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
III. Viruses as
Causes of Clonal Cancer
A. Human T-cell Leukemia Virus
and Adult T-cell Leukemia
Human T-cell leukemia virus-I (HTLV-I) was originally
discovered in a T-cell line from a leukemic patient (175). This line, termed
HUT 102, only produced virus after it had been propagated in vitro,
in the absence of the virus-suppressing immune system of the host, and
after it had been treated with mitogens and mutagens such as iododeoxyuridine,
an agent known to activate dormant retroviruses (6). Since the virus was
isolated from a cell line that came from an adult patient with T-cell leukemia,
the virus was proposed to be the cause of adult T-cell leukemia (ATL),
and hence named human T-cell leukemia virus (175, 176). However, a parallel
T-cell line, termed HUT 78, derived from another patient with T-cell leukemia,
failed to yield a retrovirus (87).
Further support for the hypothesis was derived from epidemiological
correlations between antibodies to HTLV-I and ATL in Japan and the U.S.
(3, 37, 176). Based on 30,000 blood donations, the American Red Cross has
reported that in 1986-1987 about 0.025% or 65,000 Americans were infected
with HTLV-I (3, 82), but the American T-cell Leukemia/Lymphoma Registry
had recorded in 1990 in the U.S. no more than 90 ATLs. Among these, 75
were non-Caucasians (177), a group in which HTLV-I is often endemic (178).
However, the same Registry also reports, "although most cases of ATL
are HTLV-I-associated ... many are not" (177). As in the U.S., HTLV-I-free
ATLs have been observed in Japan (1 79). A controlled study comparing the
incidence of the leukemia in HTLV-I-positive and -negative control groups
has never been published.
By definition, "The diagnosis of ATL is made from
the characteristic clinical findings, the detection of serum antibodies
to HTLV-I and, when necessary, the confirmation of monoclonal integration
of HTLV-I proviral DNA in cellular DNA of ATL cells" (180). According
to this tautology, ATL is defined and distinguished from virus-free T-cell
leukemias solely by the presence of antibody to HTLV-I or viral DNA.
In addition, HTLV-I is also postulated to cause an HTLV-I-associated
myelopathy (HAM), which is a neurological disease also defined only by
the presence of HTLV-I (3, 181).
ATL is clonal, originating from a single cell, like virus-free
T-cell leukemias. The clonality of the leukemia is defined by chromosome
abnormalities, as well as by clonal proviral integration sites (2, 176).
However, there are no specific integration sites of HTLV-I in different
leukemias (2). In leukemic cells, the virus is always latent, suppressed
by antiviral immunity, and sometimes even defective (2). It is for this
reason that the virus was originally discovered only in vitro, after
reactivation from latently infected leukemic cells grown in culture.
HTLV-I, like other non-oncogenic retroviruses (6, 54),
is naturally transmitted from mother to child with an efficiency of 22%
based on testing for antiviral antibodies (176, 182, 183). Indeed, latent
proviruses appear to be transmitted perinatally at a higher efficiency
than antibody tests indicate, because the antibody titers increase with
age (176) at a much faster rate than could be accounted for even by thousands
of sexual contacts (183). Thus, this virus, like all other retroviruses
without oncogenes (54), survives from perinatal transmission. Sex is another,
although highly inefficient, mode of transmission, depending on an average
of over 1000 sexual contacts (183).
Based on epidemiological studies from Japan, HTLV-I is
said to cause leukemia in only 1-5% of all virus carriers in a lifetime
(182). The annual incidence of the leukemia per HTLV-I carrier in Japan
is estimated to be only 1 in 1000 (182, 184). Since HTLV-I is a perinatally
transmitted retrovirus, but leukemia typically appears, if at all, only
in 50- to 60-year-olds, the latent period from infection to disease is
estimated at 55 years (176, 185).
The following epidemiological and virological arguments
cast doubt on the HTLV-I-leukemia hypothesis:
1. According to the American Red Cross, "ATL ...
as of September 1989, has not been reported in association with transfusion
transmitted HTLV-I infection," although about 65,000 Americans were
infected with HTLV-I and about 12 million blood donations are annually
transfused to millions of recipients in the U.S. (82). Thus, HTLV-I cannot
be sufficient to cause leukemia.
