VIRUSMYTH HOMEPAGE

Progress in Nucleic Acid Research and Molecular Biology
43:135-204, 1992


Latent Viruses and Mutated
Oncogenes: No Evidence
for Pathogenicity

PETER H. DUESBERG AND
JODY R. SCHWARTZ

Department of Molecular and Cell Biology
University of California at Berkeley
Berkeley, California 94720



IV. Mutated Oncogenes, Anti-oncogenes, and Cancer

A. Mutated Proto-myc Genes and Burkitt's Lymphoma

The transforming gene of the directly oncogenic avian carcinoma virus MC29 contains a specific coding region, now termed myc (217), derived from a cellular gene termed proto-myc (218). Thus, the viral myc gene is a genetic hybrid that consists of a strong retroviral promoter linked to a coding region that is a hybrid of virus- and proto-myc derived sequences (219). This viral myc gene, like synthetic hybrids in which the native proto-myc promoter is replaced with that of a retrovirus (40, 42), is expressed to about 100-fold higher levels in all virus-transformed cells in vitro and in viral tumors than the cellular proto-myc genes (220-222).

The cellular proto-myc gene, located on chromosome 8, is rearranged with immunoglobulin genes from chromosomes 2, 14 and 22 in all (29) or most (30) cell lines derived from Burkitt's lymphomas. However, direct cytogenetic studies show that chromosome 8 is rearranged in only about 50% of primary Burkitt's lymphomas (223-226). Analogous rearrangements have also been observed in the proto-myc genes of mouse plasmacytoma cell lines (1, 8, 36). The rearrangements do not alter the coding region of proto-myc genes. Most rearrangements link the proto-myc coding regions to genetic elements from cellular immunoglobulin genes in the opposite transcriptional orientation (1, 8, 36). Other rearrangements in Burkitt's lymphomas do not affect the location and structure of proto-myc on chromosome 8, but instead rearrange regions 3' from proto-myc (36, 227-232). Because both retroviral myc genes and the rearranged proto-myc genes of most, but not all, Burkitt's lymphomas differ from normal proto-myc genes in truncations 5' from the coding region, and because both were found in cancers, the viral and rearranged cellular myc genes were proposed to be equivalent oncogenes (6, 8, 29, 30).

The transcriptional activity of the rearranged proto-myc genes in lymphomas is moderately enhanced, not altered, or even suppressed in Burkitt's lymphoma cells compared to normal proliferating cells (5, 30, 36, 216, 227). It is thus nearly 100-fold lower than that of viral myc genes or proto-myc genes artificially linked to retroviral promoters (40, 42, 220-222, 233).

Moreover, rearranged proto-rnyc genes from Burkitt's lymphomas do not transform any human or rodent cells upon transfection (5, 36, 38)-even if they are artificially linked to retroviral promoters (234, 236). In efforts to develop a system that is more efficient than transfection for introducing mutated proto-myc genes into cells or animals, synthetic avian retroviruses with the coding region of the human proto-myc gene were constructed (233, 237). Since these viruses transform avian cells, it was concluded that "ungoverned expression of the gene can contribute to the genesis of human tumors" (237). However, transformation of human cells was not demonstrated. Moreover, three independent studies report that murine cells cannot be transformed by authentic avian (238) and synthetic murine retroviruses with myc genes (239, 240), signaling a restricted transforming host range of myc genes.

Several arguments cast doubt on the hypothesis that rearranged proto-myc genes of Burkitt's lymphomas are functionally equivalent to retroviral myc genes and thus oncogenic:

1. Rearranged proto-myc genes from Burkitt's lymphomas or mouse plasmacytomas lack transforming function in transfection assays, while retroviral myc genes and proto-myc genes driven by retroviral promoters are sufficient to transform at least avian primary embryo cells (40, 42, 237). This indicates that the proto-myc genes from lymphomas and viral myc genes are functionally not equivalent.

2. Since expression of rearranged proto-myc genes from lymphomas is either the same as, or similar to, that of normal proto-myc genes, and their coding regions are identical, rearranged proto-myc cannot be sufficient for lymphomagenesis. By contrast, viral myc genes are oncogenic, owing to a 100-fold higher level of myc expression.

3. Primary Burkitt's lymphomas with normal chromosome 8, and with rearrangements of chromosome 8 that occur 3' from proto-myc and thus do not affect the structure and regulation of the proto-myc gene, indicate that proto-myc translocation is not necessary for Burkitt's lymphomas.

It follows that rearranged proto-myc genes of human and animal tumors are transcriptionally and functionally not equivalent to viral myc genes, and that they are not necessary for lymphomagenesis.

In view of this, the demonstration (237) that human proto-myc transforms avian cells after it had been converted artificially to a retroviral myc gene is not relevant to its hypothetical role in human tumors. This claim is all the more questionable because even retrovirus-promoted myc genes appear unable to transform non-avian cells. Instead, such experiments model the genesis of a viral myc gene from a retrovirus and a cellular proto-myc gene by rare illegitimate recombination (37). The critical step in this process is the substitution of the weak cellular promoter by the strong retroviral counterpart (40, 42).

