophoric patterns [10, 11] or topomeric fields [12]. Alternatively, physicochemical properties (normally calculated from a 2D or 3D molecular structure) can be used to describe topological, electronic, steric, lipophilic or geometric features [13-15], with a molecule represented by a real-numbered vector. Once molecules have been characterised in this way, the similarity between a pair of them is calculated by means of a similarity coefficient, which provides a numeric quantification of the degree of resemblance between two sets of such characterisations [5]. Similarity calculations employing substructural data (both 2D and 3D) have generally used association coefficients, typically the Tanimoto coefficient, based on the numbers of fragments common and not-common to a pair of molecules. Conversely, similarity calculations employing property data have generally used distance coefficients, typically, the Euclidean distance, based on a comparison of the two dets of property values. 3. COMPOUND SELECTION METHODS The similarities or distances obtained as above provide the input to the various methods that are available for selecting a structurally diverse set of compounds. Four principal types of selection method have been described thus far in the literature: cluster-based selection; partition-based selection; dissimilarity-based selection; and optimisationbased selection. Cluster-based selection involves subdividing a set of molecules into groups, or clusters, that exhibit a high degree of both intra-cluster similarity and inter-cluster dissimilarity, and a diverse subset is then obtained by choosing one compound from each of the clusters in turn. Clustering was used in some early studies of subset selection [1,6, 16], but is increasingly being complemented, or even replaced, by the other approaches discussed below. Partition-based selection requires the identification of a small number of characteristics, these typically being molecular properties that would be expected to affect binding at a receptor site [14, 17, 18]. The range of values for each such characteristic is subdivided into a set of sub-ranges, and the combinatorial product of all possible sub-ranges then defines the set of cells that make up the partition. Each molecule is assigned to the cell that matches the set of characteristics for that molecule, and a subset is then

127 obtained by selecting one (or some small number) of the molecules from each of the resulting cells. This approach is limited to low-dimensionality datasets but is proving increasingly popular: it is exceedingly fast in operation; it facilitates the comparison of different databases; and it enables the identification of those sections of structural space that are under-represented, or even unrepresented, in a database. Cluster-based and partition-based approaches identify a diverse set of molecules by first identifying groups of similar molecules, and then picking one molecule from each cluster or cell, respectively. Dissimilarity-based approaches seek to identify a diverse subset directly, typically by iteratively selecting compounds that are as dissimilar as possible to those that have already been selected [19]. The identification of a subset that is maximally dissimilar is computationally infeasible but approximate procedures have been found to work well in practice and several different algorithms have been described (see, ^.^., [20]). Optimisation-based approaches involve defining some quantitative measure of diversity, or diversity index [2, 21], and then formulating the identification of the most diverse subset as a combinatorial optimisation problem. The first such approach was described by Martin et al, who used the theory of D-optimal designs in probably the first paper on computer methods for selecting diverse sets of compounds [22] but more recent studies have focused upon the use of simulated annealing and genetic algorithm (GA) methods, both of which appear to be well suited to product-based library design (see, e.g., [2326]). This approach to compound selection is exemplified by the GA-based SELECT program that is described in Section 5 below. 4. FILTERING METHODS Thus far, we have not considered the sorts of molecule to which the selection methods described above are to be applied. However, there is little point in selecting compounds that are unlikely to yield potential leads, and drug-likeness or drugability filtering methods are being increasingly used to focus attention on those compounds that have the greatest a priori probability of exhibiting the properties of previous leads or known drugs [27], with these filters typically being applied prior to the use of a selection procedure. The first reported statistical analysis of the characteristics of known drugs was the widely cited 'Rule of Five', and there have been several subsequent such studies (see, e.g., [28-30]. An obvious extension is to consider also sets of non-drug (or, more usually, presumed non-drug) molecules. Gillet et al. studied the distributions of global molecular properties in sets of drug and non-drug molecules [31], and described a GA to produce a bioactivity profile, weights that maximise the separation between the distributions for the two classes of molecule; the profiles are then applied to the property values for test-set compounds so as to obtain a ranking of them in decreasing order of predicted drug-likeness.

128 Neural networks and decision trees have also been used to develop classification rules based on drugs and non-drugs. The first two reports of the use of neural networks were by Ajay et al. [32] and by Sadowski and Kubinyi [33]. Here, the network is trained using sets of drugs and non-drugs, and a scoring threshold derived that can maximally discriminate between the two classes; test molecules can then be classified by calculating the score when they are presented to the network. Recent examples of work in this area are described by Frimurer et al. [34] and by Sadowski (who discusses the separation of crop-protecting and non crop-protecting compounds) [26]. In decision tree approaches, the root of the tree represents an entire dataset, and this is subdivided into two (or more) subsets depending on the value of some splitting criterion, such as the presence or absence of a particular substructural feature or a CLOGP value lying within a particular range. Potential splits are scored in some way, and the most advantageous chosen to partition the dataset; the procedure is then repeated on the resulting sub-sets, and continued until some termination condition is satisfied. Decision trees have proved to be highly popular [35, 36], with much use being made of the recursive partitioning approach of Rusinko and co-workers [37, 38], which uses a modified f-test as the splitting criterion. 5. DESIGN OF COMBINATORIAL LIBRARIES USING SELECT We have noted previously the need to select sets of compounds that are as structurally diverse as possible. A simple application of selection techniques is to select a diverse subset of an entire database, such as a company's corporate compound collection or a publicly available database such as the Available Chemicals Directory. However, this cherrypicking approach is not appropriate when combinatorial libraries, rather than individual molecules, are required: in this case, a choice needs to be made between reagent-based and product-based selection algorithms. Assume that a combinatorial synthesis involves reacting m reagents of type A with n reagents of type B to yield mn products of type AB. Then a reagent-based design procedure involves selecting a diverse m-member subset from all of the M available reagents of type A (and a similar nmember subset from all of the A^ available reagents of type B)\ product-based design, conversely, involves selecting the mn products from the fully enumerated set of MN possible products (or virtual library). Reagent-based selection is far less demanding of computational resources, but may result in some cases in libraries that are less diverse than those resulting from product-based approaches [39,40]. The identification of the maximally diverse subset of a set of objects is known to belong to the class of NP-complete problems; practical approaches to the selection of diverse subsets hence require the use of approximate approaches that are often framed, as noted previously, as a combinatorial optimisation problem. SELECT is a program we have developed for product-based selection of combinatorial libraries that takes direct account of the combinatorial constraint by means of a GA [41, 42]. This employs a multi-objective fitness function that allows many properties to be optimised simultaneously, facilitating the design of combinatorial libraries that are, by definition, synthetically efficient and that are optimised with respect not just to diversity but also to other user-defined properties of importance in designing bioactive molecules. The

129 program can also be used to design libraries that complement existing libraries and to explore different library configurations. SELECT can be used with a range of types of descriptor: thus far, we have employed both Daylight and UNITY fingerprints and MOLCONN-Z parameters (which are real numbers that have been standardised to fall in the range 0..1). It is also hospitable to a range of diversity indices, with those studied thus far including the sum-of-pairwise dissimilarities, and the average nearest neighbour distance (in each case for the moleculs in a chosen subset). The multi-component fitness function in SELECT enables it to design libraries in product space where the properties of individual molecules within these libraries are optimised simultaneously with the library's structural diversity. Specifically, the physicochemical property profiles of the libraries are optimised with respect to the profile of the same property in some reference collection, for which the experiments reported here used the World Drugs Index (hereafter WDI) database of known drugs (although any other specific collection could be used for this purpose). The fitness function is of the form WD(D) + wc(C) + WfiA/i + waAf2... where the first term, WD(Z)), describes the diversity of the library that is being designed, using one of the three diversity indices listed above. The second term, wc(C), is designed to force the library to be different from some existing reference collection; for example, it may be desirable to ensure that the library is maximally dissimilar from a library that has already been synthesised and tested. This weight can be set to zero if there are no such additional libraries that need to be considered. The remaining terms in the fitness function, WfiAfl, WfjAfI and so on, relate to physical properties of molecules that are thought to affect their ability to function as a drug (such as the molecular weight, the numbers of rotatable bonds, hydrogen donors and acceptors, and the octanol/water partition coefficient) and that can be calculated sufficiently rapid for the processing of libraries of realistic size. A physical property of the library is optimised by comparing the distribution of its values in the library with the distribution of values of the same property in the WDI. The various w terms act as weights that reflect the relative importance of each of the various components of the fitness function, thus allowing the designer to control the characteristics of the libraries that are produced. The effectiveness of this procedure will be illustrated with reference to a threecomponent library that is based on a thiazoline-2-imine template, as shown in Figure 1. Here, the Rl reactants are isothiocyanates, the R2 reactants are amines, and the R3 reactants are haloketones. A set of 10 isothiocyanates, 40 amines and 25 haloketones were selected at random to give a fully enumerated virtual library of 10000 thiazoline-2imines. These molecules were represented by Daylight fingerprints and the diversity measure used was the sum of the pairwise dissimilarities calculated using the cosine

130

R^-N=C=S

R2-NH2

X \ ^ R

^ 3

"'w'l

r^fis

D /

R2

Figure 1. Combinatorial synthesis of a thiazoline-2-imine library. coefficient, with the number of rotatable bonds and the molecular weight as the physicochemical properties of interest; other studies of this dataset are described by Gillet and Nicolotti [40]. The experiments compared the physicochemical property profiles of diverse libraries selected by analysing reactant space with the profiles of the same physicochemical properties in libraries selected from product space that are optimised on property and diversity, simultaneously. In each case, the profile is recorded in a series of 20 bins where each bin represents the percentage of compounds in the library having a given number of rotatable bonds or having molecule weight within a given range. In the case of rotatable bond profiles the bins represent the occurrence of 0, 1, 2, ... > 19 rotatable bonds, while in the case of molecular weight profiles the bins cover the following ranges: 0..49, 50..99,.. ..>950. SELECT was first used to generate diverse sets of reactants (6 isothiocyanates, 10 amines and 15 haloketones) and hence to generate a combinatorial library in reactant space containing 900 thiazoline-2-imines, for which the profiles of rotatable bonds and molecular weights were then calculated. SELECT was next run to choose an analogous 900-molecule library in product space, with the library optimised on both diversity and the rotatable bond profile, and finally in the same way but using both diversity and the molecular weight profile. In each case, the fitness function consisted of the sum of two weighted terms, the diversity term and the relevant property term. The property was included in the fitness function as the RMSD between the distribution of the property in the library represented in a chromosome and the distribution of the property in WDI, where the distributions are given as percentages. The weight assigned to diversity was 1.0 and the weight assigned to the RMSD of the property was 0.1, these weights being chosen so that the RMSD property values were approximately in the same range of values as diversity. The results of these runs are illustrated in Figure 2, where it will be seen that that simple reactant-based selection often results in libraries with poor physicochemical property profiles. The product-based selection, conversely, has enabled the construction of libraries with profiles that are much more "WDI-like" and that are thus more likely to contain bioactive compounds. Current work with SELECT is evaluating a more sophisticated optimisation procedure, called a multiple objective genetic algorithm, that removes the need to specify the weights, w, and that also provides not just one but a whole family of optimised libraries for the user to choose from. This work will be reported shortly.

131

•WDI OProduct-based selection QReactant-based selection

No of Rotatable Bonds

Figure 2a. Profiles for the numbers of rotatable bonds

•WDI ®Product-based selection OReactant-based selection

400

Molecular weight

Figure 2b. Profiles for molecular weight Figure 2. The physical property profiles of thiazoline-2-imine libraries designed using reactant-based selection (in white) are compared with libraries that are optimised in product-space (grey) and with the property profiles found in WDI (black).

ACKNOWLEDGEMENTS We thank John Bradshaw, Peter Fleming, Darren Green, Illy Khatib, Andrew Leach and Orazio Nicolotti for their contributions to the SELECT program. This work has been

132 funded by GlaxoSmithKline, with hardware and software support being provided by Daylight Chemical Information Systems, the Royal Society, Tripos Inc. and the Wolfson Foundation. The Krebs Institute for Biomolecular Research is a designated centre of the Biotechnology and Biological Sciences Research Council. REFERENCES 1. P. Willett, Ed., Special issue on Computational Methods for the Analysis of Molecular Diversity, Perspect. Drug Discov. Design 7/8 (1997) 1. 2. P.M. Dean and R.A. Lewis, Eds., Molecular Diversity in Drug Design, Kluwer, Dordrecht, 1999. 3. D.B. Boyd, D.K. Agrafiotis and E.J. Martin, Eds., Specila Issue on Combinatorial Library Design, J. Mol. Graph. Model. 18 (2000) 317. 4. P.M. Dean, Ed., Molecular Similarity in Drug Design, Chapman and Hall, Glasgow, 1994. 5. P. Willett, J.M. Barnard and G.M. Downs, J. Chem. Inf. Comput. Sci. 38 (1998) 983. 6. R.D. Brown and Y.C. Martin, J. Chem. Inf. Comput. Sci. 36 (1996) 572. 7. G.M. Downs and J.M. Barnard, J. Chem. Inf. Comput. Sci. 37 (1997) 59. 8. D.M. Bayada, H. Mamersma and V.J. van Geerestein, J. Chem. Inf. Comput. Sci. 39(1999)1. 9. T. Potter and H. Matter, J. Med. Chem. 41 (1998) 478. 10. S.D. Pickett, J.S. Mason and I.M. McLay, J. Chem. Inf. Comput. Sci. 36 (1996) 1214. 11. J.S. Mason, I. Morize, P.R. Menard, D.L. Cheney, C. Hulme and R.F. Labaudiniere, J. Med. Chem. 42 (1999) 3251. 12. R.D. Cramer, D.E. Patterson, R.D. Clark, F. Soltanshahi and M.S. Lawless, J. Chem. Inf. Comput. Sci. 38 (1998) 1010. 13. D.J. Cummins, C.W. Andrews, J.A. Bentley and M. Cory, J. Chem. Inf. Comput. Sci. 36(1996)750. 14. R.S. Pearlman and K.M. Smith, J. Chem. Inf. Comput. Sci. 39 (1999) 28. 15. D. Gorse, A. Ress, M. Kaczorek and R. Lahana, Drug Disc. Today 4 (1999) 257. 16. N.E. Shemetulskis, J.B. Dunbar, B.W. Dunbar, D.W. Moreland and C. Humblet, J. Comput.-Aid. Mol. Design 9 (1995) 407. 17. J.S. Mason, S.D. Pickett, Perspect. Drug Disc. Design 7/8 (1997) 85. 18. M.J. Bayley annd P. Willett, J. Mol. Graph. Model. 17 (1999) 10. 19. M.S. Lajiness, Perspect. Drug Disc. Design 7/8 (1997) 65. 20. M. Snarey, N.K. Terret, P. Willett and D.J. Wilton, J. Mol. Graph. Model. 15 (1997) 372. 21. M. Waldman, H. Li and M. Hassan, J. Mol. Graph. Model. 18 (2000) 412. 22. E.J. Martin, J.M. Blaney, M.A. Siani, D.C. Spellmeyer, A.K. Wong and W.H. Moos, J. Med. Chem. 38 (1995) 1431. 23. A.C. Good and R.A. Lewis, J. Med. Chem. 40 (1997) 3926. 24. M. Hassan, J.P. Bielawski, J.C. Hempel and M.J. Waldman, J. Comput.-Aid. Mol. Design 2 (1996) 64. 25. D.K. Agrafiotis, J. Chem. Inf. Comput. Sci. 37 (1997) 841.

133 26. J. Sadowski, Perspect. Drug Discov. Design 20 (2000) 17. 27. S.J. league, A.M. Davis, P.D. Leeson and T.I. Oprea, Angew. Chem. Int. Ed. 38 (1999) 3743. 28. A.K. Ghose, V.N. Viswanadhan and J.J. Wendoloski, J. Comb. Chem. 1 (1999) 55. 29. G.W. Bemis and M.A. Murcko, J. Med. Chem. 42 (1999) 5095. 30. T.I. Oprea, J. Comput.-Aid. Mol. Design 14 (2000) 251. 31. V.J. Gillet, P. Willett and J. Bradshaw, J. Chem. Inf. Comput. Sci. 38 (1998) 165. 32. W. Ajay, W. Walters and M.A. Murcko, J. Med. Chem. 41 (1998) 3314. 33. J. Sadowski and H. Kubinyi, J. Med. Chem. 41 (1998) 3325. 34. T.M. Frimurer, R. Bywater, L. Naerum, L.N. Lauritsen and S. Brunak, J. Chem. Inf. Comput. Sci. 40 (2000) 1315. 35. M. Wagener and V.J. van Geerestein, J. Chem. Inf. Comput. Sci. 40 (2000) 280. 36. S.J. Cho, C.F. Shen and M.A. Hermsmeier, J. Chem. Inf. Comput. Sci. 40 (2000) 668. 37. X. Chen, A. Rusinko, A. Tropsha and S.S. Young, J. Chem. Inf. Comput. Sci. 39 (1999) 887. 38. A. Rusinko, M.W. Farmen, C.G. Lambert, P.L. Brown and S.S. Young, J. Chem. Inf. Comput. Sci. 39 (1999) 1017. 39. E.A. Jamois, M. Hassan and M. Waldman, J. Chem. Inf. Comput. Sci. 40 (2000) 63. 40. V.J. Gillet and O. Nicolotti, Perspect. Drug Discov. Design 20 (2000) 265. 41. V.J. Gillet, P. Willett and J. Bradshaw, J. Chem. Inf. Comput. Sci. 37 (1997) 731. 42. V.J. Gillet, P. Willett, J. Bradshaw and D.V.S. Green, J. Chem. Inf. Comput. Sci. 39 (1999) 169.

This Page Intentionally Left Blank

H. van der Goot (Editor) Trends in Drug Research III © 2002 Elsevier Science B.V. All rights reserved

135

ENHANCING DRUG DISCOVERY BY ACQUISITION OF CHEMICAL DIVERSITY David Langley GlaxoSmithKline R & D , Medicines Research Centre, Stevenage, Hertfordshire SGI 2NY, UK

Introduction One of the key parameters which influences the productivity of high throughput screens, in pharmaceutical drug discovery, is the chemical diversity of the compounds that are tested. Large pharmaceutical companies have significant collections of historical compounds at their disposal and continue to make more by conventional and combinatorial chemistries. However, experience has shown that these resources are inadequate for providing progressible active molecules in all screens, and therefore compound collections require ongoing enhancement from external sources.

Recent experience collection

in growing

the

GlaxoWellcome

compound

This presentation will examine the growth of the compound collection, used for screening in GlaxoWellcome, during the period 1996 to 2000. I will discuss the internal drivers for expanding the compound collection and how we developed a rationale for compound acquisition as a means of achieving this. Another key fector during this period has been the establishment and continuing evolution of a group cf specialist compound supply companies, who have concentrated on making compounds available which are suitable for testing in biological screens. A newer trend has seen the development of a growing number of companies using various forms of library chemistries to generate compounds for screening. As high throughput screening continues to develop and the pharmaceutical companies seek to improve the productivity of the process, the compound suppliers and library companies continue in parallel to evolve their chemistries and business models to get a good fit with the needs of their customers. Various developments in computational capabilities have underpinned our ability to review hundreds of thousands of potentially available chemical structures. It is now possible to build both 2D and 3D databases of millions of chemical structures. In addition, advances in chemical software mean that processes for selecting compounds are getting more powerful. As we move forward in HTS I believe that one of the keys to success will be the utilisation of biological data to select compounds (and/or design new compounds for synthesis), focused on specific biological targets. Cutting across the considerations above is the question of how pharmaceutical companies can make best use of their in-house chemistry resources and which activities

136 they can best purchase or outsource. Many companies are taking the view that a significant proportion of their hit generation activities can be obtained externally and their own chemists are best employed at hit-to-lead stages and beyond. The exception to this would be in cases of generating focused libraries that make use of proprietary mformation. When operating a highly complex process such as HTS it is vital to obtain metrics on the performance of the various components of the system, in an attempt to identify which approaches are giving the best results. This is certainly the case in terms of the various types of compounds and samples that are used in HTS campaigns. This also applies to the selection or design processes used to generate the compoimds. Specifically, in GlaxoWellcome, we have sought to determine whether in-house or acquired compounds, selected on a diversity or targeted basis, have yielded the best results. The drivers for increasing chemical diversity During the past ten or more years, the process of drug discovery in pharmaceutical companies has undergone many radical changes. There has been a move away fixan pharmacology-based testing of small numbers of compounds (often analogues of the natural ligand), generated by conventional medicinal chemistry, towards what we now understand as high throughput screening (HTS). Advances in molecular biology and the human genome have identified many, potentially disease-related, molecular screening targets. Developments in automation and data handlmg have seen the evolution of systems that are capable of screening many tens of thousands of samples in a day. The development of HTS has led to a demand for companies to have at their disposal significant numbers of appropriate samples for screening. When feced with running HTS, particularly for screens where there is little knowledge of the biological target or where knowledge-based screenirig has been unproductive, one needs to expose the biological target to as wide a range of different chemical classes (sometimes referred to as chemotypes) as possible. One of the potential advantages of the so-called random approach to screening is that one can identify activity with novel chemotypes that had not previously been associated with the particular biological target. There is thus a considerable onus on the screening organisation to greatly enhance, on a continuing basis; the chemical diversity of it's screening collection.

We have a biological assay - what do we screen and how? As HTS has developed, companies have been faced with making choices concerning which samples to screen. There are broadly three categories of samples, which are used in HTS. There are synthetic compounds generated by conventional synthetic chemistry, synthetic compounds generated by combinatorial chemistry and natural product samples. It is my contention that the choice of samples, and how to screen them, is one of the key rate-determming steps in HTS, and one which until recently has had fer too little

137 attention paid to it. The identification of screening targets, the validation of a disease links, and the development of HTS systems are obviously all critical activities. However, it has been the case in the past that new assays were simply exposed to whatever samples were available, rather than making a concerted effort to specifically select existing and new (acquu*ed) compounds for the given assay. Decisions have also to be made about how to screen the samples of choice. Do all assays test all samples at their disposal? Although this may be attractive, as it might offer a better chance of success, there are considerable downsides such as the expense cf running large numbers of assays and there is also the data-handling problem. Even the large pharmaceutical companies wouldfindit prohibitively expensive to run all of their screens at fiill throughput. Of course there are various approaches that can reduce the cost, such as miniaturisation, pooling of compounds or multiplexing of two or more assays. The data handling and downstream issues can also be considerable. If one were running screens at a throughput of 500,000 compounds then a primary hit rate cf 1% would yield 5000 apparent "hits" which would require further consideration and/or assays. Other approaches include initially trying a targeted approach where samples are chosen specifically for particular assays. This approach will make use of knowledge of the particular assay to select samples which may be predicted to have a better than random chance of being active. This can be equally applied to acquiring new compounds, which fit the selection criteria, in addition to those already in-house. If the targeted approach does not yield useful results then it can be foUowed-up with afiillscreen. Other sample selection processes take a sequential or iterative approach. Here a set cf samples is selected for screening and based on the resuhs obtained a second set is selected using statistical and cheminformatic tools. Drawbacks of the different sample types Most of the large pharmaceutical companies have three types of sample at their disposal for screening: compounds made by conventional organic synthesis, compounds made by combmatorial chemistry and natural products. Each of the three sample types has drawbacks. Although natural products have a tremendous track record in drug discovery, and still represent a unique source of chemical diversity and novelty, in practice natural samples present considerable difficulties in screening. Natural samples are usually screened as crude extracts of microbes, plants and other life forms. The first difficulty encountered is that many screens, particularly those involving mammalian cells, are sensitive to crude extracts and find them generally toxic. Also, crude extracts are complex mixtures containing thousands of compounds, and so when an apparent "activity" is observed it is often difficult to know whether the effect is a specific one.

138 Where the activity is of sufficient interest it is necessary to undertake isolation and identification of the active component and this can be very time-consuming. Therapeutic projects are usually under time pressures and may not be willing to wait (sometimes for months) for the structures to be elucidated. Even when the active principle has been identified, many natural molecules are of high molecular weight and complex chemistry and thus medicmal chemists find them difficult to work with. Thus there is a growing trend for pharmaceutical companies to move away from natural product screening. Combinatorial chemistry began to be employed in drug discovery research in the mid1990's. There was mitially enormous expectation around the techniques and much importance was placed on the huge numbers of compounds that could be made. It was confidently forecast that combinatorial chemistry would become the sole source of lead molecules in HTS. Early efforts mostly focussed on generating relatively large (tens of thousands) member libraries, often referred to (inappropriately) as "diversity" or "hit generation" libraries. Experience has shown us that in general these approaches have not been productive and recently more effort is going into smaller more target-focused libraries. However, combinatorial chemistry (and its application in HTS) continues to develop and it's optimum deployment is still evolving. Moreover, it remains the case that the range of chemistries possible in combinatorial systems can still not match those in "hand-crafted" synthesis. Therefore, it can be argued that the diversity cf conventional organic molecules is much greater. The overall trend in sample input for HTS in recent years has seen a decline in the use of natural product samples and a steep increase in the involvement of combmatorial libraries in HTS. However, the corporate collection of compounds, made by conventional organic synthesis, has remained a key screening resource. Limitations of historical corporate compound collections Following the amalgamation of the Glaxo and Wellcome compound collections in 1995 we studied the make-up of the compounds at our disposal for screening. We found that over 80% of the compounds had been made by in-house chemistry in the two former companies. This meant that these compounds offered relatively limited chemical diversity, as they were mostly concentrated mto series of molecules, which had been made for former medicinal chemistry programmes. Moreover, it could be considered that because many of the compounds had been made for specific biological targets they were unlikely to provide hits in the many new HTS assays. Prior to the development of HTS the corporate compound collections had served the function of an archive and their origin was not as a resource for new biological assays. Thus in the early to mid-1990's companies were forced to radically re-evaluate the composition cf then- compound collections. Another, growing realisation was that significant proportions of the compounds in the collection were not "appropriate" for screening. Many of the compounds in the collection were either of high or low molecular weight, were inorganic, contained

139 reactive groups or were known to be toxic. An exercise was undertaken to electronically code the undesirable features and to apply the resulting "exclusion filters" to remove compounds inappropriate for screening. Many companies use these filters (sometimes referred to as garbage filters!) and continue to develop them to include various properties deemed to be undesirable. Another area that has received growing attention is the quality and stability cf screening compound collections, particularly in solution. As more compounds, in solid and liquid forms, are analysed this has identified some which have been removed from screening collections.

Critical importance of the corporate compound collection for screening As discussed above, there are significant problems associated with screening natural products, and the optimum role of combinatorial chemistry in HTS is still evolving. This has meant that in recent years the collection of conventional organic molecules (a combination of in-house and acquired compounds) has been the main source cf chemical diversity for screening. The corporate collection is viewed as a reliable resource and one that is well defined. Although the concentration of chemistry in specific therapeutic areas and biological targets can be seen as a disadvantage for screening against new targets, it can conversely be an advantage if screening continues in those areas. What we realised in the mid-90's was that the collection of organic compounds at our disposal was not large enough. The debate continues on numbers of compounds required in HTS in order to be successful. Numbers alone do not describe the problem. It is clearly a matter of what the compounds are, rather than how many are involved. I shall return to this theme later. However, it was apparent to us that our screening collection was relatively small and undoubtedly of limited diversity.

Rationale for compound acquisition - a need to radically expand the collection It is interesting to note that around the time in the early to mid-90's when HTS began to develop and there was a growing demand for new sources of diverse, appropriate compounds we saw the establishment of a number of new, specialist companies oflFering compounds for screening. As an aside it is interesting to reflect on which came first, HTS or the compound suppliers? I will discuss the suppliers later, but a critical point to note here was that between them they were able to offer hundreds of thousands of molecules, which were "appropriate" for screening. By "appropriate" I mean compounds that would pass the exclusion filters (discussed above) and which would be accepted into hits-to-leads programmes by medicinal chemists.

140 Another critical element in accessing compoundsfromexternal sources is their quality. All major compound suppliers now routinely conduct a minimum of LC-MS on all compounds and indeed many do NMR on all compounds. The experience cf Glaxo Wellcome with the quality of external compounds has been generally good. There are anecdotal accounts from some pharmaceutical companies of problems with quality in the early days of compound acquisition but we have not suffered these. Compound suppliers have proven themselves to be very flexible in meeting the requirements of their customers and this has particularly been the case in terms of the formats in which compounds can be provided. The decision to obtain thousands cf compoundsfromexternal sources has a number of implications for in-house compoundand data-handling resources. One example of how in-house handling can be minimised is to obtain compounds provided in deep well plates (on a weight or molar basis) which can go du*ectly onto dispensing robots. Formats for structure files and plate maps can be specified which are compatible with in-house data requirements. As with any activity the cost is another important consideration. When buying compounds for screening in large numbers, the customer can expect discounts such that the unit cost of compounds becomes relatively acceptable. Compounds are usually obtained on a non-exclusive basis with no royalties or milestone payments involved. Some people question the non-exclusive nature of screening compound acquisition, but most take the view that screening is a race and "you have to be in it to win it". Moreover, there are potentially hundreds of thousands (possibly millions) of available compounds and although there will be undoubted overlap between the compound collections of large pharmaceutical companies and their screening portfolios, there is still a very large biological and chemical playing field to operate in. Finally, in terms of building a rationale for compound acquisition, we were aware in the mid-90's that many of our competitors were acquiring compounds on a large scale and although one should not always copy the opposition likewise one can not ignore them! Compound selection processes Probably the single aspect of compound acquisition that continues to develop and unprove is our ability to select compounds. This is a detailed area and in the scope cf this presentation I can only give brief details of the kmds of approaches we have used to select compounds for acquisition. For some years compound suppliers have provided their catalogues of compounds in electronic format - typically as sd files. The approach taken in Glaxo Wellcome has been to build cumulative databases of the compounds on offer from the various suppliers. Databases were originally built with 2D Daylight software but more recently there has been a growing interest in 3D databases built with MSI Catalyst software. These databases have now reached in excess of 2 million structures. It was not long ago that there was not the computational capacity to build and operate

141 databases of that size. Coupled with the ability to build such large databases we have seen rapid advances in chemical software, coupled with ever-higher resolution mapping of biological targets. So, it has become commonplace to derive detailed pharmacophore maps of active sites, which allow for the selection and/or design cf molecules to fit the site. Overall our ability to make ever more precise compound selections continues to improve and I believe will be the most significant advance in the future when our mapping of "drug-like chemical space" may come to mirror our knowledge of the human genome.

Compound supply companies Some of the main compound suppliers are listed below. Their inclusion does not imply any endorsement from Glaxo Wellcome. The suppliers are merely listed in order to make various points. There are also several other companies not listed for space considerations. * * * * * * *

ASINEX BIONET CHEMBRIDGE CHEMDIV CHEMSTAR COMGENEX ENAMINE

* INTERBIOSCREEN * LABOTEST * MAYBRIDGE * SALOR * SPECS * CBI (ZELINSKY INSTITUTE) * 50+ OTHERS!

The origin of many of the companies is interesting. The change of political system in the former Soviet Union in the early 90's allowed access to the considerable chemical resources in the country. One of the first companies to exploit this resource was Specs who are based in the Netherlands. Other companies such as Asinex, Chemstar, Interbioscreen and CBI are based in Russia. ChemBridge and ChemDiv obtam their compounds from Russia and former Soviet Union countries but have growing organisations in the USA. All of these organisations source their compounds from university chemistry departments and various research institutes. Initially, the majority of compounds offered by these companies were "historical collections" - that is they were compounds which had been made at some point in the past. Such compounds continue to represent a significant proportion of those on offer but increasingly compounds are being newly made, either by the academic researchers and/or the companies themselves. There is a significant overlap between the databases of the companies sourcing their compounds from the former Soviet Union but this can sometimes be beneficial. Since the early days of the screening compound supply companies the geographical origin of compounds has greatly expanded. There have been a number of companies in the UK, such as Bionet and Maybridge, for some years. They have always made their own compounds and thus their databases tend to have little overlap with others. Other

142 countries with compound companies include Germany (Labotest), Hungary (Comgenex) and the Ukraine (Enamine, Iflab & Oak). The Dutch company Specs have established offices in Spain, China and India in order to source compounds on an increasingly world-wide basis. Overall the companies are accessing chemical ingenuity globally and thus allowing *their customers access to the products of the ideas of many thousands of chemists. One of the challenges facing the customer is keeping track of the growing number of compound suppliers, their business models and the most recent compounds on offer. Getting the most from external compounds In my experience the companies offering compounds for screening are staffed by highly qualified scientists who can offer significant added value to the customer. It is necessary to maintain a very active network of communications in order to derive maximum benefit from the suppliers. They should not be dealt with in the same manner as standard reagent suppliers. Indeed it is often the case that the companies can act in a collaborative mode rather than just as a supplier. For example, very useful information on compound analogues, synthesis and analytical issues can often be obtamed. In addition to the obvious wish to do more business, many of the staff in the compound supply companies are pleased to offer the benefit of their scientific knowledge. A key element m making best use of the companies is to actively collect and process their compound databases and update files. The traffic in compound files has steadily increased in recent years and it is now almost a daily occurrence to receive new files. It is essential to have a robust and efficient system for dealmg with these files. The methods of transmission of the files continue to evolve and offer some help in handling. So, for example we have movedfromreceiving databases on 20fioppydiscs to the routine use of CD's and a growing trend in downloadablefilesfromweb sites and updates sent by e-mail. There are also a number of practical issues which one needs to be aware of For example most compounds are not held in stock mdefinitely. Many customers have run into difficulties when seeking the re-supply of a compound. Experience has shown that it is best to secure an adequate sample up-front. Although in all purchases one must balance the available budget with the numbers of compounds ideally requh*ed, the quantity of material required to service the HTS programmes and maintain back-up material should it be required for hit confirmation and other further work. Exploiting real and virtual compound databases As HTS has evolved over recent years there has been a need for those in pharmaceutical companies who are involved in the activity to develop new skills and practices. Part of this process mvolves developmg an awareness of the opportunities that exist outside of the company and determining how best to make use of them. In terms of external

143 compounds I have found that it is necessary to conduct an ongoing education of the chemists and biologist engaged in HTS such that they can make best use of what is available. An example of this would be where hits-to leads chemists wish to compile some early SAR around promising hits. Irrespective of whether the hits camefrommhouse or acquired compounds it is wise to search the external, as well as the internal, database for analogues. Such analogues can be brought in cheaply and quickly and prove an effective means of determining which of the hits may be worth pursuing. In Glaxo Wellcome we sought to embrace the concept of the "available chemical world" which comprises the compounds on our shelves together with those which can be readily obtainedfromsuppliers' shelves. In order to achieve this it is essential to have software that will readily allow searching across different databases such as the corporate database, ACD, the cumulative database of screenmg compounds and any others. Compound Acquisition in GlaxoWellcome - Two Broad Themes In GlaxoWellcome we had two broad themes for compound acquisition. Over a five year period we obtained in excess of 300,000 molecules from external sources. For around 80% of the compounds they were acquired, using various selection processes, with the intention of enhancmg the diversity of the compound collection, for use in broad-based screening. So, these compounds were intended for screening in those screens that tested most or all of the collection. A growing trend, which involved the other 20% of the compounds that were acquired, involved various forms of targeted selections. In these cases we used knowledge about families of biological targets, such as kinases, proteases, GPCR's etc., to make selections of compounds which would be tested specifically against assays of that type. (In practice such compounds are often also screened against other types of assay where there is the capacity for afiillHTS campaign.) Factors that will influence future compound acquisition Pharmaceutical companies are having to mcreasingly evaluate where best to utilise their in-house chemistry resources in the drug discovery process. For example it might now be considered unproductive to employ large numbers of chemists to generate large "diversity libraries" for random screening. Whereas it would be more effective for the chemists to be making focused libraries (based on proprietary mformation) for screening against specific targets. It is probably more efficient to obtain hit generation compounds and hit libraries by external purchases and/or outsourced synthesis. As discussed above, the processes for selecting compounds are continuing to improve. As knowledge of the biological screening targets improves one can imagme future scenarios where mcreasingly specific sets of compounds can be assembled for screening with a much better than random chance of observing true hits which can be progressed

144 to leads. It will therefore be essential to have extensive real and virtual compound databases at one's disposalfromwhich to make selections. Compound acquisition in the future will need to match up to developments in HTS. Although improved selection processes may mean that relatively small numbers of compounds will need to be screened, it is likely that the move to ever-higher throughput screening will continue. Thus until a particular approach to hit seeking has been shown be unequivocally more productive we will contmue to build relatively large and diverse compound collections to give us the best chance in HTS. Future prospects for compound acquisition After the profusion of compounds that became available for screening in recent years it was thought by some that the flow of compounds could not continue at the same rate. However, in practice new compounds, from both conventional organic synthesis and combinatorial chemistry, have continued to appear with no sign of slowing down. New companies and countries of origin continue to start business. Indeed, one of the major challenges is to keep track of the many companies offering compounds for screening and to be aware of thefr business models and to ensure that the most recent compound databases and updates are routinely obtained and processed. Improving selection processes will enhance the precision with which we obtain compounds. Various computing developments will also aid the acquisition process. For example, a growing number of suppliers have a live, structure-searchable, stock available database on their web-site which incorporates "online shopping". Other developments have included the company ChemNavigator who have combined the compounds from a number of suppliers on to their website that offers a kind of "one-stop shopping." Library chemistry in GlaxoWellcome When pharmaceutical companies began to apply combinatorial chemistry to their HTS programmes they adopted various approaches in order to expand their capabilities in the area. In the case of Glaxo Welcome the company chose to acquire the company Affymax in 1995. Aflfymax were one of the pioneers in the field in the early 90's. In the years following the incorporation of Affymax, GlaxoWellcome steadily built up its mtemal combinatorial chemistry resources on it's various research sites. To begin with we chose not to access any of the commercially available libraries partly because of the large investment in Affymax and partly because we did not believe that the libraries available at that time offered any advantage over our own. In more recent years there has been a realisation that the many new library companies are offering interesting compounds, of high quality and on realistic terms. Thus we can no longer entfrely depend on in-house library chemistry. Combinatorial library companies A growing number of companies have become established in recent years that are based on the use of combinatorial chemistry to offer products and services, primarily to the

145 pharmaceutical industry. Here is a list of some of those companies. Again, as with the compound suppliers, mention of these companies does not imply any endorsement by GlaxoWellcome. * Arqule * Array * Biofocus * ChemBridge * Coelacanth * DPI (ChemRx) * Molecumetics

* Nanosyn * Oxford Asymmetry * Pharamcopeia * TransTech Pharma * Trega * Tripos * 20+ others!