2. Since viruses, as self-replicating toxins, are all
potentially fast pathogens, but leukemia is only observed about 55 years
after infection, HTLV-I cannot be sufficient for leukemia.
3. Considering that 1% of HTLV-I carriers develop ATL
per lifetime in Japan and about 0.1% (90 : 65,000) in the U.S., that the
leukemias are clonal deriving from single cells, and that each carrier
must contain at least 107 latently infected T-cells (because the limit
of provirus detection by hybridization is 1 in 1000 cells) and that humans
contain 1010 to 1011 T-cells that go through at least 420 generations in
a 70-year lifetime (see Section IV) (37, 186), then only 1 out of 102 (Japan)
to 103 (U .S.) x 107 x 420 = 1012 infected T-cells become leukemic. Thus,
HTLV-I cannot be sufficient for leukemogenesis.
4. Antiviral antibodies that completely neutralize HTLV-I
to virtually undetectable levels (2) do not protect against the leukemia.
This also indicates that HTLV-I is not sufficient for leukemogenesis.
5. Retroviruses cause either polyclonal tumors via dominant,
biochemically active oncogenes (6, 37), or possibly clonal tumors via site-specific
integration that generates active virus-cell hybrid oncogenes (31, 40,
42). Yet HTLV-I neither expresses a leukemia-specific gene product that
could function as an active oncogene, nor does it integrate at a specific
site in different "viral leukemias" (2, 187). Thus, HTLV-I cannot
be sufficient for leukemogenesis.
6. The statement of the American T-cell Leukemia/Lymphoma
Registry that "although most cases of ATL are HTLV-I associated ...
many are not" (177) and the reports of virus-negative leukemias from
Japan (179) and other countries (2) indicate that HTLV-I is not even necessary
for the disease.
7. The HTLV-I-leukemia hypothesis fails to explain the
clonal chromosome abnormalities that are consistently found in all ATLs
(2, 188)-except if one makes the additional odd assumption that HTLV-I
only transforms cells with a preexisting chromosome abnormality.
Thus, there are no virus-determined diagnostic criteria,
besides the presence of antiviral antibodies, nor are there any controlled
epidemiological and virological criteria to support the hypothesis that
HTLV-I is the cause of ATL. Therefore, ad hoc hypotheses have been
advanced proposing "a second oncogenic event, such as a chance translocation
or a second oncogenic virus ..." for viral leukemogenesis (187). Others
estimate five steps in leukemogenesis, of which HTLV-I is postulated to
be an "initiator" (185).
Since not even one transfusion-transmitted leukemia case
has been recorded in the U.S., it seems surprising that a blood test for
antibodies against HTLV-I has become mandatory for members of the American
Association of Blood Banks since February 1989. It raises the cost of each
of the approximately 12 million annual blood donations in the U.S. (82)
by $5-11 (189; Irwin Memorial Blood Bank, personal communication, 1990).
Indeed, an HTLV-I epidemiologist pointed out, "Ironically, this route
of [HTLV-I] transmission is numerically the least important," considering
the 55-year average latent period from infection to leukemia, "the
advanced age of most U.S. blood recipients, and the observation that as
many as 60% of transfusion recipients may die within approximately 3 years
of transfusion because of their underlying disease" (183). Nevertheless,
in terms of blood testing expenses, HTLV-I has reached cost-parity with
HIV, which adds another $11 test fee to each blood donation (Irwin Memorial
Blood Bank, personal communication, 1990).
An alternative hypothesis suggests that spontaneous or
perhaps radiation-induced chromosome abnormalities induce the clonal leukemias
(see Section VI). Nuclear radiation from the Hiroshima and Nagasaki bombs
is blamed for 147 leukemias (190). By proposing that one out of billions
of normal HTLV-I-infected cells is transformed by a spontaneous chromosome
abnormality, our hypothesis readily resolves the paradox of the clonal
chromosome abnormalities in all "viral" leukemias.
B. Herpes Virus,
Papilloma Viruses, and Cervical Cancer
Inspired by the SV40/adenovirus-cancer models, infection
by herpes simplex virus (HSV) was postulated in the 1970s to be the cause
of cervical cancer based on epidemiological correlations with HSV DNA (3).