Thus, there is currently only circumstantial evidence for the hypothesis that rearranged proto-myc genes play a role in Burkitt's lymphomas. This evidence includes the structural, but not functional, similarity to viral myc genes, and the approximately 50% incidence of chromosome-8 rearrangements with breakpoints near proto-myc in primary lymphomas. In view of this, rearranged proto-myc genes either may be involved in a mechanism of leukemogenesis that is not analogous to the viral model, or they may not be involved at all. Since the incidence of chromosome-8 rearrangements is higher in lymphoma cell lines than in primary lymphomas, it has been pointed out that the rearrangement may favor lymphoma cell growth in culture (225).

In efforts to link the proto-myc rearrangements with a role in tumorigenesis, despite these discrepancies with the one-gene model, it was postulated that rearranged proto-myc genes may cooperate with other genes for carcinogenesis (236, 238, 241 ). To test these ad hoc hypotheses, transgenic mice were constructed that carry rearranged proto-myc genes linked to artificial promoters and hypothetical helper genes in every cell of their bodies. However, only some of these mice developed clonal tumors late in their lives (236). This indicates that even these combinations are not sufficient for carcinogenesis. Consequently, further helper genes were postulated (236, 241).

An alternative hypothesis suggests that the appearance of certain chromosome abnormalities is sufficient for lymphomagenesis. It is consistent with this proposal that cytogenic studies have identified chromosome abnormalities in all Burkitt's lymphomas, even in those that lack rearranged proto-myc genes (224-226). The reason that a high percentage of these rearrangements include proto-myc and immunoglobulin genes may be a consequence of the natural functions of these genes in B cells, namely generating antibody diversity in which proto-myc genes may play an active or passive role.

B. Rearranged Proto-abl Genes and Myelogenous Leukemia

Human myelogenous or granulocytic leukemia develops in two stages. The first is a chronic phase that may last, on average, 3-4 years. During this phase, immature myeloblasts are overproduced in the bone marrow and appear in the blood, but may differentiate into functional cells. This hyperplastic stage is followed by a terminal blast crisis of several months, during which non-functional leukemic cells emerge (242, 243). The leukemic cells of both the chronic and terminal stages in 85-90% of patients are marked by a reciprocal translocation between chromosomes 9 and 22. The rearranged chromosome 22 is termed the Philadelphia chromosome (193). In the remaining 10-15% of cases, chromosome 22 is rearranged with other chromosomes (193, 242-245). The reciprocal translocation between chromosomes 9 and 22 substitutes the 5' end of the coding sequence of the proto-abl gene on chromosome 9 with a 5' regulatory and coding element of a gene of unknown function, termed bcr (for breakpoint cluster region), from chromosome 22 (33, 246-248). The breakpoints with regard to the proto-abl gene vary over 200 kb (249), but those within bcr fall in a range of 5.8 kb (8, 247, 248). The transcriptional activity of the proto-abl gene is virtually unaffected by the translocation (8, 246).

The proto-abl gene is the cellular precursor of the transforming gene of the murine Abelson leukemia virus (6). This virus is sufficient to cause terminal myelogenous leukemia in susceptible mice within 3-5 weeks after infection (250, 251). In this virus, the promoter and 5' coding sequence of proto-abl are replaced by retroviral counterparts. Since the 5' proto-abl coding regions are substituted by heterologous genetic elements in both the virus and the leukemias, it has been postulated that the structurally altered proto-abl gene of the leukemia is a cellular oncogene that is functionally equivalent to the transforming gene of Abelson virus (7, 8, 246, 252). However, the Abelson virus or provirus (253), but not the DNA of human leukemic cells, is capable of transforming the mouse NIH 3T3 cell line in vitro (8).

The failure of the bcr-proto-abl hybrid genes to function like the virus could be a technical problem, because the hybrid genes may not be transfectable due to their large size of over 200 kb (8, 249). Therefore, the transforming function of a cDNA transcribed from the 8.5-kb mRNA of the bcr-proto-abl was tested in murine retrovirus vectors. In such vectors, as in wild-type Abelson virus (251), the transcriptional activity of the abl gene is about 100 times that of normal or rearranged cellular proto-abl genes (252, 254, 255). One such recombinant virus induced proliferation of lymphoid mouse cells in vitro (254). Another induced clonal lymphomas when introduced into the germline of transgenic mice (255). Finally, a myelogenous leukemia was obtained by infecting bone marrow in vitro with the synthetic virus and transplanting this marrow into irradiated syngeneic mice (252). The leukemias appeared after relatively short latent periods of 9 weeks (252), almost as fast as those caused by the wild-type Abelson virus (250). The karyotype of this leukemia was not described (252).

