In this report, we used XMRV to investigate MuLV glyco-gag. XMRV is a recombinant between two endogenous proviruses in the mouse genome, preXMRV-1 and preXMRV-2, and it provides insight into the progenitor viruses that endogenized the mouse genome during evolution. The 5’ end of the XMRV genome, which is where glyco-gag is encoded, is derived from preXMRV-2. While XMRV does not encode a classical glyco-gag, there are two open reading frames in the 5’ end of the genome that could potentially encode glyco-gag activity. However, mutation of either of these potential reading frames had no measurable effect on viral release or infectivity, indicating that XMRV does not encode glyco-gag function. On the other hand, XMRV can respond to glyco-gag (albeit less than M-MuLV), since a chimeric XMRV encoding in-frame glyco-gag from M-MuLV (MXMRV) showed enhanced virus release through lipid rafts and enhanced infectivity. Expression of M-MuLV glyco-gag in trans could also enhance viral release and infectivity. Studies on the infectivities of MXMRV vs XMRV showed different effects in different human cell lines, which revealed relative restriction of XMRV in some human cells (e.g. HeLa) compared to others (e.g. 293); the restriction could be overcome by M-MuLV glyco-gag.
As shown in Figure 2, mutations designed to interfere with translation of two upstream ORFs in XMRV did not significantly affect infectivity or virus release, which allowed us to conclude that the putative 53 or 58 amino acid proteins were not important for viral replication. For one of these mutants, VP62Δgg1, there was approximately 80% reduction in infectivity. However, there was also an equivalent reduction in incorporation of viral RNA into released virus, which suggested that the reduced RNA incorporation was the cause of the reduced infectivity. For M-MuLV, there are several stem-loop structures in the region encoding glyco-gag (including the CUG initiation codon in one loop), and this region of the genome contains the RNA packaging (Psi) sequences . It is thus likely that the analogous region of XMRV is important for RNA incorporation. Use of the M-fold program [48, 49] on nucleotides 1–607 of XMRV predicted several stable stem-loops, and the nucleotide (382) mutated in VP62Δgg1 would normally be in a stem; the predicted structure of the mutant RNA showed an additional loop in this stem (T. Nitta and H. Fan, unpublished). The secondary structure of 5’ end of the XMRV genome predicted by SHAPE (selective 2’-hydroxyl acylation analyzed by primer extension) also suggested that the region corresponding to the mutation in VP62Δgg1 may form a stem structure, and that it may be involved in the formation of intermolecular loop-loop kissing interactions in RNA dimers . This supports the idea that the region encompassing the VP62Δgg1 mutation is important in packaging of viral RNA into virions.
Reports from us and others have suggested there are multiple functions for glyco-gag, although the regions of glyco-gag responsible for the different activities remain to be fully elucidated [15–19]. In our studies describing enhancement of virus release through lipid rafts for MuLV and HIV-1, the unique N-terminal 88 amino acids (expressed by pHA-gg88) were sufficient for this activity . Likewise, in this study, pHA-gg88 was sufficient to maximally enhance XMRV release from cells (Figure 3). On the other hand, additional glyco-gag sequences are required to enhance infectivity of nef-negative HIV-1 and glyco-gag negative MuLV from lymphocytes: the N-terminal 189 amino acids are required for full enhancement, and a 96 amino acid N-terminal fragment showed 60% activity . Consistent with this, we found that HA-gg88 enhanced XMRV infectivity 40-60% as well as full-length gPr80
Studies on the infectivities of MXMRV vs XMRV in different human cells indicated that some cell lines (e.g. HeLa) showed substantially higher susceptibility/infectibility for MXMRV than for XMRV (≥ 20-fold, Table 3), while for other cell lines, the two viruses had more similar titers (≤ 5-fold difference); this reflected more efficient replication of XMRV in the latter cells. Glyco-gag in virions might affect viral factors (e.g. efficient incorporation of Env into virions or change of viral core nature), which would provide a missing activity (e.g. viral attachment or entry) required to infect certain cell lines. The results of infectivities would also be consistent with restrictive cells (e.g. HeLa) expressing a factor that restricts XMRV replication, but which can be overcome by glyco-gag. Cells such as 293 would not express or express lower amounts of the factor. Comparisons between cell lines that differ in relative restriction of HIV-1 have led to the discovery of cellular restriction factors for HIV-1 replication (such as APOBEC3G and Trim5alpha [51, 52]), and our results might lead to identification of a novel restriction factor for XMRV in human cells. The experiments of Figure 6 and Tables 4 and 5 further indicated that the cell specific variation in restriction of XMRV observed here is not due to hA3 proteins, even though hA3G, and potentially hA3B and hA3F can restrict XMRV in vitro.
