Previous Article | Next Article 
Microbiology and Molecular Biology Reviews, March 2008, p. 157-196, Vol. 72, No. 1
1092-2172/08/$08.00+0 doi:10.1128/MMBR.00033-07
Copyright © 2008, American Society for Microbiology. All Rights Reserved.
Human RNA "Rumor" Viruses: the Search for Novel Human Retroviruses in Chronic Disease
Cécile Voisset,1,
Robin A. Weiss,2 and
David J. Griffiths3*
CNRS-UMR8161, Institut de Biologie de Lille et Institut Pasteur de Lille, Lille, France,1 Division of Infection and Immunity, University College London, London, United Kingdom,2 Division of Virology, Moredun Research Institute, Penicuik, Midlothian, United Kingdom3
SUMMARY INTRODUCTION STRUCTURE AND REPLICATION OF RETROVIRUSES HUMAN ENDOGENOUS RETROVIRUSES: CONFOUNDING FACTORS FOR HUMAN RETROVIRUS DISCOVERY HERVs and Disease HUMAN DISEASES WITH SUSPECTED RETROVIRAL ETIOLOGY Retroviruses in Human Cancer Retroviruses in Human Inflammatory Disease Molecular mimicry. SAgs. Epidemiology LABORATORY METHODS FOR IDENTIFYING RETROVIRAL INFECTIONS Electron Microscopy Detection of Virus Antigens Detection of Serum Antibodies to Retroviruses RT Assays Virus Culture Detection of Retroviral Sequences by PCR Bioinformatics and Genomics in Retrovirus Discovery SPECIFIC CANDIDATE HUMAN RETROVIRUSES Human Mammary Tumor Virus Retroviruses Associated with Multiple Sclerosis HTLV-1. MSRV. HERV-H(RGH-2). Human Intracisternal A-Type Particles IAPs and SS. IAPs and ICL. Human Retrovirus 5 HERV-K HERV-K and testicular cancer. HERV-K and melanoma. HERV-K and IDDM: the superantigen hypothesis. Infectious HERV-K viruses. Betaretrovirus in Bronchioloalveolar Carcinoma Xenotropic MLV-Related Virus and Prostate Cancer HUMAN INFECTION WITH SIMIAN RETROVIRUSES Human Foamy Virus Primate Betaretroviruses HUMAN RNA ''RUMOR'' VIRUSES: PROVING CAUSATION Assessing the Etiological Role CONCLUDING REMARKS ACKNOWLEDGMENTS REFERENCES
 Top NextReferences |
|---|
Summary: Retroviruses are an important group of pathogens that cause a variety of diseases in humans and animals. Four human retroviruses are currently known, including human immunodeficiency virus type 1, which causes AIDS, and human T-lymphotropic virus type 1, which causes cancer and inflammatory disease. For many years, there have been sporadic reports of additional human retroviral infections, particularly in cancer and other chronic diseases. Unfortunately, many of these putative viruses remain unproven and controversial, and some retrovirologists have dismissed them as merely "human rumor viruses." Work in this field was last reviewed in depth in 1984, and since then, the molecular techniques available for identifying and characterizing retroviruses have improved enormously in sensitivity. The advent of PCR in particular has dramatically enhanced our ability to detect novel viral sequences in human tissues. However, DNA amplification techniques have also increased the potential for false-positive detection due to contamination. In addition, the presence of many families of human endogenous retroviruses (HERVs) within our DNA can obstruct attempts to identify and validate novel human retroviruses. Here, we aim to bring together the data on "novel" retroviral infections in humans by critically examining the evidence for those putative viruses that have been linked with disease and the likelihood that they represent genuine human infections. We provide a background to the field and a discussion of potential confounding factors along with some technical guidelines. In addition, some of the difficulties associated with obtaining formal proof of causation for common or ubiquitous agents such as HERVs are discussed.
|
TopPreviousNextReferences |
|---|
Retroviruses are a large and diverse group of human and animal pathogens that cause a wide variety of diseases including many cancers and various immunological and neurological conditions (102, 176, 522). In animals, retrovirus-associated diseases have been known for well over a century. For example, pulmonary adenocarcinoma in sheep was first documented in 1825 (495), and bovine leukosis was described in 1870 (238), while filterable agents were associated with infectious anemia of horses in 1904 (499) and with erythroid leukemia of chickens in 1908 (131). Although it was several decades before the causative agents were identified as retroviruses, studies on animal systems such as these have provided an enormous amount of information on retrovirus biology and revealed some fundamental aspects of the cellular mechanisms of disease, particularly in cancer. The recognition of retroviruses as tumor-inducing agents in animals initially led to their designation as "RNA tumor viruses" (522).
Given the preponderance of retroviruses in animals, much effort has been applied to find related viruses in human disease, and so far, four infectious human retroviruses have been identified (Table 1). The first to be discovered, human T-lymphotropic virus type 1 (HTLV-1), was isolated in 1980 (399) and has been shown to cause cancer (adult T-cell leukemia [ATL]) or neurological disease (HTLV-associated myelopathy/tropical spastic paraperesis [HAM/TSP]) in a small percentage of infections (32). Shortly thereafter, a related virus, HTLV-2, was reported in a patient with hairy cell leukemia (242), although its association with disease remains tentative (14, 423). Subsequently, human immunodeficiency viruses type 1 (HIV-1) and HIV-2, which both cause AIDS, were identified in 1983 and 1986, respectively (33, 101).
|
View this table: [in a new window] |
TABLE 1. Confirmed infectious human retroviruses
|
|
View this table: [in a new window] |
TABLE 2. Putative association of human diseases with retrovirusesa
|
|
TopPreviousNextReferences |
|---|
Retroviruses are enveloped viruses with single-stranded, positive-sense RNA genomes (102, 176, 522) (Fig. 1). The defining features of retrovirus replication are reverse transcription of the viral RNA soon after the infection of a target cell and integration of the DNA product into the cellular chromosomal DNA. These two steps in replication are catalyzed by the virus-encoded enzymes reverse transcriptase (RT) and integrase (IN), respectively. Once integrated, the DNA provirus replicates by directing the synthesis of viral proteins and particles using cellular mechanisms for transcription and translation. Capsid assembly occurs either in the cytoplasm or at the plasma membrane, and new progeny virions are released by budding. The replication strategy of retroviruses is important for their role in disease because reverse transcription and integration confer unique pathogenic mechanisms including insertional mutagenesis and oncogene transduction (365). Previous classification schemes for retroviruses have been based on disease association or morphological features. Currently, seven genera are recognized and distinguished by the genetic relatedness of the RT protein (Table 3) (286).
