Although, based on accumulating results, the proposed association of XMRV infection with human disease, as well as evidence of any authentic human XMRV infection, appears increasingly unlikely, the virus itself is a replication-competent retrovirus with unknown pathogenic potential that is capable of infecting human cells (7
). Given the lack of any confirmed positive cases of human XMRV infection (29
), the in vivo
replication capacity, sequence evolution, tissue tropism, and elicited immune responses associated with XMRV infection are unknown and it is unclear what results using a variety of direct and indirect detection methods might be expected in the setting of authentic XMRV infection. Information about the natural history of XMRV infection and host immune responses to the virus would provide a framework to help interpret results suggesting evidence of human XMRV infection. In the absence of any known human infection, animal models must be relied upon to provide this information.
To address these questions in a nonhuman primate species, we intravenously infected two adult male pigtailed macaques with >1010 XMRV virions, an inoculum which likely exceeds by many orders of magnitude any viral inoculum that might be involved in a physiological human transmission. Despite this large viral inoculum, XMRV replicated only transiently to relatively low peak levels in both animals, achieving peak plasma VLs of ≤2,200 RNA copies/ml that declined to undetectable levels within 4 weeks of infection. This decline in viremia was associated with striking levels of G-to-A hypermutation in PBMC-associated vDNA, likely reflective of APOBEC-mediated viral restriction. Although plasma viremia was brief, within the first 2 to 4 weeks of infection, both animals raised robust anti-XMRV antibody responses primarily directed toward p15E, p30CA, and gp70SU that were largely maintained up to 119 dpi. In addition to these binding antibody responses, neutralizing antibodies were also elicited within the first 2 weeks of infection and maintained through the study's conclusion. In contrast, innate immune responses in LNs were only transiently upregulated in the first week of infection and rapidly diminished to baseline levels by 2 weeks postinfection, while adaptive T cell responses were essentially negligible in ICS format assays using whole XMRV virions or purified recombinant XMRV proteins as stimuli.
Our findings differ markedly from some of those made in a previous study by Onlamoon and coworkers, which examined XMRV infection of rhesus macaques (M. mulatta
). Although Onlamoon et al. also reported low, transient levels of viremia that declined to undetectable levels within 3 weeks of infection (albeit with a less sensitive qRT-PCR assay) associated with the induction of a relatively stable antibody response with neutralizing activity, there were several key differences between the results of the rhesus macaque infection study and those reported here for pigtailed macaques. First, while we observed a relatively stable pool of PBMC-associated XMRV DNA composed predominantly of G-to-A-hypermutated viral genomes that were detectable through 119 dpi, Onlamoon et al. reported a complete loss of PBMC-associated vDNA within the first month of infection (44
). Although a recent follow-up study has shown G-to-A hypermutation in PBMC-associated vDNA sequences in these infected rhesus monkeys (67
), the disappearance of vDNA in PBMC seems to suggest that factors other than APOBEC-mediated restriction may contribute importantly to control of XMRV infection in rhesus macaques. It is not clear, however, if the PCR assay used here and that used by Onlamoon and coworkers are equally capable of detecting hypermutated vDNA, which are likely less efficiently amplified due to inefficient priming, and it is possible that the apparent loss of vDNA-positive PBMC in rhesus macaques simply reflects the accumulation of hypermutated genomes.
