Advances in Cancer Research Volume 37

ADVANCES IN CANCER RESEARCH VOLUME 37

Contributors to This Volume Kari Alitalo

Jan Klein

J. Michael Bishop

Gilbert Lenoir

Carmia Borek

Arnold J. Levine

Elisabeth Gateff

Zoltan A. Nagy

George Klein

Kenneth Nilsson

Antti Vaheri

ADVANCES IN CANCERRESEARCH Edited by

GEORGE KLEIN Department of Tumor Biology Karolinska lnstitutet Stockholm, Sweden

SIDNEY WEINHOUSE Fels Research Institute Temple University School of Medicine Philadelphia, Pennsylvania

Volume 37-1982

ACADEMIC PRESS A Subsidiary of Harcourt Brace Jovanovich, Publishers

New York London Paris San Diego San Francisco SBo Paulo Sydney Tokyo Toronto

COPYRIGHT @ 1982, B Y ACADEMIC PRESS,INC. ALL RIGHTS RESERVED. NO PART OF THIS PUBLICATION MAY BE REPRODUCED OR TRANSMITTED IN ANY FORM OR BY ANY MEANS, ELECTRONIC OR MECHANICAL, INCLUDlNG PHOTOCOPY, RECORDING, OR ANY INFORMATION STORAGE AND RETRIEVAL SYSTEM, WITHOUT PERMISSION IN WRITING FROM THE PUBLISHER.

ACADEMIC PRESS, INC.

111 Fifth Avenue, New York, N e w

York 10003

United Kitigdoni Edition published by ACADEMIC PRESS, INC. (LONDON) LTD. 24/28 Oval Road, London NWl 7DX

LIBRARY OF CONGRESS CATALOG CARDNUMBER:52-1 3360 ISBN 0-12-006637-8 PRINTED IN THE UNlTED STATES OF AMERICA 82 83 84 85

9 8 1 6 5 4 3 2 1

CONTENTS CONTRIBUTORS TO VOLUME 37

. . . . . . . . .

. . .

.

.

ix

Retroviruses and Cancer Genes J . MICHAEL BISHOP I. I1. 111. IV. V. VI . VII . VIII . IX . X. XI . XI1. XI11. XIv.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Introducing Retrovirus Oncogenes . . . . . . . . . . . . . . . . . . . . Properties of Retrovirus Oncogenes and Their Products . . . . . . . . . . . The Origin of Retrovirus Oncogenes: Emergence of the Thesis. . . . . . . . The Discovery of c-oncs . . . . . . . . . . . . . . . . . . . . . . . . Characterizing c-oncs . . . . . . . . . . . . . . . . . . . . . . . . . c-oncs Are Cellular Genes . . . . . . . . . . . . . . . . . . . . . . . The Expression of c-oncs . . . . . . . . . . . . . . . . . . . . . . . . Proteins Encoded by C-ORCS . . . . . . . . . . . . . . . . . . . . . . . How Similar Are Viral Oncogenes and c-oncs? . . . . . . . . . . . . . . . Are c-oncs Members of a Multigene Family? . . . . . . . . . . . . . . . How Might Retroviruses Transduce Cellular Genes? . . . . . . . . . . . . What Is the Function of c-oncs in Normal Cells?. . . . . . . . . . . . . . The Paradox of Neoplastic Transformation by Retrovirus Oncogenes . . . . . xv. Does the Homology between Viral Oncogenes and c-oncs Dictate the Host Range of Viral Transformation? . . . . . . . . . . . . . . . . . . . . . . XVI . Do c-oncs Provide a Pathway for Oncogenesis? . . . . . . . . . . . . . . XVII . Conclusion: The Pursuit of Cancer Genes . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1 2 6 7 8 10 12 12 13 14 16 17 20 21 22 23 26 28

Cancer. Genes. and Development: The Drosophila Case ELISABETH GATEFF I . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I1. Drosophila Development . . . . . . . . . . . . . . . . . . . . . . . .

111. General Information on the Drosophilu Tumor Mutants . IV. Description of the Drosophila Tumor Mutants . . . . . V. Viruses Found in Drosophila Tumor Cells . . . . . . V I . Retroviral Oncogenes and Their Cellular Counterparts in Species . . . . . . . . . . . . . . . . . . . . . . VII . Transfection and Vertebrate Tumor Genes: A Comparison VIII . Concluding Remarks . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . V

33 34 . . . . . . . . . . 35 . . . . . . . . . . 37 . . . . . . . . . . 57 Different Animal 62 . . . . . . . . . with Drosophilu . . 66 69 . . . . . . . . . 69 . . . . . . . . .

vi

CONTENTS

Transformation-AssociatedTumor Antigens ARNOLD J . LEVINE I. I1. 111. IV. V VI . VII . VIII .

.

IX.

Introduction . . . . . . . Simian Virus 40 . . . . . . Adenoviruses . . . . . . . Epstein-Barr Virus . . . . Methylcholanthrene-Induced Abelson Virus . . . . . . . Rous Sarcoma Virus . . . . Teratocarcinonias . . . . . Conclusions . . . . . . . . References . . . . . . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . . Transformed Cells (Meth A) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . 75 . . . . 81 . . . . 86 . . . . 88 . . . . . 91 . . . . 93 . . . . 96 . . . . 97 . . . . 100 . . . . 104

Pericellular Matrix in Malignant Transformation KARIALITALO AND ANTTI VAHERI I . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111 I1. Components of the Extracellular Matrix and Their Functions . . . . . . . . 112 111. Pericellular Matrix and the Cellular Phenotype in Vitro . . . . . . . . . . . 123 Iv.

V.

VI . VII . VIII .

Malignant Transformation: Altered Biosynthesis of Matrix Components and Failure to Deposit Them . . . . . . . . . . . . . . . . . . . . . . . . Tumorigenicity. Invasion. and Metastasis . . . . . . . . . . . . . . . . . Proteins of Basement Membranes and Interstitial Matrix as Characteristics of Tumor Cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cell-Matrix Interaction and Anchorage Dependence of Normal Cells. . . . . Molecular Mechanisms of Altered Cell-Matrix Interaction in Rous Sarcoma Virus Transformation . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

128 132 138 141 143 146

Radiation Oncogenesis in Cell Culture CARMIA BOREK I. I1. I11. IV.

V

Foreword . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cell Transformation in Vitro . . . . . . . . . . . . . . . . . . . . . . Radiation Oncogenesis in Vitro . . . . . . . . . . . . . . . . . . . . . Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

159 160 161 177 220 227

CONTENTS

vii

Mhc Restriction and Ir Genes JAN KLEINAND ZOLTAN A . NAGY I. I1. I11. IV. V VI . VII . VIII . IX.

Mhc through a Keyhole . . . . . . . . . . . . . . . . . . . . . . . . How Was Mhc Restriction Discovered? . . . . . . . . . . . . . . . . . . Mhc Restriction of Cytolytic Responses . . . . . . . . . . . . . . . . . . Mhc Restriction of Regulatory Responses . . . . . . . . . . . . . . . . . Mhc Restriction of DTH and CS Responses . . . . . . . . . . . . . . . . The Puzzle of the Class I and Class I1 Gene Dichotomy . . . . . . . . . . . Is the T-cell Repertoire Individualized? . . . . . . . . . . . . . . . . . . Nature of Nonresponsiveness and the So-called Ir Genes . . . . . . . . . . The Parable of the Blind . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

234 238 242 245 262 265 270 285 309 310

Phenotypic and Cytogenetic Characteristics of Human B-Lymphoid Cell Lines and Their Relevance for the Etiology of Burkitt’s Lymphoma KENNETHNILSSON AND GEORGE KLEIN I . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I1. Lymphohlastoid Cell Lines (LCLs). . . . . . . . . . . . . . . . . . . . 111. EBV-Carrying Burkitt’s Lymphoma (BL) Cell Lines . . . . . . . . . . . . IV Basis for Distinction between EBV-Carrying Lymphoblastoid and BL Cell Lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V. EBV Genome-Negative BL Cell Lines . . . . . . . . . . . . . . . . . . VI . EBV Genome-Negative B-LeukemidLymphoma Cell Lines . . . . . . . . . VII . EBV-Carrying Non.BL. Nonlymphoblastoid Cell Lines Derived from EBV Genome-Negative LeukemidLymphomas . . . . . . . . . . . . . . . . . VIII . The Relationship of EBV-Carrying Lymphoid Cell Lines to Normal B-Cell Differentiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . IX. The Progression in Lymphoblastoid Cell Lines in Vitro and in Vim . . . . . X . The Role of EBV in Progression . . . . . . . . . . . . . . . . . . . . . XI . The Role of Chromosomal Changes in Progression . . . . . . . . . . . . . XI1. General Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

319 321 338 346 346 349 352 353 360 362 364 368 371

Translocations Involving Ig Locus-Carrying Chromosomes: A Model for Genetic Transposition in Carcinogenesis GEORGEKLEINAND GILBERT LENOIR Text . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

381 386

INDEX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . CONTENTS OF PREVIOUS VOLUMES. . . . . . . . . . . . . . . . . . . . . . .

389 393

This Pagepage Intentionally Left Blank This intentionally

left blank

CONTRIBUTORS TO VOLUME 37 Numbers in parentheses indicate the pages on which the authors' contributions begin.

KARIALITALO,' Department of Virology, University of Helsinki, 00290 Helsinki 29, Finland (111) J. MICHAEL BISHOP,Department of Microbiology and Immunology, University of California, Sun Francisco, California 94143 (1) CARMIA BOREK,Radiological Research Laboratory, Department of Radiology, Cancer Centerllnstitute of Cancer Research, and Department of Pathology, College of Physicians and Surgeons, Columbia University, New York, New York 10032 (159) ELISABETH GATEFF,Biologisches lnstitut 1 (Zoologie), Albert-Ludwigs Universitat, 7800 Freiburg i. Br., Federal Republic of Germany (33) GEORGE KLEIN,Department of Tumor Biology, Karolinska lnstitutet, S-104 01 Stockholm, Sweden (319, 381) JAN KLEIN, Max-Planck-Znstitut f u r Biologie, Abteilung lmmungenetik, 7400 Tubingen 1, Federal Republic of Germany (233) GILBERT LENOIR,International Agency f o r Research on Cancer, F-69372 Lyon, Cedex 2, France (381) ARNOLD J. LEVINE,Department of Microbiology, School of Medicine, State University of New York, Stony Brook, New York 11794 (75) ZOLTAN A. NAGY, Max-Planck-lnstitut f u r Biologie, Abteilung lmmungenetik, 7400 Tubingen 1, Federal Republic of Germany (233) KENNETH NILSSON,The Tumor Biology Group, The Wallenberg Laboratory, University of Uppsala, S-751 22 Uppsala, Sweden (319) ANTTIVAHERI,Department of Virology, University of Helsinki, 00290 Helsinki 29, Finland (111)

'Present address: Department of Microbiology and Immunology, University of California, San Francisco, California 94143.

ix

This Pagepage Intentionally Left Blank This intentionally left blank

RETROVIRUSES AND CANCER GENES J. Michael Bishop Department of Microbiology and Immunology. University of California. San Francisco, California

I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1

11. Introducing Retrovirus Oncogenes. . .............. 111. Properties of Retrovirus Oncogenes a ..................... IV. The Origin of Retrovirus Oncogenes: Emergence of the Thesis.. . . . . . . . . . . . . . .

2 6 7 8 10 12

The Discovery of c-oncs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Characterizing c-oncs . . . . . . . . . . . . . . . .................. c-oncs Are Cellular Genes ............................................... The Expression of c-oncs . . . . . . Proteins Encoded by c-oncs. .................................... How Similar Are Viral 0 Are c-oncs Members of a Multigene Family? . . . . . . . . . . . . How Might Retroviruses What Is the Function of c-oncs in Normal Cells? . The Paradox of Neoplast Does the Homology between Viral Oncogenes and c-oncs Dictate the Host Range of Viral Transformation? ................................. ............ XVI. Do c-oncs Provide a Pathway for Oncogenesis? . XVII. Conclusion: The Pursuit of Cancer Gen References ......................... .........

V. V1. VII. VIII. IX. X. XI. XII. XIII. XIV. XV.

22 23 28

I. Introduction

The possibility that viruses might be tumorigenic materialized at the turn of this century, with reports that erythroleukemia and fibrosarcomas could be induced in chickens by transmissible agents (Ellermann and Bang, 1908; Rous, 191 1). Skepticism greeted these reports, dissipated only slowly over the ensuing years, and lingers on in the continuing debate about the possible role of viruses in the genesis of human tumors. But these are needless concerns for most experimentalists : the oncogenic potential of many animal viruses is well established, and the use of tumor viruses now dominates efforts to dissect the mechanisms of tumorigenesis. Oncogenic viruses are taxonomically diverse. DNA viruses both large (herpes and adenoviruses) and small (papovaviruses), as well as many (but nor all) members of the large family of retroviruses, can induce tumors in either experimental or natural hosts (Gross, 1970). Two patterns of viral oncogenesis have emerged. Some viruses possess genetic loci (or "oncogenes") whose actions initiate and maintain the neoplastic phenotype of the infected cell. Other viruses are devoid of specific oncogenes and induce tumors by 1 ADVANCES IN CANCER RESEARCH, VOL. 37

Copyright 0 1982 by Academic Press, Inc. All rights of reproduction in any form reserved. ISBN 0-12-006637-8

2

J . MICHAEL BISHOP

more subtle means, whose particulars we are just beginning to perceive. But both forms of viral oncogenesis are united by the persistence of at least a portion of the viral genome in the host cell, either as an integral part of a host chromosome or as independently replicating units. For the moment, it appears that persistence of the viral genome is a necessary event for viral oncogenesis-to maintain the influence of an oncogene over the host cell, or to sustain the more indirect but equally malicious effects of viruses that induce tumors without benefit of an oncogene. The possibility that transient infection by viruses might trigger an irreversible sequence of events (the “hit-and-run” mechanism) has been posited repeatedly but has gained no substantive experimental support to date. Although several themes appear to unite oncogenesis by diverse viral agents, and although each family of tumor viruses offers important distinctive features to the experimentalist, it is the retroviruses that have provided the most coherent and penetrating view of tumorigenesis presently available to us. Three features of retroviruses account for this sentiment. First, the oncogenes of retroviruses have proved exceptionally accessible to definition and study, and as a consequence, they have provided our first glimpse of enzymatic mechanisms responsible for neoplastic transformation. Second, the diversity of retrovirus oncogenes has provided a rich set of oncogenic agents whose versatility far exceeds that of DNA tumor virus oncogenes, and whose tumorigenic capacities provide separate experimental models for most major forms of malignancy. Third, oncogenes appear not to be indigenous components of retrovirus genomes, but instead have been transduced from normal genetic loci of the vertebrate hosts in which retroviruses replicate. Moreover, we have reasons to believe that the vertebrate genes from which retrovirus oncogenes derive may participate in tumorigenesis induced by agents other than viruses. Thus while tracking the evolutionary origins of oncogenes, retrovirologists have been led well beyond the confines of tumor virology, to confront what may be a final common pathway of oncogenesis. The genesis and status of this confrontation are the main subjects of this essay. II. Introducing Retrovirus Oncogenes

Retrovirus oncogenes display a burgeoning diversity that is largely attributable to nature’s generosity in providing field isolates, but recent years have also witnessed provisional successes in bringing new oncogenes to view by experimental manipulations (Rapp and Todaro, 1978, 1980; Young ef a/., 1981). At least 15 retrovirus oncogenes have been identified (Table 1). Together, they are known by the generic term v-onc (for viral oncogene). Each gene is distinguished by its nucleotide sequence and is designated by a term derived from the name of the virus that bears the gene (hence, v-SYC,

3

RETROVIRUSES AND CANCER GENES

v-my, v-abl, etc.; see the table). Many of these genes are represented in more than one viral isolate, and the topographical details of kindred genes may vary from one isolate to another; a few of the genes have been obtained from more than one host species; and a number share similar oncogenic properties and/or biochemical functions (see later). The replicative unit in all retrovirus genomes is composed of three genes : gag (structural proteins of the virion); pol (reverse transcriptase); and e m (glycoproteins of the viral envelope). An oncogene may be inserted into this unit in at least four distinctive ways (Fig. 1): (1) as an independently expressed

RSV

AMV

MCV

AEV

I

0

I

1

I

2

I

3

I

4

I

5

I

6

I

7

I

8

1

9

KILOBASES

FIG.1. Styles in retroviruses genomes. Genomes of avian retroviruses are used to illustrate typical dispositions of oncogenes. ALV, avian leukosis virus: contains no oncogene; gag encodes structural proteins of the viral core; pol, reverse transcriptase; and e m , the glycoproteins of the viral envelope. RSV, Rous sarcoma virus: the oncogene src is expressed from a spliced subgenomic mRNA. AMV, avian myeloblastosis virus: the oncogene myb is expressed from a spliced subgenomic mRNA. MCV, MC29 virus: the oncogene rnyc is expressed from a genomiclength mRNA that directs uninterrupted translation from gag and myc. AEV, avian erythroblastosis virus: the oncogene erb-A is expressed from a genomic-length mRNA that directs the uninterrupted translation from gag and erb-A; the oncogene erb-B is expressed from a spliced subgenomic mRNA. Stippling denotes regions that do not encode protein; solid vertical lines denote established gene boundaries; and diagonal shading denotes uncertainty in gene boundaries.

J

10

TABLE I RETROVIRUS ONCOGENFS AND THEIRPROGENITORS ~

v-onP

Protypic virus

v-src v-jps v-yes v-rash

Rous sarcoma virus

v-myc v-erb-A v-erb-B v-myb v-re1 v-mos v-ubl v-bus v-rus V+S

v-fms v-sis

Fujinami sarcoma virus Y73 sarcoma virus University of Rochester sarcoma virus 2 Myelocytomatosis virus Avian erythroblastosis virus Avian erythroblastosis virus Avian myeloblastosis virus Reticuloendotheliosis virus Moloney sarcoma virus Abelson leukemia virus BALB sarcoma virus Harvey sarcoma virus Gardner-Arnstein feline sarcoma virus McDonough feline sarcoma virus Simian sarcoma virus

~~~

Probable species of originb

~~~~

Tyrosine phosphorylation'

+ + +

c-one

c-onr expressedd

c-src c-jp9 c-yes c-ros"

Yes ?

Chicken Chicken Chicken Chicken Turkey Mouse Mouse Mouse Rat Cat

c-myc c-erb-A' c-erb-B' c-myb c-re1 c-mos c-ubl c-bus' c-rus" c-fes

Yes Yes Yes Yes Yes

Cat

c: fms

?

Woolly monkey

c-sis

'I

Chicken/quaill Chicken Chicken Chicken

+

? 'I

J

Yes ? Yes Yes

Protein product of c-once pp60'-'" ? > 9

~

The names of viral genes are here treated as generic terms, although in most instances several separate viral isolates are known. The probable species of origin is inferred from the host in which the particular oncovirus first emerged. A plus sign denotes the existence of evidence (not necessarily definitive) for tyrosine phosphorylation by the oncogene product. Expression is defined as either detection of transcription from the c-onc or detection of a protein encoded by the locus. Question marks indicate that suitable analyses have not been completed. Some strains of RSV have been generated experimentally in quail (Wang er al., 1979), but all field isolates of v-src have come from chickens. The c-fps of chicken is apparently related to c-fes of cats (Shibuya et al., 1980). The term v-ros is a provisional designation for a newly identified oncogene in an avian sarcoma virus isolated at the University of Rochester (P. Balduzzi, personal communication). The separate domains of erb (A and B) are represented by similarly separate-but possibly linked-domains in the chicken genome. j All efforts to date have failed to detect transcription of c-mos (Frankel and Fischinger, 1976). Ir Failure to detect transcription from c-mos and features of the gene’s nucleotide sequence (Van Beveren er al., 1981) raise the possibility that c-mos is inactive unless transferred into a retrovirus genome. The c-bas of mice is apparently closely related to the c-ras of rats (S. Aaronson and E. Scolnick, personal communication). Rat DNA contains at least two small, distinct gene families related to v-ras. One family is apparently the source of Harvey and Rasheed v-rus, the other of Kirsten v-ras (DeFeo er al., 1981). a

’

vI

6

J . MICHAEL BISHOP

gene that does not impose on either the structure or function of the replicative genes and is expressed from a subgenomic mRNA (v-src of the Rous sarcoma viruses is the sole known example); (2) as an independently expressed gene that replaces part or all of a replicative gene (enu, for example) and is expressed from a subgenomic mRNA (e.g., v-myb of avian myeloblastosis virus); (3) as a fusion between v-onc and a portion of gag that is accompanied by deletions in one or more of the replicative genes (usually pol and portions of gag and env) and is expressed as a polyprotein produced from a genomic-length mRNA (e.g., v-myc of avian myelocytomatosis virus) ; and (4) as two separately expressed v-onc domains-one fused with a portion of gag, the other expressed independently, and the two together replacing portions of replicative genes. In this last instance, the yag-onc protein is produced from a genomic-length mRNA, the second onc protein from a subgenomic mRNA (e.g., erb-A and erb-B of avian erythroblastosis virus). These themes are varied in yet a further way: the same class of v-onc can be fused to gag in some viral isolates and occur as an independently expressed locus in others. With the exception of v-src, the insertion of oncogenes into retrovirus genomes creates genetic defects that preclude the production of virus unless the defective function(s) is provided by a second “helper” virus. The variety of retrovirus oncogene construction is merely tedious at first glance, but in reality it poses an important challenge because all of these configurations are thought to be products of recombination between a replicating retrovirus and the genome of its host cell (see later). The mechanisms that effect this recombination with its remarkable plurality of outcomes may have no precedent in the annals of molecular genetics. Ill. Properties of Retrovirus Oncogenes and Their Products

The oncogenes of retroviruses are genetic luxuries whose actions are highly selective. They are not required for viral replication-indeed they make no known contribution to replication; and their activities may be muted even in cells that sustain vigorous replication, and perhaps chemical expression, of the oncogenes (Graf et al., 1980; Durban and Boettiger, 1981a). This muting underlies one of the cardinal properties of retrovirus oncogenes : the specificity of their pathogenicity. Each oncogene induces tumors in only a limited and characteristic set of tissues; transformation of cells in culture follows the same selective pattern. We cannot at present explain the selectivity of oncogene actions, but the phenomenon has contributed to the view that transformation by retrovirus oncogenes is fundamentally a disturbance of differentiation. According to one prevalent view, oncogenes may act by arresting cellular development within a specific compartment of one or another developmental lineage; tumorigenesis ensues because the immature cells that constitute the compartment continue to divide, as is their nature,

RETROVIRUSES AND CANCER GENES

7

and become a continuously expanding population-a tumor composed of ostensibly normal cells (Graf and Beug, 1978). Other observers have argued that the effects of oncogenes more commonly (or perhaps inevitably) distort the phenotype of susceptible cells to a form that is not representative of a single developmental compartment (Boettiger and Durban, 1979; Durban and Boettiger, 1981b). No matter which view is correct (there appear to be elements of truth in both), the effects of oncogenes on cells, and the selectivity of these effects, point to biochemical functions that in other guises might well direct the course of normal growth and development. Inferences of this sort brook large in present efforts to interpret the evolutionary origins of retrovirus oncogenes (see later). How do oncogenes evoke the myriad changes that accompany neoplastic transformation? Efforts to answer this daunting question have produced at least dim outlines of several potentially important refrains. 1. Some retrovirus-transforming proteins (but not others; see Table I) apparently mediate the phosphorylation of tyrosine in protein substrates (Hunter and Sefton, 1980b). The best-studied example is the 60,000-MW phosphoprotein encoded by v-src (pp60'-"'), but the products of v-fps, v-yes, v-abl, and v-fes also appear to follow suit. Phosphorylation of proteins represents one of the principal devices by which cellular functions are regulated (Rubin and Rosen, 1975) and thus offers an attractive explanation for the pleiotropic effects of retrovirus oncogenes. Moreover, the findings with pp6OV-"*'and other retrovirus-transforming genes have produced unexpected further dividends by alerting cell biologists to the existence of tyrosine phosphorylation. This hitherto unknown reaction is now rapidly emerging as an important regulatory mechanism in normal as well as transformed cells. But efforts to identify cellular proteins that serve as substrates for retrovirus protein kinases have just begun, only a small number of candidate substrates have been sighted, and none of these as yet offers any explanation for the abnormal growth of transformed cells. 2. Several retrovirus-transforming proteins (including pp6OV-"*',pp2 1v-ras, and the product of v-abl) appear to act at the periphery of the cell because they are found attached to the plasma membrane of the transformed cell (Hynes, 1980). This inference is for the moment largely circumstantial, however. The techniques used to locate the transforming proteins have limited resolving power and sensitivity; trace amounts of the proteins in presently unappreciated locations might cause major effects. IV. The Origin of Retrovirus Oncogenes: Emergence of the Thesis

The virus isolated from a chicken sarcoma by Peyton Rous did not spring quickly or easily into view. Rather, an infectious tumorigenic agent was obtained from extracts of tumor tissue only after the original sarcoma had

8

J. MICHAEL BISHOP

been passaged repeatedly from one bird to another (Rous, 1911). It seems possible in retrospect that the original tumor was not the consequence of viral infection; the sarcoma virus that eventually emerged may not have been present in the tissue with which Rous began his work. The isolation of murine sarcoma viruses (Harvey, 1964; Moloney, 1966) and the Abelson murine leukemia virus (Abelson and Rabstein, 1970a,b) decades later raised these issues in a more explicit manner: the new viruses appeared during the passage of leukemia viruses in rodents, as if new capabilities for pathogenesis could be acquired from the host animal. The discovery of endogenous retroviruses in chickens (Robinson, 1978) and mice (Aaronson and Stephenson, 1976), as well as the development of inbred lines of mice whose predisposition to leukemia appeared to involve genetically transmitted retroviruses (Rowe, 1973), added appreciably to these inferences and engendered the “oncogene hypothesis” of Huebner and Todaro (1969). According to this hypothesis, carcinogens of many sorts act by inducing the expression of otherwise cyrptic retrovirus genes already resident in the genome of the target cells. The oncogene hypothesis is no longer regarded as strictly correct, but it served an important heuristic purpose by prompting experimentalists to ask whether normal cellular DNA might contain retrovirus oncogenes. We now know that vertebrate cells do harbor genetic loci homologous to retrovirus oncogenes (designated here by the generic term “c-onc”), but these loci are cellular, not viral genes, and the oncogene hypothesis has been eclipsed by even more sweeping views of the nature of these cellular genes.

V. The Discovery of c-oncs

The search for oncogenes in cellular DNA began with the use of molecular hybridization. The strategy exploited naturally occurring deletions that remove most or all of v-src (but no other viral gene) from the genome of Rous sarcoma virus (Duesberg and Vogt, 1970, 1971 ; Martin and Duesberg, 1972; Lai et al., 1973). Viral RNA bearing this class of deletions could be employed to isolate radioactive DNA (cDNA,,,) that hybridized only with nucleotide sequences encoding (or related to) src (Stehelin et a/., 1976a). The result was a reagent that provided specificity and sensitivity sufficient to detect a single genetic locus among the immense complexity of vertebrate DNA. Similar cDNAs were prepared for replication-defective murine sarcoma viruses (Scolnick et a / . , 1973, 1975; Frankel et a/., 1976), but the genetic definition of these reagents was less rigorous because suitable deletion mutants were not available for isolation of the cDNAs. As a consequence, the experimental strategies had to rely on the assumption that nucleotide sequences not present in the genome of the helper virus must perforce

RETROVIRUSES AND CANCER GENES

9

represent portions of the oncogene-an assumption that proved useful but not inevitably correct (see later). The initial findings with cDNA,,, for Rous sarcoma virus prefigured subsequent conclusions for virtually all-retrovirus oncogenes. Each family of vertebrates examined-including fish, birds, and mammals-displayed evidence of both DNA and RNA related to the src gene (Stehelin et al., 1976a; Spector et al., 1978a,b,c). The DNA related to v-src appeared to occur as only one or very few copies in each haploid portion of vertebrate genomes. Retrovirologists had obtained their first glimpse of the cellular gene we now know as c-src. The mere fact that homologous DNA could be detected across such large phylogenetic distances indicated that the genetic locus or loci in question were highly conserved during the course of evolution. More recent findings have dramatized the extent of this conservation by demonstrating homology with v-src (and several other v-ones) in the DNA of the insect Drosophila (personal communication, R. Weinberg). Conservation of c-src was also explored by evaluating the thermal stability of molecular hybrids formed between cDNA,,, and DNAs from various sources. The results indicated that the nucleotide sequences of c-src might diverge by no more than 10- 15% from fish to chicken genomes on the one hand, from chicken to human genomes on the other (Stehelin et al., 1976b; Spector et al., 1978~).The full implications of these findings were not easily sustained at the outset, largely because no assay was available for the protein product of src; nevertheless, it appeared that vertebrate species possessed a highly conserved and expressed (i.e., transcribed) gene that is closely related to a viral oncogene. The strong evolutionary conservation of this gene, and the fact that it was found to be expressed in every tissue and every species examined, indicated an essential function in cellular metabolism. Early doubts about the veracity of these deductions were obviated by the eventual identification and characterization of a protein encoded by c-src (and known as pp60c~"c).The characteristics of pp6OC-"' vindicate and extend all of the original predictions based on molecular hybridization (see later). Difficulties did arise, however, from the use of a less well-defined cDNA for the oncogene of Harvey/Kirsten murine sarcoma virus (v-ras). Initial results indicated that v-ras was related to (and presumably derived from) nucleotide sequences in the genome of an endogenous retrovirus of rats (Scolnick et al., 1973; Scolnick and Parks, 1074), a troubling deduction because it stood in striking contrast to the mounting evidence that other retrovirus oncogenes are derived from conserved cellular genes. The advent of molecular cloning to the study of retrovirus genomes quickly resolved the apparent anomaly. It now appears that the genome of Harvey/Kirsten sarcoma virus was constructed with three distinct components (Ellis et al., 1980): one derived from the murine helper virus that was used to initiate

10

J . MICHAEL BISHOP

recovery of the sarcoma virus and was isolated together with the sarcoma virus; a second derived in fact from an endogenous virus of rats; and a third, the oncogene proper, derived true to form from a cellular gene of the rat in which the sarcoma virus originally arose. The principles first enunciated for src have since been shown to be widely applicable to retrovirus oncogenes (see table) : homologues of these genes (i.e,, c-oncs) can be found in vertebrate DNA, many (but apparently not all) of which are expressed in phenotypically normal cells. The sole exception at present is the oncogene of the Spleen Focus Forming Virus, which appears to be a recombinant form of the retrovirus env gene rather than the derivative of a cellular gene (Oliff et al., 1980). All of the identified c-oncs are found in more than one vertebrate species, but the extent of evolutionary conservation varies from one c-onc to another: some are readily detectable only in closely related species; others appear to have taken form in primitive vertebrates (or even earlier in evolution) and to have evolved thereafter in concert with speciation. These variations may be only matters of degree, however ; it is now reasonable to suppose that every c-onc represents a genetic lineage that extends throughout the vertebrate phyla and, in an least some instances, farther down the phylogenetic hierarchy. The kinship between retrovirus oncogenes and cellular genes is certain. But how can we distinguish parent from progeny? Phylogenetic patterns provide a clue: in contrast to the evolutionary conservation of the cellular genes, the viral oncogenes are usually restricted to single strains of retroviruses that were isolated from particular species (although not inevitably ; see later. Moreover, the homology between viral oncogene and cellular DNA is greatest for the species in which the oncogene allegedly originated. The most straightforward interpretation of these findings is that retrovirus oncogenes are derived from cellular genes. Widespread acceptance of this scheme, and the remarkable similarity between retrovirus oncogenes and their cellular homologues (to be described later), have engendered a standard nomenclature that is followed here (Coffin et al., 1981): viral oncogenes are denoted by v, as in v-src or v-myc (see earlier, and the table); the cellular progenitors of v-oncs by c, as in c-src or c-myc (see the table). The nomenclature is only a convenience, however, and should not be construed as indicating that homologous viral and cellular genes are necessarily identical in either structure or function. The precise relationship between cellular progenitor and viral progeny has yet to be fully explored for any retrovirus oncogene. VI. Characterizing c-oms

Enumeration of c-oms by molecular hybridization and by mapping with restriction endonucleases has revealed that some may be unique loci within

RETROVIRUSES AND CANCER GENES

11

a given species. Exceptions exist, however, and may eventually prove to be the rule. (a) The DNA of chickens probably contains a second, possibly incomplete locus (a “pseudo-gene”) related to c-src (Parker et al., 1981). (b) The murine retrovirus oncogene v-ras has been isolated in at least three distinctive forms, represents by the Harvey, Kirsten, and Rasheed strains of sarcoma virus (Gross, 1970; Young et al., 1981). The Harvey v-ras may have originated from either of two genetic loci in rats, one with introns, the other without (DeFeo et al., 1981); the Kirsten v-ras apparently derives from yet another rat gene that is related only distantly to the Harvey forms of c-ras (personal communication, E. Scolnick); and the Rasheed v-ras probably emerged from the Harvey C-MS family (personal communication, E. Scolnick). The origin of the Rasheed v-ras came as a surprise because the relationship of the Rasheed oncogene to the other forms of v-YUSoriginally seemed quite distant and was perceived only by serological means, not by tests with molecular hybridization (Young e l al., 1979). The c-oncs behave as classical Mendelian loci. They occupy constant positions within the genomes of particular species (Hughes et a/., 1979a), and they segregate in a predictable fashion when breedings are analyzed with the assistance of structural polymorphisms that have been identified by restriction mapping (D. Spector and B. Vennstrom, personal communications). The loci are recognized by virtue of homology with a viral oncogene, but in most (if not all) instances, the homologous nucleotide sequences do not comprise the entire cellular gene. Three major considerations prompt this statement. First, heteroduplex analysis and restriction mapping have demonstrated that the homology between several viral oncogenes and their C-oncs is interrupted by one or more intervening sequences (or introns) in the cellular locus (Goff et al., 1980; DeFeo et al., 1981; Favera et al., 1981; Franchini et al., 1981; Parker et al., 1981; Shalloway et al., 1981). The c-onc for v-mos provides an interesting exception to this rule: the murine and human forms of c-mos display uninterrupted homology with v-mos in heteroduplex analysis (Oskarsson et al., 1980; and personal communication, G. Vande Woude). The same is true of at least one of the several rat loci representing c-ras (DeFeo et al., 1981). Second, transcription from a number of c-oncs generates RNAs that, even in their mature forms, are appreciably more complex than the homologous viral oncogene (Bishop et al., 1981). For example, the mature form of the RNA produced from c-src contains 3.9 kb (Parker et al., 1981)-a complexity almost threefold greater than that of v-src. Yet both v-src and c-src give rise to a protein of 60,000 MW (Brugge and Erikson, 1977; Collett et al., 1978; Oppermann et al., 1979). It appears that large portions of the mRNA for c-src may not be translated, and that the boundaries of this (or any other) c-onc can only be located by applying the definitions that demarcate a transcriptional unit in eukaryotic DNA (i.e., the sites of initiation and

12

J . MICHAEL BISHOP

polyadenylation). Third, many retrovirus oncogenes have been formed by fusing a portion of a replicative gene (typically, gag; see earlier) to nucleotide sequences of a c-om. It seems unlikely that this fusion always incorporates the entirety of the c-onc locus into the viral genome. in particular, sequences in the 5' domain of the c-onc may be missing from the viral oncogene. VII. C-oncs Are Cellular Genes

The oncogene hypothesis portrayed cellular oncogenes as components of retrovirus genomes-a conceptual predisposition that proved difficult to override. But we are now certain that c-ones are cellular genes, not viral genes in disguise. The conclusion rests on three major points of evidence: (1) the location of c-ones at constant genetic loci in every member of a species (in striking contrast to the diverse distribution and positioning of endogenous retrovirus genes; for example, see Hughes et at., 1979a,c); (2) the presence of intervening sequences within many of the c-oncs-a hallmark of eukaryotic genes and, again, a telling contrast with the organization of retrovirus genes; and (3) the fact that no c-one has been found within or even linked to a complete or defective provirus of an endogenous retrovirus (Hughes et d., 1979a,c; Sheiness r t al., 1980). It seems unlikely that C-OIICS were introduced into vertebrate genomes by infection of ancestral species with retroviruses ; instead, we must explain how complex cellular genes have made their way in part or entirely into the genomes of preexistent retroviruses. VIII. The Expression of c-oncs

The possibility that c-ones might be expressed in phenotypically normal cells emerged from the discovery of RNA homologous to v-src in uninfected fibroblasts of several avian species (Wang et al., 1977; Spector et a/., 1978a,b). Similar findings were made subsequently for other c-oms and other species (Sheiness and Bishop, 1979; Bishop et at., 1981 ; Roussel et al., 1979; Chen, 1980). Some c-onc's may not be expressed, however; for example, a thoroughgoing search has failed to detect transcription from c-mos (Frankel and Fischinger, 1976). Efforts to detect transcription from c-jes were also unsuccessful (Frankel et al., 1979), but a more recent report described a protein (pp92'-jeS) that may be encoded by c+s and that was found in a number of mammalian species (Barbacid et al., 1980). Transcription from c-oms has not been widely studied, but a reasonably coherent set of data are available for four c-oms representative of avian retroviruses : c-src, c-myc, c-erb, and c-myb. Several principles are apparent that may endure even after a greater variety of c-ones has been studied (Bishop et al., 1981).