 View larger version (52K): [in a new window] |
FIG. 1. Retrovirus structure and replication. (a) Genome organization. The RNA and DNA forms of a generalized retrovirus genome are shown with conserved features. R, repeated region at termini of RNA genome; U5 and U3, unique elements close to the 5' and 3' ends, respectively, of the RNA genome; PBS, primer binding site used for initiation of reverse transcription;  , encapsidation signal; PPT, polypurine tract. All infectious retroviruses have at least one splice donor (SD) and one splice acceptor (SA) site used for expression of a spliced transcript encoding env; some retroviruses have additional splice sites. During reverse transcription, the LTR is formed, which contains gene promoter and enhancer elements. At least four genes are present in all infectious retroviruses, gag, pro, pol, and env. Retroviral proteins are synthesized as large polyprotein precursors and later cleaved into the mature viral proteins matrix (MA), capsid (CA), nucleocapsid (NC), protease (PR), reverse transcriptase (RT), and integrase (IN) and into-the-surface (SU) and transmembrane (TM) glycoproteins. Specific retroviruses encode additional proteins with specialized functions in the viral life cycle or pathogenesis. (b) Comparison of proviral structures of MLV and HTLV-1 showing arrangement of ORFs for viral genes. (Panels a and b are adapted from reference 345 with permission from Elsevier.) (c) Structure of a generalized retrovirus particle indicating virus capsid containing two copies of the RNA genome associated with NC protein, viral enzymes, and a cellular tRNA molecule. The capsid is contained within the viral lipid envelope, which is associated with the envelope glycoproteins. (d) Replication. Retroviruses infect their target cells by adsorption to one or more specific cell surface receptors. Binding leads to conformational changes in the envelope and receptor molecules that trigger fusion of the viral and cell membranes. Depending on the specific virus, this may occur at the plasma membrane or within endosomes following endocytosis. Fusion releases the viral core into the cytoplasm (uncoating), and reverse transcription is initiated, during which the single-stranded RNA genome is converted into a double-stranded DNA form. This DNA subsequently becomes integrated into the chromosomal DNA of the cell to form the provirus. The expression of viral genes and proteins requires the host cellular machinery for transcription and translation, although some retroviruses also encode proteins that can regulate these processes. The cellular specificity of expression is dependent on enhancer elements located in the LTR. Assembly of retroviral capsids occurs either in the cytoplasm prior to budding (betaretroviruses and spumaviruses) or at the plasma membrane concomitant with budding (all other retroviral genera). Once released, the retroviral protease is activated, and the viral polyproteins become cleaved into their mature forms. This maturation step is required for infectivity.
|
|
View this table: [in a new window] |
TABLE 3. Classification of retroviruses
|
|
TopPreviousNextReferences |
|---|
Occasionally, retroviruses infect germ cells, and this enables them to colonize the host's germ line DNA by forming endogenous retroviruses (ERVs) (29, 30, 57, 172). This has occurred numerous times during evolution, and ERVs have been detected in all vertebrate species examined. ERVs can be recognized by their sequence similarity with modern-day infectious retroviruses even though millions of years have passed since they integrated. In some host species, exogenous, infectious counterparts are still circulating as infectious viruses, e.g., jaagsiekte sheep retrovirus (JSRV) (377) and mouse mammary tumor virus (MMTV) (2).
In humans, estimates from the genome sequencing project suggest that ERVs now comprise some 8% of our DNA (285), representing around 4,000 proviruses and thousands more solitary long terminal repeats (LTRs) (29). Human ERVs (HERVs) have been divided into three classes, classes I, II, and III, based on their sequence similarity to animal gammaretroviruses, betaretroviruses, or spumaviruses, respectively. Each class contains several families, each representing an independent integration event (172). There is no standard nomenclature for HERVs, but one system refers to the tRNA specificity of the primer binding site used to initiate reverse transcription (Fig. 1). Thus, HERV-K would use lysine and HERV-H would use histidine if they were replicating viruses. There are no HERVs related to lentiviruses or deltaviruses. Although some very short regions of sequence similarity are present in human DNA (218, 383), they appear unlikely to represent genuine ancestral infections by these retroviruses (492). Note, however, that ERVs related to lentiviruses have recently been detected in rabbits (248).
In some animal species such as mice (468), chickens (222), and pigs (380), a few ERV proviruses have retained the ability to encode replication-competent viruses. Such viruses, while transmitted in germ line DNA as endogenous elements, are also capable of infectious transmission and can therefore behave as infectious viruses, infecting either the same species or other species. An example is the endogenous feline retrovirus RD114, which was identified after it infected human rhabdomyosarcoma cells that had been passaged through the brain of a fetal kitten and which was originally thought to be a human virus (322). Similarly, foreign tissue grafts in nude mice are commonly infected by endogenous murine viruses (5, 108, 416, 486). In contrast to these animal ERVs, none of the HERVs characterized so far are capable of producing infectious particles. In vitro transmission has been reported for some HERVs (e.g., see references 97 and 354), but further work is needed to clone the infectious genomes (discussed below).
The majority of HERV proviruses contain mutations that prevent the expression of viral particles. Despite this, transcription of HERV RNA occurs in many normal human tissues and cultured cells (326, 395, 438, 464). In a few instances, viral proteins or particles may be produced (Table 4). The level and pattern of expression may be modulated under pathological conditions, particularly in cancer and inflammatory disease (293, 529). There has been considerable debate as to whether HERV expression has pathogenic consequences, and HERVs have been proposed to be etiological cofactors in numerous diseases on the basis of increased RNA and protein expression (30, 293, 498). However, as yet, there is no conclusive proof that HERVs have a causative role in any disease (discussed below).
|
View this table: [in a new window] |
TABLE 4. HERV proteins expressed in tissues and cultured cellsa
|
|
View this table: [in a new window] |
TABLE 5. Potential mechanisms for HERV-induced disease
|
|
TopPreviousNextReferences |
|---|
The association of retroviruses with cancers and inflammatory disorders of animals naturally led to a search for related agents in analogous human diseases. A great deal of evidence has now accumulated to suggest that retroviruses other than HIV and HTLV may be human pathogens. However, much of this evidence is circumstantial, and we argue that a strong case has yet to be made for an etiological role for any specific candidate retrovirus. In a few instances, nucleotide sequences from putative retroviruses have been cloned from human tissues by PCR (94, 189, 246, 385, 514, 538), but these have so far all been proven to be closely related to ERVs of either human or animal origin.