The virus is sexually transmitted and is latent in about 85% of the adult
population of the U.S. (3). Infection by intact HSV typically kills the
cell. However, defective and intact viruses that become latent do not kill
The viral DNAs in cervical cancers are defective and integrated
with cell DNA. Cervical cancers with defective HSV DNA are clonal, just
like virus-free cancers (191-194). In agreement with the SV40/ adenovirus
models, HSV does not replicate in the tumors. But, unlike the SV40/adenovirus
models, no set of viral genes is consistently present or expressed in human
cervical cancers. Therefore, the "hit-and-run" mechanism of viral
carcinogenesis was proposed (195). It holds that neither the complete HSV,
nor even a part of it, needs to be present in the tumor. Obviously, this
is an unfalsifiable, but also an unprovable, hypothesis.
Also inspired by the SV40/adenovirus models, and based
on epidemiological correlations, infection by human papilloma virus (HPV)
was postulated in the 1980s by zur Hausen to be a causative factor in cervical
and anogenital cancers (3, 191, 196).
Papilloma viruses are transmitted by sexual and other
contacts, like the herpes viruses, and are widespread or "ubiquitous"
in at least 50% of the adult population of the U.S. and Europe (3, 191).
For example, using the PCR to amplify sequences of one particular strain
of papilloma virus, 46% of 467 women in Berkeley, California, with a median
age of 22 were found to carry HPV, but none of them had cervical cancer
(199). Many other strains of HPV exist (3, 191) that could not be detected
with the assay used in this study (199). Like the SV40/ adenovirus models,
HPV does not replicate in the tumors. But, unlike these models, HPV naturally
replicates nonlytically (13), forming polyclonal warts with unintegrated
viral DNA plasmids (200).
zur Hausen reports that cervical cancers occur in less
than 3% of infected women in their lifetime, but the incidence in HPV-free
controls was not reported (191). In the U.S., the incidence of cervical
cancer in all women, with and without HPV, per 70-year lifetime is about
1% (197). In a controlled study of age-matched women, 67% of those with
cervical cancer and 43% of those without were found to be HPV-positive
(198). These cancers are observed on average only 20-50 years after infection
Different sets and amounts of viral DNA are integrated
into cell DNA of different carcinomas (191), and viral DNA is poorly expressed
in some cancers and not expressed at all in others (3, 191, 201). Moreover,
different HPV strains are found in different cancers (3, 191, 196). Viral
antigens are found in only 1-5% of carcinomas (3). Accordingly, HPV does
not replicate in the cancer cells and there are no reports of HPV-specific
histological or physiological markers that set HPV DNA-positive apart from
negative carcinomas (191). There is also no virus-specific integration
site in HPV DNA-positive cancers (191), indicating that no specific cellular
gene is activated, or that a tumor suppressor gene is inactivated by integration
of viral DNA. HPV DNA-positive tumors are clonal and carry clonal chromosome
abnormalities, just like virus-negative tumors (191-194).
The HPV-cancer hypothesis of zur Hausen proposes that
HPV encodes a "transforming factor" that is suppressed in normal
cells by a cellular interference factor (CIF). Inactivation of both CIF
alleles by mutation is postulated to result in viral carcinogenesis (191).
The low probability of developing mutations in both suppressor alleles
is said to explain the long intervals between infection and cancer. This
hypothesis correctly predicts that only a small fraction of infections
lead to cancer. It further predicts clonal tumors with active HPV DNA and
mutations in both alleles of the suppressor genes, and it predicts no effects
on the karyotypes of cells.
Howley et al. proposed that a viral protein neutralizes
the proteins of the retinoblastoma and p53 tumor suppressor genes, and
that neutralization of these suppressor proteins causes cancer (202). The
proposal is modeled after the hypothesis that retinoblastoma is caused
by a cellular cancer gene, provided that a complementary suppressor gene,
termed the retinoblastoma gene, is inactivated (see Section IV). This hypothesis
predicts polyclonal tumors.