Combinatorial library companies ~ evolving business models Combmatorial library companies and then* customers (the pharmaceutical companies) have been engaged in a learning process over the past five years. Both parties continue to seek the optimum use of combinatorial chemistry in the drug discovery process. There has therefore been an evolution of products and business models from the libraiy companies. The first libraries to be made available to customers were sold on the basis of buying all or none. Many libraries consisted of tens of thousands of compounds and the potential customer was often interested in obtaining only some of the compounds. More recently such "cherry-pickmg" (the favourite activity of those involved in compound acquisition!) is being increasingly accommodated by some suppliers. Another, growing trend is that a number of companies are oflFering focused libraries designed for screening agamst target systems such as kinases or proteases. The "diversity-based" or "hit-generation" libraries are still in existence but another trend has been for the size of these to decrease. Many pharmaceutical companies have chosen to engage library companies on an exclusive basis to make libraries based on their own or joint designs. In some cases this has mvolved abnost a total outsourcing of hit generation chemistry. Yet other models involve companies with screening facilities where if a client wishes to expose their target to the libraries then the library provider will conduct the screen in thefr shop. As with the pharmaceutical industry itself, there has been considerable consolidation among the library companies with a number merging and/or becoming involved with other organisations whose skill-set includes bioinformatics, cheminformatics, genomics, protein crystallography and ADME/tox modelling etc. The intention is to incorporate information derived from these various sources into the design of compounds, thereby offering much more medicinally robust molecules.

146 Combinatorial library companies - issues and concerns It was our perception that many of the early offerings from some library companies were not attractive molecules and that the terms and conditions for their acquisition were excessive. Many companies were making large "floppy" compounds with high molecular weights leaving little room for hits-to-leads chemistry. Moreover, initially there were relatively few chemistries available and so the range of chemotypes was limited. There were also known to be quality concerns on the purity and stability of many of the early compounds. The early library offerings from some companies were unrealistically priced, often with excessive milestone and royalty payments built in to the terms. Another area cf concern involves the requirement of many library companies for secrecy agreements in order to view their structures. Many pharmaceutical companies find this difficult to deal with as they are concerned that they may be already working on structures which appear in supplier databases.

Where next? At the time of writing (April 2001) GlaxoSmithKline is faced with the consolidation cf two very large compound collections. We then need to derive from this a screening collection and will apply rigorous filters such that only appropriate structures and samples will pass quality criteria are included. There will undoubtedly be a large collection of compounds at our disposal but we need to remember the mantra that "numbers are not everything" and it is the types of compound which is important. The aim will remain to enrich the screening collection, both by in-house synthesis and the acquisition of appropriate external compounds.

H. van der Goot (Editor) Trends in Drug Research 111 © 2002 Elsevier Science B.V. All rights reserved

147

Chemical diversity as a driving force to design and put in practice synthetic strategies leading to combinatorial libraries for lead discovery and lead optimization

Pierfausto Seneci NAD AG, Landsbergerstrasse 50, 80339 Munich, Germany

Chemical diversity is nowadays among the most discussed matters in drug discovery, and a plethora of communications, full articles and reviews dealing with it appeared recently [1-4]. The quantitative measurement of the diversity (or similarity) between two compounds, or between two sets of compounds has been made possible using computational methods [5-7]; terms such as molecular descriptors, similarity indices, pharmacophores, drug-like properties have entered the chemists' world and cannot be neglected once a new drug discovery project is started, or when key decision points in the project hfe are reached. A whole Chapter of this book is dedicated to chemical diversity from a computational perspective [8], so my analysis will rather focus on the strong influence of chemical diversity on the design and synthesis of combinatorial libraries in the drug discovery process. Let us at first try to put thing in context, i.e. examine the phases of modem drug discovery projects and see where chemical diversity and combinatorial libraries are playing a major role (Figure 1). Target identification and target validation are now crucial milestones, as the unraveling of the human genome is providing thousands of new, uncharacterized genes as potential targets for the cure of important diseases. Research laboratories able to identify and validate targets better and faster than competitors will be significantly advantaged, and combinatorial approaches and tools will provide relevant benefits at this stage [9,10]. Nevertheless, the full potential of chemical diversity and combinatorial libraries is evident in the following three steps of the process (Figure 1, bold).

148

FROM GENE TO FUNCTION: TARGET IDENTIFICATION f

FROM FUNCTION TO TARGET: TARGET VALIDATION FROM TARGET TO HIT: DIVERSITY, SCREENING, STRUCTURE DETERMINATION, HIT IDENTIFICATION FROM HIT TO LEAD: HIT OPTIMIZATION FROM LEAD TO CLINICAL CANDIDATE: LEAD OPTIMIZATION Figure 1. Modem drug discovery: The critical steps. Once a target is validated, the hit identification phase kicks off and chemical disciplines start to be heavily involved. While biologists will develop a suitable biological assay to identify compounds interacting with the target, chemists will assemble a suitable chemical collection (DIVERSITY) which will be screened on the biological assay (SCREENING) eventually leading to active compounds, or hits (STRUCTURE DETERMINATION, Figure 1). If the target is novel and poorly characterized, chemical diversity is the main driver to assemble a screening collection. In fact, a collection spanning the so-called diversity space (Figure 2, left) is more likely to identify an active compound, or hit, than a collection which is clustered in an area of the same space (Figure 2, right). Historically large multi-purpose libraries were prepared at this stage to be part of a chemical collection, and known high-throughput synthetic methods (mostly soUd-phase "one bead-one compound" libraries [11] but also automation-assisted discrete libraries [12]) were used to make them. This scenario, though, is more and more outdated as even limited available information can drive the selection of compound sets for screening towards a subset of the chemical space. Examples may arise from the nature of the target (kinase-directed libraries, 7-TM-directed libraries, and so on) or available information about its active site (pharmacophore-driven libraries). These libraries are medium-large in size and made by discrete compounds. The impact of chemical diversity in designing targeted libraries is even higher: an ideal compromise between maximal similarity of the library components with the structural model and maximal diversity among library components must be found to maximize the synthetic effort and to identify useful hits.

149

X

X X X X X

X

X

X X

X X ^

^^

X X X ^ X X X'^ Y X X ^ X y X XX

\'<x\xxx X \

X X

^ X X X X

X X

CHEMICAL SPACE: diverse library

CHEMICAL SPACE: focused library

Figure 2. Chemical space: Diverse and focused chemical libraries. A more general representation of hit identification as an interdependent high-throughput circle (Figure 3) is more suitable for a detailed analysis.

High-throughput structural characterization

Chemical diversity >100K

High-throughput screening (HTS) Figure 3. From target to hit: Diversity, screening and structural characterization. Traditionally the accent in this phase was put on the throughput, i.e. on the availability of large diversity collections (»100K), of high-throughput robotics for the handling and the screening of the diversity, and of high-throughput analytical tools for the determination of the stmcture(s) and of the quality of active compounds. As for the collections, four major sources of compounds are available:

150 • • • •

Single compounds (externally acquired or in house prepared); Natural products from living organisms; Discrete libraries (parallel synthesis, individual compounds); Pool libraries (mix and split synthesis, mixtures).

Each source has its advantages and disadvantages. A collection made of single compounds will be the most diverse, as any individual can be structurally tailored to fit in a specific portion of the chemical space; unfortunately, the time needed to build a » 1 0 0 K diverse collection of single compounds is too long to be effective, or the money needed to buy it from vendors is too much to be generally afforded. Major pharmaceutical companies possess internal collections, ranging from several hundred thousands to several millions compounds, which are extensively used for HTS; unluckily, representatives from these collections are consumed, but most of all these collections are per se biased according with the company's history (e.g., a company with strong tradition in antibacterials will have a large chunk of its collection made by (i-lactams and cefalosporins). Discrete libraries will provide more compounds faster than using classical synthetic protocols while maintaining an acceptable diversity throughout the collection and producing discrete individuals for which both screening and structural characterization are routine operations; regrettably, the synthesis of small-medium discrete Ubraries require extensive automation [13], which is expensive and requires dedicated operators. Pool libraries have the highest productivity, because only minor efforts may lead even to millions of compounds with no or little automation required; sadly, the embedded diversity in the resulting collection is very low as all the components share an identical core structure, screening becomes less rehable (false positives [14]) and structural characterization requires complex or sophisticated techniques (deconvolution [15], encoding [16]). Natural products are the best source for novel, unpredictable diversity possessing biological activity, especially when novel organisms are screened or novel selection criteria are applied [17]; unfortunately, their isolation and purification is often troublesome, and the risk to rediscover a known natural product after extensive efforts is extremely high. The above comments must not be interpreted as a criticism, but rather as a proof of the imjjortance of multiple diversity sources in a screening collection. This would in fact allow to maximize the diversity obtainable with reasonable costs and efforts, and to exploit the advantages provided by each source without experiencing their drawbacks. Another elegant and convenient approach to relevant chemical diversity is represented in Figure 4, using the same circular model. Virtual compounds are generated in siHco by computational methods (virtual collection assembly [18]), then screened in silico using a virtual model for the target (virtual screening [19]) and finally ranked in terms of virtual affinity for the model (diversity subset selection [20]). Such an approach requires significant skills in the dedicated operators, substantial hardware and software investments, and constitutes a risk as virtual rather than experimental data determine the selection of positive compounds; nevertheless, the huge advantage provided by reduction in costs, efforts and time needed to assemble the subset collection, to screen it and to characterize the positives (Figure 5) largely overcomes the above mentioned concerns.

151

Virtual diversity (millions)

Diversity subset selection

Virtual screening Figure 4. Virtual confounds, screening and subset selection.

Medium-low throughput structure determination

Diversity subset assembly (1-lOK)

Medium-low throughput screening Figure 5. Virtual-driven hit identification. Several key messages siunmarize the tendencies in chemical diversity and synthetic combinatorial libraries:

152 • • • • •

A collection must contain subsets from all diversity sources, and must evolve by continuous acquisition of novel, relevant individuals or libraries; Virtual approaches are becoming popular to decrease experimental efforts and to rationalize hit identification; Large pool primary libraries are less popular than in the past; Medium-small, high quality, modular discrete libraries are increasingly popular; Libraries inspired by natural products' structures (decoration or assembly libraries) are increasingly popular.

We will briefly present two examples referring to the last bullet points above mentioned. The first covers the synthesis of 9 modular libraries in solution derived from a common chalcone library LI (Scheme 1) and was recently reported by researchers at ArQule [21].

Scheme 1. Modular chalcone-derived libraries Ll-LlO. The medium-size library LI was assembled from available acetophenones or acetylheterocycles, and from aryl or heteroaryl carboxaldehydes after the assessment of a general experimental protocol (Scheme 2). The authors used the whole LI or a part of it to yield polyfimctional libraries L2-L10 (Schemes 3-4) using the known "libraries from

153 libraries" principle [22]. Commercially available or easily prepared monomers were used in these transformations. The libraries' size varied from 1280 to 25,600 discretes to give a >74,000 compound collection spanning a large area of the chemical diversity space.

L1

32

40

acetophenones, acetylfurans, acetylthiophenes, acetyl pyrroles

benzaldehydes, furancarboxaldehydes, thiophenecarboxaldehydes, pyrrolecarboxaidehydes

1 2 8 0 discretes a: NaOH, 4/1 EtOH/H20, rt, overnight.

Scheme 2. Synthesis of the 1280-membered discrete chalcone library LI.

L1 1280 discretes

L4 7680 discretes R3\K.^^4

0 +

L1

O -^ N

O

x^\/^M^^3 I 40 ^4

subset of 320 discretes

-

^

L5

a: NaOH, EtOH, 80°C, 12 hours; 12,800 discretes b: NaOH, 7 0 X , 8 hours; c: NaOH, EtOH, BOX, 16 hours.

Scheme 3. Synthesis of discrete libraries L2-L5.

154

L1

L6

1280 discretes

7680 discretes

" -' '^

LI 1280 discretes

O

L1 1280 discretes

L1 1280 discretes

7680 discretes

R2 +

R3lh

L1

^

I

^6

subset of 80 discretes R^^ HN

>=0 '

+

H,N

COOH "^^ 20

^<:S\. | -fl-R /

o L10 ocenn-.. * 25,600 discretes

Scheme 4. Synthesis of discrete Ubraries L6-L10.

a: NaOH, EtOH, 80°C, 16 hours; b: NaOH, EtOH, 70°C, 6 hours; c: NaOH, EtOH, 80°C, 12 hours; ^. dioxane/HjO 1/3, 80°C, 16 hours.

155 Another recent example from the Harvard Institute of Chemistry and Cell Biology [23] reported a high quality solid phase pool library of complex, natural products-like compounds obtained from the key supported synthon 1. Its synthesis from easily available precursors 2 and 3 [24,25] was inspired by known synthetic protocols in solution (Scheme 5) [26,27].

a

MeOOC. X = 0-, m- and p-\ 3 resin: water-compatible TentaGel L = photolinker

- alcohol decoration nucleophilic epoxide opening

a: PyBOP, DIPEA, NMP, rt; b: HATU, DIPEA, DMAP, DCM, rt.

Scheme 5. Synthesis of the key supported synthon 1. The opportunities provided by 1 for diversification and library generation are many (italics, Scheme 5), but only high yielding, flexible transformations accepting a wide range of monomers were desirable. A careful assessment proved the 3 boxed/bold transformations in Scheme 5 to be fully satisfactory. A complete monomer rehearsal for alkynes, amines and carboxylic acids (Scheme 6) was carried out on model substrates 4-6 using the assessed experimental protocols. This lead to the selection of the monomers Mi (30), M2 (62) and M3 (62) for library synthesis. All compounds from monomer rehearsal were prepared as discretes and fully characterized by HPLC, NMR and MS after photorelease from the beads.

156

23 monomers: >90% conversion and purity 7 monomers: >70% conversion and purity

30 selected M- monomers

54 monomers: >90% conversion and purity 8 monomers: >70% conversion and purity

87 monomer candidates

62 selected Mj monomers

c: Die. DIPEA. DIVIAP. CH2CI2, rt, 12-16 hours.

+

HOOC

98 monomer candidates 44 monomers: >90% conversion and purity 18 monomers: >70% conversion and purity

62 selected M3 monomers

Scheme 6. Monomer rehearsal for the synthesis of a large, solid phase pool library.

157 The authors prepared a small 456-member model library L l l before embarking into the whole library synthesis (Scheme 7). The synthesis was successful, as judged by the complete analytical characterization of L l l , and its biological testing on several targets highlighted some active compounds [23]. This meant drug-like properties for the components of L l l , as cellular assays were used to screen the library and penetration through the cell membrane was needed to show activity. Unfortunately, two major drawbacks were also apparent: • Deconvolution of active pools, even when containing only 8 compounds, gave rise both to non-reproducible synergistic effects and to false positives, and • The quantity of compound supported onto a single 90|im-bead was not enough to test the library on single bead-based assays.

.NH,

8 representatives

8 representatives

L11 8 representatives

456-member library 8 pools, 64 compounds/pool 90|xm Tentagel beads

Scheme 7. Synthesis of the solid phase model library L l l . The authors eventually prepared the final library L12 (Scheme 8) using larger, in house developed beads [28] and a known chemical encoding technique [29] to enable beadbased screening and structure characterization. The synthesis was again successful and a large, valuable collection of complex, natural products- and drug-like compounds was obtained.

158

^\ y

r\j^

558 compounds

^2

M2 62 representatives + skip codon

30 representatives + skip codon

Ra^ - O

34,596 compounds

px-^^OH

L12

M3

2,179,548-member encoded library 63 pools, 34,596 compounds/pool proprietary 500|im beads

63 representatives + skip codon

Scheme 8. Synthesis of the soUd phase encoded pool Hbrary L12. The hits identified from any screening campaign are by definition showing an in vitro activity on the target, but many other features need to be optimized in order to make them useful leads. These include their physico-chemical properties, their stability, their bioavailability, their selectivity and their toxicity. This phase, here named hit optimization (Figure 1), goes usually together with hit identification under the name of lead discovery; in terms of diversity and libraries of compounds, though, the focusing on the hit structure as a model in hit optimization (similarity-driven libraries) is evident and different from diversity-driven or targeted libraries used in hit identification. It should be kept in mind that several drug discovery projects actually start from hit optimization: in fact, when the structure of a known inhibitor/modulator for a target of interest is known from the literature or from any other available source this structure becomes the model onto which to build focused libraries to progress the "external hit" status to a lead. Conversely on several occasions the hit optimization phase is not successful, that is the efforts prove how the hit structural core can not be progressed to a lead; often this happens because the compound potency and its toxicity, or aspecificity are similarly modulated and leads can not be obtained. In such a case a new screening campaign using a different screening collection, is sometimes considered to find alternative hit structures.

159 Structural focusing becomes even more stringent in lead optimization, when the number of compounds prepared decreases but their characterization becomes extremely accurate to eventually identify one or more clinical candidates to be processed in the downstream events of the drug discovery process. It should be finally noted, though, that a successful project completion at this stage strongly depends on the correct selection of chemical diversity in the first screening campaign, on the selection of the most promising hits and on the rational design of focused libraries for the optimization of the selected hit structures. References [I] [2] [3] [4] [5] [6] [7] [8]

[9] [10] [II] [12]

[13] [14] [15] [16] [17] [18] [19] [20] [21] [22]

L. Weber, Curr. Opin. Chem. Biol 4, 295-302 (2000). M. Wagener and V.J. van Geerestein, J. Chem. Inf. Comput. Sci. 40, 280-292 (2000). E.A. Wintner and C.C. Moallemi, J. Med. Chem. 43, 1993-2006 (2000). P. Willett, Curr. Opin. Biotechnol 11, 85-88 (2000). L. Xue, J.W. Godden, and J. Bajorath, J. Chem. Inf. Comput. Sci. 40, 1227-1234 (2000). N. Rhodes, P. Willett, J.B. Dunbar and C. Humblet, J. Chem. Inf Comput. Sci. 40,210-214(2000). J. Xu and J. Stevenson, J. Chem. Inf Comput. Sci. 40, 1177-1187 (2000). P. Willett, Computational methods for the analysis of molecular diversity. Trends in Drug Research III, vol. 32: Pharmacochemistry library, Ed. H. van der Goot, Elsevier, Amsterdam, 2001. D.S. Thorpe, Combi. Chem. High Throughput Screening 3, 421-436 (2000). K.-Y. To, Combi. Chem. High Throughput Screening 3, 235-241 (2000). M. Lebl and V. Krchnak, Chem. Rev. 97, 411-448 (1997). V.V. Antonenko, Automation in combinatorial chemistry. Combinatorial chemistry and technology: Principles, methods and applications, Eds. S. Miertus and G. Fassina, Marcel Dekker, New York, 1999, pp. 205-232. W.J. Coates, D.J. Hunter and W.S. MacLachlan, Drug Discovery Today 5, 521527 (2000). J.-H. Zhang,, T.D.Y. Chung and K.R. Oldenburg J. Comb. Chem. 2, 258-265 (2000). C. Barnes and S. Balasubramanian, Curr. Opin. Chem. Biol. 4, 346-350 (2000). A. Furka, Combi. Chem. High Throughput Screening 3, 197-209 (2000). Harvey, A., Drug Disc. Today 5, 294-300 (2000). A.R. Leach and M.M. Hann, Drug Disc. Today 5, 326-336 (2000). L. Xue and J. Bajorath, Combi. Chem. High Throughput Screening 3, 363-372 (2000). S.D. Pickett, I.M. McLay and D.E. Clark, J. Chem. Inf Comput. Sci. 40, 263272 (2000). D.G. Powers, D.S. Casebier, D. Fokas, W.J. Ryan, J.R. Troth and D.L. Coffen, Tetrahedron 54, 4085-4096 (1998). J.M. Ostresh, G.M. Husar, S.E. Blondelle, B. Doemer, P.A. Weber and R.A. Houghten, Proc. Natl. Acad Sci. USA 91, 11138-11142 (1994).

160 [23] [24] [25] [26] [27] [28]

[29]

D.S. Tan, M.A. Foley, B.R. Stockwell, M.D. Shair and S.L. Schreiber, J. Am. Chem. Soc. 121, 9073-9087 (1999). D.A. McGowan and G.A. Berchtold, J. Org. Chem. 46, 2381-2383 (1981). D. Keirs and K. Overton, Heterocycles 28, 841-848 (1989). O. Tamura, T. Okabe, T. Yamaguchi, K. Gotanda, K. Noe and M. Sakamoto, Tetrahedron 51, 107-118 (1995). O. Tamura, T. Okabe, T. Yamaguchi, J. Kotani, K. Gotanda and M. Sakamoto, Tetrahedron 51, 119-128 (1995). M.A. Foley, R.W. King and S.L. Schreiber, Book of Abstracts, 221'' ACS National Meeting, BTEC-40, San Diego, April l'*-5*^, 2001, ACS, Washington, DC. H.P. Nestler, P.A. Bartlett and W. Clark Still, J. Am. Chem. Soc. 118, 1813-1814 (1994).

H. van der Goot (Editor) Trends in Drug Research III © 2002 Elsevier Science B.V. All rights reserved

161

Beyond G Proteins: The Role of Accessory Proteins in G Protein-Coupled Receptor Signalling Herwig Just\ Eduard Stefan^ Cornelia Czupalla^, Bemd Niimberg^, Christian Nanoff^ and Michael Freissmuth^ Institute of Pharmacology, University of Vienna, Wahringerstr. 13a, A-1090 Vienna, Austria Department of Pharmacology and Toxicology, University of Ulm, Albert-Einstein-Allee 11, D89069 Ulm, Germany Introduction The basic mechanism by which G protein-coupled receptors (GPCRs) control their cellular effectors has been clarified in the 1970s and 1980s [1,2]; the cycle of activation and deactivation comprises the receptor-mediated exchange of GDP prebound to the G protein a-subunit for GTP, the ensuing dissociation of active, GTP-liganded Ga from the Py-dimer, effector regulation by both free Ga and GPy and signal termination by the intrinsic GTPase of -Ga. Formal proof has been provided for essentially all reaction steps including the recent demonstration that dissociation and reassociation of Ga and GPY do occur in intact cells [3]. Superimposed on this reaction sequence, there is a second cycle of receptor desensitisation and resensitisation which allows for negative feed-back regulation at the level of the receptor; the individual steps include (i) phosphorylation of the agonist-liganded receptor by GRKs (G protein-coupled receptor kinases), (ii) uncoupling by binding of arrestins to the phosphorylated receptor, (iii) sequestration and intemalisation by endosomes and (iv) down-regulation by lysosomal degradation or resensitisation by dephosphorylation and recycling to the membrane [4]. These two reaction schemes, i.e. the cycle of G protein activation and deactivation and the cycle of receptor desensitisation and resensitisation, have also served as working model to analyse the regulatory input that is generated in the network of cellular signalling pathways. The basic components of the transmembrane signalling machinery are the receptor, G protein heterotrimer, and the effector. It has increasingly been appreciated, however, that these do not suffice to understand the complexity of the signal transduction process. Several additional proteins (other than G proteins) impinge on the regulatory cycle by interacting directly with receptors and G protein subunits; these include the afore-mentioned GRKs and arrestins [4]; the latter play roles in signalling other than merely turning off the signal [5]. In addition, the G protein cycle is regulated by RGS proteins, regulators of G protein signalling, which bind to GTP-liganded Ga and function as GTPase activating proteins [6,7]; signalling by Py can be quenched by phosducin and phosducin-like proteins [8]. Finally, the rate of GDP-release from Ga can be modified directly by proteins that are distinct from receptors (termed AGS, activators of G protein signalling, based on their effect in the Py-dependent pheromone pathway of yeast that was originally employed to identify them; [9-11]). Conceptually these proteins can all be placed within the framework of the central cycle, i.e. they affect the kinetics of G protein activation and/or deactivation. There are however, alternative (i.e. non-G protein regulated) effectors that bind directly to receptors (see below).

162 A list of representative accessory proteins that is not exhaustive has been compiled in Table 1. As will become evident in the subsequent sections, in many cases the distinction between components required for sorting, scaffolding proteins, regulatory proteins and effectors is arbitrary; the association of a given receptor with its interaction partner may, for instance, be required for export from the endoplasmic reticulum (ER) and for clustering this receptor at a specific site; proteins that affect receptor trafficking may also directly impinge on the mechanism of signal transduction or feed-back regulation. Receptor synthesis and sorting It is obvious that Table 1 ought to include proteins that act as adapters in the process of sorting, delivery and targeting/retention. After synthesis and insertion into the membrane of the ER, a given receptor (=cargo) is subjected to repeated sorting events; these involve a defined set of analogous steps [12]: the receptor must be concentrated in budding vesicles; this presumably involves the recruitment of coat adapter proteins and coat proteins, protein(s) that support(s) the budding of the vesicle (by pinching off the lipid bilayer), a molecular motor (to allow for movement of the vesicle along microfilaments or microtubules), a vesicular SNARE (soluble N-ethylmaleimide-sensitive factor attachment receptor; to allow for fusion with the target membrane). These steps are repeated as the receptor moves along the biosynthetic pathway from the ER through the Golgi. Surprisingly little is known about the protein-protein interactions that are required for the biosynthesis and sorting of GPCRs. It is however clear that arrestins actually not only serve to block G protein signalling but also act as scaffolds to assemble clathrincoated vesicles. Thus, arrestins are adapter proteins in the budding of endocytotic vesicles [5]. In addition, the third intracellular loop of the Pi-receptor has been found to bind endophilinsl-3 [13]; endophilin-1 is, however, also required for vesicle budding [13] because the modification of lipids by its lysophosphatidic acid (LPA) acyltransferase activity is thought to induce negative curvature of the membrane which facilitates the invagination of lipids [15]. Arrestins - and presumably endophilins - are not involved in biosynthetic trafficking (coat recruitment and vesicle budding in ERGolgi trafficking); nevertheless, these two examples illustrate that GPCRs are not only a passive cargo but can, in principle, recruit molecules required to generate a transport vesicle. It is therefore likely that GPCRs also associate with analogous molecules in the ER. Accordingly receptors that are deficient in binding appropriate components will be retained; this may be the reason why some GPCRs are constitutively retained in the ER/Golgi, if they are heterologously expressed. A prominent example is the GABAer receptor, which requires the presence of the GABAei-receptor for export from the ER [16,17]. Similarly, several mutations in the vasopressin V2-receptor (the genetic basis for X-linked diabetes insipidus) result in the retention of the protein in the ER [18]. Some of these may be due to a loss in the ability of the receptor to recruit the export machinery. Finally, in cerebellar granular cells, mGluR5 (=metabotropic glutamate receptor-5) is not efficiently exported from the ER in the absence of the Homer-1 [19], Homer (several isoforms of which exist, see [20]) is also required for clustering group I mGluRs at specific sites of the cell membrane (see below). Thus, at the very least, these observations provide an example that complexes of receptors and accessory proteins can be preassembled in the ER.

163 Table 1 Accessory proteins (=proteins other than G proteins, arrestins and GRK) that bind directly to G protein-coupled receptors (receptor abbreviated as R) Target Protein Functional effect ref. rhodopsin Tctex-l=dynein chain microtubular transport 32 mGluRl & 5 Homer targeting/retention at presynaptic 19,39,40 or dendritic membrane; clustering inGluR? PICKl retention/protein kinase C-binding 44-46 tubulin mGIuRla targeting/transport ? 52,53 sst-R-2, Shank/SSRIP Retention/clustering? 48 CLl-R 49 spinophilin retention? 50,51 D2-R, a2-R endophilins Sorting/vesicle budding (?) 13 p2-R IL8-R-B PP2A core enzyme Dephosphorylation/resensitisation/ 76 =CXCR2 signalling (?) CXCR2 RGS12 regulation of signalling/clustering 47 src signalling (MAP kinase activation) 66 p3-R signalling (ras /MAPkinase)? Grb2, Nek 65 P4-R 74 EBP50/NHERF regulation of Na^/H^-exchange PrR 36 endosomal recycling 63 signalling ATII-Rl phospholipase Cyl 64 inhibition of signalling ATII-Rl, eNOS B2-, ETB-R 62 signalling ATII-Rl JAK2 89 inhibition of signalling Ca^Vcalmodulin |Li-opioid R 79 Ca^'^-channel regulation mOluR? 81 inhibition of signalling P2-R 77,78 inhibition /G protein selectivity coupling cofactor Ai-R 72,73 Impact (?) on transcription ATF/CREB GABABI-R 99 adenosine deaminase enhanced intemalisation (?) ApR 82,87 promoted by agonist occupancy homooligomerisation PrR 88 homooligomerisation constitutive oligomers 5-opioid-R 16,17 ER-export/membrane insertion GABABI/2-R heterooligomerisation heterooligomerisation altered pharmacological specificity 83 K-/5-opioid heterooligomerisation cross-talk with ionotropic receptor 90 DI/GABAA 86 heterooligomerisation and D2/sst-R5 84,85 heterooligomerisation G protein-coupled Ai/Di 84 heterooligomerisation receptors A2A/D2 91,92 heterooligomerisation Coreceptor for axon guidance (?) A2B/DCC altered pharmacological specificity 89 adrenomedul RAMPs lin/CGRP-R Receptor delivery/targeting and retention Vesicles that are loaded with a given receptor leave the Golgi and move along tracts of microtubules (and/or along the actin cytoskeleton) and are delivered to the plasma

164 membrane. Like many other signalling components G protein-coupled receptors and G proteins are, however, in many instances not uniformly distributed over the plasma membrane. Caveolins have been proposed to play an important role in organising signalling components (for review [21]); this view, however, has been questioned [22]. Of interest, G protein subunits and receptors are co- and posttranslationally modified by lipids (myristoylation, prenylation and palmitoylation). Palmitoylation of both, receptors and Ga-subunits, is dynamic [23]. Although the precise role of these lipid modifications is not fully understood, they are likely to play a role in targeting the proteins to lipid rafts; these membrane microdomains are thought to rely, at least in part, on the propensity of certain lipids (e.g. glycosphingolipids and cholesterol) to support selforganisation and to trap lipid-modified signalling molecules [24]. In polarised cells (epithelia and neurons), however, the segregation of membrane proteins is more pronounced; in epithelia, some proteins reside exclusively in the basolateral membrane (e.g. Na^/K^-ATPase) while others are only found on the apical side (e.g. epithelial Na^-channel). Originally, it had been argued that there was a functional equivalence between membrane compartments in polarised epithelial cells and neurons with basolateral and apical membranes corresponding to somatodendritic and axonal compartments, respectively [25]. This model provided a framework to formulate hypotheses and explore common principles; e.g. in epithelial cells the tight junctional ring prevents exchange of proteins between the apical and the basolateral membrane. A similar barrier to diffusion also exists in the vicinity of the axon hillock [26]. Lipid rafts may also contribute to the polarised distribution in neurons [27]. However, these factors can only maintain the segregation of proteins; clearly, other mechanisms must be important to generate the different distribution of proteins. Because of the length of axons (up to >1 m in man) and due to extensive ramifications of dendritic trees, neurons face challenges in targeting that epithelial cells do not have to meet. For instance, some proteins must be deposited along the axon (e.g. Na"^- and K'^-channels), while others are delivered exclusively to the presynaptic specialisations (e.g. presynaptic GPCRs, Ca^^-channels, neurotransmitter transporters). Given this complex task, it is not surprising that the original proposal (apical=axonal; somatodendritic=basolateral) has repeatedly been proven to be incorrect [28]. This is also true for GPCRs [29] and presumably reflects the fact that epithelia lack components of the targeting/retention machinery. An important mechanism in the control of delivery is the selection of the appropriate motor molecule that will direct the movement of receptor-containing vesicles. In most cells, the minus ends of microtubuli originate in the perinuclear region while the plus ends point to the periphery. In axons, microtubuli are uniformly oriented with their plus ends pointing to the synapse and their minus ends to the soma. Accordingly, anterograde delivery of axonal proteins is accomplished by plus end-driven molecular motors. These are N- and M-type KIFs (=members of the kinesin family with a motor domain at the N-terminus or in the middle; [30]). In contrast, C-type KIFs and cytosolic dynein function as minus-end driven motors and are thus candidates for mediating retrograde transport and endosomal recycling. Accordingly, transport vesicles are segregated into different pools thai select different kinesins. The mechanisms, by which motor molecules are recruited, include binding to vesicular coat proteins and adapter proteins, PDZ-

165 domain containing proteins (for explanation of the acronym see below) and other scaffolds as well rab GTPases [31]. There is, however, at least one example, where the cargo, i.e. a GPCR, serves as the docking site for the motor molecule: The retinal photoreceptor rhodopsin binds directly - via its carboxy terminal extension - to a molecular motor, namely to the dynein light chain Tctex-1 [32]. This observation indicates that GPCRs can per se function as binding partners for motor molecules and thereby specify the direction of transport. Rhodopsin is apparently delivered to minus end of the microtubular rail at the base of the cilia [32]; at this point, unconventional myosin VII (=an actin-based motor) is thought to take over [33]. The example of rhodopsin also illustrates the clinical relevance of receptor transport and delivery: mutations in the C-terminus of rhodopsin have a low affinity for dynein and are mistargeted; this is associated with retinal degeneration (=retinitis pigmentosa) due to loss of photoreceptors [32]. In polarised cells, disruption of the microtubuli (e.g. by colchicine, vinca alkaloids or nocodazol) impedes the delivery of (some, but not all) GPCRs to the plasma membrane [34,35]. Similarly, disrupting cortical (=submembranous) actin filaments with latrunculin (A or B) also affects receptor recycling [36]. (It should be noted that in many cases experiments with cytochalasin D, which also destroys actin filaments, have been less revealing, presumably because cytochalasin D has only a modest impact on the cortical actin meshwork [37]). This indicates that (i) trafficking itineraries differ and (ii) that there is an interplay between kinesin (=tubulin-based) and myosin (=actin-based) motors, where a given receptor is sequentially handled by several motors. The physiological relevance is evident; during prolonged desensitisation, in many cases, receptorcontaining vesicles can be visualised in the perinuclear region (reflecting presumably endocytotic retrieval via a dynein motor). Latrunculin B-sensitive transport may allow for recycling to the plasma membrane and, hence, for resensitisation. Alternatively, a different motor molecule, delivers the receptor to the lysosomal compartment which leads to loss of receptors, i.e. desensitisation by down-regulation. In the case of the Pareceptor, this decision depends on binding of EBP50/NHERF (the ezrin binding protein of 50 kDa = NaVH^-exchange regulatory factor) to the (phosphorylated) very C-terminal end of the receptor; the association of EBP50/NHERF reroutes the receptor to endosomal recycling rather than to lysosomal degradation [36]. Receptor retention and clustering After the receptor-containing vesicle has been delivered and fused (via SNARE interaction) to its target membrane, the receptor needs to be retained at this specific site in order to maintain a polarised distribution. If this does not occur, the receptor may be redistributed [38]. In many cases that have been studied, receptors are concentrated in specialised compartments of the plasma membrane (e.g. clustered at postsynaptic densities) via interaction of their carboxy termini with scaffolding proteins that contain PDZdomains (PDZ standing for a domain found in >50 proteins but first identified in the postsynaptic density protein of 95 kDa, Drosophila discs large and the zonula occludens protein zo-1, hence the acronym). Clustering was originally reported for ionotropic receptors (e.g. the nicotinic acetylcholine receptor in the end plate of the skeletal muscle). The discovery of Homer (now termed Homer-la) represented a major breakthrough; Homer-la was originally identified as an immediate early gene [39] and proposed to contain a PDZ-do-

166 main; however, the N-terminal anchoring domain of Homer-1 (and its relatives, Homer2 and -3; see [20]) is an EVHl-domain (shared by Drosophila Enabled, VASP=vasodilator stimulated phosphoprotein, WASP= Wiskott-Aldrich syndrome protein; [40]). This EVHl-domain of Homer interacts with the carboxy terminus of group I mGluR (mGluRl and mGluRS) by recognising a proline-rich motif (PPXXFR) and this interaction is thought to target the receptor to the postsynaptic membrane. Importantly, the expression of Homer-la is regulated in a highly dynamic manner in response to synaptic activity [19,39]. Thus, in the absence of Homer, the receptor is not delivered to the synapse. The splice variants Homer-lb and -Ic differ from Homer-la by the presence of an additional C-terminal leucine zipper; depending on which splice variant is expressed, mOluRS is delivered to dendrites (in the presence of Homer-lb and -Ic), whereas expression of Homer-la also resulted in delivery to axons [19]. Thus, Homer-lb and -Ic are apparently endowed with an axonal exclusion signal. In addition, only Homer-lb and -Ic support the formation of receptor clusters via their leucine-zipper [41]. The case of Homer also illustrates the fact that it is difficult to separate the individual steps (ERexport, sorting, targeting, delivery and retention) because Homer is actually required for targeting and clustering; however, once the receptors are delivered to the appropriate sites, they remain anchored even in the absence of Homer [19]. Obviously, there must be many more proteins that target and anchor GPCRs; mGluR7, a group III mGluR, is of particular interest because it is almost exclusively targeted to presynaptic specialisations and highly enriched in the active zone [42]. The C-terminus of mGluR7 specifies axonal localisation but it does not contain a motif which supports interaction of mGluRV with Homer-la or its relatives [43]. Recently, PICKl was identified as a binding partner for mGluRV [44-46]. PICKl is a binding partner for protein kinase Ca (PKCa) and contains a PDZ that binds the very C-terminal end of mGluR7 (and, to a lesser extent, that of other group III mGluR [45]); PICKl is not required for axonal targeting of mGluR7; however, in the absence of PICKl the receptor is no longer clustered [44]. In addition, the interaction with PICKl renders the receptor less susceptible to regulation by PKCa, which phosphorylates the C-terminus of mGluRT [46]. Thus, PICKl is also likely to interfere with receptor regulation. RGS12 also has related multifunctional properties. RGS12 is a GAP (GTPase activating protein) that is specific for Gtto and Gai; in addition, RGS12 contains an N-terminal PDZ domain that specifically recognises the C-terminus of the interleukin-8 receptor B (=CXCR2) but not the interleukin-8 receptor A (=CXCR1) or other GPCRs [47]. Finally, alternative splicing generates a C-terminus that binds to the N-terminal PDZ domain and thereby supports self-association of RGS12 by generating head-to-tail RGS12 concatemers [47]. It is evident that this arrangement favours the formation of clusters. A multidomain, actin-binding protein was also found to specifically bind to the C-terminus of the somatostatin-receptor-2 (SSTR2) and hence named SSTR-interacting protein (SSTRIP [48]); this scaffolding protein contains ankyrin repeats, an SH3- and a PDZ-domain; it is highly enriched in posts)aiaptic densities, and it was also independently identified by several groups (under various names, e.g. shankl and synamon). Shank proteins (of which there are, at present, 3 isoforms) also link other GPCRs to the actin cytoskeleton of presynaptic specialisations, e.g. CLl-receptor for the spider venom a-latrotoxin [49].

