Advances in Cancer Research Volume 72

~~

Advances in

CANCER RESEARCH Volume 72

This Page Intentionally Left Blank

Advances in

CANCER RESEARCH Volume 72

Edited by

G e o r g e F. Vande Woude A B L-Basic Reseorch Progrmi National Cancer Institrite Frederick Cancer Research and Development Center I-rederick, Marylrind

G e o r g e Klein Microbiology and Ttrmor Biology Center Karolinska Institutet Stockholtn, Swedeti

ACADEMIC PRESS San Diego London Boston New York Sydney Tokyo Toronto

This book is printed on acid-free paper.

@

Copyright 0 1998 by ACADEMIC PRESS All Rights Reserved. N o part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the Publisher. The appearance of the code at the bottom of the first page of a chapter in this book indicates the Publisher's consent that copies of the chapter may be made for personal or internal use of specific clients. This consent is given on the condition, however. that the copier pay the stated per copy fee through the Copyright Clearance Center, Inc. (222 Rosewood Drive, Danvers. Massachusetts 01923). for copying beyond that permitted by Sections 107 or 108 of the U.S. Copyright Law. This consent does not extend to other kinds of copying, such as copying for general distribution, for advertising or promotional purposes, for creating new collective works, or for resale. Copy fees for pre-1998 chapters are as shown on the title pages. I f no fee code appears on the title page. the copy fee is the same as for current chapters. 0065-230>(/98 $25.00

Academic Press 11

di\'isioii of Hirnwm Br1ic.r & Cornpuny

525 B Street, Suite 1900, San Diego. California 92101-4495, USA http://www.apnet.com Academic Press Limited 24-28 Oval Road. London NW I 7DX. U K http://www.hbuk.co.uk/ap/

International Standard Book Number: 0- 12-006672-6 PRINTED LN THE UNITED STATES OF AMERICA 97 98 9 9 0 0 01 0 2 B B 9 8 7 6

5

4

3 2

I

Contents

Contributors to Volume 72 ix

FOUNDATIONS IN CANCER RESEARCH Foulds’ Dangerous Idea Revisited: The Multistep Development of Tumors 40 Years Later George Klein

I. Foulds’ Rules 2 11. Oncogenetics 3 111. Other Destabilizing Mutations Caused by Loss of DNA

Repair Functions 11 IV. Molecular Biology of Multistep Carcinogenesis 14 V. Strong Selective Pressures Favor Multiple Escapes 1 7 V1. Conclusion 19 References 20

FOUNDATIONS IN CANCER RESEARCH Cancer Cells Exhibit a Mutator Phenotype Lawrence A. Loeb 1. Introduction 26 11. Historical Perspective 28 111. Requirements for a Mutator Phenotype 3 2 IV. Origins of Mutations 36 V. Candidate Target Mutator Genes 38 VI. Microsatellite Instability 4 3 VII. Theoretical and Practical Considerations 4 7 VIII. Summary and Perspectives 51 References 5 2

V

vi

Contents

Increasing Complexity of R a s Signal Transduction: Involvement of Rho Family Proteins Roya Khosravi-Far, Sharon Campbell, Kent L. Rossman, and Channing J. Der 1. Introduction 57 11. Ras Is a Point of Convergence for Diverse Extracellular Signal-Stimulated Pathways 60 111. Ras Activation of Raf-Independent Pathways Contributes to

Ras Transformation 65 1V. Ras Mediates Its Actions through Interaction with Multiple Effectors 6 9 V. Ras Activation of a GTPase Cascade: An Involvement of R h o Family Proteins in Transformation 78 VI. Rho Family Proteins Mediate Their Actions through Interaction with Multiple Effectors 85 VII. A Search for the Missing Link between Ras and Rho Family Proteins 95 VIII. Increasing Complexity of Ras Signal Transduction: A Boon or Bust for Drug Discovery and the Development of Anti-Ras Drugs for Cancer Treatment? 96 IX. Future Directions 97 References 9 9

B-Myb: A Key Regulator of the Cell Cycle Mark K. Saville and Roger I. Watson 1. Introduction 109 11. B-Myb Structure and Functional Domains 11 I 111. Transcriptional Regulation o f B-rnyb 115 IV. Modification of B-Myb Protein in the Cell Cycle 123 V. The Requirement for B-Myb in Cell Proliferation 1 2 7 VI. B-Myb Function 132 References 137

Alterations in DNA Methylation: A Fundamental Aspect of Neoplasia Stephen B. Baylin, James G. Herman, Jeremy R. Graff, Paula M.Vertino, and Jean-Pierre lssa

I. Introduction 142 11. Brief History of DNA Methylation in Eukaryotes 142 111. The Normal Roles for Cytosine Methylation in Higher Order Eukaryotes 144 IV. Abnormalities of DNA Methylation in Neoplasia 150 V. Mechanisms Underlying the DNA Methylation Changes in Neoplastic Cells 167 VI. An Overview of Tumor Progression That Incorporates the Roles of Altered DNA Methylation 186 VII. Clinical Implications of Altered DNA Methylation in Cancer 189 References 190

Contents

Ara-C: Cellular and Molecular Pharmacology S t e v e n Grant I. 11. 111. IV. V. VI. VII. VIII.

Introduction 198 Structure 200 Metabolism 200 Mechanisms of Cytotoxicity 205 Mechanism of Resistance 210 Signaling Pathways and Oncogene Interactions 213 Modulation of ara-C Associated Toxicity 220 Conclusions 225 References 225

Index 235

vii

This Page Intentionally Left Blank

Contributors

Nuinhers in parentheses indicate the pages on iubrch the authors’ contrihictions begin.

Stephen B. Baylin, The Johns Hopkins Comprehensive Cancer Center, Department of Medicine, The Johns Hopkins Medical Institutions, Baltimore, Maryland 21231 (141) Sharon Campbell, Department of Biochemistry and Biophysics, Lineberger Comprehensive Cancer Center, University of North Carolina, Chapel Hill, North Carolina 27599 (57) Channing J. Der, Department of Pharmacology, Lineberger Comprehensive Cancer Center, Curriculum in Genetics and Molecular Biology, University of North Carolina, Chapel Hill, North Carolina 27599 (57) Jeremy R. Graff, The Johns Hopkins Comprehensive Cancer Center, The Johns Hopkins Medical Institutions, Baltimore, Maryland 21231 (141 ) Steven Grant, Departments of Medicine, Pharmacology, and Microbiology, Medical College of Virginia, Virginia Commonwealth University, Richmond, Virginia 23298 (197) James G. Herman, The Johns Hopkins Comprehensive Cancer Center, The Johns Hopkins Medical Institutions, Baltimore, Maryland 21231 (141) Jean-Pierre Issa, The Johns Hopkins Comprehensive Cancer Center, The Johns Hopkins Medical Institutions, Baltimore, Maryland 21231 (141) Roya Khosravi-Far, Department of Biology, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139 (57) George Klein, Microbiology and Tumor Biology Center (MTC),Karolinska Institutet, S-17177 Stockholm, Sweden (1) Lawrence A. Loeb, The Joseph Gottstein Memorial Cancer Research Laboratory, Department of Pathology, University of Washington, Seattle, Washington 98195-7705 (25) Kent L. Rossman, Department of Biochemistry and Biophysics, Lineberger Comprehensive Cancer Center, University of North Carolina, Chapel Hill, North Carolina 27599 ( 5 7 ) Mark K. Saville, Department of Medical Microbiology and Ludwig Institute for Cancer Research, Imperial College School of Medicine at St. Mary’s, Norfolk Place, London W2 lPG, United Kingdom (109)

ix

X

Contributors

Paula M. Vertino, The Johns Hopkins Comprehensivc Cancer Center, The Johns Hopkins Medical Institutions, Baltimore, Maryland 21 23 1 (141) Roger J. Watson, Department of Medical Microbiology and Ludwig Institute for Cancer Research, Imperial College School of Medicine at St. Mary’s, Norfolk Place, London W2 1 PG, United Kingdom (109)

FOUNDATIONS IN CANCER RESEARCH

Foulds’ Dangerous Idea Revisited: The Multistep Development of Tumors 40 Years Later George Klein Microbiology and Tumor Bio/ogy Center Karolrnska Instrtiifet Stockholm. Suneden

I. Foulds’ Rules 11. Oncogenetics

111. ’

IV. V. VI.

A. Oncogenes B. Tumor Suppressor Genes C. DNA Repair Genes D. Genes That Influence Programmed Cell Death by Apoptosis Other Destabilizing Mutations Caused by Loss of DNA Repair Functions Molecular Biology of Multistep Carcinogenesis Strong Selective Pressures Favor Multiple Escapes Conclusion References

In his recent book on evolution, Darwin 2 Dangerous Idea, Dennett (1995) alludes to natural selection, an idea that has acted as a “corrosive acid,” dissolving established societal, ideologic, and religious structures. I have always regarded Leslie Foulds as the micro-Darwin of the neoplastic microevolution. Dennett’s title provoked the association that Foulds’ (1958)analysis of tumor progression, a term first used by Rous to describe the process whereby “tumors changed from bad to worse” (Rous and Beard, 1935), might have been corrosive as well, albeit on a much smaller scale, as befits a limited evolutionary process that is, moreover, pathological and a dead end in itself. Just as Darwinian natural selection has made creationistic myths obsolete in a single stroke, the concept that tumors develop by multiple, stepwise changes has dealt a fatal blow to the idea that has dominated cancer research during the first half of the century: that a single key event is responsible for tumor development. Biochemists, immunologists, cytogeneticists, and virologists kept searching, each with the methodology characteristic of their disciplines, for the decisive event that had to be discovered before the riddle of Advancer in CANCER RESEARCH 00hS-23OX/YX $25.00

Copyrighr 0 IYYX hy Academic Press. All righrs of reprorluurion In any trirm reserved.

2

George Klein

cancer could be solved. The Warburg hypothesis of a common metabolic disturbance, the mirage of a common cancer antigen, the universal cancer virus, and general chromosomal imbalance are some examples of the all-encompassing theories that were advocated with particular fervor. The multistep theory of cancer development originated from two main sources. Mathematical analysis of the age-incidence curves indicated that most of the major human tumors arose after five to seven mutation-like changes (Armitage and Doll, 1954; Farber and Cameron, 1980). Horizontal studies o n the natural history of tumor development at the tissue level, including Foulds’ own observations (Foulds, 1969; 1975), have defined some of these steps in histopathologic and biologic terms. The idea that tumors evolve after a series of mutations that provide the cell with a selective advantage within the gradual process of its emancipation from growth control has attained a dominating position in cancer research, just as Darwinian natural selection is accepted by all biologists (Klein and Klein, 1985a; Nowell, 1976).This concept is less generally recognized by the medical profession and only to a minor extent, if a t all, by nonbiologists. Actually, it makes good sense. It is generally accepted that cell division and the processes leading to it are regulated by a complex set of receptors as well as by signal transduction and transactivation pathways. The advent of knockout mice brought the unexpected news that several known genes that participate in this regulation can be deleted without fatal harm to the organism. Looking back a t the natural history of cancer as described by Foulds (1958) from o u r present knowledge concerning the regulation of the cell cycle by multiple signals, the loss of the controlled state as it occurs during neoplastic development must proceed through multiple steps.

1. FOULDS’ RULES The core of Foulds’ ideas is factorial analysis. In 1958, Foulds stressed that the neoplastic phenotype is composed of several unit characteristics, capable of independent reassortment and liable to independent progression. Progression was defined as a series of stepwise changes in several unit characteristics. This implied that progression could proceed along alternative pathways and, therefore, that each tumor of a given type had a certain degree of biologic individuality. During the past two decades, Foulds’ postulates have been provided with molecular meaning. There is only one major departure from the original text. Darwin developed his theory of natural selection without any knowledge of

Foulds’ Dangerous Idea Revisited

3

genes. A century later, Foulds formulated the rules of tumor progression but rejected the notion that mutations, in the strict sense of the word (i.e., changes a t the DNA level), could be responsible for its steps. H e accepted (Foulds, 1969) the conclusions of our early work (Klein & Klein, 1957; 1958),which postulated that the selection of ascites convertible variants isolated from solid tumors and of H-2 haplotype loss variants from H-2 heterozygous mouse tumors was due to variation and selection. He interpreted his own observations on the focal appearance of behaviorally different tumor cell variants in a similar way. Foulds did not believe, however, that phenotypic changes in a diploid cell could be caused by mutations because most of them would be recessive and remain “silent,” owing to the presence of a normal allele. H e chose to speak about neoplastic development, with a strong emphasis on the second word. This also became the title of his major monography (Foulds 1969; 1975). He attributed the steps of progression to epigenetic changes, akin to the phenotypic switches of differentiation. He was proved largely, if not completely, wrong on this point. Epigenetic, di’ ferentiation-related changes may occur and can be induced in tumors, but they are superimposed on a chain of consecutive changes at the DNA level. Tumor progression itself can be viewed as a combination of genetic and epigenetic changes, with mutations and DNA methylation, respectively, as paradigmatic mechanisms. There is only one known tumor that may have arisen by a purely epigenetic mechanism: the mouse teratoma, investigated by Mintz two decades ago (Mintz and Fleischman, 1981). The diploid tumor she used for these studies grew progressively and killed 100% of its grafted syngeneic hosts during an observation period of 8 years. Nevertheless, it could be fully normalized by implantation into an early blastocyst, whereupon it gave rise to all normal tissues of the mouse. Unfortunately, this important experiment has not been followed up a t the molecular level.

11. ONCOGENETICS Four worlds of genes are presently known that can contribute to tumor progression by constitutive activation, mutation, or deletion. They are, in order of their discovery, oncogenes, tumor suppressor genes, DNA repair genes, and genes that influence programmed cell death by apoptosis. There is a certain functional overlapping among these categories, as pointed out later. Some genes have multiple functions that may assign them to more than one category. It nevertheless seems practical to discuss the current state of the art under these four headings. Also, the tumorigenic process can be influenced by genes that belong to other functional categories. They may act

4

George Klein

by influencing vascularization, telomerase activity, invasion and metastasis, and immune escape. They are not the subject of this review.

A. Oncogenes The oncogenes catalogued and reviewed by Hesketh (1994)are highly conserved household genes. They are involved in signaling functions that can trigger or drive the cell cycle. Their illegitimate activation by structural o r regulatory changes may stimulate the cell to continuous proliferation. Some act by producing a growth factor in a cell that is normally programmed to respond to it, not to produce it. Others encode growth factor receptors o r signal-transducing proteins, which may mutate to emit a faulty signal, like a buzzer that has gotten stuck. Still others are transcription factors that have been illegitimately activated by mutation, retroviral insertion, chromosomal translocation, or gene amplification, with the stimulated DNA replication as a direct or indirect consequence. Activated oncogenes stimulate cell division in a dominant fashion. This meets one of Foulds’ objections, that cancer-related mutations would necessarily be recessive in somatic cells. Dominance does not mean, however, that an activated oncogene can transform any target cell. On the contrary, the transformation-sensitive target cell spectrum of each oncogene shows highly idiosyncratic features. Part of this is differentiation related. When the retrovirally transduced oncogenes were first discovered, it became apparent that each of them had its characteristic lineage- and/or differentiation-dependent “window” of transformable target cells (see Klein, 1982). Particularly informative experiments were performed with temperature-sensitive mutants of vsrc and v-erb B (Boettiger et al., 1977; Weintraub et al., 1982). These oncogenes could transform target cells of several lineages into undifferentiated neoplastic cells at the permissive temperature. Even a brief stay of the transformed cells at the nonpermissive temperature released the differentiation block imposed by the oncogene, however. Each cell type remained faithful to its lineage or, in Ephrussi’s term, its “epigenotype.” The oncogenes mutated to the terminal form of the respective lineage and stopped dividing. Subsequent reexpression of the oncoprotein at the permissive temperature could no longer transform the cells. This illustrates how the release of maturation arrest by a purely epigenetic switch can revert malignant behavior. The oncogene field has other idiosyncratic features. It is not easy to explain why some proteins within a certain signal-transducing chain or within a given receptor or transcription factor family can turn into an oncogene, whereas others within the same chain or family that might have been expected to have the same potential do not appear on the list of known oncogenes. It is also peculiar that some oncogene activation events-such as the

Foulds' Dangerous Idea Revisited

5

constitutive switching on of c-myc by juxtaposition to an immunoglobulin locus in Burkitt lymphoma and rodent plasmacytomas (Klein and Klein, 1985b) or the creation of a new fusion protein by the bcrlabl translocation in chronic myelogenous leukemia (Konopka and Witte, 1985)-occur in all or most tumors of a given phenotype, whereas others are only found in a minority. B- and T-cell-derived lymphomas and leukemias are different in this respect in both mice and humans. All tumors of a given B-cell phenotype may show the same oncogene activation event, as exemplified by the Iglmyc and the Iglbcl-2 translocations in Burkitt lymphoma and in follicular lymphoma, respectively. T-cell leukemias and lymphomas of a given phenotype show a broad variety of different oncogene activation events (Rabbitts, 1994). In the human T-cell leukemias, oncogenes are mainly activated by chromosomal translocation, in contrast to murine T-cell tumors, in which activation is usually due to retroviral insertion (Jonkers and Berns, 1996). Still, the same broad variation prevails in both human and murine T cells. This difference between B and T cells may relate to the broader activability of T cells. Depending on the oncogene that has been juxtaposed to an an immunoglobulin locus by chromosomal translocation, the derived B-cell tumors differ with regard to their grade of malignancy. Activated myc is known to drive cell proliferation. It therefore makes sense that Burkitt lymphoma is a highly malignant tumor. It is also easy to rationalize that the Iglbcl-2 translocation-carrying follicular lymphomas are low grade because bcl-2 protects against apoptosis but does not provide a continuous drive to cell division (Korsmeyer, 1992). It is less easy to understand why bcl-I, now identified as cyclin D1, generates low-grade lymphomas after translocation to an ZgH locus (Weisenburger, 1991). This suggests that the activation of this single cyclin is not sufficient to provide a strong driving force for proliferation. Activation of c-myc by translocation into one of the physiologically rearranging immunoglobulin loci is thus a regular and probably rate-limiting (bottleneck)event in the development of Burkitt lymphoma and rodent plasmacytoma. In the course of solid tumor progression, c-myc or, alternatively, N-myc o r L-myc may be activated by gene amplification. This occurs only in a fraction of all tumors of a given type. In small cell lung carcinoma and neuroblastoma, myc amplification was found to be associated with the appearance of a more malignant variant (Little et al., 1983). There are many other examples of oncogene amplification. The prognosis-related amplification of c-erb B2, or m u , in breast carcinoma is particularly noteworthy among them (Slamon et al., 1987). Neither the mechanism of gene amplification nor its contribution to the malignant phenotype has been satisfactorily explained. In a particularly wellstudied model, the degree of dehydrofolate reductase amplification could be related to the level of methotrexate resistance (Biedler et al., 1983).Upon the release of the selective pressure by the drug, the amplified gene copies were

6

George Klein

lost. Extrapolating this to the case of oncogene amplification, Bishop (1983) suggested that the persistence of amplified oncogenes in growing tumor cells must mean that they provide the cell with a selective advantage. Amplification of an androgen receptor gene in hormone-independent variants of prostatic carcinoma is consistent with this concept because increased receptor expression facilitates the capture of the ligand, even if it is present only in a low concentration (Kallioniemi and Visakorpi, 1996).

B. Tumor Suppressor Genes The earliest indications suggesting the existence of genes that may antagonize tumor development came from several sources. Fusion of normal with malignant cells showed, unexpectedly, a regular suppression of tumorigenicity, as long as the chromosome complement derived from the normal parent was maintained (reviewed by Klein, 1976). Reappearance of tumorigenicity was associated with the loss of specific chromosomes from the normal parent. Another early finding was the isolation of more or less “normalized” revertants from tumor cell cultures under saturation conditions in which continued DNA synthesis of the tumor cells was made suicidal by highly radioactive thymidine or a combination of 5-bromodeoxyuridine and blue light. Although several types of revertant cells could be isolated and characterized, the interesting ones were those that maintained the original transforming oncogene in an unchanged form. For example, the viral Ki-ras gene was still active after helper rescue from the phenotypic revertants. The transformed phenotype was suppressed by one of several downstream-acting genes, exemplified by Krev-1 (Noda, 1993; Yanagihara et al., 1990). The molecular approach has shown that tumor development and progression are accompanied by numerous genetic losses. Studies o n colon, breast, prostate, kidney, lung, and brain tumors have created the inipression that gene inactivation may be more frequent than oncogene activation (Collins and James, 1993; Fearon and Vogelstein, 1990; Kallioniemi and Visakorpi, 1996; Kok et al., 1996; Kovacs et al., 1988; Lee, 1995; Yokota and Sugimura, 1993). The most meaningful functional information came from the study of two suppressor genes, RB and p53. It can hardly be a coincidence that the transforming proteins of three different DNA tumor viruses (SV40, adenovirus, and papillomavirus) have R B and p53 as their cellular targets and that they use essential domains of their transforming proteins to cancel their function (Selivanova and Wiman, 1995; Weinberg, 1995). This surprising convergent evolution among viruses of different origins may have been driven by the common requirement of resting cells to enter the cell cycle, to establish latent viral persistence. RB and p53 play important and different roles as cell cycle checkpoint con-

Foulds' Dangerous Idea Revisited

7

trols. RB has a pivotal position a t the R (restriction) point that controls the transition from C1 to S phase (Weinberg, 1995). The nonphosphorylated form of the protein stops the cycle in middle to late G1 phase. This block to cell cycle progression serves not only as a control post in proliferating cells but also as a prerequisite for the exit of cells from the cycling compartment by differentiation or by senescence. The position of the RB protein within the complex chain of events that must take place to allow the cells to transgress the cycle is well known. Triggering of the cycle by appropriate ligand-receptor interactions and subsequent signaling elicit the activation of cyclin-cyclin-dependent kinase (CDK) complexes (Hall and Peters, 1996; Hunter and Pines, 1994; Sherr, 1993). This leads, after several steps, to the inactivation of RB by phosphorylation. Phosphorylated RB releases a bound transcription factor, E2F, that transactivates numerous genes and represses some. This is obviously a gross oversimplification of a much more complex situation with many additional participants, including a family of RB-like molecules (p107, p130, and others) and several different E2F-like factors. Only RB, however, has been identified as a tumor suppressor gene, and none of the E2F family members have been identified as potential oncogenes in vivo. Somewhat surprisingly, E2F1 knockout mice were found to develop a variety of tumors, perhaps owing to the loss of the suppressor function of E2F (Yamasaki et al., 1996). The loss of RB plays a direct causative role in retinoblastoma and is also essential for the development of osteosarcomas in RB families. It can also contribute to the progression of other tumors, including sarcomas and prostatic and bladder carcinomas (Knudson, 1993). A central theme of the oncogene field repeats itself: the same gene may act in a regular bottleneck capacity in causing tumors within a certain cell lineage; whereas in tumors derived from other cell types, it contributes to some late event that occurs only in part of the tumors during their progression and that may render them more invasive, metastatic, or less responsible to local growth-regulating signals. The earliest molecular analyses of familial retinoblastomas have shown that Foulds' objection to the mutational basis of tumor development and progression is not valid for recessive mutations either. Foulds could not see how a recessive mutation that affects only one of two alleles could be expressed in a diploid cell. It turned out, however, that one RB allele is mutated in the germ line, as Knudson originally postulated (Knudson, 1971), whereas the second is cancelled by chromosome loss, mutation, deletion, or mitotic crossing over during somatic development (Cavenee et al., 1983). In about 7 0 % of the tumors, the normal allele is eliminated through the loss of the whole chromosome by nondisjunction, with or without the duplication of the remaining chromosome. Different cytogenetic mechanisms operate in different individual tumors, even if they arise in the same patient, but the loss of the normal allele is their invariable common denominator.

a

George Klein

Progression through the cell cycle can be inhibited by several other proteins as well. Some have been identified as tumor suppressors, but others have not. Some of the more prominent examples deserve special mention. DNA damage is a potent inducer of p53. This switches on p21 (also known as WAF-l), a protein that can inhibit all cyclin-CDK complexes (Xiong et a/., 1993; elDeiry et al., 1992). Although p53 is an important tumor suppressor gene, p21 is less prominent in that capacity but is occasionally found mutated in prostatic cancers (Gao et d., 1995). A transforming growth factor p-inducible protein, p27, can inhibit one or more cyclin-CDK complexes as well but is not known as a tumor suppressor gene either (Toyoshima and Hunter, 1994). In contrast, p16 and its close relative p l 5 , which specifically inhibit cyclin Dl-CDK4 complexes, were identified as suppressor genes because they have mutated or became methylated in many different tumors. The differential significance of the various cyclin- and CDK-inhibitor proteins in relation to tumor development has not been fully explained (Hall and Peters, 1996).

C. DNA Repair Genes More than half of all tumors carry mutant p53, the most frequently mutated gene in human tumors (Selivanova and Wiman, 1995). The tumorassociated p53 mutants have lost the ability to bind to specific p53 binding sites on double-stranded DNA by the middle domain of the protein molecule, or have only a strongly reduced specific DNA binding. Most are missense point mutations, suggesting a positive selection for mutant p.53, in contrast to RB, which is usually inactivated by nonsense mutation. The high frequency of p53 mutations is consistent with the broad involvement of this protein in the control of the cell cycle and apoptosis, its response to DNA damage and other inducers, and its ability to influence tumor development in several different ways. Certain p53 mutants have a direct transforming effect in vitro, even on p53 knockout fibroblasts, and can thus be regarded as oncogenes (Dittmer et al., 1993; Hann and Lane, 1995; Milner, 1994). Other mutants have no direct transforming activity but have lost their suppressor function. They no longer bind specifically to DNA and cannot arrest the cycle. Some of these mutants act as dominant negatives, owing to their ability to heterodimerize with wild-type p53 and thereby inhibit its function (Dittmer et al., 1993; Zambetti and Levine, 1993). p53 has been called the “guardian of the genome” (Kastan et af., 1991; Lane, 1992). Normal p53 is a short-lived protein that is expressed a t a low level. Damaged DNA or other p53 inducers, such as adenoviral E1A (Debbas and White, 1993; Lowe and Ruley, 1993) or elevated c-myc (Hermeking and Eick, 1994), raise the p53 level by stabilizing of the protein. The accumulated p53 arrests the cycle by the transcriptional transactivation of p21,

Foulds' Dangerous Idea Revisited

9

as mentioned earlier. DNA repair enzymes gain tine to work during this growth arrest and to correct some of the damage. If they are successful, the level of p53 declines, and cell division can proceed. RB is essential to normal development; p53 knockout mice, however, develop relatively normally but are highly tumor prone as adults (Donehower et al., 1992).Germline p53 mutations in humans are associated with the LiFraumeni multicancer syndrome (Malkin, 1993; Strong et al., 1992). Although the tumor spectra in men and mice are different, tumor incidence is believed to increase in the absence of the guardian function because the lack of the p53-mediated growth arrest response to DNA damage permits the survival of many genetic variants that would have been eliminated otherwise. Mutated p53 can thus also be considered a destabilizing gene. Other destabilizing genes are discussed in a separate section later. As already mentioned, high levels of p53 can induce growth arrest and elicit apoptosis. Different pathways are involved in these two reactions. Wildtype p53 plays an important role in a t least one major apoptotic pathway that is nonfunctional in mutant p53-carrying and p53-negative cells (Evan et al., 1995). Thus, p53, which is basically regarded as a tumor suppressor gene, can influence the probability of programmed cell death, as can a variety of activated oncogenes, including c-myc.

D. Genes That Influence Programmed Cell Death by Apoptosis More specialized genes regulate the probability of programmed cell death by apoptosis. These genes can also act as important modifiers of tumor development. The potential to elicit a nucleolytic breakdown of DNA is part of the normal developmental and functional repertoire of all organisms. It is essential for morphogenetic remodeling in lower forms. Its genetic control has been thoroughly analyzed in the nematode Caenorhabditis elegans, in which the full set of apoptosis-preventing and apoptosis-triggering genes has been identified (Horvitz et al., 1994). In vertebrates, a vitally important apoptotic function is responsible for the removal of lymphocytes that have not been called on by the cognate antigen within a limited period after they have rearranged their TCR or immunoglobulin genes. Prevention of their apoptotic destruction may favor neoplasia. The most important known apoptosis-modifying gene in mammalians, bcl-2, has been discovered as a result of its illegitimate rearrangement by chromosomal translocation in human follicular lymphoma (Tsujimoto et al., 1984). Most follicular lymphomas carry a reciprocal translocation between the IgH cluster on chromosome 14 and a gene on chromosome 18. In a minority of cases, the light-chain genes are involved on

10

George K l e i n

chromosome 2 or 22. The juxtaposed sequence from chromosome 18 has been identified by cloning the 14;18 translocation breakpoint and designated as bcl-2. It was found that bcl-2 elevates the apoptotic threshold by heterodimerizing with another protein, bax (Oltvai et al., 1993), which promotes apoptosis in the homodimeric form. Transgenic mice that carry immunoglobulin enhancer-bcl-2 constructs have chronic lymphocytosis (McDonnell et al., 1989; Strasser et al., 1991). Moreover, their B cells survive for abnormally prolonged periods in vitro. Some develop slow-growing, low malignant lymphomas late in life. Their human counterpart, follicular lymphoma, is also low grade. Occasionally, a second translocation that juxtaposes myc to an immunoglobulin locus may occur in the human lymphoma (Karsan et al., 1993; Gauwerky et al., 1988). This invariably leads to high-grade lymphoma. The same is found in crosses between bcl-2 and myc transgenic mice. It is easy to rationalize the fact that constitutive activation of a gene that protects from apoptosis but provides no drive for continuous cell division only gives rise to low-grade lymphoma, perhaps triggered by the normal microenvironment that is predestined to stimulate legitimately activated lymphocytes. It is also understandable that myc, a proliferation-stimulating transcription factor, provides an impetus for high-grade lymphoma development. It is more difficult to rationalize the fact that the gene cloned from 11;14 translocations in human low-grade lymphoma, bcl-2, now identified as cyclin D1, induces low-grade rather than high-grade tumors, as mentioned previously (Harris et al., 1995). This may imply that a single cyclin may not be sufficient, in contrast to myc, to push the cell through the whole cycle, without the aid of other factors. This is also consistent with the fact that no other cyclins were found to act as oncogenes, with the possible exception of cyclin E, which is highly expressed in some tumors at an advanced stage of progression (Dou et al., 1996). Other oncogenes that may protect cells from apoptosis include the tyrosine kinase gene v-abl, which acts through a bcl-2-independent pathway (Cotter, 1995; Fernandes et al., 1996). Proteins encoded by DNA tumor viruses (e.g., the Epstein-Barr virus-encoded membrane protein LMPl ), protect their B-cell hosts from apoptosis by elevating cellular bcl-2 expression (Henderson et al., 1991; Okan et al., 1995). As mentioned earlier, p53 is only required for certain apoptotic pathways, such as for radiation-induced apoptosis in lymphocytes. Loss of p53 elevates the apoptotic threshold of lymphocytes exposed to radiation and DNA-damaging agents such as mitomycin C and etoposide, but not to dexamethasoneinduced apoptosis. High levels of bcl-2 expression were observed to block some apoptotic pathways, but not others (Strasser et al., 1994). Protein products of the bcl-2 gene family, which includes bax, bad, and several other genes, interact to form various homodimers and heterodimers that can ei-

Foulds’ Dangerous Idea Revisited

II

ther accelerate or inhibit programmed cell death (White, 1996). Like other p53 target genes, bax is a unique p53-regulated gene. Its induction by genotoxic stress requires not only that p53 is functional but also that the cells are apoptosis “proficient” (Liebermann et al., 1995). Protection against apoptosis by p53 mutation is also important in another oncogene-related context. Activated myc can drive cells not only to proliferate but also to apoptose, depending on the cell type and the environmental conditions (Evan et al., 1992). This may be related to o r even dependent on the fact that activated c-myc can induce p53. The same domains of c-myc are required for transformation as for apoptosis. Mutations in p53 may protect cells driven by activated myc genes (e.g., Burkitt lymphoma cells) from apoptosis. Counterbalancing effects of this type may play a major role in multistage carcinogenesis. Protection of Burkitt lymphoma cells from apoptosis may be important also from another point of view. Most Burkitt lymphoma cells express CD77 (BLA), identified as a cell surface glycolipid (Wiels et al., 1981). The same unusual marker also appears on the surfaces of normal germinal-center centroblasts and centrocytes (Gregory et al., 1987).These cells migrate from the mantle zone of the lymph nodes, rearrange their immunoglobulin genes on the way, and tend to undergo massive apoptosis when they arrive to the interior of the germinal center. This probably reflects the fate of the B cells that failed to meet an appropriate activating complex on the way. A ligand, verotoxin, which binds to CD77, was found to elicit apoptosis. This suggests that CD77 may be a receptor for a natural ligand that triggers apoptosis within the germinal center. With Barbro Ehlin-Henriksson, we recently found that all 9 tested Burkitt lymphoma cell lines that expressed BLA a t a high level carried only mutant but no wild-type p53, whereas only 9 of 14 low CD77 expressors had mutated p53 in single or double dose. Selection for p53 mutations may thus protect cells with the BL phenotype from myc-driven and possibly even physiologic apoptosis. Thus, p53 can play four different roles in tumor development and progression-as an oncogene, a tumor suppressor gene, a destabilizing gene, and an apoptosis-influencing gene-even though its action in the latter capacity may be more indirect and tied to the function of p53 in cell cycle regulation.

111. OTHER DESTABILIZING MUTATIONS CAUSED BY LOSS OF DNA REPAIR FUNCTIONS In 1909, Ehrlich wrote a widely quoted sentence, often regarded as the first formulation of the immune surveillance theory. He stated that the “complex

I2

George Klein

fetal and postfetal development” must involve “many aberrations” and that these would allow tumors to develop in an “enormous frequency,” unless eliminated by the “defense forces of the organism.” Ehrlich meant immune defense. We know now, however, that the first line of defense is not immunity (except against virus-transformed cells), but DNA repair. We should have known it before. The rare recessive condition known as xeroderma pigmentosum affects a single enzyme system responsible for the excision of thymidine dimers from the DNA of keratinocytes damaged by ultraviolet light (see Lambert et al., 1995). Hundreds of carcinomas develop in all daylight-exposed areas of the skin. It has also been known for a long time that inherited chromosome fragility is associated with increased cancer frequencies in ataxia telangiectasia, Fanconi anemia, and Bloom syndrome (reviewed by Kidson, 1980). The multicancer syndrome of humans with inherited p53 mutations and of the p53 knockout mice, mentioned earlier, is a more recent example. The equipment of DNA repair enzymes is faultless in this case, but they are not given time to work properly, owing to the absence of the p53-dependent growth arrest. There was more to come. Yeast geneticists have often pointed to the possible analogies between cancer and the destabilizing mutations in yeast that affect some part of the DNA repair enzyme system (Hartwell and Kastan, 1994). Loeb has also suggested that the mutator phenotype is a prerequisite for multistage carcinogenesis (Loeb and Cheng, 1990). The fidelity of DNA replication, the mechanisms of DNA repair, and the mechanisms of chromosome segregation were suggested as the vulnerable points that may trigger or favor cancer development when damaged. More specifically, spindle errors produce aneuploidy, spindle pole errors may lead to polyploidy, and faulty replication can generate translocations, deletions, and amplification. Most recently, mismatch repair defects were identified as the causative factor in familial hereditary nonpolyposis colon cancer (HNPCC), also called Lynch syndrome (Peltomaki and de la Chapelle, 1996). At least five highly conserved mismatch repair enzymes, known from bacterial and yeast genetics, were alternatively mutated in different families with the syndrome. First, a genetic linkage between HNPCC and a site on chromosome 2p was discovered in some families. Tumors arising in the HNPCC families showed frequent microsatellite instability, characterized by slippage in repeat number. This was taken to indicate that mismatch repair defects were the cause of HNPCC. Later, the human homologue of the bacterial mismatch repair enzyme MutS was found to map to 2p, and mutations in the corresponding human gene MSH2 were shown to be responsible for HNPCC. A second HNPCC locus, MLH1, was discovered by linkage analysis on chromosome 3p. Three further loci have been identified as well. Importantly, sporadic tumors were also found to carry mutations in genes involved in hereditary HNPCC. Mutations in mismatch repair genes induce multiple replication errors

Foulds' Dangerous idea Revisited

13

(RER),leading to microsatellite instability, also called the RER-positive phenotype (Casares et al., 1995). Although most of the tumors from identified HNPCC patients were RER-positive, the frequency of RER-positive tumors largely exceeds that of HNPCC patients. Most of the known HNPCC families carry MSH2 or M L H l mutations that map to chromosome 2 and chromosome 3 , respectively. Most Finnish families have chromosome 3 mutations, indicating a founder effect, whereas chromosome 2 mutations are more common in the United States (Peltomaki and de la Chapelle, 1996). In families carrying germline mutations, the mutation in the second allele occurs during somatic development and is then associated with microsatellite instability. HNPCC families show not only an increased colon tumor incidence but also endometrial, stomach, gallbladder, pancreas, and urinary tract tumors. It is not yet clear whether all HNPCC patients carry a mismatch repair gene mutation or whether all carriers of such a mutation can be regarded as HNPCC patients. Among sporadic tumors, microsatellite instability was found in colon, stomach, and endometrial cancers. Although genetic instability is thus clearly an initiating or a contributory factor to many forms of tumor development, the idiosyncratic nature of the tumor spectra generated by the different repair gene mutations is difficult to explain. It suggests the probable existence of lineage- or differentiationdependent differences in repair gene function or, as in the case of xeroderma, in the mutagenic exposure. Genetically determined variations in the likelihood of different tissues to undergo neoplastic transformation were documented during the early history of experimental cancer research (Heston, 1959). Inbred strains of mice selected for high incidence of mammary carcinoma, lung adenoma, thymic lymphoma, and adrenal carcinoma were found to carry multiple genes that increased the probability of tumor development. These genes were shown to act at several different levels. Selective breeding of other mammalian species, which favors complex phenotypes, such as the race horse or dog breeds with complex behavioral patterns, leads to the fixation of many different genes that contribute to the ultimate phenotype. Each gene contributes to the multigenetically determined trait in an indirect way and is neither necessary nor sufficient by itself. Selective inbreeding of mice for high mammary tumor incidence has led to a similar result. Some of the known susceptibility genes promote mammary cancer development by influencing the hormonal environment. Others make the host permissive for the replication of a cancer-promoting retrovirus (MMTV). Similarly, selection for high leukemia incidence has fixed genes that favor the replication of leukemia-promoting viruses (MuLV)and other genes that impair the specific immune response of the host. Genes that influence retroviral replication or immune surveillance against retroviral proteins are not highly relevant for human carcinogenesis. A third

14

George Klein

set of proven but little-known genes that were found to influence tumor development in the high tumor strains at the level of the target tissue may be more pertinent. The existence of such genes was shown by experiments based on the transplantation of the normal tissue-mammary gland, thymus, lung, o r adrenal-from strains with high or low incidence of the corresponding tumor, into common F1 hybrid recipients. F1 hybrids derived from the crossing of two homozygous strains accept transplants from both parental strains. Having been implanted into the same host, the target tissues of a low- and a high-cancer-prone strain were thus exposed to the same hormonal, viral, and other potential carcinogens. Tumors arose with a significantly higher frequency in the tissues of the high-cancer strain than in the tissues of the lowcancer strain in all four systems. How could genes selectively influence the propensity of a given cell type t o undergo neoplastic transformation in a given host environment? Tissuespecific repair, tissue-specific apoptosis, or the scavenger mechanism might be affected. These may or may not belong to the same category of phenomena as those responsible for the previously discussed idiosyncratic features of tumor incidence in patients with certain repair defects. Another tissue-specific cancer-prone case may be of interest in the same context. We have found that the high susceptibility of the BALB/c strain to plasmacytoma induction by pristane oil is determined, at least in part, at the level of the target cell. Tumors induced in reciprocal chimeras between the susceptible BALB/c and the resistant D B N 2 strain originated exclusively from BALB/c cells in both combinations (Silva et al., 1991 ). The special susceptibility of the BALB/c cell may be due to a high propensity to generate Iglmyc translocations, the common denominator of all mouse plasmacytomas. Resistance is dominant over susceptibility, and nearly all mouse strains, except BALB/c and NZB, are resistant. It is conceivable that BALB/c may be deficient in some mechanism that protects B cells against illegitimate nonhomologous translocations that may occur during physiologic Ig-locus rearrangement. Alternatively, or in addition, the postulated BALB/c defect may affect an apoptotic or scavenger function that can eliminate cells that have committed such an error.

IV. MOLECULAR BIOLOGY OF MULTISTEP CARCINOGENESIS How does the current molecular information fit with the progression rules of Foulds? And how is the responsibility for the emancipation of somatic cells from multiple host control divided among the four worlds of genes? Are there other worlds, in addition to these four? Several mutational steps that contribute to the development and progres-

Foulds' Dangerous Idea Revisited

15

sion of colorectal, prostatic, and breast carcinomas and of malignant gliomas have been identified at the molecular level. The analyses of Vogelstein and his associates and others have defined the sequence of events in colorectal tumors (Fearon and Vogelstein, 1990). They include the mutation of the APC gene on 5q (in polyposis-related colon cancer), activation of K-ras, loss of DCC, a gene that encodes an adhesion protein, loss or mutation of p53, and further changes, reflected by aneuploidization. Some of these changes can be related to the well-defined progression steps from benign polyps through various stages of adenoma, into frank and eventually metastatic cancer. The horizontal study of individual patients has confirmed Foulds' dictum that progression can follow several alternative pathways. The development of astrocytomas into malignant glioblastoma involves a large number of losses, as shown by the loss of heterozygosis (LOH) technique and also by some oncogene activation events (James and Collins, 1992). Particularly noteworthy are the losses of the CDKN2 genes (encoding p16 and p l 5 ) that can inhibit the cyclin-CDK cascade. According to our understanding, p16 binding to CDK4 inhibits the formation of CDK4-cyclin D complexes. Low o r absent expression of p16 or CDK4 amplification favors the phosphorylation of RB, leading to its inactivation. Other changes during astrocytoma progression include the amplification of M D M 2 and of both EGF and PDGF receptors. These changes are akin to oncogene activation events, increasing the drive to cell proliferation. Documented losses involve both p53 and RB and numerous unidentified genes, as indicated by LOH. A particularly significant change, associated with transition to the highest grade of histopathologic malignancy, is a loss of heterozygosis on chromosome 10. Collins and others (Collins and James, 1993; Jiang et al., 1993; Lowe and Ruley, 1993; Strauss et al., 1995; Debbas and White, 1993; Hermeking and Eick, 1994) have pointed out that deregulation of the cyclin D-CDK-CKI-RB pathway may be an obligatory step in tumorigenesis in this and other systems. Examples include overexpression of cyclin D1 in tumors with an intact RB and p16, normal cyclin D1 expression in tumors with inactivated RB, and others. Importantly, overexpression of cyclin D1 has the same effect as loss of p l 6 , and both together have an effect similar to loss of RB. All these changes may converge in targeting the restriction point ( R ) of the cell cycle. Note that p16 and some other inhibitors may be inhibited by DNA methylation, an epigenetic change, rather than by mutation, in the course of tumor progression (Herman et al., 1995; Jones, 1996; Merlo et al., 1995). Three known susceptibility genes have been identified as being responsible for inherited breast cancer proneness: BRCAl, BRCAZ, and p53 (Bieche and Lidereau, 1995; Cannon-Albright and Skolnick 1996; Feunteun and Lenoir, 1996). The spectrum of tumors in the breast and in other organs is different in the syndromes associated with the germline mutations of these three genes and may also differ among different mutations of the same gene.

16

George Klein

BRCAl mutations are associated with both overian and female breast cancers, whereas BRCA2 mutations favor the development of only breast cancer, but in both females and males. Many other chromosome regions that may show LOH in breast tumors have been identified (Cox et al., 1994; Lee, 1995). The RB gene is mutated or deleted in about one third. RB loss appears to be associated with increased invasiveness. Amplification of the c-erb B2 oncogene is a prognostically unfavorable sign (Slamon et al., 1987). The development of prostatic cancer is also associated with numerous genetic changes (Kallioniemi and Visakorpi, 1996). A large fraction of the early-onset and up to 5 to 10% of all prostate cancer patients may carry an inherited germline mutation that facilitates the initiation of the carcinogenic process. Localized prostate cancer may already show LOH in the region. The transition to metastatic prostate cancer may be associated with a ras gene mutation or with RB loss. Another loss may affect a 16q region, known to carry the E-cadherin gene. This is consistent with the fact that the E-cadherinor catenin-mediated adhesion mechanism is changed in about half of all prostate cancers. Inactivation of p53 is relatively infrequent, whereas amplification of the myc-carrying 8q region is common in progressing tumors. An important and readily rationalizable step has been associated with the transition from hormone-dependent to hormone-independent prostatic cancer. In the recurrent hormone-insensitive tumor, the androgen receptor (AR) gene, carried by the X chromosome, may be amplified by up to 10 gene copies per cell. Hormone-independent prostatic carcinomas may also carry AR mutations, which are rare in primary tumors. Hormone therapy may selectively favor the growth of tumor cells that contain mutant o r amplified ARs genes. These and other examples suggest that the loss of known or presumptive suppressor genes may be more common in multistep carcinogenesis and tumor progression than the activation of known oncogenes. This impression may be biased, however, by the fact that genetic losses are more easily identified by the relatively simple LOH technique. Oncogenes d o not lend themselves to screening but rather require the study of each specific gene, one by one. Only gene amplification can be studied by a similarly generalized technique as gene loss, namely comparative genomic hybridization (CGH). Amplification is a relatively late event in progression, however, and affects only a fraction of tumors. I t may be safely concluded that all four key events-destabilization of the genome, activation of dominant oncogenes by mutations or regulatory changes, functional loss of cell cycle checkpoint (suppressor) genes, and mutational or regulatory changes in genes that protect cells against apoptotic death-may contribute to the development of most tumors. The long-standing prediction that the tumors of adults require five to seven mutations is holding up well. Childhood tumors may require fewer changes. The age-

Foulds' Dangerous Idea Revisited

17

incidence curve in retinoblastoma indicates only a single event in the familial cases, in which the germ line allele is already mutated, and two events in the sporadic cases, corresponding to the loss of both RB alleles, as Knudson (1971) originally postulated. Leukemias and lymphomas may be due to a smaller number of changes at all ages than solid tumors, although multistep progression has been clearly demonstrated in Ph 1 ( b c d a b f )translocation-positive chronic myelogeneous leukemia that eventually leads to acute leukemia, designated as blast crisis (Rowley and Testa, 1982). In Burkitt lymphoma, c-myc is constitutively activated in 100% of cases, owing to the juxtaposition of the protooncogene to an immunoglobulin locus by chromosomal translocation (Klein, 1989). In addition, about 60% of the Burkitt lymphoma-derived lines and at least 30% of the in uiuo tumors carry mutated p53 (Gaidano and Dalla-Favera, 1995; Ramqvist et d., 1993; Wiman et al., 1991). This may provide the myc-driven cells with a selective advantage, owing to the protective effect of p53 mutations against apoptosis. Transgenic mice that carry immunoglobulin-enhancer myc constructs provide an experimental facsimile of the Iglmyc translocation (Adams et al., 1985). Such mice develop pre-B and B lymphomas in more than 90% of cases. Most lymphomas are monoclonal, however, indicating that at least one additional change is required for tumor development. With rare exceptions, such as retinoblastoma, most relatively well-known pathways of tumor development thus appear to require mutations in several genes that lead to oncogene activation, suppressor loss, and apoptosis protection. The relative importance of changes in these categories appears to differ between different cell lineages and their developmental stages. Mutations in the first two categories have the derangement of cell cycle regulation as their common denominator. Can the gene categories mentioned account for all phenotypic features of malignant cells? Probably not. Additional properties include the ability of tumor cells to stimulate angiogenesis (Folkman, 1992; 1995), their constitutive telomerase activity (Healy, 1995), and their escape from immune responses (Klein, 1993). With regard to the former, it is particularly clear that the angiogenic switch precedes the progression to invasive cancer in most, if not all, tumor systems. This step may therefore reflect the activation of a yet unknown set of genes (reviewed by Hanahan and Folkman, 1996).

V. STRONG SELECTIVE PRESSURES FAVOR MULTIPLE ESCAPES The mammalian fetus is a protected homograft within the mother's body. Prevention of rejection is absolutely essential for the survival of the species.

18

George Klein

It is therefore not surprising that multiple mechanisms have evolved to pro-

tect the fetus. They include impairment of antigen release, protection from invading maternal lymphocytes, a certain degree of tolerance of the fetal antigens by the mother, specific blocking antibodies, and nonspecific delay of rejection responses. It is no overstatement to say that all conceivable mechanisms sought were actually found. The experimental fantasy of the immunologist appears to be more limited than the protective mechanisms that have been established by selective pressure. A similar point can be made about the escape of the tumor cell from host control. Even if only a single type of host control were analyzed in relation to a single type of tumor cell, multiple escape mechanisms could be identified. This can be exemplified by the escape of the Epstein-Barr virus-carrying Burkitt lymphoma cell from the host immune response. Each of the following mechanisms might have been sufficient to permit the in vivo growth of this cell: downregulation of all immunogenic, virally encoded proteins, except the one that must be regularly expressed to secure the maintenance of the viral episomes (Rowe et al., 1987); sequestration of the latter (EBNA1) from ordinary processing and transport (Levitskaya et al., 1995; Trivedi et a/., 1991); decreased expression of adhesion molecules (Patarroyo et al., 1988);impairment of the TAP-dependent processing mechanism (Khanna et al., 1994; Rowe et al., 1995); and decreased expression of certain major histocompatibility complex class I alleles (Imreh et al., 1995). All of these escape routes have been compounded by the fully fledged Burkitt lymphoma cell that is driven by the translocated c-myc gene and is protected from apoptosis, as already mentioned. It is particularly remarkable that all this modeling can take place in a tumor that develops predominantly in young children! Multiple escape mechanisms are also exemplified by Hanahan’s mouse model of multistage carcinogenesis (Christofori et al., 1995). Transgenic mice that carry the SV40 LT gene under the control of the rat insulin gene regulatory region were generated. This initiates the selective entry of the pancreatic p cells into a multistep pathway that leads them toward islet cell carcinoma. LT is already expressed in the developing pancreas of the 9-day embryo, but hyperplastic islets with a high proliferation index are only seen 3 to 4 weeks after birth. The insulin-like growth factor gene IGF2 is upregulated at this point, indicating that it is functionally involved in the tumorigenic process. Progression does not proceed in the absence of the IGF2 gene. Both the expressed and the imprinted allele of IGF2 are activated during tumor development. The autocrine or paracrine signaling pathway involving IGF2 and its receptors thus appears as an important factor in initiating tumor cell proliferation. The second step is induction of angiogenesis, the activation of the normally quiescent vasculature to proliferate and form new capillaries. The third stage is the development of solid tumors. The three stages are separable. Further progression is accompanied by resistance to the

Foulds’ Dangerous Idea Revisited

19

immune response against the LT-positive cells and by upregulation of telomerase activity in advanced tumors. Suppressor genes localized on mouse chromosomes 9 and 16 are lost at different stages of progression. The early loss of a gene from chromosome 16 appears to be associated with the appearance of angiogenic activity, whereas the loss from chromosome 9 is linked to the progression from the angiogenic stage to a solid tumor. Protection from apoptosis is another identifiable step that occurs independently of the development of angiogenic activity. The selective pressure, the number of changes involved in progression, and the flexibility of the preneoplastic cell population can thus hardly be overestimated.

VI. CONCLUSION Speaking about “Darwin’s dangerous idea,” Dennett ( 1995) writes: If I were to give an award for the single best idea anyone has ever had, I would give it to Darwin, ahead of Newton and Einstein and everyonc else. In a single stroke, the idea of evolution by natural selection unifies the realm of life, meaning and purpose with the realins of space and time, cause and effect, mechanism and physical law. But it is not just a wonderful scientific idea. It is a dangerous idea. . . . My admiration for Darwin’s magnificent idea is unbounded but, I, too, cherish many of the ideas and ideals that it seems to challenge, and want to protect them. There are many . . . magnificent ideas that are . . . jeopardized, it seems, by Darwin’s idea, and they. . . may need protection. The only good way to do this . . . is to cut through the smoke screens and look at the idea as unflinchingly, as dispassionately, as possible.

The same could be said about Foulds’ dangerous idea. Tumor biologists must welcome and indeed embrace complexity. The sociology of scientists may create some problems here, but the clash is much gentler than that between the evolutionists and the creationists. Some of our clinical colleagues and most of the lay public still expect us to come up with “the solution” and “the cure.” It is sometimes said that the investment into cancer research has been a waste, that science did not live up to expectations, or that cancer supports more people than it kills. For those who have followed in Darwin’s and Foulds’ footsteps, there is no return. Even though tumor development is a tiny piece of evolution, compared with our common mother, it is an evolutionary process, with many subtle, seemingly disconnected selective steps, based on an almost infinite cellular variability. Like each species, each cancer is the result of an algoritm. Fortunately, we d o not have to identify every step in each algoritm to stop the process. The appropriate differentiation-inducing signal, inhibition of vascularization, introduction of a powerful suppressor or apoptosis-

20

George Klein

promoting gene, or construction of a potent immune missile may cut the Gordian knot. At the end of the day, we can, in the words of Dennett, “assess the bargain we get when we trade in pre-Darwinian for Darwinian thinking.. . . showing how what really matters to us . . . shines through, transformed but enhanced by its passage through the Darwinian evolution.

ACKNOWLEDGMENTS I express my gratitude to Drs. Klas Wiman, Douglas Hanahan, and George Vande Woude for their critical reading of the manuscript. The editorial help of Marie Bohm is greatly appreciated.

REFERENCES Adams, J. M., Harris, A. W., Pinkert, C. A., Corcoran, L. M., Alexander, W. S., Cory, S., Palmiter, R. D., and Brinster, R. L. (1985). Nature, 318, 533-538. Armitage, P., and Doll, R. (1954). BY. 1. Cancer, 8 , 1-12. Bieche, I., and Lidereau, R. (1995).Genes Chromosomes Cancer, 14,227-251. Biedler, J. L., Melera, P. W., and Spengler, B. A. (1983). In “Chromosomes and Cancer.” (J. D. Rowley and J. E. Ullman, Eds.), pp. 117-138. Academic Press, New York. Bishop, J. M. (1983). Annu. Rev. Biochem., 52,301-354. Boettiger, D., Roby, K., Brumbaugh, J., Biehl, J., and Holtzer, H. (1977). Cell, 11, 881-890. Cannon-Albright, L. A., and Skolnick, M. H. (1996). Semin. Oncol., 23, 1-5. Casares, S., lonov, Y., Ge, H., Stanbridge, E., and Perucho, M. (1995). Oncogene, 11, 2303-2310. Cavenee, W. K., Dryja, T. P., Phillips, R. A., Benedict, W. F., Godbout, R., Gallie, B. I.., Murphree, A. L., Strong, L. C., and White, R. L. (1983). Nature, 305, 779-784. Christofori, G., Naik, P., and Hanahan, D. (1995). Nature Genet., 10, 196-201. Collins, V. P., and James, C. D. (1 993). FASEB j . , 7, 926-930. Cotter, T. G. ( 1995). Leuk. Lymphoma, 18,23 1-236. Cox, L. A., Chen, G., and Lee, E. Y. (1994). Breast Cancer Res. Treat., 32, 19-38. Debbas, M. and White, E. (1993). Genes Deu., 7,546-554. Dennett, D. C . (1995). “Darwin’s Dangerous Idea.” Simon and Schuster, New York. Dittmer, D., Pati, S., Zambetti, G., Chu, S., Teresky, A. K., Moore, M., Finlay, C., and Levine, A. J. ( I 993). Nature Genet., 4.42-46. Donehower, L. A., Harvey, M., Slagle, B. L., McArthur, M. J., Montgomery, Jr C. A., Butel, J. S., and Bradley, A. (1992). Nature 356,215-221. Dou, Q. P., Pardee, A. B., and Keyomarsi, K. ( 1 996). Nature Med., 2,254. Ehrlich, P. (1909).Ned. Tijdschr. Geneeskd., 273. el-Deiry, W. S., Kern, S. E., Pietenpol, J. A., Kinder, K. W., and Vogelstein, B. (1992). Nature Genet., 1,45-49. Evan, G. I., Brown, L., Whyte, M., and Harrington, E. (1995). Curr. Opin. Cell Biol., 7 , 825-834. Evan, G . I., Wyllie, A. H., Gilbert, C. S., Littlewood, T. D., Land, H., Brooks, M., Waters, C . M., Penn, L. Z., and Hancock, D. C. (1992). Cell, 69, 119-128.

Foulds’ Dangerous Idea Revisited

21

Farber, E., and Cameron, R. (1980).Adv. Cancer Res., 31, 125-226. Fearon, E. R., and Vogelstein, B. (1990).Cell, 61, 759-767. Fernandes, R. S., Gorman, A. M., McGahon, A., Lawlor, M., McCann, S., and Cotter, T. G. (1996).Leukemia, lO(Suppl 2), s17-s21. Feunteun, J., and Lenoir, G. M. ( 1 996). Biochim. Biophys. Acta., 1242, 177-1 80. Folkman, J. ( 1992). Semin. Cancer Biol., 3, 65-71. Folkman, J. (1995).Nature Med., 1,27-31. Foulds, L. (1958).J. Chronic Dis., 8, 2-37. Foulds, L. (1969). “Neoplastic Development 1.” Academic Press, New York. Foulds, L. ( 1975). “Neoplastic Development 11.” Academic Press, New York. Gaidano, G., and Dalla-Favera, R. (1995).Adu. Cancer Res., 67, 113-153. Gao, X., Chen, Y. Q., Wu, N., Grignon, D. J., Sakr, W., Porter, A. T., and Honn, K. V. (1995). Oncogene, 11, 1395-1398. Gauwerky, C. E., Haluska, F. G., Tsujimoto, Y., Nowell, P. C., and Croce, C. M. (1988).Proc. Natl. Acad. Sci. U.S.A., 85, 8548-8552. Gregory, C. D., Tursz, T., Edwards, C. F., Tetaud, C., Talbot, M., Caillou, B., Rickinson, A. B., and Lipinski, M. (1987).J.Immunol., 139, 313-318. Hall, M., and Peters, G. (1996). Adv. Cancer Res., 68, 67-108. Hanahan, D., and Folkman, J. (1996).Cell, 86,353-364. Hann, B. C., and Lane, D. P. (1995). Nature Genet., 9,221-222. Harris, A. W., Bodrug, S. E., Warner, B. J., Bath, M . L., Lindeman, G. J., and Adams, J. M . (1995).Curr. Topics Microbiol. Immunol., 194,347-353. Hartwell, L. H., and Kastan, M. B. (1994).Science, 266, 1821-1828. Healy, K. C. (1995).Oncol. Res., 7, 121-130. Henderson, S., Rowe, M., Gregory, C., Croom-Carter, D., Wang, F., Longnecker, R., Kieff, E., and Rickinson, A. (1991). Cell, 65, 1107-1115. Herman, J. G., Merlo, A., Mao, L., Lapidus, R. G., Issa, J. P., Davidson, N. E., Sidransky, D., and Baylin, S. B. ( 1 995). Cancer Res., 55,4525-4530. Hermeking, H., and Eick, D. (1994). Science, 265,2091-2093. Hesketh, R. (1994).Science, 265,2091-2093. Hesketh, R. ( 1 994). Academic Press, New York. Heston, W. (1959).In “Genetics and Cancer.” (Staff of the Univ of Texas M.A.H.a.T., Eds.), pp. 226-240. Univ. of Texas, Austin. Horvitz, H. R., Shaham, S., and Hengartner, M. 0. (1994).Cold Spring Harb. Symp. Quant. B i d , 59, 377-385. Hunter, T., and Pines, J. (1994).Cell, 79, 573-582. Imreh, M. P., Zhang, Q. J., de Campos-Lima, P. O., Imreh, S., Krausa, P., Browning, M., Klein, G., and Masucci, M. G. (1995).Int. J. Cancer, 62,90-96. James, C . D., and Collins, V. P. (1992).Adv. Cancer Res., 58, 121-142. Jiang, W., Zhang, Y. J., Kahn, S. M., Hollstein, M. C., Santella, R. M., Lu, S. H., Harris, C. C., Montesano, R., and Weinstein, 1. B. (1993).Proc. Natl. Acad. Sci. U.S.A., 90, 9026-9030. Jones, P. A. (1996).Cancer Res., 56,2463-2467. Jonkers, J., and Berns, A. (1996).Biochim. Biophys. Acta., 1287,29-57. Kallioniemi, 0. P., and Visakorpi, T. (1996).Adv. Cancer Res., 68,225-256. Karsan, A., Gascoyne, R. D., Coupland, R. W., Shepherd, J. D., Phillips, G. L., Horsman, D. E. (1993).L.euk. Lymphoma, 10,433-441. Kastan, M. B., Onyekwere, O., Sidransky, D., Vogelstein, B., and Craig, R. W. (1991).Cancer Res.. 51,6304-6311. Khanna, R., Burrows, S. R., Argaet, V., and Moss, D. J. (1994). Int. Immunol., 6, 639-645. Kidson, C. (1980).Clin. Haematol., 9, 141-157. Klein, G. (1976).Fed. Proc., 35,2202-2204.

22

George Klein

Klein G. (Ed.) (1982). I n “Advances in Viral Oncology.” Raven Press, New York. Klein, G. (1989). Genes Chromosomes Cancer, 1 , 3-8. Klein, G. (1993). Gene, 135, 189-196. Klein, G., and Klein E. (1957). Symp. Soc. E x p . Biol., 11, 30.5-328. Klein, G., and Klein, E. (1958).J. Cell Conrp. Physiol., 52, 126-168. Klein, G., and Klein, E. (198Sa). Nature, 315, 190-195. Klein, G., and Klein, E. (1985b). Immunol. Today, 6, 208-215. Knudson, A. G. Jr. (1971). Proc. Natl. Acad. Sci. U.S.A., 68, 820-823. Knudson,A. G. (1993). Proc. Nut/. Acad. Sci. U.S.A., 90, 10914-10921. Kok, K., Naylor, S., and Buys, C. (1996). Adu. Cancer Res., 71, 28-93. Konopka, J. B., and Witte, 0. N. (lY85). Biochim. Biophys. Acta., 823, 1-17. Korsmeyer, S. J. (1992). Annu. Rev. Immunol. 10, 78.5-807. Kovacs, G., Erlandsson, R., Boldog, F., Ingvarsson, S., Mullcr-Brechlin, R., Klein, G., and Sumegi, J. (1988). Proc. Nut/. Acad. Sci. U.S.A.. 85, 1571-1575. Lamhert, W. C., Kuo, H. R., and Lambert, M. W. (1995). Dermatol. Clin., 13, 169-209. Lane, D. P. ( 1 992). Nature, 3.58, 15-16. Lee, E. Y. (1995). Senzirz. Cancer Biol., 6, 119-12.5. Levitskaya, J., Coram, M., Levitsky, V., Imreh, S., Steigcrwald-Mullen, P. M., Klein, G., Kurilla, M. G., and Masucci, M. G. (1995). Nature, 375,685-688. Liebermann, D. A., Hoffman, B., and Steinman, R. A. (1995). Oncogene, 11, 199-210. Little, C. D., Nau, M. M., Carney, D. N., Gazdar, A. F., and Minna, J. D. ( I 983). Nature, 306, 194-1 96. Loeb, 1.. A., and Cheng, K. C. (1990). Mutut. Res., 238, 297-304. Lowe, S. W., and Ruley, El. E. (1993). Genes Dei)., 7, 535-545. Malkin, D. ( 1993). Cancer Genet. Cytogenet., 66, 83-92. McDonnell, T. J., Deanc, N., Platt, F. M., Nunez, G., Jaeger, U., McKearn, J. ,!I and Korsmeyer,S. J. (1989). Cell, 57,7948. Merlo, A., Herman, J. G., Mao, I.., Lee, D. J., Gabrielson, E., Burger, I?C., Haylin, S. B., and Sidransky, D. (1995). Nature Med., 1, 686-692. Milner, J. (1994). Semin.Cancer Biol., 5 , 21 1-219. Mintz, B., and Fleischman, R. A. (1 98 I ) . Adv. Cancer Res., 34, 21 1-278. Noda, M. ( 1993). FASEH J., 7 , 834-840. Nowell, P. C. (1976). Sciwce, 194, 23-28. Okan, I.,Wang, Y., Chen, F., Hu, L. F., Imreh, S., Klein, G., and Wiman, K. G. (1995). Oncogene, 11, 1027-1 03 I . Oltvai, Z . N., Milliman, C. I.., and Korsmeyer, S. J. (1993). Cell, 74. 609-619. Patarroyo, M., Prieto, J., Ernberg, I., and Gahmherg, C. G. (1988). In?./. Cancer, 41, 901-907. Peltomaki, P., and de la Chapelle, A. (1996). Adv. Cancer Res., 71, 94-121. Rabbitts, T. H. (1994). Ndtcrre, 372, 143-149. Ramqvist, T., Magnusson, K. P., Wang, Y., Szekcly, L., Klein, G., and Wiman, K. G. ( I 993). Oncogme, 8, 149.5-1 500. Kous, I).,and Beard, J. ( I 935). /. E x p . Med., 62,523-548. Rowe, M., Khanna, K., Jacob, C. A., Argaet, V., Kelly, A., Powis, S., Belich, M., Croom-Carter, D., Lee, S., Burrows, S. R., et a/. (1995). E m J. Iwimrrnol., 25, 1374-1384. Rowe, M., Ilowe, D. T., Gregory, C:. I)., Young, L. S., Farrell, P. J., Rupani, H., and Kickinson, A. B. ( 1 987). E M A O J., 6, 2743-27.5 I . Rowley, J. D., and Testa, J. R. (1982). Adv. Cuncer Res., 36, 103-148. Selivanova, G., and Wirnan, K. G. (1995).Adu. Cancer Res.. 66, 143-180. Sherr, C. J. (1993). CelI, 73, 1059-1065. Silva, S., Sugiyama, H., Babonits, M., Wiener, F., and Klein, G . (1991). Int. J. Cancer, 49, 224-228.

Foulds' Dangerous Idea Revisited

23

Slamon, D. J., Clark, G. M., Wong, S. G., Levin, W. J., Ullrich, A,, and McGuire, W. L. (1987). Science, 235, 177-1 82. Strasser, A,, Harris, A. W., Jacks, T., and Cory, S. (1994). Cell, 79, 329-339. Strasser, A., Whittingham, S., Vaux, D. L., Bath, M. L., Adams, J. M., Cory, S., and Harris, A. W. (1991).Proc. Natl. Acad. Sci. U.S.A., 88, 8 6 6 1 4 6 6 5 . Strauss, M., Lukas, J., and Bartek, J. (1995). Nature Med., 1, 1245-1246. Strong, L. C., Williams, W. R., and Tainsky, M. A. (1992). Am. 1. Epidemiol., 135, 190-199. Toyoshima, H., and Hunter, T. (1994). Cell, 78, 67-74. Trivedi, P., Masucci, M. G., Winherg, G., and Klein, G. (1991). Int. 1. Cancer, 48, 794-800. Tsujimoto, Y., Finger, L. R., Yunis, J., Nowell, 1.' C., and Croce, C. M. (1984). Science, 226, 1 09 7-1 0 99. Weinbcrg, R. A. (1995). Cell, 81, 323-330. Weintraub, H., Beug, H., Groudine, M., and Graf, T. (1982). Cell, 28, 931-940. Weisenburger, D. D. (1991). Leukemia, 5(Suppl I ) , 26-29. White, E. (1996).Genes Deu., 10, 1-15. Wiels, J., Fellous, M., and Tursz, T. (1981). Monoclonal antibody against a Burkitt lymphomaassociated antigen. Proc. Natl. Acad. Sci. U.S.A., 78, 6485-6488. Wiman, K. G., Magnusson, K. P., Raniqvist,T., and Klein, G. (1991).Oncogene, 6 , 1633-1639. Xiong, Y., Zhang, H., and Beach, D. (1993). Genes Deu.. 7, 1.572-1583. Yamasaki, L., Jacks, T., Bronson, R., Goillot, E., Harlow, E., and Dyson, N. J. (1996).Cell, 85, 537-548. Yanagihara, K., Ciardiello, E, Talbot, N., McCeady, M. I,., Cooper, H., Benade, L., Salomon, D. S., and Bassin, R. H. ( 1990). Oncogene. 5, 1 179-1 186. Yokota, J., and Sugimura, T. (1993). FASEB J., 7, 920-925. 7,, 855-865. Zambetti, G. P., and Levine, A. 1. (1993). FASEB I.

This Page Intentionally Left Blank

FOUNDATIONS IN CANCER RESEARCH ~~

Cancer Cells Exhibit a Mutator Phenotype Lawrence A. Loeb The Joseph Gottstern Memorid Cancer Research 1.ahoratory Department of Pathology, Unrurrsrty of Wushmgton. Box 357705 Seattle, Washington 98 19 F-770Y

1. Introduction 11. Historical Perspective 111. Requirement for a Mutator Phenotype A. General Considerations B. Number of Mutations in Human Cancer Cells C. Roles of Clonal Expansion and Somatic Selection D. Coupling of Enhanced Mutagenesis with Somatic Selection IV. Origins of Mutations A. DNA Replication and Mismatch Repair B. DNA Damage and Repair V. Candidate Target Mutator Genes A. DNA Polymerases B. Genes Governing the Cell Cycle C. Genes Involved in DNA Repair and DNA Replication D. Other Target Genes VI. Microsatellite Instability A. Microsatellites B. Microsatellite Instability in Hereditary Nonpolyposis Colon Cancer C. Microsatellite Instability in Sporadic Tumors D. Other Mechanisms for Mutation Accumulation VII. Theoretical and Practical Considerations A. Mutations During Tumor Progression B. Diagnosis C. Prognosis D. Cancer Prevention VIII. Summary and Perspectives References

This review analyzes the concept and evidence in support of a mutator phenotype in human cancer. The large number of mutations reported in tumor cells cannot be accounted for by the low mutation rates observed in normal somatic cells; rather, it must

Advances in CANCER 00hS-L30X/YX $?S.OO

RESEARCH

Copyright 0 199X by Academic Prcs. All rights of reproduction in any form rmervcd.

26

Lawrence A. Loeb

be a manifestation of a mutator phenotype present early during the tumorigenic process. The interaction hetween increased mutagenesis and clonal selection provides a mechanism for the selection of cells with increased proliferative advantage. The concept of a mutator phenotype in cancer has gained considerable support from the findings of enormous numbers of somatic mutations in repetitive sequences in human tuniors. Moreover, cell lines exhibiting microsatellite instability demonstrate an increased mutation frequency in expressed genes. A knowledge of mechanisms that generate multiple mutations in cancer cells has important implications for prevention. For many tumors, a delay in the rate of accumulation of mutations by a factor of two could drastically reduce the death rates from these tumors.

1. INTRODUCTION Evidence indicates that human cancers arise from one or a few of the 500 trillion cells in the human body. It frequently takes 20 years or more for a single cell to multiply preferentially and to constitute a mass that can be clinically detected. At the time of diagnosis, most tumors contain about 1 billion cells and exhibit a high degree of intercellular heterogeneity. Yet even these phenotypically diverse tumors display genetic footprints that reveal their clonal lineage (Fialkow, 1974). Karyotype analysis suggests that most tumors contain multiple chromosomal alterations. Although these chromosomal changes encompass large segments of the genome, they may signify an even larger number of smaller changes that may also be present in tumor DNA. A complete description of nucleotide sequence changes in various types of human cancers-their number, nature, and origins-has yet to be obtained. The genetic and phenotypic heterogeneity exhibited by most human tumors has practical implications. The heterogeneity within each tumor could account for o u r repeated failure to delineate common metabolic alterations in cancer cells as well as the failure of chemotherapeutic agents to eradicate cancers. The cellular heterogeneity within a tumor could account for the ability of tumors to develop resistance to both immunotherapy and chemotherapy. Quantitation of genetic heterogeneity within a tumor could provide a marker for the extent of tumor progression. Because genomic stability is under genetic control, mutations in genes that maintain this stability may provide new targets for cancer treatment and prevention. Two overlapping mechanisms have been proposed to account for the multiple mutations in tumor cells. We proposed that these mutations result from an increase in the rate of mutagenesis in cancer cells, that is, that cancer cells exhibit a mutator phenotype (Loeb et al., 1974). The mutator concept states that tumor evolution is driven by genetic instability with the generation of large numbers of random mutations; there is selection for clones that exhib-

Cancer Cells Exhibit a M u t a t o r Phenotype

27

it malignant properties. Also, Nowell (1976) analyzed chromosomal aber-

rations in tumors and proposed that these mutations result from multiple rounds of clonal selection. Acquired genetic variability permits stepwise selection of variant sublines and underlies tumor progression. Because these concepts were proposed some 20 years ago, it has become increasingly apparent that both increased mutagenesis and clonal selection are important factors in tumor progression. These two mechanisms are not necessarily mutually exclusive and in fact may be tightly coupled. I have argued that the large number of mutations already reported in tumor cells cannot be accounted for by the low mutation rates observed in normal somatic cells, but instead must be a manifestation of a mutator phenotype (Loeb, et al., 1974; Loeb, 1991 [expanded]). The number of mutations observed in individual tumor cells exceeds that which would be expected from the data on spontaneous mutation frequencies in normal cells, even when clonal expansion is taken into consideration. The concept of a mutator phenotype in cancer has gained considerable support from findings of enormous numbers of somatic mutations in repetitive sequences in human tumors (Ionov et al., 1993).Moreover, cell lines exhibiting microsatellite instability also display more than 100-fold increases in mutations in expressed genes (Bhattacharyya et al., 1994; Phear et al., 1996), and these mutations include single base substitutions (Malkhosyan et al., 1996).The central question may no longer be whether cancer cells exhibit a mutator phenotype, but whether the multiple mutations that accumulate in cancers are rate limiting for the carcinogenic process. Is genetic diversity a driving force in tumor progression, or is it simply one of many manifestations that underlie this process? The purposes of this review are to summarize the evidence that the progression of malignancy is associated with increased mutagenesis and to consider whether a mutator phenotype is required for the evolution of tumors. For a historical perspective on chromosomes and cancer, refer to the review by Nowell ( 1 993). The perspective in the current review is at the level of molecular biology. I consider the following concepts: ( 1 ) that cancers exhibit a mutator phenotype; (2) that DNA is turned over in human cells; (3) that there are multiple potential targets for a mutator phenotype; and (4)that the multiple mutations present in cancer cells have important theoretical and practical implications. Many other enticing phenomena, including the regulation of transcription, epigenetic changes, and cell-cell interactions, that may be equally important for the evolution of tumors (Prehn, 1994), are not yet opened to mechanistic experimentation and are not discussed here. Thus, this review is biased and personal in that it focuses on the concept of a mutator phenotype. Evidence supports the hypothesis that sometime during the evolution of a tumor, cancer cells must have exhibited a mutator phenotype. The new mol-

28

Lawrence A. Loeb

ecular techniques have made it possible to dissect the human genome and to delineate genes that are mutated during tumor progression. We hope that methods will be developed to identify random mutations in cells; it may then be possible to quantitate the number of mutations within individual cancer cells. Most important, we need to determine whether an increase in the mutation rate of cancer cells is an early event in the evolution of a tumor and whether the accumulation of mutations limits events in tumor progression.

II. HISTORICAL PERSPECTIVE Our initial hypothesis that cancer cells exhibit a mutator phenotype was based on the infidelity of DNA synthesis in vitro (Loeb et al., 1974). It was grounded on extensive laboratory experimentation that indicated that DNA polymerases and reverse transcriptases frequently incorporate noncomplementary nucleotides (Battula and Loeb, 1974; Fry and Loeb, 1986). Mutations in DNA polymerases might render them even more error prone. Mutant DNA polymerases could be a source of random mutations throughout the genome. We proposed that “infidelity of DNA replication may be responsible for tumor oncogenesis and progression” (Loeb et al., 1974). Our initial hypothesis was supported by multiple phenotypic changes occurring during tumor progression, in particular (a) the continuous evolution of new cell variants; ( b )the increasing ability of tumor cells to escape the host mechanism for regulating cellular proliferation; (c) the progressive accumulation of chromosomal aberrations; and ( d ) resistance to chemotherapeutic agents. We pointed out that infidelity of DNA replication could result from increased misincorporation either by an altered DNA polymerase or by deficits in the repair of misincorporated nucleotides (Loeb et a/., 1974).The concept, diagrammed in Figure l, is that tumor progression proceeds by enhanced mutagenesis resulting from mutations in key enzymes (DNA polymerases) that render them error prone. The concept of a positive feedback loop, in which subsequent mutations occur in many genes, including those that encode DNA polymerases, was in accord with the concept of an error catastrophe involved in aging (Murray and Holliday, 1981; Orgel, 1973). In our model, tumors would grow as a result of a mutation cascade, with the production of new mutants that exhibited a proliferative advantage. From an analysis of chromosomal changes in tumors, Nowell (1976) proposed that tumor progression occurs by clonal evolution, in which a single cell expands by stepwise selection to populate a tumor. Tumor progression results from acquired genetic variability within the original clone, allowing sequential selection of more aggressive sublines. Successive waves of clonal expansion drive tumor progression and result in multiple chromosomal

29

Cancer Cells Exhibit a Mutator Phenotype

VIRUS CHEMICAL CARCINOGEN

1 Neoplastic

Tumor Progression

I Frequency

Proliferation

4 Phenotypes with Selective Value for Proliferation in Host

]

Genotypes

I I

I

Parental

Fig. I Accumulation of genetic errors during tumor progression. The model indicates that an early event in neoplastic transformation is a mutation in DNA polymerase resulting in an increase in mutation during tumor progression. Included is a positive feedback mechanism in which additional mutations could occur in DNA polyrnerases, rendering them more error prone (Loeb, et al., 1974).

changes within individual tumors. Nowell's model for clonal evolution is diagrammed in Figure 2. Human tumors with minimal chromosome changes, such as diploid acute leukemias, are considered to be early in clonal evolution, whereas highly aneuploid solid tumors have undergone multiple clonal selections. Inherent in this model is a mechanism for drug resistance and metastasis based on the selection of mutant variants. Even though the concept of cancer exhibiting a mutator phenotype and evolving by clonal selection was introduced in 1974 to 1976, it only recently has been studied experimentally. The idea that cancer results from random mutagenic events has always met resistance, in part because of the direct implication that carcinogenesis would then be an irreversible mutagenic process-once a cancer cell, always a cancer cell. The reversion of any mutation is an infrequent event. Embryonal carcinoma cells, however, populate tissues on transplantation into a developing blastocyst and result in a mosaic mouse (Illmensee and Mintz, 1976). Thus, there are situations in which a tumor can assume a normal phenotype, but it may still harbor mutations. The attractiveness of the concept of cancers having an epigenetic origin is the

Lawrence A. Loeb

30

mm Diploid Acute Leukernia

Early Solid

E l Human Solid Malignancies

Fig. 2 Model of clonal evolution in neoplasia. Carcinogen-induced change in progenitor normal cell ( N ) produces a diploid tumor cell ( T I ,46 chromosomes) with growth advantage, permitting clonal expansion to begin. Genetic instability of T , cells leads to production of variants (illustrated hy changes in chromosome number, T, to T6). Most variants die, owing to metabolic or immunologic disadvantage (hutched circles); occasionally, one has an additional selective advantage (e.g., T,, 4 7 chrornosomes), and its progeny heconie the predominant subpopulation until a n even morc favorable variant appears (c.g., T4).The stepwise sequence in each tumor differs (being partially determined by environmental pressures on selection) and results in a different, aneuploid karyotype in each fully developed malignancy (T,) (Nowell, 1976).

hope of turning off the carcinogenic stimulus and converting cancer cells to normal cells. In the past 15 years, a major emphasis in cancer research has been on identifying pathognomonic chromosomal changes that delineate particular cancers; more recently, emphasis has been on identifying specific mutated genes within those altered chromosomal segments. This reduction approach has yielded a bewildering number of mutated genes, many of which are associated with specific cancers. Multiple dominant oncogenes and tumor suppressor genes have been identified in most cancers (Weinberg, 1991). Interestingly, in no cancer can a single mutated oncogene be found in 100% of the tumors. I have argued that this multiplicity of mutated oncogenes is a manifestation of earlier mutations that increase mutagenesis (Fig. 3 ) . The reductionist approach may still be fruitful; but, rather than trying to identify specific oncogenes, it might be more productive to focus o n a specific pathway that underlies malignancies. An attractive possibili-

Sources of a Mutator Phenotype TARGET MUTATOR GENES

EXOGENOUS Chemical X-ray U.V.

ENDOGENOUS Oxygen Replication Error

Sporadic Cancers DNA Repair Genes X.P., A.T. Helicases Bloom, Werner Syndrome DNA Polymerases Nucleotide Synthesis Cell Cycle Gene

MALIGNANT PHENOTYPE ONCOGENES Tumor Suppressor Gene Metastasis Genes

. { } Metastasize

Chromosome Segregation

Fig. 3 Mutations during tumor progression. The mutator phenotype hypothesis proposes that genes involved in maintaining chromosomal stability are mutated early during tumor progression. Mutations in these genes would result in other mutations throughout the genome. Mutation in oncogenes would be a later manifestation and result in the ability of tumors to divide, invade, and metastasize.

32

Lawrence A. Loeb

ty is that tumors evolve by random mutations and that changes observed in oncogenes represent later alterations that determine many of the properties of the tumor but may not be the principle driving force in tumor progression. Current research emphasis is on the concept that mutations in mismatch repair genes account for the multiple mutations observed in many human tumors. I would argue, however, that this is but one of the many mutator pathways that will be found to be altered early in tumor progression. I t is likely that neither a mutator phenotype nor clonal selection is adequate to account for the multiple mutations that are present in cancer cells. Instead, there is probably extensive interplay between the two processes. Repetitive selective events during the course of tumor progression may simultaneously select for cells exhibiting a mutator phenotype.

