We used a real-time PCR assay capable of detecting XMRV sequences in DNA from a very small number of infected cells, even in the presence of a vast excess (more than 10,000-fold) of uninfected cell DNA. We also performed IHC with two antisera, each specific for a different MLV protein, under conditions where the sera reproducibly stained XMRV-containing cells but not identically treated control cells. Taken together, the two assays surveyed nearly 800 prostate tumors, including microdissected tumor specimens; metastatic tumor tissue; and intermediate and high-grade primary tumors. No signs of XMRV infection were found in any of these tests. The results suggest that the prevalence of XMRV in prostate tumors may be far lower than has been reported previously.
How can our negative results be reconciled with the positive reports from other laboratories? It has been suggested that XMRV might be present in North American, but not European, prostate tumors (9
). However, our samples, like those of Schlaberg et al. (4
), were from North American men. Also, while we did not select RNase L R462Q homozygotes for analysis, the number of cases we examined was high enough to include a substantial number of these individuals. Another possibility is that XMRV was present in our samples, but we failed to detect it because the viral sequences were somewhat different from the published XMRV sequences. While little variation in XMRV sequences has been observed to date (the reported sequences are ~ 97% identical), this could potentially explain our negative PCR results. However, we used several primer sets, some against highly conserved MLV sequences, and still saw no MLV signals. Further, unlike PCR primers, the sera we used in our IHC assays are both broadly reactive, since they were generated using Mo–MLV proteins but reacted with the XMRV proteins in our positive controls (Mo-MLV and XMRV are 82 % identical at the amino-acid level). Thus it seems extremely improbable that sequence polymorphisms can explain our failure to detect XMRV by IHC.
It could also be proposed that infected cells are present at such a low level in virus-positive tumors that the samples we tested were too small to contain infected cells. (Contrary to this, Schlaberg et al. initially reported that positive samples contained 1–10 XMRV copies per 660 cells; 660 diploid cells contain ~ 5 ng of DNA, while we tested amounts ranging from 25 to 1000 ng (4
)). This might explain the negative IHC results with tissue microarrays, but seems unlikely in the >100 tumors for which we analyzed standard slides, which generally contain more than 105
Finally, another conceivable explanation for the staining seen by Schlaberg et al. (4
) is that the anti-XMRV serum used in their experiments contains antibodies directed against cellular proteins, in addition to the antibodies against XMRV proteins. The XMRV used as immunogen by Schlaberg et al. was apparently produced in human cells. It is thus difficult to exclude the possibility that human proteins were present in the virus preparation used as immunogen. HIV-1 virus particles are known to incorporate a wide variety of proteins from the virus-producing cells (13
), so that these proteins are impossible to remove from the virus; indeed, early vaccine trials with simian immunodeficiency virus were confounded by this phenomenon (14
). Incorporation of major histocompatibility complex proteins into MLV particles has also been reported (16
). We received PCa tissue sections (kindly provided by Dr Ila Singh, University of Utah) from a number of cases from specimens used by Schlaberg et al. (4
). Based on their results with the anti-XMRV antiserum, these samples were predicted to be IHC-positive. However, the sections did not stain with our MLV30 or MLV70 antisera (data not shown). While we cannot fully explain the discrepancies in staining results, Switzer et al. have also demonstrated that under immunoblotting conditions, the anti-XMRV antiserum (4
) reacts with proteins in uninfected HeLa cells (17
Many laboratories have used PCR to detect XMRV in clinical samples. However, the extraordinary sensitivity of this technique magnifies the risk of finding false positives, as well as the ability to find authentic positives. The risk is compounded by the widespread use of mice in biomedical research. Every mouse cell contains, in its DNA, ~ 100 MLV genomes, termed “endogenous viruses”. These genomes reflect past infections of germ cells and the resulting integration of the viral sequences into the mouse germline. As PCR is capable of amplifying and detecting a single molecule of viral DNA, this means that, for example, (depending, of course, on the specificity of the primers) a millionth of a microliter of mouse blood is a potential source of a positive signal in a PCR assay for MLV. Indeed, there are anecdotal reports of false-positive MLV signals ultimately traced to the use of the same microtome blade for cutting mouse and PCa sections, and to the tiny amounts of mouse DNA contaminating the mouse anti-polymerase monoclonal antibody used in commercial “hot start” PCR kits.
The existence of endogenous MLVs may be pertinent to another recent set of observations. In an attempt to reproduce the detection of XMRV in cases of CFS, Lo et al. (18
) performed PCR and reverse transcription-PCR on blood samples from CFS patients and healthy blood donors. They obtained positive signals from a high proportion of the CFS cases (and a much lower proportion of the healthy donors). However, when the PCR products were sequenced, they were found to differ from XMRV; thus these results are completely distinct from the reports of XMRV detection. In fact, the sequences match endogenous MLV sequences almost exactly. It should be emphasized that (unlike the studies reporting isolation of XMRV) this report does not include direct evidence for the presence of an infectious virus: the data consisted exclusively of amplification and detection of MLV-like sequences. Notably, the endogenous MLVs that they resemble most closely are defective MLV genomes which do not give rise to infectious MLV. While the authors provided strong experimental evidence arguing against contamination of their clinical samples with mouse DNA, this remains a possible explanation for their results.
In conclusion, the fundamental question of whether XMRV is really an infectious agent circulating in the human population is still unresolved. This question will not be settled until reproducible assays for the virus are established and validated; in turn this will require exchange of samples and testing of well-characterized standards, followed by cross-comparison of results obtained in different laboratories. Efforts in this direction are now underway at the U.S. National Institutes of Health. However, based on the data presented here, as well as that from other investigators (8
), we are doubtful that XMRV is commonly found in PCa. Over the years, many claims associating viruses with diseases have turned out to be mistaken (19
), and it is still possible that XMRV will fall into this category.
Finally, it is crucial to distinguish the question of the existence and prevalence of XMRV in the human population from the question of its causal role in PCa. In general, gammaretroviruses like XMRV induce malignant transformation by insertional mutagenesis, so that tumors induced by a gammaretrovirus are clones in which all the cells are infected (21
). This mechanism of carcinogenesis has been observed not only in laboratory animals, but also in children exposed to gammaretrovirus-derived vectors in gene-therapy trials (22
). Although some exceptions to this insertional mutagenesis mechanism have been described (24
), the viral genome is present in the transformed cells in all known cases. Thus, infection of an extremely minute fraction of the cells in some prostate tumors, even if confirmed, would seem to be incompatible with the possibility that XMRV plays a causal role in prostate tumorigenesis.