| |
Retroviruses
as Carcinogens and Pathogens:
Expectations and Reality
By Peter H. Duesberg
Cancer Research, Vol. 47, pp. 1199-1220,
(Perspectives in Cancer Research), March 1, 1987.
Abstract
Retroviruses (without transforming genes) are thought to cause
leukemias and other cancers in animals and humans because they were
originally isolated from those diseases and because experimental
infections of newborns may induce leukemias with probabilities of
0 to 90%. According to this hypothesis viral cancers should be contagious,
polyclonal, and preventable by immunization. However, retroviruses
are rather widespread in healthy animals and humans where they typically
cause latent infections and antiviral immunity. The leukemia risk
of such infections is less than 0.1% and thus about as low as that
of virus-free controls. Indeed retroviruses are not sufficient to
initiate transformation (a) because of the low percentage
of symptomatic virus carriers and the complete lack of transforming
function in vitro; (b) because of the striking discrepancies
between the long latent periods of 0.5 to 10 years for carcinogenesis
and the short eclipse of days to weeks for virus replication and
direct pathogenic and immunogenic effects; (c) because there
is no gene with a late transforming function, since all genes are
essential for replication; (d) because host genes, which
do not inhibit virus, inhibit tumorigenesis up to 100% if intact
and determine the nature of the tumor if defective; and above all
(e) because of the monoclonal origin of viral leukemias,
defined by viral integration sites that are different in each tumor.
On these bases the probability that a virus-infected cell will become
transformed is estimated to be about 10-11. The viruses
are also not necessary to maintain transformation, since many animal
and all bovine and human tumors do not express viral antigens or
RNA or contain only incomplete proviruses. Thus as carcinogens retroviruses
do not necessarily fulfill Koch's first postulate and do not
or only very rarely (10-11) fulfill the third. Therefore
it has been proposed that retroviruses transform inefficiently by
activating latent cellular oncogenes by, for example, provirus integration.
This predicts diploid tumors with great diversity, because integration
sites are different in each tumor. However, the uniformity of different
viral and even nonviral tumors of the same lineage, their common
susceptibility to the same tumor resistance genes, and transformation-specific
chromosome abnormalities shared with nonviral tumors each argue
for cellular transforming genes. Indeed clonal chromosome abnormalities
are the only known transformation-specific determinants of viral
tumors. Since tumors originate with these abnormalities, these or
associated events, rather than preexisting viruses, must initiate
transformation. Therefore it is proposed that transformation is
a virus-independent event and that clonal viral integration sites
are consequences of clonal proliferation of transformed cells. The
role of the virus in carcinogenesis is limited to the induction
of hyperplasia which is necessary but not sufficient for carcinogenesis.
Hyperplasia depends on chronic viremia or high virus expression
which are very rare in animals outside the laboratory and have never
been observed in humans. Since latent viruses, which are typical
of nearly all natural infections, are neither direct nor indirect
carcinogens, they are not targets for cancer prevention. Viruses
are also not targets for cancer therapy, since tumors are not maintained
and not directly initiated by viral genes and occur naturally despite
active antiviral immunity.
Lymphotropic retrovirus has been proposed to cause AIDS because
90% of the patients have antibody to the virus. Therefore antibody
to the virus is used to diagnose AIDS and those at risk for AIDS.
The virus has also been suggested as a cause of diseases of the
lung and the nervous system. Promiscuous male homosexuals and recipients
of frequent transfusions are at a high risk for infection and also
at a relatively high annual risk for AIDS, which averages 0.3% and
may reach 5%. Others are at a low risk for infection and if infected
are at no risk for AIDS. AIDS viruses are thought to kill T-cells,
although these viruses depend on mitosis for replication and do
not lyse cells in asymptomatic infections. Indeed the virus is not
sufficient to cause AIDS (a) because the percentage
of symptomatic carriers is low and varies between 0 and 5% with
the risk group of the carrier, suggesting a cofactor or another
cause; (b) because the latent period for AIDS is 5 years
compared to an eclipse of only days to weeks for replication and
direct pathogenic and immunogenic effects; and (c) because
there is no gene with a late AIDS function, since all viral genes
are essential for replication. Moreover the extremely low levels
of virus expression and infiltration cast doubt on whether the virus
is even necessary to cause AIDS or any of the other diseases with
which it is associated. Typically, proviral DNA is detectable in
only 15% of AIDS patients and then only in 1 of 102 to 103 lymphocytes
and is expressed in only 1 of 104 to 105 lymphocytes. Thus the virus
is inactive or latent in carriers with and without AIDS. It is for
this reason that it is not transmitted as a cell-free agent. By
contrast, all other viruses are expressed at high titers when they
function as pathogens. Therefore AIDS virus could be just the most
common occupational infection of those at risk for AIDS because
retroviruses are not cytocidal and unlike most viruses persist as
latent, nonpathogenic infections. As such the virus is an indicator
of sera that may cause AIDS. Vaccination is not likely to benefit
virus carriers, because nearly all have active antiviral immunity.
Introduction
How often have I said to you, that when
you have eliminated the impossible, whatever remains however improbable
must be the truth.
-Sherlock Holmes
The irreversible and predictable courses of most cancers indicate
that cancer has a genetic basis. In 1914 Boveri (1) proposed that
cancer is caused by chromosomal mutations. This hypothesis has since
received ample support (2-4), although a cellular cancer gene has
yet to be identified (5). In the light of the spectacular discovery
of RSV2 in 1911, which proved to be a direct, infectious carcinogen,
the hypothesis emerged that viruses may be a significant source of
exogenous cancer genes (6). The virus-cancer hypothesis has since
steadily gained support because retroviruses and DNA viruses were
frequently isolated from animal leukemias and other tumors, and occasionally
from human leukemias, in efforts to identify causative agents (7-16).
However, once discovered in tumors and named tumor viruses, most of
these viruses were subsequently found to be widespread in healthy
animals and humans (8, 12-18). Thus these viruses are compatible with
the first but apparently not necessarily with the third of Koch's
postulates3 as viral carcinogens. Only a few of the many tumor viruses
are indeed directly oncogenic, such as RSV and about 20 other types
of retroviruses (5, 13, 19, 20), and hence compatible with Koch's
third postulate. Therefore, if we want to assess the role of viruses
in cancer, there must be a clear separation between those viruses
which are directly oncogenic and those which are not.
The directly oncogenic retroviruses owe their transforming
function to a particular class of genes which are termed oncgenes
(20). These are as yet the only known autonomous cancer genes that
can transform diploid cells in vitro as well as in animals
susceptible to the particular virus (5). Since susceptible cells are
inevitably transformed as soon as they are infected, the resulting
tumors are polyclonal (13, 16). Nevertheless, directly oncogenic retroviruses
have never caused epidemics of cancer. The probable reason is that
oncgenes are not essential for survival of the virus and hence
are readily lost by spontaneous deletion or mutation (5). Indeed,
oncgenes were originally discovered by the analysis of spontaneous
onc deletion mutants of RSV (21). Moreover, because oncgenes
typically replace essential genes (except in some strains of RSV)
these viruses cannot replicate unless aided by a nondefective helper
virus (5, 13).
The vast majority of the tumor viruses are retroviruses
and DNA viruses that do not contain oncgenes. The RNA genomes
of all retroviruses without oncgenes measure only 8 to 9 kilobases
(13, 22). They all encode three major essential genes which virtually
exhaust their coding capacity. These are in the 5' to 3' map
order gag which encodes the viral core protein, pol
which encodes the reverse transcriptase, and env which encodes
the envelope glycoprotein (23, 24). Although these viruses lack oncgenes
they are considered tumor viruses, because they were originally isolated
from tumors and because experimental infections may induce tumors
under certain conditions. However, in contrast to tumors caused by
viruses with oncgenes, such tumors are always monoclonal and
induced reproducibly only in genetically selected animals inoculated
as newborns after latent periods of over 6 months (see below). Because
of the long latent periods, these retroviruses are said to be "slow"
viruses (13, 16), although their mechanism of replication is exactly
the same as that of their fast and efficient relatives with oncgenes
that transform cells as soon as they infect them (5, 19) (Table 1).
The retroviruses are also considered to be plausible natural carcinogens
because they are not cytocidal and hence compatible with neoplastic
growth and other slow diseases. Indeed, retroviruses are the only
viruses that depend on mitosis for replication (13, 25).
However, the retroviruses without oncgenes
are also the most common and benign passenger viruses of healthy animals
and humans probably because of their unique noncytocidal mechanism
of replication and their characteristic ability to coexist with their
hosts without causing any pathogenic symptoms either as latent infections,
which make no biochemical demands, or even as productive infections.
Based on the permissiveness of a host for expression and reproduction,
they have been divided into exogenous viruses which are typically
expressed and hence potentially pathogenic and endogenous viruses
which are typically latent and hence nonpathogenic (16-18). Because
they are so readily suppressed in response to as yet undefined cellular
suppressors (8, 11, 12, 16-18), endogenous viruses are integrated
as proviruses into the germ line of most if not all vertebrates (8,
13, 16-18). Nevertheless, the endogenous and exogenous retroviruses
are entirely isogenic and there is no absolute biochemical or functional
distinction between them except for their response to suppressors
of a particular host (13, 16-18) (Part I, Section A). Therefore the
association of these viruses with a given disease is not sufficient
even to suggest a causative role in it. Indeed there is as yet no
direct evidence that retroviruses play a role as natural carcinogens
of wild animals and humans. Thus the critical expectations of the
virus-cancer hypothesis, namely that RNA or DNA tumor viruses would
be direct carcinogens, that viral tumors would be polyclonal because
each virus-infected cell would be transformed, and above all that
viral carcinogenesis would be preventable by immunization, remain
largely unconfirmed.
Recently retroviruses without oncgenes have
been isolated from patients with AIDS and those at risk for AIDS and
have since been considered the cause of AIDS (26). In contrast to
other retroviruses, the AIDS viruses are thought to act as direct,
cytocidal pathogens that kill susceptible T-cells (13, 27).
Here we discuss how the retroviruses without oncgenes
fit the role of viral carcinogens or AIDS pathogens and whether these
viruses are indeed the vessels of evil they have been labeled to be.
Above all we hope to identify transformation-specific or AIDS-specific
viral and cellular determinants and functions. Since the genetic repertoire
of all retroviruses without oncgenes, including that of the
AIDS viruses (28), is exhausted by genes that are essential for virus
replication (13, 24), a hypothetical oncogenic or AIDS function would
have to be indirect or it would have to be encoded by one of the essential
genes. In the second case the virus would be oncogenic or cause AIDS
whenever it replicates. A survey of the best studied animal and human
retroviruses demonstrates that these viruses are not sufficient to
cause tumors and not necessary to maintain them. Most likely these
viruses play a role in inducing tumors indirectly. Indeed transformation
appears to be a virus-independent, cellular event for which chromosome
abnormalities are the only specific markers. Likewise the AIDS viruses
are shown not to be sufficient to cause AIDS, and the evidence that
they are necessary to cause it is debated.