The following epidemiological and biochemical arguments
cast doubt on these HPV-cancer hypotheses:
1. Random allelic mutation of suppressor genes, as postulated
by zur Hausen, predicts a few cancers soon, and more long after infection.
Since cancers only appear 20-50 years after infection, cooperation between
HPV and mutations cannot be sufficient for carcinogenesis.
2. Further, the proposal of zur Hausen that inactivation
of host suppressor genes is necessary for viral transformation is not compatible
with HPV survival. Since HPV, like all small DNA viruses, needs all of
its 8-kb DNA for virus replication (13), suppression of one or more HPV
proteins by normal cellular genes would effectively inhibit virus replication
in all normal cells. Conversely, if viral transforming proteins were not
suppressed by normal cells, virus-replicating wart cells should be tumorigenic
because all viral genes are highly expressed in virus replication (1, 13,
3. The clonality of cervical cancers rules out the Howley
4. The lack of a consistent HPV DNA sequence and of consistent
HPV gene expression in HPV DNA-positive tumors is inconsistent with the
zur Hausen and Howley hypotheses and indicates that HPV is not necessary
to maintain cervical cancer.
5. The presence of HPV in no more than 67% of age-matched
women with cervical cancer (198) also indicates that HPV is not necessary
for cervical cancer.
6. The hypothesis also fails to explain the presence of
clonal chromosome abnormalities consistently seen in cervical cancer (16,
192-194)-except if one makes the additional odd assumption that only cells
with preexisting chromosome abnormalities are transformed by HPV.
It follows that neither HPV nor HSV plays a direct role
in cervical carcinomagenesis. Moreover, the HPV-cancer hypothesis offers
no explanation for the absence of a reciprocal venereal male carcinoma.
Thus, detecting inactive and defective viral DNA from
past infections in non-tumorigenic cells with a commercial hybridization
test (Vira/Pap, Digene Diagnostics, Silver Spring, Maryland) or with the
PCR (199) seems worthless as a predictor of rare carcinomas appearing decades
later, in view of the "ubiquity" (191) of these viruses in women
and the total lack of evidence that cervical cancer occurs in women with
HPV more often than in those without. This test, at $30-150, is currently
recommended for the 7 million Pap smears that appear "atypical"
in the U.S. per year (Digene Diagnostics, personal communication, 1991).
By contrast only 13,000 cervical cancers are observed annually in both
HPV-positive and -negative women in the U.S. (197). Indeed, the test may
be harmful, considering the anxiety a positive result induces in believers
of the virus-cancer hypothesis.
An alternative cervical carcinoma hypothesis suggests
that rare spontaneous or chemically induced chromosome abnormalities, which
are consistently observed in both HPV and HSV DNA-negative and -positive
cervical cancers (192-194), induce cervical cancer. For example, smoking
has been identified as a cervical cancer risk (204). The controlled study
of age-matched women described above suggests that 52% of the women with
cervical cancer were smokers compared to only 27% of those without (198).
Indeed, carcinogens may be primary inducers of abnormal cell proliferation
rather than HPV or HSV. Since proliferating cells would be more susceptible
to infection than resting cells, the viruses would be just indicators,
rather than causes of abnormal proliferation. Activation of latent retroviruses
like HTLV-I (Section III,A) (2), herpes viruses (12), and lambda phages
(205) by chemical or radiation-induced cell damage and subsequent proliferation
are classical examples of such indicators. Indeed, Rous first demonstrated
that the virus indicates hydrocarbon-induced papillomas; it "... localized
in these and urged them on ..." and suggested that enhanced proliferation
is a risk factor for carcinogenesis (203).
According to this hypothesis, HPV or HSV DNAs in tumor
cells reflect defective and latent viral genomes accidentally integrated
into normal or hyperplastic cells, from which the tumor is derived. This
hypothesis readily reconciles the clonal chromosome abnormalities with
the clonal viral DNA insertions of the "viral" carcinomas. The
inactive and defective viral DNA in the carcinomas would be a fossil record
of a prior infection that was irrelevant to carcinogenesis.
