The term malignancy defines a complex cellular phenotype that is conventionally divided into three stages: (1) progressive multiplication in vivo-, (2) loss of coaptation resulting in the movement of cells away from the growing tumour into the surrounding tissues (invasion) ; and (3) the generation of secondary deposits elsewhere in the body (metastasis). Metastasis presupposes invasion, but progressive (and destructive) growth may occur without either. In a clinical context a tumour would not be classified as malignant unless it showed at least some evidence of invasion; but a malignant tumour may or may not metastasize. Essentially nothing is known about the genetic determinants of invasiveness. In large part this is due to the absence of any satisfactory experimental model for the study of this parameter. Systems involving the penetration of cells into avian and other embryonic membranes (Armstrong, Quigley & Sidebottom, 1982; Basson & Sidebottom, unpublished), or into artificial deposits of collagen, gelatin or other semi-solid matrices in vitro, have not proved reliable predictors of invasiveness (Basson & Sidebottom, 1985).

The analysis of metastasis has moved a little further, but is still at a very early stage. For several years the issue was confused by the use of inappropriate methods, notably the scoring of colonies developing in the lungs or other organs after intravenous injection of cell suspensions. It is now clear, and has indeed been so for many years, that the potential of tumour cells to colonize organs after intravenous injection does not predict metastatic potential under natural conditions (Stackpole, 1981). A beginning has been made in the genetic analysis of metastasis by somatic cell hybridization, and it has been shown that the genetic determinants of progressive growth can be segregated from those that determine metastasis (Sidebottom & Clark, 1983 ; Clark & Sidebottom, 1984, 1985) ; but as yet no cultural characteristic in vitro or no biochemical marker has been found that consistently cosegregates with metastatic potential. A claim has been made that metastatic potential can be conferred on non-metastatic cells by transfection with preparations of DNA from metastatic cells (Bernstein & Weinberg, 1985); and there is one study that suggests that transfection of epithelial tumour cells by a c-H-ras oncogene may enhance their ability to produce metastases (Eccles, Marshall, Vousden & Purvies, 1985). However, neither of these reports contains any information about karyoptic changes induced by the transfection and selection procedures used, so that it is not yet possible to assess how specific or how direct the effect produced by the transfected genes might be. In any case, the data so far available do not permit a review of metastasis in genetic terms. We are therefore reduced to a consideration of progressive cell multiplication in vivo. On this subject there is a rich and interesting genetic literature, and it is this that will form the basis of the present article.

The genetic analysis of progressive cell multiplication in vivo falls into four categories based on different, but overlapping, methodologies: (1) the study of ‘hereditary’ tumours (tumours for which a predisposition is inherited) ; (2) somatic cell hybridization; (3) cytogenetic investigations; and (4) the study of oncogenes (defined in the present context as cellular genes showing substantial homology with the genes of retroviruses). The conclusions drawn from the study of hereditary tumours and those reached by somatic cell hybridization are in substantial agreement. These conclusions are supported by some of the cytogenetic studies, and they are not actually contradicted by any; but they are contradicted by the interpretation given to many of the studies on oncogenes. One aim of the present article is to suggest that these latter interpretations may be inadequate ; another is to present a model in which the results obtained with all four methodologies can be accommodated.

The idea that malignancy might be a consequence of recessive mutations in somatic cells appears first to have been suggested in 1969 on the basis of cell fusion experiments in which the malignant phenotype was found to be suppressed when malignant cells were fused with non-malignant ones (Harris et al. 1969). In 1971, Ohno argued, on theoretical grounds, that the malignant phenotype was much more likely to be generated in somatic cells by recessive mutations than by dominant ones and proposed that these recessive mutations might be unmasked by genetic events that rendered them hemizygous, commonly by a loss, in aneuploid cells, of the homologous chromosome bearing the unaffected allele (Ohno, 1971). In the same year, a specific version of this general model was proposed by Knudson (1971) to account for the incidence pattern of retinoblastoma. This tumour exists in two forms, one sporadic and another in which the predisposition to form the tumour is inherited in an autosomally dominant fashion. Knudson postulated that the genetic locus involved was the same in the two forms of the disease, but that in the heritable form one of the alleles was already mutated in the germ line. A single somatic mutation involving the homologous locus in the unaffected chromosome would then be enough to generate the tumour in the predisposed individual, whereas both alleles would have to be inactivated by two separate genetic events in the sporadic cases. This model, essentially based on homozygosity or hemizygosity of recessive mutations, accommodated the incidence data reasonably well and prompted a search for more direct supporting evidence. This was in due course obtained.

It is now clear from evidence based on both cytogenetic analysis and the analysis of a genetically linked enzyme polymorphism that Knudson’s model is essentially correct (Sparkes et al. 1983; Benedict et al. 1983; Cavenée et al. 1985). The locus involved maps to band ql4 on the long arm of chromosome 13. In the cases where predisposition is heritable, the cells of the affected individual often show a visible deletion in this region, and even when no deletion is visible, evidence based on enzyme polymorphism suggests the presence of a submicroscopic deletion. Tumours arising in such individuals show either homozygosity for the mutated allele, or hemizygosity caused either by a secondary deletion in the previously unaffected allele or by elimination of the whole of the unaffected chromosome 13. An essentially similar situation is found in the sporadic cases of the disease, in which the two genetic events generating homozygosity or hemizygosity of the recessive mutation must both have been somatic. Knudson & Strong (1972) further suggested that the model proposed for retinoblastoma might also be applicable to Wilms’s tumour and, indeed, quite generally, to all cancers for which a heritable predisposition can be demonstrated. In the case of Wilms’s tumour, the prediction has again been fulfilled: a deletion in band p 13 on chromosome 11 is found in the germ line, and this recessive lesion becomes homozygous or hemizygous in the tumours (Koufos et al. 1984; Orkin, Goldman & Sallan, 1984; Fearon, Vogelstein & Feinberg, 1984). Recent work has revealed a similar situation in oat cell (small cell) carcinoma of the lung, where the homozygous deletions are in band pl4 on chromosome 3 (Minna, 1985), and in hepatoblastoma and rhabdomyosarcoma, where the deletions, as in Wilms’s tumour, are in chromsome 11 (Koufos et al. 1985). It seems very likely that similar homozygous genetic defects will be found in other tumours in which there is an overt inherited predisposition, for example, familial renal carcinoma (Pathak, Strong, Ferrell & Trindade, 1982) and several malignancies of the nervous system; but Knudson (1985) argues that there may be a cryptic genetic component in many other tumours for which an inherited predisposition has not yet been recognized. If so, then homozygosity or hemizygosity of recessive mutations might be a much more general mechanism for generating the malignant phenotype than originally envisaged. Recent work on the lethal(2)giant larvae mutation in Drosophila melanogaster has revealed that the malignant tumours of presumptive adult brain cell centres and imaginai discs that arise in this condition are determined by homozygosity or hemizygosity of recessive mutations (deletions or mutational insertions) at the l(2)gl locus (Mechler, McGinnis & Gehring, 1985). It is clear that retinoblastoma and Wilms’s tumour are not to be regarded as special cases. Three important general conclusions can be drawn from this work: (1) malignancy can be determined by mutational events at a single locus ; (2) the number of genetic events involved need not exceed two, the minimum number required to generate homozygosity or hemizygosity at that locus; and (3) the mutations are recessive and engender loss of function, not gain in function. What this function might be will be discussed at a later stage.