Yet several observations cast doubt on the hypothesis that the rearranged proto-abl gene from human chronic myelogenous leukemias is functionally equivalent to the transforming gene of Abelson virus and that it is leukemogenic:

1. The transcriptional activity of the rearranged proto-abl gene in the leukemias is about 1% of that of wild-type Abelson virus and those of the synthetic recombinant viruses. Thus, mutated cellular proto-abl genes and viral abl genes are functionally not equivalent.

2. Given estimates that chromosome translocations occur spontaneously in human cells in 1 out of 102 to 104 mitoses (37, 256, 257), it can be calculated that a brc-proto-abl rearrangement would be much more probable than chronic myelogenous leukemia. The probablity that a random reciprocal rearrangement falls within the 200-kb 5' region of proto-abl and the 5.8-kb 5' region of bcr of the 106-kb human genome is (200 : 106) x (5.8 : 106) or 10-9. Thus, 1 in 109 translocations would generate a Philadelphia chromosome. Considering that humans carry about 1010 to 1011 lymphocytes (186), which are replaced at least six times per year (53), or 420 times in an average lifetime of 70 years, a human life represents at least 1013 mitoses of lymphocytes. Making the conservative assumption that a translocation occurs in 1 out of 104 human mitoses (256, 257), about 109 (1013 : 104) lymphocytes with rearranged chromosomes are generated in a lifetime. Accordingly, every human should, by the age of 70, develop 1, and possibly 100, lymphocytes with a Philadelphia chromosome (109 : 109) and thus leukemia. However, chronic myelogenous leukemia is observed in only 1 (242) to 2.4 (197, 258) out of 100,000 per year or about 0.1% of people in a 70-year lifetime. Therefore, a rearranged proto-abl gene appears not to be sufficient for leukemogenesis.

3. Since in 10-15% of the chronic myelogenous leukemia cases proto-abl is not rearranged (193, 244), proto-abl mutation is not necessary for leukemogenesis. According to Nowell, "These variants appear to have no significance with respect to the clinical characteristics of the disease, and so it appears that it is the displacement of the sequence of chromosome 22 that is of major importance, rather than the site to which it goes" (193).

Thus, a rearranged proto-abl is functionally not equivalent to the transforming gene of Abelson virus. The rearrangement appears to be more probable than a leukemia, and is not even necessary for the leukemia. It is consistent with the first point that the proto-abl translocation is observed in the rather benign, early stage of chronic myelogenous leukemia, in which cells can differentiate into functional myeloblasts (242, 243), whereas the Abelson virus causes a terminal leukemia within several weeks.

Since the transcriptional activity of retroviral abl genes is about 100 times that of normal and rearranged proto-abl genes, and since it is not known whether even a viral abl gene can transform a human cell, the claims that "retrovirus-mediated expression of the bcr-proto-abl protein provides a murine model system for further analysis of the disease" (252) are not realistic. These claims fail to take into consideration the 100-fold transcriptional discrepancy between the retroviral and cellular abl genes and the question of whether the transforming host range of abl genes includes human cells. Therefore, synthetic proto-abl viruses are just experimental reproductions of the rare spontaneous genesis of retroviral transforming genes from normal cellular genes and retroviruses. The critical step in this process is the recombination of the coding region of a proto-onc gene with a retroviral promoter (37).

It follows that the 85-90% incidence of proto-abl rearrangements in chronic myelogenous leukemia and the structural similarity of the gene to that of Abelson virus are the only evidence to suggest that proto-abl plays a role in human leukemogenesis. In view of this, proto-abl either must be involved in human leukemogenesis by a mechanism that is not analogous to that of the viral counterpart, or it may not be involved at all.

An alternative hypothesis suggests that alterations of the normal balance of chromosomes cause the leukemia. According to this hypothesis, the Philadelphia translocation would only affect the growth control of the cell. This is consistent with the rather normal function of cells with the translocation during the 3-4 years prior to the blast crisis. In one case, a person with a Philadelphia chromosome did not develop a leukemia for at least 7 years (P. H. Fitzgerald, personal communication, 1985) (245). Indeed, the blast crisis of myelogenous leukemia is accompanied by further chromosomal abnormalities, which are observed in leukemia with and without rearranged proto-abl genes (193, 244).

C. Point-mutated Proto-ras Genes and Cancer

Two laboratories have reported that transfection of the DNA of a human bladder carcinoma cell line transforms morphologically the mouse NIH 3T3 cell line (259, 260). Subsequent cloning proved the transforming DNA to be the coding region of the proto-ras gene, the same gene from which the coding region of the ras gene of the murine Harvey sarcoma virus is derived. Sequencing indicated that the 3T3 cell-transforming proto-ras from the bladder carcinoma cells differs from normal proto-ras in a point-mutation in codon 12 that changes the native Gly to Val (23, 26, 261).