C. Hepatitis B Virus and
Epidemiological evidence indicates that chronic hepatitis
B virus (HBV) carriers in Asia have a 250-fold higher risk of developing
hepatomas than do non-carriers (3, 12, 206-208). The virus is typically
transmitted perinatally in Asia and Africa (3, 207). In over 95% of infections
in Asia and 99.9% in the U.S. and Europe the virus is completely neutralized
by antiviral immunity. In people with drug- or disease-induced immunodeficiencies
the virus remains chronically active (12). Approximately 1 out of 70 chronic
HBV carriers in Asia develop clonal hepatomas and 1 out of 300 develop
liver cirrhosis in their lifetime (3, 207). However, the liver tumors appear
only in 30- to 60-year-olds. Moreover, chronic HBV carriers in Asia are
"more likely" to develop hepatomas than those in Europe and the
U.S. (12). Inoculation of HBV into chimpanzees has failed to cause hepatomas
The virus is thought not to kill infected cells and viral
DNA is replicated as a plasmid and thus not typically integrated into the
host DNA (3, 12). However, molecular studies have detected clonal inserts
of HBV DNA randomly integrated into the cellular DNA of liver carcinoma
tissues (196, 209). Viral DNA is defective and not replicated in HBV DNA-positive
hepatomas (209), like SV40 and adenovirus DNAs in the corresponding viral
tumors. By contrast to the SV40/adenovirus models, no subset of viral DNA
is consistently found or expressed in HBV-positive tumors (209, 210). Only
11-19% of tumors in HBVpositive patients express some viral antigens, compared
to 26-61% expressing them in surrounding non-tumorous tissues (211). In
addition to clonal inserts of HBV DNA, the hepatomas carry clonal chromosome
abnormalities (16, 193, 196).
On the basis of these data, it has been proposed that
HBV causes liver carcinoma in a step-wise process that begins with antigenemia,
followed by chronic hepatitis, cirrhosis, and cancer (3, 207, 209). However,
cirrhosis is not a necessary precursor of a hepatoma (3).
The following epidemiological and biochemical arguments cast doubt on
the HBV-hepatoma hypothesis:
1. The long intervals of 30-60 years between infection
and hepatomas indicate that HBV is not sufficient to initiate carcinogenesis.
2. The evidence that HBV is naturally transmitted perinatally
also indicates that the virus is not sufficient to cause fatal diseases
such as cirrhosis and hepatomas, because the viruses that depend on perinatal
transmission for survival are not inherently pathogenic.
3. The evidence that the hepatoma risk among chronic HBV
carriers in Asia is higher than in the U.S. and Europe also indicates that
HBV is not sufficient for carcinogenesis.
4. The clonality of the HBV-positive hepatomas further
indicates that HBV is not sufficient for carcinogenesis, because only one
out of billions of chronically infected liver cells becomes tumorigenic.
5. The absence of an HBV-specific tumor marker, and of
a specific HBV DNA sequence or integration site in viral hepatomas, both
indicate that HBV is not necessary to maintain hepatomas.
6. The HBV-hepatoma hypothesis fails to explain the clonal
chromosome abnormalities of hepatomas-except if one makes the additional
odd assumption that HBV only transforms cells with preexisting chromosome
Thus, there is no convincing evidence that HBV DNA is
functionally relevant for the initiation and maintenance of hepatomas.
Its presence in the tumor could merely reflect that the tumor had originated
from one of probably many liver cells of HBV carriers that contain defective,
biochemically inactive viral DNA integrated randomly into their chromosomes
(196). Therefore, molecular analysis of HBV DNA and of HBV DNA integration
sites (210) is not likely to illuminate carcinogenesis.
However, chronically replicating HBV may function as an
indirect carcinogen in the form of a long-term source of intoxication,
inducing necrosis and tissue regeneration, a known risk factor for carcinogenesis
(1, 196, 203). This view is consistent with the higher-than-normal incidence
of hepatomas in persons with chronic HBV infection.
A competing hypothesis suggests that chronic HBV infection
may only be an indicator of a chronic nonviral intoxication and immunodeficiency.
Indeed, nonviral factors are involved in hepatomagenesis because the incidence
of the hepatomas per HBV carrier varies with different countries (12).