The overall results obtained by somatic cell hybridization are in complete agreement with the conclusion that progressive multiplication of cells in vivo involves a loss of some cell function, not a gain of function. It is now well established that when malignant cells, defined by their ability to generate progressive tumours in genetically compatible hosts, are fused with diploid fibroblasts of the same species, the resulting hybrid cells, so long as they retain certain specific chromosomes donated by the diploid parent cell, are unable to generate such tumours. When these particular chromosomes are eliminated, however, the ability to multiply progressively in vivo reappears in the hybrid cell, and it is once more able to generate a progressive tumour. This phenomenon, commonly described as the suppression of malignancy or tumorigenicity, has been extensively studied in several laboratories and has been shown to apply to crosses between malignant and non-malignant cells in the mouse, hamster and man. (For reviews, see Miller & Miller, 1983; Sager, 1985.) Apparent exceptions to this rule arise from two sources: (1) the relevant chromosomes derived from the diploid parent cell are eliminated early so that the clones are already malignant segregants of the original hybrid cells when they are first examined (in some cases there is strong selection pressure against certain chromosomes derived from the diploid cell); and (2), less frequently, the malignant parent cells contain more than two gene sets or multiple copies of some of the autosomes. (The significance of this will become apparent when the role of gene dosage in the suppression of malignancy is discussed below). While the terms ‘dominance’ and ‘recessiveness’ are not easy to apply in a genetically rigorous fashion to experiments with hybrid cells, an acceptable operational description of the overall findings would be that, in this kind of test, the genetic determinants of the malignant phenotype are recessive to those determining the normal phenotype.

Cytogenetic analysis of this phenomenon has revealed that in hybrids between malignant and diploid mouse cells, the chromosomes 4 derived from the diploid parent cell have a decisive role in suppressing the malignant phenotype (Jonasson, Povey & Harris, 1977). This is true for a wide range of different kinds of malignant cells, including tumour cells bearing and expressing retroviral oncogenes. The locus involved appears to map to the lower part of the upper half of chromosome 4, but this assignment remains to be confirmed and further refined. The critical role played by a single diploid chromosome in suppressing the malignant phenotype and, in particular, the assignation of the operative locus to chromosome 4 in mouse hybrids is strongly reinforced by a similar analysis in crosses between normal diploid human fibroblasts and anchorage-independent hamster cells transformed by exposure to a chemical carcinogen. In these crosses, a locus that maps to human chromosome 1 is apparently responsible for the suppression of the transformed phenotype (Stoler & Bouck, 1985). Since human chromosome 1 shares a large area of homology with mouse chromosome 4 (Lalley, Francke & Minna, 1978), it seems probable that the same genetic mechanism is operative in the two species. Further evidence for the role of human chromosome 1 in the suppression of tumorigenicity comes from crosses between a human fibrosarcoma (HT1080) and normal human diploid fibroblasts (Benedict, Weissman, Mark & Stanbridge, 1984). In this case, suppression occurs despite the fact that the fibrosarcoma carries a transforming N-ras oncogene. There is also evidence that, in crosses between a human cervical carcinoma cell line and normal diploid human fibroblasts, suppression of tumorigenicity is again determined by a single chromosome (in this case chromosome 11), but the cytogenetic analysis remains uncertain since no markers were available to permit discrimination between the chromosomes derived from the tumour cell and those derived from the diploid fibroblast (Stanbridge, Flandermeyer, Daniels & Nelson-Rees, 1981).

A detailed cytogenetic study of the suppression of tumorigenicity in a particularly informative set of intraspecific mouse hybrids, in which the parental origin of all the chromosomes 4 in the cells could be assigned with certainty, has provided strong evidence for the conclusion that this suppression is subject to gene dosage effects: it is reinforced by an increase in the number of diploid chromosomes 4 in the hybrids and it may be overcome by an increase in the number of chromosomes 4 derived from the malignant parent cell (Evans et al. 1982). It is for this reason that suppression of tumorigenicity may not be observed in some crosses between diploid cells and aneuploid or subtetraploid malignant cells in which multiple copies of many autosomes may be found. Similar gene dosage effects for human chromosome 1 have been described in hybrids between diploid human fibroblasts and human fibrosarcoma cells, where, again, tumorigenicity was found to be suppressed when the fibroblasts were crossed with near-diploid sarcoma cells but not when they were crossed with near-tetraploid cells (Benedict et al. 1984).