RETROVIRUSES AND CANCER GENES

13

1. Each of the c-oncs is transcribed in a variety of tissues and in every vertebrate species that has been satisfactorily examined. 2. The c-oms are not coordinately expressed as a group, and the function of each gene may be required only in certain tissues. 3. There is no evidence that transcription from c-oncs is ever coordinated with the expression of endogenous retrovirus genes (Wang et al., 1977; Spector et al., 1978b). 4. The c-oncs give rise to distinctive RNAs whose sizes have so far proved to be identical in various types of cells and even in different species. The constancy of these RNAs among widely divergent species testifies to the selective pressures that have apparently preserved the structure and function of c-oncs. As discussed already, most of the transcripts are appreciably larger than would be required to encode the nucleotide sequences that are shared by the homologous viral and cellular genes. 5. Three of the avian c-oncs give rise to single mature transcripts, in accord with the expectation that each locus represents but one gene. Transcription from c-erb provides a striking contrast, however, because at least four (and perhaps five) different mature RNAs have been identified, and the distribution of these RNAs varies among different tissues. Two of the RNAs are derived from one domain of c-erb, the others from a separate domain; this pattern mirrors the organization of v-erb, which is also composed of two independently expressed domains (v-erb-A and v-erb-B; see Table I, and Sheiness et al., 1981).

IX. Proteins Encoded by c-oncs

RNA transcribed from several c-oms has been found in polyribosomes and is therefore presumably translated into proteins (Spector et al., 1978a; personal communication, Vennstrom). The search for these proteins has not been easy : most cells contain only small amounts of c-onc mRNAs and proteins, and identification of c-onc proteins has so far turned on the development of antisera that react with the product(s) of the corresponding viral oncogenes-an undertaking that is itself unpredictable and technically demanding. Three proteins encoded by c-oncs have been identified to date : pp6OC-'"', a 60,000-MW phosphoprotein specified by c-src (Collett et al., 1978, 1979; Oppermann et al., 1979; Rohrschneider et al., 1979); ~21'-"~',a 21,000-MW protein encoded by the Harvey c-ras (Langbeheim et al., 1980); and p150c-"b', a 150,000-MW protein apparently derived from c-abl and formerly called NCPl5O (Witte et al., 1975b). Both pp6OC-"' and ~21'-*"~ have been extensively studied ; both are remarkably similar to their viral counterparts in

14

J . MICHAEL BISHOP

structure and apparent function (Collett et af., 1978, 1979; Karess et al., 1979; Oppermann et af., 1979; Rohrschneider et af., 1979; Langbeheim et af., 1980; Karess and Hanafusa, 1981); and both are found in a large variety of cells and are distributed across wide phylogenetic distances (Oppermann et al., 1979; Rohrschneider et al., 1979; Langbeheim et a f . , 1980). Much less is known of p150c-ah'(Witte et af., 1979); its size and composition are different from those of any of the proteins encoded by the several variants of v-abl; it has been found in appreciable (but very small) amounts only in thymocytes and other lymphoid cells ; its phylogenetic distribution has not been reported; and nothing is known of its function. Surprisingly, RNA transcribed from c-abf has been found widely distributed among tissues and cells of many different sorts, most of which contain no detectable p150c-"b' (personal communication, D. Baltimore). This discrepancy cannot at present be explained. X. How Similar Are Viral Oncogenes and c-oms?

Assessment of the similarities between c-oncs and their viral derivatives has taken two general forms : comparison of the nucleotide sequences that embody the genes, and comparison of the proteins encoded by the genes. Early studies with molecular hybridization raised the possibility of substantial kinship between c-onc and viral oncogene, but satisfactory tests of the issue awaited the isolation of the genes by molecular cloning on the one hand, identification and characterization of the proteins encoded by the genes on the other. With either or both of these chores now accomplished in several instances, the evidence mounts for remarkable similarity-if not identity-between the viral and cellular forms of oncogenes.

1. Coding sequences shared by viral oncogenes and cellular c-oms have so far been indistinguishable by heteroduplex analysis (Goff et af., 1980; Oskarsson et a f . ,1980; DeFeo et al., 1981; Parker et al., 1981; Favera et af., 198l), although ambiguities arise whenever the cellular locus is punctuated by introns. 2. DNA sequencing has permitted an extensive comparison of v-mos to c-mos (Van Beveren et al., 1981). The first several codons of v-mos are vestiges of the enu gene into which c-mos was inserted. Otherwise, only occasional nucleotide substitutions distinguish v-mos from c-mos. 3. The c-erb locus is extraordinarily complex, extending over at least 40 kb of chicken DNA and containing a minimum of 12 introns (personal communication, B. Vennstrom). When reduced to its exons, however, the locus displays organization similar to that of the homologous viral oncogene: each of the domains of v-erb (A and B; see table) has a separately

RETROVIRUSES AND CANCER GENES

15

expressed counterpart in chicken DNA (c-erb-A and c-erb-B) (personal communication, B. Vennstrom). 4. The viral and cellular forms of pp60"" are remarkably similar (Collett et al., 1978, 1979a,b; Karess et al., 1979; Oppermann et al., 1979; Rohrschneider et al., 1979; Karess and Hanafusa, 1981). They share size, display antigenic cross-reactivities, and yield closely related peptide maps. Both are principally affiliated with the plasma membrane of the cell (Courtneidge et al., 1980). Both are phosphorylated and have similarly disposed phosphoamino acids, with phosphoserine in the proximity of the aminoterminus (Collett et al., 1979a,b) and phosphotyrosine within a carboxyterminal domain (Hunter and Sefton, 1980a; Karess and Hanafusa, 1981). Both are protein kinases that phosphorylate tyrosine in substrate proteins (Collett et al., 1979a,b; Oppermann et al., 1979; Rohrschneider rt al., 1979; Collett et al., 1980; Hunter and Sefton, 1980a,b; Levinson et al., 1980). The two proteins can be distinguished on!y by very subtle criteria: some antisera react with the viral but not the cellular protein-presumably reflecting the fact that the antisera were raised against the viral rather than the cellular protein (Oppermann et al., 1979); the peptide maps of the two proteins differ in some ways (Rohrschneider et al., 1979; Sefton et al., 1980; Karess and Hanafusa, 1981); and the phosphotyrosine may be contained within different tryptic peptides in the two proteins (Karess and Hanafusa, 1981). It also remains possible that the viral and cellular proteins respond differently to controlling influences in the cell, and that the kinase activities of the two proteins have different substrate specificities. Definitive tests of these important potential distinctions are not presently available. 5. The possibility that v-SYCand c-src are functionally similar has received dramatic support from the demonstration that recombination between c-SYC (in chickens) and deletion mutants of v-src can reconstitute a functional oncogene (Hanafusa et al., 1977; Wang et al., 1978, 1979; Karess et al., 1979; Vigne et al., 1980; Karess and Hanafusa, 1981). In the most telling examples, the reconstituted oncogene retains a 3' terminal portion of v-src (-25% of the gene) but is otherwise apparently constructed entirely of nucleotide sequences derived from c-src (Wang et al., 1979; Karess and Hanafusa, 1981).The protein encoded by the reconstituted v-src is so similar to both pp60'-"' and pp6OC-"' that its genetic origins are difficult to discern, although peptide maps suggest that pp60'-"" of the recombinant virus is indeed a hybrid of both cellular and viral origins (Karess and Hanafusa, 1981; Vigne et al., 1980). Even these findings cannot assure us that pp60c--"r' and pp60'-"c are functionally equivalent, however; in every instance, the recombinant viral protein has derived at least 20% of its carboxyterminal domain from the parental virus in the recombination (Wang et al., 1979; Karess and Hanafusa, 1981), and it is the carboxyterminal domain of

16

J . MICHAEL BISHOP

pp6Ov-"*' that bears the protein kinase activity (Levinson et al., 1981; Oppermann et al., 1981). 6. The proteins encoded by Harvey v-rus and c-ras have been compared in considerable detail and appear to be quite similar. They each have a molecular weight of 2 1,000; they react with the same antisera; they yield related peptide maps; and they display the same biochemical function, viz., the capacity to bind guanine nucleotides with high affinity (Scolnick et al., 1979; Shih et al., 1980). They may differ in one regard, however; pp21"-r"s is phosphorylated on threonine residues, whereas phosphorylation of p2 1c-ras has yet to be detected (Shih et al., 1979a,b, 1980; Langbeheim e f al., 1980). 7. Functional similarities between V-ones and c-oncs have been demonstrated most persuasively by work with c-mos and c-rus. Both of these c-ones have been isolated by molecular cloning (c-mos from mouse DNA, c-ras from rat DNA) and coupled to a retrovirus promoter for transcription in order to favor vigorous expression. Some of the cells that receive these hybrid genetic units by transfection become transformed to a neoplastic phenotype (Oskarsson et ul., 1980; DeFeo et al., 1981), as if the c-ones might carry out the same functions as their homologous v-oncs.

-

None of the preceding examples provides a definitive demonstration of identity between viral oncogene and c-onc. But the weight of the evidence now suggests that retrovirus oncogenes encode functions also found in normal vertebrate cells. XI. Are

c-oms Members of a Multigene Family?

We presently know of at least 15 retrovirus oncogenes, each distinguished by its nucleotide sequence, and each with a corresponding oncogene (see Table I). Moreover, ostensibly similar oncogenes may be the products of related but separate cellular loci. For example, the apparently homologous oncogenes of the Harvey and Kirsten murine sarcoma viruses (v-ras), formerly believed to have originated from the same c-one, are now known to be the progeny of two different (albeit related) cellular genes (DeFeo et al., 198 1). The total number of c-oncs is therefore likely to grow as efforts to identify novel isolates of retroviruses continue. On the other hand, the number of these genes may not be inordinately large: the c-ones for src, myc, erb, myb, andfps are each represented at least twice among the handful of independently isolated avian retroviruses; c:fhs is a feline counterpart of cTfjs that appears in two strains of feline sarcoma virus (Shibuya et al., 1980; and v-bus and v-rus are derived from closely related genes in the mouse and rat, respectively (personal communication, S . Aaronson and E.

RETROVIRUSES AND CANCER GENES

17

Scolnick). The reiterative emergence of c-oncs in different viral isolates and from different species suggests that we may have the majority of these genes already in view. Whatever their number, c-oncs might constitute a family of genes whose interrelationships are akin to those found in the multigene families that encode immunoglobulins, histocompatibility antigens, and so forth. This suggestion stems from the fact that all c-oncs, however diverse in structure, give rise to viral genes with the dramatic property of oncogenicity in common. In fact, there are reasons to believe that the apparent structural diversity of c-oncs may obscure common origins and related functions: (a) the nucleotide sequences of v-src, v-mos, and v-fes are very different, yet the amino acid sequences encoded by these genes reveal significant homologies that indicate a common ancestor (Van Beveren et al., 1981 ; personal communication, .I.Stephenson); (b) several different viral oncogenes (and so far as we know, their c-oncs as well) encode tyrosine protein kinases (Collett et al., 1980; Feldman et al., 1980; Levinson et al., 1980; Van de Ven et al., 1980; Witte et al., 1980; Neil et al., 1981), and these enzymes may affect similar sets of cellular proteins (Cooper and Hunter, 1981; personal communication, T. Hunter). Given the apparent functional relationships among the identified c-oncs, it is of interest to know whether these genes might be clustered or linked in the cellular genome. The available data are in conflict. Fractionation of chicken chromosomes by rate-zonal centrifugation has located c-src on one of the smaller macrochromosomes (Padgett et al., 1977; Hughes et al., 1979b),c-myc on one of the two or three largest chromosomes (Sheiness et al., 1980), and c-erb on a chromosome of intermediate size (personal communication, B. Vennstrom). By contrast, hybridization in situ indicated that c-src, c-myc, c-myb, and c-erb are all located on one or another of the chicken microchromosomes (Tereba ef al., 1979; personal communication, A. Tereba). The discrepancies may arise from the fact that the cells used for chromosome fractionation are neoplastic and contain at least one chromosomal translocation. On the other hand, it may not be necessary that all c-oncs be genetically linked. For example, demonstrably related genes (such as a- and B-globin, and genes whose functions are coordinately induced by estrogen) are located on different chromosomes in the chicken (Hughes el al., 1979b). XII. How Might Retroviruses Transduce Cellular Genes?

By what means have c-oncs been acquired by viral genomes? Two competing answers to this question have been proposed.

18

J. MICHAEL BISHOP

1. It is possible that each retrovirus arises fully grown from the rearrangement and permutation of cellular genes due to the antics of ancestral transposable elements (Temin, 1980). This account is a restatement of Temin’s original “protovirus hypothesis” (Temin, 1974) and suggests that oncogenes may be present from the inception of certain retrovirus genomes. 2. It is more generally assumed that preexistent retroviruses assimilate c-oncs by recombination (Bishop, 1981). Several lines of evidence conform to (but do not prove) this explanation. First, large (but never complete) deletions in v-src can be repaired by recombination with c-src in chickens (see earlier). It is by no means certain, however, that the mechanism of this recombination provides a general explanation for the transduction of c-oncs. Second, the oncogenes of several murine retroviruses (v-YUS,v-mos, and v-abl) appeared during the passage of leukemia viruses in rodent hosts (Harvey, 1964; Moloney, 1966; Abelson and Rabstein, 1970a,b). We presume, but cannot prove, that here the experimentalist may have reproduced the events that give rise to v-oncs in the wild. Third, several investigators have reported deliberate and apparently successful efforts to transduce c-oncs by infection of cells in tissue culture with retroviruses that do not initially contain oncogenes (Rasheed et al., 1978; Rapp and Todaro, 1978, 1980). These efforts have produced retroviruses with varied and novel oncogenic potentials, but the genetic bases of most of these potentials has yet to be elucidated. The protovirus hypothesis can be sustained only by circumstantial evidence and arguments. But proponents of the recombination model must also confront vexing problems. In particular, suitable homologous regions that could facilitate crossing over between cellular and viral genomes have not been identified, and we need to explain how the introns of c-oncs are removed to generate the uninterrupted coding units of viral oncogenes. Figure 2 illustrates the prevailing model that attempts to meet these strictures. The model posits the chance juxtapositioning of integrated retrovirus DNA and a c-onc locus, followed by a deletion that joins the viral and cellular DNA into a single genetic element. Transcription from this element begins on the retrovirus promoter, proceeds through the c-om, and generates a hybrid RNA from which the introns of the c-onc have been removed. The viral portion of the hybrid RNA permits the formation of heterodimers (retrovirus genomes are diploid) with the genome of a superinfecting (or co-resident) retrovirus and encapsidation of the RNA into virions. Recombination occurs by a form of copy-choice when the virions enter new host cells and initiate reverse transcription from viral RNA. The crossing over is illustrated as “illegitimate”-requiring no homology between the participating nucleic acids-but is otherwise similar to previous models for

-

or

I

C-one

Deletion

Transcription and Splicing AA

Packaging

Reverse Transcription

.ct gag

pol

env

Integration

v-onc

FIG.2. A model for transduction by retroviruses. The scheme begins with an intact retrovirus provirus integrated adjacent to a cellular oncogene. The characteristic terminal redundancies of the provirus (Temin, 1980) are illustrated by black and white boxes. Exons of the c-one locus are denoted by stippled boxes. A deletion fuses the provirus and the c-one into a single genetic unit, in the process eliminating one of the proviral terminal redundancies and other portions of the viral genome; two typical outcomes are illustrated. Transcription from the hybrid unit generates polyadenylated RNA (solid lines), from which the introns of the c-onc have been removed by splicing. The hybrid RNA can form a heterodimer with the genome of another retrovirus in the same cell; the resulting diploid unit can be packaged into virus and transferred into another cell. Infection of a new host cell initiates reverse transcription from viral RNA. Beginning near the 5’ end of the genome and then jumping to the 3’ end to continue chain propagation along the length of the genome (Gilboa ef al., 1979). Copy-choice during reverse transcription can generate a new retrovirus genome contining the c-one.

20

J . MICHAEL BISHOP

recombination among retrovirus genomes (Coffin, 1979). Crucial portions of this model have received support recently from work that demonstrated the encapsidation of hybrid RNAs similar to those illustrated by the scheme in Fig. 2, and subsequent illegitimate recombination to bring the parental cellular domain of the RNA into a reconstituted (albeit defective) retrovirus genome (Goldfarb and Weinberg, 1981). Other puzzles remain as well. Is the seizure of c-ones a unique event, or might retroviruses be generalized transducing agents whose acquisition of more prosaic genes is merely less likely to be perceived? Are there selective pressures that favor the transduction of c-ones and their retention by retrovirus genomes? When oncogenes are formed by fusing c-one to a portion of a viral structural gene (as is frequently the case; see earlier), what portion of the cellular locus actually joins the viral genome, and what influence does the hybrid nature of the resulting oncogene have on oncogenicity? (The hybrid genetic structure appears not to be necessary for tumorigenesis: at least two oncogenes-v-myc and v-myb-occur as both hybrid genes and as independently expressed loci not fused with gag.) How are we to interpret the unusual nature of the oncogene for the Spleen Focus Forming Virus? Is it an exception to an otherwise pervasive rule, or does it signify that the origins of retrovirus oncogenes are more diverse and more complex than we presently realize? And what of the oncogenes of DNA tumor viruses? Efforts to trace the origins of these genes to cellular DNA have failed to date (although some uninfected cells allegedly contain a protein that is serologically related to the T antigen of SV40 virus; see Lane and Hoeffler, 1980). It appears that recombination between DNA tumor viruses and their host cells is less facile than that displayed by retroviruses, so we can rationalize the failure of DNA viruses to transduce cellular oncogenes. But to what end have the oncogenes of DNA viruses evolved, and what are the goods of which they are formed? XIII. What Is the Function of c-oncs in Normal Cells?

It has become an article of faith that c-ones serve essential purposes in uninfected cells. Why else would these genes have been conserved over long periods of evolutionary time, and why else would many of their numbers be expressed in both embryonic and adult tissues? The inference is easy to draw, but difficult to explore, and the difficulty lies less with biochemical function than with cellular physiology. Once pp60'-"' was known to be a protein kinase, the demonstration of a similar enzymatic activity associated with pp60""" followed in short order (Collett et al., 1979a,b; Oppermann c't al., 1979; Rohrschneider, ct al., 1979). But how does this enzymatic activity-or the biochemical function of any other c-one, for that matter-

RETROVIRUSES A N D CANCER GENES

21

serve the metabolism of the normal cell? The question is usually answered by reasoning that the actions of viral oncogenes mirror the functions of c-oms. The cell transformed by a retrovirus oncogene divides incessantly. Might the homologous proto-oncogene therefore be a normal effector of cell division? Many (perhaps all) retrovirus oncogenes arrest, reverse, or otherwise disturb the normal course of cellular differentiation (Graf and Beug, 1978; Boettiger and Durban, 1979; Maltzman and Levine, 1981). Might their counterparts in normal cells be regulators of growth and development, and if so, might the lineages in which they are normally active dictate the sorts of cells that are vulnerable to transformation by the homologous viral oncogenes? Experimental data that speak to these issues are sparse and enigmatic.

1 . Because v-src transforms fibroblasts, it is conceivable that expression of the cellular homologue c-src might vary in concert with changes in cell growth. To date, efforts to sustain this expectation have failed. For example, the expression of c-src remained unchanged throughout the course of experiments in which the growth of fibroblasts was first arrested for as long as 2 weeks by serum deprivation and then stimulated by restoration of serum to the growth medium (Spector et al., 1978a). 2. Efforts to discern preferential expression of c-oms in specific tissues have so far failed to yield coherent results. Some loci (such as c-src, c-myc, and c-erb) are active at low or intermediate levels across a broad spectrum of tissues, whereas the activities of others (e.g., c-myb) are more restricted in their distribution (personal communications, T. Gonda and D. Sheiness). In most instances, the distribution of activity does not conform to predictions based on the pathogenicities of the corresponding oncogenes. The most provocative finding at present comes from Scolnick and his colleagues (Scolnick et al., 1981) : primitive hematopoietic cells (pluripotential CFU-S cells)-but not cells of other origins-contain large amounts of ~ 2 1 ' - ~ ~a' , protein whose viral homologue is apparently capable of inducing erythroleukemias as well as sarcomas. XIV. The Paradox of Neoplastic Transformation by Retrovirus Oncogenes

If retrovirus oncogenes embody functions found also in normal vertebrate cells, why do the oncogenes induce abnormal phenotypes in infected cells? Two possible answers come to mind. First, transformation by retroviruses may be a consequence of dosage: the virus may overload cells with otherwise normal gene products; sustained and abundant expression of the genes, rather than anomalous properties of their products, may lie at the root of

22

J . MICHAEL BISHOP

tumorigenesis by v-oncs. Alternatively, viral oncogenes and c-oncs may differ in subtle but important ways. For example, the protein kinase activity of pp6Ov-"*' might have unique substrate specificities that could account for neoplastic transformation by Rous sarcoma virus. Although the extent of resemblance between c-oncs and viral oncogenes has yet to be fully measured, several points of evidence suggest that retrovirus oncogenes transform cells by means of dosage. 1. The dosages of v-onc products are indeed large, when compared with the amounts of c-onc products found in most cells ; fibroblasts transformed by Rous sarcoma virus contain about 100-fold more pp6OvpSrc than pp60""" (Collett et al., 1978; Oppermann et al., 1979), and similar differences have been found for other oncogene and c-onc products (Witte et al., 1979; Langbeheim et al., 1980). The vigor of oncogene expression may be attributable in large measure to the efficacy of the retrovirus promotet for transcription. 2. The relatively large dosages of oncogene products are essential to maintain the neoplastic phenotype. On occasion, presently unidentified events in the infected cell can attenuate the synthesis of retrovirus RNA and, hence, the expression of viral genes. The amounts of oncogene product fall by 10- to 100-fold,and the cell reverts to an ostensibly normal phenotype (MacPherson, 1965; Boettiger, 1974; Deng et al., 1977; Bishop et al., 1979; Prozig et al., 1979). 3. Molecular clones of two c-oncs (c-mos and w a s ) have been linked to transcriptional promoters from the murine leukemia virus genome (see earlier). The chimeric DNAs, bearing no portion of a viral oncogene, can transform fibroblasts to a neoplastic phenotype (Oskarsson et al., 1980; Blair et al., 1981; DeFeo et al., 1981; and personal communications, G. Vande Woude and E. Scolnick). Cells transformed in this manner by c-ras contain relatively large quantities of the gene product, p21'-'"" (DeFeo et al., 1981 ; and personal communication, E. Scolnick), thus sustaining the view that amplification of c-ras expression suffices to induce the neoplastic phenotype. XV. Does the Homology between Viral Oncogenes and c-oms Dictate the Host Range of Viral Transformation?

Neoplastic transformation by retrovirus oncogenes is remarkably specific : pathogenicity for specific tissues is a distinctive property of each strain of retrovirus ; and in cell culture, oncogenes display a similarly predictable and generally limited range of susceptiblecells that correlates well with the actions

RETROVIRUSES AND CANCER GENES

23

of the oncogene in infected animals (Graf and Beug, 1978). The origins of “target cell specificity,” as this phenomenon is known, are uncertain. It is first of all possible that transformation is merely a reflection of susceptibility to infection by different strains of virus. Inference speaks against this possibility: viral genes that determine the host range of infectivity (such as gag and enu) are shared among families of retroviruses whose spectra of oncogenicities are very different. Moreover, experimental data indicate that cells can be infected by and produce retroviruses bearing oncogenes without necessarily being transformed (Graf et af., 1980; Durban and Boettiger, 1981a,b). These findings have led to the more subtle suggestion that only certain cells are vulnerable to the effects of each oncogene. What factors could determine cellular vulnerability to transformation by v-oncs? We do not know, but it has been suggested that the kinship of viral oncogenes and c-oncs might be responsible : the deleterious effects of oncogene dosage might be restricted to cells in which the homologous c-onc is normally expressed and effective; alternatively, cells in which the c-onc is not usually active might be more vulnerable to the actions of the oncogene. The available (and admittedly provisional) data do not sustain these views. For example, the patterns of c-onc expression among different tissues in no way reflect the spectrum of susceptibility to the homologous viral oncogenes (personal communication, T. Gonda and D. Sheiness). XVI. Do c-oncs Provide a Pathway for Oncogenesis?

The action of viral oncogenes may provide useful analogues for the enzymatic mechanisms that give rise to and sustain the malignant phenotype. But it appears that viruses bearing oncogenes are not usually responsible for tumorigenesis in human beings (Pimentel, 1979). We should therefore look to the cell itself if we are ever to discern common origins of malignancy. In particular, we need to identify the events that spark the onset of oncogenesis, and we must determine whether a particular cellular gene (or set of genes) always mediates progression to and maintenance of the malignant phenotype. We do not know how oncogenesis originates; the matter has elicited great controversy, with some investigators arguing for mutations (Ames, 1979; Epstein and Swartz, 198l), others for chromosomal rearrangements, transpositions of DNA, or even reversible epigenetic events (Rubin, 1980; Cairns, 1981 ; Mintz and Fleischman, 1981). By contrast, the discovery of c-oms may have brought to view genes whose actions can mediate oncogenesis, once the cell has sustained an initiating lesion. We have diverse reasons to suspect the existence of such “cancer genes.”

24

J. MICHAEL BISHOP

1. A number of malignancies have appeared as heritable traits in human pedigrees (Knudson, 1981), and it has even been suggested that each of the roughly 100 types of malignancy will eventually be attributable to abnormalities affecting a specific genetic locus (Knudson, 1979, 1981). 2. Several efforts have been made to enumerate the genetic loci that might mediate neoplastic transformation by chemical carcinogens. In most instances, the results implicate no more than a few dozen or a few hundred genes as potential mediators of chemical carcinogenesis (Parodi and Brambilla, 1977). 3. DNA extracted from some lines of chemically transformed cells and from certain tumors induces neoplastic transformation when transferred into cells in culture (Shih et al., 1979a; Krontiris and Cooper, 1981; Shilo and Weinberg, 1981). Caveats are necessary: the efficiency of transformation is generally quite low; transformation occurs reliably in only a few established lines of recipient cells (particularly mouse NIH-3T3); and only a limited number of transformed cell lines or tumors have so far yielded DNA capable of inducing neoplastic transformation. But the data to suggest that the DNA from at least some forms of neoplastic cells contains stable and heritable changes that are responsible for the malignant phenotype. Moreover, provisional studies with restriction endonucleases indicate that the same domain of DNA may be affected in independent tumors of common type and/or common cause (Shilo and Weinberg, 1981). 4. If first sheared to molecular weights of -0.3-3 x lo6, even DNA from normal cells can transform NIH-3T3 cells at a very low frequency, and the transformed mouse cells in turn yield DNA that can induce transformation at much higher efficiencies (Cooper et al., 1980)-as if the original shearing of normal DNA unleashed a potentially oncogenic gene whose action is now stably established in the transformed mouse cells. Activation of the gene has been attributed to disruption of linkage between the oncogenic gene and a cis-active regulator (Cooper et al., 1980). Are c-oncs among the cancer genes of normal cells? Is the induction of their activity responsible for at least some forms of oncogenesis? Answers to these question may come eventually from surveys of c-onc expression in naturally occurring tumors. At the present, we have a single but immensely provocative clue, derived from an unexpected source-the study of tumorigenesis by avian leukosis virus (ALV). ALV has no oncogene, yet infection of the bursa in young chicks by this virus gives rise to a fatal B-cell lymphoma (Gross, 1970). The course of tumorigenesis is protracted, unlike tumorigenesis by viral oncogenes, and most of the resulting tumors give evidence of having originated from single infected cells-they arise in individual follicles of the bursa (Cooper et id.,

RETROVIRUSES AND CANCER GENES

25

1978), and they contain proviruses whose sites of integration in cellular DNA are identical in every member of the tumor cell population (Payne et al., 1981a; Nee1 et al., 1981). In many instances, the proviruses are affected by deletions that preclude the expression of viral genes in the tumor cell (Payne et ul., 1981a); thus viral proteins are not required to sustain the lymphomatous phenotype. How then does viral infection induce the tumors? A possible solution to this puzzle began to unfold with the discovery that the leukosis virus DNA is almost always inserted in the vicinity of c-myc, and as a seeming consequence of these insertions, the expression of the c-one appears to have been greatly augmented (Hayward et al., 1981). The apparent activation of c-myc has been attributed to the influence of the nearby ALV provirus, or more specifically, to the promoter for RNA synthesis contained within the provirus. And indeed, in many instances, transcription may initiate within the provirus and then proceed through the c-myc locus, generating RNA that is a composite of nucleotide sequences derived from the provirus and from c-myc (Hayward et al., 1981). But two forms of exceptions have been found that cannot be easily explained (Payne et al., 1981b). In one, the ALV provirus is inserted upstream from c-myc, but the orientation of the provirus is the opposite of that required to promote transcription into the c-myc locus. In the other form of exception, the ALV provirus is located downstream from c-myc; transcription traverses c-myc in the usual direction and then continues into a portion of the provirus. It thus appears that some insertions of viral DNA in the vicinity of c-myc activate expression of the cellular locus by means other than providing a direct promoter for transcription. This nuance should not be allowed to obscure the general significance of the findings with ALV : oncogenesis by retroviruses may not always require viral gene products ; the frequency with which ALV proviruses are found in the vicinity of c-myc implies that the site of insertion figures in the oncogenic mechanism; and it is likely (but by no means proven) that the heightened expression of c-myc provokes and/or sustains the chain of events that eventuate in lymphoid leukosis. Questions remain, however. 1. In a few ALV-induced tumors, viral DNA is not inserted near c-myc, and expression of c-myc has not been induced (Hayward et a/., 1981; and personal communication, W. Hayward). Nevertheless, single (or very few) sites have been used for integration of the ALV provirus in each of these tumors, and it is therefore possible that the insertion of viral DNA has induced the expression of nearby cellular DNA representing a c-one. 2. Is the alleged tumorigenic effect of c-myc limited to cells in the B-lymphocyte lineage? Possibly not. For example, provisional data also implicate

26

J . MICHAEL BISHOP

c-myc in the genesis of renal tumors induced by ALV (personal communication, P. Neiman). It is perhaps significant that these tumors are analogous to the renal carcinomas commonly induced by the action of v-myc (Moscovici et al., 1978). 3. Is c-myc an inevitable participant in the genesis of B-cell (and perhaps renal) tumors, whatever their initiating cause? The induction of c-myc expression has been implicated in lymphomagenesis by another retrovirus that lacks an oncogene, the chicken syncytial virus (CSV) (personal communication, H. J. Kung). Although the CSV genome is not homologous to that of ALV, the similarities of the two viruses are too great to provide a compelling test of the larger issue. 4. The viral oncogene (v-myc) derived from c-myc has never been reported to cause lymphomas (Moscovici et af., 1978). The implication of c-myc in the genesis of B-cell tumors therefore came as a surprise that remains unexplained. Recall, however, that the induction of lymphomas by ALV follows a protracted course of events that may not spring immediately from the effects of a single gene: the roles of c-myc and v-myc in oncogenesis may differ greatly. There is evidence to sustain this deduction (Cooper and Neiman, 1980). DNA from lymphoid tumors induced by ALV elicits the neoplastic phenotype when introduced into mouse fibroblasts by transfection. Contrary to expectations, however, the transformed mouse cells contain neither ALV DNA nor c-myc derived from the lymphoid tumors (personal communication, G. Cooper). Transformation of the fibroblasts must therefore be due to another cancer gene, activated in the infected B cells (perhaps by the effects of c-myc), and more effective than c-myc itself in the transformation of fibroblasts. In this scheme, c-myc can be viewed as the initiator of tumorigenesis, the cancer gene responsible for transformation of fibroblasts as a potential maintenance function. XVII. Conclusion: The Pursuit of Cancer Genes

It now appears likely that normal cells bear the seeds of their own destruction in the form of cancer genes, whose anomalous activities mediate tumorigenesis. The term cancer genes is a convenience, of course, and is viewed by some as a misnomer (Mintz and Fleischman, 1981): the loci in question may be physiologically essential constituents of the cell’s genetic apparatus that become pathogenic only when their structure or control is disturbed by oncogenic agents. The lines of inquiry surveyed in this article have engendered three strategies by which cancer genes might be identified and isolated.