167 PDZ-domains recognise the C-terminal 4 amino acids; Homer also recognises a sequence in the C-terminus; hence, currently the focus is on the C-termini of GPCRs; it is generally assumed that these specify targeting in neurons and other polarised cells. However, additional sequence elements of GPCRs, including transmembrane segments [38], may also be important for mediating retention in specific membrane compartments. In many Gi-coupled receptors, the C-terminus is short and the third intracellular loop i3 is extended. It is therefore not surprising that this segment may also serve as a site of attachment. Spinophilin (a protein that is highly enriched in dendritic spines) has, for instance, been identified as an interaction partner for the D2-dopamine receptor [50]. Spinophilin is of particular interest because it is a multifunctional protein: the Nterminus of spinophilin binds to filamentous actin (F-actin), the C-terminus contains a PDZ-domain; the middle portion of the molecule associates with PPl (protein phosphatase-1) and also binds the third intracellular loop of the D2-dopamine receptor. It is very tempting to speculate that the D2-receptor is targeted to dendritic spines (e.g. in the intermediate size spiny neurons of the corpus striatum) by spinophilin. However, the crucial evidence is still missing, i.e. it has not been possible to co-immunoprecipitate D2-receptors and spinophilin [50]. This has recently been achieved for the 3 subtypes of the a2-adrenergic receptor [51]. Interestingly, the interaction of these receptors with spinophilin is enhanced by agonist occupancy indicating that spinophilin may also regulate signalling (i.e. possibly play a role in desensitisation). It is well established that the mobility of membrane proteins is restricted by the submembranous actin cytoskeleton; the contribution of microtubuli is less clear. The C-terminus of mGluRla binds to (monomeric) tubulin [52]. Because of the abundance of microtubules, the co-purification of tubulin with a given protein is generally viewed with suspicion. Agonist-stimulation promotes tubulin depolymerisation and recruits tubulin to the plasma membrane [53]. It is however not clear, if this is related to the direct interaction of tubulin with the receptor. There are several mechanisms by which receptors may affect microtubular dynamics, e.g. in vitro both, Ga (Gtts, Gai; [54]) and GPy [55] affect tubulin polymerisation; direct effects (=tubulin depolymerisation) have also been postulated for Gaq [56,57]. In addition, GRK2 and GRK5 bind tightly to tubulin which they also recognise as substrate [58,59]. Microtubuli participate in organising ligandgated ion channels into clusters (e.g. of the glycine receptor via the bridging protein gephyrin, [60]). In contrast, there is no firm evidence that microtubuli are important in retaining GPCRs at specific sites. Finally, while it is clear that the actin cytoskeleton and microtubuli must play a role in anchoring signalling components, their role must not be overemphasised because additional scaffolding components can actually maintain clusters even after disruption of micotubuli and actin depolymerisation [61]. Association of GPCRs with (non-G protein-dependent) effectors and regulators GPCRs generate two signalling molecules (Ga and Py-dimers) that each can control several effectors. Hence, it seems counterintuitive that these receptors require additional effector mechanisms. Nevertheless, GPCRs bind and regulate alternative (i.e. non-G protein regulated) effectors, e.g. the janus kinase JAK2 (which is activated; [62]) phospholipase Cyl (which is activated; [63]), endothelial NO-synthase (which is inhibited;

168 [64]); in addition they can bind adapter molecules (Grb2 and Nek) that recognise proline-rich stretches via SH3-domains and lead to activation of the ras-cascade ([65]). Two examples illustrate the contentious nature of direct effector binding to GPCRs: (i) the non-receptor tyrosine kinase src has been reported to bind directly to the Ps-adrenergic receptor, and this reaction was considered the mechanistic basis for activation of MAP kinase (=mitogen-activated protein kinase =erk) [66]. Similarly, regulation of a neuronal cation channel (via mGluRl in hippocampal neurons) was reportedly G protein-independent (i.e. insensitive to uncaging the inhibitory GDP analogue GDPPS) but blocked by the src-inhibitor PPl [67]. However, in adipocytes, i.e. the cells in which the Ps-receptor is expressed endogenously, erk activation depends on cAMP; i.e. the signal generated by the cognate signalling pathway of the Ps-receptor (Gg/adenylyl cyclase) is necessary and sufficient to account for erk activation [68]. It is also difficult to reconcile a G protein-independent activation of src with the recent observation that src can be directly activated by Gtts and Gaj [69] (related observations have been made with other non-receptor tyrosine kinases). In addition, it has long been known that PKA can phosphorylate src (on Ser^^ in the N-terminus [70]) and that elevation of cAMP can increase src kinase activity [71]. (ii) The C-terminus of the GABABI-receptor recognises transcription factors of the ATF/CREB family; the relevance of these findings in a matter of debate; in fact, the two studies that reported on the association reached opposite conclusions, i.e. that agonist occupancy of the GABAe-receptor removed ATF4 (=CREB2) from the nucleus (and hence presumably inhibited transcription; [72]) or translocated the ATF4 from the cytoplasm to the nucleus and activated transcription [73]. It is evident that in many cases the distinction between regulatory proteins and effectors is blurred; e.g. EBP50/NHERF mediates the effect of the P2-receptor on the Na^/H^-exchanger [74] and targets internalised receptors to recycling rather than degradation [36]. Similarly, phosphorylation of the P2-receptor and subsequent binding of P-arrestin does not only target the receptor to clathrin-dependent endocytosis but also redirects the signalling pathway from activation of cAMP accumulation to activation of the MAP-kinase cascade G protein [5,75]. Receptor recycling requires that the receptor is dephosphorylated. Accordingly, binding of protein phosphatases to phosphorylated receptors is to be anticipated. Protein phosphatase 2A (PP2A) binds to the proximal portion of Cterminus of CXCR2, i.e. the region adjacent to the 7* transmembrane helix; this binding requires the core of PP2A (=regulatory 65 kDa A-subunit + catalytic 36 kDa domain) and receptor intemalisation, but it is surprisingly independent of phosphorylation [76]. It is therefore not clear, if the docking of PP2A to CXCR2 plays a role in signalling other than receptor dephosphorylatioh. There must be components that account for the difference in G protein-coupling that is observed in reconstituted systems and in intact cells; G protein-coupled receptor can display an exquisite capacity to discriminate between closely related G protein subunits, if this selectivity is assessed in intact cells; this specificity, however, is lost if the selectivity is tested in reconstitution experiments (for review see [2]). For the Ap adenosine receptor (a Gj/o-coupled receptor), a component that imposes an inhibitory constraint on receptor-G protein coupling [77] has been identified; this protein, termed

169 coupling cofactor, also enhances the complex formation between Apieceptor and Gi; in contrast, when complexed to Go, the Apreceptor is insensitive to the action of coupling cofactor [78]. Binding of additional components also allows for cross-talk, signal integration and coincidence detection. Calmodulin serves as an illustrative example; several GPCRs bind calmodulin, but the consequences differ substantially. Binding of cahnodulin to the C-terminus of mGluRTA [79] is mutually exclusive with binding of the GPy; in fact, calmodulin is required to release GPy from the receptors, which supports inhibition of the neuronal Ca^^ channels [79]. This coincidence detection limits presynaptic autoinhibition to those nerve terminals that were actively releasing glutamate because only these contain elevated Ca^^-levels and hence Ca^^-liganded calmodulin. Binding of calmodulin to two other Gi/o-coupled receptors, the |Li-opioid and the D2-dopamine receptor, inhibits signalling and thus allows for cross-talk by receptors that raise intracellular Ca^^. However, the mechanisms differ: calmodulin prevents access of Ga to the C-terminal part of i3 of the |i-receptor [80]. In contrast, calmodulin interacts with the N-terminal portion of i3 of the D2-receptor; this allows for the formation of a receptor-G protein complex but impedes G protein activation [81]. Formation of receptor homo- and heterooligomers GPCRs have the ability to form oligomers, a phenomenon originally observed with the P2-receptor [82]. Although the functional implications of oligomer formation is not fully understood, there are examples where dimerisation is important: as mentioned earlier, the GABAB rreceptor is retained in the ER and binds agonist with low affinity; concomitant expression of the GABAB2-receptor results in efficient plasma membrane insertion and formation of functional receptors that bind agonists with high affinity [16]. Opioid K- and 5-receptors can form dimers; the pharmacological specificity of theses dimers is distinct from that of each individual receptor [83]. It is highly likely that oligomerisation of GPCRs is a more general phenomenon that does not only involve closely related receptors; heterodimer formation has, for instance, been anticipated based on functional data [84]. In fact, the formation of heterodimers between Di- and Ai-receptors [85] was proposed as the basis for the mutual antagonism exerted by activation of the receptors. A similar form of (mutual) antagonism is thought to underlie the crosstalk between D2- and A2A- receptors [84]. However, it has to be pointed that these experiments did not provide formal proof for heterodimer formation because (i) co-localisation of proteins by immunofluorescence is only circumstantial evidence and (ii) because co-immunoprecipitation may still reflect the formation of a large complex, in which the two receptors are trapped due to their association with associated scaffolding proteins. The most rigorous proof is to visualise heterooligomers in intact cells by FRET (Foerster resonance energy transfer) or by related methods that rely on bioluminescence (BRET=bioluminescence resonance energy transfer) using appropriately tagged proteins (e.g. donor-acceptor pairs of cyan and yellow fluorescent protein). Because the efficiency of (quantal) energy transfer declines with the 6th power of distance, FRET can only occur if the two proteins exist in a complex (as opposed to a short-lived, non-productive collision). By using these stringent criteria, the formation of homo- and heterooligomers has been demonstrated for several receptors (see also Table 1). While heterodimer formation allows for direct cross-talk between receptors (e.g. co-activation in [86]), the physiological significance of homodimer formation remains a matter of

170 debate. A modest increase of BRET (and hence of oligomer formation) of appropriately tagged P2-receptors has been observed upon addition of agonist; this is unrelated to intemalisation because manipulations that prevent endocytosis do not affect the agonistinduced increase in energy transfer [87]. However, a study that allowed for time-resolved measurements of cell-surface receptors indicates that 5-receptors are constitutive oligomers [88] and that there is essentially no impact of agonist (or antagonist). GPCRs heterodimerise with other transmembrane proteins: RAMPs (receptor-activity modifying proteins) are single transmembrane proteins that associate with receptors for calcitonin-gene related peptides, adrenomeduUin and related agonists. The ligand specificity of a given receptor depends on the association with a RAMP [89]. The Ds-dopamine receptor (but not the closely related Di-receptor) and (ionotropic) GABAA-receptors directly associate which results in reciprocal (inhibitory) modulation of their activities [90]. The A2B-receptor has been identified as the dimeric partner of DCC {deleted in colorectal cancer= netrin-receptor); this dimerisation was reportedly essential for high-affinity binding of netrin-1 for the chemoattracting action of netrin-1 on commissural axons [91]. However, a re-analysis that used more reahstic concentrations of A2B-antagonist failed to document any requirement of the A2B-receptor for directing axon outgrowth [92]. Conclusion The diversity of additional proteins that bind to GPCRs is bewildering; in many cases, the significance of the reaction is difficult to assess because a given receptor, clearly, cannot interact with these proteins simultaneously. The structure of G protein heterotrimers and of rhodopsin is known to atomic resolution. Hence, the sizes of the two proteins can be compared. The intracellular surface of rhodopsin barely suffices to cover the necessary contact sites on transducin [93]. There is little, if any, space left to accommodate additional (bulky) proteins. The formation of receptor multimers may offer a way out; it suffices if one or a few receptors in a cluster are actually tethered to and immobilised by the scaffolding protein(s). Given their propensity to form complexes, the remaining receptors may aggregate by homo- and heterotypic aggregation. It is, in this context, interesting to note that G proteins can also form large aggregates in their basal state [94,95]. Eventually some of the G protein-independent effects of receptors that were listed in this review may turn out to be of modest physiological relevance. While the physiological relevance of some of the protein-protein interactions listed above needs to be estabhshed, it is clear that receptors interact with proteins other than G proteins. Hence, stimulation and blockage of a receptor by agonists and antagonists, respectively, is expected to yield effects other than those achieved by activating or inhibiting the down-stream G protein [96]. This is one of the arguments that supports the concept of direct G protein ligands as an opportunity for drug development [97,98]. References 1. Freissmuth M, Casey PJ, Gilman AG (1989) FASEB J, 3:2125-2131. 2. Gudermann T, Schoneberg T, Schultz G (1997) Annu. Rev. Neuroscl 20:399-427. 3. Janetopoulos C, Jin T, Devreotes P (2001) Science 291:2408-11. 4. Pitcher JA, Freedman NJ, Leflcowitz RJ (1998) Annu. Rev. Biochem. 67:653-692. 5. Miller WE, Lefkowitz RJ (2001) Curr.Opin. Cell Biol. 13:139-145.

171 6. 7. 8. 9.

Berman DM, Oilman AG (1998) J. Biol Chem. 273: 1269-1272. Wieland T, Chen CK (1999) Naunyn Schmiedebergs Arch. Pharmacol 360:14-26. Lohse MJ, Bliiml K, Danner S, Krasel C (1996) Biochem. Soc. Trans. 24:975-980. Takesono A, Cismowski MJ, Ribas C, Bernard M, Chung P, Hazard S, Duzic E, Lanier SM (1999) J. Biol Chem. 274: 33202-33205. 10. Cismowski MJ, Ma C, Ribas C, Xie X, Spruyt M, Lizano JS, Lanier SM, Duzic E (2000) J. Biol Chem. 275:23421-23424. 11. Bernard ML, Peterson YK, Chung P, Jourdan J, Lanier SM (2001) J. Biol Chem. 276:1585-1593. 12. Rothman JE, Wieland FT (1996) Science 272: 227-234. 13. Tang Y, Hu LA, Miller WE, Ringstad N, Hall RA, Pitcher JA, DeCamilli P, Lefkowitz RJ (1999) Proc. Natl Acad. Set USA 96:12559-12564. 14. Ringstad N, Gad H, Low P, DiPaolo G, Brodin L, Shupliakov O, DeCamilH P (1999) iVewrow 24:143-154. 15. Schmidt A, Wolde M, Thiele C, Fest W, Kratzin H, Podtelejnikov AV, Witke W, Huttner WB, Soling HD (1999) Nature 401:133-141. 16. Marshall FH, Jones KA, Kaupmann K, Bettler B (l999)Trends Pharmacol Sci. 20: 396-399. 17. Pagano A et al. (2001) /. Neurosci. 21: 1189-1202. 18. Oksche A, Rosenthal W. (1998) J. Mol Med. 76:326-37 19. Ango,F, Pin JP, Tu, JC, Xiao B, Worley PF, Bockaert J, Fagni L (2000) J. Neurosci. 20:8710-8716. 20. Kato A, Ozawa F, Saitho Y, Fukazawa Y, Sugijyama H, Inokuchi, K. (1998) J. Biol Chem. li:^: 23969-23975. 21. Okamoto T, Schlegel A, Scherer PE, Lisanti MP (1998) J. Biol Chem. 273:5419-22. 22. Huang C, Hepler JR, Chen LT, Gilman AG, Anderson RGW, Mumby SM (1997) Mol Biol Cell 8: 2365-2378. 23. Mumby, S.M. (1997) Reversible palmitoylation of signaling proteins. Curr. Opin. Cell Biol 9:14S-154. 24. Rietveld A, Simons K (1998) Biochim. Biophys. Acta 1376: 467-479. 25. Dotti CG, Simons,K. (1990) Cell 62: 63-72. 26. Winckler, B., Forscher, P., Melhnann, L (1999) Nature 397:698-701. 27.Ledesma MD, Brugger B, Bunning C, Wieland FT, Dotti CG (1999) EMBO J. 18: 1761-1771. 28. Craig AM, Banker G (1994) Annu. Rev. Neuroscience 17:267-310. 29. Wozniak M, Limbird LE (1998) Brain Res. 780: 311-322 30. Hirokawa,N (1998) Science 279: 519-526. 31. Klopfenstein, D.R., Vale, R.D., Rogers, S.L. (2000) Cell 103:537-540. 32. Tai AW, Chuang JZ, Bode C, Wolfrum U, Sung CH (1999) Cell 97: 877-887. 33. Wolfrum U, Schmitt A (2000) CellMotil Cytoskeleton 46: 95-107. 34. Saunders C, Limbird LE (1997) J. Biol Chem. Ill: 19035-19045. 35. Saunders C, Limbird LE (2000) Mol Pharmacol 57:44-52. 36. Cao TT, Deacon HW, Reczek D, von Zastrow M (1999) Nature 401:286-290. 37. Lamaze C, Fujimoto LM, Yin HL, Schmid SL (1997) J. Biol Chem. 272:20332-35. 38. Saunders C, Keefer JR, Bonner CA, Limbird LE (1998) J. Biol Chem. 273:2419624206. 39. Brakeman PR, Lanahan, AA, O'Brien R, Roche K, Barnes CA, Huganir RL, Worley PF (1997) Nature 386: 284-288.

172 40. Tu JC, Xio B, Yuan JP, Lanahan AA, Leoffert K, Li M, Linden DJ, Worley PF (\99S) Neuron 21: 717-726. 4L Tadokoro S, Tachibana T, Imanaka T, Nishida W, Sobue K (1999) Proc. Natl Acad. Sci. ^5y4 96:13801-13806. 42. Shigemoto R, Kulik A, Roberts JD, Ohishi H, Nusser Z, Kaneko T, Somogyi P (1996) A^arwre 378: 96-98. 43. Stowell JN, Craig AM (1999) Neuron 22: 525-536. 44.Dev KK, Nakajima Y, Kitano J, Braithwaite SP, Henley JM, Nakanishi S (2000) /. Neurosci. 20:7252-7257 45.ElFar O, Airas J, Wischmeyer E, Nehring RB, Karschin A, Betz H (2000) Eur. J. Neurosci. 12:4215-4221. 46. Boudin H, Doan A, Xia J, Shigemoto R, Huganir RL, Worley PF, Craig AM (2001) Neuron 28:485-497 47. Snow BE et al.(1998) J. Biol. Chem. 273: 17749-17755. 48.Zitzer H, Honck HH, Bachner D, Richter D, Kreienkamp HJ (1999) J. Biol Chem. 274:32997-33001 49. Tobaben S, Sudhof TC, Stahl B (2000) J.Biol.Chem. 275:36204-36210 50. Smith FD, Oxford GS, Milgram SL. (1999) J. Biol. Chem. 274:19894-19900. 51.Richman JG, Brady AE, Wang Q, Hensel JL, Colbran RJ, Limbird LE (2001) J. Biol. Chem. 276, in press 52. Ciruela F, Robbins MJ, Willis AC, Mcllhinney RA(1999) J. Neurochem. 72:346-54. 53. Ciruela F, Mcllhinney RA (2001) J. Neurochem. 76:750-757. 54. Roychowdhury S, Rasenick MM (1997) J. Biol Chem. 272: 31576-31581. 55.Roychowdhury S, Panda D, Wilson L, Rasenick MM. (1999) J. Biol Chem. 274: 13485-13490. 56.Ravindra, R, KunapuU SP, Forman LJ, Nagele RG, Foster KA, Patel SA. (1996) / Cell Biochem. 61:392-401 57. Popova JS, Rasenick MM (2000) J. Neurosci. 20:2774-2782. 58. Pitcher JA, Hall RA, Daaka Y, Zhang J, Ferguson SS, Hester S, Miller S, Caron MG, Lefkowitz RJ, Barak LS (1998) J. Biol Chem. 273:12316-12324 59. Carman CV, Som T, Kim CM, Benovic JL (1998). J. Biol Chem. 273:20308-20316. 60. Kirsch J, Wolters I, Triller A, Betz H (1993) Nature 366:745-748. 61. Allison DW, Chervin AS, Gelfand VI, Craig AM (2000) J. Neurosci. 20:4545-4554. 62. AH MS, Sayeski PP, Dirksen LB, Hayzer DJ, Marrero MB, Bernstein KE (1997) J. Biol Chem. 272:23382-23388. 63. Venema, R.C., Ju H, Venema, V.J., Schieffer, B., Harp, J.B., Ling, B.N., Eaton, DC, Marrero MB (1998)7. Biol Chem. 273: 7703-7708. 64. Marrero MB, Venema VJ, Ju H, He H, Liang H, Caldwell RB, Venema RC (1999) Biochem. J. 343:335-340. 65. Oldenhof J, Vickery R, Anafi M, Oak J, Ray A, Schoots O, Pawson T, von Zastrow M, Van Tol HH (1998) Biochemistry 37:15726-15736. 66. Cao W, Luttrell LM, Medvedev AV, Pierce KL, Daniel KW, Dixon TM, Lefkowitz RJ, Collins S (2000) J. Biol Chem. 275:38131-38134. 67. Heuss C, Scanziani M, Gahwiler BH., Gerber U (1999) Nat. Neurosci. 2:1070-1077. 68.Lindquist JM, Fredriksson JM, Rehnmark S, Cannon B, Nedergaard J. (2000) J. Biol Chem. 275: 22670-22677. 69. Ma, Y.C., Huang, J., Ali, S., Lowry, W., Huang, X.Y. (2000) Cell 102:635-46.

173 70. Yaciuk P, Choi JK, Shalloway D.(1989) Mol Cell Biol 9:2453-63. 71. Roth CW, Richert ND, Pastan I, Gottesman MM (1983) J. Biol Chem. 258:1076810773 72. Vernon E, Meyer G, Pickard L, Dev K, Molnar E, Collingridge GL, Henley JM (2001) Mol Cell Neurosci. 17:637-645 73. White JH, Mclllhinney RA, Wise A, Ciruela F, Chan WY, Emson PC, Billinton A, Marshall FH (2000) Proc. Natl Acad. Set USA 97:13967-13972. 74. Hall RA et al. (1998) Nature 392: 626-630. 75.Daaka Y, Luttrell LM, Ahn S, Delia Rocca G, Ferguson SS, Caron MG, Lefkowitz RJ(1998) J. Biol Chem. 273:685-688. 76. Fan GH, Yang W, Sai J, Richmond A. (2001) J. Biol Chem. 276, in press 77.Nanoff C, Mitterauer T, Roka F, Hohenegger M, Freissmuth M (1995) Mol Pharmacol 4S:S06-S\1 78. Nanoff C, Waldhoer M, Roka F, Freissmuth M (1997) Neuropharmacol 36:1211-19 79. O'Connor V, ElFar O, Bofill-Cardona E. Freissmuth M, Nanoff C, Karschin A, Airas JM, Betz H, Boehm S (1999) Science 286: 1180-1184. 80. Wang, D., Sadee, W., Quillan, J.M. (1999) J. Biol Chem. 274:22081-22088. Sl.Bofill-Cardona E, Kudlacek O, Yang Q, Ahom H, Freissmuth M, Nanoff C (2000) 1 Biol Chem. 275: 32672-32680. 82.Hebert TE, Moffett S, Morello JP, Loisel TP, Bichet DG, Barret C, Bouvier M. (1996) J. Biol Chem. 211: 16384-16392. 83. Jordan BA, Devi LA (1999) Nature 399: 697-700. 84. Fuxe K, Ferre S, ZoH M, Agnati F (1998) Brain Res. Brain Res. Rev. 26:258-273. 85. Gines S et al. (2000) Proc. Natl Acad. Set USA 97: 8606-8611 86. Rocheville M, Lange DC, Kumar U, Patel SC, Patel RC, Patel YC (2000) Science 288:154-157. 87. Angers S, Salahpour A, Joly E, Hilairet S, Chelsky D, Dennis M, Bouvier M (2000) Proc. Natl Acad Scl USA 97:3684-3689. 88.McVey, M, Ramsay D, Kellett E, Rees S, Wilson S, Pope AJ, Milligan G (2001) J. Biol Chem. 276:14092-14099. 89. Foord SM, Marshall FH (1999) Trends Pharmacol Scl 20: 184-187. 90. Liu F, Wan Q, Pristupa Z, Yu XM, Wang Y, Niznik H (2000) Nature 403: 274-280. 91. Corset V, Nguyen-Ba-Charvet KT, Forcet C, Moyse E, Chedotal A, Mehlen P (2000) Nature 407:747-750. 92. Stein E, Zou Y, Poo M, Tessier-Lavigne M. (2001) Science 291:1976-1982. 93. Palczewski K et al. (2000) Science 289:739-745. 94.Mixon MB, Lee E, Coleman DE, Berghuis AM, Gilman AG, Sprang SR. (1995) Science 270: 954-960. 95. Jahangeer S, Rodbell, M. (1993) Proc. Natl Acad Scl USA 90:8782-8786. 96. Hoeller C, Freissmuth M, Nanoff C (1999) Cell Mol Life Set 55: 257-270 97. Freissmuth M, Waldhoer M, Bofill-Cardona E, Nanoff C (1999) Trends Pharmacol Scl 20:237-245. 98.Niimberg B, Togel W, Krause G, Storm R, Breitweg-Lehmann E, Schunack W (1999). Eur. J. Med Chem. 34, 5-30. 99.Saura, CA, Mallol J, Canela EI, Lluis C, Franco R. (1998) J. Biol Chem. 273: 17610-17617.

This Page Intentionally Left Blank

H. van der Goot (Editor) Trends in Drug Research III © 2002 Elsevier Science B.V. All rights reserved

175

Mechanisms of action of antipsychotic drugs: the role of inverse agonism at the D2 dopamine receptor Philip G Strange School of Animal and Microbial Sciences, University of Reading, Whiteknights, Reading, RG6 6AJ, UK

The antipsychotic drugs provide the main treatment for the serious brain disorder schizophrenia. They have enabled many people to lead more normal lives but their use is associated with serious side effects. Their precise mechanism of action is still unclear and it would help in the design of new antipsychotics if this could be defined more clearly. In this article I wish to consider some of the possible mechanisms that have been proposed for antipsychotic action.

Schizophrenia Schizophrenia is a serious brain disorder affecting about 1% of the population [1]. The symptoms experienced by schizophrenics are varied but may be categorised in to two groups. The positive symptoms are those that are additions to normal behaviour and comprise symptoms such as thought disorder, perceptual disturbances e.g. hallucinations, abnormal beliefs or delusions. The negative symptoms are those that subtract from normal behaviour and comprise symptoms such as social withdrawal, poverty of thought and speech, blunted affect, reduced motor function.

Antipsychotic drugs The antipsychotic drugs were first introduced in the 1950's with the use of chlorpromazine [2]. Subsequently a very wide range of drugs with differing chemical structures has been used for the treatment of schizophrenia. For the most part these drugs treat only the positive symptoms of the disorder, having little effect on the negative symptoms. The reduction in positive symptoms is very important for patients but it would be of great use to be able to treat the negative symptoms as well. Some drugs, notably clozapine, have been reported to reduce negative symptoms [3] and there is a considerable drive within the pharmaceutical industry to generate new drugs that can reduce negative symptoms. The antipsychotic drugs also elicit side effects (extrapyramidal side effects) that are mostly motor in character. Early in treatment patients can experience parkinsonian-like side effects with symptoms such as tremor and slowness of movement. After months or

176 years of treatment patients can experience tardive dyskinesia which is associated with abnormal involuntary movements of the tongue and lips. The therapeutic effects of the antipsychotic drugs are not immediate and in fact take up to four weeks to reach a maximum level. It seems that there may be some adaptive process occurring in the brain that needs time to occur.

The role of dopamine receptors in antipsychotic action Following intensive investigation of the properties of the antipsychotic drugs it became apparent that these drugs were affecting dopamine systems [2]. For example it was shown that these drugs would inhibit the abnormal behaviour seen in experimental animals following administration of amphetamine or apomorphine, both of which mimic the actions of dopamine. It was proposed that the antipsychotics were blocking the postsynaptic receptors for dopamine. In the late 1970's it became possible to assay the different dopamine receptor subtypes (Di, D2) defined using biochemical and pharmacological tests. In 1976, two labs reported a strong correlation between the affinities of a range of antipsychotic drugs for the D2 dopamine receptor and their daily doses for treating schizophrenia [4,5]. No such correlation was seen for other receptors including the Di dopamine receptor and it was concluded that antagonist action at D2 dopamine receptors was an essential part of the mechanism of action of the antipsychotics. In the late 1980's the dopamine receptors were subjected to attention by molecular biologists and five dopamine receptor sequences were cloned (Di, D2, D3, D4, D5) [6,7]. Subsequent analysis of the pharmacological properties of the gene products has shown that these may be divided in to two sub-families that are related to the original D1/D2 subdivision based on biochemistry/pharmacology. Thus Di and D5 are related to one another and to the original Di receptor and so are termed the Dplike receptors. D2, D3, D4 are related to one another and to the original D2 receptor and are termed the D2-like receptors. This meant that the effects of antipsychotic drugs on dopamine receptors could have been at D2 or D3 or D4 receptors. D3 and D4 receptors were of particular interest in this respect as they are found largely in limbic/cortical areas of brain that are likely to be regions where antipsychotic effects would be mediated. In principle, therefore, a drug with selective actions at D3 or D4 receptors might be an antipsychotic but without the extrapyramidal side effects as these are thought to be mediated via D2 receptors in the striatum, a brain region lacking D3 and D4 receptors. The role of these D2-like receptors in antipsychotic action is unclear but some clues may be obtained by considering the affinities of different antipsychotics for these receptor subtypes. If the affinities of antipsychotic drugs for the different receptor subtypes are considered then it seems that some drugs have much lower affinities for the D4 receptor whereas all of the drugs have high affinities for D2 and D3 receptors. It seems, therefore, that occupancy of D4 may not be mandatory for antipsychotic action.

177 In the years since these subtypes were defined the pharmaceutical industry has expended much effort to make compounds selective for the different subtypes, especially the D3 and D4 receptors. A D4 selective compound (L745870) was synthesised by Merck Sharp and Dohme and in clinical trials this proved to be without effect in schizophrenia [8]. This supports the idea that the D4 dopamine receptor does not seem to play a major role in the therapy of schizophrenia. We may conclude that the principal sites of action of the antipsychotic drugs are the D2 and D3 dopamine receptors [9] but the relative role of these two subtypes is unclear. Clinical testing of selective D2 and D3 receptor agents against psychosis in schizophrenia will help elucidate this. Several other receptors have been proposed as being important for the actions of antipsychotic drugs including the serotonin receptors (5HT2 and 5HTIA) [10,11]. It has been proposed that actions of the antipsychotics at these receptor subtypes may be important for suppressing side effects or promoting additional beneficial clinical actions of the drugs.

The antipsychotic drugs as inverse agonists It had been the dogma for many years that the antipsychotic drugs were acting as antagonists at the D2-like dopamine receptors. In the 1990's, however, for several drugs that had been considered to be antagonists at other receptors, it became apparent that these were in fact inverse agonists. This was seen in assays where the drugs were able to inhibit the agonist-independent activation of receptor-associated signalling systems. For the dopamine receptors it has been shown that the antipsychotic drugs possess inverse agonism at Di, D2, D3 and D5 receptors [12-16]. The inverse agonism is best characterised for the D2 dopamine receptor so I shall focus on this receptor subtype. The first report of inverse agonism at this receptor came from studies on the D2 receptor expressed in pituitary cells [15]. The antipsychotic drug, haloperidol, was found to stimulate prolactin release from these cells. Dopamine normally inhibits release so that haloperidol was acting as an inverse agonist. Subsequently, inverse agonism has been demonstrated from the ability of the drugs to potentiate forskolin stimulated cAMP production [13] or to inhibit stimulation of [^^S]GTPYS binding [16]. A detailed examination of the inverse agonism of the antipsychotic drugs was performed using a CHO cell line expressing the native D2 dopamine receptor at high levels (4-6 pmol/mg) [13]. Using these cells it was possible to show that all of the antipsychotic drugs tested exhibited inverse agonism at the D2 receptor based on their ability to potentiate forskolin stimulated cAMP production. This was seen for different chemical classes of drug and for drugs with different clinical profiles e.g. high or low side effects. The extent of inverse agonism was similar for each drug so that they all appear to be full inverse agonists and there is a good correlation between inverse agonist potency and binding affinity. Only one substance has been found so far that is a neutral antagonist. This is the aminotetralin (+)-UH-232. This compound was found to be a

178 neutral antagonist in studies on the regulation of forskolin-stimulated cAMP production in CHO cells expressing D2 receptors, and in fact this compound would inhibit competitively both the agonist effects of dopamine and the inverse agonist effects of (+)-butaclamol. The ability to detect inverse agonism, in this system with the native receptor, was very dependent on the expression level of the receptor in the CHO cells. In order, therefore, to provide a better detection system for inverse agonism at the D2 receptor we constructed a mutant receptor with changes in the amino acid sequence in the third intracellular loop at the base of the sixth transmembrane region. This is a region that has been found for other receptors to be "hot spot" for generating constitutively active receptors i.e. receptors active in the absence of agonist and therefore better for detecting inverse agonism. For the D2 dopamine receptor the mutation was T343R [17]. The T343R mutant D2 receptor exhibited the expected properties i.e. increased agonist affinity and potency and increased agonist-independent activation. The mutant receptor also gave responses to inverse agonists at lower expression levels, where the native receptor would not respond to inverse agonists. The range of antipsychotics that are inverse agonists has been extended using the T343R mutant receptor and again all antipsychotics tested exhibit inverse agonism including a set of atypical antipsychotics such as clozapine, risperidone, olanzapine, quetiapine. Despite the increased sensitivity of the T343R mutant receptor to inverse agonists (+)-UH-232 still behaves as a neutral antagonist in this test system. As a further test of the activity of (+)-UH-232 we examined its efficacy in CHO cells that had been treated with sodium butyrate in order to increase agonist responsiveness [18]. In these cells weak partial agonists exhibit increased efficacy but (+)-UH-232 remains a neutral antagonist in this system. In some systems with high amplification (FLIPR [19], microphysiometer [20]), however, (+)-UH-232 exhibits weak agonism. We may conclude, therefore, that (+)-UH-232 is a weak agonist or neutral antagonist unlike the antipsychotic drugs that are full inverse agonists.