Ill. REQUIREMENT FOR A MUTATOR PHENOTYPE A. General Considerations Many studies indicate that the spontaneous mutation rate of human cells is about lo-’ mutations per single copy gene per cell generation (Meuth, 1990) or, for an average gene, 1.4 X l O - ’ O mutations per base pair per cell generation (Loeb, 1991). For mutations in autosomal genes to result in a change in phenotype, two independent events may be required (Phear et al., 1996). Based on the total number of cell divisions during an average human life span, we estimated that the spontaneous (background) mutation rate would produce at most two or three mutations in each normal cell. Epidemiologic studies, however, suggest that many events are required for the production of a cancer (Armitage and Doll, 1954; Renan, 1993; Stein, 1991). Although plots of the incidence of retinoblastoma and Wilms’ tumors as a function of age indicate a two-hit mechanism and confirm Knudson’s hypothesis, similar plots of age versus incidence for adult tumors suggest a large number of events. Estimates of the number of events associated with adult tumors vary: 3 for early-onset bone tumors, 6 for cancer of the cervix, and as many as 12 for cancer of the prostate (Renan, 1993). If we assume that these events correspond to mutations, then the spontaneous mutation rate can account for only two or three of these events. If it is true that cancers arise only in a fraction of the cell population (i.e., stem cells), then even fewer cancer-causing mutations could result from normal mutation rates. Furthermore, many tumors contain as many as 50 chromosomal rearrangements. This discrepancy between the rarity of mutations in normal cells and the large number of mutations in cancer cells led us to propose that acquired

Cancer Cells Exhibit a Mutator P h e n o t y p e

33

genetic instability o r a mutator phenotype must be an early event in the evolution of a tumor (Loeb, 1991).

B. Number of Mutations in Human Cancer Cells During the past 20 years, there has been an increasing number of reports on chromosomal alterations in human tumors (Balaban et al., 1986; Nowell, 1993) that could result from a mutator phenotype. First, karyotype analysis revealed multiple chromosome translocations (Erikson et al., 1992; Rowley, 1975) and deletions (Yunis and Sanchez, 1975) as well as the presence of amplified portions of chromosomes (Biedler and Spengler, 1976). Second, multiple chromosomal alterations have been reported in specific tumors. Twenty-five to 50% of colon adenocarcinomas have been shown to contain more than nine chromosomal alterations (Fearon and Vogelstein, 1990). Third, loss of heterozygosity is a common feature in tumors. It occurs as a result of the deletion of a large piece of a chromosome, and its presence in tumor DNA has been considered a marker for the deletion of tumor suppressor genes. In ovarian cancers, 23 different chromosomal segments have been shown to exhibit loss of heterozygosity a t a frequency of greater than 3 0 % (Chigira et al., 1993). Fourth, multiple mutations are present within single human cancer cells. Cancers exhibiting multiple chromosomal alterations include small cell cancer of the lung (Naylor et al., 1987), adenocarcinoma of the colon (Fearon and Vogelstein, 1990; Stanbridge, 1990),malignant melanoma (Balaban et al., 1986), and glioma (James et al., 1988). Large deletions and losses of entire chromosomes could be the result of mutations in genes that govern the overall mechanism of chromosomal segregation, including DNA helicases, topoisomerases, and genes not yet identified that are involved in recombination, spindle formation, and chromosomal partitioning during cell division. In addition to multiple chromosomal changes reported in tumors, gene amplification may also be a diagnostic marker for malignant cells. Gene amplification is an increase in the number of copies of a particular gene at a specific locus and thus in a broad sense can be classified as a mutation. Tlsty and colleagues reported that resistance to the drug PALA occurs by gene amplification. It is undetectable ( < l o p 9 ) in primary diploid cells and ranges from l o p 3 to lo-” in transformed cells (Tlsty et al., 1992; Livingstone et al., 1993). Moreover, in cell fusion experiments, it behaves as a recessive trait, suggesting that it is suppressed in normal diploid cells. The fact that gene amplification occurs at high frequency in neoplastic cells from diverse sources and is undetectable in the corresponding normal diploid cells provides strong evidence for genetic instability as a hallmark of neoplastic cells in culture (Otto et al., 1989).

34

Lawrence A. Loeb

Frequently, a sequential order of chromosomal alterations occurs during tumor progression. Vogelstein and colleagues ( 1988) delineated an ordering of mutations in colon cancer that parallels the evolution of benign polyps to invasive carcinomas. Among the genes mutated are APC (the gene defective in the hereditary adenomatous polyposis coli), K-ras (a gene involved in signal transduction), DCC (a gene that encodes a protein involved in cell adhesion), and p.53 (a transcription factor involved in apoptosis and in governing checkpoints during the cell cycle). Similar studies that assign an order to chromosomal mutations during tumor progression have been reported for gliomas (James et a/., 1988), carcinomas of the breast and lung (Birrer and Minna, 1988) and malignant melanomas (Balaban et al., 1986). These alterations have been detected cytologically and involve large segments of chromosomes. The ordering of mutations with time could reflect the selection of clones with different phenotypes during tumor progression. These mutations probably are a reflection of an even much larger number of smaller mutations, such as small deletions, rearrangements, additions, and single base substitutions, each of which could involve one or a few nucleotides (Loeb, 1991).Thus, the multiple chromosomal aberrations in cancer may be only the tip of an iceberg. As we determine the sequence of nucleotides in tumor DNA, we likely will discover tens of thousands of mutations in each cancer cell (Loeb and Cheng, 1990).

C. Roles of Clonal Expansion and Somatic Selection Clonal expansion of some cells in a tumor is likely to account for an increased number of mutations by increasing the number of cell replications during tumor progression. A mutation confering a proliferative advantage would result in an expansion of cells with that mutation within the tumor. Each successive round of clonal expansion would yield a large number of cells having N mutations, and thus increasing the probability of the N + 1 mutation. We have estimated that in the absence of clonal expansion, the spontaneous mutation rate in normal human cells can account for only two or three mutations. If we assume that each mutation confers a proliferative advantage that leads to a clonal expansion to los cells before the next mutation, then the chance of a cell getting two mutant genes within that clone would be increased from (10-7)(10-7) to (10p7)(10-7X lo5).Thus, without an increase in the mutation rates, successive rounds of clonal selection could account for as many as five o r six mutagenic events. In the absence of a mutator phenotype, clonal selection could account for the increased incidence of some cancers as function of age (Armitage and Doll, 1954). Despite this theoretical argument, it is unlikely that clonal expansion in the absence of enhanced mutagenesis accounts for the multiple mutations in

Cancer Cells Exhibit a Mutator Phenotype

35

cancer cells. First, each mutation would have to exhibit a proliferative advantage. Even a mutation in one of the two alleles of a tumor suppressor gene would have to confer a proliferative advantage to cells harboring that mutation. Second, most mutations found in cancer cells are recessive for tumorigenesis (Harris, 1988), and thus it is unlikely that a single mutation would result in a proliferative advantage. In tumor suppressor genes, usually both alleles have to be inactivated for expression of the malignant phenotype (Knudson, 1971). Third, for the mutant clones to populate the tumor rapidly, the clones must have an exceptionally high selective advantage, or they must prevent DNA synthesis in other cancer cells within the tumor. Most important, although clonal expansion could conceivably account for the five or six mutations in specific cancers, it cannot account for the much larger number of mutations that are likely to be present in many cancers. On the other hand, mutations in genes involved in maintaining chromosomal integrity could occur early during tumorigenesis, accounting for the multiple mutations in individual cancer cells. It is still conceivable that the multiple mutations in cancer cells have nothing to d o with malignant progression, but instead are a parallel occurring event (i.e., the accumulation of mutations is not rate limiting for tumor growth or tumor progression). We have postulated that a mutator phenotype in cancer cells could account for a large number of mutations. Among the random mutations generated would be some that yield a proliferative advantage and that result in clonal expansion and others that are responsible for the cancer phenotype. A mutator phenotype alone does not account for the clonal nature of most human cancers, and thus selection must also be an important determinant in tumor progression.

D. Coupling of Enhanced Mutagenesis with Somatic Selection It is likely that the large number of mutations generated in cancer cells is the result of both enhanced mutagenesis and somatic selection. Mutations may not always be disadvantageous. Competition experiments with bacteria indicate that there are conditions in which bacteria with a higher mutation frequency have a growth advantage over the wild type (Cox, 1976).The question of whether selection for mutants would concomitantly select for mutators has been addressed by Miller, Mao, and colleagues in a series of telling experiments in Escherichiu coli (Ma0 et af., 1996; Mao et ul., 1997; Miller, 1996).The selection of spontaneous Lac-positive revertants in a Lacnegative population resulted in a parallel enrichment of mutator mutants. In this system, reversion occurred by frameshift mutagenesis, and each of their mutators was deficient in mismatch repair, a system that corrects frameshift mutations. The implication is that a fraction of spontaneous mutations is the

36

Lawrence A. Loeb

result of mutator mutations; the selection for mutants would increase the fraction of mutators in the population. Moreover, these investigators demonstrated that exposure to a mutagen followed by multiple selections can result in a population of cells consisting nearly entirely of mutators. This situation could be analogous to repetitive rounds of clonal selection in tumor progression. Selection for invasion, nutrients, and drug resistance during the life of a tumor would also result in selection for mutators. As a result, somatic selection could be concatenated with the selection of mutators.

IV. ORIGINS OF MUTATIONS In cells, DNA is damaged by both exogenous and endogenous processes. DNA damage, if unrepaired, can cause mutations when the damaged DNA is copied at the time of DNA replication. The fact that spontaneous mutations occur infrequently in human cells does not necessarily imply either that DNA synthesis is exceptionally accurate or that DNA is not damaged during the life of a cell. Although rare mutations are needed in germline cells for a species to respond to environmental changes, there is no obvious scenario indicating that somatic mutations would be advantageous. One can envision an overall DNA homeostasis, encompassing a multitude of mechanisms, that evolved to govern the mutation rate in normal somatic cells. Either an increase in errors during DNA replication or a decrease in the efficiency of DNA repair could shift the balance of the system and result in an increase in mutations.

A. DNA Replication a n d Mismatch Repair During each cell cycle, every human cell replicates 6 x lo9 nucleotides of nuclear DNA. Based on a background mutation rate in human cells of about 1.4 x mutations per base pair per cell generation, we have estimated that each normal cell replicates its genome with about one error per cell generation (Loeb, 1991). The frequencies of misincorporation by DNA polymerases in vitro is much higher, however. Studies on the fidelity of DNA synthesis in vitro indicate that the frequencies of misincorporation by homogeneous eukaryotic DNA polymerases vary from 1 in 5000 for DNA polymerase-P (Kunkel, 1985; Kunkel and Loeb, 1981) to 1 in l o 7 for DNA polymerase-€ (Kunkel, 1992). Even though other proteins may function to increase the accuracy of DNA synthesis, it is unlikely that the accuracy of the DNA-replicating complex is adequate to account for the low frequency of observed spontaneous mutations in cells. Therefore, human cells must

Cancer Cells Exhibit a Mutator Phenotype

37

possess powerful mechanisms to excise noncomplementary nucleotides that are incorporated as a result of DNA polymerization, a process referred to as mismatch correction. The genetics and biochemistry of mismatch correction have been elegantly delineated in E. coli (Modrich, 1987), and homologous proteins have been identified in yeast and human cells (Modrich and Lahue, 1996), although we still lack a n understanding of the biochemistry and mechanisms to distinguish between the template and newly synthesized, error-containing strand in eukaryotic cells. The enhanced mutation rates exhibited by cancer cells with deficiencies in mismatch correction have provided the strongest evidence for a mutator phenotype.

B. DNA Damage and Repair DNA in both dividing and nondividing cells is subjected to a variety of modifications. DNA is damaged by both endogenous (Loeb and Cheng, 1990)and exogenous sources (Ames etal., 1990). Even small chemical modifications of nucleotides in DNA, such as the addition of methyl and ethyl groups to either purine or pyrimidine bases, alters the coding properties (Singer and Essigmann, 1991; Singer and Grunberger, 1983).Thus, damage to DNA, if not repaired, is likely to cause misincorporations at the time of DNA replication. Many environmental chemicals damage DNA; epidemiologic data indicate that many chemicals that damage DNA are carcinogens (Ames, 1983). In addition, normal cellular metabolic processes generate reactive chemical intermediates with the potential to damage DNA, these might also be a source of spontaneous mutations and cancers. Among such endogenously generated reactions is the hydrolysis of the glycosylic bond in DNA by water molecules. This results in depurinations, with the formation of an abasic site in DNA. Based on extrapolations from rate constants obtained in vitro, it has been estimated that depurination of DNA is an exceptionally frequent occurrence; about 10,000 depurinations occur per cell per day (Lindahl and Nyberg, 1972). The resultant abasic sites are mutagenic (Loeb, 1985; Schaaper et al., 1983). Deamination of cytosine to thymidine residues in DNA is less frequent, but the end product is a change in the nucleotide sequence of DNA (Lindahl and Nyberg, 1974). Equally extensive is damage by oxygen free radicals and related active molecules. Based on the urinary excretion of two major products of DNA damage by oxygen free radicals, thymine glycol and 8-hydroxyguanosine, it has been estimated that reactive oxygen species introduce 20,000 lesions in DNA in each cell per day (Cathcart et al., 1984; Shigenaga et al., 1989). Many of the alterations in DNA produced by oxygen free radicals have been shown to change the coding properties of nucleotides in DNA during copying by DNA polymerases (Basu etal., 1993; Cheng et al., 1992; Shibutani et al., 1991). Other sources

38

Lawrence A. Loeb

of endogenous DNA damage being investigated include: methylation (Erlich and Wang, 1981; Shen et al., 1992), attack by lipids (Chaudhary et al., 1994), and glycosylation (Bucala et al., 1984). Each of these processes is likely to result in many altered bases in DNA. Considering the multiplicity of lesions produced in cellular DNA and the compact structure of chromatin, it is unlikely that all lesions in DNA would be repaired before DNA replication. Thus, DNA damage by both endogenous and exogenous processes is likely to contribute to mutagenesis.

V. CANDIDATE TARGET MUTATOR GENES Mutations can be generated by many processes, including errors in DNA replication and DNA damage by endogenous and exogenous sources. Conversely, diminution in the efficiency of mismatch repair or DNA damage repair can also result in increased mutagenesis. Thus, a t a minimum, mutations in genes that govern DNA synthesis and DNA repair could yield cells that exhibit a mutator phenotype. In a larger sense, many other genes are involved in recombination, transcription, cell cycle control, and chromosomal segregation that ensure the genetic stability in normal cells; mutations in these genes could also confer on cells a mutator phenotype (Cheng and Loeb, 1997). As examples, I consider recent studies on a few of the potential target mutator genes.

A. DNA Polymerases Misincorporation by DNA polymerases may be a major contributor to spontaneous mutations. Presumably, misincorporation by mutant DNA polymerases could exceed the capacity of the mismatch repair system. Schaaper (1993)demonstrated that the spontaneous mutation rate of E. coli harboring antimutator DNA polymerases is two- to three-fold lower than that of wild-type cells. This implies that half of spontaneous mutations in E . coli are the result of errors by DNA polymerases. Mutations in DNA polymerases have been shown to alter the fidelity of DNA synthesis and to increase the mutation rate of cells (Beard et al., 1996; Copeland et al., 1993; Sweasy and Yoon, 1995). Most dramatically, a single substitution of an alanine for an arginine at position 283 of DNA polymerase+ decreases the fidelity and catalytic efficiency of DNA polymerase+ by 160-fold and 5000fold, respectively (Beard et al., 1996). Even though it is unlikely that an enzyme with such a marked reduction in catalytic efficiency could function in vivo, these experiments demonstrate how mutations in DNA polymerases can dramatically affect the fidelity of DNA synthesis.

Cancer Cells Exhibit a Mutator Phenotype

39

Only fragmentary studies have been reported o n DNA polymerase genes in cancer cells. Most of these studies have been on DNA polymerase-P. DNA polymerase-P is located on the short arm of chromosome 8; in cancers, this segment is subjected to loss of heterozygosity and microsatellite instability (Patel et al., 1994). Alterations in DNA polymerase-@ mRNA was detected in 5 of 6 cases of colon cancer (Wang et al., 1992),in 2 of 12 cases of prostate cancer (Dobashi et a/., 1994), and in 4 of 24 cases of bladder cancer (Matsuzaki et al., 1996). The DNA polymerase+ alterations in colorectal cancers consist of a 87-base-pair deletion in the mRNA transcript and are likely to encode an inactive enzyme as a result of an alternative slice site. If this mutation is a dominant negative, however, it might yield a mutator phenotype by preventing DNA repair by the unaltered allele. Among polymerasep mutations in bladder cancer are missense mutations that could alter the catalytic properties of the enzyme. Because the polymerase+ mutations in bladder cancer were observed in early tumors, it has been suggested that polymerase-P mutations contributed to the development of these cancers (Matsuzaki et af., 1996). Thus, polymerase-P remains a candidate target for a mutator phenotype. Mutations in other DNA polymerases that synthesize larger portions of the genome during DNA replication are more likely to generate a mutator phenotype in cancer cells. The only other polymerase mutant that has been identified is a mutation in the conserved exonuclease Ill motif of DNA polymerase-6 in DLDlIHCTlS, a human colon cell line. I t is not known if this mutation reduces proofreading o r enhances slippage by the polymerase (Minnick and Kunkel, 1996). Studies on infidelity of DNA synthesis in cancers have been limited mainly to experiments with crude extracts. These studies measure the frequency of misincorporation by DNA polymerases using either polynucleotides or plasmid DNA as a template. In early studies, we demonstrated that cell extracts from human leukemic cells misincorporated noncomplementary nucleotides more frequently when copying polynucleotides than did similar extracts from phytohemagglutinin-stimulated normal lymphocytes (Loeb et al., 1974). Recently, experiments carried out using chemically induced mouse primary thymic lymphomas also demonstrated a reduction in fidelit y using natural DNA templates (Kubota et al., 1995). Control experiments in both studies provided evidence that the increase in errors is not correlated with the presence of terminal deoxynucleotidyl transferase that adds nucleotides onto DNA without instruction from the template strand. In contrast, studies of the fidelity of DNA synthesis using the SV40 replication complex failed to reveal a difference between normal and malignant cells in culture (Boyer et al., 1995) and thus d o not support the concept of an errorprone polymerase in cancer cells. In principle, fidelity studies with purified DNA polymerases from normal and tumor tissues should provide definitive information on whether any of these enzymes is mutated. Mutant enzymes,

40

Lawrence A. Loeb

however, are likely to be thermosensitive (Munir et al., 1993), exhibit different chromatographic properties, and may be lost during extensive purification. Thus, the most definitive studies in this area will be those that identify alterations in the DNA sequence of polymerase genes in cancer cells and then demonstrate that these same alterations enhance the frequencies of misincorporation by the encoded DNA polymerase. In the absence of these studies, it is difficult to determine whether mutations in DNA polymerase result in a mutator phenotype in specific cancers.

B. G e n e s Governing the Cell Cycle Genes that control the cell cycle are likely targets for the production of a mutator phenotype (Hartwell and Kastan, 1994). Mutations in G,/S checkpoint genes allow DNA replication in the presence of unrepaired lesions (Crook et al., 1994) and thus result in increased mutagenesis. For the purpose of this review, I focus on p53. Refer to recent reviews on the cell cycle for possible mutations in other proteins (Stillman, 1996; Elledge, 1996). In human cancers, p53 has been reported to be mutated more frequently than any other single gene (Greenblatt et al., 1994). It encodes a multifunctional protein involved in DNA repair, cell cycle control, and genomic stability (Harris, 1996). Mutations in p53 have been postulated to permit cells to enter S phase before the completion of DNA repair. As a result, unrepaired lesions could cause mutations during DNA replication. Also, mutations in p53 in murine cells can destabilize the genome and result in amplification of drug-resistant genes (Livingstone et al., 1992) and presumably other genes. Even though functional studies on mutants in p53 have yielded a wealth of information, we still lack knowledge of a definite role for p53 in malignant transformation. The p53 mutation spectrum, however, has provided unique clues about the relation of DNA damage to mutations in tumors (Weinstein, 1995). In three common human cancers, p53 mutations are the same type as those of the agents causally associated with that tumor. In liver cancer, in geographic areas with a high level of exposure to aflatoxin, p53 mutations are localized to codon 249 and are predominantly G+T transversions. The same G+T transversions are produced in livers of animals exposed to aflatoxin. In human skin tumors (basal and squamous cell carcinomas and melanomas), presumably caused by ultraviolet ray exposure, p53 mutations include tandem CC+TT substitutions. These mutations have been reported only after exposure of DNA to ultraviolet irradiation (Miller, 1985) or to oxygen free radicals (Reid and Loeb, 1992). In lung cancers, three codons in p53 are mutated a t an unusually high frequency (Hollstein et al., 1996). One of these codons, 157, is not a frequent site for mutations

Cancer Cells Exhibit a Mutator Phenotype

41

in other tumors. Mutations at this hot spot in lung cancers are predominantly G+T transversions. The distribution of adducts and mutations in p53 in HeLa cells or bronchial epithelial cells after exposure to the tobacco carcinogen, benzo[a]pyrene diol epoxide, corresponds to the distribution in lung cancers. These studies imply that adduct formation in p53 is targeted by some unknown mechanism and is an early and perhaps initiating event in the carcinogenesis by these agents. Before designating mutations in p53 as the primary event in the pathogenesis of these cancers, however, we should recall the situation with the ras oncogene in methylnitrosourea-induced breast cancer. Without exception, G+A transitions were observed in the induced tumors; this mutation mimics the mutagenic specificity of methylnitrosourea in vitro (Burns et al., 1988) and thus provides compelling evidence that the tumors result from the corresponding mutation in the ras oncogene (Barbacid, 1987). The presence of the mutation in normal cells and the dependency of the tumors on hormonal stimulation, however, have brought this concept into question (Cha et al., 1996). The mutation spectrum of p53 has also provided support for a mutator phenotype in cancer. Strauss (1996, and unpublished results) analyzed the reported sequences of p53 mutations in human cancers. The percentage of silent mutations (i.e., mutations that encode an amino acid identical to the wild type) in mutated p53 is at least 20-fold greater than anticipated. There is no evidence to indicate that these silent mutations are selected, nor d o these mutations correspond to the p53 polymorphisms in the human population. Therefore, these results provide strong evidence that p53 is hypermutable, at least a t some stage during the progression of these tumors. A demonstration of an increase in silent mutations in other genes or in pseudogenes would provide exceptionally strong evidence in support of a cancer mutator phenotype.

C. Genes Involved in DNA Repair and DNA Replication Several rare inherited human diseases are characterized by deficits in DNA repair and also exhibit unusually high incidences of cancer (Cleaver and Kraemer, 1989). There appears to be direct associations among the source of exogenous DNA damage, the mutation that renders the DNA repair pathways ineffective, and the type of mutations produced in the resultant tumors. These associations may provide evidence that deficits in DNA repair and mutations are the initiating events in human cancer. The two most extensively investigated mechanisms for the repair of chemically modified DNA are nucleotide excision repair and base excision repair pathways (Hanawalt et a/., 1979). Nucleotide excision repair removes a diverse spectrum of chemically modified nucleotides in DNA. Genes involved

42

Lawrence A. Loeb

in this repair pathway are mutated in xeroderma pigmentosum, and cells harboring these mutations are sensitive to ultraviolet irradiation and carcinogens. Considering the wide spectrum of modified nucleotides repaired by this mechanism, it is surprising that susceptibility in xeroderma pigmentosum is restricted to skin cancer. Based on this anomaly, it has been suggested that many common human cancers are not caused by the DNA lesions that patients with xeroderma are unable to repair (Cairns, 1981). Base excision repair appears to be more specific for certain modifications. Mutations in this pathway have not been identified in human diseases. Studies are in progress in many laboratories to knock o u t these genes in mice and to determine if there is an increased incidence of malignancies. Despite the importance of these inherited diseases in defining the associations between DNA damage and cancer, mutations in these DNA repair genes have not been associated with common human cancers. Possible exceptions are the increased incidence of breast, lung, bladder, and stomach cancer in blood relatives of patients with ataxia telangiectasia (Swift et al., 1991). The DNA repair protein Oh-methylguanine-DNA methyltransferase (MGMT) removes highly mutagenic 06-alkyl groups from Oh-alkylguanine in DNA. It has been reported that histologically normal brain adjacent to most primary human brain tumors lacks detectable levels of this important DNA repair activity (Silber et al., 1996). These authors suggested that the mutator phenotype arising from lack of M G M T may be involved in generation of the genomic alterations characteristic of gliomas.

D. Other Target G e n e s Less well defined targets for the generation of a mutator phenotype are DNA helicases. These enzymes separate the two strands of a DNA helix ahead of the DNA replicating proteins. They are involved in a variety of DNA synthetic processes, including DNA replication, repair, transcription, and recombination. Decreased strand separation could cause pausing by DNA polymerases and increased misincorporation (Fry and Loeb, 1994). Two inherited diseases, Werner and Bloom syndromes, have been shown to contain mutations in genes that have all the signature nucleotide motifs of a helicase (Ellis et al., 1995; Yu et al., 1996). In both syndromes, cells from patients exhibit spontaneous genetic instability in culture (German et al., 1977; Fukuchi et al., 1989). Patients with Bloom syndrome are immunologically deficient and usually die of leukemia or lymphomas in early adulthood (German et al., 1977). Patients with Werner syndrome prematurely exhibit nearly all of the stigmata of aging and have a wide spectrum of unusual tumors at a young age (Epstein et al., 1966; Goto et al., 1996). Many proteins participate in DNA replication. DNA polymerases are just

Cancer Cells Exhibit a Mutator Phenotype

43

one component of a complex DNA-replicating apparatus to ensure the fidelity of DNA synthesis. Among these proteins is proliferative cell nuclear antigen (PCNA), a processivity factor for DNA polymerase-6 that has been shown to interact with human mutL (Umar et al., 1996). These findings suggest a direct linkage between complexes of proteins that replicate DNA and that repair errors in DNA replication. Based on these associations, it is likely that mutations in PCNA would alter the fidelity of DNA synthesis. In a wider context, it is reasonable to assume that many of the genes that function to maintain genomic stability in normal cells would be targets for a mutator phenotype. Among these candidate target genes are those involved in nucleotide metabolism, recombination, and chromosomal segregation as well as genes yet to be identified that determine chromosomal localization, gene copy number, and the mitotic cycle (Cheng and Loeb, 1993; Hartwell and Kastan, 1994).

VI. MICROSATELLITE I NSTABlLlTY Until recently, experimental support for the concept of a mutator phenotype has been equivocal. There was insufficient experimental evidence to establish that any of the postulated target genes are mutated in tumors or that these mutations generate a mutator phenotype. Analysis of mutation frequencies in cell lines derived from human tumors yielded variable results, as have most studies on the error rates of DNA polymerases or crude DNA replicating complexes (Loeb, 199 1).This situation has dramatically changed. Recent studies on microsatellite instability have provided strong evidence for a mutator phenotype in human cancers. Microsatellite instability has been detected in colon cancers, both sporadic and those associated with hereditary nonpolyposis coli, and it was then shown to be a general property of a variety of human malignancies.

A. Microsatellites Within the human genome are a multitude of repetitive nucleotide sequences. Among these are as many as 50,000 noncoding microsatellites, consisting of repeat units 1 to 4 bases long, with a n average of 15 to 60 tandomly linked repeat units. For the most part, they are found in introns and presumed to be noncoding. The most frequent repetitive sequence is a homopolymeric tract of adenosine residues [(A),];the second most frequent are alternating cytidine and adenosine residues [(CA),] (Weber and May, 1989). The variations in these sequences among individuals has been extensively ex-

44

Lawrence A. Loeb

ploited for gene mapping. The pattern of sequences within all tissues in an individual is invariant. Within tumors, however, there are marked variations in many of these sequences generated by slippage during copying by DNA polymerases. In vivo, most of the slippage errors by DNA polymerases are corrected by the mismatch repair system. In tumors, this mechanism is defective, and as a result, variations in the lengths of microsatellites in tumor DNA occur frequently in association with mutations in mismatch repair genes.

B. Microsatellite Instability in Hereditary

Nonpolyposis Colon Cancer In seminal papers, Peinado and colleagues ( 1 992), and Ionov and associates (1993) detected multiple deletions in repetitive nucleotide sequences in DNA by comparing DNA fingerprints in colorectal tumors using polymerase chain reaction (PCR) primers with nucleotide sequences that were arbitrarily chosen. Because this method screens a large fraction of the human genome, the authors were able to conclude that these tumors carry more than 100,000 such mutations. They surmised that these deletions may represent an inherited form of colon cancer and might be mediated by a reduction in the fidelity of DNA replication or a decrease in the efficiency of DNA repair. Moreover, the presence of these mutations in adenomas suggested that these mutations were an early event in the carcinogenic process. At the same time, it was reported that many types of repetitive sequences within the genome, including specific microsatellites, were altered in hereditary nonpolyposis colon cancer (HNPCC) (Thibodeau et al., 1993). It was recognized rapidly that expansion of repetitive sequences in DNA could be the result of defects in the mismatch repair pathway. A historical perspective of these findings has been presented (Marra and Boland, 1995). Four different mismatch repair genes have been demonstrated to be mutated in about half of the different HNPCC families. Sixty percent of these HNPCC families have a mutation on chromosome 2p (Aaltonen et al., 1993; Peltomaki etal., 1993).The chromosome 2p HNPCC protein was soon identified as a homolog of the bacterial mutS protein, involved in mismatch repair in E. coli, and was named MSH2 (human mutS homolog) (Fishel et al., 1993). Thirty percent of these HNPCC families segregate with a locus o n chromosome 3p (Lindblom et al., 1993) that has been established to encode a different mismatch repair homolog, MLH1, which is homologous to the bacterial mutL protein and the yeast MLH protein (Bronner et al., 1994; Papadopoulos et al., 1994). Also, the remaining mutations have been detected in the human mutL homologs, PMSl and PMSZ (Nicolaides et al., 1 994); PMSZ has been shown to be functionally involved in mismatch repair (Risinger et al., 1995a).In each of these

Cancer Cells Exhibit a Mutator Phenotype

45

families, one mutant allele is inherited, and the other arises as a result of differences in somatic mutation rates in different tissues. Systematic biochemical studies of E. coli have established the function of each of these genes in E . coli mismatch repair. The system has been reconstituted in vitvo with purified proteins (Modrich, 1995). mutS is involved in the recognition of the mismatched nucleotides in DNA (Modrich and Lahue, 1996), and mutL is involved in the interaction of mutS with other proteins. In human cells, there appear to be multiple components for recognition of specific mismatches as well as insertions and deletions. In E. coli, mutH recognizes the methylated sequence on the template strand and thus is responsible for determining which of the strands is repaired. The fact that there is no human homolog for mutH suggests that the signal for strand discrimination in eukaryotic cells is different from that in bacteria. Although there is some evidence that single-strand breaks act as a signal to designate the newly replicated lagging strand in eukaryotes, further studies are required. The concept that microsatellite instability is an indicator of a more general mutator phenotype (Loeb, 1994) in cancer is receiving increasing experimental support. First, alterations in repetitive sequences do not appear to be confined to introns. Human colon cancers with high rates of microsatellite instability were found to harbor mutations in repetitive sequences in transforming growth factor-p receptor (Lu et al., 1995). These mutations inactivate the receptor, allowing the escape of cancer cells from transforming growth factor-p mediated growth control. This finding links studies on a mutator phenotype with mutations in tumor suppressor genes. Second, cell lines derived from colon cancers that exhibit a high degree of microsatellite instability also exhibit a 200-fold to 600-fold increase in mutations that render them resistant to 6-thioguanine and ouabain (Bhattacharyya et al., 1994; Phear et al., 1966; Eshleman et al., 1995). Thus, microsatellite instability is linked to and provides a sensitive indicator of a mutator phenotype.

C. Microsatellite Instability in Sporadic Tumors Microsatellite instability has been reported in a large number of tumors not associated with HNPCC, including cancers of the colon (Ionov et al., 1993), endometrium (Risinger et al., 1993), breast (Patel et al., 1994), lung (Merlo et al., 1994), stomach (Dos Santos et al., 1996), ovary (Quinn et al., 1995), and prostate (Egawa et al., 1994). Even though microsatellite instability in many of these sporadic cancers is not as robust as that reported in HNPCC, the identification of this instability in multiple laboratories is impressive. A detailed compilation of the percentages of different sporadic tumors exhibiting microsatellite alterations is given in a review (Eshleman and Markowitz, 1995). In addition to sporadic tumors, there is an exceptional-

46

Lawrence A. Loeb

ly high incidence of microsatellite instability in familial gastric cancers in Japan (Akiyama et al., 1996). These families have been shown to lack mutations in the known human mismatch repair genes. Furthermore, defects in known mismatch repair genes have been detected in only about half of sporadic colorectal tumors that exhibit microsatellite instability (Huang et al., 1996; Papadopoulos et al., 1995). This implies that other changes, presumably other mutations, are responsible for the observed microsatellite instability in many of these tumors. Thus, it is likely that other genes still to be identified are involved in mismatch repair in humans, o r that other genes involved in DNA replication are responsible for these mutations (Loeb, 1994). The source of microsatellite instability in many sporadic tumors is not clear. The presence of microsatellite instability in sporadic tumors has been verified in studies of tumor cell lines. As far as is known, most of these cell lines are derived from sporadic tumors and not from HNPCC-associated tumors (Boyer et ul., 1995). Tumor cell lines have been shown to contain mutations in MSH2 (Umar et al.. 1994), GTBP (Papadopoulos et al., 1995), M L H l (Li and Modrich, 1995), and PMS2 (Risinger et al., 1995a). Cornplementation studies with cell free extracts from endometrial and colorectal cell lines indicate at least four different subsets that generate microsatellite instability (Umar et al., 1994).The finding that microsatellite instability can be complemented with wild-type mismatch repair proteins provides support that the mutations in mismatch repair genes in some of these cell lines are responsible for the associated microsatellite instability (Drummond et al., 1995; LI et al., 1994). It is possible, however, that these mutations in mismatch repair genes were produced during repetitive passages in cell culture. Nevertheless, the facts that the defect exhibited in sporadic tumors is also observed in cell lines derived from these tumors and that the defect can be corrected by the introduction of specific chromosomes containing the wildtype genes (Koi et al., 1994; Risinger et al., 1995b) establish that microsatellite instability in at least some sporadic tumors is the result of deficits in mismatch repair.

D. Other Mechanisms for Mutation Accumulation The concept of a mutator phenotype is based on the assumptions that mutations arise in dividing cells and that the mutation rate (mutations per cell generation) in these cells is insufficient to produce the large numbers of mutations found in human cancers. An alternative hypothesis has been proposed, namely, that mutations accumulate in nondividing cells (Strauss, 1992). Simply expressed, the correct denominator in this formulation is not cell generation, but time. The implication is that cancers arise in nondividing cells that have accumulated large numbers of mutations. In support of

Cancer Cells Exhibit a Mutator P h e n o t y p e

47

this concept, E. coli and yeast maintained in stationary phase exhibit high mutation frequencies in specific selectable markers (Cairns et al., 1988; Hall, 1992; Hall, 1995). Furthermore, many of the mutations that arise in stationary phase occur in runs of similar nucleotides, suggesting slippage by DNA polymerases. MacPhee (1995) has set forth the interesting hypothesis that mutations in nondividing cells can arise by mismatch repair in the absence of strand discrimination. In this situation, half of mismatches would be repaired correctly (by replacement of the altered nucleotide) and half would be converted to mutations. This is at variance with the findings of mutations in mismatch repair genes in some cancers but is compatible with cancers that exhibit microsatellite instability in the absence of mutations in mismatch repair genes.

VII. THEORETICAL AND PRACTICAL CONSI DERATI ON S A. Mutations during Tumor Progression For a mutator phenotype to be a driving force in tumor progression, it would have to be an early event. The initial mutation in a key enzyme involved in DNA replication or DNA repair could decrease the fidelity or efficiency of these processes. With each subsequent round of DNA replication, there would be an increase in mutations throughout the genome. Among these mutations would be mutations in other genes involved in maintaining genetic stability. The timing of mutations in mismatch repair genes is controversial. Some data suggest that mutations in mismatch repair genes are not early events in the progression of colorectal cancer. For example, mutations in mismatch repair genes and microsatellite instability were considered to occur during the transition from late adenoma to early carcinoma stages (Tomlinson et al., 1996). Moreover, computer simulation of a mathematical model for colon cancer suggested that selection without increased mutation rates is sufficient to explain the evolution of tumors (Tomlinson et al., 1996). In contrast, the higher prevalence of microsatellite instability in cancers from patients with multiple independent sites than in cancers from patients with solitary tumors suggests that microsatellite instability is a much earlier event (Shinmuv c l al., 1995). Shibata and colleagues (1996) microdissected different Seb;fl
48

Lawrence A Loeb

early event in tumor progression. Germline mutation in the APC gene is responsible for adenomatous polyposis coli and is also found frequently mutated in sporadic colon cancer. The detection of frameshift mutations in this gene in some (Huang et al., 1996) but not all studies (Tomilson et al., 1996) provides evidence that microsatellite instability precedes and is responsible for mutation in a gene believed to initiate colon cancer. Also in support of a mutator phenotype being an early event in tumor progression are the observations that microsatellite instability can be detected in chronic inflammatory diseases associated with a high incidence of cancer. Patients with chronic pancreatitis are at an increased risk for the development of pancreatic cancers. Pancreatic juice from patient with pancreatis or pancreatic cancer has been reported to demonstrate microsatellite instability (Brentnall et al., 1995a). Microsatellite instability was also found in 50% of cells from ulcerative colitis patients who did not exhibit early changes associated with dysplasia (Brentnall et al., 1996). Thus, microsatellite instability is observed in chronic inflammation, perhaps as a result of the inability of DNA repair to correct the large number of lesions produced by oxygen free radicals. A mutator phenotype might not be manifested in the final stages of tumor progression; genomic instability could confer a selective disadvantage to a relatively well adapted tumor. Even if a mutator phenotype is no longer expressed in a tumor, the presence of multiple mutations in that tumor would confirm its contribution to the early stages of tumor progression.

B. Diagnosis The utility of a mutator phenotype for the diagnosis of cancer is limited to mutations in known mismatch repair genes and/or to measurements of microsatellite instability. Mutations in mismatch repair genes in HNPCC and related hereditary cancers have been used to trace families with germline mutations and thus can be used to diagnose these inherited syndromes. HNPCC constitutes only 5% of cases of colon cancer, however, and the same mismatch repair mutations are found in only a limited number of tumors exhibiting microsatellite instability (Huang et d., 1996). In addition, it will be difficult to design a simple means of detecting these mutations owing to the paucity of characteristic hot spots and the multiplicity of mismatch defective genes. However, common polymorphisms in mismatch repair genes are present in the carcinoma (Brentnall et al., 1995b).Thus, the development of a microchip technology for hybridization of oligonucleotides to mismatch repair genes may have general utility for screening populations for risk of cancer. Microsatellite instability, per se, has been proposed as an assay for cancer screening. In a direct test of this approach, Mao and associates (1996) iden-

Cancer Cells Exhibit a Mutator Phenotype

49

tified microsatellite instability in urine sediments from 19 of 20 patients who were diagnosed with bladder cancer. In all of the 15 patients who underwent tumor biopsy, the microsatellite alterations detected in the urine corresponded to those in the primary tumor. Preliminary reports indicate that microsatellite instability can be ascertained using plasma from patients with head and neck cancers (Nawroz et al., 1996) and small cell lung cancers (Chen et al., 1996). Both studies used patients with advanced disease, and further efforts are needed to determine if this approach is applicable to early diagnosis. A major impediment to the use of microsatellite instability as a diagnostic marker for the presence of malignant cells is the inadequate quantitation of this assay. Size differences in microsatellites are usually determined by gel electrophoresis after PCR amplification of specific sequences. Inconsistencies in the assay might be the result of expansion during PCR amplification. There is no uniform set of microsatellites to be examined in different laboratories. This may account for some of the contradictory results that have been reported; for example, in different studies, the percentage of microsatellite-positive tumors in breast and esophageal cancers varied from 0 to 20% and from 2 to 60%, respectively (Speicher, 1995). New methods are required for quantitation and especially for the elimination of slippage during PCR amplification. Aaltonen and colleagues (1993) urged the adoption of a stringent criterion requiring alterations at two independent genomic sites to classify a tumor as positive. Even this criterion may not be adequate; it does not specify the number of microsatellites to be amplified, the number of cycles of PCR amplification, nor the fidelity of the polymerase used. Also, instability is not uniform in different microsatellites; some appear to be subjected to sequence variations more than others, and this may be tumor specific. Considering that there are hundreds of thousands of microsatellites within the human genome, it should be possible to develop sensitive methodologies for studying microsatellite instability that allow the identification of rare malignant cells or DNA in body fluids, including blood.

C. Prognosis Multiple chromosomal aberrations and loss of ploidy have historically provided an important criterion for classifying tumors as aggressive with a less favorable prognosis. The presence of multiple chromosomal aberrations has indicated that genetic instability is positively correlated with the rapidity of tumor progression. Studies on the relation between microsatellite instability and prognosis, however, have been contradictory. Surprisingly, in hereditary nonpolyposis colon cancer, the presence of microsatellite instability has been interpreted as indicative of a better prognosis (Lothe et al.,

50

Lawrence A. Loeb

1993; Thibodeau, 1993). In contrast, the presence of microsatellite instability in breast cancer is associated with multiple positive lymph nodes, large tumors, and metastasis (Paulson et al., 1996). Patients with breast cancer and microsatellite instability have a reduced survival rate. There are many explanations for this discrepancy. First, the specific microsatellites studied in the two cancers are not the same. Second, hereditary nonpolyposis colon cancer is associated with mutations in mismatch repair genes; in breast cancer, the source of microsatellite instability remains to be determined. Third, as previously stated, standardized quantitation of microsatellite instability is inadequate. Perhaps most important, the rapidity of replication in these two cancers is not the same. Breast cancers are frequently slow growing and limited by the presence of surrounding tissues. Colonic epithelium is progressively extravasated into the intestinal lumen and constantly replenished. It has been estimated that the total production of cells by the digestive tract in 30 days is equal to the total number of cells in the body (Leblond, 1965). Thus, clonal selection may occur to a greater extent in colon cancer than in many other cancers. In addition to prognosis, the quantitation of microsatellite instability is predictive of the chemosensitivity of tumors to certain chemotherapeutic agents. Studies show that, contrary to expectations, many alkylated lesions in DNA can be tolerated in the absence of mismatch repair. These studies have suggested that use of repetitive attempts to repair the alkylated nucleotides in DNA by the mismatch repair system is associated with lethality (Karran et al., 1993; Kat et al., 1993). As a result, tumors exhibiting microsatellite instability are resistant to the lethal effects of alkylating agents. Studies are needed on the extent of repair of different adducts in DNA by the mismatch repair system and on the correlation of microsatellite instability with resistance to other chemotherapeutic agents.

D. Cancer Prevention If we assume that the presence of multiple mutations in human cancers is central to the pathogenesis of these malignancies, then it becomes important to determine the sources of these mutations and to diminish their occurrence. It takes about 20 years or longer between the exposure of a person to a carcinogen and the clinical manifestations. Delaying the rate of accumulation of mutations in premalignant lesions or in early tumors provides a new approach to preventing cancer deaths. In previous publications, we have considered three human cancers in which the time interval between the exposure to the carcinogens and the clinical manifestation of the tumor either can be documented or logically estimated (Loeb and Cheng, 3 990). A delay in the rate of accumulation of mutations by a factor of two in these tumors

Cancer Cells Exhibit a Mutator Phenotype

51

would drastically reduce the death rates from these tumors in the population. A mutator phenotype implies an irreversible process characterized by the progressive accumulation of mutations during tumor progression. It may still be feasible, however, to modify this phenotype, decreasing the rate of mutation accumulation and therefore the progress of the tumor.

VIII. SUMMARY A N D PERSPECTIVES In this article, 1 have considered the hypothesis that cancer cells exhibit a mutator phenotype. The presence of multiple chromosomal alterations in cancer cells and the evidence for microsatellite instability in tumors strongly support this concept. Most mutations in cancer cells have been identified on the basis of chromosomal rearrangements involving thousands o f nucleotides. A much larger number of smaller changes in the nucleotide sequence may also be present in tumors. Such a large number of mutations may not be the result of a normal mutation rate, but instead may be a manifestation of a mutator phenotype in cancer cells. The concept that cancer cells exhibit a mutator phenotype brings forth a series of new questions: Are these errors the result of misincorporations by DNA polymerases? Are there mutant, error-prone DNA polymerases in tumors? Could deficits in other proteins, such as DNA binding proteins o r helicases, enhance the frequency of DNA replication errors? Multiple genes are involved in maintaining chromosomal stability in normal cells. Do mutations in these genes result in a mutator phenotype? From an analysis of the spectrum of mutations in a cancer, can we surmise the nature of the mutator protein? New methodologies are being developed that should allow us to quantitate definitively the frequency of nucleotide changes in DNA cancer cells. This quantitation is difficult because o f the size of the human genome and the likelihood that most mutations in cancer cells are randomly distributed. This quantitation is important, however, because it may serve as a barometer of individuals’ exposure to environmental mutagens and also allow us to stage the extent of tumor progression molecularly and to predict the probability of metastasis.

ACKNOWLEDGMENTS I thank Drs. 1:A. Kunkel, T. A. Brentnall, R. Prehn, A. Skandalis, and A. Jackson for critical comments, and M. Whiting for editing. 6. S. Strauss generously provided unpublished results on pS3 silent mutations. This work was supported by a n Outstanding Investigator Grant

52

Lawrence A. Loeb

(CA39903) from the National Cancer Institute, and by a Program Project Grant (AG017SI) from the National Aging Institute.

REFERENCES Aaltonen, L. A., Peltoniaki, P., Leach, F. S., Sistonen, P., Pylkkanen, L., Mecklin, J.-P., Jarvinen, H., Powell, S. M., Jin, J., Hamilton, S. R., Petersen, G. M., Kinder, K. W., Vogelstein, B., and de la Chapelle, A. ( 1 993). Science, 260, 812-816. Akiyama, Y., Nagasaki, H., Nihei, Z., Iwama, T., Nomizu, T., Utsunomiya, J., and Yuasa, Y. (1996).]pn.]. Cancer Res., 87,595-601. Ames, B. N. (1983). Science, 221, 1256-1263. Ames, B. N., Profet, M., and Gold, 1.. S. ( 1 990). Pror. Natl. Acud. Sci. U.S.A., 87, 7782-7786. Armitage, P., and Doll, R. (1954). Br. 1. Cancer, 8, 1-12. Balaban, G. B., Herlyn, M., Clark, W. H., Jr., and Nowell, P. C. (1986). Cancer Genet. Cytogenet., 19, 113-122. Barbacid, M. ( 1 987). Annu. Rev. Biochem., 56, 779-827. Basu, A. K., Wood, M. L.,Niedernhofer, L. J., Ramos, 1.. A., and Essigmann, J. M. (1993). B i o chemistry, 32, 12793-1 2801. Battula, N., and Loeb, L. A. ( 1 974).]. Bid. Chem., 249,4086-4093. Beard, W. A., Osheroff, W. P., Prasad, R., Sawaya, M. R., Jaiu, M., Wood, T. G., Kraut, J., Kunkel, T. A., and Wilson, S. H. (1996).]. B i d . Cheni., 271, 12141-12144. Bhattacharyya, N. P.,Skandalis, A., Ganesh, A., Groden, J., and Meuth, M. (1994). Proc. Natl. Acad. Sci. U.S.A., 91, 6319-6323. Biedler, J. L., and Spengler, B. A. (1976). Science, 191, 185-187. Birrer, M. J., and Minna, J. D. (1988). Semin. Oncol., 15, 226-235. Boyer, J. C., Umar, A., Risinger, J , I., Lipford, J. R., Kane, M., Yin, S., Barrett, J. C., Kolodner, R. D.,and Kunkel, T. A. ( 1 995). Cancer Res., 55,6063-6070. Brentnall, T. A., Chen, R., Lee, J. G., Kimmey, M. B., Bronner, M. P., Haggitt, R. C., Kowdley, K. V., Hecker, L. M., and Byrd, D. R. (199Sa). Cancer Res., 55,42644267. Brentnall, T. A., Rubin, C. E., Crispin, D. A., Stevens, A., Batchelor, R. H., Haggitt, R. C., Bronner, M. P., Evans, J. P., McCahill, L. E., Bilir, N., Boland, C. R., and Rabinovitch, P. S. (1995b). Gastroenterology, 109, 151-156. Brentnall, T. A., Crispin, D.A., Bronner, M. P., Cherian, S. P., Hueffed, M., Rabinovitch, P. S., Rubin, C. E., Haggitt, R. C., and Boland, C. R. (1996). Cancer Res., 56, 1237-1240. Bronner, C. E., Baker, S. M., Morrison, P. T., Warren, G., Smith, L. G., Lescoe, M. K., Kanc, M., Earabino, C., Lipford, J., Lindblom, A., Tannergard, P., Bollag, R. J., Godwin, A. R., Ward, D.C., Nordenskjold, M., Fishel, R., Kolodner, R., and Liskay, R. M. (1994). Nature, 368,258-261.

Bucala, R., Model, P., and Cerami, A. (1984). Proc. Nut/. Acad. Sci. U.S.A., 81, 105-109. Burns, P. A., Gordon, A. J., and Glickman, B. W. (1988). Carcinogenesis, 9, 1607-1610. Cairns, J. (1981). Nature, 289, 353-357. Cairns, J., Overbaugh, J., and Miller, S. (1988). Nature, 33.5, 142-145. Cathcart, R., Scheiers, E., Saul, R. I.., and Ames, B. N. (1984). Proc. Nut/. Acad. Sci. U.S.A., 81,5633-5637.

Cha, R. S., Guerra, L., Thilly, W. G., and Zarbl, H. (1996). Carcinogenesis, 17, 2519-2524. Chaudhary, A. K., Nokuho, M., Reddy, G. R., Yeola, S. N., Morrow, J. D.,Blair, I. A., and Marnett, L. J. (1994). Science, 265, 1580-1582. Chen, X. Q., Stroun, M., Magnenat, J.-I-., Nicod, L. P., Kurt, A.-M., Lyautey, J., Lederrey, C., and Anker, P. (1996). Nature Med., 2, 1033-1035.

Cancer Cells Exhibit a Mutator Phenotype

53

Cheng, K. C., Cahill, D. S., Kasai, H., Nishimura, S., and Loeb, L. A. (1992)./. B i d . Chen?., 267,166-172. Cheng, K. C., and Loeb, L. A. (1993).Adv. Cancer Res., 60, 121-156. Cheng, K. C., and Loeb, L. A. ( 1 997). In “Genetic Instability and Tumorigenesis” (M. B. Kastan, Ed.), 221, pp. 5-18, Springer-Verlag, Berlin. Chigira, M., Arita, S., and Watanabe, H. (1993).Int. /. Oncol., 2, 931-935. Cleaver, J. E., and Kraemer, K. H. (1989).In “Metabolic Basis of Inherited Disease” (C. R. Scriver, A. L. Beudet, W. S., S., and D. Valle, Eds.), pp. 2949-2971. McGraw-Hill, New York. Copeland, W. C., Lam, N. K., and Wang, T. S. (1993)./. B i d . Chem., 268,11041-11049. Cox, E. C. (1976).Annu. Rev. Genet., 10, 135-156. Crook, T., Marston, N. J., Sara, E. A., and Vousden, K. H. (1994). Cell, 79, 814-827. Dobashi, Y., Shuin, T., Tsuruga, H., Uemura, H., Torigoe, S., and Kubota, Y. (1994). Cancer Res., 54,2827-2829. Dos Santos, N. R., Seruca, R., Costancia, M., Seixas, M., and Sobrinho-Simoes, M. (1996). Gastroenterology, 110, 38-44. Drummond, J. T., Li, G.-M., Longley, M. J., and Modrich, P. ( 1 995). Science, 268, 1909-1 91 2. Egawa, S., Uchida, T., and Suyama, K. (1994).Cancer Res., 55,2418-2421. Elledge, S. J. (1996).Science, 274, 1665-1672. Ellis, N. A., Groden, J., Ye, T.-Z., Straughen, J., Lennon, D. J., Ciocci, S., Proytcheva, M., and German, J. ( 1 995). Cell, 83, 655-666. Epstein, C. J., Martin, G. M., Schultz, A. L., and Motulsky, A. G. (1966). Medicine, 4.5, 177-22 1. Erikson, P., Gao, J., Chang, K.-S., Look, T., Whisenant, E., Raimondi, S., Lasher, R., Trujillo, J., Rowley, J., and Drabkin, H. (1992). Blood, 80, 1825-1831. Erlich, M., and Wang, R. Y.-H. (1981).Science,212, 1350-1357. Eshleman, J. R., Lang, E. Z., Bowerfind, G . K., Parsons, R., Vogelstein, B., Wilson, J. K. V., Veigl, M. L., Sedwick, W. D., and Markowitz, S. D. (1995). Oncogene, 10, 33-37. Eshleman, J. R., and Markowitz, S. D. (1995). Cur% Opin. Oncol., 7, 83-89. Fearon, E. R., and Vogelstein, B. (1990).Ce11,61, 759-767. Fialkow, P. J. (1974).N. Engl. /. Med., 291, 26-35. Fishel, R., Lescoe, M. K., Rao, M. R. S., Copeland, N. G., Jenkins, N. A., Garber, J., Kane, M., and Kolodner, R. (1993).Cell, 75, 1027-1038. Fry, M., and Loeb, L. A. (1986). “Animal Cell DNA Polymerases.” C R C Press, Boca Raton, FL. Fry, M., and Loeb, L. A. (1994). P r x . Natl. Acad. Sci. U.S.A., 91,4950-4954. Fukuchi, K., Martin, G. M., and Monnat, R. J., Jr. (1989). Proc. Nutl. Acad. Sci. U.S.A., 86, 58934897. German, J., Roe, A. M., Leppert, M. F., and Ellis, N. A. (1977).Proc. Natl. Acad. Sci. I1.S.A.. 91,6669-6673. Goto, M., Miller, R. W., Ishikawa, Y., and Sugano, H. ( 1 996). Cancer Epidemiol. Biomarkers Prev., 5, 239-246. Greenblatt, M. S., Bennett, W. P., Hollstein, M., and Harris, C. C. (1994). Cancer Res., 54, 48~5-4878.\ Hall, B. G. (1992). Proc. Natl. Acad. Sci. U.S.A., 89,4300-4303. Hall, B. G. (1995). Proc. Natl. Acad. Sci. U.S.A., 92,5669-5673. Hanawalt, P.C., Cooper, P. K., Ganesan, A. K., and Smith, C. A. ( 1 979). Ann. Rev. Biochem., 48,783-836. Harris, H. (1988).Cancer Res., 48, 3302-3306. Harris, C. C. ( 1996). Science, 262, 1980-1 98 I . Hartwell, L. H., and Kastan, M. B. (1994).Science, 266, 1821-1827. Hollstein, M., Shomer, B., Greenblatt, M., Soussi, T., Hovig, E., Montesano, R., and Harris, C. C. (1996). Nucleic Acids Res., 24, 141-146.

54

Lawrence A. Loeb

Huang, I., Papadopoulos, N., McKinley, A. J.. Farrington, S. M., Curtis, 1.. I., Wyllie, A. H., Zheng, S., Willson, J. K. V., Markowitz, S. S., Morin, P., Kinzler, K. W., Vogelstein, 13., and Dunlop, M . G. (1996).Pro(. Natl. Acad. Sci. U.S.A., 93, 9049-9054. Illmensee, K., and Mintz, B. (1976). Proc. Natl. Acad. Sci. U.S.A., 73, 549-553. lonov, Y., I’einado, M. A., Malkhosyan, S., Shibata, D., and Perucho, M. ( 1 9 9 3 ) .Nature, 363, 5 5 8-56 1 . James, (1. I)., Carlbom, E., Dumanski, J. B., Hansen, M., Nordenskiold, M., Collins, V. P., arid Cavcnee, W. K. (1988). Cancer Res., 48, 5546-5551. Karran, P., MacPhcrson, I)., Ceccotti, S., Dogliotti, E., Griffin, S., and Bignami, M. (1993). J. B i ~ l Chetll., . 268, 15878-15886. Kat, A,, Thilly, W. G., Fang, W.-M., Longley, M. J., Li, G.-M., and Modrich, I? (1993).Proc. Nail. Acad. Sci. U.S.A., 90, 6424-6428. Knudson, A. G., Jr. (1971). Proc. Natl. Acad. Sci. U.S.A., 68, 820-823. Koi, M., Umar, A., Chauhan, D. P., Cherian, S. P., Carethers, J. M., Kunkel, T. A., and Holand, C. R. (1994). Carrcer Res., 54,4308113 12. Kuhota, Y., Murakami-Muroftishi, K., Shimada, Y., Ogiu, T., and Oikawa, 1’.( I 995). Caiicer Rcs., 5 5 , 3777-3780. Kunkel, T. A . (1985). 1. Biol. Chent., 260, 5787-5796. Kunkel, T. A . (1992).1. B i o l . Chent., 267, 18251-1 8254. Kunkcl, T. A,, and Loeh, 1.. A. (1981).Scienc-e, 213, 765-767. Lehlond, C. P. (1965).A m . / . Anat., 116, 1-27. Li, C.-Y., Yandell, D. W., and l.irtle, J. B. (1994). M o l . Cell Biol., 14, 4373-4379. Li, G.-M., and Modrich, P. ( I 995). Proc. Ntztl. Acud. Sci. U.S.A., 92, 1950-1954. Lindahl, T., and Nyhcrg, B. ( 1972). Biocbetrzrstry, 11, 36 10-36 18. L.indahl, T., arid Nyberg, 13. ( 1974). Rioc..hemisrry, 13, 3405-,3410. Lindhlom, A,, Tannergard, I?, Werclius, B., and Nordenskjold, M. ( 1 9 9 3 ) . Nature Genet., 5, 279-282. Livingstone, 1.. K., Whirc, A., Sprouse, J., Livanos, E., Jacks, T., and Tlsty, T. I). (1992).Cell, 70, 923-9.3.5. Loeb, I.. A. (1985). Cell, 40, 4831184. L.oeh, 1.. A. ( 199 1 ). Cirrrcer Res., 51, 3075-3079. I.oeh, L. A. ( 1994). C m c e r Res.. 54, 5059-506.3. h e b , 1.. A., and Cheng, K . C.( 1990).Mritirt. Res., 238, 297-304. l.oeh, L. A., Springgatc, C. F., and Battula, N. ( 1974). Catrcer Res., 34, 2.3 I 1-2321 . Lothe, R. A., Peltomaki, P., Meling, G. I., Aaltnnen, 1.. A., Nystrom-Lahti, M., Pylkkanen, L., Hcimdal, I.., Anderscn, -1. I., Moiler, P., liognuin, T. O., Fossa, S. D., Haidorsen, T., 1.angmark, F., Brogger, A., de la Chapelle, A,, and Borresen, A.-1.. ( 1993). ( ~ m r c e iRex, 5.3, s849-5 8s 2. Lu, S.-L., Akiyama, Y., Nagasaki, H., Saitoh, K., and Yuasa, Y. ( 1 9 9 5 ) .Rrochem. Hiophys. Kes. C(JI~WZU 216, I ? .452-457. , MacPhee, D. <;. (199.5). C‘mtccr Res., 5 5 , 5489-5492. Malkhosyan, S., McCarty, A., Sawai, H., and Perucho, M. (1996).Mtrtat. Rrs., 316, 249-259. Mao, E. I., Lane, L., Lee, I., and Miller, J. t-1. ( 1997).J. Rclcteriol.. 179, 4 17-422. Man, I,., Schoenberg, M. P., Scicchitano, M., Erozan, Y. S., Merlo, A., Schwab, D., and Sidransky, D. ( 1 9 9 6 ) .Scicwre, 271, 6.59-662. Marra, G.,and Boland, C. K. ( l 9 9 S ) . / . Natl Con. Inst., 87, I 1 14-1 125. Matsuzaki, J., Dohashi, Y., Miyamoto, H.. Ikeda, I., Fujinami, K., Shuin, T., and Kuhota, Y, (1996).M o l . G r c i n o g . , 15, 38-4 Mcrlo, A., Mahry, M., Gahrielson, ., Vollmer, K., Hasylin, S. B., and Sidransky, D. ( 1994). Cheer Res., 54, 2098-2101. Meuth, M. ( 1990). Bi(JCkiI?l. BiofIhyS. A c t r ~ ,1032, 1-1 7.

Cancer Cells Exhibit a Mutator Phenotype

55

Miller, J. H. (1985).]. Mol. E d . , 182, 45-68. Miller, J. H. (1996).In “Cancer Surveys” (T. Lindahl, Ed.), 28, pp. 141-153. Cold Spring Harbor Laboratory Press, Fairview, New York. Minnick, D. T., and Kunkel, T. A. (1996).In “Cancer Surveys” (T. Lindahl, Ed.), 28, pp. 3-20. Cold Spring Harbor Laboratory Press, Plainview, New York. Modrich, P. (1987).Annu. Rev. Biochem., 56,435-466. Modrich, P. ( I 995). Philosoph. Transact. Royal Soc., 347, 89-95. Modrich, P., and Lahue, R. (1996).Annu. Rev. Biochem., 65, 101-133. Munir, K. M., French, D. C., and Loeb, L. A. (1993). Proc. Natl. Acud. Sci. U.S.A., 90, 40 1 2 4 0 1 6. Murray, V., and Holliday, R. (1981).J. Mol. Biol., 146, 55-76. Nawroz, H., Koch, W., Anker, P., Stroun, M., and Sidransky, D. (1996). Nature Med., 2, 1035-1 037. Naylor, S. I-., Johnson, B. E., Minna, J. D., and Sakaguchi, A. Y. (1987).Nature, 329,451-454. Nicolaides, N. C., Papadopoulos, N., Liu, B., Wei, Y. F., Carter, K. C., Ruben, S. M., Rosen, C. A., Haseltine, W. A., Fleischmann, R. D., Fraser, C. M., Adams, M. D., Venter, J. C., Dunlop, M. G., Hamilton, s. R., Petersen, G. M., de la Chapelle, A., Vogelstein, B., and Kinzler, K. W. (1994).Nature, 371,75-80. Nowell, I? C. ( 1 976). Science, 194,23-28. Nowell, P. C. (1993).Adv. Cancer Res., 62, 1-17. Orgel, L. E. (1973).Nature, 243, 441-445. Otto, E., McCord, S., and Tlsty, T. D. ( 1989).J. &Jl. Chem., 264, 3390-3396. Papadopoulos, N., Nicolaides, N. C., Liu, B., Parsons, R. E., Lengauer, C., Palombo. F., D’Arrigo, Markowitz, S., Wilson, J. K. V., Kinzler, K. W., Jiricny, J., and Vogelstein, B. (1995). Science, 268, I91 5-1 9 17. Papadopoulos, N., Nicolaides, N. C., Wei, Y.-F., Ruben, S. M., Carter, K. C., Rosen, C. A., Haseltine, W. A., Fleischmann, R. D., Fraser, C. M., Adams, M. D., Venter, J. C., Hamilton, S. R., Petersen, G. M., Watson, P., Lynch, H. T., Peltomaki, P., Mecklin, J.-P., de la Chapelle, A., Kinzler, K. W., and Vogelstein, B. (1994). Science, 263, 1625-1 629. Patel, U., Grimdfest Broniatowski, S., Gupta, M., and Banerjee, S. ( 1994). Oncogene, 9, 3695-3700. Paulson, T. C., Wright, F. A., Parker, B. A., Russack, V., and Wahl, G. M . (1996). Cancer Res., 56,4021-4026. Peinado, M. A., Malkhosyan, S., Velazquez, A., and Perucho, M. ( I 992). Proc. Natl. Acad. Sci. U.S.A., 89, 10065-10069. Peltomaki, P., Aaltonen, L. A., Sistonen, P., I’ylkkanen, L., Mecklin, J.-P., Jarvinen, J., Green, J. S., Jass, J. R., Weber, J. L., Leach, F. S., Petersen, G. M., Hamilton, S. R., de la Chapelle, A., and Vogelstein, B. (1993).Science, 260, 810-812. Phear, G., Bhattacharyya, N. P.,and Meuth, M. (1996).Mol. Cell Biol., 16,6516-6523. Prehn, R. T. [ 1994). Cancer Res., 54, 5296-5300. Quinn, A. G., Healy, E., Rehman, I., Sikkink, S., and Rees, J. L. (1995).1. Invest. Dermatol., 104,309-312. Reid, T. M., and h e b , L. A. (1992).Cancer Res., 52, 1082-6. Renan, M. J. (1993).Mol. Carcinog., 7, 139-146. Risinger, J . I., Umar, A., Barrett, J. C., and Kunkel, T. A. (199Sa). J . E d . Chem., 270, 18183-18 186. Risinger, J. I., Uinar, A., Boyer, J . C., Evans, A. C., Berchuck, A., Kunkel, T. A., and Barrett, I. C. (1995b). Cancer Res., 55,5664-5669. Risinger, J. L., Berchuck, A., Kohler, M. F., Watson, P., Lynch, H. T., and Boyd, J. (1993).Cancer Res., 53, 5100-5103. Rowley,J. D. (1975).Proc. Natl. Acad. Sci. U.S.A., 72, 152-156.

56

Lawrence A. Loeb

Schaaper, R. M. (1993).Genetics, 134, 1031-1038. Schaaper, R. M., Kunkel, T. A., and Loeb, L. A. (1983). Proc. Natl. Acad. Sci. U.S.A., 80, 487-49 I . Shen, J. C., Kideout, W. M., and Jones, P. A. (1992). Cell, 71, 1073-1080. Shibata, D., Navidi, W., Salovaara, R., Li, Z.-H., and Aaltonen, L. A. (1996). Nature Med., 2, 676-68 I . Shibutani, S., Takeshita, M., and Grollman, A. 0. (1991). Nature, 349, 4 3 1 4 3 4 . Shigenaga, M. K., Cirneno, C. J., and Ames, B. N. (1989). Proc. Natl. Acad. Sci. U.S.A., 86, 9697-970 I . Shinmtira, K., Sugimura, H., Naito, Y., Shields, P. G., and Kino, I. (1995). Carcinogenesis, 16, 2989-2993. Silber, J. R., Blank, A,, Bohola, M. S., Mueller, B. A,, Kolstoe, D. D., Ojeniann, G. A., and Berger, M. S. (1996). Proc. Natl. Acad. Sci. U.S.A., 93, 6941-6946. Singer, B., and Essigmann, J. M. (1991). Carcinogenesis, 12, 949-955. Singer, B. and Grunberger, D. ( 1983).The molecular biology of mutagens and carcinogens, pp. Plenum Press, New York. Speicher, M. K. (1995). Oncol. Res., 7,267-275. Stanhridge, E. J. (1990). Science. 247, 12-1 3. Stein, W. D. (1991). Adu. Cancer Res., 56, 161-213. Stillman, B. (1996). Science, 274, 1659-1664. Straws, B. S. (1992). Cancer Res., 52, 249-253. Strauss, B. S. ( 1996). Proc. Internat. Symp. Biodefense Mcchanisms, Kodansha Scientific Ltd., Tokyo. Sweasy, J. B., and Yoon, M. S. ( I 99.5). Mol. Gem Genet., 248,217-224. Swift, M., Morrell, D., Massey, R. B., and Chase, C. I.. (1991). N . Engl. ]. Med., 325, 1831-18.36. Thibodeau, S. N., Bren, G., and Schaid, D. (1993). Science, 260, 816-819. Tlsty, T. D., White, A., and Sanchez, J. (1992). Science, 255, 1425-1427. Tomlinson, I. P. M., Novelli, M. K.,and Bodmer, W. F. ( 1 996). Proc. Natl. Acad. Sci. U.S.A., 93,14n00-14~03 ,Thomas, D. C., Nguyen, D. C., Risinger, J. I.,Boyd, J., Ionov, Y., I’erucho, M., and Kunkel, T. A. (1994).]. B i d . Chem., 269, 14367-14370. Umar, A., Buermeyer, A. B., Simon, J. A., Thomas, D. C., Clark, A. B., Liskay, R . M., and Kunkel, T. A. ( I 996). Cell, 87, 65-73. Vogelstein, B., Fearon, E. R., Kern, S. E., Hamilton, S. R., Preisingcr, A. C., Leppert, M., Nakamura, Y., White, R., Smits, A. M. M., and Bos, J. I.. (1988). N . Engl. I . Med., 319, 525-532. Wang, L., Patcl, U., Ghosh, L., and Banerjee, S. (1992). Cancer Res., 52, 4824-4827. Weber, J. I.., and May, P. E. (1989). A m . ] . Hum. Genet., 44,388-396. Weinherg, R. A. (1991). Science, 254, 1138-1 146. Weinstein, I. B. (1995).Cancer Preu., 768, 3 0 4 0 . Yu, C.-E., Oshinia. J., Fu, Y.-H., Wijsman, E. M., Hisama, F., Alisch, R., Matthews, S., Nakura, J., Miki, T., Ouais, S., Martin, G. M., Mulligan, J., and Schellenberg, C. D. (1996).Scierne, 272, 258-262. Yunis, J. J., and Sanchez, 0. (1975).Humangenetik, 27, 167-172.

Increasing Complexity of R a s Signal Transduction: Involvement of Rho Family Proteins Roya Khosravi-Far, I Sharon Kent L. Rossrnan,2 a n d Channing J . Der3v4s5 Department of Biology, Massachusetts Institute of rechnolr)gy Cufnhridge, MA 02 I 39; 2Departmrnt of Biochemtstry and Biophysrcs 'Department of Pharmacdogy J14inehergerComprehensirv C m c r r Center 'Curriculum tn Genetics and Molecuhr Btohgy Ut?ri,ersrtyof North C:nro/ina. Chapel Hill, N C 27599

I. Introduction 11. Ras Is a Point of Convergence for Diverse Extracellular

Signal-Stimulated Pathways A. Ras Activation of the Raf+MEK+MAPK Kinase Cascade B. Raf Is a Key Effector of Ras Ill.Ras Activation of Raf-Independent Pathways Contributes to Ras Transformation IV. Ras Mediates Its Actions Through Interaction With Multiple Effectors A. A Growing Roster of Candidate Effectors of Ras B. Structural Requirements for Ras-Effector Interactions V. Ras Activation of a GTPase Cascade: Involvement of Rho Family Proteins in Transformation A. A Requirement for Rho Family Proteins for rus Transformation B. Rho Family Proteins Are Regulators of Diverse Cellular Processes VI. Rho Family Proteins Mediate Their Actions Through Interaction With Multiple E.ffectors A. h Plethora of Candidate Effectors of Rac and CDC42 B. A Plethora of Candidate Effectors of Rho C. Structural Requirements for Rho Family Protein-Effector Interactions VII. A Search for the Missing Link Between Ras and Rho Family Proteins VIII. Increasing Complexity of Ras Signal Transduction: A Boon or Bust for Drug Discovery and the Development of Anti-Ras Drugs for Cancer Treatment? IX. Future Directions References

The initial discovery that YUS genes endowed retroviruses with potent carcinogenic properties, and the subsequent determination that mutated rus genes were present in a wide variety of human cancers, prompted a strong suspicion Advances in CANCER 00hS-L30X/YX $?S.OO

RESEARCH

Copyright 0 199X by Academic Prcs. All rights of reproduction in any form rmervcd.

58

Roya Khosravi-Far e l al.

that the growth-promoting actions of mutated Ras proteins contribute to their aberrant regulation of growth-stimulatory signaling pathways. In 1993, a remarkable convergence of experimental observations from genetic analyses of Drosophila melanogaster, Saccharomyces cerevisiae, and Caenorhabditis eleguns, as well as biochemical and biologic studies in mammalian cells, came together to define a clear role for Ras in signal transduction. What emerged was an elegant linear signaling pathway in which Ras functions as a relay switch that is positioned downstream of cell surface receptor tyrosine kinases and upstream of a cytoplasmic cascade of kinases, including the mitogen-activated protein kinases (MAPKs; Fig. 1).Activated MAPKs in turn regulate the activities of nuclear transcription factors. Thus, a signaling cascade every component of the signaling cascade between the cell surface and the nucleus was defined and conserved in worms, flies, and humans. This was remarkable achievement in our efforts to appreciate how the aberrant function of Ras proteins may contribute to the malignant growth properties of the cancer cell. However, the identification of this pathway has proved to be just the beginning, rather than the culmination, of our understanding of Ras in signal transduction. We now appreciate that this simple linear pathway represents but a minor component of a complex signaling circuitry. Ras signaling has emerged to involve a complex array of signaling pathways, in which cross-talk, feedback loops, branch points, and multicomponent signaling complexes are recurring themes. The simplest concept of a signaling cascade, in which each component simply relays the same message to the next, is clearly not the case. This review summarizes two major emerging themes that have refocused our perceptions of how Ras functions in signal transduction. First, it has become clear that the Raf-1 serinehhreonine kinase is not the sole downstream effector of Ras. Earlier proclamations that Ras served simply as an activator of Raf were clearly premature. Instead, what has emerged is that Ras uses a multitude of functionally diverse effector targets. Second, the functions of certain Ras-related proteins that constitute part of the Rho family proteins have been shown to be important for Ras transformation. This, together with evidence implicating these small guanosine triphosphates (GTPases) downstream of Ras, has prompted considerable effort in understanding how the function of Rho family proteins contribute to Ras signaling and transformation. In particular, it has become apparent that each Rho family protein, like Ras, also uses a plethora of functionally diverse downstream effector targets. This revelation is not surprising considering that Rho family proteins are not solely regulators of the organization of the actin cytoskeleton. Instead, Rho family proteins are also involved in signaling pathways that activate kinase cascades, regulate gene expression, and control cell cycle progression and cell proliferation. In this review, we summarize our current understanding of Ras signal transduction, with emphasis on the involvement of Rho family proteins and on how Rho family proteins may contribute to Ras function.

increasing Complexity of Ras Signal Transduction: Involvement of Rho Family Proteins

D’

59

W -

IGene Expression1

\

Fig. 1 Ras regulation o f a cascade of kinases. Ras serves as a GDPIGTP-related binary switch that resides at the inner surface of the plasma membrane and acts as a relay for extracellular ligand-stimulated signals to cytoplasmic-signaling cascades. A linear pathway, in which Ras functions downstream of receptor tyrosine kinases (RTKs, such as the EGFR rcccptor) and upstream of a cascade of serinehhreonine kinases (Raf+MEK+MAPK), provides a complcte link between the cell surface and the nucleus.

Targeting components of the Ras signaling pathways has been proposed as one approach for the development of anti-Ras drugs for cancer treatment. The realization that our comprehension of the complexities of Ras signaling is limited, even simple-minded, may discourage any serious consideration that this is a fruitful direction for drug discovery a t this time. Thus, some believe that such efforts are premature and must await a complete knowledge of this signaling circuitry. Evidence that intervention of Ras signaling a t mul-

60

Roya Khosravi-Far el al.

tiple points can significantly impact the ability of Ras to cause cellular transformation, however, argues that a full unveiling of all the complexities of Ras signaling will not be required before successful targeting of Ras signaling pathways for drug discovery can be achieved.

II. Ras IS A POINT OF CONVERGENCE FOR DIVERSE EXTRACELLULAR SIGNAL-STIMULATED PATHWAYS The three human ras genes (H-ras, N-ras, and K-ras) encode four 188 to 189-amino acid (p21)proteins that function as guanosine diphospate/guanosine triphosphate (CDP/GTP) regulated switches (H-Ras, N-Ras, K-Ras4A, and K-Ras4B) (Barbacid, 1987; Bourne et al., 1990; Boguski et al., 1993; Quilliam et al., 1995). The two forms of K-Ras diverge solely in their COOH-terminal25 amino acids as a consequence of alternate exon use. Ras proteins are positioned at the inner surface of the plasma membrane, where they serve as binary molecular switches to transduce extracellular ligand-mediated stimuli into the cytoplasm, to control signal transduction pathways that influence cell growth, differentiation, and apoptosis (Satoh and Kaziro, 1992; Khosravi-Far and Der, 1994b). Ras biologic activity is controlled by a regulated CDP/GTP cycle (see Fig. 1). The intrinsic G D P K T P exchange and GTP hydrolytic activity of Ras are low. Instead, GDP/GTP cycling is accelerated and regulated by two types of regulatory proteins. Guanine nucleotide exchange factors (GEFs; Ras-GRFlmCDC25, SOS1/2) promote formation of the active GTP-bound state, and Ras-GTPase-activating proteins (GAPS;p120 GAP, NF1-GAPheurofibromin) promote formation of the inactive GDP-bound state (Boguski and McCormick, 1993; Quilliam et al., 1995).The single amino acid substitutions a t 12,13, or 61 that unmask Rastransforming potential create mutant proteins that are insensitive to GAP stimulation (Bos, 1989; Gibbs, 1984). Consequently, these oncogenic Ras mutant proteins are locked in the active, GTP-bound state, leading to constitutive, deregulated activation of Ras function.

A. Ras Activation of the Raf+MEK+MAPK Kinase Cascade Ras is point of convergence for many signaling pathways (Khosravi-Far and Der, 1994b). Ras proteins are activated transiently in response to a diverse array of extracellular signals, such as growth factors, cytokines, hormones, and neurotransmitters that stimulate cell surface receptors. The latter include receptor tyrosine kinases (RTKs), non RTK-associated receptors,

Increasing Complexity of Ras Signal Transduction: Involvement of Rho Family Proteins

6I

and G-protein-coupled seven-transmembrane receptors (SRs) (Satoh and Kaziro, 1992; Khosravi-Far and Der, 1994b). The best characterized Rasmediated signal transduction pathway involves the activation of peptide mitogen-stimulated RTKs, such as the epidermal growth factor (EGF) receptor (EGFR; see Fig. 1) (Egan and Weinberg, 1993; Khosravi-Far and Der, 199413).The EGF-stimulated EGFR undergoes autophosphorylation of specific tyrosine residues in its cytoplasmic domain, which creates phosphotyrosy1 binding sites for the Src homology 2 (SH2) and/or phosphotyrosyl binding (PTB) domains for the Shc and/or Grb2 adaptor proteins (Williams, 1992; Mayer and Baltimore, 1993a; Schlessinger, 1993). Shc becomes autophosphorylated on association with activated RTKs, which creates recognition sites for the SH2 domain of Grb2. Because Grb2 is stably associated with the Ras-GEF (Son of Sevenless [SOS]),the Shc/Grb2- or Grb2-mediated translocation of the SOS to the plasma membrane, where Ras resides, leads to the transient elevation of Ras GTP levels (Feig, 1994; Schlessinger, 1993). Shc, Grb2, and SOS provide the link between many types of activated cell surface receptors and Ras (Williams, 1992; Schlessinger, 1993). Activated Ras relays its signal downstream through a cascade of cytoplasmic proteins (see Fig. 1). Substantial biologic, biochemical, and genetic evidence has implicated the Raf-l serinehhreonine kinase as a critical effector of Ras function (Moodie and Wolfman, 1994). A key observation was that biologically active, but not inactive, Ras forms a high-affinity complex with Raf 1 (Van Aelst et al., 1993; Moodie et al., 1993; Zhang et al., 1993; Warne et al., 1993; Vojtek et al., 1993). The Ras-Raf association promotes a translocation of the normally cytoplasmic Raf protein to the plasma membrane, where subsequent events lead to the activation of its kinase function. These events are complex and incompletely understood (Morrison and Cutler, 1997). On activation, Raf then phosphorylates and activates two MAPK kinases (MAPKKs; also called MEKl and MEK2). The MEKs directly associate with the COOH-terminal catalytic domain of Raf 1 and are phosphorylated by Raf (Crews and Erikson, 1993). Activated MEKs, which function as dual-specificity kinases, phosphorylate tandem threonine and tyrosine residues (TEY motif) in two MAPKs (also referred to as extracellular signal-regulated kithem nases; ERKs), designated p42MAPK/Erk2and ~ 4 4 ~ * ~ ~ / Etor activate kl, (Crews et al., 1992). Once activated, MAPKs translocate to the nucleus, where they phosphorylate and activate a variety of substrates, including the Elkl nuclear transcription factor (Marais et al., 1993). Elkl forms a complex with serum response factor (SRF) at the serum response DNA element (SRE) present in many promoters, such as the c-fos promoter (Treisman, 1996a). MAPKs also activate other kinases, such as the p90-ribosomal S6 kinase (RSK) serinehhreonine kinase, to regulate protein synthesis (Blenis, 1993). Much of the information that has allowed the delineation of this Ras signaling pathway has come from genetic studies of the fruit fly Drosophila

62

Roya Khosravi-Far e l of.

melanogaster, the nematode Caenorhabditis elegans, the baker’s yeast Saccharomyces cerevisiue, and the fission yeast Schizosaccharomyces pombe (Fig. 2). Drosophila eye development is controlled by a signaling pathway that begins with ligand stimulation of the Sevenless RTK, leading to the activation of a Ras homolog, which in turn activates an MAPK cascade (Rubin, 1991). A near-identical scheme is seen for the pathway that regulates proper development of the vulva in C. eleguns, which involves an RTK (Let23)-mediated pathway that activates the Ras homolog, Let-60 ( Sternberg and Horvitz, 1991). S . pombe Rasl also activates a MAPK cascade (Nishida and Gotoh, 1993). In fact, many of the early clues that identified the pos-

Mammals

C. elegans

Drosophila

pq 1-

1(

Growth and Differentiation

1

1-

t

t

t

i

i

.)

1 pLZl

4

4

I EYe Development

INIPK-11 .) Vulva Development

Fig. 2 Conservation of 3 Kas signaling cascade in mammals and simple eukaryotes. Like mammalian cells, homologs of Kas also serves as downstream components of RTK-mediated developmcntal pathways i n Drosophi/u melurtoguster and C. eleguns. Ilrosophi/a melanop~tcv Drasl is a regulator of thc Sevenless KTK pathway that is important f o r proper eyc development. C. elega~zsLet-60 is a regulator of the Kossllxt-23 RTK pathway that is essential for proper vulva development.

Increasing Complexity of Ras Signal Transduction: Involvement of Rho Family Proteins

63

itive (CDC25) and negative (IRA) regulators of Ras GDP/GTP cycling came from RAS studies in S . cerevisiae (reviewed in Tamanoi, 1988; Broach, 1991). The remarkable evolutionary conservation of this Ras signaling cascade reflects the central role of Ras in diverse organisms.