RETROVIRUSES A N D CANCER GENES

27

1. The role of retrovirus c-oncs in tumorigenesis can be explored by testing for their anomalous expression in spontaneously occurring tumors. There are uncertainties in this exercise. How is normal expression of c-oms to be defined? The cells from which many tumors arise are either poorly defined or too rare to be easily studied. Do we have the entire family of c-oncs in hand? It seems unlikely that we do, given the rate at which the repertoire of retrovirus oncogenes has expanded of late (Bishop, 1981), and the indications that even more v-oms can be found if the issue is forced by experimental ingenuity (Rapp and Todaro, 1978, 1980). And how certain can we be that the families of c-oncs and cancer genes are mutually inclusive? It is conceivable, for example, that some cancer genes will elude transduction by retroviruses, or if transduced, will fail to act as v-oms and thus go unrecognized. 2. The findings with ALV reveal for the first time how viruses devoid of oncogenes might cause tumors. The induction of c-onc expression by the integration of ALV DNA is a form of insertional mutagenesis, and other integrative viruses are in principle capable of the same. Potential examples abound: the nephroblastoma retrovirus of chickens (Watts and Smith, 1980); the thymic leukemia (Rowe, 1973) and mammary carcinoma (Bentvelzen and Hilgers, 1980) viruses of mice; the leukemia viruses of cats (Essex, 1975), cows (Burny et al., 1978), and gibbon apes (Gallo and Wong-Staal, 1980); and further afield, leukemogenesis by SV40 virus (Diamandopoulos, 1978), the hepatitis B virus that has been implicated in the genesis of hepatic carcinoma (Szmuness, 1978), and oncogenesis by herpes viruses (Biggs et al., 1972). If any of these sundry viruses acts by means of insertional mutagenesis, the site(s) of its integration in the DNA of the tumor cell may finger a cancer gene (or genes) whose effects are restricted to the tissue(s) in which the virus induces tumors. Virologists are in hot pursuit of these possibilities, hoping to expand the catalogue of cancer genes and to gain insight into the tissue specificity of their actions. 3. DNA-mediated transfection offers a third and versatile strategy whose reach may prove to be broader than the virological approaches just described. Any cancer gene whose action can be perceived in previously normal recipient cells becomes accessible to subsequent isolation by molecular cloning. Informal accounts indicate that first successes with this strategy are in the immediate offing. Success with any or all of these schemes will be heady business, no doubtcause for celebration and optimism. But daunting problems will remain. First, we are not likely to gain the full variety of cancer genes from the present strategies : none of the schemes seems geared to yield recessive

28

J . MICHAEL BlSHOP

effectors of oncogenesis, and none seems likely to unveil regulatory loci whose mutation might unleash the activity of an otherwise normal cancer gene. Second, we would be naive if we were to regard a single genetic locus as the full explanation for the genesis of a spontaneously occurring tumor. Most if not all human tumors appear to arise from several successive events (Prehn, 1976). Third, we must invent means by which to distinguish genes that initiate oncogenesis from those that maintain the neoplastic phenotype. The distinction is not necessarily inherent in any of the current strategies for identifying cancer genes (although the immediacy with which transformation follows DNA-mediated transfections appears to betoken maintenance rather than initiating functions). Fourth, the isolation of cancer genes offers no guarantee that we can deduce the mechanisms of their action. Here the lessons of tumor virology-which served so well to launch the search for cancer genes-may again provide seminal clues. Caveats such as these cannot blunt the zest with which cancer genes are now pursued, and they cannot detract from the object lessons the pursuit has produced. The study of viral agents far removed from human concerns appears to have wrought the long-elusive conjunction between tumor virology and human oncology. The issue is not whether viruses might cause human tumors (as perhaps they may), but rather how much viral oncology can teach us of the mechanisms by which human tumors arise. ACKNOWLEDGMENTS 1 thank B. Cook for assistance and R. Swanstrom for conceiving and preparing Fig. 2. Portions of this article will also appear as part of a chapter prepared for a book on RNA tumor viruses, to be published by the Cold Spring Harbor Laboratory, Cold Spring Harbor, New York. Work in my laboratory is supported by grants from the National Cancer Institute and the American Cancer Society.

REFERENCES Aaronson, S. A., and Stephenson, J. R. (1976). Biochim. Biophys. Acta 458,323-354. Abelson, H. T., and Rabstein, L. S. (1970a). Cancer Res. 30,2208-2212. Abelson, H. T., and Rabstein, L. S. (1970b). Cancer Res. 30,2213-2222. Ames, B. N. (1979). Science 204,587-593. Anderson, S . M., Hayward, W. S., Neel, B. G., and Hanafusa, H. (1980). J . Virol. 36,676-683. Barbacid, M., Beemon, K., and Devare, S. G. (1980). Proc. Narl. Acad. Sci. U.S.A. 77, 51585162. Bentvelzen, P., and Hilgers, J. (1980). In “Viral Oncology” (G. Klein, ed.), pp. 31 1-355. Raven, New York. Beug, H., and Graf, T. (1980). Virology 100,348-356. Biggs, P. M., De-The, G., and Payne, L. N. (1972). Inr. Agency Res. Cancer. Bishop, J. M. (1981). Ce/l23,5-6.

RETROVIRUSES AND CANCER GENES

29

Bishop, J. M., Courtneidge, S, A., Levinson, A. D., Oppermann, H. Quintrell, N., Sheiness, D. K., Weiss, S. R., and Varmus, H. E. (1979). Cold Spring Harbor Symp. Quant. Biol. 44, 919-930. Bishop, J . M., Gonda, T., Hughes, S. H., Sheiness, D. K., Stubblefield, E., Vennstrom, B., and Varmus, H. E. (1981). In “Mobilization and Reassembly of Genetic Information” (Twelfth Miami Winter Symp.), pp. 261-273. Blair, D. G., Oskarsson, M., Wood, T. G., McClements, W. L., Fischinger, P. J., and Vande Woude, G. G. (1981). Science 212,941-943. Boettiger, D. (1974). Virology 62,522-529. Boettiger, D., and Durban, E. M. (1979). Cold Spring Harbor Symp. Quanr. Biol. 44 , 12491254. Brugge, J. S., and Erikson, R. L. (1977). Nature (London) 269,346-347. Burny, A., Bex, F., Chantrenne, H., Cleuter, Y., Dekegel, D. Ghysdael, J., Kettmann, R., Leclercq, M., Leunen, J., Mannerickx, M., and Portetelle, D. (1978). Ado. Can. Res. 28, 25 1-31 I . Cairns, J. (1981). Nature (London) 289, 353-357. Chen, J. H. (1980). J. Virol. 36, 162-170. Coffin, J . M. (1979). J . Gen. Virol. 42, 1-26. Coffin, J . M., Varmus, H. E., Bishop, J. M., Essex, M., Hardy, W. D., Jr., Martin, G. S., Rosenbert, N. E., Scolnick, E. M., Weinberg, R. A , , and Vogt, P. K. (1981). J. Virol., in press. Collett, M. S., Brugge, J. S., and Erikson, R. L. (1978). Cell 15, 1363-1370. Collett, M. S., Erikson, E., and Erikson, R. L. (1979a). J . Virol. 29,770-788 Collett, M. S., Erikson, E., Purchio, A. F., Brugge, J. S., and Erikson, R. L. (1979b). Proc. Narl. Acad. Sci. U.S.A. 76,3159-3163. Collett, M. S., Purchio, A. F., and Erikson, R. L. (1980). Nature (London) 285, 167-168. Cooper, G. M., and Neiman, P. E. (1980). Nature (London)287,659-660. Cooper, G . M., Okenquist, S., and Silverman, L. (1980). Nature (London)284,418-421. Cooper, J. A , , and Hunter, T. (1981). Mol. Cell. Biol. 1,394-407. Cooper, M. D., Payne, L. N., Dent, P. B., Barmeter, B. R., and Good, R. A. (1978). JNCI, J . Natl. Cancer Inst. 41, 373-389. Courtneidge, S. A., Levinson, A. D., and Bishop, J. M. (1980). Proc. Natl. Acad. Sci. U . S . A . 77,3783-3787. DeFeo, D., Gonda, M. A., Young, H. A., Chang, E. H., Lowy, D. R., Scolnick, E. M., and Ellis. R. W. (1981). Proc. Natl. Acad. Sci. U.S.A.78,3328-3332. Deng, C.-T., Stehelin, D., Bishop, J. M., and Varmus, H. E. (1977). Virology 76,313-330. Diamandopoulos, G. T. (1978). Fed. Proc. Fed. Am. Soc. Exp. Biol. 37,2082-2084. Duesberg, P . H., and Vogt, P. K. (1970). Proc. Natl. Acad. Sci. U.S.A. 67, 1673-1680. Durban, E. M., and Boettiger, D. (1981a). Proc. Natl. Acad. Sci. U.S.A. 78, 3600-3604. Durban, E. M., and Boettiger, D. (1981b). Cell, in press. Ellennan, V., and Bang, 0. (1908). Zentralbl. Bukteriol. Abr. 146, 595-609. Ellis, R. W., DeFeo, D., Maryak, J. M., Young, H. A,, Shih, T. Y., Chang, E. H., Lowy, D. R., and Scolnick, E. M. (1980). J . Virol. 36,408-420. Epstein, S. S., and Swartz, J. B. (1981). Nature (London) 289, 127-130. Essex, M. (1975). Ado. Cancer Res. 21, 175-248. Favera, R. D., Gelmann, E. P., Gallo, R. C., and Wong-Staal, F. (1981). Nature (London) 292, 31-35. Feldman, R. A., Hanafusa, T., and Hanafusa, H. (1980). Cell 22,757-766. Franchini, G . , Even, J., Sherr, C. J., and Wong-Staal, F. (1981). Nature (London) 290,154-156. Frankel, A. E., and Fischinger, P. J. (1976). Proc. Narl. Acad. Sci. U.S.A. 73,3705-3709. Frankel, A. E., and Fischinger, P. J. (1977). J . Virol. 21, 153-160.

30

J . MICHAEL BISHOP

Frankel, A. E., Neubauer, R . L., and Fischinger, P. J . (1976). J. Virol. 18,481-490. Frankel, A. E., Gilbert, J. H., Prozig, K. J., Scolnick, E. M., and Aaronson, S. A. (1979). J. Virol. 30,821-827. Gallo, R. C., and Wong-Staal, F. (1980). In “Viral Oncology” (G. Klein, ed.), pp. 390-431. Raven, New York. Gilboa, E. Mitra, S. W., Goff, S., and Baltimore, D. (1979). Cell 18,93-100. Goff, S. P.,Gilboa, E., Sitte, 0. N., and Baltimore, D. (1980). Cell 22,777-786. Goldfarb, M. P., and Weinberg, R. A. (1981). J. Virol38, 136-150. Graf, T., and Beug, H . (1978). Biochim. Biophys. Acra 516,269-300. Graf, T., Beug, H., and Hayman, M. J . (1980). Proc. Narl. Acad. Sci. U.S.A. 77,389-393. Gross, L. (1970). “Oncogenic Viruses.” Pergamon, Oxford. Hanafusa, H., Halpern, C. C . , Buchhagen, D. L.,and Kawai, S. (1977). J. Exp. Med. 146, 1735-1747. Harvey, J . J. (1964). Narure (London) 1104-1 105. Hayward, W. S., Ned, B. G . , and Astrin, S. M. (1981). Nature (London) 290.475-480. Hueber, R. J., and Todaro, G. J. (1969). Proc. Nail.Acad. Sci. U.S.A. 64, 1087-1094. Hughes, S. H., Payvar, F., Spector, D., Schimke, R. T., Robinson, H . L., Payne, G. S., Bishop, J. M., and Varmus, H. E. (1979a). Cell 18,347-360. Hughes, S. H., Stubblefield, E., Payvar, F., Engel, J. D., Dodgson, J. B., Spector, D., Cordell. B.,Schimke,R.T.,andVarmus,H. E.(1979b).Proc. Narl. Acad.Sci. U.S.A.76,1348-1352. Hughes, S. H.. Vogt, P. K., Stubblefield, E., Robinson, H., Bishop, J. M., and Varmus, H . E. (1979~).Cold Spring Horbor Symp. Quanr. Bid. 44, 1077- 1089. Hunter, T., and Sefton, B. M . (1980a). Proc. Nail. Acad. Sci. U.S.A. 77, 1311-1315. Hunter,T., and Sefton, B. M. (1980b).In ”Molecular Aspects ofCellular Regulation” (P. Cohen and S. Van Heznigen, eds.), pp. 5-42. Elsevier, Amsterdam. Hynes, R. 0. (1980). Cell. 21,601-602. Karess, R. E., and Hanafusa, H. (1981). Cell24, 155-164. Karess, R. E., Hayward, W. S., and Hanafusa, H . (1979). Proc. Nail. Acud. Sci. U.S.A. 76, 3154-31 58. Knudson, A. G . (1979). New Engl. J . Med. 301,606-607. Knudson, A. G. (1981). Annu. Symp. Fundam. Cancer Res., 3 3 4 M. D . Anderson Hosp. Tumor Inst. pp. 453-462. Krontiris, T. G . ,and Cooper, G . M. (1981). Proc. Nurl. Acad. Sci. U.S.A. 78,1181-1184. Lai, M. M.-C., Duesberg, P. H., Horst, J., and Vogt, P. K. (1973). Proc. Nail. Acnd. Sci.U.S.A. 70,2266-2270. Lane, D. P., and Hoeffler, W. K. (1980). Nature (London) 288, 167-170. Langbeheim, H.. Shih, T. Y., and Scolnick, E. M. (1980). Virologv 106,292-300. Levinson, A. D., Oppermann, H., Varmus, H. E., and Bishop, J. M. (1980). J . Biol. Chem. 255, I 1973- I 1980. Levinson, A. D., Courtneidge, S. A., and Bishop, J. M. (1981). Proc. Null. Acad. Sci. U.S.A. 78, 1624- 1628. MacPherson, I. A. (1965). Science 143, 1731-1733. Maltzman, W., and Levine, A. J. (1981). In “Advances in Virus Research” (M. A. Lauffer, F. B. Bang, K. Maramorosch, and K. M. Smith, eds.), Vol. 26, pp. 65-116. Academic Press, New York. Martin, G. S., and Duesberg, P. H. (1972). Virology 47,494-497. Mintz, B., and Fleischman, R. A. (1981). A h . Cancer Res. 34,211-278. Moloney, J. B. (1966). In “Conference on Murine Leukemia,” pp. 139-142. (Natl. Cancer Inst. Monograph No. 22.) US.Public Health Service, Bethesda, Maryland. Moscovici, C., Alexander, R. W., Moscovici, M. G., and Vogt, P. K . (1978). Proc. ICREWEMBO Workshop Puuiu, September 1977 pp. 45-56.

RETROVIRUSES AND CANCER GENES

31

Neel, B. G., Hayward, W. S., Robinson, H. L., Fank, J., and Astrin, S. M. (1981). Cell 23, 323-334. Neil, J. C., Ghysdael, J., and Vogt, P. K. (1981). Virology 109,223-228. 018, A., Linemeyer, D., Ruscetti, S., Lowe, R., Lowy, D. R., and Scolnick, E. (1980). J . Virol 35,924-936. Oppermann, H., Levinson, A. D., Varmus, H. E., Levintow, L., and Bishop, J. M. (1979). Proc. Natl. Acad. Sci. U.S.A. 76, 1804- 1808. Oppermann, H., Levinson, A. D., and Varmus, H. E. (1981). Virology 108,47-70. Oskarsson, M., McClements, W. L., Blair, D. G., Maizel, J. V., and Vande Woude, G. F. (1980). Science 207, 1222-1224. Padgett, T. G., Stubblefield, E., and Varmus, H. E. (1977). Cell 10,649-657. Parker, R. C., Varmus, H . E., and Bishop, J. M. (1981). Proc. Natl. Acad. Sci. U.S.A.,in press. Parodi, S., and Brambilla, G. (1977). Mutat. Res. 47, 53-74. Payne, G. S., Courtneidge, S. A., Crittenden, L. B., Fadley, A. M., Bishop, J. M., and Varmus, H . E . (1981a). Cell23,311-322. Payne, G . S., Bishop, J. M., and Varmus, H. E. (1981b). Nature (London) 295,209-213. Pimentel, F. (1979). Biochim. Biophys. Acta 560, 1669-216. Prozig, K. J., Robbins, K. C., and Aaronson, S . A. (1979). Cell 16,875-884. Prehn, R. T. (1976). Adu. Cancer Res. 23,203-236. Rapp, U. R., and Todaro, G . J. (1978). Science 201,821-823. Rapp, U . R.,and Todaro, G . J. (1980). Proc. Natl. Acad. Sci. U S A . 77,624-628. Rasheed, S . , Gardner, M. R.,and Huebner, R. J. (1978). Proc. Natl. Acad. Sci. U.S.A. 75, 2972-2976. Reynolds, F. R., Jr., Van de Ven, W. J. M., and Stephenson, J. R. (1980). J . Biol. Chem. 255, 11040-11048. Robinson, H. L. (1978). Curr. Top. Microbiol. Immunol. 83, 1-36. Rohrschneider, L. R., Eisenman, R. N., and Leitch, G . R. (1979). Proc. Natl. Acad. Sci. U.S.A. 76,4479-4483. Rous, P. (1911). J. Exp. Med. 13,397-411. Roussel, M., Saule, S., Lagrou, C., Rommens, C., Beug, H., Graf, T., and Stehelin, D. (1979). Nature (London) 281,452-455. Rowe, W. P. (1973). Cancer Res. 33,3061-3068. Rubin, C. S., and Rosen, 0. M. (1975). Annu. Rev. Biochem. 44,831-887. Rubin, H. (1980). JNCI J. Natl. Cancer Insr. 64,995-1000. Scolnick, E. M., andParks, W. P. (1974). J. Virol. 13, 1211-1219. Scolnick, E. M., Rands, E., Williams, D., and Parks, W. P. (1973). J . Virol. 12,458-463. Scolnick, E. M., Howk, R. S., Anisowicz, A., Peebles, P. T., Scher, C. D., and Parks, W. P. (1975). Proc. Natl. Acad. Sci. U.S.A. 72,4650-4654. Scolnick, E. M., Papageorge, A. G., and Shih, T. Y.(1979). Proc. Natl. Acad. Sci. U.S.A. 76, 5355-5359. Scolnick, E. M., Weeks, M. O., Shih, T. Y.,Ruscetti, S. K., and Dexter, T. M. (1981). Mol. Cell. Biol. 1,66-74. Sefton, B. M., Hunter, T., and Beemon, K. (1980). Proc. Natl. Acad. Sci.U.S.A.77,2059-2063. Sefton, B. M., Hunter, T., Ball, E. H., and Singer, S. J. (1981). Cell 24, 165-174. Shalloway, D., Zelenetz, A. D., and Cooper, G. M. (1981). Ce1124, 531-542. Sheiness, D., and Bishop, J. M. (1979). J. Virol. 31, 514-521. Sheiness, D. K., Hughes, S. H., Varmus, H. E., Stubblefield, E., and Bishop, J. M. (1980). Virology 105,415-424. Sheiness, D., Vennstrom, D., and Bishop, J. M. (1981). Cell 23, 291-300. Shibuya, M., Hanafusa, T., Hanafusa, H., and Stephenson, J. R. (1980). Proc. Null. Acad. Sci. U.S.A. 77,6536-6540.

32

J . MICHAEL BISHOP

Shih, C., Shilo, B.-2.. Goldfarb, M. P., Dannenberg, A., and Weinberg, R . A. (1979a). Proc. Natl. A d . Sci. U.S.A. 76,5714-5718. Shih, T. Y., Weeks, M. O . ,Young, H. A,, and Scolnick, E. M. (1979b). Virology 96,64-79. Shih,T. Y., Papageorge, A. G., Stokes, P. E., Weeks, M. O . ,and Scolnick, E. M. (1980). Narure (London) 287,686-691. Shilo, B.-Z., and Weinberg, R . A. (1981). Nature (London) 289,607-609. Souza, L. M., Strornrner, J. N., Hilyard, R. L., Komaromy, M . C., and Baluda, M. A. (1980). Proc. Natl. Acud. Sci. U . S . A .77,5177-5181. Spector, D., Baker, B., Varmus, H. E., and Bishop, J. M. (1978a). Cell 13,381-386. Spector, D., Smith, K., Padgett, T., McCombe, P., Roulland-Dussoix, D., Moscovici, C., Varrnus, H. E., and Bishop, J . M. (1978b). Cell. 13,371-379. Spector, D., Varmus, H . E., and Bishop, J . M. (1978~).Proc. Nail. Acad. Sci. U.S.A. 75,41024106.

Stehelin, D., Guntaka. R. V., Varmus, H. E., and Bishop, J. M . (1976a). J . Mol. Biol. 101, 349-365.

Stehelin, D., Varmus, H. E., Bishop, J. M., and Vogt, P. K. (1976b). Nature (London) 260, 170173.

Szmuness, W. (1978). In “Progress in Medical Virology” (J. L. Melnick, ed.), pp. 40-69. Karger, Basel. Temin, H. M. (1974). Annu. Rev. Genet. 8, 155- 178. Ternin, H. M. (1980). Cell 21,599-600. Tereba,A., Lai, M. M. C..and Murti, K . G. (1979). Proc. Nail. Acad. Sci. U.S.A.76,6486-6490. Todaro, G. J . , and Huebner, R. J. (1972). Proc. Nail. Acad. Sci. U.S.A. 69, 1009-1015. Van Beveren, C., Gallesshaw, J. A., Jonas, V., Berns, A. J. M., Doolittle, R. F., Donoghue, D. J., and Verma, 1. M. (1981). N a m e (London) 289,258-262. Vande Von, W. J. M., Reynolds, F. H., Jr., and Stephenson, J. R. (1980). Virology 101,185- 197. Vigne, R., Neil, J. C., Breitman, M. L., and Vogt, P. K . (1980). Virology 105, 71-85. Vogt, P. K. (1971). Virology 46,939-946. Wang, L. H., Halpern, C. C., Nadel, M., and Hanafusa, H. (1978). Proc. Nail. Acad. Sci. U.S.A. 75,5812-5816.

Wang, L. H., Moscovici, C., Karess, R. E., and Hanafusa, H. (1979). J . Virol. 32,546-556. Wang, S. Y., Hayward, W. S . , and Hanafusa, H., (1977). J . Virol. 24,64-73. Watts. S. L., and Smith, R. E. (1980). Infect. fmmun. 27,501-512. Witte, 0 .N., Rosenberg, N . E., and Baltimore, D. (1979). Narure (London) 281,396-397. Witte, 0 .N . , Dasqupta, A,, and Baltimore, D . (1980). Nature (London) 283,826-831. Wang,T. C., and Lai, M . M. C. (1981). Virology 111,289-293. Young, H. A., Shih, T . Y., Scolnick, E. M., Rasheed, S., and Gardner, M. B. (1979). Proc. Nail. Acud. Sci. U.S.A. 76,3523-3527. Young, H. A., Rasheed, S., Sowder, R., Benton, C. V., and Henderson, L. E. (1981). J . Virol. 38.286-293.

CANCER, GENES, AND DEVELOPMENT: THE Drosophila CASE Elisabeth Gateff Biologisches lnstitut I (Zoologic), Albert-Ludwigs Universitat. Freiburg, Federal Republic of Germany

1. Introduction .

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

11. Drosophilu De .............................................. 111. General Information on the Drosophilu Tumor Mutants . . .

IV. Description of the Drosophilu Tumor Mutants . . . . . A. Embryonic Tumor Mutants .... B . Larval Tumor Mutants ...................... V. Viruses found in Drosophilu Tumor Cells . . . . . . . . . . ........... V1. Retroviral Oncogenes and Their Cellular Counterparts in Different Animal Species . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V11. Transfection and Vertebrate Tumor Genes: A Comparison with Drosophilu . . . . . VIII. Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

33 34

62 66

I. Introduction

At the outset of this article I wish to express my strong conviction, based on my experience with Drosophila tumors of genetic origin (Gateff, 1978a,b,c), that in all instances cancer has a genetic base and that cancer is a problem of cell differentiation and the maintenance of this state. In Drosophila, 25 of a total of 5000 genes have been found to cause malignant or benign neoplasms in the mutated state. These genes play apparently crucial roles in the differentiation of specific cell types and tissues. Thus in order to understand the primary event(s) leading to malignant neoplastic transformation, we will have to gain insight into the genetics of normal development. The recent revival of interest in genetic factors as causative agents in cancer has been stimulated by findings that on the genome of a wide variety of vertebrates, including humans, homologous sequences to at least a dozen retroviral oncogenes exist (Bishop, 1981;Wong-Staal et al., 1981;Vennstrom and Bishop, 1982; Barnekow et al., 1982). Shilo and Weinberg (1981b) detected on the Drosophila genome, sequences homologous to five acute RNA tumor virus oncogenes. Moreover, evidence that the genomes of different mammals contain genes capable of transforming NIH-3T3 mouse cells stems from transfection studies with DNA from spontaneously, virally, and chemically transformed cells (Shih et al., 1979), and DNA from normal cells (Cooper et al., 1980). 33 Copyright Q 1982 by Academlc Press, Inc. ADVANCES 1N CANCER RESEARCH, VOL. 37

All rights of reproduction in any form reserved. ISBN 0-12-006637-8

34

ELlSABETH GATEFF

The genetic base for the melanomas and other malignant tumors in platyfish/swordtail hybrids are by now well established (Anders and Anders, 1978). Finally, it has been known for many years that certain human cancers are associated with gene mutations (Knudson, 1978; Knudson and Meadows, 1978; Lynch, 1976). Based on all these findings, Bishop (1981) proposed that on the genome of vertebrates a family of genes exists whose members control specific differentiation steps during development. However, when mutated or abnormally regulated, these genes lead to malignant neoplastic transformation. The Drosophilu tumor system has shown exactly this for quite some time (Gateff and Schneiderman, 1967, 1969; Gateff, 1978a,b,c). In this system, genes responsible for the proper differentiation of a specific cell type, when mutated, give rise to benign or malignant neoplasms. In the light of the aforementioned exciting developments, it may well turn out that Drosophilu will not remain just a “case” but will help to unravel the function of these developmental genes during normal development and in malignant neoplastic transformation. Maybe the unifying principle in malignant neoplastic transformation can be shown to be a gene family with important, basic functions during development, which for that reason has been conserved throughout evolution in all animal systems from humans down to the sponges. This article describes the different benign and malignant neoplasms found in 25 Drosophilu mutants and discusses these mutants in the light of the new findings just mentioned. II. Drosophila Development

Drosophila development involves one embryonic, three larval, one pupal, and one adult stage. Except for the early part of embryonic life, at all other developmental stages two cell systems exist: (1) terminally differentiated cells incapable of dividing and (2) dividing cells. In view of the well-established fact that only dividing cells express and propagate the neoplastic state, this implies that only the dividing cells can be expected to grow as tumors at any developmental stage. During the first part of embryonic life, all cells divide and thus can all become neoplastically transformed. Thereafter, the majority of cells stop dividing and differentiate into larval structures. The larva grows after hatching exclusively by cell enlargement, polytenization, and polyploidization ; as a result, in larval cells no tumorous growth can ensue. Tumors originating from larval cells proper can be expected during the first half of embryonic life, when the cells determined for larval cell lineages are still dividing.

CANCER, GENES, AND DEVELOPMENT

35

In the embryo, in addition to the large majority of differentiating larval cells, small nests of cells are set aside that remain diploid and capable of dividing. These nests of cells represent the primordia for the adult integument, the so-called imaginal discs, the presumptive adult optic centers of the fly in the larval brain, the primordial gonial cells in the female and male gonads, and the blood cells. From this it follows that in the larva, tumorous growth can develop only in the adult primordia as just listed. During adult development in the pupa, oogonia, and spermatogonia, myoblasts, blood cells, and a variety of primordial imaginal cells-from which the various adult organs arise-divide and are candidates for neoplastic transformation. Finally, in the adult only gonial cells and blood cells divide, thus they can be expected to become neoplastic. With this background in mind I will proceed to describe the various tumor mutants. 111. General Information on the Drosophila Tumor Mutants

At present 25 tumor mutants located on three of the four Drosophila chromosomes are known (Table I). They originated either spontaneously or after mutagenesis with ethyl methane sulfonate (Gateff, 1978a,b,c); the latter are indicated in Table I by an asterisk. All mutants are recessive and, with the exception of the gonial cell mutants, lethal. The lethality falls either at the end of embryonic life or at the time of the larval-pupal transformation. Apart from Notch*, which represents a deficiency, all remaining mutants are considered single gene mutations by classical genetic criteria. Future gene cloning and sequencing, however, may show small deletions or insertions of transposable elements within one or the other gene. The individual lethal mutant alleles are propagated in the heterozygous balanced state (for the genetic composition of the balancer chromosomes, see Lindsley and Grell, 1968), whereas the homozygous or hemizygous animals express the tumorous condition. The strains breed through and show full penetrance of the tumorous condition (see later). Depending on the developmental stage at which the tumor becomes obvious, embryonic, larval, and adult tumors can be distinguished (see earlier). Based on the cell type(s) or the tissue(s) affected by the mutation, we recognize in the embryo mixed tumors; in the larva, malignant neuroblastomas, intermediate imaginal disc neoplasms with a compact and invasive mode of growth, and malignant blood cell neoplasms; and in the adult, benign gonial cell neoplasms (Table I). In all mutants, tumorous growth interrupts normal differentiation at a specific developmental stage of a particular cell type(s), which results in unrestrained, mostly lethal neoplastic growth (apart from gonial cell neoplasms; Table I). The following criteria

36

ELISABETH CATEFF

TABLE I Drosophila melanogasier MUTANTS SHOW~NG MALIGNANT AND BENIGN NEOPLASMS Tumor mutants on chromosome number Designation of neoplasm Embryonic tumors Malignant neuroblastorna

Intermediate imaginal disc neoplasm with compact lethal mode of growth

Intermediate imaginal disc neoplasm with invasive, lethal mode of growth Malignant blood cell neoplasm

Benign gonial cell neop1asm

I

I1

DJ"(l)Not~h.1-3.0"; shibire'"- 1-52.2' lerhal( 1 )2,-'*" ; lerhal(l)2269,-'."

111

',

Fourteen alleles of lethal(2) giant larvae; Df(2)lgl net, 2-00'.''; lerhal(2)1542,-c~" lethal(1)2,-'*"; Fourteen alleles of leiha/(I )2269,-'1° ; lethal(2)giantlarvae lethal(1)disc large-1, (two o f them ts'; 1-36'."; lerhal(1)benign Df(2)lgl net, 2-OOj; wing imaginal disc lethal(2)giant neoplasm, 1-34'; disc,-'.'; lerhal lerhal(l)lpr-2, (2)1542,-'.' 1-36.2'*"; shl'"', 1-52.2" lerhal(1)disc large-I, 1-36dJ. lerhal(1)benign imaginal disc neoplasm, 1-34"; lerhal(1)disc. large-2, 1-24.9" lethal( I)malignant lethal(2)malignant blood neoplasm, 1-39"; blood neoplasm,-'*" tumorous-lethal, 1- 34.5/."

benign(2)gonial cell Two alleles of./emale neoplasm, between sterile(l)231,in D f BI 38L-B170RB;,female 13.0 and 48.5'; sterile(l)l621, 1-1 1.7h; female srerile(2) fused, 1-59.5' of Bridges, 2-5" ; ( I 0 alleles) 5 alleles of narrow, 2-83"

lethal(3)giant larvae,-'*"; lethal(3)brain tumoJE-I,-c~" lethal(3)giant larvae,-'*"; lelhal(3)brain tumor's,-c*o

lethal(3)malignant blood neoplasmI , between h-th"; lethal(3)malignanr blood neopIasm-2.-'~" -

' Mohr (1919); Gateff and Schneiderman (1974); * Grigliatti et al. (1973); MacMorrisSwansonandPoodry(l981); Williams(i981);'-, notlocated;d Kisseral. (1978);'Stewart eta/. (1972); Convin and Hanratty (1976); Gans er al. (1975); King (1979); Gollin and King (1981); King er al. (1957); 12 I(2)gl alleles and Df(2) Iglnet Golubovsky and Sokolova (1973); Gateff eral. (1977); Ir Bryant and Schubiger (1971); ' W. P. Hanratty(persona1 communication); Koch andKing(1964);" Kingand Bodenstein(1965)." Mutantsisolated byGateff(l978a,b,c).