The relevance of the inverse agonism of the antipsychotic drugs to their clinical effects It seems that all of the antipsychotic drugs possess inverse agonism at the D2 dopamine receptor. It is important to ask whether this is relevant to their clinical effects or whether it is a curiosity of the systems used to detect the activity. One way to approach this question would be to identify a neutral antagonist for the D2 receptor and assess its clinical effects. (+)-UH-232 seems to a suitable compound to use in this assessment as it has activity close to that of a neutral antagonist at the D2 receptor as outlined above. This compound has been used in a trial in schizophrenia and it was shown to be without antipsychotic activity [21]. Although these data relate only to one compound they at least support the proposal that the inverse agonism of the antipsychotic drugs is important for their clinical effects. A link, however, with the extrapyramidal side effects of the drugs cannot be excluded.

179 If this is correct then we need to ask what the mechanisms of the antipsychotic effect might be in relation to the inverse agonism exhibited by the drugs. Here we need to bear in mind that the effects of the drugs are not immediate and in fact take several weeks to reach a maximal level. There are many possible effects of the chronic treatment of drugs on neuronal systems but the delayed time course of the effects of the antipsychotic drugs suggest that there is some adaptive process occurring in the relevant neuronal system. One suggestion has been that the balance between the different dopamine neuronal systems is altered upon chronic antipsychotic treatment in favour of certain mesocortical neurones [1]. This would then change the balance of activity between different brain regions. One, well described, effect of the chronic administration of antipsychotics to experimental animals is an increase in the number of D2 dopamine receptors in the brain [22,23]. It has been shown that the chronic administration of dopamine agonists to animals will lead to a down regulation of D2 dopamine receptors. It has been assumed, therefore, that the up-regulation of D2 receptors following antipsychotic treatment is due to blockade of the access of dopamine to its receptors. It may, however, be that the effect is a reflection of the inverse agonist property of the drugs. There is some evidence from other G protein coupled receptor systems to show that such up-regulation might be a response to the inverse agonism of the drugs rather than receptor blockade alone [24]. Neutral antagonists, where these have been tested in parallel with inverse agonists, fail to elicit up-regulation. The neutral antagonists would still prevent access of agonist to the receptors and hence the lack of up-regulation with these compounds supports the idea that up-regulation is a reflection of inverse agonism. It has been shown that the antipsychotic drugs will cause up-regulation of D2 dopamine receptors expressed in recombinant cells [25]. There can be no dopamine present in these systems so the drugs cannot be preventing access of agonist to the receptors. In this case the up-regulation may provide an index of the effects of the drugs directly on the receptor. The effects of the drugs may, therefore, be a reflection of the conformational change that these compounds elicit in the receptor. We have examined a range of drugs for their effects on D2 receptors expressed in CHO cells and preliminary data suggest that the antipsychotic drugs tested do cause D2 receptor upregulation whereas (+)-UH-232 does not [26]. These observations provide a further suggestive link between the inverse agonism of the antipsychotic drugs and their clinical effects. Therefore a plausible theory of antipsychotic action may be proposed as follows. In this, the antipsychotic drugs are inverse agonists and this property is important for their therapeutic effects. The long-term use of the drugs changes the sensitivity of certain synapses in the brain and thus achieves an antipsychotic effect. It is still unclear as to which synapses in the brain are affected, although some suggestions have been made above. It is also unclear as to whether the up-regulation of D2 dopamine receptors in the brain is part of this change in synaptic efficacy.

180 References 1. Strange PG (1992) Brain Biochemistry and Brain Disorders, OUP 2. Leysen JE and Niemegeers CJE Handbook of Neurochemistry 9 (1985) 331 3. Kane JM. CNS Drugs 7 (1997) 347-348 4. Seeman P, Lee T, Chau-Wong M amd Wong K Nature 261 (1976) 717 5. Creese I, Burt DR and Snyder SH Science 194 (1976) 546 6. Civelli, O, Bunzow JR and Grandy DK. Ann. Rev. Pharmacol.Toxicol. 32 (1993) 281 7. Missale C, Nash SR, RibinsonSW, Jaber M and Caron MG Physiol Rev 78 (1998) 189 8. Bristow LJ, Kramer MS, Kulagowski J, Patel S, Regan CI and Seabrook GR. Trends Pharmacol Sci 18(1997)186 9. Strange PG Pharmacol Rev 53 (2001) 119 10. Meltzer HY, Park S and Kessler R Proc Natl Acad Sci USA 96 (1999) 13591 11. Millan MJ J Pharmacol Exp Therap 295 (2000) 853 12. Charpentier S, Jarvie KR, Severynse DM, Caron MG and Tiberi M J Biol Chem 271(1996)28071 13. Hall DA and Strange PG Brit J Pharmacol 121 (1997) 731. 14. Griffon N, Pilon C, Sautel F, Schwartz JC and Sokoloff P J Neural Trans 103 (1996) 1163 15. Nilsson CL and Eriksson F. J Neural Trans 92 (1993) 213 16. Kozell LB and Neve KA Mol Pharmacol 52 (1997) 1137 n.Wilson J, Fu D, Lin H, Javitch JA and Strange PG J Neurochem 77 (2001) 493 18. Marston DM and Strange PG, Brit J Pharmacol 131 (2001) 5P 19. Pauwels PJ, Finana F, Tardif S, Wurch T and Colpaert FC J Pharmacol Exp Therap 297 (2001)133 20. Coldwell MC, Boyfield I, Brown AM, Stemp G and Middlemiss DN Brit J Pharmacol 127(1999) 1135 21.Lahti AC, Weiler M, Carlsson A and Tamminga CA. J Neural Trans 105 (1998) 719. 22. Creese I and Sibley DR Ann Rev Pharmacol Toxicol 21 (1981) 357 23. Lidow MS, Williams GV and Goldman-Rakic PS Trends Pharmacol Sci 19 (1998) 136 24. Leurs R, Smit M, Alewijnse AE and Timmerman H Trends Biochem Sci 23 (1998) 418 25. Sibley DR and Neve KA (1997) in The dopamine receptors. Neve KA and Neve RL eds Humana Press, Totowa, New Jersey pp383-424 26. Marston DM, Kennedy M and Strange PG unpublished observations

H. van der Goot (Editor) Trends in Drug Research III © 2002 Elsevier Science B.V. All rights reserved

181

Agonist channeling of a2-adrenoceptor function Karl E.O. Akerman, Johnny NSsman, Tomas Holmqvist and Jyrki P. Kukkonen Department of Physiology, Division of Cell Physiology, Uppsala University, BMC, Box 572, SE-75123 Uppsala, Sweden

Introduction The a2-adrenergic receptors (a2-ARs) control the function of different organs via central and peripheral effects (Table 1). Ligands for these receptors have several therapeutic applications includmg anaesthesia and treatment of hypertension and glaucoma. Potential new indications may include obesity and psychiatric disorders. The central effects are thought to mainly be a result of presynaptic mhibition of noradrenaline release. However, a2-ARs are localised at postsynaptic sites as well. The mechanisms involved in the peripheral actions of a2-AR ligands are less well understood. Their effects can be both inhibitory and stimulatory depending on the organ in question [1].

Table 1. Physiological effects of ai adrenoceptor activation Central effects • Reduction in plasma adrenaline and noradrenaline • Inhibition of neurotransmitter release • Sedation • Food intake • Hypotermia • Antinociception

Peripheral effects Inhibitory Stimulatory • Insulin secretion • Vascular tonus (e.g. • Vasopressin secretion coronary & pulmonary • Tyroxine secretion venal tonus) • Saliva secretion • TSH secretion • Gut movement (ileus) • Growth hormone • Lipolysis secretion p Glycogenolysis • Platelet aggregation • Water absorption (gut)

The variety of responses elicited by these receptors indicates that they may possess the ability to activate multiple signal transduction pathways [2]. The response elicited by the receptors depends on the tissue where these receptors are expressed as well as the subtype of receptor [3-10]. The classical cellular response observed upon activation of a2-ARs is an inhibition of sthnulated cAMP production [11]. However, changes in cAMP alone do not explain the physiological actions of a2-ARs. Three subtypes of a2ARs, a2A, 0C2B and a2c, have been identified by cloning and pharmacological tools.

182 Many of the systemic eflfects of a2-AR ligands are due to actions on presynaptic receptors, which inhibit transmitter release from in neuronal cells via inhibition cf voltage-gated N- or P/Q-type Ca^^ channels [12-14]. Like the inhibition of cAMP production this response is sensitive to pertussis toxin and therefore likely to be mediated by Gi/o-type G proteins. Stimulation of cAMP accumulation Activation of a2-AR has also been shown to cause an increase rather than a decrease of cAMP production both through recombinantly expressed [5,9,10,15] and endogenous a2-ARs [16-18]. In certain cells this stimulatory response is seen only at high agonist concentrations or if Gi-type G proteins are inactivated by pertussis toxin-treatment [6,10]. In many cell types the primary response seems to be stimulation of cAMP production and in many cases particularly the a2B-AR subtype has been implicated in the stimulatory coupling [10,19,20]. With all the subtypes, there are some agonists (e.g. clonidine, oxymetazoline, UK 14,304), though different for each subtype, that prefer coupling to an inhibitory response and have only a very weak stimulatory action [20-22]. The phenomenon where certam agonists can selectively activate specific signal pathways has been termed agonist trafficking (of receptor signals) [23]. The basis for this is that different active receptor conformations will preferentially activate specific G proteins. Adenylyl cyclase (AC) can be stimulated directly through GSOL and directly or indirectly by other messengers [24] (Table 2). In many studies the stimulation cf cAMP production by a2-ARs and other Gj-coupled receptors has thus been interpreted as Gs coupling. However, many other possibilities exist such as effects via G protein p y subunits (Gjiy)? Ga^^ and protein kinase C (PKC). The effect of receptor activation would be dependent on the transduction mechanisms it utilises in the cells in question and on the respective AC expressed in the cells studied. In PC-12 cells, for instance, chelation of intracellular Ca^^ with BAPTA reduces the a2-AR-stimulated cAMP accumulation [5]. Table 2. Activation of adenylyl cyclase isoforms by different effectors Isoform activator Inhibitor Typel Type III, VIII Type II, IV, VII Type V, VI Type IX

Gsa, Ca'* G^a, Ca^"^ G,a, PKC, GpY

Gsa Gsa

GpY, Gitt

Gia Gia (?) Gitt, Ca^^

Gia

To test whether py subunits would be involved in the stimulation of cAMP accumulation via the human a2B-AR in Sf9 cells the effects of Gpy and P-ARK (which functions as a Gpy scavenger) on forskolin-stimulated cAMP accumulation was tested (Fig. 1). While p-ARK enhanced forskolin-stimulated cAMP accumulation the GPy coexpression considerably reduced cAMP accumulation. This means that the adenylyl cyclase of Sf9 cells is inhibited by Py rather than stimulated. The a2B-AR-stimulated AMP accumulation in Sf9 cells can thus not be due to Py. The basal and forskolin-

183 stimulated cAMP accumulation are unaffected by stimulation of protein kinase C while a considerable stimulation of coexpressed type II adenylyl cyclase can be observed (manuscript in preparation). Ca^"*" enhances the forskolin-stimulated cAMP accumulation [25]. However, stimulation of cAMP accumulation is seen also in cells depleted of intracellular Ca^"*" [10]. Thus, the typical second-messenger- or CPy-niediated effects on adenylyl cyclase can be excluded in the case of the a2B-AR.

p-ARK

3-1

^2.5!

control

0

3

21 —

0 j = .

0.5^ n J

n basal forsk

r^ basal

forsk

n basal

forsk

Fig. 1. Effect of P-ARK and Gpy expression on basal and forskolin-stimulated (forsk) cAMP accumulation in Sf9 cells. The methods used are essentially those described in references [10,22,25]. Cells were infected with recombinant baculovirus harboring the genes for either PARK (gift from Dr. A. DeBlasi, Consorzio Negri Sud, Santa Maria Imbaro, Chieti, Italy) or pi and Y2 (gifts from Dr. T. Haga University of Tokyo, Tokyo, Japan) subunits in the same recombinant virus. The infection time was 26 hours. Activated ai-ARs have been shown to coprecipitate Gs proteins [6]. Mutagenesis studies on a2A-AR and a2B-AR suggest that the stimulatory response can be specifically modified by structural changes in the second intracellular loop [22] and in the Nterminal or C-terminal portion of the third intracellular loop [26]. Certain basic residues in the C-terminal portion are important for the stimulatory coupling [27]. The approximate positions of these domains in the intracellular loops of rhodopsin [28] are illustrated in Fig. 2. These findings, together with the fact that certain agonists can preferentially couple to inhibition of adenylyl cyclase, suggest that the inhibition and stimulation of adenylyl cyclase via a2-ARs require different structural determinants or are induced by different conformations of the receptor.

184

Fig. 2. The approximate position of domains important for stimulatory coupling super imposed on the intracellular surface of rhodopsin deduced from its crystal structure . The outline of the Cterminus second (i2) and third (i3) loops as well as the approximate positions of helices III, IV, V and VI are marked. Note that the i3 loop of a2-ARs is much longer than the corresponding loop in rhodopsin. The areas where stimulation of cAMP accumulation by a2-ARs are affected by mutagenesis are dotted black. The ability of different a2-AR agonist to enhance cAMP accumulation as compared to their ability to activate coexpressed Gi, determined by ^^S-GTPyS binding, is shown in Fig. 3. Many of the ligands, like UK14,304 and clonidine, showed a weaker ability to stimulate cAMP accumulation while they were almost as effective as noradrenaline in coupling to Gi. With Gi coexpression these ligands inhibited rather than stimulated cAMP accumulation (Fig. 4). A small inhibiton was seen even with those ligands, which showed a high degree of stimulation when the Gi expression levels were further increased with longer incubation times. These results demonstrate that Gi can override the stimulatory coupling. They also indicate that the ability of the different ligands to couple to Gi is similar but the coupling to Gs is ligand-dependent. Thus when the receptor is activated by these ligands its ability to couple to stimulation of cAMP production is weaker than the coupling to inhibition. This would suggest that the Gs "affinity" of the receptor conformation induced by these ligands is lower. The results also suggest that Gi coexpression considerably attenuates the stimulatory coupling.

185 OH

H

Y

HO

adrenaline

CH3

HjC

^

(CH^C^

noradrenaline

^ oxymetazoline

CH3

N

^^xL ^CH

HO

dexmedetomidine

a-methyl-noradrenaline

guanabenz

NH2

t> ^"CP" BHT933

UK 14,304

clonidine

r 1.2 O

3

5'

crq

o

o

N

C

I O

7^ O

•«

N

£

o

B o

0)

^

C

s

to/)

^ 0

m en

^ ^

H

m T-H

;D

G^

UJ «

S X T3

Fig. 3 Effect of different a2-AR agonists on cAMP accumulation (empty bars) and S-GTPyS binding (gray bars) to co expressed Q (Qa and Gpy). The methods used are essentially those described in references [10,22,25]. The data is normalised with respect to noradrenaline.

186

u s

0.3

o

.s o

I o

c3

N

2

*| o

s s

T3

o

X

S 1 d

S


O

0.1 H

n

A


^a

O o

;T^

O

-0.1 H

-0.3

T N C 0)

-^ o

en en ON

(D C

"2 S 73

Fig. 4. Effect of Gi coexpression at 26 (light grey) and 38 hours of infection (dark grey) on agonist-mediated cAMP accumulation (a2B adrenoceptors expressed in Sf9 cells). Otherwise conditions are the same as in Fig. 3. The data suggest that the two G proteins Gi and G^ interact with the same receptor. A model where two G proteins Ga and Gb competes for the same site of the receptor was developed (Fig. 4). This would not necessary mean that they compete for the same site but that coupling to one G protein excludes the coupling to another. The agonist activation of the receptor was modelled according to the equation:

r^l ^ [agonist][Rl^,_ ^ ^"^ [agonist] + EC50-R

187 i[R]aci = concentration of activated receptors; [R]act-max = maximum concentration of receptors that can be activated; EC50-R = the concentration of agonist producing the halfmaximal receptor activation). The efficacy could be regulated by changing the [R]act-max in an analogous fashion as in [25] and EC50-R. This activated receptor acts as an activator for two G proteins, Ga and Gb. The activation of Gb can be described according to the equation

[^]„JGJ„ [GbL=[^L+EC 50-G. 1 +

G

50

y

and the activation of Ga in a corresponding way. The effect of Ga on Gb activation is contained within the term (l+[Ga]max]/Ga-i5Q). The effect of the Ga expression level on the response via Gb is shown in Fig. 5 at different relative affinities of the two G proteins for the receptor. When the affinity for the Gb site is reduced Ga will attenuate the effect of G5.

Gj^/G^, affinity for receptor 1 10

0

0

1 10

0

1 10

100

0

_•. 10

expression level (G,JG^) Fig 5. Simulation of the consequences of overexpression of one G protein G^ on the activity of another G protein Gb with the same or lower affinity for the same receptor.

188 Thus if the conformations induced by ligands such as UK 14,304 and clonidine have a lower "affinity" for Gs (=Gb) then Gi (=Ga) overexpression would lead to a considerable attenuation of the Gs response.

Coupling to Ca mobilisation In many cells a2-ARs have been shown to couple to production of inositol-1,4,5trisphosphate [4,29,30] and Ca^"" mobilisation [3,30-35]. In Sf9 cells only the a2B receptor seems to be able to cause any significant Ca^"^ mobilisation [30]. The agonist profile for this response is similar to that of stimulation of cAMP accumulation, e.g. UK 14,304 and clonidine are very weak agonists with respect to Ca^"*" mobilisation. Fig. 6 shows that Gi coexpression considerably attenuates the Ca^"^ mobilisation via a2B but Gs has little or no effect. Coexpression with Py subunits alone caused a small reduction in the noradrenaline-induced Ca^^ elevation.

[Ca^-^li /nM 400 n

1

300 200 100 0

^5^ O^ O'v <^^

.Q^ Fig. 6 Effect of Gs, Gi and py coexpresion on 100 nM noradrenaline-stimulated Ca^"^ mobilisation in cells expressing the a2B receptor. The methods used are essentially those described in references [10,22,25,]. Ca^^ measurement were performed using fura-2 as described in [30].

This result is in line with the effect of Gi coexpression on the stimulatory cAMP response and suggest that the Ca^"^ mobilisation occurs through coupling to another Gprotein with lower "affinity" for the receptor than Gi. Ca^"^ mobilisation via a2-AR activation has previously been observed in many other cells such as CHO or HEK293 cells expressing recombinant receptors. The response

189 seen in these cells is usually pertussis toxin sensitive and hence mediated by Gi/o proteins unlike the response in Sf9 cells. The properties of pertussis toxin sensitive Ca^"^ mobilization are listed in table 3. Table 3. Pertussis toxin sensitive Ca^^ mobilisation • • • • •

Is seen with recombinant Gi coupled receptors expressed in CHO, HEK, COS-7 etc as well as with native receptors in HEL erythroleukemia cells All a2 receptor subtypes are about equally effective Sensitive to inhibitors of PLCP Associated with an increase in IP3 Sensitive to cotransfection with (Jy scavengers like (J-ARK

The pathway for activating Ca^"^ mobilization in CHO cells is complex since it requires cooperating responses via other endogenously activated receptors such as purinergic P2Y receptors [35]. Thus a possible mechanism for this type of coupling is activation of PLCp by a cooperative action of Gq like proteins and py subunits from Gi/o-

It is uncertain whether Ca^"*" mobilisation in Sf9 cells and the pertussis toxin-sensitive Ca^^ mobilisation represent the same phenomenon since in Sf9 cells Gi inhibited rather than promoted the response. Interestingly the agonist profile for Ca^^ mobilisation via the a2B receptor is very similar to that seen in CHO cells both with respect when the maximal response and EC50 values are compared (Fig. 7). In HEL human erythroleukemia cells, which have been used as models for platelets, there is a very robust Ca^"*^ response to ai-AR activation via the endogenous a2A. In these cells synthetic compounds like imidazolines show a very similar ability to activate Gj coupling and Ca^"*" mobilisation but the catecholamines are much more prone to induce Ca^"^ mobilisation [36].

190

A . Maximum response 0) v> c

A •

150

catecholamine imidazolines best fit througii (0,0) D-Medetomidine I Oxymetazoline

CO

O

100

(0

o c

50

2 CO

0

50

100

CHO (nomialised Ca^ response)

1000 1

A #

catecholamine imidazolines best fit

100

2.4:1 Oxymetazoline

Noradr Clonidine

CO

10

10

100

1000

CHO (nM) Fig. 7 Comparison of the ability of catecholamine and imidazoline ligands to induce Ca mobilisation via the a2B receptors in Sf9 and CHO cells. Methods are described in references [30, 35].

191 Conclusions a2, adrenoceptors, in particular a2B, can couple to several signal pathways. The mutagenesis studies indicate that Gs and possibly Gq -type proteins couple to another site or conformation of the receptor than Gj. The ligand-specific coupling suggest that the conformation induced by these ligands is different as compared to native ligands. The UK 14,304-induced receptor conformation couples equally well Gi as the noradrenalineinduced conformation but less efficiently to Gs. The model with a different apparent affinity for the interaction of the secondary G protein for the receptor fits well with the experimental data. The outcome at the effector level will thus depend on the "affinity" of the G protein in question for the receptor as well as the G protein expression level. The more rigid synthetic ligands are thus more prone to promote coupling to the GJGo pathway. The more flexible natural ligands allow coupling to several pathways. Even more rigid ligands could be useful as probes for testing conformation-signal coupling. Coupling to different pathways could well explain the variety of effects of a2-AR agonists seen in different tissues

Acknowledgements r The original working this presentation was funded by The Medical Research Council of Sweden (MFR K98-14X-12205-02B), The Cancer Research Fund of Sweden, The Claes Groschinsky Foundation and The Ake Wiberg Foundation, the Academy of Finland, TEKES and The Sigrid Juselius Foundation.

References [I] R.R. Ruffolo, A.J. Nichols, J.M. Stadel, and J.P. Hieble, Annu. Rev. Pharmacol. Toxicol. 33 (1993) 243. [2] G. Milligan, Trends Pharmacol. Sci. 14 (1993) 239. [3] M.C. Michel, L.F. Brass, A.Williams, G.M. Bokoch, V.J. LaMorte, and H.J. Motulsky, J. Biol. Chem. 264 (1989) 4986. [4] S. Cotecchia, B.K. Kobilka, K.W. Daniel, R.D. Nolan, E.Y. Lapetina, M.G. Caron, R.J. Lefkowitz, and J.W. Regan, J. Biol. Chem. 265 (1990) 63. [5] E. Duzic and S.M. Lanier, J. Biol. Chem. 267 (1992) 24045. [6] M.G. Eason, H. Kurose, B.D. Holt, J.R. Raymond, and S.B. Liggett, J. Biol. Chem. 267 (1992) 15795. [7] E.E. MacNulty, S.J. McClue, I.C. Carr, T. Jess, M.J. Wakelam, and G. Milligan, J. Biol. Chem. 267 (1992) 2149. [8] T.K. Aburto, C. Lajoie, and K.G. Morgan, Circ. Res. 72 (1993) 778. [9] C.C. Jansson, A. Marjamaki, K. Luomala, J.M. Savola, M. Scheinin, and K.E.O. Akerman, Eur. J. Pharmacol. 266 (1994) 165. [10] C.C.Jansson, M. Karp, C. Oker-Blom, J. Nasman, J.M. Savola, and K.E.O. Akerman, Eur. J. Pharmacol. 290 (1995.) 75. [II] J.W. Lomasney, S. Cotecchia, R.J. Lefkowitz, and M.G. Caron, Biochim.

192

[12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28]

[29] [30] [31] [32] [33] [34] [35] [36]

Biophys. Acta 1095 (1991) 127. M. Gollasch, J. Hescheler, K. Spicher, F.J. Klinz, G. Schultz, and W.Rosenthal, Am. J. Physiol. 260 (1991) C1282. S.Y. Song, K., Saito, K. Noguchi, and S. Konishi, Pflugers Arch. 418 (1991)592. A. Surprenant, D.A. Horstman, H. Akbarali, and L.E. Limbird, Science 257 (1992)977. S.B. Jones, S.P. Halenda,. and D.B. Bylund, Mol. Pharmacol. 39 (1991) 239. S. Ullrich, and C.B. Wollheim,. J. Biol. Chem. 259 (1984) 4111. S.B. Jones, M.L. Toews, J.T. Turner, and D.B. Bylund,. Proc. Natl. Acad. Sci. USA 84 (1987) 1294. S. Mhaouty, J. Cohen-Tannoudji, R. Bouet-Alard, I. Limon-Boulez, J. Maltier, J. P. and C. Legrand, J. Biol. Chem. 270 (1995) 11012. D.J. Pepperl, and J.W. Regan, Mol. Pharmacol. 44 (1993) 802. K. Pohjanoksa, C.C. Jansson, K. Luomala A. Marjamaki, J.M. Savola and M. Scheinin Eur. J. Pharmacol. 335 (1997) 53. M.G. Eason, M.T. Jacinto, and S.B. Liggett, IVlol. Pharmacol. 45 (1994) 696. J. Nasman, C.C. Jansson, and K.E.O. Akerman, J. Biol. Chem. 272 (1997) 9703. T. Kenakin, Trends Pharmacol. Sci. 16 (1995) 232. R.K. Sunahara, C.W. Dessauer, and A.G. Gilman, Annu. Rev. Pharmacol. Toxicol. 36 (1996) 46. J. Nasman, J.P. Kukkonen and K.E.O. Akerman, Insect Biochem. Mol. Biol. (2001) In press M.G. Eason, and S.B. Liggett, J. Biol. Chem. 271 (1996) 12826. S.M. Wade, W.K. Lina, K.-L. Lan, D.A. Chung, M. Nanamori and R.M. Neubig. Mol.Pharm58 (1999) 1005. K. Palczewski, T. Kumasaka, T. Hori, C.A. Behnke, H. Motoshima, B.A. Fox, I. Le-Trong, D.-C. Teller, T. Okada, R.-E. Stenkamp, M. Yamamoto, M. Miyano Science. 289 (2000) 739. M.O.K. Enkvist,H. Hamalainen, C.C. Jansson, J.P. Kukkonen, R. Hautala, M.J. Courtney,, and K.E.O. Akerman, J. Neurochem. 66 (1996) 2394. C.I. Holmberg, J.P. Kukkonen, A. Bischoff, J. Nasman, M.J. Courtney, M.C. Michel, and K.E.O. Akerman, Eur. J. Pharmacol. 363 (1998) 65. A.K. Salm, and K.D. McCarthy, Glia 3 (1990) 529. W. Erdbrugger, P. Vischer, H.J. Bauch,. and M.C. Michel, J Cardiovasc Pharmacol 22 (1993) 97. G.W. Dorn, K.J.. Oswald, T.S. McCluskey, D.G. Kuhel, and S.B. Liggett, Biochemistry 36 (1997) 6415. J.P. Kukkonen, A. Renvaktar, R. Shariatmadari, and K.E.O. Akerman, J. Pharmacol. Exp. Ther. 287, (1998) 667. K.E.O. Akerman, J. Nasman, P.E. Lund, R. Shariatmadari, and J.P Kukkonen, FEBS Lett. 430 (1998) 209. J.P. Kukkonen, C.C. Jansson, and K.E.O. Akerman , Br. J. Pharmacol. 132 (2001)1477.

H. van der Goot (Editor) Trends in Drug Research III © 2002 Elsevier Science B.V. All rights reserved

193

Antibacterials as Wonder Drugs and How Their Effectiveness Is Being Compromised Dasantila Golemi,§ Laurent Maveyraud,* Jalal Haddad, Wenlin Lee, Akihiro Ishiwata,§ Kazuyuki Miyashita,§ Lionel Mourey,+ Sergei Vakulenko, Lakshmi Kotra, Jean-Pierre Samama,-!- and Shahriar Mobashery,§* •••Groupe de Cristallographie Biologique, Institut de Pharmacologic ct de Biologic Structurale du CNRS, 205 route dc Narbonnc, 31077-Toulousc ccdcx, France ^Department of Chemistry and Institute for Drug Design, Wayne State University, Detroit, MI 48202-3489 USA.

Treatment of infections by bacterial pathogens became possible only in the twentieth century. Indeed, virtually all classes of antibacterials that are in clinical use presently were discovered and developed in the span of a mere 30 or so years. The advent of antibiotics, along with immunization, has had profound consequences on the quality of life. Life expectancy increased in the United States by 30 years from the beginning to the end of the twentieth century, as a direct consequence of these developments [1]. These successes created an air of complacency. There was the perception that bacterial infections were infinitely curable. This view was strengthened by the late 1960s into 1970s, in part because of the discoveries of the various classes of antibiotics, but also by the fact that medicinal chemists appeared to be able to improve these molecules of natural origins by synthetically altering them. These synthetic molecules were often more potent that their parental natural products, and at times the breadth of activities were also increased to create pharmaceuticals that were being hailed as miraculous in the clinic. The euphoria lead to complacency, such that by the late 1980s and earlier 1990 many pharmaceutical companies decided to leave the field of antibiotics. Whereas the number of available antibiotics during that period was considerable, the extensive use of antibiotics in human health, in animal husbandry and in agriculture had provided ample opportunities for the microbial organisms to evolve and adjust to the challenge. Indeed, in the period between 1973 and 1996, discoveries of no less than 30 new infectious agents were made [2]. There is a companion list to this of the organisms that are being described as "re-emerging" microbes [3]. These were organisms that we were able to treat in the prior years, but the various resistant forms are more difficult to treat presently. The reasons for these microbial successes in the face of clinical challenge by antibiotics are several fold. Bacteria are capable of reproducing rapidly, often with doubling times of less than half an hour under optimal growth conditions. Replication in bacteria takes place with a vanishingly small rate of 10"^^ for mutation, because bacteria enjoy mechanisms for correction of mutations in their genomes. Nonetheless,

194 considering the high densities that bacteria can attain—10^ cells/mL in infections of blood and 10^ cells/mL in infections of tissues—and their large genome sizes, the opportunities for creation of random mutations throughout the genome are considerable. It has been estimated that there may be 10^-10^ mutant variants in a culture of actively growing bacteria [4]. This estimate does not take into account the rate of hypermutation in stressed bacteria, such as bacteria exposed to antibiotics, which might enhance these values by three orders of magnitude. Some of these mutations will be lethal to the organisms, others will be silent, yet there may be some that actually might have survival benefit to the organism in the presence of the antibiotic challenge. These advantageous mutations would be selected during exposure to antibiotics. In light of the fact that susceptible competing organisms will be wiped out by the antibiotic exposure, the selected mutant organisms would have opportunity to survive and thrive in the presence of the antibiotic. This is a simplistic view of the process of selection of resistance, but it is instructive for the purpose of our discussion. There are indeed other means for selection of resistant organisms, such as horizontal transfer of genetic materials from one organism to another. There exist bacteria that augment their genetic materials by picking up sequences of nucleic acids from the environment, then there are viral infections that have the potential to transfer genetic materials from one bacterium to another, among others [5], [6]. Resistance to any antibiotic is inevitable and it usually happens in 1-4 years [7]. The case of penicillin G is of special interest since resistance to it was first documented prior to its first large-scale clinical use [8]. There is now documented resistance to the latest antibiotic of a novel class, linezolid, the first member of the oxazolidinones, which became clinical in the summer of 2000 [9]. Resistance in this case emerged in less than a year. A total of 51 microbial genomes have been sequenced in their entirety as of May 2001, and many more are in the process of being sequenced [10]. It is argued that there may be 20-200 critical targets for development of broad-spectrum antibiotics against bacteria [11], [12], [13], [14]. It is likely that the current efforts in development of new antibiotics against some of these targets may bear fruit in the near future. Yet, it is remarkable that the analyses of the sequences of these genomes reveal that as many as 30-40% of these genes have not been attributed any functions yet. Furthermore, many of the functions that have been attributed to the remaining gene products have not been verified by experiments. The demonstration of functions and determination of the structures of these proteins will be critical to begin to understand how bacteria live. These tasks are time consuming and will keep scientist busy for many years to come. It is noteworthy that the functions and structures of many of the targets demonstrated to be critical for antibiotics have not been elucidated either. For example, the vast majority of antibiotics work by binding to the bacterial ribosome, whereby the function of the ribosome in protein biosynthesis is disrupted. The low-resolution images of bacterial ribosome became available for the first time in 1999 [15], [16], [17], [18], [19], [20]. The many functional attributes of the components of the ribosome await scientific discovery and scrutiny. Another important target for antibacterial is the biosynthetic machinery for assembly of the bacterial cell wall. This is a uniquely bacterial entity and involves multiples of enzymes [7], [21], [22]. In principle, each of these enzymes could be the target of antibacterials. Many of these steps take place in

195 the cytoplasm. Lipid II, the immediate building block for assembly of cell wall is transferred to the periplasmic space, where it is polymerized into the A^-acetyl glucosamine (NAG)-N-acetyl muramic (NAM) acid repeating backbone. A bacterial pentapeptide is appended to the NAM unit. The point of cross-linking is via two of these peptide moieties, as described below.

I I ! ^ [GS^

k "

O L-Ala

CO2 n.r,... D-Glu

" O J - N ^D-Ala HAP -^) 0^\ T DAP X>2~ E-OH + p

O

CO2"

Nu-E

H N.^a)02

CO2 0(?

•' co2-oyKl^

'CO2 strand #2 ^ I QlcNAcI

l^cc strand #1

H H "

O

»

r I MurNAcp^V^N"^ ^002 strand #2

Figure 1. The crosslinking of the peptidoglycan stands in the bacteria cell-wall synthesis. The cross-linking of the bacterial cell wall is catalyzed by transpeptidases, members of the penicillin-binding protein (PBP) family. These enzymes pursue an active site-serine strategy in their catalytic reaction. As depicted in Figure 1, the active site serine in these enzymes reacts with the amide carbonyl group of the penultimate DAla in the peptidoglycan structure, giving rise to an acyl-enzyme species. On binding of

196 the second strand of peptidoglycan to the active site, the side chain amine is promoted to approach the ester carbonyl to give the amide bond in the product Transpeptidases are the targets of p-lactam antibiotics, such as penicillins and cephalosporins. Tipper and Strominger [23] had proposed that the conformation of acyl-D-Ala-D-Ala portion of peptidoglycan mimics the conformation of the backbone of the p-lactam antibiotics. As such, the antibiotic would acylate the active site serine of these enzymes, and the acyl-enzyme species resists deacylation, resulting in the bacterial cell death. We proposed that this property of the cephalosporin nucleus could be used to design a novel molecule (1) that would acylate the active site of these enzymes, as would the first strand of the peptidoglycan. However, we incorporated in the design of the molecule features that would mimic the incoming second strand of peptidoglycan (2). On acylation of the active site of the transpeptidation, the complex would have features of the two peptidoglycan strands sequestered in the active site, as would the two strands of the peptidoglycan in the course of the cross-linking reaction. Essentially, we would have a snapshot of the two strands of the peptidoglycan just prior to the cross-linking event.

mimics the first strand of peptidoglycan after acylation of the active site serine

^

mimics the approaching second strand of peptidoglycan

PBP E-OH

CO2 -^ ^ ^^NH 1+ CO2 ^ O0.^^h ""V^ H"

"

O

I

A

N"^C02

Synthesis of compound 1 has been reported [24]. This compound indeed modifies the active site serine of the bifunctional transpeptidase/DD-peptidase from species Streptomyces R61. The structure of the complex that was solved at 1.2 A resolution provided a detailed knowledge of the specific interactions in the active site of this bacterial enzyme that leads to the cross-linking event, as reported elsewhere [24]. This structure was the building point for the generation of a model for the two full peptidoglycan strands up to the NAG-NAM repeat units. The energetics of the model were evaluated in the course of dynamics simulations and energy minimization procedures [24]. These models provide information on the binding sites for the two substrates and they reveal two grooves on the surface of the enzyme that are the likely locations for the binding of the polymeric NAG-NAM repeat backbones [24]. Paul Ehrlich prophetically stated that "drug resistance follows the drug like a faithful shadow" [11]. There are seven major classes of antibacterials that are currently in clinical use. These are p-lactams, fluoroquinolones, aminoglycosides, glycopeptides, macrolides, tetracyclines and sulfonamide [25]. Cases of resistance to all are known.

