In this study, we investigated the putative association between XMRV and prostate cancer using a combination of microarray, PCR, FISH, serological, and deep sequencing approaches. XMRV was not detected in a new set of 39 prospectively collected prostate tumors (both with or without RNAse L mutations) by PCR assays performed independently in 3 different laboratories or ViroChip microarray. Moreover, XMRV was not detected in archival VP62 tissue previously found to be XMRV-positive 
. These negative findings were supported by the failure to detect XMRV sequences in 19 of the newly collected prostate cancer samples and archival VP62 tissue by FISH. In addition, we failed to detect antibody responses to XMRV in plasma samples from the 39 patients with prostate cancer, a finding that was also observed recently in another study 
Taken together, the data presented here strongly suggest that there is no association between XMRV infection and prostate cancer, regardless of RNAse L status.
In the original 2006 study by Urisman, et al. as well as a 2010 study by Arnold, et al. 
, XMRV was detected in a small proportion of nonmalignant stromal cells by FISH. It is unclear as to why XMRV was detected by FISH in these previous studies but not in the current study, which included re-analysis of archival VP62 tissue. Here we used a direct-labeled, full-length plasmid XMRV probe with a high label incorporation rate 
, which produced a clear punctate staining pattern in 22Rv1 cells by FISH (). Of note, this novel probe design was able to detect human papillomavirus (HPV) in cervical cancer cells harboring 1 to 2 copies of integrated HPV-16 per cell as well as in cervical cancer tissue sections (data not shown). Thus, if integrated XMRV was present in the tissues examined in this study, it should have been detected by FISH. It is likely that the low frequency of XMRV FISH-positive prostate cells observed in previous studies 
represent non-specific binding artifacts.
To investigate whether the discovery of XMRV may have resulted from inadvertent laboratory contamination, we re-analyzed available archival RNA extracts from prostate cancer samples taken from the original 2006 study by Urisman, et al
. By microarray and PCR analysis, the previous findings that a subset of these samples harbored XMRV sequences was replicated (; ). Furthermore, unbiased deep sequencing analysis of 3 XMRV-positive samples (VP35, VP42, and VP62), revealed that the entire viral genome was present (). Failure to detect mouse mitochondrial or IAP sequences in these 3 samples also support the contention that these samples harbor XMRV and not related mouse endogenous gammaretroviruses.
One of the findings arguing against laboratory contamination as a possible source of XMRV has been the degree of sequence variation observed between XMRV genomes, up to 2% in the gag
. Although the reported diversity is extremely low for retroviruses in general 
, certain retroviruses such as HTLV-1 can exhibit comparably low rates of natural sequence variation, with strains in the wild that are 96–99% identical 
. Nevertheless, the SNP data generated from deep sequencing reveal that the consensus sequences of the XMRV VP35, VP42, and VP62 genomes are in fact identical to each other and to the consensus 22Rv1-associated XMRV strain (). Thus, previously reported sequence diversity between different strains in the 2006 study by Urisman, et al.
and presumably in other fully- or partially-sequenced XMRV genomes appears to arise from Taq polymerase errors introduced during PCR and/or sequencing 
, and not from natural genetic variation.
Notably, we found evidence of XMRV infection of a 2003 LNCaP prostate cancer cell line by deep sequencing (). The consensus sequence of the XMRV genome in these LNCaP cells was found to be identical to the 22Rv1 XMRV consensus sequence. Both of these cell line-associated XMRV genomes were found to exhibit a lower degree of intra-strain variation than previously reported for XMRV from 22Rv1 cells 
, with only 19 SNPs detected in the 22Rv1-associated XMRV genome at the 3% frequency cutoff by deep sequencing, and only 25 SNPs in the LNCaP-associated genome (; Table S1
). It is therefore striking that the three most common SNP variants identified in LNCaP- and 22Rv1-associated XMRV by deep sequencing, A790G, A4264G, and C8122G, are also present in the 3 prostate cancer-associated XMRV genomes. In conjunction with the 100% consensus sequence identity shared among cell line and prostate cancer-associated XMRV genomes (), these findings suggest a high likelihood that a viral contamination event had occurred.