I. Retroviruses and Cancer
A. Retroviruses Are Not Sufficient for Transformation Because Less
Than 0.1% of Infected Animals or Humans Develop Tumors
Avian lymphomatosis virus was originally isolated from leukemic chickens
(29). However, subsequent studies proved that latent infection by
avian lymphomatosis viruses occurs in all chicken flocks and that
by sexual maturity most birds are infected (30-32). Statistics report
an annual incidence of 2 to 3% lymphomatoses in some inbred flocks.
Yet these statistics include the more common lymphomas caused by Marek's
virus (a herpes virus) (33, 34). The apparent paradox that the same
virus is present in most normal and healthy animals (30) but may be
leukemogenic in certain conditions was resolved at least in descriptive
terms by experimental and congenital contact infections. Typically
experimental or contact infection of newborn animals that are not
protected by maternal antibody would induce chronic (31, 32) or temporal
(35, 36) viremia. The probability of such animals for subsequent lymphomatosis
ranges from 0 to 90% depending on tumor resistance genes (Section
C). However, infection of immunocompetent adults or of newborn animals
protected by maternal antibody and later by active immunity would
induce latent, persistent infections with a very low risk of less
than 1% for lymphomatosis (32).4 Thus only viremic animals are likely
to develop leukemia at a predictable risk.
Viremia has a fast proliferative effect on hemopoietic
cells and generates lymphoblast hyperplasia (Fig. 1) (32, 36, 37).
Hyperplasia appears to be necessary but not sufficient for later leukemogenesis
because it does not lead to leukemia in tumor-resistant birds (36)
(Section C) and because removal of the bursa of Fabricius, the major
site of lymphoproliferation, prevents development of the disease (9,
32).
The murine leukemia viruses were also originally isolated
from leukemic inbred mice (9) and subsequently detected as latent
infections in most healthy mice (8, 13, 16, 17, 38). Indeed, about
0.5% of the DNA of a normal mouse is estimated to be proviral DNA
of endogenous retroviruses, corresponding to 500 proviral equivalents
per cell (18). Nevertheless leukemia in feral mice is apparently very
rare. For instance low virus expression, but not a single leukemia,
was recorded in 20% of wild mice (38) probably because wild mice restrict
virus expression and thus never become viremic and leukemic. However,
in an inbred stock of feral mice predisposed to lymphoma and paralysis,
90% were viremic from an early age, of which 5% developed lymphomas
at about 18 months (39).
Experimental infections of newborn, inbred mice with
appropriate strains of murine leukemia viruses induce chronic viremias.
Such viremic mice develop leukemias with probabilities of 0 to 90%
depending on the mouse strain (Section C). However, if mice that are
susceptible to leukemogenesis are infected by the time they are immunocompetent
or are protected by maternal antibodies if infected as neonates, no
chronic viremia and essentially no leukemia are observed (although
a latent infection is established) (41). Thus leukemogenesis depends
on viremia (40) as with the avian system. However, viremia is not
sufficient, because certain tumor-resistant strains do not develop
leukemia even in the presence of viremia (42) (Section C). Again viremia
has an early proliferative effect on lymphocytes which has been exploited
to quantitate these viruses in vivo within 2 weeks by the "spleen-weight"
or "spleen-colony" assay (18, 43-47). This hyperplasia of lymphocytes
is necessary for leukemogenesis, because the risk that an infected
animal will develop leukemia is drastically reduced or eliminated
by thymectomy, which is a major source of cells for prospective leukemogenesis
(9).
The AKR mouse is a special example in which spontaneous
expression of endogenous virus and the absence of tumor resistance
genes inevitably lead to viremia at a few weeks after birth and, in
90% of the animals, to leukemia at 6 to 12 months of age (9, 41, 48).
This also shows that endogenous viruses can be just as pathogenic
or leukemogenic as exogenous viruses if they are expressed at a high
level. Likewise, endogenous avian retroviruses are leukemogemic in
chickens permissive for acute infection (49, 50).
The evidence that mammary carcinomas are transmissible
by a milk-borne virus, MMTV, indicates that the virus is an etiological
factor (51, 52). However, the same virus is also endogenous but not
expressed in most healthy mice (16, 53). Since no mammary tumors have
been reported in wild mice the natural incidence must be very low,
but in mice bred for high incidence of mammary carcinomas it may rise
to 90% (13, 16, 54, 55). As with the leukemia viruses, the risk for
tumorigenesis was shown to depend on a high level of virus expression
from an early age and on the development of hyperplasias that are
necessary but not sufficient for carcinogenesis (56, 57). For example,
BALB/c mice that express over 100 µg virus per ml milk all develop
tumors after latencies of over 12 months, but mice that express 3
µg or less virus per ml develop no tumors at all (54, 58).
Feline leukemia virus was originally isolated from
cats with lymphosarcoma (59) and subsequently from many healthy cats.
It is estimated that at least 50 to 60% of all cats become naturally
infected by feline leukemia viruses at some time during their lives
(60, 61). However, only about 0.04% of all cats develop leukemia on
an annual basis (62), which is thought to be caused by these viruses
(13, 61, 63). Most natural infections cause transient virus expression
which is followed by an immune response, after which little virus
is expressed (60, 64, 65). Such infections do not induce leukemias
at a predictable rate (61). However, 1 to 2% of the naturally infected
cats become chronically viremic (66). About 28% of the viremic cats
develop leukemias after latent periods of 2 years. Thus viremia indicates
a high risk for the development of leukemia (66). Viremia may result
from a congenital infection in the absence of maternal antibody or
from a native immunodeficiency. As in the avian and murine systems,
experimental infection of newborn, immunotolerant cats produces early
viremia and runting diseases and late leukemias at a much higher incidence
than natural infections (63, 64, 67, 68). The gibbon ape leukemia
virus was also initially discovered in leukemic apes and was later
isolated from healthy gibbons (13, 69). Again, only chronically viremic
gibbons were shown to be at risk for leukemia (70).
The bovine and human retroviruses associated with
acute leukemias are always biochemicaIly inactive or latent (Section
D). Viremia, which is frequently associated with a leukemia of congenitally
or experimentally infected domestic chickens, cats, or inbred mice,
has never been observed in the bovine or human system. Accordingly
bovine and human leukemia viruses could be isolated from certain leukemic
cells only after cultivation in vitro away from the suppressive
immune system of the host (71, 72). In regions of endemic bovine leukemia
virus infection 60 to 100% of all animals in a herd were found to
contain antiviral antibody (73, 74). However, the incidence of leukemia
was reported to range only from 0.01 to 0.4% (16, 73). Experimental
infections with cell-free virus have not provided conclusive evidence
for viral leukemogenesis. As yet only 1 of 25 animals infected with
bovine leukemia virus has developed a leukemia 7 years after inoculation
(73). Additional inoculations of 20 newborn calves did not cause a
single leukemia within 5 years, although all animals developed antiviral
antibody.5 However, 50% of newborn sheep inoculated with bovine leukemia
virus developed leukemia about 4 years later (75). These sheep were
probably more susceptible to the bovine virus than cattle, because
they would lack maternal antibody to the virus. Indeed they could
have been transiently viremic, because antibody was detected only
4 months after inoculation (75).
HTLV-I or ATLV was originally isolated from a human
cell line derived from a patient with T-cell leukemia (71). It replicates
in T-cells (27) and also in endothelial cells (76) or fibroblasts
(77). The virus was subsequently shown, using antiviral antibody for
detection, to be endemic as latent, asymptomatic infections in Japan
and the Caribbean (27). Since virus expression is undetectably low
not only in healthy but also in leukemic virus carriers, infections
must be diagnosed indirectly by antiviral antibody or biochemically
by searching for latent proviral DNA (Section D). Due to the complete
and consistent latency, the virus can be isolated from infected cells
only after activation in vitro when it is no longer controlled
by the host's antiviral immunity and suppressors. Therefore the
virus is not naturally transmitted as a cell-free agent like other
pathogenic viruses, but only congenitally, sexually, or by blood transfusion,
that is, by contacts that involve exchange of infected cells (13,
27).
It is often pointed out that functional evidence for
the virus cancer hypothesis is difficult to obtain in humans because
experimental infection is not possible and thus Koch's third postulate
cannot be tested. However, this argument does not apply here since
naturally and chronically infected, asymptomatic human carriers are
abundant. Yet most infections never lead to leukemias and none have
been observed to cause viremias. Moreover, not a single adult T-cell
leukemia was observed in recipients of blood transfusions from virus-positive
donors (13, 78, 79), although recipients developed antiviral antibody
(81).
The incidence of adult T-cell leukemia among Japanese
with antiviral immunity is estimated to be only 0.06% based on 339
cases of T-cell leukemia among 600,000 antibody-positive subjects
(78). Other studies have detected antiviral antibody in healthy Swedish
blood donors (268) and in 3.4% of 1.2 x 106 healthy Japanese blood
donors (79). Further, it was reported that 0.9% of the people of Taiwan
are antibody positive, but the incidence of the leukemia was not mentioned
(80).
In conclusion, the tumor risk of the statistically
most relevant group of retrovirus infections, namely the latent natural
infections with antiviral immunity, is very low. It averages less
than 0.1% in different species, as it is less than 1% in domestic
chickens, undetectably low in wild mice, 0.04% in domestic cats on
an annual basis, 0.01 to 0.4% in cattle, and 0.06% in humans. Thus
the virus is not sufficient to cause cancer.
Moreover, since the viruses associated with all human
tumors and most natural tumors of animals are latent and frequently
defective (Section D), it is difficult to justify the claims that
these viruses play any causative role in tumorigenesis. Indeed nearly
all healthy chickens, mice, cats, cattle, and humans carry endogenous
and exogenous retroviruses that are latent and hence neither pathogenic
nor oncogenic (12, 16-18, 78, 82). Latent infections by cytocidal
viruses, such as herpes viruses, are likewise all asymptomatic (83).
Nevertheless it may be argued that only a small percentage
of retroviral infections are expected to be oncogenic because only
a small percentage of all other viral or microbial infections are
pathogenic. However, the low percentage of symptomatic infections
with other viruses and microbes reflects the low percentage of acute
infections that have overwhelmed host defense mechanisms, but not
a low percentage of latent infections that cause disease. Thus there
is no orthodox explanation for the claims that some murine and avian,
most feline, and all bovine and human leukemias (Section D) are the
work of latent viruses.
Even the view that retroviruses cause leukemia or
carcinoma directly in productive infections is debatable, because
indeed highly productive infections are frequently asymptomatic. For
example, despite chronic acute viremias certain chickens, mice, or
cats, inoculated experimentally or by contact as immunotolerant newborns,
do not develop leukemia (see above and Section C). Further, no malignant
transformation has ever been observed in cultured cells that are actively
producing retroviruses, and the probability that an infected cell
of an animal will become transformed is only 10-11 (Section
F). This low probability that a productively infected cell will become
transformed is a uniquely retrovirus-specific reason for asymptomatic
infections. It is for this reason that retroviruses without onc gene
can be asymptomatic for cancer even in acute, productive infections
of animals (30, 31, 36, 42, 66, 70), although they may then cause
other diseases (Section B).