197 Indeed, resistance to any one class by itself is a relatively rare occurrence, and one actually sees often resistance to multiple classes of antibiotics in pathogens. There are five common strategies that microorganisms have evolved for resistance to antibacterials. These are resistance due to reduced permeability into the organisms (plactams, fluoroquinolones, and folate inhibitors), altered target site (p-lactams, aminoglycosides, tetracyclines glycopeptides, fluoroquinolones, and folate inhibitors), resistance enzymes (p-lactams, aminoglycosides, macrolides, chloramphenicol), efflux mechanisms (tetracyclines and fluoroquinolones), and target by-pass (sulfonamide and trimethoprim) [25]. Of these, the only two that are amenable to intervention by medicinal chemists are the strategies by efflux mechanisms and by resistance enzymes. In essence, if one were to inhibit the efflux pumps or the resistance enzymes, the viability of the original antibiotic could be restored. The Mobashery laboratory has had a long-standing interest in enzymes of antibacterial resistance. We have devised strategies to either inhibit or circumvent these enzymes over the years. One recent effort centered on developing new classes of antibiotics that would bind to the validated targets that are known for the existing classes of antibiotics. In one such work, we commenced by investigating the structure of the complex of paromomycin, an aminoglycoside antibiotic, bound to the acyltransfer site ("A site") of bacterial ribosome [26]. We retained as essential elements two rings of paromomycin for binding to the RNA structure. These two rings made important electrostatic interactions with the RNA moieties and produced the core of the structure within the A site that was retained. The remainder of the structure of paromomycin was eliminated in silico. We explored a number of three-dimensional compound data banks for molecules that would bind to the space that was within the A site, which was created by the elctronic elimination of the functionalities that were removed. This exercise was carried out using the program DOCK, which sampled 380,000 such molecules. The top 100 scoring molecules that were shown to have affinity for binding at the subsites contiguous with those of the two retained rings were scrutinized closely. Several revealed the possibility of attachment to the retained core of the two sugars. A number of the molecules were synthesized and they were shown to bind to the A site in in vitro experiments. We also showed that some of the synthetic molecules were not susceptible to the action of three of the common enzymes of resistance to aminoglycosides. The project disclosed here is one step in the incremental process necessary to move away from the structure of aminoglycosides, while retaining the ability to bind to a validated target in bacteria for antibiotics, namely the A site of the bacterial ribosome. The use of aminoglycoside antibiotics has facilitated evolution of a number of enzymes that structurally modify these antibiotics, whereby the affinity for binding to the bacterial ribosome is lost [27], [28], [29], [30]. This is indeed a very common mechanism for resistance to these antibacterials. The most common mechanism for resistance to these antibiotics is by the activity of aminoglycoside 3 ' phosphotransferases [APH(3')s]. These enzymes transfer the y-phosphoryl group of ATP to the 3'-hydroxy 1 of aminoglycosides. This modification interrupts the essential aforementioned electrostatic interactions between the ribosomal RNA and aminoglycoside, disfavoring the binding process. These enzymes are usually expressed in high copy numbers and often operate at the diffusion limit [27]. Despite the fact that

198 they are cytoplasmic [27], as is the target ribosome, they confer resistance extremely effectively. It occurred to us that if we could manipulate the structure of the antibiotic such that the 3' carbon were to bear a geminal dihydroxy species (a hydrated ketone), we may be able to reverse the process of drug phosphorylation. The concept is disclosed in Figure 2. Compound 3 was synthesized from kanamycin A in 11 steps [31]. The mixture of the hydrated version of compound 3 (compound 4) to the ketonic version was 3:1 at neutral pH and at room temperature. We demonstrated that compound 4 underwent phosphorylation by APH(3')s, followed by the spontaneous release of the inorganic phosphate. Hence, a "futile cycle" ensues, in which ATP is hydrolyzed. It is noteworthy that the minimal inhibitor concentration of kanamycin A, the parental compound to 4, went up by 1000-fold in the organisms that produced the APH(3')s over the background strain that did not express the enzyme. The similar situation increased the MIC for compound 4 by a mere four-fold [31]. If ATP could be set up for recylcing, as shown in Figure 2, the process of phosphorylation and elimination of phosphate could go on for as long as there exists NADH. Once exhausted, the process could be resumed by the addition of NADH, and as expected, the event was dose dependent [31]. The major cause of resistance to p-lactam antibiotics comes about as a consequence of the catalytic action of p-lactamases. As depicted in Figure 3, these enzymes hydrolyze the P-lactam moiety of these antibiotics, the product of the reaction lacks antibiotic property. Whereas the first p-lactamase was discovered in the early 1940s, prior to the beginning of the use of penicillins in the clinic, the discoveries of the novel forms of these enzymes did not take place until much later in the 1980s and 1990s [32], [33]. There are over 340 of these enzymes known to date and they fall into four classes, classes A, B, C, and D. Class B enzymes are zinc-dependent, whereas the remaining three classes are active-site-serine enzymes.

199

A

NH2

0^^0^g;^NH2

-a

OH

}H

'^-^^ y( NH2

> NH2

/

NH2

^8h ^

ATP

-oV„-

a.Po-

OH

ADP

-Xf

pyruvate kinase

o

NAD"^

lactate dehydrogenase

PEP (excess)

B

NADH

^i^OH

80 ^xM NADH

/

72^MNADH

/ e

0.4 40 ^AM NADH

en

/

CD

16 H,M NADH

0.?.

/

o

1000

2000

3000

4000

Time (s) Figure 2. (A) The coupled spectrophotometric assay of pyruvate kinase and lactate dehydrogenase recycles ATP, which is consumed by the APH(3')-mediated phosphorylation of 4. (B) The gradual disappearance of the chromophore for NADH, as a function of phosphorylation of 4, was monitored by the decline in absorbance at 340 nm. The process is continuous until PEP (180-fold excess) is entirely depleted (details in ref 20). Each run was initiated/reinitiated by the addition of the given quantity of NADH to the recycling mixture.

200

There is clear evidence by now that p-lactamases and PBPs are related to each other [34], [35]. The issue of kinship of this family of enzymes is addressed elsewhere in greater depth [35]. But, it would appear that nature has selected four distinct mechanisms for resistance to P-lactam antibiotics [36]. The zinc-dependent enzymes are distinctly different, but so are the serine enzymes. In the case of class A enzymes, the hydrolytic water approaches the acyl-enzyme species from the a direction and the water molecule is promoted by the invariant residue Glu-166 as the general base [37], [38], [39], [40]. In the case of the class C enzymes the water is promoted from the opposite P direction. It is believed that the collective environment of the hydrolytic water promotes it in the reaction. We and others have shown that there is an electrostatic interaction with the nitrogen of the acyl-enzyme species, hence substrateassisted catalysis applied here [41], [42]. The mechanism of class D enzymes is interesting. These enzymes utilize a carbamylated lysine residue in the active site for both the acylation and deacylation chemistries [43]. Hence, the catalytic events in class D enzymes, in contrast to enzymes of classes A and C, are synunetrical [44]. Only a mere handful of proteins are known to have carbamylated lysine in their structures.

p-lactamase

CO2

>-'

N—1^^

further degradation

p-lactamase

CO2

Figure 3. Hydrolysis of P-lactams by P-lactamases

These mechanistic differences underscore the clinical difficulties in fighting the deleterious action of p-lactamases. It is desirable to have inhibitors that work against more than one class of enzyme. This expectation has not been fulfilled to date, and indeed the examples of inhibitors that would inhibit two classes of these enzymes are extremely rare [45], [46], [47]; none have made it to the clinic to date. There exist combinations of three inhibitors for class A P-lactamases, clavulanate, sulbactam or tazobactam, with penicillins that are being used in fighting infections [32]. These inhibitors all work by the same mechanism in inhibiting the enzymes of class A [32]. Over the past few years, variants of these class A enzymes have been discovered that are capable of performing their function as resistance enzymes, but they have become resistant to inhibition by these clinically used inhibitors

201 [48], [49], [50], [51]. Since the mechanism of inhibition by all three inhibitors is the same, there is the potential that resistance to inhibition by one inhibitor might impart the same property to others. Therefore, there is a need for additional classes of inhibitors that inhibit the enzyme by distinct mechanisms. We have reported on additional types of inhibitors for p-lactamases recently. The chemistry in each case takes advantage of the reactivity of the active site serine for the onset of inhibition. Monobactam inhibitors are exquisitely effective in inhibition of class A plactamases, and they are competitive non-covalent inhibitors of class C enzymes [52], [53], [54], [55]. These inhibitors acylate the active site serine rapidly, followed by the elimination of the good living group. There are four possibilities for the inhibitory species (6-9), two of which, 6 and 7 have been observed crystallographically.

/^^3

cJrrk ,0>O.Ts Ser-70 |-^^

PHa

0=( ^^0-Ts 3 H Ser-70 5 — ' ^ p-elimination

/)H3

r—J^ 0=(

O "S 0-Ts Ser-70

NH2

Ser-70|-^

tautomerization

/=\ ^

0==(^NH2

Ser-701—^

6

8 hydrolysis /

hydrolysis tautomerization

of\ Ser-70 \ - ^ 7

^"3

0==PbH Ser-70

\—^ 9

AT O 0-Ts 12 Ser-70

202 Compound 10 is of special interest. This compound inhibits the class A TEM1 P-lactamases within seconds and resists deacylation of the inhibitory species for days [53]. The x-ray structure revealed the presence of the iminium species 11 for the inhibitory species [53]. Compound 12 proved extremely effective in inhibition of the broad-spectrum class A enzyme NMCA from Enterobacter cloacae [55]. X-Ray analysis revealed that the carbonyl species of 12 exists in two conformations after acylation of the enzyme. One ensconced in the active site oxyanion hole, the other with the carbonyl flipped out of this space [55]. We explored the process of this conformational change in the active site by computational simulations. Within the time scale of picoseconds, one sees dynamics motion of the inhibitor from one species to the other, a process that repeats itself, indicative of the facility of the transition from one species to the other. It was also demonstrated that the clinically used carbapenem antibiotic imipenem also undergoes the same motion of the acyl-enzyme species in an out of the oxyanion hole [56]. Another class of inhibitors for the class A enzymes are based on the structure of penicillanate. These compounds acylate the enzyme active site and the acyl-enzyme species resists deacylation. In essence, the hydroxyalkyl moiety at the 6a position prevented the approach of the hydrolytic water from that direction in the acyl-enzyme species. Therefore, these moelcules acylate the enzyme and resist deacylation for a number of hours [40], [57], [58].

COf-

O

cor

o

-N02

r ^ N

1 further degradation

In the U.S. alone 50 million pounds of antibiotics were used in 1996, and the trends over the past decades would indicate that these numbers increase on an annual basis. A key general feature of antibiotics is that they arfe often not metabolized in the body as they are eliminated from patients. Hence, these biologically active molecules are introduced into the sewer system and ultimately to the environment with plenty of exposure to bacterial. Antibiotics have the ability to select for resistance well after their

203

elimination from patients.lt occurred to us that one could design antibiotics that could self-destruct after use in humans. Many such antibiotics can be envisioned, but as a proof of concept, we have recently reported on cephalosporin 13 [59]. This compound unmasks a hydrazine derivative from its side chain on exposure to either visible or ultraviolet light. The so-generated hydrazine destroys the p-lactam moiety of the antibiotic, rendering it inactive (15). Cephalosporin 13 has activity against both Gramnergative and Gram-positive bacteria. It also serves as substrate to P-lactamases, so it functions like a typical p-lactam antibiotic. Development of antibiotics is a labor-intensive process and expensive. It is incumbent upon us to make certain that their clinical viability is extended as long as possible. To begin to understand the processes that are necessary in development of antibiotics and occurrence of mechanisms of resistance, it is necessary that these be studied at the molecular level. This manuscript described some of our efforts in this direction. ACKNOWLEDGEMENTS The work in France was funded in part by the Programme de Recherche Fondamentale en Microbiologic (MENRT) and CNRS. The work in the USA was supported by grants from the National Institutes of Health (AI33170 and GM61629) and the National Science Foundation (SM).

REFERENCES [I] [2] [3] [4] [5] [6] [7]

[8] [9] [10] [II] [12] [13] [14]

J. Lederberg, Science 288 (2000) 287. The World Health Report 1996. Fighting Disease, Fostering Development. World Health Organization, Geneva, 1996. J. Lederberg, R. E. Shope, Eds., Emerging Infections: Microbial Threats to Health in the United States, National Academy Press, Washington, D. C , 1992. S. Mobashery, E. Azucena, Encyclopedia Life Sci. London. Nature Publishing Group, UK, 2000 (www.els.net). J. Davies, Trends Curr. Biol. 9 (1999) M2. 283. K. J. Poole, J. Pharm. Pharmacol. 53 (2001). K. K. Wong, D. L. Pompliano, Resolving the Antibiotic Paradox: Progress in Understanding Drug Resistance and Development of New Antibiotics: Peptidoglycan biosynthesis: Unexploited antibacterial targets within a familiar pathway, Eds. B.P. Rosen, and S. Mobashery, New York: Plenum Press, 1998. W. M. M. Kirby, Science, 99 (1944) 452. R. D. Gonzales, P. C. Schreckenberger, M. B. Graham, S. Kelkar, K. DenBesten, and J. P. Quinn, The Lancet, 357 (2001) 1179. http://www.ncbi.nlm.nih.gov L. P. Kotra, S. Vakulenko, S. Mobashery, Microbes Infect. 2 (2000) 651. A. E. Allsop, Curr. Opin. Microbiol. 1 (1998) 530. D. T. Moir, K. J. Shaw, R. S. Hare, G. F. Vovis, Antimicrob. Agents Chemother. 43(1999)439. D. J. C. Knowles, F. King, Resolving the Antibiotic Paradox: Progress in

204

[15] [16] [17]

[18] [19] [20] [21] [22] [23] [24] [25] [26] [27]

[28] [29] [30] [31] [32]

[33] [34] [35] [36]

Understanding Drug Resistance and Development of New Antibiotics: The Impact of Bacterial Genomics on Aantibacterial Discovery, Eds. B.P. Rosen, S. Mobashery, Plenum Press, New York, NY, 1999. J. H. Gate, M. M. Yusupov, G. Zh. Yusupova, T. N. Earnest, H. F. NoUer, Science, 285 (1999) 2095. G. C. Agalarov, G. Sridhar Prasad, P. M. Funke, C. D. Stout, J.R. Williamson, Science, 288 (2000) 107. B. T. Wimberly, D. E. Brodersen, W. M. Clemons Jr., R. J. Morgan-Warren, A. P. Carter, C. Vonrhein, T. Hartsch, and V. Ramakrishnan, Nature, 407 (2000) 327. M. M. Yusupov, G. Zh. Yusupova, A. Baucom, K. Lieberman, T. N. Earnest, J. H. D. Gate, and H. F. Noller, Science, 292 (2001) 883. J. M. Ogle, D. E. Brodersen, M. C. William, Jr., M. J. Tarry, A. P. Carter, V. Ramakrishnan, Science, 292 (2001) 897. S. C. Blanchard, and J. D. Puglisi, Proc. Natl. Acad. Sci. U.S.A. 98 (2001) 3720. C. T. Walsh, Nature, 406 (2000) 775. C. Goffin, J. M. Ghuysen, Microbiol. Mol. Biol. Rev. 62 (1998) 1079. D. J. Tipper, and J. L. Strominger, Proc. Natl. Acad. Sci. U.S.A. 54 (1965). W. L. Lee, M. A. McDonough, L. P. Kotra, Z. H. Li, N. R. Silvaggi, Y. Takeda, J. A. Kelly, S. Mobashery, Proc. Natl. Acad. Sci. U.S.A. 98 (2001) 1427. L. P. Kotra, D. Golemi, S. Vakulenko and S. Mobashery, Chemistry & Industry (2000) 339, J. Haddad, L. P. Kotra, B. Llano-Sotelo, E. F. Azucena Jr., M. Liu, S. B. Vakulenko, C. S. Chow and S. Mobashery, (unplublished results). G. D. Wright, A. M. Berghuis, S. Mobashery, Resolving the Antibiotic Paradox: Progress in Understanding Drug Resistance and Development of New Antibiotics: Aminoglycoside Antibiotics: Structures, Functions and Resistance, Eds. B.P. Rosen, S. Mobashery, Plenum Press, NY, 1998. L. P. Kotra, J. Haddad, S. Mobashery, Antimicrob. Agents Chemother. 44 (2000) 3249. E. Azucena Jr., I. Graspas and S. Mobashery, J. Amer. Chem. Soc. 119 (1997) 2317. G. D. Wright, Curr. Opin. Microb. 2 (1999) 499 J. Haddad, S. Vakulenko, S. Mobashery, J. Am. Chem. Soc. 121 (1999) 11922. K. Bush and S. Mobashery, Resolving the Antibiotic Paradox: Progress in Understanding Drug Resistance and Development of New Antibiotics: How pLactamases Have Driven Pharmaceutical Drug Discovery: from Mechanistic Knowledge to CHnical Circumvention. Eds. B.P. Rosen and S. Mobashery, Plenum Press, NY, 1998 K. Bush, Clin. Met. Dis. 32 (2001) 1085. J. A. Kelly, A. P. Kuzin, P. Charlier, and E. Fonze, Cell. Mol. Life Sci. 54 (1998) 353. I Massova and S. Mobashery, Antimicrob. Agents Chemother. 42 (1998) 1. L. P. Kotra, J. P. Samama, S. Mobashery, Bacterial Resistance to Antimicrobials, Mechanisms, Genetics, Medical Practice and Public Health: Structural Aspects of

205

[37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52] [53] [54] [55] [56] [57] [58] [59]

p-Lactamase Evolution, Eds. A. Lewis, Salyers, H. Haber, R. G. Wax, Marcel Dekker, Inc, 2001 (in press). O. Herzberg. and J. Moult, Science, 236 (1987) 694. J. R. Knox, and P. C. Mowes, J. Mol. Biol. 220 (1991) 435. N. C. Strynadka, H. Adachi, S. E. Jensen, K. Johns, A. Sielecki, C. Betzel, K. Sutoh, and M. N. James, Nature, 359 (1992) 700. K. Miyashita, I. Massova, P. Taibi and S. Mobashery, J. Am. Chem. Soc. 117 (1995)11055. A. Bulychev, I. Massova, K. Miyashita, and S. Mobashery, J. Am. Chem. Soc. 119(1997)7619. A. Patera, L. C. Blaszczak, B. K. Shoichet, J., Am. Chem. Soc. 122 (2000) 10504 L. Maveyraud, D. Golemi, L. P. Kotra, S. Tranier, S. Vakulenko, S. Mobashery, J. P. Samama, Structure, 8 (2000) 1289. D. Golemi, L. Maveyraud, S. Vakulenko, S. Tranier, A. Ishiwata, L. P. Kotra, J. P. Samama, S. Mobashery, J. Am. Chem. Soc. 122 (2000) 6132. A. Bulychev, I. Massova, S.A. Lerner, S. Mobashery, J. Am. Chem. Soc. 117 (1995) 4797. T. Nagase, D. Golemi, A. Ishiwata, S. Mobashery, Bioorg. Chem. (2001) (in press). J. D. Buynak, V. R. Doppalapudi, G. Adam. Bioorg. Med. Chem. Lett. 10 (2000) 853. S. B. Vakulenko, B. Geryk, L. P. Kotra, S. Mobashery, S. A, Lerner, Antimicrob. Agents Chemother.43 (1998) 1542. E. B. Chaibi, D. Sirot, G. Paul, R. Labia, J. Antimicrob. Chemother. 43 (1999) 447. R. A. Bonomo, and L. B. Rice, Front. Bioscience. 4 (1999) e34. C. Therrien, R. C. Levesque, FEMS Microb. Rev. 24 (2000) 251. A. Bulychev, M.E. O'Brien, L Massova, M.Teng, T.A. Gibson, M.J. Miller, S. Mobashery, J. Am. Chem. Soc. 117 (1995) 5938. P. Swaren, I. Massova, J. Bellettini, A. Bulychev, L. Maveyraud, L. P. Kotra, M. J. Miller, S. Mobashery, J. P. Samama, J. Am. Chem. Soc. 121 (1999) 5353. A. Bulychev, J. R. Bellettini, M. O'Brien, P. J. Crocker, J. P. Samama, M. J. Miller, S. Mobashery, Tetrahedron, 56 (2000) 5719. L. Mourey, L. P. Kotra, J. Bellettini, A. Bulychev, M. O'Brien, M.J. Miller, S. Mobashery, J. P. Samama, J. Biol. Chem. 274 (1999) 25260. L. Maveyraud, L. Mourey, L. P. Kotra, J. D. Pedelacq, V. Guillet, S. Mobashery and J. P. Samama, J. Am. Chem. Soc. 120 (1998) 9748. L. Maveyrad, I. Massova, C. Birck, K. Miyashita, J. P. Samama, S.Mobashery, J. Am.Chem. Soc. 118 (1996) 7435. L. Mourey, K. Miyashita, P. Swaren, A. Bulychev,J. P. Samama, S. Mobashery, J. Am. Chem. Soc. 120 (1998) 9383. W. L. Lee, Z. H. Li, S. Vakulenko, S. Mobashery, J. Med. Chem. 43 (2000) 128.

This Page Intentionally Left Blank

H. van der Goot (Editor) Trends in Drug Research III © 2002 Elsevier Science B.V. All rights reserved

207

PSEUDOMONAS AERUGINOSA QUORUM SENSING: A TARGET FOR ANTIPATHOGENIC DRUG DISCOVERY Everett P. Greenberg Department of Microbiology, University of Iowa, Iowa City, Iowa 52242 USA

It was once held that most bacteria function only as individuals designed to compete with one another and to multiply rapidly under appropriate conditions. This concept has given way to the view that like animals, bacteria can communicate with each other and form communities that represent more than the sum of the individuals [1-3]. Bacteria use chemicals to signal each other, and to coordinate their activities. Many Gram-positive bacteria use small peptides in signaling one another [4, 5], Gram-negative bacteria appear to use small molecule signals of various sorts [1, 6-9]. Perhaps the best-studied signaling system is the Gram-negative acyl-homoserine lactone (acyl-HSL) system. This type of bacterial cell-to-cell communication was first discovered in the context of microbial ecology but it is now evident that acyl-HSL signaling is important in plant and animal (including human) diseases. Acyl-HSL signaling is a dedicated communication system that is used by bacteria to control specific genes in response to population density. Acyl-HSLs are small molecule signals with no other known function. These chemical signals are produced by specific enzymes, and they are detected by specific receptors. Because acyl-HSL signaling provides a mechanism by which a bacterial species can monitor its' own population density, this type of signaling and other signaling systems that achieve the same purpose have been termed quorum sensing systems [10]. Acyl-HSL signals are generated by the activity of a single enzyme. The enzyme uses as substrates, 5-adenosylmethionine, and an intermediate of fatty acid biosynthesis, acyl-acyl carrier protein [11-15]. The enzyme is generally a member of the Luxl family of acyl-HSL synthases. Different Luxl homologs generate different acyl-HSLs. Thus the Pseudomonas aeruginosa Rhll primarily catalyzes the synthesis of A^-butyryl-HSL (C4-HSL) and the P. aeruginosa LasI directs the synthesis of N-(3-oxododecanoyl)-HSL (30C12-HSL). The acyl side-chain length and the substitutions on the side chain provide signal specificity. Acyl side chains of these signals can be fully saturated, or they can have hydroxyls or carbonyls on the third carbon. Acyl-HSLs with side chain lengths of 4 to 16 carbons have been identified [7], Short-chain signals like C4HSL diffuse freely through the cell membrane [16, 17], and 30C12-HSL partitions into cells, presumably in the membrane. This signal can diffuse into the surrounding environment but export is enhanced by the mexAB-oprM, and perhaps other efflux pumps [17, 18]. Regardless, the cellular concentration of an acyl-HSL is defined by the environmental concentration, and environmental

208 concentrations can rise only when there is a sufficient population of the signal producing bacterium. The specific receptors for acyl-HSL signals are members of the LuxR family of transcriptional regulators. LuxR family members have been proposed to consist of 2 domains, a C-terminal DNA-binding domain, and an N-terminal acyl-HSL-binding domain [19]. Quite often the two regulatory genes (the R and I genes) are linked but not always. The orientation of the two genes with respect to each other is variable. Acyl-HSL quorum sensing was first discovered to control the luminescence of Vibrio fischeri, a bacterium that forms a mutualistic Ught organ symbiosis with certain marine animals [20, 21]. Here quorum sensing is critical to the symbiosis. Acyl-HSL signaling is critical for virulence of the plant pathogen Erwinia carotovora [22], and for virulence of P. aeruginosa in mouse models of lung [23] and bum infections [24], in invertebrates [25-27], and in plants [28]. In this report P. aeruginosa serves as a model for the role of bacterial communication in community behaviors important in pathogenesis. That the P. aeruginosa quorum sensing system controls virulence makes it a target for ani-pathogenic drug development. Quorum Sensing in Pseudomonas aeruginosa P. aeruginosa can be isolated from soil and water. It is also an opportunistic pathogen of humans, other animals, and plants. One of the reasons P. aeruginosa is a successful opportunistic pathogen is that it produces a battery of secreted virulence factors. These virulence factors include exoproteases, siderophores, exotoxins, and lipases. Many of these virulence factors are regulated by quorum sensing [1, 29, 30]. Of what advantage to P. aeruginosa is quorum sensing control of virulence factors? First it is economical to produce extracellular factors only after a critical population has been achieved. A mass of cells is required to produce sufficient quantities of these factors to influence the surrounding environment. Furthermore, in the host, timing of the deployment of virulence factors may be critical. The pathogen can amass without displaying its virulence factors, and then the pathogen can mount a surprise attack in which the arsenal of virulence factors is deployed in a coordinated and overwhelming fashion. Genetic studies have revealed two quorum-sensing systems in P. aeruginosa. Both of these systems have linked R and I genes. They are the LasR-I and RhlR-I quorum sensing systems [31-36]. In addition, the recently completed P. aeruginosa genome sequencing project has revealed a third LuxR homolog that is adjacent to a cluster of quorum sensing controlled (qsc) genes [37]. However, a third Luxl homolog is not evident from the sequence, and the function of the third LuxR homolog is as yet unknown. LasR is a transciptional regulator that responds primarily to the Lasl-generated signal, 3-OC12-HSL, and RhlR is a transcriptional regulator that responds best to the Rasl-generated, C4-HSL. The current model for the quorum sensing in P. aeruginosa is as follows: at low population densities LasI produces a basal level of 3-OC12HSL. As density increases, 3-OC12-HSL builds to a critical concentration at which it interacts with LasR. This LasR-3-OC12-HSL complex then activates transcription of a number of genes including rhlR [29, 32, 36, 38, 39]. The

209 activation of rhlR by LasR results in a quorum sensing regulatory cascade, in which activation of the rhl system requires an active las system. RhlR responds best to the Rhll-generated C4-HSL. RhlR then activates expression of genes required for production of a variety of secondary metabolites such as hydrogen cyanide and pyocyanin [29]. A DNA sequence with dyad symmetry called a /wx-box-like sequence can easily be identified in the promoter regions of many quorum-sensing controlled (qsc) genes [10, 37, 40, 41]. By analogy to other acyl-HSL quorum sensing systems we deduce that the lux-box like sequences function as binding sites for LasR and RhlR. It is not yet clear how RhlR and LasR discriminate between their respective binding sites. In fact many genes show partial activation with either LasR or RhlR and the appropriate acyl-HSL [for example see 30, 37]. One explanation for this is that binding site discrimination is less than perfect and either LasR or RhlR can bind with varying efficiency to any lux box-like element. However, lux box-like sequences are not apparent in the promoter regions of all qsc genes. This suggests that LasR or RhlR may also bind to yet to be identified sequences, or that some qsc genes are controlled by LasR or RhlR indirectly. A recent study used a random mutagenesis approach to identify 39 gene that were highly regulated (minimum 5-fold induction, maximum 740-fold induction) by quorum sensing [37]. The genes were divided into 4 different classes, two of which respond to 3-OC12-HSL, and two of which required both C4-HSL and 3-OC12HSL for maximum induction. The qsc genes map throughout the P, aeruginosa chromosome, confirming the view that quorum sensing in this bacterium represents a global regulatory system [29]. The 39 genes revealed by the random mutagenesis study represent only a subset of the qsc genes in P. aeruginosa. It was estimated that as many as 4% of the roughly 6,000 P, aeruginosa genes are controlled by quorum sensing [37]. One report indicates that transcription of rpoS, a gene encoding an RNA polymerase subunit involved in expression of stationary phase factors is activated by RhlR and C4-HSL [42]. This raises the possibiUty that many genes may be controlled indirectly rather than directly by quorum sensing. It is also an enticing hypothesis because it lends itself to the idea that one specific cue that enables a cell to anticipate stationary phase is crowding. Unfortunately, quorumsensing control of rpoS transcription is an example for which there is limited evidence. It is also an example for which there are low levels of induction at best (3-fold). In fact recent investigations suggest that quorum sensing may have no significant influence on rpoS transcription in P. aeruginosa [43]. Biofilms and quorum sensing Bacteria often tend to attach to surfaces and form communities enmeshed in a self-produced polymeric matrix. These communities are called a biofilm [2, 44]. P. aeruginosa is often found in naturally occurring biofilms. Under the appropriate laboratory conditions, P. aeruginosa forms characteristic biofilms that can be several hundred micrometers thick. Development of a mature biofilm proceeds through a programmed series of events [2]. After attachment, cells multiply to form a layer on a solid surface. Individuals in the layer then exhibit a surface motiUty called twitching. Twitching is dependent on

210

Type IV pili. As a result of twitching motility small groups of P. aeruginosa called microcolonies form. Microcolonies then differentiate to form a mature biofilm. Microcolonies in a mature biofilm have tower and mushroom-shaped architectures. The cells in these structures are encased in an extracellular polysaccharide matrix. Water channels that allow the flow of nutrients into and waste products out of the biofilm innervate these structures. There is a significant physiological heterogeneity within biofilms. This heterogeneity in physiological activity makes studying biofilms with traditional molecular microbiological techniques difficult. Bacteria in these mature biofilms are phenotypically resistant to microbiocidal agents including antibiotics. Thus biofilms cause many different types of chronic or persistent bacterial infections [2]. Recent studies have linked quorum sensing and biofilm maturation [45]. This is a particularly gratifying finding because quorum sensing functions to control gene expression in groups of bacteria, and biofilms are just that, organized groups of bacteria. A mutation in lasl has a dramatic affect on biofilm maturation. Lasl mutants are incapable of 30-C12-HSL synthesis, and the development of Lasl mutant biofilms is arrested after microcolony formation but prior to maturation of the microcolonies into thick structured assemblages. Thus Lasl mutant biofilms appear flat and undifferentiated. The normal biofilms architecture can be restored to the mutant by addition of the Lasl-generated quorum sensing signal 30-C12-HSL. A Rhll mutant exhibits normal biofilm development and architecture. The 30-C12-HSL-responsive qsc genes involved in biofilm maturation remain unknown. ACKNOWLEDGEMENTS Work on quorum sensing in the author's laboratory has been funded by the NIH, NSF, Office of Navel Research, and the US Cystic Fibrosis Foundation. Similar accounts of quorum sensing have been published elsewhere.

REFERENCES [1] Fuqua, C , and Greenberg, E. P. (1998) Curr, Opin. Microbiol. 1,183-189. [2] Costerton, J. W., Stewart, P., and Greenberg, E. (1999) Science 284, 13181322. [3] Salmond, G. P. C , Bycroft, B. W., Stewart, G. S. A. B., and Williams, P. (1995) MoL Microbiol 16, 615-624. [4] Dunny, G. M., and Leonard, B. A. {l991)Ann, Rev. Microbiol. 51, 527-564. [5] Novick, R. P., and Muir, T. W. (1999) Curr. Opin. Microbiol. 2,40-45. [6] Fuqua, W. C , and Winans, S. C. (1994b) / Bacteriol. 176,2796-2806. [7] Fuqua, W. C , Winans, S. C , and Greenberg, E. P. (1996) Annu. Rev. Microbiol. 50,121'15l. [8] Pesci, E. C , Milbank, J. B., Pearson, J. P., Kende, A. S., Greenberg, E. P., and Iglewski, B. H. (1999) Proc. Natl. Acad. Sci. USA 96,11229-11234. [9] Bassler, B. L. (1999) Curr. Opin. Microbiol. 2,582-587.

211

[10] Fuqua, W. C , Winans, S. C , and Greenberg, E. P. (1994a) J. Bacteriol. 176, 269-275. [11] Schaefer, A. L., Val, D. L., Hanzelka, B. L., Cronan, J. E., Jr., and Greenberg, E. P. (1996) Proc. Natl Acad. ScL USA 93, 9505-9509. [12] Parsek, M. R., Val, D. L., Hanzelka, B. L., Cronan, J. E., Jr., and Greenberg, E. P. (1999) Proc. Natl. Acad. Sci. USA96,4360-4365. [13] More, M. I., Finger, D., Stryker, J. L., Fuqua, C , Eberhard, A., and Winans, S. C. (1996) Science 272, 1655-1658. [14] Hanzelka, B., Parsek, M. R., Val, D. L., Dunlap, P. V., J. E. Cronan, J., and Greenberg, E. P. (1999) /. Bacteriol. 181, 5766-5770. [15] Hoang, T. T., Y. Ma, R.J. Stem, M.R. McNeil, and Schweizer, H. P. (1999) Gene 237, 361-371. [16] Kaplan, H. B., and Greenberg, E. R (1985) J. Bacteriol. 163,1210-1214. [17] Pearson, J. P., Van Delden, C , and Iglewski, B. H. (1999) J. Bacteriol. 181, 1203-1210. [18] Evans, K., Passador, L., Srikumar, R., Tsang, E., Nezezou, J., and Poole, K. (1998) J. Bacteriol. 180, 5443-5447. [19] Stevens, A. M., and Greenberg, E. P. (1998) in Cell-Cell Signaling in Bacteria (Dunny, G., and Winans, S. C , Eds.) pp 231-242, ASM Press, Washington, DC. [20.] Nealson, K. H., and Hastings, J. W. (1979) Microbiol. Rev. 43,469-518. [21.]Ruby, E. G. (1996) Anna. Rev. Microbiol. 50, 591-624. [22] Pirhonnen, M., Flego, D., Heikiheimo, R., and Palva, E. T. (1993) EMBO J. 12, 2467-2476. [23] Tang, H. B., DiMango, E., Bryan, R., Gambello, M., Iglewski, B. H., Goldberg, J. B., and Prince, A. (1996) Infect. Immun. 64, 37-43. [24] Rumbaugh, K. P., Griswold, J. A., and Hamood, A. N. (1999) J. Burn Care Rehabil. 20,42-49. [25] Tan, M. W., Rahme, L. G., Sternberg, J. A., Tompkins, R. G., and Ausubel, F. M. (1999) Proc Natl Acad Sci USA96, 2408-2413. [26] Tan, M. W., Mahajan-Miklos, S., and Ausubel, F. M. (1999) Proc. Natl. Acad. Sci. f / 5 ^ 96, 715-720. [27] Mahajan-Miklos, S., Tan, M. W., Rahme, L. G., and Ausubel, F. M. (1999) Ce//96,47-56. [28] Rahme, L. G., Stevens, E. J., Wolfort, S. F., Shao, J., Tompkins, R. G., and Ausubel, F. M. (1995) Science 268, 1899-1902. [29] Pesci, E. C , and Iglewski, B. H. (1997) Trends in Microbiol. 5,132-135. [30] Van Delden, C , and Iglewski, B. H. (1998) Emerg Infect. Dis. 4, 551-560. [31] Brint, J. M., and Ohman, D. E. (1995) J. Bacteriol. 177, 7155-7163. [32] Gambello, M. J., Kaye, S., and Iglewski, B. H. (1993) Infect. Immun. 61, 1180-1184. [33] Latifi, A., Winson, K. M., Foglino, M., Bycroft, B. W., Stewart, G. S. A. B., Lazdunski, A., and WiUiams, P. (1995) Mol. Microbiol. 17, 333-344. [34] Ochsner, U. A., Koch, A. K., Fiechter, A., and Reiser, J. (1994) J. Bacteriol. 176, 2044-2054. [35] Ochsner, U. A., and Reiser, J. (1995) Proc. Natl. Acad. Sci. USA92, 64246428.