As noted above, retroviruses were first discovered as tumor-inducing agents in animals (131, 424). Early studies in mice and chickens defined two general mechanisms by which retroviruses induce tumors; oncogene capture and insertional mutagenesis (102, 176, 365, 501). Oncogene capture involves the insertion of cellular proto-oncogene sequences into the retroviral genome by recombination during reverse transcription. The resulting virus is usually defective and requires coinfection with a helper (wild-type) virus for transmission. Oncogenesis is a result of the constitutive expression of the captured oncogene under the control of the viral LTR. Such viruses are known as acute-transforming viruses, reflecting the rapid growth of tumors in animals in which they arise. Insertional mutagenesis arises by the chance integration of retroviruses at a site in the host DNA adjacent to a cellular proto-oncogene. The promoter and enhancer elements in the viral LTR then activate the expression of the oncogene, thereby enhancing cellular proliferation, which may ultimately lead to the development of a tumor. Viruses that induce tumors in this way are replication competent and are known as cis-activating retroviruses. In this case, the tumors generally take much longer to develop than with acutely transforming viruses.
In addition to these two well-characterized oncogenic mechanisms, a small number of retroviruses that encode their own oncogenic protein or that directly stimulate cells through signaling motifs contained within the Env proteins have been described. Examples include the Tax protein of HTLV-1, which promotes cellular proliferation by activating the expression of a number of cellular genes (184), and the Env protein of JSRV, which activates signaling pathways including the MEK/extracellular signal-regulated kinase and Akt protein kinase cascades (288). Thus, retroviruses have a variety of strategies for inducing cellular proliferation and cancer, and novel human retroviruses might use any of these, or they might use entirely new strategies.
The discovery of HTLV-1 and its role in ATL confirmed that retroviruses can be oncogenic in humans. HIV-1 is indirectly implicated in cancer by creating an immunosuppressive environment that permits the growth of opportunistic tumors (65). In addition, studies on B- and T-cell lymphomas in AIDS patients have suggested that oncogene activation by insertional mutagenesis might be another mechanism by which HIV-1 could cause cancer (206, 448), although direct tumorigenesis by HIV in this way appears to be very rare.
While both HIV and HTLV can be oncogenic, neither is closely related to the large groups of well-characterized oncogenic gammaretroviruses and alpharetroviruses of animals. It has been suggested that human cancers could potentially be caused by the cross-species transmission of these retroviruses from animals (547). Such zoonotic viruses may not necessarily induce tumors in their natural host or replicate efficiently in human cells. Numerous studies in the 1960s and 1970s searched for retroviruses in human cancers (245, 523). In some cases, the viruses identified turned out to be cell culture contaminants of animal viruses, such as HL-23 virus, later found to be gibbon ape leukemia virus, and "HeLa" virus, which was actually Mason-Pfizer monkey virus (MPMV) (523). The provenance of some other viruses is still unexplained. More recently, PCR has been used to search for retrovirus sequences expressed in human tumors, and these have been found either to be HERVs (69, 82, 294, 440) or to be closely related to animal viruses (350, 414, 497, 514). HERVs have been implicated in a variety of tumors including melanoma (82, 354), germ cell tumors (30), breast cancer (516), and leukemia (66, 339), but it is unclear whether the increased expression of these elements precedes the cancer or whether it is a result of altered gene regulation in the tumor cells.
As with studies of cancer, a search for retroviruses in human inflammatory diseases followed the description of related conditions during infection with some animal retroviruses. For example, murine models for glomerulonephritis, a feature of systemic lupus erythematosus (SLE), were previously thought to be triggered by immune complexes containing retroviral antigens (158). Similarly, small ruminant lentiviruses cause a chronic inflammatory disease that can involve several tissues including lung, brain, and joints (528). In addition, transgenic mice expressing human retroviral proteins such as HTLV-1 Tax and Env exhibit inflammatory symptoms reminiscent of human disease (185, 233). However, the tissues involved and the severity of the inflammation that occurs in these models depend to some degree on the exact structure of the transgene, so it is difficult to be certain how relevant such models are to human disease.
HTLV-1 and HIV-1 can both cause inflammatory symptoms, and this has reinforced the concept that other retroviruses might have a role in human inflammatory disease. A subgroup of individuals infected with HIV-1 develop a salivary gland inflammation similar to that seen in Sjögren's syndrome (SS), known as diffuse inflammatory lymphocytosis syndrome (229). Additional features of inflammatory disease such as autoantibody production, arthropathy, and vasculitis also occur in patients infected with HIV-1 (reviewed in references 163 and 249). Inflammatory reactions in HTLV-1 infection are even more striking, and while those of HAM/TSP are the most clinically overt, HTLV-1 is also associated with SS, arthropathy, uveitis, polymyositis, and myelitis in up to 5% of infections (367).
These clinical observations and animal models demonstrate that human retroviruses can cause inflammatory reactions and have led many workers to investigate other groups of patients for evidence of retrovirus expression (reviewed in references 103, 163, and 498). In addition, a number of models have been proposed to describe how retroviruses might trigger autoimmunity. These include general mechanisms such as lymphocyte activation and the upregulation of major histocompatibility complex (MHC) molecules and proinflammatory cytokines. However, the potential direct effects of retroviral proteins acting through molecular mimicry or as superantigens (SAgs) have received the greatest amount of attention (154, 163, 263, 498).
Molecular mimicry. The concept of virus pathogenesis due to molecular, or antigenic, mimicry has been around for several decades and is characterized as an immune response to an infectious agent that cross-reacts with a host antigen (reviewed in reference 154). Despite the establishment of immune tolerance by the removal of self-reactive T lymphocytes during thymic maturation, it is clear that some self-reactive T cells persist in healthy individuals (243). Similarly, self-reactive B lymphocytes are also present. Molecular mimicry implies that these self-reactive T cells are activated in susceptible individuals following infection with a pathogen encoding a protein with a shared epitope. Subsequently, additional self-reactive lymphocytes might then be activated through epitope spreading, thereby exacerbating the autoimmune pathology (154). A number of diseases represent good candidates for initiation by molecular mimicry; however, while there is tantalizing evidence to support these examples, at present, they remain unproven (40, 154).