B. Raf Is a Key Effector of Ras Considerable evidence supports the critical role of Raf and the MAPK cascade in Ras function. For example, dominant negative mutants of Rafl, MEK, and MAPKs have been shown to impair Ras-transforming activity (Kolch et al., 1991; Schaap et al., 1993; Cowley et al., 1994; Westwick et al., 1994; Qui et al., 1995a; Khosravi-Far et al., 1995). Similarly, loss of function mutations in Raf, MEK, or MAPK homologs disrupt these Ras-mediated developmental pathways in flies and worms (reviewed in Satoh et al., 1988). Conversely, constitutively activated mutants of Rafl cause a transformed and tumorigenic phenotype that is indistinguishable from that caused by oncogenic mutants of Ras, when analyzed in rodent fibroblast transformation assays (Bonner et al., 1985; Stanton et al., 1989; Leevers et al., 1994; Stokoe et al., 1994). Constitutively activated mutants of MEK also cause morphologic and growth transformation of NIH 3T3 cells (Alessi et al., 1994; Mansour et al., 1994). Gain of function mutations in homologs of Raf, MEK, or MAPK can overcome the loss of Ras function in the fruit fly and nematode (Dickson et al., 1992; Sprenger et al., 1993). Constitutively activated Raf can overcome the loss of Ras function caused by dominant negative mutants of Ras [e.g., Ras(17N)] or by the Y 13-259 anti-Ras neutralizing monoclonal antibody (Smith et al., 1986; Feig and Cooper, 1988). Finally, when it was demonstrated that targeting Rafl to the plasma membrane alone was sufficient to unmask its transforming activity fully (Leevers et al., 1994; Stokoe et al., 1994), it was suggested that the biochemical function of Ras was simply to promote the membrane association and activation of Raf. When taken together, these observations supported the likelihood that the Raf+MEK+MAPK cascade was both necessary and sufficient for Ras function. Despite the apparent linear nature of the Ras+Raf+MEK+MAPK signaling cascade, there was mounting evidence that it represented a mere subset of a complex array of signaling interactions a t several levels. For example, the inability of Ras interaction with Raf alone to promote the activation of its kinase function indicated that other signals were needed to converge a t the point of Raf. The observation that Ras interacts with two distinct NH,-terminal regions of Rafl suggests that Ras promotes more than just membrane translocation of Raf; it may also facilitate the subsequent events that lead to Rafl activation (Fig. 3) (Brtva et al., 1995; Drugan et al., 1996; Hu, C. et al., 1995). Other components that may contribute to Rafl activa-

64

Roya Khosravi-Far el a / .

tion include at least two isoforms of 14-3-3 proteins, the Hsp9O and p50 molecular chaperones, phospholipids, serine/threonine and tyrosine kinases, and the kinase suppressor of Ras (KSR) (reviewed in Morrison et al., 1997).The fact that 14-3-3 can exist as a homodimer and is known to complex with a variety of proteins (e.g., BCR, Cdc25, A20, PKC) suggests that it may act as an adaptor protein and adds further complexity to the mechanism of Raf activation (Braselmann et a/., 1995; Conklin et al., 1995; Vincenz et al., 1996). Thus, 14-3-3 may bring in other proteins that promote Raf activation or may promote Raf-Raf dimerization (Farrar et al., 1996; Luo et al., 1996). Therefore, the connection between Ras and Raf alone is not simply linear and requires multicomplex formation to complete Raf activation. As discussed later, this theme of effector-mediated multicomplex formation also may he extended to other Ras effectors and to effectors of Rho family proteins. Other observations also revealed the oversimplification of a Raf+MEK+MAPK linear cascade. Raf can be activated by Ras-independent mechanisms, for example, by the Src tyrosine kinase (Fabian et al., 1994).Similarly, protein kinase C (PKC) can phosphorylate and activate Raf directly (Morrison, 1994; Sozeri et al., 1992; Kolch et al., 1993). MAPKs can be activated by Ras-independent mechanisms. Integrin-mediated activa-

Fig. 3 Ras-mediated activation of Raf is a complex multistep process. Although Ras-GTP activation of Raf is mediated by causing the translocation of Raf from the cytosol to the inner surface of the plasma membrane, other events are required t o activate Kaf kinase function. Ras, through interaction with a second distinct binding site in the cysteine-rich domain of Raf (Cys), may also promote other events that allow the subsequent activation steps to occur. 14-3-3 interacts with multiple regions of Raf and may serve as a negative regulator. Other possible components that may be involved in Raf activation include KSR, phosphatidylserine lipids, p50 and HspSO, and both tyrosine and serinehhreonine kinases. The requirement for multiple signals to activated Raft underscores an emerging theme that effectors may promote complex formation at the level of Ras, rather than simply forming a linear signaling pathway.

Increasing Complexity of Ras Signal Transduction: Involvement of Rho Family Proteins

65

tion of MAPKs can occur by a process independent of Ras (Chen, Q. et al., 1996). MEK-transforming activity, however, may still be dependent on Ras function (Cowley et al., 1994). Furthermore, MAPKs can phosphorylate MEK, although the functional consequences are not clear. Raf and MEK can cause activation of the p70 S6 kinase, but apparently independent of p42/p44 MAPK activation (Lenormand et al., 1996). Taken together, each signaling protein in this cascade is likely to have additional targets, and each component may be activated by other upstream components. Furthermore, feedback loops and autocrine mechanisms may also exist. For example, Raf activation of a second MAPK pathway, involving the Jun NH,-terminal kinase (JNK; also known as SAPK) was found to be mediated by an autocrine loop involving the EGF receptor (Minden et al., 1994). Most important, observations have begun to establish that Raf is not the sole downstream target of Ras and that Ras function may require its association with a surprisingly large number of other effectors (Boguski and McCormick, 1993) (see discussion that follows).

Ill. Ras ACTIVATION OF Raf-INDEPENDENT PATHWAYS CONTRIBUTES TO Ras TRANSFORMATION A considerable body of evidence supports the hypothesis that Raf is not the sole effector of Ras function and that Raf-independent pathways are critical for mediating Ras function. First, genetics studies in S. pombe revealed the involvement of a t least two distinct downstream effector-mediated signaling pathways that facilitate full Rasl function (Fig. 4) (Marcus et al., 1995).One Rasl effector is Byr2, a MEKWRaf homologue (Van Aelst et al., 1993; Wang et al., 1991). Byr2 is a component of a protein kinase cascade that includes Byrl (MEK) and Spkl (MAPK); this cascade regulates agglutination, conjugation, and sporulation. Scdl represents a second effector and functions as an exchange factor for the S. pombe CDC42 homolog, Cdc42 (Chang et al., 1994). Scdl regulates a pathway that includes CDC42 and Shkl (PAK homolog) and controls cell morphology (Marcus et al., 1995).Finally, although the best characterized effector of S. cerevisiae RAS2 is not a Raf-related molecule, but instead is adenylate cyclase, a second Ras-dependent pathway was suggested by the fact that loss of RAS, but not of adenylate cyclase, is lethal (Toda et al., 1985). RAS2 also signals by a CDC42/STE20 MAPK cascade to induce filamentous growth (Mosch et al., 1996). Second, in addition to p42 and p44 MAPKs/ERKs, activated Ras induces activation of the JNWSAPK and p38/HOG MAPK cascades, which in turn activate the Elkl, c-Jun, and ATF2 nuclear transcription factors (Fig. 5 ) (Dtrijard et al., 1994). Although these kinase cascades are analogous to the

66

Roya Khosravi-Far e l al.

J

a (RhoGEF)

0 Cdc42

a (MAPK)

Mating Responses, Meiosis and Sporulation

Cell Morphology

Fig. 4 S. pombr Rasl regulates mating and cell morphology through the activation of two distinct signaling pathways. Rasl regulates sexual differentiation (conjugation and sporulation) and agglutination through activation of a pathway that is analogous to the Raf-+MEK-+MAPK cascade. Ras 1 regulation of cell morphology is regulated by a distinct pathway that bifurcates at the level of Ras and involves a pathway leading to the Cdc42 Rho family protein. Scdl (an exchange factor for Cdc42) exhibits properties of a Ras effector and is a regulator of the a pathway that includes CDC42 and Shkl (I’AK homolog).

Raf-+MEK+MAPK cascade (reviewed in Treisman, 1996b; Kyriakis and Avruch, 1996), they define parallel pathways that are stimulated independent of Raf activation (Minden et al., 1994; Olson et al., 1995). However, whereas ERK activation is associated with growth stimulatory responses, JNK and p38 activation is typically associated with stress responses that result in apoptosis (Xia et al., 1995; Verheij et al., 1996; Chen, Y. R. et al., 1996; Yang et al., 1997). For example, MEKKl activates both JNK and p38 and causes apoptosis (Johnson et al., 1996). Similarly, activation of JNKs is associated with apoptosis in PC12 pheochromocytoma, U937 leukemia, and other cell types (Xia et al., 1995; Verheij eta/., 1996; Chen, Y. R. et al., 1996; Yang et al., 1997). We observed, however, that inhibition of JNK activation

Increasing Complexity of Ras Signal Transduction: Involvement of Rho Family Proteins

67

impaired Ras transformation, suggesting a growth-promoting role for these two MAPK cascades (Clark et al., 1997b).Jun function is also required for Ras transformation (Granger-Schnarr et al., 1992; Johnson, R. et al., 1996). Therefore, JNK activation may promote different cellular consequences that depend on the coordinate activation of other pathways. JNK activation by stress stimuli alone may cause apoptosis, whereas JNK activation under other conditions may synergistically enhance the growth-promoting action of activated MAPKs (Xia et al., 1995). Alternatively, JNK activation may possess cell-type-specific actions (Dtrijard et al., 1994; Xia et al., 1995; Verheij et al., 1996; Chen, Y. R. et al., 1996; Su et al., 1994) and serve a protective function, at least in fibroblasts, thereby contributing to cell transformation. Ras stimulates yet a third distinct cascade that causes activation of a Fos nuclear kinase (FNK) and Fos function (Deng and Karin, 1994). Third, transformation studies in several epithelial cell lines, including RIEl rat intestinal epithelial cells, showed that Ras activation of the Raf+MEK+MAPK pathway alone was not sufficient to cause complete transformation (Oldham et al., 1996). Whereas constitutively activated mutants of Ras and Raf show equivalent transforming potencies when assayed in rodent fibroblast transformation analyses, activated Raf failed to cause morphologic and growth transformation of RIEl cells. Instead, Ras activation of Raf-independent pathways, which led to the induction of an EGFRdependent autocrine growth pathway, was found to be critical for oncogenic

I

t

I

Gene Expression

Fig. 5 Ras interacts with multiple effectors. In addition to the three Raf serindthreonine kinases (Rafl, A-Raf, and B-Raf), a number of other proteins share properties of a Ras effector and bind preferentially to active Ras-GTP. Interaction is dependent on an intact core effector domain.

68

Roya Khosravi-Far e1 a/.

Ras transformation of these cells. Other evidence for the inability of Raf to promote full Ras function was provided by a study that showed that Ras, but not Raf, could trigger the morphologic changes and actin reorganization associated with cellular hypertrophy (Thorburn et al., 1993). Fourth, constitutively activated mutants of two members of the Ras branch of Ras-related proteins (TC2UR-Ras2 and R-Ras; 5 5 % identity with Ras) can cause tumorigenic transformation of rodent fibroblasts (Fig. 6 ) (Graham et al., 1994; Chan et al., 1994; Saez et al., 1994; Cox et al., 1994). Although TC21 and R-Ras share complete sequence identity with the core Ras effector domain sequence (residues 32 to 40) and can interact with an isolated Ras-binding domain present in the NH2-terminus of Rafl, these proteins failed to interact with and activate Raf kinases in vivo or to activate ERKs directly (Graham et al., 1996; Huff et al., 1997). In particular, we have found that activated TC21 exhibits the same differentiation-inducing (PC12

Ras TC21 R-Ras RaplA

Effector Interaction

Fig. 6 Ras effector interaction sequences. Multiple, distinct sequences of Ras proteins are involved i n interactions with Raf and other candidate Ras effector targets. The core Ras sequence domain (residues 3 2 to 40) may represent a common interaction site for a l l Ras-CTP-binding proteins and shows complete identity with the equivalent sequences of TC21/R-Ras2, R-Ras, and Rap proteins. Sequences that flank this core sequence most likely determine the specificity of effector binding because differences are seen in the closely related hut biologically distinct Rap and R-Ras proteins. Other sequences that may influence effector interactions have been identified in the switch I1 domain and in residues 92 through 106.

Increasing Complexity of Ras Signal Transduction: Involvement of Rho Family Proteins

69

cells) or differentiation-inhibiting (C2 myoblasts) phenotype that has been observed with activated Ras (Graham and Der, unpublished observation). Therefore, Raf-independent pathways are clearly sufficient to mediate the transforming and differentiating functions of TC2 1 and, presumably, Ras. Although the Raf-independent pathways that promote transformation by these two Ras-related proteins are not known, it is suspected that Ras also activates these same Raf-independent pathways. This hypothesis is based in part on the fact that TC21 and R-Ras can interact with many of the RasGTP binding partners that have been identified (discussed later). Finally, important and direct evidence for the involvement of multiple signaling pathways downstream of Ras comes from the identification of Ras effector domain mutants that are defective in interaction and activation of Raf (White et al., 1995). These Raf-binding defective mutants, designated Ras( 12V, 37G) and Ras(12V, 40C), no longer activate Raf or MAPKs but retain the ability to bind to other putative Ras targets (Khosravi-Far et al., 1996).Surprisingly, although these effector mutants did not cause significant morphologic transformation when stably expressed in NIH 3T3 cells, they promoted a growth-transformed phenotype (growth in low serum, colony formation in soft agar, and tumor formation in nude mice) similar to that caused by oncogenic Ras. These mutants caused a transformed phenotype distinct from that caused by Raf, which was similar to that caused by constitutively activated mutants of Rho family proteins. Furthermore, coexpression of these Ras-effector domain mutants with activated Raf resulted in potent synergetic transforming activity. The non-Raf effectors that mediate the transforming actions of these two mutants have not been established. Because Ras(12V, 40C) caused a transformed morphology similar to Rho proteins, retained the ability to activate JNK (Khosravi-Far et al., 1996),and induced membrane ruffling in REF52 cells, which could be blocked by dominant negative Racl (Joneson et al., 1996b), an effector pathway leading to Racl may be important for the transforming activity of this mutant protein. A second study suggested that Ras(l2V, 40C) may retain the use of phosphoinositide 3-kinase (PI3K) as an effector, although these authors did not observe any focus-forming activity with this mutant (Marte et al., 1996). Finally, two studies suggested that Ras(l2V, 37G) may mediate its actions through RalGDS (White et al., 1996; Kauffmann-Zeh et al., 1997).

IV. Ras MEDIATES ITS ACTIONS THROUGH INTERACTION WITH MULTIPLE EFFECTORS The existence of Raf-independent Ras signaling pathways is further insinuated by the expanding roster of candidate Ras effectors that has emerged during the past several years (see Fig. 5 ) (Van Aelst et al., 1994; Marshall,

70

Roya Khosravi-Far ef al.

1996). Like Rafl , these structurally and functionally diverse proteins show preferential binding to the active GTP-bound form of Ras, and this interaction requires an intact core Ras effector domain (see Fig. 6). These candidate effector molecules have been identified by a variety of experimental approaches, including biochemical, yeast two-hybrid library screening, and yeast functional screening. It is likely that additional effectors will be identified. In this section, we summarize our current understanding of the involvement of these candidate Ras effectors in mediating Ras function. We also speculate on the structural requirements for Ras interaction with such a diverse spectrum of functionally divergent targets.

A. A Growing Roster of Candidate Effectors of Ras Although p120 GAP and NFllneurofibromin clearly function as negative regulators of Ras function (Boguski and McCormick, 1993),p l 2 0 GAP represented the first candidate Ras effector (Adari et al., 1988; Calks et al., 1988). Apparent evidence against an effector function for Ras GAPs, however, came from studies of Ras effector domain mutants that were impaired in p120 o r N F l interaction yet that retained strong focus-forming activity (Marshall and Hettich, 1993). These observations, together with evidence establishing Raf as a true Ras effector, signaled the demise of the notion that GAPs serve as effectors of Ras. Because Raf-binding-deficient mutants of oncogenic Ras retain transforming potential (Khosravi-Far et al., 1996), it has become apparent that multiple effectors are required for full Ras function. Therefore, the loss of binding of any one effector alone may not cause a significant alternation to Ras-transforming activity. Thus, these effector domain mutagenesis results alone do not rule out an effector role for GAPs. Instead, a significant body of evidence has established intriguing, but incomplete, support for p120 GAP as a Ras effector (Tocque et al., 1997). Evidence for p120 GAP effector function was derived from studies that evaluated the activity of NH,-terminal fragments of p120 GAP that lack the catalytic GTPase stimulatory domain and Ras-binding sequences. The NH,terminus of p120 GAP contains an SH3 domain flanked by two SH2 domains, a pleckstrin homology (PH) domain, a calcium-binding domain, and a site of interaction with phospholipids. In one study, a Ras-pl20 GAP complex was required for inhibition of muscarinic receptor-activated potassium channel opening, whereas an NH,-terminal fragment of p120 GAP could cause this independent of Ras (Yatani etal., 1990; Martin et al., 1992). In a separate study, an antibody directed against the SH3 domain of p120 GAP blocked Ras-induced germinal vesicle breakdown, but not MAPK activation, when assayed in Xenoptrs oocytes (Pomerance et al., 1996; Hartwell, 1992). Similarly, coexpression of a NH,-terminal fragment was shown to block

Increasing Complexity of Ras Signal Transduction: Involvement of Rho Family Proteins

7I

oncogenic H-Ras-transforming activity (Clark et al., 199313; Habets et al., 1994). Furthermore, this fragment blocked Ras activation of JNK but not of MAPKs. Another study found that overexpression of a p120 GAP NH2-terminus fragment altered cell morphology and organization of the actin cytoskeleton (McGlade et al., 1993). These observations, together with others (reviewed in Tocque et al., 1997), suggest that p120 GAP may serve dual roles as both a negative regulator of Ras and as a scaffolding protein that promotes formation of a Ras-GTP-dependent signaling complex with Ras. A working model proposes that Ras-GTP binding to the COOH-terminal catalytic domain would expose the protein-ligand-binding motifs pres-ent in the NH2-terminus, and that the subsequent association of components with the NH,-terminus would activate a signaling function. The identification of a number of p120 GAP-binding proteins (e.g., p190 Rho-GAP, G3BP) provide support for such a scenario (Ellis et al., 1990; Parker et al., 1996; Yamanashi et al., 1997; Carpino et al., 1997). Evidence for N F l as a Ras effector is less compelling and is based largely on detection of N F l function beyond its role as Ras-GTPase stimulator. Like p120 GAP, a significant portion of the 220-kd NF1 protein is unrelated to its GTPase catalytic function and, consequently, may be involved in an effector function. Clues to such a function are limited, but they suggest an effector function involved in growth suppression rather than promotion. For example, no elevation in Ras-GTP was seen in a number of NFI-deficient tumors (Andersen et al., 1993; The et al., 1993; Johnson et al., 1993), yet introduction of full-length NF1 into these cells resulted in severe reductions in growth (Johnson et al., 1994).These observations suggest that a loss of NF1 function, other than as a Ras GTPase stimulator, contributes to the development of these tumors. In further support of this hypothesis, two separate studies showed that NF1 overexpression inhibited the transformed growth properties of cells that expressed constitutively activated Ras mutants. This would be unexpected if NF1 served simply as a GAP because the GTPase activity of mutant Ras is not stimulated by NFI. First, Johnson and colleagues (1994) showed that overexpression of neurofibromin drastically reduced (five-fold) viral H-Ras, but not Raf, focus-forming activity. Second, Li and White (1996) found that the introduction of full-length NF1 into H C T l l 6 colon carcinoma cells, which harbor mutated K-Ras, suppressed soft agar growth in vitro and tumor formation in nude mice. NF1 overexpression was found to reduce MAPK activation, and introduction of activated Raf could overcome the NF1 suppression (Li and White, 1996). Thus, these investigators suggested that only modest elevations (two-fold to three-fold) of NF1 expression can block Ras activation of Raf. Taken together, a role for NF1 as an effector for Ras functions other than transformation is suggested. Such a role may also explain, in part, why Ras mutants impaired in N F l binding retain potent transforming activity. Thus, N F l may serve as an effector that

72

Roya Khosravi-Far el al.

mediates the differentiating actions of oncogenic Ras seen in certain cell types, such as PC12 pheochromocytoma cells (Bar-Sagi and Feramisco, 1985; Noda et al., 1985). In summary, the jury is still out with regard to an effector function for Ras GAPS. p120 CAP- or NF1-deficient mice have been described, and both lead to death during embryogenesis (Henkenmeyer et al., Bollag et al., 1996). The analysis of whether Ras-transforming potential is impaired in fibroblasts derived from the embryos of these animals should provide further assessment of whether Ras-GAPS are important effectors of Ras transformation and may identify the signaling pathways that they regulate. The Ral guanine nucleotide dissociation stimulator RalGDS) and two closely related proteins (RGL and RGL2/RIf) represent intriguing candidate effectors of Ras that may link Ras with other Ras-related proteins. Members of this family have been identified repeatedly by yeast two-hybrid library screening searches for candidate effectors of Ras and Ras-related proteins (Rap, R-Ras, and TC21/R-Ras2) (Kikuchi et al., 1994; Hofer et al., 1994; Spaargaren and Bischoff, 1994; Wolthuis et al., 1996; Peterson et al., 1996; Lopez-Barahona et al., 1996). RalGDS was identified originally as a CDC25related protein (yeast Ras GEF), but was found to function instead as a GEF for the two Ras-related proteins RalA and RalB (85% identical to each other) (see Fig. 5 ) (Albright et al., 1993).The COOH-terminal noncatalytic domains of RalGDS, RGL, and RGL2/Rlf interact with the effector domain region of Ras in a GTP-dependent manner in vitro. These RalGDS family proteins can compete with Rafl for binding to the Ras effector domain region in vitro. Ras association with RalCDS has also been observed when overexpressed in COS cells (Kikuchi and Williams, 1996). Furthermore, the GTPase activity of both p120 GAP and NFI-GAP were inhibited by RalGDS (Kikuchi et al., 1994), suggesting that RalGDS, like Rafl, interacts with Ras in a GTP-dependent manner through its effector domain. Although RalGDS and related proteins also bind to Ras-related proteins, transient expression assays in COS cells showed that Ras, but not RaplA or R-Ras, enhanced the Ral-CEF activity of RalGDS (Urano et al., 1996). This observation underscores an important caution with regard to the interpretation of data showing the ability of a particular protein to bind in vitro. Whether binding occurs in vivo and whether this leads to stimulation of effector function needs to be established to validate a true interaction. Evidence that RalGDS may serve as a positive regulator of Ras transformation comes from several independent observations. First, coexpression of isolated Ras-binding domains of RGL and RGL2/RIf inhibited Ras-transforming activity but not Raf-transforming activity, in NIH 3T3 cells (Okazaki et al., 1996; Peterson et al., 1996). Second, although constitutively activated Ral alone does not cause transformed foci, one study reported that its coexpression enhanced Ras transformation, whereas dominant negative Ral

Increasing Complexity of Ras Signal Transduction: Involvement of Rho Family Proteins

73

impaired Ras focus-forming activity (Urano et al., 1996).These effects, however, were not dramatic, and a second study did not find significant regulation of Ras-transforming activity by mutant RalA o r RalB proteins (White et al., 1996). We also found no significant regulation of Ras focus-forming potential by mutants of RalB (Jordan and Der, unpublished observation). Third, coexpression of RalGDS cooperated synergistically with activated Rafl to induce transformation of NIH 3T3 cells (White et al., 1996). Although RalGDS alone showed no focus-forming activity, we found that coexpression of RalGDS caused a three-fold enhancement of R-Ras, but not of Ras or TC21, focus-forming activity (Brtva and Der, unpublished observation). It was not established whether the growth-promoting activity seen with RalGDS was due to its activation of Ral proteins. It is conceivable that RalGDS activation of Ral proteins may in turn lead to activation of either CDC42IRac (discussed later) or phospholipase D (Jiang et al., 1995). RalGDS, however, may also have functions distinct from its Ral GEF activity. PI3K is a lipid kinase that phosphorylates phosphoinositides at the 3' position of the inositol ring (Carpenter and Cantley, 1996), it has also been implicated as a Ras effector (see Fig. 5).PI3K is composed of a p l l 0 catalytic and a p85 regulator subunit, and both are members of a family of related proteins. Recombinant p l 1 0 (01 and p) showed high-affinity interaction with the CTP-bound form of recombinant Ras through the Ras effector domain (Rodriguez-Viciana et al., 1994; 1996).Furthermore, dominant negative Ras inhibited platelet-derived growth factor (PDGF) activation of PI3K. In addition, introduction of activated Ras, but not Raf, into COS cells resulted in a significant induction of inositol phosphates, demonstrating the convergence of Raf and PI3K pathways a t the level of Ras. Taken together, these observations suggest strongly the involvement of PI3K as a downstream effector of Ras function (Rodriguez-Viciana et al., 1994). R-Ras, but not RaplA, has also been shown to bind to and activate PI3K (Marte et al., 1996). Finally, in addition to the a and p isoforms, a novel p l l 0 isoform designated p l l 0 C! was also identified during a yeast two-hybrid library screen for Ras-interacting proteins (Vojtek et al., 1993; Vojtek and Cooper, unpublished observation). One downstream target of activated PI3K is protein kinase B (PKB; also Akt; see Fig. 5) (Franke et al., 1995; Klippel et al., 1996; Marte et al., 1996). A study using Ras effector domain mutants implicated PI3WPKB in oncogenic Ras suppression of Myc-induced apoptosis in Rat 1 fibroblasts, whereas Ras activation of the Raf+MEK+MAPK pathway was found to promote the Myc apoptotic response (Kauffmann-Zeh et al., 1997). Because oncogenic Ras also potentiated the apoptotic response, the Raf-mediated pathway may be dominant over the PI3K pathway. On the other hand, several observations implicated PI3K as an upstream activator of Ras function. First, phosphorylation of PDGF receptor at skes

74

Roya Khosravi-Far el nl.

that are involved in p85 binding is required for the activation of Ras by PDGF (Fantl et al., 1992). Second, an activated form of p l l 0 caused a small elevation in the level of activated Ras-GTP in vivo (Hu, Q. et al., 1995). One possible resolution to these apparently contradictory roles of PI3K is that, depending on the specific extracellular stimulus o r cell type, PI3K may act as either an upstream activator or a downstream mediator of Ras function. Another candidate effector of Ras is the mitogen-activated protein kinase kinase kinase 1 (MEKKl), a serine-threonine kinase that is an upstream activator of SEK (a MAPKK), which in turn activates JNK/SAPK (see Fig. 5 ) (Lange-Carter et al., 1993; Yan et al., 1994). MEKKl can be activated in response to a variety of extracellular stimuli, ultraviolet radiation, and activated Ras. Evidence for its role as a Ras effector comes from observations that MEKKl binds directly to GST-Ras( 12V) in a GTP-dependent manner through its COOH-terminal kinase domain in vitro (Russell et al., 1995).This binding was blocked by a Ras effector peptide. Whether MEKKl is a true Ras effector remains to be critically tested. Because overexpression of MEKK 1 causes apoptosis (Johnson, N. L. et al., 1996), it appears unlikely that it would be an important positive effector for Ras-transforming activity. AF6/Rsb1 is yet another candidate Ras effector that was identified by yeast two-hybrid library screening for Ras-binding proteins and by affinity chromatography purification of Ras-GTP-binding proteins (see Fig. 5 ) (Van Aelst et al., 1994; Kuriyama et al., 1996). AF6 shows a high degree of sequence similarity to Drosophila Canoe, which is assumed to function in signaling pathways downstream from the Notch receptor, acting as a regulator of cellular differentiation (Hunter, 1997).The p l 8 0 AF6 protein was also identified independently as one of up to 10 fusion partners of the MLL protein associated with chromosome translocation events in human leukemias (Prasad et al., 1993). The MLL/AF6 chimeric protein is the gene product of a reciprocal translocation t( 6;l l)(q27;q23) associated with a subset of human acute lymphoblastic leukemias (Prasad et al., 1993; Tanabe et al., 1996; Taki et al., 1996). Transcripts for the reciprocal AF6/MLL fusion have not been detected (Tanabe et al., 1996; Taki et al., 1996). Interestingly, all MLL/AF6 fusion proteins contain the identical AF6 residues, 36 to 1612. In vitro analyses showed that the NH,-terminal domains of AF6 and Canoe interacted specifically with GTP-bound form of Ras, and that this interaction interferes with the binding of Ras to Raf (Kuriyama et al., 1996). AF6 appears to be expressed in a variety of tissues. A function for AF6 has not been determined; however, limited sequence homology was observed between AF6 and proteins that may be involved in signal transduction at special cell-cell junctions. Yeast studies looking for genes that interfere with Ras function identified the Ras interactionlinterference gene 1 (Rinl ) that suppressed activated RAS2-induced cyclic AMP activation (Han and Colicelli, 1995). This inter-

Increasing Complexity of Ras Signal Transduction: involvement

of Rho Family Proteins

75

ference is at the level of RAS because Rinl cannot suppress the activity of components downstream of Ras in the adenylyl cyclase pathway. Additionally, in vitro binding and yeast two-hybrid analysis demonstrated that Rinl directly interacted with RAS2 as well as the mammalian H-Ras. This interaction was GTP-dependent and required a n intact Ras effector domain. Furthermore, Rinl competed with Rafl in binding to Ras in vitro (Han and Colicelli, 1995). Although overexpression of Rinl alone in NIH 3T3 cells showed n o growth-promoting activity, coexpression of Rinl with activated R-Ras showed a synergistic enhancement of focus-forming activity (Brtva, Colicelli, and Der, unpublished observation). Thus, Rinl may be involved in positive growth regulation. The 5 isoform of protein kinase C (PKCS) has also been implicated as a Ras effector. First, PKCC was required for Ras-induced maturation in Xenopus oocytes (Dominguez et al., 1992) and for serum-stimulated mitogenic signaling in mammalian cells (Berra et al., 1993). Second, the regulatory NH,-terminal fragment of PKCC showed preferential binding to Ras-GTP in vitro, and a peptide corresponding to Ras residues 17 to 44 blocked this interaction (Diaz-Meco et af., 1994). Furthermore, PDGF stimulation promoted PKCC association with Ras in vivo. Finally, dominant negative HRas( 17N) blocked PDGF-stimulated activation of PKCC. Taken together, these observations suggest that PKCC may serve as a positive regulator of Ras growth stimulation. In summary, a growing family of candidate Ras effectors has emerged; however The precise roles, if any, of these Ras-GTP-binding proteins in mediating Ras downstream signal transduction and transformation remain to be established. Some of these proteins may be involved in mediating normal Ras function, whereas others may be critical mediators of Ras-transforming activity. A subset of these proteins may actually be negative regulators of Ras function by preventing the interaction of Ras with bona fide Ras effectors. Finally, a portion of these candidate effectors may serve not as true Ras effectors, but rather as effectors for one o r more of the closely related TC21, R-Ras, o r Rap proteins. Analysis of the role of these proteins in the function of Ras and Ras-related proteins is the focus of ongoing investigations in many laboratories.

B. Structural Requirements

for Ras-Effector Interactions The current roster of candidate Ras effectors is composed of a diverse collection of structurally and functionally distinct proteins. All recognize the GTP-bound form of Ras by the core Ras effector domain, which suggests the possible existence of a common Ras-GTP recognition element. We demon-

76

Roya Khosravi-Far e1 of.

strated that the cysteine-rich domain of Rafl served as a second Ras-binding site and contained a small eight-amino-acid sequence that exhibited strong sequence similarity with sequences present in a subset of other candidate Ras effectors (Ras-GAPS and RalCDS) (Clark et al., 1995). Mutagenesis of these residues in Rafl, however, did not impair Ras-binding (Clark, Drugan, and Campbell, unpublished observation). Thus, this sequence is not a consensus Ras-binding sequence, as we had originally suggested. A distinct Ras-GTP-binding motif has been proposed based on limited sequence and predicted structural similarity between RalCDS and AF6, designated the RA domain (Ponting and Benjamin, 1996). The RA domain is present in a number of candidate Ras effectors (RalGDS, AF6/Canoe, Rinl ) , but not in others (Ras-GAPS, PI3K, MEKKI). Although the alignment presented in this work is intriguing, it has not been demonstrated whether the RA domain is a Ras-binding motif. Thus, it remains to be determined whether candidate Ras effectors share a common primary o r tertiary structure that defines their shared Ras-binding properties. One additional complexity to this issue is the likelihood that the various Ras-binding proteins interact with distinct and multiple sequences in Ras (see Fig. 6 ) (Wittinghofer and Nassar, 1996). Because a common feature of all of these proteins is the loss of binding due to mutations in the core Ras effector domain (32 to 40), this region may represent a shared binding region important for all Ras effectors. The conformation of Ras-CTP differs from that of Ras-GDP in two distinct regions, designated switch I (Ras residues 30 to 37) and switch I1 (Ras residues 59 to 76) (Tong et al., 1989; Milburn et al., 1990; Krengel et al., 1990). Because switch I overlaps with the Ras effector domain sequences, it is not surprising that this region is important for effector interaction. Mutagenesis analyses have shown, however, that residues flanking the core effector region (spanning Ras residues 26 to 45) can also influence Ras-transforming activity and Ras interaction with candidate effectors (Shirouzu et al., 1994; Fujita-Yoshigaki et al., 1995). Similarly, mutagenesis of switch I1 residues also support their importance in interaction with at least some candidate effectors (Moodie et al., 1995; Hwang et al., 1996; Drugan et al., 1996). For example, Y64GIS65G or Y64G/Y71C double mutants of H-Ras were impaired in complex formation with NF1 or PI3K (Moodie et al., 1995). Furthermore, mutation of residues 60 or 64 also perturbed Ras interaction with Rafl (Drugan etal., 1996).The Q61L oncogenic mutation increases affinity for both GAPS (Vogel et al., 1988). Whether switch I1 is important for interaction with all Ras effectors is not known. Mutation of yet other residues (e.g., residues 92 o r 106) suggests a third region of contact between Ras and Ras-GAPS and possibly other effectors (Morcos et al., 1996; Parrini et al., 1996; Yoder-Hill et al., 1995). Furthermore, COOH-terminal farnesylation of Ras appears to be important for

Increasing Complexity of Ras Signal Transduction: Involvement of Rho Family Proteins

77

high-affinity interaction with Raf, suggesting that the Ras COOH-terminus may also be involved in effector interaction (Hu, C. et al., 1995; Luo et af., 1997). Thus, the original definition of the Ras effector domain (Sigal et al., 1986) is clearly a n oversimplification. For example, the determination of the crystal structure of the catalytic domain of p120 GAP suggests that it interacts with four distinct regions of Ras (Scheffzek et al., 1996). This is consistent with mutagenesis studies suggesting that interactions between Ras and p120 GAP differ from those between Ras and Rafl. Therefore, at least four distinct regions of Ras appear to be involved in effector interactions. The fact that Ras effectors possess similar, but not identical, requirements to Ras can be exploited to generate mutants of Ras that are selectively impaired in their interaction with only a subset of effectors (White et al., 1995; Joneson et al., 1996b; Khosravi-Far et al., 1996; Akasaka et al., 1996). Such approaches have proved invaluable for defining the specific contribution of Raf to Rastransforming activity. Ras-related proteins may shed some light on the Ras sequences and effectors important for Ras-transforming activity. Although several members of the Ras branch of the Ras superfamily share complete identity with Ras residues 32 to 40 and can also bind to many Ras-GTP-binding proteins, they exhibit distinct biochemical and biologic properties. For example, although Ras and TC2 1/R-Ras2 share indistinguishable biologic properties, TC21 is not an activator of the R a f b M E K j M A P K pathway (Graham et al., 1996). R a p l N K r e v l binds to Raf, GAPS, RalGDS, and other candidate effectors of Ras (Zhang et af., 1993; Frech et al., 1990; Hata et al., 1990; Kikuchi et al., 1994; Hofer et al., 1994; Wolthuis et al., 1996; Spaargaren et al., 1994; Peterson et al., 1996; Lopez-Barahona et al., 1996), but does not activate Raf or cause transformation. Instead, its Ras-inhibiting properties (Kitayama et al., 1989; 1990) have been attributed to its ability to compete with Ras for key effectors important for Ras transformation by forming nonproductive complexes that are sequestered at intracellular membranes away from Ras (Beranger et al., 1991; Sat0 et al., 1994; Cook et al., 1993). Alternatively, RaplA may use a distinct subset of Ras-GTP-binding proteins to regulate distinct signaling pathways (Yoshida et al., 1992). Although R-Ras exhibits transforming properties in NIH 3T3 cells (Cox et al., 1994; Saez et al., 1994), R-Ras may regulate cellular processes (apoptosis, integrin-mediated cellular adhesion) distinct from Ras (Zhang et al., 1996). Consistent with this, constitutively activated Ras or TC21, but not R-Ras, was able to reverse the growth inhibition due to the loss of Ras function in fibroblasts (Huff et al., 1997). Finally, another Ras-related protein, designated Rheb (Clark et al., 1997a), has also been shown to interact with Raf (Clark et al., 1997a). Similar to Rap1A, Rheb exhibits no transforming potential, but instead antagonizes Ras signaling and transformation, possibly by formation of a nonproductive complex with Raf and other Ras effectors.

78

Roya Khosravi-Far ef al.

V. Ras ACTIVATION OF A GTPase CASCADE: INVOLVEMENT OF Rho FAMILY PROTEINS IN TRANSFORMATION Although it is clear that Ras must use effectors other than Raf, the precise nature of these Raf-independent pathways remains to be delineated fully. Several lines of evidence, however, have implicated the involvement of the Rho family of proteins as one of the downstream mediators of Ras function. One of the first suggestions of a link between Ras and Rho proteins was provided by the observation that a p120 GAP-associated protein, p190, which is a GAP primarily for RhoA, but is also active on CDC42 and Rac (Settleman et al., 1992). Thus, p120 GAP may serve as an effector that facilitates Ras regulation of Rho protein function. In this section, we summarize additional observations that support a connection between Ras and Rho family proteins and provide a brief overview of the key features of this branch of the Ras superfamily of small GTPases.

A. A Requirement for Rho Family Proteins

in Ras Function and Transformation Genetic analysis i n S . pombe Ras ( R a s l ) identified two distinct Rasl effector-mediated activities (see Fig. 4). One of these involves Rasl interaction with Byr2, which is a MEK kinase homolog (Wang et al., 1991; Van Aelst et al., 1993; Masuda et al., 1995). The other involves Rasl interaction with Scdl (Chang et al., 1994). Scdl is a putative Rho exchange factor and may in turn regulate the function of the Rho family protein CDC42sp. Interestingly, Scdl binds directly to activated Ras and possesses properties of a Ras effector, thus directly linking Ras activation to activation of Rho family proteins. Whether a specific mammalian Rho-GEF is analogous to Scdl and can function as a Ras effector remains to be determined. Finally, the scheme shown in Figure 4 is an oversimplification because Scdl is likely to promote a multiprotein complex formation, together with the SH3 domain-containing Scd2 protein, with GTP-complexed Rasl (Chang et al., 1994). Further evidence for the involvement of Rho family members in mediating oncogenic Ras function came from microinjection studies in Swiss 3T3 cells by Ridley, Hall, and colleagues (1992a; 1992b) (Fig. 7). These studies demonstrated that oncogenic Ras activates Racl, which in turn activates RhoA, as measured by changes in actin cytoskeletal organization. Activated Rac stimulates the formation of lamellipodia to cause membrane ruffling and to stimulate RhoA function. Activated RhoA in turn controls the formation of actin stress fibers and focal adhesions. Thus, in addition to the activation of kinase cascades, oncogenic Ras also activates a cascade of small GTPases.

increasing Complexity of Ras Signal Transduction: involvement of Rho Family Proteins

79

Fig. 7 Rho family proteins are regulators of signaling pathways that regulate the organization of the actin cytoskeleton. CDC42, Racl, and RhoA function as downstream components of signaling pathways initiated by ligand-stimulated G-protein-coupled serpentine receptors (SRs) or RTKs. Two GTPase cascades have been identificd. One involves oncogenic Ras activation of Racl, then Racl activation of RhoA. A second involves CDC42 activation of Rncl and RhoA. CDC42, Racl, and RhoA each modulate distinct changes in actin organization.

Consequently, the morphologic transformation seen in Ras-transformed cells may be caused, in part, by deregulation of Rho family protein function. Several studies have demonstrated the requirement for Rho family proteins in oncogenic Ras-transforming activity. Coexpression of dominant negative mutants of Racl, RhoA, RhoB, o r CDC42 reduced oncogenic Rastransforming activity when analyzed in rodent fibroblasts (Qiu et al., 1995a; Khosravi-Far et al., 1995; Prendergast et al., 1995; Qiu et al., 1995b; Qiu et al., personal communication). On the other hand, coexpression of constitutively activated Racl and RhoA with activated Rafl showed cooperative transforming activity. Similarly, coexpression of constitutively activated Racl and RhoA with oncogenic Ras also caused a greatly enhanced morphologic transformation. Finally, the constitutively activated mutants Racl, RhoA, RhoB, and CDC42 each caused growth transformation of NIH 3T3 or Rat1 rodent fibroblasts (Qiu et al., 1995a; KhosraviFar et al., 1995; Prendergast et al., 1995; Qiu et a/., 1995b). Taken together, these observations suggest that the function of Rho family proteins is essential for full Ras-transforming activity. However, a demonstration that Ras-transformed cells exhibit constitutively elevated GTP-bound levels of a specific Rho family protein has not been described. Thus, which, if any, of the known Rho family proteins are clearly targeted by oncogenic Ras remains to be established. An involvement of Rho family proteins in Ras function prompts at least three key questions. First, how does Ras connect with and regulate specific

80

Roya Khosravi-Far e l al.

Rho family protein function? Second, what aspect of Rho protein function contributes to Ras transformation? Third, what are the key effectors of Rho protein function, and what signaling pathways mediate their actions? We are clearly in the early days of approaching the answers to these questions. The following sections first provide an overview of Rho family proteins and their functions and then summarize our current level of understanding of these questions.

B. Rho Family Proteins Are Regulators

of Diverse Cellular Processes Rho family proteins are members of a major branch of the Ras superfamily of small GTPases and share about 30% of their amino acid identity with Ras (Fig. 8 ) (Madaule and Axel, 1985). Like Ras, Rho family proteins function as GTP/GDP-regulated switches that cycle between active GTP- and an inactive GDP-bound forms (Fig. 9) (Ridley, 1996).This cycle is regulated by three distinct classes of regulatory proteins. First, Rho GEFs (also referred to as Dbl homology proteins) serve as activators and stimulate the replacement of GDP by GTP (Whitehead et al., 1997). More than 20 Dbl homology proteins have been identified, and most were initially discovered as transforming proteins in NIH 3T3 focus-forming assays (e.g., Dbl, Vav). Dbl family proteins are presumed to cause transformation by causing constitutive upregulation of Rho protein function. Second, at least 10 Rho-GAPS have also been identified that serve as negative regulators of Rho family protein function and stimulate their intrinsic GTPase activities (e.g., p l 9 0 Rac/Rho-GAP) (Cerione and Zheng, 1996). BCR and ABR are multifunctional regulators and serve as both Rho-GEFs and Rho-GAPS (Chuang et al., 1995). In addition, some Rho-GAPS may also serve as downstream effectors (e.g., N-chimaerin) (Kozma et al., 1996). Finally, Rho-GDP-dissociation inhibitory factor (Rho-GDI) was originally identified to inhibit GDP dissociation of Rac (Ueda et al., 1990), but subsequently was shown to interfere with both intrinsic and GAP-stimulated GTP hydrolysis of Rac and CDC42 (Chuang et al., 1993; Hart et al., 1992). At least 11 mammalian Rho family proteins have been identified: RhoA, RhoB, RhoC, RhoD, RhoE, RhoG, Racl, Rac2, CDC42, TC10, and TTF) that share significant (ranging from 50 to 9070)amino acid identity with each other (Ridley, 1996). Two mammalian isoforms of CDC42 have been identified from brain or placenta. In light of the fact that the substrates for many of the Dbl family proteins have not been identified, it is anticipated that more members of the Rho family will be identified. Much of our knowledge of Rho family protein function has been derived primarily from the studies of Racl, RhoA, and CDC42. One report demonstrated that RhoD exhibits a function distinct from these proteins (Murphy et al., 1996). Sequence comparisons

increasing Complexity of Ras Signal Transduction: Involvement of Rho Family Proteins

8I

0.043 H

-

RhoB

-

-

RhoC RhoA

-El0

-

- CDC42(P) - CDC42(B) RhoG

-

Rac2

- Racl Fig. 8 The Rho family of small GTPases constitutes a major branch of the Ras superfarnily. Primary sequences of Rho family proteins were aligned to the H-Ras primary sequence using the dynamic algorithm alignment program CLUSTALW (Thompson et al., 1994) and then were used to construct the Rho dendrogram. Using CLUSTALW, a distance matrix was prepared from the multiple sequence alignments. The branch lengths in the dendrogram are proportional to the estimated divergence along each branch.

suggest that RhoE and TTF may also share functional relationships, whereas TClO and RhoC are anticipated to be most related in function to CDC42 and Racl/2, respectively (see Fig. 8). Thus, sequence and functional differences allow the definition of a t least five distinct Rho subfamilies: (1)CDC42 and TClO; (2) Racl, RaQ, and RhoC; ( 3 )RhoA, RhoB, and RhoC; (4) RhoE

a2

Roya Khosravi-Far el al.

STP

GDP

-

f

EXTRACELLULARSTIMULI

Actin Cytoskeletal Organization Gene Expression Cell Cycle Progression

-

Rho GEFs: Dbl Homology Proteins Dbl, Vav, Ect2, Tiam-1, BCR, etc. Rho GAPS - BCR, p190, N-chimaerin, etc. Fig. 9 Rho family proteins tunction as GDI’/GTP-regulated switches. A variety of extracellular stimuli regulate Rho family protein GDPlGTP cycling through modulation of Rho GEF, GAP, o r GI11 function. Activated Rho family proteins exhibit diverse cellular functions that include regulation of actin organization, gene expression, and cell cycle progression.

and TTF; and (5) RhoD. Analyses of TC10, RhoG, RhoD, and RhoE function may result in further subdivision of these Rho family proteins. The first clues to Rho protein function were provided by studies using C3 toxin from Clostridium botulinum, an inhibitor of RhoA, B, and C function (Chardin et al., 1989) and by microinjection analyses using mutant Rho family proteins (Ridley et al., 1992b; Ridley and Hall, 1992a; Nobes and Hall, 1995). These studies demonstrated that Rho family proteins induce unique morphologic changes that involve rearrangements of F-actin, which in turn influence cell shape, cell motility, and cell-cell interactions. In addition to the induction of lamellipodia and membrane ruffling (Racl) (Ridley et al., 1992b) or the formation of stress fibers and focal adhesions (RhoA) (Ridley and Hall, 1992a),constitutively activated CDC42 caused induction of filopodia, which are finger-like cytoplasmic extensions that may be involved in the recognition of the extracellular environment (Nobes and Hall, 1995). Furthermore, it has been suggested that activated Rho stimulates contractility, driving the formation of stress fibers, focal adhesions and elevation of tyrosine phosphorylation (Chrzanowska-Wodnicka and Burridge, 1996; Burridge and Chrzanowska-Wodnicka, 1996). Transient expression of activated RhoD in a variety of cell types caused rearrangements of the actin cytoskeleton and cell surface and regulated endosome motility and distribution (Murphy et al., 1996).

IncreasingComplexity of Ras Signal Transduction: Involvement of Rho Family Proteins

83

As described previously, oncogenic Ras causes activation of a GTPase cascade that includes Rac and Rho. Similarly, in a distinct pathway, constitutively activated CDC42 was found to activate Rac and, consequently, RhoA (Nobes and Hall, 1995). GTPase cascades that involve Rho family proteins have also been described in yeast (reviewed in Chant and Stowers, 1995). The precise links between these GTPase cascades have not been established. Based on the involvement of Dbl family-related proteins (CDC24) in the activation of CDC42 in yeast pathways (Zheng et al., 1994b), however, it is logical that Dbl family proteins serve as intermediates between GTPases. Like Ras, Rho family proteins also serve as CDP/GTP-regulated relay switches that transmit extracellular ligand-mediated signals that promote changes in actin structures (see Fig. 7). It has been observed previously that lysophosphatidic acid (LPA), bombesin, and bradykinin each bind to seventransmembrane, C-protein-coupled cell surface receptors to activate the formation of filopodia, membrane ruffles, and stress fibers. Rho family proteins were subsequently implicated in these processes (Ridley et al., 1992b; Ridley and Hall, 1992a).For example, LPA stimulation of actin stress fibers formation and the assembly of focal adhesion complexes in Swiss 3T3 cells were blocked by C3 toxin (Ridley and Hall, 1992a). PDGF stimulation of its receptor, a transmembrane tyrosine kinase, causes membrane ruffling, which was blocked by dominant negative Rac (Ridley etal., 1992b).Dominant negative CDC42 blocked the formation of filopodia induced by bradykinin (Nobes and Hall, 1995). Finally, insulin-induced and hepatocyte growth factor-induced membrane ruffling in KB human epidermoid carcinoma cells were dependent on Rac and Rho, respectively (Nishiyama et al., 1994). Thus, like Ras proteins, Rho family proteins serve as key intermediates in a multitude of signaling pathways. Rho family proteins are also associated with processes that involve the actin cytoskeleton in other organisms. The regulation of cytoskeletal organization by Rho family members is evident from genetics studies in yeast and Dvosophila. Yeast CDC42 has been demonstrated to coordinate polarization of the actin cytoskeleton during cell division by budding and involves a cascade of other small GTPases (BUDl/RSRl and R H O proteins) (reviewed in Chant and Stowers, 1995). In Drosophila, Dracl is required to assemble actin at adherens junctions of the wing disk epithelium, whereas Dcdc42 is involved in the regulation of polarized cell shape during various stages of wing disk development (Eaton et al., 1995). Studies have also implicated a requirement for Rho family proteins during cytokinesis in sand dollars (Mabuchi et al., 1993), Xenopus sp. (Kishi et al., 1993; Drechsel et a/., 1996), and Dictyostelium. (Larochelle et al., 1996). In addition to their involvement in regulation of cytoskeletal organization, Rho proteins are also regulators of gene expression (Fig. 10). First, constitutively activated Racl and CDC42 are activators of the JNKs (also

Roya Khosravi-Far el a/.

Fig. 10 Rho family proteins are regulators of gene expression. Rho family proteins regulate the activity of a variety of nuclear transcription factors. RhoA, Racl, and CDC42 activate SRF, which forms a complex with ternary complex factors (Elkl and SAPI) at the serum response DNA element found in the promoter sequences of c-fos and other genes. RacllCDC42 (and RhoA in some cells) activate JNK, which in turn can activate the Jun, ATF2, and Elk1 nuclear transcription factors. Activated c-Jun-ATF2 heterodimers stimulate AP 1-like DNA elements in various promoters that regulate c-jun and other genes. Racl activates NFKB by as yet unknown pathways.

SAPKs), but not the p42 and p44 ERKs (Coso et al., 1995; Minden et al., 1995; Olson et al., 1995). JNK in turn activates the ATF2 and Jun nuclear transcription factors. ATF2 and Jun can dimerize with other transcription factors to stimulate transcription from promoters containing APl and related DNA sequences (e.g., the c-jun promoter) (Karin, 1995). RhoA can also activate JNK in some cells (Teramoto et al., 1996b). Second, RhoA, Racl, and CDC42 have been shown to activate the serum response factor (SRF) (Whitmarsh et al., 1995). SRF cooperates with ternary complex factors (TCFs; Elkl and SAP1) and the serum response DNA elements found in certain promoters, such as the c-fos promoter (Marais et al., 1993). TCFs are activated by the Raf+MEK+MAPK pathway (see Fig. 1 ) . Interestingly, Rac and CDC42 activation of SRF is not dependent on Rho, indicating that Rho family proteins use distinct pathways to activate SRF (Hill et al., 1995). Thus, the CDC42+Rac-+Rho cascade that regulates actin structure is clearly distinct from pathways that regulate SRF. Finally, activated Racl, but not CDC42, activates NFKB (Sulciner et al., 1996). NFKBelements are found in a wide variety of promoters (Baeuerle and Baltimore, 1996). Ras activation of JNK and NFKBhas been shown to be mediated, in part, by activation of Rac (Minden et al., 1995; Sulciner et al., 1996). A third function of Rho family proteins (RhoA, Racl, and CDC42) is their requirement for cell cycle progression through the G I phase of the cell cycle (see Fig. 9) (Olson et al., 1995). For example, C3 toxin treatment and inhibition of Rho function caused Swiss 3T3 cells to accumulate in the G I phase of the cell cycle (Yamamoto et al., 1993). Furthermore, microinjection of

Increasing Complexity of Ras Signal Transduction: Involvement of Rho Family Proteins

85

constitutively activated mutants of Ras, RhoA, Racl, and CDC42 caused G , progression and stimulation of DNA synthesis in quiescent Swiss 3T3 cells (Olson et al., 1995). The mechanism by which Ras and Rho proteins regulate cell cycle progression is not yet understood but is thought to involve stimulation of D-type cyclins by Ras and AP1 proteins (Albanese et al., 1995). We showed that Racl and RhoA stimulated transcription of cyclin D1 (Westwick et al., 1997). Thus, it is possible that Rho family proteins are mediators of Ras-induced cell cycle progression. Finally, in agreement with their involvement in mediation of Ras-induced cellular transformation, expression of constitutively activated mutants of Racl, RhoA, or RhoB alone has been shown to cause tumorigenic transformation of NIH 3T3 and Ratl rodent fibroblasts (Qiu et al., 1995a; KhosraviFar et al., Qiu et al., 1995b; Prendergast et al., 1995). Constitutively activated CDC42 is growth inhibitory in NIH 3T3 cells (Khosravi-Far and Der, unpublished observation) but can cause tumorigenic transformation of Ratl cells (Qiu et al., personal communication). These observations indicate that members of the Rho family of proteins are also involved in regulation of cellular proliferation. Thus, positive regulation of cell growth by Rho family proteins is likely to contribute to the transforming actions of Ras and to the transforming activity of the Dbl family of oncogene proteins (Whitehead et al., 1997; Khosravi-Far et al., 1994a) as well as to some heterotrimeric G proteins (G,, and G, 3 ) and G-protein-coupled seven-transmembrane (serpentine) receptors (Mas and XGR) (Martin, Zohn, and Der, unpublished observations). Taken together, these observations point to the Rho family proteins as critical components of a Raf-independent signaling pathway important for Ras transformation. O u r knowledge of the details of this GTPase cascade remains limited. How Ras proteins connect to Rho proteins and which Rho family members are involved are still poorly understood. In addition, Rho proteins regulate diverse cellular processing and signaling events. The emerging picture is that Rho family proteins, like Ras, mediate their actions through interaction with multiple effectors. Therefore, which Rho effector function is important in promoting Ras transformation is not known. In the following two sections, we summarize our knowledge of how Ras regulates Rho family proteins and describe possible effectors that mediate Rho family protein function.

VI. Rho FAMILY PROTEINS MEDIATE THEIR ACTIONS THROUGH INTERACTION WITH MULTIPLE EFFECTORS Although Racl, RhoA, RhoB, and CDC42 function have been shown to be necessary for Ras transformation, the precise contribution of these small GTPase to Ras transformation remains to be elucidated. It was originally

86

Roya Khosravi-Far ef al.

postulated that Rho family protein-induced actin cytoskeletal alterations contributed to the morphologic transformation caused by Ras. Aside from actin regulation, however, it is now known that Rho proteins control gene expression, cell cycle progression, and cellular proliferation. Therefore, the contribution of Rac, Rho, and CDC42 to Ras transformation is not likely simply due to changes in actin organization. Indeed, whereas Ras-transformed cells are devoid of well-developed stress fibers, Rac- and Rho-transformed cells retain significant stress fibers and focal adhesions (Khosravi-Far et a/., 1995). Thus, defining the contribution of Rho family proteins to Ras transformation is not anticipated to be a straightforward endeavor. The next steps in this process will be to define the downstream effectors of Rac and Rho, to determine the functions that they regulate, and finally, to establish the contribution of each Rac/Rho function to Ras function. Like Ras, Rho family proteins mediate their actions through interaction with multiple effector targets. Many GTP-dependent Racl-, RhoA-, and CDC42-binding proteins have already been identified, and it is likely that many remain to be discovered. Our understanding of their contributions to Rho family protein function is limited. In the following sub-sections, we provide a brief overview of their structures and possible functions. We also summarize observations from recent studies using effector domain mutants o f Racl that begin to define the precise function of specific effectors. Finally, we conclude with a discussion of how a spectrum of structurally and functionally diverse proteins all share the common property of serving as effectors of a particular Rho family protein. We have categorized the Rho-binding proteins into two general groups based on their binding specificities: (1) those that bind Rac and/or CDC42, but not Rho, and ( 2 ) those that bind Rho. Much of these analyses has been limited to the study of interaction with Racl, RhoA, or CDC42. Therefore, possible effector interactions with other Rho family members may complicate these simple classifications. Furthermore, many of the binding analyses have been performed in vitro or with yeast two-hybrid binding analyses. Whether they represent physiologically relevant interactions in vivo remains to be determined for many of these proteins. Answers to several questions will aid in this determination. First, the ability of a particular small GTPase to bind to an isolated domain in vitro may be misleading. Therefore, interaction with the full-length protein needs to be established in vivo. Second, whether binding leads to the activation of effector function will be important to establish.

A. A Plethora of Candidate Effectors of Rac and CDC42 Unlike the situation with candidate effectors of Ras, clear consensus-binding sequences for Rho family-binding proteins have been identified for at

Increasing Complexity of Ras Signal Transduction: Involvement of Rho Family Proteins

87

least some of these proteins. Burbelo and colleagues (1995) have described a minimal region of 16 amino acids required for binding of CDC42 and/or Rac, designated the CDC42/Rac/Interactive/Binding (CRIB) motif. Using this motif in a search of the GenBank data base, these authors identified more than 25 proteins from a wide variety of eukaryotic species that contain the CRIB motif. This motif is found in known CDC42/Rac-binding proteins, such as the three p21 (CDC42/Rac)-activated kinase (PAKs) isoforms and the ACK tyrosine kinase. Other CRIB-containing proteins include the Wiscott-Aldrich Syndrome protein (WASP),several mixed-lineage kinase (MLK) 3 homologs, a human protein MSE55, and a C. elegans protein DRP2. Further analyses demonstrated that all four interacted with GTP-bound forms of CDC42 and/or Racl. Therefore, CRIB may define the binding domain of a t least a large group of Rac/CDC42 effector molecules. Generally, Rac and CDC42 share a number of binding partners that do not bind RhoA, which may be related to the overall sequence homology between Rac and CDC42 compared with Rho (see Fig. 8). Some Rac-specific (e.g., p67P""" and POR1) and some Rac/CDC42-binding proteins have been identified that lack the CRIB motif (e.g., pp70 S6 kinase, the p85 subunit of PI3K, IQGAPs). Clearly, Rac-binding motifs other than the CRIB motif will be found. Among the Rac-binding proteins that have been identified, the three mammalian PAK serinekhreonine kinase isoforms have attracted the most interest (Manser et a/., 1994; Manser et al., 1995; Teo et al., 1995; Bagrodia e t a l . , 1995; Martin et al., 1995). By virtue of its striking homology with the S. cerevisiae protein Ste20, which is implicated in G-protein-associated pheromone signaling to a MAP kinase cascade, PAK has been considered a likely candidate for a link from Rac and CDC42 to mammalian MAPKs (Simon et al., 1995; Zhao e t al., 1995). Support for this possibility was provided by experiments using Xenoptls sp. oocyte extracts, in which Ste20, a related protein from S. pombe (Shkl), and PAK were shown to activate JNWSAPK (Polverino et al., 1995). Additionally, the intrinsic kinase activit y of PAK was activated by RacKDC42 binding (Manser et al., 1994), and overexpression or constitutive activation of PAK showed enhanced activation of JNK or p38 (Zhang et al., 1995; Frost et al., 1996). In contrast, dominant negative mutants of PAK inhibited Racl activation of JNK/p38 (Zhang et al., 1995). Other evidence, however, argue-that PAK is not an effector for Rac/CDC42 activation of JNK. First, identification of Rac effector domain mutants that n o longer bound to nor activated PAK yet retained the ability to activate JNK and p38 demonstrated that PAK was not required for Rac activation of JNK (Westwick et al., 1997). Second, MLK3 was shown to link Rac/CDC42 to JNK activation (Teramoto et al., 1996a). MLK3 (also known as PTK1) is a member of a family of related kinases (MLK1 and MLK2) of unknown function. MLK3 RNA is expressed in most tissues and cell lines (Ezoe e t al., 1994;

88

Roya Khosravi-Far el

al.

Ing et al., 1994). These investigators found that MLK3 associated with activated Racl and CDC42, but not with RhoA, in vivo, and that coexpression of MLK3 (and not PAK) caused a synergistic enhancement of Racl and CDC42 activation of JNK (Teramoto et al., 1996a). Overexpression of MLK3 alone caused activation of JNK, but not of p38 or ERK. Thus, MLK3, rather than PAK, may serve as the Rac/CDC42 effector that leads to activation of JNK. Rac effector domain mutant studies also eliminated PAK as the effector for directing Rac activation of SRF or induction of transformation (Westwick et al,. 1997) and for induction of lamellipodia (Lamarche et al., 1996; Joneson et al., 1996a; Westwick et al., 1997). Instead, a correlation was found that implicated PAK as a possible effector for mediating Rac stimulation of transcription from the cyclin D , promoter (Westwick et al., 1997). PAK was found to be dispensable, however, for Rac-mediated progression through G I phase (Lamarche et al., 1996). Thus, Rac upregulation of cyclin D, alone may not be sufficient for its regulation of cell cycle progression. Furthermore, the Racinduced pathways leading to the regulation of lamellipodia and activation of JNK, SRF, and cyclin D , appear to be mediated by distinct effectors (Westwick et al., 1997). Therefore, which, if any, of the known Rac functions are important for Rac transformation and thus Ras transformation is unresolved. The 70-kd S6 kinase (p7OSbK)also appears to be an important effector of Rac and CDC42 function. It is activated by diverse mitogenic stimuli, including growth factors, cytokines, and activated oncogenes (Chou et al., 1995). Evidence supports an important role for p70ShKin the progression of cells from G , to S phase of the cell cycle. Chou and Blenis showed that catalytically inactive p70SbK interacted with GTP-complexed Racl and CDC42, but not with RhoA, in vitro (Chou and Blenis, 1996).Constitutively activated mutants of Racl and CDC42, but not of RhoA, stimulated ~ 7 0 " ~ activity in vivo. Effector domain mutations abolished this activity. Dominant negative CDC42( 17N) or Racl( 17N) blocked EGF- and PDGF-induced ac. to the Ras-Raf interaction, CDC42 associativation of ~ 7 0 " ~Analogous tion with p7OShKin vitro alone did not activate its kinase activity. Thus, CDC42/Rac interaction with p70ShKmay facilitate subsequent events that lead to full kinase activation. It is not known what aspect of Rac function is mediated by p70SbK. Racl and CDC42 have been shown to complex the p85 subunit of PI3K in vitro and to coprecipitate with PI3K activity in vitro and in vivo (Zheng et al., 1994a; Tolias et al., 1995). This interaction was found to be GTP-dependent, and a CDC42(35A) effector domain mutant showed impaired p85 binding (Zheng et al., 1994a). CDC42 stimulated immunoprecipitated PI3K activity two- to four-fold. Thus, although the sequence homology of p85 with other Rho GAPS (e.g., BCR, N-chimerin, and p190 Rho GAP) suggested that it acted as a negative regulator of Rho function, no GAP function has

Increasing Complexity of Ras Signal Transduction: Involvement of Rho Family Proteins

89

been demonstrated for p8.5. Instead, this GAP-related domain may serve to promote PI3K association with Rac/CDC42 as an effector in regulating reorganization of the actin cytoskeleton. Interestingly, another Rho GAP (Nchimerin) has also been shown to exhibit properties of a RadCDC42 effector that mediates actin cytoskeletal events (Kozma et al., 1996). Two Rac-binding proteins have been implicated in mediating Rac regulation of the actin cytoskeletal organization. In one study, Van Aelst and colleagues (1996), using yeast two-hybrid screening, identified a partner of Racl (POR1) protein that interacted with GTP-complexed Racl, but not with CDC42, RhoA, or H-Ras. The 34-kd PORl protein showed no significant sequence identity with any known proteins and no conserved motifs besides the presence of a putative leucine zipper. The Racl(35A)effector domain mutant that is impaired in membrane ruffling also showed impaired binding to PORl. Truncated PORl inhibited Racl ( 12V)-induced membrane ruffling, whereas full-length PORl acted to synergistically enhance activated HRas( 12V) membrane ruffling activity. No synergistic interaction, however, was seen between Racl and PORl. In another study, yeast two-hybrid library screening with activated RhoA(63L) isolated a previously identified RhoAbinding protein, p160RoCK(discussed later) (Ishizaki et al., 1996) that also interacts with activated forms of Racl and CDC42, but not with R-Ras (Lamarche et al., 1996). The Racl(37A) effector domain mutant no longer interacted with ROCK and failed to induce lamellipodia or G, phase progression, suggesting that ROCK may mediate these two activities of Racl. The WASP protein contains the CRIB motif and was identified as a CDC42binding protein that bound weakly to R0c1, but not to RhoA (Symons et al., 1996; Aspenstrom et al., 1996; Kolluri et al., 1996). The WASP gene is mutated in WAS patients. These patients with Wiskott-Aldrich Syndrome possess an X-linked recessive disorder characterized by thrombocytopenia, which is characterized by recurrent infections due to T- and B-cell function and eczema. Abnormalities seen in T and B cells suggested defects in the organization of the actin cytoskeleton (Kirchhausen and Rosen, 1996). WASP overexpression caused the formation of WASP clusters that were highly enriched in polymerized actin in porcine aortic endothelial cells, and clustering was inhibited by dominant negative (CDC42(17N), but not by R a c l ( 1 7 N ) (Symons et al., 1996). Additionally, coexpression of WASP inhibited CDC42( 12V) and Racl( 12V)-induced changes in actin cytoskeleton organization. Furthermore, a region of homology between WASP and other proteins (VASP and Mena) involved in the organization of actin cytoskeleton and control of microfilament dynamics has been identified (Gertler et al., 1996). Interestingly, this domain in VASP and Mena has been shown to interact directly with components of focal adhesions. These results point to WASP as an effector that mediates the activity of CDC42/Rac in F-actin polymerization (Symons et al., 1996), and reduced WASP expression may

90

Roya Khosravi-Far e1 a/.

contribute to the cytoskeletal abnormalities seen in WAS-affected males (Kolluri et al., 1996). WASP is also recognized by the Nck adaptor protein through its Src homology 3 domains (Rivero-Lezcano et al., 1995; Quilliam et al., 1996), and Nck also complexes with other Rho family-binding proteins (PAKI, PAK3, and PRK2) (Bagrodia et al., 1995; Quilliam et al., 1996; Bokoch et al., 1996; Galisteo et al., 1996). Nck associates with activated receptor tyrosine kinases (Li et al., 1992; Lee et al., 1993). Therefore, it may serve as an adaptor that links extracellular stimuli to Rho family proteins. IQGAP1 and the closely related IQGAP2 (62% identity) have been shown to bind Racl and CDC42, but not RhoA (Brill et al., 1996; Kuroda et a[., 1996). No CRIB motif is present in these proteins. IQGAP was identified originally as a human protein that harbored potential calmodulin-binding IQ motifs upstream of a sequence related to the catalytic domain of Ras GAPS (Weissbach et al., 1994). It was isolated independently by affinity chromatography analyses for GTP-dependent CDC42-binding proteins (Kuroda et al., 1996). IQCAPl RNA is highly expressed in lung, kidney, and placenta, whereas IQGAP2 is expressed predominantly in the liver (Brill et al., 1996).Although both proteins show strong homology with the Ras-GAP of S. pombe, no stiniulation of Ras, Racl, or CDC42 GTPase activity has been detected (Brill et al., 1996; Kuroda et al., 1996). Because IQGAPs harbor a potential actin-binding domain and accumulate at insulin- or Racl -induced membrane ruffling areas, it has been postulated that IQGAPs serve as effectors that mediate the actin cytoskeletal events of CDC42 and Racl (Kuroda et al., 1996). Otherwise, it is not known whether IQGAPs are physiologically relevant effectors of R a c K D C 4 2 , and if so, what effector function they might facilitate. A determination of whether IQGAP binding correlates with the ability of Racl effector domain mutants to induce lamellipodia will help to assess this possibility. Other candidate effectors of Rac include ROKa, a previously described RhoA-binding protein (Leung et al., 1996b); POR2, a protein of unknown function (Van Aelst et al., 1996); and tubulin (Best et al., 1996). Citron is a protein of unknown function that was identified as a RhoC-binding protein in a yeast two-hybrid library screen; it was found to bind to Racl, but not to CDC42, in vitro (Madaule et al., 1995). In summary, multiple candidate effectors o f Rac/CDC42 function exist, and more are likely to be identified. The existence of Rac/CDC42-binding proteins that lack the CRIB motif indicates that at least two, if not more, Racand/or CDC42-binding motifs will be identified. It is not clear, which, if any, of these Rac-binding proteins are physiologically relevant effectors. Furthermore, which effectors mediate Racl regulation of actin organization, gene expression, and regulation o f cell proliferation and invasion remain to be characterized. Our evaluation of Rac effector domain mutants failed to implicate any known Racl function with transforming potential (Westwick et al., 1997). One interpretation is that multiple pathways promote Rac growth regulation and that no one pathway alone is necessary. Alternatively, some as yet

IncreasingComplexity of Ras Signal Transduction: Involvement of Rho Family Proteins

91

undiscovered Rac function may be responsible for Rac-mediated transformation. Regardless of which is the correct scenario, the Rac functions required to promote oncogenic Ras transformation remain to be defined.

B. A Plethora of Candidate Effectors of Rho A growing number of Rho-binding proteins have also been identified. Two distinct Rho-binding sequence motifs have been described for a subset of Rho-binding proteins (Reid et al., 1996; Fujisawa et al., 1996). Both Rho effector binding motifs are distinct from the CRIB motif and are associated with proteins that contain predicted coiled-coil domains (Cohen and Parry, 1994). These coiled-coil domains may provide the basis for the multimerization of these proteins with Rho and other signaling molecules in a complex with Rho. Two closely related serinehhreonine kinases that show strong sequence similarity to the kinase domain of PKCs (PKN/PRKl and PRK2) and two novel proteins lacking kinase domains (Rhotekin and Rhophilin) appear to share a distinct Rho effector binding motif (designated class I, REM1) (Fig. 11A). PRKl and PRK2 were originally identified by a PCR-mediated approach to identify sequences that encoded novel PKC-related human proteins (Palmer et al., 1995). The encoded 120 to 130-kd proteins identified by this approach contain an NH,-terminal basic sequence encompassing GAXN (a putative pseudosubstrate domain); Rho-binding sequences, followed by amphipathic helix and leucine zipper domains; and a COOH-terminal kinase domain (see Fig. 11A). The 71-kd Rhophillin protein also contains an NH,-terminal class I Rho effector binding motif; however, it lacks a kinase domain or any other known catalytic sequence. Aside from the NH,-terminal REMl motif, two proline-rich motifs and putative SH3-binding domains are present in the COOH-terminus. Thus, Rho-binding proteins with class I binding motifs also typically contain other putative protein-protein or protein-lipid binding motifs, suggesting that these effectors may serve to promote targeting and/or multiprotein complex formation with Rho proteins. PRKl was identified independently by affinity chromatography analysis for RhoA-binding proteins and designated protein kinase N (PKN) (Watanabe et al., 1996). Interestingly, LPA resulted in Rho-dependent activation of PKNPRKl, supporting a role for this kinase in mediating Rho function. The observation that staurosporine, a potent inhibitor of PKC (and possibly of PKNPRK1 ), blocks formation of focal adhesions suggests that PKN may mediate this action of Rho proteins. Additionally, PKN has been shown to interact with and phosphorylate the head-rod domain of neurofilament proteins, suggesting its role in the regulation of neurofilament protein assembly (Mosch et al., 1996). Hence, PKN may be a potential mediator of the growth factor-induced stimulation of intermediate filaments. PRK2 was isolated in-

PKN PR K2

PKN 3 2 1 0 1 PRKZ 12-11:

Rhophilln

11 I U i

\ \

\

A Rhophilin

B

Citron

N

I

-_ - - _ -_ _ _ _ - -_ - - -

N

C

cc

Fig. I 1 Consensus binding sequences define different classes of Rho family candidate effectors. The domain structure of several known RhoA-binding proteins can be grouped into two distinct classes that are similar in their minimal Rho-bindingsequences. (A)The class I Rho effector-binding motif (REM) is shared by the PKN/PRKl and PRKZ serinehhreonine kinases and by a novel protein that lacks kinase domains (Rhophilin). Shown f o r I'KNIPRK2 are ( 1 ) three predicted homologous coiled-coil (CC) domains referred to as homologous regions la, lb, and Ic ( H R l a , H R l b, HRlc), ( 2 ) a V,,/HK2 domain that shows homology between PKN and PRK2 and that is highly similar to sequences surrounding the NH,-terminal V0 and pseudosubstrate sites of I'CKE and PKCq, and ( 3 )the COOH-terminal kinase domain. The Rho-binding sequence (shaded region) is located in the NH,-terminal C C domain of PKNPRKZ and Rhophilin. The multiple sequence alignment of class I REM shows a highly conserved sequence followed by a heptad repeat, indicated by a solid square (D)above the sequences. The first position o f the heptad repeat is not predicted to be present in Rhophilin. Strictly conserved residues are indicated with an asterisk ("), and positions of high amino acid similarity are indicated by a solid circle ( 0 )below the sequences. (B) Proteinscontaining the class I1 REM include Rho-kinase, p160R"':K, and ROKalp. These related kinases possess an NH,-terminal kinase domain, followed by a predicted large central C C domain and a pleckstrin homology (PH) domain that encompasses a cysteine-rich domain (CRD). The Rho-binding sequence (shaded box) is located a t the COOH-terminus of the central C C domain. The multiple sequence alignment of class I1 REMs contain a region of strict sequence conservation and are indicated by an asterisk ("). Citron shows a similar domain structure to Rho-kinase, p160"0"K, and ROKa/P, but it lacks a kinase domain. The NH,-terminal half of the protein contains a large (about 87.5 residues) predicted C C domain. Like the other class I1 proteins, the C C domain is followed by CRD and PH domains. Citron also contains a prolinerich site (PR) that is a putative SH3 domain-binding motif. The putative Rho-binding sequence (shaded box) is found in the extreme COOH-terminus of the CC domain. Apart from similarities in domain structure, no strong sequence homology was observed between the Citron and the class I or I1 REMs.

Increasing Complexity of Ras Signal Transduction: Involvement of Rho Family Proteins

93

dependently during an expression library screen for proteins that bound to the SH3 domains of the Nck SH2/SH3 adaptor protein (Quilliam et al., 1996). It was also shown to bind RhoA, but not Racl or CDC42, and coexpression of PRK2 with activated RhoA caused synergistic enhancement of RhoA-dependent activation of SRF expression. Rhophilin, and to a lesser degree Rhotekin, share NH,-terminal sequence homology with the PKN/PRKl and PRK2 class I Rho-binding motif, Rhotekin was identified by yeast two-hybrid screening for RhoC-binding proteins (Reid et al., 1996). It also interacted with GTP-complexed RhoA and RhoB, but not with Racl or CDC42. Rhophilin was isolated by yeast two-hybrid screening of a mouse embryo cDNA library for RhoA-binding proteins (Watanabe et al., 1996). Rhophilin showed strong GTP-dependent interaction with RhoA, little with RhoC, and less with RhoB. It also shares a putative NH,-terminal coiled-coil domain with PKN/PRKl and PRK2. A family of novel serinekhreonine kinases share a distinct, second Rho effector binding motif (designated class IIREM2) (Fig. 11B). ROKa/Rho-kinase (Leung et al., 1996b), ROKP/ROCK-I1 (Leung et al., 1996a), and p160R0"K (Ishizaki et al., 1996) all share an NH,-terminal kinase domain that has strong sequence identity with the kinase domain of myotonic dystrophy kinase (MDK) (Ishizaki et al., 1996), a central putative coiled-coil domain, and a COOH-terminal PH domain (Mayer et al., 1993b) that is split by a cysteine-rich, zinc-dependent folding domain. The Rho- and Racl-, but not CDC42-, binding protein Citron (Madaule et al., 1995), contains a related NH,-terminal coiled-coil domain that includes the Rho-binding domain but lacks the kinase domain. The Rho-binding sequence, however, exhibits weaker sequence homology to the class I1 motif present in these kinases. Hence, Citron may possess a Rho-binding element distinct from the class I1 motif. The 183-kd Citron protein also contains a COOH-terminal cysteinerich motif, a PH domain, and a proline-rich, putative SH3-binding motif. Thus, like some Ras effectors, Citron may also function as a scaffolding protein for Rho. Whether the class I1 Rho-binding motif is also important for the interaction of these kinases with Racl has not been addressed. Roka was identified in expression screening for the RhoA-binding protein serinekhreonine kinase (Leung et al., 1996b). ROKa showed a cytoplasmic location, and coexpression of RhoA( 14V) caused an increased association of ROKa with the pellet fraction, suggesting that RhoA promoted its translocation to membranes. Injection of DNA encoding full-length ROKa caused the formation of stress fibers and focal adhesions in HeLa cells (Leung et al., 1996a).A closely related protein with 64% overall identity to ROKa and with 90% identity to the ROKa kinase domain was subsequently isolated (ROKP; also ROCK-11). In vitro binding analyses using purified recombinant proteins showed that both ROKa and ROKP bound RhoA, RhoB, and RhoC, but not Racl or CDC42 (Leung et al., 1996a), although subsequent yeast two-hybrid analysis also showed interaction with Racl (Joneson et al., 1996a).

94

Roya Khosravi-Far el al.

Narumiya and colleagues identified p l 60K""K (a Rho-associated coiledcoil containing protein kinase) as a novel RhoA-binding protein serinekhreonine kinase that showed 44% identity with the kinase domain of MDK (Ishizaki etal., 1996).RhoA was shown to associate with p160''"CK in vivo, and this association promoted weak enhancement of its autophosphorylating activity. Finally, Matsui and colleagues (1996) identified the p l 6 4 Rhokinase by affinity chromatography analyses for RhoA-GTP-binding proteins. Rho-kinase is likely a splice variant of ROKa because they differ only at their NH,-termini, with Rho-kinase containing an additional nine amino acids. Because the kinase activities of these proteins show only limited stimulation by association with RhoA, RhoA binding may facilitate other events that lead to full kinase activation. Another RhoA-GTP-binding protein, myosin phosphatase, together with Rho kinase, provides a mechanism for RhoA regulation of stress fiber formation (Kimura et d . , 1996).The myosin-binding subunit (MBS) of myosin phosphatase was identified as a RhoA-binding protein by affinity chromatography analysis of bovine brain membrane extracts. RhoA binding and activation of Rho kinase results in Rho kinase phosphorylation of MBS. This results in the inactivation of myosin phosphatase activity, leading to activation of myosin through a net increase in myosin light-chain phosphorylation by other protein kinases, ultimately causing the formation of stress fibers. The COOH-terminal sequences of MBS that bind to RhoA have structural similarities (polybasic region followed by leucine zipper-like motif) to PKN/PRKl, but no detectable class I1 RhoA-binding motif was observed (Kimura et al., 1996). Because Citron, Rhotekin, and Rhophillin lack known catalytic functions, essentially nothing has been described for their functions. At least three possible mechanisms for the function of these proteins can be envisioned. They may either be directly involved in mediating some aspect of Rho function that does not require a catalytic function by promoting their translocation to various components of cytoskeleton, or they may somehow negatively regulate the function of Rho proteins by forming an inactive complex and preventing the accessibility to other downstream effectors. The third possible function is that they may act as scaffolding proteins that regulate the coordinate activities of Rho proteins and their downstream effectors in a multimeric structure.

C. Structural Requirements for Rho Family Protein-Effector Interactions In summary, each Rho family protein is likely to use multiple effector-mediated pathways to cause its diverse array of cellular effects. Like Ras, the

Increasing Complexity of Ras Signal Transduction: Involvement of Rho Family Proteins

95

regions of Rho family proteins important for effector interaction also appear to be complex. The region analogous to the core Ras effector sequences (Ras residues 32 to 40) is important for all Rho family protein interaction with effectors. All Rho family proteins share a n additional 13-amino acid insert sequence between Ras residues 123 and 124, which distinguishes them from the other members of the Ras superfamily (Chardin et al., 1993), and there is evidence that this sequence may be important for interaction with at least some effectors (Freeman et al., 1996). Studies of chimeric proteins between Racl and RhoA also identified a second, carboxyl terminal region of Racl (residues 143 to 175) important for Rac effector function. Both amino and carboxyl terminal regions of Racl have been shown to be required for PAK and p67phoXbinding as well as for Rac-induced lamellipodia formation (Diekmann et al., 1995). In agreement with their observations, we also observed that the Rac7”Rho chimera did not bind PAK or induce lamellipodia, but instead induced stress fiber formation (Westwick et al., 1997). Furthermore, we observed a loss of JNK activation, suggesting that the effector required for JNK activation also requires interaction with carboxyl terminal Rac sequences. Thus, the effector-interacting sequences of Rho family proteins are also composed of multiple, distinct sequences.

VII. A SEARCH FOR THE MISSING LINK BETWEEN R a s A N D R h o FAMILY PROTEINS The studies presented above clearly establish a critical role for Rho proteins in mediating Ras-induced cytoskeletal alterations, cellular transformation, gene expression, and cell cycle progression. However, the components that transmit the signal from activated Ras to Rho proteins remain to be determined. Evidence that supports the involvement of known Ras-binding proteins as candidate effectors linking Ras with Rho proteins is summarized in this section. At least five candidate effectors of Ras display properties that support their possible role in linking Ras to Rho family proteins (see Fig. 5 ) . First, similar to S . pombe Ras (Chang et al., 1994), a mammalian Rho-GEF (Whitehead et al., 1997) that is analogous to yeast Scdl may function as an effector and directly link Ras with specific Rho family proteins. Although more than 20 candidate mammalian Rho-GEFs, or Dbl family proteins, have been identified, none has been implicated as a Ras-GTP binding protein (Whitehead et al., 1997). Nevertheless, by analogy to Ras proteins, it is anticipated that a Rho GEF will still be an important intermediate between Ras and Rho proteins. Second, Ras interaction with a p120 GAP-p190 Rho-GAP complex may lead to activation or inactivation of Rho family proteins (Settleman et

96

Roya Khosravi-Far el al.

al., 1992). Overexpression of p190 GAP has been shown to cause inhibition of stress fiber formation (Ridley et al., 1993). I t is not known, however, whether Ras association influences p l 9 0 Rho-GAP function. Indirect evidence for such a role is provided by the observation that dominant negative mutants of p120 GAP block Ras activation of JNKs but not MAPKs (Clark et al., 1997b). Third, three groups independently identified a Ral-binding protein (RalBPl/RLIPl/RIPl) that interacted with Ral in a GTP-dependent manner, required an intact Ral effector domain, and contains a Rho-GAP domain (Cantor et al., 1995; Jullien-Flores et al., 1995; Park and Weinberg, 1996). This protein may serve as a link between Ras and Rho proteins (Feig et al., 1996). Ras association with RalGDS may lead to the activation of Ral. Activated Ral then associates with RalBPl/RLIPl/RIPl and alters the activity of certain Rho family members. This is an intriguing model that remains to be validated. This Ral-binding protein showed strongest GAP activity toward CDC42, and to a lesser degree toward Rac, but not toward RhoA. Ral association and regulation of RalBPl GAP activity also have yet to be demonstrated in vivo. Fourth, Ras interaction and activation of MEKKl may explain how Ras mediates activation of JNK and p38, although this pathway would not be expected to involve a Rho family protein. Finally, although PI3K has been shown to be a candidate effector of Ras by interacting with Ras in a GTP-dependent manner, it is also thought to be required for activation of Rac proteins (Hu et al., 1995). Furthermore, constitutive activated mutants of PI3K activate JNK (Klippel et al., 1996),although a second study did not show this activation (Marte et al., 1996). Despite these five possibilities, the exact mechanism by which signals from activated Ras proteins are transmitted to cause activation of Rho family proteins still remains to be characterized. A direct demonstration that specific Rho family proteins are constitutively activated in Ras-transformed cells also remains to be shown.