The observations made on hybrid cells are thus in accord with those made on retinoblastoma and other hereditary tumours: the malignant phenotype appears to be generated by recessive mutations at a particular locus and appears to be suppressed by the product of the homologous normal locus. Knudson (1985) calls such homologous loci ‘anti-oncogenes’, but the function of these genes is not, of course, to suppress cancer any more than the function of oncogenes is to induce it. ‘Anti-oncogenes’ are simply normal cellular genes carrying out normal cellular functions. But what might these functions be? Recent histological studies on hybrids in which malignancy has been suppressed have been especially informative in this regard. It has been shown that when a normal diploid cell is fused with a malignant one, it suppresses the malignant phenotype by imposing on the hybrid cell its own pattern of terminal differentiation. When tumour cells are fused with diploid fibroblasts, the hybrid cells, when injected into the animal, assume an increasingly elongated fibrocytic morphology, synthesize a collagenous extracellular matrix, as normal fibroblasts do in forming scar tissue, and stop multiplying (Stanbridge & Ceredig, 1981; Harris, 1985). But segregants from which the specific suppressive chromosome derived from the diploid fibroblast has been eliminated, do not assume an elongated fibrocytic morphology, do not synthesize any stainable collagenous extracellular matrix and continue to multiply. Similarly, hybrids in which malignancy is suppressed by fusion of the tumour cells with normal diploid kératinocytes undergo squamous differentiation and keratinization in vivo, whereas malignant segregants derived from these hybrids do not (Peehl & Stanbridge, 1982). In the suppression of malignancy by cell fusion we are thus dealing with an imposed pattern of terminal differentiation bringing cell multiplication to a stop. But this encourages the conclusion that the tumour cells themselves continue to multiply progressively in vivo because they cannot execute the pattern of terminal differentiation that is proper to them. In other words, the recessive defect that generates progressive cell multiplication in vivo is a defect of differentiation. In cell fusion experiments, this defect may be overcome either by the imposition of a foreign pattern of terminal differentiation on the malignant cell, as occurs when non-fibroblastic tumour cells are fused with normal fibroblasts, or by the restoration of the malignant cell’s own pattern of terminal differentiation. The latter effect has been observed with anaplastic fibrosarcoma cells, which do not synthesize an extracellular matrix in vivo, but which can be induced to do so and to undergo terminal fibrocytic differentiation when fused with normal diploid cells of a non-fibroblastic type (Harris, 1985). The relationship between terminal differentiation and progressive cell multiplication in vivo is discussed further below.

The role of homozygous or hemizygous deletions in the genesis of hereditary tumours has already been considered; but cytogenetic studies have revealed, in certain other malignant tumours, two other types of chromosome abnormality that are thought to involve quite different genetic mechanisms : translocations and gene amplifications. While a wide variety of translocations may be found sporadically in many malignant tumours, some malignancies of the haemopoietic system are characterized by the consistent presence of specific translocations involving constant chromosome regions. These translocations have been intensively studied in both mouse and man and have been reviewed recently by Rowley (1984) and Klein & Klein (1985). The biological interest of these translocations has been immensely enhanced by the realization that the break-points involve bands in which known oncogenes have been located and in three cases (two human, one murine) it has been demonstrated that the translocation moves an oncogene to a new site. The human cases are Burkitt’s lymphoma, in which the myc gene on chromosome 8 is translocated to the vicinity of one of the three immunoglobulin loci on chromosomes 14, 2 or 22, and chronic myeloid leukaemia, in which the abl gene on chromosome 9 is translocated to chromosome 22. In the mouse, plasmacytomas show essentially homologous translocations to those seen in Burkitt’s lymphoma: the myc gene on chromosome 15 is moved to the vicinity of immunoglobulin gene loci on chromosomes 12 or 6. Since nothing is known of the function of the genetic area to which the abl gene is translocated in chronic myeloid leukaemia, most investigations have so far centred on the myc translocations in Burkitt’s lymphoma and mouse plasmacytoma.

An attractive idea proposed by Klein was that the translocation of the myc gene to an immunoglobulin gene region, where a high level of transcription was operative, would impose a higher than normal level of transcription on the translocated gene, and that this would ensure continued stimulation of cell multiplication. In its simplest form, however, this idea has become untenable. Although a higher than normal level of transcription was claimed for some Burkitt’s lymphoma cell lines (Erickson et al. 1983), it was not clear that the appropriate normal controls had been chosen or, indeed, what they should be ; and other reports on different cell lines soon showed that an enhanced level of transcription of the translocated myc gene was by no means the rule, either in Burkitt’s lymphoma or in mouse plasmacytoma (Klein & Klein, 1985). Nucleotide sequence determination then revealed that, in some cases, the translocated myc gene had undergone mutation, rearrangement or partial deletion; but usually no structural change could be detected in the translocated gene. These findings reduced the original model to a suggestion that perhaps the translocated gene, although specifying a normal product, was not regulated properly. If so, then it would have to be assumed that the transcription of the translocated gene was regulated by cts-acting mechanisms, for example, changes in chromosome structure. Evidence in support of this view was provided by the claim that only the translocated gene was transcribed in Burkitt’s lymphoma, while the untranslocated homologue was silent (Nishikuraet al. 1983). However, almost at once it was shown, in a different Burkitt’s lymphoma cell line, that the untranslocated myc gene was transcribed (Rabbitts, Forster, Hamlyn & Baer, 1984), although apparently at a lower level than the translocated gene (Feo et al. 1985). On the other hand, it has been shown, in experiments in which Burkitt’s lymphoma cells were fused with fibroblasts, that, like several other oncogenes that have been tested in this way, transcription of the translocated myc gene can be regulated by trans-acting elements (Nishikura et al. 1984).

Despite the hazards inherent in generating models from such variable data, it is worth considering what the genetic consequences of these different claims might be. If the translocated myc gene is mutated or structurally altered in such a way as to make it unlikely that a normal functional gene product is made, and the untranslocated gene is silent (because it, too, carries a cryptic mutation, or for any other reason), then the cell is without a functional myc gene and is formally equivalent to a cell that has undergone a homozygous or hemizygous recessive mutation. If the product of the translocated gene is functionally normal, then the variability found in the rate of transcription of this gene makes it unlikely that an elevated level of the myc gene product can be the essential determinant of the malignant phenotype. If the product of the translocated gene is abnormal, but is in some way functional, while that of the untranslocated gene remains normal, then the consequences would depend on whether the abnormal gene product interfered with the action of the normal product. If it did, then one would expect to observe dosage effects of various kinds. But all these models are called into question by recent data on the behaviour of the myc protein itself. It has been shown that the level, rate of synthesis and rate of turnover of the myc protein is no different in Burkitt’s lymphoma cells carrying a translocated myc gene than in chronic leukaemia cells carrying no translocated or amplified myc gene (Hann, Thompson & Eisenman, 1985). Moreover, the levels of myc protein do not vary significantly throughout the cell cycle, as might be expected if the myc gene were involved in stimulating the entry of the cells into the phase of DNA synthesis. This study on the myc protein is supported by an independent study showing a similar lack of variation of myc messenger RNA throughout the cell cycle (Thompson, Chaloner, Neiman & Groudine, 1985). Nonetheless, it is difficult to avoid the conclusion that the myc translocation in Burkitt’s lymphomas and in mouse plasmacytomas must confer some selective advantage on these particular cells, as it would otherwise be difficult to explain its consistent presence. What this selective advantage might be and how it operates are at the moment unclear. One idea that appears not to have been considered is that the importance of the myc translocations might lie not in their effects on the myc gene, but in their effects on the further differentiation of the antibody-forming cell. The possibility seems worth exploring that the myc translocations act by impeding the progress of the antibody-forming cell into its normal fully differentiated state, the non-dividing memory cell. It remains an open question whether the translocation is involved in initiating the process that leads to malignancy or whether it takes place at a later stage in the growth of the tumours conferring such a strong selective advantage that cells bearing it become the dominant, or even exclusive, cell type. There is, in any case, nothing in the studies on Burkitt’s lymphomas or mouse plasmacytomas that provides convincing evidence of a genetically dominant mode of action for the cytogenetic abnormalities found in this form of malignancy.