Intoxication could induce tissue regeneration and immune suppression, a
classical precondition for opportunistic virus infections (see HPV in Section
III, B). According to this view, the hepatoma would be caused by a rare
virus-independent mechanism that generates chromosome abnormalities in
one of many normal cells with HBV DNA inserts. This hypothesis would readily
resolve the presence of the clonal chromosome abnormalities in all "viral"
hepatomas. The defective and inactive viral DNAs in the hepatomas would
be a fossil record of a prior infection that was irrelevant to carcinogenesis.
D. Epstein-Barr Virus and
In the early 1960s, Burkitt suggested that a B-cell tumor,
now called Burkitt's lymphoma, which occurs in 1 out of 10,000 Central
African children per year between 4 and 14 years of age, was caused by
a virus (3, 12). Although not detectable in biopsies of lymphoma patients,
a virus was found with the electron microscope in lymphoma cells grown
in culture away from the suppressive immune system of the host (212). The
Epstein-Barr virus (EBV) has since been postulated to be the cause of Burkitt's
lymphoma (3, 8, 12).
In Central Africa, infection with the virus occurs perinatally
in the first months of life in almost 100% of the population (3, 207).
In the U.S. and Europe, infection occurs typically during or after puberty
in about 50% of healthy adults (3, 213). However, the incidence of lymphomas
with EBV in these countries is only less than 1 in 106 per year (3). Moreover,
only 30% of otherwise indistinguishable lymphomas express EBV antigens
(3). In America, Burkitt's lymphomas free of EBV DNA were described in
1973 (214). In China, EBV is also said to cause nasopharyngeal carcinoma
in adults (1, 3).
During a primary infection, the virus may induce transient,
polyclonal lymphoproliferative diseases, such as mononucleosis, if a large
percentage of lymphocytes are infected prior to immunity. After antiviral
immunity is established, the virus remains chronically associated with
the host in a latent form (3, 12). During the chronic state of infection,
viral DNA is detectable with the PCR in about 1 out of 105 lymphocytes
(213) and viral antigens in only about 1 out of 107 lymphocytes (12).
In lymphomas, the virus is also suppressed, producing
but a few viral antigens (3), as the history of its discovery had first
indicated. Burkitt's lymphomas are clonal, deriving from single cells that
carry characteristic chromosome translocations that often rearrange the
proto-myc gene (see Section IV). Since EBV, like other herpes viruses,
generally does not integrate into the host chromosome (1, 3), the time
of infection of tumor cells (e.g., whether infection occured before, during,
or after tumorigenesis) cannot be determined.
The EBV-lymphoma hypothesis suffers from numerous epidemiological
and biochemical inconsistencies:
1. The clonality of the lymphomas that emerge from a single
tumorigenic cell among billions of non-tumorigenic EBV-infected cells indicates
that EBV is not sufficient for tumorigenesis.
2. The long intervals between infection and carcinogenesis, averaging
10 years in Africa, and the incidence of only 1 lymphoma per 10,000 infected
persons also indicate that EBV is not sufficient to initiate tumorigenesis.
3. The lymphoma incidence varies over 100-fold between African and European
or American EBV carriers, also indicating that EBV cannot be sufficient
to cause a lymphoma.
4. The lack of a lymphoma-specific EBV function in symptomatic carriers
indicates that EBV is not necessary to maintain lymphomas.
5. The existence of EBV-free Burkitt's lymphomas in American and European
patients indicates most directly that EBV is not even necessary for Burkitt's
Thus, EBV appears neither necessary nor sufficient for
lymphomagenesis. Nevertheless, it has been argued that EBV plays at least
an indirect role in lymphomagenesis, because only a minority of susceptible
cells from EBV-positive patients are infected in vivo, but virtually
all lymphoma cell lines in culture are infected by the virus (215, 216).
However, this could be an artifact of studying cells in culture, because
the virus would spread unimpaired by immunity from a few infected, normal,
or lymphoma cells to all lymphoma cells that survive in culture.
Since about 100% of the Central African and 30-50% of
the American population carries latent EBV, and since EBV-negative Burkitt's
lymphomas exist, it is likely that the correlations between EBV and tumors
are accidental rather than causal. In view of this, an alternative hypothesis
has been advanced, which holds that altered cellular proto-myc genes
are the cause of Burkitt's lymphoma (see Section IV).