V111. INCREASING COMPLEXITY OF R a s SIGNAL TRANSDUCTION: A BOON OR A BUST FOR DRUG DISCOVERY AND THE DEVELOPMENT O F ANTI-Ras DRUGS FOR CANCER TREATMENT? A promising approach for targeting Ras for cancer treatment involves the use of farnesyltransferase inhibitors (FTIs) (Cox and Der, 1997). FTIs block Ras function by preventing its posttranslational modification by the farnesyl isoprenoid. Although evidence supports the potential use of FTIs for cancer treatment, some concerns exist because FTIs block all Ras function. Thus, because Ras protein function is believed to be central to so many cel-

Increasing Complexity of Ras Signal Transduction: Involvement of Rho Family Proteins

97

lular processes, targeting only a subset of Ras functions by downstream intervention may provide significant advantages. Therefore, another approach for targeting Ras protein function involves the development of inhibitors that prevent activated Ras form relaying its downstream signal by interfering at different points in its signaling cascades (Gibbs and Oliff, 1994; Der et al., 1996). For example, components of the Raf+MEK+MAPK cascade have been targeted by ongoing efforts at a number of pharmaceutical companies. The revelations indicating the complexity of Ras signaling pathways required for transformation are complex, however, and these efforts may be premature. Instead, we must wait until full appreciation can be achieved of the involvement of Ras in the signaling circuitry that regulates cell growth and differentiation. Although the unknown complexities of Ras signaling pose serious concerns with regard to identifying the most appropriate and key downstream components for intervention, there is also evidence that intervention at any number of points may be promising. Ras transformation of rodent fibroblasts can be antagonized by dominant negative mutants of Raf, MEK, o r MAPK (Kolch et al., 1991; Schaap et al., 1993; Cowley et al., 1994; Westwick et al., 1994; Qiu et al., 1995a; Khosravi-Far et al., 1995); by dominant negative Rho family proteins (Qiu et al., 1995a; Khosravi-Far et al., 1995; Qiu et al., 1995b; Prendergast et al., 1995); and by blocking the action of specific transcription factors, including Myc, Fos, Jun, Ets, and NFKB(Sklar et al., 1991; Granger-Schnarr et al., 1992; Langer et al., 1992; Johnson, N. L. et al., 1996; Westwick, Der, and Baldwin, unpublished observation). These observations provoke an apparent paradox. If mutants of Ras that fail to activate the Raf+MEK+MAPK pathway can still cause transformation, then why does blocking this kinase cascade still result in potent inhibition of Ras transformation? One possible resolution to this paradox is that full Ras transformation requires the action of multiple pathways. Impairment of any one of several key pathways alone, however, will significantly impair Rastransforming function and, consequently, may have a significant impact on reversing the malignant and invasive growth properties of cancer cells. Thus, although many components of Ras signaling remain to be identified, and their precise role in Ras transformation established, we may already have sufficient knowledge to target Ras signaling to initiate these drug discovery efforts.

IX. FUTURE DIRECTIONS A convergence of experimental observations from genetic studies in Drosophila, C. eleguns, and yeast, together with biologic and biochemical

98

Roya Khosravi-Far ef al.

studies in mammalian cells, was required to define the Ras+Raf+MEK+MAPK+Elkl signaling pathway. Thus, a similar convergence of observations will be required to delineate fully the precise signaling pathways and components that connect Ras with Rho family proteins. Additionally, it will be critical to determine which members of the Rho family are directly involved in Ras function and whether oncogenic Ras transformation triggers their constitutive activation. What specific aspects of Ras transformation are triggered by Rho and the other Ras effector-mediated signaling events will be important to establish. Finally, analysis of the interaction of Ras effector domain mutants with each of the candidate Ras effectors, in conjunction with the analysis of biologic and biochemical properties of each of these mutants, should provide valuable information about the role of each of the Ras effectors in mediating normal and transforming Ras function. Furthermore, the current pace of information generated by analyzing the functional roles of Rho family proteins has indicated that these proteins not only play a critical role in regulation of cytoskeletal organization but also are critical regulators of gene expression and cell cycle progression that may be important for cellular proliferation. A roster of candidate Rho family effectors have been identified in the past year. However, most, if not all, of these effectors appear to regulate different aspects of Rho-induced cytoskeletal organization. The next few years may be interesting times for identifying the other Rho effector molecules that regulate Rho-induced gene expression and cell cycle progression and to define what Rho effector-mediated pathways are mediators of Ras effects. The frequent association (30%) of mutant Ras with human cancers has prompted considerable efforts to identify pharmacologic approaches to antagonize Ras function for cancer treatment. Thus, defining the complexity of downstream signaling pathways important for mediating normal and oncogenic Ras function may have significant ramifications for the direction of such efforts and for the identification of pharmacologic agents that may antagonize functions downstream of Ras important to Ras transformation. Because Ras plays such a critical role in so many diverse cellular processes, these approaches may be more advantageous than drugs that directly antagonize Ras function.

ACKNOWLEDGMENTS We thank Arie Abo, Jon Cooper, Frank McCormick, Rong-Guo Qiu, Marc Symons, and Anne Vojtek for communicating unpublished observations and Jennifer Parrish for excellent assistance in the preparation of the text, references, and figures. O u r studies are supported by NIH grants to CJD (CA42978, CA52072, CA55008 and CA67771) and SC (CA64569 and CA70308).

Increasing Complexity of Ras Signal Transduction: Involvement of Rho Family Proteins

99

REFERENCES Adari, H., Lowy, D. R., Willumsen, B. M., Der, C. J., and McCormick, F. (1988). Science, 240, 5 18-52 1. Akasaka, K., Tamada, M., Wang, F., Kariya, K., Shima, F., Kikuchi, A., Yamamoto, M., Shirouzu, M., Yokoyama, S., and Kataoka, T. (1996).J . Biol. Chem., 271,5353-5360. Albanese, C., Johnson, J., Watanabe, G., Eklund, N., Vu, D., Arnold, A., and Pestell, R. G. (1995). J. Biol. Chem., 270,23589-23597. Albright, C. F., Giddings, B. W., Liu, J., Vim, M., and Weinberg, R. A. (1993). EMBO J., 12, 339-347. Alessi, D. R., Saito, Y., Campbell, D. G., Cohen, P., Sithanandam, G., Rapp, U., Ashworth, A. Marshall, C. J., and Cowley, S. (1994). E M B O J.,13, 1610-1619. Andersen, L. B., Fountain, J. W., Gutmann, D. H., Tarle, S. A,, Glover, T. W., Dracopoli, N. C., Housman, D. E., and Collins, F. S. (1993).Nature Genet., 3, 118-126. Aspenstrom, P., Lindberg, U., and Hall, A. (1996).Curr. Biol., 6, 70-75. Baeuerle, P. A,, and Baltimore, D. (1996). Cell, 87, 13-20. Bagrodia, S., Taylor, S. J., Creasy, C. L., Chernoff, J., and Cerione, R. A. (1995).J. Biol. Chem., 270,22731-22737. Bar-Sagi, D., and Feramisco, J. R. (1985). Cell, 42, 841-848. Barbacid, M. (1987).Annu. Rev. Biochem., 56, 779-827. Beranger, F., Goud, B., Tavitian, A., and de Gunzburg, J. (199 1). Proc. Natl. Acad. Sci. U.S.A., 88,1606-1610. Berra, E., Diaz-Meco, M. T., Domingucz, I., Municio, M. M., Sanz, L., Ixzano, J., Chapkin, R. S., and Moscat, J. (1993). Cell, 74, 555-563. Best, A,, Ahmed, S., Kozma, R., and Lim, L. (1996).J. Biol. Chem., 271,3756-3762. Blenis, J. (1993).Proc. Natl. Acad. Sci. U.S.A., 90, 5889-5892. Boguski, M. S., and McCormick, F. (1993). Nature, 366,643-654. Bokoch, G. M., Wang, Y., Bohl, B. P., Sells, M. A., Quilliam, L. A., and Knaus, U. G. (1996). J. Biol. Chem., 271,25745-25749. Bollag, G., Clapp, D. W., Shih, S., Adler, F., Zhang, Y. Y., Thompson, P., Lange, B. J., Freedman, M. H., McCormick, F., Jacks,T., and Shannon, K. (1996).Nature Genet., 12,144-148. Bonner, T. I., Kerby, S. B., Sutrave, P., Gunnell, M. A,, Mark, G . , and Rapp. U. R. (1985). M o l . Cell. Biol.,5 , 1400-1407. Bos, J . L. ( 1 989). Cancer Res., 49, 4682-4689. Bourne, H. R., Sanders, D. A., and McCormick, F. (1990). Nature, 349, 117-126. Braselmann, S., and McCormick, F. ( 1995). EMBO J., 14,4839-4848. Brill, S., Li, S., Lyman, C. W., Church, D. M., Wasmuth, J. J., Weissbach, L., Bernards, A., and Snijders, A. J. (1996). Mol. Cell. Biol., 16, 4869-4878. Broach, J. R. (199 1 ). Trends Genet., 7,28-33. Brtva, T. R., Drugan, J. K., Ghosh, S., Terrell, R. S., Campbell-Burk, S., Bell, R. M., and Der, C. J. (IYYS).]. Biol. Chem., 270,9809-9812. Burbelo, P.D., Drechsel, D., and Hall, A. (IYYS).]. Biol. Chem., 270, 29071-29074. Burridge, K., and Chrzanowska-Wodnicka, M. ( I 996). Annu. Rev. Cell Dev. Biol., 12, 463519.

Cales, C., Hancock, J. F., Marshall, C. J., and Hall, A. (1988). Nature, 332, 548-551. Cantor, S. B., Urano, T., and Feig, L. A. (1995). M o l . Cell. Biol.,15,457811584. Carpenter, C. L., and Cantley, I.. C. (1996). Curr. Opin. Cell Biol., 8, 153-158. Carpino, N., Wisniewski, D., Strife, A., Marshak, D., Kobayashi, R., Stillman, B., and Clarkson, B. (1997).Cell, 88, 197-204. Cerione, R. A,. and Zheng, Y. (1996). Curr. Opin. Cell Biol., 8, 216-222.

I00

Roya Khosravi-Far el al.

Chan, A. M. L., Miki, T., Meyers, K. A., and Aaronson, S. A. (1994).Proc. Nut/. Acad. Sci. U.S.A., 91,7558-7562. Chang, E. C., Barr, M. Wang, Y., Jung, V., Xu, H.-P., and Wigler, M. H. (1994). Cell, 79, 131-141. Chant, J., and Stowers, I.. (1995).Cell, 81, 1-4. Chardin, P., Boquet, P., Madaule, P., Popoff, M. R., Rubin, E. J., and Gill, D. (1989). E M 1 3 0 I., 8, 1087-1092. Chardin P., Dickey, B. E, Birnbaumer L. (Eds). (1993).“GTPases in Biology I . ” Berlin: SpringerVerlag, 12, pp. 159-176. Chen, Q., Lin, T. H., Der, C. J., and Juliano, R. 1.. (1996). /. Biol. Chem., 271, 18 122-1 8127. Chen, Y. R., Meyer, C. F., and Tan, T.-H. (1996).I. B i d . Chem., 271,631-634. Chou, M. M., and Blenis, J. (1995).Curr. Opin. Cell Biol., 7, 806-814. Chou, M. M., and Blenis, J. (1996).Cell, 85, 573-583. Chrzanowska-Wodnicka, M., and Burridge, K. (1996)./. Cell Biol., 133, 1403-1415. Chuang, T.-H., Xu, X., Kaartinen, V., Heisterkamp, N., Groffen, J., and Bokoch, G. M. (1995). Proc. Nut/. Acad. Sci. U.S.A., 92, 10282-10286. Chuang, T. H., Xu, X., Knaus, U. G., Hart, M. J., and Bokoch, G. M. (1993)./. Biol. Chem., 268,775-778. Clark, G. J., Der, C. J., Dickey, B. F., Birnhaumer, L. (Eds). (1993a). “GTPase in Biology 1.” Berlin: Springer-Verlag, 18, pp. 259-288. Clark, G. J., Drugan, J. K., Terrell, R. S., Bradham, C., Der, C. J., Bell, R. M., and CampbellBurk, S. (1995).Proc. Natl. Acud. Sci. U.S.A.. 93, 1577-1581. Clark, G. J., Kinch, M. S., Rogers-Graham, K., Sebti, S. M., Hamilton, A. D., and Der, C. J. (1997a)./. B i d . Chem., 272, 10608-10615. Clark, G. J.. Quilliani, 1.. A., Hisaka, M. M., and Der, C. J. (1993b). Proc. Natl. Acad. Sci. U.S.A., 90,4887-4891. Clark, G. J., Westwick, J. K., and Der, C. J. (1997b3.J. B i d . Chem., 272, 1677-1681. Cohen, C., and Parry, D. A. D. ( 1 994). Science, 263,488-489. Conklin, D. S., Galaktionov, K., and Beach, D. (1995). Proc. Nut/. Acud. Sci. U.S.A., 92, 7892-7896. Cook, S. J., Rubinfeld, B., Albert, I., and McCormick, F. (1993). EMBOI., 12, 3475-3485. Coso, 0. A., Chiariello, M., Yu, J.-C., Teramoto, H., Crespo, P., Xu, N., Miki, T., and Gutkind, J. S. (1995). Cell, 81, 1137-1146. Cowley, S., Paterson, H., Kemp, P., and Marshall, C. J . (1994). Cell, 77, 841-852. Cox, A. D., Brtva, T. R., Lowe, D. G., and Der, C. J. (1994).Oncogene, 9,3281-3288. Cox, A. D., and Der, C. J . (1997). Biochimica. Biophysica. Acta.. in press. Crews, C. M., Aiessandrini, A., and Erikson, R. I.. ( 1 992). Cell Growth Differ., 3, 135-142. Crews, C. M., and Erikson, R. L. (1993).Cell, 74,215-217. Deng,T., Karin, M. (1994).Nature, 371, 171-175. Der, C. J., Cox, A. D., Sehti, S. M., and Hamilton, A. D. ( 1996).Anti-Cancer Drugs, 37, 165-172. Derijard, B., Hibi, M., Wu, I.-H., Barrett, T., Su, B., Deng, T., Karin, M., and Davis, R. J. ( 1 994). Cell, 76, 1025-172. Diaz-Meco, M. T., Lozano, J., Municio, M. M., Berra, E., Frutos, S., Sanz, I>., and Moscat, J. (1994)./. Biol. Chem., 269,31706-31710. Dickson, B., Sprenger, F., Morrison, D., and Hafen, E. (1992).Nature, 360, 600-603. Diekmann, D., Nohes, C. D., Burhelo, P. D., Abo, A., and Hall, A. (1995). E M B O /. 14, 5338-5349. Dominguez, I., Diaz-Meco, M. T., Municio, M. M., Berra, E., Garcia de Herreros, A,, Cornet, M. E., Sanz, L., and Moscat, J. (1992).Mol. Cell. Biol., 12, 3776-3783. Drechsel, D. N., Hyman, A. A., Hall, A., and Glotzer, M. (1996). Curr. E d . , 7, 12-23. Drugan, J. K., Khosravi-Far, R., White, M. A., Der, C. J., Sung, Y.-J., Huang, Y.-W., and Camphell, S. 1.. (1996)./. Biol. Chem., 271, 233-237.

Increasing Complexity of Ras Signal Transduction: Involvement of Rho Family Proteins

101

Eaton, S., Auvinen, P., Luo, L.,Jan, Y. N., and Simons, K. (1995).J . Cell Biol., 131, 151-164. Egan, S. E., and Weinberg, R. A. (1993).Nature, 365,781-783. Ellis, C., Moran, M., McCormick, F., and Pawson, T. (1990).Nature, 343, 377-380. Ezoe, K., Lee, S.-T., Strunk, K. M., and Spritz, R. A. (1994). Oncogene, 9, 935-938. Fabian, J. R., Vojtek, A. B., Cooper, J. A,, and Morrison, D. K. (1994).Proc. Natl. Acad. Sci. U.S.A., 91, 5982-5986. Fantl, W. J., Escohedo, J. A., Martin, G. A., Turck, C. W., del Rosario, M., McCormick, F., and Williams, L. T. (1992). Cell, 69, 413-423. Farrar, M. A,, Alherola-[la, J., and Perlmutter, R. M. (1996). Nature, 383, 178-1 8 1. Feig, L. A., and Cooper, G. M. (1988).Mol. Cell. Biol., 8 , 3235-3243. Feig, L. A. ( 1994). Curr. Opin. Cell Biol., 6 , 204-2 11. Feig, L. A., Urano, T., and Cantor, S. (1996).Trends Biochem. Sci., 21,438-441. Franke, T. F., Ynag, %-I., Chan, T. O., Datta, K., Kazlauskas, A., Morrison, D. K., Kaplan, D. R., and Tsichlis, P. N. (1995).Cell, 81, 727-736. Frech, M., John, J., Pizon, V., Chardi, P., Tavitian, A., Clark, R., McCormick, F., and Wittinghofer, A. ( 1990). Science, 249, 169-1 71. Freeman, J. L., Aho, A., and Lamheth, J. D. (1996).J. Biol. Chem., 271, 19794-19801. Frost, J. A., Xu, S., Hutchison, M. R., Marcus, S., and Cobb, M. H. (1996).Mol. Cell. Biol., 16,3707-3713. Fujisawa, K., Fujita, A,, Ishizaki, T., Saito, Y., and Narumiya, S. (1996).J . Biol. Chem., 271, 23022-23028. Fujita-Yoshigaki, J.. Shirouzu, M., Ito, Y., Hattori, S., Furuyama, S., Nishimura, S., and Yokoyama, S. (1995)./. B i d . Chem., 270, 466111667. Galisteo, M. L., Chernoff,J., Su, Y.-C., Skolnik, E. Y., and Schlessinger,J. ( 1996).]. Biol. Chem., 271,20997-21000. Gertler, F. B., Niehuhr, K., Reinhard, M., Wehland, J., and Soriano, P. (1996).Cell, 87,227-239. Gibbs, J. B., Ellis, R. W., and Scolnick, E. M. (1984). Proc. Natl. Acad. Sci. U.S.A., 81, 57045708. Gibhs, J. B., and Oliff, A. (1994). Cell, 79, 193-198. Graham, S. M., Cox, A. D., Drivas, G., Rush, M. R., D’Eustachio, P., and Der, C. J. (1994). Mol. Cell. B i d , 14,4108-41 15. Graham, S. M., Vojtek, A. B., Huff, S. Y., Cox, A. D., Clark, G. J., Cooper, J . A., and Der, C. J. (1996). Mol. CellBiol., 16, 6132-6140. Granger-Schnarr, M., Benusiglio, E., Schnarr, M., and Sassone-Corsi, P. (1992). Proc. Natl. Acad. Sci. U.S.A., 89, 4236-4239. Habets, G. G. M., Scholtes, E. H. M., Zuydgeest, D., van der Kammen, R. A., Stam, J. C., Berns, A., and Collard, J. G . (1994). Cell, 77,537-549. Han, L., and Colicelli, J. (1995).Mol. Cell. Biol., 15, 1318-1323. Hart, M. J., Maru, Y., Leonard, D., Witte, 0.N., Evans, T., and Cerione, R. A. (1992).Science, 258,812-815. Hartwell, L. (1992).Cell, 71, 543-546. Hata, Y., Kikuchi, A,, Sasaki, T., Schaher, M. D., Gihbs, J. B., and Takai, Y. (1990).J. Biol. Chem., 265,7104-7107. Henkenmeyer, M., Rossi, D. J., Holmyard, D. P., h i , M. C., Mbamalu, G., Harpal, K., Shih, T. S., Jacks, T., and Pawson, T. (1995).Nature, 377,695-701. Hill, C. S., Wynne, J., and Trcisman, R. (1995). Cell, 81, 1159-1 170. Hofer, F., Fields, S., Schneider, C., and Martin, G . S. (1994). Proc. Natl. Acad. Sci. U.S.A., 91, 11089-1 1093. Hu, C., Kariya, K., Tamada, M., Akasaka, K., Shirouzu, M., Yokoyama, S., and Kataoka, T. ( 1 995).]. Biol. Chem., 270, 30274-30277. Hu, Q., Klippel, A., Muslin, A. J., Fantl, W. J., and Williams, L. T. (1995). Science, 268, 100-102.

I02

Roya Khosravi-Far el al.

Huff, S. Y., Quilliam, 1.. A,. Cox, A. D., and Der, C. J. (1997).O n c o g m e , 14, 133-143. Hunter, T. ( 1997). Cell, 88, ,333-346. Hwang, M.-C. C., Sung, Y.-J., and Hwang, Y.-W. (1996).J . Biol. Chem., 271, 8196-8202. Ing, Y. I.., l.etrng, I. W. t.., Heng, H. El. Q., Tsui, I..-C., a n d Lassam, N. J. ( 1 994). O m ~ ) ~ e ~ i e , 9, 1745-1 750. Ishizaki, T., Maekawa, M., Fujisawa, K., Okawa, K., Iwamatsu, A,, Fujita, A,, Watanabe, N., Saito, Y., Kakizuka, A., Morii. N., ;ind Narumiya, S. ( 1996). EMRO J., 15, 1885-1893. Jiang, €{., L.uo, J.-Q., Urano, T., Fmnkel, P., Lu, Z., Foster, D. A., and Feig, L.. A. (199.5).Natrrre, 378, 409-412. Johnson, M. R., DeClue, J. E., Felzmann, S., Vass, W. C., Xu, G., White, R., and t.owy, D. R . (1994). M o l . Cell. B i d , 14, 641-645. Johnson, M. R., Look, A. T., IM;lue, J. E., Valentine, M. B., and Lowy, D. R. (1993).Proc. N d .Acad. Sci. U.S.A., 90, 5539-554.3. Johnson, N. t,.? Gardner, A. M., Diener, K. M., Langc-Carter, C., Gleavy, J., Jarpe, M. B., Minden, A,, Karin, M., Zon, t.. I., and Johnson, G. t.. (1996)./. Riol. Chern., 271, 3229-3237. Johnson, R., Spiegelman, B., Hanahan, I)., and Wisdom, R. (1996). Mol. Cell. Hiol., 16, 4.50445 I I . Joncson, T., McDonough, M., Bar-Sagi, D., and Van Aelst, 1. ( I996a). Sciemr, 274, I 374- 1377. Joneson, T., White, M. A., Wigler, M. H., and Bar-Sagi, D. ( 19961)). Scieiice, 271, 8 10-8 12. Jullien-Florcs, V., Dorseuil, O., Romero, I . , l.etourneur, F.. Saragosti, S., Bcrger. R., l'avitian, A., Gacon, C., and Camonis, J. H. ( 1995).J. B X J ~<:hem., . 270, 22473-22477. Karin. M. ( l 9 9 S ) . J .Biol. Cbem., 270, 16483-16486. Kauffmann-Zeh, A., Rodriguez-Viciana, P., Ulrich, E., Gilbert, C., Coffer, I?, Downward, and J., E V ~ G. , ( I 997). Nosirre, 385,544-548. Khosravi-Far, R., Chrzanowska-Wodnicka, M., Solski, P. A., Eva, A., Burridge, K., and Der, c.J. (1994a).M o l . cell. B i ~ l . ,14, 6848-6857. Khosravi-Far, R., and Ikr, C;. J. (199413).Cuncer Metastrrsis Reit,, 13, 67-89. Khosravi-Far, R., Solski, P. A,, Kinch, M. S., Burridge, K., and Der, C. J . ( 1995).Mol. Cell. h l . , 15,6443-6453. Khosravi-Far, R., White, M. A., Wcstwick,J. K., Solski, P. A,, Chrzanowska-Wodnicka, M., Van Aclst, I.., Wigler, M. H., and Der, C. J. (1996). M o l . Cell. Biol., 16. 3923-3933. Kikuchi, A., Demo, S. D., Ye, Z.-H., Chcn, Y.-W., and Williams, L. T. (1994). Mol. Cell. Biol., 14,7483-749 I . Kikuchi, A., and Williams, L. T. (1996).J . Rrol. Chem., 271, 588-594. Kimura, K., Ito, M., Ammo, M., Chihara, K., Fukata. Y., Nakafuku, M., Yamamori, B., Geng. I., Nakano, T., Okawa, K., Iwaniatsu, A., and Kaibuchi, K. (1996).Science, 273, 245-248. Kirchhausen, T., and Rosen, F. S. (1996). Ctrrr. Riol., 6, 676-678. Kishi, K., Sasaki, T., Kuroda, S., Itoh, T., and Takai, Y. (1993).J . Cell BiCJl., 120, 1187-1 195. Kitayama, H., Matsuzaki, T., Ikawa, Y., arid Noda, M. (1990). Proc. Nut/. A c i d . Sci. U.S.A., 87,4284-4288. Kitayama, H., Sugimoto, Y., Matsuzaki, T., Ikawa, Y., and Noda, M. ( 1 989). Cell, 56, 77-84. Klippel, A,, Reinhard, C., Kavanaugh, M., Apell, G., Escohedo, M.-A., and Williams, 1. T. (1996). Mol. Cell. Biol., 16, 41 17-4127. Kolch, W., Heidecker, G., Kochs, G., Hunimcl, R., Vahidi, H., Mischak, H., Finkenzeller, G., Marmt!, D., and Rapp, U. R. (1993). Nature, 364, 249-252. Kolch, W., Heideckrr, G., I.loyd, P., and Rapp, U. R. (1991).Nntrrre, 349,426428. Kolluri, R., Tolias, K. F., Carpenter, C. I.., Rosen, li. S., and Kirchhausen, T. (1996). Proc. Narl. Acad. Sci. U.S.A., 93, 5615-5618. Kozma, R., Ahmcd, S., Best, A., and l.im, L. (1996). Mol. Cell. Bid., 16, 5069-5080. Krengel, U., Schlichting, I.., Scherer, A., Schumann, R., Frech, M., John, J., Kabsch, W., Pai, E. F., and Wittinghofer, A. (1990). Cell, 62, 539-548.

Increasing Complexity of Ras Signal Transduction: Involvement of Rho Family Proteins

I03

Kuriyama, M., Harada, N., Kuroda, S., Yamamoto, T., Nakafuku, M., Iwamatsu, A., Yamamoto, D., Prasad, R., Croce, C., Canaani, E., and Kaibuchi, K. (1996).J.B i d . Chem., 271,607-610. Kuroda, S., Fukata, M., Kobayashi, K., Nakafuku, M., Nomura, N., Iwamatsu, A., and Kaibuchi, K. (1996).J.Biol. Chem., 271, 23363-23367. Kyriakis, J., and Avruch, J. (1996).Bioessays, 18, 567-577. Lamarche, N., Tapon, N., Stowers, L., Burbelo, P. D., Aspenstrom, P., Bridges, T., Chant, J., and Hall, A. (1996).Cell, 87, 519-529. Lange-Carter, C. A., Pleiman, C. M., Gardner, A. M., Blumer, K. J., and Johnson, G. L. (1993). Science, 260, 31 5-3 19. Langer, S. J., Bortner, D. M., Roussel, M. F., Sherr, C. J., and Ostrowski, M. C. (1992). M o l . Cell. Biol., 12, 5355-5362. Larochelle, D. A., Vithalani, K. K., and De Lozanne, A. (1996).J.Cell Biol., 133, 1321-1329. Lee, C.-H., Li, W., Nishimura, R., Zhou, M., Batzer, A. G., Myers, M. G., Jr., White, M. E, Schlessinger, J., and Skolnik, E. Y. (1993).Proc. Natl. Acad. Sci. U.S.A., 90, 11713-11717. Leevers, S. J., Paterson, H. F., and Marshall, C. J. (1994). Nature, 369,4111114. Lenormand, P., McMahon, M., and PouyssCgur, J. (1996).J.Biol. Chem., 271, 15762-15768. Leung, T., Chen, X.-Q., Manser, E., and Lim, L. (1996a). M o l . Cell. Biol., 16, 5313-5327. Leung, T., Manser, E., Tan, L., and Lim, L. (1996b).J. Biol. Chem., 270,29051-29054. Li, W., Hu, P., Skolnik, E. Y., Ullrich, A., and Schlessinger, J. (1992). Mol. Cell. Biol., 12, 5824-58 33. Li, Y., and White, R. (1996).Cancer Res., 56,2872-2876. L6pez-Barahona, M., Bustelo, X. R., and Barbacid, M. (1996).Oncogene, 12,463-470. Luo, L., Hensch, T. K., Ackerman, L., Barbel, S., Jan, L. Y., and Jan, Y. N. (1996).Nature, 379, 837-840. Luo, Z., Diaz, B., Marshall, M. S., and Avruch, J. (1997).M o l . Cell. Biol., 17, 46-53. Mabuchi, I., Hamaguchi, Y., Fujimoto, H., Morii, N., Mishima, M., and Narumiya, S. (1993). Zygote, 1,325-331. Madaule, P., Furuyashiki, T., Reid, T., Watanabe, G., Morii, N., and Narumiya, S. ( 1 995).FEES Lett., 377, 243-248. Manser, E., Chong, C., Zhao, Z.-S., Leung, T., Michael, G., Hall, C., and Lim, L. (1995).J. Biol. Chem., 270, 25070-25078. Manser, E., Leung, T., Salihuddin, H., Zhao, Z.-S., and Lim, L. (1994). Nature, 367, 40-46. Mansour, S. J., Matten, W. T., Hermann, A. S., Candia, J. M., Rong, S., Fukasawa, K., Vande Woude, G. F., and Ahn, N. G . (1994). Science, 265, 966-970. Marais, R., Wynne, J., and Treisman, R. (1993).Cell, 73, 381-393. Marcus, S., Polverino, A., Chang, E., Robbins, D., and Cobb, M. H. (1995).Proc. Natl. Acad. Sci. U.S.A., 92, 6180-6184. Marshall, C. J. ( 1996). C u m Opin. Cell Biol.,8 , 197-204. Marshall, M. S., and Hettich, L. A. (1993). Oncogene, 8,4251131. Marte, B. M., Rodriguez-Viciana, P., Wennstrom, S., Warne, P. H., and Downward, J. ( 1996). Curr. Biol.,7, 63-70. Martin, G. A., Bollag, G., McCormick, F., and Abo, A. (1995).E M B O J . , 14, 1970-1978. Martin, G. A., Yatani, A., Clark, R., Conroy, L., Polakis, P., Brown, A. M., and McCormick, F. (1992).Science, 255, 192-194. Masuda,T., Kariya, K.-i., Shinkai, M., Okada, T., and Kataoka, T. (199S).J.Biol. Chem., 270, 1979-1982. Matsui, T., Amano, M., Yamamoto, T., Chihara, K., Nakafuku, M., Ito, M., Nakano, T., Okawa, K., Iwamatsu, A., and Kaibuchi, K. (1996).E M B O J . , 15, 2208-2216. Mayer, B. J., and Baltimore, D. (1993a). Trends Cell Biol., 3, 8-13. Mayer, B. J., Ren, R., Clark, K. L., and Baltimore, D. (1993b). Cell., 73,629-630. McGlade, J., Brunkhorst, B., Anderson, D., Mbamalu, G . , Settleman, J., Dedhar, S., RozakisAdcock, M., Chen, L. B., and Pawson, T. (1993).E M B O J., 12, 3073-3081.

104

Roya

Khosravi-Far el al.

Milburn, M. V., Tong, I.., DeVos, A. M., Brunger, A., Yamaizunii, Z., Nishimura, S., and Kim, S.-H. ( 1 990). Science, 247, 939-945. Minden, A,, Lin, A., Claret, F.-X., Abo, A., and Karin, M. (1995).Cell, 81, 1147-1 157. Minden, A,, Lin, A., McMahon, M., Lange-Carter, C., Derijard, B., Davis, R. J., Johnson, C;. L., and Karin, M. (1994).Science, 266, 1719-1723. Moodie, S. A., Paris, M., Villafranca, E., Kirshmeier, P., Willumsen, B. M., and Wolfman, A. ( 1995). Oncogene, 11,447-454. Moodie, S. A., Willumsen, B. M., Weber, M. J., and Wolfman, A. (1993). Science, 260, 1658-1661. Moodie, S. A., and Wolfman, A. (1994).Trends Genet. 10, 14-18. Morcos, P., Thapar, N., Tusneem, N., Stacey, D., and Tarnanoi, F. (1996).M o l . Cell. Biol., 16, 2496-2503. Morrison, D. K. (1994).Science, 2 6 6 , 5 6 4 7 . Morrison, D. K., and Cutler, R. E., Jr. (1997).Curr. Op. Cell Biol., 9, 174-179. Mosch, H.-U., Roberts, R. L., and Fink, G. R. (1996). Proc. Natl. Acad. Sci. U.S.A., 93, 5352-5356. Murphy, C., Saffrich, R., Grummt, M., Gournier, H., Rybin, V., Rubino, M., Auvinen, P., Liitcke, A., Parton, R. G., and Zerial, M. (1996).Nature, 384,427-432. Nishida, E., and Gotoh, Y. (1993).Trends Biochem. Sci., 18, 128-131. Nishiyama, T., Sasaki, T., Takaishi, K., Kato, M., Yaku, H., Araki, K., Matsuura, Y., and Takai, Y. (1994). Mol. Cell. Biol., 14, 2447-2456. Nobes, C. D., and Hall, A. (199.5).Cell, 81, 53-62. Noda, M., KO, M., Ogura, A., Liu, D.-G., Amano, T., Takano, T., and Ikawa, Y. (1985).Nature, 318, 73-75. Okazaki, M., Kishida, S., Murai, H., Hinoi, T., and Kikuchi, A. (1996). Cancer Res., 56, 23 87-23 92. Oldham, S. M., Clark, G. J., Gangarosa, I.. M., Coffey, R. J., Jr., and Der, C. J. (1996).Proc. Natl. Acad. Sci. U.S.A., 93, 6924-6928. Olson, M. F., Ashworth, A., and Hall, A. (1995).Science, 269, 1270-1272. Palmer, R. H., Ridden, I., and Parker, P. J. (1995).Ettr. J. Biochem., 227, 344-351. Park, S.-H., and Weinberg, R. A. (1996). Oncogene, 11,2349-2355. Parker, F., Maurier, F., Delumeau, I.,Duchesne, M., Faucher, D., Debussche, I.., Dugue, A,, Schweighoffer, F., and Tocque, B. (1996).Mol. Cell. Biol., 16, 2561-2569. Parrini, M. C., Bernardi, A., and Parmeggiani, A. (1996).EMBO J., 15, 1107-1 1 1 1. Peterson, S. N., Trabalzini, L., Brtva, T. R., Fischer, T., Altschuler, D. L., Martelli, P., Lapetina, E. C., Der, C . J., and White 11, G. C. (1996).J. Biol. Chem., 271, 29903-29908. Polverino, A., Frost, J.. Yang, P., Hutchinson, M., Neiman, A. M., Cobb, M. H., and Marcus, S. (1995).J. Biol. Chem., 270, 26067-26070. Pomerance, M., Thang, M. N., Tocque, B., and Pierre, M. (1996). Mol. Cell. Biol., 16, 31 79-31 86. Ponting, C. P., and Benjamin, D. R. (1996).Trends Biochem. Sci., 21,422-42.5. Prasad, R., Gu, Y., Alder, H., Nakamura, T., Canaani, O., Saito, H., Huebner, K., Gale, R. P., Nowell, P. C., Kuriyama, K., Miyazaki, Y., Croce, C. M., and Canaani, E. (1993). Cancer Res., 53, 5624-5628. Prendergast, G. C., Khosravi-Far, R., Solski, P. A., Kurzawa, H., Lebowitz, P. F., and Der, C. J. (1995). Oncogene, 10, 2289-2296. Qiu, R.-G., Chen, J., Kirn, D., McCormick, F., and Symons, M. (199Sa).Nature, 374,457-459. Qiu, R.-G., Chen, J.. McCormick, F., and Symons, M. (199Sb).Proc. Natl. Acad. Sci. U.S.A., 92,1178 1-1 1785. Quilliam, L. A., Khosravi-Far, R., Huff, S. Y., and Der, C. J. (1995). Bioessays, 17, 395404.

Increasing Complexity of Ras Signal Transduction: Involvement of Rho Family Proteins

I05

Quilliam, L. A., Lambert, Q . T., Westwick, J. K., Mickelson-Young, L. A., Sparks, A. B., Kay, B. K., Jenkins, N. A., Gilbert, D. J., Copeland, N. G., and Der, C. J. (1996).]. Biol. Chem., 271,28772-28776. Reid, T., Furuyashiki, T., Ishizaki, T., Watanabe, G., Watanahe, N., Fujisawa, K., Morii, N., Madaule, P., and Narumiya, S. (1996). /. Btol. Chem., 271, 13556-13560. Ridley, A. J. (1996).Curr. Biol., 6,1256-1264. Ridley, A. J., and Hall, A. (1992a). Cell, 70, 389-399. Ridley, A. J., Paterson, H. F., Johnston, C. L., Diekmann, D., and Hall, A. (1992b). Cell, 70, 401 -4 10. Ridley, A. J., Self, A. J., Kasmi, F., Paterson, H. F., Hall, A., Marshall, C. J., and Ellis, C. ( 1 993). EMBO]., 12,5151-5160. Rivero-Lezcano, 0. M., Marcilla, A., Sameshima, J. H., and Robbins, K. C. (199.5).Mol. Cell. Biol., 15, 5725-5731. Rodriguez-Viciana, P., Warne, P. H., Dhand, R., Vanhaesebroeck, B., Gout, I., Fry, M. J., Waterfield, M. D., and Downward, J. (1994).Nature, 370,527-532. Rodriguez-Viciana, P., Warne, P. H., Vanhaesebroeck, B., Waterfield, M. D., and Downward, J . (1996).EMBOY., 15,2442-2451. Rubin, G. M. (1991). Trends Genet. 7, 372-377. Russell, M., Lange-Carter, C. A., and Johnson, G. L. (1995).]. Biol. Chem., 270,11757-1 1760. Saez, R., Chan, A. M.-L., Miki, T., and Aaronson, S. A. (1994).Oncogene, 9,2977-2982. Sato, K. Y., Polakis, P. G., Haubruck, H., Fasching, C. L., McCormick, F., and Stanbridge, E. J. (1994). Cancer Res., 54, 552-559. Satoh, T., and Kaziro, Y. (1992). Cancer Biol., 3, 169-177. Schaap, D., van der Wal, J., Howe, L. R., Marshall, C. J., van Blitterswijk, W. J. (1993).]. Biol. Chem., 268,20232-20236. Satoh, T., Nakamura, S., Nakafuku, M., and Kaziro, Y. (1988). Biochim. Biophys. Acta., 25; 949(1),97-109. Scheffzek, K., Lautwein, A., Kabsch, W., Ahmadian, M. R., and Wittinghofer, A. (1996). Nature, 384,591-596. Schlessinger, J. (1993).Trends Biochem Sci., 18, 273-275. Settleman, J., Narasimhan, V., Foster, L. C., and Weinberg, R. A. (1992).Cell, 69, 539-549. Shirouzu, M., Koide, H., Fujita-Yoshigaki, J., Oshio, H., Toyama, Y., Yamasaki, K., Fuhrman, S . A., Villafranca, E., Kaziro, Y., and Yokoyama, S. (1994).Oncogene, 9, 2153-2157. Sigal, 1. S., Gibbs, J. B., D’Alonzo, J. S., and Scolnick, E. M. (1986). Proc. Natl. Acad. Sci. U.S.A., 83,4725-4729. Simon, M.-N., De Virgilio, C., Souza, B., Pringle, J. R., Abo, A., and Reed, S. 1. ( 1 995). Nature, 376,702-70.5. Sklar, M. D., Thompson, E., Welsh, M. J., Liebert, M., Harney, J., Grossman, H. B., Smith, M., and Prochownik, E. V. (1991).Mol. Cell. Biol., 11, 3699-3710. Smith, M. R., DeGudicibus, S. J., and Stacey, D. W. (1986).Nature, 320, 540-543. Sozeri, O., Vollmer, K., Liyanage, M., Frith, D., Kour, G., Mark, G. E., and Stabel, S. (1992). Oncogene, 7,2259-2262. Spaargaren, M., and Bischoff, J. R. (1994). Proc. Natl. Acad. Sci. U.S.A., 91, 12609-12613. Sprenger, F., Trosclair, M. M., and Morrison, D. K. (1993).Mol. Cell. Biol., 13, 1163-1172. Stanton, V. P., Jr., Nichols, D. W., Laudano, A. P., and Cooper, G. M. (1989).Mol. Cell. Biol., 9,639-647. Sternberg, P. W., and Horvitz, H R. (1991).Trends Genet., 7,366-370. Stokoe, D., Macdonald, S. G., C adwallader, K., Symons, M., and Hancock, J. F. (1994).Science, 264,1463-1467. Su, B., Jacinto, E., Hibi, M., Kallunki, T., Karin, M., and Ben-Neriah, Y. (1994). Cell., 77, 727-736.

I06

Roya Khosravi-Far e l a / .

Sulciner, D. J.. Irani, K., Yu, 2.-X., Ferrans, V. J., Goldschmidt-Clermont, P., and Finkel, T. (1996).M o l . Cell. Biol., 16, 71 15-7121. Symons, M., Derry, J. M . J., Karlak, B., Jiang, S., Lemahieu, V., McCormick, F., Francke, U., and Aho, A. (1996). Cell, 84,723-734. Taki, T., Hayashi, Y., Taiiiwaki, M., Seto, M., Ueda, R., Hanada, R., Suzukawa, K., Yokota, J., and Morishita, K. (1996).Oncogene, 13,2121-2130. Tamanoi, F. (1988).Biochim. Biophys. Acta, 948, 1-15. Tanahe, S., Zeleznik-Le, N. J., Kobayashi, H., Vignon, C., Espinosa, R. I., LeBeau, M. M., Thirman, M. J., and Rowley, J. D. (1996). Genes Chromosomes Cancer, 15, 206-216. Teo, M., Manser, E., and Lini, L. (1995).1. B i d . Chem., 270, 26690-26697. Teramoto, H., Coso, 0 . A,, Miyata, H., Miki, T., and Gutkind, J. S. (1996a).J. B i d . Chem., 271,27225-27228. Terarnoto, H., Crespo, P., Coso, 0. A,, Igishi, T., Xu, N., and Gutkind, J. S. (1996h).J. Biol. Chem., 271,25731-25734. The, I., Murthy, A. E., Hannigan, G. E., Jacohy, L. B., Menon, A. G., Gusella, J. F., and Bernards, A. (1993).Nature Genet., 3, 62-66. Thompson, J. D., Higgins, D. G., and Gibson, T. J. (1994).Nucl. Acids Res., 22, 4673-4680. Thorburn, A., Throburn, J., Chen, S.-Y., Powers, S., Shuheita, H. E., Feramisco, J. R., and Chien, K. R. ( 1993).J. B i d . Chem., 268, 2244-2249. Tocque, B., Delumeau, I., Parker, F., Maurier, F., Multon, M.-C., and Schweighoffer, F. ( 1 997). Cell Signaling, 9, 153- 158. Toda,T., Uno, I., Ishikawa, T., Powers, S., Kataoka, T., Broek, D., Cameron, S., Broach, J., Matsumoto, K., and Wigler, M. ( 1985). Cell, 40, 27-36. Tolias, K. F., Cantley, L. C., and Carpenter, C. L. (1995).J. B i d . Chem., 270, 17656-17659. Tong, L., de Vos, A. M., Milhurn, M. V., Jancarik, J., Noguchi, S., Nishimura, S., Miura, K., Ohtsuka, E., and Kim, S.-H. ( 1 989). Nature, 337, 90-93. Treisman, R. (1996a).Semi??.Cancer Biol., 1, 47-58. Treisman, R. ( 1 996b). Curr. Opin. Cell Biol., 8, 205-2 15. Ueda, T., Kikuchi, A,, Ohga, N., Yamamoto, J., and Takai, Y. (1990)./. Biol. Chem., 265, 9373-9380. Urano,T., Emkey, R.,and Feig, L. A. (1996).EMBOJ.. 15, 810-816. Van Aelst, L., Barr, M., Marcus, S., Polverino, A., and Wigler, M. ( 1 993). Proc. Natl. Acad. Sci. U.S.A., 90, 62 13-62 17. Van Aelst, L., Joneson, T., and Bar-Sagi, D. (1996). EMBO /., 15, 3778-3786. Van Aelst, L., White, M. A., and Wigler, M. H. (1994).Cold Spring Harbor Symp. Quant. Biol., 59, 181-186. Vcrheii, M., Hose, R., Lin, X. H., Yao, B., Jarvis, W. D., Grant, S., Birrer, M . J., Szaho, E., Zon, L. I., Kyriakis, J. M., Haimovitz-Friedman, A., Fuks, Z.,and Kolesnick, R. N. (1996).Nuture, 380, 75-79. Vincenz, C., and Dixit, V. ( 1 996).J. B i d . Chem., 271,20029-20034. Vogel, U. S., Dixon, R. A. F., Schaher, M. D., Diehl, R. E., Marshall, M. S., Scolnick, E. M., Sigal, 1. S., and Gibhs, J . €5. (1988).Nature, 335, 90-93. Voitek, A. B., Hollenberg, S. M., and Cooper, J. A. (1993).Cell, 74, 205-214. Wang, Y., Xu, H.-P., Riggs, M., Rodgers, L., and Wigler, M. (1991). Mol. Cell. Biol., 11, 3554-3563. Warne, P. H.. Viciana, P. R., and Downward, J . (1993). Nature, 364, 352-355. Watanahe, G., Saito, Y., Madaule, P., Ishizaki, T., Fujisawa, K., Morii, N., Mukai, H., Ono, Y.. Kakizuka, A,, and Narumiya, S. ( 1 996). Science, 271,645-648. Weisshach, L., Settleman, J., Kalady, M. F., Sniiders, A. J., Murthy, A. E., Yan, Y.-X., and Bernards, A. (1994)./. Biol. Chein., 269, 20517-20521. Westwick, J. K., Cox, A. D., Der, C. J., Cohb, M. H., Hihi, M., Karin, M., and Brenner, D. A. (1994).PrfJC. Natl. Acad. Sci. U.S.A., 91, 6030-6034.

Increasing Complexity of Ras Signal Transduction: Involvement of Rho Family Proteins

I07

Westwick, J. K., 1-ambert, Q. T., Clark, C. J., Symons, M., Van Aelst, L., Pestell, R. G., and Der, C. J. (1997). M d . Cell. Bin/., 17,1324-1335. White, M. A., Nicolette, C., Minden, A., Polverino, A., Van Aelst, L., Karin, M., and Wigler, M. H. ( 1995).Cell, 80, 533-54 1. White, M. A., Vale, T., Camonis, J. H., Schaefer, E., and Wigkr, M. H. ( 1 996)./. B i d . Chenr., 271, 16439-16442. Whitehead, 1. P., Campbell, S., Rossman, K. L., and Der, C.J. (1997).Biochem. Biophys. Acta, 1332,FI-F23. Whitmarsh, A. J., Shore, P., Sharrocks, A. D., and Davis, R. J. (1995). Science, 269,403407. Williams, L. T. ( 1 992).Curr. Opin. Cell B i d , 2, 601-603. Wittinghofer, A., and Nassar, N. (1996).Tretzds Biochem. Sci., 21,488-491. Wolthuis, R. M.,Bauer, B., van’t Veer, I.. J., de Vries-Smits, A. M., Cool, R. H., Spaargarcn, M., Wittinghofer, A,, Burgering, B. M., and Bos, J. L. (1996).Oncogene, 13,353-362. Xia, Z., Dickens, M., Raingeaud, J., Davis, R. J., and Greenberg, M. E. (1995). Science, 270, 1326-1331. Yamamoto, M., Marui, N., Sakai, T., Morii, N., Kozaki, S., Ikai, K., Imamura, S., and Narumiya, S. (1993).Oncogene, 8, 1449-1455. Yamanashi, Y., and Baltimore, D. ( 1 997).Cell, 88,205-21 1. Yan, M., Dai, T., Deak, J. C., Kyriakis, J. M., Zon, 1.. I., Woodgctt, J. R., and Templeton, D. J. Nature, 372, 798. (1994). Yang, X., Khosravi-Far, R., Chang, H. Y., and Baltimore, D. (1997).Cell, 89, 1067-1076. Yatani, A., Okabe, K., Halenbeck, R., Polakis, P., McCormick, F., and Brown, A. M. ( I 990). Cell, 60, 769-776. Yoder-Hill, J., Goluhic, M., and Stacey, D. W. (1995). /. Riol. Cltmz., 270,27615-27621. Yoshida, Y., Kawata, M., Miura, Y., Musha, T., Sasaki, T., Kikuchi, A,, and Takai, Y. (1992). Mol. Cell. B i d , 12,3407-34 14. Zhang, S., Han, J., Sells, M. A., Chcrnoff, J., Knaus, U. G., Ulevitch, R. J., and Bokoch, G. M. ( 1 995).J. Biol. Chem., 270,23934-23936. Zhang, X., Settleman, J., Kyriakis, J. M., Takeuchi-Suzuki, E., Ellcdge, S. J., Marshall, M. S., Bruder, J. T.,Rapp, U. R., and Avruch. J. ( 1993).Nature, 364, 308-3 13. Zhang, Z., Vuori. K., Wang, H.G., Reed, 1. C., and Ruoslahti, E. (1996).Cell. 85, 6 1-69. Zhao, Z.-S., L ~ i i g T., , Manser, E., and Lim, L. (1995). Mol. Cell. Biol.. 15, 5246-5257. Zheng, Y., Bagrodia, S., and Cerione, R. A. ( 1994a)./. Biol. Chem., 269, 18727-1 8730. Zheng, Y., Cerione, R., and Bender, A. ( 1 994h).1. Biol. Chenz., 269,2369-2372.

This Page Intentionally Left Blank

BeMyb: A Key Regulator of the Cell Cycle Mark K. Saville and Roger 1. Watson Department of Medical Mil robiology and Ludwig Institute for Cancer Researc h lmperinl College School of Medicine at Tt. Miiryi Norfolk Place, London W2 I PG. Unitcd Kingdom

I. Introduction 11. B-Myh Structure and Functional Domains A. Evolutionary Conservation of B-Myb

111.

IV.

V.

VI.

B. DNA-Binding Domain C. Transcription Activation Domains D. Negative C-Terminal Domain Transcriptional Regulation of B-niyh A. Transcriptional Regulation of B-myh B. Developmental Regulation of B-myh Expression (1. Mechanism of B-myh Transcriptional Regulation D. B-myb Transcription as a Control Point in the Cell Cycle Modification of B-Myh Protein in the Cell Cycle A. Hyperphosphorylation of B-Myb in S Phase B. Cyclin-Dependent Activation of 6-Myb Transactivation Function Requirement f o r B-Myb in Cell Proliferation A. Effects of Ablating B-myb Expression B. Effects of B-myh Overexpression C. Suppression of Cell Proliferation Blocks by B-Myb B-Myb Function A. Mechanisms of Transcriptional Control by B-Myb B. Possible B-Myb Target Genes C. Conclusions References

1. INTRODUCTION The precise balance between tissue differentiation and regeneration needed to maintain organ size and integrity in higher eukaryotes requires a fine measure of control over cell division. Stem cells that repopulate the tissue are predominantly in a state of quiescence and are presumed to enter the cell division cycle only on stimulation by extrinsic factors, such as hormones and cytokines. Progress through the initial G , phase of the cell cycle is then deAdvances

in

CANCER Rt.SFARCH

0065-2.30X/YR $25.00

I10

Mark K. Saville and Roger I Watson

pendent o n continuous signaling by growth factors until a restriction point is reached in late G , phase (Pardee, 1989). Thereafter, removal of growth factors is unable to halt entry into S phase and the consequent DNA replication. As with unicellular eukaryotes, such as yeast, the basic regulators of the cell cycle in higher eukaryotes are the cyclins and their associated enzymatic workhorses, the cyclin-dependent kinases (Cdks). The properties of these complexes have been reviewed recently (Hall and Peters, 1996), and we touch on only their more salient features here. The D cyclins are expressed first on emergence from the G , quiescent state and act together with their catalytic Cdk4 or Cdk6 partners to move cells through G I phase (Baldin et al., 1993; Quelle et al., 1993). Progress through the Gl/S transition is also dependent on later expression of cyclin E, which acts in concert with Cdk2 (van den Heuvel and Harlow, 1993). Evidence that D cyclins and cyclin E have different roles in this transition is provided by the finding that although expression of each individually shortens G I , simultaneous expression of both has an additive effect on accelerating entry into S phase (Resnitzky and Reed, 1995). Cyclin A expression is induced in S phase and is continuously required for maintenance of DNA synthesis (Girard et al., 1991; Pagano et al., 1992), whereas cyclin B orchestrates the GJM transition. The most critical role of cyclin D- and E-dependent kinases is removal of an inhibitory block in G I by direct phosphorylation of the retinoblastoma protein, pRb. In its hypophosphorylated state, pRb interacts with a number of different transcription factors; significantly, hyperphosphorylation mediated by D cyclins, cyclin E, and their associated Cdks reverses these associations. On current evidence, by far the most important interaction of pRb is made with the E2F transcription factor, which is a heterogeneous entity composed of heterodimers between members of the E2F and DP protein families (La Thangue, 1994).There is clear evidence that E2F is a major regulator of the G,/S transition (reviewed in Lam and La Thangue, 1994; Muller, 1995), and ectopic expression of the E2F-1 component is able to drive certain cells into the S phase of the cycle (Johnson et al., 1993). The remarkably frequent genetic alterations to the cyclin D/pRb pathway found in human cancers (Hall and Peters, 1996) emphasize the importance of this regulatory mechanism to the maintenance of normal cell proliferation. The ability of E2F to orchestrate the Gl/S transition is thought to reflect its role in the control of transcription of a number of cellular genes required for entry into and passage through S phase. These genes include some whose products are involved directly in DNA biosynthesis, such as dihydrofolate reductase, thymidine kinase, and DNA polymerase-cY (Blake and Azizkhan, 1989; Pearson et d., 1991; Ogris et al., 1993), and others that play regulatory roles during S phase, such as cyclin E, cyclin A, the pRB-related protein p107 and the E2F1 component of E2F activity (Hsiao et al., 1994; Johnson et al., 1994; Neumann et al., 1994; Ohtani et al., 1995; Schulze et al., 1995; Zhu et al., 1995).

B-Myb: A Key Regulator of the Cell Cycle

111

Of particular relevance to this review is the observation that E2F is also a regulator of B-myb gene transcription (Lam and Watson, 1993). B-myb was first identified by screening human cDNA libraries at low stringency with a c-myb probe (Nomura et ul., 1988). An additional member of this gene family, A-myb, was also identified in this screen. Extensive homology was evident among the predicted amino acid sequences of all three proteins, implying that the Myb family have closely related cellular functions. It is therefore of interest to note that c-myb is the cellular progenitor of the avian retrovirus oncogene, v-myb, and has been ascribed a regulatory role in both cell proliferation and differentiation (reviewed in Luscher and Eisenman, 1990; Weston, 1990; Lyon et ul., 1994). This role presumably reflects the activity of c-Myb as a sequence-specific transcription factor (Nishina et al., 1989; Weston and Bishop, 1989). A number o f parallels can be drawn between the properties of E2F-1 and B-Myb; that is, their expression is regulated similarly in the cell cycle by E2F activity, they are both sequence-specific transcription factors, and both have characteristics of oncogene products (Sala and Calabretta, 1992; Xu et al., 1995). The purpose of this review is to discuss the evidence that, like E2F1, B-Myb may also be a critical regulator of the cell cycle.

11. B-Myb STRUCTURE AND FUNCTIONAL DOMAINS

A. Evolutionary Conservation of B-Myb The three known members of the mammalian myb gene family display a number of conserved regions. They appear to have evolved from a primordial myb gene concurrently with the emergence of vertebrate species. Thus, the fruitfly Drosophilu melunoguster has a single myb gene (Katzen and Bishop, 1996), whose product shows equivalent regions of homology with all three mammalian Myb family members; whereas distinctive B-myb, c-myb, and A-myb genes are present in the amphibian Xenopus luevis and in chickens (Bouwmeester et ul., 1992; Foos et al., 1992). Mouse and Xenopus B-Myb proteins show homology throughout much of their length (Fig. 1).The largest region of nonidentity is located between amino acids 190 and 286 in mouse B-Myb; this region also has the greatest number of mismatches between mouse and human B-Myb (Lam et al., 1992). The extent of sequence conservation over this extended evolutionary timeframe implies a high degree of conservation of domain function. As expected, sequence conservation between different members of the Myb family in a particular mammalian species is less than that between BMyb proteins in distantly related species. The three related Myb proteins are of similar size and have a number of regions of homology indicating con-

Mark K. Saville and Roger 1. Watson Mouse A-Myb

Xenopus 8-Myb

Fig. I Dot matrix comparison of mouse B-Myb with Xenopus sp. B-Myb, mouse A-Myh, and mouse c-Myb. Their respective amino acid positions are indicated on the X and Y axes.

served domains (see Fig. 1 ) . B-Myb is more similar to A-Myb than to c-Myb (see Fig. l ) ,and this is due in part to an additional region of homology with A-Myb (B-Myb amino acids 308 to 332) that is absent in c-Myb. Moreover, sequence comparisons show that the major c-Myb species is actually a truncated derivative of an archetypal myb protein, and a variant c-Myb protein (p85 c-Myb) encoded by an incompletely spliced mRNA contains an additional 121 amino acids with further homology to both B-Myb and A-Myb (Lyon et al., 1994). Determination of the B-Myb functional domains has been based largely o n previous studies of c-Myb, and it is convenient to discuss B-Myb protein structure by reference to these earlier studies.

B. DNA-Binding Domain The most conserved region within the Myb proteins is located near the amino-terminus and corresponds to the c-Myb DNA binding domain (Fig. 2). The c-Myb and A-Myb proteins show 90% amino acid identity in this region, whereas homology between c-Myb and B-Myb is only 75% (see Fig. 1). The DNA-binding domain is composed of three imperfect reiterations of 51 t o 52 amino acids (termed R1, R2, and R3), the length and arrangement of which are perfectly conserved among all three proteins. Transduction of the c-myb gene by the AMV and E26 avian retroviruses resulted in deletion of much of R1 (Klempnauer et al., 1983), but it is clear that only R2 and R3 are absolutely essential for sequence-specific DNA binding (Howe et al., 1990; Oehler et al., 1990). Sophisticated NMR and other structural studies have revealed that the c-Myb R2 and R3 repeats each contain a variant he-

B-Myb: A Key Regulator of t h e Cell Cycle

I I3

C-Myb

(636aa)

B-Myb (704 aa) Fig. 2 Comparison of the c-Myb and B-Myb domain structures. The c-Myb DNA-binding, transactivation (Tx),leucine zipper (LZ),and conserved region (CR) domains are indicated. The percentage of amino acid (aa) identity for the DNA-binding and conserved regions is shown in the B-Myb domains.

lix-turn-helix motif involved in DNA recognition (Ogata et al., 1995). The structure formed by the three R3 helices is rigid, whereas that in R 2 undergoes conformational fluctuations and only forms a stable structure when bound to the specific target DNA (Ogata et al., 1996). Studies of B-Myb show a similar picture (Carr et al., 1996), with strict conservation of the amino acids that comprise the distinctive helical structures of R 2 and R3. Myb proteins bind DNA as monomers (Howe et al., 1990); cooperative binding of R2 and R 3 within the major groove to the consensus Myb binding site (MBS; consensus sequence CRAACNG) ensures the necessary sequence specificity (Ogata et al., 1995). The precise function of R1 has not been defined, but it has been suggested that it may interact nonspecifically with DNA to stabilize binding (Ogata et al., 1995). It is notable that B-Myb binds DNA with similar sequence specificity to cMyb (Mizuguchi et al., 1990; Howe and Watson, 1991), although some distinct preferences for particular nucleotides flanking the core binding site were observed (Howe and Watson, 1991). It is likely, therefore, that the different Myb proteins interact with the regulatory regions of an overlapping range of target genes, and it is clear that when considering B-Myb function, the influence of c-Myb and A-Myb cannot be ignored.

C. Transcription Activation Domains The c-Myb transactivation domain has been localized to a weakly acidic sequence (minimally, amino acids 275 to 325) lying downstream of the DNA-binding domain (Sakura et al., 1989; Weston and Bishop, 1989). A C-terminally deleted c-Myb protein containing only the DNA-binding and transactivation domains efficiently transactivates an SV40 early promoter

I14

Mark K. Saville a n d Roger 1. Watson

regulated by six upstream MBS (the SV40/MBS reporter). The c-Myb transactivation domain shows partial homology with an equivalent region of AMyb, but no homology with B-Myb is evident (see Fig. 1). Despite lack o f sequence conservation, it is clear that an equivalent B-Myb region (amino acids 205 to 338) is required for transactivation of the SV40/MBS reporter (Nakagoshi et al., 1993). This B-Myb region is referred to as the acidic domain, because overall, it has a weakly negative charge. In contrast to c-Myb, a C-terminal B-Myb truncation that leaves intact the DNA-binding domain and the acidic domain is not sufficient for activation of the SV40/MBS reporter (Nakagoshi et al., 1993). Additional C-terminal sequences required for activation of the reporter map to a conserved region (CR: amino acids 468 to 545), which demonstrates significant homology both within the Myb family and with the evolutionarily distant Drosophila sp. myh protein (Tashiro et al., 1995). Two nuclear localization signals located in the B-Myb C-terminus are also required for efficient transactivation activity (Takemoto et al., 1994). In contrast to the work described previously, other studies failed to demonstrate activation of MBS-regulated promoters by B-Myb (Foos et al., 1992; Watson et al., 1993); indeed, B-Myb competitively inhibited c-Myb transactivation function o n the mim-1 and SV40/MBS promoters. A possible resolution of these conflicting results is suggested by the finding that B-Myb transactivation activity is cell type specific (Tashiro et al., 1995). Comparison of permissive and nonpermissive cell types showed that transactivation function correlates with the presence of specific cellular proteins able to interact in uitro with the B-Myb CR (Tashiro et al., 1995).The identity of these cellular proteins has yet to be determined, and their actual contribution to B-Myb function has therefore not been established. As discussed later (section VI.A), transcriptional regulation by B-Myb is complex and may depend on a number of factors. It is also pertinent to note that in the initial studies of B-Myb functional domains, transactivation was not dependent o n specific MBS binding (Nakagoshi et al., 1993); therefore, the nature of the activity measured i s open to question.

D. Negative C-Terminal Domain The C-terminal half of the c-Myb protein, downstream of the activation domain, is implicated in negative regulation of protein function through inhibitory effects on DNA-binding, transactivation, and transformation (Kalkbrenner et al., 1990; Hu et al., 1991; Dubendorff et al., 1992). Three distinct negative regulatory elements have been identified in this region, including a putative leucine zipper (Kanei-Ishii et al., 1992) and the recently described EVES motif (Dash et al., 1996). It has been suggested that the

B-Myb: A Key Regulator of the Cell Cycle

I I5

leucine zipper and EVES motif inhibit c-Myb DNA binding, respectively, by promoting formation of an inactive homodimer (Nomura et al., 1993) or by making intramolecular contact with the DNA-binding domain (Dash et al., 1996). Significantly, neither of these motifs is found in B-Myb. The third negative regulatory element spans the c-Myb CR (Sakura et al., 1989; Dubendorff et al., 1992), however, the most conserved part of this sequence is not essential for inhibition of c-Myb transactivation function (Dubendorff et al., 1992). This suggests that the B-Myb CR may not be an exact functional counterpart of this c-Myb domain, indeed, the B-Myb CR acts positively rather than negatively on transactivation function (Nakagoshi et al., 1993). The initial characterization of B-Myb functional domains suggested that this protein differs fundamentally from c-Myb in that its C-terminus has no inhibitory effects on transactivation activity (Nakagoshi et al., 1993). Subsequent studies demonstrated that whereas wild-type B-Myb is unable to bind DNA in in vitro assays, latent activity could be progressively uncovered by successive deletions of the C-terminus (Watson et al., 1993). In addition, we found that elimination of B-Myb sequences downstream of the CR markedly potentiates transactivation activity (Lane et al., 1997). Both latter findings suggest that the extreme B-Myb C-terminus does contain a negative regulatory domain (NRD).It may be significant that, in contrast to previous studies, Lane and colleagues ( 1997) measured B-Myb transactivation mediated through specific DNA-binding to the promoter MBS. Using transfected cell extracts, it is now apparent that B-Myb truncated downstream of the CR is competent to bind to a n MBS oligonucleotide, whereas binding cannot be demonstrated with the wild-type protein (R. Watson, unpublished observations). It is probable, therefore, that removal of the C-terminal NRD enhances transactivation mediated through specific DNA binding, but this domain is irrelevant when transactivation activity is measured on reporters in which MBS are either absent or dispensable.

111. TRANSCRIPTIONAL REGULATION OF B-my6

A. B-my6Transcription in Somatic Cells It has been recognized since the earliest description of B-my6 (Nomura et al., 1988) that expression of this gene is widespread in different cell types. Subsequent studies have confirmed that B-myb mRNA is detectable in every one of the numerous cell lines in which it has been examined (Golay et al., 1991; Reiss et a/., 1991; Lam et al., 1992; Kamano et al., 1995), suggesting that it may in fact be transcribed in all cell lineages. In this respect, B-my6 stands apart from the other members of the family because it is evident that

I I6

Mark K . Saville and Roger 1. Watson

expression of both c-my6 and A-my6 demonstrates marked tissue tropism (Westin etal., 1982; Nomura etal., 1988; Golay etal., 1996).Although transcription of B-my6 may be ubiquitous within cell lineages, it is in no sense a “housekeeping gene” because several studies have shown a strong positive correlation between cell proliferation and B-myb expression (Golay et al., 1991; Reiss et al., 1991; Lam et al., 1992; Marhamati and Sonenshein, 1996). This association can be seen most readily on mitogenic stimulation of quiescent cells, for example, serum-deprived murine fibroblasts and human peripheral blood T lymphocytes. In these cell types, B-my6 expression is completely downregulated during quiescence; on mitogenic stimulation, induction of B-my6 transcription was found to be delayed until late G , phase, reaching maximal mRNA levels in S phase, and thereafter declining as cells approached the subsequent G , phase (Golay etal., 1991; Reiss et al., 1991; Lam et al., 1992). A number of lines of evidence indicate that B-my6 transcription is determined by the stage of the cell cycle rather than by the presence of mitogens per se. In this respect B-my6 appears to be a true cell cycle-regulated gene unlike, for example, c-myc, whose transcription is determined by mitogenic stimulation rather than by the cell cycle clock (Thompson et al., 1985).Thus, by synchronizing cells in M phase with nocodazole, it was apparent that Bmy6 mRNA levels fluctuated on release into cycle, declining in G , and reaching maximal levels in S phase, despite the constant presence of growth factors (Lam et al., 1992). Moreover, during differentiation of both hema-topoietic and neuroblastoma cell lines, it is clear that the decrease in B-my6 mRNA levels closely parallels the progressive loss of cell proliferation rather than correlating with the differentiation state (Lam et al., 1992; Raschella et al., 1995; Bies et al., 1996). In contrast to B-my6, transcription of the other members of the my6 family is not rigorously directed to late G , and S phases. Indeed, the high levels of c-my6 expression observed in immature hematopoietic cells are associated with invariant mRNA levels during the cell cycle (Thompson et al., 1986; Catron et al., 1992). In more mature hematopoietic cells and in certain nonhematopoietic cells, in which c-my6 mRNA is less abundant, c-my6 transcription is regulated in the cell cycle with kinetics superficially similar to Bmy6 (Thompson et al., 1986; Brown et al., 1992; Olson et al., 1993). Even so, induction of c-my6 transcription appears to be more closely associated with the mitogenic signal than with the stage of the cell cycle (Golay et al., 1991; Lyon et al., 1994; Golay et al., 1992). A-my6 expression shows no correlation with that of B-my6. Indeed, A-my6 is transcriptionally active in quiescent T cells, and induction into cycle results in a gradual decrease in transcription levels (Golay et al., 1991). All three my6 genes are expressed in certain cells, most notably peripher-

B-Myb: A Key Regulator of t h e Cell Cycle

I I7

al blood T lymphocytes and tonsillar B lymphocytes (Golay et al., 1991). Their different kinetics of expression suggests that the individual Myb proteins either perform distinct roles o r d o similar tasks at different phases of the cell cycle. Taking into consideration that B-my6 is expressed in cells of all lineages but that transcription of c-my6 and A-my6 is markedly cell type specific, it may be suggested that B-Myb plays a general role in G,/S transition or during S phase, whereas c-Myb and A-Myb have more specific roles, perhaps in the differentiation of particular cell lineages.

B. Developmental Regulation of B-my6 Expression A general role for B-my6 in cell proliferation is supported by studies of its expression during embryonic development (Sitzmann et al., 1996). Thus, B-my6 is the only member of the my6 family expressed during early stages (i.e., before day 10)of murine embryogenesis (Sitzmann et al., 1996). In the frog, Xenopus laevis, B-my6 gene expression is also detectable at the earliest embryonic stage (Bouwmeester et al., 1992). Moreover, B-my6 is expressed widely in mouse embryonic as well as extraembryonic tissues, whereas c-my6 and A-my6 expression is restricted to particular tissues (Sitzmann et al., 1996). Expression of B-my6 within particular tissues of the developing embryo correlates precisely with histone H 4 expression (a marker of cell proliferation). For example, B-my6 is abundantly expressed in the highly proliferative dorsal part of the day-10 neural tube, whereas the fully differentiated motorneurons in the ventral region, which have ceased cell division, d o not express B-my6 (Sitzmann et al., 1996). Although studies of mouse embryogenesis reinforce the notion that B-Myb plays a role in cell division, these events appear to be uncoupled during spermatogenesis in the adult mouse (Sitzmann et al., 1996). Thus, the highest levels of B-my6 expression in testis were found in cells undergoing meiosis, whereas mitotic stem cells located in the periphery of the seminiferous tubules showed insignificant expression. As expected, terminally differentiated cells also lacked B-my6 expression. It may be significant that A-my6 is highly expressed in testis (Mettus et al., 1994; Trauth et al., 1994) and that abundant expression of this gene was apparent in the proliferating stem cells (Mettus et al., 1994). It may be suggested, therefore, that in this specialized organ, A-Myb takes over the putative general role of B-Myb in the proliferating stem cell population. Xenopus B-Myb (XB-Myb) is a maternal protein and is consequently also expressed during meiosis (Bouwmeester et al., 1994). XB-Myb is expressed throughout embryogenesis. In the adult frog, XB-Myb is most abundant in blood cells, with lower expression detectable in testis, intestine, and spleen

I I8

Mark K . Saville and Roger I . Watson

(Bouwmeester et al., 1992). This may reflect a particular role for XB-Myb in these organs or, alternatively, may be an indication of their relatively high indices of cellular proliferation.

C. Mechanism of B-my6 Transcriptional Regulation Analysis of B-myb transcription using the run-on assay, whereby transcripts initiated in vivo are extended and labeled in isolated nuclei, revealed that extinction of B-myb expression in quiescent fibroblasts and the subsequent induction on restimulation with serum are largely dependent on the relative transcription rates (Lam et al., 1992). Thus, B-myb was transcribed a t extremely low levels directly after release of quiescent fibroblasts into G , , whereas during S phase, this rate was about 10-fold greater (Lam et al., 1992). Fluctuations in c-myb transcription rates during hematopoietic cell differentiation have been ascribed to a transcription elongation rather than to an initiation mechanism (Bender et al., 1987; Watson, 1988). Run-on analyses, however, revealed no evidence for premature termination of B-myb transcripts, indicating that control was actually effected a t the level of transcription initiation (Lam and Watson, 1993). Isolation of the mouse and human B-myb prornotcrs has allowed a more detailed study of the mechanism of cell cycle-regulated expression of this gene (Lam and Watson, 1993; Lam et al., 1995). These promoters were found to confer appropriate cell cycle-regulated expression when linked t o a luciferase reporter in transfected cells; the promoter activity was negligible in quiescent transfected cells, was maintained at a low level immediately after serum stimulation, and was induced 5- to 10-fold at the G,/S transition. Both mouse and human promoters are extremely GC-rich and lack an obvious TATA box. Consistent with this class of promoter, transcription is initiated at multiple sites, spanning about 80 nucleotides in the mouse promoter (Lam and Watson, 1993). Deletion analyses indicate that promoter activity per se is dependent on sequences that lie upstream of the multiple transcription initiation sites. These upstream sequences are surprisingly poorly conserved between the mouse and human promoters, although short interspersed regions of homology are present (Lam and Watson, 1993; Lam et al., 1995). In contrast, the transcription initiation sites and the 5’ mRNA noncoding sequences are highly conserved. The transcription initation region contains a conserved E2F transcription factor-binding site, which is perfectly conserved in the mouse and human Bmyb promoters (Lam and Watson, 1993; Lam et al., 1995).Significantly, mutation of the E2F site was found to abolish cell cycle control of B-myh promoter activity, resulting in constitutively high levels of reporter activity when measured in transient transfection assays (Lam and Watson, 1993; Lam et

B-Myb: A Key Regulator of t h e Cell Cycle

1 I9

al., 1995). These findings further emphasized the importance of E2F in controlling the induction of gene expression at the G,/S transition, but they were paradoxical in respect to the prevailing view of E2F as a cell cycle-regulated transcriptional activator. The clear implication is that B-myb promoter activity is regulated through the E2F site by transcriptional repression in G, and G I phases and derepression a t the G,/S boundary. This mechanism is compatible with the observation that pRb actively represses transcription through E2F binding sites in an artificial promoter (Weintraub et al., 1992). Although B-myb was the first example of a cellular gene controlled in this manner, subsequent studies have revealed that a number of other cell cycle genes are also regulated by transcriptional repression in G,, and G I phases (reviewed in Zwicker and Miiller, 1997). A number of lines of evidence implicate the pRb-related proteins p107 and p130 as critical components of repressor complexes mediating E2F regulation of B-myb transcription. First, transcriptional repression correlates with the presence of characteristic E2F complexes in Go to early G , phase that contain one or other of these proteins (Lam et al., 1994; Bennett et al., 1996). In most quiescent and terminally differentiated cells, p107 expression is turned off, and p130/E2F complexes are the predominant species (Cobrinik et al., 1993; Lam et al., 1995). In certain cells, however, p107/E2F complexes are predominant under these conditions (Bennett et al., 1996; Richon and Venta-Perez, 1996).Second, derepression of B-myb transcription occurs concurrently with modification of the G,/G, complexes (Lam and Watson, 1993), evidenced by decreased mobility in electrophoretic mobility shift assay (EMSA).The altered mobility results from binding of cyclin E/Cdk2, and later cyclin NCdk2, to the pocket proteins in these complexes (Lees et al., 1992). Third, B-myb transcription is derepressed on coexpression of tumor virus oncoproteins, such as the adenovirus E l A and human papillomavirus (HPV) E7 proteins, which bind to the pocket region of the pRb family and disrupt their interactions with E2F (Lam and Watson, 1993; Lam et al., 1994). Significantly, the ability of HPV E7 mutant proteins to deregulate Bmyb transcription in NIH 3T3 cells correlates precisely with their ability to bind p l 0 7 and p l 3 0 rather than pRb (Lam et al., 1994). Derepression of B-myb transcription in late G I phase is presumably precipitated by the activity of G , cyclin-dependent kinases on the pRb protein family, but the precise mechanism is far from obvious. In part, the gaps in our understanding reflect the difficulty in performing unambiguous experiments in cell cycle regulation of transcription. For example, although there is evidence that ectopic expression of D-type cyclin kinase activity in quiescent fibroblasts induces B-myb transcription (results quoted in Johnson, 1995), it is unclear whether this is a direct effect of phosphorylation of p130 in the G, E2F complex or a downstream consequence of the cells being stimulated to enter the cell cycle. To resolve this issue, it is critical to understand

120

Mark K. Saville and Roger I . Watson

in detail how transcriptional repression is imposed. A significant step in this direction was provided by genomic footprinting experiments showing that the B-myh promoter E2F site is occupied in quiescent murine NIH 3T3 fibroblasts (Zwicker et al., 1996). After serum restimulation, the site was found to remain protected at 4 hours but became unoccupied at 8 hours, concurrently with induction of B-my6 transcription in late G , phase. The obvious interpretation of these results is that release of proteins binding the E2F site in Go to early G I phase, presumably an inhibitory p130/E2F complex, removes a block on B-my6 transcription. The footprinting data are somewhat paradoxical, however, given that E2F binding activity, in the guise of free E2F heterodimers uncomplexed to pRb family proteins, actually becomes more abundant in S phase. The results could be explained if p130/E2F o r p107/E2F complexes bound more tightly to the site than does free E2F, but in uitro binding assays actually revealed no significant difference in avidity (Zwicker et al., 1996). A possible resolution of this paradox is suggested by a more detailed mutagenesis study of the B-myh promoter (Bennett et al., 1996). This revealed that a site immediately downstream of the E2F site, designated the downstream repression site (DRS), was additionally required for promoter repression in Go phase. Significantly, the magnitude of derepression obtained with DRS mutations was similar to that seen previously with E2F site mutations. Moreover, the effect of mutating both sites was not additive, indicating that these sites act cooperatively to mediate repression (Bennett et al., 1996). DRS mutations affected neither in uitro binding of E2F to the adjacent site nor transactivation of the B-myh promoter by transfected E2F-1 (Bennett et al., 1996), suggesting that repression involves a functional interaction between the Go E2F complex and another factor binding to the DRS. Similarity has been noted between the bipartite B-myh transcriptional control region and conserved elements, designated the cell cycle-dependent element (CDE)and the cell cycle genes homology region (CHR), regulating cell cycle-dependent transcription of the cdc25C, cdc2, and cyclin A genes (Liu et al., 1996). Intriguingly, transcription of these latter genes is also repressed in Go phase and derepressed at the G,/S transition (Zwicker et al., 1995); however, unlike with B-my6, maximal transcription is not reached until later in the cell cycle (late S phase for cdc2 and cyclin A and G, phase for cdc2SC). The cdc25C CDE sequence is identical to the GC-rich core of the B-my6 E2F site, and the downstream CHR shows some sequence similarity with the DRS (Fig. 3). Strikingly, the relative spacing within the bipartite E2F/DRS and CDJYCHR elements is identical. Similarity between these control elements suggests that common factors repress B-myh and cdc25C transcription in quiescent cells, but mutagenesis studies show that the nucleotide requirements for repression through the E2F and CDE sites are distinct (Zwicker et al., 1995; Bennett et al., 1996). Further studies indicated that

121

B-Myb: A Key Regulator of t h e Cell Cycle

E2F

DRS

B-wb CDC25C CCNA

CDE

CHR

Fig. 3 Comparison between the bipartite regulatory elements (E2F and DRS) required for transcriptional repression of B-myb in Go to early G , phase, with those (CDE and CHR) that perform a similar role in the promoters of the CDCZSC, cyclin A ( C C N A ) ,and CDC2 genes.

the cdc25c CDE binds weakly to E2F (Zwicker et al., 1995) and that the Bmy6 DRS cannot functionally replace the cdc25C CHR (Liu et al., 1996). Although a distant relation between factors interacting respectively with the E2F/DRS and CDHCHR sites cannot be ruled out, evidence favors the possibility that these control mechanisms are actually distinct. Initial attempts to detect DRS-binding proteins in cell extracts made from quiescent fibroblasts were unsuccessful (Bennett et al., 1996). Using more sensitive EMSA conditions, we have seen a novel complex that appears to involve simultaneous interaction with both the B-my6 E2F and DRS sites (P. Farlie, S. Catchpole, and R. Watson, manuscript in preparation). Antibody supershift experiments suggest that the novel complex involves the typical Go p130/E2F complex, but the identity of the protein putatively binding the DRS is unknown. The presence of the novel complex in cell extracts correlates well with the times a t which B-my6 transcription is repressed because it persists for 8 hours after serum stimulation of quiescent cells but is absent from cells in S phase. Other workers have purified a protein from HeLa cells that interacts in the minor groove with the B-my6 DRS (Liu et al., 1996); however, there is no indication that this factor is cell cycle regulated. It remains to be determined whether either of these binding proteins is actually implicated in transcriptional repression, but the presence of additional interactions a t the DRS allows a possible resolution of the paradoxical finding that the B-my6 E2F site is occupied only in Go and early G, (Zwicker et al., 1996). It can be speculated that binding of a corepressor protein to the DRS stabilizes binding of the adjacent Go E2F complex (Fig. 4). Why this functional interaction is lost a t the G,/S transition remains a matter for further

I22

Mark K . Saville and Roger 1. Watson

Fig. 4 A model to explain the requirements for both the E2F and DRS elements in repression of B-myb transcription at G,, to early C , phase. At this stage of the cell cycle, the E2F site is occupied by well-characterized pl30/EZF (and in certain cells, p l 0 7 / E 2 F ) complexes. It is proposed that binding of the ELF complcxes is stabilized by simultaneous occupation of the adjacent DKS by an uncharacterized DRF factor, resulting in transcriptional repression. Phosphorylation of p130 and p107 in late G , to S phase releases it from EZF; the fate of the putative DRF is unknown. EZFlDP heterodimers can still interact transiently with the E2F site. Transcription of Hmyh is now derepressed.

investigation, but it would not be a major surprise if it were found that this event is controlled directly by the G I cyclins.