Many examples have been described of gene amplification in malignant tumours, both in the form of supernumerary chromosomes and as tandem expansions of particular genes commonly identified in cytological preparations as homogeneously staining regions. Among the genes that may be amplified in this way are a number of oncogenes. For example, amplification of c-myc has been described in some breast carcinomas (Capon et al. 1983; Kozbor & Croce, 1984), colon carcinoma (Alitalo et al. 1983), neuroblastoma (Schwab et al. 1983), retinoblastoma (Lee, Murphree & Benedict, 1984) and leukaemia (Nowell et al. 1983); and amplification of either c-myc or rv-myc has been found in some 20 % of human lung cancer cell lines (Little et al. 1983). However, in no case is amplification of a particular oncogene invariably or consistently associated with any particular kind of malignant tumour. Nor is gene amplification limited to malignant cells: amplification of an unmodified c-H-ras oncogene has recently been demonstrated in normal diploid human fibroblasts during their limited replicative life-span in vitro (Srivastava, Norris, Schmooker-Reis & Goldstein, 1985). It is therefore very unlikely that amplification of oncogenes is the initiating event in the generation of the malignant tumour. Most of the evidence bearing on this point suggests that this amplification is a later effect of selection pressure applied to the growing cell population. How the amplified oncogene confers a selective advantage is not at all clear, but it seems that the selection pressure is not limited to malignant cells multiplying in vivo but can operate on normal diploid cells growing in vitro.

There is now a massive literature, punctuated by excited editorials from Nature, on the occurrence and behaviour of cellular genes that show homology with retroviral genes, and the subject is constantly being reviewed. Only one aspect of this literature will be considered in any depth here : the evidence that some of these oncogenes act to produce malignancy in a genetically dominant fashion. The question is of great theoretical importance, for the experiments in which morphological transformation in vitro or tumorigenicity in vivo are conferred on non-transformed and non-tumorigenic cells by transfection of oncogenes constitute the only substantial body of work that appears to contradict the conclusions reached from the study of hereditary tumours and from somatic cell hybridization. Because clones of transformed and sometimes tumorigenic cells arise in populations of untransformed cells after transfection with identifiable genes taken from malignant cells, the assumption is commonly made that these genes must be operating in a dominant fashion. This assumption is precarious. To begin with, it overlooks the complexity of the changes that may be produced in the recipient cells by the transfection and subsequent selection procedures. If it were certain that the only change produced in the genotype of the recipient cell by the transfection procedure was the addition of a single copy of the transfected oncogene, then the conclusion that this gene acted in a dominant fashion would, at least provisionally, be plausible. But it is, in fact, certain that this is not the case. Karyological investigation of clones transformed by transfected oncogenes invariably shows that the karyotype of the transformed cells differs substantially from that of the untransformed cells (Gilbert, Evans & Harris, unpublished). The recipient cells have not only received the oncogene, they have in the process also undergone other, often multiple, genetic changes. Against this background of complex modifications in genotype, conclusions about the dominance of the mode of action of the transfected gene are meaningless, unless it can be shown that these complex modifications are irrelevant. To do this, it would be necessary to back-select against the transfected oncogene and demonstrate that all clones that have eliminated the transfected oncogene have reverted to a non-transformed and non-tumorigenic phenotype. A systematic investigation of this question is at present being undertaken (Gilbert & Harris, unpublished); but, in the interim, it is of interest to consider the available evidence.

It has been claimed that if NIH3T3 cells are transformed by transfection with an N-ras oncogene, the continued presence of the oncogene is required to maintain the transformed phenotype (Murray et al. 1983); but this assumption is based on the analysis of a single flat revertant in which the transfected oncogene was not found. On the other hand, in a more extensive set of experiments, it has been shown that flat revertants of NIH3T3 cells transformed by the Si-ras oncogene may or may not retain the transfected gene (Noda, Selinger, Scolnick & Bassin, 1983); and there is evidence that such revertants may retain and express the K-ras oncogene but still be resistant to its transforming action (Norton et al. 1984). Moreover, it has been shown that the lymphomas produced by the Abelson virus eliminate the Abelson virus genes on continued growth in vivo, but remain tumorigenic (Grunwald et al. 1982). This is also true for tumours induced by the avian leukosis virus (Payne et al. 1981), and for Chinese hamster fibroblasts transformed and rendered tumorigenic by transfection with the c-H-ras oncogene (Lau et al. 1985). In all these cases the results support the view that the transfected oncogene acts in an indirect way by inducing stable secondary changes in the genotype of the recipient cell, a mechanism that has been dubbed ‘hit and run’. But even if this were not the case, the evidence for a dominant mode of action of mutated or transfected oncogenes would still be weak. It has been shown in three different malignant tumours carrying mutated K-ras oncogenes (from colon (Capon et al. 1983), bladder and lung (Santos et al. 1984)) that the cells in-each case have eliminated the normal unmutated K-ras allele, a hemizygous state obviously reminiscent of the findings in hereditary tumours. And even in the prototype case, transformation of NIH3T3 cells by transfection with the H-ras oncogene, it appears that the normal mouse allele is only transcribed at a very low level in the transfected cells (Capon et al. 1983), a situation that may not be very different from that previously discussed in the section dealing with myc translocations in Burkitt’s lymphoma. It is, in any case, clear that the expression of at least some oncogenes can be suppressed when malignant cells bearing them are fused with non-malignant cells (Marshall, 1980; Dyson, Quade & Wyke, 1982; Benedict et al. 1983; Dyson, Cook, Searle & Wyke, 1985; Craig & Sager, 1985).