'

J

CANCER, GENES, AND DEVELOPMENT

37

were applied to characterize the neoplasms: (1) autonomous and lethal growth in situ as well as after transplantation, (2) loss of the capacity for differentiation, (3) invasiveness, and (4) loss of structure. In some cases the capacity to grow in vitro was used as an additional criterion (Gateff, 1978a). The following section deals with the various tumor mutants.

IV. Description of the Drosophila Tumor Mutants

A. EMBRYONIC TUMORMUTANTS A systematic search for embryonic lethal mutants causing tumorous growth has not yet been undertaken. However, two X-linked mutants exist that show malignant neoplastic growth. These are the Notch' and shibire'"' (shi'"') mutants (Table I). The Notch' mutant, known since 1919, affects the wing morphology in the heterozygous state. In the hemizygous male embryo the ventral ectoderm fails to differentiate (Poulson, 1937, 1940, 1945; for normal embryonic development, see Poulson, 1950; Fullilove and Jacobson, 1978). Instead, neuroblasts are formed exclusively, which differentiate into nervous structures. This mode of growth is nonautonomous when tested in the wild type (Gehring, 1973). However, some implants derived from the anterior portion of older Notch' embryos yielded malignant neoplastic growth after implantation and growth in wild-type female adults (Gateff and Schneiderman, 1974). The lethal growth was propagated for 14 transfer generations before being discontinued. The transplantable tumor consisted of a variety of cells that resembled neuroblasts, glial cells, ganglion-mother cells, and blood cells. Thus it appears that the Notch' tumor originated from undifferentiated cells belonging to two germ layers, the ectoderm (neuroblasts, glial cells) and the mesoderm (blood cells). Lehmann et al. (1981) have studied further mutants that affect early neurogenesis and exhibit phenotypes similar to that of Notch'. It remains to be shown whether they too exhibit tumorous growth. The interest in this locus has extended into the molecular realm. S. Artavanis-Tsakanas (personal communication) has obtained DNA sequences that appear to derive from the Notch' locus. However, the final proof is not yet available. The second embryonic mutant shP' is well known for its various effects on embryonic and larval development (Poodry et al., 1973). Recently, MacMorris-Swanson and Poodry (1 98 l), demonstrated that heat pulses of 4 to 6 hr given between 3.5 and 14 hr of embryonic development resulted in

FIG. 1. Comparisons of wild-type and mutant whole-mount and histological preparations. (a) Whole mounts of fully grown third instar wild-type (A) and l(2)g/4(B) larvae. Note the bloated and transparent appearance of the mutant larva and the region of the tumorous brain and imaginal discs located between the two arrows. (b) Whole mounts of mature wild-type (A) and /(2)gI4 (B) brain (b) and ventral ganglion (vg) with associated imaginal discs (i). Compare the shape of the wild-type imaginal discs with the clumped and shapeless tumorous l(2)ql4 imaginal discs; note further the enlarged and deformed /(2)gp brain hemispheres and the smaller, round wild-type brain hemispheres. (c) Frontal histological section through the two wild-type brain hemispheres (b) and the ventral ganglion (vg). The black lines show the border between the larval portion of the brain (I) and the developing presumptive adult optic centers (aoc). Within the adult optic centers one recognizes the inner and outer formation centers (ifc, ofc), made up of large neuroblasts, and in between the developing outer, middle, and

CANCER, GENES, AND DEVELOPMENT

39

disorganized growth and lack of differentiation in the embryo. The heattreated lethal shi‘“’ embryos represent masses of cells that, after transplantation into wild-type female flies, grow into invasive, lethal tumors. The tumor has been subcultured for several transfer generations. However, it is not yet determined which cell types are transformed. Nevertheless, the shP’ mutant shows that between 3.5 and 14 hr of embryonic development the gene is active and responsible for molecular events that apparently have an impact on the differentiation and morphogenesis of one or more cell types or tissues in the embryo. It further shows that the shitsl gene product is, if made at all, irreversibly damaged at the restrictive temperature.

B. LARVAL TUMORMUTANTS

1. Mutants Developing Madignant Neuroblastomas Table I shows seven mutants that, during the last larval instar, develop malignant neuroblastomas in the presumptive adult optic centers of the larval brain. As a result the animals die either before or at the larval-pupal transformation. The mutant larvae can easily be recognized by their larger size and their more or less transparent bodies (Fig. la). Anatomical investigations show an overgrown central nervous system and, in some of the mutants, tumorous imaginal discs as well (Fig. 1 b). The wild-type larval central nervous system consists of two brain hemispheres and a ventral ganglion (Fig. 1b). Associated with the central nervous system are three pairs of imaginal discs, the eye antennae, and the first and second leg discs (Fig. 1b,A). It was mentioned earlier (Section 11) that in the embryo two developmental systems become determined : the larval,

inner optic glomeruli (og, mg, ig), indicated by their respective neuropiles (n) (arrows). f. Fat body; in, integument; 0, oesophagus; rg, ring gland; s, salivary gland. (d) Frontal histological section through the two brain hemispheres (b) and the ventral ganglion (vg) of a l(Z)g14 larva. The brain hemispheres are enlarged, deformed, and have lost the wild-type organization. Tumorous optic neuroblasts (nb) and ganglion-mother cells (gm)are seen in abnormal array. They have invaded the larval portion of the brain where remnants of the neuropile (n) can be recognized. The ventral ganglion, in contrast, shows normal arrangement of the neuropile (n) and the cellular cortex. f, Fat body; s, salivary gland. (e) Wild-type host in the premortal stage which had been injected with a small piece of 4 2 ) g P brain (A). Compare the “bloated phenotype” with the phenotype of the noninjected wild-type control fly (B). (f) Histological cross section through the abdomen of a tumor-bearing wild-type host fly. The tumorous neuroblasts and ganglionmother cells (arrows) have invaded and destroyed the ovaries and all other organs except for the gut (9). The tumorous growth is opposing closely the integument (in). All histological preparations 5 pm; Hematoxylin and eosin.

40

ELlSABETH GATEFF

which differentiates; and the presumptive adult, which remains undifferentiated until later during adult development in the pupa, when it forms the adult structures. Thus in the embryo, 13-20 large neuroblasts, located in the posterolateral region of each brain hemisphere, are destined to give rise to the adult optic centers (Gateff, 1978a; Hofbauer, 1979). After the hatching of the larva from the egg, the optic neuroblasts divide equally, giving rise to new neuroblasts. During the second and third larval instars they divide unequally, producing more of their own, and in addition, small optic ganglion-mother cells. The ganglion-mother cells divide an undetermined number of times before differentiating into optic neurons. (For the development of the adult optic centers in the larval brain, see Hofbauer, 1979; and for the entire nervous system see Kankel er a/., 1980.) At the end of larval life, in the posterolateral region of each brain hemisphere one finds the presumptive adult optic centers showing the inner and outer formation centers, consisting of large neuroblasts, and in between the anlagen of the developing outer, middle, and inner optic glomeruli (Fig. lc). After pupariation and during adult development, they differentiate into the optic centers of the fly. In the mutants, this differentiation process is interrupted at the stage of the optic ganglion-mother cell. The inability of this cell type to differentiate into optic neurons causes a perpetual growth and an accumulation of neuroblasts and ganglion-mother cells. As a result the mutant brains are 1.5 to 2 times larger than their wild-type counterparts (Fig. l b ; compare Fig. lc and d). They show little normal organization (Fig. Id). The mutant neuroblasts and ganglion-mother cells invade the cellular cortex and the neuropile of the larval brain portions and destroy their organization. Because in all mutants the brain tumors originate from optic neuroblasts, they are classified as neuroblastomas. Affected by the invasive growth of the optic neuroblasts and ganglionmother cells are also the neurosecretory cells and their axons. The interrupted neurosecretory axonal pathways cause the accumulation of large amounts of neurosecretory granules in the cells (Akai, 1972). Thus the target cells for the neurohormone, e.g., the prothoracic gland cells in the ring gland, receive no information as to when to release ecdysone, which initiates puparium formation. For this reason in all neuroblastoma mutants, puparium formation is either blocked or delayed (Gateff et al., 1977). The tumors developing in situ in the larval brain show typical characteristics of malignant neoplasms, such as continuous fast growth, lack of differentiation, invasiveness, and lethality to the host. This malignant mode of growth is also observed after transplantation. When small pieces of mutant brains are implanted into the body cavity of wild-type female adult flies they behave autonomously, and grow in 7-14 days in an invasive, lethal

CANCER, GENES, AND DEVELOPMENT

41

way. Figure le compares a tumor-bearing host in the premortal stage exhibiting the typical “bloated syndrome” to an uninjected control. Histological preparations of a similar host show that the tumorous cells have invaded and destroyed most of the host tissues (Fig. If). All these findings show that the described tumor type represents a malignant neuroblastoma. The most thoroughly studied of all the seven mutants listed in Table I is the fetlzal(2)giant larvae (l(2)gl) mutant. More than 78 l(2)gl alleles are known at the present time. Four of these have originated spontaneously or were induced in laboratory strains (Gateff and Schneiderman, 1969; Lindsley and Grell, 1968). The discovery of 78 independent alleles in isolated natural populations from the Soviet Union (Golubovsky and Sokolova, 1973; Golubovsky, 1978) demonstrate that the Drosophila tumors are not solely a laboratory phenomenon. Furthermore, the occurrence of a lethal giant larvae mutation in Drosophila hydei (Z. Srdid and H. Gloor, personal communication; SrdiC and Frei, 1980) indicates that on the genomes of other Drosophila species, and maybe also other dipteran species, homologous genes may exist that are instrumental in the differentiation of the optic neuroblast and that in the mutated state cause malignant neuroblastomas. The expression of the mutant phenotype in 14 l(2)gfalleles and a deficiency were studied by Gateff et al. (1977). This study showed large variations in the expression of the neuroblastoma phenotype, and thus of the severity of the tumorous condition. In one of the alleles no neuroblastomas at all were detected, as revealed by the size of the brain and the inability of brain pieces to grow in a neoplastic way in wild-type adult hosts. In the remaining alleles the growth of the adult optic neuroblasts in the wild-type adult environment varied from extremely fast to slow. The time of action of the l(2)gl and f(2)g14 alleles falls around the middle of embryonic life (Gateff and Schneiderman, 1974). Work with temperaturesensitive (ts) 1(2)91alleles confirmed this finding and established, in addition, a second ts period during larval life (Sokolova and Golubovsky, 1978; T. L. Gough, personal communication). Conditional mutants have been of great value for the study of the time of gene activity and the mode of action of the gene product. A total of four ts tumor mutant alleles have been found : (1) the already discussed embryonic tumor mutant shitsl, (2) two l(2)gl alleles (W. P. Hanratty and P. Adler, personal communication), and (3) a new ts tumor mutant causing malignant neuroblastomas and imaginal disc tumors (E. Gateff, unpublished). The two ts l(2)gl alleles will be discussed in Section IV,B,2 because they cause only imaginal disc tumors. The new ts tumor mutant lethal(3)malignant brain tumor‘”-l [l(3)rnbtfS-1 ; see Table I] is located on the third chromosome. At the permissive temperature (22”C), all animals show wild-type development (Fig. 2A, B, and C). Shift-up to the restrictive temperature (29°C) at the

42

ELISABETH GATEFF

onset of development results in brain and imaginal disc tumor in 100% of the animals, and their death at the time of the larval-pupal transformation (Fig. 2A). Shifts to the restrictive temperature during larval life showed deleterious effects on the imaginal disc and the fertility, but no tumorous growth was encountered (Fig. 2a). Thus the ts period for tumor development falls during embryonic life. To pinpoint this period, a series of shift-up experiments were performed at 2-hr intervals during embryonic life (Fig. 2B). They showed that 2-, 4-,6-, 8-, lo-, and 12-hr-old embryos, when shifted to the restrictive temperature and kept there, develop malignant neuroblastomas at the end of larval life (Fig. 2B). Embryos shifted to the restrictive temperature when 14 to 18 hr old develop into larvae, some of which pupariate but die as prepupae caused by the damage to the imaginal discs. Still older embryos treated the same way either die as larvae, prepupae, or pupae. Similar results were obtained in shift-up experiments during the first and second larval instars. Temperature treatment of late, third instar larvae has no influence on adult development, but renders all adults sterile. The sterility

I

Adult anlmala sterlle -t-----,

/I

22°C

I

I

1

Ji

II

2

4

I

I

/+ x----x Animals die as larvae or prepupae X- / h l m a l s die as larvae prepupae or Pupae

/I

--//

22°C

29oc hr 220c AELO

t I t t t t t t t t ! ! ! /'Z&'J' rF1 6

8

10

12

14

16

18

20

22

24

26

28

In all cases normal adult development

162 Pupariation

FIG.2. Shift-up (arrow u p ; A and B) and shift-down (arrow down; C) experiments, performed with the ts mutant lerhal(3)brain tumofs--I (43)bfs-1).In the shift-up experiments (A and B), adults were allowed to lay eggs for 2 hr at the permissive temperature (22'C), after which the adults were removed. The eggs and larvae, respectively, were shifted to the restrictive temperature a t the indicated hours after egg laying (hr AEL). In the shift-down experiments (C), eggs were collected as in A and B, at 2 hr intervals at the restrictive temperature, and were brought down to the permissive temperature at various hr AEL. (For further details see text.)

CANCER, GENES, AND DEVELOPMENT

43

is caused by tumorous egg chambers (see Section IV,C) and arrest of spermatide differentiation. The determination for tumor development thus falls between 0 and 12 hr of embryonic development. In order to test whether the gene product was made at the restrictive temperature, a series of shift-down experiments were performed during embryonic and larval life (Fig. 2C). These experiments showed that in all shiftdown experiments the mutant effect was abolished, which suggests that at the restrictive temperature the gene product was made, but was not functional. However, at the permissive temperature, the gene product regains its function and normal development ensues. The foregoing experiments allow the following conclusions. (1) The 1(3)mbt"-I gene product acts in various cell types at different developmental stages in a specific manner. (2) The determination for tumor development takes place during the first 12 hr of embryonic life. (3) At the permissive temperature (22"C), the wild-type 1(3)mblr"-1allele synthesizes the gene product during the first 12 hr of embryonic life in the adult optic neuroblasts and the presumptive imaginal disc cells (see Section IV,B,2). (4) At the restrictive temperature (29"C), the 1(3)mbtrS-1gene product is also synthesized, but it is nonfunctional, possibly because of a change in its tertiary structure. (5) The nonfunctional gene product can be rendered functional again by a shift-down to the permissive temperature (22°C) at any time during larval development. (5) The shift-up and shift-down experiments show further that for normal development to proceed, the functional gene product synthesized during the early part of embryonic life has to be present throughout the second part of embryonic and most of larval life. 2. Mutants Developing Imaginal Disc Tumors Twelve imaginal disc tumor mutants have been isolated (Table I). All tumors show intermediate qualities between truly benign and malignant neoplasms. Most of them grow in a compact, lethal manner (Fig. 3c); however, a small number also show invasiveness in situ (Fig. 3d; Table I). The imaginal discs are the primordia for the different parts of the adult integument, such as the head or the thorax with all its appendages. They become determined quite early in the embryo as small groups of 10-50 cells located in distinct regions of the blastoderm. Throughout embryonic life they remain mitotically quiescent and begin to grow only after the larva hatches from the egg. The imaginal discs attain their final size and shape at the end of larval life and differentiate during adult development in the pupa into different adult integumental structures: For further details on wildtype imaginal disc development and pattern formation, consult Gehring (1978), Bryant (1978), and Poodry (1980).

FIG. 3. Whole-mount, histological, and ultrastructural preparations of wild-type and tumorous imaginal discs from mature third instar larvae. (a) Whole mount of wild-type wing (wid), third leg (lid), and haltere imaginal disc (hid), connected to the trachea, t (A) and the /(/)d.ly-2counterparts. which are clumped and fused (B). (b) Wild-type wing imaginal disc showing a folded epithelium and the cells arranged in monolayers. (c) I(l)bwn wing imaginal disc lacking the typical folding pattern of the wild-type wing imaginal discs (compare with b). Instead, the tissue is made u p of regions consisting of small folds and regions with compact cellular arrangement (arrow). (d) /(l)bw,i brain (b)and ventral ganglion complex (vg), showing regions of imaginal disc (i) invasion (arrows) into the neuropile (n) and the cellular cortex. (e) Cross section through the abdomen of a wild-type host fly bearing a compactly growing tumor (midn, arrow), derived from an implanted piece of a l(Z)y/4 imaginal disc. The ovaries

CANCER, GENES, AND DEVELOPMENT

45

The tumorous pattern of growth of the mutant imaginal discs manifests itself during the last larval instar. The mutant imaginal discs are made up of abnormally shaped clumps of cells. (Compare in Fig. 3a the wild-type with the mutant imaginal discs.) Their histology exhibits drastic abnormalities from that of the wild type (Fig. 3b and c). The wild-type imaginal disc epithelium consists of a folded monolayer of columnar cells (Fig. 3b). The apical secretory surfaces of the elongated cells exhibit microvilli (Fig. 30, and the basal surfaces form a basal lamina. Complex junctional complexes connect the cells laterally to each other (Poodry, 1980). The mutant imaginal disc epithelium shows very little of the aforementioned organization. The cells either are clustered or form small folds (Fig. 3c), and show less contacts among each other (Fig. 3g). In some of the imaginal disc tumor mutants the imaginal discs invade the larval central nervous system, where they cause the destruction of the cellular cortex and the neuropile (Table I ; Fig. 3d; Gateff, 1978b,c). A main characteristic of malignant growth in general is the capacity of the tumorous cells to grow autonomously after transplantation into a wild-type host. Small pieces of mutant imaginal discs injected into the body cavity of wild-type flies grow in a fast autonomous lethal fashion. All tumors attain a ball-like shape after 7-14 days of growth., and consist to a large extent of cells in clustered arrangements (see previous discussion; Fig. 3e). Such tumorous imaginal disc tissue sublines grow in the adult for numerous transfer generations. In contrast, pieces of wild-type imaginal discs grow in a regulated way and never cause the death of the adult host. A further property of tumorous imaginal discs is their incapacity to differentiate into cuticular patterns when exposed to the hormones at metamorphosis in a wild-type larva that is ready to pupariate. Here they respond to ecdysone by cessation of cell division; however, none of the complex processes of chitin biosynthesis or the secretion of a cuticular pattern ensues. On the contrary, pieces of wild-type imaginal discs differentiate inevitably into disc-specific cuticular patterns. Therefore, the lack of the capacity to differentiate and the disordered, lethal growth in situ as well as after transplantation into a wild-type host allow the classification of the imaginal disc tumors as lethal benign or lethal invasive tumors (Gateff, 1978a,b,c). (ov) on the tumor side are destroyed. ib,Fat body; g, gut; int, integument. ( f ) Fine-structure preparation of a few wild-type imaginal disc cells (idc) in close lateral contact with each other, showing toward the lumen (I) an apical region with microvilli (mv). Within the cytoplasm, vacuoles (v) and low-opacity droplets (Id)can be observed. In one of the nuclei a few virus-like particles (vlp) are seen. (g) Electron micrograph of portion of an /(2)gI4 imaginal disc, showing in the lumen (I) the tumorous cells (mid) with numerous cytoplasmic processes (cp) and invading blood cells (bc). In the upper left are peripodial membrane-like cells (pml). All histological preparations 5 pm thick and stained with hematoxylin and eosin. The ultrastructure preparations, uranylacetate-osmium tetroxide. (Courtesy of Dr. H. Akai).

46

ELlSABETH GATEFF

The oldest and most thoroughly studied of the eight mutants is the 1(2)gl mutant, with its more than 78 alleles and a deficiency. The 1(2)91and 1(2)g14 alleles were the first to be investigated for their neoplastically growing imaginal discs (Gateff and Schneiderman, 1969, 1974). Of the more than 78 alleles found by Golubovsky and Sokolova (1973) and Golubovsky (1978), Gateff et al. (1977) compared the morphology and developmental capacities of 14 alleles and the deficiency. Just as in the case of the neuroblastoma (see Section IV,B, I), so for the imaginal disc tumors found in the different 1(2)91 alleles, large variations in size, morphology, and the degree of malignancy were found. In some alleles the imaginal discs were very large and proved highly malignant in the transplantation test. In other alleles they were smaller or not present at all. This variation in the expression of the tumorous phenotype in the different alleles seems to point to the possibility that the 1(2)91 locus is a complex one. It may be that most alleles are smaller or larger deletions. Thus, depending on the number of deleted gene sequences, a more or less extreme tumor phenotype may ensue. Gateff and Schneiderman (1969, 1974) determined in transplantation experiments the time of the f(2)yP gene activity around the tenth hour of embryonic life. Anterior portions of l(2)g14 embryos aged 10 hr and older, devoid of the brain-ventral ganglion complex, when implanted into the body cavity of wild-type flies grew tumorous imaginal discs. Corresponding anterior integumental portions from wild-type control embryos showed normal imaginal disc growth. With the help of a ts allele, 1(2)91'"', Hanrathy and Gough (1982) demonstrated that the ts period for the differentiation of cuticular structures extends from larval hatching to the early third instar. In addition, the data suggest a possible ts period during embryonic life. Sokolova and Golubovsky (1978) studied the ts period of two l(2)91heteroallelic combinations. In one of the combinations the ts period was diphasic. At 17°C there is an early embryonic and third larval instar ts period. At 29°C a ts period was found during the first larval stage and the second during the prepupal stage. In the second allelic combination only the third larval stage proved to be sensitive to low temperature. The aforementioned studies indicate that the 1(2)91gene function is required for normal disc development and for the maintenance of the differentiated state throughout embryonic and postembryonic life. W. P. Hanratty and P. Alden (personal communication) studied the developmental capacities of the imaginal discs of the I(2)gl'" allele in a heteroallelic combination with the 1(2)91 deficiency. At the restrictive temperature, imaginal discs of normal as well as of tumorous morphology were found. Brain tumors seem not to occur in this allele. Implanted into wildtype adult hosts, the mutant imaginal discs with tumor morphology proved fully committed to neoplastic growth irrespective of the rearing temperature.

CANCER, GENES, AND DEVELOPMENT

47

Similarly, intact f(2)gPs1imaginal discs with normal morphology subcultured in wild-type hosts at the permissive temperature maintained their normal morphology. At the restrictive temperature, however, they gave rise to lethal neoplasms. Moreover, when mutant discs of normal morphology were cut into two pieces and the pieces were subcultured at the permissive temperature, in contrast to intact imaginal discs, they grew in a lethal neoplastic way. This indicates that the information for neoplastic development is present in both imaginal disc fragments. The f(2)gP' allele shows that the frequency of abnormal imaginal discs is directly related to the length of the exposure to the restrictive temperature. The mutant allele is, however, leaky because the animals, continuously grown at the restrictive temperature, show tumorous imaginal discs in 69.4%, whereas the remainder exhibit normal morphology and developmental capacities. The new autosomal lethal mutant l(3)bt'"-1 (E. Gateff, unpublished ; Table I ; Fig. 2) discussed already in Section IV,B,l is fully penetrant at the restrictive as well as the permissive temperature. It develops imaginal disc tumors in addition to neuroblastomas in the brain, and shows, like the previously discussed 1(2)91'"' allele, imaginal discs with normal and tumorous morphology and developmental capacities. The shift-up and shift-down experiments demonstrate that also for tumor development of the imaginal discs, the ts period falls during the first half of embryonic life. The shift-up and shift-down experiments reveal that in order for normal development to ensue, the gene product-which apparently is synthesized in the optic neuroblasts and the imaginal disc primordial cells during early embryogenesis is needed continuously for the rest of embryonic and larval life, up until the early third instar. Constant exposure to the restrictive temperature renders the gene product nonfunctional, and thus the development of tumors in the brain and the imaginal discs ensues. The f(3)bt"-1 gene product is, however, synthesized not only during early embryogenesis. It must also be synthesized in the neuroblasts and the imaginal disc cells during larval life because shifts-up during large parts of larval life cause lethality in the pupa and during adult development. Anatomical investigation shows morphological abnormalities but no tumors in the imaginal discs and in the optic centers of the brain. For the 1(2)91'"' allele, similar conclusions were drawn (see previous discussion). Maybe in all these mutants the gene product is synthesized at different times during embryonic and postembryonic development, and may be the presence of the gene product in the optic neuroblasts and the imaginal disc cells is an essential prerequisite for all following developmental steps up until the early third larval instar. Gene cloning experiments of the 1(2)gl gene that are now under way in a few laboratories (0.Schmidt, personal communication) should give, in a

FIG.4. Whole mounts and histological preparations of mutant blood cells, hematopoietic organs, and tumor-bearing hosts. (a) Schematic drawing of wild-type (A) and 4l)mbn (B) brain (b) and ventral ganglion (vg) with the dorsal heart vessel (h) and the attached normal (ho) and tumorous hematopoietic organs (tho). Note the enlarged, tumorous, hematopoietic organs in B. (b) Longitudinal histological section through the first hematopoietic lobe of a fully grown wild-type larva exhibiting primordial blood cells. (c) Longitudinal histological section through the first hematopoietic lobe from a mature /(Z)rnbn larva. Note the size difference between the wild-type and mutant hematopoietic lobes and the cells in the /(l)mbn hematopoietic organ. (d) A group of /(l)rnbn blood cells from the hemolymph showing extensive cytoplasmic processes. c, Crystal cell; I, lamellocyte; p, plasmatocyte. (e) Histological section through secondary nests of hematopoiesis (arrows) in the body cavity of a 43)rnbn mature larva consisting of proplasmatocytes. ( f ) Phase contrast of /(3)mbn plasmatocytes @) and giant lamellocytes (I)

CANCER, GENES,

AND DEVELOPMENT

49

few years from now, answers to some of the questions concerned with the role of the genes described here during normal development and in malignant neoplastic transformation. The ts mutant s W 1 (Table I) shows extensive pleiotropic effects on development (Poodry el al., 1973; MacMorris-Swanson and Poodry, 1981). In contrast to the tumor developing in the shitsl embryo after exposure to high temperature (see Section III), tumorous imaginal discs in situ are not observed after restrictive temperature treatment. However, tumorous growth ensues during the first subculture of the eye-antenna1 disc in the wild-type adult abdomen at the restrictive temperature (Williams, 1981). Wild-type imaginal discs, in contrast, never show tumorous growth in a comparable transplantation experiment. The tumorous growth pattern of the implanted disc is established during the first 48 hr of in uivo subculture. Thereafter, the tumorous state is irreversible and restrictive culture conditions are not required for its maintenance during further transfer generations. Apparently in the larva an external factor or factors are present that prevent tumor development, even though the eye-antenna1 disc cells are competent for neoplastic growth, which may not exist in the adult. In this sense the tumorous growth of the imaginal discs in uivo is not autonomous. However, once established, tumorous growth ensues autonomously at the permissive temperature as well.

3. Mutants Developing Hematopoietic Tumors The third type of tumor found in the larva develops in the hematopoietic organs and involves the primary blood cells. Five such mutants are known at present (Table I). A brief outline of the hematopoiesis in the wild-type larva will precede the description of the mutants. The wild-type larval hematopoietic organs, the so-called lymph glands, represent four to seven pairs of lobes located along the dorsal heart vessel (Stark and Marshall, 1930; Gateff, 1977; Fig. 4a). The two blood cell types found in the hemolymph, the plasmatocytes and crystal cells (Rizki, 1957), originate, respectively, from proplasmatocytes and procrystal cells located in the hematopoietic organs (Gateff, 1977; Shrestha, 1979; Shrestha and

from the hemolymph. (g) Longitudinal section through l(3)mbn larva in the premortal stage, filled with millions of blood cells, which have destroyed most of the organs. int, Integument; g, gut. (h) Histological cross section through the abdomen of a wild-type fly that had been injected with a small piece of a l(l)mbrr hematopoietic tumor. The ovary on the left side has been replaced almost completely by tumorous blood cells, whereas the right ovary (ov) is still intact. The only egg chamber (e) on the left is surrounded and invaded by the tumorous blood cells (arrows). f, Fat body; g, gut; int, integument. All histological sections 5 pm thick and stained with hematoxylin and eosin (further details in text).

50

ELISABETH GATEFF

Gateff, 1982a; Rizki et al., 1980). The differentiation of plasmatocytes into plasmatocytes and further into podocytes and lamellocytes is closely related to (1) the formation of the primary lysosomal system and (2) the acquisition of the capacity for melanine biosynthesis (Shrestha, 1979; Shrestha and Gateff, 1982a). Crystal cell maturation, on the other hand, involves the synthesis of the crystalline material, whose molecular nature is not yet established, and crystal formation. While the larva grows, the primordial blood cells divide in the hematopoietic organs. A small number of them are released into the hemolymph at the end of each larval instar. During the larval-pupal transformation, however, all primordial blood cells present in the hematopoietic organs enter the hemolymph, where they play a role in the digestion of the disintegrating larval tissues. Plasmatocytes can take on variable shapes in the hemolymph, which are referred to as podocytes and lamellocytes (Rizki, 1957; Fig. 4d). During defense reactions they phagocytose and encapsulate foreign bodies (Nappi, 1975). The very fragile crystal cells function most probably in hemolymph coagulation but may possess further unknown functions in defense reactions. In all five mutants shown in Table I, the malignantly transformed blood cells are the plasmatocytes, not the crystal cells. Characteristic for the blood tumor mutants are the much enlarged hematopoietic organs (Fig. 4a,b, and c) and the increased blood cell counts in the hemolymph (Gateff, 1978a,b,c; Hanratty and Ryerse, 1981). In contrast to the wild type, where blood cells are released into the hemolymph at particular times and in controlled amounts, in the third instar mutant larvae, blood cells are formed constantly in the hematopoietic organs and are released continuously into the hemolymph. The freed tumorous blood cells invade the organs of the larva, eventually destroying them. This can be seen in Fig. 4g, where in the extreme case of the mutant lethal(3) malignant blood neoplasm-1 (l(3)mbn-I), the body cavity of the larva is entirely filled with blood cells. Most larval organs, such as the muscles, the fat body, and the imaginal discs, have disappeared. Histological sections through l(3)rnbn-1 larvae show secondary nests of hematopoiesis distributed throughout the body cavity (Fig. 4e). The majority of the cells in this mutant are immature and resemble proplasmatocytes. However, 5 to 10% of the free blood cells engage in an abnormal differentiation yielding giant plasmatocytes, podocytes, and lamellocytes (Fig. 4f). Sometimes giant lamellocytes accumulate and melanize, forming melanotic masses. The giant plasmato-, podo-, and lamellocytes show in their cytoplasm enormous amounts of primary and secondary lysosomes, when compared to their wild-type counterparts and the undifferentiated tumorous blood cells (R. Shrestha and E. Gateff, unpublished). Proplasmatocytes are predominantly present within the enlarged l(3)rnbn-1

CANCER, GENES, AND DEVELOPMENT

51

hematopoietic organs. Procrystal and crystal cells are also found but in considerably smaller numbers. The blood tumor mutants designated as f(2)rnbnand f(3)mbn-2(Table I) exhibit a similar tumorous phenotype. Thus it appears that in these mutants the differentiation of proplasmatocytes into plasmato-, podo-, and lamellocytes is largely inhibited or abnormal. As a result proplasmatocytes continue to divide, which causes their considerable accumulation in the hemolymph and the enlargement of the hematopoietic organs. In contrast to the three blood tumor mutants just discussed, in the mutants f(l)mbn(Gateff, 1978a,b,c; Shrestha, 1979; Shrestha and Gateff, 1982b) and Turn’ (Hanratty and Ryerse, 1981;Table I), blood cell differentiation is not impaired. The enlarged mutant hematopoietic organs show, in addition to primordial blood cells (e.g., prohemocytes and proplasmatocytes), also differentiated plasmato-, podo-, and lamellocytes (Shrestha and Gateff, 1982b; compare Figs. 4b and c). During the third larval instar, mature blood cells are continuously released from the hematopoietic organs into the hemolymph, where they invade and destroy the larval tissues. The mutant blood cells exhibit increased amounts of primary and secondary lysosomes and show more phagocytic vacuoles when compared to the wild type (Shrestha and Gateff, 1982b). The malignant qualities of the mutant blood cells can also be demonstrated in the transplantation test. Small pieces of f(l)rnbn or Turn’ hematopoietic organs implanted into the body cavity of wild-type flies grow in an autonomous lethal, invasive fashion for numerous transfer generations (Shrestha and Gateff, 1982b; Hanratty and Ryerse, 1981; Fig. 4h). The I(l)rnbn tumors in uiuo consist predominantly of proplasmatocytes. However, just as in the in siru situation in the hematopoietic organs, one finds differentiated podocytes and lamellocytes interspersed. In contrast to f(l)rnbn,in the Turn’ mutant, extensive melanizations in siru as well as after transplantation into adult hosts can be seen (Hanratty and Ryerse, 1981). This indicates that in this mutant melanin synthesis, a normal repertoire of this blood cell line (see previous discussion), is abnormally regulated. In the wild type, melanization of lamellocytes takes place only in defense reactions after they have encapsulated a foreign body such as a nematode (Nappi, 1975). In this respect the Turn’ mutant resembles melanotic “pseudotumor” mutants in which, for unknown reasons, lamellocytes also encapsulate the fat body or accumulate and subsequently melanize (Sparrow, 1978). Small melanotic “pseudotumors” can also be observed in f(l)rnbn;however, the ability of the lamellocytes to melanize is not expressed after transplantation. The melanotic masses consist of melanized lamellocytes, which are considered dead cells. The capacity for malignant growth after transplantation applies to the last two blood tumor mutants. The

52

ELISABETH GATEFF

tumorous blood cells in the three mutants first discussed do not grow after transplantation in a wild-type adult host. Thus all five blood tumor mutants cause malignant transformation of the plasmatocyte blood cell line, whereas the crystal cells are normal. Furthermore, the individual mutant either affects the capacity for differentiation of the proplasmatocytes into plasmato-, podo-, and lamellocytes (l(2)mbn; I(3)mhn-I; 1(3)rnbn-2), or the mutated genes seem to be involved in the control of the demand for blood cells in the hemolymph and their supply by the hematopoietic organs (I(l)rnhn;Turn'). Dehn (198 1) compared the defense reactions of wild-type I(l)mhn and I(3)rnbn-l blood cells against spores of Aspergillus niger and Methnrrhizium unisopliae. After injection of 30 spores per larva he found a lower mortality rate in the mutants than in the wild-type larvae, which shows that the spores were more efficiently encapsulated by the mutant than by the wild-type lamellocytes. The fact that in both mutants lamellocytes are present in increased amounts may explain this result. Many spores were covered with granulae, pointing to the possibility of a humoral immune response in addition to the encapsulation reaction. Comparing the mortality rates of the two mutants after spore injection, l(l)mbn larvae showed lower mortality rates and thus higher efficiency in rendering the injected spores harmless, than did the I(3)rnbn-l mutant. This finding agrees well with the observation that in I(1)rnbn larvae most free blood cells and many cells in the hematopoietic organs are fully differentiated and apparently functional. In the l(3)nzbn-I mutant, in contrast, the majority of the cells are immature and thus nonfunctional. The availability of large numbers of blood cells in the mutants faciliated the establishment of tumorous blood cell lines in uirro derived from the l(2)rnbn and l(3)rnbn-l mutants. The fine structure of the blood cell lines has been studied by Shrestha (l979), and a variety of properties by Gateff et al. (1980). The cells in uirro show high phagocytic activity during the stationary phase and contain two viruses in addition to the well-known virus-like particle (vlp; see later).