Thus retrovirus infections are not only asymptomatic
due to latency and low levels of virus infiltration, like all other
viruses, but are also asymptomatic due to a particular discrepancy
between acute and productive infection and oncogenesis. To answer
the question of why some viremic animals do and others do not develop
leukemia and why tumors appear so late after infection (Section B),
both tumor resistance genes (Section C) and the mechanism of transformation
must be considered (Section H).
B. Discrepancies between the Short Latent Period of Replication and
the Long Latent Periods of Oncogenesis: Further Proof That Virus Is
Not Sufficient for Cancer
Here we compare the kinetics of virus replication and direct pathogenic
and immunogenic effects with the kinetics of virus-induced transformation.
If retroviral genes were sufficient to induce cancer, the kinetics
of carcinogenesis would closely follow the kinetics of virus replication.
Kinetics of Replication and of Early Pathogenic
and Immunogenic Effects. The eclipse period of retrovirus
replication has been determined to be 1 to 3 days in tissue culture
(Table 1) using either transforming oncgenes as markers or
the appearance of reverse transcriptase or interference with other
viruses or plaque formation for viruses without oncgenes (13,
16) (see below). The incubation period following which retroviruses
without oncgenes induce viremia in animals is 1 to several
weeks (see below). The retroviruses with oncgenes cause cancer
essentially with the same kinetics namely within 1 to several weeks
(9, 13, 14, 16) (Table 1). In immunocompetent animals antiviral immunity
follows infections with a lag of 2 to 8 weeks.
In animals, retroviruses without oncgenes can
be directly pathogenic if they are expressed at high titers. For instance,
avian retroviruses may cause in newborn chickens diseases of polyclonal
proliferative nature like osteopetrosis, angiosarcoma, hyperthyroidism
(84-87), or hyperplastic follicles of B-cells in the bursa of Fabricius
(36, 37) after latencies of 2 to 8 weeks. The same viruses may also
cause diseases of debilitative nature such as stunting, obesity, anemia,
or immunodeficiency after lag periods of 2 to 8 weeks (88, 89). Infections
of newborn mice that cause viremia also cause polyclonal lymphocyte
hyperplasias, splenomegaly, and immunosuppression several weeks after
infection (47) (Section A). The early appearance of hyperplastic nodules
in mammary tumor virus-infected animals prior to malignant transformation
has also been proposed to be a virus-induced, hyperplastic effect
(56, 57). Infection of newborn kittens with feline leukemia virus
causes early runting effects and depletion of lymphocytes within 8
to 12 weeks (64, 67, 68) followed by persistent viremia in up to 80%
of the animals (90). In experimentally infected adult animals mostly
transient (85%) and only a few persistent (15%) viremias are observed
(64, 68, 90). Likewise primate retroviruses such as Mason-Pfizer virus
(91) or simian AIDS virus (92) or STLV-III virus (93) may cause runting,
immunodepression, and mortality several weeks after inoculation if
the animals do not develop antiviral immunity. These early and direct
pathogenic effects of retroviruses without oncgenes depend
entirely on acute infections at high virus titers and occur only in
the absence of or prior to antiviral immunity.
Retroviruses have also been observed to be directly
pathogenic by mutagenesis via provirus integration of cellular genes
(13, 16, 94, 95). Given about 106 kilobases for the eukaryotic genome
and assuming random integration, a given cellular gene would be mutated
in 1 of 106 infected cells (see Sections E and F). Therefore this
mechanism of pathogenesis would play a role in vivo only if
mutagenesis were to occur at a single or few cell stage of development
(94) or if such a mutation would induce clonal proliferation, as is
speculated in Section E.
Certain direct, cytopathic effects of retroviruses
without oncgenes are also detectable in vitro within
days or weeks after infection, although malignant transformation has
never been observed in cell culture. For example, the avian reticuloendotheliosis
viruses fuse and kill a fraction of infected cells during the initial
phase of infection (96, 97). Certain strains of avian retroviruses
form plaques of dead primary chicken embryo cells in culture within
7 to 12 days post infection. This effect is probably based on cell
fusion and has been used as a reliable virus assay (45, 98). The plaque
assays of murine leukemia viruses on XC rat cells (99) and on mink
cells (100) or of feline, bovine, and simian viruses on appropriate
cells (101-104) also reflect fast cytopathic effects involving fusions
of infected cells (45). Cell fusion of human lymphocytes in vitro
is also typical of HTLV-I (105, 106) and of AIDS virus (27) (see Part
II). Cells are thought to be fused in vitro by cross-linking
through multivalent bonds between viral envelope antigens and cellular
receptors, a process that requires high local concentration of virus
particles (13, 16, 27, 45, 105). The fusion effect is not observed
in chronic acute or latent infections of animals or humans or in chronically
infected cell lines cultured in vitro. Therefore it appears
to be predominantly a cell culture artifact, possibly resulting from
interaction between virus receptors of uninfected cells with viruses
budding from the surface of adjacent cells. This has been directly
demonstrated by inhibition of HTLV-I-mediated fusion with antiserum
from infected individuals (105). Thus as direct pathogens the retroviruses
are not "slow" viruses, as they are frequently termed with regard
to their presumed role in carcinogenesis. The "lentiviruses"
that are considered models of slow viral pathogenesis (13), but not
carcinogenesis, are no exception. Recently an ovine lentivirus known
as visna or maedi virus was shown to cause rapid lymphoid interstitial
pneumonia in newborn sheep, several weeks after infection (269). This
study pointed out that the virus, if expressed at high titer, is directly
and rapidly pathogenic. Slow disease may reflect persistent virus
expression at restricted sites.
Late Oncogenesis. Since retroviruses
without oncgenes do not transform cells in culture, all measurements
of the latent period of viral oncogenesis are based on studies of
infected animals or humans (Table 1). Typically, the latent periods
are dated from the time of virus infection and thus are somewhat presumptuous,
in that the assumption is made that tumors, if they appear, were initiated
by the virus.
The latent period between experimental or congenital
infection and lymphomatosis in chickens ranges from 6 months to several
years (13, 16, 30, 32, 36, 107). In mice congenitally or experimentally
infected with murine leukemia viruses, leukemia takes 6 to 24 months
to appear (9, 39, 42, 108). The latent period of mammary carcinomagenesis
in mice infected by milk-transmitted MMTV ranges from 6 to 18 months
and typically requires several pregnancies of the mouse (16, 54).
Longer latent periods of up to 24 months are observed in mice that
do not express virus in their milk (55, 109).
The latent period between experimental infection and
leukemia is 8 and 12 months in most cats, but only 2 to 3 months in
some (62, 66, 90). (The early tumors may have been hyperplasias or
tumors induced by feline sarcoma viruses.) The latent period estimated
between natural virus infection and leukemia is estimated to be 2
to 3 years in cats that express virus and about 2 to 6 years in cats
that do not express virus (63, 66, 110). By contrast, induction of
antiviral immunity occurs within several weeks after infection (64,
67).
Bovine leukemia virus-associated leukemias are never
seen in animals less than 2 years old and appear at a mean age of
6 years (16). The only experimental bovine lymphosarcoma on record
appeared 7 years (73) and some experimental ovine leukemias appeared
4 years (75) after virus inoculation. By contrast, antibody to viral
core and envelope proteins appears 4 and 9 weeks after infection (73).
Experimental infection of gibbon apes generated leukemia after a latent
period of 1 year compared to only 2 weeks for the appearance of antiviral
immunity (16, 70).
The latent period for the development of human T-cell
leukemia in HTLV-I positive cancers has been estimated at 5 to 10
years based on the lag between the onset of leukemia and the first
appearance of antiviral antibodies of proviral DNA (13, 111, 112).
More recently, the latent period of HTLV-I has been raised to record
heights of 30 (270) and 40 years (271). By contrast, the latent period
of infection and subsequent antiviral immunity was determined to be
only 50 days based on seroconversion of the recipients of HTLV-I-positive
blood transfusions (81).
The 5- to 40-year latencies claimed for leukemogenesis
by HTLV-I are perhaps the most bizarre efforts in linking a virus
with a disease. If correct this means either that an infected T-cell
becomes leukemic by the time it is 5 to 40 years old or that one of
its offspring becomes leukemic in the 50th to 500th generation, assuming
an average generation time of a month (176). Clearly the role of the
virus in such a process, if any, must be highly indirect. Since all
viral genes are essential for replication (13, 204), there is nothing
new that the virus could contribute after one round of infection or
24 to 48 hours. This is specifically relevant for HTLV-I and bovine
leukemia viruses which are biochemically inactive not only during
the long latent period but also during the lethal period of the disease
(Sections A and D).
The monumental discrepancies between the long latent
periods from 6 months to 10 years for leukemogenesis compared to the
short latent periods of several weeks for virus replication or direct
pathogenic and immunogenic effects are unambiguous signals that the
viruses are not sufficient to initiate leukemia and other tumors (Fig.
1). The viruses are fast and efficient immunogens or pathogens but
are either not or are highly indirect carcinogens.
Transformation in vitro by HTLV-I in
30 to 60 days? Immortalization of primary human lymphocytes
infected by HTLV-I or ATLV or simian retroviruses in vitro
has been suggested to be equivalent to leukemogenic transformation
in vivo (13, 27, 113, 114). If correct, this would be the only
example of a retrovirus without oncgenes capable of malignant
transformation in vitro. The assay infects about 5 x 106 primary
human lymphocytes with HTLV-I. However, less than one of these cells
survives the incubation period of 30 to 60 days, termed "crisis"
because the resulting immortal cells are monoclonal with regard to
the proviral integration site and because only 4 of 23 such experiments
generate immortal cells (115). Since no virus expression is observed
during the critical selection period of the immortal cell and since
some immortalized cells contain only defective proviruses (115), immortalization
is not a viral gene function. Further it is unlikely that the integration
site of the provirus (Sections E, G, and H) is relevant to the process
of immortalization, since different lines have different integration
sites (115). Indeed, spontaneous transformation or immortalization
of primary human lymphocytes has been reported applying this assay
to simian viruses (113). It follows that immortalization in culture
of cells infected by HTLV-I is an extremely rare, perhaps spontaneous
event.
There are several indications that in vitro
immortalization and leukemic transformation are different events and
that both do not depend on HTLV-I: (a) the latent period
for immortalization is 30 to 60 days, while that of leukemogenesis
is estimated to be 5 to 10 years; (b) in vitro immortalized
cells are diploid (116), while all leukemic cells have chromosome
abnormalities (Section G); (c) leukemic cells do not express
virus (Section D) while immortalized cells do (115); (d) cells
that are clonal with regard to viral integration sites are not necessarily
leukemic, because normal T-lymphocytes monoclonal with regard to HTLV-I
integration were observed in 13 nonleukemic Japanese carriers (112);
(e) finally immortalized cell lines with defective viruses
(115) or no viruses (113) indicate that immortalization is a virus-independent,
spontaneous event. The evidence that cat, rat, and rabbit cells are
immortalized, although they are presumably insusceptible to the human
virus (13), endorses this view. It would appear that HTLV-I is directly
involved neither in immortalization nor in transformation (Sections
A, B, G, and H). Instead the assay appears to be a direct measure
of cell death of human lymphocytes, due in part to HTLV-I-mediated
fusion in vitro (105, 106), and of rare spontaneous immortalization.
