Progress in Nucleic Acid Research and Molecular Biology
Latent Viruses and Mutated
Oncogenes: No Evidence
H. DUESBERG AND
Department of Molecular and Cell Biology
University of California at Berkeley
Berkeley, California 94720
A. Latent Viruses as Harmless Passengers
The inactive viruses associated with fatal diseases such
as AIDS, hepatitis C, cervical cancer, T-cell leukemia, hepatoma, Burkitt's
lymphoma, and encephalitis are all not disease-specific. They are common,
like HSV, HPV, EBV, and HBV (3, 12), or rare, like HIV and HTLV-I (54),
in healthy persons. The long "latent periods" and the low incidence
of "viral" disease among virus carriers indicate that such infections
are typically not pathogenic. Although the term "latent period"
implies that the virus becomes active thereafter, even this is almost never
true (see Section II and III). During the presumably virus-caused diseases,
including AIDS, cervical cancer, T-cell leukemia, hepatoma, or panencephalitis,
the virus remains typically inactive, leaving pathogenic functions to unnamed
cofactors. And there is no cofactor that has been found only during the
disease but not prior to it. It is hardly surprising that latent viruses
or fragments of their DNAs are still there if their host develops a nonviral
disease. Thus, the latent viruses are innocent bystanders or "passengers,"
rather than drivers, in nonviral disease processes (159).
B. Drugs as Alternatives
to Hypothetical Viral Pathogens
The great triumphs in the pursuit of microbial and viral
pathogens in the last 100 years have eclipsed, and even led to the ridicule
of, alternative, less spectacular, explanations of disease, such as pathogenic
drugs and toxins (15, 324-326). Although we are in the middle of a drug-use
epidemic in America, the pathology and epidemiology of recreational drugs,
and even of some medical drugs such as AZT, are virtually unstudied by
the scientific community (155).
The drug-AIDS hypothesis, described in Section II, A,
2, is one example of how drug use could cause AIDS diseases (54, 60, 103).
Psychoactive drugs and medical drugs could explain diseases caused by the
depletion of many cells, such as the depletion of T-cells in AIDS or of
hepatocytes in hepatitis C, much better than can dormant viruses. Indeed,
both of these diseases are observed primarily in drug addicts (54, 103,
160). Drug toxicity is also much more compatible with the restriction of
these diseases to risk groups, as, for example, AIDS, which is almost exclusively
restricted to users of recreational drugs and anti-HIV drugs such as AZT
(like lung cancer is to smokers).
Exogenous toxins could also explain the actions of putative
viral tumors, such as nitrite inhalants causing Kaposi sarcomas and AZT
causing lymphomas (69, 103), smoking possibly causing cervical cancer (198,
204), nutritional toxins causing hepatomas, and radiation possibly causing
T-cell leukemia (190) (see Section III). Toxins would also provide a plausible
explanation for the lack of contagiousness of these "viral" diseases.
The cumulative effects of drug or nutrient toxicity over time are compatible
with the appearance of these diseases relatively late in life and at unpredictable
intervals after infection by presumed viral causes. By contrast, viruses
as self-replicating toxins all cause diseases soon after infection. In
light of this theory, hypothetical linkages between infection by a virus
and a subsequent onset of disease via long and unpredictable latent periods
of up to 55 years would dissipate, because infection and pathogenesis are
C. Mutated Genes and Latent
Viruses as Trivial Genetic Scars of Cancer Cells
The spontaneous or virus-induced mutations in tumor cells
are also not disease-specific. For example, point-mutated proto-ras
genes have been observed in chemically induced skin hyperplasias of laboratory
mice (280) and in spontaneous liver hyperplasias of B6C3FI mice (281) that
all spontaneously revert to normal. Further, they have been observed in
reversible skin hyperplasias of humans (282, 327) and in human hemopoietic
hyperplasias (238, 284). Moreover, a recent study of transgenic mice concluded
that "... expression of the mutant [proto-Ha-ras] gene via
its own promoter at the normal chromosomal locus is nontransforming"
(R. Finney and J.M. Bishop, 7th Annual Oncogene Meeting, Frederick, Maryland,
1991, personal communication). In addition, point-mutations and all other
mutations affecting hypothetical tumor suppressor genes are not tumor-specific.