To prove the hypothesis that an XMRV-infected cell line had contaminated the prostate cancer samples in the 2006 Urisman, et al.
study, we analyzed available RNA extracts using a novel technique referred to as mitochondrial RNA (mtRNA) profiling. Unlike profiling strategies involving whole or partial genome sequencing of mitochondrial DNA 
, here the ~16.5 kb mitochondrial genome is assembled from only RNA-derived deep sequencing reads. By mitochondrial SNP analysis, direct evidence of contamination from LNCaP mitochondrial sequences in the VP35 and VP42 samples was detected, with surprisingly high minority SNP frequencies ranging from 4.0% to 50.6% (). Importantly, VP35 and VP42 were among the first 10 samples processed in the original 2006 study by Urisman, et al.
, and the only two samples positive for XMRV from both total and polyA RNA (, , and 
). Taken together, these findings imply that the initial contamination event, involving VP35 and/or VP42, occurred very early in the course of the 2006 study.
In light of the data presented here, we have generated a model for how XMRV contamination was introduced into the prostate cancer samples analyzed by Urisman, et al.
(). An XMRV-infected LNCaP cell line in the laboratory at the Cleveland Clinic inadvertently contaminated RNA from VP35 and VP42 during RNA extractions, which comprised part of the initial set of 10 samples that were extracted on the same day (, “SET #1″). The LNCaP cell line, in turn, had been likely infected with XMRV from 22Rv1 cells in the same laboratory in which 22Rv1 cells were previously used, or in another laboratory at the Cleveland Clinic that was working with both cell lines and had initially provided the LNCaP cells for analysis. It should be emphasized, however, that only after 2009 was XMRV known to be present in 22Rv1 cells or in any other cell line 
. In fact, as a necessary precaution, all cell lines circulating in the laboratory were tested in 2004 for XMRV and all tested negative by RT-PCR with the exception of a different but related MLV from an aliquot of LNCaP whose genome at the time was fully sequenced (“MLV-LNCaP”). Based on this analysis, it was mistakenly deduced that XMRV could not have originated from LNCaP or another cell line in the laboratory. Interestingly, in the current study, we were unable to recover sequence from this related MLV in a fresh aliquot of 2003 LnCaP cells by deep sequencing, and instead, found only XMRV (; Fig. S1
). After contamination of the VP35 and/or VP42 sample(s) by XMRV-infected LNCaP, polyA RNA extracts from other prostate cancer samples then became cross-contaminated. The detected association of XMRV with the RNAse L R462Q variant (QQ) may have resulted in part from an increased proportion of QQ samples analyzed (11 of 19; 58%) relative to the QQ genotype frequency in prostate cancer cases of approximately 15% 
. Notably, after detection of 8 XMRV-positive samples (out of 19) by ViroChip and PCR, subsequent PCR screening of an additional 67 prostate cancers yielded only one additional positive sample, VP184 
, which in hindsight may represent nested PCR contamination.
Proposed Model for Laboratory Contamination by XMRV.
In summary, our findings do not support any association between XMRV infection and prostate cancer, and by extension indicate that XMRV has never replicated outside of the laboratory setting. The initial discovery linking XMRV to prostate cancer in 2006 arose from laboratory contamination of clinical samples by an XMRV-infected LNCaP cell line. In turn, the LNCaP cells were most likely previously infected by 22Rv1, from which XMRV almost certainly originated through in vivo
passaging of the CWR22 xenograft in mice 
. Nevertheless, the discovery of XMRV in 2006 accelerated research that has now established the virus as a genuine infectious agent with a unique biology and as-yet undefined pathogenic potential. Important features of XMRV biology include (1) tropism for a variety of cell lines, including prostate cancer DU145 and LNCaP cells 
, and human neural cell types 
, (2) adaptations that promote growth in prostate epithelium and human-derived prostate cancer cell lines including an androgen response element in the promoter region 
and downregulation of APOBEC3G 
, and (3) cellular effects with potential oncogenic properties including increased tumor aggressiveness mediated by downregulation of p27 
and differential regulation of several microRNAs 
. A study of XMRV-induced apoptosis of SY5Y human neuroblastoma cells identified its receptor, Xpr1, as a novel atypical G-protein-coupled receptor (GPCR) 
. Finally, XMRV was found to establish both acute and chronic infections in mice and two species of non-human primates