N-Linked Glycosylation Protects Gammaretroviruses against Deamination by APOBEC3 Proteins

ABSTRACT

Retroviruses are pathogens with rapid infection cycles that can be a source of disease, genome instability, and tumor development in their hosts. Host intrinsic restriction factors, such as APOBEC3 (A3) proteins, are constitutively expressed and dedicated to interfering with the replication cycle of retroviruses. To survive, propagate, and persist, retroviruses must counteract these restriction factors, often by way of virus genome-encoded accessory proteins. Glycosylated Gag, also called glycosylated Pr80 Gag (gPr80), is a gammaretrovirus genome-encoded protein that inhibits the antiretroviral activity of mouse A3 (mA3). Here we show that gPr80 exerts two distinct inhibitory effects on mA3: one that antagonizes deamination-independent restriction and another one that inhibits its deaminase activity. More specifically, we find that the number of N-glycosylated residues in gPr80 inversely correlates with the sensitivity of a gammaretrovirus to deamination by mouse A3 and also, surprisingly, by human A3G. Finally, our work highlights that retroviruses which have successfully integrated into the mouse germ line generally express a gPr80 with fewer glycosylated sites than exogenous retroviruses. This observation supports the suggestion that modulation of A3 deamination intensity could be a desirable attribute for retroviruses to increase genetic diversification and avoid immune detection. Overall, we present here the first description of how gammaretroviruses employ posttranslational modification to antagonize and modulate the activity of a host genome-encoded retroviral restriction factor.

IMPORTANCE APOBEC3 proteins are host factors that have a major role in protecting humans and other mammals against retroviruses. These enzymes hinder their replication and intensely mutate their DNA, thereby inactivating viral progeny and the spread of infection. Here we describe a newly recognized way in which some retroviruses protect themselves against the mutator activity of APOBEC3 proteins. We show that gammaretroviruses expressing an accessory protein called glycosylated Gag, or gPr80, use the host's posttranslational machinery and, more specifically, N-linked glycosylation as a way to modulate their sensitivity to mutations by APOBEC3 proteins. By carefully controlling the amount of mutations caused by APOBEC3 proteins, gammaretroviruses can find a balance that helps them evolve and persist.

INTRODUCTION

Retroviruses are exceptional pathogens in that they permanently modify the genome of their host upon infection. Proviral integration can lead to deleterious insertions in the coding sequence of genes and thereby alters the sequence, stability, splicing, and function of host mRNAs. Additionally, because the genomes of retroviruses encode an active promoter and enhancer sequences, they can also influence the expression of nearby host genes, which can lead to diseases such as cancer (1). In response to this potential threat to their genome, vertebrates express several intrinsic antiretroviral restriction factors that are dedicated to the prevention of infection, replication, release, and spread of retroviruses (2). There are several retroviral restriction factors that operate in parallel in mammals; the most notable are BST-2, TRIM5-α, SAMHD1, and APOBEC3 (A3) proteins, as well as several other factors that are dependent on an interferon response for expression (3). To persist, retroviruses therefore need to concomitantly develop countermeasures to all these restrictions factors, and these are often in the form of accessory proteins or genetic substitutions at sites of interactions with restriction factors (1, 4, 5). HIV-1 and HIV-2 are among the retroviruses that have been the most successful at avoiding restriction by the host, whereby BST-2 is defeated by expression of the viral accessory protein Vpu (6, 7), SAMHD1 is defeated by Vpx (8, 9), A3 (A3F, A3G, A3D, and A3H) is defeated by Vif (10), and TRIM5α is defeated by the virus evolving target sequences that avoid its capsid being recognized by this restriction factor (11–14). HIV is not the only retrovirus that successfully counteracts the effects of restriction factors. Some murine leukemia viruses (MLVs) have also developed ways to avoid detection and restriction by host retroviral restriction factors (15–18).

