In this study, we have examined the prevalence of XMRV in patients with or without PCa at Mayo Clinic. We were unable to find XMRV sequences or anti-XMRV antibodies in our patients, most of whom are from the mid-west area of the USA, indicating that there is no or very low prevalence of XMRV in this region. Moreover, we were unable to confirm the correlation between XMRV infection and PCa, higher tumor grade or RNASEL R462Q mutation.
A high prevalence of XMRV has been reported in patients with PCa and chronic fatigue syndrome (CFS) in the USA [3, 7, 8], but similar studies in Europe have failed to detect XMRV [10–12]. It has been suggested that geographical differences might explain this striking variation in XMRV prevalence  but our results, as well as recent US studies that also find no evidence for XMRV [9, 21], appear to rule this explanation out. In this regard, it is notable that previous studies to identify XMRV in patients with PCa or chronic fatigue syndrome have relied on very sensitive PCR detection methods. Because of the high similarity between patient associated XMRV/MLV and endogenous MLV sequences and the striking discordance between studies, it has been suggested that PCR-positive results might be attributed to unintentional detection of contaminating mouse DNA in human specimens [6, 22–24]. It is notable that Lo et al.  detected polytropic and modified polytropic MLV sequences, but not XMRV, in blood samples from chronic fatigue patients (Figure 1). These authors were unable to identify the samples as contaminated using mouse mitochondrial PCR. In our study, real-time PCR and nested PCR identified 6 of 150 samples as positive for MLV. However, the amplified sequences were closely related to known endogenous MLV proviruses, rather than XMRV. In fact one patient sample (#52) contained two independent MLV sequences. This might be interpreted as evidence for evolution of the virus in the patient but closer analysis reveals that one of the sequences is identical to a known endogenous modified polytropic sequence whilst the other is a single nucleotide different from a known mouse endogenous xenotropic MLV. This, therefore, suggests either infection of this patient with two independent MLVs or PCR contamination with mouse DNA as a source. As all of the MLV PCR-positive samples contained detectable levels of mouse mitochondrial DNA, we conclude that the amplified sequences originated from mouse DNA that somehow contaminated the study samples.
In order to confirm that the viral sequences were amplified from endogenous MLV in mouse genomic DNA, but not replicating MLV in human tissue, we attempted to determine viral integration sites. We first used the protocol described by Kim et al.  but failed to amplify DNA sequences containing the partial XMRV LTR. We then designed universal primers to recognize LTRs from XMRV and endogenous and exogenous MLVs , as well as a series of primers specific for the viral sequences identified in our clinical samples. Unfortunately, we were not successful, likely due to low viral copy numbers in the clinical samples. Very recently, Robinson et al.  and Oakes et al.  reported similar observations; all XMRV PCR-positive specimens contained detectable levels of mouse mitochondrial or endogenous retroelements (IAPs). Together with our data, these findings highlight the difficulty of avoiding DNA contamination in clinical samples and the risk of testing contaminated samples as XMRV-positive by sensitive PCR detection assays. As a possible source of contamination, Sato et al.  demonstrated that a commercially available hot-start PCR enzyme contained mouse DNA. We used several enzymes and obtained similar results. Thus, it is unlikely that the contaminating mouse genome originated from a PCR kit. Since we could amplify the viral sequences from multiple aliquoted DNA samples, they appeared to be contaminated before or during the DNA isolation step, most likely during tissue sectioning on a microtome.
XMRV antigen-positive cells have been detected in prostatic stromal fibroblasts  or in malignant prostatic epithelium . Our IHC study using two different antisera showed conflicting results. The goat anti-MLV antibody found no viral antigens in clinical samples, while the rabbit anti-XMRV antibody used in the study by Schlaberg, Singh and colleagues  detected antigen-positive cells in prostatic epithelium. Strikingly, the goat anti-MLV serum did not stain the cells, which were IHC-positive by the anti-XMRV rabbit serum, in serial sections of the same tissue. The rabbit antiserum also found antigen-positive cells in PCR-negative sections, confirming the observations of Schlaberg and colleagues who reported frequent detection of IHC-positive samples in PCR-negative tissues . Importantly, both the rabbit and goat antibodies detected XMRV in experimentally infected cells with high sensitivity (Figure 3). Together, these observations strongly suggest that the rabbit antiserum is detecting a non-viral antigen sporadically expressed by tumor cells in the tissue section. We conclude that our PCa samples do not have XMRV antigen-expressing cells that are detectable by IHC.
We recently reported that Mus pahari mice elicit potent XMRV-specific humoral immune response upon XMRV infection . At a serum dilution of 1:640, antisera from infected animals almost completely blocked XMRV infection . Similarly, an animal study using XMRV-infected rhesus macaques and sensitive ELISA detection assays showed that infected animals rapidly develop antibodies against XMRV proteins, including gp70 (Env), p15E (transmembrane), and p30 (CA) . These results indicate that XMRV is strongly immunogenic in these animals. In contrast, we were unable to detect strong XMRV-specific neutralizing antibodies in our 360 patients, age 50-70, with or without PCa. This observation further suggests a lack of XMRV in our cohorts. It is possible, although less likely, that XMRV is not immunogenic in humans or that XMRV-specific immune response might have disappeared in these relatively elderly patients.