They are detected singly and in all combinations, including mutated proto-ras,
in benign colon adenomas at about the same rates as in malignant carcinomas
Proto-abl translocations are seen in functional
granulocytes that are overproduced during the chronic, hyperplastic phase
of myelogenous leukemia (242, 243) (see Section IV). Hormone-dependent
mammary hyperplasias with int genes mutated by integrated MMT proviruses
have been described (see Section IV). Also, the DNA of hepatitis B virus
has been detected and is expressed in non-tumorigenic liver cells more
consistently than in hepatomas (196, 211). Inactive and defective HPV DNA
is routinely detected in non-tumorigenic tissues with the commercial Vira/Pap
test or with the PCR (199). And HTLV-I is almost only detected in normal
rather than leukemic carriers (see Section III). Further, viable transgenic
mice with mutated proto-abl, proto-myc, and proto-ras,
and even with hypothetically cooperating combinations of proto-myc
and proto-ras, have been constructed, and some are commercially
bred ("OncoMouse-TM shortens the path to knowledge ...," Dupont
Co., Wilmington, DE, 1990) (236). This argues either for even more cofactors
or for other mechanisms altogether.
Thus, spontaneous and viral mutations of tumor cells are
not disease-specific. These findings confirm the above calculations that
the probability of these mutations is much higher than the incidence of
cancer and that carcinogenesis even among hyperplasias is still a very
rare event. In view of this, we agree with Bishop that "the nomenclature
for the affected genes [oncogenes] is unfortunate, since it is based largely
on occasional [presumed] pathogenic aberrations...." (9).
Nevertheless, since clonal tumors have been observed to
emerge from hyperplasias and transgenic animals at a higher-than-normal
rate, their mutated genes and latent viruses could play roles in carcinogenesis
that are not analogous to those played by the biochemically active models
that led to their discovery. For example, they could alter growth control
genes and thus generate hyperplasias. However, even this is speculation
because the mutations and latent viruses are not consistently found in
hyperplastic cells, with the exception of HPV in papillomas (13) (see Section
III). Therefore, they must be presumed innocent until proven guilty (326).
In view of this, we propose that the mutations and latent
viruses that are found in tumor cells are trivial genetic scars that were
picked up by non-tumorigenic somatic cells during many generations of growth
in the presence of mutagens or viruses. Because of detection and reporting
biases in favor of disease, the mutations and latent viruses would be reported
more often in diseased than in healthy carriers. Further, the mutations
and viruses would be more readily and more often observed in cancer cells
than in non-tumorigenic somatic tissues, because cancers are clonal populations
of cells (192, 193, 328) that provide multiple copies of identical mutations,
biological equivalents of the PCR. In contrast, such mutations, including
latent and fragmented viral DNAs, would not be detectable in mutationally
"heterogenous" populations of normal cells, unless individual
cells were cloned or their nucleic acids were amplified.
Since many of these somatic mutations could be incompatible
with normal fetal development, they would not be seen in the germline (329)
and thus not in an average normal cell. The many congenitally and genetically
transmitted animal (6) and human retroviruses, including HTLV-I and HIV
(54), would be notable exceptions. Apparently, retroviruses are so harmless
that they can be accepted as parasites even during development (2).
The hypothesis correctly predicts the same mutations and
latent viruses in non-tumorigenic somatic cells and in tumors that emerged
from these cells, as, for example, the proto-ras and other mutations
or the many "tumor" viruses that are shared by tumorigenic and
nontumorigenic cells. Further, the hypothesis correctly accounts for the
"too many mutations in human tumors" observed by Loeb (47), perhaps
those that were considered irrelevant for carcinogenesis by Heidelberger
("I don't care if cells are 90% transformed, I am only interested
in the last 10%.") (330). In view of this hypothesis, the latent viruses
and nonactivating mutations of cellular genes in cancer cells would be
D. Cancer by Somatic Gene
The clonality, irreversibility, and predictable course
of most cancers all indicate that cancer has a genetic basis. Yet an autonomous
cellular cancer gene, or a complement of interdependent ones, that can
be activated by the statistically cheap mutations observed in hypothetical
oncogenes and anti-oncogenes is improbable on the following grounds.
(i) Nothing could be more terminal for a multicellular
organism than a battery of latent cancer genes that are as easy to "activate"
as the over 50 putative cellular oncogenes that have been named or the
unnamed oncogenes that are said to be activated by inactivation of suppressor
genes (6, 8, 9, 331). The activation of just one dominant oncogene would
be sufficient to initiate a clonal cancer and thus to kill the organism.