C. Tumor Resistance Genes That Inhibit Tumorigenesis but not Virus
Replication
If the virus were a direct and specific cause of tumorigenesis, one
would expect that all individuals who are permissive for infection
would also be permissive for viral tumors. However, this does not
appear to be so. For example certain inbred lines of chicken like
line 7 (117, 118) or line SC (35, 107) are highly susceptible to induction
of lymphomatosis by avian retroviruses, whereas line 151 (32, 119,
120) is highly susceptible to induction of erythroblastosis by the
same avian retroviruses. By contrast other lines like line 6 (118,
121), line FP (107), or line K28 (122) are either completely or highly
resistant to these leukemias but are just as susceptible to virus
infection and replication as the tumor-susceptible lines (32, 117,
118, 122, 123). Indeed, both the lymphoma-susceptible SC chickens
and the resistant FP chickens develop early viremias and hyperplastic
B-cell follicles, but only 50% of the SC chickens develop lymphomas
(35, 36). Lymphoma resistance is dominant, indicating that tumor suppressors
are encoded (120, 124). The same genes also appear to impart resistance
to Rous sarcoma (124). By contrast resistance to erythroblastosis
is recessive (Section E).
Analogous tumor resistance genes have been observed
in mouse strains. For instance, resistance of C57BL mice to radiation
leukemia virus-induced leukemia (125) or of AKR x BALB/c mice to AKR
virus-induced leukemia (40) is controlled by the H-2D gene,
which is dominant for resistance. Inoculation of the virus into adult
C57BL mice caused polyclonal B- and T-cell hyperplasia from which
most animals died after 4 to 5 months. However, no leukemia was observed
(47). Clearly the tumor resistance genes of the C57BL mice do not
suppress virus replication but apparently proliferation of transformed
cells. Likewise the S1 and the Fv-2 genes of mice inhibit leukemogenesis
but not replication of Friend leukemia virus (13, 16, 126). The fates
of DBA/2 and ST/b mice inoculated neonatally with AKR virus are another
example. After expressing virus for at least 8 months (41), only ST/b
mice show a high incidence (about 80%) of leukemia between 8 and 12
months of age, whereas DBA/2 mice show a lower incidence (about 30%)
but only at 2 to 3 years of age.6
Futhermore, not a single lymphoma developed during
a period of 1 year in chronically viremic CBA/N mice, inoculated as
newborns with Moloney leukemia virus, signalling an absolute resistance
to leukemogenesis (42, 46). By contrast, about 90% of viremic AKR
mice develop leukemia (40, 48). The wide range of susceptibilities
to virus-induced leukemia among different mouse strains inoculated
with AKR virus, as originally observed by Gross (9), probably also
reflects postinfection tumor resistance genes in addition to genes
conferring resistance to virus infection and expression (16).
The over 100-fold variation (from less than 1% to
90%) in the incidence of mammary carcinomas among mice that are susceptible
to the mammary tumor virus and also contain endogenous MMTVs also
reflects host genetic factors that govern resistance to tumorigenesis
(16, 54, 55, 58, 127-129). One set of resistance genes governs virus
expression, as for example the sex of the host, because almost only
females secrete virus and develop tumors (13, 16). Another set governs
resistance to carcinogenesis because virus-induced hyperplasia does
not necessarily lead to mammary tumors (56, 57).
Resistance to tumorigenesis in animals which are permissive
for virus replication indicates that tumors contain nonviral, cellular
determinants or tumor antigens. Moreover defects of tumor resistance
genes rather than viral genes determine tumor specificity since the
nature of the tumor induced by a given virus depends on the host and
not on the virus. This lends new support to the conclusion that viruses
are not direct causes of tumorigenesis.
D. Tumors without Virus Expression, without Complete Viruses, or without
Viruses: Proof That Virus Is Not Necessary to Maintain Transformation
If the retroviruses encode transformation-specific functions,
one would expect that viral genes are continuously expressed in viral
tumors. However, only 50% of virus-induced avian lymphomas express
viral RNA (130). In many clonal lymphomatoses of chickens only incomplete
or truncated proviruses are found. These defective proviruses lack
the 5' half of the genome and hence are unable to express any viral
gene (36, 50, 131, 132).
Moreover neither exogenous nor active endogenous retroviruses
can be detected in some lymphomas. One rare study that investigated
lymphomatosis in lymphomatosis virus-free chickens found that 10 of
about 2000 (0.5%) chickens of line 7 died from lymphomas that were
indistinguishable from viral lymphomas at the ages of 6 to 18 months
(49, 121). Thus the incidence of lymphoma in virus-free chickens is
very similar if not the same as that of chickens infected by lymphomatosis
virus with antiviral immunity (less than 1%) (Section A). Since almost
all chickens contain multiple endogenous retroviruses (16, 133), it
may be argued that these viruses were responsible for the leukemias
in animals free of exogenous virus. However, the evidence that endogenous
viruses were latent in leukemic as in nonleukemic birds indicated
that the endogeous retroviruses were not involved in these spontaneous
lymphomas (121). The existence of endogenous viruses in the lymphoma-resistant
chickens of line 6 supports this view (121, 133). In fact, it has
been argued that endogenous viruses protect by interference against
infection by exogenous variants (13, 16, 134).
A few cases of mouse T-cell lymphomas with defective
leukemia viruses have also been observed (135-137). These findings
indicate that murine leukemia can also be maintained without expression
of retroviral genes.
Expression of mammary tumor virus appears also not
necessary to maintain tumors, because no viral antigens (138) and
no virus expression are detectable in many virus-positive mammary
tumors (9, 52, 139) and because defective proviruses are observed
in some tumors (140). Moreover, in mice which lack mammary tumor virus
altogether, mammary tumors were observed that cannot be distinguished
from virus-positive tumors, indicating that the virus is not necessary
to initiate mouse mammary tumors (141). However, in the absence of
virus expression, mammary carcinomas develop at lower incidence and
after longer latent periods (9, 16, 52, 139-142).
Among virus-positive feline leukemias, some contain
only defective proviruses, as in the avian system (143-145). However,
about 25 to 35% of all feline leukemias are free of virus, viral antigens
(67, 68, 110), and proviral DNA (143-145). This is significantly higher
than the percentage of virus-free avian lymphomas. In some virus-free
leukemias, the presumably lymphotropic virus is believed to be in
other cells of the cat (65).
In provirus-positive natural bovine and experimental
ovine leukemias expression neither of virus nor of viral RNA have
been detected (75, 146). This result is at odds with the proposal,
based on in vitro evidence, that the virus encodes a protein
that activates virus transcription and expression of latent cellular
transforming genes (147). In addition, the 5' half of bovine leukemia
provirus is absent from 25% of bovine leukemias (146, 148). This entirely
prevents expression of all viral genes. Other investigators have described
that 30% of bovine leukemias are virus free (72).
The proviruses of HTLV-I associated with human T-cell
leukemias are also consistently latent. For instance, no expression
of viral antigens (149) and no transcription of viral RNA are observed
in freshly isolated leukemic T-cells from (5 of 6) HTLV-I positive
patients with human T-cell leukemia (150, 151). Again this is incompatible
with the in vitro evidence for a viral transcriptional activator
that was proposed to activate virus expression and expression of latent
cellular transforming genes (152, 153) (Section H). Moreover, about
10% of the ATLV- or HTLV-I-positive adult T-cell leukemias from Japan
contain only defective viruses (77, 151,154). Since the 5' half
of the viral genome was reported to be missing no viral gene expression
is possible (77, 151, 155). Further, a minority of Japanese ATL patients
appears to be free of ATLV, based on the serological assays that are
used to detect the virus (156, 157). A recent analysis found 5 virus-free
cases among 69 Japanese ATL patients, who lacked both HTLV-I provirus
and antiviral immunity (158). Comparisons among T-cell leukemias in
Italy found only 2 of 68 (159) or 3 of 16 (160) otherwise identical
cases to be HTLV-I positive. A survey from Hungary found 2 of 326
leukemias antibody positive (161). Other studies from the United States
and Italy describe HTLV-I-free T-cell leukemias that share chromosome
abnormalities with viral leukemias (Section H). Thus, the ratio of
nonviral to viral T-cell leukemias in humans outside Japan appears
to be even higher than that of nonviral to viral feline and bovine
leukemias.
Since retrovirus expression is not observed in many
virus-positive leukemias and since only defective viruses are associated
with some leukemias it follows that viral gene products are not necessary
to maintain these leukemias. These tumors must be maintained by cellular
genes (Section H). The occurrence of "viral" leukemias of chicken,
mice, cats, cattle, and humans despite antiviral immunity (Section
A) supports this conclusion. This conclusion is also consistent with
the evidence that about 30% of the natural feline and bovine leukemias
as well as many human and some avian leukemias and murine mammary
carcinomas are virus free, yet these tumors cannot be distinguished
from viral tumors by any criteria other than the virus (Section H).
E. Transformation Not Dependent
on Specific Proviral Integration Sites
Since retroviruses without oncgenes are not sufficient to cause
tumors and do not encode transformation-specific functions (Sections
A-C) but may nevertheless induce experimental tumors (Section A),
several hypothetical mechanisms of viral carcinogenesis have been
proposed that each require a specific interaction with the host cell
(Section H). One of these postulates that retroviruses without oncgenes
activate latent cellular cancer genes, termed proto-oncgenes,
by site-specific proviral integration (13, 16, 130, 162). The proposal
is based on structural analogy with retroviral oncgenes, which
are hybrids of sequences derived from retroviruses and proto-oncgenes
(5, 19, 20). It is termed downstream promotion hypothesis (130) because
the promoter of the 3' long terminal repeat from the provirus is
thought to promote transcription of a proto-onc gene downstream.
It is consistent with this hypothesis that leukemias
and other tumors from retrovirus-infected animals and humans are typically
all monoclonal with regard to the integration sites of the provirus
in the host chromosome. However, if one compares different monoclonal
tumors of the same cell lineage, different integration sites are found
in each individual tumor. This has been documented for retroviral
lymphomas of chickens (37, 131, 132), mice (13, 163, 164), cats (143-145),
cattle (146, 148), and humans (13, 151, 154, 155, 165) and also for
mammary tumors of mice (13). It is unlikely that the mutant genes
generated by provirus integrations are transforming genes, because
they are not specific and not known to have transforming function
upon transfection. Instead the clonal proviral integration sites of
individual tumors appear to be the consequence of clonal proliferation
of a single transformed cell from which the clonal tumor originated
(Section G).
Relevance of Preferred Integration Regions.