D. B-my6 Transcription as a Control Point in t h e Cell Cycle Transcription of B-myh is extremely responsive to positive and negative modulators of E2F activity, raising the possibility that this constitutes a sensitive point for control of the cell cycle. Thus, on one hand, ectopic expression of the HPV E7 protein or the E2F-1 transcription factor leads to overexpression of cellular B - m y h as well as its inappropriate transcription early in G , phase (Lam et al., 1994; DeCregori et al., 1995); on the other hand, induction of p53 tumor suppressor protein expression or treatment of susceptible cells with transforming growth factor+, (TGF-PI ) results in downregulation of B - m y h transcription (Lin et al., 1992; Satterwhite et al., 1994). HPV E7 and E2F-1 are able to induce DNA synthesis in quiescent fibroblasts

B-Myb: A Key Regulator of t h e Cell Cycle

123

(Sato et al., 1989; Johnson et al., 1993), whereas p53 and TGF-P, inhibit proliferation of many cell types (Mercer et al., 1990; Moses et al., 1990), presumably through their effects on activity or expression of the cyclin-dependent kinases. Although not uniquely sensitive, the strict concordance between B-myb expression and cell growth, allied with the fact that it is a member of a protooncogene family, are sufficient cause to question whether B-myb may be a key component of the intracellular signal transduction pathways that link periodic cyclin expression with coordinated gene expression during the cell cycle.

IV. MODIFICATION OF B-Myb PROTEIN IN THE CELL CYCLE A. Hyperphosphorylation of B-Myb in S Phase As expected from the regulation of B-myb mRNA abundance in the cell cycle, B-Myb protein levels are relatively low in serum-starved Swiss 3T3 fibroblasts, whereas serum stimulation results in a substantial increase in BMyb expression coincident with the onset of S phase (Robinson et al., 1996). A small fraction of the total B-Myb protein is apparently stable, persisting in serum-starved cells for some time despite downregulation of the mRNA to undetectable levels (Robinson et al., 1996). This protein has a n apparent molecular weight of 110 kd (Robinson et al., 1996), significantly greater than the 85 kd predicted. Aberrant gel electrophoretic mobility does not appear to be an indication of posttranslational modification because B-Myb made in a number of systems (i.e., by in vitro translation and in bacterial and baculovirus expression systems) shares this property. The increase in B-Myb expression at the onset of S phase is accompanied by the appearance of an additional, more slowly migrating form of B-Myb, running a t a n apparent molecular weight of 112 kd (Robinson et al., 1996).Treatment of the S phase 112-kd species with phosphatase reduces its apparent molecular weight to 110 kd, strongly suggesting that B-Myb is specifically modified by phosphorylation o n entry of cells into S phase. The presence of a phosphorylated B-Myb form with reduced electrophoretic mobility (hyperphosphorylated B-Myb) appears to be restricted entirely to the S phase of the cell cycle. For example, B-Myb was shown to be ostensibly unmodified in cells synchronized in M phase by nocodazole treatment, and after release from the drug block, hyperphosphorylated B-Myb did not appear until subsequent entry into S phase (Robinson et al., 1996). This behavior differs from that of c-Myb, which was found to be hyperphosphorylated in nocodazole-arrested chicken BK3A lymphoma cells

I24

Mark K. Saville a n d Roger 1. Watson

(Luscher and Eisenman, 1992); the hyperphosphorylated c-Myb form was then rapidly lost o n release into the cell cycle. Although this difference may apply to mature cells, B-Myb is hyperphosphorylated during the meiosis that accompanies oocyte maturation in Xenopus sp. oogenesis (Bouwmeester et al., 1994) and also at the blastula stage of embryogenesis. Hyperphosphorylated XB-Myb was not detected in gastrula-stage embryos o r in later stages, however, and it has been suggested that the loss of this modification may be related to the addition of G , and G, phases to the embryonic cell cycle at the mid-blastula transition (Bouwmeester et al., 1994). The kinase responsible for S-phase-specific hyperphosphorylation of BMyb has yet to be identified definitively, although its regulation in the cell cycle clearly implicates the Cdks. On current evidence, the cyclin A/Cdk2 and possibly cyclin E/Cdk2 complexes are the most likely candidates. Thus, the timing of B-Myb hyperphosphorylation in the cell cycle correlates well with the kinetics of cyclin E/Cdk2 and cyclin A/Cdk2 activation in late G , and S phase and S phase, respectively, but occurs later than would be expected if cyclin D-dependent kinases are involved (Hall and Peters, 1996). Coexpression of cyclin A/Cdk2 with B-Myb by baculovirus vector transfer into Sf9 insect cells resulted in a reduction in the electrophoretic mobility o f B-Myb, consistent with the hyperphosphorylated form seen naturally during S phase in mammalian cells; in contrast, coexpression of cyclin E/Cdk2 and cyclin Dl/Cdk4 in these cells had n o effect on B-Myb mobility (Robinson et al., 1996). This experiment demonstrated that B-Myb is a substrate for cyclin A/Cdk2, but it has not yet been established that the sites phosphorylated by this kinase in insect cells are the same as those modified in S phase. It is unwise, moreover, to exclude the possibility that B-Myb is a substrate for cyclin E/Cdk2 because, for example, the influence of activating (CAK) kinases is difficult to predict in the Sf9 cell background. Indeed, both cyclin A and cyclin E were able to induce a partial B-Myb mobility shift when overexpressed by transient transfection in exponentially growing Saos-2 osteosarcoma cells (Sala et al., 1997), although it is possible that modification was an indirect consequence of effects on the cell cycle because coexpression of these genes promoted cells into S phase (see section V.B). Similar studies using U-2 OS cells produced almost complete conversion of B-Myb to the electrophoretically less mobile form by cotransfection with cyclin A and Cdk2 (Lane et al., 1997), further strengthening the conclusion that this complex may be implicated in the appearance of hyperphosphorylated B-Myb during S phase. It is possible that B-Myb is phosphorylated at certain sites throughout the cell cycle; such modifications would not have been detected if they did not cause an obvious electrophoretic mobility change. In this respect, it is pertinent to note that c-Myb appears to be a substrate not only for C d k l (previously designated Cdc2), presumed to be responsible in part for its modifica-

B-Myb: A Key Regulator of t h e Cell Cycle

I25

tion in M phase (Luscher and Eisenman, 1992), but also for casein kinase I1 (CKII) and mitogen-activated protein (MAP) kinases. The functional consequences of c-Myb phosphorylation by these kinases is not entirely clear (for review, see Ness, 1996). Modification of c-Myb at serines 11 and 12 by CKII has been reported to decrease its DNA binding and transactivation activities (Luscher et al., 1990; Oelgeschlager et al., 1995), but analyses of c-Myb mutations at these sites have not consistently supported this notion (Dini and Lipsick, 1993).c-Myb is also phosphorylated by p42 MAP kinase in vitro a t sites in its C-terminal (Aziz et al., 1993), including serine 528, which is also phosphorylated in intact cells. Replacement of this serine with alanine results in enhanced transactivation by c-Myb (Aziz et al., 1995), suggesting that phosphorylation negatively regulates activity. The CKII and MAP kinase sites identified in c-Myb are not conserved in B-Myb, and there is no indication that these enzymes have similar activities on B-Myb.

B. Cyclin-Dependent Activation of B-Myb Transactivation Function As detailed previously, c-Myb phosphorylation events characterized have all been associated with negative regulation of transactivation function. It seemed unlikely that the S-phase-specific phosphorylation of B-Myb would inhibit function because it occurs precisely at that stage of the cell cycle at which expression of this protein is upregulated as a consequence of the E2Fdependent transcriptional mechanism. Indeed, recent studies strongly indicate that Cdk-dependent B-Myb phosphorylation markedly enhances transactivation function (Lane et al., 1997; Sala et al., 1997).Thus, in human U-2 0s osteosarcoma cells, the ability of B-Myb to stimulate transcription of a cotransfected reporter gene whose promoter is regulated by three copies of a mim-1 gene MBS is strongly enhanced by cotransfection with cyclin A (Lane et al., 1997). Evidence that Cdk2 activity is required or this effect was provided by the findings that a dominant negative Cdk2 blocked cyclin Amediated enhancement of B-Myb activity and that expression of Cdk2 in concert with cyclin A further increased B-Myb transactivation function to a significant extent. These incremental effects o n transcription correlated closely with the proportion of transfected B-Myb modified by hyperphosphorylation. Similar results were also obtained in another study using human Saos-2 osteosarcoma cells, in which cotransfection of B-my6 with cyclin A synergistically activated transcription through the HIV-1 LTR, SV40 early, and CDC2 gene promoters (Sala et al., 1997). Curiously, high-affinity MBS in these promoters was not required for transactivation by cyclin Astimulated B-Myb, suggesting either that B-Myb bound to other sites or acted through an indirect mechanism. B-Myb was found to have little inherent

I26

Mark K. Saville a n d Roger 1. Watson

transactivation activity in Saos-2 cells, and this was correlated with low endogenous cyclin A levels in this line (Sala et al., 1997). Furthermore, cyclin E also potentiated B-Myb activity, although generally this was less active than cyclin A (Sala et al., 1997). It could not be established unequivocally in the cotransfection experiments described previously that Cdk-mediated hyperphosphorylation directly enhances B-Myb function because it is possible that the critical target of this enzyme is another component of the transcriptional machinery (potentially a coactivator). An indication that the effect is direct, however, has come from experiments showing that B-Myb immunoprecipitated from cells coexpressing cyclin A was more active in promoting in vitro transcription from the HIV-1 LTR promoter than that selected from cells expressing B-Myb alone (Sala et al., 1997). It is most likely, therefore, that the hyperphosphorylated B-Myb has enhanced transactivation function, although it cannot be absolutely excluded that an activating protein was coimmunoprecipitated with B-Myb. We can only speculate on how Cdk-mediated B-Myb hyperphosphorylation could enhance transactivation function. There are 22 potential Cdk phosphorylation sites in murine B-Myb, with the minimum specificity serine/ threonine-proline (S/T-P) (Fig. 5); 1 7 of these sites are conserved in human B-Myb. There is no indication of the number or identity of the sites actually modified in vivo during S phase or in vitro by cyclin A/Cdk2. Most of the conserved Cdk sites are clustered in a region centered on the CR (amino acids 396 to 581 ).Because B-Myb transactivation function is enhanced by either hyperphosphorylation or removal of a C-terminal NRD (Lane et al., 1997), it is possible that hyperphosphorylation negates the inhibitory effects of the NRD. Experimental evidence does not favor this possibility, however. Thus, C-terminally truncated B-Myb can still be activated by cyclin NCdk2, albeit to a lesser extent than the wild-type protein (Lane et al., 1997). Moreover, although the major effect of removing the NRD is to enhance DNA-binding, hyperphos-phorylation does not appear to have a similar effect ( S . Lane and R. Watson, unpublished observations). Consistent with these findings, hyperphosphorylation was found to enhance B-Myb even when transactivation function was measured on promoters that have no MBS (Sala et al., 1997). We speculate, therefore, that the role of hyperphosphorylation is to enhance interactions of B-Myb with coactivating proteins. A likely target of these putative interactions is the B-Myb CR. In summary, the weight of evidence indicates that B-myb expression is controlled by the cyclins at two levels during the cell cycle. An increase in B-Myb protein expression occurs, presumably through enhanced transcription of the B-my6 gene as a consequence of modification of repressor EZF-pocket protein complexes by the G I cyclins. In addition, the ability of B-Myb to

I27

B-Myb: A Key Regulator of the Cell Cycle

Potential Cdk phosphorylation sites

B-Myb DNA-binding Acidic Domain

CR

NRD

Fig. 5 The position of potential Cdk-phosphorylation sites (minimum specificity serine/ threonine-proline (SR-P) in m o w c B-Myb. Sites conserved in human B-Myb are represented by closed lollipops. The location of the C-terminal negative regulatory domain (NRD) is indicated.

transactivate responsive genes may also be enhanced through its phosphorylation by cyclin MCdk2.

V. REQUIREMENT FOR B-Myb IN CELL PROLIFERATION

A. Effects of Ablating B-my6 Expression The participation of B-Myb in the cell cycle has been studied using two approaches: the ablation of cellular B-my6 expression by antisense RNA or DNA techniques, or its enhancement by ectopic B-my6 expression. An obligate requirement for B-Myb in cell cycling is suggested by the finding that introduction of B-my6 antisense oligodeoxyribonucleotides into human myeloid o r lymphoid hematopoietic cell lines inhibits their proliferation (Arsura et al., 1992). In control experiments, antisense c-myb oligonucleotids were found to have a similar effect, suggesting independent requirements for B-Myb and c-Myb in proliferation of these cells. Although differentiation of one of the myeloid cells lines studied, U937, results in downregulation of Bmy6 expression, antisense B-my6 treatment did not induce the expression of differentiation markers (Arsura et al., 1992), indicating that effects on proliferation are dissociated from at least the terminal stages of the differentiation program. Although these experiments gave n o indication of the stage of the cell cycle at which B-my6 antisense oligonucleotides block proliferation, a further study showed that incorporation of tritiated thymidine was severely reduced in a human glioblastoma cell line labeled immediately after microinjection with B-my6 antisense oligonucleotides (Lin et d.,1994).These findings indicate that this treatment had a n immediate inhibitory effect on entry of cells into S phase, suggesting that the proliferation block was imposed at the G,/S transition. A limitation of the antisense oligonucleotide experiments described is a

I28

Mark K. Saville and Roger I . Watson

lack of evidence that endogenous B-myb expression was actually ablated. Using a sensitive polymerase chain reaction technique, B-myb mRNA levels were found to be unaffected in B-myb antisense-treated U937 cells after 24 hours, although similar treatment with c-myb antisense oligonucleotides caused severe reduction in c-myb expression (Arsura et al., 1992). This discrepancy may indicate that B-my6 mRNA is stable and that the primary effect of this treatment is inhibition of translation. Further studies are required to resolve this issue. Caution in the interpretation of antisense olgonucleotide experiments has also been precipitated by the results of some recent studies using c-myb and c-myc antisense oligonucleotides that revealed inhibitory effects related more to the pharmacologic properties of these agents than t o their specific antisense effects (Burgess et al., 1995; Chavany et al., 1995). In mitigation, the B-myb antisense oligonucleotides employed (Arsura et al., 1992; Lin et al., 1994) did not contain the two features implicated in spurious nonantisense effects: a dG quartet and phosphorothioate backbone. An essential role for B-Myb in cell proliferation is also supported by experiments in which B-myb antisense expression plasmids were transfected into mouse BALB/c 3T3 fibroblasts and human neuroblastoma cells (Sala and Calabretta, 1992; Raschella et al., 1995). Expression of antisense B-myb transcripts in these cells was found to reduce cloning efficiency dramatically on selection for a cotransfected drug-resistance marker. This presumably reflects inhibition of proliferation in the transfected cells, although other effects, such as an increase in apoptosis, cannot be excluded. Some caution should be exercised in the interpretation of these results because formation of double-stranded B-myb RNA may inhibit cellular proliferation through other mechanisms, perhaps involving interferon induction. It can be concluded that antisense experiments support the concept that Brnyb expression is required for cellular proliferation, but this case has yet to be made watertight. Further progress in this area will be made when the results are known of B-myb gene knockout experiments in progress in a number of laboratories. In the absence of these genetic data, it is pertinent to note that the unique Drosophila sp. myb gene, presumed to represent a progenitor of all three vertebrate rnyb genes, provides an essential function during embryonic and imaginal development (Katzen and Bishop, 1996). Isolation of recessive mutant alleles in Drosophila sp. my6, which fortuitously proved to be temperature sensitive, revealed multiple phenotypic defects in homozygotes at the restrictive temperature. This finding, together with the fact that my6 is expressed widely in Drosophila sp. tissues, is consistent with a broad role for this function in cell proliferation. Assuming this ubiquitous function applies also to mammalian species, it is significant that c-myb mouse knockouts display defects in development of only the hematopoietic cell compartment. It may follow that B-Myb and A-Myb have essential roles in other cell lineages.

B-Myb: A Key Regulator of t h e Cell Cycle

129

B. Effects of B-my6 Overexpression The effects of ectopic expression of B-Myb in cell lines are also consistent with its putative role in promoting cell growth. In contrast to the parental cells, stably transfected BALBk 3T3 mouse fibroblast lines constitutively expressing B-myb are able to maintain extended growth in medium that contains low serum concentrations, albeit with a reduced doubling time compared with cells grown under normal serum conditions (Sala and Calabretta, 1992). Moreover, these cells adopt a partially transformed phenotype and are able to grow more efficiently in soft agar (Sala & Calabretta, 1992). Similar experiments in other mouse fibroblast lines (i.e., NIH 3T3 and Swiss 3T3 fibroblasts), however, failed to reveal any significant effects of B-myb overexpression on growth either in low serum conditions or in soft agar (Robinson, 1995; Robinson etal., 1996). Overexpression of B-myb in human T98G glioblastoma cells also was unable to prevent cell cycle arrest on serum deprivation (Sala et al., 1996a). The reasons for the disparity in responses of various cell lines to ectopic B-myb expression are unclear but may be related to the previously noted tissue tropism exhibited for B-Myb transactivation function (see section 1I.C) or other differences in the growth requirements of these cell lines. Additional evidence that B-myb overexpression can promote cell proliferation, at least in certain cells, was obtained by studying stable transfectants of the M 1 myeloid leukemia cell line (Bies et al., 1996). In the parent cells, interleukin-6 (IL-6) treatment induces terminal differentiation, leading to a gradual decrease in cell proliferation over a period of several days. In response to IL-6, c-myb mRNA levels are immediately downregulated, whereas B-myb mRNA abundance decreases much later, just before terminal differentiation and growth arrest (Bies et al., 1996). Constitutive expression of B-myb was found to prevent IL-6-induced cell cycle arrest completely in M , cells. Although the IL-6-treated Ml/B-Myb cells appeared immature, several differentiation markers were induced to near-normal levels (Bies et al., 1996), suggesting that B-myb overexpression is unable to prevent commitment to maturation but blocks the terminal differentiation step that is associated with cessation of cell proliferation. Apoptosis that accompanies IL-6stimulated terminal differentiation of M 1 cells is also blocked by B-myb overexpression; however, a broad role for B-Myb in abrogating cell death cannot be assumed because apoptosis is actually accelerated in MUB-Myb cells in response to TGF-P, (Bies and Wolff, 1995). Further evidence that B-Myb has a regulatory role in the cell cycle was provided by experiments showing that its overexpression affects the cell cycle distribution of certain asynchronously growing lines. This effect was most evident in human T98G glioblastoma cells, in which transient transfection with B-myb significantly increased the proportion of cells in the S phase of

I30

Mark K. Saville and Roger 1. Watson

the cell cycle (Sala et al., 1996a). In the other lines tested, human Saos-2 and U-2 0s osteosarcoma cells, transient transfection with B-my6 alone had little effect on cell cycle distribution. In these lines, nonetheless, B-my6 was found to synergize with cyclin A to promote cells into S phase (Lane et al., 1997; Sala et al., 1997), suggesting that in Saos-2 and U-2 0s cells, B-Myb needs to be activated by phosphorylation to drive proliferation. The ability of transfected B-Myb alone to promote S phase in T98G cells may be explained by the finding that they express much higher levels of endogenous cyclin A than d o Saos-2 cells (Sala et al., 1996a). Although it is most probable that the requirement for synergistic action of B-Myb and cyclin A t o promote S phase reflects the enhancement of B-Myb transactivation function by hyperphosphorylation (see section IV.B), the possibility that these proteins are in fact acting on complementary pathways cannot be discounted. The mechanism by which B-my6 overexpression promotes S phase in the responsive cell lines is not yet understood. In principal, it could d o so by shortening the cell cycle, but the time taken for B-my6-transfected T98G cells to enter S phase after release from G, (about 16 hours) was identical to that taken by control cells (Sala et al., 1996a), clearly suggesting that ectopic BMyb did not truncate G I . The major difference observed between B-mybtransfected T98G and control cells is that a significantly greater proportion of the former were in S phase at both the 20-hours and 24-hour time points after serum stimulation. There are a number of explanations for this phenomenon. First, the proportion of cells that fail to exit Go on serum stimulation may be decreased, implying that B-my6 overexpression reduces arrest arising from contact inhibition. Second, the B-my6-transfected cells may enter S phase more synchronously, perhaps because ectopic B-Myb increases the probability of initiation of DNA replication once a minimum G , period has been served. Third, the B-my6-transfected cells may actually be arrested in S phase and therefore accumulate at this stage of the cell cycle; this arrest could not be permanent, however, because the B-my6-transfected T98G cells eventually return to G , phase (Sala et al., 1996a). These various possibilities are amenable to experimental analysis, and further study in this area will be critical to understanding the role of B-Myb in cell cycle control.

C. Suppression of Cell Proliferation Blocks by B-Myb Growth-promoting signals emanating from Cdk activity are subject to counterbalance by inhibitory factors that lead to cell cycle arrest a t distinct checkpoints. An important group of negative regulatory proteins are the Cdk inhibitors, which bind directly to either the cyclin/Cdk complex o r Cdk itself, resulting in inhibition of cell proliferation in response to signals such as DNA damage or cell-cell contact (reviewed in Hall and Peters, 1996). For

B-Myb: A K e y Regulator of the Cell Cycle

131

example, ~ 2 1 ~ ;(abbreviated ~ ~ " as p21) is transcriptionally regulated by the pS3 tumor suppressor protein and may account for the ability of p53 to inhibit cell proliferation in response to DNA damage. Similarly, p27K1P'accumulates when cells are arrested by TGF-P, or by contact inhibition. An important target of the Cdk inhibitor-mediated inhibitory effect is the pRb family of proteins, which remain or return to a hypophosphorylated state when Cdk activity is precluded. This results in accumulation of pocket protein-E2F complexes, which repress transcription of cell cycle-regulated genes, such as B-myb. Overexpression of B-my6 has been found to overcome some of the cell proliferation blocks outlined previously, perhaps accounting for its potential to promote cells into S phase (see section V.B). In the first report on this property, B-myb overexpression was shown to bypass p53-induced G, phase arrest of T98G and Saos-2 cells (Lin et al., 1994). This activity required an intact B-Myb DNA-binding domain and was partially dependent on the acidic domain, implying that enhancement of B-Myb target gene expression overcomes the proliferation block. I t is intriguing that B-myb-overexpressing cells were able to enter S phase despite the accumulation of high levels of p21 in response to p53 induction, particularly because this was associated with inhibition of cyclin E-(and presumably cyclin A ) dependent Cdk activity (Lin et al., 1994). It can be assumed that Cdk-mediated phosphorylation of the ectopically produced B-Myb would be inhibited in the presence of high levels of p21 C D K N 1 , and these results are therefore not easily reconcilable with later reports that cyclin A synergizes with B-Myb to promote S phase in Saos2 cells (Sala et al., 1997). Further studies are required to determine the effects of pS3 induction on B-Myb modification and transactivation function. B-myb overexpression has also been found to overcome partially a pl07mediated proliferation block of Saos-2 cells but was completely ineffective in response to pRb-mediated growth arrest (Sala et al., 1996a). The p107 protein contains two domains that are independently able to arrest cell growth ( Z h u et al., 1995). One domain contains a region of homology with p21 and is required for interaction with and inhibition of cyclin N C d k 2 and cyclin E/Cdk2 complexes, whereas the other binds and inhibits members of the E2F family of transcription factors. In contrast to p107, pRb is not able to form stable complexes with cyclin N C d k , and its ability to arrest growth correlates only with binding to E2F (Hiebert et al., 1992). It is therefore possible that B-Myb overcomes pl07-mediated growth arrest imposed through the Cdk-binding domain but not the E2F-binding domain, and this may explain why the effect on cell proliferation is only partial (Sala et a!., 1996a). This hypothesis is consistent with the ability of B-Myb to bypass pS3-mediated growth arrest. It will be of further interest to determine whether B-Myb can directly overcome proliferation blocks imposed by the Cdk inhibitors. The ability of B-Myb to overcome proliferative blocks could reflect its role

132

Mark K. Saville and Roger 1. Watson

in transcriptional activation of genes required for cell cycle progression that act downstream of the p107 and p53/Ckd inhibitor-regulated G I checkpoint. This could imply that B-my6 is a critical target for these negative regulatory proteins and that its downregulation is necessary for the checkpoint to operate. Consistent with this notion, induction of p53 in T98G cells results in downregulation of B-my6 mRNA (Lin et al,. 1992). Moreover, ectopic p107 expression represses B-my6 promoter activity when measured in transient transfection assays (Sala et al., 1996a). Data indicating that B-Myb and p107 can directly interact (Sala et af., 1996b), a t least when p107 is overexpressed, suggest another explanation for the capacity of B-Myb to counter cell proliferation blocks. Interaction with B-Myb could prevent p107 from associating with and inhibiting other target proteins, such as E2F/DP heterodimers, that positively regulate cell cycle progression. This interaction could also account for the ability of B-Myb to overcome a p53 cell cycle block because the critical downstream target of p53 may actually be p107. Thus, it could be predicted that p53 overexpression would induce p21, which in turn would maintain p107 in the active hypophosphorylated state by inhibiting cyclin/Cdk activity. Indeed, p107 may be a particularly important target for p53 in Saos-2 cells because they lack functional pRb. The cell cycle block imposed by p107 (and perhaps p53 acting through inhibition of p107 phosphorylation) may be mediated in part by directly blocking B-Myb transactivation function. In support of this model, cotransfected p107, but not pRB, was found to inhibit B-Myb transactivation function measured on a human B-my6 promoter reporter (Sala etaf., 1996b) This activity is dependent on the p107 pocket region and does not require the cyclin/Cdk-binding domain. This binding requirement is therefore reminiscent of the pl07/c-Myc interaction (Beijersbergen et af., 1994), rather than the plO7/Spl interaction, which does not require an intact pocket region (Datta et af., 1995). It may be further hypothesized that B-myb overexpression overcomes a p107 cell proliferation block because it titrates out inhibition of endogenous B-Myb activity by pl07-binding. Further evidence needs to be obtained of an in vivo complex between endogenous p107 and B-Myb proteins for this hypothesis to gain credence. It also remains to be determined whether p107 can inhibit transactivation of other cellular genes by B-Myb.

VI. B-MYB FUNCTION A. Mechanisms of Transcriptional Control by B-Myb Transcriptional control by B-Myb is not straightforward. By reference to c-Myb activity, it was anticipated that B-Myb would activate expression

B-Myb: A K e y R e g u l a t o r of t h e Cell Cycle

I33

through binding to MBS in the transcriptional control regions of target genes. In fact, most studies have demonstrated that B-Myb transactivation is not dependent on MBS (Foos et af., 1993; Nakagoshi et al., 1993; Watson et af., 1993; Sala et af., 1997), although MBS-specific transactivation of a herpes simplex virus thymidine kinase promoter has been observed in certain cells (Lane et al., 1997). Additionally, other studies have indicated that B-Myb is a repressor of gene transcription, for example, by competitively inhibiting c-Myb transactivation of MBS-regulated promoters (Foos et al., 1992; Watson et al., 1993) and by actively repressing type I collagen gene promoters in smooth muscle vascular cells (Marhamati and Sonenshein, 1996). In the latter study, repression of a synthetic promoter was shown to be MBS dependent; however, whether repression of type I collagen gene promoters was mediated by specific DNA-binding to this consensus site was not clear. A number of explanations for MBS-independent transactivation of gene expression by B-Myb come to mind. First, B-Myb could interact directly with an alternative binding site; however, although initial studies indicated that bacterially expressed B-Myb, but not c-Myb, additionally binds to the sequence AGAAANYG (Mizuguchi et al., 1990), other groups have been unable to confirm these results using other sources of B-Myb protein (Howe and Watson, 1991; Bouwmeester et al.,1994). Second, B-Myb could interact with promoter sequences indirectly by bridging through another transcription factor. As an example of this mechanism, E2F-1 can transactivate a promoter by “piggy-backing’’ through Spl bound to its cognate DNA recognition site (Lin et a/., 1996). The B-Myb DNA-binding domain is still required for MBS-independent transactivation (Nakagoshi et al., 1993);although this may suggest that direct DNA binding is required for this activity, a hydrophobic surface in this region has been hypothesized to serve as a protein interaction domain (Ogata et af., 1995; Ness, 1996). Third, B-Myb may interact directly with the basal transcription complex, although this is unlikely because B-Myb shows specificity for the promoters that it activates. Fourth, B-Myb may compete out a specific negative regulator of promoter activity; for example, interaction of B-Myb with p107 may prevent this pocket protein from inhibiting other activating transcription factors. Much additional work needs to be done to identify the precise mechanisms by which B-Myb regulates transcription. Another important question to address is, Which B-Myb activities are necessary for it to function as a cell cycle regulator?

B. Possible B-Myb Target Genes B-Myb overexpression has been reported to activate transcription from the promoters of several cellular genes in transient transfection assays. These in-

I34

Mark K . Saville and Roger 1, Watson

clude c-myc (Nakagoshi et al., 1992), B-myb itself (Sala et al., 1996b), DNA polymerase-cx (Watson et al., 1993), HSP70 (Foos et al., 1993), and cdc2 (Sala et a/., 1997). In each example, there is n o conclusive evidence that activation is mediated through MBS, and it has been demonstrated that activation of the latter two promoters does not require this binding site. Although it is apparent that some of these genes are intimately involved in cell cycle control, definitive evidence that their deregulation accounts for the positive action of B-Myb o n cell proliferation requires additional experimental approaches. In this respect, it is relevant that B-myb overexpression in Balb/c 3T3 fibroblasts is associated with an increase in the abundance of cyclin D 1 and cdc2 mRNAs (Sala and Calabretta, 1992); it is still uncertain, however, whether this is the direct result of B-Myb transactivation o r a downstream consequence of deregulation of other cell cycle-control targets. Moreover, there is little indication that B-Myb regulates expression of these genes in normal cells because cyclin D1 mRNA levels d o not increase in S phase (Matsushime et al., 1991), the period of maximal B-Myb activity, and cdc2 appears to be cell cycle regulated by E2F o r CDE/CHR-driven mechanisms (Dalton, 1992; Zwicker et af., 1995). Functionally important MBS have been identified in the transcriptional control regions of many other genes (reviewed in Ness, 1996). Potential binding sites in published gene sequences have also been catalogued based on systematic binding affinity measurements (Deng et al., 1996). Such sites could bind all three known Myb proteins, and the specific role of B-Myb in regulating transcription of genes in which these sites are found is yet unknown. It is entirely possible that different Myb proteins may have opposite effects on gene expression, and it will be important to develop systems in which to characterize the particular contributions of the individual members. Studies of human c-myb expression indicate that two closely linked MBS located 5’ o f this gene mediate negative regulation of transcription in T-cell lines; although in myeloid lines, these sites are nonfunctional (Guerra et al., 1995). In T-cell nuclear extracts, both B-Myb and c-Myb proteins were identified that interacted in in vitro binding assays with these 5’ c-my6 sites. These findings have certain similarities with the observation that B-Myb downregulates promoter activity of type I collagen genes (Marhamati and Sonenshein, 1996). These studies serve to emphasize that B-Myb has the potential to act as both an activator and a repressor of transcription.

C. Conclusions B-my6 is subject to two putative cyclin-dependent regulatory mechanisms during the cell cycle: upregulation of mRNA abundance by derepression of E2F/DRS-regulated transcription at the G,/S boundary, and phosphorylation of the protein during S phase. As illustrated in Figure 6, this results in high

B-myb transcription repressed

p b

B-mybtranscription derepressed

-=

Fig. 6 Scheme showing the involvement of the cyclidcdks in regulation of B-Myb expression and function. Transcriptional repression of B-myb transcription by p130/E2F or p 1 0 7 E 2 F complexes in Go to early G , phase is ablated by phosphorylation of the pocket proteins by cyclin D- and possibly cyclin E-dependent Cdks. The B-Myb protein synthesized as a result of B-myb mRNA induction is a putative substrate for cyclin A/Cdk2 and possibly cyclin WCdk2. This hyperactivates B-Myb function, which is required for S-phase progression.

I36

Mark K. Saville and Roger 1. Watson

levels of hyperactivated protein, specifically in S phase. There are several interesting features of this dual mechanism. First, it ensures that in quiescent cells, there is effectively no active B-Myb because transcription is greatly downregulated and the little protein that may be produced is not activated by Cdk-mediated phosphorylation. Second, it allows for a huge increase in active B-Myb as cells go from the G , phase into S phase. Third, in cycling cells, the B-Myb produced in S phase may be partially inactivated by dephosphorylation at later stages and only hyperactivated again on entry into the subsequent S phase; this may be relevant in restricting B-Myb function to S phase because both the B-myb mRNA and protein are fairly stable (Marhamati & Sonenshein, 1996; Raschella etal., 1996; Robinson et al., 1996). Strict regulation of active B-Myb expression may be required because its inappropriate function, as seen in the overexpression experiments (section V.B), may deregulate cell cycle controls. Regulation at the transcriptional level also has benefits to the economy of the cell because it prevents wasteful B-my6 transcription during periods of quiescence when B-Myb presumably has no function. The precise role of B-myb in cell cycle regulation remains unknown. As a transcriptional activator, we may envisage an essential function for B-myh in regulation of other genes whose products are involved in synthesis of the components of nucleotide biosynthesis and DNA replication. Equally, as a transcriptional repressor, we may envisage a role for B-my6 in extinguishing expression of negative regulators of cell proliferation. To gain further insight into this, it will be necessary to devise systems to study more critically alterations in cellular gene transcription elicited by ectopic B-myb expression. As a starting point, it may be useful to determine whether potential Myb target genes, identified through a database search for putative MBS (Deng et al., 1996),are transcriptionally deregulated under such conditions. However, BMyb has the potential to activate transcription from promoters that lack MBS, and these genes cannot be predicted from database searches. By mutagenesis of the DNA-binding domain, it should be possible to determine whether interaction of B-Myb with its cognate-binding site is necessary for it to operate in cell cycle control. The realization that B-Myb is subject to controls on both its abundance and activity, that this directs maximal activity to S phase, and that B-Myb function appears to be necessary for cell proliferation lead us to conclude that it plays a central role downstream of the cyclins in controlling the basic mechanisms that coordinate passage through the cell cycle.

ACKNOWLEDGMENTS We are grateful to Paul Farrell for critically reading the manuscript and to other members of the laboratory for their unpublished contributions to o u r research quoted here. Mark Saville is supported by a Cancer Research Campaign grant.

8-Myb: A Key Regulator of the Cell Cycle

137

REFERENCES Arsura, M., Introna, M., Passerini, F., Mantovani, A., and Golay, J. (1992). Blood, 79, 2708-2716. Aziz, N., Miglareses, M. R., Hendrichson, R. C., Shabanowitz, J., Sturgill, T. W., Hunt, D. F., and Bender, T. P. (1995).Proc. Natl. Acad. Sci. U.S.A., 92,6429-6433. Aziz, N., Wu, J., Dubendorff, J. W., Lipsick, J. S., Sturgill, T. W., and Bender, T. P. (1993).Oncogene, 8,2259-2265. Baldin, V., Lukas, J., Marcote, M. J., Pagano, M., and Draetta, G . (1993). Genes Dev., 7, 8 12-82 1. Beijersbergen, R. L., Hijmans, E. M., Zhu, L., and Bernards, R. (1994). EMBO I., 13, 4080-4086. Bender, T. P., Thompson, C. B., and Kuehl, W. M. (1987).Science, 237, 1473-1476. Bennett, J. D., Farlie, P. G., and Watson, R. J. (1996).Oncogene, 13,1073-1082. Bies, J., Hoffman, B., Amanullah, A., Giese, T., and Wolff, L. (1996).Oncogene, 12, 355-363. Bies, J., and Wolff, L. (1995).Cancer Res., 55, 501-504. Blake, M. C., and Azizkhan, J. C. (1989).Mol. Cell. Biol., 9, 4994-5002. Bouwmeester, T., Guehmann, S., el-Baradi, T., Kalkbrenner, F., van-Wijk, I., Moelling, K., and Pieler, T. ( 1 992). Mech. Dev., 37, 57-68. Bouwmeester, T., van-Wijk, I., Wedlich, D., and Pieler, T. (1994). Oncogene, 9, 1029-1038. Brown, K. E., Kindy, M. S., and Sonenshein, G . E. (1992).]. Biol. Chem., 267,4625-3460. Burgess, T., Fisher, E., Ross, S., Bready, J., Qian, Y., Bayewitch, L., Cohen, A., Herrera, C., Hu, S., Kramer, T., Lott, F., Martin, F., Pierce, C., Simonet, L., and Farrell, C. (1995).Proc. Natl. Acad. Sci. U.S.A., 92,4051-4055. Carr, M. D., Wollborn, U., Mclntosh, P. B., Frenkiel, T. A., McCormick, J. E., Bauer, C. J., Klempnauer, K. H., and Feeney, J. (1996).Eur.]. Biochem., 235,721-735. Catron, K. M., Purkerson, J. M., Isakson, P. C., and Bender, T. P. (1992).1. lmmunol., 148, 934-942. Chavany, C., Connell, Y.,and Neckers, L. (1995). Mol. Pharmacol., 48, 738-746. Cobrinik, D., Whyte, P., Peeper, D. S., Jacks, T., and Weinberg, R. A. (1993). Genes Dev., 7, 2392-2404. Dalton, S. (1992).EMBO]., 11, 1797-1804. Dash, A. B., Orrico, F. C., and Ness, S. A. ( 1 996). Genes Dev., 10, 1858-1 869. Datta, P. K., Raychauduri, P.,and Bagchi, S. (1995).Mol. Cell. Biol., 15,5444-5452. DeGregori, J., Kowalik, T., and Nevins, J . R. (1995).Mol. Cell. Biol., 15,4215-4224. Deng, Q.-L., Ishii, S., and Sarai, A. ( I 996). Nucl. Acids Res., 24, 766-774. Dini, P. W., and Lipsick, J. S. (1993).Mol. Cell. Biol., 13, 7334-7348. Dubendorff, J. W., Whittaker, I.. J., Eltman, J. T., and Lipsick, J. S. (1992). Genes Dev., 6, 2524-2535. Foos, G., Grimm, S., and Klempnauer, K. H. (1992).EMBO I., 11,4619-4629. Foos, G., Natour, S., and Klempnauer, K. H. (1993). Oncogene, 8, 1775-1782. Girard, F., Strausfeld, U., Fernandez, A., and Lamb, N. J. C. (1991).Cell, 67, 1169-1179. Golay, J., Capucci, A., Arsura, M., Castellano, M., Rizzo, V., and Introna, M. (1991). Blood, 77,149-158. Golay, J., Cusmano, G., and Introna, M. (1992).]. lmmunol., 149, 300-308. Golay, J., Luppi, M., Songia, S., Palvarini, C., Lombardi, L., Aiello, A., Delia, D., Lam, K., Crawford, D. H., Biondi, A., Barbui, T., Rambaldi, A., and Introna, M. (1996).Blood, 87, 1900-1911. Guerra, J., Withers, D. A., and Boxer, L. M. ( 1 995). Blood, 86, 1873-1 880. Hall, M., and Peters, G. (1996). Advances Cancer Res., 68, 67-108. Hiebert, S. W., Chellappan, S. P., Horowitz, J. M., and Nevins, J. R. (1992). Genes Dev., 6, 177-1 85.

I38

Mark K . Saville and Roger I. Watson

Howe, K. M., Reakes, C . F., and Watson, R. J. (1990). E M B O I. 9,, 161-169. Howe, K. M., and Watson, R.J. (1991). Nuclric Acids Res., 19, 3913-3919. Hsiao, K.-M., McMahon, S. L., and Farnhani, P. J. (1994). Genes Dev., 8, 3526-1537. Hu, Y. L., Ramsay, R. G., Kanei, I. C., Ishii, S., and Gonda, T. J. (1991). Oncogene, 6, 1549-1 553. Johnson, D. G. (1995). Oncogene, 11, 1685-1692. Johnson, D. G., Ohtani, K., and Nevins, J . R. (1994). Genes Dev., 8, 1514-1525. Johnson, D. G., Schwarz, J. K., Cress, W. I)., and Nevins, J. R. (1993). Nature, 365, 349-352. Kalkhrenner, F., Guehmann, S., and Moelling, K. (1990). Oizcogene, 5, 657-661. Kamano, H., Burk, B., Noben-Trauth, K., and Klempnauer, K. H. (1995). Oncogene, 11, 2.575-2582. Kanei-lshii, C., MacMillan, E. M., Nomura, T., Sarai, A., Ramsay, R. G., Aimoto, S., Ishii, S.. and Gonda, T. J. (1992).Proc. Natl. Acad. Sci. U.S.A., 89, 3088-3092. Katzen, A. L., and Bishop, J. M. (1996). Proc. Natl. Acad. Sci. U.S.A., 93, 13955-13960. Klempnauer, K.-H., Ramsay, G., Bishop, J. M., Moscovici, M. G., Moscovici, C., McGrath, J. P., and Levinson, A. D. (1983). Proc. Natl. Acad. Sci. U.S.A., 33, 345-355. La Thangue, N. B. (1994). Trends Biochem. Sci., 19, 108-1 14. Lam, E. W., Bennett, J. D., and Watson, R. J. (1995). Gene, 160,277-281. Lam, E. W., Morris, J. I)., Davies, R. Crook, T., Watson, R. J., and Vousden, K. H., (1994). E M B O J., 13, 871-878. Lam, E. W., Robinson, C:., and Watson, R. J. (1992).Oncogene, 7. 1885-1890. Lam, E. W., and Watson, R. 1. (1993). EMBO J., 12, 2705-2713. Lam, E. W.-F., and La Thangue, N. B. ( 1 994). Curr. Opin. Cell Biol., 6, 859-866. Lane, S., Farlie, P., and Watson, R. ( 1 997). Oncogene, 14, 2445-2453. Lees, E., Faha, B., Dulic, V., Reed, S. I., and Harlow, E. (1992). Genes Dev., 6, 1874-1885. Lin, D., Fiscella, M., O’Connor, P. M., Jackman, J., Chen, M., Luo, L. L., Sala, A,, Travali, S., Appella, E., and Mercer, W. E. (1994). Proc. Natl. Acad. Sci. U.S.A., 91, 10079-10083. Lin, D., Shields, M. T., Ullrich, S. J., Appella, E., and Mercer, W. E. ( I 992). Proc. Natl. Acad. Sci. U.S.A., 89, 9210-92 14. Lin, S.-Y., Black, A. R., Kostic, D., Pajovic, S . , Hoover, C;. N., and Azizkhan, J. C. (1 996). Mol. Cell. Biol., 16, 1668-1675. Liu, N., Lucibello, F. C., Zwicker, J., Engeland, K., and Miiller, R. (1996). Nucl. Acids. Res., 24,2905-2910. Luscher, B., Christenson, E., Litchfield, D. W., Krehs, E. G., and Eisenman, R . N. ( 1 990). N a ture. 344, 5 17-522. Luscher, B., and Eisenman, K. N. ( 1 990). Genes Dev., 4,223.5-2241. Luscher, B., and Eiseniaan, R. N. (1992).J. Cell Biol., 118, 775-784. Lyon, J., Robinson, C., and Watson, R. (1994).Crit. Rev. Oncog., S, 373-388. Marhamati, D. J., and Sonenshein, G. E. (1996).J . B i d . Chem., 271,3359-3365. Marsushime, H., Roussel, M. F., Ashmun, R . A., and Sherr, C. J. (1991). Cell, 6S, 701-713. Mercer, W. E., Amin, M., Sauve, G. J., Appella, E., Ullrich, S. J., and Romano, J. W. (1990). Oncogene, 5,973-980. Mettus, R. V., Litvin, J., Wali, A., Toscani, A., Latham, K., Hatton, K., and Rcddy, E. P. ( 1 994). Oncogene, 9,3077-3086. Mizuguchi, G., Nakagoshi, H., Nagase, T., Nomura, N., Date, T., Ueno, Y., and Ishii, S. (1990). J . B i d . Chem., 265,9280-9284. Moses, H. L., Yang, E. Y., and Pictenpo1,J. A. (1990). Cell, 63, 245-247. Miiller, R. (1995).Trends Genet., 11, 173-178. Nakagoshi, H., Kanei-lshii, C., Sawazaki, T., Mizuguchi, G., and Ishii, S. ( 1 992). Oncogene, 7, 1233-1 240. Nakagoshi, H., Takemoto, Y., and Ishii, S. (1993).]. Biol. Chem., 268, 14161-14167.

B-Myb: A Key Regulator of t h e Cell Cycle

139

Ness, S. A. (1996).Biophysica Actu, 1288, F123-Fl39. Neumann, E., Flemington, E. K., Sellers, W. R., and Kaelin, W. G. ( 1 994). Mol. Cell. B i d , 14, 6607-66 15. Nishina, Y., Nakagoshi, H., Imamoto, F., Gonda, T. J., and Ishii, S. (1989).Nucleic Acids Res., 17, 107-117. Nomura, N., Takahashi, M., Matsui, M., Ishii, S., Date, T., Sasamoto, S., and Ishizaki, R. ( I 988). Nucleic Acids Res., 16, 11075-1 1089. Nomura, T., Sakai, N., Sarai, A., Sudo, T., Kanei-lshii, C., Ramsay, R., Favier, D., Gonda, T. J., and Ishii, S. (1993).]. Biol. Chem., 268, 21914-21923. Oehler, T., Arnold, H., Biedenkapp, H., and Klempnauer, K . H. (1990).Nucleic Acids Res., 18, 1703-1710. Oelgeschlager, M., Krieg, J., Liischer-Firzlaff, J.-M., and Luscher, B. ( 1 995). Mol. Cell. Biol., 15,5966-5974. Ogata, K., Kanei-Ishii, C., Sasaki, M., Hatanaka, H., Nagadoi, A., Enari, M., Nakamura, H., Nishimura, Y., Ishii, S., and Sarai, A. ( 1 996). Nature Struct. Biol., 3, 178-1 87. Ogata, K., Morikawa, S., Nakamura, H., Hojo, H., Yoshimura, S., Zhang, R., Aimoto, S., Ametani, Y.,Hirata, Z., Sarai, A., Ishii, S., and Nishimura, Y. (1995).Nature Struct. Biol., 2,309-320. Ogris, E., Rothender, H., Mudrak, I., Pichler, A., and Wintersberger, E. (1993)./. Virol., 67, 1765-1771. Ohtani, K., DeGregori, J., and Nevins, J. R. (1995). Proc. Natl. Acad. Sci. U.S.A., 92, 12146-1 2150. Olson, J. E., Winston, J. T., Whitlock, J. A., and Pledger, W. J. (1993).J. Cell Physiol., 154, 333-342. Pagano, M., Pepperkok, R., Verde, F., Ansorge, W., and Draetta, C. (1992). E M B O I. 11, , 96 1-97 1 . Pardee, A. B. (1989).Science, 246, 603-608. Pearson, B. E., Nasheuer, H.-P., and Wang, T. S.-F. (1991). Mol. Cell. Biol., 11, 2081-2095. Quelle, D. E., Ashmun, R. A., Shurtleff, S. A., Kato, J., Bar-Sagi, D., Roussel, M. F., and Sherr, C. J, (1993). Genes Deu., 7, 1559-1571. Raschella, G., Negroni, A., Pucci, S., Amendola, R., Valeri, S., and Calabretta, B. (1996).Exp. Cell Res., 222, 395-399. Raschella, G., Negroni, A., Sala, A., Pucci, S., Romeo, A., and Calabretta, B. (1995)./. B d . Chem., 270, 8540-8545. Reiss, K., Travali, S., Calabretta, B., and Baserga, R. (1991)./. CeNPbysiol., 148, 338-34.3. Resnitzky, D., and Reed, S. 1. (1995).M o l Cell. Biol.,IS, 3463-3469. Richon, V. M., and Venta-Perez, G. ( I 996). Cell Growth Differ, 7 , 3 1 4 2 . Robinson, C . (1995). PhD thesis, University of London. Robinson, C., Light, Y., Groves, R., Mann, D., Marias, R., and Watson, R. (1996).Oncogene, 12,1855-1 864. Sakura, H., Kanei, I. C., Nagase, T., Nakagoshi, H., Gonda, T. J., and Ishii, S. (1989).Proc. Natl. Acad. Sci. U.S.A., 86,5758-5762. Sala, A., and Calabretta, B. (1992).Proc. Natl. Acad. Sci. U.S.A., 89, 10415-10419. Sala, A., Casella, I., Bellon, T., Calabretta, B., Watson, R. J,, and Peschle, C. (1996a)./. Biol. Chem., 271,9363-9367. Sala, A., De Luca, A., Giordano, A., and Peschle, C. (1996b). 1. Biol. Chem., 271, 28738-28740. Sala, A., Kundu, M., Casella, I., Engelhard, A., Calabretta, B., Grasso, L., Paggi, M. G., Giordano, A., Watson, R. J., Khalil, K., and Peschle, C. (1997). Proc. Natl. Acad. S c i U.S.A., 94, 532-536. !jato, H., Furuno, A., and Yoshike, K . (1989). Virology, 168, 195-199.

I40

Mark K . Saville and Roger 1. Watson

Satterwhite, D. J., Aakre, M. E., Gorska, A. E., and Moses, H. I.. ( 1 994). Cell Growth Differ, 5,789-799. Schulze, A., Zerfass, K., Spitovsky, D., Middendorp, S., Berges, J., Helin, K., Jansen-Durr, P., and Henglein, B. (1995).Proc. Natl. Acad. Sci. U.S.A.. 92, 11264-1 1268. Sitzmann, J., Nohen-Trauth, K., Kamano, H., and Klempnauer, K. H. (1996).Oncogene, 12, 1889-1894. Takemoto, Y., Tashiro, S., Handa, H., and Ishii, S. (1994).FEES Lett., 350, 5.5-60. Tashiro, S., Takemoto, Y., Handa, H., and Ishii, S. (1995).Oncogene, 10, 1699-1707. Thompson, C. B., Challoner, P. B., Neiman, P. E., and Groudine, M. (1985). Nature, 314, 363-367. Thompson, C. B., Challoner, P. B., Neiman, P. E., and Groudine, M. (1986). Natzrre, 319, 3 74-3 80. Trauth, K., Mutschler, B., Jenkins, N. A., Gilbert, D. J., Copeland, N. G., and Klempnauer, K. H. (1994). E M B O J . , 13,5994-6005. van den Heuvel, S., and Harlow, E. (1993).Science, 262,2050-2054. Watson, K. J. ( 1988). Oncogene, 2,267-272. Watson, R. J., Robinson, C., and Lam, E. W. (1993).Nucleic Acids Res., 21, 267-272. Weintraub, S. J., Prater, C. A., and Dean, D. C. (1992). Nature, 358,259-261. Westin, E. H., Callo, R. C., Arya, S. K., Eva, A., Souza, L. M., Baluda, M., Aaronson, 5. A., and Wong-Staal, F. (1982). Proc. Natl. Acad. Sci. U.S.A., 79, 2194-2198. Weston, K. M. (1990).Semin. Cancer Biol., 1, 371-382. Weston, K., and Bishop, J. M. (1989).Cell, 58, 85-93. Xu, G., Livingston,D. M.,and Krek, W. (1995).Proc. Nutl. h a d . Sci. U.S.A., 92,1357-1361. Zhu, L., Enders, G., Lees, J. A., Beijersbergen, R. L., Bernards, R., and Harlow, E. (1995). EMBOJ., 14, 1904-1913. Zhu, L., Zhu, I.., Xie, E., and Chang, L.-S. (1995). Mol. Cell Biol., 15, 3552-3562. Zwicker, J., Liu, N., Engeland, K., Lucibello, F. C., and Muller, R. (1996). Science, 271, 1595-1597. Zwicker, J., Lucihello, F. C., Wolfraim, L. A., Gross, C., Truss, M., Engeland, K., and Muller, R. (1995).L M B O J . , 1 4 . 4 5 1 4 4 5 2 2 . Zwicker, J., and Miiller, R. (1997). Trends Genet., 13, 3-6.

Alterations in DNA Methylation: A Fundamental Aspect of Neoplasia Stephen B. Baylln,'p2 JamesG. Herman,' Jeremy R. Graff,' Paula M. Vertino,' and Jean-Pierre h a ' ‘TheJohns Hopkins Comprehensive Cancer Center and 2Department of Medicine T l ~ e/ohm Hopkins Medical Instrtutions Baltimore. Maryland 2 12 3 1

1. Introduction 11. Brief History of DNA Methylation in Eukaryotes 111. Normal Roles for Cytosine Methylation in Higher-Order Eukaryotes A. Lessons Learned From Prokaryotes B. Role of DNA-Methyltransferase in Establishing DNA Methylation Patterns in Higher-Order Eukaryotes C. Anatomy of DNA Methylation in Higher-Order Eukaryotes I v. Abnormalities of DNA Methylation in Neoplasia A. DNA Hypomethylation in Cancer B. General Aspects of Regional DNA Hypermethylation in Cancer C. C p C island Hypermethylation Associated with Transcriptional Inactivation of Specific Genes in Neoplastic Cells V. Mechanisms Underlying the DNA Methylation Changes in Neoplastic Cells A. Role of DNA-Methyltransferase Activity Changes in Neoplasia B. What Additional Mechanisms Underlie the Aberrant Methylation of C p C Islands i n Neoplasia? VI. An Overview of Tumor Progression That Incorporates the Roles of Altered DNA Methylation VII. Clinical implications of Altered DNA Methylation in Cancer References

~~

~~~

~~~

Neoplastic cells simultaneously harbor widespread genomic hypomethylation, more regional areas of hypermethylation, and increased DNA-methyltransferase (DNAMTase) activity. Each component of this “methylation imbalance” may t!.:idamentally contribute to tumor progression. The precise role of the hypomethylation I S unclear, but this change may well he involved in the widespread chromosomal alterations in tumor cells. A main target of the regional hypermethylation are normally unmethylated C p C islands located in gene promoter regions. This hypermethylation correlates with transcriptional repression that can serve as an alternative to coding region mutations for inactivation of tumor suppressor genes, including p16, PIS, VHL, and E-cad. Each gene can be partially reactivated by demethylation. and the selective advantage for loss of gene function is identical to that seen for loss by classic mutations. H o w abnormal methyl-

Advance5 in CANCFK RESEAKCH

0065-LIOXIYE $ 2 5 . 0 0

Lopyrighr 0 1998 b y AcadciiiiL h e r \ . All rights ot rcproduction in any form rcwrvcd.

I42

Stephen B. Baylin P I al.

ation, in general, and hypermethylation, i n particular, evolve during tumorigenesis are just beginning to be defined. Normally, unmethylated CpC islands appear protected from dense methylation affecting immediate flanking regions. In neoplastic cells, this protection is lost, possibly by chronic exposure to increased DNA-MTasc activity andlor disruption of local protective mechanisms. Hypermethylation of some genes appears to occur only after onset of neoplastic evolution, whereas others, including the estrogen receptor, become hypermethylated in normal cells during aging. This latter change may predispose to neoplasia because tumors frequently are hypermethylated for these same genes. A model is proposed wherein tumor progression results from episodic clonal expansion of heterogeneous cell populations driven by continuous interaction between these methylation abnormalities and classic genetic changes.

1. INTRODUCTION During the past several years, there has been incremental interest in the role of abnormal DNA methylation in neoplasia because it has become apparent that changes in patterns and control of this process may play an integral role in tumorigenesis. This recent focus on an “epigenetic” mechanism was initially viewed as radical by some in the face of the prevailing thought that all of the critical steps in the origins and progression of cancer are mediated by classic genetic alterations. However, the potential role for methylation changes is increasingly supported not only by direct studies of neoplastic cells but also from increasing understanding of the essential contributions of DNA methylation to normal cell function. The purpose of this review is to place into perspective the evidence for a fundamental role of altered DNA methylation in the genesis and progression of neoplasia. We first review patterns and function of DNA methylation in normal cells and then describe the methylation changes in cancer, the consequences of these changes for specific events in tumor progression, and the mechanisms that may underlie them. We stress the notion that, with respect to the full spectrum of changes in DNA that drive the process of tumorigenesis, the epigenetic and genetic are integrally and dynamically linked.

II. BRIEF HISTORY OF DNA METHYLATION IN EUKARYOTES Since the serendipitous recognition some 50 years ago (for review, Weissbach, 1993) of the presence of methylated cytosines in DNA, this nucleotide modification has generated much interest and controversy. It was soon recognized that cytosine methylation is extensive in the genomes of species as diverse as bacteria, plants, and humans but is absent in organisms such as yeast, Cuenorhabditis eleguns, and Drosophilu sp. (for review, Bird, 1995).

DNA Methylation and Neoplasia

I43

The fact that these latter three species can replicate DNA and process it to facilitate all aspects of cellular function throughout embryogenesis and adult life raised questions about whether DNA methylation is a vital process in multicellular organisms. Until recently, this dilemma has had significant impact in slowing the pace of DNA methylation research with regard to both normal and pathologic states, especially when compared to the explosion of information emerging in other areas of genetic and molecular biology investigation. Another significant problem for understanding the role of DNA methylation in eukaryotes, especially for the molecular biologist interested in cancer, arises when one considers the difference in origin of methylcytosine in the genome from that of the four bases that constitute the standard genetic machinery. It was predicted some 40 years ago, and confirmed 10 years later (for review, Weissbach, 1993), that as a “fifth base,” methylcytosine is formed postreplicatively in DNA by addition of a methyl group to a cytosine already incorporated into previously synthesized DNA. Hence, methylation constitutes a modification of DNA and a technically epigenetic process. Furthermore, this fifth base has much greater potential than the other four to be removed, by demethylation to cytosine, from the genome during normal DNA replication. This is a particularly important point for considering the role of DNA meth-ylation in cancer because the aberrant presence o r absence of methylated sites in neoplastic cells cannot be viewed as a classic mutation, absolutely locked into the genome during the DNA replicative process. Despite these conundrums, work during the past decade has established an essential role for DNA methylation in higher-order eukaryotes. A critical step in this elucidation has been the identification of the major enzyme, DNA-methyltransferase (DNA-MTase), which catalyzes the addition of methyl groups to cytosine in vertebrate species. The mouse gene encoding this protein was cloned by Bestor and colleagues in 1988, and homologs have been subsequently identified in humans, chickens, and sea urchins (Aniello et al., 1996; Tajima et al. 1995; Yen et al., 1992; Yoder et al., 1996). Acquisition of the murine gene led to the key study in establishing a n essential role for DNA methylation in higher-order eukaryotes, in that homozygous deletion of the gene in mice produces embryonic lethality (Li et al., 1992). As discussed later, studies in this same gene knock-out model have also given functional context to the long-recognized inverse relation between DNA methylation and gene expression (Bird, 1992; Cedar, 1988; Razin, and Riggs, 1980; Razin and Cedar, 1991). This issue has become particularly important for the role of altered DNA methylation in neoplasia. Despite the growing evidence that DNA methylation plays a vital role in eukaryotes, much remains to be elucidated about the exact position of this process and the mechanisms involved in control of gene expression and other DNA functions. If one theme could be cited to summarize current under-

I44

Stephen B. Baylin e l af.

standing of why methylation is present in the higher-order eukaryotic genome, it might be the concept that this DNA modification is integrally tied to chromatin organization and function (Bestor, 1990; Bird, 1995). In fact, Bestor, Bird and others have hypothesized that the marked species variation in presence and extent of DNA methylation may be best explained in this context. The idea is that with increasing complexity of the genome during evolution, compensatory mechanisms became required for organization of the DNA into functionally inactive and active regions with respect to gene transcription (Bestor, 1990; Bird, 1995). Such mechanisms have been hypothesized to facilitate loss of gene expression no longer required for function of the organism, or dangerous to the organism, and continued expression of those genes that are important (Bestor, 1990; Bird, 1995). This potential interaction between patterns of DNA methylation and chromatin arrangement is a vital concept for viewing implications of the altered methylation patterns in neoplastic cells and will be revisited repeatedly in the discussions that follow.

111. NORMAL ROLES FOR CYTOSINE METHYLATION IN HIGHER-ORDER EUKARYOTES Elucidation of the normal role of DNA methylation in eukaryotes has been an elusive goal because, as noted previously, this DNA modification varies widely in its extent, and even existence, over a wide range of species and may play vastly different roles in different biologic settings. In this section, we briefly review, including consideration of species variation, current understanding of why DNA methylation is important for normal cellular function.

A. Lessons Learned From Prokatyotes Although this review deals almost exclusively with DNA methylation in humans, the goal of understanding the normal function of the process in this species is well served by considering, briefly, the extensive research conducted with bacteria. In these lower forms, both adenine and cytosine can be methylated to subserve complex functions with regard to DNA replication and arrangement (Noyer-Weidner et al., 1993; Riggs, 1990). Because only cytosine methylation occurs in humans, it is this process in bacteria that may hold some clues to function in higher organisms. Unlike for the single enzyme defined in vertebrates, a series of DNA-MTases, many with sequence specificity, can catalyze cytosine methylation in bacteria (Noyer-Weidner et al., 1993). For many of the sites involved, the responsible enzyme may car-

DNA Methylation and Neoplasia

145

ry both methylase and endonuclease activities. Alternatively, the site is recognized by a pair of enzymes, one that cleaves and one that methylates DNA. The apparent role for these dual activities is to provide for a n “immune” system, whereby the bacteria are protected from effects of foreign DNA insertion. In this process, endonuclease sites in inserted phage DNA are cleaved by the endonuclease activity, whereas these same sites in the parent DNA are protected by virtue of being methylated. Phage function can then survive only by acquiring similar methylation a t the same sites (Noyer-Weidner et al., 1993). This participation of DNA methylation in a n immune response to eliminate foreign DNA, and/or the function of such sequences, has been postulated to be conserved through evolution (Bestor, 1990), and indeed there is evidence for this. For example, Neurospora eliminate unwanted reiterated genes by a process that combines extensive cytosine methylation and relatively rapid appearance of point mutations throughout the repeated sequence region (Selker et al., 1987; Selker, 1993a; Selker et al., 1993b). Likewise, in rodents and humans, inserted viral sequences can become methylated in association with silencing of associated genes (for review, Doerfler, 1993; Stewart et a/., 1982), and transgenes in mice are not infrequently silenced in the same manner (for review, Sapienza et al., 1987; Sasaki et al., 1993; Surani et al., 1988; Swain et al., 1987). Thus, recognition by methylation of foreign DNA could be a normal role for this modification in humans, and certain methylation changes in neoplastic cells, discussed later, might also result from attempts to modify altered DNA sequences or structure.

B. Role of DNA-Methyltransferase in Establishing the

DNA Methylation Patterns in Higher-Order Eukaryotes As noted previously, emerging evidence during the past decade indicates that one overriding function of DNA methylation in vertebrates may be to facilitate organization of the genome into active and inactive regions with respect to gene transcription. In this regard, broadly speaking, there are two types of DNA methylation. First, new patterns of methylation, o r so-called de novo methylation, must be established at critical periods of embryogenesis and a t some points in the differentiation of adult cells (Monk, 1987; Monk et al., 1990; Razin and Cedar, 1993; Razin and Shemer, 1995). In vertebrates, only one DNA-MTase has been molecularly identified that catalyzes this de nouo methylation, and this is the same enzyme, as described later, that mediates maintenance of established DNA methylation patterns. DNA methylation has been extensively studied in mouse embryogenesis, and dynamic changes characterize the early stages (Monk et al., 1987; Monk, 1990; Razin and Cedar, 1993; Razin and Shemer, 1995). A wave of overall

I46

Stephen B. Baylin el al.

genomic demethylation is complete between the eight cell and blastula stages and is followed by dramatic remethylation of the genome thereafter. Most of the adult patterns of DNA methylation are established by late embryonic development. As noted in a previous section, homozygous knockout studies of DNA-MTase in the mouse suggest that this enzyme and the changing patterns of methylation that it establishes and/or maintains during early embryogenesis are vital for development and survival of the embryo (Li et al., 1992). One caveat to our understanding of how embryonic DNA methylation patterns are established is that more than one DNA-MTase may participate. Investigators have reported that low levels of DNA methylation persist during passaging of cultured cells from embryos of mice, which have a total homozygous knockout of the gene encoding the known vertebrate DNAMTase (Lei et al., 1996; Tucker et al., 1996). Additionally, vital sequences inserted into these cells continue to exhibit the same type of de novo methylation as is seen in control cells with wild-type DNA-MTase activity, albeit at a much reduced level (Lei et al., 1996). The exact nature of the one or more additional DNA-MTase enzymes involved in the residual methylation in the knockout cells remains to be determined with isolation of the responsible genes. The role of such an enzyme in adult cells, or in neoplasia, must than be elucidated. The process of DNA methylation in normal adult cells of vertebrates, which is most relevant to the changes associated with most forms of human cancer, largely involves maintaining the general patterns characteristic of all cells and selected patterns associated with specific cell types (for review, Bird, 1992; Cedar, 1988). This so-called maintenance methylation is, again, catalyzed by the single DNA-MTase characterized, and the properties of this enzyme are well suited to this role. The highly homologous DNA-MTase found in all vertebrate species studied has a marked preference for hemimethylated rather than unmethylated DNA (Bestor et al., 1988; Yoder et al., 1996). This property may hclp protect against the establishment of extraneous sites of DNA methylation in adult cell genomes and facilitates preservation of normal sites. During DNA replication in adult cells, DNA-MTase can thus recognize the normally methylated CpG sites in the parent strand and catalyze addition of a methyl group to the cytosine in the corresponding CpG site of the daughter strand. Active localization of the enzyme to sites of DNA replication in dividing cells (Leonhardt et al., 1992) may facilitate this maintenance role of DNA-MTase. The structure of the DNA-MTase protein that facilitates the above roles of the enzyme is well conserved in sea urchins, chickens, mice, and humans (Aniello et al., 1996; Tajima et al., 1995; Tucker et al., 1996a, 1996b; Yen et al., 1992; Yoder et al., 1996). The enzyme is an approximately 190-kd protein that, in eukaryotes, apparently arose from fusion of a protein that

DNA Methylation and Neoplasia

147

has the structure common to the bacterial cytosine methyltransferases, with a larger peptide that plays key modulatory roles. The portion resembling the bacterial enzymes is located in the carboxy one third of the protein and has a series of 10 motifs, including a proline-cysteine dipeptide core for the catalytic site, which are well conserved in position and amino acid content with the bacterial cytosine methyltransferases ( Bestor, 1988; Bestor et al., 1988; Yen et al., 1992). This carboxyl region of eukaryotic DNA-MTase acts as a potent de novo DNA-MTase when removed from the larger N-terminus region (Bestor, 1992). In addition to rendering the eukaryotic protein a more potent hemi-methylase than de novo methylase, the N-terminus region appears to contain the sequences that target DNA-MTase to DNA replication foci in dividing cells (Leonhardt et al., 1992). Little is known about the eukaryotic signal transduction pathways that control the expression of the DNA-MTase gene. The major known phenotypic ramification of DNA-MTase regulation is the increased enzyme activity that occurs with onset of DNA synthesis and that performs maintenance DNA methylation (Adams et al., 1993; El-Deiry et al., 1991; Szyf et al., 1985; Szyf et al., 1991). Posttranscriptional control appears to be important for increases in mRNA that mediate this increased enzyme activity in murine cells (Szyf et al., 1991).A weak basal promoter, thought to be upstream from the transcriptional start site of the murine DNA-MTase gene (Rouleau et al., 1992), has been reported to be upregulated by stimulation of the ras signal transduction pathway (MacLeod et al., 1995). Recent studies of the murine and human DNA-MTase genes (Tucker et al., 1996b; Yoder et al., 1996), however, now place the actual transcription start site far upstream from this area, and the transcriptional control of eukaryotic DNA-MTase expression must continue to be extensively explored.

C. Anatomy of DNA Methylation in Higher-Order Eukaryotes If the major role of DNA methylation in higher-order eukaryotes is to help organize chromatin into transcriptionally active and inactive regions, then the distribution of candidate C p C sites for methylation is as important as the role of DNA-MTase activity. During evolution, the sequence CpG has been progressively eliminated from the genome such that, in humans and other higher-order eukaryotes, this dinucleotide is present at only 5 to 10% of its predicted frequency (for review, Antequera and Bird, 1993a, 1993b; Bird, 1995). Methylation appears to have played a critical role in this process in that most CpC sites lost appear to represent the conversion, through deamination, of methylcytosines to thymidines (Antequera and Bird 1993a, 1993b; Bird, 1995). In humans, most vertebrates, and plants with complex

148

Stephen B. Baylin el al.

genomes, most remaining CpG sites, some 70 to 80%, contain methylated cytosines (Antequera and Bird 1993b; Bird, 1995). These methylated regions are typical of the bulk chromatin that constitutes most nontranscribed DNA and of the body of many genes. The chromatin pattern associated with this methylated DNA represents, in histone composition and nucleosomal arrangement, the late replicating DNA, which is relatively inaccessible to transcription factors (Tazi and Bird, 1990). In contrast to most of the higher eukaryotic genome, smaller regions of DNA, termed CpG islands, ranging in size from 0.5 to about 4 to 5 kb and occurring on the average every 100 kb (Antequera and Bird, 1993b; Bird, 1986, 1992, 1995), have maintained the expected frequency of CpG content. Mathematical criteria employed to define a CpG island include a G + C content of greater than 6 0 % and a ratio of CpG to GpC of at least 0.6 (Antequera and Bird, 1993b). Most frequently, but not always, these islands are associated with the 5’ regulatory regions of genes. This may reflect the fact that maintained C p C frequency is important to the process of gene transcription. Hence, over evolution, these areas were not methylated and, therefore, not depleted through the C to T transitions associated with deamination of methylcytosine (Antequera and Bird, 1993b; Bird, 1995).In fact, save for inactivated genes on the inactive X-chromosome of females in most mammalian species, and on the parentally silenced alleles of a selected number of imprinted genes on autosomal chromosomes, the CpG islands in 5’ gene regulatory areas are completely unmethylated (for review, Antequera and Bird, 1993b).The specific mechanisms that protect most CpG islands from methylation are only beginning to be understood and are reviewed later in the discussion of the role of C p C island methylation in neoplasia. To understand, fully, DNA methylation changes in cancer, it is imperative to realize that these patterns of normal DNA methylation broadly reflect two types of gene 5’ regulatory regions in the eukaryotic genome (Table I). About half of all genes in species such as mouse and human, or some 40,000 to 50,000 genes, have 5’ CpG islands (Antequera and Bird, 1993a, 1993b). These are largely the “housekeeping” genes, which have a broad tissue pattern of expression, but many relatively tissue-specific genes are also represented (Antequera and Bird, 1993a, 1993b; Bird, 1992). It is important to note that the unmethylated status that characterizes the CpG islands associated with most of these genes is independent of whether a given gene is being actively transcribed. Thus, differential methylation status is not a regulatory feature for control of expression of these genes in normal cells. That the lack of methylation is important for active transcription of these genes is reflected by the facts that (1) chemically induced demethylation of CpG islands associated with X-chromosome inactivated genes leads to partial reactivation (for review, Migeon, 1994; Singer-Sam and Riggs, 1993; Wolf et al., 1984); and (2) some silenced imprinted alleles with normally densely

I49

D N A Methylation and Neoplasia

Table I Differences in Methylation Status and Function in Gene Promoters With and Without CpC Islands Promoter Region CpG Islands Virtually always unmethylated, regardless of active transcriptional status of gene, in all normal tissues. Exceptions include inactivated genes o n the inactive female X-chromosome in most species and the inactive allele for selected imprinted genes on autosomal chromosomes. The unmethylated state is associated with open nucleosomal configuration characteristic of transcribed genes, including early replication timing, relative lack of histone H1, and presence of acetylated histones. Methylation, as a function of increasing numbers of CpC sites involved, is associated with long-range changes in chromatin, which mediate a closed, nucleosomal conformation, delayed replication timing, and transcriptional repression. ~~

~

~

~

CpG Sites in Promoter Regions Without CpG Islands Variably methylated, often in a tissue-specific pattern, and inversely related to the transcriptional status of the gene Rather than modulating long-range chromatin effects, methylation status of specific sites in, or near, transcription factor recognition motifs can affect factor binding and modulate gene expression.

methylated CpC islands are actively transcribed in mouse embryos, which have homozygous deletion of the DNA-MTase gene and virtual absence of genomic methylation (Li et al., 1993). The remaining half of mouse and human genes do not have 5 ’ region C p C islands. Importantly, the individual CpC sites in the regulatory areas of these genes can be methylated in normal cells and often in tissue-specific patterns that reflect the transcription status of the genes (for review, Bird, 1992; Cedar, 1988). In many instances, these CpG sites are not methylated if the gene is actively expressed and methylated in cells with little or no transcription of the gene. It is critical for considering the implications of altered DNA methylation patterns in cancer to understand the differential manner in which methylation affects the transcriptional status of genes with and without CpC islands (see Table I). In those with CpC islands, lack of methylation is highly correlated with a n early timing for replication of the DNA region (Kitsberg et al., 1993; Selig et al., 1992) and with the chromatin pattern of actively transcribed genes, as characterized by an open nucleosome configuration, a reduced amount of histone H1 in the regions, and presence of acetylated histones (Tazi and Bird, 1990).The ability of methylation to silence genes with CpG islands, as studied for X-chromosome inactivated genes (Riggs and Pfeifer, 1992; Singer-Sam et al., 1993) and gene C p C islands transfected into

I50

Stephen B. Baylin ef of.

cells in the methylated versus unmethylated state (see for example, Cedar, 1988; Keshet et al., 1986; Paroush et al., 1990), appears largely to reflect conversion of the overall chromatin pattern from an open to a closed configuration. In turn, this chromatin modulation is highly correlated with increasing numbers of CpG sites methylated within a given CpG island (Pfeifer et al., 1990). In CpG island-containing genes, accessibility of selected transcription factors to specific DNA sequences could be negatively influenced by the methylation status of individual CpG sites. This mechanism, however, appears less important than the increased methylation density and the resultant long-range chromatin effects that favor exclusion of positive transcription factors from the regulatory promoter region. For selected genes, such as many on the inactive X-chromosome and silenced alleles for imprinted genes, which normally have methylated CpG islands, the resultant chromatin effects may well be mediated, in part, by a series of recently defined proteins, termed methylcytosine-binding proteins (MCBPs), which bind preferentially to methylated as opposed to unmethylated DNA (Boyes and Bird, 1991; Meehan et al., 1989; Nan et al., 1993, 1996; Tate et al., 1996). Some MCBPs bind proportionally to the density of methylated CpG dinucleotides in a defined region like a CpG island, and these may play an important role in determining the different chromatin patterns associated with different states of DNA methylation. Indeed, mice highly chimeric for ES cells containing mutations of a X-linked gene for one such MCBP demonstrate similar embryonic lethality to that seen with knockout of the DNA-MTase gene (Tate et al., 1996). In contrast to these genes, those without 5’ CpG islands may have their state of transcription modulated more often by methylation of individual CpG sites in the promoter region that are near, or directly within, transcription factor-binding motifs. Examples of such site-specific methylation affecting transcription include a sequence in the GK-interferon promoter (Melvin et al., 1995; Young et al., 1994;) and the cyclic AMP response element-binding protein (CREB) region in the P-globin gene promoter (IguchiAriga and Schaffner, 1989). In each case, methylation of a CpC site in the transcription factor-binding sequence blocks accessibility of the transcription factor.

IV. ABNORMALITIES OF DNA METHYLATION IN NEOPLASIA For the past 15 to 20 years, abnormal patterns of DNA methylation have been recognized as a constant, but poorly understood, molecular change in human neoplasia. With respect to the normal regulation of DNA methy-

DNA Methylation and Neoplasia

151

lation discussed previously, at least three alterations, constituting an imbalance of the process, are features of tumor progression. Transformed cells of virtually all types often have, simultaneously, widespread loss of methylation from normally methylated sites, increased total activity of DNA-MTase, and more regional areas of hypermethylated DNA (previous reviews by Baylin et al., 1991; Jones, 1996; Laird and Jaenisch, 1994). Each of these changes has potentially profound consequences for DNA function and structure, and the interplay between them may be particularly important for neoplastic progression. In the discussions that follow, both the individual components of the imbalance of DNA methylation in cancer and the possible dynamic links between them are stressed.

A. DNA Hypomethylation in Cancer One of the first alterations of DNA methylation to be recognized in neoplastic cells was a decrease in overall genomic levels of this modification (see, for example, Gama-Sosa et al., 1983). In a number of experimental models of carcinogenesis, this decrease in numbers of methyl groups appears to begin early in tumor progression and before the appearance of frank tumor formation (Christman et d., 1993; Pogribny et al., 1995; Wainfan and Poirier, 1992). A possible direct role for DNA hypomethylation in the neoplastic process has been proposed from experimental data that, in rodents, depletion of methyl donor groups by dietary restriction of S-adenosyl-methionine results in liver carcinogenesis and in DNA hypomethylation, which precedes tumor development (Christman et al., 1993; Pogribny et al., 1995; Wainfan and Poirier, 1992). Despite the clear association of DNA hypomethylation with both experimentally derived and spontaneous tumors, the exact ramifications of this change for steps in tumor progression are poorly understood. A possible link to changes in patterns of gene expression, and specifically to potential activation of oncogenes, emerged from the observations of Feinberg and Vogelstein (1983a, 1983b) that the hypomethylation in human tumors could directly involve specific genes, such as K-ras, in lung and colon carcinomas. Subsequently, hypomethylation patterns have been described for a number of oncogenes in human and experimentally produced tumors in other species (see, for example, Bhave et al., 1988; Christman et al., 1993; Miyamura et al., 1993; Wainfan and Poirier, 1992).As for the overall decrements in DNA methylation in tumor progression, these gene-localized events can also occur before the appearance of frank malignancy, such as in the benign colon polyps, which can serve as precursor lesions to colon carcinoma (Goelz et al., 1985). Although gene-specific hypomethylation can certainly be a potential fac-

I52

Stephen B. Baylin ef a / .

tor in aberrant increases in gene expression in neoplastic cells, this may actually not be a common association. This is best appreciated when the tumor-associated hypomethylation changes are put into perspective with respect to the position of the hypomethylation within the structure of a given gene. Changes in DNA methylation within the body of genes often correlate poorly with levels of gene expression. As previously noted, the best correlation between patterns of DNA methylation and regulation of gene expression involves CpG sites located within the two types of promoters outlined in Table I. For the genes with promoter region CpG islands, CpG sites would not be methylated in normal cells and thus would not be targets for hypomethylation in transformed cells. For the remainder of genes that have limited numbers of promoter region CpC sites, many of which have tissuespecific expression patterns, hypomethylation of only selected sites would be expected to participate in activating gene expression, and then only if specific trans-acting transcription factors are available in the cell. These characteristics of normal promoter region structure and methylation patterns probably explain why the frequent overall DNA hypomethylation in neoplastic cells has only rarely been associated (see, for example De Smet et al., 1996) with aberrant activation of expression of specific genes. If changes in gene expression are not a common consequence of hypomethylation in neoplastic cells, then what role might this DNA alteration be playing in tumor progression? Although this is not yet clear, it appears possible that the structural integrity of chromosomes could be profoundly affected. As previously outlined, the bulk chromatin, which is important for cellular packaging of DNA and structural considerations, is heavily methylated. Satellite DNA-containing regions of chromosomes are prime examples, and replication timing of such areas appears dependent on methylated DNA content (Selig et al., 1988). When decreases in overall genomic methylation are induced by 5-AZA-cytidine, some cells appear to become transformed (see, for example, Rainier and Feinberg, 1988; Walker and Nettesheim, 1986), and disjunctive abnormalities of chromosome division during cell replication have been observed (see, for example, Schmid et a/., 1983). Although many more studies are required to establish the precise consequences of DNA hypomethylation in neoplastic cells, it is possible that this change could play a central role in the gross chromosome alterations, such as changes in ploidy, nondisjunction, and so forth, that are the hallmark of cell transformation.

B. General Aspects of Regional DNA

Hypermethylation in Cancer The same neoplastic cells that often harbor the overall genomic hypomethylation (discussed earlier) frequently have more limited regions of

DNA Methylation and Neoplasia

I53

dense hypermethylation. As would be mandated by the starting situation in the normal cells from which tumors arise, this change would have to involve the only 2 0 % or so of CpG sites that are unmethylated. This fact, in turn, would predict that the normally unmethylated CpG islands in gene 5’ regulatory regions would be a prime target for aberrant hypermethylation in tumor cells. Indeed, work during the past 10 years suggests that such hypermethylation is a hallmark of the cancer cell and is the most profound association between altered DNA methylation and changes in gene expression in neoplasia. As indicated from the previous discussion, dense methylation within a 5’ regulatory region has the potential to participate in chromatin events that result in transcriptional silencing of an involved gene. The concept that such events might be an important association with aberrant gene expression in cancer is not new. Holliday and others have noted that promoter region hypermethylation can serve as an “epimutation” to silence genes involved in situations of auxotrophy in cultured cells (Holliday and Jeggo, 1985; Holliday, 1990a, 1990b, 1991). These data led to the prediction that key genes in tumor cells could be similarly affected. In the middle and late 1980s, our group serendipitously detected that a CpG island in the promoter region of the calcitonin gene at chromosome 1l p , which was unmethylated in all normal tissues tested, was densely methylated in human solid tumors (Baylin et al., 1986), leukemias (Baylin et al., 1987), and cells transformed with various viruses, including Epstein-Barr virus, SV40, and human T-cell lymphotrophic virus type 1 (Baylin et al., 1987; de Bustros et al., 1988). Although the role for this particular gene in tumor progression seemed to be obscure, we also discovered that multiple other CpG-rich regions on this same chromosome arm, which is known to harbor multiple candidate tumor suppressor genes (Loh et al., 1992), were simultaneously hypermethylated in these situations (de Bustros et al., 1988). We hypothesized that this 1l p region behaved as a “hot spot” for CpC island methylation in neoplasia and that this DNA change could be an important potential mechanism for inactivation of tumor suppressor genes (de Bustros et al., 1988). Subsequent association, in multiple tumor types, of hypermethylation of a CpG-rich area located in a candidate tumor suppressor region represented by frequent loss of heterozygosity (LOH) at chromosome 17~13.3(Makos et al., 1993a, 1993b, 1992) strengthened our conviction that this theory might be plausible. During this period, Antequera and colleagues recognized that multiple CpG islands, some associated with genes, were hypermethylated in immortalized human and murine cells and postulated that as many as half of the C p C islands in the genome might be so altered in such cells (Antequera et al., 1990; Antequera and Bird, 1993a). They theorized that this process could contribute to the immortalized phenotype by silencing expression of genes responsible for control of normal cell differentiation and/or inhibition

I54

Stephen B. Baylin e l al.

of cell growth, a process they termed methylation-associated gene inactivation (MAGI)(Antequera and Bird, 1990). A specific example of this concept emerged from the data that the muscle differentiation control gene, myo-D, which was discovered through its activation by chemical demethylation (Chen and Jones, 1990; Jones et al., 1990a, 1990b), contained a 5’ region CpC island that is methylated in immortalized, but not normal, murine fibroblasts (Jones etal., 1990b). In contrast, Antequera and colleagues (1990) also postulated, and presented some data to support the concept, that C p C island methylation of genes vital for cell growth and/or viability would be selected against in immortalized cells. All of these data from our group and others, although suggestive that C p C island hypermethylation could be an important mechanism for loss of gene function in neoplasia, provided n o direct proof for this concept. The critical steps of providing specific examples of functional involvement for genes essential to the genesis and/or progression of cancer was required. During the past 2 years, a growing list of such genes has emerged (Table 11). The following sections outline how study of these genes is clarifying the participation of C p C island hypermethylation in gene inactivation in tumor progression and describe early studies aimed at defining the molecular mechanisms underlying the hypermethylation process.