Further transfection analyses with the 3T3-cell-transformation assay detected point-mutated proto-ras genes in less than 1% to about 20% of most common human tumors (1, 6, 36, 262) and in up to 40% of colon cancers (28, 263, 264). The proto-ras genes from these tumors were each from a closely related group that includes the Harvey, Kirsten, and N-ras genes. Like the Harvey gene, the Kirsten proto-ras gene is named after a sarcomagenic murine retrovirus with a coding region of that gene (6). Regardless of point-mutation, proto-ras expression is enhanced 2- to 10-fold in about 50% of tumors compared to normal control tissues (44, 262, 265, 266). Transcription of normal proto-ras is also enhanced in normal proliferating cells (36), as, for example, 8-fold in regenerating rat liver cells (267).

1. The Original ras-Cancer Hypothesis Postulates a First-Order Mechanism of Transformation

The observations that point-mutated proto-ras genes from human and some animal tumors transform mouse 3T3 cells became the basis for the hypothesis that point-mutations of proto-ras genes cause cancer (23, 26, 27). The hypothesis derived additional support from the observation that the ras genes of Harvey and Kirsten sarcoma viruses also differ from normal proto-ras in point-mutations in codon 12 (5, 39, 268). The hypothesis assumes that point-mutations confer to proto-ras genes dominant transforming function that is equivalent to that of sarcomagenic retroviral ras genes (268). Further, it assumes that the 3T3-cell transformation assay measures a preexisting function of mutated cellular proto-ras genes. Consequently, point-mutated proto-ras genes were termed "dominantly acting oncogens" (4, 5, 9, 46, 259, 260, 269). Subsequently, other proto-onc genes, such as proto-myc (270, 271) and proto-src and the src genes of Rous sarcoma virus (275), and even genes that are not structurally related to retroviral oncogenes, such as certain anti-oncogenes (see Section IV, E), were also proposed to derive transforming function from point-mutations (1, 5, 6, 9, 46, 272-274).

Numerous observations designed to test the ras point-mutation-cancer hypothesis indicate that point-mutation is not sufficient for carcinogenesis:

1. Point-mutated proto-ras genes from tumors do not transform diploid embryo cells from rodents or humans, as retroviral ras genes do (238, 276). However, upon simultaneous transfection with other viral oncogenes or cellular genes linked to viral promoters, proto-ras genes transform embyro cells (234, 235, 238). This indicates that point-mutation is not sufficient to convert proto-ras to a gene that can transform normal cells.

2. Numerical arguments based on relative probabilities of point-mutations versus cancer also indicate that point-mutated proto-ras genes are not sufficient for carcinogenesis. The probability of point-mutations is 10-9 per nucleotide and per mitosis in eukaryotic cells (37, 38, 47, 277). Since eukaryotes carry about 109 nucleotides per cell (278) and consist of 1011 (mice) to over 1014 (humans) cells, mice carry 102 (1011 : 109) and humans carry 105 (1014 : 109) cells with the specific point-mutation that changes Gly to Val in codon 12 of proto-ras at any time (37, 38). Since the average cell is replaced about 100 times during a human lifetime of 70 years (37, 277), this number must be multiplied by 100. Moreover, since at least 50 different point-mutations in at least five different codons confer transforming function to proto-ras in the 3T3 assay (39, 279), mice would contain 5 x 103 and humans have 5 x 106 such cells.

3. Further, the existence of point-mutated proto-ras genes in nontumorigenic, hyperplastic tissues (see Section VI) (280-284) and in transgenic mice (236, 241; R. Finney and J.M. Bishop, 7th Annual Meeting on Oncogenes, Frederick, Maryland, 1991, personal communication) indicates that these mutations are not sufficient for carcinogenesis.

4. Point-mutation is not necessary for the transforming function of Harvey and other murine sarcoma viruses, as mutants without point mutations in ras and synthetic retroviruses with normal proto-ras coding regions are almost as oncogenic as those with point-mutations (41, 44, 285). This indicates that viral ras genes derive transforming function from other virus-specific elements (39, 41, 44) and suggests that point-mutation may not be sufficient for proto-ras genes to transform.

5. In primary tumors, point-mutated proto-ras genes are expressed at nearly the same level as normal proto-ras genes (36, 44, 262, 264, 280, 286). By contrast, point-mutated proto-ras genes in cells transformed by transfection are expressed like viral ras genes, which is at a level at least 100-fold higher than native proto-ras genes (44, 234, 235, 262, 280, 286, 287). Thus, the 3T3-cell-transfection assay creates proto-ras expression artifacts that are transcriptionally about 100 times more active than native proto-ras genes from tumors. Their activity is similar to that of retroviral ras genes.

It appears then that a point-mutated but intact cellular proto-ras gene is not sufficient for carcinogenesis. Further, it follows that the transfection assay does not measure a genuine function of pointmutated proto-ras genes as they exist in tumors, but measures that of an expression artifact created during the transfection assay. An analogous functional artifact has been observed upon transfection of an antioncogene (see Section IV, E) (287a).