For retroviruses, the development of computer programs for comparing protein sequences led to a number of predicted epitopes shared between viral and host proteins, including some known autoantigens (160, 163, 320, 355, 378, 403). Epitopes identified in this way do not always prove to be biologically significant (154, 355), but for some retroviruses, the presence of cross-reacting antibodies has been established in a number of studies. For example, the CA (p30) proteins of mammalian gammaretroviruses such as feline leukemia virus and murine leukemia virus (MLV) share common epitopes with the U1 small nuclear ribonucleoprotein-associated 70K autoantigen (403) and DNA topoisomerase I (320), which are autoantigens in SLE and systemic sclerosis, respectively. Molecular mimicry between a C-terminal epitope of the HTLV-1 Tax protein and a neuron-specific ribonuclear antigen has also been demonstrated (283). Patients with HAM/TSP have antibodies that bind to this epitope on both proteins, as do monoclonal antibodies to the Tax peptide. Similarly, there is antigenic mimicry between the HIV-1 TM (gp41) protein and human leukocyte antigen class II molecules (178). Antibodies recognizing the cross-reactive epitopes are predicted to contribute to the functional impairment of CD4+ T lymphocytes in HIV patients (179).
These findings provide a basis for retrovirally induced autoimmunity through antigenic mimicry, although whether such mimicry extends to T-cell-mediated autoimmunity is currently unclear. Due to their similarity with exogenous retroviruses, HERV proteins have been cited as being potential autoantigens (103, 163, 498). While it is true that some individuals (with and without disease) do have T cells and antibodies that recognize HERV proteins (24, 60, 159, 208, 404), currently, no HERV protein is a recognized autoantigen in any disease.
It is worth mentioning that an alternative outcome of antigenic mimicry is that the cross-reactive epitope on a pathogen may be recognized as "self" by the immune system, effectively creating a "hole" in the immune response. An example is the HIV-1 TM (gp41) protein, some epitopes of which mimic sites on phospholipids such as cardiolipin and phosphatidylserine (201). Since these are ubiquitous host antigens, antibodies to these epitopes are rarely produced in HIV-1-infected individuals. Such mimicry may therefore be responsible for partially protecting these cross-reactive HIV epitopes from immune response, thereby impairing neutralization of the virus and contributing to pathogenesis.
SAgs. SAgs are a class of immune-stimulating proteins encoded by some bacteria and viruses that activate T lymphocytes in a non-antigen-restricted manner by interacting directly with the Vβ chain of the T-cell receptor (90, 315, 526). This results in massive polyclonal T-cell activation and cytokine release. The activation of T cells in the absence of specific antigen may lead to anergy, and so the longer-term consequences of SAg activity can be the peripheral depletion of a specific T-cell Vβ subset and/or local tissue proliferation of the same T-cell subset (2). This ability to dysregulate the immune system presents SAgs as potential mediators of inflammatory and autoimmune disease.
SAgs have been described in a number of bacteria, e.g., Streptococcus pyogenes and Staphylococcus aureus, where they cause diseases such as toxic shock syndrome and food poisoning (289). Several viruses have been proposed to encode SAgs, but the best-characterized example is the SAg of MMTV, which has a central function in virus dissemination in the early stages of infection (1, 2, 90). SAg activity has also been linked with HIV, rabies virus, the human herpesviruses Epstein-Barr virus (EBV), and human cytomegalovirus, but in these cases, the SAg genes and proteins have not been identified (534). In addition, the presence of a SAg-encoding retrovirus has sometimes been invoked to explain inappropriate T-cell activation in autoimmune diseases where there is evidence for the deletion of specific Vβ T-cell subsets (104, 163, 400). The controversial description of a SAg encoded by a specific HERV-K provirus (105, 463, 472) has provided a model by which various unrelated viruses might induce SAg activity on infection (discussed below). Interestingly, in the MMTV system, SAg activity and T-cell activation do not commonly elicit autoimmune inflammatory symptoms.
Novel human viruses that have been discovered in recent years have been found in diseases where there was good epidemiological support for an infectious cause (250). For example, in the mid-1980s, it was obvious that an additional infectious agent was present in non-A, non-B hepatitis, and this justified the search that led to the discovery of hepatitis C virus (91). In contrast, many of the chronic immunological and neurological conditions that have been linked to retroviral infection do not have strong epidemiological evidence for a simple contagious etiology. Instead, these diseases are proposed to have a multifactorial etiology requiring the interaction of a number of genetic and nongenetic causative factors, with infection representing just one environmental component (e.g., see references 173, 181, and 200). Therefore, disease may be a rare outcome of infection in susceptible individuals. For some diseases, this model is supported by studies of monozygotic twins and patterns of disease incidence in migrant populations (265). Although infectious agents represent attractive targets for environmental factors, their role remains speculative in the absence of a clearer understanding of how all the contributory factors are linked. Interestingly, for a number of diseases where an infectious retrovirus has been sought, HERVs (i.e., genetic factors) have been identified (94, 105, 336, 384).
For diseases where there are no epidemiological data supporting a role for a virus, a "hit-and-run" mechanism in which acute infection with a specific virus elicits a chronic pathological response that persists after the original infection has been cleared might be proposed. Possible mechanisms include the activation of autoreactive T cells by antigenic mimicry (154) and the initiation of tumorigenesis that no longer requires the presence of the viral oncogene once the tumor is established (10). There is some experimental evidence supporting hit-and-run mechanisms, e.g., in adenovirus and herpesvirus transformation (364, 445) and in the development of chronic inflammatory disease following paramyxovirus infection (216); however, as yet, there are no confirmed examples in human disease. Subacute sclerosing panencephalitis in measles virus infection might perhaps be proposed as one example, although a defective form of the virus persists in the brain. Gastric adenocarcinoma induced by Helicobacter pylori is another possible example, but this too can persist. It is currently unclear whether hit-and-run mechanisms apply to retroviral infections. Because retroviruses integrate into the host DNA and therefore persist for the lifetime of that cell, the likelihood of them eliciting disease by a hit-and-run mechanism appears to be lower than that for other microbes.
|
TopPreviousNextReferences |
|---|
A wide range of techniques has been used to identify and characterize novel retroviruses. In addition to traditional methods such as cell culture, electron microscopy (EM), and serological assays, RT activity has been exploited as a biochemical marker of retroviral infection and is routinely used to quantify retrovirus production in vitro. The introduction of PCR provided new opportunities for the identification and analysis of virus infections. In addition, advances in bioinformatics and microarray technologies have provided further novel and potentially high-throughput approaches for new virus discovery. The relative merits and disadvantages of these techniques were reviewed elsewhere previously (303, 523). Here, we describe some aspects of particular relevance to novel retrovirus discovery and highlight problems that can arise due to the presence of ERVs.