Perhaps enough has been said to convince the reader that the case for a genetically dominant mode of action for oncogenes has yet to be established; but whatever the formal genetic situation may turn out to be, the mode of action of these genes in generating tumours remains a fascinating subject. In a recent paper, the suggestion has been made that mutated, or otherwise functionally abnormal oncogenes, may act by impeding the process of terminal differentiation in the cell (Harris, 1985). A good deal of evidence already exists that at least some of the normal cellular homologues of oncogenes are involved in a critical way in various forms of differentiation (Jacob, 1983). In a recent study with the c-fos oncogene, it has been shown that transfection of the normal mouse or human gene into an embryonic carcinoma cell line that does not normally differentiate in vitro can induce various elements of the differentiated state (Müller & Wagner, 1984; Rüther, Wagner & Müller, 1985). If one supposes that the normal cellular homologue of an oncogene acts to initiate a particular programme of differentiation, or is involved in that programme in some decisive way, then it is not difficult to envisage that a mutated or otherwise functionally abnormal form of that gene might impede that programme of differentiation, especially if it is present in a hemizygous or homozygous condition, or if, for some reason, the altered gene product were made in much greater amounts than the normal product. In the case of transfection experiments, an aberrant effect might in some cases be achieved even with a structurally normal gene, for any one such gene might be involved in only one specific programme of differentiation (for example, the erb gene in erythropoiesis), but it might be transfected to a cell of a quite different lineage (for example, the NIH3T3 fibroblast). This could have the consequence that the recipient cell’s own programme of differentiation might be impaired by the operation of quite inappropriate signals. This idea finds experimental support in several experimental systems in which it has been shown that a variety of different oncogenes introduced into the cell do indeed impede the process of differentiation. For example, transformation of fibroblastic cells by Rous sarcoma virus (Arbogast et al. 1977; Vaheri et al. 1978), by simian virus 40 (Krieg et al. 1980; Triieb, Lewis & Carter, 1985) or by transfection with the H-ras or v-mos oncogene (Liau, Yamada & de Crombrugghe, 1985; Schmidt, Setoyama & de Crombrugghe, 1985) produces in all cases a severe impairment of the production of one or more components of the extracellular matrix. Transfection of bronchial epithelial cells by the H-ras oncogene renders them incapable of undergoing squamous differentiation (Yoakum et al. 1985) ; and several different oncogenes similarly impair the differentiation of myogenic cells (Falcone, Tato & Alemà, 1985). The great attraction of the idea that the genesis of malignancy is causally linked to a defect in the process of normal differentiation is that it can accommodate without conflict all the evidence provided by the four different categories of investigation that have been discussed: the data from hereditary tumours, somatic cell hybridization, cytogenetics and oncogenes. But one is still entitled to ask how the defect in the process of differentiation induces progressive cell multiplication in vivo.

The least taxing approach to this question is to assume that cells, both prokaryotic and eukaryotic, are so constituted that they will continue to multiply, given an adequate supply of nutrients and co-factors, until the process of differentiation induces them to stop. Continuous multiplication is, in this model, envisaged as the natural steady state, and cessation of multiplication a restriction imposed on the system. Any event that produces a stable heritable block to the process of differentiation would then ipso facto result in continuous cell multiplication. This model is not contradicted by the observed effects of growth factors, for it is perfectly plausible to regard growth factors as agents that do impede the process of differentiation, perhaps, in some cases, by preventing the exit of cells from the cycle of DNA synthesis. These ideas have been discussed in greater detail elsewhere (Harris, 1985). If they are true, they make the prospect of finding agents that might discriminate between the multiplication of malignant cells and that of their normal homologues rather remote, but they should encourage the search for physiologically tolerable agents that can induce the terminal differentiation of malignant cells.