212 [36] Passador, L., Cook, J. M., Gambello, M. J., Rust, L., and Iglewski, B. H. (1993) Science 260,1127-1130. [37] Whiteley, M., Lee, K. M., and Greenberg, E. P. (1999) Proc. Natl. Acad. Sci. USA 96,13904-13909. [38] Gambello, M. J., and Iglewski, B. H. (1991) J. Bacteriol. 173, 3000-3009. [39] Seed, P. C , Passador, L., and Iglewski, B. H. (1995) J. Bacteriol. Ill, 654659. [40] Pearson, J. P., Pesci, E. C , and Iglewski, B. H. (1997) J. Bacteriol. 179, 5756-5767. [41] Pesci, E. C , Pearson, J. P., Seed, P. C , and Iglewski, B. H. (1997) J. Bacteriol. 179, 3127-3132. [42] Latifi, A., Foglino, M., Tanaka, K., Williams, P., and Lazdunski, A. (1996) Mol. Microbiol. 21,1137-1146. [43] Whiteley, M., Parsek, M., and Greenberg, E. P. Submitted for Publication. [44] Reimmann, C , Beyeler, M., Latifi, A., Winteler, H., Foglino, M., Lazdunski, A., and D, D. H. (1997) Mol. Microbiol. 24, 309-319. [45] de Kievit, T., Seed, P. C , Nezezon, J., Passador, L., and Iglewski, B. H. (1999) J. Bacteriol. 181, 2175-2184. [46] Albus, A. M., Pesci, E. C , Runyen-Janecky, L. J., West, S. E., and Iglewski, B. H. (1997) J. Bacteriol. 179, 3928-3935. [47] Costerton, J. W., Lewandowski, Z., Caldwell, D. E., Korber, D. R., and Lappin-Scott, H. M. (1995) Ann. Rev. Microbiol. 49, 711-745. [48] Davies, D. G., Parsek, M. R., Pearson, J. P., Iglewski, B. H., Costerton, J. W., and Greenberg, E. P. (1998) Science 280,295-298. [49] Stickler, D. J., Morris, N. S., McLean, R. J., and Fuqua, C. (1998) Appl. Environ. Microbiol. 64, 3486-90.

H. van der Goot (Editor) Trends in Drug Research III © 2002 Elsevier Science B.V. All rights reserved

213

Discovery and Development of New Anti-Bacterial Drugs Ian Chopra, Lars Hesse and Alexander O'Neill Antimicrobial Research Centre and Division of Microbiology, School of Biochemistry and Molecular Biology, University of Leeds, Leeds LS2 9JT, United Kingdom

INTRODUCTION The ability to treat bacterial infections with chemotherapeutic agents, introduced vy^ith the discovery of penicillin and prontosil in the 1930s (Figure 1), represents one of the most important medical achievements of the twentieth century. Indeed, the rapid advances made in the discovery of new antibiotics and other antibacterial agents during the so-called "golden" period between 1940 and the mid-1960s (Figure 1) led to widespread optimism that bacterial infections could be completely conquered. Figure 1. Discovery of antibacterial agents Empiric screening has been based on the identification of antibacterial agents by their ability to inhibit bacterial growth. Synthetic approaches comprise chemical modification of existing drug classes to improve their properties e.g. circumvention of resistance mechanisms to earlier members of the class. Only representative antibacterial agents are indicated. Empiric screening Cycloserine Erythromycin Ethionamide Isoniazid Metronidazole Pyrazinamide Rifamycin —Trimethoprim Vancomycin Virginiamycin Chlortetracycline •

1930

^

±

1940

- ^

•^ ^

^

Imipenem

Semi-synthetic penicillins & cephalosporins Oxazolidinones Defensins

Ethambutol Fusidic acid -^ Mupirocin Nalidixic acid

1960

Glycylcyclines

- ^ ^ Newer cartiapenems

Newer aminoglycosides > - Fluoroquinolones

\ 1950

Rifabutin Rifalazil Rifapentine

Semi-synthetic glycopeptides Semi-synthetic streptogramlns

Minocycline

- ^ Chloramphenicol Neomycin Polymixin Streptomycin — Thiacetazone

Newer macrolides & ketolides

- Rifampicin

^

Cephalosporin Cy

Penicillin Prontosil

Synthetic approaches

1970

1980

\

I

1990

2000

214 This period of optimism is captured by the famous remark made in 1969 by the US Surgeon General who testified to Congress that "the time has come to close the book on infectious disease" [1]. However, even from the very earliest period of the antibiotic era the potential for the emergence of drug resistant bacteria has been recognised (Figure 2), [2]. Figure 2. Emergence of antibacterial resistance (1944 - 2001). The first significant reports of resistance in clinical isolates are indicated.

Unfortunately the selection of organisms resistant to antibacterial agents has continued to the present day (Figure 2) and the next millennium has arrived with the dramatic emergence of resistance to antibacterial agents in all significant bacterial pathogens [1,3-6]. In some cases, organisms multiply resistant to virtually all chemotherapeutically useful antibacterial agents have been identified. The latest developments in this increasingly alarming situation have been the emergence of vancomycin intermediate resistant strains of Staphylococcus aureus (so-called "VISA" strains) [6] and linezolid resistant strains of Enterococcus faecium [7]. As noted by Cookson [8], the emergence of VISA strains may herald the appearance of high-level vancomycin-resistant strains of S. aureus. Such an event would have serious consequences for the control of nosocomial staphylococcal infections. The rapid emergence of linezolid resistance in the enterococci is equally disturbing since linezolid, a member of the new oxazolidinone class of antibiotics, was only approved by the US Food and Drug Administration (FDA) in April 2000 [6]. Although sensible measures to limit antibiotic usage to valid therapeutic indications and to reduce the spread of resistant organisms are of value in limiting the emergence of resistant organisms, the resistance problem continues to require renewed effort by the

215 pharmaceutical industry to create products that will prevent or treat infections caused by antibiotic-resistant pathogens [3,4]. Currently there are four principal drug discovery approaches: • Expansion of knov^n drug classes to include organisms resistant to earlier members of the class • Protection of known classes by resistance mechanism inhibitors • Re-evaluation of earlier pharmacophores • Discovery or design of new agents through rational selection of novel targets underpinned by genomics This chapter considers these approaches for the discovery of new antibacterial agents presenting the advantages and disadvantages of each strategy. Strategies will be illustrated by specific examples. Since the emphasis for future research will be the discovery of novel agents active against new molecular targets, it will be important to pursue approaches that minimise the emergence of bacterial resistance to newly discovered agents. This chapter also addresses this issue. 1.

EXPANSION OF KNOWN DRUG CLASSES TO INCLUDE ORGANISMS RESISTANT TO EARLIER MEMBERS OF THE CLASS

L1 Overview and examples Since the mid-1970s industrial approaches to the development of new antibacterial agents have been dominated by this paradigm (Figure 1). Indeed, in recent years, the oxazolidinones and defensins represent the only new classes of antibacterial agents to be developed. Expansion of known drug classes has been the principal strategy adopted by the pharmaceutical industry to combat the emergence of bacterial resistance to antibacterial agents. Essentially, this strategy depends on the synthesis of repertoires of new analogues related to known antibacterial agents to create structural modifications that circumvent the resistance phenotype. Early examples of this approach include development of semi-synthetic penicillins such as methicillin and the isoxazolyl penicillins which were developed on the basis of their stability to staphylococcal penicillinases [4]. More recent examples include the third generation cephalosporins, exemplified by agents such as cefotaxime (introduced in 1981) and ceftazidime (introduced in 1985), which v/ere developed on the basis of their stability to the TEM-1 and SHV-1 P-lactamases which are broadly dispersed amongst clinical isolates of Gram-negative bacteria [4]. The strategy has continued to the present day [9-13]. For example the glycylcyclines, exemplified by tigilcycline (GAR936) (Figure 3) which is currently in Phase II clinical trials, represent a new class of tetracycline analogues [12]. Glycylcyclines exhibit activity against bacteria expressing resistance to earlier tetracyclines by efflux and ribosomal protection mechanisms [12,13].

216 Figure 3. Structure of 9-t-butylglycyIamido minocycline (GAR-936;tigilcycline) N(CH3)2

N(CH3)2

1.2 Limitations of developing analogues of existing drug classes Expansion of existing drug classes to meet the clinical challenges imposed by resistant organisms has undoubtedly led to the introduction of a number of clinically successful agents (Figure 1). However, this approach can now only be considered at best a temporary solution to the problem of resistance. Unfortunately, the existence of resistance mechanisms to earlier members of the drug class often provides the organisms with a "head-start" for mutational adaptation by which expression of resistance to the newest member of the class also rapidly emerges. Notable examples include the relatively limited number of amino acid changes in the TEM-1 and SHV-1 P-lactamases required to confer resistance to the third generation cephalosporins [13] and the observation, at least under laboratory conditions, that single-site mutations in genes encoding tetracycline efflux pumps confer resistance to members of the new glycyIcycline group of antibiotics [13,14]. 2. PROTECTION OF KNOWN CLASSES BY RESISTANCE MECHANISM INHIBITORS 2.1 Overview and examples The bacterial enzymes that degrade or modify antibiotics, as well as the efflux systems that remove antibiotics from bacteria are themselves potential targets for drug action. The objective of this approach is the introduction of combination products which contain an antibiotic and a specific inhibitor that protects the antibiotic from enzymatic inactivation or removalfromthe cell [4]. The concept of using resistance mechanism inhibitors in conjunction with antibiotics has been most successfully applied in the area of P-lactamase inhibitor-P-lactam combinations [4]. The p-lactamase inhibitor clavulanic acid was the prototype molecule that established the value of this strategy. It exhibits weak antibacterial activity, but binds with high affinity and essentially irreversibly to many bacterial p-lactamases [4]. Clavulanic acid thus protects p-lactam antibiotics from destruction and is available commercially in combination with amoxycillin and ticarcillin (Table 1). Other Plactamase inhibitors, e.g. tazobactam and sulbactam, have also been introduced as therapeutic agents in combination with p-lactam antibiotics (Tablel) [4,15,16].

217 Table 1. Commercially available P-lactam/p-lactamase inhibitor combinations. (i.v. = intravenous; i.m. = intramuscular) Inhibitor

P-lactam antibiotic

Clavulanic acid

Amoxycillin

Augmentin

Oral, i.v.

Clavulanic acid

Ticarcillln

Timentin

I.v.

Tazobactam

Piperacillin

Zosyn

I.v.

Sulbactam

Ampicillin

Unasyn

I.m., I.v.

Antibiotic efflux is now recognised as a major mechanism of bacterial resistance to antibiotics. Some efflux pumps selectively extrude specific antibiotics (e.g. tetracyclines), while others, classified as multidrug resistance (MDR) pumps, mediate efflux of a variety of structurally diverse compounds with differing antibacterial modes of action [17,18]. In keeping with the paradigm of p-lactamase inhibitor-P-lactam combinations the discovery and development of efflux pump inhibitors would lead to products containing an antibiotic and an inhibitor that would prevent efflux of the drug from the bacterial cell. A number of new research programmes have recently been initiated in several pharmaceutical companies to identify and develop appropriate bacterial efflux pump inhibitors. These programmes have already identified a number of promising research leads. Limitations of space prevent complete discussion here. However, readers are referred to a recent comprehensive review [18] for further information. 2.2. Limitations associated with resistance mechanism inhibitors Although the existing P-lactamase inhibitor-p-lactam combinations have gained widespread acceptance as valuable therapeutic agents [15,16], their spectrum of activity encompasses only some of the clinically relevant P-lactamase enzymes (Table 2 ). Thus the three inhibitors developed to date are at best only weak inhibitors of the molecular class B and C p-lactamases. The deficiencies in the spectrum of activity of existing plactamase inhibitors has led to intensive research efforts to discover new inhibitor classes that also encompass the molecular group B and C enzymes. To date, these approaches have only identified reaearch leads. Readers are referred to the recent excellent review by Payne et al [16] for further details on the discovery of new plactamase inhibitors. A further limitiation of existing commercially available plactamase inhibitors, has been the development of inhibitor-resistant p-lactamase variants in clinically relevant organisms [16].

218 Table 2. Clinically important ^-lactamases and activity of serine p-lactamase inhibitors. Based on reference [16]. Molecular class Class A

Class B

Class C

Class D

P-lactamase type TEM-1

Clinical significance of enzymes

SHV-1

Most commonly found plactamases

Extended spectrum TEM/SHVs

Resistance to 3rd and 4th generation cephalosporins

Narrow spectrum carbapenemases

Carbapenem resistance

Broad spectrum carbapenemases (metallo enzymes)

Resistance to majority of plactams

Stably derepressed mutants

Resistance to 3rd generation cephalosporins

Plasm id-mediated Class C

As above, but with capacity for mobility

Extended spectrum OXA enzymes

Resistance to 3rd generation cephalosporins

Inhibition (ICjo, uM) by | Clavulanic Tazobactam Sult>actam 1 acid <1

<1

<1

<1

<1

<1

<1

1-5

5-20

>20

>20

>20

>20

6-20

5-20

>20

5-20

5-20

Variable

<1

Variable

The resistance inhibitor approach has undoubtedly produced important achievements in the area of p-lactamase inhibitors and may also subsequently deliver antibiotic efflux pump inhibitors for combination therapies. However, there has been little success in applying the strategy to inhibition of other antibiotic-inactivating enzymes [4]. 3

RE-EVALUATION OF EARLIER PHARMACOPHORES

3.7. Overview and examples The discovery and successful development of the antibiotics currently in use led to termination of interest in many other antibiotic classes discovered during the period 1950-1970 [4]. These under-exploited agents include the avilamycins, kirromycins, lankacidins, sideromycins, pleuromutilins, indolmycin and several natural product RNA polymerase inhibitors. A number of these agents appear to have structures and modes of action that are distinct from those of antibiotics in current use, suggesting that crossresistance with agents already in use may be minimal [4]. Therefore, re-evaluation of these older agents with a view to developing them for use against organisms resistant to current agents may be worthwhile. We have recently started to re-evaluate a group of RNA polymerase inhibitors discovered in the period 1950-1980 [19,20]. Since this work has only recently been completed it is probably unfamiliar to most readers. We have therefore decided to highlight it in this review. 3.2 RNA polymerase inhibitors Bacterial DNA-dependent RNA polymerase (RNAP) mediates the transcription cycle and represents a major point of regulation for prokaryotic gene expression. In addition to being indispensable for cell viability, this multi-subunit enzyme complex possesses other features that potentially render it an attractive target for antibacterial inhibitors. However, to date it remains underexploited, with only a single class of RNAP inhibitor (the rifamycins) in clinical use [19].

219 Bacterial core polymerase is a stable, non-covalent assembly of five polypeptide chains with the subunit composition P'pa^a"co [21]. The enzyme has been most extensively characterised in E. coli (Table 3). The roles of the individual subunits have been defined and functional sites, or regions important for subunit interactions, have been identified [22]. Promoter-specific initiation of transcription requires both core RNAP (E) and a sigma factor (a), which together comprise the holoenzyme. Thus, RNAP comprises a number of functionally distinct and essential sites, which may be amenable to inhibition by antimicrobial drugs. Table 3. The subunits of E.coli DNA-dependent RNA polymerase holoenzyme.

P' P

rpoC rpoB

Polypeptide MW(kDa) 155 151

Number in enzyme 1 1

c'' a

rpoD rpoA rpoZ

70 36.5 10

1 2 1

Subunit

Gene

CO

Function DNA binding DNA binding & catalysis of RNA synthesis Promoter-specific initiation Assembly/ Promoter recognition Assembly and stability of RNAP

Only one of these sites on RNAP has to date been exploited for antibacterial chemotherapy. The semi-synthetic rifamycin, rifampicin (rif), binds close to the active site of the catalytic p subunit [23] and inhibits initiation of RNA synthesis [24]. Resistance to the drug frequently arises by point mutations in the rpoB gene which encodes the p subunit [19]. On the basis of cross-resistance with rif*^ alleles, other rifamycins (e.g. rifabutin, rifapentine, rifalazil) and members of the structurally-related ansamycin class of which the rifamycins form part, likely bind at, or near to, the rifbinding site [25]. As rifampicin resistance has already arisen in the clinical setting, the derivation of further rifamycin/ansamycin compounds represents a strategy of questionable utility for the development of novel RNAP inhibitors. In order to select effective RNAP inhibitors that will not be compromised by preexisting rifamycin-resistance alleles in the clinic, it is essential that they bind to regions of RNAP distinct from the rif-binding site, or at least be unaffected by mutations that confer rifampicin resistance. A number of secondary metabolites, discovered in the last 50 years but not developed, appear to fulfill these criteria. These include the zwittermicins, myxopyronins, holomycin, thioloutin, coallopyronins and ripostatins (Figure 4) [19,20,26-28]. Although these inhibitors have yet to be classified into functional groups based on their site(s) of interaction with the enzyme, it is evident that they must exploit binding sites in the enzyme to cause inhibition of function, and that in some instances this site does not constitute the rifampicin-binding site. Nevertheless, although these compounds are likely to interact with RNAP at sites distinct from the rifamycins the possibility of advancing some of them as drug candidates may be compromised by lack of prokaryotic specificity (Table 4)

220 Figure 4. Structure of various RNA polymerase inhibitors. Zwittermicin A (a), Myxopyronin A (b), Holomycin (c), Thiolutin (d), Corallopyronin A (e), Ripostatin A (f).

OH

(a)

^ 2

NH2

(b) NHCOiDhb

(c)

(d) cr

(e)

(f)

^

'

r

N-cH,

^cHj

NHCQ2CH3

221 Table 4. Natural product bacterial RNAP inhibitors with targets apparently distinct from that of rifamycins. Antibiotic Class Ripostatins

Corallopyronins Myxopyrrohins

Pyrrothines Thiolutin Holomycin Zwittermicins

Specificity^ P

^

Point of inhibition Initiation

P

Elongation

P

Elongation?

P/E P P/E

Elongation Elongation 7

Comments No apparent crossresistance with the sorangicins

Reference _

.

27 Structurally similar to the corallopyronins. No reported in vivo activity.

Mutations in the p' gene (rpoC) confer resistance to the drug

27

28 20 26

^ P, prokaryotic; E, eukaryotic 3.3 Limitations of exploring earlier pharmacophores Although exploration of earlier antibiotic discoveries can reveal agents worthy of further consideration as chemotherapeutic agents, there are difficulties with this approach. For example the identification of most agents with antibacterial activity in the period 1950-1970 relied upon empiric screening with intact bacteria (i.e. growth inhibition screens) (Figure 1). This gives rise to the possibility that many of the earlier discoveries are not truly specific antibacterial agents i.e. their antibacterial activities reflect cytotoxic behaviour. This difficulty is illustrated in the case of several bacterial RNAP inhibitors discussed above that lack true prokaryotic specificity. Furthermore, even if prokaryotic specificity is evident the fact that an agent exhibits no apparent structural relationship to an existing class is not necessarily a guarantee for the absence of cross-resistance to existing agents. For example, in the context of the RNAP inhibitors discussed above it is evident that rif^ genotypes mediate cross-resistance to other RNAP inhibitors that are structurally distinct from the rifamycins, including streptolydigin and the sorangicins (Figure 5), suggesting binding sites displaying either partial or complete overlap with that of rifampicin [19].

222

Figure 5. Structures of various RNA polymerase inhibitors demonstrating cross resistance with each other. Rifampicin (a), Streptolydigin (b), Sorangicin A (c) CH,

(a)

CH

(b)

(C)

CH3

HO

CH3

223 4 DISCOVERY OR DESIGN OF NEW AGENTS THROUGH RATIONAL SELECTION OF NOVEL TARGETS UNDERPINNED BY GENOMICS 4.1 Overview and advantages In recent years a wealth of bacterial genome sequence data has become available to assist in the search for new drug targets. The application of this information to drug discovery, in conjunction with many other advances in molecular genetics and bioinformatics has been extensively reviewed in recent years, both by us and others [4,29-33]. Although genomic methods are of major importance for antibiotic discovery, in the light of these comprehensive reviews we will restrict ourselves to only a limited discussion in this chapter. Genomic approaches, in contrast to earlier empiric screening methods (particularly growth inhibition screens), permit a more focused approach to drug discovery. Furthermore, through parallel advances in human genome sequencing, prokaryotic selective targets can be chosen and the antibacterial spectrum of a potential drug can be predicted based upon the distribution of a particular target gene in a range of bacterial pathogens. Genomic based methods and associated advances in molecular genetic techniques and crystallographic analysis will also permit an increasing number of targets to be subjected to structure-based drug design approaches. A further advantage of genome-based methods will be the identification and development of targets with potential for multiple blockade [3,13]. Simultaneous inhibition of more than one target renders the emergence of resistance less likely since mutations are required in all targets to confer resistance to the drug. This approach is most likely to be successful with groups of essential bacterial enzymes that are mechanistically related where it may be possible to design or screen for a single inhibitor of more than one member of the enzyme class. Examples of related enzyme systems that might be amenable to multiple blockade with a single agent are the mycobacterial enzymes involved in the biosynthesis of the mycolic acid and arabinogalactan cell envelope components of these organisms, and the tRNA synthetases, two-component signal transduction systems and muramyl peptide ligases which are all potential broad-spectrum drug targets. The muramyl peptide ligases encoded by murD, murE and inurF which are involved in the early stages of peptidoglycan synthesis represent particularly attractive targets for the development of an antibacterial drug with multiple blockade properties [13]. Each of these enzymes is mechanistically related (i.e. they are all non-ribosomal ATPdependent peptide synthases), they have an essential role in peptidoglycan synthesis (i.e. they are lethal targets), and they utilise D-amino acids or we^o-diaminopimelic acid, which having no eukaryotic counterpart, provides good prospects for discovering a selective bacterial inhibitor. Furthermore the enzymes are soluble which has allowed acquisition of structural data [13,34] thereby permitting application of structure-based inhibitor design methods.

224 4.2 Disadvantages Despite the great advantages for drug discovery offered by genome-based approaches, there are some limitations [33]. Strictly speaking these are not limitations associated with the generation of genomic sequence data per se, but relate to limitations of downstream screening or structure-based methods. Thus genomic approaches which result in the development of cell-free screens may result in the identification of inhibitors which cannot reach their target site in the organism and structure-based methods, while also suffering from this disadvantage, require X-ray refraction of the target to <2 angstroms resolution and powerful computer resources to identify a potential inhibitor that may, in practice, be difficult to synthesise. 5.

CONCLUSIONS

The twentieth century witnessed the discovery and development of many chemotherapeutic agents for the treatment of bacterial infections. Indeed, these developments must be regarded as some of the most significant medical achievements of the last century. The era of antibacterial chemotherapy has enabled physicians to treat infection rather than simply offer palliative care. Unfortunately, as we enter the new millennium many of our existing antibacterial agents are under threat from the widespread emergence of bacterial resistance. New agents are needed to counter this threat. This chapter has discussed the four major research areas within which such agents are likely to be identified or discovered. In view of a number of general limitations reached with the first approach (expansion of known drug classes), it is likely that greater emphasis will, in future, be placed on approaches 2-4. Even though new drug classes will be identified, we should not underestimate the ability of pathogenic bacteria to adapt to new selective pressures imposed by the introduction of new agents. Therefore the discovery and development of new drugs with a minimal potential for emergence of future resistance is of paramount importance. 6.

ACKNOWLEDGEMENTS

The work of the authors described here has been supported by grants to IC from SmithKline Beecham Pharmaceuticals, Smith and Nephew Research and Intrabiotics Pharmaceuticals. 7.

REFERENCES

[1] M.L. Cohen, Nature 406: (2000) 762. [2] E.P. Abraham and E. Chain, Nature 146: (1940) 837. [3] L.L. Silver and K.A. Bostian, Antimicrob. Agents Chemother. 37: (1993) 377. [4] I. Chopra, J. Hodgson, B. Metcalf and G. Poste, Antimicrob. Agents Chemother. 41.(1997)497. [5] A. Morris, J.D. Kellner and D.E. Low, Curr. Opin. Microbiol. 1: (1998) 524 [6] F.C. Tenover, J.W. Biddle and M.V. Lancaster, Emerg. Infect. Dis. 7: (2001) 327. [7] R.D. Gonzales, P.C. Schreckenberger, M.B., Graham, S. Kelkar, K. DenBesten and J.P. Quinn, Lancet 357: (2001) 1179.

225 [8] B.D. Cookson, Culture 19:(1998) 1. [9] D.T.W. Chu, Curr. Opin. Microbiol. 2: (1999) 467. [10] V.J. Lee, G.H. Miller and M. Yagisawa, Curr. Opin. Microbiol 2: (1999) 475. [11] M. Fines and R. Leclercq, Curr. Opin. Anti-Infect. Invest. Drugs 1: (1999) 443. [12] A.P. Johnson, Curr. Opin. Anti-Infect. Invest. Drugs 2: (2000) 164. [13] I. Chopra, Culture 20: (1999) 3. [14] M. Tuckman, P.J. Petersen and S.J. Projan, Microbiol. Drug Res. 6: (2000) 277. [15] L.M. Bush and C.C. Johnson, Inf. Dis. Clin. N. Am. 14: (2000) 409. [16] D.J. Payne, W. Du and J.H. Bateson, Exp. Opin. Invest. Drugs 9: (2000) 247. [17] RNikaido, Curr. Opin. Microbiol. 1: (1998) 516. [18] O. Lomovskaya and W. Watkins, J. Mol. Microbiol. Biotechnol. 3: (2001) 225. [19] A. O'Neill, B. Oliva, C. Storey, A. Hoyle, C. Fishwick and I. Chopra, Antimicrob. Agents Chemother. 44: (2000) 3163. [20] B. Oliva, A. O'Neill, J. M. Wilson, P. J. O'Hanlon and I. Chopra, Antimicrob. Agents Chemother. 45: (2001) 532. [21] L. Minakhin, S. Bhagar, A. Brunning, E. A. Campbell, S. A. Darst, R. H. Ebright and K. Severinov, Proc. Natl. Acad. Sci. (USA) 98: (2001) 892. [22] K. Severinov, Curr. Opin. Microbiol. 3: (2000) 118. [23] K. Severinov, A. Mustaev, E. Severinov, M. Kozlov, S. A. Darst and A. Goldfarb, J. Biol. Chem. 270: (1995) 29428. [24] D. D. Anthony, E. Zeszotek and D. A. Goldthwait, Proc. Natl. Acad. Sci. (USA) 56: (1966) 1026. [25] B. Yang, H. Koga, H. Ohno, K. Ogawa, M. Fukuda, Y. Hirakata, S. Maesaki, K. Tomono, T. Tashiro, and S. Kohno, J. Antimicrob. Chemother. 42: (1998) 621. [26] E. V. Stabb and J. Handelsman, Mol. Microbiol. 27: (1998) 311. [27] H. Reichenbach and G. Hoffle, Myxobacteria as producers of secondary metabolites, p. 149: Drug Discovery from Nature, Eds. S. Grobley and R. Thiericke, Springer-Verlag, Berlin. [28] G. G. Khachatourions and D. J. Tipper, J. Bacteriol. 119: (1974) 795. [29] D. J. C. Knowles, Trends Microbiol. 5: (1997) 379. [30] A. E. AUsop, Curr. Opin. Microbiol. 1: (1998) 530. [31] D. T. Muir, K. J. Shaw, R. S. Hare and G. F. Vovis, Antimicrob. Agents Chemother. 43: (1999) 439. [32] M. Y. Galperin and E. V. Koonin, Curr. Opin. Biotech. 10: (1999) 571. [32] T. Block and R. Hare, Curr. Opin. Microbiol. 3: (2000) 522. [33] J. Rosamond and A. Allsop, Science 287: (2000) 1973. [34] Y. Yan, S. Munshi, B. Leiting, M. S. Anderson, J. Chrzos,and Z. Chen. J. Mol. Biol. 304: (2000) 435.

This Page Intentionally Left Blank

H. van der Goot (Editor) Trends in Drug Research III © 2002 Elsevier Science B.V. All rights reserved

227

Discovery of Small Molecule Insulin Mimetics as Potential Novel Antidiabetic Agents BeiB. Zhang R80W250, Merck Research Laboratories, P.O. Box 2000 Rahway, New Jersey 07065 E-mail: bei [email protected] Summary Insulin elicits diverse biological responses by binding to its specific receptor. A non-peptidyl fungal metabolite (L-783,281) was identified in a screen for small molecules that activate the human insulin receptor (IR). An active analog with enhanced IR kinase activation potency and selectivity over related receptors was also characterized. These compounds acted as insulin mimetic agents in several biochemical and cellular assays and demonstrated antidiabetic activity in rodent models of diabetes. Thus, small molecule IR activators exert insulin mimetic and sensitizing effects in cells and in animal models of diabetes. These results have implications for the future development of new therapies for diabetes mellitus. Pharmaceutical intervention aimed at augmenting insulin receptor function may ultimately prove beneficial as novel therapeutic option in patients with diabetes.

Introduction Diabetes is a chronic metabolic disorder affecting ~ 5% of the population in the industrialized nations. The more prevalent form, type 2 diabetes, accounts for more than 90% of cases. The pathogenesis of type 2 diabetes is complex, involving progressive development of insulin resistance and a relative deficiency in insulin secretion, leading to overt hyperglycemia [1]. Currently available therapies for type 2 diabetes include insulin and the following oral agents: sulfonylureas, metformin, a-glucosidase inhibitors such as acarbose, thiazolidinediones including pioglitazone and rosiglitazone. These agents are used as monotherapy in newly diagnosed patients. In more advanced patients, such drugs are frequently used in combination to achieve better glycemic control [2]. Each of the above oral agents suffers from generally inadequate efficacy (as monotherapy) and a number of serious adverse effects [2]. As a consequence, there continues to be a high demand for new oral antidiabetic drug with novel mechanisms.

228 Insulin elicits a diverse array of biological responses by binding to its specific receptor [3]. The insulin receptor (IR) is a heterotetrameric protein consisting of two extracellular a subunits and two transmembrane 6 subunits. The binding of the ligand to the a subunit of IR stimulates the tyrosine kinase activity intrinsic to the P subunit of the receptor. Extensive studies have indicated that the ability of the receptor to autophosphorylate and phosphorylate intracellular substrates is essential for its mediation of the complex cellular responses of insulin [4-7]. Insulin receptors trans phosphorylate several immediate substrates (on Tyr residues) including insulin receptor substrate (IRS) proteins 1 - 4 , She, and Gab 1, each of which provide specific docking sites for other signaling proteins containing Src homology 2 (SH2) domains [8]. These events lead to insulinmediated activation of glucose transport and glycogen synthesis through activation of downstream signaling molecules including phosphatidylinositol-3kinase (PI-3-kinase) and Akt (or PKB) [9] [10]. Insulin is essential for maintaining glucose homeostasis and regulating carbohydrate, lipid, and protein metabolism [11]. Alterations of IR in specific tissues and/or IRS 1 and 2 function via genetic manipulation have been shown to produce insulin resistance and diabetes in mice [12-19]. Decreased cellular responses to insulin or perturbation of the insulin signaling pathways are associated with a number of pathological states. Mutations in insulin receptor gene which lead to alterations of receptor synthesis, degradation, and function have been described in patients with several uncommon syndromes associated with severe insulin resistance [20]. The molecular basis for insulin resistance that proceeds, or is associated with, common forms of Type 2 diabetes remains poorly understood. However, several studies have shown modest decreases in insulin receptor number attributed to downregulation in response to hyperinsulinemia in tissues or cells from NIDDM patients [21,22]. Substantial decreases in insulinstimulated receptor tyrosine kinase activity and an even more substantial defect in receptor-mediated IRS phosphorylation or PI-3-kinase activation have been described using samples of tissue (e.g. muscle or fat) from rodents or human subjects with Type 2 diabetes [23-25]. Pharmaceutical intervention aimed at augmenting insulin receptor function may ultimately prove beneficial in patients with diabetes. ^Abbreviations: IR, insulin receptor; TK, tyrosine kinase; GST, glutathion-Stransferase; WGA, wheat germ agglutanin; IRS, insulin receptor substrate; PAGE, polyacrylamide gel electrophoresis; CHO, Chinese hamster ovary; SDS, sodium dodecyl sulfate; Compound 1, 2,5-Dihydroxy-3-[7-(3-methyl-but-2-enyl)-lHindol-3-yl]-6-[2-(l,l-dimethyl-allyl)-lH-indol-3-yl]-[l,4]benzoquinone; compound 2, 2,5-Dihydroxy-3-(l-methyl-lH-indol-3-yl)-6-phenyl[l,4]benzoquinone.

229 Discovery of Small Molecule Insulin Receptor Activators I. Lead identification A cell-based 96-well screening assay was designed to identify small molecule insulin receptor activators. This assay utilizes CHO cells (CHO.IR) which overexpress human insulin receptor [26]. Following incubation of intact cells with insulin or test compounds, insulin receptors are immunopurified using a monoclonal antibody specific for insulin receptor. Activation of insulin receptor is determined by measuring the receptor tyrosine kinase activity towards an exogenous substrate (polyGluiTyr 4:1). This assay was employed to screen synthetic chemicals and natural product extracts in an effort to identify novel compounds capable of activating insulin receptor. Such an approach is advantageous over a receptor-binding assay since small molecules that compete with insulin binding are likely to be antagonists for the receptor. Furthermore, the assay was designed to identify compounds that work in intact cells either by mteracting directly with the receptor or by indirectly stimulating receptor activation (e.g., a specific phosphatase inhibitor). By employing the cell-based insulin receptor activation assay, a large collection of synthetic compounds and natural product extracts were screened. This effort led to the identification of small molecule compound (Compound 1), L-783,281, isolated from a fungal culture [27]. Compound 1 induced activation of insulin receptor tyrosine kinase activity in CHO.IR cells with EC50 of 5 uM (Table 1). To identify a more potent and selective analog of L-783,281, derivatives of this compound were synthesized and tested. One of these derivatives (compound 2) increased the insulin receptor tyrosine kinase activity in these cells with an EC50 of 300 nM, reflecting a greater than 10-fold improvement in the potency compared to compound 1 (Table 1). In contrast, a closely related synthetic analog (compound 3) or a natural product analog (compound 4) were not effective in activating IRTK in the same assay at concentrations up to 100 |LiM [28,29], suggesting that the effect of compound 1 and compound 2 are likely due to specific activation of the insulin receptor.

Compound 1 Compound 2 Compound 3 Compound 4 EC50 = 5 MM 0.3 MM > 100 MM > 100 MM Table 1. Structure of compounds and activity in cell-based IRTK activation assay. CHO.IR cells were serum starved for 2 hr and then treated with insulin or test compounds for 20 min. The lysates of cells were prepared and receptors were captured with anti-IR antibody and IRTK activity was measured using [y ^^P]ATP and polyGlu:Tyr (4:1) as substrates. EC50 values are shown.