EM is of great value for directly visualizing candidate novel viruses in human tissues and cultured cells. Retroviruses have been grouped into four morphological types, denoted A-, B-, C-, and D-type particles (42, 111), and there have been numerous reports of retroviral infection in humans based on EM evidence (Fig. 2). Putative virions have been described in tissues or blood from patients with SS (161, 540), multiple sclerosis (MS) (194, 390), breast cancer (137), psoriasis (230), myeloproliferative diseases (66), malignant melanoma (25, 50), benign osteopetrosis and sporadic para-articular osteoma (267), and thymomas (16). Similarly, retrovirus-like particles (RVLP) have been found in body fluids from patients with rheumatoid arthritis (RA) (469), SLE (480), and epidemic neuropathy (418) and in cell cultures established from individuals with testicular tumors (46, 71) and chronic fatigue syndrome (192). In addition, RVLPs have also been observed in healthy human tissues and fluids including placenta (241, 504) and breast milk (138, 338) and in cultured cells (126, 328, 348). Although reported as being retroviruses, such particles do not always have the expected morphology or size of retroviruses; for example, particles described in RA (469) have an apparent diameter of 200 nm, which is roughly double that expected for retroviruses, and their identity as such is therefore doubtful.
 View larger version (80K): [in a new window] |
FIG. 2. Retrovirus-like particles described in diseased human tissues and cultured cells. (a) LM7 particles from leptomeningeal cells from MS induced by ICP0 protein of herpes simplex virus type 1. (Reprinted from reference 393 with permission of the publisher.) (b) Particles in cultured lymphocytes from MS. (Reprinted from reference 194 with permission from Elsevier.) (c) Particles in SS salivary gland (see arrows). (Reprinted from reference 540 with permission of the publisher.) (d) HICRV in ICL. Bar, 0.5 µm. (Reprinted from reference 193 with permission.) (e) Virus-like particles in human milk. (Reprinted from reference 429 by permission from Macmillan Publishers Ltd.) (f) Particles in PBC. (Reprinted from reference 538 with permission of the publisher. Copyright 2003 National Academy of Sciences, U.S.A.) (g) Particles in myeloproliferative disease. Bar, 100 nm. (Reprinted from reference 66 with permission from Elsevier.) (h) HERV-K in teratocarcinoma-derived cell lines labeled with gold anti-HERV-K Gag. (Reprinted from reference 61 with permission from Elsevier.)
|
100 nm diameter, if cut transversally, might appear as a 100-nm-diameter extracellular RVLP. Another potential disadvantage of using EM for virus discovery is that this method cannot detect latent retroviral infection where few or no particles are being produced. Indeed, the technique is generally rather insensitive, and several hours of scrutinizing sections may be required to identify virions even for a moderately productive infection. Despite these issues, micrographs of sections taken directly through a retrovirus particle can be very convincing, particularly if they are stained with gold-labeled antibodies to the suspected virus. This approach has been used successfully to confirm the identity of HERV-K particles produced by teratocarcinoma cell lines (61). A similar approach may be successful in determining the identity of RVLPs in other human cell lines and tissues under either pathological or normal physiological conditions (339).
Several investigators have attempted to detect the presence of retroviral proteins in human tissues by immunohistochemistry using antibodies raised against other retroviruses. A number of studies in the 1970s found antigens related to mammalian betaretroviruses and gammaretroviruses in healthy human tissues, in some cancers, and in inflammatory diseases (reviewed in reference 523). More recent examples include an antigen that is reactive with a monoclonal antibody to the HTLV matrix protein (MA) (p19) in salivary gland tissue from a patient with SS (444) and an antigen in some human lung adenocarcinomas that is reactive with an antiserum to the CA protein of JSRV (115). It has been hypothesized on the basis of such data that a related antigen, and possibly a retrovirus, is present. While it is true that cross-reactive epitopes are present in Gag and (less commonly) Env proteins, the significance of such reactivity is unclear unless the antigen is purified or otherwise identified, and this has generally not been achieved. Therefore, while the detection of cross-reactive antigens with antiretroviral antibodies can provide support for data obtained by other means, this technique is not very informative when used alone. Recently developed proteomics technologies that facilitate the identification of peptides and proteins isolated from complex mixtures may permit the antigens detected in this way to be characterized more rapidly in the future.
The detection of serum antibodies to viruses is a well-established method for determining prior or current infection. Serological assays can be employed either in large cross-sectional studies to monitor the prevalence of infection in a population or in longitudinal studies to track the humoral immune response through the course of infection. For genuine human retroviral infections, this has been applied very successfully, and patients infected with HTLV or HIV generally produce strong antibody responses to Gag and Env proteins and to regulatory proteins such as Tat and Tax. Antibodies to Pol proteins are also common (269, 292). While many retrovirus infections of animals also elicit strong humoral responses, some do not (217, 374), and for a putative novel human retrovirus, it is therefore not possible to predict with certainty whether antibodies would be produced in infected individuals, especially if the virus is closely related to a HERV.
Several reports described antibodies to animal retroviruses in humans, e.g., to MMTV in breast cancer patients (530), to MLV in patients with psoriasis and in healthy individuals (335), and to bovine leukemia virus in blood donors (78) (also see reference 523). However, the epitopes responsible have been characterized in relatively few studies, and in some of the older reports, this seroreactivity turned out to be to carbohydrate antigens present on the viruses due to their production in nonhuman cell lines (31, 455). Antibodies to HERV proteins have also been described, particularly in patients with testicular tumors (60, 430) and melanoma (82), but such antibodies can also be present in healthy individuals (24, 208).
Antibodies reactive with HIV or HTLV Gag (CA, p24) are common in patients with autoimmune diseases such as MS, SS, and SLE (369, 407, 477, 478). In general, anti-Env antibodies are not present in these individuals, although sera from patients with mixed connective tissue disease are reported to contain neutralizing antibodies that block HIV infection (127). These groups of patients have no other markers that would indicate a genuine infection with HIV or HTLV, so it has been proposed that this seroreactivity reflects the expression of another, cross-reactive retrovirus, which might be exogenous or endogenous. Alternatively, such antibodies could be elicited by cross-reactive epitopes on nonretroviral host antigens such as ribonucleoprotein antigens, or they may simply represent low-affinity antibodies generated by nonspecific B-cell activation, a common feature of some autoimmune diseases.