A3 proteins constitute a family of cytidine deaminases that convert deoxycytidines into deoxyuridines in single-stranded DNA (for a review, see reference 19). These proteins have a major role in intrinsic defenses against foreign naked DNA, some DNA viruses, and retroelements but mostly against retroviruses. Humans and primates express seven A3 proteins (A3A, A3B, A3C, A3DE, A3F, A3G and A3H), with A3G being the most potent against retroviruses. Deamination occurs primarily on single-stranded minus-strand viral DNA during reverse transcription, resulting in C-to-U transition mutations. Newly generated uracils then direct the incorporation of adenines on the plus-strand DNA, thereby generating G-to-A mutations. Intense deamination or hypermutation results in inactivating substitutions and premature stop codons in viral genes. A3 proteins can also, under some experimental conditions, inhibit retrovirus infection by multiple deamination-independent mechanisms (20–28).

Mice, in contrast to humans and primates, express APOBEC3 (mouse A3 [mA3]) proteins only from a single gene (29, 30). Although mA3 is able to potently restrict and hypermutate HIV and simian immunodeficiency virus in a Vif-independent manner in vitro (29, 31), it displays a variable and often much weaker ability to restrict murine retroviruses in vivo and in vitro (32–35). This variability is likely due to mA3's prevailing inability to hypermutate most murine retroviruses, and therefore, mA3 does not genetically inactivate circulating viruses. The most striking example is the Moloney MLV (M-MLV) gammaretrovirus, which is modestly restricted by mA3 in the absence of detectable G-to-A hypermutation, despite efficiently packaging mA3 into virions (33, 36–40). However, M-MLV becomes much more sensitive to restriction by mA3 when expression of the glycosylated viral Pr80 Gag (gPr80) protein is suppressed (15, 18). This is reportedly caused by gPr80 blocking access of mA3 to the reverse transcription complex in viral cores (15).

The gPr80 protein of M-MLV appears as a multiband protein with an apparent molecular mass ranging from 80 to 100 kDa on SDS-polyacrylamide gels (41, 42). This glycoprotein is expressed from an alternative in-frame CUG start codon 264 bases upstream of the primary AUG initiation codon of the structural polyprotein Pr65^Gag, resulting in 88 additional amino acids at the N terminus (43). Glycosylated Pr80 is then further processed by cellular proteases that cleave it into a 40-kDa amino-terminal fragment (containing the gPr80 leader polypeptide and viral matrix [MA] and p12 proteins) and a 55-kDa carboxy-terminal moiety (containing the viral capsid [CA] and nucleocapsid [NC]) (44, 45). Although the N-terminal 40-kDa moiety can associate with the cell membrane as a type II integral membrane protein, it is also present in secreted viral particles (43, 45, 46). While the primary function of gPr80 has not been fully elucidated, its described roles include facilitating late-stage viral release from infected cells, increasing viral core stability and integrity, and improving viral spreading and pathogenesis in vivo (15, 38, 47–50).

AKV MLV is an endogenous murine gammaretrovirus that is highly similar in genomic sequence to M-MLV. However, AKV, in contrast to M-MLV, is sensitive to deamination by mA3 (33). In this study, we mapped AKV's sensitivity to deamination and restriction by mA3 and human A3G (hA3G) to the pr80 gene sequence of the virus. We identified three putative N-linked glycosylated sites in gPr80 of M-MLV but only two in that of AKV. Our biochemical and cell-based analyses show that the number of glycosylated sites in gPr80 inversely correlates with the level of resistance to deamination. Abolishing gPr80 expression also resulted in hypermutated virus, indicating that gPr80 and the N-linked glycans attached to it are together an integral part of the resistance mechanism. Additionally, we demonstrate that genetically modified mouse gammaretroviruses with fewer glycosylated sites in gPr80 are mutated by endogenous mA3 expressed in murine primary splenocytes. However, to our surprise, these mutated viruses remained infectious and capable of replication. These results highlight the important contribution of the host's posttranslational modification machinery to helping gammaretroviruses modulate their sensitivity to deamination by mA3. This could constitute a novel strategy used by gammaretroviruses to increase their genetic diversity and avoid immune detection.

MATERIALS AND METHODS

Mice.

All breeding and manipulations performed on animals were conducted in accordance with the Ontario Animals for Research Act and were approved by the University of Ottawa Animal Ethics Committee (protocol number ME-133). A3-deficient (mA3^−/−) mice (mA3-knockout [KO] mice) were backcrossed 12 times to a C57BL/6 mouse background. These mice have the A3 gene disrupted by the insertion of a neomycin resistance cassette in exon 3 (51). C57BL/6 (mA3 wild-type [WT]) and mA3-KO mice were maintained in the barrier unit of the University of Ottawa Animal Care Facility.