Other key differences between this study and that of Onlamoon et al. were noted when comparing analyses of various host tissues. In our pigtailed macaques, we identified vDNA in LNMC by X-SCA at early and late postinfection time points; however, longitudinal LN sections were vRNA negative when probed with our ISH riboprobe cocktail, the specificity of which was verified using in vitro
-infected pigtailed macaque cells (B; see Fig. S1 in the supplemental material). These results were consistent with our observations on PBMC vis-à-vis plasma viremia and suggest that the vDNA-positive cells detected by X-SCA in LNs were not productively infected. Conversely, Onlamoon et al. reported that LNs from infected rhesus macaques contained productively infected cells detectable by IHC at both early and late (>140 dpi) postinfection time points (44
). Since XMRV was initially linked to human prostate cancers, we also evaluated viral replication in prostate tissue. By X-SCA, we detected very low levels (<15 copies/106
cells) of vDNA in snap-frozen prostate tissue pieces collected at necropsy, and no vRNA-positive cells were detectable in prostate tissue sections by ISH. Since a small number of contaminating blood cells could be the source of this low-level vDNA, these data indicate very limited or no XMRV infection in prostate tissue with no productive infection at 119 dpi. Although Onlamoon and coworkers noted fewer productively infected cells in prostate tissue at late postinfection time points than at early postinfection time points, they reported that vRNA+
cells could still be detected by ISH in prostate tissue at >140 dpi (44
). These findings on rhesus macaques suggest the establishment of a chronic viral infection, characterized by the stable presence of productively infected cells in LNs and other tissues, concomitant with the apparent loss of vDNA+
PBMC and undetectable plasma viremia, suggesting that once the virus makes it into LNs and other tissues, it continues to replicate but becomes trapped and does not reseed the peripheral blood compartment despite the apparent presence of suitable target cells. This conclusion is in stark contrast to the results we report here for pigtailed macaques, wherein early rounds of viral replication lead to the seeding of PBMC and LNMC with a stable pool of archived, hypermutated, likely nonfunctional viral genomes. In addition to the difference in macaque species, differences in the sensitivity and specificity of the assays employed could also contribute to the discordant results.
In the absence of any confirmed human XMRV infection cases, it is unclear whether infection of pigtailed macaques accurately models human XMRV infection. There is, however, evidence to suggest that several features of pigtailed macaque infection may mirror what would occur in an XMRV-infected human. Previous work has shown that XMRV infection of human PBMC in vitro
results in a nonspreading, restricted infection with viral production showing a dose-dependent plateau effect with extensive G-to-A hypermutation in cell-associated vDNA, likely mediated by APOBEC proteins (7
). Infection of pigtailed macaque PBMC in vitro
showed kinetics and replication patterns strikingly similar to those reported for human cells (A), and we observed extensive G-to-A hypermutation in vivo
(). It therefore might be expected that XMRV would be similarly hypermutated and restricted during any in vivo
human infection, leading to comparably transient low-level viremia. The potential importance of APOBEC-mediated hypermutation in limiting viral replication is underscored by the rapid decline of plasma VLs to undetectable levels in our infected animals in the face of persistent vDNA-positive cells in blood and LNs, demonstrating that a total elimination of infected cells, by either immunological or viral lytic mechanisms, was not responsible for the disappearance of viremia, but rather that these cells contained defective viral genomes that did not express viral gene products. In agreement with this notion, vDNA in LNMC was first detected by X-SCA just after peak viremia and was still detectable at necropsy; however, longitudinal LN sections were vRNA negative by ISH throughout this study. Although these vDNA+
cells in blood and LNs persisted at 119 dpi, innate immune responses in LNs were only transiently detectable in the first week of infection, in stark contrast to the continual LN immune activation associated with chronic SIV infection (), and cellular adaptive immune responses were negligible, consistent with the view that the hypermutated proviral genomes were rendered nonfunctional and did not express viral gene products. Taken together, these results suggest that APOBEC-mediated hypermutation is likely largely responsible for the limited viral replication we observed in our infected animals. These results may have broader implications concerning the potential for gammaretroviral infection of humans, where such concerns have been raised in particular regarding the transfer of porcine gammaretroviruses to immunosuppressed humans in the setting of xenotransplantation (66
). Because gammaretroviruses lack the accessory genes possessed by other retroviruses, such as HIV and SIV, they are ill equipped to counteract intrinsic host restriction factors like the APOBEC3 proteins. As shown here, these primate host restriction factors can be remarkably effective at inhibiting gammaretroviral replication in vivo
and may explain, at least in part, why gammaretroviruses have not been significant pathogens in humans, including immunosuppressed individuals, despite close human association with their animal hosts.