By comparison, activation of a hypothetical death gene would kill only
a single cell.
Indeed, since each of these oncogenes is thought to be
activated via point-mutations, truncations, and virus insertions and since
the probability of such mutational events is as high as 1 in 106 per mitosis
and gene, and is as high as 1 in 109 per mitosis and nucleotide (see above
estimates for proto-ras, proto-abl, and rb) (37, 38,
47, 277), multicellular organisms such as humans, with about 1016 cells
per average 70-year lifespan, would generate at least 50 x 1016 : 109
= 5 x 108 cancer cells per lifetime. This number would be even higher if
multiple mutational sites for the activation of specific oncogenes and
for the inactivation of specific anti-oncogenes were considered (6, 8).
It would be further enhanced by the multiplicity of certain oncogenes that
exist as large families, including proto-myc and proto-ras
Nevertheless, the numerology of mutations could be reconciled
with the real incidence of cancer by postulating adequate numbers of cooperating
mutations, as has been attempted in the case of colon cancer (see Section
IV). However, this would be analogous to the invention of more and more
Ptolomaic epicycles by geocentrists, in the face of Galileo's challenge
that the earth was not the center of the solar system. Naturally, the relevance
of these efforts to carcinogenesis would depend on functional proof.
(ii) Based on the only proven examples of "mutated
cellular" oncogenes, the retroviral oncogenes, a cellular gene would
have to become about 100-fold more active than normal to become a cancer
gene. However, the odds of truly activating a gene about 100-fold over
the level for which it has been optimized during 3 billion years of evolution
by spontaneous mutation, must be much lower than the odds of the presumably
"activating" point-mutations or truncations or virus insertions
that are observed in the hypothetical proto- and anti-oncogenes of tumors.
The rare, accidental recombinants with imported retroviral promoters, which
in turn have been optimized during virus evolution to override cellular
controls, are as yet the only known examples of oncogenic mutations (37).
The odds for activating a cellular gene 100-fold by spontaneous
mutations would be particularly low for the many interdependent genes that
must determine "how cells govern their replication...." (7),
the presumed natural function of proto- and anti-oncogenes (7, 287a). According
to Bishop, mutational "damage" to the "relays in regulatory
circuitry" (proto-onc genes) and "governors of proliferation"
(anti-oncogenes) is considered a "gain of function" sufficient
to produce cancer (9). These oncogenic functions are postulated to be "dominant
because ... evil overrides good" (9). However, "damage"
of the kind observed in putative oncogenes naturally inactivates genes
causing diseases such as sickle cell anemia and hemophilia (320, 332).
Such damage is a loss of function and thus recessive, because the remaining
"good" gene overrides "evil." Ironically, the same
kind of somatic mutations or damages to genes thought to "activate"
oncogenes are said to perform conventional gene inactivations when they
Indeed, it is one of the most common misconceptions that
cancer is a consequence of unrestricted growth, because unrestricted growth
produces benign hyperplasias, not cancer. According to Cairns, "It
is a common mistake to assume that cancer cells multiply faster than the
normal cells from which they were derived .... The fact is that the cells
of most cancers divide at about the normal rate, and some even less frequently
than their normal counterparts, but they are able to increase in number
because a greater proportion of the cells' progeny remain in the dividing
pool than is normally allowed" (277).
(iii) There is no functional proof for cellular oncogenes,
because according to Stanbridge "... despite intensive efforts to
transform normal human fibroblasts or epithelial cells with varying combinations
of activated cellular oncogenes, the results have been uniformly negative"
(269). Moreover, their presence, unlike that of related viral oncogenes,
does not determine the character of a given type of tumor. Likewise, unmutated
anti-oncogenes fail to revert tumor cells to normal, and mutated anti-oncogenes
fail to distinguish tumors from those in which they are normal (see Section
The somatic mutation hypothesis owes much of its popularity
to the fact that, in the 1960s and 1970s, many carcinogens were found to
be mutagens (335, 338), although substantial non-correlations between carcinogens
and mutagens were also noted (335, 337). In the 1980s, the hypothesis derived
further notoriety from the consensus that proto-onc genes and anti-oncogenes
are the critical targets among the anonymous genes that are mutated by
carcinogens (9, 287a). Says Weinberg: "Mutations that potentiate the
activities of proto-oncogenes create the oncogenes that force the growth
of tumor cells. Conversely, genetic lesions that inactivate suppressor
genes liberate the cell ... yielding the unconstrained growth of the cancer
cell" (287a). However, not even one of the many somatic mutations
observed to date in cancer cells has been shown to function as a cancer
gene. According to Pitot: "... that carcinogens are mutagenic or may
be converted to mutagens is important but not direct evidence for the genetic
origin of neoplasia" (16).