4. Mutants Developing Germ-line Tumors The germ-line precursors of Drosophila, the so-called pole cells, have been already set aside before blastoderm formation. During gastrulation they migrate into the respective female or male gonads, where they become primary oogonia (Mahowald and Kambysellis, 1980) or primary spermatogonia (Lindsley and Tokuyasu, 1980), respectively. During adult development and in the adult, oogonia and spermatogonia differentiate into eggs and sperm, respectively (King, 1970; Mahowald and

53

CANCER, GENES, AND DEVELOPMENT

TABLE I1 MUTANTSCAUSING BENIGNTUMORS IN THE MALEAND FEMALE GERMLINEOF Drosophila melanogaster Designation of mutant Female sterile ( 2 ) of Bridges (,fi(2)E) Tow alleles : female sterile ( 1 ) 231 G ( f i ( 1 )231 G ) and female sterile (1)231 M (js(1)231 M ) Narrow (nw), 5 alleles

Sex

affected

Y ?

0

Fused (,fu), 7 alleles

Y

Female sterile ( 1 ) 1621 (fS(1)1621) lethal(3)brain tumor"-l

P

Benign gonial cell neoplasm (bgcn)

?

9.6

6

Differentiation step affected Early cystocyte division Early cystocyte division

Early cystocyte division Late cystocyte division Late cystocyte division Late cystocyte division Differentiation of cystoblasts into cystocytes and of primary spermatocytes into meiotic spermatocytes Differentiation of primary spermatocytes into meiotic spermatocytes

References Koch and King (1 964) ; King (1970) Cans er al. (1975); King er al. ( 1978); King (1979); King and Buckles (1978); Mohler (1977) King (1970); King and Bodenstein (1965) King er al. (1 957) ; Smith and King (1966); King (1970) Gollin and King (I98 1)

E. Gateff (unpublished) Gateff (1981, 1982)

Lifschytz (1978)

Kambysellis, 1980; Lindsley and Tokuyasu, 1980). Hundreds of sex-specific genes are involved in these differentiation processes (Mohler, 1977; King and Mohler, 1975; Romrell, 1975); mutations in these genes cause either male or female sterility. Genes acting in both the female and the male germ lines are exceedingly rare. In fact, only one such mutant is known (Gateff, 1981, 1982; Table 11). The aforementioned relationship of genes affecting either of the two sexes to genes acting in both sexes is seen also in the mutants causing germ-line tumors. Table I1 shows eight such mutants. Of these, seven affect either the female or the male germ line and only one affects both sexes.

54

ELISABETH GATEFF

Before we proceed with the discussion of the developmental effects of the various tumor mutants on oogenesis, a brief description of wild-type egg maturation should help the reader to understand the steps at which germ-line differentiation is interrupted in the various ovarian tumor mutants. Each of the two Drosophilu ovaries consists of approximately 16 ovarioles. The ovarioles are sausage-like and show distally the germaria and proximally the vitellaria (Fig. 5a). At the distal end of the germaria, two or three oogonia divide and give rise to cystoblasts, which in turn produce in four consecutive mitoses cysts made up of 16 cells. The cells in such cysts are interconnected by a system of ring canals. The cysts become enveloped by follicle cells and enter the vitellaria, where one of the 16 cells develops into the oocyte and the remaining 15 differentiate into nurse cells. For further details on wild-type oogenesis consult King (1970) and Mahowald and Kambysellis (1 980). Table 11 shows the different mutants and the differentiation steps controlled by the various genes. Seven female sterile mutants are listed. Six of them affect cyst formation. In three of the mutants, early cystocyte divisions are abnormal, whereas in the remaining three the late cystocyte divisions are affected. In contrast to the wild type, where cystocyte divisions are incomplete and the two daughter cystocytes remain connected with ring canals, in the mutants a high percentage of the cystocytes show complete cytokinesis, which accounts for the accumulation of free single cystocytes within the egg chambers. However, in some cases cystocytes perform incomplete cytokinesis and remain connected with each other by ring canals. Thus, in addition to the single cystocytes, cysts of two, four, and more cystocytes form. Some of them even differentiate into pseudonurse cells (King, 1970). Young follicles from female sterile mutants implanted into wild-type female hosts grow in the same tumorous fashion as they do in situ (King and Bodenstein, 1965). All ovarian tumors are benign and autonomous. In contrast to the six female sterile mutants just discussed, the recessive mutant benign goniul celZ neoplasm (bgcn) affects an earlier differentiation step, namely, the differentiation of cystoblasts into cystocytes (Gateff, 1981, 1982; Table 11). As a result, cysts and egg chambers never form. The ovaries of newly emerged flies show germaria filled with many single cells (Fig. 5b). Upon rupturing the peritoneal sheath of the germarium in a drop of Ringer's solution, the single cells leave the germarium and in phase contrast closely resemble cystoblasts (Fig. 5c). They exhibit large nuclei with prominent nucleoli and can often be seen in mitosis. As the flies age, they enter the vitellaria, which, in contrast to the wild type, are devoid of follicles (Fig. 5b). Seven days after eclosion the vitellaria contain millions of cystoblastlike cells (Figs. 5d, e, and f). The cells are not connected with each other, and when freed from the vitellaria they show the same phenotype as the cystoblasts in Fig. 5c. Follicle cells are also present. They line the vitellaria and

FIG.5. Histological comparison between the wild-type and bgcn adult gonads. (a) Longitudinal section through two ovarioles from a freshly eclosed wild-type fly. Each ovariole shows a germarium (9) and two young follicles (f) in the vitellarium (v). The ovarioles are surrounded by a peritoneal sheath (t). Each germarium consists of three regions: region 1 contains the mitotically active oogonia ( 0 ) and cystoblasts; region I1 is inhabited by cystocytes (cc) in various stages of cyst formation; region 111 shows a terminal cyst (tc) surrounded by follicle cells (fc). (b) Longitudinal section of two germaria (g) from a freshly eclosed bgcn fly. Note the empty vitellaria (v). delineated by the peritoneal sheath (t) and the germaria (g), filled with loosely arranged cystoblasts. In the posterior region of one of the germaria, follicle cells (fc) can be seen. (c) Phase contrast of the cells derived from a byrn germarium. (d) Whole mount of 7-dayold mutant adult ovary (mov) and 10-hr-old wild-type ovary (wtov). Compare the mutant ovary, devoid of follicles, with the wild-type ovary exhibiting young follicles (arrow). The mutant ovarioles are filled with millions of cystoblast-like cells. (e) Cross section through the abdomen of a 7-day-old bgcn fly showing the two ovaries (ov), each with I5 ovarioles (ovl) embedded in fat body (fb)and surrounding the gut (gt). (f) Detail from (e) showing a few ovarioles in cross section. The ovarioles are lined by a single layer of follicle cells (fc). The entire lumen is filled with tumorous cystoblasts (cb). (g) Longitudinal section through the distal end of a wild-type testis. Various stages of spermatogenesis, such as primary spermatocytes @s) and spermatid bundles (sb) can be observed. (h) Oblique section through bgcn testis showing primary spermatocytes. Histological sections 5 pm; Hematoxylin and eosin.

56

ELISABETH GATEFF

are easily distinguishable from the larger cystoblasts by their smaller size and cuboidal shape (Figs. 5b and f). The cells in the germaria of young bgcn flies were tested for their developmental capacity through the implantation of the germaria into the body cavity of wild-type female flies. After 14 days the implants had grown into spheres 40-50 times larger than the original implants. The enlargement was due to millions of cytoblasts growing in them. In order to test whether the female germ line is already determined for abnormal growth earlier in development, gonads from mid-third-instar bgcn larvae (100 hr old) were tested for their mode of growth by implantation into wild-type larvae of comparable age, which were allowed to develop into adults. These experiments showed that the primary gonial cells from the larval gonad develop autonomously in the wild-type host, giving rise to tumors similar to those seen in situ, consisting exclusively of cystoblasts (Gateff, 1982). These experiments indicate that the bgcn gene must act sometime before the third larval instar. The fact that in this mutant both sexes develop tumors strongly indicates that the bgcn gene must function at or before the time of the segregation of the female and male germ lines. After the entry of the primary germ cells into the prospective gonads, separate female and male specific gene sets seem to be turned on that are instrumental either in oogenesis or spermiogenesis. Before discussion the benign tumors developing in the testis of the autosomal bgcn and the x-linked f(455 mutants (Table II), a short compendium of wild-type spermiogenesis should give the reader a basis for comparison with the mutant phenotype. Wild-type spermiogenesis begins at the apical portion of the testis where pregonial cells divide unequally, giving rise to stem cell and primary spermatogonia. The primary spermatogonia produce, in four successive divisions, cysts of 16 primary spermatocytes, which remain in contact with ring canals similar to the ones in cystocyte cysts of the female (King, 1970). After a growth phase, primary spermatocytes become secondary spermatocytes, which undergo meiosis and yield cysts of 64 interconnected spermatids. The spermatids in a cyst engage synchronously in a complex differentiation process ending with mature sperms (Lindsley and Tokuyasu, 1980). Figure 5g demonstrates some of the differentiation stages in wild-type testes. Compared to the female, where six ovarian tumor mutants have been found, in the male only two such mutants are known (Table 11). In both mutants, the differentiation of the male germ line is interrupted at the early primary spermatocyte stage before the onset of the growth phase. As a result, the primary spermatocytes engage in supernumerary cell divisions resulting in cysts of 300 to 400 primary spermatocytes, some of which seem to remain connected by ring canals (H. Akai and E. Gatef€, unpublished

CANCER, GENES, AND DEVELOPMENT

57

results). Figure 5h is a section through a bgcn testis that shows exclusively primary spermatocytes (compare with Fig. 5g). The male sterile mutants just discussed reveal two nonallelic developmental genes, both of which, in the mutated state, impair the further differentiation of primary spermatocytes. The resulting continuous mitotic activity of the primary spermatocytes leads to tumorous cysts. The two genes act at different points during development, bgcn functioning in both sexes before the segregation of the male from the female germ lines, and /(1)55 only in the male after the segregation event. In the preceding section, concerned with mutations causing tumors in various larval cell types, we discussed four tumor types involving (1) the adult optic neuroblasts, (2) the imaginal discs, (3) the primordial phagocytic blood cells, and (4)the male and female gonial cells. Except for the l(Z)mbn and Tum' blood tumor mutants, in all the other mutants, differentiation of a particular cell type is arrested. Thus the described mutants are developmental mutants, and the mutated genes that cause the tumorous condition are instrumental in controlling cell-specific differentiation steps. The following sections will relate different aspects of the Drosophilu tumors to recent developments in the molecular biology of retroviruses and the identification of putative vertebrate genes involved primarly in malignant neoplasms. V. Viruses Found in Drosophila Tumor Cells

Seven viruses, belonging to various families, have been found in Drosophilu laboratory and natural populations and cell lines in vitro (Brun and Plus, 1980). Of these, two viruses are present in tumorous blood cells cultured in uitro: (1) a reovirus named F, (Gateff et al., 1980; Haars et a/., 1979) and (2) a picornavirus designated Drosophila C virus (DCV; Plus et al., 1975). In addition, in all tumors whose fine structure has been studied, large numbers of virus-like particles (vlp) have been found primarily in the nucleus and less so in the cytoplasm (reviewed by Gateff, 1978a; see later). The two viruses just named almost certainly have no direct relationship to the neoplastic transformation of the blood cells. However, a study by Plus and Golubovsky (1981) of the resistance to DCV in 15 l(2)gl stocks, derived from isolated, natural populations of the Soviet Union (Golubovsky and Sokolova, 1973; Golubovsky, 1978), should be mentioned. DCV is the most common virus of Drosophila laboratory and natural populations. The above l(2)gl strains were found to be free of DCV. Upon infection with the virus, Plus and Golubovsky encountered considerable variation in the resistance to the virus among the different l(2)gl strains. However, on the whole they demonstrated a higher resistance to the virus in the 42)gl strains

58

ELISABETH GATkI-F

than in all other wild-type and mutant strains studied. Finally, the resistance to the virus seems to act through some cytoplasmic factor(s). A comparison of the resistance to DCV with the degree of malignancy in the different l(2)gl alleles (Gateff et al., 1977) revealed no correlations. In physiologically stressed cells, such as aged cells, cells treated with carcinogens or X rays, or mutant cells, virus-like particles (ulps) were regularly found (for review see Gateff, 1978a). They are especially common in tumorous cells in situ, in subculture in uiuo and in uirro (Akai et d., 1967; Shrestha, 1979; Shrestha and Gateff, 1982b). Vlps are round to oval with a diameter of 37 nm, showing a single envelope and a core that is often empty. They occur singly, in groups, or in crystalline array, primarily in the nucleus, and in older or dying cells, also in the cytoplasm. Two recent papers report on particles of 36- and 40-nm size, which resemble ulps morphologically. The first paper, by Scott et al. (1980), describes a new type of virus designated HPS-1. The virus contains a single segment of double-stranded RNA and shows two coat proteins (1 20,000 and 200,000 daltons). It bands at 1.41 gm/cm3 in CsCl. The particle was isolated from the supernatant of Schneider cell line 2, cultured in Dulbecco modified Eagle’s medium supplemented with 10% calf serum and 0.5% lactalbumin hydrolysate. The origin of this particle and its relation to the ulp will have to be demonstrated. The second paper by Heine et al. (1980), shows a 40-nm particle in the cytoplasm of Schneider cell line 1 cells. The particle is found in a subcellular fraction banding at 1.22 gm/ml in sucrose. This fraction was associated with reverse transcriptase activity, high-molecular-weight polyadenylated RNA, and the above particle. The authors concluded that the Drosophilu cells in vitro contain retroviruses, similar to but not identical to A-type particles found in mammalian and avian cells. This conclusion awaits further substantiation. Thus the existence of retroviruses in Drosophilu is not yet established. However, evidence is accumulating that suggests a relationship between the transposable elements of eukaryotes and the integrated retroviral proviruses of mammals and birds (Finnegan et al., 1978; Shimotohno er al., 1980; Sutcliff et al., 1980; van Beveren et ul., 1980; Dhar et al., 1980; Gafner and Phillippsen, 1980; Levis er ul., 1980; Dunsmuir et al., 1980; Scherer and Pirrotta, 1981; Scherer er al., 1982; see also article by Spradling and Rubin, 1981). Copia and other copia-like transposable elements in Drosophilu represent families of moderately repetitive gene sequences with an average length of 5 to 8.5 kb, repeated 10 to 100 times per genome, which code for abundant polyadenylated mRNAs (Table 111). The mRNAs are found predominantly in the nucleus and only a small proportion, as expected, in the cytoplasm as

TABLE 111 CHARACTERISTICS OF SIX Copia-LIKE TRANSP~SABLE ELEMENTS FOUND ON Internal segment

Direct terminal repeat

2

1 (1)DNA sequence of internal segment in kb

Designation of transposable element

5.0"*b

Copia 412 29 7 mdglk mdg3 8104

7.0" 6.5b*f 7.28 5.5h,' 8.7'

(2)Number of base pairs in direct terminal repeat 276' 481' 412/ 30O-40Og 269h*J 429'

3 (3,4)Number of base pairs in terminal inverted repeat 13/17 8/10' 3f ND 15/1gh,' 3'

THE

Drosophila GENOME

Host DNA

4

5 (5)Number of base pairs in flanking sequences at point of insertion" 5d

4' 4f ND 5 5'

Mean number Percentage and classes of POlY(A) mRNA 3% 5 kb, 2 kb 0.6 ND 0.7-0.8 0.8 8 and 0.8-1.2

of hybridization sites on the giant chromosomesh

Copies per haploid genome

Percentage of total DNA

35

20-60'

ND'

30 30 20-30 20-30

40b 30b 20-308 15hj 80-95'

ND ND 0.7 0.3 0.5

+ chromocenter

100

a Finnegan et al. (1978);b Potter ef al. (1979); ' Levis et al. (1980); Dunsmuir et al. (1980); ' A. A. Bayev, K. Will, and D. J. Finnegan (personal communication); M. W. Young and G. M. Rubin (personal communication); Ilyin et al. (1980b); Bayev er al. (1980); Scherer et al. (1982); j Ilyin et al. (1980a); ' The fragments described in earlier publications as Dm225 and Dm234 are parts of the complete mdgl element; ND, no data; Columns 1-4 and 7 from Spradling and Rubin (1981); different for each insertion site; " the number and distribution of hybridization sites vary widely among the different wild-type and mutant stocks (Strobe1 er al., 1979; Ilyin et al., 1978; Ananiev et al., 1978; Young, 1979). f

'

60

ELISABETH GATEFF

RNP particles (for review see Spradling and Rubin, 1981). Approximately 30 families of copia-like sequences have been detected (Rubin et a/., 1981). They comprise about half of the middle repetitive DNA sequences (6%). The members of a family show similar nucleotide sequences. Between the families, on the other hand, very little sequence homology exists. Six copiiilike elements have been cloned, sequenced, and analyzed by in situ hybridization to giant salivary gland chromosomes (Table Ill; Rubin et id., 1981). The hybridization sites are dispersed throughout the genome. Furthermore, the distribution pattern varies from stock to stock, even after short periods of separate rearing, which suggests that the elements are mobile within the genome. Characteristic for all six transposable elements are direct terminal repeats, which are flanked by short inverted repeats. A further peculiarity exists in the site of insertion of the transposon into the host genome, where a few base pairs become duplicated to give rise to flanking direct repeats (Table 111). In these features copia-like elements resemble bacterial transposons (Kleckner, 1977; Nevers and Saedler, 1977; Calos and Miller, 1980), the transposable element Tyl of yeast (Cameron et a/., 1979), and the integrated retroviral genomes (Bishop, 1978; Shimotohno et a/., 1980). The mature DNA of retroviral proviruses in the nucleus of virus-infected cells has been shown to represent covalently closed circles (Rotenberg et al., 1977; Shank and Varmus, 1978; Clayman et al., 1979). Small circular DNA has been isolated from Drosophib embryos and cells cultured in oitro (Stanfield and Helinski, 1976; Stanfield and Lengyel, 1979). The analogy between retroviral proviruses and the transposable copiu element is carried even further by the finding in in oitro cultured cells of copia DNA, in covalently closed circles containing either one or two terminal direct repeats (Flavell and Ish-Horowicz, 198l), and which resembles closely integrated proviruses of vertebrate retroviruses. Recently, Gehring and Par0 (1980) found a copin element associated with a large transposone (TE) (more than 200 kb), carrying the genes whiteapricot, roughest, and uerticuls, which encompasses four bands on the polytene chromosomes. This TE transposes with a frequency of and has been mapped in genetic experiments to more than 100 positions scattered over the genome (Ising and Ramel, 1976). Because copia is a movable element, it presumably causes the observed transpositions of the three aforementioned genes. One of the effects of transposition in bacteria is the induction of mutations (Starlinger and Saedler, 1976; Kleckner, 1977; Calos and Miller, 1980). lsing and Ramel (1976) reported that many transpositions of the aforementioned TE induce nonallelic, recessive-lethal mutations that persist after

CANCER, GENES, AND DEVELOPMENT

61

the loss of the TE from the particular site. The present author has investigated some of these lethal stocks such as TE 41 and TE 42 for putative tumorous growth, with negative results. However, two recently reported transpositions of the above TE to the aouter most tip of the left arm of the second chromosome, such as TE 75 (Ising and Block, 1981) and TE 133 (G. Ising, personal communication), showed in crosses to various /(2)g/ alleles in the heterozygous condition the typical 1(2)91 tumor phenotype (E. Gateff, G. Ising, and 0. Schmidt, unpublished results). The results indicate that the two TEs have inserted either into the l(2)gllocus or into its regulatory region. This fortunate coincidence creates a realistic possibility of cloning the /(2)91 genomic sequences with the help of the above TE. In another case of high transposition frequency of the white locus, Rasmuson et al. (1980) found that their transposon shows a DNA segment that is homologous to the intercalary heterochromatic DNA, cloned in the plasmid mdgl (Table 111). Also here, the observed transpositions may be caused by the mdgl transposable element. M. M. Green (personal communication) observed recently increased mutation rates on the 1(2)g/locus as a result of the insertion of a transposable MR element on the left arm of the second chromosome (for discussion of MR see Green, 1980). The accumulating data on transposable elements in Drosophih are highly significant in relation to their mutagenic activities (Green, 1980). Future cloning experiments of one or the other tumor gene described here may reveal an insertion of a transposon into the gene sequence, thus proving it responsible for the mutant effect. Nevertheless, the main question is whether copia-like sequences are involved in regulating normal development. The fact that the copy number and the points of insertion of the different copia-like elements vary considerably from stock to stock (Ilyin et a/., 1978) and also from individual to individual (Strobe1 et a/., 1979) makes it unlikely that they have some important role in normal development. It seems more likely that they are evolutionary ballast that obeys the cellular control mechanisms. In stress conditions, however, in which the control processes in the cells become sloppier, they may be transcribed and translated at higher rates and thus may enhance the pathological condition. In this respect it would be of prime interest to affirm their putative retroviral nature and their possible relation to the previously discussed vlps. I suggest, on purely speculative grounds, that the ulps are the visible products of the various copia-like elements. In the wild type, ulps are not assembled because their transcription and translation are regulated. In pathological states, however, particles are made.

62

ELISABETH GATEFF

VI. Retroviral Oncogenes and Their Cellular Counterparts in Different Animal Species

Genetic factors have been considered by many as causative agents in the induction of malignant growth (Knudson, 1978; Knudson and Meadows, 1978; Lynch, 1976; Anders and Anders, 1978; Gateff, 1978a,b,c; Ponder, 1980; Arrighi et af., 1981; and others). However, it was not until recently that this view gained optimistic approval. The finding that directed the spotlight toward the genome as the initiation site of malignant growth came from molecular and genetic studies on retroviral oncogenes (v-om) and from transfection experiments with DNA from normal and tumorous tissues. Oncogenes are essential for malignant neoplastic transformation, but they play no role in viral structure and replication. Because oncogenes are not a constitutive part of retroviruses, it was important to learn where oncogenes come from. Several hypotheses suggested that oncogenes may be transduced cellular genes (Hieger, 1961; Huebner and Todaro, 1969). Indeed, in the 1960s a number of novel defective murine sarcoma and acute leukemia viruses were isolated from animal tumors inoculated with murine leukemia virus (MuLV). This was the first indication that the V-ORCS may be transduced cellular genes (Harvey, 1964; Moloney, 1966; Kirsten and Mayer, 1967; Abelson and Rabstein, 1970). Molecular hybridization experiments provided further compelling evidence for this possibility. By now about 20 independent isolates of oncogenic retroviruses have proved to contain sequences homologous to genomic host sequences (Vennstrom and Bishop, 1982). Moreover, the cellular sequences incorporated into the viral genome are expressed in cells transformed by that virus. The best-studied viral oncogene, the src gene of Rous sarcoma virus (RSV), shows complete conservation in all avian species investigated (Spector et a/., 1978). At least part of the gene has been detected in a variety of vertebrates other than birds (Stehelin er af., 1976a,b; Hanafusa er af., 1977; Spector er al., 1978; Wang et af., 1980; Oskarsson er al., 1980). For other retroviruses, such as the myelocytomatosis virus (MC29V), the myeloblastosis virus (AMV) and the erythroblastosis virus (AEV), homologous genomic sequences have also been demonstrated in birds and other vertebrates (Sheiness and Bishop, 1979; Roussel er af., 1979). Less well conserved in evolution are the corresponding c-onc sequences to the viral oncogenes of Moloney murine sarcoma virus (MSV), Harvey sarcoma virus (HaSV), and feline sarcoma virus (FeSV), because they may exist only in mammals (Fishinger, 1980). Sheiness and Bishop (1979) examined the possibility of a cellular genetic element in the avian myelocytomatosis virus (MC29V) responsible for

CANCER, GENES, AND DEVELOPMENT

63

tumorigenesis. The phylogenetic origin of the v-om of MC29V, containing 1500 nucleotides, was analyzed by hybridizing the v-onc to genomic DNA of birds and other vertebrate species such as cow, mouse, and salmon. In all of them, nucleotide sequences related to v-onc DNA were found. Chickens for instance, the natural host of the virus, harbor at least 90% of the 1500 nucleotide-long viral DNA sequence in their genome. According to their evolutionary divergence from chickens, the other species showed correspondingly decreased sequence complementarity with the viral oncogene. A single, constant locus of the c-MCV oncogene is located on each chicken genome. The locus shows no linkage with c-src (Sheiness et al., 1980). The replication-defective avian leukemia viruses (ALV), which require helper viruses for propagation in cultured cells, are highly tumorigenic. Roussel et al. (1979) found three new oncogenes in them and demonstrated cellular homologues. Depending on the target cells they transform, they were designated as v-erb (erythroblasts), v-mac (macrophages), and v-myb (myeloblasts). Vennstrom and Bishop (1982) studied, by restriction endonuclease mapping and heteroduplex analysis, the structure of the cellular genomic locus corresponding to the AEV oncogene, c-erb, and c-erb, . They found that the cellular homologue to the v-erb, consists of approximately 750 nucleotides with three introns, and the c-erb, of 2000 nucleotides and 11 introns. The flanking genomic sequences were unrelated to the nononcogenic part of the AEV. With the help of immunological methods, Baltimore et al. (1979) identified a protein in mouse cells that appears to be the cellular counterpart to the viral P120 transforming protein of Abelson murine leukemia virus (Ab-MuLV). A single and constant locus on the human genome, homologous to the transforming gene v-sis of simian sarcoma virus, was described recently by Dalla Favera et al. (1 98 1). Of particular interest for this discussion are the spontaneous melanomas, occurring with high frequency in hybrids between the teleosteans platyfish (Xiphophorus maculatus) and swordtail (Xiphophorus helleri), which are a model system for the study of genetic factors in cancer (Gordon, 1958; Anders and Anders, 1978). Fish from wild populations growing for many generations in the laboratory never develop melanomas or other tumors, and are insensitive to carcinogens. The hybrids here discussed, however, show spontanously malignant tumors of all sorts and are susceptible to carcinogens (Anders and Anders, 1978).The genetic information responsible for the malignant neoplastic growth of pigment cells is mediated by a gene designated as “tumor gene” (Tu),which is repressed in wild-type animals by a system of regulator genes (R).In the hybrids, however, regulation of the Tu fails, and instead of normal differentiation malignant growth ensues.

64

ELISABETH GATEFF

In this system, Barnekow et al. (1982) detected the expression of endogenous pp6OSrcphosphoprotein (Collett and Erikson, 1978; Erikson er ul., 1980) and found associated kinase activity in brain extracts from different Xiphophorus species, hybrids, malignant, and nonmalignant tissues. Thus the conclusion was drawn that on the Xiphophorus genome a cellular src gene (c-src) is present that codes for the 60K phosphoprotein with kinase activity. Moreover, the expression of c-src, as identified by the pp60"-""associated kinase activity in melanomas of different genotypes and degree of tumorigenicity, were compared (Schartl et al., 1982). That study showed a correlation between the expression of c-src and Tu. Therefore, cells containing Tu express more pp60'-"c than cells lacking Tu. A correlation was also found between an elevated pp60"-"*'-associated kinase activity and the degree of Tu expression, which may indicate that Tu and c-src are regulated coordinately. Schwab (1981) reported the existence of v-src homologous sequences on the genome of Xiphophorus fish hybrids as detected in hybridization experiments. Further, he has been able to transform NIH-3T3 mouse cells in uirro by transfecting with c-src genomic DNA (see later). All the previous data demonstrate that animal genomes contain a group of evolutionarily highly conserved genes homologous to the different oncogenes. The conservation of these genes may have been necessitated by their fundamental role in specific cell types engaged in particular developmental programs (see later). Drosophila exhibits genomic sequences homologous to the v-src, v-ras, v-myc, vTfes,and v-ah1 oncogenes (Shilo and Weinberg, 1981b). Some of the Drosophila tumor genes may turn out to represent the corresponding homologous sequences to one or the other viral oncogene. All Drosophila tumor mutants represent developmental mutants. Therefore, the wild-type gene product must control some key differentiation step, as judged by the devastating effects caused to the animals by the mutant gene. Similarly, the various c-onc appear to interfere with particular pathways of differentiation of the infected target cells (Graf rt ul., 1980; see later) and resemble in this respect Drosophila tumor genes. The state of differentiation of macrophages is differentially affected by acute leukemia viruses, such as avian myeloblastosis virus (AMV) and avian cytomatosis virus (MC29V), whereas RSV fails to induce functional alterations in the macrophages, despite the production of infectious virus (Durban and Boettiger, 1981). Thus retroviral oncogenes show target-cell specificity in uiuo as well as in uitro. Infection of cells, determined for the erythrocyte, macrophage, or myelocyte differentiation pathways, with the corresponding avian erythroblastosis virus (AEV), MC29V, respectively, AMV causes acute leukemias. Graf ef ul. (1980) showed that the different viruses replicate

CANCER, GENES, AND DEVELOPMENT

65

and produce the respective transformation proteins, not only in their target cells but also in hematopoietic nontarget cells. There, however, they do not cause malignant transformation. These results support the notion that the respective transforming proteins block the differentiation of their target cells by competitively inhibiting the action of homologous cellular differentiation proteins. Chen et nl. (1981), on the other hand, demonstrated that in the case of the feline sarcoma virus (FeSV), cells of embryonal layers other than the mesoderm could also be transformed, and they expressed antigenically indistinguishable transformation-associated polyproteins designated as gag-x. Thus alternative pathways of neoplastic transformation must also exist. A number of nonallelic Drosophilu genes were found to induce, in the mutated state in a particular cell type, one and the same tumor phenotype. For instance, six independent mutant alleles, representing different loci, cause malignant neuroblastoma (see Section IV,B, 1 ; Table I), which implies that the six wild-type gene products control six developmental steps leading from an optic neuroblast to a fully differentiated optic neuron. A mutation in any one of these six genes results in a neuroblastoma. The situation is analogous in the case of the different retroviral oncogenes. The various v-ones causing sarcomas differ in their nucleotide sequence and thus in their function. For instance, the src genes of both the Fuginanii sarcoma virus and RSV cause sarcomas in chicken fibroblasts despite their differing genomic structure (Bishop, 1982). Thus, as in Drosophilu (see earlier), different cellular homologues to retroviral oncogenes also seem to be involved in the control of specific differentiation steps in one and the same cell type. Experiments using temperature-sensitive RSV showed that the transformation as well as the maintenance of the transformed state of fibroblasts depends on the continuously functioning v-src gene product pp60"" present in excessive amounts (for review, see Bishop, 1982). In Drosophila, in the case of the l(2)gl alleles and the l(2)gl deficiency, the mechanism seems to be different. Here, the absence of a gene product appears to be responsible for the malignant transformation and not its excessive synthesis, because the deficiency of the gene locus shows the most extreme tumor phenotype. Similarly, in the case of the 1(3)mbt'"-l mutant the lack of the functional gene product from the beginning of embryonic life up until the middle of the third larval instar is responsible for the brain and imaginal disc tumors. Retroviral studies suggest that functional cellular genes can be subject to different regulation, depending on whether they are in the cellular or the

66

ELISABETH GATEFF

viral genome. The excessive production of a cellular protein by the viral gene causes malignant transformation (Hunter and Sefton, 1980). This new, exciting, and simple concept of cancer development will have to allow for other types of genetic mechanisms as well, because there are also slow-transforming RNA tumor viruses that do not possess oncogenes, but despite this cause malignant neoplastic transformation. Nee1 ef al. (1981) demonstrated for the avian and murine leukemia viruses that they activate a cellular gene through the integration of the provirus adjacent to the c-myc, the cellular counterpart of the MC29 viral oncogene. The transcription of the gene begins at the proviral promoter, which results in enhanced gene expression and malignant neoplastic transformation (Hayward et al., 1981). The studies just discussed provide compelling support for the existence in animals of phylogenetically highly conserved genes that are causally related to the malignant neoplastic transformation of cells. However, because most cancers are not of viral origin, the question arises whether the postulated cellular oncogenes are primarily involved also in cancers lacking viruses. The approach taken by a group of investigators and its relation to the Drosophila tumor system is discussed next.