230 11. In Vitro and In Vivo Activation of Insulin Signaling Pathway Compound 1 and compound 2 induced tyrosine phosphorylation of the IR P subunit and IRS-1 in CHO.IR cells, as evidenced by anti-phosphotyrosine inmiunoblotting. Activation of these proximal insulin signaling molecules was coupled to stimulation of other components of the signaling cascade. Thus, compound 1 stimulated PI 3-kinase activity [30,31] and both active compounds stimulated phosphorylation of Akt kinase on the activating Ser-473 residue [32] in CHO.IR cells. Moreover, the effect of insulin and compound 2 on activation of Akt can be blocked by pretreatment with wortmannin (an inhibitor of PI-3 kinase), suggesting the activation of a PI-3 kinase-dependent pathway [33,34]. Compound 3 did not stimulate phosphorylation of IR P subunit, IRS-1, or Akt, suggesting that it was not effective in activating the insulin signal transduction pathway. Compound 1 induced a classical insulin-like effect by acutely stimulating glucose uptake in rat primary adipocytes (2.6 fold at 10 jiM) and in isolated soleus muscle from lean mice (2.4 fold at 2 jiM). These results indicated that the small molecule insulin mimetic compound was capable of modulating multiple steps of insulin signal transduction pathway. In order to determine whether the compounds are capable of modulating activation of insulin receptors in vivo, we examined the IRTK activity in liver extracts prepared from mice. Lean normal mice (db/+) were treated with compound 1 (at 50 and 150 mg/kg) or vehicle for 2 hours and then given an injection of saline or insulin via the tail vein prior to preparation of liver extracts. In vehicle treated groups, high dose of insulin (2 U/kg) induced a ~ 6 fold increase in the hepatic IRTK activity. Treatment with compound 1 resulted in significant increase in basal IRTK activity to a level that was comparable to -- 30% of that stimulated by injection of moderate dose of insulin (0.4 U/kg). Furthermore, compound 1 potentiated insulin activation of IRTK in the liver. Similar insulin sensitizing effect was also observed in mice treated with compound 2 (at 10 mg/kg). In studies with compound 2, insulin stimulated IRTK activities in treated groups were 177% and 143% of those in the vehicle groups for db/+ and db/db mice, respectively (p < 0.05, n = 8 - 10 in each group). In contrast, treatment with compound 3 (at 10 mg/kg) had no effect on insulin-stimulated IRTK activity. These data demonstrate the ability of compounds 1 and 2 to activate and enhance insulin activation of IRTK in vivo, which may account for the antidiabetic effects of this class of compounds. III. Selectivity for Insulin Receptor Insulin receptor belongs to a super family of receptor tyrosine kinases with high degree of sequence homology in the tyrosine kinase domain [35,36]. This family includes the insulin receptor and the receptors for many growth factors such as insulin-like growth factor I (IGFI), epidermal growth factor (EOF), and platelet-derived growth factor (PDGF). Activation of receptor tyrosine

231 kinases leads to a wide variety of beneficial biological effects ranging from metabolic regulation, embryonic development and tissue regeneration. These receptor tyrosine kinases also play pivotal roles in pathological conditions including diabetic retinopathy, atherosclerosis, and deleterious neoplastic transformation. Thus, a critically important aspect in the search of small molecules capable of activating IRTK is to determine the selectivity of such compounds. Although these tyrosine kinases are highly homologous, small variations in primary sequences have provided bases for identification of potent and exquisitely selective inhibitors [37-42]. In order the determine the specificity of test compounds for insulin receptor vs. other selected homologous receptors, parallel cell-based assays were established and used to counter screen insulin receptor activators against IGFIR, EGER, and PDFGR. In CHO cells overexpressing IGFIR (CHO.IGFIR), compound 1 (at 10 |iM) did not stimulate IGFIR or IRS-1 tyrosyl phosphorylation and no other compound-mediated tyrosyl protein phosphorylation was evident, suggesting that the compound is selective for IR vs. IGFIR activation. In subsequent studies using CHO.IGFIR cells, EGFR receptor overexpressing cells (CHO.EGFR), and quantitative tyrosine kinase (IGFIR) or anti-phosphotyrosine ELISA (EGFR) assays, compound 1 was shown to induce weak IGFIR and EGFR activation; 100 \xM concentrations were required to achieve 50% efficacy of either IGFI or EGF. Based on these data, compound 1 has > 10-fold selectivity for insulin receptor and it may provide a scaffold for building in greater selectivity [27]. Indeed, in all these cell lines, compound 2 failed to activate IGF-1, EGF, or PDGF receptors at concentrations up to 30 |LiM, representing an -100 fold selectivity for insulin receptor vs. the other homologous receptors [29]. IV. Direct Interaction with Insulin Receptor To establish that the effect of compound 1 observed in the cell-based assays is due to direct activation of IR, experiments were conducted to study the mechanism of action of the compound. Several lines of evidence suggested that compound 1 acted to directly activate the IR intracellular P-subunit (tyrosine kinase domain) [27]. Firstly, transfected CHO cells (CHO.IRR/IR) which overexpress chimeric receptors composed of the insulin receptor intracellular domain fused to the non-homologous insulin receptor-related receptor (IRR) extracellular domain [43] were used to show that compound 1 (but not insulin) could still activate receptor tyrosine kinase activity in intact cells. Secondly, Compound 1 did not displace radiolabeled insulin binding to IRs expressed in intact CHO.IR cells, nor was the affinity of insulin for the receptor affected. Thirdly, direct in vitro incubation with compound 1, but not compound 4, was able to increase tyrosine kinase activity of recombinant IR tyrosine kinase protein. Furthermore, insulin receptor was partially purified from CHO.IR cells using WGA affinity chromatography. When the partially purified IR was incubated in the presence y-^^P-ATP and a 12-mer IR peptide substrate, compounds I and 2

232 stimulated IR kinase activity in a dose-dependent manner as measured by increased incorporation of radiolabel into the substrate peptide. The mechanism for activation of insulin receptor tyrosine kinase has been a subject of intensive investigation. High-resolution structural information has been obtained through crystallographic studies of IR kinase domain [44,45]. Based on crystal structures of the unphosphorylated low* activity form, as well as the phosphorylated, active form of the IR, a model of cis-inhibition and transactivation of the receptor was proposed. The unliganded receptors exist in the autoinhibitory conformation that prevents access of ATP and substrate to the active site. Upon autophosphorylation of Tyrll58, Tyrll62 and Tyrll63 in the activation loop, the IR kinase undergoes a major conformational change resulting in unrestricted access of ATP and substrate to the active site and full activation of the kinase. More recently, the three dimensional (3D) structure of insulin receptor bound to insulin was determined by electron cyromicrospcopy [46]. The 3D reconstruction of the quaternary structure reveals that the both a subunits are involved in insulin binding and that the two P subunits are poised for transautophosphorylation. These structural studies have provided molecular basis of activation of IR. In order to ascertain whether interaction of compound 1 with IRTK domain could cause potential alterations in the conformation of the protein, recombinant IR intracellular domain protein (48 kDa) was incubated with the compound followed by partial protease digestion. This approach revealed that the compound induced a change in protease sensitivity of IRTK as illustrated by altered proteolytic pattern of the protein [27]. Further experiments demonstrated that interaction of compound 1 with the IR kinase domain apparently altered the conformation of the protein in the region encompassing the ATP binding site(s). This could conceivably lead to activation of the kinase by partially relieving the cis-inhibition of the enzyme. Structure biology studies will be necessary to further elucidate the mechanism of action. V. Antidiabetic Activity in Rodent Models of Diabetes The results of in vitro experiments demonstrated that compounds 1 and 2 are insulin mimetic agents with selectivity for insulin receptor in cell based assays, suggesting that they might exhibit substantial glucose lowering effects in vivo. This hypothesis was tested using db/db and ob/ob mice; obese animal models of type 2 diabetes characterized by severe insulin resistance and hyperglycemia. The purpose of the in vivo studies was to determine the potential ability of the compounds to correct the diabetic state in the mouse models without resulting in undesirable side effects. Oral administration of compounds 1 and 2 to diabetic db/db mice was shown to be efficacious for lowering of elevated blood glucose. Single-dose oral administration of compound 1 resulted in dose-dependent glucose lowering with > 50% transient correction of hyperglycemia achieved at 25 mg/kg/day (over 3-6

233

hours; food withheld) [27]. Likewise, single dose oral administration of compound 2 (5 mg/kg) resulted in significant lowering of blood glucose (over 2-4 hours; food withheld), achieving ~ 50% transient correction of hyperglycemia. However, treatment of db/db mice with compound 3 (at 30 mg/kg) did not alter the elevated blood glucose levels [29]. This finding is consistent with the inability of this compound to activate IR in the in vitro assays and demonstrates a correlation between activation of IR and glucose lowering with this class of insulin receptor activators. Having established that compound 2 can lower glucose upon acute treatment, the ability of compound 2 to correct hyperglycemia in db/db mice after long-term treatment was also determined. Mice were treated with a daily dose of compound 2 or 3 for 8 days. At day 8, the extent to which compound 2 treatment resulted in correction of hyperglycemia was 35% at 1 mg/kg and 76 % at 10 mg/kg, respectively. Under similar conditions compound 3 (at 10 mg/kg) was not efficacious in lowering glucose in these animals. Furthermore, similar treatment with compound 2 at 1 or 10 mg/kg had no significant effect on blood glucose levels in normal lean mice. In addition, when compound 2 was administered to ob/ob mice with extreme hyperinsulinemia and mild hyperglycemia, significant reductions in glucose insulin levels were observed. In streptozotocin-induced diabetic mouse model, compound 2 potentiated insulin's effect on decreasing hyperglycemia [29]. Furthermore, compound 2 was able to maintain or improve glucose disposal in the presence of reduced insulin levels in the non-diabetic Sprague Dawley rats without causing hypoglycemia [29]. These data indicate that the small molecule insulin receptor activators function as efficacious antidiabetic agents in rodent models of diabetes. Conclusion As we enter the new millennium, we are faced with the challenge of rising prevalence of diabetes and related metabolic disorders. The recently completed United Kindom Prospective Diabetes Study (UKPDS) has highlighted the importance of antihyperglycemic agents in tight glycemic control and intervention of microvascular complications. New approaches for the treatment of diabetes represent an urgent medical need. The discovery of insulin receptor activators demonstrated that small, nonpeptidyl molecules are capable of mimicking the in vitro and in vivo function of a protein hormone by interacting with and activating its receptor. It is worth noting that small molecule agonists have been identified for other peptitdyl hormone receptors such as erythropoeitin and granulocytecolony-stimulating factor receptors [47-49]. These agonists interact with the extracellular ligand binding domain and induce dimerization of the monomeric receptor subunits leading to activation of the receptors. Earlier studies have indicated that protein tyrosine phosphatase (FTP) inhibitors can also act as insulin mimetic agents in vitro and in vivo [50,51]. Recent studies with FTP IB knockout

234 mice further validated that PTPIB plays an important role in modulation of insulin signaling and that specific inhibitors of PTPIB represent a potential treatment for diabetes or other metabolic disorders [52,53]. Thus, further exploration of agents that specifically activate insulin receptor will hopefully lead to safe and efficacious drugs that will become mainstays of next generation therapeutics of diabetes.

Acknowledgement The author is grateful to colleagues at Merck Research Laboratories who contributed to the work summarized in this review article.

References [I] [2] [3] [4] [5] [6] [7] [8] [9] [10] [II] [12] [13] [14] [15] [16] [17]

Saltiel, A. R. (2001) Cell 104, 517-529 DeFronzo, R. (1999) Ann Inter Med 17(13), 281-303 White, M. F., and Kahn, C. R. (1994) J Biol Chem 269, 1-4 Kasuga, M., Karlsson, F. A., and Kahn, C. R. (1982) Science 215, 185187 Rosen, O. M., Herrera, R., Olowe, Y., Petruzzelli, M., and Cobb, M. H. (1983) Proc Natl Acad Sci USA 80, 3237-3240 Yu, K. T., and Czech, M. P. (1984) / Biol Chem 259(8), 5277-86 Ellis, L., Morgan, D. O., Clauser, E., Roth, R. A., and Rutter, W. J. (1987) Mol Endocrinol 1, 15-24 White, M. F., and Yenush, L. (1998) Curr Top Microbiol Immunol 228, 179-208 Kohn, A. D., Summers, S. A., Birnbaum, M. J., and Roth, R. A. (1996) / Biol Chem 271(49), 31372-8 Alessi, D. R., and Cohen, P. (1998) Curr Opin Genet Dev 8(1), 55-62 Rosen, O. M. (1989) Diabetes 38,1508-1511 Bruning, J. C , Winnay, J., Cheatham, B., and Kahn, C. R. (1997) Mol Cell Biol 17(3), I5l3'2l Bruning, J. C , Winnay, J., Bonner-Weir, S., Taylor, S. I., Accili, D., and Kahn, C. R. (1997) Cell 88(4), 561-72 Bruning, J. C , Michael, M. D., Winnay, J. N., Hayashi, T., Horsch, D., Accili, D., Goodyear, L. J., and Kahn, C. R. (1998) Mol Cell 2(5), 559-69 Kulkarni, R. N., Winnay, J. N., Daniels, M., Bruning, J. C , Flier, S. N., Hanahan, D., and Kahn, C. R. (1999) / Clin Invest 104(12), R69-75 Kulkarni, R. N., Bruning, J. C , Winnay, J. N., Postic, C , Magnuson, M. A., and Kahn, C. R. (1999) Cell 96(3), 329-39 Lauro, D., Kido, Y., Castle, A. L., Zarnowski, M. J., Hayashi, H., Ebina, Y., and Accili, D. (1998) Nat Genet 20(3), 294-8

235 [18] [19] [20] [21] [22]

[23]

[24] [25] [26] [27]

[28]

[29]

[30] [31] [32]

[33] [34] [35] [36] [37]

Taylor, S. I. (1999) Cell 97(1), 9-12 Kido, Y., Burks, D. J., Withers, D., Bruning, J. C , Kahn, C. R., White, M. F., and Accili, D. (2000) / Clin Invest 105(2), 199-205 Taylor, S. I. (1992) Diabetes 41, 1473-1490 Olefsky, J. M. (1976) J Clin Invest 57, 1165-1172 Caro, J. F., Itoop, O., Pories, W. J., Meelheim, D., FUckinger, E. G., Thomas, F., Jenquin, M., Silverman, J. F., Khazanie, P. G., and Sinha, M. K. (1986) / Clin Invest 78,249-258 Caro, J. F., Sinha, IM. K., Raju, S. M., Ittoop, O., Pories, W. J., FUckinger, E. G., Meelheim, D., and Dohm, G. L. (1987) / Clin Invest 79, 1330-1337 Goodyear, L. J., Giorgino, F., Sherman, L. A., Carey, J., Smith, R. J., and Dohm, G. L. (1995) J Clin Invest 95,2195-2204 Kerouz, N. J., Horsch, D., Pons, S., and Kahn, C. R. (1997) J Clin Invest 100(12), 3164-72 Zhang, B., Szalkowski, D., Diaz, E., Hayes, N., Smith, R., and Berger, J. (1994) Journal of Biological Chemistry 269, 25735-25741 Zhang, B., Salituro, G., Szalkowski, D., Li, Z., Zhang, Y., Royo, I., Vilella, D., Diez, IVL. T., Pelaez, F., Ruby, C , Kendall, R. L., IVlao, X., Griffin, P., Calaycay, J., Zierath, J. R., Heck, J. V., Smith, R. G., and Moller, D. E. (1999) Science 284(5416), 974-7 Liu, K., Xu, L., Szalkowski, D., Li, Z., Ding, V., Kwei, G., Huskey, S., Moller, D. E., Heck, J. V., Zhang, B. B., and Jones, A. B. (2000) J Med C/i^m 43(19), 3487-94 Qureshi, S. A., Ding, V., Li, Z., Szalkowski, D., Biazzo-Ashnault, D., Xie, D., Saperstein, R., Brady, E., Huskey, S., Shen, X., Liu, K., Xu, L., Salituro, G. M., Heck, J. V., Moller, D. E., Jones, A. B., and Zhang, B. B. (2000) J Biol Chem, Endemann, G., Yonezawa, K., and Roth, R. A. (1990) J Biol Chem 265, 396-400 Levy-Toledano, R., Taouis, M., Blaettler, D. H., Gorden, P., and Taylor, S. I. (1994) J Biol Chem 269, 31178-31182 Franke, T. F., Yang, S. I., Chan, T. O., Datta, K., Kazlauskas, A., Morrison, D. K., Kaplan, D. R., and Tsichlis, P. N. (1995) Cell 81(5), 727-36 Cross, D. A. E., Alessi, D. R., Vandenheede, J. R., McDowell, H. E., Hundal, H. S., and Cohen, P. (1994) BiochemJ. 303, 21-26 Shepherd, P. R., Nave, B. T., and Siddle, K. (1995) Biochem J 305, 2528 Ullrich, A., and Schlessinger, J. (1990) Cell 61,203-212 Hubbard, S. R. (1999) Prog Biophys Mol Biol 71(3-4), 343-58 Levitzki, A., and Gazit, A. (1995) Science 267(5205), 1782-8

236 [38]

[39] [40]

[41] [42] [43] [44] [45] [46] [47]

[48]

[49]

[50] [51] [52]

[53]

Fry, D. W., Kraker, A. J., McMichael, A., Ambroso, L. A., Nelson, J. M., Leopold, W. R., Connors, R. W., and Bridges, A. J. (1994) Science 265(5175), 1093-5 Fry, D. W., and Bridges, A. J. (1995) Curr Opin Biotechnol 6(6), 662-7 Mohammadi, M., Froum, S., Hamby, J. M., Schroeder, M. C , Panek, R. L., Lu, G. H., Eliseenkova, A. V., Green, D., Schlessinger, J., and Hubbard, S. R. (1998) Embo J 17(20), 5896-904 Mohammadi, M., Schlessinger, J., and Hubbard, S. R. (1996) Cell 86(4), 577-87 Mohammadi, M., McMahon, G., Sun, L., Tang, C , Hirth, P., Yeh, B. K., Hubbard, S. R., and Schlessinger, J. (1997) Science 276(5314), 955-60 Zhang, B., and Roth, R. A. (1992) J Biol Chem 267(26), 18320-8 Hubbard, S. R., Wei, L., Ellis, L., and Hendrickson, W. A. (1994) Nature 372,746-754 Hubbard, S. R. (1997) Embo J 16(18), 5572-81 Luo, R. Z., Beniac, D. R., Fernandes, A., Yip, C. C., and Ottensmeyer, F. P. (1999) Science 285(5430), 1077-80 Livnah, O., Stura, E. A., Johnson, D. L., Middleton, S. A., Mulcahy, L. S., Wrighton, N. C., Dower, W. J., Jolliffe, L. K., and Wilson, I. A. (1996) Science 273(5274), 464-71 Johnson, D. L., Farrell, F. X., Barbone, F. P., McMahon, F. J., Tullai, J., Hoey, K., Livnah, O.^ Wrighton, N. C., Middleton, S. A., Loughney, D. A., Stura, E. A., Dower, W. J., Mulcahy, L. S., Wilson, I. A., and Jolliffe, L. K. (1998) Biochemistry 37(11), 3699-710 Tian, S. S., Lamb, P., King, A. G., Miller, S. G., Kessler, L., Luengo, J. I., Averill, L., Johnson, R. K., Gleason, J. G., Pelus, L. M., Dillon, S. B., and Rosen, J. (1998) Science 281(5374), 257-9 Drake, P. G., and Posner, B. I. (1998) Mol Cell Biochem 182(1-2), 79-89. Goldfme, A. B., Simonson, D. C , Folli, F., Patti, M. E., and Kahn, C. R. (1995) Mol Cell Biochem 153(1-2), 217-31 Elchebly, M., Payette, P., Michaliszyn, E., Cromlish, W., Collins, S., Loy, A. L., Normandin, D., Cheng, A., Himms-Hagen, J., Chan, C. C , Ramachandran, C , Gresser, M. J., Tremblay, M. L., and Kennedy, B. P. (1999) Science 283(5407), 1544-8 Klaman, L. D., Boss, O., Peroni, O. D., Kim, J. K., Martino, J. L., Zabolotny, J. M., Moghal, N., Lubkin, M., Kim, Y. B., Sharpe, A. H., Stricker-Krongrad, A., Shulman, G. I., Neel, B. G., and Kahn, B. B. (2000) Mol Cell Biol 20(15), 5479-89

H. van der Goot (Editor) Trends in Drug Research III © 2002 Elsevier Science B.V. All rights reserved

237

Expression Databases for Pharmaceutical Lead Optimisation L. Mike Fumess, Incyte Genomics Ltd, 100 Hills Road, Cambridge. CB2 IFF. UK mfumess @ incyte.com +441223 454900

In the past few years, expression technologies have moved to the forefront of genomic research. These technologies are now also being rapidly adopted by pharmaceutical and biotechnology companies, primarily for use in target discovery, but more recently in Lead Optimisation. This has seen the growth of areas such as pharmacogenomics (the changes in gene expression patterns seen when biological systems are treated with pharmacological agents) and toxicogenomics (the changes seen with toxins).

If we look first at the process of drug discovery and development as it currently stands, it is a long, complicated and expensive process, as can be seen below in Figure 1.

Figure 1 - The Drug Discovery and Development Process Compounds

Preclinical Testing (Lab + Animal Tests) Phase I (20-80 healthy volunteers- Safety + Dosage)

• •

Phase II (100-300 volunteers- Efficacy + Side Effects) I Phase m I (1000-3000 volunteers- ADR + Long Term) HHH|Food & Drug Admin. B ^ M Review + Approval Years

1^ 10

12

16

Adapted from PhRMA Report 1998

This is one representation of the process, outlining the initial identification of small numbers of drug targets from the whole genetic complement of an organism, and then following the identification and optimisation of compounds which act therapeutically against these targets. In this view of the process, we can get a rough idea of where the major fallout in the process occurs in discovery. By either more rapidly reducing the

238

number of compounds that fail in the process, or minimising the number of compounds likely to fail later in the process, we can hopefully reduce costs from clinical failures, by identifying adverse effects earlier in the process, and at the same time, increase knowledge of the mechanism of action of a drug. To put this in context, it has been estimated that it costs around $500 million to bring a drug to market (1). DiMasi et al, estimate that 70% of the costs of drug development are associated with failed compounds, and that a 2% increase in those that succeed in passing on from phase I to the clinic could offer a 10% saving in drug development costs (1,2).

If we look further into the time component of the process, there are two main issues exclusivity time and patent lifetime. Exclusivity has been an issue that is now becoming more prevalent. When the beta blocker Inderal was introduced in 1968, it took 10 years before its competition reached the market in the guise of Lopressor. However, if we look at some of more recent drugs like Invirase, competition was on the market in 6 months, reducing exclusivity time, and therefore market share and revenues. If we look at some recent figures on prescription sales in the US alone, we can see that an average drug makes over $lm per day, and in the case of drugs such as Prilosec, a proton pump inhibitor used in the treatment of ulcers, this rises up to $ 11m per day, so even a few months extra exclusivity can make a huge financial difference in the returns for a company(3).

If we now look at where the compounds filed as NCEs fail during development, the 3 main categories are lack of efficacy, animal toxicity, and human toxicity, which together make up almost 75% of the total. In the latter 2 cases, we have failure due to toxicity in animals, which may or may not reflect toxicity in humans, as best estimates put the accuracy of these models at 70% (4), and adverse effects in man, which is reflected in this 70% success rate.

So, any system or systems we can use to optimise the understanding of a drugs functions, both efficacious and adverse, should allow us a greater success rate in passing drugs further through the screening process.

Figure 2 shows a schematic of the goal we are working towards. The ideal would be to screen a group of compounds of known efficacies and/or toxicities, in a 'reference' cell and/or tissue. The mRNA, representing all the transcribed genes from a sample, would then be hybridised to one or more microarrays to identify which genes are differentially regulated between compounds, or compound classes. This data could then be used to construct a central core of reference data which can be used to identify gene expression patterns associated with specific phenotypes (e.g. hepatotoxicity, beta adrenergic inhibition, etc). New compounds that have been identified from high throughput screening could then be run against the reference cells/tissues, and across microarrays. Once run, the data generated could be compared to the reference data set to see which compounds it shows most similarity to, and which phenotypes it would appear to be

239 likely to be express. In the short term, this would most likely be used to screen out compounds with a high likelihood of showing serious toxicities, but longer term it may help understand the mode of action of drugs in detail, and even identify new therapeutic apphcations of compounds.

In this brief review, microarray technology will not be discussed, as there are a number of reviews already available on the subject (5-11). However, one key issue is to know which genes are the most relevant or important for any particular application, whether that is looking for novel drug targets, identifying disease-specific genes, or understanding mechanisms of toxicity. Many academic and commercial sources generate arrays with a specific focus, such as 'brain arrays' or 'toxicology arrays'. These have usually been generated by using DNA isolated from a specific tissue, or from historic literature data on which genes play a key role in certain cell functions. The limitation with this approach is that until recently, we only knew a tiny fraction of the genes expressed in humans and other species. Even today, while there are more accurate estimates of how many genes are present in several species, only 10-30% of these have any form of functional annotation.

240

The X-axis represents the drugs treatments, showing left to right carboxymethylcellulose, clofibrate, fenofibrate, and gemfibrozil, by dose and then by time. The y-axis lists different gene transcripts.drugs

Our approach to this problem has been to try and represent every gene transcript on microarrays. For example, in humans there are -60-80,000 gene transcripts represented in the public databases, and a further -60,000 in Incyte's proprietary databases. We have used these extensive sequence databases, primarily generated from human, rat and mouse samples, to identify representative, unique sequences of DNA covering each transcript. These are then being arrayed on glass slides, using fully sequence verified reagents.

But what about the most important components - the biology and chemistry used to generate the samples put on to the microarrays ? We have begun to address this by using common in vivo (rat) and in vitro (HepG2 C3A cells, Amphioxus Cell Technologies) model systems. As far as a standard protocol design, discussions with a wide variety of academic and industrial toxicologists, drug metabolism groups, and discovery biologists, there was very little consensus as to what the 'ideal' protocol should look like. The model we have used to date is exactly that ~ a model. We started with a compromise between detailed studies on a few compounds, and superficial analysis over a large number of compounds, to try to allow us to empirically determine what an 'ideal' protocol should look like.

241 To this end, for the in vivo work, the model we used included 2 male and 2 female rats for each drug at 12 different conditions (4 doses and 3 times). Doses were based on literature values for Maximum Tolerated Dose, and scaled down appropriately, based on some initial dosing studies. Times were based on a compromise between practicality and some historical data run at Incyte. We kept the times short, which in many cases means we may not see significant pathological changes during the study, but we do see an enormous amount of gene changes. Blood chemistry was collected to give extra indications of changes between individual animals, as with these small groups, the statistical value would be limited.

Figure 3 shows an example of a comparison data set generated by comparing liver from vehicle (carboxymethylcellulose) dosed animals against those from 3 different fibrates clofibrate, fenofibrate and gemfibrozil. In this, and all future illustrations, the intensity of the green represents the increase in expression of a gene transcript in the compoundtreated versus vehicle, and for red represents relative decrease in expression. The data shown in these analyses was cropped electronically, to leave only data that showed at least a 2-fold increase or decrease in gene expression in at least one condition, and run through k-means clustering. In this instance, the resulting data table was then coloured using a macro in Microsoft Excel.

Two main downregulated clusters are seen, which comprise 107 genes, of which, 55 are not seen in the public sequence databases, and 20-30 of the remainder have no known function assigned to them. In general though, clofibrate and fenofibrate have a greater impact on down regulation of genes than gemfibrozil that generally has little effect. In addition, the fenofibrate response decreases over the 72 hour time course, whereas clofibrate increases. In addition, there are 5 main upregulated clusters seen. In total, these comprise 119 genes, of which 14 were Incyte-unique, and a further 34 are of unknown ftinction. Even by eye, you can begin to see patterns, such as increases in effect with increasing dose in some clusters, and the presence and/or absence of certain clusters between different drugs. Some of the key known genes identified in these clusters include UDP glucoronyl transferase, cholesterol 7 a hydroxylase, 2,3 oxidosqualene cyclase, aldehyde dehydrogenase, peroxisomal membrane protein PMP70, peroxisomal long chain acyl CoA thiolase, carnitine octanoyl transferase, P450IVB2. and a number of other P450s.We can also see patterns of genes that change in an almost exclusively gender-dependent fashion near the bottom of the figure. In addition to simple gender differences, there are examples of genes where the expression differences in response to compounds are clearly stronger in one gender.

As for the in vitro component of the work, there has been a lot of discussion as to whether a cell line should be mechanistically relevant to the compound being tested. If one was to do so, and say, test a cardiovascular drug in a cardiomyocyte, and test a neurological drug in a neuronal cell, while you gain more relevant mechanistic information, you have complications in comparing the data sets with each other, as you

242 have variability from both the variations in pharmacological conditions and the variation between cell types.

Figure 4 -- In vitro microarray data from compound treatments iN

2,263 microarray hybridisations were included, representing around 29 different treatments, each with multiple time and dose points, run over up to 6 different microarrays. The data was generated as a text file which was then imported into Spotfire® DecisionSite 6.0 for some further analyses. This particular image was generated using the HeatMap tool in, ordering the genes by microarray type and then by cluster number. The pale grey areas represent data points that are absent. The Y-axis corresponds to approximately 7,500 different gene transcripts spread across 6 different microarray types (l=Human Drug Target, 2=Human Foundation 1, 3=Human Foundation 2, 3=Human Foundation 4, 5=Human Foundation 5, and 6=Human Unigene microarray, all generated by Incyte Genomics).

We took the approach that for a reference data set, the cell line should be both stable and reproducible over time, which though it sounds a trivial task, turns out not to be so.A clonal HepG2-derived cell line, called HepG2 C3A (Amphioxus Cell

243

Technologies, Houston,Tx.) was used as the 'reference' cell line. This cell line has several advantages over most cell lines in that it is clonal, so we have a homogenous cell population. The cells are also grown in such a way that they retain a number of key functions of adult liver cells including metabolic functions, secretion of proteins seen in adult liver cells, and P450 inducibility. The cells are grown in a serum-free medium to minimize any batch-to-batch variation, and reach a mature state 5-7 days from initial passage. Initial studies comparing gene expression changes between 7 day old, and 14 day old cells (the window during which cells were dosed) showed very little difference, indicating that any changes we see in compound-treated cells should be primarily from the effects of the compound on the cells, and not from the cells themselves.

As with the rat studies, all the C3A data was electronically cropped to generate a file with compound-treated C3A microarray data that had shown a greater than 2-fold change up or down, when compared to controls (Figure 4). More details of the genes on the microarrays can be found at http://www.incyte.com.

If we just look at the data for the endocrine compounds on the Human Drug Target microarray (Incyte Genomics Inc.), we can take the default k-means clustered data, and re-order them by fold-change, we can readily identify smaller sets of genes which appear to correlate with endocrine-related compounds, and not with other compounds used in the study. When we take the top 100 upregulated and 100 downregulated genes from these endocrine-related data sets, we can do a direct comparison between all the compounds (Figure 5). Interestingly, 40 of the top 45 most changed genes are all novel genes with no annotation associated.

We can do a similar analysis using the Distinction Calculation feature in the Spotfire® Array Explorer 3.0 software. The basic premise is to choose two data sets that we wish to define

244

Figure 5 - Endocrine-related gene expression patterns Ref

W777i

^

Ref

Y///y/yyy//////yyyy/yy/yy/m The *Ref lane is the compound against which all the expression data was re-ordered, and the bars at the bottom of each table identify which compounds fall within the 'endocrine' (white) and 'non-endocrine' (hatched) groups. differences between, and then to calculate which genes are most differently regulated between these two groups. Using this tool, each set of compound data was compared to all other compounds, to generate distinction measurements that show which gene changes most separate each drug from the others in this study (Figure 6). From this we can see that, although the majority of endocrine samples do look similar by eye, there are discrete compound-specific differences seen, even on just one microarray, between all the compounds.

245

Figure 6 - comparison of compounds by distinction correlation

1

3

9

11

13 15

17

19

2 1 23

25

HeatMap showing the comparison of individual compound distinction coefficients. The distinction coefficients were generated by comparing all data from one compounds versus all other data, over the -2,400 genes on the Human Drug Target microarray. The figure shows the genes ordered by the distinction values for the compound on the farthest left (budesonide in this case).

l=beclomethasone, 2=betamethasone, 3=budesonide, 4=danazol, 5=dexamethasone, 6=medroxyprogesterone, 7=inifepristone, 8=prednisone, 9=progesterone, 10=acetaminophen, ll=catechol(all), 12=catechol(+CMC), 15=clofibrate(+water), 14=clofibrate(all), 13=catechol(+water) 18=fenofibrate(all), 17=fenofibrate(+water), 16=clofibrate(+CMC), 21 =gemfibrozil(+water), 20=gemfibrozil(all), 19=fenofibrate(+CMC), alpha, 24=LY294002, 25=3 22=gemfibrozil(+CMC), 23=interferon methylcholanthrene.

While these tools are giving us the ability to rapidly generate enormous amounts of data, the biggest issue is still how we handle data on this scale. Currently, enormous efforts are being made in the field of bioinformatics to try and find ways to make sense of this data. As we begin to add in physiological parameters and pathology data, and then link

246 the data with chemical data and clinical data, we have a great task lying before us. To quote from T.S.Elliot, " Where is the knowledge we lost in information" (from The Rocks'). Hopefully in the next couple of years, we can begin to identify some of the knowledge in the information we all have available to us.

Acknowledgements

Thanks to Kenny Pollock for the extensive analyses of the in vivo data, Paul Laub for his help with the data cropping and preliminary analysis tools, and the many other staff at Incyte Genomics who made this work possible. Finally, a special thanks to Kate, Elizabeth and Charlotte, who put up with my evenings at home working on this article.

References [1] DiMasi,J.A., et al. Cost of innovation in the pharmaceutical industry. J. Health Econ. 10(1991)107.

[2] DiMasi,J.A., et al. Research and development costs for new drugs by therapeutic category: a study of the US pharmaceutical industry. Pharmacoeconomics 7 (1995) 152.

[3] Getz,K.A., et al. Breaking the development speed barrier: assessing successful practices of the fastest drug developing companies. DIJ 34 (2000) 725.

[4] 01son,H., et al. Concordance of the toxicity of pharmaceuticals in humans and in animals. Regul. Toxicol. Pharmacol. 32 (2000) 56.

[5] Brown,P., et al. Exploring the new world of the genome with DNA microarrays. Nature Genet. 21 suppl:33 (1999) 30.

[6] Evertsz,E., et al. Technology and Applications of Gene Expression Microarrays. pl49-166. In M. Schena (Ed.), Microarray Biochip Technology. (BioTechniques Books Division, Natick, MA)

247 [7] Fumess,L.M., et al. Expression databases - resources for pharmacogenomic R&D. Pharmacogenomics 1 (2000) 281.

[8] Kane,M.D., et al. Assessment of the sensitivity and specificity of oligonucleotide (50mer) microarrays. Nuc. Acids Res. 28 (2000) 4552.

[9] RinningerJ.A., et al. Differential gene expression technologies for identifying surrogate markers of drug efficacy and toxicity. DDT 5 (2000) 560.

[10] Schena,M., et al. Microarrays: biotechnology's discovery platform for functional genomics. TIBTECH 16 (1999) 301.

[11] Yue,H., et al. An evaluation of the performance of cDNA microarrays for detecting changes in global mRNA expression. Nuc. Acids Res. 29 (2(X)1) e41.

This Page Intentionally Left Blank

H. van der Goot (Editor) Trends in Drug Research III © 2002 Elsevier Science B.V. All rights reserved

249

New Developments in the Pharmacology of Cannabinoids Roger G. Pertwee Department of Biomedical Sciences, Institute of Medical Sciences, University of Aberdeen, Aberdeen AB25 2ZD, Scotland, UK.