The exogenous MuLV that endogenized to form preXMRV-2 presumably could replicate in mice at the time of endogenization. Indeed the preXMRV-2 provirus contains full gag, pol and env genes, although it cannot encode functional glyco-gag. It was therefore of interest to investigate whether other endogenous proviruses in the mouse genome could encode glyco-gag or not. Endogenous MuLV proviruses have been grouped into ecotropic, xenotropic, and polytropic subgroups based on sequence homologies and whether their predicted Env proteins can mediate infection of murine and non-murine cells  (Additional file 5 Table S1). The polytropic ERVs have been subdivided into Pmvs and Mpmvs, and xenotropic ERVs (Xmvs) have been further subdivided into three clades (A, B and C) based on nucleotide sequence comparisons . As shown and described in Figure 7, based on the 5’ (glyco-gag) regions, XMRV and preXMRV-2 grouped with clade A Xmvs and also the Pmvs and Mpmvs. None of these viruses could encode glyco-gag, indicating that the exogenous progenitors for all of these viruses could replicate in mice in the absence of glyco-gag. On the other hand, clade B and C Xmvs would be predicted to encode glyco-gags (Figure 7). It was interesting that replication-competent MuLVs with glyco-gags highly similar or identical to those of clade B and C xenotropic proviruses have been isolated – i.e. the leukemogenic SL3-3 MuLV (derived from the endogenous ecotropic MuLV AKV) for clade C and the xenotropic MuLV DG-75 for clade B . Thus the exogenous progenitors for the clade B and C xenotropic proviruses encoded glyco-gag at the time of endogenization.
The results of Figure 7 raise the question as to when glyco-gag developed during evolution of the murine gammaretroviruses. At first glance the shape of the phylogenetic tree in Figure 7 might suggest that glyco-gag was present in more ancient gammaretroviruses, and that it was subsequently lost in the progenitors for clade A Xmvs
Pmvs, and Mpmvs. Laboratory mice are hybrids between European mice (rich in Pmv and Mpmv proviruses) and Asian mice (rich in Xmvs) [40, 54], and the cluster of ERVs lacking glyco-gag largely derived from the western European M. m. domesticus. Arguments exist for whether glyco-gag negative MuLVs preceded glyco-gag positive viruses, or vice versa. A glyco-gag – like protein (upstream ORF in-frame with the equivalent of Pr65
) is encoded by gammaretroviruses of several other mammalian species (feline leukemia virus , Gibbon ape leukemia virus [M. Pizzato, personal communication] and koala retrovirus ). Moreover the putative glyco-gag coding sequence of FeLV is also present in two endogenous FeLV-related proviruses in the cat genome (T. Nitta and H. Fan, unpublished), so glyco-gag was present in the primordial FeLV genome at the time it endogenized in cats. This conservation would support development of glyco-gag before radiation of gammaretroviruses to these different species, although we have not tested if these putative glyco-gags have equivalent function.
On the other hand, if glyco-gag negative MuLVs developed from glyco-gag positive viruses, what would be the driving force for loss of glyco-gag? One possibility could be that mice developed a restriction factor against glyco-gag, so that loss of glyco-gag could have facilitated MuLV replication in these mice. Nevertheless, exogenous MuLVs (glyco-gag positive) replicate efficiently in laboratory mice suggesting they carry no such factor, and inactivation of glyco-gag reduces infectivity in vivo . It was interesting that the 5’ region of the glyco-gag negative proviruses (and XMRV) have CUG codons and immediately downstream sequences that would encode amino terminal highly conserved glyco-gag amino acids (Figure 8). This would be more consistent with loss of glyco-gag from initially glyco-gag positive viruses.