Although the search for specific proviral integration sites in viral
tumors has met with no success, preferred integration regions were
observed in three systems, namely in erythroblastoses and lymphomas
of chicken strains predisposed to these tumors and in mammary tumors
of mice bred for susceptibility to this tumor (13, 16). For instance,
in erythroblastosis-prone 15I chickens that suffer 80% erythroblastosis
upon infection (120), integration upstream of proto-erb was
observed in 90% (119) and 45% (120, 122) of erythroblastoses. Proto-erb
is a proto-onc gene because it is the cellular progenitor of the transforming
gene of avian erythroblastosis virus (13, 19). This region-specific
integration appears to activate proto-erb transcription compared
to certain normal controls (119). However, there are as yet no data
on activation of proto-erb translation in leukemic cells. Unexpectedly
45% of the erythroblastoses observed in 15I chickens contained viruses
with transduced proto-erb (122). The outstanding yield of proto-erb
transductions in this line of chicken compared to others (5, 19) (Section
H) suggests an altered proto-erb gene, perhaps already flanked
by defective proviral elements which would permit transduction via
homologous recombination. It is consistent with this view that in
15I chickens susceptibility to erythroblastosis is dominant (120),
while typically resistance to tumors is dominant in chickens and mice
(Section C).
Further in about 85% of the viral lymphomas of lymphomaprone
chicken lines (Section C) transcription of the proto-myc gene
is activated compared to certain controls (130). Proto-myc
is a proto-onc gene because it is the cellular progenitor of the transforming
genes of four avian carcinomas viruses, MC29, MH2, CMII, and OK10
(5, 13, 19). Transcriptional myc activation ranges from 300- to 500-fold
in some lymphoma lines (RP) to 30- to 100-fold in most primary lymphomas
(85%) down to undetectable levels in a few (6%) primary lymphomas
(130). However, the activation of proto-myc translation, compared
to normal fibroblasts, was estimated as only 7-fold in one RP lymphoma
line and even lower in three other lines (166). Assuming that the
same ratios of transcriptional to translational activation apply to
all lymphomas, activation of myc translation would be only 1- to 2-fold
in most lymphomas, hardly enough to explain carcinogenesis. In 5 to
15% of the lymphomas there is no detectable transcriptional activation
of proto-myc and the retroviruses appear to be integrated outside
of and in random orientation relative to the proto-myc genes
(50, 105, 130, 132, 167, 168, 169).
Thus, in lymphomas, proto-myc transcription
is frequently but not always activated whereas proto-myc translation
appears to be barely, if at all activated. It is not known whether
translation of proto-erb is activated in viral erythroblastoses.
By contrast viral myc and erb genes are efficiently translated in
all virus-transformed cells (5, 13, 16, 19, 20). Moreover in contrast
to the hypothetical lymphoma specificity of activated proto-myc,
viral myc genes typically cause carcinomas and viral erb genes cause
sarcomas in addition to erythroblastosis (5, 13).
Integration of mostly intact murine leukemia viruses
into or upstream of proto-myc is also observed in mouse and
rat lymphomas. But since it occurs only in 10 (170, 171) to 65% (172)
of the cases analyzed, it is not necessary for lymphomagenesis. Moreover
provirus integration near murine proto-myc is also not sufficient
for leukemogenesis. Virus integrated near proto-myc was found
in 15% of the hyperplastic thymus colonies of AKR mice that appeared
35 days after infection with MCF virus. These colonies were not tumorigenic
(172). However, more malignant lymphomas develop from cells with provirus
integrated near myc than from other cells, because in 65% of the lymphomas
virus was integrated in proto-myc.
There are also preferred regions of provirus integration
for MMTV in carcinomas of mice, termed int-1 in C3H mice and
int-2 in BR6 mice (13, 16). The int loci or genes are
considered to be proto-oncgenes only because they are preferred
MMTV integration sites. They have not been progenitors of viral oncgenes
and there is no direct evidence that they can be activated to cellular
cancer genes. Moreover transcriptional activation of int is
observed only in some tumors (173) and there is no evidence for viral-int
hybrid mRNAs (140). It is also not known whether the int loci
are coding. The two int loci are totally unrelated to each
other and map on different chromosomes (174). Integration within the
int regions is neither site nor orientation specific with regard
to the int loci (13). Integration at int loci is also
not necessary for carcinogenesis, because integration in int-1
is found in only a fraction (20 of 26) of C3H tumors (173) and in
int-2 only in a fraction (22 of 45) of BR6 tumors (140). Further
integration in int-1 was found in benign hyperplastic nodules
that did not become malignant, proving that it is also not sufficient
for carcinogenesis (56, 57).
The hypothesis that region-specific integration generates
hybrid transforming genes that are equivalent to viral oncgenes
is inadequate on several counts. (a) Region-specific
integration is not necessary for transformation, because in most systems
(human, bovine, feline) it is not observed and in all others it is
not obligatory. (b) It is also not sufficient for carcinogenesis
based on the particular cases of clonal murine leukemia virus integration
into proto-myc that did not cause leukemia (172), clonal MMTV
integration into int-1 that did not cause mammary carcinomas
(56, 57), and monoclonal HTLV-I infections that did not cause T-cell
leukemia (112). The nonleukemic proto-myc integration is incompatible
with the model purporting that activated proto-myc is like
the inevitably transforming viral myc genes (5). (c)
The prediction that native proviral-cell DNA hybrids have transforming
function, like the related retroviral onc gene models, is unconfirmed.
Attempts to demonstrate transforming function of proviral-proto-myc
hybrids from chicken lymphomas were negative but led to a DNA with
transforming function termed B-lym (13, 175). A plausible reason is
that the myc RNAs initiated from upstream viral promoters are poor
mRNAs because they start with intron sequences that are not part of
normal mRNA and cannot be spliced out, since there is no splice donor
downstream of the 3' viral long terminal repeat (Section H). (d)
The prediction that the probability of all infected cells to become
transformed should be the same as that of region-specific integration
is also unconfirmed on the basis of the following calculations (5).
The proto-myc, -erb, or int regions that are
preferential proviral landing sites in viral tumors measure about
2 and 40 kilobases, respectively (13). Since the chicken chromosome
contains about 1 x 106 kilobases and the mouse chromosome contains
about 3 x 106 kilobases, and since provirus integration is random
(13, 16), about 2 in 106 or 1 in 105 infections should generate a
tumor cell, if region-specific integration were the mechanism of carcinogenesis.
Yet the probability that an infected cell will initiate a monoclonal
tumor is only about 10-11 (Section F). In addition, the
latent period of tumorigenesis would be expected to be short because
there are at least 108 target cells of the respective lineages and
many more viruses to infect them (Section F). Moreover, given the
long latent periods of carcinogenesis, polyclonal rather than monoclonal
tumors would be expected from integrational carcinogenesis. It may
be argued that this discrepancy reflects the work of tumor resistance
genes. However, post infection resistance genes that suppress tumor
formation by the viral derivatives of proto-myc or erb, like
MC29 or avian erythroblastosis virus, have never been observed in
vivo or in vitro. Clearly, since tumor resistance genes
do not function in vitro it would be expected that at least
2 of 106 cells infected in vitro would be transformed by activation
of proto-myc and 2 by activation of proto-erb. However,
no transformation by leukemia viruses has ever been observed in
vitro (Section B).
In view of this, it is more likely that region-specific
integration may provide proliferative advantages to hyperplastic cells
or may initiate hyperplasia by activating or inactivating growth control
genes rather than being the cause of malignancy. This proposal predicts
that integration into proto-myc and proto-erb precedes
tumorigenesis (Fig. 1 ).
It is consistent with this proposal that murine leukemia
virus integration into proto-myc (172) and MMTV integration
into int-1 (56, 57) occur prior to carcinogenesis and thus are not
sufficient for carcinogenesis. This proposal predicts also that the
chicken lines that are susceptible to lymphoma or erythroblastosis
lack genes that check hyperplasia of lymphocytes or erythroblasts.
It is consistent with this view that the same retroviruses cause either
lymphomatosis or erythroblastosis or no tumors in different chicken
lines. The exclusive (but not absolute) usage of only one of two different
int loci by MMTV, namely int-1 in carcinomas of C3H
mice and int-2 in BR6 mice, is also more likely to reflect
strain-specific activation or inactivation of proliferative controls
than two entirely different transforming genes that would nevertheless
generate indistinguishable carcinomas.
F. The Probability that a Virus-Infected Cell
Will Become Transformed is Only 10-11
To calculate the probability that a retrovirus-infected cell will
become transformed, we must consider the ratio of symptomatic to asymptomatic
carriers, the clonality of the viral tumors, and the long latent periods
of oncogenesis. (a) The ratio of symptomatic to asymptomatic
carriers with latent infections and antiviral immunity averages less
than 10-3 (Section A), but that of viremic animals susceptible
to transformation may reach 0.9 (Section C). (b) Since monoclonal
tumors emerge from at least 108 B- or T-cells (176), the probability
of an infected cell in an animal to become the progenitor of a clonal
leukemia is only about 10-8. This calculation assumes that
all of these cells are infected. This is certainly true for the mice
that carry AKR virus, radiation leukemia virus (82), or inducible
mammary tumor virus (75, 142) in their germ line, and is probably
the case in congenitally infected viremic chickens, cats, gibbons,
and mice (12, 16, 31, 39, 63, 66, 70). In fact in viremic animals,
the hyperplastic effect of the virus would have enhanced the number
of prospective tumor cells to at least 109 (Sections A and B). Even
if only a fraction of susceptible cells are infected in animals or
humans with latent infections and antiviral immunity, the number of
infected cells per host is estimated to be at least 106 to account
for the immune response (Section B, and Refs. 13, 16, 27, 31, and
63) or the proviruses that are used to diagnose latent virus infection
(Section D). Proviruses cannot be detected biochemically unless they
are present in at least 1 of 100 cells. (c) Finally,
the probability of an infected cell to become transformed in an animal
is a function of the number of generations of infected cells that
occur during the latent period of the disease. Given latent periods
of 6 to 120 months (Section B) and assuming an average life span of
1 month for a susceptible B- or T-cell (176), about 10 to 100 generations
of infected cells are required to generate the one transformed cell
from which a clonal tumor emerges. The corresponding probability that
a generation of cells will develop a clonal tumor would be 10-1
to 10-2. Considering the proliferative effect of the virus
on hemopoietic target cells in viremic animals, this may again be
a conservative estimate. Indeed, a mitotic rate of 1 day has been
assumed for B-cells of lymphomatosis virus-infected chickens (177).
Thus the probability that a virus-infected, hemopoietic
cell will become transformed in an individual with a latent infection
and antiviral immunity is about 10-3 x 10-6
x 10-2 = 10-11, and that in a viremic individual
without tumor resistance genes is about the same, namely 0.9 x 10-9
x 10-2 = 10-11. Therefore the increased risk
of viremic animals to develop leukemia must be a direct consequence
of the hyperplasia of prospective tumor cells (Section A) (Fig. 1).