1. Introduction Cannabis sativa is the source of a set of more than sixty oxygen-containing aromatic hydrocarbon compounds called cannabinoids and one of these, A^tetrahydrocannabinol (A^-THC), is the mam psychotropic constituent of this plant [1]. A^-THC is also of interest because it is one of just two cannabinoids to be licensed fcr medical use. Thus, as the oral preparation dronabinol (Marinol), it is available in the USA for the suppression of nausea and vomiting provoked by anticancer drugs and for the reversal, through appetite stimulation, of body weight loss experienced by AIDS patients. The other cannabinoid that it is permissible to use clinically is nabilone (Cesamet), a synthetic analogue of A^-THC that is also given by mouth and that is licensed for use in the UK, again to suppress nausea and vomiting produced by cancer chemotherapy. Because cannabinoids have high lipid solubility and low water solubility they were long thought to owe their pharmacological properties to an ability to perturb the phospholipid constituents of biological membranes. However, all this changed in the late 1980's with the discovery of specific cannabinoid receptors. The existence of endogenous ligands for these receptors ("endocannabinoids") has also been demonstrated, suggesting that cannabinoid receptors have physiological as well as pharmacological significance. Cannabinoid receptors and their endogenous ligands constitute the "endocannabinoid system". 2. The endocannabinoid system Mammalian tissues contain at least two types of cannabinoid receptors, CBi and CB2 (see references 2 and 3 for reviews). Both receptor types are coupled through Gi/o proteins, negatively to adenylate cyclase and positively to MAP kinase. CBi receptors are also coupled through Gi/o proteins to ion channels, positively to A-type and inwardly rectifying potassium channels and negatively to N-type and P/Q type calcium channels and to D-type potassium channels. There is also evidence that CBi receptors are negatively coupled to M-type in rat hippocampal CAl pyramidal neurones and to voltage gated L-type calcium channels in cat cerebral arterial smooth muscle cells. Experiments with CBi and CB2-transfected cells have revealed other signalling mechanisms for cannabinoid receptors [3]. However, the physiological significance of these remains to be established. CBi receptors are present in the central nervous system and also in some peripheral tissues including pituitary gland, immune cells, reproductive tissues, gastromtestinal tissues, sympathetic ganglia, heart, lung, urinary bladder and adrenal gland [2]. Many

250 CBi receptors are to be found at central and peripheral nerve terminals and an important function of these receptors is to suppress the release of a range of excitatory and inhibitory neurotransmitters [3]. For example, CBi receptors are known to mediate inhibition of evoked release of acetylcholine from neurones in rat hippocampal slices and guinea-pig intestinal tissue and of dopamine in rat striatal slices and guinea-pig retinal discs. CBi receptors have also been found to mediate inhibition of evoked release of noradrenaline, 5-hydroxytryptamine, y-ammobutyric acid, glutamate and aspartate in various brain areas or in the peripheral nervous system. Much less is known about the role of CB2 receptors although it is very likely that this includes immunomodulation as CB2 receptors are expressed mainly by immune cells, particularly B-cells and natural killer cells [2, 3]. One important role of CB2 receptors may be to regulate cytokine release in health or disease [3, 4]. If this is true, then a common property of CBi and CB2 receptors would be the ability to modulate ongoing release of various chemical messengers, CBi receptors from neurones and CB2 receptors from immune cells. Within the bram, the distribution of CBi receptors is heterogeneous, brain areas that express this receptor type including the cerebral cortex, hippocampus, caudate-putamen, substantia nigra pars reticulata, globus pallidus, entopeduncular nucleus, cerebellum, periaqueductal grey, rostral ventromedial medulla, superior coUiculus and certain nuclei of the thalamus and amygdala [2, 3, 5]. This distribution pattern accounts for several prominent pharmacological properties of CBi receptor agonists, for example their ability to impair cognition and memory and to alter the control of motor function. It also accounts for the ability of these agonists to produce analgesia in humans and antinociception in anunal models both of acute pam and of tonic pain induced by nerve damage or by the injection of an inflammatory agent. More specifically, as detailed elsewhere [3], CBi receptors that mediate the analgesic/antinociceptive effects of cannabmoids seem to be located not only in the brain but also on the terminals of neurones that project from the brain stem to the spinal cord and/or on intrmsic spinal neurones. There are also CBi receptors at the central and peripheral terminals of primary afferent neurones, both on C-fibres and on larger diameter Ap/A5-fibres. The presence of significant numbers of CBi receptors on these larger diameter primary afferent fibres helps to explain the eflBcacy shown by CBi receptor agonists against signs of neuropathic pain in animals since this kind of pain is thought to be elicited in part by abnormal spontaneous discharges of myelinated AjJ- and A8-fibres. CB2 receptors, and possibly other types of cannabmoid receptors yet to be characterized, may also contribute to the analgesic/antinociceptive effects of cannabinoids [3]. The most important endocannabinoids so fer identified are arachidonoylethanolamide (anandamide), 2-arachidonoyl glycerol (2-AG) [6-9] and probably also arachidonoyl glyceryl ether (noladin ether), the discovery of which has only just been announced [10]. Of these, anandamide behaves as a partial cannabinoid receptor agonist with margmally greater CBi than CB2 affinity but much less CB2 than CBi eflBcacy [11]. The pharmacological properties of 2-arachidonoyl glycerol and arachidonoyl glyceryl ether have been less well characterized. The available data suggests that both are cannabinoid receptor agonists, that the affinity of 2-AG for CBi and CB2 receptors is similar to that of anandamide and that noladin ether has significantly higher affinity for CBi receptors (Ki= 21.2 nM) than for CB2 receptors (Ki > 3 ]xM) [10, 11]. Anandamide

251 and 2-AG both serve as neurotransmitters or neuromodulators as there is evidence that they are synthesized by neurones ("on demand"), that they can undergo depolarizationinduced release from neurones and that once released they are rapidly removed from the extracellular space by a membrane transport process yet to be ftilly characterized [7, 1215]. Once within the cell, anandamide is thought to be hydrolysed to arachidonic acid and ethanolamine by the microsomal enzyme, fatty acid amide hydrolase (FAAH) [7, 14, 16]. 2-AG can also be hydrolysed enzymically, both by FAAH and by intracellular lipases [7, 17]. Mechanisms underlying the release and fate of noladin ether remain to be identified. There is firm evidence that anandamide activates not only cannabinoid receptors but also vanilloid VRl receptors, a property not shared by non-eicosanoid CBi or CB2 receptor agonists. The existence of an SR144528-sensitive non-CB2 cannabinoid receptor ('CB2-like' receptor) has also been proposed [18]. The evidence for this receptor type is based on the observation that even though pahnitylethanolamide lacks significant affinity for CBi or CB2 receptors [11, 19], its ability to produce signs of antihyperalgesia m the mouse formalin paw test is readily attenuated by the CB2-selective antagonist/inverse agonist, SR144528 but not by the CBi-selective antagonist/inverse agonist, SR141716A [18]. The existence of CB2-like receptors in the mouse vas deferens has also been proposed [20]. There is also evidence for the presence in vascular endothelium of an SR141716A-sensitive non-CBi, non-CB2, non-vanilloid receptor that is unresponsive to established non-eicosanoid CB1/CB2 receptor agonists but that can be activated both by the eicosanoid cannabinoids, anandamide and methanandamide, and by certain classical cannabinoids that do not act through CBi or vanilloid receptors ("abnormal cannabidiol" and its more potent analogue, 0-1602) [21, 22]. Interestingly, two effects of abnormal cannabidiol, hypotension and mesenteric vasodilation, were found to be antagonized by the non-psychotropic classical cannabinoid, cannabidiol. So too was anandamide-induced mesenteric vasodilation. Finally, the existence in the brain of non-CBi, non-CB2, SR141716A-insensitive G protein-coupled receptors for anandamide has recently been proposed to explain results obtained from experiments with CBi knockout [23]. 3. Inhibitors of endocannabinoid membrane transport or enzymic hydrolysis The discovery that the actions of anandamide and 2-AG are terminated by tissue uptake and intracellular enzymic hydrolysis has led to the development of inhibitors of these processes. The first of these to have been developed is A^-(4-hydroxyphenyl) arachidonylamide (AM404). When administered to rats by itself, AM404 increases plasma levels of anandamide and shares the ability of this endocannabinoid to decrease locomotor activity, depress plasma levels of prolactin and alter tyrosine hydroxylase activity in the hypothalamus (increase) and substantia nigra (decrease) [24-26]. The inhibitory effect of AM404 on locomotor activity is susceptible to antagonism by SR141716A [24, 25]. AM404 does not, however, elicit two other typical responses to CBi receptor agonists in rats: catalepsy and signs of analgesia in the hot plate test [24]. At concentrations at which it inhibits the membrane transport of endocannabmoids, AM404 bmds both to CBi receptors and to vanilloid VRl receptors. However, whilst it is known to activate vanilloid receptors, there are no reports that AM404 behaves as a

252 CBi receptor agonist or antagonist. A recently developed analogue of AM404, VDM11, retains the ability to inhibit endocannabinoid membrane transport but shows markedly less eflBcacy than AM404 as a vanilloid receptor agonist [27]. However, like AM404, VDM-11 does bind to CBi receptors at inhibitory concentrations [27]. The compound that has been most widely used to inhibit the enzymic hydrolysis of endocannabinoids in non-clmical experiments is the non-specific serine protease inhibitor, phenyhnethylsulphonyl fluoride. However, inhibitors with much greater potency are now available. Among these are two irreversible inhibitors of FAAH, palmitylsulphonyl fluoride (AM374) and stearylsulphonyl fluoride (AM381). Both of these inhibitors show good separation between potency for FAAH inhibition and ability to bind to CBi receptors [28]. AM374 potentiates both anandamide-induced inhibition of evoked [^H]acetylcholine release in rat hippocampal slices [29] and anandamideinduced suppression of rat operant lever pressing and open field locomotor activity [30]. Even more potent inhibitors of FAAH are to be found in a series of a-keto bicyclic heterocycles with alkyl or phenylaUcyl side chains. These inhibit the enzyme competitively, some with Ki values m the picomolar or low nanomolar range [31]. Other pharmacological properties of these inhibitors, for example their ability to interact with cannabmoid or vanilloid receptors or to potentiate endocannabinoids have yet to be reported. 4. Ligands for CBi and CBi receptors There are several established cannabinoid receptor agonists that bind more or less equally well to CBi and CB2 receptors, the best known examples being A^-THC, the Pfizer compound, CP55940 and the Sterlmg Winthrop compound, WIN55212-2, which has only marginally greater CB2 than CBi affinity [11, 32]. However, since the discovery of cannabinoid receptors, agonists with significant selectivity for CBi or CB2 receptors have emerged, hnportant CBi-selective agonists include the anandamide analogues, methanandamide, 0-689, arachidonyl-2'-chloroethylamide (ACEA) and arachidonylcyclopropylamide (ACPA) [11, 33]. Of these both ACEA and ACPA share the susceptibility of anandamide to enzymic hydrolysis whilst methanandamide and O689 are metabolically more stable than anandamide. This is presumably because methanandamide and 0-689 are protectedfromenzymic hydrolysis by the presence of a methyl substituent on the T or 2 carbon whilst anandamide, ACPA and ACEA are not. In line with this hypothesis, it was recently shown that the addition of a methyl group to the 1' carbon of ACEA markedly decreases the susceptibility of this molecule to FAAH-mediated hydrolysis [34]. This structural change also reduces the affinity of ACEA for CBi receptors by about 14-fold. The best CBi-selective agonists to have been developed so fer are all structural analogues of THC. They include L-759633, L759656, JWH-133 and HU-308 [32, 35, 36]. The discovery of cannabinoid receptors also prompted a search for selective CBi and CB2 receptor antagonists. The most promising compounds so fer developed are both Sanofi compounds. These are the CBi-selective SR141716A and the CB2-selective SR144528 [11, 37, 38]. There is convincing evidence, however, that both these compounds are not "silent" antagonists. Thus, as well as attenuating effects of CBi or CB2 receptor agonists, both agents can by themselves elicit responses in some

253 cannabinoid receptor-containing tissues that are opposite in direction from those elicited by CBi or CB2 receptor agonists. Whilst some of these "inverse cannabimimetic effects" may be attributable to a direct antagonism of responses elicited at cannabinoid receptors by released endocannabinoids, there is evidence that this is not the only possible mechanism and that SR141716A and SR144528 are in fact both inverse agonists [36, 38-41]. Thus both agents may produce inverse cannabimimetic effects in at least some tissues by somehow reducing the constitutive activity of cannabinoid receptors (the coupling of these receptors to their effector mechanisms that it is thought can occur in the absence of exogenously added or endogenously produced agonists). Two cannabinoid receptor ligands that are closer to being silent cannabinoid receptor antagonists at CBi and/or CB2 receptors are 6'-azidohex-2'-yne-A*-THC (0-1184) and 6-iodopravadoline (AM630) [42, 43]. 0-1184 behaves as a high-affinity low-eflBcacy agonist at CBi receptors and as a high-affinity low-efficacy inverse agonist at CB2 receptors. AM630 is a potent CBa-selective antagonist/inverse agonist which resembles 0-1184 appears in having less inverse efficacy at CB2 receptors than SR144528. AM630 also interacts with CBi receptors, albeit with significantly less potency. Results from several investigations when taken together suggest that AM630 has mixed agonistantagonist properties and that it is a low-affinity partial CBi agonist [11, 36, 44-46]. There is also one report that it can behave as a low-potency inverse agonist at CBi receptors [47]. 5. The therapeutic potential of cannabinoids Several potential therapeutic applications have been suggested for CBi receptor antagonists/inverse agonists [48]. These include appetite suppression [49-51], the reduction of L-Dopa-induced dyskinesia in patients with Parkinson's disease [52], the management acute schizophrenia [53] and the amelioration of cognitive/memory dysfunctions associated with disorders such as Alzheimer's disease [54]. The prospect of exploiting CBi receptor antagonists/inverse agonists for clinical purposes remains particularly attractive to the pharmaceutical industry as such agents do not produce the unwanted central effects for which CBi receptor agonists are so renowned. Even so, there is growing evidence that in addition to their recognized uses in the clmic as appetite stimulants and anti-emetics, CBi receptor agonists may have therapeutic potential as neuroprotective agents through CBi-mediated inhibition of glutamate release [55], as anticancer agents [56, 57] and for the management of glaucoma [58], pain [3, 59] and various kinds of motor dysfunction that include the muscle spasticity/spasm/tremor associated with multiple sclerosis and spinal cord injury, the tics and psychiatric signs and symptoms of Tourette's syndrome and the dyskinesia that is produced by L-Dopa in patients with Parkinson's disease [60-64]. Of these potential therapeutic targets for CBi receptor agonists, pain and motor disorders associated with multiple sclerosis and spinal cord injury, are currently attractmg particular attention. Some cannabinoids that do not act through CBi or CB2 receptors also have pharmacological properties that may come to be exploited in the clinic. One of these is (+)-ll-hydroxy-A^-THC-dimethylheptyl(HU-211, dexanabinol) which, as discussed in greater detail elsewhere [65], shows potential as a neuroprotective agent and also for the management of neuropathic pain and septic shock. The psychotropically inactive plant

254 cannabinoid, cannabidiol, also has neuroprotective activity [66]. Another cannabinoid that merits special mention is r,r-dimethylheptyl-A^-THC11-oic acid (DMH-A^-THC-11-oic acid). This is a cannabinoid receptor ligand with therapeutic potential as an anti-inflammatory/analgesic [3, 67]. In spite of its ability to interact with CBi and CB2 receptors [68], it remains possible that DMH-A^-THC-11oic acid produces its anti-inflammatory effects by suppressing eicosanoid synthesis in inflamed tissue through inhibition of inducible cyclooxygenase (COX-2), an enzyme that is thought to be activated during inflammation and to be responsible for the production of eicosanoids that act as inflammatory mediators [69]. DMH-A*-THC-11oic acid is more potent as an inhibitor of COX-2 than of COX-1 (constitutive cyclooxygenase), raising the possibility that it may be able to relieve inflammation without producing gastrointestinal or kidney toxicity [67, 69, 70]. It is unwanted effects of this kind that limit the clinical usefiibiess of less selective non-steroidal antiinflammatory drugs that inhibit COX-1 more or less as effectively as they inhibit COX2. 6. Future Clinical Research One challenge for future clinical research is to develop cannabinoid formulations and modes of administration that produce more reliable cannabinoid absorption than has so far been realisable, at least by the oral route. Possible solutions are to develop improved oral formulations or to use other routes for cannabinoid delivery, for example administration by rectal suppository [71], by aerosol/vapour inhalation, by injection, by skin patch or by the sublingual or intrathecal route, all modes of administration that avoid first-pass metabolism of the absorbed drug. Some success has ateady been achieved by GW Pharmaceuticals in phase I studies with a sub-lingual cannabinoid spray (personal communication from Dr. Geoffrey Guy). The emergence of better modes of cannabinoid administration should be facilitated by the development of a centrallyactive water-soluble cannabmoid [72]. The availability of better delivery systems fir cannabinoids should facilitate the gathering of more conclusive clinical data both about the efficacy of cannabinoids and about then- unwanted effects than has hitherto been achievable. Another important area for future clmical research must be the development of strategies that maxunize separation between the sought-after therapeutic effects of CBi receptor agonists and the unwanted effects of these drugs, particularly their psychotropic effects. One strategy may be to use agents that activate the endogenous cannabinoid system indirectly by increasing extracellular levels of endocannabinoids through inhibition of their membrane transport or enzymic hydrolysis. This approach is based on the expectation that drugs with this kind of action would be more selective than direct CBi receptor agonists. This is because they are unlikely to affect all parts of the endocannabinoid system at one time, producing instead effects only at sites where ongoing production of endocannabinoids is taking place. The success of this strategy depends on whether endogenous cannabinoids are released to a greater extent at sites at which they produce sought-after effects than at sites at which they provoke unwanted effects; In line with this possibility is evidence obtained from experiments using an autoimmune model of multiple sclerosis that is set up by mjecting Biozzi ABH mice

255 subcutaneously with an emulsion of mouse spinal cord homogenate in Freund's complete adjuvant. These showed that inhibitors of endocannabinoid membrane transport (AM404) or enzymic hydrolysis (AM374) share the ability of direct cannabinoid receptor agonists to ameliorate spasticity in CREAE mice and that spastic CREAE mice have elevated concentrations of the endocannabinoids anandamide and 2arachidonoyl glycerol in their brains and spinal cords [73]. There is also evidence fix)m animal experiments that peripheral inflammatory pain triggers increased release of anandamide in at least one area of the brain at which cannabinoids can induce signs of analgesia in animals, the periaqueductal grey area of the midbrain [74]. Because the unwanted central effects of cannabmoids are probably mediated largely or entirely by CBi receptors within the brain, a second strategy for reducing the unwanted effects of a CBi receptor agonist, at least for pain relief, might be to inject a CBi receptor agonist directly into the spinal cord. This approach exploits the presence of CBi receptors on nociceptive neurones within the spinal cord and also takes account of the ability of CBi receptor agonists to induce antinociception in anunals when these agents are administered intrathecally. The same strategy is sometimes adopted to reduce the incidence of adverse responses of multiple sclerosis patients to the anti-spasticity agent, baclofen. Yet another strategy would be to administer a cannabinoid in combination with a second agent that augments only the sought-after effects of the cannabinoid. Indeed, there is evidence fi-om animal experiments that synergistic interactions can occur between cannabinoids and opioids for analgesia [3, 75] and between cannabinoids and benzodiazepines for depressant effects on motor function [76, 77]. As there are claims by multiple sclerosis patients that cannabis can relieve their symptoms at dose levels that do not induce a 'high', another strategy may be to admmister an agonist (partial agonist) with a reduced ability (efficacy) to activate CBi receptors. This approach assumes that it should be possible to develop a partial agonist that has sufficient efficacy to relieve muscle spasticity/spasm and pam but insufficient efficacy to produce a full range of cannabhnimetic psychotropic effects even when it occupies all available CBi receptors. A possible lead compound is 0-1184 (see section 4). 7. References 1. Pertwee, R.G., 1988, Pharmacol. Ther., 36, 189. 2. Pertwee, R.G., 1997, Pharmacol. Then, 74, 129. 3. Pertwee, R.G., 2001, Prog. NeurobioL, 63, 569. 4. Molina-Holgado, E., Guaza, C , Borrell, J. and Molina-Holgado, F., 1999, Biodrugs, 12, 317. 5. Herkenham, M., Lynn, A.B., Johnson, M.R., Melvin, L.S., de Costa, B.R. and Rice, K.C., 1991, J. Neurosci., 11, 563. 6. Devane, W.A., Hanus, L., Breuer, A., Pertwee, R.G., Stevenson, L.A., Griffin, G., Gibson, D., Mandelbaum, A., Etinger, A. and Mechoulam, R., 1992, Science, 258, 1946. 7. Di Marzo, V., Melck, D., Bisogno, T. and De Petrocellis, L., 1998, Trends Neurosci., 21, 521. 8. Mechoulam, R., Fride, E. and Di Marzo, V., 1998, Eur. J. Pharmacol., 359, 1.

256 9. Sugiura, T. and Waku, K., 2000, Chem. Phys. Lipids, 108, 89. 10. Hanus, L., Abu-Lafi, S., Fride, E., Breuer, A., Vogel, Z., Shalev, D.E., Kustanovich, I. and Mechoulam, R., 2001, Proc. Natl. Acad. Sci. USA, 98, 3662. 11. Pertwee, R.G., 1999, Curr. Med. Chem., 6, 635. 12. Di Marzo, V., 1999, Life Sci., 65, 645. 13. Hillard, C.J. and Jarrahian, A., 2000, Chem. Phys. Lipids, 108, 123. 14. Maccarrone, M., van der Stelt, M., Rossi, A., Veldink, G.A., Vliegenthart, J.F.G. and Agro, A.F., 1998, J. Biol. Chem., 273, 32332. 15. Piomelli, D., Beltramo, M., Glasnapp, S., Lin, S.Y., Goutopoulos, A., Xie, X.Q. and Makriyannis, A., 1999, Proc. Natl. Acad. Sci. U.S.A., 96, 5802. 16. Ueda, N., PufTenbarger, R.A., Yamamoto, S. and Deutsch, D.G., 2000, Chem. Phys. Lipids, 108, 107. 17. Khanolkar, A.D. and Makriyannis, A., 1999, Life Sci., 65, 607. 18. Calignano, A., La Rana, G., Giuflfrida, A. and Piomelli, D., 1998, Nature, 394, 277. 19. Lambert, D.M., DiPaolo, F.G., Sonveaux, P., Kanyonyo, M., Govaerts, S.J., Hermans, E., Bueb, J.-L., Delzenne, N.M. and Tschirhart, E.J., 1999, Biochim. Biophys. Acta, 1440, 266. 20. Griflfm, G., Fernando, S.R., Ross, R.A., McKay, N.G., Ashford, M.L.J., Shu-e, D., Hufl&nan, J.W., Yu, S., Lainton, J.A.H. and Pertwee, R.G., 1997, Eur. J. Pharmacol., 339, 53. 21. Jdrai, Z., Wagner, J.A., Varga, K., Lake, K.D., Compton, D.R., Martm, B.R., Zimmer, A.M., Bonner, T.I., Buckley, N.E., Mezey, E., Razdan, R.K., Zimmer, A. and Kunos, G., 1999, Proc. Natl. Acad. Sci. U.S.A., 96, 14136. 22. Kunos, G., J^rai, Z., Batkai, S., Goparaju, S.K., Ishac, E.J.N., Liu, J., Wang, L. and Wagner, J.A., 2000, Chem. Phys. Lipids, 108, 159. 23. Di Marzo, V., Breivogel, C.S., Tao, Q., Bridgen, D.T., Razdan, R.K., Zimmer, A.M., Zimmer, A. and Martin, B.R., 2000, J. Neurochem., 75, 2434. 24. Beltramo, M., Rodriguez de Fonseca, F., Navarro, M., Calignano, A., Gorriti, M.A., Grammatikopoulos, G., Sadile, A.G., Giuflfrida, A. and Piomelli, D., 2000, J. Neurosci., 20, 3401. 25. Giuf&ida, A., Rodriguez de Fonseca, F., Nava, F., Loubet-Lescouli6, P. and Piomelli, D., 2000, Eur. J. Pharmacol, 408, 161. 26. Gonzalez, S., Romero, J., de Miguel, R., Lastres Becker, I., Villanua, M.A., Makriyannis, A., Ramos, J.A. and Femdndez Ruiz, J.J., 1999, Life Sci., 65, 327. 27. De Petrocellis, L., Bisogno, T., Davis, J.B., Pertwee, R.G. and Di Marzo, V., 2000, FEBS Lett., 483, 52. 28. Deutsch, D.G., Lin, S., Hill, W.A.G., Morse, K.L., Salehani, D., Arreaza, G., Omeir, R.L. and Makriyannis, A., 1997, Biochem. Biophys. Res. Communs., 231,217. 29. Gifford, A.N., Bruneus, M., Lin, S., Goutopoulos, A., Makriyannis, A., Volkow, N.D. and Gatley, S.J., 1999, Eur. J. Pharmacol, 383, 9. 30. Cervone, K., Keene, A., Lin, S.Y., Makriyannis, A. and Salamone, J.D., 1999, Proc. Soc. Neurosci., 25, 2077.

257 31. Boger, D.L., Sato, H., Lemer, A.E., Hedrick, M.P., Fecik, R.A., Miyauchi, H., Wilkie, G.D., Austin, B.J., Patricelli, M.P. and Cravatt, B.F., 2000, Proe. Natl. Acad. Sci. USA, 97, 5044. 32. Pertwee, R.G., 2000, Expert Opin. Invest. Drugs, 9, 1553. 33. Hillard, C.J., Manna, S., Greenberg, M.J., Di Camelli, R., Ross, R.A., Stevenson, L.A., Murphy, V., Pertwee, R.G. and Campbell, W.B., 1999, J. Pharmacol. Exp. Ther., 289, 1427. 34. Jarrahian, A., Manna, S., Edgemond, W.S., Campbell, W.B. and Hillard, C.J., 2000, J. Neurochem., 74, 2597. 35. Hanus, L., Breuer, A., Tchilibon, S., Shiloah, S., Goldenberg, D., Horowitz, M., Pertwee, R.G., Ross, R.A., Mechoulam, R. and Fride, E., 1999, Proc. Natl. Acad. Sci. USA, 96, 14228. 36. Ross, R.A., Brockie, H.C., Stevenson, L.A., Murphy, V.L., Templeton, F., Makriyannis, A. and Pertwee, R.G., 1999, Br. J. Pharmacol., 126, 665. 37. Rinaldi-Carmona, M., Barth, F., H^auhne, M., Shire, D., Calandra, B., Congy, C , Martinez, S., Maruani, J., N^liat, G., Caput, D., Ferrara, P., Soubrie, P., Breliere, J.C. and Le Fur, G., 1994, FEBS Lett., 350, 240. 38. Rinaldi-Carmona, M., Barth, F., Millan, J., Derocq, J.-M., Casellas, P., Congy, C, Oustric, D., Sarran, M., Bouaboula, M., Calandra, B., Portier, M., Shu-e, D., Breliere, J.-C. and Le Fur, G., 1998, J. Pharmacol. Exp. Ther., 284, 644. 39. Bouaboula, M., Perrachon, S., Milligan, L., Canat, X., Rinaldi-Carmona, M., Portier, M., Barth, F., Calandra, B., Pecceu, F., Lupker, J., Maffrand, J.-P., Le Fur, G. and Casellas, P., 1997, J. Biol. Chem., 272, 22330. 40. Coutts, A.A., Brewster, N., Ingram, T., Razdan, R.K. and Pertwee, R.G., 2000, Br. J. Pharmacol., 129, 645. 41. MacLennan, S.J., Reynen, P.H., Kwan, J. and Bonhaus, D.W., 1998, Br. J. Pharmacol, 124, 619. 42. Ross, R.A., Brockie, H.C., Fernando, S.R., Saha, B., Razdan, R.K. and Pertwee, R.G., 1998, Br. J. Pharmacol., 125, 1345. 43. Ross, R.A., Gibson, T.M., Stevenson, L.A., Saha, B., Crocker, P., Razdan, R.K. and Pertwee, R.G., 1999, Br. J. Pharmacol., 128, 735. 44. Hosohata, K., Quock, R.M., Hosohata, Y., Burkey, T.H., Makriyannis, A., Consroe, P., Roeske, W.R. and Yamamura, H.I., 1997, Life Sci., 61, PL115. 45. Hosohata, Y., Quock, R.M., Hosohata, K., Makriyannis, A., Consroe, P., Roeske, W.R. and Yamamura, H.I., 1997, Eur. J. Pharmacol., 321, Rl. 46. Pertwee, R.G., Fernando, S.R., Nash, J.E. and Coutts, A.A., 1996, Br. J. Pharmacol., 118,2199. 47. Landsman, R.S., Makriyannis, A., Deng, H., Consroe, P., Roeske, W.R. and Yamamura, H.I., 1998, Life Sci., 62, PL109. 48. Barth, F., 1998, Expert Opin. Ther. Patents, 8, 301. 49. Amone, M., Maruani, J., Chaperon, F., Thiebot, M-H., Poncelet, M., Soubri6, P. and Le Fur, G., 1997, Psychopharmacol., 132, 104. 50. Colombo, G., Agabio, R., Diaz, G., Lobina, C , Reali, R. and Gessa, G.L., 1998, Life Sci., 63, PLl 13. 51. Simiand, J., Keane, M., Keane, P.E. and Soubrie, P., 1998, Behav. Pharmacol.,

258 9, 179. 52. Brotchie, J.M., 1998, Mov. Disord., 13, 871. 53. Poncelet, M., Bamouin, M-C, Brelidre, J-C, Le Fur, G. and Soubri6, P., 1999, Psychopharmacol., 144, 144. 54. Terranova, J.-P., Storme, J.-J., Lafon, N., PeriOj A., Rinaldi-Carmona, M., Le Fur, G. and Soubrie, P., 1996, Psychopharmacol, 126, 165. 55. Shen, M. and Thayer, S.A., 1998, Mol. Pharmacol., 54, 459. 56. De Petrocellis, L., Melck, D., Palmisano, A., Bisogno, T., Laezza, C , Bifulco, M. and Di Marzo, V., 1998, Proc. Natl. Acad. Sci. U.S.A., 95, 8375. 57. Galve-Roperh, I., Sdnchez, C , Cort6s, M.L., Gomez del Pulgar, T., Izquierdo, M. and Guzman, M., 2000, Nature Med., 6, 313. 58. Green, K., 1998, Arch. Ophthabnol., 116, 1433. 59. Martin, B.R. and Lichtman, A.H., 1998, Neurobiol. Dis., 5, 447. 60. Hemming, M. and Yellowlees, P.M., 1993, J. Psychopharmacol, 7, 389. 61. Muller-Vahl, K.R., Kolbe, H., Schneider, U. and Emrich, H.M., 1999, Forsch. Komplementarmed., 6 (Suppl 3), 23. 62. Muller-Vahl, K.R., Schneider, U., Kolbe, H. and Emrich, H.M., 1999, Amer. J. Psychiat., 156, 495. 63. Pertwee, R.G., 1999, CNS Drugs, 11, 327. 64. Pertwee, R.G., 2001, Pharmacol Ther., Accepted for publication. 65. Shohami, E. and Mechoulam, R., 2000, Drug Dev. Res. 50, 211. 66. Hampson, A.J., Grimaldi, M., Axelrod, J. and Wink, D., 1998, Proc. Natl. Acad. Sci. U.S.A., 95, 8268. 67. Burstein, S.H., 1999, Pharmacol. Ther., 82, 87. 68. Rhee, M.-H., Vogel, Z., Barg, J., Bayewitch, M., Levy, R., Hanus, L., Breuer, A. and Mechoulam, R., 1997, J. Med. Chem., 40, 3228. 69. Zurier, R.B., Rossetti, R.G., Lane, J.H., Goldberg, J.M., Hunter, S.A. and Burstein, S.H., 1998, Arthritis and Rheumatism, 41, 163. 70. Burstein, S.H., Friderichs, E., K6gel, B., Schneider, J. and Selve, N., 1998, Life Sci, 63, 161. 71. Brenneisen, R., Egli, A., ElSohly, M.A., Henn, V. and Spiess, Y., 1996, Int. J. Clin. Pharmacol Ther., 34, 446. 72. Pertwee, R.G., Gibson, T.M., Stevenson, L.A., Ross, R.A., Banner, W.K., Saha, B., Razdan, R.K. and Martin, B.R., 2000, Br. J. Pharmacol, 129, 1577. 73. Baker, D., Prce, G., Croxford, J.L., Brown, P., Pertwee, R.G., Makriyannis, A., Khanolkar, A., Layward, L., Fezza, F., Bisogno, T. and Di Marzo, V., 2001, FASEB J., 15, 300. 74. Walker, J.M., Huang, S.M., Strangman, N.M., Tsou, K. and Saftudo-Pefta, M.C., 1999, Proc. Nat. Acad. Sci. (USA), 96, 12198. 75. Welch, S.P. and Stevens, D.L., 1992, J. Pharmacol Exp. Ther., 262, 10. 76. Pertwee, R.G., 1992, Marijuana/Cannabinoids. Neurobiology and Neurophysiology, p. 165: Eds, L. Murphy and A. Bartke, CRC Press, Boca Raton. 77. Richter, A. and Loscher, W., 1994, Eur J Pharmacol, 264, 371.

259

AUTHOR INDEX Akerman, K.E.O., 181 Angeli, P., 95 Bruggink, A., 21 Bymaster, P.P., 97 Chopra, L, 213 Crawley, J., 97 Czupalla, C , 161 Deng, C , 97 Duttaroy, A., 97 Egebjerg, J., 1 Felder, C.C., 97 Freissmuth, M., 161 Furness, L.M., 237 Gaviraghi, G., 95 Gillet, VJ., 125 Golemi, D., 193 Gomeza, J., 97 Greenberg, E.P., 207 Greenwood, J., 1 Haddad, J., 193 Heck, AJ.R., 81 Hervieu, G., 115 Hesse, L., 213 Hogner, A., 1 Holmqvist, T., 181 Hubbart, R.E., 53 Ishiwata, A., 193 Just, H., 161 Kastrup, J.S., 1 Kieboom, A.P.G., 39 Kliewer, S., 67 Kotra, L., 193 Kouwijzer, M., 57 Krogsgaard-Larsen, P., 1 Kukkonen, J.P., 181 Langley, D., 135 Lee, W., 193

Maier, C.S., 81 Makita, R., 97 Maveyraud, L., 193 Mestres, J., 57 Miyakawa, T., 97 Miyashita, K., 193 Mobashery, S., 193 M0ller, E.H., 1 Mourey, L., 193 Nanoff, Chr., 161 Nasman, J., 181 Niirnberg, B., 161 O'Neill, A., 213 Pertwee, R.G., 249 Previtali, S., 115 Quattrini, A., 115 Reetz, M.T., 27 Samama, J., 193 Scheideler, M.A., 115 Schoevaart, R., 39 Seneci, P., 147 Shannon, H., 97 Stefan, E., 161 Stensb0l, T.B., 1 Strange, Ph.G., 175 Vakulenko, S., 193 Van Boeckel, C.C.A., 13 Vogensen, S.B., 1 Wess, J., 97 Willett, P., 125 Yamada, M., 97 Zaratin, P.F., 115 Zhang, B.B., 227 Zhang, L., 97 Zhang, W., 97

This Page Intentionally Left Blank

261

SUBJECT INDEX a2-adrenergic receptor, 181 accessory proteins, 161 agonist channeling, 181 AMPA receptor ligands, 1 AMPA receptor, 1 analgesia, 97 anandamide, 249 antibacterial compounds, 193, 213 antibacterial resistance, 213 antidiabetic agents, 227 antipathogenic drug discovery, 207 antipsychotic action, 175 antithrombotics, 13 arixtra, 13

dual antithrombotic, 13 dynamics simulations, 57

p-lactam antibiotics, 193 p-lactamase inhibitors, 213 biocatalysts, 21 biofilms, 207 biological diversity, 95 biomolecular interactions, 81 biosynthesis, 21, 39 biotransformation, 21

G-proteins, 161 gene expression pattern, 237 genistein, 57 genomic approaches, 213 glucose isomerase, 39 glyoxalase I, 81

Ca^"^ mobilisation, 181 cAMP accumulation, 181 cannabinoid receptors, 249 cannabis, 249 chemical diversity, 95, 135, 147 combinatorial libraries companies, 135 combinatorial libraries, 125, 147 compound selection, 135 computational methods, 125 D2 dopamine receptor, 175 database, 237 diethylstibestrol, 57 directed evolution of enzymes, 27 DNA binding domain, 67 dopamine receptor, 175 drug discovery, 95, 207, 237 drug resistance, 193 drug-receptor interactions, 81

enantioselective enzymes, 27 endocannabinoid membrane transport, 249 endocannabinoids, 249 enzyme catalyzed conversion, 39 estradiol, 53, 57 estrogen receptor, 53, 57 estrogen response modulators, 53 expression database concept, 237 expression of orphan GPCR's, 115 food intake, 97

heparin, 13 high-throughput screening, 27, 135, 147 high-throughput structural characterization, 147 hypothermia, 97 inflammation, 115 inhibitors of P-lactamases, 193 insulin mimetics, 227 insulin receptor, 227 insulin signaling pathway, 227 inverse agonism, 175 kainic acid receptor agonists, 1 lead discovery, 147 lead optimization, 147, 237 Hgand binding domain, 53 mass spectrometry, 81 molecular biology, 21 molecular diversity, 125 molecular docking, 57

262 molecular similarity, 125 muramyl peptide ligases, 213 muscarinic autoreceptors, 97 muscarinic receptor subtypes, 97 neuropathies, 115 nuclear receptors, 67 organic synthesis, 21, 39 orphan GPCR's, 115 orphan nuclear receptors, 67 pentasaccharide, 13 peptidoglycan, 193, 213 peroxisome proliferator-activated receptors, 67 phytase, 39 progesterone, 57 protein-membrane interactions, 81 Pseudomonas auruginosa, 207 quorum sensing, 207

raloxifene, 53 receptor clustering, 161 receptor signalling, 161 receptor synthesis, 161 recombinant AMPA receptor binding domain, 1 resistance mechanism inhibitors, 213 reverse cholesterol transport, 67 reverse endocrinology, 67 RNA polymerase inhibitors, 213 schizophrenia, 175 selection of combinatorial libraries, 125 solid phase chemistry, 147 stereoselective synthesis, 27 steroid hormone receptors, 57 tremor, 97 vancomycin, 81 virtual screening, 147