Alitalo
,
K.
,
Schwab
,
M.
,
Lin
,
C. C.
,
Varmus
,
H. E.
&
Bishop
,
J. M.
(
1983
).
Homogeneously staining chromosomal regions contain amplified copies of an abundantly expressed cellular oncogene (c-tnyc) in malignant neuroendocrine cells from a human colon carcinoma
.
Proc. natn. Acad. Sci. U.SA
.
80
,
1707
1711
.
Arbogast
,
B. W.
,
Yoshimura
,
M.
,
Kefalides
,
N. A.
,
Holtzer
,
H.
&
Kaji
,
A.
(
1977
).
Failure of cultured chick embryo fibroblasts to incorporate collagen in to their extracellular matrix when transformed by Rous sarcoma virus
.
J. biol. Chem
.
252
,
8863
8868
.
Armstrong
,
P. B.
,
Quigley
,
J. P.
&
Sidebottom
,
E.
(
1982
).
Transepithelial invasion and intramesenchymal infiltration of the chick embryo chorioallantois by tumor cell lines
.
Cancer Res
.
42
,
1826
1837
.
Basson
,
C. T.
&
Sidebottom
,
E.
(
1985
).
An in vitro study of tumour invasion to compare metastatic and non-metastatic cells
.
In Treatment of Metastasis: Problems and Prospects
(ed.
K.
Hallam
&
S. A.
Eccles
), pp.
243
246
.
London
:
Taylor and Francis
.
Benedict
,
W. F.
,
Murphree
,
A. L.
,
Banerjee
,
A.
,
Spina
,
C. A.
,
Sparkes
,
M. C.
&
Sparkes
,
R.
(
1983
).
Patient with 13 chromosome deletion: evidence that the retinoblastoma gene is a recessive cancer gene
.
Science
219
,
973
975
.
Benedict
,
W. F.
,
Weissman
,
B. E.
,
Mark
,
C.
&
Stanbridge
,
E. J.
(
1984
).
Tumorigenicity of human HT1080 fibrosarcoma X normal fibroblast hybrids: chromosome dosage dependency
.
Cancer Res
.
44
,
3471
3479
.
Bernstein
,
S. C.
&
Weinberg
,
R. A.
(
1985
).
Expression of the metastatic phenotype in cells transfected with human metastatic tumor DNA
.
Proc. natn. Acad. Sci. U.SA
.
82
,
1726
1730
.
Capon
,
D. J.
,
Seeburg
,
P. H.
,
Mcgrath
,
J. P.
,
Hayflick
,
J. S.
,
Edman
,
U.
,
Levinson
,
A. D.
&
Goeddel
,
D. V.
(
1983
).
Activation of Ki-nzs 2 gene in human colon and lung carcinomas by two different point mutations
.
Nature, Land
.
304
,
507
513
.
Cavenée
,
W. K.
,
Hansen
,
M. F.
,
Nordenskjold
,
M.
,
Kock
,
E.
,
Maumenee
,
I.
,
Squire
,
J. A.
,
Phillips
,
R. A.
&
Gallie
,
B. L.
(
1985
).
Genetic origin of mutations predisposing to retinoblastoma
.
Science
228
,
501
503
.
Clark
,
S. R.
&
Sidebottom
,
E.
(
1984
).
Selection of metastatic variants on the basis of clonal morphology in vitro
.
Invasion, Metastasis
4
(
Suppl. 1
),
1
11
.
Clark
,
S. R.
&
Sidebottom
,
E.
(
1985
).
Heterogeneity, stability and selection in metastatic cells
.
In Treatment and Metastasis: Problems and Prospects
(ed.
K.
Hallam
&
S. A.
Eccles
), pp.
393
396
.
London
:
Taylor and Francis
.
Craig
,
R. W.
&
Sager
,
R.
(
1985
).
Suppression of tumorigenicity in hybrids of normal and oncogene-transformed CHEF cells
.
Proc. natn. Acad. Sci. U.S A
.
82
,
2062
2066
.
Dyson
,
P. J.
,
Cook
,
P. R.
,
Searle
,
S.
&
Wyke
,
J. A.
(
1985
).
The chromatin structure of Rous sarcoma proviruses is changed by factors that act in trans in cell hybrids
.
EMBOJ
.
4
,
413
420
.
Dyson
,
P. J.
,
Quade
,
K.
&
Wyke
,
J. A.
(
1982
).
Expression of the ASV src gene in hybrids between normal and virally transformed cells : specific suppression occurs in some hybrids but not others
.
Cell
30
,
491
498
.
Eccles
,
S. A.
,
Marshall
,
C. J.
,
Vousden
,
K.
&
Purvies
,
H. P.
(
1985
).
Enhanced spontaneous metastatic capacity of mouse mammary carcinoma cells transfected with c-H-ras
.
In Treatment of Metastasis: Problems and Prospects
(ed.
K.
Hallam
&
S. A.
Eccles
), pp.
385
389
.
London
:
Taylor and Francis
.
Erikson
,
J.
,
Ar-Rushdi
,
A.
,
Drwinga
,
H. L.
,
Nowell
,
P. C.
&
Croce
,
C. M.
(
1983
).
Transcriptional activation of the translocated c-myc oncogene in Burkitt lymphoma
.
Proc. natn. Acad. Sci. U.SA
.
80
,
820
824
.
Evans
,
E. P.
,
Burtenshaw
,
M. D.
,
Brown
,
B. B.
,
Hennion
,
R.
&
Harris
,
H.
(
1982
).
The analysis of malignancy by cell fusion. IX. Re-examination and clarification of the cytogenetic problem
.
JC Cell Sci
.
56
,
113
130
.
Falcone
,
G.
,
Tato
,
F.
&
Alemà
,
S.
(
1985
).
Distinctive effects of the viral oncogenesmyc, erb,fps and src on the differentiation program of quail myogenic cells
.
Proc. natn. Acad. Sci. U.SA
.
82
,
426
430
.
Fearon
,
E. R.
,
Vogelstein
,
B.
&
Feinberg
,
A. P.
(
1984
).
Somatic deletion and duplication of genes on chromosome 11 in Wilms’ tumours
.
Nature, Lond
.
309
,
176
178
.
Feo
,
S.
,
Ar-Rushdi
,
A.
,
Huebner
,
K.
,
Finan
,
J.
,
Nowell
,
P. C.
,
Clarkson
,
B.
&
Croce
,
C. M.
(
1985
).
Suppression of the normal mouse c-myc oncogene in human lymphoma cells
.
Nature, Lond
.
313
,
493
495
.
Grunwald
,
D. J.
,
Dale
,
B.
,
Dudley
,
J.
,
Lamph
,
W.
,
Sugden
,
B.
,
Ozanne
,
B.
&
Risser
,
R.
(
1982
).
Loss of viral gene expression and retention of tumorigenicity by Abelson lymphoma cells
.
J. Virol
.
43
,
92
103
.
Hann
,
S. R.
,
Thompson
,
C. B.
&
Eisenman
,
R. N.
(
1985
).
c-myc oncogene protein synthesis is independent of the cell cycle in human and avian cells