Cells.

Human embryonic kidney epithelium (293T) cells and mouse embryonic fibroblasts (NIH 3T3 cells) were cultured in HyClone Dulbecco modified Eagle medium (high-glucose medium) supplemented with 10% decomplemented fetal bovine serum (FBS), 100 U/ml penicillin, and 100 μg/ml streptomycin and propagated at 37°C in a 5% CO₂ incubator. Mouse splenocytes were prepared by homogenizing the spleens of neonatal mouse pups 3 to 5 days of age by enforced passage through a 70-μm-mesh size nylon cell strainer as previously described (33).

Viruses and APOBEC expression vectors.

The pMOV-eGFP expression vector encoding replicative M-MLV and the pAKV-NB-eGFP viral plasmid encoding replicative AKV MLV have been described before (33). Expression vectors for the Flag-tagged C57BL/6 mA3 delta exon 5 allele (referred to throughout as mA3), Flag-tagged human APOBEC expression vectors (hA2 and hA3G), and their respective catalytically inactive mutants (mA3 [E73A] and hA3G [E257A]) have been described before (33, 52). Hybrid viruses were generated by replacing DNA sequences in M-MLV by orthologous sequences from AKV (see Fig. 3A). Viruses with point mutations in the pr80 gene sequence (for M-MLV, N113Q, N480Q, and N505Q; for AKV, N113D and S507N) were made by using a QuikChange XL site-directed mutagenesis kit (Agilent Technologies) according to the manufacturer's specifications. M-MLV (CTA) and AKV (CTA), which do not express gPr80, were generated by replacing the CTG alternative initiation codon in proviral plasmid DNA by CTA using site-directed mutagenesis.

FIG 3.

M-MLV–AKV hybrid viruses reveal that deamination resistance maps to the gag gene. (A) Schematic illustrations of the M-MLV, AKV, and hybrid proviruses. The segments used to construct the three hybrid viruses are indicated. Hybrid viruses are composed of M-MLV into which orthologous segments of AKV have been inserted. Hyb, hybrid; IRES, internal ribosome entry site; PBS, primer binding site; PPT, polypurine tract. (B) Analysis of spread of hybrid virus infection produced in the presence of the various A3 proteins. Infection results are from at least three independent transfections with triplicate infection values for each. Results are presented as the mean ± standard deviation. (C) Comparison of the level of infection with the hybrid viruses relative to that with M-MLV and AKV produced in the presence of mA3 or mA3 (E73A) at 48 h postinfection. Values were normalized individually to the value for infection with virus produced in the presence of hA2 at 48 h. Error bars represent standard deviations. Statistical significance was determined using a two-tailed unpaired Student t test; P values are indicated on the graph. N.S., not significant. (D) 3D-PCR analysis performed on genomic DNA extracted from infected NIH 3T3 cells from the 48-h time point. Representative gels from 3 independent assays are shown.

Viral infection and spreading assays.

The procedures used for the viral infection and spreading assays are graphically presented in Fig. 1.

FIG 1.

Flow charts describing the infection assays. (A) In vitro viral spreading assay. Viruses were produced by plasmid cotransfection in 293T cells and were harvested after 48 h. NIH 3T3 cells were infected with an MOI of 1 in respect to the amount of control viruses produced in the presence of hA2. Infection was measured at 24-h intervals for 72 h. (B) Ex vivo viral spreading assay in primary murine splenocytes. Stably infected NIH 3T3 cells were first generated. Viruses released from these cells were then used to infect primary murine splenocytes. Splenocytes from WT and mA3^−/− (KO) C57BL/6 mice were infected with an MOI of 0.2. Cells were washed at 24 h after infection, and viruses were harvested 96 h later and used to infect NIH 3T3 cells with an MOI of 0.025. Infection was monitored every 24 h for 72 h.

(i) In vitro assay.