C. CpC Island Hypermethylation Associated With Transcriptional Inactivation of Specific Genes in Neoplastic Cells I . RETINOBLASTOMA GENE The retinoblastoma (Rb) gene was the first classic tumor suppressor gene in which CpC island hypermethylation was detected. The change has been noted in tumor DNA from about 10% of patients with the nongenetic o r sporadic form of retinoblastoma (Greger etal., 1994; Sakai etal., 1991).The specific consequences of this finding were not initially readily apparent because the large size of the gene made concomitant inactivating mutations hard to rule out and because studies in only one cell line have correlated gene expression and the methylation patterns (Greger et al., 1994). A subsequent report, however, revealed that in vitro methylation of the promoter region of Rb directly blocked promoter activation by a trans-activating transcription factor (Ohtani-Fujita et al., 1993). Interestingly, although the Rb gene is commonly mutated in many forms of human cancer, promoter region hypermethylation has been reported only in retinoblastomas.

Table II Genes Associated With Promoter Region CpG Island Hypermethylation in Human Tumors ~

Gene

Locus

Tumors With Methylation

Comments

Evidence for Selective Advantage

Documented Tumor Suppressor Genes Rb

13q14.2

Retinoblastoma

Transcription status incompletely studied; n o studies of reactivation

VHL

3p25

Renal carcinoma

Clear cell histology only

P16'"K4A

9p21

Most common solid tumors and lymphomas

p151NK4h

9p2 1

Primary acute leukemias and Burkitt lymphoma

Both deletions of p16 and hypermethylation occur only in tumor with wild-type retinoblastoma (Rb) Hypermethylation infrequent in nonhematologic tumors

E-cadherin

16q22.1

Bladder, breast, colon, liver Colon

hMLH 1

Methylation seen in three sporadic tumors

Hypermethylation occurs in gene for which mutations are known to initiate the tumor type Both mutations and hypermethylation occur only in renal cancer Accounts for loss of cyclin DlRb pathway in tumor cells with wildtype Rb Hypermethylation may corelate with loss of growth-inhibitory response to TGF-P Loss of function contributes to invasive and/or metastatic phenotype Loss causes microsatellite DNA instability

Probable or Candidate Titmor Suppressor Genes ER

6q25

Breast, colon, lung tumors; leukemia

Age-related event in normal colon

Loss of hormonal responsiveness in breast cancer and possible growth suppression in multiple tumor types

(continues)

Table I1 Continued Gene

Locus

Tumors With Methylation

Comments

Evidence for Selective Advantage

Located in a frequent LOH region; member of zinc-finger transcription Fatty acid-binding protein, for which role not yet known

Inserted gene decreases G418 selection of cultured tumor cells Inserted gene causes growth inhibition of cultured breast cancer cells

Present in virtually 100% of tumors; high biomarker potential for prostate cancer Glioblastoma cell lines only

Not known

HICl

17~13.3

MDGI

lp35

Brain, breast, colon, renal tumors Breast cancer

GST-P

llq13

Prostate cancer

06-MGMT

10q24

Brain cancer

Calcitonin

Ilpl5

Carcinoma, leukemia

Concordantly hypermethvlated with other 11 p genes

myo-D

llpl5

Bladder tumors

Frequently seen in immortalized murine cells

Other Genes

Loss of function would confer resistance to multiple types of drugs None known-may be occurring in regional change affecting other 1 1 p tumor suppressor genes None known-same regional changes as for calcitonin

Data compiled from multiple references discussed in the text except for the MDGI (mammary-derived growth inhibitor) gene (Huynh ef a/.. 1996), for the Oh-MCMT (0-6-methylguanine DNA methyltransferase) gene (Costello ef a/., 1994; Costello eta/.. 1994b; Pieper er a/., 1996), and the h,MLHI gene (Kane ef a/., 1997).

DNA Methylation and Neoplasia

I57

2. VON HIPPEL-LINDAU GENE The von Hippel-Lindau (VHL) gene, located at chromosome 3p, was the first classic tumor suppressor gene for which a full paradigm was developed for association of transcriptional inactivation with CpG island hypermethylation (Herman et al., 1994). Germline mutations in the coding region of the VHL gene are responsible for the autosomal dominant inheritance of the von Hippel-Lindau syndrome, in which patients develop clear cell renal cancer and other tumors (Latif et al., 1993). In addition, somatically acquired mutations in one copy of the VHL gene, coupled with loss of the second allele, characterize 60% of the more common sporadic form of clear cell renal carcinomas (Gnarra etal., 1994). In another 20% of these sporadic tumors, the retained VHL allele has no coding region mutations but harbors dense hypermethylation of a typical CpG island spanning the transcription start site (Herman et al., 1994). Our studies of the VHL gene have provided guidelines, summarized in Table 111, for linking tumor suppressor gene inactivation with promoter region hypermethylation. First, by virtue of having defined coding region mutations in the germline of patients with a n inherited tumor syndrome and L O H in tumors from such patients, VHL is a bona fide tumor suppressor gene. Second, renal cancers with VHL gene hypermethylation have no coding region mutations within this gene (Herman et al., 1994). Third, such tumors express no transcripts for the gene, whereas those with coding region mutations d o (Herman et al., 1994). Fourth, direct involvement of the hypermethylation in loss of gene expression was demonstrated. Treatment of cultured renal tumor cells harboring a hyperrnethylated VHL gene with the demethylating agent, 5-deoxy-AZA-cytidine, resulted in restoration of active transcription (Herman et al., 1994). Finally, and extremely important, the selective advantage that VHL gene inactivation confers to renal cancer cells Table 111 Prerequisite Characteristics to Define Tumor Suppressor Genes Inactivated in Association With CpG Island Hyperrnethylation Gene with defined tumor suppressor function Dense methylation in tumor DNA (defined by examination of at least multiple methylationsensitive restriction sites) of a typical CpG island in the 5’ region of the gene (usually encompassing the 5’ flanking region and the transcription start site), which is not methylared in corresponding normal tissue Absence of coding region mutations in tumors in which gene is hypermethylated Absence of gene expression in tumor at the mRNA transcript level Ability to reactivate, a t least partially, gene expression in tumor cells by demethylation of gene with 5-AZA-cytidine treatment Evidence that selective advantage for loss of gene function is the same for either hypermethylation or inactivating mutations

I58

Stephen B. Baylin rl ul.

appears t o be identical for either mutations or hypermethylation of the genc. Among sporadic tumor types, mutations of VHL are selective for renal carcinoma. Similarly, VHL promoter region hypermethylation has also been found only for this cancer type (Herman et al., 1994). These principles, learned from study of the VHL gene, have now been defined for each of the bona fide tumor suppressor genes discussed later and listed in Table 11. 3 . INVOLVEMENT IN SOLID TUMORS OF CYCLINDEPENDENT KINASE INHIBITOR GENES, p16 AND p15

Studies of the genes, p16 and p l 5 , located near one another on chromosome 9 ~ 2 1 have , taught us much about how combining assessment of promoter region methylation with classic mutation analyses can reveal the full scope of tumor suppressor activity for a given gene. The original work began with the p16 gene, which encodes for a constitutively expressed cyclindependent kinase inhibitor that plays a vital role in control of the cell cycle by the cyclin D-Rb pathway (Serrano et al., 1993). This protein blocks activity of the cyclin D-dependent kinase (CDK), which can inactivate the Rb protein by phosphorylation (Strauss et al., 1995; Weinberg, 1995). Thus, loss of p16 can result in cellular inability to activate the Rb gene, with resultant loss of cellular capacity to block cell cycle progression (Strauss et al., 1995; Weinberg, 1995). The p16 gene resides in an area of 9p that is frequently lost, homozygously, in many types of cancer, and this finding created much excitement about the possibility that p l 6 could be the most frequently altered tumor suppressor gene in neoplasia (Kanib et al., 1994a; Nobori et al., 1994). Furthermore, more recent studies indicate that homozygous deletion of the p16 gene i n mice is a highly tumorigenic event (Serrano et al., 1996).Controversy arose, however, concerning the true importance of p16 gene alterations in human cancer. The homozygous deletions s o common in culture proved difficult t o detect in primary tumors, possibly for technical reasons and because native tumors may be heterogeneous for cell subpopulations harboring this change (Cairns et al., 1994; Okamoto et al., 1994; Spruck, et a/., 1994). Also, classic inactivating point mutations are not common for the p16 gene in tumors, even though clear examples were found in the germline and in tumors from some kindreds with familial forms o f melanoma (Kanib et al., 1994b; Hussussian et al., 1994) and in 38% of pancreatic adenocarcinomas (Caldas et al., 1994). In this setting, studies of promoter region methylation status have played a major role in solidifying the concept that the p16 genc is a major tumor suppressor gene. Among most solid tumor types studied, the incidence for hypermethylation of a CpG island in the 5 ’ region of the p16 gene in noncultured tumors ranges from 20 to 67% (Fig. 1 and Table IV) (GonzalezZulueta et al., 1995; Herman etal., 1995; Lo et al., 1995; Merlo et al., 1995;

DNA Methylation

and Neoplasia

159

Otterson et al., 1995; Reed et al., 1996; Shapiro et al., 1995b). As for the homozygous deletions of this gene, the incidence of this event in cell cultures of the same tumor types is generally even higher, ranging up to 90% for colon cancer (see Fig. 1 and legend). In cell culture, all of the components of the paradigm outlined in Table 111 for designating tumor suppressor genes inactivated in association with hypermethylation were fulfilled in the first (Merlo et al., 1995) and subsequent studies (Gonzalez-Zulueta et al., 1995; Otterson et al., 1995). When the incidence for hypermethylation of p l 6 in solid tumors, taken from those studies, is added to that for homozygous deletions of the gene, an extraordinary picture emerges for involvement of p16 function in noncultured solid tumors (ranging from about 30% in breast and colon tumors to near 9 0 % in head and neck carcinomas, see Table IV). Some features of the p16 gene story in solid tumors bear special mention and have revealed surprising aspects of the role of hypermethylation in tumor suppressor gene inactivation. First, the selective advantage for tumor progression of losing the p16 gene must be considered in the context of the entire cyclin D-Rb pathway. Cells appear to need only one step altered in this cell cycle control route to escape inhibition of cell growth. Thus, in any given tumor, amplification of cyclin D or CDK or, most commonly, loss of p16 function or inactivating mutations of Rb almost always occur independently of one another (Strauss et al., 1995). In this context, it is important to note that the presence of either coding region deletions or promoter region hypermethylation of p16 are now known to be inversely related to the presence of Rb gene mutations in multiple tumor types (Herman et al., 1995; Merlo et al., 1995; Okamoto et al., 1994; Otterson et al., 1994, 1995; Shapiro et al., 1995a; Washimi et al., 1995), and this is especially well studied for human lung cancer cell lines (Merlo et al., 1995; Otterson et al., 1994, 1995; Shapiro et al., 1995b). As outlined in Table IV, because p16 function may be so frequently lost in solid tumors, the critical cyclin D-RB pathway is often disrupted in noncultured tumors and probably seldom is intact in cultures of epithelial cancers. Second, investigations of p l 6 gene methylation have revealed that this promoter region change can function in some tumor types as the inactivating event for both alleles of the gene and also constitute the only molecular lesion associated with loss o f the cyclin D-Rb pathway. The best example of both of these situations is colon cancer, in which allelic loss of the 9p region, homozygous deletions of the p16 gene, and Rb gene mutations are all virtually absent (Cairns et al., 1995; Kamb et al., 1994a). Yet, 30 to 4 0 % of primary colon cancers have distinctly hypermethylated p l 6 alleles, and virtually all cell cultures of this tumor type have this gene aberrantly methylated (Herman et al., 1995). This loss of p16 gene and cyclin D pathway function in colon cancer would never have been expected from classic examination, alone, for mutational changes in either the Rb or p16 genes.

Normal Familial melanoma, pancreatic carcinoma Many Cancers

Occasional Carcinomas Pediatric T-ALL Carcinoma, NH lymphoma Adult AML, ALL Pediatric AML, BALL BurkiWs lymphoma

DNA Methylation and Neoplasia

161

Third, studies of methylation patterns in the p16 gene have further clarified the role of this protein compared with other candidate tumor suppressor genes located in the same region of chromosome 9p. In most solid tumors, the predominant structural lesion of chromosome 9p21 is deletion of an area harboring not only the p16 gene but also the nearby p l 5 gene (see Fig. 1). (Jen et al., 1994; Kamb et al., 1994b). This latter gene is another CDK inhibitor with extensive homology to p16 (Hannon and Beach, 1994). Unlike the p16 gene, however, the p l 5 gene is less constitutively expressed and is induced during the growth-inhibitory effects of transforming growth factor-f3 (TGF-P) in several cell systems (Hannon and Beach, 1994).The frequent inclusion of both the p l 5 and p16 genes in the 9p21 deletion area has made it difficult to know which gene loss may be more important to tumor progression, although most evidence has pointed to the p16 gene. When deletions include only one of the genes, p16 is almost always the gene lost (Hebert etal., 1994). Also, although point mutations are not frequent in the p16 gene, they d o occur, but these have not been documented in the p l 5 gene (see, for example, Stone et al., 1995). Studies of methylation patterns for the p l 5 and p16 genes in solid tumors have further pointed to the more important role for the p16 gene. Both genes have typical CpG islands around their 4

Fig. 1 Most frequent modes of p l S and p16 gene inactivation in major classes of human neoplasms. Lines 1 through 8 depict the status of the two genes in each tissue class. A break in the line indicates presence of a homozygous deletion, the closed squares indicate exons, a break in an exon indicates a point mutation, and M indicates hypermethylation o f the promoter region CpG island. The arrows depict the transcription start site and an active state of transcription for each gene except when hypermethylation is present. The data in the figure are compiled from the multiple references given and discussed in the text and in Table IV. Comments for each line are as follows: 1. The normal status of the genes and their location near one another on 9p21 are shown, including complete absence of methylation of the 5’ C p C islands. 2. Point mutations for p16 are infrequent in human tumors and occur in the germline of some kindreds with familial melanoma and in 38% of pancreatic carcinoma nude mouse xenografts. 3 . One of the most common lesions in human tumors, especially cell cultures of the common solid tumors, is homozygous deletion of both genes. 4. Isolated homozygous deletion of p l 6 , seen occasionally in multiple tumor types. 5 . Isolated hypermethylation o f p l S with homozygous deletion of p16, a pattern quite specific (about SO%) for pediatric T-cell acute lymphocytic leukemia (T-ALL). 6 . Isolated hypermethylation of p l 6 , the most common hypermethylation pattern found for this gene in most solid tumor types (see Table IV). 7 . Isolated hypermethylation of PIS, a frequent pattern in subtypes of hematologic malignancies, including adult acute myelogenous leukemia (AMI., SSYo), pediatric AMI. (67%), adult ALL (710/), and pediatric B-cell ALL (48%)). 8. Concomitant hypermethylation of both p l S and p16, a pattern frequently seen only in Burkitt lymphoma (50%).

I62

Stephen B. Baylin e l al.

Table IV Inactivation of the p16 Gene by Homozygous Deletion ( H D ) or With CpG Island Hypermethylation in Some Human Solid Tumors

Cell Lines (YO)

Primary Tumors (YO) Tumor Type

HD

Methylated

Inactivated

HD

Methylated

Inactivated

Brain Breast Colon Head and neck Nasopharyngeal Non-small cell lung Renal

30-50 0 67 35 2s

25 31 30-40 20-2s 22 2s >

55-75

7s

50 0 30 67 44 50

5 31 92 30 ? 22-64 2s

no

?

?

30-40 87-92 57

SO ?

ni 92 60 ?

66-100 75

Data compiled from the following references: hrain (Cosrelloetal., 1996; M e r l o e t d . , 1995). hreast (Herman eta!., 1995; Herman e t a / . , 1996b). colon (Cairns et d . , 199s; Herman e t a / . , 1995, 1996h), head and neck (Merloetal., 1995; Reedetal., 1996),nasopharyngenl ( L o e t a / . , 199.5, 199h),non-smallcell lung(Merlo e t a / . , 1995; Otterson et a/., 1994, 1995; Shapiroetol., I Y Y S h ) , renal (Herman eta/.. 199.5). See also N o hori e t a / . , 1994 and Kamh e t a / . . 19943, 1994h.

transcription start sites, but despite residing within 15 kb of one another, only the p16 gene is frequently aberrantly methylated when copies of both genes are retained in solid tumors (see Fig. 1)(Herman et al., 1995, 1996b). Only occasional CpG island methylation of the p15 gene is seen in lung cancer cell lines, and almost always with concomitant hypermethylation of the p16 gene (Herman et al., 1996b).Isolated hypermethylation of p l 5 has also been detected in two of multiple examined glioblastomas (Herman et al.. 1996b). Finally, studies of the p16 gene in an animal model of lung carcinogenesis have provided further evidence for the importance of this gene in tumor progression and for the role of CpG island hypermethylation in inactivation of the gene. In the rat and mouse, exposure to various tobacco-related carcinogens and radiation leads to lung tumors that are essentially identical to human lung adenocarcinoma (Dragani etal., 1995). In the rat model, in both carcinogen settings, the incidence of p l 6 gene CpG island hypermethylation (45%) and of homozygous deletions of the gene (40”/) are extremely high in cultures of the lung tumors (Swafford et al., 1997). This conservation of p16 gene loss across species, and for involvement of hypermethylation in the process, emphasizes the importance of these events in tumorigenesis. Furthermore, data from study of the rat lung tumors firmly indicate that the frequent CpG island hypermethylation of p16 in cultured tumor cells originates in cells from the parent noncultured tumors. In the rat model system, in every paired case examined, there was concordancy for presence of methylated p16 gene alleles in primary tumor and the culture established

DNA Methylation and Neoplasia

I63

from that particular tumor. In contrast, methylated alleles were infrequently found in tumors for which the corresponding cultures had homozygous deletions of the gene (Swafford et al., 1997). 4. THE p16 A N D p15 G E N E S IN LEUKEMIA A N D OTHER H EMATOPOlETlC MALIGNANCIES In tumors of hematopoietic origin, the patterns and mechanisms of inactivation for the p16 and p l 5 genes differ dramatically from those in solid tumors and are determined to a remarkable degree by the differentiation origins of tumor cell type within this single cell lineage system. Again, the data show that CpC island methylation does not silence these genes only in cell culture because virtually all of the information about hematopoietic tumors has evolved from studies of primary tumor cells. Unlike for the solid tumors, the loss of p l 5 gene function is a dominant event in tumors derived from the hematopoietic cell system, and C p C island hypermethylation is by far the dominant mechanism involved (see Fig. 1). This is most dramatically apparent from large studies of patients with both the adult and childhood types of acute myelogenous leukemia (AML), in which 85% and 65%, respectively, of the noncultured tumors have this change in the p l 5 gene (Herman et al., 1996b; Herman et al., 1997). No homozygous deletions or mutations of either the p l 5 or p16 genes have been found in this setting, and the p16 gene is not simultaneously hypermethylated (Herman et al., 1996b; Herman et al., 1997).In established leukemia cell cultures, the parameters summarized in Table I1 have been met for the p l 5 gene (Herman et al., 1996b). In this regard, the selective advantage for loss of this gene in leukemogenesis bears special mention. As mentioned previously, this CDK inhibitor appears to play a role in TGF-p-mediated growth inhibition responses (Hannon and Beach, 1994), and normal hematopoietic stem cells are known to be particularly sensitive to such inhibition (for example, Keller et al., 1989).Furthermore, in mice engineered for homozygous disruption of the TGF-P gene, which die soon after birth, a profound lymphoid hyperplasia develops with features of a preleukemic state (Shull et al., 1992). Finally, there is an excellent correlation between those leukemia cell lines known to harbor a hypermethylated p l 5 gene and those that lack a growth inhibition response to TGF-P (Herman et al., 1996b). The tumor cell type specificity for inactivation of either the p16 or p l 5 gene and the mechanism involved are also striking in other types of hematopoietic malignancies. The pattern seen for AML predominates in adult/acute lymphocytic leukemia (ALL), and the sole lesion is solitary hyper-methylation of the p l 5 gene (see Fig. 1). A much more complex situation exists in pediatric ALL, however, especially when one examines tumors as a function of B-versus T-lymphocyte origin. Inactivation of the p l 5 gene

164

Stephen B. Baylin ef of.

continues to be frequent in both tumor types, but now, homozygous deletion, as well as hypermethylation, is a common mechanism (Herman et af., 1997). Furthermore, although involvement of p16 is infrequent in the B-cell tumors, occurring only occasionally and by homozygous deletion, loss of this gene function is a dominant feature of the T-cell tumors. In fact, all of the latter tumors have inactivation of either the p l 5 or p16 gene, and over half have inactivation of both genes. Intriguingly, in the T-cell tumors with lesions involving both genes, the p16 gene is always deleted, and the p l 5 gene is most often hypermethylated (see Fig. 1). Yet additional patterns emerge when lymphomas are examined. Neither the p l 5 nor p16 gene is frequently altered in low-grade non-Hodgkin’s lymphoma (NHL), and hypermethylation of the p16 gene alone was the only molecular event observed (see Fig. 1). In contrast, in high-grade NHL tumors, the p15 gene is still infrequently altered, but more than 80% of these tumors have hypermethylation of the p16 gene (see Fig. 1). Moreover, in Burkitt’s lymphoma, a tumor with a high proliferative index, both the p15 and p16 genes are frequently simultaneously hyperrnethylated (see Fig. l ) , and methylation of one gene or the other is virtually a constant event in this neoplasm. Finally, studies of the p16 and p l 5 genes shed light on one other aspect of hematopoietic malignancies. Chronic niyelogenous leukemia (CML) resembles acute leukemia only in the advanced stage of blast crisis, and may have a different initial cell of origin. We have detected n o involvement of either the p l 6 or p l 5 genes by mutation or hypermethylation in any stage of CML. (Herman e t al., 1997). This finding may be helpful in delineating the different routes of development of the acute and chronic leukemias and how involvement of the cyclin D-Rb pathway differs for these diseases.

5. E-CADHERIN GENE The family of cadherin genes, many located at chromosome 16q, are essential regulators of cell growth and differentiation by their role in promoting homotypic cell-to-cell adhesion (Mareel et al., 1995). The protein that plays a major role in this activity in epithelial cells, E-cadherin (E-cad), is increasingly recognized as a major invasion and metastasis suppressor gene in multiple solid tumor types. This gene is located in a chromosome 16q area that frequently undergoes LOH, especially in breast carcinomas (see, for example, Sato et al., 1990). There is a heterogeneous decrease in expression of this gene in breast, prostate, and other tumors, which is often reported to be most prominent in higher-grade, more poorly differentiated tumors (Graff et al., 1995; Oka et al., 1993; Umbas et al., 1992), and in vitro studies have

DNA Methylation and Neoplasia

165

well documented the increased invasive behavior of cells lacking the E-cad gene and restoration of invasion suppression after gene reinsertion (Vleminckx et al., 1991). The mechanisms underlying loss of E-cad gene expression are unclear because coding region mutations in this gene are uncommon, occurring primarily in gastric carcinomas (Becker et al., 1994) and lobular carcinomas of the breast (Berx et al., 1995; Kanai et al., 1994). Studies in cell culture have suggested that loss of expression may be associated with decreased transcription of the gene (Hennig et al., 1995). Most recently, studies from two groups revealed that this transcriptional change is tightly associated in culture lines of breast, prostate, and colon tumors with hypermethylation of a CpG island spanning the transcription start site of the gene (Graff et a/., 1995; Yoshiura et al., 1995). The hypermethylation change can also be detected in primary breast and prostate cancers (Graff et al., 1995). The criteria outlined in Table I1 for associating hypermethylation with E-cad inactivation have been met (Graff et al., 1995; Yoshiura et al., 1995), and this gene provides another example in which transcriptional silencing is dominant over coding region mutations for loss of suppressor gene function.

6 . ESTROGEN RECEPTOR GENE Lack of estrogen receptor (ER) gene function, usually without evidence of gene mutation, has long been recognized as a critical aspect of loss of hormone responsiveness in breast carcinoma (Clarke et al., 1992; Danforth, 1992). Less well recognized has been the evidence that this gene may have important tumor suppressor properties in multiple cancer types, including breast cancer, in which reinsertion of the gene into ER breast cancer cell lines actually slows growth (Jiang and Jordan, 1992; Zajchowski et al., 1993). A similar response has been seen in culture lines of other tumor types, such as HeLa (Maminta etal., 1991), Chinese hamster ovary (CHO)(Kushner et al., 1990), osteosarcoma (Watts et al., 1989), neuroblastoma (Ma, 1993), and colon cancer cells (Issa et al., 1994). The ER gene is widely expressed outside of the breast and may be an important modulator of cell growth and differentiation in the brain and other tissues (Keefe et al., 1994). We have found a tight association between multiple tumor types and hypermethylation of a 5’ region CpG island in the ER gene (Issa et al., 1994; Issa et al., 1996c; Ottaviano et al., 1994). There is tight correlation for this event with lack of ER gene expression in ER breast cancer cell lines, and the gene can be reactivated by demethylation (Ferguson et al., 1995). Perhaps most intriguing is the association of ER gene hypermethylation with colon and other tumor types. Every colon polyp and carcinoma examined has this change, and insertion of an ER gene into colon carcinoma cells profoundly

166

Stephen B. Baylin ef nl.

slows their growth ( h a et al.. 1994). This hypermethylation of the ER gene is a conserved event in mice, rats, and humans for lung tumors (Issa et al., 1996a) and leukemias (Issa et al., 1 9 9 6 ~ )Furthermore, . in humans, a tight relation exists between advancing age and appearance of hypermethylated ER gene alleles in normal colonic mucosa (Issa et al., 1994). This association may provide a pivotal link between the well-recognized association of aging and the increasing incidence of colon neoplasia, discussed in more detail later.

7. HYPERMETHYLATED-IN-CANCER G E N E (HIC- I ) HIC-1 encodes for a protein with features of the ZiN (zinc finger N-terminal) or Poz (Podzinc finger) subfamily of zinc finger transcription factors (Albagli et al., 1995; Bardwell and Treisman, 1994). HICl was cloned by our group (Wales et al., 1995) as an exercise to determine whether detection of hypermethylated CpG islands might help identify candidate tumor suppressor genes in a chromosome region with a high frequency of LOH. The 5' flanking and entire coding regions of HICl are embedded in an unusually dense CpG island in a region of chromosome 1 7 ~ 1 3 . 3that frequently is deleted independently of the more centromerically located p53 gene in t u mors such as brain, breast, and other cancers (Makos et al., 1992, 1993a, 1993b; Wales et al., 1995), and loss of expression correlates with this status in cell lines for breast, colon, and lung cancers and primary brain tumors (Wales et al., 1995). HICl remains a candidate tumor suppressor gene, but several features make such a role attractive. First, at least two other genes in the Zin gene family, most of which have transcriptional repression activity (Albagli et al., 1995; Bardwell and Treisman, 1994), were recognized by virtue of their inclusion in translocations in human tumors (Chen et al., 1995; Kerchaert et al., 1993; Migliazza et al., 1995; Weis et al., 1994). Second, attempts to reexpress H I C l in cell lines of brain, breast, and colon cancer have resulted in reduced selection o f G418-resistant clones and general failure of surviving clones to express the gene (Wales et al., 1995). Finally, the 5' region of HICl contains a putative p53-binding site, and overexpression of the p53 gene can markedly augment HICl expression in at least one colon cancer cell line (Wales et al., 1995). Thus, in some cells, H l C l could function in the p53 pathway, and future studies should clarify the precise contribution of this gene to neoplasia.

8. OTHER GENES HYPERMETHYLATED IN CANCER There is a growing list of genes (see Table 11) that are hypermethylated in cancer, although their precise role in tumorigenesis is less well defined than

D N A Methylation and Neoplasia

I67

for those genes discussed in the previous section. The aberrant methylation patterns for several of these may turn out to provide biomarker systems, even independent of ascertaining the functional consequences of inactivation of the genes during tumor progression. A prime example is the glutathione-Stransferase-n (GST-n) gene, which appears to be specifically hypermethylated, and transcriptionally silenced, in virtually every primary prostate cancer examined (Lee et al., 1994).This methylation change appears to separate benign from malignant prostate lesions and is perhaps the most constant molecular change found for this common form of cancer.

V. MECHANISMS UNDERLYING THE DNA METHYLATION CHANGES IN NEOPLASTIC CELLS A growing body of data now indicates a potentially critical role for abnormalities of DNA methylation, and especially aberrant methylation of gene promoter regions in tumorigenesis. Therefore, it is important to understand the mechanisms responsible for these changes in neoplastic cells. O u r understanding of this biology is just beginning. Research into the mechanisms involved has been increasing, however, and as discussed in the following sections, distinct clues to key components of the process are emerging.

A. Role of DNA-Methytransferase Activity Changes

in Neoplasia I . INCREASED DNA-METHYLTRANSFERTASE ACTIVITY IN NEOPLASTIC CELLS

As noted previously, one component of the DNA methylation imbalance in neoplastic cells is an apparent increase in DNA-MTase activity. Initial data for this change emerged from the studies of Kautiainen and Jones (1986), who found multifold higher levels of DNA-MTase activity in multiple types of transformed vs. normal cultured murine and human cells. This was a surprising finding considering that overall genomic levels of DNA methylation were often reduced in the same lines of neoplastic cells. The authors concluded that there might be hot spots of increased methylation, resulting from the increased DNA-MTase activity (Kautiainen and Jones, 1986). This concept, of course, would fit well with the findings for increased CpG island methylation in tumor cells, and this possibility is further addressed later. These early data for tumor cell cultures were confirmed in our later work, in

3.0

-

I I

I I

I

I

: I

. 2,.I c

i2S

gt 2.0

.

I I I I I

--

6

I 0

I

1

I I I

9)

.t, Y

: 1

A

1.0

I

0

' i

I I

\ : .

I

.

\

0

a

d

I I I I I I I I

I I

.

I

I

I

Senescing FibroMlPg

0

I

:\

:

Y0E.g

.

\

J

:*

I

.

I

i

I

Fibroblasts

*

1:

0

I I I

I

/ *

\: - 1

>

-

gCrkh .

I

is d

/

I I I

E

I I

I

. I

6'

I

Senescence

Pre-Crisi Fibroblasts

Immortnlized Fibroblasts

Fig. 2 Summary of experimental data for relations of DNA-MTase activity to fibroblast aging and infection with SV40. The graph depicts the fold changes in DNA-MTase activity relative to 1.O as the level in normal young cells (about passages 0 to 2 5 ) , as a function of cell passage for fetal human fibroblasts (young, about passage 25; senescent, about passage 50) and the same cells infected with SV40 virus at about passage 30 (infected cells in the precrisis stage from about passage 50 to 70; crisis phase cells from about passage 70 to 90; immortalized cells from about passages 130 to 200). The data are from Vertino et al., 1994, as discussed in the text, and relate closely to findings for overall DNA methylation by others in similar studies (Matsumura et al., 1989a, 1989b).

DNA Methylation and Neoplasia

I69

which DNA-MTase gene mRNA levels were found to be substantially higher in cultured cancer cells of several types than in cultured nontumorigenic cells (El-Deiry et al., 1991). Matsumura and coworkers (1989a, 1989b) have noted a fundamental difference in the capacity of normal versus transformed human fibroblasts to methylate DNA, which fits well with the previous data. In nontransformed fibroblasts in culture, overall levels of genomic methylation fall as the cells approach a final state of senescence. When young cells are infected with the SV40 virus, however, they bypass the normal cell senescence point and maintain an increased life span during which overall levels of DNA methylation d o not decrease. Moreover, in these pretransformed cells and in those that actually survive a later crisis period and emerge as immortalized cells, it is more difficult to inhibit active DNA methylation with the drug 5-AZA-cytidine than in normal fibroblasts (Matsumura et al., 1989b). More recently, we have mapped the expression of DNA-MTase mRNA and activity in this same experimental system and revealed a pattern, summarized in Figure 2, which well fits with the previous data (Vertino et al., 1994).In normal senescing fibroblasts, levels of DNA-MTase gene expression fall to 5 0 % of those seen in young fibroblasts. When the cells are infected with SV40, however, these levels fail to fall because the cells bypass the senescence checkpoint and enter an extended replication period before entering cell crisis. At crisis, levels of DNA-MTase expression are actually as high or higher than those in young normal fibroblasts, and these high levels are maintained in those few cells that escape crisis and become immortalized (Vertino et al., 1994). An association between increased DNA-MTase activity and early events leading toward the transformed state has been further reinforced by two recent studies. First, nontumorigenic immortalized hybrid cells, which have been produced from fusion of normal human fibroblasts with malignant fibrosarcoma cells that have a seven-fold increased DNA-MTase activity, retain the high levels of the parent tumor cells (Kuerbitz and Baylin, 1996). This finding, in addition t o those from SV40 work described earlier, indicate that increases in DNA-MTase activity can occur before appearance of the fully transformed phenotype. Second, in a strain of mice genetically susceptible to lung carcinogenesis, but not in a resistant strain, exposure to the major tobacco carcinogen, NNK, leads to a three-fold increase in DNA-MTase activity, specifically in the lung type I1 pneumocytes that later give rise to adenocarcinomas (Belinsky et al., 1996).The levels initially peak within 3 days, return to baseline by 7 days, and then increase again several weeks later in the early benign lung tumors and progressively thereafter as these form carcinomas (Belinsky et al., 1996). These findings reinforce the concept that DNA-MTase gene expression begins early in, and is thus in a position to contribute to, tumorigenesis. Findings for DNA-MTase gene expression in natural settings for human

170

S t e p h e n B. Baylin el nl

tumor progression have paralleled some of the previously described findings for experimental systems. In initial studies of expression of the human DNAMTase gene in the colon, using reverse transcriptase polymerase chain reaction (RT-PCR) analyses, we found evidence for progressively higher gene expression from normal mucosa to premalignant adenomas to frank carcinomas (El-Deiry et ul., 1991). Additionally, in this study, we found higher DNA-MTase transcript levels in normal-appearing mucosa from patients with colon neoplasms than from patients without these tumors (El-Deiry et al., 1991). The PCR procedures in this first study may have led to overestimates of the differences between each progression stage (15-fold for cancer versus normal mucosa from the same patient). Other workers more recently have shown 1.5- to +fold increases in mRNA for DNA-MTase in colon cancer versus adjacent colon (Lee et al., 1996; Schmutte et al., 1996). Our own subsequent study of DNA-MTase activity in these same settings again revealed increasing levels from normal mucosa to polyps to carcinoma and an average increase of about three-fold in colon cancers versus normal mucosa from the same patient (Issa et d.,1993). The factors underlying increased DNA-MTase activity in tumor progression have not been clarified. As mentioned earlier, there are reports that control regions within the murine DNA-MTase gene mediate increased transcription by response to activation of the ras signal transduction pathway (MacLeod et al., 1995). The region of the DNA-MTase gene identified as the rus responsive area, however, initially thought to be in the 5' flanking promoter, actually lies far downstream of the transcription start site and has weak basal transcriptional activity (Tucker etal., 1996b; Yoder et al., 1996). Thus, further studies are required to understand fully the role for the ?as pathway, or for other pathways, in modulating DNA-MTase activity during the various steps in cell transformation. As mentioned earlier, DNA-MTase activity is regulated as a function of DNA synthesis and cell cycle activity, and some have considered that this relation might be instrumental for producing any differences seen between normal and neoplastic cells. In fact, in a recent study, Lee and colleagues (1996) challenged, on this basis, the significance of the reported increases in DNAMTase gene transcripts in human colon tumor progression. When they use RNAase protection assays for DNA-MTase expression, they found only a 1.8- to 2.5-fold increase between normal colon and colon tumors (Lee etal., 1996). Furthermore, these differences were abolished if they normalized the data for expression of histone H4, which is a measure of S-phase activity of cells. These authors concluded that any differences between the tumor and normal cells are, then, solely due to the increased numbers of proliferating cells in the cancer (Lee et al., 1996). Although the results of the above study must be taken carefully into account, we think there is much evidence that such a simple relation between

DNA Methylation and Neoplasia

171

increased numbers of proliferating cells in tumor and normal tissue cannot account for much experimental data linking increased DNA-MTase and neoplasia. First, in our hands and others (Kautiainen and Jones, 1986), both in cell culture systems and in natural tumors, the fold increases for DNA-MTase activity in neoplastic cells are often too high for simple attribution to increased cell proliferation. In fact, in many tumors, such as colon cancer, the proliferative index is only marginally, or not at all, increased compared with corresponding normal cells or tissues (El-Deiry et al., 1991; Issa et al., 1993). Furthermore, when we employed in situ hybridization techniques for visualization of DNA-MTase mRNA in the murine lung cancer model discussed previously, virtually every type I1 cell, from the stages of hyperplasia to frank carcinoma, had markedly increased staining compared with normal cells (Belinsky et al., 1996). No previous data would predict, at a given point in time, such universal participation of entire cell populations in the modestly increased proliferative indices found for neoplastic tissue. Indeed, we found n o such measurable proliferative increase in the type I1 cells that had responded early to NNK exposure with a n increase in DNA-MTase activity (Belinsky et al., 1996). Given all of the above data, can some of the controversies about the increase of DNA-MTase in neoplasia be reconciled? It is well possible that entrance of cells from a resting Go stage into cell cycle could be fundamental t o initiating increased DNA-MTase activity in the process of neoplastic evolution. As discussed previously, however, in immortalized and transformed cells, this increase appears to persist, and even accumulate, when such cells are not engaged in cycling activity and S-phase-mediated DNA synthesis. This fact may underlie the accumulating evidence, discussed later, that the increases in DNA-MTase gene expression in tumorigenesis have a profound effect on DNA and play a fundamental role in the neoplastic process rather than serving as a simple index of cell proliferation.

2. WHAT IS THE EVIDENCE THAT INCREASES IN DNA-METHYLTRANSFERASE PLAY A FUNDAMENTAL ROLE IN TUMOR PROGRESSION? Although, as presented previously, there is a growing body of evidence associating increased DNA-MTase activity with the neoplastic state, it remains to be proved formally that this change is directly responsible for mediating specific steps in tumor progression. A major conundrum that must be resolved is why, frequently, the cancer cell genome simultaneously harbors overall hypomethylation and increased DNA-MTase activity. Several possibilities have been entertained in this regard. First, there could be a decreased efficiency of maintenance methylation because of chromatin changes that block the normal access of DNA-MTase to DNA. In this regard, the increased

172

Stephen B. Baylin el al.

DNA-MTase might actually be stimulated, early in tumorigenesis, by the decreased overall genomic methylation, but the increase is insufficient to restore normal sites of methylation. Support for such a concept comes from the fact that folate deficiency induces an evolution of hepatic carcinogenesis during which both DNA hypomethylation and increased DNA-MTase activity occur very early (Christman et al., 1993; Pogribny et al., 1995). Alternatively, there is some evidence that active mechanisms are increased in tumor cells t o demethylate DNA directly, as opposed to loss of methyl groups resulting from simple decreases in maintenance methylation (Syzf et al., 1995). In this regard, an RNA species with apparent demethylation capacity has been reported (Weiss et al., 1996) and could play a role through increased activity in neoplastic cells. These possibilities and others remain to be verified, but for now, the explanation for the coexisting DNA-MTase increases, and genomic hypomethylation in tumor cells remains a mystery. Despite this dilemma, several studies suggest that the levels of increased DNA-MTase activity found in neoplastic cells can profoundly affect DNA and could well be involved with regional hypermethylation. Multiple lines of experimental evidence indicate that acute, but modest, increases in DNAMTase activity can transiently increase overall DNA methylation. First, twoto three-fold increases of DNA-MTase activity in murine NIH 3T3 cells expressing an exogenous DNA-MTase gene are accompanied by significant increases in overall DNA methylation (Wu et al., 1993). Second, in the previously discussed murine model for carcinogen-induced lung carcinoma, increased overall DNA methylation accompanies the transient increases in type I1 pneumocyte DNA-MTase activity induced by acute exposure to NNK (Belinsky et al., 1996). Third, studies from the Jaenisch laboratory in mouse cells homozygously deleted for the endogenous DNA-MTase gene (Tucker et al., 1996b) again reflect significant effects of modestly increasing DNA-MTase activity. In these cells, constitutive expression of the originally characterized murine cDNA for DNA-MTase from a strong exogenous promoter is poorly tolerated. In those few surviving clones in which the cDNA was expressed, however, levels of DNA-MTase activity much lower than those seen in wild-type cells were sufficient to effect restoration of much of the methylation normally seen in endogenous repeat viral sequences (Tucker et al., 199613). In addition to these studies, there are also experimental data indicating how the more long-term effects of increased DNA-MTase activity may profoundly effect DNA and influence tumor progression. The work suggests that during tumor progression, transformed cell types could be selected that have only a modest increase for DNA-MTase expression, whereas cells with higher increases d o not survive. For example, expression of a de novo bacterial cytosine DNA-MTase in murine cells is toxic except when low levels are produced from a leaky, inducible promoter or from a constitutive pro-

DNA Methylation and Neoplasia

I73

moter in occasional selected cell clones (Wu et al., 1996). Similarly, constitutive overexpression of the human DNA-MTase gene is not well tolerated in human cells (Vertino et al., 1996). Also, even though it is not particularly difficult to select for murine NIH 3T3 cells expressing a constitutively regulated exogenous murine DNA-MTase gene cDNA, all clones obtained had not more than two to three times the activity of the parent cells (Wu et al., 1993). This increase is comparable to those discussed previously for colon carcinoma and other types of cancer cells. Thus, increases in DNA-MTase activity of more than three-fold may lead to cell death, whereas cells with more modest DNA-MTase gene overexpression are selected. Importantly, in the previously discussed murine systems, these latter cells have become tumorigenic and have evidence for either overall or regional increases in DNA methylation (Wu et al., 1993, 1996; Vertino et al., 1996). These and other discussed overall relations between various levels of DNA methylation and experimental manipulation of DNA-MTase activity are all summarized in Figure 3. The associations depicted may provide insight into how DNA methylation changes are of fundamental importance for tumor progression. Relative to the data in Figure 3, there is an important caveat for the biologic importance of the studies suggesting that cells may only tolerate modest increases in DNA-MTase activity. In the previously mentioned study of Tucker and colleagues (1996b), although expression of the short original cDNA for murine DNA-MTase was not tolerated by cells, expression of the longer DNA-MTase gene, now known to encode for the entire DNA-MTase protein of humans and mice (Tucker et al., 1996b; Yoder et al., 1996), was well tolerated and restored wild-type DNA-MTase activity and DNA methylation levels. The authors attribute this tolerance to inclusion of the new 5’ sequences in the protein (Tucker et al., 1996b). It is critical to note, however, that the longer DNA-MTase gene was inserted in a genomic fragment rather than as a cDNA driven from an exogenous constitutive promoter. In this setting, expression was achieved solely from an endogenous promoter located either in the 5’ flanking region of the genomic piece or in the homologous DNA-MTase site where this fragment was documented to have inserted (Tucker et al., 1996b). Thus, tolerance for the gene could well be due to the lower levels of expression for this genomic construct than found in those achieved with the constitutively driven shorter cDNA. If, as we are suggesting, increased DNA-MTase activity contributes directly to cell transformation, what are the underlying mechanisms? Although this remains to be determined, experimental evidence is accruing to suggest that this change could contribute to the CpG island hypermethylation associated with transcriptional inactivation of key tumor suppressor genes. First, when human fibroblasts are infected with SV40, and DNA-MTase activity levels fail to decrease as cells bypass the senescence checkpoint, aberrant CpG island methylation begins to appear by the time cells enter the crisis pe-

DNA Methylation and Neoplasia

I75

riod (Vertino et al., 1994). This change is further apparent in those cells that survive crisis, and the immortalized progeny have multiple densely methylated CpG islands (Vertino et al., 1994). Second, in SV40-transformed human fibroblasts engineered to overexpress an exogenous human DNA-MTase gene, selected C p C islands become progressively hypermethylated over successive numbers of cell passage (Vertino et al., 1996). In general, the islands affected were those that showed some initial hypermethylation in the parent transformed cells. Thus, the single human DNA-MTase enzyme characterized could, during the prolonged course of neoplastic evolution, be permissive for both aberrant de nova methylation events and their perpetuation and extension. Alternatively, the initial de nova methylation events could be mediated, during tumorigenesis, by another DNA-MTase, such as that proposed by the Jaenisch group (Lei et d., 1996; Tucker et al., 1996a). This methylation could then be readily extended by the increases in the well-characterized DNA-MTase because this enzyme is known to perform de N O V O methylation more efficiently in regions of DNA in which fully methylated CpC sites already exist (Tollefsbol and Hutchison, 1995). All of the data linking increased DNA-MTase activity with tumor progression may underly findings from the Jaenisch laboratory, which have focused much recent attention on the role of DNA methylation changes in cancer. These workers performed a genetic experiment in which phenotypically normal mice heterozygous for deletion of the DNA-MTase gene and having 5 0 % of the wild-type levels of DNA-MTase activity, were bred with mice harboring a mutant allele of the APC gene (Laird et al., 1995). This latter mutation, when combined with somatic loss of the other APC allele, results in progressive formation of adenomas throughout the gastrointestinal tract

-

Fig. 3 A model summary of experimental data, discussed in the text, relating relative levels of DNA-MTase activity to cell phenotype, in systems manipulated to overexpress o r undcrexpress the DNA-MTase gene. The graph assumes a valuc of 1.0 for a normal parent cell and relative values for experimental maneuvers that raise o r lower steady-state enzyme activity from this value. As discussed in detail in the text, levels below about 10 to 15% are shown to be associated with cell death, such as is seen in mouse embryos homozygous for knockout of the DNA-MTase gene (Li et a/., 1992), or after attempts to differentiate cultured embryonic stein cells completely lacking the gene (Tucker et d,1996a). Levels from about 15 to 2 0 % arc depicted associated with chromosome instability changes, as reported in studies of cells cultured in 5-AZA-cytidine (see text). Cells with levels from 25 to 100% are assumed to grow with a normal phenotype (Li et a/.. 1992; Tucker e l al., 1996a). Manipulations that chronically increase DNA-MTase activity up to 3-fold are associated with cell transformation (Wu et nl., 1993, 1996), whereas increases above this range are highly correlated with cell death (Vertino et ul., 1996; Wu et al., 1996). The asterisk refers to the fact that some rare ccll subpopulations survive levels higher than threc-fold and have CpG island hypermethylation and/or a transformed phenotype (Vertino et al., 1996).

I76

Stephen B. Baylin el al.

of newborn animals. Offspring mice harboring both the APC and DNAMTase gene mutations had a 50% reduction of intestinal adenomas during a 6-month period following birth, and these lesions could be essentially eliminated if treatment with the demethylating agent, 5-deoxy-AZA-cytidine, was combined with the DNA-MTase allelic deletion. The mechanisms underlying this striking reduction of tumor progression are not yet apparent, although two leading possibilities have received much discussion (Hopkin, 1995; Laird et al., 1995). The Jaenisch group initially favored the proposal that the lowering of DNA-MTase activity reduces rates of DNA mutations during adenoma formation in the newborn animals (Hopkin, 1995; Laird et al., 1995). The rationale underlying this theory is that the reduced levels of DNA-MTase activity in the heterozygous knockout mice might result in appearance of less methylated C p C sites, which could undergo deamination to thymidines. Also, in vitro, a bacterial DNAMTase has been shown to directly cause deamination of cytosine to form uracil when levels of the methyl donor, S-adenosylmethionine, are extremely limiting (Shen et al., 1992). When repaired, these uracils would be converted to thymidines. Such low levels of S-adenosylmethionine, however, have not been found in vivo (Schmutte et al., 1996), and subsequent studies of the APC and DNA-MTase heterozygous animals have shown no evidence for decreased mutational rates (Jackson-Grusby et al., 1997). The second leading possibility for the results from the Jaenisch laboratory is that lowering of DNA-MTase activity might reduce rates of accrual for aberrantly methylated CpG islands and thus reduce the rate of tumor suppressor gene inactivation over the course of adenoma formation (Hopkin, 1995; Laird et al., 1995). We think this is the most attractive possibility, but one that must await verification from more formal testing in the Jaenisch and other models. In summary, a growing body of evidence is pointing to an important role for increased DNA-MTase activity in tumorigenesis. Much work remains to document this possibility and to elucidate the exact mechanisms involved. The evidence has been crystallized by the findings of the Jaenisch group, but the generality of their findings must be studied in other systems. The intestinal tract of the mouse undergoes profound differentiation changes 2 weeks after birth (Crane, 1968; Koldovsky, 1969), and some of the effects of lowering DNA-MTase on adenoma formation in this setting might be linked to alterations in this process. Studies of adult carcinogenesis models will then be particularly important before the results from the Jaenisch group can be placed into perspective for neoplasia in general. The next few years should see much research activity regarding the role of DNA-MTase in neoplasia, and the possibility of targeting this enzyme as an anticancer strategy is discussed at the end of this chapter.

DNA Methylation and Neoplasia

177

B. What Additional Mechanisms Underlie the Aberrant

Methylation of CpG Islands in Neoplasia? Among the methylation changes that occur in the evolution of neoplasia, hypermethylation of CpG islands is now best functionally linked to tumor progression, so further discussion of the mechanisms underlying this process seem warranted. Although, as noted previously, increases in DNA-MTase activity may play a critical permissive role in the appearance of this change, it is likely that other factors are equally as important, or in a gene-specific manner, more important. These potential events are considered in this section. I . DOES CpG ISLAND HYPERMETHYLATION TRULY REFLECT A DE NOVO PROCESS IN TUMOR PROGRESSION?

For classic gene mutations in noninherited neoplasms, one usually does not question whether the specific mutation preexisted in the normal cells from which the neoplasm arose. The special nature of DNA hypermethylation as an inactivating event, however, particularly the potential reversibility of the process, raises the issue of whether some of the patterns now associated with neoplasia could reflect clonal expansion of cells normally harboring such methylation. For example, could stem cells in selected tissues, which may be the initial targets of cell transformation, normally contain densely methylated CpG islands in the promoter regions of genes that are subsequently unmethylated in all the more mature progeny of these precursor cells? Such a mechanism might help maintain the associated gene in a silent state before required activation for steps in cell differentiation. We have addressed this possibility for several genes by using a sensitive new PCR technique that can monitor, in DNA from small numbers of cells, the methylation status of minor allele populations in complex cellular backgrounds. This procedure, which we have termed methylation-specific PCR (MSP) (Herman et al., 1996a), depends on initial exposure of DNA to sodium bisulfite under conditions in which all unmethylated cytosines are deaminated to form uracil, but methylated cytosines are resistant to this change (Frommer et al., 1992; Myohanen et al., 1994).During subsequent PCR amplification of the treated DNA, the uracils are converted to thymidine. The PCR primers are designed to detect these sequence differences and specifically to amplify sequences that did or did not initially have selected CpG sites methylated (Herman et al., 1996a). This technique can detect the methylation of 1 among 1000 unmethylated alleles (Herman et al., 1996a). Using MSP analysis for study of the p l 5 and p16 genes, which are frequently hypermethylated in multiple types of hematopoietic malignancies, we did not detect this change in highly purified stem cell populations from the bone mar-

178

S t e p h e n B. Baylin P I a/.

row of normal individuals (Herman et al., 1997). Similarly, the MSP technique detected no hypermethylated alleles for the E-cad gene CpG island in purified normal breast epithelial cells (Graff et al., 1997), even though hypermethylation of this gene is a frequent event in breast carcinomas. We concluded from these initial studies that there are tumor suppressor genes for which C p C islands truly become de nouo methylated during the process of tumor progression.

2. AGING IS ASSOCIATED WITH PROMOTER REGION CpG ISLAND METHYLATION IN SELECTED G E N E S In contrast to the previous findings for several tumor suppressor genes, we have detected a class of genes that appear to have hypermethylated patterns in neoplasia, which may reflect the normal cell populations, giving rise to the involved tumors. These genes may provide one fundamental link between cancer and perhaps the most important risk factor for the development of most human neoplasms, aging. The role of aging in cancer has generally been attributed to cumulative exposure to carcinogens over time as well as to the long period required for the development of a clinically detectable tumor in uiuo. Physiologic aging, however, is accompanied by profound changes in gene expression that cannot be explained on the basis of accumulated mutations (for review, Holliday, 1991; Smith and Pereira-Smith, 1996). Because, as discussed previously, DNA methylation changes have great potential to affect gene expression by epigenetic mechanisms, altered DNA methylation patterns have long been suspected to play a role in aging and the associated increased incidence of neoplasia (Holliday, 1987). Many of the patterns observed fit well with components of the methylation imbalance observed in neoplasia, and hypermethylation changes in gene promoter regions are proving most intriguing. First, both overall genomic hypomethylation and hypomethylation of specific genes have been associated with aging cells. Aging human fibroblasts in culture have progressive loss of methylcytosines as they approach senescence (Holliday, 1986; Wilson and Jones, 1983), and this decrement has even been proposed as a “counting mechanism” used by the cells to monitor the number of divisions allotted over their lifetime (Holliday, 1990b). It is has been suspected that this hypomethylation could play a role in altered gene expression associated with aging because this change has been reported for genes such as c-myc (Ono et al., 1990) and actin (Slagboom et al., 1990). The position of the change relative to gene regulatory regions, however, has generally not been well correlated, and the role of methylcytosine loss in aging-associated changes in gene expression remains unknown. Second, regional areas of hypermethylation have also been associated with aging cells (Schmookler Reis and Goldstein, 1982; Swisshelm et al., 1990; Uehara et al., 1989), although until recently, the implications of this change

DNA Methylation and Neoplasia

I79

during the aging process and its link to cancer have not been clear. Hints that these changes could help predispose cells to neoplasia had emerged from experimental systems. As mentioned earlier, Matsumura and colleagues (1989a, 1989b) have observed that induction of human fibroblasts by SV40 to escape the senescence checkpoint is associated with abrogation of the methylcytosine loss seen during senescence of normal cells. Subsequently, we have shown that this SV40 effect is paralleled by the emergence of hypermethylation changes in selected promoter region CpG islands (Vertino eta/., 1994). Most important, however, has been the association, in the promoter regions of specific genes, of hypermethylation patterns common to aging cells and neoplasms. Our initial observations of these associations stemmed from the serendipitous observation that, in colon neoplasms, while all benign and malignant tumors examined displayed hypermethylation of the ER gene promoter and a decreased expression of this gene relative to normal colon, distinct partial methylation could also be observed in adjacent, normal-appearing colonic tissue (Issa et a/., 1994). O n further investigation, we determined that this methylation in normal colon was an age-related event, almost undetectable in young individuals and progressively more prominent in older individuals (Issa et ~ l . ,1994). Because 100% of colonic tumors examined had extensive ER methylation, including small adenomatous polyps, and the aging changes are greatest in colon regions where tumor incidence is highest (Issa et al., 1994), we hypothesized that the cell of origin of these tumors was one in which the ER gene was hypermethylated. Based on these data, and o n the fact that introduction of the ER gene into colon cancer cells resulted in marked growth suppression (Issa et al., 1994), it may be envisioned that the aging-associated methylation of the ER gene provides a selective growth advantage to subpopulations of colon mucosa cells that harbor this change. We discuss later how such cells could constitute the field defect that may predispose groups of cells to the most common forms of colonic neoplasia. The prevalence of these types of age-related gene methylation changes remains to be determined. Our initial observations from studying randomly selected genes suggest that such hypermethylation is a relatively rare event. Genes such as p16, which are hypermethylated in 20 to 4 0 % of all colonic tumors, are not detected by MSP to have any methylation in normal colon from older people (Herman, unpublished data). Nevertheless, the age-related events seen for the ER gene in normal colon d o occur for other genes, and the observations for one of these, the insulin-like growth factor-2 (IGF2) gene, possibly even hold clues to the events that underlie the process. The IGF2 gene is a fetal growth factor that is imprinted in rodents and in humans, and is silent on the maternal allele and expressed exclusively from the paternal allele (Barlow, 1995). This gene is highly expressed in several forms of adult tumors, including colon cancer, and this growth factor is

I80

Stephen

B. Baylin el al.

thought to play a potential role in the evolution of these neoplasms (Christofori et al., 1994). In humans, the IGF2 gene, located on chromosome 1 lp, has four promoters (P1 to P4), with four different leader exons (for review, Sussenbach et al., 1993). P2 through P4, which are relatively close to each other, are contained in a CpG island and are coordinately imprinted. By contrast, P1, which is located more than 20 k b upstream of P2, is not contained in a CpG island and is expressed biallelically in human tissues, suggesting that it escapes imprinting (Vu and Hoffman, 1994). In adult humans, the liver is the major source of IGF2 production, and P1 is preferentially used in this tissue. The imprinted status of IGF2, in part, may be determined by interaction with enhancers for the H19 gene, which resides nearby on chromosome 1l p . The H19 gene is reciprocally imprinted to IGM and is thus expressed exclusively from the maternal chromosome (for review, Barlow, 1995). The gene also encodes for an RNA species only, which could play a key role in imprinting of IGF2 (Barlow, 1995). We hypothesized that in addition to the previously proposed regulation, promoter methylation might also play a role in the allele-specific expression of IGF2. We investigated methylation around P2 to P4 in various tissues, in aging colon, and in various neoplastic tissues and found that, in young people, the P2 to P4 CpG island is methylated only on the silenced maternal allele (Issa et al., 199613).Strikingly, P2 to P4 methylation increases progressively with age, and in the colon, this increased methylation can be accounted for, in part, by methylation of the CpG island on the previously unmethylated paternal allele. Most common adult neoplasms studied had even more extensive P2 to P4 methylation, and in tumor cell lines, hypermethylation at P2 to P4 resulted in a marked reduction of IGF2 transcription from P3. By contrast, P l y the upstream promoter that is not contained in a CpG island, was used in most cell lines examined (Issa et al., 1996b). In this setting, we then have the ironic situation in which promoter hypermethylation is associated with possible overexpression of a gene but in which a promoter switch is associated with this event. We believe that these age-related methylation changes in IGF2 are particularly important not only for their potential to enhance tumor growth but also because they provide a potential clue to the mechanism of age-related hypermethylation. The data suggest that areas of parentally determined allele-specific methylation may be at special risk for this phenomenon because, during aging, methylation could spread from the initially methylated allele to the opposite allele. Such spreading of methylation across alleles previously has been suggested to account for a process in plants termed cosuppression (Flavell, 1994), in which a transgene, which has become methylated and inactivated, results in age-dependent methylation and inactivation of the homologous endogenous gene. To explain this phenomenon, it has been proposed that the methylated transgene and the endogenous gene pair, transiently, by homologous recombination and that DNA-MTase recognizes the paired strands as hemimethylated DNA and spreads methylation from the

DNA Methylation and Neoplasia

181

transgene to the endogenous gene. Such homologous pairing has been observed at imprinted loci in humans (Lasalle and LaLande, 1996). In addition, methylation spread across alleles has now been shown to occur in Ascobolus sp., in which it appears to be mechanistically related to homologous recombination (Colot et al., 1996).Interestingly, homology-dependent spreading of methylation with age may not always require prior methylation of one allele. Many human genes have pseudogenes which are inactive and usually hypermethylated. If these pseudogenes share significant homology to the 5’ region of an expressed gene, then one can envision homologous pairing and spreading of methylation during the aging process.

3. D O OTHER RISK FACTORS FOR TUMOR DEVELOPMENT PLAY A ROLE IN C p C ISLAND HYPERMETHYLATION? In addition to aging, many other events contribute directly to tumor development. Environmental exposure and specific carcinogens are perhaps the most prominent of these. One group has shown that the metal carcinogen, nickel, which localizes to the cell nucleus, can evoke reversible tumorigenic stimuli that are apparently associated with hypermethylation changes surrounding an inserted DNA construct containing a selectable marker (Lee et al., 1995). These authors postulated that the metal evokes an alteration in chromatin structure associated with the methylation change and that this is a model for how carcinogens might lead to inactivation of certain tumor suppressor genes. In keeping with this model, our studies in collaboration with the Belinsky laboratory have shown that different carcinogens evoke strikingly different patterns of ER gene hypermethylation in murine and rodent models of lung carcinogenesis (Issa et al., 1996a). In this setting, the major tobacco metabolite, NNK, causes tumors with a low incidence of ER gene hypermethylation, whereas this change is virtually universal in the same tumor type induced by irradiation or when arising spontaneously in the animals. In contrast, in rats, there is a high incidence of p16 gene hypermethylation in the NNK-induced tumors (Swafford et al., 1997). Thus, different carcinogens may evoke specific changes in different regions of the genome, which may play a major role in the hypermethylation changes under discussion. Likely, defining these interactions will become a n active area of research during the next few years. 4. ARE THERE MUTATIONAL CHANGES IN G E N E PROMOTER REGIONS THAT UNDERLY CpG ISLAN D HYPE RM ET HY LATlO N ? The aberrant promoter area CpG island methylation changes in tumor suppressor genes occur without concomitant coding region mutations in the involved genes. However, could it be that mutations in the promoter regions

I82

S t e p h e n B. Baylin el nl.

themselves might be responsible for structural changes in DNA that incite the aberrant methylation patterns? If so, then the entire process would evolve from a genetic change. This is not an easy question to dissect. If acting through long-range chromatin changes, the mutations involved could be located at some distance from the CpG island regions that actually become aberrantly methylated. What can be said is that the CpG island regions that directly harbor the methylation changes in the VHL and p16 tumor suppressor genes contain no mutations in tumor cells (Herman et al., 1994, 1996a). These studies must be expanded before DNA mutational changes can be ruled o u t as a fundamental cause of aberrant methylation in cancer. As mentioned previously, areas of unusual DNA structure appear to be preferentially methylated by DNA-MTase (Christman et al., 1995; Smith et a[., 1991), and this may hold clues for the aberrant patterns seen in cancer. Findings for other diseases stress the importance of such investigations and further suggest that concentration on repeat sequences may prove rewarding. For example, the inherited mental retardation syndrome, fragile X, is caused by abnormal expansion of a promoter region triplet repeat. This event clearly precedes aberrant methylation of the CpG island containing the repeat sequences, and the disease phenotype results from associated loss of expression of the involved gene (Bell et al., 1991; Pieretti et al., 1991). 5. WHAT LOCAL FACTORS ARE ASSOCIATED WITH PROTECTION O F CpC ISLANDS IN NORMAL CELLS, A N D ARE THESE ALTERED IN NEOPLASTIC CELLS?

One of the most intriguing questions in the biology of DNA methylation is, What normally protects CpG islands from methylation? As previously discussed, we have provided experimental evidence that prolonged exposure o f cells to increased DNA-MTase activity can cause initiation and/or spreading of CpG island methylation (Vertino et al., 1996). In the fibroblast model used, however, only selected regions showed this change, and some of the tumor suppressor genes hypermethylated in other cell types, such as VHL. and p16, did not become hypcrmethylated (Vertino et al., 1996). Although the DNA-MTase increases may be permissive for the CpC island events, local factors that normally maintain islands free of methylation could become disrupted, on a cell type- and gene-specific basis, in neoplasia. Elucidation of such factors in normal cells is a priority of ongoing research, although what little is known has been gleaned largely from studies of one gene, APRT, in hamsters and mice. Three laboratories, using gene insertion studies in cell culture and/or transgenic mice, have ascertained that three consensus and one nonconsensus Sp1 binding sites located just 5’ to the transcription start site of the APRT gene are important for protecting a typical CpG island in the region from becoming niethylated (Brandeis et d., 1994; Macleod et d., 1994; Mumma-

Fig. 4 A model for the anatomy of a typical human gene 5’ region CpG island and the methylation status in normal and neoplastic cells. Data represent average structural content and distances compiled from analyses of database sequences for the human VHL, E-cad, APRT, GST (glutathione-S-transferase)-K, and TIMP (tissue inhibitor of metalloproteinase)-2 genes. (A) The typical 5’ C p C island in normal cells is shown as an unmethylated CpC-rich area of about 1.5 kb extending from the immediate 5’ flanking region, through the transcription start site and the first exon, and well into the first intron. The unmethylated island (blue area, labeled protection) is embedded between 5’ and 3’ flanking areas, which are densely methylated (all gene regions outside protected area). Each gene analyzed contained two Alu repeats within 1.0 kb of the 5’ portion of the island, and the VHL gene has an Alu within 0.5 k b of the 3’ end. These Alu repeats are heavily methylated in the VHL and E-cadherin genes, as analyzed by the MSP procedure and by bisulfite genomic sequencing (see text). The average position for clusters of SPl sites toward the borders of the island is depicted. DNA-MTase activity is shown to be excluded from the “protected” island and accessible to the flanking regions. (B) A hypermethylated C p C island is shown, as would be typical for our analyses of the E-cadherin gene in many breast cancers and the VHL gene in certain renal cancers. The entire island is methylated, as shown by absence of the blue central area of protection. DNA-MTase now has access to the island because the depicted relative increase in activity has overidden the 5’ and 3’ borders that normally protect against methylation and/or abnormalities of local features of these borders have led to disruption of the protective function.

Fig. 5 A hypothetical model for progression of a generic tumor that takes into account genetic changes and altered DNA methylation. Details are found in the text, and salient features of the model are as follows. (A) A normal epithelial cell system, such as the normal colon, in an older person. Two cells have age-related CpG hypermethylation in the ER gene. ( B ) This change participates in a growth advantage for the involved cells, which leads to their forming a “field effect” for colon cancer predisposition. Exposure of this field to carcinogens and/or chronic injury leads, in some cells, to a relative increase in DNA-MTase activity (green nuclei) and to a transient increase in overall DNA methylation (red circle around the nucleus). (C) In one such cell, an initiating mutation occurs, and the heightened sensitivity of the cell to this event, by virtue of the ER gene change, leads to clonal expansion into the preneoplastic area of dysplasia. In the dysplasia, the cells retain the relative increase in DNA-MTase activity, but now have either normal o r reduced levels of overall DNA methylation (absence of red circle around the nuclei). Heterogenous cells now appear in the dysplastic lesion, which have, individually, hypermethylation of the p16 gene and a mutation of the pS3 gene. ( D )There are phenotypic consequences for these events, however, only when they occur together in a single cell (bottom right of the lesion), and this cell leads to clonal expansion into carcinoma in sztu. (E) This carcinoma develops a heterogenous pattern for hypermethylation and inactivation of the E-cadherin gene (cell at the left, bottom of the carcinoma), and this event results in evolution o f invasion of the basement membrane and metastatic capacity of the carcinoma.

DNA Methylation and Neoplasia

I83

neni et al., 1995). The CpC-rich Sp-l-binding motif is contained in most CpG islands (Antequera and Bird, 1993a). The APRT gene CpG island is protected from methylation as long as the two most 3’ of the four Sp-1 sites are intact (Mummaneni et al., 1995).Interestingly, although such deletion of the more 5 ’ Sp-1 sites leads to loss of protection from methylation, this does not interfere with transcriptional activity when the unmethylated proximal promoter is used to drive a reporter gene (Dush et al., 1988). Mummaneni and colleagues (1993) have designated a region just 5 ’ to the APRT Sp-1 sites, and to the island in which they reside, as a “de novo methylation center,” which can initiate methylation and from which the nearby island is protected. Their data suggest the presence of czs-acting elements that, when preferentially methylated by the DNA-MTase enzyme, can facilitate spreading of methylation bidirectionally. Deletion studies reveal that the putative methylation center resides in an about 800-base-pair (bp) area that harbors tandem arrays of B-1 elements (Mummaneni et al., 1993), the murine counterpart to human Alu repeat elements (Quentin, 1994). Interestingly, Alu elements, independent of location, appear t o be heavily methylated in all somatic cell types (Hellman-Blumberg et al., 1993; Kochanek et ul., 1990), and these sequences have been proposed to initiate sites of normal DNA methylation in the p53 gene (Magewu and Jones, 1994). Taking all of these findings into account, how does the anatomy of CpG islands of tumor suppressor genes that become methylated in human cancers fit with the APRT model? In this regard, we have compared the VHL and Ecad genes to the human APRT, TIMP2, and GST-ITgenes and derived a strikingly generic anatomy (Fig. 4 and see Graff et al., 1997). Each gene has two Alu repeat elements located within 1.0 k b 5’ of the CpG islands that span the transcription start sites, and the VHL gene has another Alu repeat located within 500 bp 3’ to the CpG island. At the 5 ’ and 3’ ends of each island, within the area where the CpG density is beginning to decline, there are multiple Sp-1 sites. Using a combination of the new MSP PCR technique (Herman et al., 1996a) and bisulfite genomic sequencing (Myohanen et al., 1994), we have mapped, in normal tissues and culture lines of tumors, the methylation status of 80 or more CpG sites located within the CpG islands of the VHL and E-cad genes and in the regions flanking the islands (Graff et a/., 1997 and summarized in Fig. 4). In all instances, even in normal cells, the Alu sequences were heavily methylated, as were adjacent CpG sites, extending up to the Sp-1 sites at both ends of the islands (see Fig. 4). The island regions within the Sp-1 sites were completely free of methylation in the normal tissues and in tumors expressing the genes. In contrast, in cultured breast tumors previously shown to have a nonexpressing and hypermethylated E-cad gene, and in a renal cancer with a similar situation for the VHL gene, all the island CpG sites, as well as the island-flanking regions, were extensively methylated (Graff et al., 1997). We conclude from these studies that the CpG islands of human genes may

I84

Stephen B. Baylin el af.

have a generic anatomy characterized by sharp boundaries a t both the 5’and 3’ edges, which normally separate unmethylated sequences within the islands from densely methylated flanking regions often containing Alu repeat sequences, In addition, multiple Sp-1 sites are located at both the 5’ and 3’ boundaries in a strategic position to potentially protect the island from spread of methylation from the flanking regions. Interestingly, in the studies of the APRT gene previously discussed, only the Sp-1 sites at the 5 ’ edge o f the CpC island were tested for protective function. Determining precisely what molecular mechanisms normally allow the unmethylated CpG islands to coexist with nearby methylated areas must be the goal of intensive research during the next years. If this anatomy helps define the protection boundaries for the unmethylated status of a typical C p C island, how does this protection break down during the evolution of a neoplasm? There is little experimental data to address this critical question. Certainly, experimental data indicate that during tumor progression, chronically increased DNA-MTase activity can override the normal protective mechanisms that maintain the unmethylated status of these regions (Vertino et al., 1996). We have looked a t the dynamics of how this process occurs for the E-cad CpG island in our fibroblast model for overexpression of the human DNA-MTase gene. In this setting, E-cad hyper-methylation appears to spread, over 30 to 40 cell passages, inward from both the 5’ and 3’ borders, to involve the transcription start site at the center of the island as the last event (Graff et al., 1997). These data fit well with those of others indicating that de nova methylation of CpC-rich regions evolves by a spreading process that requires multiple cell cycles (Toth et al., 1989). The data also indicate how prolonged exposure to increased DNA-MTase activity during tumor progression could exert constant pressure to disrupt the boundaries that normally protect C p C islands from methylation. As mentioned earlier, defects other than DNA-MTase dysregulation may contribute to CpC island hypermethylation in neoplasia and could well involve the local mechanisms that constitute the C p C island protective borders. One attractive target would be loss of binding proteins like those that may interact with the Sp-1 sites at the island borders. If these were deficient in amount or function, the CpC islands might be rendered vulnerable to methylation. Again, this mechanism could represent a genetic origin to aberrant methylation if such proteins were disrupted by mutational events. Although the proteins involved with protection of C p C islands from methylation are yet to be defined, some may be transcription factors, such as Sp-1. These proteins could help to maintain chromatin structure around promoter areas, so that the access of DNA-MTase to the island is precluded. Proteins known to be essential for chromatin structure, such as histones, are now known to share structural homologies to other proteins that simultaneously

DNA Methylation and Neoplasia

I85

function to modulate transcription and chromatin structure around gene promoters (Felsenfeld, 1996; Wolffe, 1994). In this regard, one particular variant of histone H1, H l e , has been shown to have binding affinity for CpG-rich DNA and to block DNA-MTase activity in vitro when bound to such sequences (Santoro et al., 1995; Zardo et al., 1996).The authors of this work actually propose that this histone is a candidate factor to explain the protection of C p C islands from methylation. If alterations in transcription factors and/or constituents vital to active transcription could play a primary role in the aberrant C p C island methylation in neoplasia, then this would raise a vital issue regarding the associated loss of gene function, that is, which comes first, the loss of transcription o r the appearance of aberrant C p C island methylation? Another way of phrasing the question would be, does promoter region methylation start the process of gene inactivation or simply reflect it? Of course, this same question has plagued understanding of the role of promoter region methylation in normal cells. The answer to this conundrum in both the normal and cancer cell settings is probably that either scenario can be operative and that the sequence of events may be cell type- and gene-specific. Experimentally, we know, as Antequera and Bird (1993a) have pointed out, that methylation may both respond to chromatin changes and initiate them. Thus, CpC islands methylated in vitro can form condensed chromatin when placed in cells (Keshet et al., 1986). In contrast, inactivation of genes on the X-chromosome of females appears to precede the hypermethylation of C p C islands in the corresponding gene promoters, and the latter event may “lock in” the transcriptionally inactive state (Lee et al., 1996). Likely, this dual role of methylation will emerge for individual gene inactivation events in cancer cells. In some instances, during tumor progression, aberrant methylation may be the initial event, and loss of gene transcription may then ensue. In other situations, a decrement in transcription factors that normally maintain transcriptional activity may precede aberrant methylation of certain genes. We and others have some evidence for this type of transcriptional deficit from studies of the E-cad gene in cultures of human breast and prostate cancer. Transient transfection of reporter constructs for the promoter of this gene into cell lines that have hypermethylation of the endogenous gene consistently yields less activity than in cell lines in which the endogenous gene is unmethylated (Graff et al., 1995; Hennig et al., 1995). Although, in our hands, the degree of reduction in the reporter gene activity does not quantitatively account for the complete lack of native E-cad transcripts in the lines with the hypermethylated endogenous gene, the results reflect a decrement in one o r more trans-acting factors important to E-cad expression. In such settings, definition of the proteins involved, and the mechanisms underlying loss of their transcriptional function may be most revealing for our understanding of aberrant C p C island methylation in tumor cells.

186

S t e p h e n B. Baylin ef a / .

6 . SUMMARY The molecular mechanisms underlying aberrant CpG island methylation in neoplasia are complex and may involve multiple simultaneously occurring events. A model can be envisioned (see Fig. 4B)wherein, during tumor progression, a combination of chronic exposure of cells to increased DNAMTase activity, loss of region-specific demethylation activity, and/or loss of binding proteins that help maintain CpG island methylation boundaries lead to promoter region hypermethylation and gene inactivation. Identifying these steps precisely will be an essential target of future research endeavors.

VI. AN OVERVIEW OF TUMOR PROGRESSION THAT INCORPORATES THE ROLES OF ALTERED DNA METHYLATION The data reviewed previously indicate that changes in DNA methylation should be considered a vital aspect of tumor progression. If so, it may be helpful to view the potential contribution of each component of altered DNA methylation in a working model of neoplastic evolution (Fig. 5). We hope that the hypotheses entailed may help to guide future investigations of the methylation changes in neoplasia. One of the cardinal features of our proposed progression model is that the imbalance of DNA methylation begins at the earliest stages of the neoplastic process. For example, in a previously discussed mouse lung tumor model, increased DNA-MTase activity and a transient increase in overall genomic methylation accompany the earliest cellular responses to carcinogen exposure (see Fig. SB).This overall methylation increase could help initiate the promoter region hypermethylation events that influence tumor suppressor gene expression. Despite these initial methylation increases, the steady state for early neoplasms is often one of overall genomic hypomethylation in association with a continued increase in DNA-MTase activity (see Fig. SC). Thus, from the initial neoplastic stages and beyond, the components of DNA methylation imbalance are in place and available to participate in alterations in chromosome structure and gene expression throughout tumor progression. Among the components of the DNA methylation imbalance in neoplasia, a direct role for the regional hypermethylation that targets CpC islands is now, perhaps, best appreciated. In this regard, as outlined earlier, there appear to be two classes of hypermethylated genes that have to be considered. Some genes appear to undergo CpG island hypermethylation only during neoplastic evolution, whereas in others, the change may preexist in normal cells as a function of the aging process. As depicted in Figure 5, both of these

D N A Methylation and Neoplasia

I87

gene classes may contribute to the earliest steps in neoplasia. For the age-related changes, for example, evidence points to the possibility that colon tumors may initiate in cells that harbor hypermethylation of the ER gene. In Figure 5A and B, such cells are hypothesized to be particularly sensitive to initial carcinogen effects and to respond with a selective growth advantage that provides a field effect for tumor potential. In contrast to genes hypermethylated during aging, the classic tumor suppressor genes examined appear to acquire this change only after the neoplastic process initiates. What is the timing of these events during tumor progression? This is an important question for which information is only beginning to emerge. Studies suggest, however, that this CpG island methylation often begins early and with a pattern of cellular heterogeneity (see Fig. 5C through SE), which is critical to our progression model. Evidence for such heterogeneity emerged early in our studies of the calcitonin gene on chromosome 1Ip, where we found both hypermethylated and normal alleles in leukemic cells proven to be a clonal population by molecular analyses (Baylin et al., 1987). This same phenomenon was later recognized for the p l 5 gene in these same neoplasms (Herman et al., 1996b). More recently, in ongoing studies of breast cancer progression, hypermethylated alleles for the ER and E-cad genes can be clearly detected by the new MSP technique in 25 to 35% of ductal carcinoma in situ (DCIS) lesions (Graff et al., unpublished results). This incidence increases in the next stage of invasive tumors. Similarly, in lung cancer progression, hypermethylated alleles of the p l 6 gene have been detected in preinvasive lesions in patients who simultaneously harbored established lung carcinomas with this same change (Herman et d.,unpublished results). This situation for human lung cancers is paralleled by findings in a rat model for lung carcinogenesis, in which hypermethylated p16 alleles were detected at a stage of epithelial hyperplasia that preceded frank adenoma and subsequent adenocarcinoma development (Belinsky et al., unpublished results). These early methylation changes can subsequently progress during tumorigenesis in several ways. First, the density of methylation within the involved gene can increase over time, as observed for the candidate tumor suppressor gene, HIC1, between the stages of benign adenomatous colon polyps and frank carcinomas (Makos et al., 1992). Second, cells with the early methylation changes may be selected for with time as the increasing methylation density becomes associated with loss of key gene function. This selection may be facilitated, especially if these gene expression changes coincide in the same cell with other genetic and/or epigenetic events that favor clonal expansion (see Fig. 5B through E). For example, in breast tumor progression, we are now finding that concordancy for ER and E-cad gene hypermethylation may track with the invasive and/or metastatic phenotype (Graff et al., unpublished data).

I88

Stephen B. Baylin el al.

The concept for progressive involvement of CpG island hypermethylation during tumor progression is similar to that for accrual of classic gene mutations (Kinder and Vogelstein, 1996).The role of cellular heterogeneity, however, so accepted for DNA changes such as chromosome alterations (Nowell, 1976), is less often considered for specific gene mutations because we often think of these latter events as leading to immediate selective advantage, which triggers rapid clonal expansion. In fact, much evidence suggests that key gene mutations reside in cell subpopulations a t stages of tumor progression. For example, p53 gene mutations exist in subpopulations of cells early in progression of tumors such as brain cancer (Sidransky et a/., 1992) and in other, more recently studied tumor types (Ziegler et al., 1994). Similarly, as previously discussed, homozygous deletions of the p16 gene can be heterogeneous within tumor cell populations. Thus, both promoter region hypermethylation and gene mutations may exist in subclones of tumor cells and not result in a dominant progression event until these changes act cooperatively to initiate clonal expansion. Each individual molecular change, however, is still critical because the involved cell population is rendered particularly sensitive to expansion when additional events are superimposed. The model in Figure 5 summarizes this discussion by depicting a cooperative interaction between CpC island hypermethylation events and classic mutational changes that drives tumor progression through clonal expansion of heterogeneous cell populations. Note that after an initiating mutation occurs in a cell harboring an age-related hypermethylation change (see Fig. SB), clonal expansion ensues to produce a dyspastic or preneoplastic lesion in which all cells carry the molecular alterations (i.e., ER gene hypermethylation and the first mutation) that have occurred to this point (see Fig. SC). At this dysplastic state, however, two types of cellular heterogeneity arise. Subpopulations of cells have developed a pS3 mutation, and a separate set have evolved hypermethylation of the p16 gene promoter region (see Fig. 5C). There are no selective advantages from these molecular changes, however, until both coincide in the same cell. Such cooperation for inactivation of these two genes has been suggested in a recent study (Kinoshita et al., 1996). Allelic inactivation of the .opposite copy of the mutated o r hypermethylated genes must, of course, also occur. The cell now harboring two mutational changes (the initiating mutation, plus a pS3 mutation) and two hypermethylation changes (in the ER and p l 6 genes) clonally expands to produce carcinoma in situ (see Fig. 5C and D). In the next critical stage of tumor progression (see Fig. 5D and E), a carcinoma with metastatic potential evolves, again through expansion of a cell subpopulation. A cell within the carcinoma in situ develops a reduction in E-cad gene expression in conjunction with promoter region hypermethylation (see Fig. SD), and the progeny exhibit decreased homotypic cell-to-cell adhesiveness and capacity to invade (see Fig. 5E).

DNA Methylation and Neoplasia

189

Thus, this hypothetical model embodies the classic concepts of cellular heterogeneity in tumor progression (Nowell, 1976) and stresses that, from the earliest time points, cooperative interaction between genetic and epigenetic events drives critical stages. Although aspects of this model must be continually verified or refuted by experimental approach, we think the dynamics depicted fit nicely with existing data. It is important to note that cellular heterogeneity for any molecular change can appear at virtually every stage of tumor evolution. In the model, the timing for the DNA changes shown is only one possibility and is not meant to imply an absolute order. In fact, even for a single tumor type, the model predicts that, for example, cells harboring changes favoring invasive and/or metastatic potential could arise a t virtually any progression stage shown. The timing for resultant phenotypic consequences would, again, depend on the number of changes simultaneously affecting a given call.