Such artifacts could be generated during transfection by substituting by illegitimate recombination the native proto-ras regulatory elements by artificial promoters derived from carrier and helper gene DNA (44). Indeed, transformation of primary cells by cellular proto-ras genes depends on the presence of added viral helper genes or on other cellular genes linked to viral promoters (234, 235, 288, 289), or on the presence of retroviral promoters alone (44). This recombination process is entirely analogous to the generation of retroviral ras genes, in which coding regions of normal proto-ras genes are recombined by transduction with heterologous retroviral promoters that enhance the transcription over 100-fold compared to proto-ras (37, 38, 43, 44). In addition, transfection generates concatenated DNA multimers, an artificial gene amplification that would also enhance the dosage of ras transcripts (290-293).

The probable reason that proto-ras genes from tumors transform 3T3 cells, but not primary cells, is that mouse NIH 3T3 cells are much more readily transformed by exogenous genes, as well as spontaneously (294), than are embryo cells (238). Thus, the weak promoters acquired from random sources during transfection are sufficient to convert proto-ras genes with point-mutations to 3T3-cell transforming genes, but not to genes capable of transforming primary cells.

The reason that point-mutated, but rarely normal, proto-ras genes (261) are detected by transfection assays is that point-mutations enhance about 10- to 50-fold the transforming function imparted by heterologous promoters on proto-ras genes (39, 44, 285, 295). Thus, proto-ras genes derive their transforming function from heterologous promoters, and certain point-mutations merely enhance this transforming function.

2. Ad Hoc ras-Cancer Hypotheses Postulating Second- and Higher-Order Mechanisms of Transformation

In view of the evidence that native, point-mutated proto-ras genes detected in some tumors are not equivalent to viral ras genes and not sufficient for carcinogenesis, ad hoc ras-cancer hypotheses have been advanced proposing that cellular ras genes with point-mutations depend on helper genes for carcinogenesis (6, 28, 46, 236, 238). However, the hypothetical helper genes have not been identified in most tumors, except for colon cancers.

In the case of colon cancer, it has been postulated that point-mutated Kirsten and N-ras genes depend on the mutation of at least three tumor suppressor genes for transforming function (28, 46, 272). Yet the incidence of these mutations in colon cancers is not convincing proof for their postulated function for the following reasons.

Among primary colon cancers, about 40% carry point-mutated Kirsten ras genes (28, 263, 264) and some others contain point-mutated N-ras genes (28). In addition 70% of all carcinomas carry deletions or mutations in the presumed tumor suppressor gene DCC (deleted in colon cancer) located on chromosome 18, 75% in the presumed suppressor gene p53 located on chromosome 17 and 30% in the presumed suppressor gene APC (adenomatous polyposis coli) on chromosome 5 (28). Thus, only about 6% (0.4 x 0.7 x 0.75 x 0.3) of the colon cancers studied carry the genetic constellation postulated for colon cancer. About 87% carry various combinations of these mutations, and 7% carry none of the mutations (28). In addition, recent evidence indicates that mutations on chromosome 5 are scattered over several hypothetical suppressors or anti-oncogenes (296). Despite these radical mutational differences among colon carcinomas, the carcinomas do not differ from each other in any known histological or biological properties. In addition, all of these mutations alone, and even together, are also observed in benign colon adenomas (see Section VI) (28). Other tumors with point-mutated proto-ras genes are also histologically and morphologically indistinguishable from counterparts without these mutations (262, 297).

In view of such poor correlations and the absence of ras-specific tumor markers, a functional test is the only method to prove the hypothesis that point-mutated proto-ras genes have transforming function in conjunction with helper genes. However, the only functional test currently available is the 3T3-cell-transfection assay, which generates helper-independent proto-ras expression artifacts. Thus, the hypothesis that point-mutated proto-ras genes play a role in carcinogenesis is based only on circumstantial evidence, namely, structural, but not functional, similarity to viral ras genes. In addition, it is based on epidemiological evidence that mutated genes are more common, or are observed more commonly, in tumors than in normal cells (see Section VI). Moreover, the assumption that mutation of p53 is obligatory for carcinogenesis has not been confirmed in a recent study that generated developmentally normal mice without p53 genes (319a) (see Section IV, E).

It follows either that unrearranged, point-mutated proto-ras genes are oncogenic by a second- or higher-order mechanism of carcinogenesis that is not analogous to the first-order mechanism of viral ras genes and of the transfection artifacts of proto-ras genes, or that they are not relevant to carcinogenesis. Since constellations of mutated proto-ras and helper genes that are tumor-specific have not been found, there is currently no evidence for a role in carcinogenesis.