VII. Transfection and Vertebrate Tumor Genes: A Comparison with Drosophila

The coordination of gene activities in a cell resulting in what we call development is still largely obscure. For instance, during Drosophifaoogenesis, between 200 and 600 genes must be involved (King and Mohler, 1975). Most of these genes transcribe, however, only very few mRNA molecules per cell and are thus not yet amenable for investigation. These genes are, however, the gears that act time and tissue specifically, and drive the cells into their final differentiation. Attention is beginning to focus on this large gene family, some of which, when abnormally regulated or mutated, cause a halt of differentiation and continuous malignant growth (see Section IV). The study of this kind of genes in a complex genetic system, such as that of humans or of any other animal that does not possess a well-known genetic makeup, meets with immense difficulties. Nevertheless, a method is available for probing genes of the rare mRNA type (Weinberg, 1981). The method of gene transfer or transfection allows the study of genes in a cellular environment away from the regulatory mechanisms of the host cells in which they normally function. The combination of this method with an in uirro transfection focus assay has given further support for the notion that the genome of vertebrates contains

CANCER, GENES, A N D DEVELOPMENT

67

potential transforming genes (Anderson et al., 1979; Cooper et al., 1980; Shih et al., 1979; Cooper and Neiman, 1980; Krontiris and Cooper, 1981 ; Shih et al., 1981; Shilo and Weinberg, 1981a; Lane et al., 1981; for review see Weinberg, 1981; Willecke and Schafer, 1982). Transformation by transfection of NIH-3T3 mouse cells with genomic DNA is assumed to result from the activation of a cellular gene with transforming potential. This may occur, for instance, if sequences needed for the regulation of a cellular transforming gene are removed by the shearing of the DNA, and the gene is thus expressed differently (Cooper et al., 1980; Krup-Kovalinkova and van den Berg, 1980). Blair et al. (1981) showed data that agree with the foregoing interpretation. Clones containing long terminal repeats (LTR) from Moloney sarcoma virus (M-MSV) provirus to c-mos, the cellular homologue of the M-MSV specific sequence v-mos, transform cells as efficiently in the transfection focus assay as cloned v-mos together with the LTR of the virus. The experiments suggest that LTR-like elements on the genome may activate cell sequences whose abnormal expression leads to malignant transformation. Transfection with high-molecular-weight DNA from three human cell lines shows that in each case different DNA sequences are responsible for the transformation of the individual tumor cell lines (Murray et al., 1981). Similar results were obtained by Lane er al. (1981) with high-molecularweight DNA from ( a ) mammary tumor virus-induced mouse mammary carcinomas, (b) carcinogen-induced mouse mammary tumors, and (c) a spontaneous human mammary tumor cell line. All but one tumor DNA exhibited high transformation efficiency. The transfection studies just mentioned, performed with the DNA of a variety of mouse, chicken, and human tumors that had been induced either by carcinogens or by viruses, suggest that during transfection, tissue-specific cellular genes become abnormally regulated. At the moment only speculations can be made concerning the role of these genes in the cells. A reasonable suggestion, which has been expressed by all investigators in the field, is that this family of genes is involved in growth control and/or the establishment of the differentiated state and its maintenance. The Drosophila tumor genes have shown just this for more than 10 years. All Drosophila genes causing tumors in the mutated state are developmental genes acting specifically in one or, at the most, two cell types in which they surely must have important functions during normal development (Gateff, 1978a,b,c). In relation to the above transfection studies, the mutagenic effect of exogenous DNA in Drosophila deserves to be mentioned. DNA of different viral and nonviral origin induces increased rates of germ-line mutations when fed to male flies. These mutations are restricted to a small number of

68

ELISABETH GATEFF

specific loci (for review see Gershenson, 1980). Moreover, the mutation spectrum is different from that induced by conventional mutagens, most of which cause gross genomic rearrangements. Exogenous DNA, in contrast, shows exclusively single gene mutations or at the most small deletions that cannot be detected on the giant chromosomes. The visible and recessive-lethal mutations involve only a few specific loci. For example, on chromosome I1 only about 12 loci were mutated, each of which showed a number of alleles, whereas mutagens induce hundreds of mostly nonallelic recessive-lethal mutations. Transfection with Drosophila DNA, on the other hand, showed many nonallelic lethal mutations. The mutants induced by exogenous DNA are often unstable, frequently reverting to the wild type. Reversions taking place in the germ line produce wild-type flies, whereas reversions in somatic cells give rise to mosaic animals. It was further found that only high-molecular-weight DNA is mutagenic and that premeiotic germ cells are more susceptible to the foreign DNA than postmeiotic germ cells (Gershenson, 1980). Transfection of mutant Drosophilu embryos with wild-type DNA yielded repair of the mutant phenotype (Fox and Yoon, 1966). The genetic analysis of 3 of the 51 “transformed” stocks showed that the introduced wild-type allele inserted itself on the chromosome at the proper site without the excision of the mutant allele (Fox et a/., 1971). Furthermore, DNA from the aforementioned transformed stocks induced second-step “transformation” at a frequency resembling that of the original transformation (Fox et al., 1975). The transfection studies with exogenous and homologous DNA in Drosophilu show, in parallel with the above transfection experiments, a high degree of specificity of the mutagenic effect. Unfortunately, the cause for the developmental arrest in the different lethal alleles was not investigated. Another similarity seems to exist between the specific behavior of exogenous DNA in the vertebrate and Drosophila transfection studies just mentioned. In both, no species and tissue barriers seem to exist. Shih et a/. (1981) demonstrated, for instance, that DNAs obtained from human, rabbit, and mouse bladder carcinoma and mouse glioma lines are able to induce transformation of NIH-3T3 mouse fibroblasts in uitro upon transfection. Thus the transforming genes not only act across tissues, but they also seem to overcome the species barrier. The situation is similar in the case of the DNAs of two human bladder carcinoma cell lines, which induce transformation in mouse fibroblasts (Krontiris and Cooper, 1981). Nevertheless, one discrepancy arises when one compares these observations to the tissue specificity of the Drosophika genes (see Section 1V) and of the c-mar, c-erb, and c-myc oncogenes described by Beug rt al. (1979), Graf er a/. (1980), and Durban and Boettiger (1981). Thus caution will have

CANCER, GENES, AND

DEVELOPMENT

69

to prevail in drawing conclusions before the molecular basis of the observed phenomena is unveiled. VIII. Concluding Remarks

Among the approximately 5000 genes on the Drosophila genome, 25 are presently known to be causally involved in the development of malignant and benign neoplasms (Table I). The wild-type alleles of these genes control particular pathways of development in specific tissues or cells at particular times (see Section VI). The cells and tissues involved include the adult optic neuroblasts, the imaginal discs, and the primordial blood cells in the larva, the male and female gonial cells in the adult, and neuroblasts, blood cells, and other not yet identified cell types in the embryo. The mutations originated either spontaneously or after treatment of animals from wild-type laboratory populations with mutagens. However, the phenomenon is not solely a laboratory one; more than 78 l(2)gl alleles have been found in natural Drosophila populations. Moreover, at least one other Drosophila species, D. hydei, exhibits a homologous Z(2)gl locus, which in the mutated state shows a similar tumor phenotype. Normal and ts alleles have been isolated. Thus Drosophila demonstrates unequivocally that single mutated genes cause malignant or benign neoplastic transformation of specific cell types by interfering with their differentiation. Two research fields, (1) retroviral oncogenes and (2) transfection with DNA from normal and tumorous tissues, have provided compelling evidence that this may be true for higher animals as well (see Sections V, VI, VII, and VIII). ACKNOWLEDGMENTS The skilled technical assistance of D. Willer and I. Brillowski are gratefully acknowledged. I am also grateful to Professor K. Sander for providing the research facilities. The research was generously supported by the Deutsche Forschungsgemeinschaft, SFB 46. M y thanks to S. Chambers for typing the manuscript.

REFERENCES Abelson, H. T., and Rabstein, L. S. (1970). Cuncer Res. 30, 2213-2222. Akai, H. (1972). Cell4,38-41. Akai, H . , Gateff, E., Davis, L. E., and Schneiderman, H. A . (1967). Science 157,810-813. Ananiev, E. V., Gwozdev, V. A,, Ilyin, Yu. V., Tchurikov, N. A,, Georgiev, G . P. (1978). Chromosomu 70, 1- 17.

70

ELISABETH CATEFF

Anders, A., and Anders, F. (1978). Biochim. Biophys. A r i a 516,61-95. Anderson, P., Goldfarb, M . P., and Weinberg, R . A. (1979). Cell 16,63-75. Arrighi, F. E., Rao, P. N., and Stubblefield, E. (eds.). (1981). “Genes, Chromosomes and Neoplasia.” Raven, New York. Baltimore, D., Shields, A., Otto, G., Goff, S.. Besmer, P., Witte, O., and Rosenberg, N . (1979). Cold Spring Harbor Symp. Quanr. Biol. 44,849-854. Barnekow, A., Schartl, M., Anders, F., and Bauer, H. (1982). Cancer Res., submitted. Bayev, A. A., Jr., Krayev, A. S., Lyubomirskaya, N. V., Ilyin, Y. V. Skryabin. K . G . , and Georgiev, G . P. (1980). Nucl. Acid. Res. 8, 3263-3273. Beug, H., Kirchbach, A., Doederlein, G . ,Cinscience, J. F., and Graf, T. (1979). Cell 18,375-390. Bishop, J. M. (1978). Ann. Rev. Biochem. 47,35-88. Bishop, J. M. (1981). Cell23,5-6. Bishop, J. M. (1982). In “RNA Tumor Viruses.’’ Cold Spring Harbor Laboratory, Cold Spring Harbor, New York (in press). Blair, D. G., Oskarsson, M., Wood, T. G . , McClements, W. L., Fischinger, P. J. and Vande Woude, G. G . (1981). Science 212,941-943. Brun, G . , and Plus, N . (1980). In “The Genetics and Biology of Drosophila” (M. Ashburner and T. R. F. Wright, eds.). Vol. 2D, pp. 627-702, Academic Press, New York. Bryant, P. J. (1978). In: “The Genetics and Biology of Drosophila”(M. Ashburner and T. R. F. Wright, eds.), Vol. 2C, pp. 230-326. Academic Press, New York. Bryant, P. J., and Schubiger, G. (1971). Develop. B i d . 24,233-363. Calos, M. P..and Miller, J . H. (1980). Ce1120, 579-595. Cameron, J. R., Loh, E. J., and Davis, R.W. (1979). Cell 16,739-751. Chen, A. P., Essex, M., Shadduck, J. A., Niederkorn, J. Y.. and Albert D. (1981). Proc. Nail. Acad. Sri. U.S.A.78,3915-3919.

Clayman, C. H., Mosharrafa, E. T., Anderson, D. L., and Faras, A. J. (1979). Science 206, 582-584.

Collett, M. S., and Erikson, R . L. (1978). Proc. Nail.Acad. Sci. U.S.A. 75,2021-2024. Cooper, G . M., and Neiman, P. E. (1980). Narure (London) 287,656-659. Cooper, G . M., Okenquist, S., and Silverman, L. (1980). Naiure (London)248,418-422. Convin, H. O., and Hanratty. W. P. (1976). Mol. Gen. Genei. 14,345-347. Dalla Favera, R., Gelman, E. P., Gallo, R. C., and Wong-Staal, F. (1981). Nature (London) 292,31-35.

Dehn, A. (1981). Masters Thesis, Department of Biology, Freie Universitat Berlin, Federal Republic of Germany. Dhar, R., McClements, W. L., Enquist, L. W., and Vande Woude, G. F. (1980). Proc. Nail. Acad. Sci.U.S.A. 77,3937-3941.

Dunsmuir, P., Brorein, W. J., Simon, M. A., and Rubin, G. M . (1980). Cell 21, 575-576. Durban, L. M., and Boettiger, D. (1981). Proc. Nail. Acad. Sci. U . S . A . 78,3600-3604. Erikson, R. L., Purchio, A. F., Erikson, E., Collett, M. S., and Brugge, J. S. (1980). Cell Biol. 87,319-325.

Finnegan, D. J., Rubin, G . M., Young. M. W., and Hogness, D. S. (1978). ColdSpring Harbor Symp. Quant. Biol. 42, 1053-1063. Fishinger, P. (IY80). In “Molecular Biology of RNA Tumor Viruses” (E. J. Stephensoh. ed.). pp. 163-198. Academic Press, New York. Flavell, A. J., and Ish-Horowicz, D. (1981). Narure (London)292,591-595. Fox, A. S., and Yoon, S. B. (1966). Genetics 53, 897-91 I . Fox, A. S., Yoon, S. B., and Gelbart, W. M. (1971). Proc. Nail. Acud. Sci. U.S.A. 68,342-346. Fox, A. S., Parzen, S. D., Salverson, H., and Yoon. S. B. (1975). Genei. Rex. 2, 137-147.

CANCER, GENES, AND DEVELOPMENT

71

Fullilove, S. L., and Jacobson, A. G . (1978). In “The Genetics and Biology of Drosophila” (M. Ashburner and T. R. F. Wright, eds.), Vol. 2C,pp. 106-144.Academic Press, New York. Gafner, J., and Philippsen, P. (1980).Nature (London) 286,414-418. Cans. M., Audit, C., and Masson, M. (1975).Genetics 81,683-704. Gateff, E. (1977).Ann. Parasirol. Hum. Comp. (Paris) 52,8143. Gateff, E.(1978a).In “The Genetics and Biology of Drosophila.” (M. Ashburner and T. R. F. Wright, eds.), Vol. 2B, pp. 181-275.Academic Press, New York. Gateff, E.(1978b).Science 200, 1448-1459. Gateff, E. (1978~). Biol. Rev. 53, 123-168. Gateff, E. (1981).Drosophila Inform. Serv. 56, 191. Gateff, E. (1982).In “The Proceedings of the IX Congress of the International Society of Developmental Biologists” (M. M. Burger, ed.), Part B, pp. 621-632. Liss, New York. Gateff, E., and Schneiderman, H. A. (1967).Am. Zool. 7,760. Gateff, E.,and Schneiderman, H. A. (1969).Nail. Cancer Insr. Monogr. 31,365-397. Gateff, E.,and Schneiderman, H. A. (1974). Wilhelm Roux’s Arch. Entwicklungsmech. Org. 176,23-65. Gateff,E.,Golubovsky,M. D.,andSokolova, K.B.(1977).Drosophila Inform. Ser.52,128-129. Gateff, E.,Gissmann, L., Shresha, R., Plus, N., Pfister, H., Schroder, J., and zur Hausen, H. (1980).In “Invertebrate Systems in Vitro” (E. Kurstak, K. Maramorosch, and A. Dubendorfer, eds.), pp. 51 7-533. Elsevier North-Holland Biomedical Press, Amsterdam. Gehring, W. (1973).In “Mechanisms ofDevelopment” (F. Ruddel, ed.), pp. 103- 128.Academic Press, New York. Gehring, W. J. (1978).In “TheGeneticsand BiologyofDrosophila”(M. AshburnerandT. R. F. Wright, eds.), Vol. 2C,pp. 51 1-544.Academic Press, New York. Gehring, W. J., and Paro, R. (1980).Cell 19,897-904. Georgiev, G . P.(1969).J . Theorei. B i d . 25,473-490. Gershenson, S. M. (1980).In “Well-Being of Mankind and Genetics, Proceedings of the XIV International Congress of Genetics” (M. E. Vartanian, ed.), Vol. 1, book two, pp. 96-1 15. MIR Publishers, Moscow. Collin, M. S . , and King, R. C. (1981).Develop. Genet. 2,203-218. Golubovsky, M. D. (1978).Drosophila Injorm. Sero. 53, 179. Golubovsky, M. D., and Sokolova, K . (1973).Drosophila Inform. Serv. 50, 124. Gordon, M. A. (1958).Ann. N . Y.Acad. Sci.71, 1213-1222. Graf, T., Beug, H., and Hayman, M. J. (1980).Proc. Narl. Acud. Sci. U.S.A. 77,389-393. Green, M. M. (1980).Annu. Rev. Genet. 14, 109-120. Grigliatti, T. A,, Hall, L., Rosenbluth, R., and Suzuki, D. T. (1973).Mol. Gen. Genet. 120, 107-114. Haars, R., Zentgraf, H., Gateff, E., and Bautz, F. A. (1979).Virologie 101, 124-130. Hanafusa, H., Halpern, C. C., Buchhagen, D. L., and Kawai, S. (1977).J . Exp. Med. 146, 1735-1 747. Hanratty, W. P., and Cough, T. L. (1982).Genet. Res., submitted. Hanratty, W. P., and Ryerse, J. S. (1981).Deoel. Biol. 83,238. Harvey, J. Y. (1964).Narure (London) 204,1104-1 105. Hayward, W. S . , Neel, B. G . , and Astrin, S . M. (1981).Nature (London) 290,475-480. Heine, C. W., Kelly, D. C., and Avery, R. J. (1980).J . Gen. Virol. 49,385-395. Hieger, 1. (196I). “Carcinogenesis.” Academic Press, New York. Hofbauer, A. (1979).Ph.D. Thesis, Fakultat fur Biologie, Albert-Ludwigs-Universitat, Freiburg i. Br., FRG.

12

ELISABETH GATEFF

Huebner, R. J., and Todaro, G . J. (1969). Proc. Nail. Acad. Sci. U . S . A .64, 1087-1094. Hunter,T.,and Sefton, B. (1980). Proc. Nail. Acad. Sci. U . S . A .77, 1311-1315. Ilyin, Y. V., Chmeliauskaite, V. G., Ananiev, E. V., and Georgiev, C. P. (1980a). Chromosoma 81,27-53. Ilyin, Y. V., Chmeliauskaite, V. G., Ananiev, E. V., Lyubomirskaya, N. V., Kulguskin, V. V., Bayev, A. A., Jr., and Georgiev, G . P. (1980b). Nucl. Acid Res. 8, 5333-5346. Ilyin, Y.V., Tehurikov, N. A., Ananiev, E. V., Ryskov, A. P., Yenikopolov, G . N., Limborska, S. A., Maleeva, N. E., Gwozdev, V. A., Georgiev, G . P. (1978). Cold Spring Harbor Synip. Quani. B i d . 42,959-969. Ising, G., and Block, K. (1981). Cold Spring Harbor Symp. Quant. Biol. 45, 527-544. Ising, G., and Ramel, C. (1976). In “The Genetics and Biology of Drosophila” (M. Ashburner and E. Novitski, eds.), Vol. IB, pp. 947-954, Academic Press, New York. Kankel, D. R.. Ferrus, A., Garren, S. H., Harte, P. J., and Lewis, P. E. (1980). In “The Genetics and Biology of Drosophila,“ (M. Ashburner and T. R. F. Wright, eds.), Vol. 2D, pp. 295363. Academic Press, New York. King, R. C . (1970). In “Ovarian Development in Drosophila melanogaster.” pp. 33-98. Academic Press, New York. King, R. C. (1979). Int. J . Insect Morphol. Embryol. 8,297-309. King, R. C . , and Bodenstein, D. (1965). Z . Naturforsch. 20, 292-297. King, R. C., and Buckles, B. D . (1980). Drosophila It+wm. Seru. 55,74475. King, R. C., and Mohler, J . D . (1975). In “Handbook of Genetics” (R. C . King, ed.), Vol. 3, pp. 757-791, Plenum, New York. King, R. C., Burnett, R. G., and Staley, N. A. (1957). Groivih 21,239-261. King, R. C., Bahn, M., Horowitz, R., and Rarramendi, P. (1978). Int. J . Insect Morphol. Embryol. 7, 359-378. Kirsten, W. H., and Mayer. L. A. (1967). JNCI, J . Nail. Cuncer Insi. 39, 311-335. Kiss, I., Scabad, J.. and Major, J. (1978). Mol. Gen. Genet. 164,77-83. Kleckner, N. (1977). Cell 11, 11-23. Knudson, A. G . (1978). In “Tumors of Early Life in Man and Animals” (L. Severi, ed.), pp. 765-775. Perugia Quadrenial International Conference on Cancer, Monteluce, Italy. Knudson, A. G., and Meadows, A. T . (1978). In “Cell Differentiation and Neoplasia” (F. Grady, ed.), pp. 88-92. Raven, New York. Koch, E. A., and King, R. C. (1964). Growth 28,325-369. Krontiris, T. G., and Cooper, G. M. (1981). Proc. Nail. Acud. Sci. U . S . A .78, 1181-1 184. Krup-Kovalinkova, V., and van den Berg, K. J. (1980). Naiure (London) 287,353-354. Lane, M.-A., Sainten. A.. and Cooper. G . M. (1981). Proc. Nail. Acad. Sci. U . S . A . ,in press. Lehmann, R., Dietrich. U., Jiminez, F., and Campos-Ortega, J. A. (1981). Roux’s Arch. 190, 226-229. Levis, R., Dunsmuir, P., and Rubin, G. M. (1980). CeN21, 581-588. Lifschytz, E. (1978). Develop B i d . 66, 571-578. Lindsley, D. L., and Grell, E. H. (1968). Genetic variations of Drosophilu melanogaster. Carnegie Inst. Wash. Publ. No. 627. Lindsley, D. L., and Tokuyasu, K. T. (1980). In “The Genetics and Biology of Drosophila” (M. Ashburner and T. R. F. Wright, eds.), Vol. 2D, pp. 226-287. Academic Press, New York. Lynch, H . T., ed. (1976). “Cancer Genetics,” Charles C. Thomas Publ., Springfield. MacMorris-Swanson. M., and Poodry, C. A. (1981). Devel. Bid. 84,465-470. Mahowald, A. P., and Kambysellis, P. (1980). In “The Genetics and Biology of Drosophila” (M. Ashburner and T. R. F. Wright, eds.), Vol. 2D, pp. 141-209. Academic Press, New York.

CANCER, GENES, A N D DEVELOPMENT

73

Mohler, J . D. (1977).Genetics 85,259-272. Mohr, 0 .L. (1919).Genetics 4,275-282. Moloney. J. B. (1966).Nail. Cancer Inst. Monogr. 22, 139-142. Murray, M. J., Shilo, B.-Z., Shih, C., Cowing, D., Hsu, H. W., and Weinberg, R. A. (1981). Cell 25, 355-361. Nappi, A. J., (1975).In “Invertebrate Immunity” (K. Maramorosch and R. E. Shope, eds.), pp. 293-327. Academic Press, New York. Neel, B. G., Hayward, W. S., Robinson, H. L., Fang, J. M., and Astrin, S. M. (1981). Cell23, 323-334. Nevers, P., and Saedler, H.(1977). Nature (London) 268, 109-115. Oskarsson, M., McClements, W. L., Blair, D. G., Maizel, J. V., and VandeWoude, G. F.(1980). Science 207,1222-1224. Parodi, S., and Brambilla, G. (1977). Mutation Res. 47,53-74. Plus, N., and Golubovsky, M. D. (1981).Genetika 12,227-231. Plus, N., Croizier, G., Jousset, F. X., and David, J. (1975).Ann. Microhiol. (Paris)126, 107-1 17. Ponder, B. A. J. (1980).Biochim. Biophys. Acia 605,369-410. Poodry, C. A.(1980).In “The Genetics and Biology of Drosophila” (M. Ashburner and T. R. F . Wright, eds.), Vol. 2D, pp. 407-432.Academic Press, New York. Poodry, C. A,, Hall, L., and Suzuki, D.T. (1973).Develop. Biol. 32,373-386. Potter, S . , Brorein, W. J., Dunsmuir, P., and Rubin, G . M. (1979).Cell 17,415-427. Poulson, D. F. (1937).Proc. Natl. Acad. Sci. U.S.A. 23, 133-137. Poulson, D. F. (1940).J . Exp. 2001.83,271-325. Poulson, D. F. (1945).Am. Nai. 79,340-363. Poulson, D. F. (1950). In “Biology of Drosophila” (M. Demerec, ed.), pp. 209-244. Hafner, New York. Rasmuson, B., Montell, I., Rasmuson, A., Svahlin, H., and Westerberg, B.-M. (1980).Mo/. Gen. Gener. 177,567-570. Rizki, M. T. M. (1957).J . Morphol. 100,437-458. Rizki, M. T. M., Rizki, M. R., and Grell, E. H. (1980). Wilhelm Roux’s Arch. 188,91-99. Romrell, L. J. (1975).In “Handbook of Genetics” (R. C. King, ed.), Vol. 3, pp. 735-745. Plenum, New York. Rotenberg, E., Smothen, D., Baltimore, D., and Weinberg, A. (1977).Nature (London) 269, 122- 126. Roussel, M., Saule, S., Lagrou, C . , Rommens, C., Beug, H., Graf, T., and Stehelin, D. (1979). Nature (London) 281,452-455. Rubin, G.M.,Brorein, W. J., Jr., Dunsmuir, P., Flavell, A. J., Levis, R., Strobel, E., Toole, J. J., and Young, E.(1981).Cold Spring Harbor Symp. Quant. Biol. 45,619-628. Schartl, M., Barnekow, A,, Bauer, H., and Anders, F. (1982). Cancer Res., submitted. Scherer, G., and Pirrotta, V. (1981). Eur. J . Cell Biol. 24,20. Scherer, G., Tschudi, C., Perera, J., Delius, H., and Pirrotta, V. (1982).Cell, submitted. Schwab, M. (1981).Abstract IX. Congress of the International Society of Developmental Biologists. Basel, Switzerland. Scott, M. P., Fostel, J. M., and Pardue, M. L. (1980).Cell 22,929-941. Shank, P. R., and Varmus, H. E. (1978).Virology 25, 102-1 14. Sheiness, D. K., and Bishop, J. M. (1979).J . Virol. 31,514-521. Sheiness, D. K.,Hughes, S. H., Varmus, H. E., Stubblefield, E., and Bishop, J. M . (1980). Virology 105,415-424. Shih, C., Shilo, B.-Z., Goldfarb, M. P., Dannenberg, A., and Weinberg, R. A. (1979). Proc. Nail. Acad. Sci. U.S.A. 76,5714-5718. Shih. C.,Padhy, L. C . , Murray, M., and Weinberg, R. A.(1981).Nature(Londnn)290,261-264.

74

ELISABETH GATEFF

Shilo, B.-Z., and Weinberg, R. A. (1981a). Nature (London) 289,607-609. Shilo, B.-Z., and Weinberg, R. A. (1981b). Proc. Natl. Arad. Sci. U.S.A. 75,6789-6792. Shimotohno, K., Mizutani, S., and Ternin, H. M. (1980). Nature (London) 285,550-554. Shrestha, R. (1979). Doctoral Thesis, Fakultat fur Biologie, Albert-Ludwigs-Universitat, Freiburg i. Br., FRG. Shrestha, R., and Gateff, E. (1982a). Develop. Growth Dtfler. 24,64-82. Shrestha, R., and Gateff, E. (1982b). Develop. Growth Dtfler. 24, 83-98. Smith, P. A., and King, R. C. (1966). JNCI, J . Narl. Cancer Inst. 36,455-463. Sokolova, K., and Golubovsky, M. D. (1978). Drosophila Inform. Sew. 53, 195-196. Sparrow, J. C. (1978). In “The Genetics and Biology of Drosophila,” (M. Ashburner and T. R. F. Wright, eds.), Vol. 2B, pp. 277-307. Academic Press, New York. Spector, D . H., Varmus, H. E., and Bishop, J. M. (1978). Proc. Natl. Acad. Sci. U.S.A. 75, 4102-4106. Spradling, A. C., and Rubin, G . M.(1981). Annu. Rev. Gener. IS, 219-251. Srdid, Z., and Frei, H. (1980). Differenriation 17, 187-192. Stanfield, S., and Helinski, D . R. (1976). Ce/l9,333-345. Stanfield, S., and Lengyel, J. A. (1979). Proc. Narl. Acad. Sci. U.S.A. 76,6142-6146. Stark, M. B., and Marshall, A. K. (1930). J . Am. Inst.Homeopatho1. 23, 1204-1206. Starlinger, P., and Saedler, H. (1976). Curr. Top. Microbiol. Immunol. 75, 111-152. Stehelin, D., Guntaka, R. V., Varrnus, H. E., and Bishop, J. M. (1976a). J . Mol. Biol. 101, 349-365. Stehelin, D., Varmus, H. E., Bishop, J. M., and Vogt, P. K . (1976b). Nature (London) 260, 170- 173. Stewart, M., Murphy, C., and Fristrorn, J. (1972). Develop. Eiol. 27,71-83. Strobel, E., Dunsmuir, P., and Rubin, G . M.(1979). Cell 17,429-439. Sutcliffe, J. G., Shinnick, T. M., Verma, 1. M., and Lerner, R. A. (1980). Proc. Narl. Acad. Sci. U.S.A. 77,3302-3306. van Beveren, C., Goddard, J. G., Berns, A., and Verma, 1. M. (1980). Proc. Natl. Acad. Sci. U.S.A. 77,3307-331 1. Vennstrom, B., and Bishop, J. M. (1982). Cell, submitted. Wang, L.-H., Snyder, P., Hanafusa, T., and Hanafusa, H. (1980). J. Virol. 35,52-64. Weinberg, R. A. (1981). Eiochim. Biophys. Acra 651,25-35. Willecke, K . , and Schafer, R. (1982). In “Somatic Cell Genetics” (C. T . Caskey, ed.), Plenum, New York (in press). Williams, J. M. (1981). Drosophila Inform. Serv. 56, 158-161. Wong-Staal, F., Dalla Favera, R., Franchini, G., Gelmann, E. P., and Gallo, R. C. (1981). Science 213,226-228. Young, M. W. (1979). Pror. Narl. Acad. Sci. U.S.A. 76,6274-6278.

TRANSFORMATION-ASSOCIATEDTUMOR ANTIGENS Arnold J. Levine Department of Microbiology, School of Medicine, State Universityof New York. Stony Brook, New York

I. Introduction .............................. .... A. Why Do Tumors Express Tumor Antigens? ............................. B. Are These Tumor Antigens Useful in Sorting out the Primary Events of Tumorigenesis? . . . . . . . . . . .............. ..................... C. Tumor Antigens as Diagnostic Tools D. Viral and Cellular Associated Tumor ................. 11. Simian Virus 40 . . . . . . . . . . . . . . ...........................

B. Small t-Antigen-Associated Proteins, 56K and 32K ....................... 111. Adenovimses. ............................................... IV. Epstein-Barr Virus . . . . . ................................ ells (Meth A) ......... V. Methylcholanthrene-Indu VI. Abelson Virus . . . . . . ............................... VlI. Rous Sarcoma Virus ................................ VIII. Teratocarcinomas ................................... IX. Conclusions. ................................ A. A Classification of Tumor Antigens ....................................

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

References . . . . . . .