.
Nature, Lond
.
314
,
366
369
.
Harris
,
H.
(
1985
).
Suppression of malignancy in hybrid cells: the mechanism
.
J. Cell Sci
.
79
,
105
117
.
Harris
,
H.
,
Miller
,
O. J.
,
Klein
,
G.
,
Worst
,
P.
&
Tachibana
,
T.
(
1969
).
Suppression of malignancy by cell fusion
.
Nature, Lond
.
223
,
363
368
.
Jacob
,
F.
(
1983
).
Expression of embryonic characters by malignant cells
.
Ciba Fdn Symp
. vol.
96
, pp.
4
27
.
London
:
Pitman
.
Jonasson
,
J.
,
Povey
,
S.
&
Harris
,
H.
(
1977
).
The analysis of malignancy by cell fusion. VII. Cytogenetic analysis of hybrids between malignant and diploid cells and of tumours derived from them
.
J. Cell Sci
.
24
,
217
254
.
Klein
,
G.
&
Klein
,
E.
(
1985
).
Evolution of tumours and the impact of molecular oncology
.
Nature, Lond
.
315
,
190
195
.
Knudson
,
A. G.
(
1971
).
Mutation and cancer: statistical study of retinoblastoma
.
Proc. natn. Acad. Sci. U.SA
.
68
,
820
823
.
Knudson
,
A. G.
(
1985
).
Hereditary cancer, oncogenes and antioncogenes
.
Cancer Bes
.
45
,
1437
1443
.
Knudson
,
A. G.
&
Strong
,
L. C.
(
1972
).
Mutation and cancer: a model for Wilms’ tumour of the kidney
.
J. natn. Cancer Inst
.
48
,
313
324
.
Koufos
,
A.
,
Hansen
,
M. F.
,
Copeland
,
N. G.
,
Jenkins
,
N. A.
,
Lampkin
,
B. C.
&
Cavenée
,
W. K.
(
1985
).
Loss of heterozygosity in three embryonal tumours suggests a common pathogenic mechanism
.
Nature, Lond
.
316
,
330
334
.
Koufos
,
A.
,
Hansen
,
M. F.
,
Lampkin
,
B. C.
,
Workman
,
M. L.
,
Copeland
,
N. G.
,
Jenkins
,
N. A.
&
Cavenée
,
W. K.
(
1984
).
Loss of alleles at loci on human chromosome 11 during genesis of Wilms’ tumour
.
Nature, Lond
.
309
,
170
172
.
Kozbor
,
D.
&
Croce
,
C. M.
(
1984
).
Amplification of the c-myc oncogene in one of five human breast carcinoma cell lines
.
Cancer Res
.
44
,
438
441
.
Krieg
,
T.
,
Aumailley
,
M.
,
Dessau
,
W.
,
Wiestner
,
M.
&
Müller
,
P.
(
1980
).
Synthesis of collagen by human fibroblasts and their SV40 transformants
.
Expl Cell Res
.
125
,
23
30
.
Lalley
,
P. A.
,
Francke
,
U.
&
Minna
,
J. D.
(
1978
).
Homologous genes for enolase, phosphogluconate dehydrogenase, phosphoglucomutase and adenylate kinase are syntenic on mouse chromosome 4 and human chromosome Ip
.
Proc. natn. Acad. Sci. U.S A
.
75
,
2382
2386
.
Lau
,
C. C.
,
Gadi
,
I. M.
,
Kalvonjian
,
S.
,
Anisowicz
,
A.
&
Sager
,
R.
(
1985
).
Plasmid-induced ‘hit-and-run’ tumorigenesis in Chinese hamster embryo fibroblast (CHEF) cells
.
Proc. natn. Acad. Sci. USA
.
82
,
2839
2843
.
Lee
,
W.-H.
,
Murphree
,
A. L.
&
Benedict
,
W. F.
(
1984
).
Expression and amplification of the N-myc gene in primary retinoblastoma
.
Nature, Lond
.
309
,
458
460
.
Liau
,
G.
,
Yamada
,
Y.
&
De Crombrugghe
,
B.
(
1985
).
Coordinate regulation of the levels of Type 111 and Type 1 collagen mRNA in most but not all mouse fibroblasts
.
J. biol. Chem
.
260
,
531
536
.
Little
,
C. D.
,
Nau
,
M. M.
,
Carney
,
D. N.
,
Gazdar
,
A. F.
&
Minna
,
J. D.
(
1983
).
Amplification and expression of the c-myc oncogene in human lung cancer cell lines
.
Nature, Lond
.
306
,
194
195
.
Marshall
,
C. J.
(
1980
).
Suppression of the transformed phenotype with retention of the viral ‘src’ gene in cell hybrids between Rous sarcoma virus-transformed rat cells and untransformed mouse cells
.
Expl Cell Res
.
127
,
373
384
.
Mechler
,
B. M.
,
Mcginnis
,
W.
&
Gehring
,
W. J.
(
1985
).
Molecular cloning of lethal(2) giant larvae, a recessive oncogene oí Drosophila melanogaster
.
EMBOJ
.
4
,
1551
1557
.
Miller
,
D. A.
&
Miller
,
O.
(
1983
).
Chromosomes and cancer in the mouse: studies in tumors, established cell lines, and cell hybrids
.
Adv. Cancer Res
.
39
,
153
182
.
Müller
,
R.
&
Wagner
,
E. F.
(
1984
).
Differentiation of F9 teratocarcinoma stem cells after transfer of c-fos proto-oncogenes
.
Nature, Lond
.
311
,
438
442
.
Murray
,
M. J.
,
Cunningham
,
J. M.
,
Parada
,
L. F.
,
Dautry
,
F.
,
Lebowitz
,
P.
&
Weinberg
,
R. A.
(
1983
).
The HL-60 transforming sequence: a ras oncogene coexisting with altered myc genes in hematopoietic tumors
.
Cell
33
,
749
757
.
Nishikura
,
K.
,
Ar-Rushdi
,
A.
,
Erikson
,
J.
,
Watt
,
R.
,
Rovera
,
G.
&
Croce
,
C. M.
(
1983
).
Differential expression of the normal and of the translocated human c-myc oncogenes in B cells
.
Proc. natn. Acad. Sci. USA
.
80
,
4822
4826
.
Nishikura
,
K.
,
Ar-Rushdi
,
A.
,
Erikson
,
J.
,
Dejesus
,
E.
,
Dugan
,
D.
&
Croce
,
C. M.
(
1984
).
Repression of rearranged p gene and translocated c-myc in mouse 3T3 cells X Burkitt lymphoma cell hybrids
.
Science
224
,
399
402
.
Noda
,
M.
,
Selinger
,
Z.
,
Scolnick
,
E. M.
&
Bassin
,
R. H.
(
1983
).
Flat revenants isolated from Kirsten sarcoma virus-transformed cells are resistant to the action of specific oncogenes
.
Proc, natn. Acad. Sci. USA
.
80
,
5602
5606
.
Norton
,
J. D.
,
Cook
,
F.
,
Roberts
,
P. C.
,
Clewley
,
J. P.
&
Avery
,
R. J.
(
1984
).