Therefore, we propose that other events, such as chromosome abnormalities, which are consistently found in colon carcinomas with and without mutated oncogenes or anti-oncogenes (28, 47, 192), may cause colon cancers. The clonal mutations in proto-ras and hypothetical helper genes could reflect the origin of tumor cells from non-tumorigenic somatic cells with the same mutations (see Section VI).

D. int Genes with Integrated Mouse Retroviruses and Mouse Mammary Carcinomas

Mouse mammary tumor virus (MMTV) is one of the many endogenous retroviruses that are genetically and perinatally transmitted but hardly ever expressed by most strains of mice (5, 6, 298). However, inbred female mice of the C3H and GR strains express high titers of mammary tumor virus in their milk. Approximately 90% of the female offspring of C3H mice develop mammary tumors between the ages of 7-10 months (299, 300). Foster-nursing of C3H offspring by virus-free mothers of other strains reduces the risk of tumors to 20-40% and delays their appearance to 18 to 24 months (299, 301). However, wild mice foster-nursed by a C3H mother fail to develop mammary tumors (over a spontaneous background of 3% at 2 years of age), although they are infected by the virus (302, 303).

Virus replication at high titers enhances reversible, hormone-dependent mammary hyperplasias that are poly- or oligoclonal (304). Out of these hyperplasias, clonal tumors emerge that are hormone-independent (304, 305). Thus, infection by milk-borne virus initiates virus replication, hyperplasias, and frequently tumorigenesis at an earlier age compared to spontaneous virus activation and tumorigenesis-but only in certain inbred strains of mice. The virus is replicating in both early and late tumors (305). The tumors are clonal, defined by specific virus integration sites and chromosome abnormalities (2). Since only one out of millions of virus-producing mammary cells becomes tumorigenic, tumorigenesis may be virus-independent, or may be due to virus-mediated activation, or inactivation of a cellular gene, in which case, site-specific provirus integration must be postulated.

Site-specific integrations in mammary tumors were originally observed in three different mouse strain-specific loci, termed int-1 (34), int-2 (306), and int-3 (307). In C3H mice, the provirus is primarily observed in int-1, in BR6 mice in int-2, and in some feral mice in int-3. Subsequently, "numerous" (305) int loci were observed in mouse mammary carcinomas (308). In light of the observation that proto-myc genes in certain chicken leukemias are transcriptionally acitvated by retrovirus insertion (31) these int loci have been postulated to be cellular proto-onc genes that are activated to cancer genes by the promoter of integrated MMT provirus (1, 6, 8, 34, 306, 308). This hypothesis is compatible with the clonality of the mammary tumors.

Integration sites within a given int locus are spread over 20 kb and occur in both transcriptional orientations (1, 2, 8). Viral integrations into int loci are also observed prior to tumorigenesis in hormone-dependent hyperplasias (304, 309). Only 1-10 copies of int RNAs are found in tumor cells that express int genes (310). By comparison, synthetic (39, 40, 42, 44) and natural (31) retrovirus-promoted proto-onc genes make about 103 to 104 copies of RNA per transformed cell. In many viral mammary tumors, the int loci are not expressed, and in some tumors the int loci are expressed but the MMT provirus is not integrated at or near int. For example, int loci are expressed in only 2 out of 9 clonal tumors of GR mice (304), and int loci are expressed in only 19 out of 46 clonal tumors of C3H mice (305). It has also been reported that int loci are expressed in tumors in which MMTV is integrated at non-int sites. Accordingly, there is no report of int-specific tumor markers.

In view of this, several arguments cast doubt on the int-activation hypothesis:

1. Since "numerous" int loci are observed in mammary carcinomas, and since integrations are scattered over 20 kb within a given locus and occur in both orientations, MMTV integration into int loci cannot be sufficient for carcinogenesis based on the following numerical arguments. Given random retrovirus integration (6) and 1 x 106-kb DNA per mouse genome (278), and assuming only five 20-kb int loci, about 1 out of every 104 (5 x 20 out of 106) infected mammary cells should become tumorigenic. Thus, tumors should appear very soon after infection. Since this is not the case, MMTV integration cannot be sufficient for carcinogenesis.

2. Since MMT proviruses integrate into int genes prior to tumorigenesis, provirus-mutated int genes cannot be sufficient for tumorigenesis.

3. Since wild mice are susceptible to the virus and produce the same hormones as the inbred mice that develop mammary carcinomas, even the virus-hormone package is not sufficient for tumorigenesis.

4. Provirus integration into different int loci in different strains of mice indicates that integration is host-directed. Therefore, the virus is not sufficient for site-specific integration and thus for tumorigenesis, if site-specific integration proves to be relevant for tumorigenesis.

5. Since int loci are not expressed in many viral mammary tumors, transcriptional activation of int genes by any mechanism is not necessary for carcinogenesis. It is consistent with this view that proviruses are integrated into int genes in both directions and integration sites are spread over 20 kb, but retroviral promoters activate transcription in only one direction and only over limited distances (42).