75

15 80 80 80 81 82 84 86 88 91 93 96 97 100

100 103 104

1. Introduction

A. WHYDo TUMORS EXPRESS TUMORANTIGENS? Tumor antigens are experimentally defined as chemical moieties, associated with a tumor in an animal, that elicit an immune response in the host. The tumor antigen may reside on the cell surface and result in transplantation rejection of the tumor, in which case it may be termed a tumorspecific transplantation antigen (TSTA). Alternatively, the tumor antigen may be localized intracellularly and induce an immune response by virtue of cell death or tissue necrosis. Some tumor antigens will induce antibody production in animals bearing a syngeneic tumor graft or even with an autochthonous tumor graft. There are, however, several different reasons a 15 ADVANCES IN CANCER RESEARCH, VOL. 37

Copyright 0 1982 by Academic Press, Inc. All rights of reproduction in any form reserved. ISBN 0-12-006637-8

76

ARNOLD J . LEVINE

tumor antigen may elicit an immune response in spite of the fact that the tumor tissue is syngeneic with the host :

1. Alteration of a normal molecule on the surjace of the cell can produce a new antigenic determinant. Some types of transformed cells, like human adenocarcinomas, appear to contain glycolipids that lack the terminal sugar residue (glucosamine) found in the gangliosides of normal cells (Tal, 1965; Hakomori, 1971; Burger, 1971). In this case tumorigenic cells contain lower levels of the glycosyl transferases responsible for the addition of terminal sugars to glycolipids of the membrane (Grimes, 1970), and so the altered terminal sugar groups of these glycolipids become antigenic (Hakomori and Murakami, 1968). The appearance of Forsman-like antigens in some transformed cells can be explained in this way (Robertson and Black, 1969). Similarly, the increased levels of proteases (Unkeless et al., 1973; Ossowski et al., 1973) or glycosidases detected in transformed cells (Warren, 1969) may expose to the immune system antigenic determinants that previously were buried in the cell membrane (Uhlenbruck et al., 1970; Burger, 1969). Thus glycoproteins or glycolipids, which are altered because the tumorigenic cell has decreased or enhanced enzymatic activities (Levine, 1973), can then serve as tumor antigens. 2. Several types of tumors have been shown to express proteins or glycoproteins that are predominantly found in,fetal tissue. The increased expression of alpha fetoprotein in humans or animals with hepatomas (Abelev et a/., 1979) or of carcinoembryonic antigen (Neville and Laurence, 1974) in some colon carcinomas is presently utilized in diagnostic tests in humans. Serum from multiparous females can react with some tumor antigens (Coggin and Ambrose, 1979), or immunization of animals with stage-specific fetal tissues can activate a cell-mediated immunity, resulting in tumor rejection upon subsequent challenge of that animal (Coggin and Ambrose, 1979). Antigens which are expressed solely or predominantly durung fetal development may indeed be recognized as foreign by a mature immune system and therefore qualify as tumor antigens. The debate over whether to classify these antigens as reexpression of stage-specific fetal gene markers or an antigens associated more correctly with cell proliferation in either the fetus or adult animal, still continues (Hirai, 1979; Stillman and Sell, 1979). The available evidence suggests that adult liver can synthesize high levels of alpha fetoprotein even in the absence of DNA synthesis (i.e., growth) (Watanabe et a f . , 1975). On the other hand, the so-called embryonic form of thymidine kinase (Bull et al., 1974) clearly is related to the requirement for cell replication and is regulated by or with the cell cycle or cell growth (Postel and Levine, 1975). A number of isozymes such as glucose-ATP phosphotransferase, alkaline phosphatase, pyruvate kinase, and aldolase have both specificity of tissue

TRANSFORMATION-ASSOCIATED TUMOR ANTIGENS

77

distribution and adult versus fetal forms (Stillman and Sell, 1979). As such, they certainly qualify as tumor-specific enzyme markers and could under the appropriate conditions be tumor antigens. 3. Increased concentrations of normal cellular protein in transformed cells : Recently a wide variety of transformed cell lines have been shown to express high levels of a cellular protein termed p53 (53,000 MW) (Lane and Crawford, 1979; Linzer and Levine, 1979; DeLeo et al., 1979) which is found in much lower levels in nontransformed cells in culture (Linzer et al., 1979). Syngeneic mice hyperimmunized with a chemically transformed cell line (Meth A) produce antibodies to p53 (DeLeo et al., 1979). Similarly, BALB/c mice bearing tumors derived from an injection of SV40-transformed BALB/c-3T3 cells also produce anti-p53 antibodies (Linzer and Levine, 1979). Both the Meth A- and SV40-transformed BALB cell lines contain higher levels of p53 (10- to 100-fold higher) than normal cells (DeLeo et al., 1979; Linzer and Levine, 1979). This brings up the possibility that increased levels of a normal cellular protein can now act as an antigen and result in antibody production. In the case of SV40-transformed cells, the p53 protein is physically complexed with a viral protein, the SV40 large T antigen (Lane and Crawford, 1979; Linzer and Levine, 1979). 4. It is possible that the viral-cellular protein multimeric complex presents new antigenic determinants to the immune system that result in antibody production against the p53 antigenic determinants. This carrier-haptene (viral-cellular protein complex)-like concept could also result in the creation of a new tumor antigen (Linzer and Levine, 1979). Alternatively, increased levels of a cellular protein normally found in low levels might also trigger an immune response. 5. Perhaps the best understood case of why tumors present new antigens to the immune system is when the tumor cells contain new or rearranged genetic injormation, peculiar to the tumor cell. This phenomenon happens under normal circumstances, as immunoglobulin V and C genes create new idiotypic determinants which then may be recognized as foreign by B or T cells. Indeed this type of network theory of the immune system has been suggested to play a role in normal regulation of antibody production (Jerne, 1971). The rearrangement of genetic information, mediated by insertion elements (IS elements), has been suggested as a major cause of human cancer, possibly promoted by carcinogens or environmental factors (Cairns, 1981). The movement of a genetic element could either create a new cellular gene product or activate the synthesis of a gene product normally not produced in an adult animal. In the best documented case, with avian long latency leukemia viruses, a virus or a part of a virus (IS element-like) can activate new cellular genes by virtue of its integration into the cellular genome in the vicinity of that cellular gene (Nee1 et al., 1981; Payne et al., 1981). This is

78

ARNOLD J . LEVINE

presently believed to be the mechanism underlying the causes of long latency leukemias caused by virus infection in chickens (Nee1 et a/., 1981 ; Payne et al., 1981). The clearest case of why a new tumor antigen is expressed in a tumor results from the acquisition of new genetic information introduced by viruses. The virus may carry a copy of genetic information either identical or closely related to a normal cellular gene, as is the case with some of the RNA tumor viruses (Tooze, 1981). Alternatively, the virus may introduce new genetic information into a tumorigenic cell that is not detectably related to any genes in the cellular genome, as is the case with the DNA tumor viruses (Tooze, 1981). Rous sarcoma virus contains a gene (pp60v-src)that is closely related, but not identical, to the cellular gene (pp6OcpSrc) from which it likely was derived. The viral or v-src gene product has a number of peptides which differ from the cellular or c-src protein (Karess and Hanafusa, 1981). 6. Some chickens bearing sarcomas induced by this virus produce antibodies to pp60v-src,perhaps because of these direrences in the viral and cellular src proteins. However, it is clear that antibodies are not only made to the differences between viral and cellular src gene products in such animals. Newborn rabbits injected with the appropriate Rous sarcoma virus strain develop tumors and make antibodies against the pp60 chicken v-src gene product (Brugge and Erikson, 1977). These tumor-bearing sera (TBR sera) not only recognize the viral pp60'-"" but react with cellular chicken or more important rabbit c-src found in normal cells (Collett et al., 1978). Thus the rabbit produces antibodies to antigenic determinants common to its own pp60"-"", perhaps because a variation of its own cellular protein was presented to the immune system. In this context, the levels of pp60'-src in a tumor cell are much higher than that found in a normal cell (pp60c-src) (Collett et al., 1978), and these enhanced levels may also play a role, as appears to be the case with the p53 protein. 7. The D N A tumor viruses like simian virus 40 (SV40), adenovirus, or Epstein-Barr virus (EBV) introduce a new genetic information into the tumor cells and express viral proteins that are clearly and classically foreign antigens to the immune system of the host. These tumor antigens, found on both the cell surface (TSTA) (Tevethia and Tevethia, 1977) and intracellularly, can result in transplant rejection that is specific to the virus but not individual tumors or even the species of the tumor host cell. These are the most clearly understood examples of why tumors express foreign antigens and the host then mounts an immune response. This brief summary demonstrates that the origins of tumor antigens peculiar to tumor tissue may be quite diverse: (1) alterations of normal cellular components as a result of changes in the tumor cell physiology; (2)

TABLE I CLASSIFICATION OF TUMOR ANTIGENS' Tumor antigen classes Unique to a tumor

Transforming agent SV40 (2.5 kb early region genes)

Unique to a class of tumors

Specific to a transforming agent

1. > p53 levels 2. T-p53 complex

-

1. >p53 levels 2. Elb-58K-p53 complex

-

42K,28K 2. Elb*-58K,15K 1. EBNA, 48K, 65-72K 2. LDMA* (in cirro)

1. > p53, levels 2. 48K-p53 complex

1. p120*

1. >p53 levels

1. pp60'-s'c

1. >p53 levels

1. Large T antigen*

-

Common to many tumors

2. Small t antigen 3. t-56K-36K complex Adenovirus (2.9 kb early region genes Ela, Elb)

-

EBV (10% of viral genome expressed in transformed cells)

-

A k l s o n virus (I 5.5 kb, derived from the cellular genome)

-

RSV ( 21.8 kb, derived from the cellular genome)

-

1. Meth A*

Methylcholanthrene carcinogen (Meth A) Teratocarcinoma (genetic predisposition)

-

1. Ela-54K,52K,48K,

Developmental antigen* (B cells)

1. Gt-l,2,(B6)* (EC cells)

-

1. >p53 levels

-

1. > p53 levels

~

A summary of the tumor antigens derived from the seven different systems considered in the text. The tumor antigens are classified (Section IX, A) in one of four categories and the antigens involved or responsible for tumor transplantation rejection are indicated by an asterisk (*). These TSTAs can be localized in any of the three restricted classes of tumor antigens. To date there are no good examples of TSTA activities common to all tumors. A-placed in a box does not imply that an antigen is absent. Indeed SV40-transformed cells have been shown to express stage-specific fetal antigens (Coggin and Ambrose, 1979). In this case, the-indicates the poorly characterized nature of these fetal antigens. and therefore the unclear position of these antigens in this classification. For Abelson virus-transformed cells both the developmental antigen and p120 are on the cell surface and one or both of these antigens may be responsible for tumor transplant rejection.

80

ARNOLD J . LEVINE

reexpression of fetal- or growth-related proteins in an adult animal; (3) increased levels of a normal cellular protein in tumor tissue; (4) physical association of normal cellular protein and a foreign protein in a complex resulting in antibody to the normal cellular protein (carrier-haptene-like) ; ( 5 ) presentation of a slightly altered cellular protein (pp60v-src)resulting in antibody production to the normal cellular protein (pp6OcpS");(6) rearrangement of the genetic information in tumor cells by IS elements or viruses (e.g., avian leukosis virus, ALV) could result in a new antigen being produced or the reexpression of a cellular protein not normally found in the adult animal ; (7) The acquisition of foreign genetic information (DNA viruses) in tumor cells leading to the expression of viral (foreign) proteins that trigger an immune response (see Table).

B. ARETHESETUMOR ANTIGENS USEFUL IN SORTING OUT THE PRIMARY EVENTS OF TUMORIGENESIS? One can see from this discussion that tumor antigens may be a fingerprint or a reflection of what is happening at the genomic level in a tumorigenic cell. Antibodies produced in tumor-bearing animals can identify a virus present in tumor cells, localizing new cellular genes which are turned on because of genetic rearrangements or proliferation of that cell. Altered carbohydrate structures of cellular glycolipids or glycoproteins may reflect the physiological changes of a tumor cell, which may then be useful in understanding control of cell division. It is equally true, however, that some of these antigens may be irrelevant to either the primary or secondary causes of the tumor or even the control of tumor growth. Sorting out the important variables involved in tumor antigen expression becomes one of the essential tasks of present-day studies. C. TUMORANTIGENS AS DIAGNOSTIC TOOLS Whether tumor-specific antigens reflect primary events in tumor formation or not, they can be useful in therapy or diagnosis of the tumor. Monoclonal antibodies to tumor-specific antigens have been used in attempted therapy (Sears et al., 1982a) diagnosis (Koprowski er al., 1981), or for localizing metastatic lesions. Only further studies will determine if these approaches will have practical significance in tumor management at the clinical level (Herlyn et al., 1982; Sears et al., 1982b). D. VIRALAND CELLULAR ASSOCIATED TUMOR ANTIGENS At the basic science level the study of tumor antigens is leaving the descriptive phase of characterizing proteins or glycolipids expressed in tu-

TRANSFORMATION-ASSOCIATED TUMOR ANTIGENS

81

morigenic cells. Antibodies generated by tumor-bearing animals are being used to identify functional properties of tumor antigens like pp60"-"c and its role as a protein kinase (Collett and Erikson, 1978; Levinson et al., 1978) and not just an antigen. As these tumor antigens are better characterized, they have been found to be associated with cellular or viral proteins. These complexes of viral and cellular proteins detected in cell extracts presumably reflect functional activities or interactions in a tumorigenic cell. The isolation, characterization, and functional significance of these protein complexes is under intensive study at this time. This article will focus on cellular proteins that have been found to be associated with viral tumor antigens in cell extracts. The biochemical and functional properties of the viral proteins have been, in some cases, well characterized by the availability of viral mutants. We can say with some confidence that these viral proteins play a primary role in cellular transformation (see table). As such, the cellular proteins that interact with these viral tumor antigens may well play a central role in transformation. In addition, this article will include two examples of nonviral transformed cells: methylcholanthrene-induced transformation and teratocarcinomas. These examples are included for two reasons: (1) these tumors and transformed cells share a cellular tumor antigen (p53) in common with the viral transformed cells; and (2) they represent examples of tumor-specific transplantation antigens (Meth A antigen, Gt-1 antigen) detected by transplantation rejectien methods. This class of antigen, for which antibody is usually not available as an experimental tool, nevertheless provides new insights into the approaches and results of tumor immunology. Thus tumor antigens, detected by both the humoral and cellular responses of the immune system, will be discussed in this article (see table). II. Simian Virus 40

Simian virus 40 (SV40) encodes the genetic information for two gene products expressed at early times after productive infection or in transformed cells (see table). The gene A product, or large tumor antigen (T antigen) is about 94,000 MW (94K) in SDS-polyacrylamide gels (Tegtmeyer, 1974). The second protein, or small t antigen, has a molecular weight of 17,000 (17K) (Prives et al., 1977; Crawford et al., 1978) and shares common amino acid sequences with large T antigen at its N-terminal end (Paucha et al., 1978). The nucleotide sequences that contain the information for the small t antigen are found beteen 0.67 and 0.54 map units on the genome (Berk and Sharp, 1978), whereas the genetic information for the large T antigen is located in two noncontinuous segments of the SV40 genome between 0.67-0.59 and 0.54-0.14 map units (Berk and Sharp, 1978). Temperaturesensitive and deletion mutants in the large and small tumor antigen genes (Tegtmeyer et al., 1975; Rundell et al., 1977; Shenk et al., 1976) have been

82

ARNOLD J . LEVINE

utilized to demonstrate that both proteins are required for a fully transformed phenotype (Tegtmeyer, 1975; Brugge and Butel, 1975; Kimura and Itagaki, 1975; Martin and Chou, 1975; Osborn and Weber, 1975; Sleigh et al., 1978). Viral DNA derived from restriction enzyme fragments of the genome located between 0.14 and 0.72 map units will transform cells in culture (van der Eb, 1979), indicating that the early viral genes (T and t antigens) are sufficient for the production of a fully transformed cell phenotype. A. LARGET-ANTIGEN-ASSOCIATED PROTEIN, p53 Animals bearing SV40-induced tumors produce antibodies to both the large and small tumor antigens (Black et al., 1963; Tegtmeyer, 1974; Prives et al., 1977), which have been utilized to detect these proteins by immunoprecipitation reactions. When SV40-transformed mouse cells were labeled with [35S]methionineand the soluble proteins were incubated with antisera from animals bearing SV4O-induced tumors, three proteins were reproducibly immunoprecipitated : the large T antigen (94K), small t antigen (17K), and a cellular protein, termed p53 (53K) (Linzer and Levine, 1979; Lane and Crawford, 1979). It was shown that p53 is a cellular protein because the same protein was detected in embryonal carcinoma cells not infected with SV40 (Linzer and Levine, 1979). Nontransformed cells in culture (primary or 3T3 cells) contained lower levels (1/10-1/100th) of p53 than detected in the homologous transformed cell line (Linzer et al., 1979). Indeed a wide variety of murine-transformed cells were found to express high levels of p53 (De Leo et al., 1979; Maltzman et al., 1981) independent of the transforming agent. Homologous proteins related to p53 were detected in SV40-transformed cells derived from human, monkey, mouse, rat, and hamster cells (Linzer and Levine, 1979; Kress et al., 1979; Melero et al., 1979; Smith et al., 1979; Simmons et al., 1980; Gurney et al., 1980), and in human cells derived from a wide range of human tumors (Crawford et a]., 1981; Dippold et al., 1981). Thus p53 is a cellular protein expressed in high levels in transformed cells, independent of the transforming agent, and a homologous protein can be detected in many different species of host cells (see table). In SV40-transformed cells, the cellular p53 tumor antigen is physically complexed with the SV40 large tumor antigen (Lane and Crawford, 1979; Linzer and Levine, 1979). Monoclonal antibodies directed against the SV40 large T antigen, and which do not cross-react with p53, immunoprecipitate the large T antigen and coimmunoprecipitate the 'associated p53 protein (Linzer and Levine, 1979). Conversely, monoclonal antibodies to p53 immunoprecipitate that protein and coimmunoprecipitate the viral large T antigen (Gurney et al., 1980; Maltzman et al., 1981). The T antigen-p53

TRANSFORMATION-ASSOCIATED TUMOR ANTIGENS

83

complex sediments rapidly and in a heterogeneous fashion (McCormick and Harlow, 1980), suggesting a high-molecular-weight, multimeric mixture of at least the SV40 T antigen and p53 proteins. In transformed mouse cells most but not all of the SV40 T antigen can be found in association with p53, while all of the cellular p53 molecules reside in this complex (McCormick and Harlow, 1980; Oren et al., 1981). Because serum from animals bearing SV4O-induced tumors contain antibody to the SV40 large T antigen (Tegtmeyer, 1974), it remained possible that the tumor-bearing animals produced antibody to the SV40 large T antigen, which then coimmunoprecipitated the associated p53 protein. In this case no antibody would be made against p53 itself. That this was not the case was demonstrated by utilizing sera from animals bearing SV40-induced tumors to immunoprecipitate p53 from embryonal carcinoma cells which do not contain the SV40 T antigen (Linzer and Levine, 1979). In spite of one report to the contrary (Melero et al., 1979), the SV40 T antigen and p53 have no detectable cross-reacting antigenic determinants. Thus many animals bearing SV40-induced tumors produce antibodies to a normal cellular protein, p53 (Linzer et al., 1979). The possible reasons for this have been discussed in the previous section (1,A). Nontransformed cells as well as transformed cells express p53. The difference between these cells is that most transformed cell lines examined to date (but not all transformed cell lines) express 10- to 100-fold higher levels of p53 than their corresponding nontransformed cell counterpart (Linzer et al., 1979; Oren et af., 1981). In SV40-transformed cells the higher levels of p53 detected require a functional viral large T antigen (Linzer et al., 1979). SV40ts.4 mutants, defective in the large T antigen, regulate, in a temperaturedependent fashion, both the initial increase in the levels of p53 after virus infection and the maintenance of high levels of p53 in transformed cells (Linzer et al., 1979). Thus the SV40 large T antigen, known to be required for maintaining the transformed state (Tegtmeyer, 1975), also regulates the levels of a p53 cellular protein in these cells. Because many different transformed cell lines express high levels of p53 the mechanisms regulating the level of this protein in SV40-transformed cells are of some interest. SV40 T antigen could regulate the amount of p53 produced at the transcription level of the p53 gene, RNA processing, mRNA stability, translation of p53 mRNA, or even at post translational steps such as protein modification or stability of p53 protein (turnover). To investigate these possibilities the levels of p53 mRNA in SV4O-transformed (SV3T3 cells) and nontransformed cells (3T3 cells) were compared. Messenger RNA from 3T3 cells and SV40-transformed 3T3 cells (SVT2) were translated in a reticulocyte in uitro translation extract. Translation of p53 mRNAs was specifically measured by immunoprecipitation of the radioactively labeled p53 protein synthesized in uitro. The peptide maps of p53,

84

ARNOLD J . LEVINE

synthesized in v i m or in vitro using mRNA from the same cell line, were similar (Oren et a/., 1981). The levels of the p53 immunoprecipitated protein synthesized in oitro were proportional to the amount of mRNA (over 1-10 pg) added to the reticulocyte reaction mixture. Therefore, the levels of p53 synthesized in vitro accurately measure the level of translatable p53 mRNA in a preparation extracted from different cell lines. When p53 mRNA levels in nontransformed 3T3 cells and SV40-transformed 3T3 cells were compared in this way, both cell types had about equal amounts of translatable p53 mRNA (Oren et ul., 1981). This was in spite of the fact that SV3T3 cells contain as much as 100-fold more p53 protein than the 3T3 cells when the protein was labeled with [3sS]methionine in vivo (in cell culture) (Linzer et al., 1979; Oren et al., 1981). The reason for this appears to be that the levels of p53 in 3T3 and SV3T3 are regulated at a posttranslational step. When 3T3 and SV3T3 cells were pulse labeled with [35S]methionineand then the cell culture was placed in unlabeled medium (chase period), the half-life of p53 protein was found to be 20-60 min in 3T3 cells and more than 22 hr in SV3T3 cells (Oren ef a/., 1981). Thus it appears that the rate of synthesis (and mRNA levels) of p53 is similar in 3T3 and SV3T3 cells, but p53 protein stability differs dramatically in these two cell lines, leading to differing levels of p53. Because antibody to p53 is used to detect and measure the levels of this protein in all these experiments, it remains possible that the p53 protein is not actually degraded with time, but is altered so as not to react with the antibody, or fails to be extracted from cells and would no longer be detected by this assay (Oren et al., 1981). There are at least two possible reasons that p53 stability (or reactivity) is enhanced in SV3T3 cells compared to 3T3 cells : (1) there is some suggestion that posttranslational protein modifications of the p53 protein from SV3T3 cells differ from those of other cells (Maltzman et a/., 1981) and this could alter the protein turnover of p53 in SV3T3 cells; or ( 2 ) SV40 T antigen, which regulates the levels of p53 in a transformed cell (Linzer el a/., 1979), is physically associated with p53 in SV3T3 cells and this protein complex could alter the stability or turnover of p53 in transformed cells (Oren et ul., 1981).

B.

SMALL t-ANTIGEN-ASSOCIATED PROTEINS,

56K

AND

32K

Deletion mutants in the SV40 small t-antigen gene have been employed to study the functions of this protein. Although the small t antigen is not required for virus replication in cell culture (Shenk et al., 1976), it is essential for efficient transformation of growth-restricted cells (Bouck et a/., 1978; Sleigh et al., 1978; Martin et a/., 1979). Microinjection of small t antigen into cells in culture results in the disorganization of actin cables in these

TRANSFORMATION-ASSOCIATED TUMOR ANTIGENS

85

cells (Graessmann et al., 1980), and this observation is consistent with the predominantly cytoplasmic location of small t antigen in infected and transformed cells. The SV40 small t antigen in both productivity infected and transformed cells is physically associated with two cellular proteins of 56,000 and 32,000 MW (Yang et al., 1979). If unlabeled small t antigen is added to an extract of [3 'Slmethionine-labeled uninfected cell proteins, and antibody to the small t antigen is employed to immunoprecipitate it, the labeled 56,000 (56K) and 32,000 (32K) MW proteins associate in v i m with the small t antigen and coimmunoprecipitate along with it (Yang et al., 1979). The same small t-antigen-56K-32K complex is detected in SV40-infected and -transformed cell extracts (Yang et al., 1979). Small t-antigen deletion mutants fail to produce a small t-56K-32K complex and these two cellular proteins are not detected by this immunoprecipitation assay (Yang et al., 1979). These experiments also tend to eliminate a role for the large T antigen, which is not altered by these small t-antigen deletion mutants, in associating with the 56K-32K proteins. The 56K cellular protein complexed with small t antigen is most likely different from the cellular p53 antigen that is associated with SV40 large T antigen. The small t-antigen deletion mutants fail to produce a t-56K-32K complex but have a normal large T-p53 complex (Linzer et al., 1979). In addition, p53 is a nuclear antigen, in both normal and SV40-transformed cells, whereas the 56K and 32K proteins have been localized in the cytoplasm (Rundell and Yang, 1981). Finally, antisera to the 56K and 36K antigens failed to immunoprecipitate p53 from embryonal carcinoma or chemically transformed cell lines (Rundell and Yang, 1981). Thus it is likely that both the SV40 large T and small t antigens associate with different cellular proteins and this presumably reflects a functional interaction. The small t antigens from the related papovaviruses, BK and polyoma, also form complexes with the cellular 56K and 32K proteins (Rundell et d., 1981). Virtually all of the antisera from animals bearing SV40-induced tumors contain antibody directed against the large T antigen. A smaller subset of these sera will immunoprecipitate the small t antigen. Some tumor sera also contains antibodies directed against the cellular p53 antigenic determinants (Linzer and Levine, 1979), and in rare cases antisera will contain antibodies that immunoprecipitate the 56K and 32K proteins from uninfected cell extracts (Rundell et al., 1981). Antibodies may be made against these cellular proteins because (a) they are complexed with viral proteins and therefore act as a stronger immunogen, (b) the cellular protein (p53) is present in higher than normal concentrations in tumor cells, (c) the proteins in normal cells and tumor cells may differ (modification for example) in some as yet

86

ARNOLD J . LEVINE

undetected manner, or (d) abnormal tumor tissue necrosis exposes these proteins or fragments of these proteins to the immune system in a form not normally observed. Ill. Adenoviruses

The human adenovirus DNA sequences derived from the left-hand 8- 12% of the viral genome contain the genetic information that is sufficient for transformation of cells in culture (Graham et al., 1975; Sambrook et al., 1975; Flint et al., 1976). This portion of the viral genome has been subdivided into two regions, termed Ela and Elb, based on the two discrete transcriptional units encoded therein (Berget e f al., 1977; Chow et al., 1977) and two genetic complementation groups defined by host range mutants (Harrison et al., 1977; Jones and Shenk, 1979; Ross et al., 1981b). The E l a transcripts produce at least five proteins with molecular weights of 54,000, 52,000, 48,000,42,000, and 28,000 (Harter and Lewis, 1978).The E l b transcripts encode the information for a 58,000 and a 15,000 MW set of proteins (Gilead et al., 1976; Levinson and Levine, 1977a,b; Schrier et al., 1979; Ross et al., 1980a,b). Some or all of the Ela gene products are required to turn on the E l b gene products and other early adenovirus genes in a positive fashion (Berk et al., 1979; Jones and Shenk 1979; Ross et al., 1980a). One or more of the E l b gene products are required for viral DNA replication (Jones and Shenk, 1979), and modulate down the levels of adenovirus early proteins produced by the E2 and E3 viral genes (Ross et al., 1980a). Both Ela and Elb viral gene products are required for efficient transformation of cells in culture (Graham et al., 1978; Jones and Shenk, 1979) and are expressed in transformed and tumorigenic cells (Ross eta/., 1980b). For this reason, these proteins have been termed the adenovirus tumor antigens (see table). In adenovirus-transformed mouse, rat, and hamster cells, the El b-58K protein is physically associated with the same cellular p53 protein that is found in the SV40 T antigen-p53 complex (Sarnow et al., 1982). Several monoclonal antibodies to the p53 protein (Gurney et al., 1980; Coffman and Weissman, 1981) immunoprecipitate p53 and coimmunoprecipitate the adenovirus El b-58K protein. Conversely, monoclonal antibodies to the El b-58K protein immunoprecipitate that protein and coimmunoprecipitate p53 in that complex (Sarnow, unpublished results). Partial peptide maps of the murine p53 proteins that were complexed with either the SV40 large T antigen or the adenovirus El b-58K protein, demonstrated that these p53 proteins were either closely related or identical (Sarnow et al., 1982). Thus, based on both immunological and chemical (peptide maps) evidence, the SV40 large T antigen and the adenovirus El b-58K protein are both physically associated with the same cellular protein, p53, in their respective transformed murine cell lines (Sarnow et al., 1982).

TRANSFORMATION-ASSOCIATED TUMOR ANTIGENS

87

The adenovirus El b-58K-p53 complex and the SV40 large T-antigen-p53 complex share several properties. Both complexes are heterogenous in their size and mass (sedimentation rate is 25-30 S) and appear to be composed of multimeric combinations of at least these two proteins (McCormick and Harlow, 1980; Sarnow et al., 1982). In both the SV40- and adenovirustransformed cells, all of the p53 detected in these cells is in the form of a high-molecular-weightcomplex, whereas a variable proportion of the SV40 and adenovirus T antigens have been detected in a free or noncomplexed form (McCormick and Harlow, 1980; Sarnow et al., 1982). In the case of SV40-infected and -transformed cells, the increased levels and stability of p53 may be due to its physical association with T antigen (Oren et al., 1981), and so no free form of p53 protein would be expected. The enhanced levels of p53 in the adenovirus-transformed mouse cells, when compared to untransformed 3T3 cells, could also be due to the association of p53 with the Elb-58K viral tumor antigen. This has yet to be clearly demonstrated, however. The SV40 large T antigen and the adenovirus Elb-58K protein are both modified by the addition of phosphate to these proteins (Tegtmeyer, 1975; Levinson and Levine, 1977a,b). The phosphorylation of these two proteins is heterogeneous with some molecules containing more moles of phosphate per molecule than others (Tegtmeyer et al., 1977; Walter and Flory, 1979; Levinson and Levine, 1977a,b). The SV40 large T antigen associated with p53 appears to be more heavily phosphorylated than the free T-antigen species detected in the cell (Fanning et al., 1981). On the other hand, the adenovirus Elb-58K protein associated with p53 appears to contain less phosphate per molecule than the free or noncomplexed form of this E l b protein (Sarnow et al., 1982). These results suggest that posttranslational modifications of the viral T antigens could play a role in the association of viral proteins with cellular proteins and presumably influence their function. The association of SV40 and adenovirus tumor antigens with the cellular p53 protein is dramatically influenced by the species of host cell employed for these studies. In mouse, rat, or hamster cells the T antigen-p53 complexes are very stable and are readily demonstrated. In adenovirus-infected or -transformed human cells, most or all of the Elb-58K protein is not complexed to p53 (Sarnow et al., 1982). In SV40-infectedor -transformed human or monkey cells the T antigen-p53 complex either dissociates readily in uitro or higher levels of free T antigen and p53 are detected by immunoprecipitation of these soluble extracts (Gurney et al., 1980; Levine, unpublished results). In SV40-transformed human cells (SVSO), the stability of the T-antigen-p53 complex in crude extracts is significantly weaker than in SV40-transformed mouse cells (Levine, unpublished results), even though in freshly prepared lysates virtually all of the p53 is complexed to T antigen

88

ARNOLD J . LEVINE

in both human and mouse cells. It appears likely then that the T-antigen-p53 complexes from human and monkey cell extracts have a lower stability than when the same viral T antigen is present in murine or rat cells. Because SV40 and adenoviruses have evolved for their replication in monkey and human cells respectively, the evolutionary significant host is of primate origin. This suggests that the low-stability, association-dissociation reaction, is a more significant one and reflects some function. It also remains possible that the unusually tight association of T antigen with p53 in mouse or rat cells leads to an abnormal situation, i.e., transformation rather than replication. This type of reasoning could explain why a lytic virus for one species can induce tumors in a different organism. The SV40 large T antigen and the adenovirus Elb-58K protein have several functions in common. Both proteins are involved in viral DNA replication (Tegtmeyer, 1972; Jones and Shenk, 1979), and both proteins can modulate early viral gene expression (Tegtmeyer, 1975; Ross et al., 1980a). In addition, the SV40 large T antigen can stimulate cellular DNA synthesis (Tjian et al., 1978) and modulate up cellular transcripts in infected or transformed cells (T. Schutzbank, R. Robinson, and A. Levine, unpublished results). The viral T-antigen-p53 complexes observed in transformed cells could act as intermediates in the DNA replication and/or gene regulation process reflecting the phenotypes of these viral tumor antigens. IV. Epstein-Barr Virus

The virion DNA obtained from different strains of Epstein-Barr viruses (EBV) or different cell lines transformed by EBV, is heterogeneous with regard to the number of repeated DNA sequences, deletions, or insertions and restriction enzyme polymorphisms (Raab-Traub et al., 1978;Dambaugh et al., 1980). For that reason, the major features of EBV DNA will be outlined with emphasis placed on describing the properties of the B95-8 virus isolate which has been extensively studied (Dambaugh et al., 1980; Skare and Strominger, 1980; Kintner and Sugden, 1979). The genome from this virus is about 115 x lo6 daltons (Raab-Traub et al., 1978; Dambaugh et al., 1980),linear, double-stranded DNA. Accordingly, molecular weight map coordinates from 0 to 115 have been assigned to identify regions of the genome. Both ends of the genome contain a variable number of direct tandem repeats (TRor terminal repeats) of about 500 base pairs each (Given et al., 1979), located at 0-1 map units and at 114-1 15 map units. About 10 x lo6 daltons from one end (10-26 map units) is located a second variable number of tandem internal repeats (IRor internal repeats). The I, region contains a 3000 base pair repeated sequence which is distinct in its polynucleotide sequence from the TR repeats (Dambaugh et al., 1980). These I , sequences

TRANSFORMATION-ASSOCIATED TUMOR ANTIGENS

89

then divide the remainder of the viral genome into two regions, a 9 x lo6dalton short unique DNA region ( U s , 1-10 map units) and a 80 x lo6dalton long unique region (U,, from about 26 or 28 to 110 or 114 map units). The B95-8 DNA differs from several other EBV isolates (P3HR1, W91, and AG876) in that it contains a 9 x 106-dalton (1 1,000 base pair) deletion in the UL region (at about 93-95 map units) (Dambaugh et al., 1980). This deleted DNA in B95-8 contains some nucleotide homology with B95-8 DNA located near the ZR region at about 28 map units (Dambaugh et al., 1980). Thus the UL regions of some EBV isolates contain regions of homology at 28 and 94 map units. It should be recalled that the TR and IRsequences can vary in size and so the length and map coordinates given herein are approximate. In transformed lymphocytes the EBV DNA is present in a circular episoma1form (Lindahl ef al., 1976) in copy numbers varying from 2 to 200 per cell (zur Hausen et al., 1970; Lindahl et al., 1974; Nonoyama and Pagano, 1973). Similarly, cell lines developed from biopsies of nasopharyngeal carcinomas contain circular EBV episomal DNA (Kaschka-Dierich et al., 1976). In lymphocytes transformed by EBV (IB-4 cell line), about 30% of the EBV circular genome is transcribed to produce detectable levels of RNA (King et al., 1980). Of this, only 10% of the EBV genome produces RNA species that are polyribosome associated (van Santen et al., 1981). At least six polyadenylated cytoplasmic RNAs have been identified and derive from (1) the ZR-UL region (5-30 map units), (2) the TR region between 110 and 03 map units of the circular genome (possibly UL-TR-US sequences), and (3) a U, transcript at 63-66 map units. Less abundant RNAs have also been detected from additional regions of the genome (van Santen et al., 1981). Similar transcriptional patterns have been described in several Burkitt tumor cell lines (King et al., 1981). Two low-molecular-weight RNAs (180 nucleotides long), called EBER-1, 2, which are detected as ribonucleoprotein particles in the cytoplasm (Lerner et al., 1981), have been mapped to the Us segment of the genome (van Santen et al., 1981;Lerner et al., 1981). Based on these studies, it is clear that only a small segment (about 10%) of the EBV genome encodes information expressed in the transformed or tumorigenic cell. Employing antisera from patients with either Burkitt’s lymphoma or infectious mononucleosis, several EBV-associated antigens have been described. An EBV nuclear antigen, termed EBNA (Pope et al., 1969;Reedman and Klein, 1973), is invariably associated with transformed cells, Burkitt’s lymphoma, and nasopharyngeal carcinoma. An EBNA species has been purified utilizing an acid-fixed nuclear binding assay (Luka et al., 1977), and the native antigen appears to be a high-molecular-weight (170,000-230,000 MW) basic protein derived from monomeric subunits of 48,000 MW (Luka et al., 1978; Baron and Strominger, 1978). When EBNA is purified utilizing