In tumor-resistant animals the probability that the infected cell
will become transformed may be the same, but the resistance genes
would prevent proliferation of the transformed cells (Section C and
H). The apparent probability that virus-infected, nonhemopoietic cells
will become transformed must be lower in both susceptible and resistant
animals, because the incidence of solid tumors is much lower than
that of leukemia (9, 32).
G. Clonal Chromosome Abnormalities Are the Only Transformation-Specific
Markers of Retrovirus-Infected Tumor Cells: Causes of Transformation?
The evidence that viral tumors are monoclonal (Section E) and that
leukemogenesis by retroviruses (without oncgenes) is highly
dependent on tumor resistance genes, which are different from genes
that determine susceptibility to the virus, suggest virus-independent
steps in carcinogenesis (Section C). Indeed clonal chromosome abnormalities
of virus-positive mammalian tumors provide direct evidence for cellular
events that may be necessary for carcinogenesis. (Avian cells have
not been studied because of their complex chromosome structure.)
For example, trisomies of chromosomes 15 have been
observed frequently in viral T-cell leukemias of mice (16). In addition
translocations between chromosomes 15, 17, and others have been recorded
(108, 178-180, 272). In mammary carcinomas of mice, a chromosome 13
trisomy was observed in 15 of 15 cases including inbred GR and C3H
mice (which contain MMTV) and outbred Swiss mice (which probably also
contain the virus) (181). Clonal chromosome abnormalities have also
been observed in 30 of 34 bovine leukemias examined (16, 182) as well
as in ovine leukemias induced by bovine leukemia virus (75).
A recent cytogenetic analysis of human adult T-cell
leukemias (ATL) from Japan showed that 10 of 11 cases had an inversion
or translocation of chromosome 14 (183). Rearrangements of other chromosomes
have been detected in 6 of 6 (184), 12 of 13 (116), and 8 of 9 cases
of HTLV-I-positive leukemias (185). Thus over 90% of virus-positive
T-cell leukemias have chromosome abnormalities. A survey of all viral
T-cell leukemias analyzed shows rearrangements of chromosome 14 in
26% and of chromosome 6 in 29% (186, 187).
The chromosome abnormalities of these viral leukemias
and carcinomas are as yet the only known determinants that set apart
transformed from normal virus-infected cells. Since the chromosome
abnormalities are clonal, the origin of the tumor must have coincided
with the origin of the chromosome abnormality. Therefore chromosome
abnormalities or closely associated events must be directly relevant
to initiation of tumorigenesis. They could either be, or coincide
with, a single step mechanism of transformation or with one of several
steps in transformation, as postulated in the case of the Philadelphia
chromosome (188). It is consistent with this view that chromosome
abnormalities are found in all virus-infected tumors analyzed.
However, heterogeneity among the karyotypes of individual
human or murine leukemias of the same lineage (16, 179, 182, 189,
190, 272) and thus heterogeneity of mutation support the view that
chromosome abnormalities are coincidental with rather than causal
for transformation. Yet this view does not take into consideration
that together with the microscopic alterations, other submicroscopic
mutations may have occurred that could have initiated the disease
(108). It is consistent with this view that tumor cells contain in
addition to microscopic karyotype changes submicroscopic deletions,
detectable as restriction enzyme site polymorphisms (191). Some of
these mutations may be functionally equivalent to the truncation-recombination
mechanism that activates the docile proto-oncgenes of normal
cells to the oncgenes of directly oncogenic retroviruses (5,
192). Thus specific karyotypic changes may only be the tip of the
iceberg of multiple chromosomal mutations, referred to as "genequake,ś
which must have occurred in the same cell. One or several of these
could have initiated the tumor. Chromosome recombination sites are
also postulated to be cellular transforming genes of virus-negative
tumors, as for example in Burkitt's lymphoma (5) or in human leukemia
with the Philadelphia chromosome (193).
If chromosomal abnormalities are necessary for transformation
of cells infected by retroviruses without oncgenes, chromosomal
abnormalities would not be expected in tumors caused by retroviruses
with directly transforming oncgenes. This has indeed been confirmed
for tumors caused in mice by Rous sarcoma virus (194) or by Abelson
leukemia virus (195) which have normal karyotypes (Table 1).
The clonality of retrovirus-positive tumors is then
defined in two different ways: by a retroviral integration site (see
Section E), and by a chromosome abnormality (see Fig. 1). Each of
these two clonal chromosome alterations could then mark the origin
of the tumor, while the other must have preexisted. Since the tumors
originate late after infection and probably from a virus-infected,
normal cell, the clonal retroviral integration site would appear to
be a direct consequence of clonal proliferation of a cell transformed
by a chromosome alteration. Indeed chromosome abnormalities are typical
of tumor cells but not of virus-infected normal cells. This view is
consistent with the evidence that retrovirus integration does not
cause transformation and that transformation is not dependent on specific
integration sites. It is also highly improbable that chromosome abnormalities
are caused by the virus, because they are not found in virus-infected
normal cells and because they are also characteristic of virus-negative
tumors (Section H). The clonal retroviral integration sites in viral
tumors the chromosomes of which have not been analyzed, as for example
avian, feline, and simian leukemias, may indeed signal as yet undetected
clonal chromosome abnormalities.
H. Virus-Independent Transformation in Virus-Positive and -Negative
Tumors
Several hypotheses postulate that retroviruses play a direct role
in carcinogenesis. One reason is that viruses, seemingly consistent
with Koch's first postulate, are associated with tumors although
frequently in a latent or defective form. In addition it appears consistent
with Koch's third postulate that experimental infections with retroviruses
may induce leukemia under certain conditions (see Sections B and C).
However, none of these hypotheses provides an adequate explanation
for the fact that retroviruses are not sufficient to initiate (Sections
A to C) and not necessary to maintain (Sections D and E) transformation
and do not encode a transformation-specific function. Moreover none
of these hypotheses can explain why transformation is initiated with
a clonal chromosome abnormality (Section G) and why tumor specificity
is determined by the host rather than the virus (Sections C and E).
The shortcomings of three of these hypotheses are briefly reviewed
here.
1. The Oncogene Hypothesis. Huebner
(8) and others (9, 82) have postulated that retroviruses (without
oncgenes) are direct carcinogens that include oncogenes, hence
the term "oncogene hypothesis" (8). The hypothesis was based
on abundant positive correlations between retrovirus expression and
cancer incidence in laboratory mice and domestic chickens, which indeed
suggested direct viral etiology in apparent accord with Koch's
third postulate. The hypothesis generalized that either import of
retroviruses from without, or activation of latent viruses from within,
is the direct cause of spontaneous, chemically induced, or physically
induced tumors (8, 9, 82). However, the hypothesis failed to account
for the long latent periods of oncogenesis and for complete tumor
resistance by certain animals that are highly susceptible to the virus
and for host genes that would determine tumor specificity (Section
C). Above all the hypothesis failed to account for the monoclonality
and the chromosome abnormalities of the resulting tumors.
2. The Hypothesis That Latent Cellular Cancer
Genes Are Activated by Provirus Integration. This hypothesis
has been introduced in Section E. It holds that retroviruses act as
direct, albeit inefficient carcinogens by generating hybrid transforming
genes from proviruses joint with cellular proto-oncgenes. Excepting
the specific cases described in Section E, this mechanism makes four
clear predictions, namely: (a) that different transforming
genes exist in each tumor, because each has a different proviral integration
site (Section E); (b) that therefore a large number of tumor
resistance genes exist in tumor-resistant animals (Section C); (c)
that provirus-cell hybrid genes are expressed to maintain transformation;
and (d) that virus-transformed cells exist without chromosome
abnormalities, analogous to cells transformed by retroviruses with
oncgenes (Section G).
None of these predictions is confirmed, (a)
Contrary to the expectation for many different transforming genes,
all virus-positive tumors of a given lineage are phenotypically highly
uniform (Section A). Even virus-free tumors are distinguishable from
virus-positive tumors of the same lineage only by the presence of
viruses. Examples are the identical pathologies and pathogeneses of
viral and nonviral murine leukemias (196-198), chicken B-cell lymphomas
(121), human T-cell leukemias (158, 161, 186), and mouse mammary tumors
(11, 139, 141, 142) (Section D). (b) Contrary to expectation
only a small set of cellular resistance genes controls the development
of viral tumors in chicken or mice (13, 16) (Section C). Moreover
apparently the same resistance genes of chickens of line 6 suppress
viral and nonviral lymphomas, and even lymphomas induced by Marek's
virus (124). By contrast chickens of line 7 that lack these genes
are equally susceptible to both (121) (Section D). Mice provide parallel
examples such as in the CBA strain, which is resistant to spontaneous
(9) as well as to viral (46) leukemia (Section C). (c) Contrary
to expectation for virus-cell hybrid transforming genes, proviruses
are latent or defective and biochemically inactive in many animal
and all bovine and human leukemias (Section D). (d) Contrary
to expectation for viral carcinogenesis all virus-positive murine,
bovine, and human tumors analyzed have chromosome abnormalities. Further,
similar chromosome abnormalities in viral and nonviral tumors again
suggest common cellular transforming genes. For instance, the same
chromosome 15 trisomy is observed in murine leukemias induced by viruses,
chemicals, or radiation (180, 190, 199-201,272). In addition virus-positive
and virus-free human T-cell leukemias have common abnormalities in
chromosomes 14 and 16 (160, 183, 186, 187, 189, 202, 203). Since all
human T-cell leukemias and all bovine leukemias have chromosome abnormalities
but not all are infected by viruses (Sections D and G), it would appear
more likely that the viruses are coincidental passengers rather than
causes of the disease.
3. The Hypothesis that Latent Cellular Cancer
Genes Are Transactivated by Viral Proteins. This hypothesis
postulates that certain retroviruses directly activate latent cellular
transforming genes with a specific viral protein. This has been proposed
for bovine leukemia virus and human HTLV-I based on in vitro
models (147, 152, 153) (see Section D). However, the hypothesis is
unlikely for the following reasons. Since the putative transactivation
protein of HTLV-I is essential for replication (204), all cells in
which the virus replicates would be expected to be transformed. This
is clearly not the case. Further this gene cannot be relevant for
transformation since bovine and human leukemias in particular do not
express viral RNA or protein or cannot express RNA or protein because
of defective proviruses (Section D). In addition this hypothesis also
fails to account for the chromosome abnormalities found in all viral
bovine and human leukemias (Section G). Finally both the proviral
insertion and the transactivation hypotheses fail to explain the inevitably
long latent periods of viral tumorigenesis (Section B).
Therefore it is proposed that transformation is a
virus-independent event that must be due to cellular genes (Fig. 1).
These genes would be generated by chromosomal mutations for which
chromosome abnormalities are a macroscopic indicator. This explains
the clonal chromosome abnormalities that could not be predicted by
any of the virus-cancer hypotheses. In a given lineage of cells the
number of cellular genes convertible to transforming genes must be
limited since they cause highly uniform tumors which can be suppressed
by a small set of resistance genes.