6. Since the same tumors are observed with and without integration into int genes, site-specific integration is not necessary for carcinogenesis, because "clonal, hormone-independent tumors . . . seem to be the result of mutations that are unrelated to int activation" (304).

7. The retroviral int-activation hypothesis fails to account for the clonal chromosome abnormalities of all virus-positive tumors that have been characterized (2)-except if one makes the additional odd assumption that MMTV only transforms cells with preexisting chromosome abnormalities.

It thus appears that the MMTV plays only an indirect role in tumorigenesis as one of several factors that enhance mammary hyperplasias, a known risk factor for carcinogenesis (1, 203). This role is similar to that of other highly expressed animal retroviruses in leukemogenesis. For example, inbred viremic mice and chickens have been described that develop virus-induced hyperplasias from which clonal lymphomas or leukemias emerge (2). Alternatively, high levels of retrovirus expression may just signal a heritable loss of intracellular suppressors which could themselves predispose to overgrowth and thus favor carcinogenesis (2). The incidence of 20-40% carcinomas in foster-nursed C3H mice compared to a background of 3% in other laboratory mice (303) supports this view. This would be analogous to the activation of other retroviruses in cells induced to proliferate by genetic damage from chemicals or radiation (see Section III).

In view of this, we propose that mammary carcinogenesis is a rare, spontaneous event initiated by chromosome abnormalities that occur in one out of millions of virus-infected cells. This hypothesis would explain clonal viral integration sites as accidental consequences of the clonal chromosome abnormality that created a tumor cell from a normal virus-infected cell. It would also explain why the carcinomas are not distinguished by the type of int gene that is mutated. The int-loci would be strain-preferred provirus integration regions that are not relevant to tumorigenesis.

E. Constitutive Oncogenes, Mutated Anti-oncogenes, and Cancer

There are heritable and spontaneous retinoblastomas (45). Cytogenetic analyses of both have observed that chromosome 13 is either missing or deleted in 20 to 25% (311,312). In addition, other chromosome abnormalities have been observed in all retinoblastomas (311, 312). On this basis, it was proposed that retinoblastoma arises from the loss of a tumor suppressor or an anti-oncogene, now termed rb, that is part of chromosome 13 (45). In the familial cases, the loss of one rb allele would be inherited and the second one would be lost due to spontaneous mutation. In the spontaneous cases, somatic mutations would have inactivated both loci. In the retinoblastomas with microscopically intact chromosomes 13, submicroscopic mutations were postulated.

This anti-oncogene hypothesis predicts that normal cells would constitutively express oncogenes that render the cell tumorigenic if both alleles of the corresponding suppressor are inactivated. The hypothesis further predicts that the suppressor genes must be active at all times in normal cells. In 1986, Weinberg et al. (313) cloned a human DNA sequence that was missing or altered in about a third of 40 retinoblastomas and in 8 osteosarcomas. Therefore, the gene encoded in this sequence was termed the rb gene. Reportedly, the rb gene was unexpressed in all retinoblastomas and osteosarcomas, even in those without rb deletions (313). The rb gene measures almost 200 kb, includes 27 exons and encodes, from an mRNA of 4.7 kb, a 110,000-dalton protein (8, 278).

An analysis of 34 primary retinoblastomas undertaken to test the hypothesis found deletions of the rb gene in only 4 of 34 tumors analyzed and transcripts of the rb gene were found in 12 out of 17 retinoblastomas and in 2 out of 2 osteosarcomas, casting doubt on the deletion hypothesis (314). The remaining tumors had apparently normal rb genes. However, subsequent studies of retinoblastomas have observed point-mutations and small submicroscopic deletions in rb genes that did not have macrolesions (273, 274, 315, 316). For example, both Weinberg et al. (273) and Lee et al. (274) reported a point-mutation in a splice sequence of the rb gene. In view of this, it is now believed that point-mutations or other minor mutations of the rb genes are sufficient for tumorigenesis (273, 315). However, Gallie et al. reported point-mutations and deletions of rb genes in only 13 out of 21 tumors (315). In an effort to develop a functional assay, a DNA copy of the mRNA of the rb gene was cloned into a retrovirus; infection by this virus inhibited the growth of a retinoblastoma cell line in vitro (274, 317). However, two recent studies show that an intact, synthetic rb gene fails to inhibit tumorigenicity of human retinoblastoma and breast cancer cells in nude mice (318, 318a).