Reactivity to a single retroviral protein provides only weak evidence of the presence of a retrovirus since this could be due to low-affinity "nonspecific" binding. However, reactivity to multiple viral proteins or to multiple epitopes on the same protein suggests that a viral antigen is actually driving antibody production and gives greater support to the case for infection. Additional characterization of antibody reactivity can also be persuasive in favor of infection. In the Borna virus system, human infection was cast into doubt by suggestions that positive detection was due to PCR contamination (461). Subsequently, epitope-mapping studies found low titers but high-avidity antibodies in patients with schizophrenia, and this raised the profile of human infections again (48). Approaches such as this might be useful if applied to putative retroviral infections. Similarly, longitudinal studies of antibody titer, if correlated with disease severity, can provide additional support for an association between virus replication and disease (257).
Although antiretroviral antibodies have been described by several groups, cell-mediated immune responses to retroviral antigens in autoimmune patients have been examined only very recently. This is surprising since inflammatory reactions in some of the diseases, notably RA, MS, and psoriasis, are thought to be largely T-cell driven. One report described T cells that are specific for HERV-K peptides in patients with seminoma and in healthy individuals (404), but it is not yet clear whether this response has any physiological significance in disease outcome.
Since its discovery in 1970, the demonstration of RT activity has often been used as a biochemical marker for retroviral infection (523). Initial assays monitored the synthesis of radiolabeled DNA using synthetic homopolymeric primer-template complexes, and it is possible that these assays did not adequately distinguish retroviral RT activity from that of cellular enzymes such as DNA-dependent DNA polymerases (298, 509) and terminal deoxynucleotidyl transferase (87). These early assay formats were also used to discriminate the RT activity of different groups of retroviruses; for example, gammaretroviruses preferentially use Mn2+ over Mg2+ as a cation, while alpharetroviruses and lentiviruses work more efficiently with Mg2+ (503). Modified versions of this assay, such as the "simultaneous detection" of viral 70S RNA, were devised to provide additional evidence that the detected activity was retroviral in origin (432), although this assay was not adopted as a universal standard. In the context of retrovirus discovery, RT assays have been important because the detection of genuine RT activity provides a basis for additional characterization of the putative agent; e.g., the detection of RT activity in T-cell cultures from patients with lymphadenopathy and ATL was a significant step in the discovery of HIV-1 (33) and of HTLV-1 (399).
In the PCR age, RT assays have been updated. The polymerase-enhanced RT (PERT) assay is an RT-PCR-based technique that detects the presence of RT activity with up to 106-fold-greater sensitivity than conventional assays (402, 450) and is reported to be capable of detecting a single particle per assay (450). False-positive results due to cellular DNA polymerase activity were initially an issue here also, but recent adaptations have largely overcome this problem (298, 509). PERT assays have been employed to detect RT activity in diseases such as SS (189), MS (99, 195), and motor neuron disease (11). Positive results could be taken as evidence of the presence of a retrovirus, but so far, none of these examples has been confirmed as a genuine infectious virus. In some studies (11, 99, 189, 328, 385), RT assays have been performed on virus preparations that have been purified on sucrose density gradients or by ultracentrifugation, procedures thought to remove cellular contaminants. However, such preparations may also contain high levels of cellular material (Fig. 3) (also see reference 338). Because there are several cellular sources of genuine RT activity, including HERVs, non-LTR retrotransposons (41, 121, 319), and telomerase (87), the detection of RT activity alone cannot be relied upon to formally prove the presence of a retrovirus. Conversely, the absence of such activity in a specimen does not necessarily mean that a retrovirus is absent because it may be present in latent form.
 View larger version (105K): [in a new window] |
FIG. 3. Electron micrographs of cell supernatants purified by sucrose density gradient centrifugation. Culture supernatants from EBV-transformed human B lymphocytes (a, b, d, and e) and HTLV-1-infected MT-2 cells (c and f) were concentrated by ultracentrifugation and then recentrifuged through a 10 to 60% sucrose density gradient. Fractions with a density typical of retroviruses (1.15 to 1.18 g/ml) were pooled, fixed in 2.5% paraformaldehyde-0.4% glutaraldehyde, and embedded in Epon resin. Ultrathin sections were stained with 1% uranyl acetate for 1 h and 1% lead citrate for 4 min and analyzed by transmission EM. Multiple vesicular particle-like structures can be observed. The origin of these structures is unclear, although they may be derived from cellular components such as polyribosomes, exosomes, and apoptotic blebs. Alternatively, because these cells were transformed with EBV, it is also possible that these structures are viral in origin (470). Panel f shows immunogold labeling of a Unicryl-embedded section with an anti-HTLV CA antiserum. Bar, 200 nm (C. Voisset, B. Mandrand, and G. Paranhos-Baccalà, unpublished data).
|
The ability to grow a virus in cultured cells in vitro confirms that it is a genuine infectious replicating agent, and the rescue of infectious virus from disease tissue can provide convincing evidence of its existence as a real entity. On the other hand, there are several well-documented examples of human cell lines that have become contaminated with animal retroviruses. These viruses, such as HL26 virus, HeLa virus (see above), and ESP-1 virus (actually MLV), were some of the first contributions to the list of human RNA rumor viruses (reviewed in reference 523). Virus culture has been of enormous value for human retrovirology, as both HTLV-1 and HIV-1 were first identified in cultured lymphocytes from infected patients (33, 399). In contrast, some candidate human retroviruses have so far proved to be recalcitrant to culture, usually producing so little virus that it has been difficult to convincingly demonstrate transmission (97, 112, 161, 189, 388, 538). The lack of transmissibility in vitro could have numerous explanations but may suggest that the viruses are in fact defective particles encoded by HERVs, which could be activated by the conditions in the culture system, for example, in mixed cell cultures where cell lines are exposed to primary tissue biopsies that may express an undefined profile of cytokines. Indeed, several HERVs have been shown to be upregulated by proinflammatory cytokines (239, 311).