Retrovirus-independent transformation resolves the
apparent paradox that tumors occur very seldom in typical natural
infections of wild animals and humans, and then only long after infection,
and despite viral latency and antiviral immunity. It is also consistent
with virus-independent transformation that the probability that an
individual virus-infected cell will become transformed is only 10-11
and that this probability is the same in a viremic chicken with a
virus-induced hyperplasia, as in a normal chicken with a latent infection
and antiviral immunity (Section F). The low probability of virus-independent
transformation also explains directly why cells infected by retroviruses
are not transformed in culture, namely because not enough cells can
be maintained for a long enough time to observe spontaneous transformation.
Virus-independent transformation is also compatible with tumor resistance
genes that do not inhibit viral replication or growth of normal virus-infected
cells. In addition it is consistent with the notion that defects of
cellular resistance genes rather than viral genes determine tumor
specificity (Section C).
The role of the virus in tumorigenesis is then limited
to the induction of hyperplasia by activating cellular proliferative
functions either from within or from without via viral antigens or
virus-induced growth factors (13, 16, 46). For this purpose the virus
must be expressed at a high titer or it must have infected a large
number of cells, if insertional mutagenesis of proliferative genes
were involved (Section E). This may be similar to the mechanism whereby
DNA viruses induce transformation, as for example Epstein-Barr virus
which is thought to induce Burkitt's lymphoma. Exactly like their
retroviral counterparts, all Burkitt's lymphomas have chromosome
abnormalities but not all contain the virus (5). Thus the role of
the retrovirus in carcinogenesis is as indirect as that of chemical
or physical carcinogens.
Alternatively a latent retrovirus may itself be subject
to activation by physical, chemical, or spontaneous events that can
induce hyperplasias and cancer (8, 12, 82) (Fig. 1). The physically
activated radiation leukemia virus (82) or the chemically activated
endogenous retroviruses of mice or chickens (12, 16) are examples.
It is uncertain whether under these conditions the retrovirus is just
an indicator or an intermediate of proliferative activations that
may lead to carcinogenesis because comparable studies with virus-free
strains of animals are not available. The physically or chemically
inducible phages or herpes viruses may in turn be models for this
(11, 83).
Little is known about the nature of the hyperplastic
cell. The existence of viral hyperplasias in tumor-resistant animals
indicates that the hyperplastic cell is not neoplastic (Section C).
Most hyperplastic cells are polyclonal with regard to proviral integration
sites (118) and are likely to have a normal karyotype, as has been
shown in some cases (47) (Section C). Hyperplastic cells with normal
karyotypes have also been observed as precursors of radiation leukemia
in mice (205). Nevertheless the evidence for clonality with regard
to a proviral integration site in T-cell hyperplasias (172) and mammary
hyperplasias (56, 57) of mice and in T-cells of healthy humans (112)
indicates clonal, possibly virus-induced alterations that are not
sufficient for carcinogenesis. One could speculate then that hyperplastic
cells fall into two classes, those which respond to viral antigens
delivered from within or without (42) and those which respond to growth
control genes altered by provirus integration (Section E).
Notable exceptions to virus-independent transformations
are infections that generate retroviral transforming genes. However,
the probability of generating a retrovirus with an onc gene is clearly
much lower than integration into a cellular gene (10-6;
Section E) and even significantly lower than virus-independent transformation
(10-11, Section F) (273). Only about 50 such viral isolates
have been recorded in history (5, 13, 19). [The frequent erb transductions
from the chicken 15I line are an exception to this rule (Section E).]
The generation of these viruses requires two rare illegitimate recombinations
to transduce a transformation-specific sequence from a cell into a
retrovirus vector (5, 19, 20, 273). However, one illegitimate recombination
that unites the 5' promoter, translational start sequence, and
splice donor of a retrovirus with a transformation-specific sequence
from a cellular proto-onc gene would be enough to generate a functional
virus-cell recombinant onc gene that cannot be replicated. Tumors
caused by such genes are presently unknown. They will be harder to
diagnose but are probably more frequent than the rare, natural tumors
containing complete retroviruses with oncgenes (273).
This raises the question of why orthodox integration
of a provirus within a proto-onc gene, like proto-myc, is not
observed to transform infected cells in vivo or in vitro
with the predicted probability. Based on the calculations described
in Section E, this probability should be about 1 in 104 considering
that about 20 proto-oncgenes are known from 20 viral oncgenes
(5, 13, 19). A possible answer is that proviruses abutting proto-oncgenes
from the proviral ends rather than from within, as in viral oncgenes
(273), provide neither new downstream translational starts nor splice
donors for those coding regions of the proto-oncgenes that
are separated from their native start signals by the inserted provirus.
Nevertheless they can provide efficient downstream promoters (130)
of RNAs that may not be translatable.
I. Are Retroviruses a Basis for Cancer Prediction, Prevention, or
Therapy?
In assessing the tumor risk of a retrovirus-infected animal or human,
latent infections must be clearly separated from chronic, acute, or
viremic infections. The control of virus expression in a given host
is a product of three factors: the virus; the host cell; and the animal.
The viral factor is defined by viral genes and promoters (13, 16,
206). The cellular factor is defined by genes that encode viral receptors
and unknown suppressors (8, 9, 11-13, 16-18, 82). The animal factor
is defined by antiviral immunity.
By far the most common natural retrovirus infections
are latent, chronic infections that persist in animals and humans
in the presence of antiviral immunity presumably only in a limited
number of cells (38, 40, 90, 207).4 The leukemia risk of this statistically
most relevant group of natural infections averages about less than
0.1% in different animal species (Section A). It is possibly the same
as, but certainly not much higher than, that of uninfected controls
(Sections A and D). Thus latent viruses offer no targets for tumor
prevention. The low probability that an immunocompetent individual
will develop chronic viremia and hence leukemia also suggests that
retroviruses carrying therapeutic genes are not a significant risk
as leukemogens.
By contrast, the leukemia risk of a viremic animal
that survives the early pathogenic effects of the infection (Section
B) can be high, barring tumor-resistance genes (Sections A and C).
It ranges between 0 and 90% in different lines of chicken or strains
of inbred mice and averages about 30% in domestic cats. However, outside
the laboratory chronic viremias are very rare and have never been
recorded in humans. They result either from congenital infections
in the absence of maternal antibody (Section A) or from rare, native
immunodeficiency (66).
Thus a predictable tumor risk depends entirely on
high virus expression and virus-induced hyperplasia. This risk can
be reduced or prevented by limiting or blocking lymphoblast hyperplasia
as for example by bursectomy or thymectomy (Section A). Alternatively,
inoculation of newborn AKR mice with antiviral antibody was observed
to suppress viremia and subsequent leukemia in 68% (208). It would
appear more practical, however, to breed or select animals with genes
that confer resistance either to the virus or to tumorigenesis or
to both.
Above all, neither active nor latent viruses offer
targets for tumor therapy, since tumors are not maintained and are
not directly initiated by viral genes, and also occur despite active
antiviral immunity.
Clearly the cell is the more complex variable in the
as yet poorly defined interaction between retroviruses and cells that
leads to hyperplasia and then to carcinogenesis. In view of the evidence
for cellular, transforming genes in viral tumors and for cellular
genes that determine resistance to hyperplasia and tumorigenesis,
further progress in understanding and treating virus-induced cancer
will depend on identifying cellular determinants of carcinogenesis
and the function of hyperplasia and tumor resistance genes.
II. Retroviruses and AIDS
The isolation in 1983 of a retrovirus from a human patient with lymphadenopathy,
a typical symptom of AIDS, led to the proposal that the virus, now
termed lymphadenopathy-associated virus, is the cause of AIDS (26).
Related viruses, termed HTLV-III, ARV, or HIV (209), have since been
isolated from about one-half of the AIDS patients that have been sampled
(210-214). In the United States about 26,000 AIDS cases and 15,000
AIDS fatalities have been reported between 1981, when the disease
was first identified (215), and October 1986 (216). Women represent
only 7% of the AIDS cases in the United States (216). The number of
AIDS cases reported in the United States has increased from about
100 per 6-month period in 1981 to about 5,000 during the last three
6-month periods from January 1985 (216). At the same time the case-fatality
rate has declined from a high of 88% in 1981 to 32% in 1986 (216).
In absolute numbers the known deaths have declined from a high of
2,600 in the first 6 months of 1985 to 1,800 in the first 6 months
of 1986. This suggests either that the virulence of the disease is
dropping or that other diseases were diagnosed as AIDS. Recently the
virus was also suggested to cause disease of the brain and of the
nervous system (230, 255, 268, 274) and lymphoid interstitial pneumonia
(275).
Antibody to the virus is found in about 90% of AIDS
patients and correlates with chronic latent infection by the virus
(217-221). Because of the nearly complete correlation between AIDS
and immunity against the virus, the virus is generally assumed to
be the cause of AIDS (13, 27). Accordingly, detection of antiviral
antibody, rather than virus, is now most frequently used to diagnose
AIDS and those at risk for AIDS (27, 217-224). This is paradoxical,
since serum antibody from AIDS patients neutralizes AIDS virus (225-227)
and since antiviral immunity or vaccination typically protects against
viral disease. It is even more paradoxical that a low antibody titer
is equated with a low risk for AIDS (228, 229).
Unlike all other retroviruses, AIDS viruses are thought
to be direct pathogens that kill their host cells, namely T-lymphocytes
(13, 27), and possibly cells of the brain (230, 255). This view is
compatible with the phenotype of AIDS, the hallmark of which is a
defect in T-cells (13, 27, 215), and with experimental evidence that
many but not all viral isolates induce cytopathic fusion of T-lymphocytes
under certain conditions in vitro (Section D). Further it is
compatible with neurological disease (231, 232, 255). However, cell
killing is incompatible with the obligatory requirement of mitosis
for retrovirus replication (16, 25) and with the complete absence
of cytocidal effects in all asymptomatic infections in vivo
(Section D).
A. Infections with No Risk and Low Risk for AIDS Indicate That the
Virus Is Not Sufficient to Cause AIDS
Since their original discoveries in AIDS patients, the virus and more
frequently antibody to the virus have also been demonstrated in a
large group of asymptomatic persons (212, 214). The virus has been
estimated to occur in about 1 to 2 x 106 or about 0.5 to 1% of all
Americans (223, 224). In the United States persons at high risk for
infection include promiscuous homosexual and bisexual men, of whom
17 to 67% are antibody positive; intravenous drug users, of whom 50
to 87% are positive; and hemophiliacs, of whom 72 to 85% are positive
according to some studies (13, 218, 223). On the basis of this particular
epidemiology, it was concluded that the virus is not transmitted as
a cell-free agent like pathogenic viruses but only by contacts that
involve exchange of cells (13, 27).
In these virus-infected groups the annual incidence of AIDS
was found to average 0.3% (224) and to reach peak values of 2 to 5%
(218, 223, 233). However even in these groups there are many more
asymptomatic than symptomatic virus carriers.