VII. CLINICAL IMPLICATIONS OF ALTERED DNA METHYLATION IN CANCER In addition to the discussed biologic consequences of altered DNA methylation for tumorigenesis, the advances in understanding of these processes justify considering how the data may prove useful to the diagnosis and treatment of cancer. Particularly, the newer findings for inactivation of specific genes in association with CpG island methylation and the potential direct role of increased DNA-MTase activity in tumor progression bear comment. The fact that the CpC island methylation events for many involved genes may represent a process specific to neoplastic cells makes this change attractive to consider for development of molecularly based biomarkers. With the advent of PCR procedures, such as MSP (Herman et al., 1996a),that can sensitively detect the methylation status of specific CpG sites in rare alleles, DNA from small clinical specimens, processed at the time of routine pathology examination, can be analyzed for methylation changes. Aberrantly hypermethylated alleles could potentially detect stages of early tumor evolution, or such analyses of lymph nodes or margin tissues removed a t the time of initial surgery could provide a sensitive index of micrometastases. Finally, PCR analyses of posttreatment samples, such as bone marrow from patients with leukemia, could provide a sensitive means for assessing residual tumor or predicting disease recurrence. Tests of these hypotheses should be forthcoming in the next several years. The findings from the Jaenisch group that lowered levels of DNA-MTase activity reduce the incidence of adenoma formation in mice with APC gene mutations (Laird et al., 1995), combined with the evidence that increased

I90

Stephen B. Baylin ef RI.

DNA-MTase activity may be characteristic of multiple types of carcinogenesis, have legitimized consideration of targeting inhibition of this enzyme as an anticancer strategy. In fact, the demethylating agent 5-AZA-cytidine and its congener, 5-deoxy-AZA-cytidine, have been tried clinically. Some response rates have been seen in diseases such as leukemia (Gattei et al., 1993; Mandelli, 1993; Petri et al., 1993; Silverman et al., 1993; Willemze et al., 1993; Zagonel et al., 1993). Because these drugs incorporate into DNA and may have effects other than inducing demethylation, the mechanisms underlying antitumor effects are not known. It is now possible to test whether reactivation of specific genes through relief of CpG island hyperrnethylation might correlate with any observed therapeutic effects. Also, new and more specific ways to inhibit the enzyme are being studied. Promising cell-differentiating and tumor cell-inhibiting activity, bo:h in vitro and in vivo, has been reported with use of antisense approaches (MacLeod and Szyf,l995; Ramchandan et al., 1997; Szyf et al., 1992). The next several years should see testing of multiple approaches aimed at the possibility that a reduction of DNA-MTase activity may be useful for tumor treatment. Preclinical testing can take advantage of several excellent murine and rodent carcinogenesis models to study tumor responses, to outline the mechanisms involved, and to delineate the most sensitive stages of tumor progression. Such investigations should help determine whether decreasing DNA-Mtase activity is most useful for prevention, early intervention, and/or treatment of established tumors and should justify the building interest in the alterations of DNA methylation that appear to be a fundamental hallmark of the neoplastic state.

ACKNOWLEDGMENTS We gratefully acknowledge all of the colleagues over the years who have contributed to studies i n our own laboratories and the many workers in the DNA methylatioii field whose work is cited in this review. Of course, space limitations preclude referencing many other papers that have provided important data, and we apologize for those exclusions. We thank Tammy M. Means for invaluable secretarial services. Portions of work performed in the authors’ laboratories were supported by National Cancer Institute grants #CA43318, #CA.54396, and #SPOCA58184 and b y a grant from the Tobacco Research Council (#1987B). J.G.H. is recipient of a Valvano Fouindation Research Award and J.P.I. of a Kiminel Foundation Award.

REFERENCES A d a m , K. 1.. P., Lindsay, El., Reale, A., and et a/. ( 1993). / , I “DNA Methylation: Molecular Biology and Biological Significance” (J. I? Jost, and H. P. Saluz, Eds.), pp. 120-144. I3asel, Rirkhausrr Verlag.

DNA Methylation and Neoplasia

191

Albagli, O., Dhordain, P., Deweindt, C., et a/., (1995). Cell Growth Differ, 6 1 193-1 198. Aniello, F., Locascio, A., Fucci, L., ct al. (1996). GenBank. Z50183. Antequera, E, and Bird, A. (1993a).In “DNA Methylation: Molecular Biology and Biological Significance” (J. P. Jost, and H. P. Saluz Eds.), pp, 169-185. Basel, Birkhauser Verlag. Antequera, F., and Bird, A. (1993h).Proc. Natl. Acad. Sci. U.S.A. 90, 11995-11999. Antequera, F., Boyes, J., and Bird, A. (1990).Cell. 62, 503-514. Bardwell, V. J., and Treisman, R. (1994).Genes Dew/. 8, 1664-1677. Barlow, D. P. (1995).Science. 270, 1610-1613. Baylin,S. B., Fearon, E. R., Vogelsrein, B.,etal. (1987).Blood. 70,412-417. Baylin, S. B., Hoppener, J. W. M., de Bustros, A., et al. (1986). Cancer Res. 46, 2917-2922. Baylin, S. B., Makos, M., Wu, J. J., et a/ . (1991).Cancer Cells. 3, 383-390. Becker, F., Atkinson, M. J., Reich, U., etal. (1994).Cancer Res., 54, 3845-3852. Belinsky, S. A., Nikula, K. J., Baylin, S. B., ef al. (1996). Proc. Natl. Acad. Sci. U.S.A. 93, 4045-4050. Bell, M. V., Hirst, M. C., Nakahori, Y., et al. (1991).Cell. 64, 861-866. Berx, G., Cleton-Jansen, A., Nollet, F., e t a / . (199.5).E M B O J . 14, 6107-61 15. Bestor, T. H. (1988).Gene 74, 9-12. Bestor, T. H. (1990). Philos. Trans. R. Soc. I-ond. 326, 179-1 87. Bestor, T. tH. (1992). E M B O J . 11, 2611-2617. Bestor,T., Laudano, A., Mattaliano, R., e t a / . (1988).J.M o l . Biol. 203, 971-983. Bhave, M. R., Wilson, M. J., and Poirier, I.. A. ( I 988). Carcinogenesis 9, 343-348. Bird, A. (1992). Cell 70,543. Bird, A. P. (1986).Nature 321, 209-213. Bird, A. P. (1995).Trends Genet. 11,94-100. Boyes, J., and Bird, A. (1991). Cell 64, 1123-1 134. Brandeis, M., Frank, D., Keshet, I., et al. (1994). Nature 371,4351138. Cairns, P., Mao, I.., Merlo, A., et al. ( 1994). Science 265, 4 15-4 17. Cairns, P., Polascik, T. J., Ehy, Y., eta/. (1995).Nature Genet. 11, 210-212. Caldas, C., Hahn, S. A., da Costa, 1.. T., e t a / . (1994). Nature Genet. 8, 27-32. Cedar, H. (1988).Cell 53,311. Chen, J., and Jones, P. A. (1990).Cell Growth Differ. 1, 383-392. Chen,Z., Brand, N. J.,Chen,A.,etal. (1995).E M B O J . 12, 1161-1167. Christman, J. K., Sheikhnejad, G., Dizik, M., et al. (1993). Carcinogenesis 14, 551-557. Christrnan, J. K.. Sheikhnejad, G., Marasco, C. J., et a/. (1995).Proc. Nut/. Acad. Sci. U.S.A. 92,7347-7351. Christofori, G., Naik, P., and Hanahan, D. (1994).Nature 369, 414-41 8. Clarke, R., Dickson, R. B., and Lippman, M. E. (1992).Crit. Res. Oncol. Hematol. 12, 1-23. Color, V., Maloisel, L., and Rossingnol, J. (1996).Cell 86, 855-864. Costello, J. F., Berger, M. S., Huang, H. S., eta/. (1996).Cancer Res. 56, 2405-2410. Costello, J. F., Futscher, B. W., Kroes, R. A., et al. (1994a). Mol. Cell. B i d . 14, 65 15-652 I . Costello,J. F., Futscher, B. W., Tano, K., etal. (1994b).J. Biol. Cltem. 269, 17228-17237. Crane, R. K. ( 1 968). In “Handbook of Physiology” (C. F. Code, and W. Heidel, Eds.), p. 2535. Baltimore, Williams & Wilkins. Danforth, D. N. (1992).Crit. Reu. Oncol. Hematol. 12, 91-149. de Bustros, A., Nelkin, B. D., Silverinan, A., et al. (1988). Proc. Nut/. Acad. Sci. U.S.A. 85. 5693-5697. De Smet, C., De Backer, O., Faraoni, I., et a/. (1996). Proc. Natl. Acad. Sci. U.S.A. 93, 7149-7153. Doerfler, W. ( 1993). In “DNA Methylation: Molecular Biology and Biological Significance” (J. P. Jost, and H. P. Saluz, Eds.), pp. 262-299. Basel, Birkhauser Verlag. Dragani, T. A,, Manenti, G., and Pierotti, M. A. (199.5).Adv. Cancer Res. 67, 83-1 13. Dush, M. K., Briggs, M. R., Royce, M. E., e t a / . (1988).Nucleic Acids Res. 16, 8509-8524.

192

Stephen B. Baylin e1 al.

El-Dciry, W. S., Nclkin, B. D., Celano, P., ef a/. ( I99 I ) . /'roc. Nut/. Acad. Sci. U.S.A. 88, 3470-,3474. Fcinherg, A. P., and Vogelstein, B. ( I 983a). Biockeni. Bi(~phys.Res. Conzrnun. 111, 47-54. Feinberg, A. P., and Vogelstein, B. ( I 9831~).Nature 301, 89-92. Felsenfeld, G. ( 1996). Cell 86, 13-1 9. Ferguson, A. T., Lapidus, R. G., Baylin, S. B., eta/. ( 1995). Cancer Res. SS, 2279-2283. Flavell, R. B. (1994). /'roc. Natl. Acad. Sci. U.S.A. 91, 3490-3496. Frommer, M., McDonald, 1. E., Millar, D. S., et a/. (1992). Proc. Nut/. Acad. Sci. U.S.A. 89, 1827-1 8 3 I . Gaina-Sosa, M. A., Slagel, V. A., Trcwyn, R. W., et a/. (1983). Nuc/eic Acids Res. 1 1 , 6883-6894. Gattei, V., Aldinacci, I)., Petti, M. C., et al. ( I 993). Letrkeniia 7, 42-48. Gnarra, J. R., Tory, K., Weng, Y.,et a/. ( 1994). Nature Genet. 7, 85-90. Goelz, S. E., Vogelstein, B., Hamilton, S. R., et a/. (1985). Science. 228, 187-190. Gonzalez-Zulueta, M., Bender, C. M., Yang, A , et a/. (199.5). Cancer Res. 55, 4531453.5. Graff, J. R., Herman, J. G., Lapidus, R. G., rt a/. ( 199.5). Cancer Res. 55, 5 195-5 199. Craff, J. K., Herman, J. G.,Myiihanen, S. et a/. (1997).J. Biol. Chew, in press. Greger, V., Debus, N., h h i n a n n , D., e t a / . (1994). Human Genet. 94, 491-496. Hannon, G. J. and Beach, D. (1994). Nature 371, 257-261. Hehert, J., Cayuela, J. M., Berkeley, J., et a/. ( I 994). Blood 84, 4038-4044. Hellman-Blumberg, U., Hinrz, M. F., Gatewood,J. M., et a/. (1993). Mol. Cell. B i d . 13, 4523-4530. Hennig, G., Behrens, J., Truss, M., et a/. ( 1 995). Oncogene 11, 475-484. Herman, J. G., Civin, C. I., Issa, J. P., et a/. (1997). Cancer Res. 57, 837-841. Herman, J. G., Craff, J. R., Mydhinen, S., et a/. (1996a). Proc. Natl. Acad. Sci. U.S.A. 93, 9821-9826. Herman, J. G., Jen, J., Merlo, A., et a/. (1996b). Cancer Res. S6, 722-727. Herman, J. C., Latif, F., Weng, Y., et rtl. (1994). Proc. Nail. Acad. Sci. U.S.A. 91, 9700-9704. et a/. (1995). Cancer Res. SS, 4525-4530. Herman, J. G.,Merlo, A., Mao, l.., Holliday, R. (1986). Exp. Cell. Res. 166, 543-552. Holliday, K. (1987). Science 238, 163-1 70. Holliday, I<. ( 1990a). Neuj Bio/ogist 2, 7 19-726. Holliday, R. ( I990b). \'bi/os. Trans. R . Soc. Lond. €3. Biol. Sci. 326, 329-338. Holliday, R. (1991). Mutat. Res. 250, 351-363. Holliday, R., and Jeggo, I? A. (1985).Cancer S w u . 4, 557-581. Hopkin, K. (1995l.J. N.I.H. Res. 7, 26-28. Hussussian, C. J., Struewing, J. P., Goldstein, A. M., e t a / . (1994). Nature Genet. 8, 15-21. Huynh, H., Alpert, L., and Pollak, M. (1996).Cancer Res. 56, 4865-4870. Iguchi-Ariga, S. M., and Schaffner, W. (1989). Genes Dev. 3, 612-619. h a , J. J., Baylin, S. B., and Belinsky, S. A. ( I 996a). Cancer Res. 56, 3655-3658. Issa, J. P., Ottaviano, Y. L., Celano, I>., et a/. ( I 994). Nature Genet. 7, 536-540. h a , J. J., Vertino, P. M., Boehrn, C. D., et a/. (1996b). Proc. Not/. Acad. Sci. U.S.A. 93, 11757-1 1762. Issa,J. P., Vertino, P. M., Wu, J., e t a / . (1993).J. Nut/. Cancer Inst. 8.5, 1235-1240. Issa, J. I?, Zehnbauer, B. A., Civin, C. I., et a/. ( 1 9 9 6 ~ )Cancer . Res. 56, 973-977. Jackson-Grusby, L., Laird, P. W., Magge, S. N., etal. (1997). Froc. Natl. A c d . Sci. U.S.A. 94, 468 1-4685. Jen, J., Harper, J. W., Bigner, S . H., e t a / . ( I 994). Cancer Res. 54, 6353-6358. Jiang, S., and Jordan, V. C. (1992)./. Natl. Cancer Inst. 84, 580-591. Jones, P. A. ( 1 996). Cancer Res. 56, 2463-2467. Jones, P. A., Wolkowicz, M. J., Harrington, M. A., et a/. (l990a).Philos. Trans. R. Soc. h n d . B. Biol. S c i 326, 277-284.

DNA Methylation and Neoplasia

193

Jones, P. A,, Wolkowicz, M. J., Rideout, W. M., etal. (1990b). Proc. Nut/. Acad. Sci. U.S.A. 87, 61 17-6121. Kamb, A., Gruis, N. A., Weaver-Feldhaus, J., et al. ( 1994a). Science. 264,436-440. Kamb, A., Shattuck-Eidens, D., Eeles, R., et al. (1994b). Nature Genet. 8,23-26. Kanai, Y., Oda, T., Tsuda, H., et al. (1994).J p n . J. Cancer Res. 85, 1035-1039. Kane, M. F., Loda, M., Gaida, G. M., et a/. (1997). Cancer Res. 57, 808-811. Kautiainen, T. L., and Jones, P. A. (1986).J. Biol. Chem. 261, 1594-1598. Keefe, D., Garcia-Segura, L. M., a n d Naftolin, F. ( 1994). Neurohiol. Aging. 15,495-497. Keller, J. R., Sing, G. K., Ellingsworth, L. R., et al. (1989).J. Cell. Biochem. 39, 175-184. Kerckaert, J., Deweindt, C., Tilly, H., etal. (1993). Nature Genet. 5, 66-70. Keshet, I., Lieman-Hurwitz, J., and Cedar, H. (1986). Cell 44, 535-543. Kinoshita, I., Dosaka-Akita, H., Mishina, T., e t a / . (1996). Cancer Res. 56, 5557-5562. Kinder, K. W., and Vogelstein, B. (1996).Cell 87, 159-170. Kitsberg, D., Selig, S., Brandeis, M., et al. ( 1 993). Nature 364, 459-463. Kochanek, S.,Toth, M., Dehmel, A., etal. (1990).Proc. Natl. Acad. Sci. U.S.A. 87,8830-8834. Koldovsky, 0 . (1969). “Development of Functions of the Small Intestine in Mammals and Man.” New York, Karger. Kuerbitz, S. J., and Baylin, S. B. (1 996). Cell Growth Differ. 7, 847-853. Kushner, P. J., Hort, E., Shine, J., et al. (1990). Mol. Endocrinol. 4, 1465-1473. Laird, P. W., Jackson-Grusby, L., Fazeli, A., e t a / . (1995). Cell 81, 197-205. Laird, P. W., and Jaenisch, R. (1994).Hum. Mol. Genet. 3, 1487-1495. Lasalle, J. M., and LaLande, M. (1996).Science 272,725-728. Latif, F., Tory, K., Gnarra, J., et a/. (1993). Science 260, 1317-1 320. Lee, J. T., Strauss, W. M., Dausman, J. A., et a/. (1996). Cell 86, 83-94. Lee,P. J., Washer, L. L.,Law,D. J.,etal. (1996).Proc. Natl. Acad.Sci. U.S.A. 93,10366-10370. Lee, W. H., Morton, R. A., Epstein, J. I., et al. (1994). Proc. Nut. Acad. Sci. U.S.A. 91, 11733-11737. Lee, Y., Klein, C. R., Kargacin, B., eta/. (1995). Mo/. Cell. Biol. 15, 2547-2557. Lei, H., Oh, S. P., Okano, M., et a/. (1996). Development 122, 3 195-3205. Leonhardt, H., Page, A. W., Weier, H., e t a / . (1992). Cell 71, 865-873. Li, E., Beard, C., and Jaenisch, R. (1993). Nature 366, 362-365. Li, E., Bestor, T. H., and Jaenisch, R. (1992). Cell 69, 915-926. Lo, K. W., Cheung, S. T., Leung, S. F., et a/. ( 1996). Cancer Res. 56, 272 1-2725. Lo, K. W., Huang, D. P., and Lau, K. M. (1995). Cancer Res. 55,2039-2043. Loh, W. E., Scrable, H. J., Livanos, E., et al. (1992). Proc. Natl. Acad. Sci. U.S.A. 89, 1755-1759. Ma, Z. (2. (1993). Proc. Nut/. Acad. Sci. U.S.A. 90, 3740-3744. MacLeod, A. R., Rouleau, J., and Szyf, M. (1995).J. Biol. Chem. 270, 11327-1 1337. MacLeod, A. R., and Szyf, M. ( 1 995).J. Biol. Chem. 270, 8037-8043. Macleod, D., Charlton, J., Mullins, J., et a/. (1994). Genes Deu. 8, 2282-2292. Magewu, A. N., and Jones, P. A. (1994).Mo/. Cell. B i d . 14,4225-4232. Makos, M., Nelkin, B. D., Chazin, V. R., eta/. (1993a). Cancer Res. 53, 2715-2718. Makos, M., Nelkin, B. D., Lerman, M. I., et al. (1992). Proc. Nut/. Acad. Sci. U.S.A. 89, 1929-1 933. Makos, M., Nelkin, B. D., Reiter, R. E. eta/. (1993b). Cancer Res. 53,2719-2722. Maminta, M. L., Molteni, A., and Rosen, S. T. (1991). Mol. Cell Endocrinol. 78, 61-69. Mandelli, F. (1993). Leukemia. 7, 1-2. Mareel, M., Bracke, M., and Van Roy, F. (1995). Cancer Detec. Prev. 19, 451-464. Matsumura, T., Hunter, J. L., Farooq, M., eta/. (1989a). Exp. Cell. Res. 184, 148-157. Matsumura, T., Malik, F., and Holliday, R. (1989b). Exp. Gerontol. 24, 477-481. Meehan, R. R., Lewis, J. D., McKay, S., et a/. (1989). Cell 58, 499-507. Melvin, M. J., McGurn, M. E., Bort, S. J., et a/. ( 1 995). Eur. J.Immunol. 25, 4 2 6 4 3 0 .

I94

Stephen B. Baylin ef al.

Merlo, A., Herman, J . C., Mao, l.., e t a / . (1995). Nature Med. 1, 686-692. Migeon, B. R. ( 1 994). Trends Genet. 10, 230-235. Migliazza, A., Martinotti, S., Chen, W., et ul. (1995). Proc. Nut/. Acad. Sci. U.S.A. 92, 12520-I2524. Miyamura, Y., Ohtsu, H., Niwa, O., et a/. (1993). Gerontology. 39, 3-10. Monk, M. (1990). Phil. Trans. R. Soc. Lond. 326, 299-312. Monk, M., Bouhelik, M., and Lehnert, S. (1987). Development 99, 371-382. Mummaneni, P., Bishop, P. L., and Turker, M. S. (1993)./. Biol. Chem. 268, 5.52-558. Muinmaneni, P., Walker, K. A., Bishop, P. l.., et ul. ( 1995). J . B i d . Chew. 270, 788-792. MyBhlnen, S., Wahlfors, J., and Janne, J. (1994). D N A Seq. 5, 1-8. Nan, X., Meehan, R. R., and Bird, A. (1993). Nrrclezc Acids Rcs. 2 1 , 4 8 8 6 4 8 9 2 . Nan, X., Tare, P., l.i, E., rt ul. (1996). MoI. Cell. B i d . 16, 414-421. Nobori, T., Miura, K., Wu, I). J., et ul. ( 1 994). Nature 368, 753-756. Nowell, P. C. (1976). Science 194, 23-28. Noyer-Weidner, M., and Trautner, 7. A. (1993). In “DNA Methylation: Molecular Biology and Biological Significance” (J. P. Jost, and I+. P. Saluz, Eds.), pp. 39-1 08. Basel, Birkhauser Verlag. Ohtani-Fujita, N., Fujita, T., Aoike, A., et ul. (1993). Oncogene 8, 1063-1067. Oka, H., Shiozaki, H., Kohayashi, K., e t d . (199.3). Cancer Res. 53, 1696-1701. Okamoto, A., Demctrick, I). J., Spillarc, E. A,, et ul. (1994). Proc. Nut/. Acud. Sci. U.S.A. 91, I 1045-1 1049. Ono, T., Yamamoto, S., Kurishita, A,, et a/. (1990). Mut. Rrs. 237, 239-246. Ottaviano, Y. L,Issa, J. P., Parl, F. F., et ill. ( I 994). Cuncrr Rrs. 54, 25.52-2.555. Otterson, G. A,, Khleif, S. N., Chen, W., e t u l . (1995). Onc-ogrne 11, 121 1-1216. Otterson, G. A., Kratzkc, R. A., Coxon, A., e t a / . (1994). Oncogene 9, 337.5-3378. Paroush, Z., Keshet, I., Yisraeli, J., et ‘11. (1990). Cell63, 1229-1237. Petri, M. C., Mandelli, F., Zagonel, V., e t a / . (1993). I.errketniu 7, 3 6 4 1 . Pfeifer, G. P., Steigerwald. S. D., Hansen, R , et a/. (1990). Proc. Natl. Acad. Sri. U.S.A. 87, 82.524256. Pieper, R. O., Patel, S., Ting, S. A., rt ul. ( 1 996). J. Bio/. (:hem. 271, I39 16-1 3924. Pieretti, M., Zhang, I:., Fu, Y., el a/. ( 199 1). Cell 66, 8 17-822. Pogrihny, I. P., Poirier, 1.. A., and James, S. J. (1995). (~urcirrogrriesis16, 2863-2867. Quentin, Y. ( 1994). Ntrcl. Acids Res. 25, 2222-2227. Rainier, S., and Feinhcrg, A. P. (1988). Proc-. Nut/. Arud. Sci. U.S.A. 85, 6384-6388. Ramchandani, S., MacLeod, A. R., Pinard, M., et ill. (1997). Pror. Natl. Arud. Sci. U.S.A. 94, 684-689. Razin, A., and Cedar, H. ( I 991 ). Mirrohiol. Rev. 55, 45 1-458. Razin, A., and Cedar, H. ( I 993). 171 “DNA Methylation: Molecular Biology and Biological Significance” (J. P. Jost, and H. P. Saluz, Eds.), pp. 341-357. Basel, Birkhauser Verlag. Razin, A., and Riggs, A. 11. (1980).Science 210, 604-609. Razin, A., and Shemer, R. (1995). Humun Mol. Genet. 4, 1751-1755. Reed, A. L., Califano, I., Cairns, P., rt ul. (1996). Cancer Res. 56, 3630-3633. Riggs, A. 1). ( I 9 9 0 ) . Phi/. Trans. R. SOC. h n d . 326, 285-297. Riggs, A. D., and Ptciter, C. P. (1992). Trends Genet. 8, 169-174. Rouleau, J., Tanigawa, C., and Szyf, M. ( 1992)./. Biol. Chenz. 267, 7368-7377. Sakai, T., Toguchida, J., Ohtani, N., et al. ( 1 9 9 1 ) .Anr J. Hum. Grrret. 48, 880-888. Santoro, R., D’Erme, M., Mastrantonio, S., eta/. (199.5). Biocheni. J. 305, 739-744. Sapienza, C.,Peterson, A. C,.Rossant, J., et al., (1987).Nutrrre. 328, 2.5 1-254. Sasaki, H., Allen, N. D., and Surani, M. A. (1993). In “DNA Methylation: Molecular Biology and Biological Significance” (J. P. Jost, and H. P. Saluz., Eds.), pp. 469-486. Basel, Birkhauser Verlag.

DNA Methylation and Neoplasia

195

Sam, T., Tanigami, A., Yamakawa, K., eta/. ( 1 990). Cancer Res. 50, 71 84-71 89. Schmid, M., Grunert, D., Haaf, T., et al. (1983). Cytogenet. Cell Genet. 36, 553-561. Schmutte, C., Yang, A. S., Nguyen, T. T., et a/. ( 1 996). Cancer Res. 56, 2375-238 1. Selig, S., Ariel, M., Goitein, R., et (11. (1988).EMBOJ. 7, 419-426. Selig, S., Okumura, K., Ward, D. C., et a/. (1992). EMBOJ. 11, 1217-1225. Selker, E. U. (1993a).E X S 64,212-217. Selker, E. U., Fritz, D. Y., and Singer, M. J. ( 1 993h). Science 262, 1724-1 728. Selker, E. U., Jensen, B. C., and Richardson, G . A. (1987).Science 238,48-53. Serrano, M., Hannon, G. J., and Beach, D. ( 1993). Nature 366, 704-707. Serrano, M., Lee, H. W., Chin, L., et a/. ( 1 996). Cell 85, 27-37, Shapiro, G. I., Edwards, C. D., Kobzik, L., et al. (199Sa).Cancer Res. 55, 505-509. Shapiro, G. I., Park, J. E., Edwards, C. D., e t a / . (1 99Sb). Cancer Res. 55, 6200-6209. Shen,J., Rideout, W. M., and Jones, P. A. (1992).Cell 71, 1073-1080. Shmookler Reis, R. J., and Goldstein, S. (1982).Nucleic Acids Res. 14, 4293-4304. Shull, M. M., Ormsby, I., Kier, A. B., et a/. (1992).Nature 359, 693-699. Sidransky, D., Mikkelsen, T., Schwechheimer, K., e t a / . ( 1 992). Nature 355, 846-847. Silverman, L. R., Holland, J. F., Weinherg, R. S., et al. (1993).Leukemia 7, 21-29. Singer-Sam, J., and Riggs, A. D. (1993). In “DNA Methylation: Molecular Biology and Biological Significance” (J. P. Jost, and H. P. Saluz, Eds.), pp. 358-384. Basel, Birkauser Verlag. Slagboom, P. E., de Leeuw, W. J., and Vijg, J. (1990).Mech. Ageing Dev. 53,243-257. Smith, J. R., and Pereira-Smith, 0. M. (1996).Science 273, 63-67. Smith, S. S., Kan, J. I.. C., Baker, D. J., e t a / . (1991).J. Mol. Biol. 217, 39-51. Spruck, C. H., 3rd, Gonzalez-Zulueta, M., Shihata, A., et a/. (1994). Nature 370, 183-1 84. Stewart, C. L., Stuhlmann, H., Jahner, D., et a/. (1982). Proc. Nut/. Acad. Sci. U.S.A. 79, 409841 02. Stone, S., Dayananth, P., Jiang, P., et a/. (1995). Oncogene 11, 987-991. Strauss, M., Lukas, J., and Bartek, J. (1995).Nature Med. 1, 1245-1246. Surani, M. A., Reik, W., and Allen, N. D. (1988).Trends Genet. 4 , 5 9 4 2 . Sussenhach, J. S., Rodenburg, R. J., Scheper, W., et a/. (1993).Adu. E x p . Med. B i d . 343,63-71. Swafford, D. S., Middleton, S. K., Palmisano, W. A., et a[. (1997). Mol. Cell. B i d . 17, 13661374. Swain, J. L., Stewart, T. A., and Leder, P. ( 1987). Cell 50, 71 9-727. Swisshelm, K., Disteche, C. M.,Thorvaldsen, J.. etal. (1990).Mutat. Res. 237, 131-146. Szyf, M., Bozovic, V., and Tanigawa, G. (1991).J . Biol. Chem. 266, 10027-10030. Szyf, M., Kaplan, F., Mann, V., et al. (1985).J. Biol. Chem. 260 86534656. Szyf, M., Rouleau, J., Theherge, J., e t a / . (1992).J. Biol. Chem. 267, 12831-12836. Szyf, M., Theberge, J., and Bozovic, V. (1995).J. B i d . Chem. 270, 12690-12696. Tajima, S., Tsuda, H., Wakabayashi, N., e t a / . (1995).J. Biochem. 117, 1050-1057. Tare, P., Skarnes, W., and Bird, A. (1996).Nature Genet. 12,205-208. Tazi, J., and Bird, A. ( 1990). Cell 60, 909-920. Tollefsbol, T. O., and Hutchison, C. A. (1995).J. Biol. Chem. 270, 18543-18550. Toth, M., Lichtenberg, U., and Doerfler, W. (1989). Proc. Nut/. Acad. Sci. U.S.A. 86, 3728-3 732. Tucker, K. L., Beard, C., Dausniann, J., et al. (1996a). Genes Dev. 10, 1008-1020. Tucker, K . L., Talhot, D., Lee, M. A., et al. (1996b). Proc. Natl. Acad. Sci. U.S.A. 93, 12920-12925. Uehara, Y., Ono, T., Kurishita, A., et al. (1989).Oncogene. 4, 1023-1028. Umbas, R., Schalken, J. A., Aalders, T. W., e t a / . (1992). Cancer Res. 52, 5104-5109. Vertino, P. M., Issa, J. P., Pereira-Smith, 0. M., e t a l . (1994)Cell Growth Differ. 5, 139.5-1402. Vertino, P. M., Yen, R. C., Gao, J., et al. (1996).Mol. Cell Biol. 16,4555-4565. Vleminckx, K., Vakaet, L., Mareel, M., etal. (1991).Cell 66, 107-119.

I96

Stephen B. Baylin el of.

Vu, T. H., and Hoffman, A. R. (1994).Nature 371,714-717. Wainfan, E., and Poirier, L. A. (1992).Cancer Res. 52, 2071-2077. Wales, M. M., Biel, M. A., el Deiry, W., et al. ( 1995). Nature Med. 1, 570-577. Walker, C., and Nettesheim, P. (1986). Cancer Res. 46, 6433-6437. Washimi, O., Nagatake, M., Osada, H., e t a / . (1995).Cancer Res. 55, 514-517. Watts, C. K. W., Parker, M. G., and King, R. J. B. (1989).j. Steroid Biochem. 34,4831190. Weinberg, R. A. (1995).Cell 81,323-330. Weis, K., Rambaud, S., Lavau, C., et al. ( 1 994). Cell 76, 345-356. Weiss, A., Keshet, I., Kazin, A., et a/. ( 1996). Cell 86, 709-718. Weisshach, A. ( 1 993). In “DNA Methylation: Molecular Biology and Biological Significance” (J. P.Jost, and H. P. Saluz, Eds.), pp. 1-10. Basel, Birkhauser Verlag. Willemze, R., Archimbaud, E., and Muus, P. (1993).Leukemia. 7, 49-50. Wilson, V. L., and Jones, P. A. ( I 983). Science 220, 1055-1057. Wolf,S. F., Jolly, D. J.. Lunnen, K. D.,etal. (1984).Proc. Natl. Acad. Sci. U.S.A. 81,2806-2810. Wolffe, A. P. (1994).Dett Genet. lS, 463-470. Wu, I., Issa, J.-P., Herman, I., et al. ( I 993). Proc. Natl. Acad. Sci. U.S.A. 90, 8891-8895. Wu, J. J. Herman, J. G., Wilson, G., et al. (1996).Cancer Res. 56, 616-622. Yen, R. W., Vertino, P. M., Nelkin, B. D., et a / . (1992). Nucleic Acids Res. 20, 2287-2291. Yoder, J. A., Yen, R. C . , Vertino, P. M., e t a / . (1996).1. Biol. Chem. 271,31092-31097. Yoshiura, K., Kanai, Y., Ochiai, A.,etal. (1995).Proc. Natl. Acad. Sci. U.S.A. 92,7416-7419. Young, H. A. Ghosh, P., Ye, J., e f a l . (1994).j. Immunol. 153, 3603-3610. Zagonel, V., Re, G. I.., Marotta, G., et a / . ( 199.3). Leukeniia 7, 30-35. Zajchowski, D. A., Sager, R., and Wehster, L. (1993).Cancer Res. 53, 5004-501 1. Zardo, G., Santoro, R., D’Frme, M., rt a / . (1996). Biorhem. Biophys. Res. Comm. 220, 102-1 07. Ziegler, A., Jonason, A. S., Leffell, D. J., e t a / . ( 1994). Nature 372, 773-776.

AramC: Cellular and Molecular Pharmacology Steven Grant Departments of Medicine, Pharmacology, and Microbiri/ogy Medical College of Virginia Virginia Commonruealth University Richmovd, VA 2 3298

1. Introduction 11. Structure 111. Metabolism A. Transport B. Phosphorylation C. Deamination D. ora-CPT IV. Mechanisms of Cytotoxicity A. Inhibition of DNA Polymerases B. DNA Incorporation C. Endoreduplication D. Generation of Reactive Oxygen Intermediates E. Perturbation in Membrane Lipids and Glycoproteins F. DNA Fragmentation and Apoptosis V. Mechanisms of Resistance A. Reduced Transport B. Decreased Activity of Deoxycytidine Kinase C. Enhanced Deamination D. Increased Rates of ara-CTP Dephosphorylation E. Increased dCTP Pool Sizes F. Alterations in CTP Synthetase G. Cytokinetic Resistance H. Defects in the Cell Death Pathway IV. Signaling Pathways and Oncogene Interactions A. c-jun B. NF-KP C. pRb D. c-myc E. 6 e l - 2 F. p21 (WAFl/CIPl) G. E2F H. 1>34CdC' 1. Stress- and Mitogen-activated Protein Kinases J. PKC-S K. c-a61 L. Histone H1

Advances in CANCER RE5EARCH O O ~ S - ~ X / Y ~

$zm

Copyright 0 1998 hy Academic Press. All rights of reproducrion In Jny furm r e w v e d .

I98

Steven Grant

M. ras N. ECKl 0. CycE-ti I K VII. Modulation of ara-C-associated Cytotoxicity A. Biochemical Modulation R. Hematopoietic Growth Factors C. Modulation of Signal Transduction-Apoptotic Pathways V111. Conclusions References

The antiinetaholite cytosine arabinoside (ara-C) represents a prototype of the nucleoside analog class of antineoplastic agents and remains one of the most effective drugs used in the treatinelit of acute leukemia as well as other hematopoietic malignancies. The ability of ara-C to kill neoplastic cells is regulated at three distinct but interrelated levels. First. the activity of ara-C depends on conversion to its lethal triphosphate derivative, ara-CTP, a process that is intluenced by multiple factors, including nucleoside transport, phosphorylation, dcaniination, and levels of competing metabolites, particularly dC:Il? Second, the antiproliferative and lethal effects of ara-C are linked to the ability of ara-CTP to interfere with one o r more DNA polymerases as well as the degree to which it is incorporated into elongating D N A strands, leading to DNA fragmentation and chain termination. Finally, the fate of the cell is ultimately determined by whether a threshold level of ara-C-mediated D N A damage is exceeded, thereby inducing apoptosis, or programmed cell death. The latter process is influenced hy components of various signal transduction pathways (c.g., PKC) and expression of oncogenes (e.g., I d - 2 , c-Jun), perturbations i n which may significantly alter ara-C sensitivity. A better understanding of these factors could eventually lead to the development of novel therapeutic strategies capable of overcoming ara-C resistance and improving therapeutic efficacy.

1. INTRODUCTION The antimetabolite cytosine arabinoside (1-P-D-arabinofuranosylcytosine; ara-C) represents a member of the class of nucleoside analog antineoplastic agents. The antimetabolites are generally classified into two major subcategories. The first includes agents that interfere with enzymes of de novo pyrimidine or pyrimidine biosynthesis. Typical examples of this class are methotrexate, an inhibitor of dihydrofolate reductase, and N-phosphonacetyl-l-aspartate (PALA), a transition-state inhibitor of aspartate transcarbamylase. The nucleoside analogs represent the second major class of antimetabolites. These agents, which differ structurally in a critical way from naturally occurring purine and pyrimidine nucleosides, are incorporated into newly synthesized DNA or RNA using salvage enzyme pathways. Examples include, aside from ara-C, the pyrimidine analogs 5-fluoruracil (5FU) and 5-AZA-cytidine, and the purine analogs 6-mercaptopurine and 2’-

Ara-C: Cellular and Molecular Pharmacology

I99

chlorodeoxyadenosine. In fact, a number of antimetabolites can be viewed as members of both of these major subcategories. For example, 5-FU is converted intracellularly to 5-fluorodeoxyuridylate (FdUMP),a potent inhibitor of thymidylate synthase (Santi et al., 1974). In addition, 5-FU, after conversion to its triphosphate derivative (5-FUTP), is incorporated into RNA, leading to disruption of function (Kufe et al., 1981a). As discussed subsequently, ara-C may act analogously by inhibiting DNA polymerase and by being incorporated into elongating DNA strands. In both cases, the relative contribution of these two events (e.g., inhibition of de novo pyrimidine biosynthesis versus analog incorporation into nucleic acids) to cell death remains the subject of debate. It is also conceivable that simultaneous inhibition of de novo pathways and disruption of nucleic acid function stemming from analog incorporation is responsible for, o r contributes to, cell lethality. For a variety of reasons, ara-C has remained the focus of intense interest from both basic and clinical standpoints. Ara-C is arguably the most effective agent in the treatment of acute myelogenous leukemia in humans (Mastrianni et al., 1992) and is incorporated into virtually all standard induction regimens for this disease (Lister et al., 1987). It is also active in other hematologic malignancies, including acute lymphoblastic leukemia (ALL) (Stryckmans et al., 1987) and non-Hodgkin’s lymphoma (Peters et al., 1987). During the past decade, the feasibility of administering ara-C as a high-dose bolus infusion has become firmly established (Heinemann and Jehn, 1990). Such regimens achieve plasma ara-C levels 2 logs or more higher than those obtained with standard-dose therapy (Peters et al., 1988). In addition, a subset of patients with refractory or relapsed disease respond to a high-dose araC regimen (Herzig et d.,1983). Unfortunately, remissions obtained with high-dose ara-C tend to be short-lived, and ultimately, patients relapse with highly resistant disease refractory to all subsequent forms of therapy (Estey, 1996). Consequently, a clear rationale exists for efforts to understand the basis for ara-C action and to design strategies capable of circumventing resistance. There are other compelling reasons for continuing attempts to define the basis for the lethal actions of ara-C. The past decade has witnessed a dramatic increase in interest in a process referred to as apoptosis, or programmed cell death (Wyllie et al., 1980). It is now widely recognized that ara-C, as well as a host of other cytotoxic agents, may kill neoplastic cells by triggering this process (Kaufmann, 1989). In addition, it has become apparent that the susceptibility of cells to apoptosis depends on the status of genes involved in regulating various components of the cell death machinery (Green et al., 1996). These, in turn, are influenced by signal transduction events, resulting in a complex system of interacting signals that ultimately determine whether a cell exposed to a cytotoxic agent is destined to die (Gajewski and Thompson, 1996). Considerable information is available

200

Steven Grant

concerning alterations in gene expression and perturbations in signal transduction pathways accompanying ara-C-mediated cytotoxic injury. Elucidation of the mechanisms responsible for translation of early ara-C-related cell injury events into activation of the cell death machinery could have broad implications for efforts to understand the mode of action of diverse antineoplastic agents. The purpose of this review is to summarize available information pertinent to ara-C with respect to its cellular metabolism, mode of action, mechanisms of resistance, and relation to molecular events liked to signal transduction pathways and apoptosis. Extensive clinical information relating to ara-C is available elsewhere and is not reviewed here. Other broad-based reviews have appeared previously (Pallavicini, 1984; Rustum and Raymakers, 1992).

11. STRUCTURE The structures of ara-C and that of its naturally occurring congener 2’-deoxycytidine are shown in Figure 1. In contrast to the 2’-hydoxyl group of the ribose sugar of 2‘-deoxycytidine, which occupies the cis position, placement of the 2’-hydroxyl group in the trans configuration gives rise to an arabinosyl sugar backbone. Early studies demonstrated that the 2’-trans hydroxyl group of the arabinosyl moiety interferes with rotation of the cytosine ring structure around the glycosidic bond (Adler et al., 1968). The trans position of the 2’-hydrogen also permits a strong bond to be formed with the 5’-oxygen of the cytosine ring (Chwang and Sundaralingam, 1973). Differences in the relation between the pyrimidine ring and the arabinosyl sugar are likely to contribute to the differential sensitivity of ara-C and deoxycytine to enzymatic digestion (Nagyvary et al., 1968). It is presumed that the altered reactivity of the 2’-hydroxyl moiety of the arabinosyl sugar is responsible for perturbations in the interaction between 1-p-D-arabinofuranosyletyosine 5‘-triphoshate (ara-CTP) and DNA polymerase-a (Yoshida et al., 1977).

111. METABOLISM Ara-C represents a prodrug in that it must be converted to its lethal triphosphate derivative (ara-CTP) to exert its cytotoxic effects (Furth and Cohen, 1968; Graham and Whitmore, 1970). Like all nucleoside analogs, ara-C is acted o n by salvage pathway enzymes, a feature that it shares with

Ara-C: Cellular and Molecular Pharmacology

20 1

NH2 I

OH 2’-deoxycytidine Fig. I

NH2 I

OH

ara-C

The Structure of 2’-deoxycytidine and its 2’-hydroxy derivative, ara-C.

the naturally occurring nucleoside 2’-deoxycytidine. In fact, ara-C and its derivatives compete with 2’-deoxycytidine and corresponding nucleotides at each step in the metabolic pathway. For this reason, 2’-deoxycytidine is a potent antagonist of ara-C action and interferes with its activity both in vitro and in vivo (Bhalla et al., 1987; Evans and Mengel, 1964). Ara-C is transported across cell membranes by a facilitated nucleoside diffusion mechanism, after which it is converted to its nucleotide derivative by the pyrimidine salvage pathway enzyme, deoxycytidine kinase. After subsequent phosphorylation by a mono-phosphate and diphosphate kinase, ara-C is converted to its active form, ara-CTP. Ara-CPT inhibits DNA polymerase-a and p, and is incorporated into elongating DNA strands, where it may occupy either intranucleotide or chain terminus positions (Mikita and Beardsley, 1988). Opposing these anabolic processes are the degradative enzymes cytidine deaminase and deoxycytidylate deaminase, which convert ara-C and ara-CMP, respectively, to inactive, ara-U, derivatives. Finally, the intracellular disposition of ara-C is influenced by multiple feedback mechanisms that regulate various steps in the pyrimidine biosynthetic pathway. Notably deoxycytidine triphosphate (dCTP) is a potent feedback inhibitor of deoxycytidine kinase, the enzyme catalyzing the rate-limiting step in ara-C metabolism (Momparler and Fischer, 1968). Consequently, ara-C activation and ultimately, ara-C lethality, depend on a complex interplay between ara-C anabolism, degradation, and intracellular levels of endogenous purine and pyrimidine nucleosides. A schematic diagram of ara-C metabolism and the factors influencing this process is shown in Figure 2.

202

ara-C

Steven Grant

- - -ara-C

0 dCTP

are-CMP

0dTTP dCyd

dCyd

dCK

*dCMP

ara-CTP

-

DNA pol

DNA

-dCTP

Fig. 2 The metabolism of ara-C and 2’-cteoxycytidine. The enzyme catalyzing the rate-limiting step in ara-C metabolism, dCK, is under negative feedback regulatory control by dCTP and is stimulated by d7TP. The antimetabolites ara-C and ara-CMP arc deaminanted by cytidine deaminase and dCMP deaminase, respectively. DNA polymerase-a (DNA pol), which is inhihired by ara-CTP, is responsible for incorporation of ara-CTP residues into DNA.

A. Transport Ara-C (and other nucleosides) are transported across the cell membrane by an energy-dependent nucleoside diffusion mechanism (Plagemann et al., 1978; Kessel et al., 1967). At low extracellular ara-C concentrations (i.e., 1 OM or lower), transport represents the rate-limiting step in the intracellular accumulation of this nucleoside, whereas a t high concentrations, phosphorylation predominates (White et al., 1987). In one study, leukemic blasts obtained from patients who did not respond to ara-C displayed low numbers of nucleoside transport-binding sites (Wiley et al., 1982). In primary acute myelogenous leukemia (AML) blast specimens, the number of nucleosidebinding sites varies considerably, ranging from fewer than 1000 to more than 30,000 (Wiley et al., 1989,1985;Jamieson et al., 1990).Furthermore, a correlation has been reported between the proliferative rate of leukemic blasts and nucleoside-binding sites (Wiley et al., 1992),although this has not been a universal finding (Powell et al., 1991).Attempts have been made to incorporate modulation of ara-C transport in the design of multidrug regimens. For example, administration of the nucleoside transport blocker dipyridamole after, but not before, ara-C led to synergistic antileukemic effects in human promyelocytic leukemic cells (HL-60),presumably by blocking araC efflux (Chan, 1989). Conversely, pretreatment of primary leukemic blasts displaying low numbers of nucleoside-binding sites with either VP-16 or VM-26 blocked subsequent ara-C transport and metabolism (White et al., 1985). Most recently, blasts from leukemic patients primed with recombinant granulocyte-macrophage colony-stimulating factor (rCM-CSF) exhibited an increase in proliferative characteristics as well as nucleoside-binding sites (Wiley et al., 1994), raising the possibility that this strategy might be capable of increasing the intracellular accumulation of ara-C in these cells.

Ara-C: Cellular and Molecular Pharmacology

203

B. Phosphorylation Once inside the cell, ara-C is converted to its nucleoside monophosphate derivative, ara-CMP, by the pyrimidine salvage pathway enzyme, deoxycytidine kinase (dCK).This process represents the rate-limiting step in ara-C metabolism (Plagemann et al., 1978; Kufe and Spriggs, 1985). Phosphorylation of ara-C also serves to trap drug intracellularly by preventing drug efflux. Deoxycytidine kinase is relatively nonspecific in its actions, and it is known to phosphorylate, in addition to ara-C, multiple nucleosides and nucleoside analogs, including deoxyadenosine, deoxyguanosine, chlorodeoxyadenosine, fludarabine, and difluorodeoxycytidine, among others (Brockman et al., 1980; Carson et al., 1980; Hertel et al., 1990; Keating et al., 1989). Deoxycytidine kinase is under allosteric regulatory control by deoxyribonucleoside triphosphates (Durham and Ives, 1970). Of these, the most potent inhibitor of dCK activity is dCTP; conversely, dTTP is a strong positive regulator of enzyme activity (Durham and Ives, 1970). Because intracellular levels of dCTP are to a large extent influenced by the activity of ribonucleotide reductase (Moore and Hulbert, 1984),attempts have been made to exploit this phenomenon by combining ara-C with inhibitors of this enzyme, including thymidine (Streifel and Howell, 1981), deoxyguanosine (Ross et al., 1984), and most recently, fludarabine (discussed later) (Gandhi and Plunkett, 1988). Deoxycytidine kinase obtained from human granulocytes exhibits a Km (K maximum; enzyme constant) for ara-C of 25.6kM, compared with 7.8 + M for deoxcytidine, moreover, levels of deoxycytidine kinase decrease as cells undergo maturation (Coleman et al., 1975). Human deoxycytidine kinase (dCK) has been cloned (Chottiner et al., 1991; Huang et al., 1989) and demonstrated to encode a 30.5=kd protein capable of phosphorylating dCyd, deoxyadenosine, and deoxyguanosine. T lymphoblasts resistant to ara-C and dideoxycytidine exhibit low levels of dCK mRNA, consistent with the view that this protein is responsible for the phosphorylation of ara-C (Chottiner et al., 1991). The relation between leukemic cell dCK activity and ara-C responsiveness remains to be resolved. Loss of dCK activity is a common mechanism of araC resistance in vitro (Bhalla et at., 1984) and has been related to mutations within the deoxycytidine gene (Owens et al., 1992). Attempts to correlate levels of dCK activity in leukemic blasts, o r indirect measures of activity (i.e., ara-C DNA incorporation; ara-CTP formation) with responsiveness to therapy have not been successful (Ross et al., 1986). Nevertheless, several studies have shown that transfection of both hematopoietic and nonhematopoietic malignant cell lines with with retroviral vectors containing dCK cDNA substantially increases their susceptibility to ara-C and other nucleoside analogs (Hapke et al., 1996; Stegmann et al., 1995). Conversion of ara-CMP to the lethal triphosphate derivative ara-CTP involves successive phosphorylations by dCMP kinase and dCDP kinase, re-

204

Steven Grant

spectively. These low-affinity enzymes are relatively abundant in cells and, consequently are not rate limiting in the formation of ara-CTP (Hande and Chabner, 1978).

C. Deamination Opposing the anabolism of ara-C are two deaminases that convert ara-C and ara-CMP to inactive ura-U derivatives. The first of these is cytidine deaminase, which is present in granulocytes, plasma, liver, and red blood cells (Caminier and Smith, 1965). I t converts ara-C to ara-U and in normal granulocytes exhibits a greater affinity for its natural substrate (cytidine) than for ara-C (K,,, 1.1 x lo-' versus 8.8 x l o p 5 )(Chabner etal., 1974). The gene for cytidine deaminase has been cloned and encodes a 16.3=kd protein exhibiting high levels of enzyme activity (Laliberte and Momparler, 1994). Although an early report suggested that high levels of cytidine deaminase in leukemic blasts correlated with resistance to ara-C therapy (Steuart and Burke, 1971), this finding has not been substantiated. Attempts to combine ara-C with cytidine deaminase inhibitors, such as tetrahydrouridine, appear to alter ara-C pharmacokinetics, but it is unclear whether this strategy leads to an improvement in the therapeutic index (Kreis et al., 1992). Ara-CMP may also be converted to an inactive form, ara-UMP, by deoxycytidylate deaminase (Mancini and Cheng, 1982), which is also responsible for the deamination of other nucleoside analogs, including 5-FU (Caradonna and Cheng, 1981). Although some studies have shown that the deoxycytidylate inhibitor dTHU potentiates ara-C metabolism in lymphoblastic leukemic cell lines (Ellims et al., 1983; Fridland and Verhoef, 1987), it is unclear whether this effect specifically stems from inhibition of deoxycytidylate deaminase o r cytidine deaminase (Grant et al., 1991 ).An alternative possibility is that dTHU, by inhibiting dCMP deaminase and thereby the synthesis of dUMP (Maley and Maley, 1971), increases intracellular concentrations of dCTP (Heineman and Plunkett, 1989). This, in turn, reduces the formation of ara-CTP through feedback inhibition of deoxycytidine kinase (Liliemark and Plunkett, 1986). Consistent with this notion is the observation that cells deficient in deoxycytidylate deaminase activity display increase dCTP pools and resistance to ara-C (de Saint Vincent and Buttin, 1981). Consequently, deoxycytidylate deaminase may play two potentially opposing roles in the regulation of ara-C-mediated cytotoxicity: (1) direct, through deamination of ara-CMP; and (2) indirect, through depletion of dCTP pools. The net effect of deoxycytidylate deaminase on ara-C metabolism would therefore depend on which of these processes predominated, which may in turn be cell type specific.

Ara-C: Cellular and Molecular Pharmacology

205

D. ara-CTP The formation of ara-CTP represents a requisite step in ara-C-mediated toxicity. This metabolite acts in at least two ways: (1)as an inhibitor of DNA polymerases, and (2)after incorporation into elongating DNA strands, as an inhibitor of DNA elongation and initiation. Ara-CTP is a competitive inhibitor of DNA polymerase-a, although a relatively weak one (Graham and Whitmore, 1970; Momparler, 1969; Furth and Cohen, 1968; Chu and Fischer, 1965). In fact, the comparable affinity of ara-CTP and dCTP for DNA polymerase-a and the ability of relatively low concentrations of ara-CTP to inhibit DNA synthesis (Graham and Whitmore, 1970) suggest that factors other than inhibition of this enzyme are primarily responsible for ara-C actions. When present a t high intracellular concentrations, ara-CTP inhibits DNA polymerase=P and, consequently, DNA repair (Dunn and Regan, 1979; Hiss and Preston, 1977; Fram and Kufe, 1982; Miller and Chinault, 1982). It is possible that such actions may contribute to synergistic interactions between high-dose ara-C and other DNA-damaging agents (Ohno et al., 1988). Once formed, ara-CTP may be dephosphorylated, which limits the intracellular accumulation of this metabolite (Abe et al., 1982; Jamieson et al., 1987). In studies involving animals (Rustum, 1982) and humans (Preisler et al., 1979), the intracellular retention of ara-CTP has been shown to correlate with response to ara-C therapy. This phenomenon may stem from the fact that cells that exhibit higher ara-C concentration -x time ratios display higher levels of ara-C DNA incorporation (Kufe and Spriggs, 1985).

IV. MECHANISM OF CYTOTOXICITY A. Inhibition of DNA Polymerases The mechanisms by which ara-C exerts its lethal effects remain to be elucidated fully. Although initial studies suggested a link between ara-C-mediated cell killing and DNA polymerase inhibition, other findings made it difficult to attribute lethality to this phenomenon (Momparler, 1969). For example, although ara-CTP is a competitive inhibitor of DNA polymerasecx (Furth and Cohen, 1968; Graham and Whitmore, 1970; Matsukage et al., 1978a) and to a lesser extent, DNA polymerase+ (Yoshida et al., 1977; Dunn and Reagan, 1979; Ohno et al., 1988; Matsukage et al., 1978; Miller and Chinault, 1982), the effects of ara-C on cell death, as opposed to DNA synthesis inhibition, are generally irreversible on removal of the drug. In addition, as noted earlier, ara-C-associated inhibition of DNA synthesis occurs

206

Steven Grant

at ara-C concentrations far lower than those required to inhibit either polymerase (Graham and Whitmore, 1970). Together, these observations strongly suggest that factors other than, o r in addition to, inhibition of DNA polymerase are responsible for the lethal actions of ara-C.

B. DNA Incorporation A substantial body of evidence indicates that incorporation of ara-C into DNA represents a major mode of ara-C-mediated lethality. Although araCTP acts as a n inhibitor of DNA polymerase-a, it also serves as a substrate for this enzyme and competes with dCTP for incorporation into elongating DNA strands (Major et al., 1981; Kufe et al., 1981b; Major et al., 1982; Zahn et al., 1972). Strong evidence has been presented indicating that the degree of ara-C incorporated into leukemic cell DNA correlates highly with araC-associated lethality, reflected by the loss of clonogenic potential (Major et al., 1981). There are several mechanisms by which incorporation of ara-C into DNA can interfere with DNA synthesis. Early studies suggested that incorporation of ara-C residues into DNA strands interferes with DNA chain initiation (Fridland, 1977a,b). In addition, the presence of an ara-CMP residue at the 3' DNA chain terminus interferes with chain elongation by DNA polymerase-a and p (Mikita and Beardsley, 1988; Ohno et al., 1988). The position of ara-C residues within the DNA strand, and consequently the degree of chain termination, depends to a large extent on the concentration of ara-C. For example, at relatively low concentrations, ara-C slows but does not halt DNA synthesis, and the residues appear primarily in internucleotide positions (Kufe et al., 1984b). In this setting, ara-C acts as a relative chain terminator, as had previously been proposed (Cozzarelli, 1977). Studies in cell-free systems confirm that incorporation of ara-CMP residues in internucleotide positions leads to a profound slowing of DNA chain elongation (Perr h o and Mekosh, 1992). In contrast, high concentrations of ara-C lead to an increasing proportion of ara-C residues at the chain terminus (Major et a/., 1981; Kufe et al., 1984). Under these circumstances, ara-C functions as an absolute chain terminator. The ability of DNA chain elongation to proceed in the presence of ara-C residues presumably requires excision repair, possibly by DNA polymerase-P (Miller and Chinault, 1982). It is plausible that high concentrations of ara-CTP inhibit this process, resulting in the appearance of ara-C residues at the chain terminus. Although ara-C may inhibit initiation of DNA synthesis at the replicon initiation site, slowing of DNA elongation within replicons has been reported (Heintz and Hamlin, 1983). One of the consequences of ara-C-mediated interference with DNA chain elongation is the accumulation of nascent DNA replication intermediates, including Okazaki fragments as well as long and short subgenomic-length DNA. Using an alkaline elution technique, Ross and colleagues (1990)

Ara-C: Cellular and Molecular Pharmacology

207

demonstrated that exposure of human leukemic cells to ara-C resulted in inhibition of thymidine (dThd) incorporation into full-length DNA but not into Okazaki fragments or low-molecular-weight nascent DNA. These findings suggest that ara-C acts primarily to inhibit nascent DNA chain elongation rather than to induce chain termination or to block initiation. Consistent with this notion, the ability of vitamin D, to increase ara-C-mediated lethality in human leukemic cells has been attributed to interference with DNA replication intermediates (Studzinski et al., 1991). More recently, the development of a DNA replication complex model (MRC) has provided further insights into the mechanism by which ara-C interferes with DNA synthesis and induces cell death. The M R C represents a multiprotein complex consisting of various components, including DNA polymerase-a and &DNA primase, topoisomerases I and 11, proliferating nuclear cell antigen (PCNA),and DNA ligase I, among others (Wu et al., 1994). Studies employing this model indicate that, as in the case of intact cells (Ross et al., 1992; Dijkwel et al., 1978), ara-CTP residues are initially incorporated into short (nascent)DNA fragments, followed by incorporation into larger DNA species, where they reside in internucleotide positions (Wills et al., 1996). These events are accompanied by a general slowing of the DNA replicative process and eventually cell death.

C. Endoreduplication Studies by Woodcock and colleagues suggest that incorporation of ara-C residues into DNA leads to initiation of DNA synthesis in already replicated segments (rereplication or endoreduplication (Woodcock, 1987; Woodcock and Cooper, 1979, 1987; Woodcock et al., 1982, 1976). One possible mechanism by which this might occur is blocking of DNA replication forks by ara-C, followed by excision of ara-C residues, allowing reinitiation of DNA synthesis in segments that had already been replicated in the same S phase (Karon and Benedict, 1971). The aberrant rereplication may induce cell death by a variety of mechanisms, including inappropriate segregation of the genetic material and/or the formation of nonclonogenic giant cells (Woodcock, 1987). The induction of amplified DNA sequences by ara-C treatment has also been observed in the case of other lethal metabolites, including methotrexate (Schimke, 1984).

D. Generation of Reactive Oxygen Intermediates In addition to its capacity to interfere with DNA replication, there is evidence that ara-C may also exhibit toxicity through the generation of reactive oxygen intermediates (ROI). For example, after exposure to ara-C,

208

S t e v e n Grant

leukemic cells undergo a sequence of events, including an increase in free radical production, an initial increase followed by a decline in glutathione-stimdating hormone (GSH) levels, and ultimately a loss of membrane integrity (Hu et al., 1995a). In contrast, cells overexpressing the antiapoptotic protein Bcl-2, which may protect against free radical-associated toxicity (Hockenberry et al., 1993), also showed an increase in ara-C-mediated free radical formation, but without the same depletion of GSH pools (Hedley and McCulloch, 1995b). The relative contributions to cell lethality of ara-C-induced free radical generation compared with interference with DNA synthesis remains to be determined.

E. Perturbations in Membrane Lipids and Clycoproteins Exposure of cells to ara-C results in alterations in the synthesis of membrane lipids and glycoproteins. For example, ara-CMP and ara-CTP interfere with the actions of sial yltransferases, which are involved in the synthesis of important membrane constituents, the sialylglycoproteins (Klohs et al., 1979; Hawtrey et a/., 1974). In addition, ara-C alters intracellular concentrations of several lipid second messengers, which may in turn modify the susceptibility of cells to drug-induced apoptosis (discussed later). In this context, exposure of leukemic cells to ara-C results in the generation of ara-CDP choline through reversal of the cholinephosphotransferase reaction, rather than through catalysis of ara-CTP triphosphate by CTP:phosphocholine cytidiylytransferase (Kucera and Capizzi, 1992). The former reaction leads to generation of ara-CMP and diradylglycerol, the target of which is protein kinase C (PKC), perhaps accounting for the increase in PKC activity observed in leukemic cells exposed to ara-C (Kharbanda et al., 1991). In addition to generating diradylglycerol, ara-C stimulates the formation of ceramide (Strum et al., 1994), a potent inducer of apoptosis in leukemic cells (Obeid et al., 1994; Jarvis et al., 1994b). Because PKC activation is known to oppose apoptosis in hematopoietic cells (Lotem et al., 1991), it is conceivable that the lethal actions of ara-C may depend, at least in part, on its relative effects on the PKC and sphingomyelin pathways. Further complicating the issue is the finding that PKC-a has been implicated in phosphorylation of ara-C by dCK, raising the possibility that ara-C may potentiate its own metabolism by perturbing lipid-signaling pathways (Wang and Kucera, 1994). A model of the putative cytotoxic and cytoprotective actions of ara-C-stimulated lipid second messengers and their relation to the potential antiapoptotic actions of PKC are illustrated in Figure 3 .

209

Ara-C: Cellular and Molecular Pharmacology

PMA,Bryo (Chronic)

PKC inhibitors

-

Bcl-2 dephosphorylation (?)

DRG

ara-CDP Cho

@ /csramide

ara-C (DNA)

dCK

ara-C

ara-C

--

+ Ptd Cho ara-CMP

/

ara-CTP

Fig. 3 Interactions between ara-C and signaling pathways involved in the regulation of apoptosis. There is evidence that PKC acts to oppose drug-induced cell death, an effect that can be antagonized by enzyme downregulation (e.g., by chronic exposure to certain PKC activators) o r by direct inhibition. PKC may exert this effect through alterations in intracellular pH, activation of various oncogenes, and possibly modulation of Bcl-2 phosphorylation. Exposure of leukemic cells to ara-C leads to generation of diradylglycerol (DRG), which may oppose cell death, and ceramide, which may promote this process. Ptd Cho, phosphatidylcholine; ara-CDP Cho, ara-CDPcholine; Bryo, bryostatin 1; PMA, phorbol myristate acetate; IER, immediate early response genes.

F. DNA Fragmentation and Apoptosis Ara-C may also kill cells directly by damaging genomic DNA. Ara-C induced strand breaks in both previously synthesized and replicating DNA (Fram and Kufe, 1982). The induction of genomic DNA damage by ara-C may occur in S-phase cells as well as in cells at the GJG, interface (Jones et al., 1976; Karon et al., 1972). It is uncertain whether this phenomenon is distinguishable from the inhibitory effects of ara-C on DNA synthesis. Studies on induction of cell death in neoplastic cells by cytotoxic agents, including ara-C, have focused on the process of apoptosis. Apoptosis represents an energy-dependent process in which a cell commits to a program of self-destruction (Wyllie et al., 1980). It is characterized by certain stereotypical morphologic features, including cell shrinkage, nuclear condensation, and the formation of membrane-bound nuclear remnants, referred to as apoptotic bodies (Walker et al., 1988). It is also frequently, but not invariably, accompanied by degradation of genomic DNA into integer multiples of 180- to 200-base-pair fragments, presumably owing to cleavage at vulnera-

210

Steven Grant

ble internucleosomal cleavage sites by one o r more of a family of Ca2+lMg2+dependent endonucleases (Duke et al., 1983).Apoptosis can be distinguished from necrosis, a passive process of cell death characterized by cell swelling, random lysosomal DNA degradation, and loss of osmoregulation (Wyllie, 1981). The ability of ara-C to trigger apoptosis in human leukemia cells is well documented (Kaufmann, 1989; Gunji et al., 1991). A key question still to be resolved is whether induction of apoptosis by ara-C stems from one or more of the actions cited previously or instead results from ara-C-mediated perturbations in diverse signaling cascades, described later. An additional possibility is that ara-C-induced apoptosis may be triggered by a combination of these events, that is, by concomitant interference with DNA replication in conjunction with activation of a DNA damage-response pathway.

V. MECHANISMS OF RESISTANCE Neoplastic cells may become resistant to ara-C by a wide variety of mechanisms. Consequently, it has been difficult to identify a predominant mode by which such cells escape the lethal effects of ara-C action. Because these mechanisms are not mutually exclusive, it is entirely possible that cells may employ more than one strategy to evade ara-C killing. Listed subsequently are biochemical alterations noted in leukemic cells resistant to ara-C after either in vitro or in vivo exposure.

A. R e d u c e d Transport As noted earlier (Wiley et al., 1982,1992), the number of nucleoside-binding sites on leukemic myeloblasts is frequently low. As a result, nucleoside transport may represent the limiting factor in ara-C metabolism under some condition. At ara-C concentrations of 3 0 p M or more, however, phosphorylation becomes the rate-limiting step in ara-C metabolism (White et al., 1987), and it is doubtful that a decrease in transport-binding sites would confer resistance in this setting. As noted previously, correlations between ara-C resistance in leukemia and nucleoside-binding sites has been observed by some investigators (Wiley et al., 1992)but not others (Powell et al., 1991 ).

B. Decreased Activity of Deoxycytidine Kinase Loss of dCK represents the most frequent mechanism by which leukemic cells develop resistance to ara-C both in vitro and in vivo (Drahovsky and

Ara-C: Cellular and Molecular Pharmacology

21 1

Kreis, 1970; Chu and Fisher, 1962; Bhalla et al., 1984). In at least some cases, loss of the enzyme stems from mutations within the dCK gene (Owens et al., 1992). Decreased activity of dCK prevents intracellular trapping of araCMP and conversion of ara-C to its lethal triphosphate derivative. Attempts to correlate clinical resistance to ara-C in leukemia with decreased dCK activity in blast cells have not led to definitive results (Smyth et al., 1976), although a suggestion of lower levels in nonresponding patients has been reported in some studies (Colly et al., 1987; Tattersall et al., 1974). Nevertheless, retroviral and adenoviral transfection of ara-C-resistant leukemic cells (Stegmann et al., 1995) and glioma cell lines (Manome et al., 1996) with dCK cDNA has resulted in significant increase in therapeutic activity in preclinical models.

C. Enhanced Deamination Conversion of ara-C to ara-U by cytidine deaminase prevents the formation of ara-CTP and abrogates the lethal effect of ara-C. High levels of cytidine deaminase (Stueart and Burke, 1971) and/or cytidine kinase-to-cytidine deaminase ratios (Colly et al., 1987) in leukemic blasts have been correlated with low response rates in a limited number of studies but these findings have not been universal (Smyth et al., 1976). High concentrations of ara-U may enhance ara-C-induced cytoxicity through a self-poteniating mechanism (Yang, et al., 1985). Uncertainty regarding the ability of the deaminase inhibitor tetrahydrouridine to enhance the therapeutic index of ara-C (Kreis et al., 1992) raises the possibility that factors other than, or in addition to, deamination of ara-C contribute to clinical resistance.

D. Increased Rates of ara-CTP Dephosphorylation As noted previously ara-CTP may be dephosphorylated by as yet poorly characterized dephosphorylases (Abe et al., 1982; Jamieson et al., 1987). Correlations between remission durations and intracellular retention of araCTP (Rustum, 1982; Rustum and Preisler, 1979) suggest that cells that actively dephosphorylate ara-CTP may be more resistant to therapy than cells that retain the lethal metabolite.

E. Increased dCTP Pool Sizes Because dCTP potently antagonizes the formation of ara-CTP (Liliemark and Plunkett, 1986) and also competes for incorporation into DNA, expan-

212

Steven Grant

sion of intracellular dCTP pools also antagonizes ara-C-mediated lethal effects. This can occur through alterations in CTP synthetase (de Saint Vincent and Buttin, 1981) or dCMP deaminase activity (Weinberg et al., 1981; Bianchi et al., 1987; de Saint Vincent et al., 1980). Interestingly, Chiba and colleagues (1990) reported that blasts from two patients with ara-C-resistant leukemia displayed increased dCTP pools and decreased activity of dCMP deaminase compared with pretreatment levels, suggesting that this mechanism of resistance may play a role in treatment failure, a t least in some patients.

F. Alterations in CTP Synthetase A number of Chinese hamster ovary (CHO) cell lines selected for resistance to ara-C displayed base-pair substitutions rendering CTP synrhetase unresponsive to allosteric regulation, leading in turn to expanded intracellular dCTP pools (Whelan et al., 1993).This mechanism of ara-C resistance, however, could not be identified in primary human leukemic myeloblasts (Whelan et al., 1994).

G. Cytokinetic Resistance Ara-C is primarily toxic to cells actively engaged in DNA synthesis, that is, those in the S phase of the cell cycle (Young and Fisher, 1968; Karon and Chirakawa, 1970). This presumably reflects the fact that ara-C DNA incorporation occurs during S phase and that the activity of deoxycytidine kinase tends to be higher in replicating cells. Ara-C dose-response curves typically show a plateau phenomenon, in that the percentage of cell kill with higher drug concentrations is limited by the maximal fraction of cells in S phase during the interval of exposure (Bhuyan et al., 1972; Grdina et al., 1980; Pallavicini et al., 1982). Consequently, tumor cells exhibiting a low proliferative rate, and more specifically, a low S-phase fraction, would be expected to be resistant to ara-C-mediated toxicity. In acute leukemia, however, patients whose blasts had a long cell cycle time experienced longer remissions than their counterparts whose blasts divided more rapidly (Raza et al., 1990), perhaps reflecting the relative biologic indolence of the former cells. Strategies designed to overcome the theoretical problem of cytokinetic resistance to ara-C have generally involved attempts to recruit cells into cycle, by either drug scheduling (Burke et al., 1982; Colly and van Bekkum, 1982; Pallavicini et al., 1982; Ubezio et al., 1994), pharmacologic means (Greenberg and Waxman, 1976), or the use of hematopoietic growth factors (discussed later). The ability of any of these strategies to improve the therapeutic efficacy of ara-C remains to be established.

Ara-C: Cellular and Molecular Pharmacology

213

H. Defects in the Cell Death Pathway Defects in distal cell death (i.e., apoptotic pathway) may confer resistance to an unusually broad range of chemotherapeutic agents. For example, a mutant U937 cell line resistant to apoptosis displayed significant resistance to

both topoisomerase inhibitors (e.g., VP-16) and ara-C (Kataoka et al., 1994).The lack of sensitivity of mutant cells could not be attributed to drugspecific factors because equivalent biochemical lesions occurred in the mutant and parental lines. A more defined lesion has been identified in the chronic myelogenous leukemia (CML)-derived human erythroleukemia cell line K562, which exhibits resistance to multiple chemotherapeutic agents (McCahon et al., 1994). In these cells, expression of the ber-abl p2'0 tyrosine kinase appears to be responsible for decreased susceptibility to drug-induced apoptosis. In the case of ara-C, alterations in expression of Bcl-2 and related proteins has been the most commonly encountered perturbation in the apoptotic pathway conferring drug resistance (discussed later). For example, human leukemic cells genetically modified to overexpress Bcl-2 regularly display resistance to ara-C-induced apoptosis (Miyashita and Reed, 1993; Manome et al., 1993). This resistance is partial, indicating the presence of Bcl-2-independent pathways of cell death that are capable of being triggered by ara-C. The potential significance of Bcl-2-mediated resistance to ara-C-induced bell death is underscored by the observations that patients whose blasts exhibit high levels of Bcl-2 respond poorly to ara-C-containing regimens (Campos et al., 1994) and that antisense oligonucleotides directed against Bcl-2 increase the susceptibility of primary human leukemic blast cells to ara-C-induced apoptosis in vitro (Keith et al., 1995).

VI. SIGNALING PATHWAYS AND ONCOGENE I NTERACTION S Considerable information exists concerning the genes involved in the response of neoplastic cells to cytotoxic agents, particularly ara-C. In a sense, drugs like ara-C can be considered environmental stresses and, as such, capable of triggering a complex cascade of genes involved in the cellular response to genomic injury. Key questions to be resolved are whether these events represent an integral component of the cell death process, and whether modulation of cellular stress responses have a significant impact o n cell survival. If these are the case, it is likely that efforts to perturb gene expression could be employed to increase the susceptibility of neoplastic cells to ara-C and possibly other antineoplastic agents.

214

Steven Grant

The protooncogene c-jun encodes a protein (c-Jun) that heterodimerizes with the product o f the c-fos oncogene to form the AP-1 transcription complex, which is involved in the transcriptional activation of diverse genes involved in cell proliferation, differentiation, and transformation (Angel and Karin, 1991). In cases of growth factor deprivation-associated apoptosis, direct evidence implicates c-jun in the cell death response (Collata et al., 1992; Hamm et al., 1995). Exposure of leukemic cells to concentrations of ara-C sufficient to induce apoptosis has been shown to be associated temporally with induction of c-jun and c-fos (Kharbanda etal., 1990,1991; Gunji etal., 1991) as well as another member of the c-jun family, jun-D (Kharbanda et al., 1993). Moreover, induction of c-jun by ara-C has been shown in some cases to be dependent on protein kinase C in that it can be blocked by PKC inhibitors (Kharbanda et al., 1991). Induction of a PKC-like activity in leukemic cells by ara-C is also associated with an increase in mitogen-activated protein (MAP) kinase activity (Kharbanda et al., 1994). Exposure of leukemic cells to ara-C leads to AP-3 activation, further implicating c-jun in the cellular response to ara-C-associated injury (Brach et al., 1992). Altered phosphorylation of AP-1 and SP-1 transcription factors has also been reported in human myeloid leukemia cells resistant to ara-C and 1,25-dihydroxyvitamin D, (Kolla and Studzin-ski, 1994). Despite presumptive evidence linking c-jun to the ara-C-related DNA damage response, however, it is not clear whether increased c-jun expression represents a response to araC o r instead constitutes an integral component of the cell death process. Thc ability of PKC inhibitors to promote ara-C-induced apoptosis despite antagonizing c-jun upregulation has been taken as indirect evidence supporting the former possibility (Bullock et al., 1995).More direct evidence for this view emerged from a study in which a c-jun dominant negative mutant human myeloid leukemia cell line displayed unperturbed sensitivity to ara-Cinduced apoptosis (Grant et al., 1996). Interestingly, these cells were relatively resistant to the anti proliferative and differentiation-inducing effects of low concentrations of ara-C, suggesting a differential role for c-jun in the regulation of leukemic cell maturation and apoptosis.

B. NF-KB NF-KB, like AP-1, is a transcription factor involved in the induction of cytokine, immunoniodulatory, metalloproteinase, and other genes. Treatment of human myeloid leukemia cells with ara-C has been shown to increase NFKB MRNA levels and to induce DNA binding (Brach et al., 1992b). The DNA polymerase inhibitor aphidicolin increases stimulation of NF-KB DNA-binding activity by ara-C, an event associated with increased suscep-

Ara-C: Cellular and Molecular Pharmacology

215

tibility to apoptosis (Kuwakado et al., 1995).It is unclear how the latter finding is related to reports indicating that NF-KB acts as a potent inhibitor of apoptosis induced by tumor necrosis factor (TNF) and some other chemotherapeutic drugs (Wang et al., 1996; Antwerp et al., 1996; Beg and Baltimore, 1996; Liu et al., 1996).

C. pRb The retinoblastoma protein (pRb) is intimately involved in cell cycle regulation and plays a key role in progression of cells across the G ,/S boundary (Weinberg, 1995). Phosphorylation of pRb leads to release of the E2F transcription factor, resulting in the transcription of numerous genes involved in cell cycle progression (Ikeda et al., 1996). In contrast, hypophosphorylated pRb binds to and inactivates E2F, contributing to G I arrest. Cytotoxic agents, such as ara-C, have been shown to induce alterations in the expression of pRb, a phenomenon unrelated to perturbations in other cell cycle-related genes (i.e., c-myc) (Yen and Varvayanis, 1992). It has also been shown that DNA-damaging drugs, including ara-C, induce hypophosphorylation of pRB in p53-negative cells, perhaps contributing to their ability to trigger G , arrest and apoptosis (Dou et al., 1995a; Dou and Lui, 1995). The ability of such agents to induce pRb hypophosphorylation has been attributed to stimulation of pRb phosphatase activity (Dou et al., 1995). Dephosphorylation of pRb by ara-C does not occur in deoxycytidine kinase-deficient cells resistant to ara-C-associated opoptosis (Dou and Lui, 1995). Induction of apoptosis by ara-C in leukemic cells leads, in addition to hypophosphorylation of pRb, to proteolytic cleavage of the protein by an Interleukin-P converting enzyme (ICE)-likeprotease (An and DOU,1996).It is unclear, however, whether this phenomenon serves as a critical regulator of the cell death process or merely represents a by-product of activation of the proteolytic cascade.

The protooncogene c-myc is a transcription factor involved in cell cycle progression (Zimmerman and Alt, 1990). It has been directly implicated in progression of cells through the G,/S transition phase (Dang, 1991) as well as in the induction of apoptosis (Evan etal., 1992). Relatively little is known about the relations that exists between c-myc and ara-C-mediated lethality. In contrast, it has been shown that downregulation of c-myc in human leukemia cells represents a critical event in differentiation induction by low concentrations of ara-C (Baker et al., 1994). Growth inhibition by agents that induce reversible leukemic cell differentiation (e.g., hydroxyurea) are not associated with decreased expression of c-myc, whereas induction of terminal dif-

216

Steven Grant

ferentiation by ara-C is accompanied by marked downregulation of c-myc (Gomez-Casares et al., 1993). Thus, low concentrations of ara-C may mimic the actions of differentiation inducers, such as phorbol esters, with respect to effects on c-myc (Filmus and Buick, 1985). Potentiation of the differentiation effects of ara-C by leukemia inhibitory factor (LIF) resulted in further reductions in c-myc expression (Brach et al., 1993). In contrast, pretreatment of leukemic cells with vitamin D, reduced ara-C-mediated c-myc downregulation (Moore et al., 1992). Relatively few studies have examined the effects of high ara-C concentrations on c-myc expression in leukemic cells, although downregulation was observed in one study (Tang et al., 1994).

E. bcl-2 The protooncogene bcl-2 encodes a member of a family of homologous proteins that either potentiate or antagonize programmed cell death. Of these, Bcl-2, Bcl-xL, A l , Mcl-1, and Bag-1 exert antiapoptotic actions, whereas Bax, Rel-xs, Bak, and Bad promote cell death (Yang and Korsmeyer, 1996). As noted previously, leukemic patients whose blasts express high levels of Bcl-2 respond poorly to chemotherapy (Campos et al., 1994), and reduction in Bcl-2 expression by antisense oligonucleotides increases the sensitivity of leukemic blasts to ara-C in vitro (Keith et d., 1995). In in uitro systems, increased expression of Bcl-2 protects leukemic cells from ara-Cmediated toxicity (Miyashita and Reed, 1993; Hu et al., 1996; Ibrado et al., 1996; Manome et al., 1993). Similar protection has been observed in cells exhibiting overexpression of Bcl-xL (Ibrado et al., 1996; Bullock et al., 1992).The intracellular metabolism of ara-C and its initial effects on DNA fragmentation are identical in cells expressing low and high levels of the Bcl2 protein, suggesting that Bcl-2 primarily regulates the more distal steps in the ara-C-induced cell death pathway (Bullock et al., 1996). Although the mechanism by which Bcl-2 prevents ara-C-induced cell death in human leukemic cells remains to be determined, Bcl-2 and Bcl-xL have been shown to block staurosporine-associated apoptosis by preventing activation of C. Elegans death (CED)-like proteases, such as Yama/CPP32/apopain (Chinnaiyan et al., 1996). Interestingly, several studies indicate that the same mechanism may account for the protective effects of Bcl-2 and Bcl-xL with respect to ara-C-induced apoptosis in human myeloid leukemia cells (Ibrad o et a/., 1996; Datta et al., 1996).

F. p21 (WAFIKIPI) The protein p 2 1 (WAFZICIPI) is an inhibitor of cyclin-dependent kinases and has been identified as a downstream target of p53 responsible for G , ar-

Ara-C: Cellular a n d Molecular Pharmacology

21 7

rest in cells sustaining DNA damage (El-Diery et al., 1994). It has also been shown to be induced in p53-null cells undergoing maturation in response to differentiating agents such as phorbol esters (Jiang et al., 1994). Relatively little is known about the possible role of p21 in ara-C-induced apoptosis. The failure of ara-C to induce p21 argues against its participation in ara-C mediated cell death, albeit indirectly (Steinman et al., 1994). Studies indicate that loss of p21 increases the susceptibility of neoplastic cells to DNA damaging agents, such as doxorubicin, presumably by uncoupling G , arrest and mitosis and/or by limiting the potential for DNA repair (Waldman et al., 1996, 1995). It is unknown whether similar considerations apply to ara-Cinduced cytoxicity, although preliminary evidence suggests that this may not be the case (Freemerman et al., 1997).

G. E2F The transcription factor E2F plays a critical role in the G,/S transition (Shan et al., 1996). the underphosphorylated form of the retinoblastoma protein binds to and inactivates E2F, resulting in G, arrest (Chellapan, 1994).Premature or inappropriate entry of cells into S phase, that is, through dysregulated expression of E2F, leads to induction of apoptosis (Shan and Lee, 1994; Shan etal., 1996). It has been shown that a fibroblast cell line expressing a mutant form of E2F exhibits a lengthening of S phase and increased sensitivity to the S-phase-specific agent camptothecin (Logan et al., 1995). It is likely that such cells would also display increased susceptibility to ara-C, although this possibility remains to be explored. If so, this finding would represent an example of how perturbations in cell cycle regulatory proteins might regulate responses to cell cycle-specific drugs such as ara-C.