In sum, the gene mutation hypothesis of cancer is numerically
and evolutionarily implausible and is functionally unconfirmed. Similar
conclusions were reached by Rous (203, 333) and Rubin (334) after studying
oncogenic viruses and cancer for over 50 and 30 years, respectively. Rous
concluded: "A favorite explanation has been that oncogens (Rous' term
for carcinogens) cause alterations in the genes of the ordinary cells of
the body ... somatic mutations as these are termed. But numerous facts,
when taken together, decisively exclude this supposition" (203). "A
hypothesis is best known by its fruits. What have been those of the somatic
mutation hypothesis? ... It acts as a tranquilizer on those who believe
in it, and this at a time when every worker should feel goaded now and
again by his ignorance of what cancer is" (333). Likewise, Cairns
"... suggests that most human cancers are not caused by conventional
mutagens ..." (335).
E. Chromosome Abnormalities
as Causes of Cancer
But if there are no cellular genes that are converted
to cancer genes by somatic mutations, cancer would have to be caused by
normal cellular genes. Perhaps a cell could become transformed by gross
numerical imbalances of normal genes, e.g., via chromosome abnormalities,
just as a computer could be rendered uncontrollable by deleting, duplicating,
and misplacing intact chips, or by altering the operating software. To
test this hypothesis, it would be necessary to determine how probable such
abnormalities are compared to cancer and whether abnormalities exist that
Indeed, chromosome abnormalities are the oldest, and,
as yet, the only consistent observation made on cancer cells. It was postulated
by Boveri in 1914, prior to the discovery of DNA and point-mutations, that
cancer would be caused by abnormal chromosomes (194, 336). The clonal origin
of tumors, the stemline concept predicted by Boveri and defined by Winge
in 1930 (336), is the strongest support for the view that clonal chromosome
abnormalities are the causes, rather than consequences, of carcinogenesis.
This abnormal chromosome-cancer hypothesis would explain
why chromosome abnormalities are consistently found in tumors with or without
mutated cellular oncogenes and with or without latent viruses.
The hypothesis predicts that diploid cancers that differ
from normal cells only in mutated oncogenes or anti-oncogenes are not observed,
because certain chromosome abnormalities instead of somatic mutations of
specific genes are carcinogenic. Tumor progression would be a consequence
of further discontinuous chromosome abnormalities. The hypothesis would
readily resolve the paradox that all "viral" tumors presumably
caused by HTLV-I, HBV, HPV, HSV, and MMTV have clonal chromosome abnormalities.
By contrast, all virus-cancer hypotheses would have to make the odd assumption
that only cells with preexisting chromosome abnormalities are transformed
by these "tumor" viruses.
Our hypothesis also explains why "... despite intensive
efforts to transform normal human fibroblasts or epithelial cells with
varying combinations of activated cellular oncogenes, the results have
been uniformly negative" (269). In addition, the hypothesis explains
why mutated proto-onc genes and anti-oncogenes do not distinguish
tumors by their presence. According to our hypothesis, accidental somatic
mutations generated by chromosome translocations, such as rearranged proto-myc
or proto-abl genes, would be as irrelevant to carcinogenesis as
other mutations of specific genes, such as point-mutated ras genes.
Further, the hypothesis would explain why transgenic mice with activated
oncogenes are breedable and why retinoblastoma cells remain carcinogenic
for mice, even if they are infected by a retrovirus that overexpresses
its presumed suppressor, rb anti-oncogene (see Section IV). Our
hypothesis would also resolve the discrepancy between the rather high probability
and incidence of mutation or "activation" of proto-onc
genes compared to the much lower probability and incidence of cancer (see
Section IV) (37, 337).
We have previously proposed another alternative to the
oncogene hypothesis. It holds that cancer genes are generated by substituting
the normal promoters of proto-onc genes via rare illegitimate recombinations
by strong heterologous promoters from viruses or from cellular genes (37).
As yet, the retroviral oncogenes are the only proven examples of this hypothesis
(40, 43, 44). The relevance of this hypothesis to virus-free tumors depends
on whether the cell contains promoters that are as strong as those of viruses.