Hypermutation of viral genomes also likely explains our unsuccessful efforts to rescue virus from differentially stimulated PBMC collected at the study's conclusion using an LNCaP-based reporter cell line (DERSE.LiG-puro) (data not shown). Given the extensive G-to-A hypermutation identified by X-SGS in cell-associated vDNA (), it is likely that the proviral genomes were rendered incapable of producing infectious progeny. Although vRNA in plasma was not hypermutated, reflecting the fact that only nonhypermutated proviruses produced progeny, we were also unable to rescue infectious virus from plasma samples collected at several postinfection time points (see Fig. S2 in the supplemental material). While we did not expect to rescue plasma virus at time points at which plasma viremia was undetectable, we note that we were also unable to rescue virus from plasma samples taken at the time of known peak viremia. These results suggest that the isolation of replication-competent XMRV from human blood samples may be difficult or unlikely, particularly when working with samples from unknown postinfection time points that are not likely to represent peak viremia, and they are in contrast to those of previous studies that reported an ability to isolate infectious XMRV from human PBMC and plasma samples similarly using LNCaP target cells (34
). While these disparities might be explained by methodological differences, it seems that XMRV would have to be less susceptible to human APOBEC3-mediated hypermutation in order for virus to be isolatable from PBMC; however, previous work has shown that XMRV is highly susceptible to APOBEC3-mediated hypermutation in human cells (7
). To successfully rescue virus from human plasma, particularly without the ability to select peak viremia time points, XMRV would likely have to replicate to higher sustained levels in humans than those observed in our pigtailed macaques or exist in human plasma in immune complexes that have a less detrimental effect on viral infectivity.
If the course of XMRV infection in pigtailed macaques does reflect the natural history of XMRV infection in humans, we would expect serological assays and PCR assays to detect vDNA in PBMC to have the greatest likelihood of detecting XMRV infection in clinical samples, due to the greater stability of these parameters following XMRV infection in vivo. In contrast, we would expect assays to detect vRNA in plasma and assays to rescue culturable virus to be less likely to detect XMRV infection due to transient positivity and lack of sensitivity, respectively. One of the potential shortcomings of previous studies designed to identify XMRV infection in human samples using PCR, serology, IHC, ISH, virus rescue, or other means has been the lack of bona fide in vivo-derived positive-control samples. Therefore, in addition to characterizing the natural history of XMRV infection in vivo, we initiated these studies with pigtailed macaques to generate and make available positive-control samples for future analyses. We have generated an infectious, 22Rv1-derived XMRV stock and matched highly purified 1,000×-concentrated stocks and determined their virus contents by qRT-PCR and quantitative p30 densitometry. Macaque antiserum raised against infection with this 22Rv1 virus was collected at necropsy and stored for future use. We show here that this serum is XMRV specific, with clear Western blot assay reactivity for p15E, p30CA, and gp70SU at a 1:2,000 dilution (), detectable p15E reactivity at a 1:20,000 dilution (), and negligible background reactivity with microvesicles prepared from human cells (see Fig. S3 in the supplemental material).
The reagents we have generated here should facilitate future studies aimed at examining XMRV infection in in vivo-derived samples. (1×- or 1,000×-concentrated XMRV produced from 22Rv1 cells and serum samples from XMRV-infected pigtailed macaques are available on request from Julian W. Bess, Jr. [bessjw/at/mail.nih.gov].) Though XMRV infection of pigtailed macaques doubtless does not perfectly model human XMRV infection, these results suggest that a small, physiological viral inoculum, in the setting of functional APOBEC proteins, will likely replicate only briefly and to extremely low levels, at best, and might therefore be very unlikely to induce any significant pathogenesis.