In 2005, we initiated a study to examine the expression level of an endogenous human betaretrovirus, HERV-K18, in chronically ill CFS patients versus healthy controls. For this purpose, we accumulated a library of DNA samples from CFS patients which has allowed us to investigate the possible association of XMRV with this disease [3
]. We initiated our studies on XMRV using a TaqMan qPCR assay for a region in XMRV pol
that is unique to XMRV and does not detect any sequences in genomic DNA from laboratory strains of inbred mice [2
]. None of the samples from either CFS patients or healthy controls was positive in this assay, although we were able to detect a signal from two XMRV-infected lymphoblastoid cells (cell line WPI-1282) in a background of DNA from up to 106
human LnCaP cells. In our hands, the qPCR assay is 10-fold less sensitive than the nested XMRV gag
PCR assay when tested on the same XMRV-positive cell line, since the latter can detect a signal in DNA from <1 cell. This difference is a consideration for the negative results we obtained as the sensitivity of the qPCR assay may not have been adequate for the detection of minute amounts of XMRV. We are not aware of any other group who has used this technique for the detection of XMRV in the DNA of freshly isolated PBMC. However, Danielson et al
. recently reported that they could only detect XMRV sequences, using XMRV env
, but not gag
, primers [21
In contrast to the qPCR results, we were able to readily detect XMRV using the nested PCR originally described by Urisman et al.
], and we found many more positive samples in our healthy control cohort, compared to the CFS cohort. Of possible relevance for the interpretation of these findings may be the fact that the samples from the two cohorts were prepared years apart, although all in the same laboratory using somewhat different protocols and reagents. It is also important to point out that individual DNA samples remained reproducibly positive or negative on repeat examination rendering the possibility of random contamination of the PCR assays very unlikely. Furthermore, each assay contained positive and negative controls which were 100% correlative; i.e
., the DNA from the XMRV-infected cell line was always positive and the no-template control or LnCaP DNA was always negative. Thus, it is unlikely that contamination occurred at the time of setting up the PCR reactions.
To further understand the origin of the positive PCR signals, we determined the DNA sequences of the gag
PCR products. In most cases, it was only possible to obtain unique sequences from PCR products after dilution of the input DNA to an extent where single molecules were amplified, since initial studies showed that most of the positive samples contained mixtures of closely related sequences. In this way, we obtained 15 different sequences from a total of 37 single PCR products. When compared to the collection of endogenous MLV sequences extracted from the sequenced mouse genome [18
], these sequences included examples from all parts (XMV, PMV, and MPMV) of the resulting neighbor-joining tree, as well as a cluster of three sequences identical (in this region) to the VP42 isolate of XMRV. With regard to the latter result, it is of significance that no VP42 plasmid, nor VP42-containing cell line, nor isolated DNA, was present in the Huber laboratory that could have resulted in contamination (WPI-1282 contains VP62 which differs by one base change in the region analyzed). The genomic DNA from the three healthy volunteers who had XMRV VP42 sequences also contained other MLV sequences. Thus, it is not possible for us to distinguish which one of the retroviruses stemmed from mouse DNA contamination; i.e.
, it is formally possible that VP42 is an actual human retrovirus. It is also possible that it is an endogenous provirus, not present in the sequenced C57Bl/6 genome, but present in the mouse species responsible for the sequences observed [19
]. In the former case, the presence of VP42 in DNA from healthy control samples, but not CFS patients, would indicate that this virus is spread randomly through the human population, with no particular link to CFS. Further analyses are required to clarify this issue.
The presence of mixtures of MLV sequences, all closely related to known endogenous MLVs [17
], in many of the DNA samples tested is not easily reconciled with infection of human hosts with the corresponding viruses (reviewed in [14
]). Two assays specific for murine DNA, for mitochondrial cox2
and IAP sequences, were used to test the possibility that there might be trace amounts of mouse DNA contaminating some of the samples. Consistent with this idea, we found that each DNA sample that was positive for XMRV/MLV also was positive for mouse DNA by the IAP assay, while >50% of XMRV/MLV-negative samples were positive for mouse DNA which is particularly striking in the CFS group. Again, these results were confirmed in repeat experiments and never deviated in subsequent analyses, suggesting that contamination happened either during collection of blood, isolation of PBMC, or during the preparation of the DNA from the PBMC. We interpret these data that possible contamination with mouse DNA is ubiquitous, but the level seemed to vary significantly from batch to batch of sample preps, although all experimental procedures were carried out in the same facility. In particular, although samples collected at both times showed signs of contamination, the level of contamination in the normal controls collected in 2009-2010 was noticeably greater than in the CFS samples from 2005. To date, we have not been able to pinpoint a specific reagent or laboratory vessel for being consistently positive for mouse DNA, but preliminary experiments implicate both fetal calf serum (FCS) and phosphate buffered saline (PBS), although large variations in the surmised amount of contaminating mouse DNA were observed from bottle to bottle. All blood samples were collected in heparin tubes rendering the anti-coagulant also a likely suspect for mouse DNA contamination. However, a comparison of parallel blood collections from the same healthy individual in heparin, Na-citrate and EDTA tubes did not support this hypothesis. In this particular set of samples only one DNA aliquot from Na-citrate-collected blood was positive for mouse DNA (results not shown).
Currently there are highly discordant reports in the literature about the prevalence of XMRV in CFS and prostate cancer patients (reviewed in [12
]). The original publication on CFS patients reported that almost 70% of these patients, but less than 5% of healthy individuals, harbor this virus [3
], and that infectious virus and antiviral antibodies could be detected in blood from these patients. Several reports have appeared in the literature since then contesting these findings [4
], while a recent publication claimed that 80% of CFS patients, but not healthy controls, contained endogenous MLV-like sequences, but were negative for mouse mitochondrial DNA [16
]. The sequences from CFS patients identified in this latter paper were distinct from the XMRV of the original reports. A plausible explanation for these discrepant results has not been put forward to date [13
], but it is worth pointing out that the sequences identified in the latter report were similar to the ones we found in the present study. Endogenous MLVs are abundant in all laboratory mouse strains [17
], as well as in wild Mus
] and are carried by some human cell lines that have been propagated in vivo
in nude mice [20
]. Thus, extreme precautions have to be taken to exclude contamination with mouse DNA or DNA from any abundant MLV-producing cell line.