This study was immediately challenged by other reports, initially from Europe [16–18] and later from the US [19–22] that could not confirm the findings. The potential association between a new pathogen and CFS and the implied spread of a previously unrecognized gammaretrovirus in humans were nevertheless of great concern. Despite all warnings that the link between CFS and XMRV could not be confirmed, many patients suffering from this debilitating disease and desperately hoping for a cure started taking anti-retroviral drugs. However, the increasing number of reports failing to find a retroviral etiology for CFS and other disorders [23–26] cast further doubt on the reliability of the original findings. The authors and enthusiasts of the original CFS study [3] attempted to dismiss such negative results on the grounds of inadequate diagnostic procedures or inappropriate selection of patients [27], demanding that the methods used in the original paper be meticulously replicated. However, a recent study [28] again found no evidence for XMRV despite using virtually identical PCR, serology and virus isolation methodologies to investigate samples from some of the same CFS patients found to be positive in the initial study [3].

Paprotka and colleagues therefore characterized XMRV infection in early and late passages of the prostate cancer xenograft that eventually gave rise to the cell lines 22Rv1 and CWR-R1. They found no evidence for ‘human’ XMRV in early passages of the tumor xenograft but found instead two XMRV-related proviruses with complementary regions of high homology to prototypic XMRV, sequences that they named preXMRV-1 and preXMRV-2. However, later passages of the xenograft and both resulting cell lines contained the XMRV prototype sequence closely matching that reported in humans (Figure 1). This clearly demonstrates that the virus was not present in the original human tumor but was acquired later.

What, then, was the source of preXMRV-1 and preXMRV-2 in the xenografts and where did ‘human’ XMRV form? By the time of testing, the xenografts had presumably become infiltrated with murine cells and the preXMRV-1 and preXMRV-2 proviruses were therefore amplified from traces of genomic murine DNA in the xenograft samples. This was confirmed by analyzing tissues from the strains of nude mice likely to have been used in the 1992–1999 xenograft experiments. The complementary XMRV precursor proviruses (preXMRV-1 and -2) were indeed found in two of these strains (but no ‘human’ XMRV) and the authors therefore concluded that in the mid-1990s recombination between the two parental MLVs in a xenografted nude mouse gave birth to XMRV. While the replication-defective preXMRV-1 is closely related to xenotropic endogenous MLV, pre-XMRV-2 resembles the polytropic or modified-polytropic type. The fact that preXMRV-1 (xenotropic) contributed the envelope protein to ‘human’ XMRV might explain why the new virus could replicate in the human xenograft but not in the murine host itself. Comparing the sequences of preXMRV-1 and preXMRV-2 with that of XMRV, the authors suggest that a virus particle containing both preXMRV-1 and preXMRV-2 genomic RNA must have formed. This particle then infected a cell and six reverse transcriptase template switching events during minus-strand DNA synthesis resulted in the mosaic sequence known as XMRV (Figure 2). This scenario is completely plausible as it is known that the RNAs of two related proviruses can occasionally be incorporated into a single virion and that reverse transcriptase usually switches between the two RNA templates an average of four times per cycle [29].

Furthermore, Knox and colleagues [1] showed that XMRV and other xenotropic MLVs are inactivated by sera from both healthy donors and CFS patients, probably by the action of human complement. Finally, many groups, including that of Knox et al, have demonstrated the high probability of false-positive PCR results being the reason for the earlier reports linking XMRV to disease [1,20,28,37,38]. Contamination of patient samples, cells in culture or lab reagents with murine DNA, plasmid DNA containing XMRV sequences or even the virus itself [39] can easily occur if stringent precautions are not taken. Finally, the lack of variation between sequences supposedly obtained from different individuals [1,40] argues against the presence of XMRV in humans as this is not consistent with our current knowledge of retrovirus replication and spread.