Other infected groups appear to be at no risk for AIDS. In Haiti and
in certain countries in Africa antibody-positive individuals range
from 4 to 20% of the population, whereas the incidence of AIDS is
estimated at less than 0.01% (223, 229, 234). Several reports describe
large samples of children from Africa who were 20 (228) to 60% (221)
antibody positive and of female prostitutes who were 66 to 80% antibody
positive (221,235), yet none of these had AIDS. Among male homosexuals
and hemophiliacs of Hungary about 5% are AIDS virus positive, yet
no symptoms of AIDS were recorded (161). Among native male and female
Indians of Venezuela 3.3 to 13.3% have antiviral immunity, but none
have symptoms of AIDS (236). Since these Indians are totally isolated
from the rest of the country, in which only one hemophiliac was reported
to be virus positive (236), the asymptomatic nature of their infections
is not likely to be a consequence of a recent introduction of the
virus into their population. Thus it is not probable that these infections
will produce AIDS after the average latent period of 5 years (Section
B).
Since the percentage of virus carriers with symptoms
of AIDS is low and in particular since it varies between 0 and 5%
depending on the AIDS risk group of the carrier, it is concluded that
the virus is not sufficient to cause AIDS and that it does not encode
an AIDS-specific function. The virus is also not sufficient to cause
neurological disease, since it has been detected in the brains of
persons without neurological disease and of healthy persons who had
survived transient meningitis (230-232).
Thus the virus appears only rarely compatible with
Koch's third postulate as an etiological agent of AIDS. It may
be argued that the asymptomatic infections reflect latent infections
or infections of only a small percentage of susceptible cells, compared
to presumably acute infections with symptoms of AIDS. However, it
is shown in Section C that infections of neither symptomatic nor asymptomatic
carriers are acute; instead both are equally latent and limited to
a small percentage of susceptible cells.
Further, the observations that some virus carriers
are at high and others at essentially no risk for AIDS directly argue
for a cofactor (218, 237) or else for a different cause for AIDS.
The strong bias against women, because only 2.5% (479 of 17,000 cases)
of the sexually transmitted AIDS cases in the United States are women
(216), is a case in point. The virus-positive but AIDS-negative children
and prostitutes of Africa (221) or Indians from Venezuela (236) are
other examples.
B. Long Latent Period of AIDS Incompatible with Short Latent Period
of Virus Replication
The eclipse period of AIDS virus replication in cell culture is on
the order of several days, very much like that of other retroviruses
(238). In humans virus infection of a sufficient number of cells to
elicit an antibody response appears to take less than 4 to 7 weeks.
This estimate is based on an accidental needle-stick infection of
a nurse, who developed antibody 7 weeks later (239), and on reports
describing 12 (240) and 1 (232) cases of male homosexuals who developed
antibody 1 to 8 weeks after infection. During this period a mononucleosis-like
illness associated with transient lymphadenopathy was observed. In
contrast to AIDS (see below), this illness appeared 1 to 8 weeks after
infection and lasted only 1 to 2 weeks until antiviral immunity was
established. The same early mononucleosis-like disease, associated
with lymphocyte hyperplasia, was observed by others in primary AIDS
virus infections (234). This is reminiscent of the direct, early pathogenic
effects observed in animals infected with retroviruses prior to the
onset of antiviral immunity (Part I, Section B).
By contrast the lag between infection and the appearance
of AIDS is estimated from transfusion-associated AIDS to be 2 to 7
years in adults (220, 223, 241, 242) and 1 to 2 years in children
from infected mothers (220, 223). The most likely mean latent period
was estimated to be 5 years in adults (220, 223). Unexpectedly, most
of the AIDS virus-positive blood donors identified in transfusion-associated
AIDS transmission did not have AIDS when they donated blood and were
reported to be in good health 6 years after the donation (220). Likewise
there is evidence that individuals shown to be antibody positive since
1972 have not developed AIDS (228). Further, 16 mothers of babies
with AIDS did not have AIDS at the time of delivery but three of them
developed AIDS years later (276). This indicates that the latent period
may be longer than 5 years or that AIDS is not an obligatory consequence
of infection.
In view of the claim that the virus directly kills
T-cells and requires 5 years to cause disease, we are faced with two
bizarre options: Either 5 year old T-cells die 5 years after infection
or the offspring of originally infected T-cells die in their 50th
generation, assuming a generation time of one month for an average
T-cell (176). It may be argued that the virus is biochemically inactive
during the first five years of infection and then activated by an
unknown cause. However, AIDS virus is biochemically inactive even
during the acute phase of the disease (Section C). Moreover it would
be difficult for the retrovirus to become acute five years after it
had induced chronic antiviral immunity.
Because of the 5 year latency between infection and
AIDS, the virus has been likened to the lentiviruses (277), a group
of animal retroviruses that is thought to cause debilitating diseases
only after long latent periods (13) (Part I, Section B). However,
recently an ovine lentivirus, the visna or maedi virus of sheep, was
shown to cause lymphoid interstitial pneumonia in 2 to 4 weeks if
expressed at high titer (269). [The same disease is believed to be
caused by AIDS virus in humans (see below)]. Therefore lentiviruses
are not models for retroviruses that are only pathogenic after long
latency (Part I, Section B).
Based on the 5-year latent period of the disease and
on the assumption that virus infection is sufficient to cause AIDS,
one would expect the number of AIDS cases to increase to 1 to 2 x
106 in the United States in the next 5 years. The virus has reportedly
reached its present endemic level of 1 to 2 x 106 in the United States
(223, 224) since it was introduced there, presumably, less than 10
years ago (27). Yet the spread of AIDS from 1981 to 1986 has not followed
the spread of virus with a latent period of 5 years. Instead, recent
statistics (see above) indicate no further increases in the number
of AIDS cases and a significant decline in the number of AIDS fatalities
in the United States (216, 244).
Clearly, the long lag between infection and AIDS and
the large number of virus-positive cases in which as yet no AIDS is
observed, even after long latent periods, lead to the conclusion that
the virus is not sufficient to induce AIDS and does not encode an
AIDS-specific function. Indeed, this conclusion is directly supported
by genetic evidence against a viral AIDS gene. Deletion analysis has
proved that all viral genes are essential for replication (28, 245),
which requires not more than 1 or 2 days, yet AIDS follows infection
only with an average lag of 5 years and even then only very rarely.
C. Levels of AIDS Virus Expression and Infiltration Appear Too Low
to Account for AIDS or Other Diseases
If AIDS viruses were pathogenic by killing susceptible lymphocytes,
one would expect AIDS to correlate with high levels of virus infiltration
and expression, because uninfected cells would not be killed by viruses
nor would unexpressed or latent viruses kill cells. As yet no report
on virus titers of AIDS patients has appeared, despite the record
interest in the epidemiology and nucleic acid structure of this virus
(13, 27, 223). In view of the consistent antiviral immunity of AIDS
patients and the difficulties in isolating virus from them (213),
the virus titers are probably low. Titers have been said to range
between only 0 and 102 per ml blood (213).8
Proviral DNA has been detected in only 15% (9 of 65)
AIDS patients; in the remaining 85% the concentration of provirus,
if present, was apparently too low for biochemical detection (246).
Moreover, among positive samples less than 1 in 102 to 103 lymphocytes
contained the provirus (246). Viral RNA was detected in 50 to 80%
of AIDS blood samples. However, among the positive samples, RNA was
found in only less than 1 of 104 to 105 presumably susceptible lymphocytes
(247). The relatively high ratios of provirus-positive (10-2
to 10-3) to viral RNA positive cells (10-4 to
10-5) of AIDS patients indicate latent infections. Further
there is no evidence that the virus titer or the level of virus infiltration
increases during the acute phase of the disease. It is probably for
this reason that cells from AIDS patients must be propagated several
weeks in culture, apart from the host's immune system, before either
spontaneous (210-214) or chemically induced (248) virus expression
may occur. Further, the AIDS virus is completely absent from the Kaposi
sarcoma (27, 246), which is associated with 15% (216) to 30% (249)
of AIDS cases and is one of the most characteristic symptoms of the
disease.
Similar extremely low levels of virus infiltration
and expression were also recorded in AIDS virus-associated brain disease
(274). Likewise, in interstitial lymphoid pneumonia less that 0.1%
of lung cells expressed viral RNA (275).
Indeed there is evidence that even latent virus may
not be necessary for AIDS, since 85% of AIDS patients lack proviral
DNA (246) and since over 10% of AIDS patients have been observed to
lack antiviral immunity (214, 221, 222, 234). Further, in a study
from Germany 3 of 91 AIDS patients were found to be virus free, based
on repeated negative efforts to detect antibody or to rescue virus.9
It is concluded then that the AIDS virus infects less
than 1%, and is expressed in less than 0.01%, of susceptible cells
both in carriers with or without AIDS. This raises the question of
how the virus could possibly be pathogenic and responsible for immunodeficiency
or other diseases. For instance even if the virus were to claim its
10-4 or 10-5 share of T-cells that express viral
RNA every 24 to 48 h, the known eclipse period of retroviruses, it
would hardly ever match or beat the natural rate of T-cell regeneration
(176).
All other viruses function as direct pathogens only
if they are biochemically active and expressed at high levels. For
instance, the titers that correlate with direct pathogenicity for
avian retroviruses are 105-12 (31, 35, 250)4 and they are 104-7 for
murine retroviruses (12, 38, 40, 42, 251) (Section B). Hepatitis viruses
reach titers of 1012-13 when they cause hepatitis (15), and latent
infections are not pathogenic (83). Further, the very low levels of
AIDS virus expression in vivo are difficult to reconcile with
reports based on in vitro studies with synthetic indicator
genes that the AIDS virus encodes a potent transcription-stimulating
protein (28, 153, 245). Clearly such activators are not at work in
vivo.
The extremely low virus titers of symptomatic and
asymptomatic carriers also explain why infection by the virus in the
United States is essentially limited to contacts that involve transmission
of cells (244) rather than being transmitted as a cell-free, infectious
agent like pathogenic viruses. For instance, among 1750 health care
workers with exposure to AIDS, only 1 or 2 were found to be antibody
positive (252). Another study failed to find a single antibody-positive
person among 101 family contacts of 39 AIDS patients, all of whom
had lived in the same household with an AIDS patient for at least
3 months (253).
D. AIDS Viruses Not Directly Cytocidal
The AIDS viruses are reported to display in culture a fast cytocidal
effect on primary T-cells within 1 to 2 months after infection (13,
27, 254). The cytocidal effect was shown to involve cell fusion (27,
238, 254). The effect is thought to reflect the mechanism of how the
virus generates AIDS after a latent period of 5 years (27, 254).
This is debatable on several grounds: (a) above
all, the in vitro assay cannot account for the large discrepancy
between the short latent period of cell death in vitro and
the 5-year latent period of the disease; (b) T cell fusion
is not observed in vivo in chronic, asymptomatic virus carriers
and not in prospective AIDS patients during the long latent period
of the disease (255), although virus expression is not lower than
during the acute phase of AIDS; (c) T-cell killing is also
not observed in T-cell lines in vitro (27) and not in primary
lymphocytes under appropriate conditions (238). Further primary lymphocytes
infected by AIDS virus were shown to double every 5 days in cell culture
|