Expression of Kirsten murine sarcoma virus in transformed nonproducer and revertant NIH/3T3 cells: evidence for cell-mediated resistance to a viral oncogene in phenotypic reversion
.
J. Virol
.
50
,
439
444
.
Nowell
,
P.
,
Finan
,
J.
,
Favera
,
R. D.
,
Gallo
,
R. C.
,
Ar-Rushdi
,
A.
,
Romanczuki
,
H.
,
Selden
,
J. R.
,
Emanual
,
B. S.
,
Rovera
,
G.
&
Croce
,
C. M.
(
1983
).
Association of amplified oncogene c-myc with an abnormally banded chromosome 8 in a human leukaemia cell line
.
Nature, Lond
.
306
,
494
497
.
Ohno
,
S.
(
1971
).
Genetic implication of karyological instability of malignant somatic cells
.
Physiol. Rev
.
51
,
496
526
.
Orkin
,
S. H.
,
Goldman
,
D. S.
&
Sallan
,
S. E.
(
1984
).
Development of homozygosity for chromosome lip markers in Wilms’ tumour
.
Nature, Land
.
309
,
172
174
.
Pathak
,
S.
,
Strong
,
L. C.
,
Ferrell
,
R. E.
&
Trindade
,
A.
(
1982
).
Familial renal cell carcinoma with a 3 ; 11 chromosome translocation limited to tumor cells
.
Science
217
,
939
941
.
Payne
,
G. S.
,
Courtneidge
,
S. A.
,
Crittenden
,
L. B.
,
Fadly
,
A. M.
,
Bishop
,
J. M.
&
Varmus
,
H. E.
(
1981
).
Analysis of avian leukosis virus DNA and RNA in bursal tumors: viral gene expression is not required for maintenance of the tumor state
.
Cell
23
,
311
322
.
Peehl
,
D. M.
&
Stanbridge
,
E. J.
(
1982
).
The role of differentiation in the suppression of tumorigenicity in human hybrid cells
.
Int.J. Cancer
30
,
113
120
.
Rabbitts
,
T. H.
,
Forster
,
A.
,
Hamlyn
,
P.
&
Baer
,
R.
(
1984
).
Effect of somatic mutation within translocated c-myc genes in Burkitt’s lymphoma
.
Nature, Lond
.
309
,
592
597
.
Rowley
,
J. D.
(
1984
).
Biological implications of consistent chromosome rearrangements in leukemia and lymphoma
.
Cancer Res
.
44
,
3159
3168
Rüther
,
U.
,
Wagner
,
E. F.
&
Müller
,
R.
(
1985
).
Analysis of the differentiation-promoting potential of inducible c-fos genes introduced in to embryonal carcinoma cells
.
EMBO J
.
4
,
1775
1781
.
Sager
,
R.
(
1985
).
Genetic suppression of tumor formation
.
Advan. Cancer Res
.
44
,
43
68
.
Santos
,
E.
,
Martin-Zanca
,
D.
,
Reddy
,
E. P.
,
Pierotti
,
M. A.
,
Della Porta
,
G.
&
Barbacid
,
M.
(
1984
).
Malignant activation of a K-ras oncogene in lung carcinoma but not in normal tissue of the same patient
.
Science
223
,
661
664
.
Schmidt
,
A.
,
Setoyama
,
C.
&
De Crombrugghe
,
B.
(
1985
).
Regulation of a collagen gene promoter by the product of viral tnos oncogene
.
Nature, Land
.
314
,
286
289
.
Schwab
,
M.
,
Alitalo
,
K.
,
Klempnauer
,
K.-H.
,
Varmus
,
H. E.
,
Bishop
,
J. M.
,
Gilbert
,
F.
,
Brodeur
,
G.
,
Goldstein
,
M.
&
Trent
,
J.
(
1983
).
Nature, Lond
.
305
,
245
248
.
Sidebottom
,
E.
&
Clark
,
S. R.
(
1983
).
Cell fusion segregates progressive growth from metastasis
.
Br. J. Cancer
47
,
399
406
.
Sparkes
,
R. S.
,
Murphree
,
A. L.
,
Lingua
,
R. W.
,
Sparkes
,
M. C.
,
Field
,
L. L.
,
Funderburk
,
S. J.
&
Benedict
,
W. F.
(
1983
).
Gene for hereditary retinoblastoma assigned to human chromosome 13 by linkage to esterase D
.
Science
219
,
971
973
.
Srivastava
,
A.
,
Norris
,
J. S.
,
Schmooker-Reis
,
R. J.
&
Goldstein
,
S.
(
1985
).
c-H-ras 1 protooncogene amplification and overexpression during the limited replicative life-span of normal human fibroblasts
.
J biol. Chetn
.
260
,
6404
6409
Stackpole
,
C. W.
(
1981
).
Distinct lung-colonizing and lung-metastasizing cell populations m B16 mouse melanoma
.
Nature, Lond
.
289
,
798
800
.
Stanbridge
,
E. J.
,
Flandermeyer
,
R. R.
,
Daniels
,
D. W.
&
Nelson-Rees
,
W. A.
(
1981
).
Specific chromosome loss associated with the expression of tumorigenicity in human cell hybrids Somat
Cell Genet
7
,
699
712
Stanbridge
,
E. J.
&
Ceredig
,
R.
(
1981
).
Growth-regulatory control of human cell hybrids in nude mice
41
,
573
587
.
Stoler
,
A.
&
Bouck
,
N.
(
1985
).
Identification of a single chromosome in the normal human genome essential for suppression of hamster cell transformation
.
Proc. natn. Acad. Sci. U.SA
.
82
,
570
574
.
Thompson
,
C. R.
,
Challoner
,
P. B.
,
Neiman
,
P. E.
&
Groudine
,
M.
(
1985
).
Levels of c-myc oncogene mRNA are invariant throughout the cell cycle
.
Nature, Lond
.
314
,
363
366
.
Trüeb
,
B.
,
Lewis
,
J. B.
&
Carter
,
W. G.
(
1985
).
Translatable mRNA for GP140 (a subunit of Type VI collagen) is absent in SV40 transformed fibroblasts
.
J. Cell Biol
.
100
,
638
641
.
Vaheri
,
A.
,
Kurkinen
,
M.
,
Lehto
,
V.-P.
,
Linder
,
E.
&
Timpl
,
R.
(
1978
).
Codistribution of pericellular matrix proteins in cultured fibroblasts and loss in transformation: Fibronectin and procollagen
.
Proc. natn. Acad. Sci. U.SA
.
75
,
4944
4948
.
Yoakum
,
G. H.
,
Lechner
,
J. F.
,
Gabrielson
,
E. W.
,
Korba
,
B. E.
,
Malan-Shibley
,
L.
,
Willey
,
J. C.
,
Valerio
,
M. G.
,
Shamsuddin
,
A. M.
,
Trump
,
B. F.
&
Harris
,
C. C.
(
1985
).
Transformation of human bronchial epithelial cells transfected by Harvey ras oncogene
.
Science
227
,
1174
1179
.