To produce replicative viruses in the in vivo assay (Fig. 1A), 3 × 10⁵ 293T cells were seeded in 6-well plates and grown for 24 h until they reached 50 to 60% confluence. Before transfection, the medium was replaced. Cotransfections were done with 800 ng of viral expression plasmid and 100 ng of A3 expression plasmids, using the GeneJuice transfection agent (Novagen) according to the manufacturer's instructions. Virus-containing supernatant was harvested at 48 h after transfection. At 24 h prior to infection, 1 × 10⁵ NIH 3T3 target cells were seeded in 12-well plates and incubated for 24 h. At the time of the infection, NIH 3T3 cell medium was replaced by fresh decomplemented medium containing Polybrene (Sigma-Aldrich) at a concentration of 8 μg/ml. Virus-containing supernatants were cleared by centrifugation, and the concentration of viral p30 CA protein was evaluated by enzyme-linked immunosorbent assay (ELISA; QuickTiter ELISA kit; Cell Biolabs Inc.). The amount of p30 CA protein yielding a multiplicity of infection (MOI) of 1 for virus produced in the presence of A2 was calculated by infection titration. Similar amounts of p30 protein were then used for virus produced in the presence of mA3, hA3G, and their catalytically inactive mutants. The cells were spin infected at 800 × g for 1 h. Infected target cells were grown for 24, 48, and 72 h before being harvested and split into two fractions: one for enhanced green fluorescent protein (eGFP) reporter gene expression using a CyAn ADP flow cytometer (Beckman Coulter) and the other for hypermutation analysis by high-resolution melt (HyperHRM), three-dimensional PCR (3D-PCR), and direct DNA sequencing.

(ii) Ex vivo assay.

Viruses harvested from stably infected NIH 3T3 cells were used as stock for the ex vivo assay (Fig. 1B). At 1 h prior to infection, activated splenocytes were counted and seeded in activation medium (33). Splenocytes were spin infected at an MOI of 0.2. At 24 h postinfection, infected cells were washed with 1× phosphate-buffered saline, pH 7.4, resuspended in complete RPMI with 10 μg of lipopolysaccharide, and incubated for 96 h for virus production. Virus-containing supernatants were passed through 0.45-μm-pore-size cartridge filters, and viral titers (in number of transducing units [TU]/ml) were determined on NIH 3T3 cells by limiting dilution using flow cytometry. The secondary infections (viral spreading assays) were performed on NIH 3T3 cells, using an MOI of 0.025. Cells were analyzed by flow cytometry for eGFP expression at 24, 48, and 72 h postinfection. Genomic DNA (gDNA) was extracted from infected splenocytes at 96 h postwash and from the NIH 3T3 cells at each time point.

Viral APOBEC packaging assays.

Virus-containing supernatants were filtered and then purified by ultracentrifugation through a sucrose cushion as previously described (22). Viral lysates were then processed for immunoblot analysis.

Western blotting.

Details on sample preparation for Western blotting can be found elsewhere (22). Blots were probed with anti-Flag (Sigma-Aldrich), anti-β-tubulin (Abcam), and anti-p30 (clone R187; ATCC). Detection of M-MLV-expressed gPr80 was performed using an anti-p30^CA antibody kindly provided by Hung Fan (University of California, Irvine). Detection of recombinant gPr80-V5 was performed using an anti-V5 polyclonal antibody (catalog number V8137; Sigma).

3D-PCR analysis.

3D-PCR was performed in a two-step protocol using PrimeSTAR high-fidelity polymerase (TaKaRa). A first-round PCR was performed to amplify a 717-bp eGFP amplicon using primers GFP-717 FWD and GFP-717 REV (53). A second-round gradient PCR targeting a 279-bp nested fragment within the eGFP sequence was then performed using primers R279-FWD and GFP-REV. PCR cycles were 94°C for 50 s, followed by 30 cycles of a denaturation gradient from 91.1 to 88.0°C for 50 s, annealing at 56°C for 30 s, and an extension at 72°C for 1 min and a final extension of 72°C for 5 min. Samples that amplified at the lowest denaturation temperature were column purified, cloned using TA cloning (Promega), and sequenced (at the Nanuq Sequencing Facility at McGill University and the Genome Quebec Innovation Center). Deamination intensity graphs were generated by illustrating the number of amplicons that could be detected in a 7-well 3D-PCR gradient set. Confirmation of the presence of a hypermutation in an A3G or mA3 context was done by DNA sequencing (Table 1). Our experimental conditions were optimized so that 3D-PCR performed on integrated proviral DNA of virus alone or virus produced in the presence of catalytically inactive mutant A3 proteins consistently yielded amplification only at the two highest denaturing temperatures of the gradient.