A lack of transmissibility in culture may also be due to the lack of a suitable cell substrate, and a judicious (or fortuitous) choice of host cell is therefore important. Failure of replication can occur for many reasons, e.g., the absence of an appropriate surface receptor, transcription factor, or other required cellular component (reviewed in reference 175). Alternatively, infection may be restricted by the presence of dominant cellular factors that interfere with viral replication. Recent work has led to the characterization of a number of these restriction factors in mammalian cells, which can exhibit cell type or species specificity or be selective against particular retrovirus strains (reviewed in reference 177). Such proteins include members of the APOBEC family, which target reverse transcription and result in highly mutated virus genomes, and TRIM (tripartite motif) proteins, which interfere with virus trafficking at an early postentry stage. Some restriction factors in sheep and mice are derived from ERVs (44, 349). Additional restriction factors are also present in mammalian cells (433) but require further characterization. A more detailed understanding of retroviral restriction factors and their interactions with other mechanisms of innate immunity to infection, such as interferon (IFN) response pathways, may lead to improved culture systems for those viruses that do not currently grow well in vitro.
PCR is a powerful technique for studying novel viral infections. The combination of sensitivity and specificity allows the detection of rare viral sequences amid a large excess of host nucleic acid, and the ability to sequence the amplified fragments has permitted numerous studies on molecular epidemiology and phylogeny. PCR can be used in the initial stages of identification of a retroviral genome and in later studies of molecular epidemiology to determine the degree of association with disease. A crucial advantage of PCR over the methods described above is that the amplicons can be sequenced, and the identity of the product can be unequivocally determined.
A variation of PCR for pathogen discovery is the use of consensus (degenerate) primers targeted at sequence motifs conserved between several microbes, which allows researchers to search for related sequences in diseases with a suspected infectious etiology (250). This general strategy has been used successfully to identify a number of novel microbes including bacteria, e.g., Tropheryma whippelii (409), and viruses, e.g., herpesviruses (413) and papillomaviruses (197). Several groups have described degenerate oligonucleotides that are applicable for identifying retroviral genomes (Table 6). Some of these primer sets target the highly conserved pol gene and can detect a broad range of retrovirus genera, but others are selective for a specific retroviral genus, allowing reduced degeneracy and higher sensitivity in the amplification reaction.
|
View this table: [in a new window] |
TABLE 6. Degenerate PCR primers based on conserved motifs in retroviruses
|
In comparison with other virus families, the identification of a new human retrovirus by PCR presents particular difficulties due to the high number of HERV sequences in the human genome. Where genuinely new retroviral sequences are obtained, care must be taken that they do not result from ERVs, either human (385) or nonhuman (190). This is particularly true for degenerate primer PCR because HERVs share the same conserved motifs as exogenous retroviruses. Therefore, while this has been a powerful technique for the characterization of specific ERV families in several species (316, 380, 447), putative exogenous retroviral sequences cloned in this way from diseases such as MS, diabetes, and SS have later also been found to be endogenous retroviral sequences (105, 189, 385). To address this issue, primer sets have been devised for betaretroviruses, gammaretrovirus, and deltaretroviruses that are specifically designed to exclude the amplification of HERVs (80). The use of these primers in attempts to identify novel retroviruses in human lymphoma has so far been unsuccessful (80), but their use in other disease contexts may yet be fruitful. However, should an exogenous retrovirus closely related to a HERV be circulating in human populations, it would of course be excluded by these primers.
One refinement to the degenerate PCR strategy has been to purify and concentrate retrovirus virions prior to degenerate RT-PCR (66, 96, 105, 189, 246, 379, 385, 452). Purification may be achieved by centrifuging the homogenized sample through a density gradient such that retroviruses migrate to their typical buoyant density of between 1.16 and 1.18 g/ml. These gradients are traditionally prepared with sucrose, but other compounds such as iodixanol can also be employed (337). The purpose of the gradient procedure is to physically separate encapsidated viral RNA (which migrates through the gradient) from unpackaged soluble RNA (typically defective HERV RNA), which is not expected to enter the gradient to an appreciable extent. Similar purifications have been employed prior to the detection of RT activity using the PERT assay (11, 99, 189, 328). Nuclease treatment can also be used to degrade extraneous cellular DNA and RNA prior to amplification (164, 440), because in theory, encapsidated RNA genomes should be protected.
The purification of retroviruses on density gradients is thought to remove cellular contaminants, and it has generally been accepted that any sequence amplified from a gradient was packaged inside purified virions. However, it is unclear whether this is a valid assumption because material migrating at 1.16 to 1.18 g/ml might also include polyribosomes, microsomal vesicles, or other cellular components, which can migrate through the gradient (Fig. 3). Large complexes such as these could potentially protect cellular nucleic acids (including HERV sequences) from nuclease digestion. Thus, to confirm specificity, density gradients of control tissues and cultures must be analyzed in parallel. The choice of control tissue may also be important in this respect. Cell death in disease lesions or cultured cells can result in the release of large quantities of free DNA and RNA into the extracellular environment, providing a template for amplification by degenerate primers. Therefore, to determine disease specificity, the ideal control tissues for such experiments would have a similar level of cell death and not simply be "normal" tissues.
Computerized analysis of virus sequence data has long been an essential tool in the characterization of viruses (428). However, the availability of whole-genome sequence databases now means that bioinformatics methods can be used directly in the identification of new retroviral elements without any laboratory experimentation. For example, a novel family of endogenous gammaretrovirus sequences was identified by sequence analysis of murine genomic and expressed-sequence-tag databases (70). This was achieved entirely in silico with the BLASTN program using MLV as a query sequence. Other groups have devised algorithms for interrogating sequence databases by searching for sequence motifs that are conserved among retroviral elements (236, 323, 505), and these may be useful in identifying novel human retroviruses in sequence data.
The availability of the complete human genome sequence is also helping to better define the repertoire of HERVs. Turner and coworkers described two HERV-K proviruses, HERV-K113 and HERV-K115, which are present in around 30% and 15% of humans, respectively, and are therefore insertionally polymorphic alleles (494). The HERV-K113 provirus is of particular interest since it contains complete open reading frames (ORFs) for the viral proteins and may therefore be replication competent (although the Env protein appears to be nonfusogenic in vitro) (119). Whether the presence of either provirus is associated with any specific disease can be assessed by PCR of the integration sites on patient DNA, since the flanking sequences are known (79, 346, 375). In addition, since both proviruses were first identified in the genome of a single individual, it appears likely that additional polymorphic HERV-K proviruses exist in humans, and this is being confirmed as the genomes of more individuals are analyzed (38, 221, 300). Because polymorphic HERVs are more likely to have recently integrated, these are perhaps the most likely HERV proviruses still to be biologically active and pathogenic.