90

ARNOLD J . LEVINE

mild procedures, the 48,000 MW subunit copurifies with a 53,000 MW protein (Luka et al., 1980). The 48,000 MW protein was only detected in EBVpositive cell lines, whereas the 53,000 MW protein was detected in a number of transformed cell lines with or without EBV present (Luka et al., 1980). The 53,000 MW protein, found in an oligomericcomplex with the 48,000MW EBNA protein, was shown to be similar to the p53 cellular antigen derived from the Meth A murine cell line (Jornvall et af., 1982). The Meth A-p53 and the SV40 T-antigen-p53 associated proteins are known to be similar or identical from their peptide maps (Maltzman et al., 1981). Thus it appears that the 48,000 MW EBNA can also be associated (and copurify) with a p53 protein, as is the case with SV40 and adenovirus tumor antigens. Recently, Strnad et al. (1981) have utilized human sera to identify an EBNA antigen by fluoroimmunoelectrophoresis and radioimmunoelectrophoresis. They detected a 65,000 MW antigen in Burkitt’s lymphoma cells but not in EBV-negative B-cell lines. The partially purified 65,000 MW antigen was an effective immunoadsorbant for EBNA antibodies in human sera (Stmad et al., 1981). Different EBV-positive cell lines showed some molecular weight heterogeneity (65,000-73,000 MW) for this antigen (Strnad et al., 1981). The relationship of this 65,000 MW EBNA to the 48,000 MW protein purified by Luka et al. (1980) is at present unclear. The 48,000 MW EBNA could be a proteolytic breakdown product of the larger EBNA. Alternatively, because EBNA was originally defined and detected using polyclonal human antisera, EBNA could well be a large number of different proteins. When monoclonal antibodies to human p53( R. Thomas, C. Sullivan, and A. Levine, unpublished results) are used to immunoprecipitate p53 from an EBV-transformed lymphocyte cell line (Raji), the p53 protein is detected along with another protein of 65,000 MW (L. Kaplan and A. Levine, unpublished results). At present it is not clear whether this second protein is coimmunoprecipitated in a complex with p53 or the monoclonal antibody reacts directly with it. The molecular weight of this protein (65,000), found in the anti-pS3 immunoprecipitate, agrees very well with the EBNA antigen detected in these same cells (Raji) by Stmad et al. (1981). Taken together, all of these results (Luka et al., 1980; Jornvell et al., 1982; Strnad et al., 1981 ; L. Kaplan and A. Levine, unpublished) suggest that EBNA moieties may be physically complexed, in an oligomeric form, with the same p53 antigen associated with the SV40 and adenovirus tumor antigens. Other EBV-associated antigens, such as the early antigen (Henle et al., 1970; Gergely et al., 1971), viral capsid antigen (VCA) (zur Hausen, 1976), membrane antigens (Klein et al., 1968), lymphocyte-detected membrane antigen (Svedmyr and Jondal, 1975), .and an 81,000 MW DNA-binding protein (Strnad et al., 1981), all have been reported. Like EBNA, all of these

TRANSFORMATION-ASSOCIATED TUMOR ANTIGENS

91

antigens were detected using polyclonal human sera or T lymphocytes. As such it is difficult to determine how many different proteins these antigens represent or, in all cases, whether or not the EBV genome encodes the genetic information for these antigens. Although antigens like EBNA are clearly correlated with the presence of the EBV genome, and have not been reported in the absence of the viral genome, no formal proof has been presented to date to show that EBV DNA encodes these antigens. The main reason for this ambiguity is that specific antisera to defined antigenic moieties of EBV proteins is lacking and the apparent levels of both viral transcripts (van Santen et al., 1981) and antigens (Strnad et al., 1981) are quite low in EBVtransformed cells. In addition, no reliable EBV mutants are available to correlate genetic alterations with the viral encoded proteins or functions. V. Methylcholanthrene-InducedTransformed Cells (Meth A)

Murine sarcomas, produced by some polycyclic hydrocarbons like methylcholanthrene, often express tumor-specific transplantation antigens (Klein et al., 1960; Old et al., 1962).A remarkable feature of these chemically transformed cells is that each independently derived tumorigenic cell line expresses an antigen unique to that cell line as measured by tumor transplantation immunity tests in syngeneic mice. For example, two independently derived methylcholanthrene-inducedBALB/c sarcomas, Meth A and CMS4, contain cell surface antigens that provide transplant immunity only when a BALBjc mouse is immunized with the homologous tumor (De Leo et al., 1977, 1978). Cytotoxic antisera prepared against Meth A cells reacted with the Meth A cell line but not 19 other BALB/c sarcomas (De Leo et al., 1977, 1978). Similarly, cytotoxic antibodies directed against the CMS4 sarcoma cell line reacted with only 2 of 20 BALB/c sarcomas. The Meth A and CMS4 antigens were not related to Moloney leukemia virus antigens and were not detected in a wide variety of normal or transformed cell types (De Leo et al., 1977, 1978). Thus chemically transformed cells express a cell surface antigen which is unique or specific to the cell line of origin. These unique antigens are readily identified by tumor transplantation tests in syngeneic mice or in some cases by cytotoxic antibodies (De Leo et al., 1977, 1978). Chemically transformed cell lines such as Meth A also express a nuclear antigen (p53) that, in contrast to the unique antigen, is common to a wide variety of transformed cell lines (De Leo et al., 1979). BALB/c mice immunized with the BALB/c Meth A sarcoma produce antibodies to a p53 antigen. The peptide map of the methionine-containing tryptic peptides of the Meth A p53 and the p53 associated with SV40 T antigen were very similar or identical (Maltzman et al., 1981). Monoclonal antibodies prepared against p53 detected (by immunofluorescence) this protein in 13 diverse murine-

92

A R N O L D J . LEVINE

transformed cell lines but not in normal mouse fibroblasts, 3T3 cells, bone marrow cells, thymus cells, or embryo cells (Dippold et al., 1981). This monoclonal antibody cross-reacted with the homologous human p53 antigen, providing positive immunofluorescence in melanoma, renal cancer, breast cancer, and Burkitt’s lymphoma cell lines. Some normal human (but not the mouse) cell cultures were also positive by immunofluorescence (with this monoclonal antibody) for the presence of p53. Primary human kidney epithelium cultures showed positive nuclear immunofluorescence that declined in intensity over serial passage in culture (Dippold et al., 1981). Similarly, primary human fetal brain cells and primary human skin fibroblasts contained lower levels of p53, which were no longer detectable by the second passage in cell culture (Dippold et al., 1981). Dippold et al. (1981) interpret these results in support of the possibility that p53 in normal cells is related to active cell division and may be involved in the regulation of cell division. Milner and Milner (198 1) have come to a similar conclusion. Four hours after mitogenic stimulation of mouse spleen lymphocytes, increased levels of p53 were synthesized. This led to the suggestion that p53 is involved in the Go to G, transition or mitogenic transition (Milner and Milner, 1981) of lymphocytes. The fact that these studies, with normal human kidney cell cultures, were carried out with an immunofluorescent assay and a monoclonal antibody to p53 (Dippold et al., 1981) presents some possible problems. First, p53 is present and detectable in normal or primary mouse cultures at 1 to 10%of the levels observed in murine-transformed cells (Linzer et al., 1979). The fluorescent antibody test is difficult to quantitate, and so comparisons of normal and transformed cells by this test do not usually provide good quantitative results. Second, monoclonal antibodies can cross-react with unrelated proteins that may fortuitously share an antigenic determinant with the immunogen. For that reason, it would be useful to provide additional data (molecular weight of the protein, for example), which is not a feature of the fluorescent antibody assay. In spite of these difficulties, it may well be that primary normal mouse and human cell cultures differ in p53 levels and the high levels in human cells are a reflection of cell growth or cell cycle declining in subsequent culture passages. When p53 from either Meth A cells or SV40-transformed cells is immunoprecipitated with a monoclonal antibody (Dippold et al., 1981), the immunoprecipitates contain an associated phosphotransferase activity (Jay et ul., 1981). This associated protein kinase phosphorylates p53 in immunoprecipitates from Meth A extracts, and p53 and SV40 T antigen in immunoprecipitates from SV40-transformed cell extracts (Jay et al., 1981). The phosphorylation of these proteins is at serine and threonine residues (Jay el al., 1981). This result is potentially attractive in light of the reports that

TRANSFORMATION-ASSOCIATED TUMOR ANTIGENS

93

SV40 large T antigen (Tjian and Robbins, 1979; Griffin et al., 1979) and the adenovirus tumor antigens (Lassam et al., 1979) have an associated protein kinase activity in immunoprecipitated reaction mixtures. In the case of SV40 it has been possible to separate the SV40 T-antigen immunological activity from the protein kinase activity by conventional biochemical purifications of these proteins (Tjian and Robbins, 1979; Gracherio and Hager, 1979). If p53 was indeed a phosphotransferase itself, or was tightly associated with a protein kinase activity, this would explain the SV40 and adenovirus results. However, immunoprecipitates could well contain contaminating protein kinases, and the control experiments employed to eliminate contaminating enzyme activities do not totally rule out this possibility. Although p53 might have an associated protein kinase activity (Jay et al., 1981), better criteria will be required to prove that p53 itself has an intrinsic phosphotransferase activity. The presence of high levels of p53 in chemically transformed cells brings up some interesting questions. In nontransformed murine 3T3 cells, p53 is synthesized and then degraded rapidly with a half-life of 20-60 min (Oren et al., 1981). In SV40-transformed cells, the viral large T antigen both regulates the levels of p53 in transformed cells (Linzer et al., 1979) and is physically complexed with p53 (Lane and Crawford, 1979; Linzer and Levine 1979). The high levels of p53 in SV40-transformed cells could result from the inability of p53 to be degraded or turned over (Oren et al., 1981), because it is in a multimeric T-antigen-p53 complex. The half-life of p53 and T antigen in SV40-transformed cells is greater than 22 hr (Oren et al., 1981). If this explanation is correct, then what regulates the high levels of p53 in Meth A-transformed cells, which have no viral tumor antigens? Is p53 in Meth A cells associated with a cellular protein that stabilizes it and prevents rapid protein turnover? The search for p53-associated proteins in nonviral transformed cells, or in normal cells at specific stages of the cell cycle, has just begun. If successful, it may lead to clues concerning the function of p53 in normal and transformed cells. VI. Abelson Virus

Abelson murine leukemia virus (A-MuLV) is a defective retrovirus which induces an acute nonthymic lymphoproliferative disease in mice (Abelson and Rabstein, 1970). The virus can transform some fibroblast cell lines in culture (Scher and Siegler, 1975) and bone marrow-derived lymphoid cells (Rosenberg et al., 1975; Rosenberg and Baltimore, 1976). These Abelson virus-transformed lymphocytes resemble immature B cells and in some cases produce immunoglobulin chains (Premkumar et al., 1975). The Abelson-transformed lymphocytes express three different tumor antigens :

94

ARNOLD J . LEVINE

(1) a viral encoded protein called p120 (Reynolds et al., 1978; Witte et al., 1978); (2) a differentiation-specific antigen (Risser et al., 1978); and (3) the same p53 cellular antigen observed in a large number of transformed cell lines (Rotter et al., 1980). The defective genome A-MuLV encodes the genetic information for a 120,000-160,000 (p120) MW protein consisting of an N-terminal portion of the viral gag protein linked to a unique gene segment derived from the host (mouse) genome (Reynolds et al., 1978; Witte et al., 1979). Thus the Abelsontransforming protein is a fusion product between the viral gag genes and a portion of a cellular gene. Some deletion mutants in this protein fail to transform cells, suggesting that the p120, virus encoded protein is important in the transformation process (Rosenberg and Witte, 1980). Abelson virus was first isolated from BALB/c mice (Abelson and Rabstein, 1970), and so the murine genomic contribution to this virus is from a BALB/c genetic background. Abelson virus-transformed lymphocytes from C57BL/6 mice (Ll-2-MuLV tumor cells) produce antibodies in C57BL/6 mice against the p120 gag fusion product of the virus (Witte et al., 1979). These antibodies react with both the viral gag component and the host cell protein component (Witte et al., 1979). With this transformed cell line, in C57BL/6 mice, the tumor regresses and a strong p120 sera is obtained. The reason anti-p120 antibodies are directed against the cellular antigenic determinants in pl20 could be (1) the possible allogeneic differences between the BALB/c genetic material in the virus and the similar C57BL/6 gene, (2) the increased levels of p120 in A-MuLV lymphocytes compared to normal lymphocytes, (3) The anti-gag antibody response acting to enhance the immune response against the cellular antigeneic determinants, (4) the tumor rejection enhancing the humoral response, or (5) a combination of these factors. The p120 antigen has been localized both at the cell surface (Rotter et al., 1981) and associated with the detergent-insoluble cellular matrix fraction of A-MuLV transformed cells (Boss et al., 1981). A second, and most likely distinct, cell surface antigen found on A-MuLVtransformed lymphocytes appears to be a differentiation-specific antigen (Risser et al., 1978). C57BL/6 mice, immunized with A-MuLV-transformed C57BL/6 lymphocytes, produce antibodies that react with a cell surface antigen on A-MuLV lymphocytes and a subset of normal BALB/c bone marrow and fetal liver lymphocytes. This developmental antigen was not detected in C57BL/6 normal bone marrow or fetal liver cells, and so these antigenic determinants also behave as alloantigens in the BALB/c and C57BL/6 mice (Risser et al., 1978). The expression of this developmental antigen in normal cells correlates with the leukemogenic target cells (preB cells) of this virus (Rosenberg and Baltimore, 1976). In addition, the developmental alloantigen has been mapped by genetic crosses at or near

TRANSFORMATION-ASSOCIATED TUMOR ANTIGENS

95

the Av-2 locus which determines the host range (susceptibility) for A-MuLV leukemogenesis (Risser et al., 1978). The virus’ ability to induce leukemogenic changes in a limited set of pre-B cells that may express this developmental antigen and the genetic determination of susceptibility to this virus (Au-2)-both of which map at or near the gene for this differentiation antigen -suggests an important role of this tumor antigen in leukemogenesis. The third tumor antigen expressed at high levels in Abelson-transformed lymphocytes is the p53 protein (Rotter et al., 1980). BALB/c mice immunized with syngencic BALB/c A-MuLV-transformed lymphocytes produce hightiter anti-p53 antibodies. The p53 antigen from A-MuLV-transformed cells is similar or identical (by immunological criteria and peptide maps) to the p53 antigen associated with the SV40 large T antigen (Rotter et al., 1981). A monoclonal antibody to p53 used in these studies, RA3-2C2 (Coffman and Weissman, 198I), immunoprecipitated p53, and no detectable associated proteins, from A-MuLV-transformed cells. The RA3-2C2 monoclonal antibody also detected a cell surface antigen on a subset of normal mouse B lymphocytes (Coffman and Weissman, 1981) and on the cell surface of A-MuLV-transformed pre-B lymphocytes (Rotter et a[., 1981). The binding of the RA3-2C2 monoclonal antibody to the cell surface of some normal and A-MuLV-transformed lymphocytes appears to be due to a fortuitous crossreaction between an antigenic determinant on p53 and a B-cell developmental or differentiation antigen (D. Baltimore, personal communication). Other monoclonal antibodies to different p53 antigenic determinants (Gurney et al., 1980) do not recognize these cell surface antigens (D. Baltimore, personal communication), and so it appears that p53 in Abelson-transformed lymphocytes has an intracellular location as determined by cell fractionation studies (Rotter et al., 1981). These studies provide a good example of the potential problems of utiliqing monoclonal antibodies for immunofluorescent studies. Without a second criterion, like the molecular weight of the antigen under study, cross-reactions, which may or may not be significant, can provide false positives (see Section V). The Abelson-encoded p120 protein has an associated protein kinase activity in immunoprecipitates, transferring a phosphate moiety to tyrosine residues of the acceptor protein (Witte et al., 1980). P53 is a phosphoprotein but contains only phosphoserine residues and no detectable phosphotyrosine (Rotter et al., 1981), and it is therefore unlikely that the p53 antigen is a substrate for the A-MuLV kinase. Under the conditions employed to test the p120 protein for kinase activity, the p53 protein failed to act as a protein kinase when immunoprecipitated from Abelson-transformed cells (Rotter et al., 1980). This is not in agreement with the report of a p53-associated protein kinase detected in other cell lines (Meth A, SV40 transformed) (Jay et al., 1981).

96

A R N O L D J . LEVINE

VII. Rous Sarcoma Virus

Rous sarcoma virus (RSV) is a nondefective avian retrovirus. The virus genome is composed of three genes-gag, pol, and enu, required for replication and production of virions-and a fourth gene, pp60'-"*', required for transformation but not virus production. The pp6OV-"' gene was likely derived from a normal chicken cell gene (Stehelin et al., 1976), pp6OcpSrc, but is not identical to that cellular gene (Beemon et al., 1979). Pp60'-"*' is a phosphoprotein of 60,000 MW with both phosphoserine (within N-terminal half) and phosphotryosine (within C-terminal half) residues. The majority of this protein appears to be bound to the inside (cytoplasmic) portion of the plasma membrane (Courtneidge et al., 1980; Krueger et al., 1980). Pp60'-"r' has protein kinase activity, as first detected using immunoprecipitates of this protein. In this reaction the rabbit heavy chain of the immune immunoglobulin was phosphorylated at a tyrosine residue after the addition of labeled ATP (Collett and Erikson, 1978; Levinson et al., 1978; Hunter and Sefton, 1980). Purified pp60'-"' retains this protein kinase activity for a number of substrates tested (Erikson et al., 1979; Levinson et al., 1980). When radiolabeled pp6OV-"" is immunoprecipitated from infected and transformed chicken cells by antibodies directed against pp6OvpSrc, two other proteins are coimmunoprecipitated with pp60'-"' by virtue of the fact that they are associated with this transforming protein (Hunter and Sefton, 1980; Brugge et al., 1981; Opperman et al., 1981a). The molecular weights of these cellular proteins are 90,000 (90K) and 50,000 (50K), and both proteins are phosphorylated. The 90K protein contains only phosphoserine, whereas the 50K protein has both phosphoserine and phosphotyrosine residues when isolated from the transformed cell. In the uninfected cell the 50K protein contains only phosphoserine residues, suggesting that this protein is phosphorylated at a tyrosine residue upon transformation with RSV (Brugge and Darrow, 1982). The association of phosphotyrosine residues in the 50K protein with RSV infection is confused by the observation that temperature-sensitive mutants in the src gene result in phosphorylation of 50K at tyrosine residues even at the nonpermissive temperature (Brugge et al., 1981). Though these apparent contradictory results are not fully explained to date, it remains possible that a temperature-sensitive src gene product could be suppressed in its phenotype while complexed with these (50K and 90K) cellular proteins. The pp6OV-"*'-50K-90K complex represents about 5% of the intracellular pp6OV-"' detected in the transformed cell (Brugge et al., 1981). This complex sediments with an apparent molecular weight of about 200,000, while free pp6OV-"' (95%) sediments as a monomer (Brugge et a/., 1981). Pp60'-"' is first synthesized on free ribosomes (not associated with the rough endoplasmic reticulum) and is then phosphorylated

TRANSFORMATION-ASSOCIATED TUMOR ANTIGENS

97

at its serine residue. Shortly after synthesis the pp60'-"' is detected as the pp60'-"*'-50K-90K complex for a brief period. The pp60'-"' monomer then leaves the complex and moves to the membrane (Brugge et a/., 1981). Thus the temperature-sensitive pp60'-"" mutants could still phosphorylate the 50K protein in the pp60'-""-50K-90Kcomplex (suppression of the mutation by complex formation) and then act as a temperature-sensitive mutation in a monomer form at the membrane. The 50K protein is present at low concentrations in both normal and transformed cells, and comparisons of 50K protein from avian and mammalian sources (both bound to pp60'-"') show little or no evolutionary conservation of the peptides derived from this protein. By contrast, the 90K protein is present in high concentrations in normal and transformed cells as expected for a structural protein. In addition, the peptide maps of the 90K protein from avian and mammalian sources are very similar. Curiously, the synthesis of the 90K is greatly stimulated by elevated temperatures (it is a heat shock protein) (Opperman et al., 1981b) or other noxious agents. Like the 50K protein, the 90K protein has a cytoplasmic localization in the cell (Brugge, unpublished results). Based on its electrophoretic mobility in SDS-polyacrylamide gels, partial peptide maps, and localization in the cell, the 50K protein does not appear to be identical with the p53 antigen that is associated with the SV40 T antigen (D. Linzer, J. Brugge, and A. Levine, unpublished results). Although the function of the pp60'-""-50K-90K complex is not yet clear, this complex has also been detected in defective avian retrovirus-transformed cells, such as Fujinami sarcoma virus and Y73 virus (J. S. Brugge, personal communication). The transforming gene products of these viruses are protein kinases, but are chemically unrelated to the RSV pp60'-"'. VIII. Teratocarcinornas

Teratocarcinomas are tumors derived from developmentally immature pluripotent cells (Stevens, 1975; Pierce, 1975). The tumors are composed of malignant stem cells, called embryonal carcinoma cells (EC cells), and a variety of benign cell and tissue types that are differentiated from the stem cell population (Kleinsmith and Pierce, 1964; Kahn and Ephrusii, 1970; Rosenthal et al., 1970; Mintz and Illmensee, 1975). The origin of these tumors has been extensively studied in the mouse, where defined inbred strains develop either testicular (Stevens and Little, 1954) or ovarian (Stevens, 1975) teratomas with high frequency. Thus there is a genetic predisposition for the production of these tumors. Transplantable teratocarcinomas (Stevens, 1958, 1970) and cloned EC cell lines (Bernstine et al., 1973) have been utilized to demonstrate two

98

ARNOLD J. LEVINE

classes of tumor antigens associated with these tumors. Tumor transplantation rejection between allogeneic mouse strains has identified three genetic loci that encode the information for teratocarcinoma transplant rejection (Avner et al., 1978; Siegler et al., 1979; Shedlovsky et al., 1981 ; Levine and Teresky, 1981). At least some of these antigens appear to be limited to teratocarcinomas (Levine and Teresky, 1981). The second antigen detected at high levels in EC cell lines is the p53 antigen (Linzer and Levine, 1979) common to many tumors and transformed cells. Teratocarcinomas or EC cell lines derived from 129/sv mice produce tumors in syngeneic 129/sv mice but are rejected in a variety of allogeneic mice (BALB/c, for example) (Avner et al., 1978; Siegler et al., 1979). The tumor rejection was not due to differences at the H-2 locus and the basis for tumor rejection was lymphocyte mediated (Avner et al., 1978; Siegler et al., 1979). Genetic crosses between tumor-susceptible mice and tumor-rejecting mice permitted the identification of three teratocarcinoma transplantation loci called Gt-1, Gt-2, and Gt-(B6) (Shedlovsky et a/., 1981; Levine and Teresky, 1981). Gt-1 was localized about 4 centimorgans from the H-2 locus between the Tlocus and H-2K region of this chromosome (Siegler et al., 1979; Shedlovsky et al., 1981 ;Levine and Teresky 1981).Gt-2 has also been mapped to chromosome 17, about 17 centimorgans on the H-2D side of this locus (Shedlovsky et al., 1981). Gt-(B6) has not been mapped to a linkage group. It appears likely that these Gt loci encode transplantation rejection antigens expressed on EC cells. These stem cells, like the early mouse embryo, do not contain detectable H-2 antigens (Forman and Vitetta, 1975; Isa, 1976). When EC cells differentiate they acquire the H-2 antigens (Artz and Jacob, 1974; Nicholas et al., 1975; Stein et al., 1975) and are thought to lose the Gt antigens (Levine and Teresky, 1981). These observations, the strength of the Gt transplantation rejection antigens, and the proximity of Gt-1 and 2 to the H-2 locus, has led to the suggestion that the Gt loci are the embryonic analog of the major histocompatibility locus in mice (Shedlovsky et al., 1981). If this explanation is correct, the Gt antigens would be fetal antigens and are recognized as foreign by virtue of their restricted gene expression in early embryos. In addition, these tumor antigens would be restricted to developmentally immature stem cells in a tumor. Embryonal carcinoma cell lines often express very high levels of the same p53 antigen found to be associated with the SV40 large tumor antigen (Linzer and Levine, 1979). In SV40-transformed cells, p53 levels are regulated by the SV40 T antigen (Linzer et al., 1979), which is found in a multimeric protein complex with p53 (Lane and Crawford 1979; Linzer and Levine, 1979). In nontransformed 3T3 cells, p53 is synthesized and degraded rapidly (leading to low levels), whereas p53 in SV40-transformed cells has a much longer half-life and is quite stable (Oren et al., 1981). This has led to the hypothesis

TRANSFORMATION-ASSOCIATED TUMOR ANTIGENS

99

that p53 protein is stabilized, avoiding proteolytic degradation by virtue of its association with SV40 T antigen or a T antigen-induced protein modification of p53 (Oren et al., 1981). If this explanation is correct, then it affords little help in understanding why p53 levels are also high in cells transformed by other agents, with no viral T antigen present. To address this problem, Oren et al. (1982) have examined the regulation of p53 levels in an EC cell line called F9 (Bernstine et al., 1973) and its progeny cells induced to differentiate by treatment of F9 cells with retinoic acid (Strickland and Mahdavi, 1978; Strickland et al., 1980). The F9 EC cells are fully transformed and highly tumorigenic. When treated with retinoic acid and dibutyryl cyclic AMP, these cells differentiate into benign and nontransformed endoderm-like cells. Accompanying this reversion of the malignant or transformed state is a 5- to 10-fold decrease in the steady-state levels of p53 antigen (Oren et al., 1982). If these endoderm cells are infected with SV40, the viral T antigen is synthesized and the p53 levels increase in the endoderm cells (Oren et al., 1982). Pulse-chase experiments, to determine the stability of p53 protein in EC cells and endoderm-like cells, indicated that in both cell types the half-life of p53 was the same, about 3.5 hr. This was in dramatic contrast to nontransfoimed 3T3 cells (20-60 min half-life of p53) and SV-40-transformed cells (a half-life of more than 22 hr) (Oren et al., 1981, 1982). Thus the stability of the p53 protein in F9 and its progeny endoderm cells was similar in spite of the fact that F9 cells contain 5- to 10-fold higher levels of p53. When total cytoplasmic RNA extracted from either F9 and endoderm cells was employed in a reticulocyte in vitro protein translation system, the RNA from endoderm cells synthesized 6- to 10-fold less p53 in vitro than comparable RNA concentrations derived from F9 cells (Oren et al., 1982). Over a wide range of mRNA concentrations (dose-response experiment), comparable levels of RNA from F9 cells always translated 6- to 10-fold more p53 in vitro, than the mRNA from endoderm cells. Thus it appears that the levels of p53-translatable mRNAs decline upon differentiation of F9 cells to endoderm cells, and this may account for the decreased levels of p53 in these nontransformed cells (Oren et al., 1982). These results are in contranst to the nontransformed 3T3-transformed SV3T3 cell system,where both cell types have roughly equal levels of p53-translatable mRNA but protein turnover regulates the intracellular levels of p53 (Oren et al., 1981). Thus, depending on the system under study, p53 levels may be regulated by the amount of p53-translatable mRNA or a posttranslation event. The decline of p53 levels during differentiation of F9 cells in culture may be analogous to the detection of p53 at defined stages of embryogenesis in the mouse (Mora et al., 1980). Mora et al. (1980) have shown that p53 levels decrease as mouse development proceeds. It is difficult, however, based on

100

ARNOLD J . LEVINE

the studies published to date, to quantitate the relative levels of p53 found in stage-specific embryos versus the levels of this antigen in normal and transformed cells. For example, using immunofluorescence and immunoprecipitation of [35S]methionine-labeled proteins, Dippold et al. (1981) failed to detect p53 in mouse embryo cell cultures. Steady-state levels of p53, measured by a radioimmune assay, and utilizing cell lines, embryonic tissue, normal and tumor tissue, will be required before quantitative results permit definitive conclusions as to the relative levels of expression of p53 in embryonic tissue. IX. Conclusions

A. A CLASSIFICATION OF TUMOR ANTIGENS Based on the examples of different tumors examined in this article, a useful classification of tumor antigens can be proposed. Four distinct classes of tumor-specific antigens may be recognized: (1) Antigens that are unique to a particular tumor have been detected in sarcomas and transformed cell lines produced by chemical carcinogens (e.g., Meth A, CMS4) (DeLeo et al., 1977, 1978). (2) Antigens that are unique to, or at least preferentially expressed in, a class or specific type of tumor. Examples of this category of tumor antigen are the Gt-1 and 2 antigens of teratocarcinomas (Shedlovsky et al., 1981 ; Levine and Teresky, 1981), the differentiation-specific antigen of Abelson-transformed lymphocytes (Risser et al., 1978), carcinoembryonic antigen in colon carcinomas, and alpha fetoprotein detected in hepatocellular carcinomas (Abelev et al., 1979). (3) Antigens that are specific for the transforming agent employed to produce the tumor. The DNA and RNA tumor viruses provide the best examples of this category of tumor antigen. (4) Antigens that are common to many different tumors such as p53, which is expressed in high levels in transformed cells of all the examples considered in this article (Table I). 1. Antigens Unique to a Tumor Antigens that are unique to a particular tumor, like Meth A, could arise in a number of ways. If exposure of cells to carcinogens results in genetic rearrangements, via insertion elements or recombination of genomic DNA, that could give rise to a large number of unique antigens much like the immunoglobulin V gene-C gene genetic rearrangements. It is even possible that V gene idiotypes are themselves the unique antigenic moieties of carcinogen-transformed cells. A second source of gene families that encode

TRANSFORMATION-ASSOCIATED TUMOR ANTIGENS

101

membrane antigens would be the glycoprotein structural genes of endogenous RNA retroviruses of the mouse. Recombination between these genes could produce a cell surface antigen with multiple antigenic specificities. Although the nature of the Meth A-like antigens remains to be identified, the proposed mechanisms, like the immunoglobulin system, could generate considerable diversity. Because immunoglobulins are also B-cell surface receptors that trigger cell division upon antigenic stimulation, tumor antigens with similar functions are attractive candidates for Meth A-like antigens. There are, however, other ways to generate antigenic diversity and no evidence that these tumor-specific antigens play a causal or primary role in tumorigenesis. 2. Antigens Unique to a Class o j ’ Tumors

Antigens that are unique to the class or type of tumor under study are usually fetal or developmental antigens. The tissue specificity of alpha fetoprotein, carcinoembryonic antigen, or the Abelson developmental specific antigen reflect this fact. The close genetic linkage between A-MuLV susceptibility (Au-2locus) in mice and the locus responsible for the expression of this pre-B cell antigen (Risser et al., 1978) suggests that this tumor antigen could play a role in the causation or rejection of A-MuLV tumors. This class of tumor antigen is presumably responsible for the rejection of some tumors by animals previously immunized with fetal tissue (Coggin and Ambrose, 1979). At the minimum, however, these antigens have been useful in the diagnosis and classification of tumors. Antigens that are unique to a class or type of tumor may also be a reflection of the immature nature of the tumorigenic target cells for viruses or carcinogens. In this case the “fetal antigen” or “developmental antigen” of the tumor results from the selection and growth of a few target cells in the host. If this hypothesis is correct, then the tumor antigen identifies the rare cell susceptible to carcinogens or viruses, and the normal developmental progression of these cells has been stopped. Pp60-”“ is an example of a viral tumor antigen that can arrest the development of myoblasts (Kaighn et al., 1966), retinal cells (Pessac and Calothy, 1974), epithelial cells (Ephrussi and Temin, 1960), chondroblasts (Pacifici et al., 1977), and melanoblasts (Boettiger et al., 1977). Indeed developmental maturity could preclude the tumorigenic or malignant state, as is the case with teratocarcinomas (Pierce, 1975). 3. Antigens Unique to the Transforming Agent

Viral transformation or tumorigenesis provides the clearest example for both the origin of tumor antigens and their essential role in the transfor-

102

ARNOLD J . LEVINE

mation or tumorigenic process. It is well established that the SV40 large and small tumor antigens, some of the adenovirus tumor antigens, RSV pp6OV-'"', and A-MuLV p120, are required for cellular transformation and probably tumorigenesis (see Table I). The persistence of the viral genome in a tumor and the expression of these viral proteins in the transformed cells reflect the role of these viral proteins in maintaining the transformed phenotype or the tumor. As such, the mechanisms that underlie viral carcinogenesis are among the best understood at this time. In addition, it has become clear that the DNA and RNA tumor viruses differ from each other in both the origin of their " o m genes" and in the mechanism of action of the ''onc gene products" (tumor antigens). RNA tumor viruses acquire their onc genes, like the pp60'-"*' or A-MuLV p120 (Stelelin et al., 1976), from the cellular genome. Pp6OV-"' and pl20 are protein kinases that phosphorylate acceptor proteins at specific tyrosine residues (Collett and Erikson, 1978; Levinson et al., 1978; Witte et a/., 1980). The SV40- and adenovirus-transforming genes do not appear to share nucleotide sequence homology with cellular DNA. When protein kinase activities have been detected in association with SV40 or adenovirus tumor antigens, these kinase activities (Lassam et al., 1979; Griffin et a/., 1979) phosphorylate serine residues on the acceptor protein, and these activities may well result from contamination of the tumor antigen preparation (Tjian and Robbins, 1978; Gracherio and Hager, 1979). Thus the protein kinases of some RNA tumor viruses like RSV pp6OV-"' would be expected to act by modifying proteins in the cell, and that ultimately gives rise to the transformed phenotype. In this case, an inappropriately high concentration of pp6OV-"', protein kinase in the cell, yields an aberrant phenotype because most or all cells already have low levels of pp60"-"" (Collett et al., 1978). One the other hand the SV40 large T antigen is responsible for the initiation of viral DNA replication (Tegtmeyer, 1972) and the modulation down of viral gene expression (Tegtmeyer, 1975). These regulatory events are mediated by the binding of SV40 T antigen to virusspecific nucleotide sequences at or near the origin of viral DNA replication (Shortle et a/., 1979). The SV40 large T antigen is also involved in the stimulation of cellular DNA synthesis (Tjian et al., 1978) and the modulation of cellular gene expression (Postel and Levine, 1975), possibly by acting directly on cellular DNA sequences. If the action of pp6OV-"*'and SV40 T antigen result in similar alterations of cellular gene expression (transformation), the mechanisms by which these two proteins regulate the transformed phenotype appear to be quite different. These differences in the modes of action of SV40 T antigen and p60'-"' are further emphasized by the nuclear (SV40 T antigen) versus plasma membrane (pp60'-"') localization of these two tumor antigens in the transformed cell.

TRANSFORMATION-ASSOCIATED TUMOR ANTIGENS

103

4. Antigens Common to Many Tumors P53, the tumor antigen common to many different transformed cells, is detected in elevated