Clearly, the point-mutation hypothesis of the rb gene would never have emerged if the original chromosome deletion hypothesis had been confirmed. It advanced the anti-oncogene hypothesis into a virtually inexhaustible reservoir of hypothetical cancer genes: Any gene with any mutation in each of both alleles in a cancer cell could be a tumor suppressor or anti-oncogene. According to Weinberg, "... one can cast a broad net for tumor suppressor loci by using a large repertoire of polymorphic DNA markers to survey ... for repeated instances of LOH (loss of heterozygosity). Indeed, this genetic strategy has revolutionized the research field" (287a). Over a dozen deleted or point-mutated anti-oncogenes are now considered to cause osteosarcomas, breast cancer, bladder cancer, lung cancer, colon cancer, Wilms' tumor, and neuroblastoma, in addition to retinoblastoma (8, 9, 46, 287a, 317). For example, a point-mutation in one of three genes of a colon cancer cell would signal an inactivated hypothetical colon cancer suppressor gene (272, 296). Further, the range of the rb suppressor gene has since been extended to other cancers, including small cell lung, bladder, prostate, and breast carcinomas, and osteosarcoma (8, 317).

The anti-oncogene hypothesis has been difficult to prove because (a) the oncogenes that are said to be suppressed have not been named or identified (269) and will be difficult to assay because all normal cells or animals should suppress them with the corresponding antioncogenes, and because (b) transfection of an intact rb gene (274, 318) has failed to revert transformed cells to normal and to suppress their tumorigenicity (274, 318, 318a). Likewise the hypothetical colon cancer suppressor gene p53 has failed to revert transformed cells to normal (319) and its complete absence has not affected the normal development of mice (319a). Nevertheless, 74% of these p53-free mice developed lymphomas and sarcomas at six months that probably derived from single cells, rather than through a systemic transformation as the anti-oncogene hypothesis would have predicted (319a).

At this time, the hypothesis suffers from the following short-comings:

1. The probability of point-mutations and minor mutations in both alleles of the rb gene appears much higher than the cancers they are said to cause. Since the rb gene has 27 exons and each exon is flanked by at least four essential splice nucleotides, at least 108 (4 x 27) point-mutations could inactivate the rb gene. In addition, one can assume that point-mutations of at least 10% of the 928 amino acids of the 110,000-dalton rb protein would inactivate the gene (8). Thus, at least 200 point-mutations should be able to inactivate rb. Since 1 in 109 human cells contain any possible point-mutation of the human genome (see Section IV,C), about 1 in 5 x 106 would contain an inactive rb gene, and 1 in (5 x 106)2 or 2.5 x 1013 would contain two inactive rb genes in the same cell. This number would be even higher if other mutations such as minor deletions and chromosome nondysjunctions were included.

Chromosome nondysjunctions are estimated to occur in 1 out of 104 human cells (320, 321). The probability of generating a retinoblastoma cell from a point-mutation in one rb gene and a missing chromosome 13 would be 1 in 5 x 106 x 104 or 1 in 5 x 1010. Thus, every adult human consisting of about 1014 cells would contain at least 1 and possibly 5 x 103 cells in which both rb genes are inactivated, and would develop over 100 to 100,000 such cells in a lifetime of 70 years, which represents about 1016 cells (37, 277). Since inactive rb genes are now said to cause retinoblastomas, osteosarcomas, small cell lung, breast, and bladder carcinomas, etc., and the corresponding tissues represent over 20% of the human body, one would expect at least 20% of humans to develop such a tumor per year.

Moreover, 1 in 5 x 106 cells of every person with one inherited rb mutation should have defects in both rb alleles due to secondary mutation and 1 in 104 cells due to chromosome nondysjunction. A recent review on tumor suppressor genes reports exactly the same probabilities for rb mutations as we do (287a). Thus, all persons with an inherited rb deletion should develop retinoblastomas and other cancers. Since this is not the case (45), point-mutation or deletion of both rb alleles cannot be sufficient for carcinogenesis.

2. Since neither deletion nor minor mutation of rb genes is observed in all retinoblastomas or other specific tumors, rb deletion or mutation is not necessary for tumorigenesis.

3. The relevance of the growth inhibitory function of the artificial retrovirus with an rb coding region to the putative tumor suppressor function of rb is unclear for several reasons: (a) Expression from a retroviral promoter enhances the rb protein concentration at least 100-fold above physiological levels (274) and thus may not be relevant to its normal function. Similar reservations are expressed by Weinberg: "... many genes ... will antagonize growth when they are forced on a cell by ... gene transfer, but this provides no testimony as to whether these genes are normally used by the cell to down-regulate its own proliferation...." (287a). (b) Recently, elevated rather than reduced rb expression was observed in tumor cells (322). (c) Human retinoblastoma cell lines and breast cancer lines transfected with intact and artificially overexpressed rb genes are tumorigenic in nude mice, indicating that the rb gene does not suppress tumorigenesis by retinoblastoma and mammary carcinoma cells (318, 318a).

It follows that deletion or mutation affecting both alleles of the rb and p53 genes is not sufficient and probably not necessary for carcinogenesis since the same retinoblastomas and colon cancers occur in the presence and absence of these genes. An alternative hypothesis suggests that the many chromosome abnormalities associated with retinoblastomas (311,312), other tumors with rb mutations (194) and colon cancers are to blame for carcinogenesis (see Section VI).

Continu


VIRUSMYTH HOMEPAGE