The approaches described above are useful for characterizing ERVs, but they are not generally applicable for detecting infectious exogenous viruses present in only a small population of somatic cells. However, the human genome sequence has also provided new strategies for the discovery of infectious viruses by facilitating the rapid identification of sequences obtained from large-scale "shotgun cloning" experiments. Weber and colleagues have described a general method for identifying novel virus genomes in which sequences obtained from randomly sequenced cDNA libraries are searched using the BLAST algorithm against sequence databases to filter out all host sequences, known infectious agents, and artifactual sequences (519). The few sequences remaining after this "computational subtraction" can be reasonably assumed to represent novel candidate infectious agents and further validated for their function in disease using laboratory methods. So far, this technique has been successfully applied only in control experiments, and a limitation is that it is heavily dependent on sequencing, requiring upwards of 10,000 clones to be sequenced to have a reasonable chance of detecting a rare cDNA (539). However, related approaches that include an enrichment step in which putative viruses are first concentrated by ultracentrifugation and then treated with nucleases to deplete the pellets of extraneous nonviral nucleic acid have been described. Sequence-independent PCR ("random" PCR) is then applied to amplify any remaining DNA and RNA, which is characterized by shotgun cloning, sequencing, and database comparison. This strategy has been used to identify novel human parvoviruses, coronaviruses, and polyomaviruses in human tissue samples (8, 9, 167, 240, 500). Figure 4 summarizes the various approaches that might be taken. Application of these techniques to other diseases could deliver further new viruses, possibly including retroviruses, although an exogenous retrovirus with high similarity to HERVs might be excluded by the screening procedure.
 View larger version (26K): [in a new window] |
FIG. 4. Identification of novel viral sequences. A generalized scheme summarizing the approaches taken by a number of groups is shown (e.g., see references 9, 240, 500, 519, and 539). Sequence data can be generated experimentally or collected directly from expression sequence databases. Bullet points indicate alternative procedures at each stage. EST, expressed sequence tag.
|
 View larger version (9K): [in a new window] |
FIG. 5. Scheme for recovery of viral nucleic acid from microarray spots. Hybridized viral sequences were physically scraped from a DNA microarray spot using a tungsten wire probe mounted on a micromanipulator, while the spots were visualized under fluorescence microscopy. Subsequently, the virus was identified by nucleic acid amplification, cloning, and sequencing. (Reprinted from reference 512 with permission.)
|
|
TopPreviousNextReferences |
|---|
As we have described above, there is a wealth of circumstantial evidence implicating retroviruses other than HIV and HTLV in human disease. Although in some instances, specific agents have not been identified, in recent years, several putative retroviruses have emerged as candidate human pathogens. In this section, we discuss these specific examples in some detail, including those that have been characterized by PCR and gene sequencing.
The suggestion that a retrovirus related to MMTV might be involved in breast cancer in humans is one of the longest-running controversies in human retrovirology (313, 523). MMTV is a betaretrovirus that causes mammary tumors in mice by the insertional activation of proto-oncogenes (1, 84). The virus is transmitted in breast milk and infects dendritic cells and B lymphocytes in the gut of the suckling pup. MMTV encodes a SAg in its 3' LTR, which, when presented to cognate T lymphocytes in the gut lymphoid tissue, triggers their activation (90). The specific subset of T cells activated depends on the particular strain of MMTV, and the resulting T-cell response leads to the activation of the infected B cell, which proliferates sufficiently that the virus is carried to the mammary gland, where it infects the mammary epithelium. SAg-dependent activation of T cells is essential for infection of the mammary gland since replication in B cells is otherwise inefficient, and transmission to mammary tissue is unsuccessful (202).
Once the mammary epithelium is infected, estrogen-driven activation of the MMTV LTR mediates the mammary-specific activation of the virus and its accumulation in milk. Mammary tumors are induced by the insertional activation of cellular proto-oncogenes (84). The genes activated by MMTV were originally designated as int loci and later identified as members of the Wnt, notch, and fibroblast growth factor (fgf) families. In addition to its role in mammary tumors, a variant of MMTV with a rearranged U3 sequence in the viral LTR is associated with T-cell lymphomas (26). A functional SAg gene is not required for lymphoma production (353).
A link between MMTV and breast cancer in humans was first proposed following the detection of B-type virus particles in healthy human milk by EM (137, 138, 338). Similar particles were also detected in a number of cell lines derived from breast tumors (20, 155, 252, 459) and directly in tumor tissue biopsies (137). Subsequently, several laboratories attempted to characterize this putative virus. RT activity and 70S RNA were detected using a "simultaneous detection" assay (136, 170, 432), and molecular hybridization studies appeared to identify MMTV-related sequences in RNA from breast tumors (21, 459), although the specificity of these reactions is uncertain given that HERVs had not been characterized at that time. In addition, breast tissue was reported to contain antigens related to the major core protein (CA, p27) of MMTV and MPMV and to SU (gp52) of MMTV (253, 332, 460, 542). Human infection by MMTV was further supported by the detection of anti-Gag and anti-Env antibodies in sera from breast cancer patients (215, 313, 530).
Despite these data, others could not reproduce the detection of MMTV-related nucleic acid (51) or antigens (88, 203), and many viewed the evidence as unconvincing at best (425). One problem was that much of this work did not sufficiently address the question of disease association, and a comparison of data from malignant and healthy tissue revealed inconsistent findings; e.g., B-type particles were identified in healthy milk, but antigen reactivity and RT activity were specific to tumors (for a more detailed description of the early work on this subject, see reference 523).
Despite these inconsistencies, by the mid-1980s, there was a good deal of circumstantial evidence to support the idea of a human breast cancer virus, but rigorous proof had not been obtained. The advent of PCR provided an opportunity to revisit the question with more specific and sensitive probes. In 1995, Wang and colleagues reported the detection of MMTV-like env sequences in 38.5% of unselected breast cancer DNA samples but in only 2% of normal breast tissues (514). The sequenced PCR products had 95 to 99% similarity with murine MMTV viruses. Subsequent RT-PCR analysis indicated that these sequences are expressed in 66% of MMTV-positive tumors and are never expressed in normal tissue (513). The validity of these data has been questioned (49, 314, 531, 546), but at least two other groups obtained similar positive PCR amplifications (132, 145). In one study, all regions of the MMTV