H. p34cdc2 The cyclin-dependent kinase ~ 3 4 ' ~is' involved ~ in the progression of cells through the G,/M phase of the cell cycle (Nigg, 1993). Under conditions in which inappropriate activation of ~ 3 4 ' ~occurs, '~ such as following exposure of lymphoma cells to fragmentin-2, the cells undergo an apoptotic program mimicking a mitotic catastrophe (Shi et d., 1994). Unscheduled activation of p3FdC2has also been reported in HL-60 cells exposed to various DNA damaging agents (Shimizu et al., 1995).Treatment of HL-60 cells with ara-C has been shown to be associated with tyrosine phosphorylation of the ~ 3 4 ' ~kinase, '~ resulting in a reduction in enzyme activity (Yuan et al., 1995). This effect could conceivably influence the susceptibility of cells to apoptosis either by altering the cell cycle traverse or by directly modulating the apoptotic threshold.

218

Steven Grant

I. Stress- and Mitogen-Activated Protein Kinases The response of cells to genotoxic stress induced by DNA damaging agents, such as ara-C, involves induction and phosphorylation of the immediate early-response genes c-fos and c-jun (Kharbanda et al., 1990, 1991, 1994; Brach et al., 1992a). Although PKC may be involved in this phenomenon (Kharbanda et a/., 1991), other evidence implicates the p44/42 mitogen-activated protein kinase (MAPK) in these events (Kharbanda et a/., 1994). Activation of MAPK leads to phosphorylation of serine residues in the amino-terminal transactivation domain of c-Jun, leading to alterations in its transactivation potential (Kharbanda et a/., 1994).The relative contributions of PKC and MAPK in ara-C-associated phosphorylation and activation of c-Jun remains to be determined, as does the functional significance of these events. Attention has focused on the potential role of the stress-activated protein kinases (SAPK),otherwise known as Jun kinases (JNK), in the response of cells to genotoxic stress. The SAPKs are triggered by various stimuli, including ultraviolet light and TNF. These kinases, which phosphorylate the amino-terminus of c-Jun, are distinct from the MAPK (Kyriakis et a/., 1994). Studies indicate that exposure of leukemic cells to ara-C leads to activation of SAPKs, amino-terminal phosphorylation of c-Jun, and binding of SAPK to adapter protein Grb2 (Saleem et al., 1995). Although the functional significance of these findings remains to be determined, it may be relevant that interruption of the SAPK/JNK cascade has been shown to antagonize TNFand cerarnide-induced apoptosis in human leukemia cells (Verheij et al., 1996). It is conceivable that a similar phenomenon may occur in the case of ara-C, but it is equally possible that activation of SAPK/JNK may simply represent a response to ara-C-mediated DNA damage. Alternatively, activation of stress-related kinases may shift the balance away from survival pathways, such as that associated with extracellular receptor kinase (ERK), leading to cell death (Cuvillier eta/., 1996).

1. PKC-6 In addition to proteolytic activation of ICE-like proteases, induction of apoptosis by ara-C in human leukemia cells is associated with cleavage and activation of the isoform PKC-6 (Emoto et al., 1996), an event possibly linked to induction of c-Jun/AP-1 activity (Kharbanda et al., 1994). As in the case of pRb, however, it is unclear whether cleavage of PKC-6 represents a cause or consequence of ara-C-induced cell death.

Ara-C: Cellular and Molecular Pharmacology

219

K. c-a61 The reciprocal translocation of the c-a61 protooncogene from chromosome 22 to the breakpoint cluster region (BCR) of chromosome 9 represents the hallmark of CML. The chimeric c-a6NBCR encodes a 210-kd fusion protein that has been implicated in the resistance of the CML cell line K562 to apoptosis induced by diverse chemotherapeutic agents (McGahon et al., 1994).The BCWu6l translocation has been reported to lead to prolongation of GJM phase arrest in human hematopoietic cells exposed to DNA-damaging agents, presumably increasing the opportunity for DNA repair, and to confer resistance of CML progenitor cells to ara-C (Bedi et al., 1995). The protooncogene c-a61 has also been implicated in the stress response of leukemic cells to ara-C induced DNA damage. For example, leukemic cells deficient in c-a61 are resistant to ara-C-mediated activation of SAPK (Kharbanda et al., 1995). The impact of disruption of this pathway on the lethal actions of ara-C, however, remains to be determined. Transfection of K562 cells with the proapoptotic protein Bcl-x, has been shown to increase their susceptibility to ara-C-induced apoptosis and differentiation, suggesting that perturbations in members of the Bcl-2 family can circumvent c-abl-associated resistance to cell death (Ray et al., 1996).

L. Histone HI Treatment of human myeloid leukemia cells (HL-60) with ara-C has been found to reduce expression of histone H 1 mRNA by both transcriptional and posttranscriptional mechanisms (Datta et al., 1992). The functional significance of this finding is uncertain.

M. ras Activation of the ras oncogene represents the most common gain-of-function mutation in human cancers (Bos, 1989). Transfection of rodent fibroblasts and human mammary HBL 100 cells with c-Ha-rus conferred resistance to ara-C, an event attributed to decreased activity of dCK (Riva et al., 1995). O n the other hand, screening of human tumor cell lines from the National Cancer Institute’s In Vitro Antineoplastic Drug Screen revealed a strong correlation between mutations in N-ras or Ki-rus and enhanced sensitivity to ara-C (as well as to certain topoisomerase I1 inhibitors) (Koo et d., 1996). The latter finding may account for the observation that leukemic patients whose blasts harbor ras mutations appear to respond more favorably

220

Steven Grant

to ara-C-containing chemotherapeutic regimens (Neubauer et al., 1994; Liu,

1990).

N. EGRl The early growth response-1 ( E G R l ) gene is induced by proliferative, differentiation-inducing, and genotoxic stimuli. Exposure of human leukemic cells (HL-525) to cytotoxic concentrations or ara-C resulted in induction o f EGRl through transcriptional activation (Kharbanda et al., 1993a). As in the case of c-iun, it is unclear whether induction of EGRl by ara-C represents a component of the cell death response or merely a response to DNA damage.

0. CycE-HIK The cyclin E-H1 kinase (CycE-H1K) has been implicated in G , phase arrest in cells sustaining DNA damage (Sherr and Roberts, 1994). Exposure of sensitive but not dCK-deficient human leukemia cells to ara-C resulted in a dramatic increase in CycE-H1 kinase activity in association with the induction of apoptosis (Dou et al., 199513). It is possible that activation of CycEH1K represents a late apoptosis checkpoint in leukemic cells exhibiting araC-induced DNA damage.

VII. MODULATION OF ARA-C ASSOCIATED CYTOTOXICITY In view of its clinical activity, particularly in hematologic malignancies, many efforts have been directed a t increasing the efficacy of ara-C against neoplastic cells and/or overcoming ara-C resistance. In this regard, three major approaches have been taken. The earliest of these, termed biochemical modulation, involves combination of ara-C with agents that either potentiate its metabolism, interfere with its catabolism, or selectively antagonize its lethal actions toward normal cells. The second approach involves combining ara-C with hematopoietic growth factors (HGFs). This strategy is based on the concept that HGFs may either recruit neoplastic cells into the susceptible S phase of the cell cycle, augment ara-C metabolism, or increase araC-mediated apoptosis. The most recent approach, which on the verge of clinical evaluation, entails combination of ara-C with agents that act through signal transduction pathways. This strategy is based on the premise that interruption of such pathways may increase the susceptibility of cells to druginduced cytotoxicity by facilitating activation of the apoptotic cascade.

Ara-C: Cellular and Molecular Pharmacology

22 I

A. Biochemical Modulation The strategy of biochemical modulation of ara-C action has been reviewed (Grant, 1990). Most of these efforts have been directed at combining ara-C with inhibitors of de novo pyrimidine biosynthesis, with the goal of reducing intracellular levels of dCTP. Reductions in dCTP pools have two major effects on ara-C metabolism: allosteric activation of dCK leading to enhanced ara-CTP formation, and reduced competition for incorporation of ara-C residues into DNA. Synergistic interactions with ara-C have been observed in the case of thymidine (Cohen and Ullman, 1985; Grant et al., 1980), hydroxyh r e a (Streifel and Howell, 1981; Colly and van Bekkum, 1982), 3-deazauridine (Mills-Yamamoto et al., 1978), PALA (Grant et al., 1983b), deoxyguanosine (Ross et al., 1984) and imidazole pyrazole (IMPY) (Grant et al., 1983a), among others. Potentiation of ara-C action by such agents has not been a universal finding, however. For example, reductions in uridine triphosphate (UTP) pools (i.e., by PALA) may have a net negative regulatory role on ara-C phosphorylation by decreasing the availability of phosphate donors (White and Hiner, 1987). Moreover, the ability of agents such as PALA to potentiate ara-C metabolism in primary human blasts appears limited (Noordhuis et al., 1996). Another key question that remains is whether modulation of ara-C metabolism in neoplastic cells by pyrimidine antagonists, assuming it occurs in vivo, preferentially spares normal elements. Initial clinical studies addressing this issue have not provided definitive answers. Interest has focused on interactions between ara-C and the purine analog fludarabine, based on preclinical evidence of synergism between these agents (Gandhi and Plunkett, 1988). Pretreatment of leukemic cells with fludarabine increases the generation of ara-CTP, a consequence of inhibition of ribonucleotide reductase by fludarabine triphosphate (Gandi et al., 1993). Potentiation of ara-C-lethality by fludarabine may also stem from the effects of the latter agent on DNA chain termination (Huang et al., 1990) o r inhibition of nucleotide excision repair (Sandoval et al., 1996). Clinical studies examining in vivo interactions between these agents are underway, and initial results appear interesting (Estey et al., 1993). An alternative approach has been to combine ara-C with agents that interfere with drug catabolism. Chief among these have been deaminase inhibitors, such as tetrahydrouridine and deoxytetrahydrouridine (Kreis et al., 1988, 1992; Grant et al., 1991). As noted previously, THU has been shown to potentiate ara-C pharmacokinetics in vivo (Kreis et al., 1992), but whether a gain in therapeutic index results remains to be determined. A third strategy is based on the premise that neoplastic cells are more dependent on de novo than on salvage pyrmidine biosynthetic pathways than are their novel counterparts. The naturally occurring nucleoside 2’-deoxycytidine protects cells and intact animals (Evans and Mengel, 1964) from

222

Steven Grant

ara-C-induced cytotoxicity. It may also increase the therapeutic index of high-dose ara-C against certain murine tumors (i.e., L1210) (Buchman et al., 1979). In in vitro studies, 2‘-deoxycytidine preferentially protected normal compared with leukemic human myeloid progenitors from ara-C-induced toxicity (Bhalla et al., 1987). A clinical study demonstrated that administration of 2‘-deoxycytidine protected humans from otherwise lethal doses of high-dose ara-C (Grant et al., 1993). As in all other biochemical approaches to ara-C modulation, it is unclear whether coadministration of ara-C and 2’-deoxycytidine will lead to a net gain in the therapeutic index of ara-C. Finally, based upon in vitro and in vivo evidence of antileukemic synergism when ara-C is administered prior to the protein synthesis inhibitor, L-asparaginase (Schwartz et al., 1982), Capizzi and coworkers have developed clinical protocols combining these agents in patients with refractory leukemia (Capizzi and Powell, 1987). The basis for the observed scheduledependent synergism between these agents is unclear, but may result from interference with the synthesis of proteins critical for cellular recovery from ara-C-mediated DNA damage.

B. Hematopoietic Growth Factors Hematopoietic growth factors are glycoproteins that promote the survival, proliferation, and differentiation of hematopoietic progenitor cells in vitro and presumably in vivo. The concept of combining HGFs with ara-C, particularly in leukemia, is based on several rational considerations. First, leukemic cells, like their normal counterparts, are also responsive to the actions of HGFs. Thus, administration of HGFs before ara-C might serve a “priming” function, recruiting leukemic cells into the susceptible S phase of the cell cycle. In fact, a number of in vitro studies have shown that growth factors such as GM-CSF, interleukin-3, and G-CSF increase the sensitivity of leukemic cells to ara-C-induced cytotoxicity (Lista et al., 1988; Bhalla et al., 1991; Cannistra et al., 1989; Tafuri et al., 1994) and that in some cases this effect may be related to cytokinetic perturbations. Second, administration of HGFs to leukemic cells before ara-C leads to alterations in deoxyribonucleotide (dCTP)and ara-C metabolism, events that might selectively enhance the antileukemic efficacy at ara-C. For example, exposure of human leukemic cells to GM-CSF and ara-C resulted in higher ara-CTP-to-dCPT ratios than those observed in their normal counterparts (Bhalla et al., 1988). HGFs may also increase ara-C incorporation into leukemic cell DNA (Hiddeman et al., 1992). Finally, exposure of leukemic cells to HGFs in conjunction with an inhibitor of DNA synthesis, such as ara-C, may potentiate apoptosis through a mechanism analogous to that invoked in the case of cell death occurring in cells inappropriately expressing c-myc (Askew et al.,

Ara-C: Cellular and Molecular Pharmacology

223

1991).Thus, simultaneous administration of an antiproliferative signal (araC) in conjunction with a proliferative signal (HGF) may lead to apoptosis by the “conflict” model (Evan and Littlewood, 1994). In support of this concept, a study demonstrated that the GM-CSF/IL-3 fusion product (PIXY 321) enhanced the susceptibility of human leukemia cells to ara-C-related apoptosis (Bhalla et al., 1992). It is equally possible that potentiation of araC induced apoptosis in leukemia by HGFs stems from a combination of one or more of these actions. Several theoretical and practical considerations, however, could limit the success of this general strategy. First, leukemic blasts respond in an extremely heterogeneous manner to HGFs (Van der Lely et al., 1994), and it easy to envision that in growth factor-unresponsive clones, either preexisting or selected, ara-C sensitivity would not be increased. Second, in view of the ability of HGFs to stimulate the proliferation of leukemic blasts (Lemoli et al., 1991), it is at least theoretically possible that HGFs could exacerbate the underlying leukemia. In fact, in one in vitro study, the ability of HCFs to increase drug-induced sensitivity was outweighed by a stimulatory effect on residual unaffected cells (Sundman-Engberg et al., 1996). With one possible exception (Estey et al., 1992), however, there is n o evidence that administration of HGFs in conjunction with chemotherapy adversely affects clinical outcome. Another potential problem could arise if HGFs were unable to increase the S-phase fraction of leukemic blasts in vivo, thereby limiting potentiation of ara-C sensitivity (Lacombe et d., 1996). Moreover, in vivo administration of interleukin-3 has been shown to decrease the ex vivo sensitivity of blasts from a subset of leukemic patients to ara-C (Gore et al., 1995). In addition, the effects of HGFs on the cell cycle traverse of primary human leukemic blasts may be modest and weakly correlated with increased sensitivity to ara-C (Smith et al., 1995).Finally, HGFs are known to inhibit apoptosis in leukemic cells after certain stimuli, such as growth factor deprivation (Williams et al., 1990); and in at least some cases, HGFs have been reported to decrease, rather than increase, drug-induced apoptosis (Lotem and Sachs, 1992).As a result, the net effect of HGFs on ara-C sensitivity may ultimately depend on which of the many biochemical and cytokinetic consequences of growth factor administration predominate in the cells of an individual patient. It is also conceivable that HGFs may exert heterogeneous effects on the ara-C sensitivity of cells within the same patient.

C. Modulation of Signal TransductionApoptotic Pathways As noted previously, resistance to ara-C (and other neoplastic agents) can arise from distal defects in the cell death pathway. A number of efforts to overcome this form of resistance have focused on PKC, a Ca2+-and lipid-de-

224

Steven Grant

pendent serine-threonine kinase intimately involved in cell proliferation, differentiation, and stimulus-response coupling (Nishizuka, 1984).The bulk of evidence indicates that activation of PKC has the net effect of inhibiting apoptosis. For example, tumor-promoting phorboids, such as phorbol myristate acetate (PMA), antagonize apoptosis induced by growth factor deprivation (Lotem et al., 1991) as well as certain antineoplastic drugs (e.g., VP-16) (Solary et al., 1993). Conversely, PKC inhibitors are potent inducers of apoptosis in human leukemia cells (Jarvis et al. 1994b; Bertrand et al., 1994). The mechanism by which PKC opposes apoptosis is unknown but may stem from modulatory effects on proapoptotic and antiapoptotic signaling pathways (Cuvillier et al., 1996), induction of changes in intracellular pH (Rajotte et al., 1992), or phosphorylation of Bcl-2 o r related proteins (Chen and Faller, 1995). Although earlier studies indicated that PKC inhibitors (i.e., H7) could antagonize c-jun upregulation associated with ara-C-induced apoptosis (Kharbanda etal., 1991), it was not determined a t that time what effect such agents had on apoptosis. Several subsequent studies revealed that administration of agents such as staurosproine increased the susceptibility of human leukemic cells to ara-C-mediated cell death (Grant et al., 1994; Bullock et al., 1992). Thus, instead of directly modulating the actions of ara-C, PKC inhibitors appear to potentiate ara-C-related cell death by interrupting signaling pathways that contribute to cell survival. An analogous strategy has been employed involving the macrocyclic lactone PKC activator, bryostatin-1. Byrostatin-1 exhibits a spectrum of activity distinct from that of tumor-promoting phorboids (Kraft et al., 1986) and, on chronic administration, is an extremely potent downregulator of PKC activity (Isakov et al., 1993). Preincubation of human leukemia cells with bryostatin-1, under conditions in which differentiation was not induced, led to a facilitation of ara-C-induced apoptosis and synergistic antileukemic actions. (Grant et al., 1992). This effect was temporally related to reduction in PKC activity (Jarvis et al., 1994a). It is therefore possible that interruption of the PKC pathway, either through downregulation (i.e., by bryostatin-1) o r by direct inhibition (i.e., by PKC inhibitors) augments ara-C-mediated cytotoxicity through a common mechanism (see Fig. 3 ) . Molecular approaches have also been used to modulate the apoptotic response of leukemic cells to ara-C. As noted previously, antisense oligonucleotides directed against hcl-2 were shown to increase the sensitivity of leukemic blasts to ara-C-induced apoptosis (Keith et al., 1995). Moreover, resistance of chronic myeloid leukemia cells to ara-C-mediated apoptosis, a possible consequence of p2 1Ohcr’abl expression, can be circumvented by transfection of cells with Bcl-xS (Ray et al., 1996). Thus, perturbations in proteins regulating distal events in the apoptotic pathway can be employed to increase the susceptibility of leukemic cells to ara-C and presumably other cytotoxic agents.

Ara-C: Cellular and Molecular Pharmacology

225

Numerous other cellular signaling pathways might be suitable targets as modulators of ara-C action. For example, protein tyrosine kinases are intimately involved in cell survival and proliferation (Levitzki and Gazit, 1995); moreover, protein tyrosine kinase inhibitors are potent inducers of apoptosis (Markovits etal., 1994). In this regard, it has been shown that quercetin and other protein kinase inhibitors potentiate ara-C-induced lethality (Teofili et al., 1992), presumably by enhancing the cell death response. In view of the large number of pharmacologic inhibitors of signal transduction pathways already available or under development, this strategy should provide virtually unlimited opportunities for modulation of ara-C-related cytotoxicity.

VI 11. CONCLUSIONS During the past several decades, much has been learned about the mechanisms responsible for the lethal actions of ara-C as well as about the factors governing the response of neoplastic cells to this antimetabolite. As in the case of most antineoplastic agents, however, identification of the basis for either drug resistance or antitumor selectivity remains elusive. It is now clear that the response of cells to ara-C may depend on relatively specific processes (e.g., nucleoside transport, phosphorylation, deamination) as well as o n less specific downstream factors that determine whether the cell death pathway is engaged after ara-C-mediated DNA damage (e.g., Bcl-YBax levels, activity of PKC). In addition to direct cytotoxic effects, ara-C induces perturbations in the expression of various oncogenes as well as lipid second messengers (e.g., diradylglycerol, ceramide), each of which may play a role in regulation of the cell death response. This, in turn, adds another layer of complexity to models attempting to define the basis for the lethal actions of ara-C. Nevertheless, as the molecular and signal transduction-related events that regulate cell death become better understood, it is likely that such insights will shed further light on the mechanisms underlying the effectiveness of ara-C. They may also pave the way for the development of entirely novel therapeutic strategies designed to overcome ara-C resistance and to improve antileukemic activity.

REFERENCES Abe, I., Saito, S., Hori, K., Suzuki, M., and Sato, H. (1982). Cancer Res. 42, 2846-28.51. Adler, A. I., Grossman, I., a n d Fasman, G. D. (1968).Biochemistry 7, 3826-3834. An, B., and Dou, Q. P. (1996). Cancer Res. 56, 438-442.

226

Steven Grant

Angel, P., and Karin, M . (1991).Biochim. Biophys. Acta. 1072, 129-157. Antwerp, D. J., Martin, S. J., Kafri, T., Green, D. R., and Verma, I. M. (1996). Science 274, 787-789. Askew, D. S., Ashmun, R. A., Simmons, B. C., and Cleveland, J. L. (1991). Oncogene 6, 1915-1 922. Baker, S. J., Pawlita, M., Leutz, A,, and Hoelzer, D. (1994). Leukemia 8, 1309-1317. Bedi, A., Barber, J. P., Bedi, G. C., El-Diery, W. S., Sidransky, D., Vala, M Hilton, J., and Jones, R. J. (1995).Blood 86, 1148-1 158. Beg, A. A., and Baltimore, D. (1996).Science 274,782-784. Bertrand, R., Solary, E., O’Connor, P., Kohn, K. W., and Pommier, Y. ( 1 994). Exp. Cell. Res. 211,314-321. Bhalla, K., Baker, M., Cole, I., MacLaughlin, W., Arlin, Z., and Grant, S. (1987). Blood 70, 568-571. Bhalla, K., Birkhofer, M., Arlin, Z., Grant, S., Lutzky, J., and Graham, G. (1988).Leukemia 2, 8 10-8 13. Bhalla, K., Holladay, C., Arin, Z., Grant, S., Ibrado, A., and Jasiok, M. (1991). Blood 78, 2674-2679. Bhalla, K. R., Nayak, R., and Grant, S. (1984).Cancer Res. 44, 5029-5037. Bhalla, K., Tang, C., Ibrado, A. M., Grant, S., Tourkina, E., Holladay, C., Hughes, M., Mahoney, M. E., and Huang, Y. (1992).Blood 80,2883-2890. Bhuyan, B., Scheidt, L. G., and Fraser, T. J. ( 1 972). Cancer Res. 32,398-407. Bianchi, V., Pontis, E., and Reichard, P. (1987).Mol. Cell Biol. 7, 421 8 4 2 2 4 . Bos, J. I,. ( 1 989). Cancer Res. 49,46824689. Brach, M. A., Herrmann, F., and Kufe, D. W. (1992a).Blood 79, 728-734. Brach, M. A., Kharbanda, S. M., Herrman, F., and Kufe, D. W. (l992h). Mol. fharmacol. 41, 60-63. Brach, M. A., Mertelsmann, R. H., and Hermann, F. (1993).Cancer Invest. 11, 198-21 1 . Brockman, R. W., Cheng, Y-C., Schahel, F. M., and Montgomery,J. A. (1980).Cancer Res. 30, 36 10-36 15. Buchman, V. M., Belyanchikova, N. K., and Mkheidze, D. M. ( 1 979). Cancer Chemother Pharmacol. 3,229-234. Bullock, G., Ray, S., Reed, J. C., Krajewski, S., Ihrado, A. M., Huang, Y., and Bhalla, K. ( 1 996). Leukemia 10, 1731-1741. Bullock, G., Kay, S., Reed, J.. Miyashita, T., Ihrado, A. M., Huang, Y., and Bhalla, K. ( 1 992). C h . Cancer Res. 52, 5-10. Bullock, G., Ray, S., Reed, J., Stanislaw, K., Ihrado, A. M., Huang, Y., and Bhalla, K. (1995). Blood 86,2047-2054. Burke, P. J., Vaughan, W. P., Karp, S. E., and Saylor, P. L. (1982). Med. f e d . Oncol. Suppl. 1 , 201-208. Caminier, G. W., and Smith, C. G. (1965).Biochem. PharinaroL 14, 1405-1416. Campos, L., Rouault, J. P., Sabido, O., Oriol, P.,Rouhi, N., Vasselon, C., Archimbaud, E., Magaud, J. p., and Guyotat, D. (1994).Blood 84, 595-600. Cannisrra, S. A., Groshck, P., and Griffin, J. D. (1989). Leukemia 3, 328-334. Capizzi, R. L., Powell, B. L. (1987).Seminars in Oncology (Suppl 1)14, 40-50. Caradonna, S. J., and Cheng, Y-C. (1981).Mol. Pharmacol. 18,513-520. Carson, D. A., Wasson, D. B., Kaye,]., Ullman, B., Martin, D. W., Robins, R. K., and Montgomery, J. A. (1980).Proc. Natl. Acad. Sci. U.S.A. 77,6865-6869. Chabner, B. A., Johns, D. G., Coleman, N., Drake, J. C., and Evans, W. ( 1974).J. Clin. Invest. 53,922-931. Chan, C. K. ( 1989). Cancer Res., 49, 2656-2660. Chellapan, S. P. (1994).Mol. Cell Differ. 3, 201-220. Chen, Y. C., Faller, D. U. ( 1 995). Oncogene 11, 1487-1 498.

Ara-C: Cellular and Molecular Pharmacology

227

Chiba, P., Tihan, T., Szekeres, T., S a l m o n , Kraupp, M., Eher, R., Koller, U., and Knapp, W. (1990). Leukemia 11,761-765. Chinnaiyan, A. M., Orth, K., O’Rourke, K., Duan, H., Poirer, G. G., and Dixist, V. M. (1996). 1.Biol. Chem. 271,4573-4576. Chottiner, E. G., Shewach, D. S., Datta, N. S., Ashcraft, E., Gribben, D., Ginsburg, D., Fox, I. H., and Mitchell, B. S. (1991).Proc. Natl. Acad. Sci. U.S.A. 88, 1531-1535. Chu, M. Y., and Fisher, G. A. (1962). Biochem. Pharmacol. 11,423-430. Chu, M. Y., and Fisher, G. A. (1965). Biochem. Pharmacol, 14,333-341. Chwang, A. K., and Sundaralingam, M. (1973).Nature New Biol. 243,78-80. Cohen, A., and Ullman, B. (1985).Cancer Chemother. Pharmacol. 31,2841-2844. Coleman, C. N., Stoller, R. G., Drake, J. C., and Chabner, B. A. ( 1 975). Blood 46, 79 1-803. Collata, F., Polentarutti, M., Sironi, M., and Mantovani, A. (1992). 1. Biol. Chem. 267, 18278-18283. Colly, L. P., Peters, W. G., Richel, D., Arentsen-Honders, M. W., Starrenburg, C. W., and Willemze, R. (1987).Semirz. Oncol14,247-261. Colly, I.. P., Richel, D. J., Arentsen-Honders, M. W., Kester, M. G., ter-Riet, P. M., and Willemze, R. A. ( 1992). Hematology 65,26-32. Colly, L. P., and van Bekkum, A. (1982).Med. Pediatr. Oncol. 1, 209-219. Cozzarelli, N. (1977).Ann. Rev. Biochem. 46, 641-668. Cuvillier, O., Pirianov, G., Kleuser, B., Vanek, P. G., Coso, 0 . A., Gutkind, J. S., and Spiegel, S. (1996). Nature 381, 800-803. Dang, V. V. ( 199 I ) . Biochim. Biophys. Acta. 1072, 103-1 13. Datta, R., Banach, D., Kojima, H., Talanian, R. V., Alnemri, E. S., Wong, W., and Kufe, D. W. (1996).Blood 88, 1936-1943. Datta, R., Kharbanda, S., and Kufe, D. W. (1992).Mol. Pharmacol. 41, 64-68. de St. Vincent, and Buttin, G. (1979).Somat. Cell Mol. Getlet. 5, 162-167. de Saint Vincent, B. R., and Buttin, G. (1981).Biochim. Biophys. Acta 610, 352-359. de Saint Vincent, B. R., Dechamps, M., and Buttin, G. (1980).1.B i o l . Chem. 255, 162-167. Dijwel, P. V., and Wanka, F. (1978). Biochim. Biophys. Acta. 520,461-471. Dou, (2. P., An, B., and Will, P. L., (1995a). Proc. Natl. Acad. Sci. U.S.A. 92, 9019-9023. Dou, Q. P., An, B., and Yu, C. (1995b).Biochem. Biophys. Res. Commun. 214, 771-780. Dou, Q. P., and h i , V. W. Y. ( I 995). Cancer Res. 55,5222-5225. Drahovshy, D., and Kreis, W. ( 1970). Biochem. Pharmacol. 19,940-947. Duke, R. C., Chervenak, R., and Cohne, J. J. (1983). Proc. Nut/. Acad. Sci. U.S.A. 80, 6361-6365. Dunn, W. C., and Regan, J. D. ( 1 979). Mol. Pharmacol. 15, 367-374. Durham, J. P., and Ives, D. H. (1970)./. Biol. Chem. 245,2276-2284. El-Diery, W. S., Harper, J. W., O’Connor, P. M., Velculescu, V. E., Canman, C. E., Jackman, J., Peitenpol, J. A,, Burrell, M., Hill, D. E., Wang, Y., Wiman, K. G., Mercer, W. E., Kastan, M. B., Kohn, K. W., Elledge, S. J., Kinzler, K. W., and Vogelstein, B. (1994). Cancer Res. 54, 1 1 69-1 174. Ellims, P., Kao, A. Y., and Chabner, B. A. (1983).Mol. Cell Biochem. 57, 185-192. Emoto, Y., Kisaki, H., Manome, Y., Kharbanda, S., and Kufe, D. (1996).Blood 87,1990-1996. Estey, E. (1996).Leukemia 10 (Suppl 2), 39-43. Estey, E., Plunkett, W., Gandhi, V., Rios, M. B., Kantarjian, H., and Keating, M. J. ( 1993). l.ymphoma 9,343-350. Estey, E., Thall, P. F., Kantarjian, H., O’Brien, S., Koller, C. A., Beran, M., Gutterman, J., Deisseroth, A,, and Keating, M. (1992). Blood 79,2246-2255. Evan, G., and I.ittlewood, T. D. ( 1 994). Curr. Opin. Genet. Dev. 3, 44-49. Evan, G. I., Wyllie, A. H., Gilbert, C. S., L.ittlewood, T. D., Land, H., Brooks, M., Waters, C. M., Penn, t.. Z., and Hancock, D. C. (1992). Cell 69, 119-128. Evans, J. S., and Mengel, G . D. (1964).Biochem. Pharmacol. 13,989-994.

228

Steven Grant

Filnius,J., and Buick, R. N. (1985). Cancer Res. 45, 822-825. Fram, R. J., and Kute, I). W. ( I 982). Cancer Res. 4 2 , 4 0 5 0 4 0 5 3 . Freemerman, A. J., Vrana, 1. A., Tombes, R. M., Jiang, H., Chellappan, S. P., Fisher, P. B., and Grant, S. (1997).1,eukemia. 11, 504-513. Fridland, A. (1977a). Biochrm. Biophys. Res. Commtrn. 74, 72-78. Fridland, A. ( I977b). Biochemistry 16, 5308-53 12. Fridland, A., and Verhoef, V. ( 1987). Semin. Oncol. 14, 262-268. Furth, J . J., and Cohen, S. S. ( 1 968). Cancer Res. 28,206 1-2067. Gajewski. T. F., and Thompson, C. R. ( 1 996). Cell 87, 589-592. Gandhi, V., Estey, E., Keating, M. J., and Plunkett, W. (1993)./. Clin. Oncol. 1 1 , 116-124. Gandhi, V., and Plunkerr, W. (1988). Cancer Res. 48, 329-334. Gomez-Casares, M. T., Delgado, M. D., Lerga, A., Crespo, P., Quincoces, A. F., Richard, C., and Leon, J. (1993). Ixukemia 7, 1824-1 83.3. Gore, S. D., Wcng, L.,Jones, R.J., Cowan, K., Zilcha, M., I’iantdosi, S., and Burke, P. J. (1995). Clin. Cancer Rrs. 1 , 295-303. Graham, F. C., and Whitmore, G. F. (1970). Cancer Kes. 30, 2636-2644. Grant, S. ( 1 990). Phurmacol. Tlwr. 48, 29-44. Grant, S., Raker, M., and Bhalla, K. (1993). Leukemia 12, 1933-1938. Grant, S., Bhalla, K., and McCrady, C. (1991). 1.euk. Res. 15, 205-213. Grant, S., Bhalla, K., Rauscher, F., and Cadman, E. (1983a). Cancer Res. 43, 5093-5100. Grant, S., Freemerman, A. J., Birrer, M. J., Martin, H. A., Turner, A. J.. Szabo, E. Chelliah, J., and Jarvis, W. I). (1996). Cell Growth Differ. 7, 603-613. Grant, S., Jarvis, D., Swerdlow, P., Turner, A., Traylor, R., Wallace, H., Lin, P-S., Petit, G. R., and Gewirtz D. A. ( 19Y2). Cancer Rrs. 52, 6270-6278. Grant, S., Lehman, C., and Cadman, E. ( 1 980). Cancer Res. 40, 1.525-153 1. Grant, S., Rauscher, F., Jakuhowski, A,, and Cadman, E. (1983b). Cancer Res. 42,4007-4013. Grant, S., Turner, A., Bartimole, T. M., Nelms, P., Joe, V. C., and Jarvis, W. D. (1994). Onco/. Res. 6, 87-99. Grdina, D. J., Sigdestadt, C . I?, and Peters, L. J. ( 1 980). Br. 1. Cancer 42, 677-683. Green, D. R., McCahon, A., and Martin, S. J. ( I 996). I . Cell. BicJCbem. 60, 33-38. Greenberg, M. L., and Waxman, S. ( 1976). Eur. /. Cancer 12, 6 17-623. Gunji, H., Kharhanda, S., and Kufe, D. (1991 ). Cancer Res. 51, 741-743. Hamm, J., Bahij, C., Whitfield, J., I’farr, C. M., I.alleniand, D., Yaniv, M., and Rubin, L. L. (1995).Neuron 14, 927-939. Hande, K. R., and Chahner, B. A. (1978). Cancer Rrs. 38,579-585. Hapke, D., Stegmann, I? A., and Mitchell, B. S. ( 1996). Cancer Res. 56, 2343-2347. Hawtrey, A. O., Scott-Burden, T., and Robertson, G . ( I 974). Nature 252, 58-60. Hedley, W., and McCulloch, E. A. ( 1995). Leukemia 9, 789-798. Heinemann V, and Jehn Ll. ( I 990). Leukemia 4, 790-796. Heinemann, V., and Plunkett, W. ( 1989). Biochem. Pharmacol. 3 8 , 4 1 15-4 12 I. Heintz, N. H. and Hamlin, J . L. ( 1983). Biochemistry 22, 3557-3562. Hertel, L. W., Roder, G. B., Kroin, J. S., Rinzel, S. M., Poore, G. A,, Todd, G. C., and Grindey, G. B. (1990). Cancer Res. 5 0 , 4 4 17-4424. Herzig, R. H., Wolff, S. N., Lazarus, I., Phillips, G. I.., Karanes, G., and Herzig, G. P. (1983). Blood 62,361-369. Hiddemann, W., Kiehl, M., Zuehlsdorf, M., Busemann, C., Schleyer, E., Wormann, R., and / . 3 1-37. Buchner, T. (1992). Setnrn. O ~ C ( J19, Hiss, E. A., and Preston, R. J. (1977). Biochim. Biophys. Acta. 478, 1-8. Hockenberry, D. M., Oltavi, Z . N., Yin, X-M., Millman, C. I-., and Korsmeyer, S. J. (1993). Cell 75,241-251. Hu, Z-R., Minden, M. D., and McCulloch, E. A. (199Sa). Leukemia 9, 1667-1673.

Ara-C: Cellular and Molecular Pharmacology

229

Hu, Z. B., Minden, M. D., and McCulloch, E. A. (1996). Leukemia 10,410-416. Hu, Z. B., Yang, C;. S., Li, M., Miyanioto, N., Minden, M. D., and McCulloch, E. A. (199Sb). Leukemia 9, 789-798. Huang, P., Chubb, S., and Plunkett, W. (1990).J . Biol. Chem. 265, 16617-16625. Huang, S-H., Tomich. J. M., Wu, H., Jong, A., and Holcenberg, J . (1989). J. Biol. Chem. 264, 14762-14768.

Ibrado, A. M., Huang, Y., Fang, G., Liu, L., and Bhalla, K. (1996). Cancer Res. 56,4743-4748. Ikeda, M., Jakoi, L., and Nevins, J. R. (1996). Proc. Natl. Acad. Sci. U.S.A. 93, 3215-3220. Isakov, N., Calron, D., Mustelin, T., Pettit, G. R., and Altman, A. (1993). J . Immunol. 150, I 195-1 204. Jamieson, G. P., Finch, L. R., Snook, M., Wiley, J. S. (1987). Cancer Res. 47, 3130-3135. Jamieson, G. P., Snook, M. B., and Wiley, J. S. ( I 990). Leuk. Res. 14, 475-479. Jarvis, W. D., Gerwitz, D. A., Povirk, L., Turner, A., Traylor, R., Pettit, G. R., and Grant, S. ( 1994a). Biochem. Pharmacol. 47, 839-852. Jarvis, W. D., Kolesnick, R. W., Fornari, F. A,, Traylor, R. S., Gewirtz, D. A., and Grant, S. (3994b). Proc. Natl. Acad. Sci. U.S.A. 91, 73-77. Jarvis, W. D., Turner, A. J., Povirk, L. F., Traylor, R. S., and Grant, S. ( 1994c). Cancer Res. 54, 1707-1 714. Jiang, H., Lin, J., Zao-zhong, S., Collart, F. R., Huberman, E., and Fisher, P. B. ( I 994). Oncogene 9, 3397-3406. Jones, P., Baker, M., and Benedict, W. ( 1976). Cancer Res. 36,3789-3797. Karon, M., and Benedict, W. F. (1971). Science 178,62-65. Karon, M., Benedict, W. F., and Rucker, N. (1972). Cancer Res. 32, 2612-2615. Karon, M., and Chirakawa, S. ( I 970). Cancer Res. 29, 687-696. Kataoka, S., Naito, M., Tornida, A., and Tsuruo, T. (1994). Exp. Cell Res. 215, 199-205. Kaufmann, S. H. (1989). Cancer Res. 49,5870-5878. Keating, M. J., Kantarjian, H., Talpaz, M., Redrnan, J., Koller, C., Barlogie, B., Velasquez, W., Plunkett, W., Freireich, E. J., and McCredie, K. B. ( 1989). Blood 74, 19-25, Keith, F. J., Bradhury, D. A., Zhu, Y. M., and Russel, N. H. (1995).Leukemia 9, 131-138. Kessel, D., Hall, T. C., and Wodinsky, 1. (1967). Science 156, 1240-1241. Kharbanda, S., Datta, R., and Kufe, D. (1991). Biochemistry 30, 7947-7952. Kharbanda, S., Emoto, Y., Kisaki, H., Saleern, A,, and Kufe, D. (1994). Mol. Pharmacol. 46, 67-72.

Kharhanda, S., Huberman, E., and Kufe, D. (1993a). Biochem. Pharmacol. 45, 205.5-206 I . Kharbanda, S., Pandey, P., Ren, R., Mayer, B., Zon, I.., and Kufe, D. (199S).J. Biol. Chem. 270, 30278-3028 1.

Kharbanda, S., Saleem, A., Rubin, E., Sukhatme, V., Blenis, J., and Kufe, D. (l993b). R m chemistry 32, 9 137-9 142. Kharbanda, S. M., Sherman, M. L., and Kufe, D. W. ( 1990)./. Clin. Invest. 86, 15 17-1 523. Klohs, W. D., Bernacki, R. J., and Kortnyk, W. (1979). Cancer Res. 39, 1231-1238. Kolla, S. S., and Studzinski, G. P. (1994). Cancer Res. 54, 3418-1421. Koo, H-M., Monks, A., Mikheev, A., Rubinstein, L. V., Gray-Goodrich, M., McWilliams, M. I., Alvord, W. G . , Ole, H. K., Gazdar, A. F., Paull, K. D., Zarbi, H., and Vande Woude, G. F. (1996). Cancer Res. 56, 5211-5216. Kraft, A. S., Smith, J. B., and Berkow, R. L. (1996). Proc Nut/ Acad Sci U.S.A. 83, 1334-1338. Kreis, W., Budman, D. R., Chan, K., Allen, S. L., Schulman, P., Lichtrnan, S., Weiselberg, L., Schuster, M., Freeman, J., Akerman, S., Attas, L., and Vinciguerra, V. ( I 992). Leukemia 5, 99 1-9 98.

Kreis, W., Chan, K., Budman, D. R., Schulrnan, P., Allen, S., Weiselberg, L., Litchman, S., Henderson, V., Freeman, J., Deere, M., Andreef, M., and Vinciguerra, V. ( 1988). Cancer Res. 48, 1337-1 342.

230

Steven Grant

Kucera, G. I.., and Capizzi, R. L. (1992). Cancer Res. 52, 3886-389 1. Kufc, I). W., and F.gan, E. M. (1981a).Biocbem. Phurmacol. 30, 129-133. Kufe, D. W., Major, P. P., Egan, M., and Beardsky, P. (198 I b).J. B i O / . Chem. 2.55, 8997-9000. Kufe, D. W., Munroe, I)., Herrick, D., Egan, E., and Spriggs, 1). (1984a).Mid. Pharmacol. 26, 128-134. Kufe, D. W., and Spriggs, 1). R. (1985). Srnrin. O ~ C C12, J ~(Suppl . # 3 ) , 34-48. Kufe, D., Spriggs, D., Egan, E. M., and Munroe, I). (1984b).Blood 64, 54-58. Kuwakado, K., Kubota, M., Bessho, R., Katoaoka, A., Usami, I., Lin, Y. M., Okuda, A,, a n d Wakazono, Y. (199.5).I m k . Res. 19,645-6S0. Kyriakis, J. M., Banergee, P., Nikolakaki, E., L h i , Rubin, E. A., Ahniad, M . F., Avruch, J., and Woodgett, J. R. (1994).Nature (Lond) 369, 156-160. Lacombe, F., Puntous, M., Dumain, I?, Cony-Makhoul, P., Belloc, F., Bernard, P., and Boisseau, M., Reiffcrs, J. (1996). Leuk. Res. 20,481-489. Laliherte,J., and Momparler, R. L. (1994). Cancer Res. 54, 5401-5407. Leinoli, R. M., Gulati, S. C., Strife, C., Lambck, C., I’crez, A., and Clarkson, B. D. (1991). Leukemia 5. 386-39 1. Levitzki, A,, and Gazit, A. (199.5).Science 267, 1782-1788. Liliemark, J. O., and Plunkerr, W. M. (1986).Cancer Res. 46, 1079-1083. Lisra, P., Porcu, P., Avanzi, G. C., and Pegoraro, L. (1988).Br. J . Haematol. 69, 121-123. Lister, T. A., Rohatiner, A. Z . S., Bassan, R., Gregory, W., Willis, L., Barnett, M. J., Waxman, J., Richards, T. S., Gancsan, T. S.,Tucker, J., Malik, S., Biruls, R., and Guess J. (1987).Semirz Oncol2(Suppl) 53-55, Liu, E. T. (1990).The role of ras gcne mutations in myeloproliferativc disorders. Clinics Lab. Med. 10, 797-807. Liu, Z . , Hsu, H., Gocddel, D. V., and Karin, M. (1996).Cell 87, 565-576. Logan, T. J., Evans, D. L., Mercer, W. E., Bjornsti, M-A., and Hall, D. J. (1995).Cancer Res. 55.2883-2891. Lotem, J., Cragoe, E. J., and Sachs, 1. (199 I ) . Blood 78, 953-960. Lotem, J., and Sachs, I.. (1992).B h i d 80, 1750-1757. Major, P. P., Egan, E. M., Beardsley, G. P., Minden, M. E., and Kufe, D. W. (1981).Proc. Natl. Acad. Sci. U.S.A. 78, 3235-3239. Major, P. P., Egan, E., Herrick, D. J., and Kufe, D. W. (1982). BffJCbem.Pharmacol. 31, 2937-2940. Maley, F., and Maley, G.F. (1971).Arch. Biochem. 144, 723-729. Mancini, W. R., and Cheng, Y-C. (1982).Mol. Pharmucol. 23, 159-164. Manome. Y., Weichselbaum, R. R., Kufe, D. W., and Fine, H. A. (1993). Oncol. Res. 5 , 139-144. Manome, Y., Wen, P. Y.,Dong, Y., Tanaka, T., Mitchell, 8. S., Kufe, D. W., and Fine, H. A. ( 1996). Nature Med. 2, 567-573. Markovits, J., Larscn, A. K., Segal-Bendirdjian, E., Fosse, P., Saucier, J. M., Gazit, A., Levitzki, A., Umezawa, K., and Jacqucrmin-Sablon, A. (1994). Biochem. Pharmacol. 48,549-560. Mastrianni, D. M., Tung, N. M., and Tcnern, D. G. (1992).Am. J. Med. 92,286-292. Matsukage, A., Ono, K., Ohashi, A. Takahashi,T., Nakayama, C., and Saneyoshi, M. (1978a). Cancer Res. 38, 3076-3079. Matsukage, A., Takahashi, T., Nakayama, C., and Snacyoshi, M. (1978b). J. Biochem. 83, 1511-151.5.

McCahon, A., Bissonnette, R., Schmitt, M., Cotter, K. M., Green, D. R., and Cotter, T. G . (1994).B h d 8 3 , 1179-1187. Mikita, T., and Beardsley, G. P. ( 1988). Biochemistry 27, 4698-4705. Miller, M., and Chinault, (1982).J. Biol. Chem. 257, 10204-10209. Mills-Yamamoto, C., Lauzon, G. J., and Paterson, A. R. P. (1978).Biochem. Pharmacol. 27, 1 8 1-1 86.

Ara-C: Cellular and Molecular Pharmacology

231

Miyashita, T., and Reed,J. C. (1993). B l o o d 8 1 , 151-157. Mornparler, R. ( 1 969). Biochem Biophys. Res. Commun. 34,465-471. Mornparler, R. l.., and Fischer, G. A. (1968).J. B i d . Chem. 243,4298-4304. Moore, D. C., Carter, D. L., and Studzinski, G. P. (1992).]. Cell Physiol. 151, 539-548. Moore, E. C., and Hulbert, R. 8 . ( 1 984). Pharmacol. Ther. 24,439-447. Nagyvary, J.. Provenzale, K., and Clark, J. M. (1968). Biochetn. Biophys. Res. Commun. 31, 508-514. Neubauer, A., Dodge, R. K., George, S. L., Davey, P. R., Silver, R. T., Schiffer, C. A., Mayer, K. J., Ball, E. D., Wurster-Hill, D., Bloomfield, C. D., and Liu, E. T. (1994). Blood 83, 1603-161 1. Nigg, E. A. (1 993). Curr. Opin. Cell Biol. 5, 187-1 93. Nishizuka, Y. (1984). Nature (Land.) 308,693-698. Noordhuis, P., Kazernier, K. M., Kasperrs, G. J., and Peters, G. J. (1996). I x u k . Res. 20, 127-134. Obeid, 1.. M., Linardic, C. M., Karolak, L. A., and Hannun, Y. A. (1994). Science 259, 1769-1 771. Ohno, Y., Spriggs, D., Matsukage, T., Ohno, T., and Kufe, D. (1988). Cancer Res. 48, 1494-1 498. Owens, J. K., Shewach, D. S., Ullrnan, B., and Mitchell, B. S. (1992). Cancer Res. 52, 23 89-23 93. Pallavicini, M. G. (1984). Pharmucol. Ther. 25, 207-238. Pallavicini, M. G., Gray, J. W., and Folstad, L. J. (1982). Cancer Res. 42, 3 125-3 13 1. Perrino, F. W., and Mekosh, H. L. (1992).]. Biol. Chem. 267,23043-23051. Peters, W. G., Colly, L. P., and Willemze, R. (1988). Blur 56, 1-1 I . Peters, W. G., Willernze, R., and Colly, L. P. (1987). Senzin. Hematol. 2, (Suppl I ) , 86-91. Plagernann, P. G. W., Marz, R., and Wolheuter, R. M. (1978). Cancer Res. 38, 978-989. Powell, B. L., White, J. C., Gregory, B. W., Brockschrnidt, J. K., Rhinehardt-Clark, A., Lyerly, E. S., Chorely, H. M., and Capizzi, R. M. (1991). Leukemia 5, 598-601. Preisler, H. D., Rusturn, Y., and Priore, R. L. (1979). Cancer Res. 39,42-49. Rajotte, D., Haddad, P., Harnan, A., Cragoe, E. J., and Hoang, T. ( 1 992). /. Biol. Chem. 267, 99 80-9987. Ray, S., Bullock, G., Nunez, G., Tang, G., Ibrado, A. M., Huang, Y., and Bhalla, K. (1996). Cell Growth Differ. 7, 1617-1623. Raza, A,, Preisler, H. D., Day, R., Yasin, Z., White, M., Lykins, J., Kukla, C., Barcos, M., Bennett, J., Browrnan, G., Goldberg, J., Grunwald, H., Larson, R., Vardiman, J., and Vogler, R. (1990). Blo0d76,2191-2197. Riva, C., el-Khyari, S., Rustum, and Barra, Y. (1995). Anticancer Res. 15, 1297-1 302. Ross, D. D., Akrnan, S. A., Joneckis, C. C., Yang, E., and Bachur, N. R. (1984). Cancer Res. 44, 1530-153s. Ross, D. D., Chen, S-R., and Cuddy, D. P. (1990). Cancer Res. 50,2658-2666. Ross, D., Cuddy, D. P., Cohen, N., and Hensley, D. (1992). Cancer Chemother. Pharmacol. 31, 61-70. Ross, D. D., Thompson, B. W., Joneckis, C. C., Akman, S. A., and Schiffer, C. A. (1986). Blood 68,76-82. Rustum, Y. M. (1982). Cancer Res. 38, 543-549. Rustum, Y. M., and Preisler, H. D. (1979). Cancer Res. 39,42-49. Rusturn, Y. M., and Raymakers, R. A., (1992). Pharmacol. Ther. 56, 307-321. Saleem, A., Datta, R., Yuan, Z-M. Kharbanda, S., and Kufe, D. (1995). Cell Growth Differ. 6, 1651-1658. Sandoval, A., Consoli, U., and Plunkett. (1996). Clin. Cancer Res. 2, 1731-1741. Santi, D. V., McHenry, C. S., and Sornmer, A. (1974). Biochemistry 13, 471-480. Schirnke, R. T. (1984). Cancer Res. 44, 1735-1742.

232

Steven Grant

Schwartz, S. A., Morgcnstcrn. B., Capizzi, R. 1.. ( 1 982). Cancer Research 42, 21 91-2197. Shan, B., Durfee, T., and Lee, W-H. (1996). /’roc. Nut/. Acad. Sci. U.S.A. 93. 679-684. Shan, B., Farmer, A. A,, and Lee, W-H. (1996).Cell Growth Differ. 7, 689-697. Shan, B., and Lee, W-ti, ( 1994). M o l . Cell Biol. 14, 8 166-8 173. Sherr, C. J., and Roberts, J. M. (1994). Genes Deic 9, 1 149-1 163. Shi, L., Nishioka, W. K., Th’ng, J., Bradbury, E. M., Litchfield, D. W., and Greenberg, A. H. (1994).Science 263. 1143-1 145. Shimizu, T., O’Chmor, P. M., Kohn, K. W., and Pommier, Y. ( 1 99.5). Cancer Res. 55,228-23 I . Smith, M. A., Singer, C. R., Pallister, C. J . , and Smith, J. G. (199.5). Br. J . Haematol. 90, 767-773. Sniyth, J. F., Robins, A. B., and Leese, C. 1.. (1976). Eur. J . Cancer 12, .567-573. Solary, E., Bertrand, R., Kohn, K., and Pomniier, Y. (1993). Blood 8 1 , 1.359-1368. Stegmann, A. P. A., Honders, W. H., Willemze, R., van Haperen, R., and Landegent,J. E. ( 1 995). Blood 85, 1188-1 194. Steinmann, R. A., Hoffman, R., Iro, A,, Guillouf, C., Liebermann, I>. A., and El-Houseini, M. E. ( 1994). c)?lCOgPIJL’9, 3.389-3396. Steuart, C. D., and Burke, P. J. (1971). Nature Neiu Riol. 233, 109-1 10. Streifel,J. A., and Howell. S. B. (1981). l’roc. Natl. Acad. Sci. U.S.A. 7 8 , 5’132-5136. Strum, J. C., Small, G. W., Pauig, s. B., and Daniel, L. W. (1994). /. BiOl. Chem. 269, 15493-15497. Studzinski, G. P., Reddy, K. B., Hill, H. Z., and Bhandal, A . K. (1991). Cancer Res. 51, 34.5 1-3455. Stryckmans, P., DeWitte, T., Bitar, N., Marie, J. P., Suciu, S., Sohbu, G., Debusschcr, L., Bury, 1 . . Peetermans, M., Andrien, J. M., Fiere, D., Bron, I)., Dekker, A., and Zittoun, R. (1987). Seviin. Oncol. Z(Supp1 I ) , 67-72. Sundman-Engberg, B., Tidefelt, U., and Paul, C;. (1996). Eur. J . Haematol. 56, 1-6. Tafuri, A., Lemoli, R. M., Chen, R., Gulati, S. C., Clarkson, 6. D., and Andreef, M. (1994). Leukemia 8 , 749-757. Tang, C.,Huang, Y., Ponnathpur, V. S., Ray, S., Mahoney, M. E., Bullcok, G., Ibrado, A. M., and Bhalla, K. (1994). L e d . Lynzphoma 15,445-4.5 I . Tattersall, M. H. N., Ganeshagaru, K., and Hoffbrand, A. V. (1974). l3r.J. Huematol. 27,39-46. Teofili, L., Pierelli, I.., lovino, M. S., Leone, C., Scarnbis, G., de Vincenzo, R. Bencdetti, P. P., Menichella, G., Macri, E., Piantelli, M., Ranelletti, E O., and Larocca, L. M. ( I 992). / a r b . R ~ s 16,497-505. . Ubezio, P., Tagliabue, G., Schechter, R., and Ague, Z. (1994). Cancer Res. 54, 6446-64.51. Van der Lely, De Witte, T., Wesscls, J., Raymakers, R., Muss, P., and Preijers, F. ( 1 994). Ann. Hematol. 68, 22.5-232. Waldman, T., Kinzler, K. W., and Vogelstein, B. (1995).Cancer Res. SS, 5187-5190. Waldman, T., Lengauer, C., Kinzler, K. W., and Vogclstein, B. (1996). Nature 381, 713-716. Walker, N. I., Harmon, B. V., Gobe, G. C., and Kerr, J. E R. (1988). Methods Achieu t x p . rutho/. 13, 18-54. Wang, C-Y., Mayo, M. W., and Baldwin, A (1996). Science 274, 784-787. Wang, L. M., and Kucera, G. L. (1994). BloCh?WJ. Biophys. Acta. 1224, 161-167. Weinberg, G., Llllman, B., Martin, D. W. (1981). /’roc-. Nutl. Acad. Sci. U.S.A. 78,2447-2451. Weinberg, R. A. (1995). Cell81, 323-330. Whelan, J., Phear, G., Yniauchi, M., and Meuth, M. (1993). Nature Genet. 3, 317-322. Whelan, J., Smith, T., I’hear, C., Rohatiner, A., Lister, A., and Meuth, M. (1994). lxukemiu 8, 264-265. White, J. C.,and Hincs, L. H. ( 1 987). Cancer Res. 47, 1820-1 824. White, J. C., Hines, I.. H., and Rathmell, .I.P. (1985). Cancer Res. 45, 3070-3075. White, J. C., Rathmell, J. P., and Capizzi, R. L. (1987)./. Clin. Invest. 79, 380-387.

Ara-C: Cellular and Molecular Pharmacology

233

Wiley, J. S., Cebon, J. S., Jamieson, G. P., Szer, J., Gibson, J., Woodruff, R. K., McKendrick, J. J., Sheridan, W. P., Biggs, J. C., Snook, M. B., Brocklebank, A. M., Rallings, M. C., and Paterson, A. R. P. (1994). Leukemia 8, 181-1 85. Wiley, J. S., Jones, S. P., Sawyer, W. H., and Paterson, A. R. P. (1982).J. Clin. Invest. 69, 47949. Wiley, J. S., Snook, M. B., and Jamieson, G. P. (1989).Br. J. Haematol. 71, 203-207. Wiley, J. S., Snook, M. B., and Jamieson, G. P. (1992).Leukemia 6,486-487. Wiley, J. S., Taupin, J.. Jamieson, G. P., Snook, M., Sawyer, W. H., and Finch, L. R. (1985).J. Clin. Invest. 75, 632-642. Williams, G. T., Smith, C. A., Spooncer, E., Dexter, T. M., and Taylor, D. R. (1990).Natrtre 343,76-79. Wills, P., Hickey, R., Ross, D., Cuddy, D., and Malkas, L. (1996). Cancer Chemother. Pharrnocol. 38, 366-372. Woodcock, D. M. (1987).Semin. Oncol. 14,251-256. Woodcock, D. M., Adams, J. K., and Cooper, 1. A. (1982).Cancer Res. 42,4744-4752. Woodcock, D. M., and Cooper, I. A. (1981). Cancer Res. 41, 2483-2490. Woodcock, D. M., and Cooper, I. A. (1979).Exp. Cell Res. 123, 157-166. Woodcock, D. M., Fox, R. M., and Cooper, I. A. (1976).Cancer Res. 36, 3789-3797. Wu, Y., Hickey, R., Lawlor, K., Wills, P., Yu, F., Ozer, H., Starr, R., @an, J. Y., Lee, M., and Malkas, I.. ( I 994).J. Cell Biochem. 54, 3 2 4 6 . Wyllie, A. H., Kerr, J. F. R., and Currie, A. R. (1980).Int. Rev. Cytol. 68, 251-306. Wyllie, A. H. (1981).In “Cell Death in Biology and Pathology.” pp. 9-34. Bowen, 1. D., Lockshin, R. A. (Eds.). London, Chapman and Hall. Yang, J - L . Cheng, E. H., Capizzi, R. L., Cheng, Y-C., and Kute, T. ( 1 985).J. Clin. Invest. 75, 141-148. Yang, E., and Korsmeyer, S. J. (1996).Blood 88, 3 8 6 4 0 I . Yen, A,, and Varvayanis, S. (1992). In Vitro Cell Dev. Biol. 28, 669-672. Young, R . S. K., and Fisher, G. A. (1968).Biochem. Biophys. Res. Commun. 32, 23-29. Yoshida, S., Yamada, M., and Masaki, S. (1977).Biorhim. Biophys. Acta, 144-150. Yuan, Z-M., Kharbanda, S., and Kufe, D. (1995).Biochemistry 34, 1058-1063. Zahn, R., Muller, Foster, W., Madnof, A,, and Beyer, R. (1972).Eur. J. Cancer 8 , 391-396. Zimmerman, K., and Alt, F. W. ( 1 990). Crit. Rev. Oncog. 2, 75-95.

This Page Intentionally Left Blank

Index

A Actin, stress fiber formation, 82-83 Activation MAPK cascade, 6 1-65 oncogenes, 4-6 AF6IRsb1, candidate Ras effector, 74 Aging, associated with promoter CpG island methylation, 178-1 81 Alleles, inethylation spreading across, 180-181 Alu repeat elements, adjacent CpG island sites, 183-1 84 Anatomy CpG islands in tumor suppressor genes, 182-184 DNA methylation in higher-order eukaryotes, 147-1 5 0 Angiogenesis, separate from solid tumor development, 18-1 9 Animal models DNA methylation changes in cancer, 175-1 76 lung carcinogenesis, p l 6 gene studies, 162-163, 187 Antisense oligonucleotides, B-myb, 127-128 APC gene, frameshift mutations, 4 8 Apoptosis ara-C role, 209-21 0, 223-225 associated with JNK activation, 66-67 distal pathway defects, 21 3 genes regulating, 9-1 1 hematopoietic growth factor effect, 223 APRT gene, CpG islands, protection mechanisms, 182-1 84 1-P-D-Arabinofuranosylcytosine, see ara-C I -p-D-Arabinofuranosylcytosine 5’-triphosphate, see ara-CTP

ara-C cytotoxicity mechanism, 205-210 modulation, 220-225 resistance mechanisms, 2 10-2 1 3 signaling pathways and oncogene interactions, 213-220 structure and metabolism, 200-205 ara-C monophosphate, conversion to araCTP, 203-204 ara-CTP dephosphorylation, 21 1 formation, 205

B Base excision repair, 4 2 bcl-2 apoptosis-modifying, 9-1 0 interaction with ara-C, 215-216 Biochemical modulation, ara-C-associated cytotoxicity, 221-222 BLA, see CD77 B-Myb DNA-binding domain, 112-1 13 evolutionary conservation, 111-1 12 hyperphosphorylation in S phase, 123-125 negative C-terminal domain, 1 14-1 15 possible target genes, 133-134 requirement in cell proliferation, 127-1 3 2 transactivation function, cyclin dependence, 125-127 transcription activation domains, 1 13-1 14 transcriptional control by, 132-1 33 B-myb under cyclin-dependent regulatory mechanisms, 134-136

235

236 exprcssion ablation, 127-128 developmental regulation, I 17-1 18 overexpression, 129-1 30 transcription as cell cycle control point, 122- I23 in somatic cells, 115-1 17 transcriptional regulation, 11 8-122 Breast cancer, susceptibility genes, 15-16 Breeding, selective, leading to gene fixation, 13 Bryostatin- I , effect on ara-(;-induced apoptosis, 224 Burkitt’s lymphoma cellular cxpression of C1177, 1 1 mutated p5.3, 17

C c-ubl, interaction with ara-C, 2 19 E-Cadherin gene CpG islands, 183-1 85 loss of expression in tumors, 164-165 Caenorhuhditis eleguns, Ras signaling cascade, 62-63 Cancer development, multistep theory, 1-20 DNA hypomethylation, 15 1-1 5 2 DNA methylation changcs, 148-1 SO clinical implications, 189-1 9 0 genes hypermethylared in, 166-1 6 7 prevention, mutator phenotype role, 50-5 I Cancer cells genome, increased DNA-methyltransferase activity, I 7 1-1 72 number of mutations, 33-34, 5 I Carcinogenesis, multistep, molecular biology, 14-17 CD77, expressed by Burkitt’s lymphoma cells, I 1 CDC42 candidate effectors, 86-9 I constitutively activated, 79, 82-85 effect on actin cytoskeleton, 82-83 CDC42/Rac/interactive/binding motif, 87, 89-90 Cdk, see Cyclin-dependent kinase Cdk inhibitors genes p15 and p16, role in solid tumors, 158-163 as negative regulatory proteins, 130-1 32

Index

Cell cycle checkpoint control, tumor suppressor gene role, 6-8 control point, B-myb transcription as, 122-123 genes governing, 40-41 stage, and B-my6 transcription, I 16-1 17 Cell cycle genes, CDE and CHR, 120-121 Cell death pathway, defects, 213 Cell lines M I myeloid leukemia, IL-6 effect, 129 NIH 3T3, 172-1 73 RIEI, transformational studies, 67-68 Cell proliferation, requiremcnt for B-Myb, 127-132 Cellular heterogeneity, in tunior progression, 187-189 Chain terminator, ara-C role, 206-207 Chromatin pattern, transcribed genes, 149-150 Chromosomal aberrations, multiplicity i n tumors, 26-30, 33-34 Chromosomes 2 and 3, MSH2 mutation mapped to, 12-13 1 1p, CpG island methylation, IS3 Citron, Rho-binding element, 93-94 c-iun, interaction with ara-C, 214 Clinical implications, altered DNA methyl;ition in cancer, 189-1 90 Clonal evolution, in tumor progression, 28-29 Clonal expansion and promoter hypermethylation, 187188 role in multiple mutations, 34-35 Clonal selection, multiple rounds, 2 7 c-Myb conserved region, I 15 transactivation domain, 11 3-1 1 4 c-myc interaction with ara-C, 2 15-2 16 juxtaposition to immunoglobulin locus, 5 Colon, age-related hypermethylation patterns, 179-1 80 Colon carcinoma, see also Hereditary nonpolyposis colon cancer DNA-methyltransferase transcript levels, 170-171 estrogen receptor gene hypermethylation, 165-1 66

Index Colorectal tumors, sequence of events, 15 C p C islands aberrant methylation in neoplasia, 177-1 86 associated with gene S’regulatory regions, 148-1 50 hypermethylation in fibroblasts, 175 in neoplastic cells, 154-167 I 1p region, 153 protection in normal cells, 182-1 85 unmethylated, 14 1-14,? CRIB motif, see C:DC42/Rac/interactive/binding motif CTP synthase, alterations, 212 Cyclin-cyclin-dependent kinase complex, inhibition, 8 Cyclin D, and loss of p16 gene, 159 Cyclin-dependent kinase critical role in cell cycle, 110 growth-promoting signals, 130-132 Cyclin E/Cdk2 complex, in B-Myb hyperphosphorylation in S phase, 124 Cyclin E-H 1 kinase, interaction with ara-C, 220 Cyclins, B-Myb transactivation function dependent on, 125-127 Cytidine deaminase, in ara-C deamination, 204 Cytokinetic resistance, ara-C, 212 Cytosine, methylation in higher-order eukaryotes, l44-lS0 Cytosine arabinoside, see am-<: Cytotoxicity, ara-C mechanism, 205-2 10 modulation, 220-225

D Dbl family proteins, effect on Rho protein function, 80 Deamination, ara-C, 204 enhanced, 21 1 Deoxycytidine kinase ara-C phosphorylation, 203-204 loss, and ara-C resistance, 210-211 Deoxycytidine triphosphate pool sizes, 204 increased, 2 1 1-2 12 reduced, 221-222 role in ara-C metabolism, 201

237 Deoxycytidylate inhibitor, dTHU, effect o n ara-C metabolism, 204 Dephosphorylation, ara-CTP, 21 1 Diagnosis, mutator phenotype role, 48-49 Diagnostic marker, gene amplification as, 33 DNA fragmentation, ara-C role, 209-210 hypomethylation in cancer, 1.51-152 incorporation of ara-C, 206-207 regional hypermethylation in cancer, 152-1 54 DNA-binding domain, B-Myb, 112-1 13 DNA damage induction of p53, 8-9 in mutation origin, 37-38 DNA helicases, targets for mutator phenotype, 42 DNA methylation abnormalities in neoplasia, 150-1 6 7 altered clinical implications in cancer, 189-1 90 role in tumor progression, 186-1 89 anatomy in higher-order eukaryotes, 147-1 SO changes in neoplastic cells, underlying mechanisms, 167-186 in eukaryotes, history, 142-144 pattern establishment, DNA-methyltransferase role, 145-147 in prokaryotes, 144-14.5 DNA-methyltransferase increased activity in neoplastic cells, 167- 17 1 role DNA methylation pattern establishment, 145-1 4 7 tumor progression, 171-1 7 6 DNA polymerase p, mutations in cancer cells, 3 9 DNA polymerases candidate target mutator genes, 3 8 4 0 inhibitory effects of ara-C, 205-206 misincorporation, 36-40 DNA repair, in mutation origin, 37-38 DNA repair genes candidate targets for mutator phenotype, 41-42 role in tumor progression, 8-9 DNA replication, in mutation origin, 36-37 DNA replication complex model (MRC), 207

index

DNA replication genes, candidatc targets for mutator phenotype, 41-42 DNA synthesis, infidelity, 28, 39-40 Downstream repression site, B-myb E2F, 120-121 Drosophila melanogaster, MAPK cascade activation, 62 Drug discovery, Raf-+MEK+MAPK kinase cascade role, 97

E Early event, mutator phenotype as, 47-48 Early growth response- 1 gene, interaction with ara-C, 220 E2F transcription factor interaction with ara-C, 2 I7 role in G , / S transition, 110-1 1 1 , 119-122 tumor suppressor function, 7 Embryogenesis, B-myb expression throughout, 117-1 18 Endoreduplication, ara-C, 207 Epigenetic changes, superimposed in DNAlevel changes, 3 Epithelial cells, transformational studies, 67-68 Escape mechanisms, tumor cell from host control, 17-1 9 Escherichia coli, mismatch repair genes, 44-45 Estrogen receptor gene, CpG island hypermethylation, 165-166 Eukaryotes DNA methylation, history, 142-144 higher-order, cytosine methylation roles, 144-150 Evolutionary conservation B-Myb, I 1 1-1 I2 DNA-methyltransferase, 146-1 47

Farnesyltransferase inhibitors, blocking of Ras function, 96-97 Fibroblasts C p C island hypermethylation, 175 DNA methylation ability, 169 Filopodia, induction by CDC42, 82-83 Fludarabine, interaction with ara-C, 221 Foulds, L., multistep theory of tumor development, 1-20

G Gene amplification diagnostic marker for malignant cells, 33 oncogene activation, 5-6 Gene knockout, B-myh, 128 Genes breast cancer susceptibility, 15-16 in DNA repair and replication, 41-42 governing cell cycle, 40-41 hypermethylated during aging, 186-187 promoter regions aging-associated C p C island methylation, 178-1 8 1 mutational changes, 18 1-1 82 regulating apoptotic cell death, 9-1 1 with and without C p C islands, 148-150, 152 Genome cancer cell, increased DNA-methyltransferase activity, 171-1 72 eukaryotic, presence of DNA methylation, 143-1 44 Clutathione-S-transferase-a gene, hypermethylated in cancer, 167 Glycoprotein, membrane, ara-C effect, 208 Crb2 adaptor protein, Src homology 2 domain, 61 GTPase cascade, activation by Ras, role of Rho family proteins, 78-85 G+T transversions, p53 mutations, 40-4 I

H Hematopoietic growth factors, combined with ara-C, 220,222-223 Hereditary nonpolyposis colon cancer microsatellite instability, 4 4 4 5 mismatch repair mutations, 12-13, 48 Heterogeneity cellular, in tumor progression, 187-1 89 within tumors, 26 HlCl gene, candidate tumor suppressor gene, 166 Histone H I , interaction with ara-C, 219 HNPCC, see Hereditary nonpolyposis colon cancer Hypermethylared-in-cancer gene, see H l C l Hypermethylation C p C islands in fibroblasts, 175

Index

in neoplastic cells, 154-167 and tumor progression, 177-186 DNA, regions in cancer, 152-154 Hyperphosphorylation, B-Myb in S phase, 123-125 Hypomethylation, DNA, in cancer, 151-152

Immunoglobulin locus, oncogene juxtaposed to, 5 , 10, 1 7 Inactivation gene, promoter region methylation role, 185 tumor suppressor genes, 157-1 59, 176 Insulin-like growth factor-2 gene expression in tumors, 179-1 8 1 in tumorigenic process, 1 8 Interleukin-6, effect on M1 cells, 129 IQGAPs, candidate effectors of Racl/CDC42,90

I

JNK, see Jun NH,-terminal kinase Jun NH,-terminal kinase cascade, activation, 65-67 transcription factor activation, 83-84

L Lesions, DNA, oxygen free radical-induced, 3 7-3 8 Leukemia blasts, nucleoside transport-binding sites, 202 hematopoietic growth factors plus ara-C, 222-223 p16 and p l 5 genes, 163-164 Lipid membrane, ara-C effect, 208 Loss of heterozygosity, in tumor progression, 15-16 Lung carcinogenesis, p l 6 gene C p C island hypermethylation, 162-1 63, 187 Lymphoma follicular, hcl-2 rearrangement in, 9-1 0 p l 5 and p l 6 alterations, 164

M Malignancy, hematopoietic am-C role, 199 p16 and p l 5 genes, 163-164

239 MAPK, see Mitogen-activated protein kinase MEK, phosphorylation by MAPK, 6 5 MEKK1, interaction with Ras, 9 6 Membrane ruffling Racl(12V)-induced, 89-90 role of Rho family proteins, 82-83 Membranes, lipid and glycoprotein, ara-C effects, 208 Metabolism, ara-C, 200-205 Methylcytosine, genornic origin, 143 Microsatellite instability HNPCC, 44-45 HNPCC tumors, 12-13 nucleotide sequence variations, 43-44 sporadic tumors, 45-46 Misincorporation, by DNA polymerases, 36-40 Mismatch repair, in mutation origin, 36-37 Mismatch repair genes in HNPCC, 44-45,48 mutations, 12-13 Mitogen-activated protein kinase cascade, activation, 61-65 interaction with ara-C, 21 8 Mitogen-activated protein kinase kinase 1, candidate Ras effector, 7 4 Mixed-lineage kinase, MLK3, association with CDC42 and Racl, 87-88 Molecular biology, multistep carcinogenesis, 14-17 Mutagenesis enhanced, coupled with somatic selection, 35-36 switch 11, 76 Mutants, Ras effector domain, 69-70, 98 Mutation rate, spontaneous, 32 Mutations accumulation mechanisms, 46-47 rate delay, 50-5 1 destabilizing, caused by loss of DNA repair functions, 11-14 gene promoter regions, 181-1 8 2 multiple clonal expansion role, 34-35 somatic selection role, 35-36 in tumor evolution, 27-28 numbers, in cancer cells, 33-34, 51 origins, 36-38 p53, in cell subpopulations, 188 reversion, 29-30

240 series, providing cell with selective advantage, 2, 6 during tunior progression, 47-48 Mutator phenotype in cancer prevention, SO-S 1 candidate target mutator genes, 38-43 diagnostic role, 48-49 evidence from microsatellite instability, 43-47 historical perspective, 28-32 origins of mutations, 36-38 prognostic role, 49-50 requirement for, 32-36 timing and diversity of mutations, 47-48 Myb hinding site, transactivation independent of, 133-134 Myosin phosphatase, RhoA-G7'1'-hinding protein, 94

N Natural selection, Darwinian, 1, 19-20 Negative regulatory domain, B-Myb, 114-1 15, I26 Neoplasia CpG island aherrant methylation, 177-186 DNA methylation ahnormalities, 150-1 6 7 Neoplastic cells ara-C resistance, 2 10-2 13 CpG island protection mechanisms, 182-1 8.5 increased DNA-methyltransferase activity, 167-172 Neurofibromin, see N l l N F I , candidate Ras effector, 71-72 NFKB activation, 84 interaction with ara-C, 214-21.5 NH,-terminal fragment, p l 2 0 GAP, 70-71 NIH 37'3 cells, DNA-methyltransferase activity increase, 172-1 73 Nucleotide excision repair, 41-42 Nucleotide sequence, variations in microsatellites, 43-44

0 Oligonucleotides, antisense, H-myh, 127-1 28 Oncogenes activation, 4-6

Index interactions with ara-C, 21 3-220 mutated, multiplicity in tumors, 30-32 Oncogenet ics DNA repair genes, 8-9 genes influencing apoptotic cell death, 9-1 1 oncogenes, 4-6 tumor suppressor genes, 6-8 Overexpression H-njyh, 129-130 NFI, 71 Oxygen free radicals, induction of DNA lesions, 37-38

P PI5 in leukemia and hematopoietic malignancies, 163-164 role in solid tumors, 1.58-163 PI6 hypermethylared alleles, I 8 7 in leukemia and hematopoietic malignancies, 163-164 role i n solid tumors, 158-1 63 p2 I , interaction with ara-C, 2 16-2 17 p34LL'c2, interaction with ara-C, 2 1 7 p38, activation, 65-66 pS3

hinding site on HICI, 166 cell cycle block, 131-132 loss, effect on apoptotic threshold, 10-1 1 mutants, transforming effect, 8-9 mutated, iii Burkitt's lymphoma, I 7 mutations in cell subpopulations, 188 mutation spectrum, 40-4 1 role in cell cycle checkpoint control, 6-8 p70'hK, effector of Rac and CDC42, 88 p85, complex with Racl and CDC42, 88-89 p107 complex with EZF, 119-120 mediated proliferation block, 13 1-132 p130, complex with E2F, 119-121 pl60""' K, RhoA-binding protein, 94 PAK, see p2 1 (CDC42/Kac)-activated kinasc p2 1 (CI)C42/Rac)-activated kinase, activation of INK, 87-88 PCR amplification, microsateltires, 49 p I20 GAP candidate Ras effector, 70-71 interactions with Ras regions, 77

Index

Phenotype mutator, cancer cells exhibiting, 2.5-5 1 neoplastic, 2-3 RER-positive, 13 Phosphorylation, ara-C, 203-204 P13K candidate Ras effector, 73-74 coprecipitation with Rac I and CDC42, 88-89 Platelet-derived growth factor, activation of R ~ s 73-74 , PORl protein, effect on Racl activity, 8 9 pRb, interaction ara-C, 2 15 transcription factors, I10 p190 Rho-CAP, interaction with Ras, 9596 PRK 1, see Protein kinase N PRK2, candidate effector of Rho, 9 1-93 Prognosis, mutator phenotype role, 49-50 I’rokaryotes, DNA methylation, 144-145 Proliferative cell nuclear antigen, in DNA synthesis fidelity, 4 3 Promoter regions aging-associated CpC island methylation, 178-1 8 I mutational changes, 18 1-1 8 2 Prostatic cancer, AR gene mutations, 16 Protection, CpG islands, 182-185 Protein kinase C8, interaction with ara-C, 218 Protein kinase C, role ara-C-induced apoptosis, 224 ara-C phosphorylation, 208 Protein kinase Cc, candidate Ras effector, 75 Protein kinase N, candidate Rho effector, 9 1-93 Proteins, 14-3-3, 64

R Rac, candidate effectors, 86-9 1 Ra f key effector of Ras, 63-65 pathways independent of, role in Ras function, 65-69 Rafl, activation, 63-64 Raf+MEK+MAPK kinase cascade activation by Ras, 60-63 role in drug discovery, 97

24 I Ral-binding protein, linking Ras to Rho family proteins, 96 RalCDS, candidate Ras effector, 72-73 Ral guanine nucleotide dissociation stimulator, see RalGDS Ras activation GTPase cascade, role of Rho family proteins, 78-85 Raf-independent pathways, 65-69 Raf+MEK+MAPK kinase cascade, 60-63 candidate effectors, 70-75 interactions with effectors, structural requirements, 75-77 linear signaling pathway, 58-60 link to Rho family proteins, 95-96 Raf as key effector, 63-65 signal transduction, 96-97 m s , interaction with ara-C, 2 19-220 Kas-G7’l’ase-activating proteins, see NFI ; p l 2 0 CAP Ras-GTI’ recognition element, 75-76 RR, role in cell cycle checkpoint control, 6-7 Reactive oxygen intermediates, ara-C-generated, 207-208 Receptor tyrosine kinase, role in Ras activation, 60-62 Regulatory proteins, negative, Cdk inhibitors 35, 130- 132 Replication errors, multiple, 12-13 Resistance, neoplastic cells to ara-C, 2 10-2 1 3 Retinohlastoma, age-incidence curve, I7 Retinohlastoma gene CpC island hypermethylation, I54 loss, 7-8 RGLURIf, candidate Ras effector, 72-73 Rheb, interaction with Raf. 77 Rho, candidate cffectors, 91-94 RhoA binding proteins, 94 coexpression with oncogenic Ras, 79 Rho family proteins interaction with multiple effectors, 85-95 regulation of cellular processes, 80-85 role in Ras transformation, 5 8 , 78-80 Rho-GEF, linking Ras to Rho family proteins, 95 Rhophilin, candidate effector of Rho, 9 1-93 Rhotekin, candidate effector of Rho, 9 3

242 Rin 1 gene, role in growth regulation, 74-7.5 Risk factors, tumor development, role in CpC island hypermethylation, 18 I ROKa, Rho-binding element, 93-94 R-Kas, constitutively activated, 68-69

S SAPK, see Jun NH,-terminal kinase Selective advantage p l 6 gene loss in tumor progression, 159, 161 provided by series of mutations, 2, 6 Senescence, related geriomic methylation levels, 169 Serum response factor, activation, 84 Signaling pathways, ara-C role, 213-220 Signal transduction modulation, ara-C role, 223-225 Ras functions, 58-60, 96-97 Somatic cells, B-myh transcription, I 15-1 17 Somatic selection, coupled with enhanced mutagenesis, 35-36 Sp- I , CpC island regions, 183-1 84 S phase, B-Myb hyperphosphorylation in, 123-125 Src homology 2 domain, Grh2 adaptor protein, 6 I Stress-activated protein kinase, interaction with ara-C, 218 Stress fibers, actin, formation, 82-83 Switch 11, mutagenesis, 76

T Target genes, B-Myb, 133-1 34 Target mutator genes, candidate, 38-43 TC21, constitutively activated, 68-69, 77 Temperature, permissive, for target cell transformation, 4 Time, role in mutation accumulation, 46-47 Tissues, C p C island hiallelic expression, 180 Tolerance, DNA-methyltransferase gene, 173 Transcription, B-myh as cell cycle control point, 122-123 regulatory mechanisms, 118-122 in somatic cells, 115-1 17 Transcriptional control, by B-Myb, 132-1 33 Transcriptional status, genes with and witho u t CpC islands, 148-150 Transformation increased DNA-methyltransferase activity role, 173-1 75

index

Kas Raf-independent pathway role, 65-69 Rho family proteins role, 78-80 Transforming growth factor p, mediated growth inhibition, I63 Transgenic mouse hcl-2, 10 carrying SV40 LT gene, 18-1 9 Transport, ara-C, 202 reduced, 2 I0 Tumor progression altered DNA methylation role, 186-1 89 and CpG island hypermethylation, 177- I 86 Foulds’ rules, 2-3 increased DNA-methyltransferase activity, 170-1 76 mutations during, 47-48 selective advantage of p16 gene loss, 159, I61 Tumors childhood, 16-1 7 colorectal, sequence of events, 15 solid, Cdk inhibitor gene role, 158-163 sporadic, microsatellite instability, 45-46 Tumor suppressor genes CpG island anatomy, 182-1 84 HlCl gene as candidate, 166 inactivation, 157-159, 176 role in cell cycle checkpoint control, 6-8

U Uracils, repaired, conversion to thymidines, 176

v von Hippel-Lindau gene, CpG island hypermethylation, 1.57-158

W WAFI/CIPI, interaction with ara-C, 2 1 6-2 17 WASP gene, CRIB motif, 89-90 Wiskott-Aldrich syndrome, mutated WASP gene, 89-90

X Xeroderma pigmentosum, l2-13,42

Y Yeast, MAPK cascade activation, 62-63