TABLE 1.

Mutation analysis of proviral DNA by direct sequencing^a

Virus	3D-PCR band sequenced	No. of clones analyzed	No. of clones with G-to-A mutations	Total no. of mutationsb		% mutations in a 5′-YC context
				G to A	Other
Virus produced with transfected mA3
M-MLV	2	5	1	1	1	100
M-MLV (N113Q)	3	6	6	31	2	71
M-MLV (N505Q)	3	5	5	15	1	75
M-MLV (N113Q/N505Q)	4	5	5	41	0	59
M-MLV (CTA)	4	5	4	26	2	69
Hybrid 1	2	5	1	1	3	100
Hybrid 2	3	5	5	9	4	67
Hybrid 3	3	5	4	12	2	75
AKV	3	5	5	30	0	67
AKV (N113D)	4	5	5	34	0	77
AKV (S507N)	2	8	3	3	2	67
AKV (CTA)	4	5	5	27	0	70
Virus from Spl.
M-MLV	3	10	7	7	2	29
M-MLV (N113Q/N505Q)	4	8	8	59	3	98
AKV	4	11	10	27	4	74
AKV (S507N)	3	6	2	2	2	50
Virus from 3T3 (Spl. derived)
M-MLV	3	8	0	0	4	0
M-MLV (N113Q/N505Q)	4	7	6	18	3	78
AKV	4	25	23	62	9	84
AKV (S507N)	3	12	6	6	7	84

^aThe clones contain the first 279 bp of the eGFP reporter gene. Viruses were produced in the presence of transfected mA3 or endogenous mA3 in murine splenocytes. Clones displaying identical deamination profiles were excluded from the calculations. Spl., proviral sequences from infected wild-type C57BL/6 mouse splenocytes; 3T3, proviral sequences from NIH 3T3 cells infected with viruses released from splenocytes.

^bMutations compiled on the coding strand of the proviral DNA.

HyperHRM analyses.

HyperHRM analyses were carried out as previously published (53). Briefly, proviral DNA was amplified from the gDNA of virus-infected cells using primers R648-FWD and GFP-REV to generate a 648-bp amplicon. Amplicons derived from each infection condition were then cloned, and colonies positive for an insert were diluted in 250 μl of water, of which 8 μl was used for high-resolution melting analysis in a 96-well plate format. To generate standard curves for mutation quantification purposes, bacterial clones containing defined numbers of G-to-A mutations were used. Amplification by quantitative PCR was then carried out and immediately followed by a melting curve analysis in which DNA amplicons were gradually heated from 72°C to 95°C and fluorescence values were acquired at 0.025°C intervals. Determination of the number of mutations in each clone was achieved by applying the algorithm previously described (53). Unmutated clones were excluded from the calculation of mutation frequency presented in Tables 1 and 2. DNA sequencing of randomly selected proviral clones was performed as a quality control and to ensure that the results were not biased by clonal amplification of a unique sequence (data not shown).

TABLE 2.

HyperHRM analysis of editing of M-MLV, AKV, and glycosylation mutants by mA3

Virus	No. of clones analyzed	No. of mutated clones	% clones mutated	Total no. of mutations	Predicted G-to-A mutation frequency (no. of mutations/kb)
M-MLV	80	3	4	9	0.17
M-MLV (N505Q)	66	29	44	117	2.74
M-MLV (N113Q/N505Q)	71	30	42	145	3.15
AKV	50	20	40	100	3.09
AKV (N113D)	43	32	74	232	8.33
AKV (S507N)	37	2	5	5	0.16

RESULTS

Murine gammaretroviruses display different sensitivities to restriction and deamination by APOBEC3.

M-MLV and AKV are two murine gammaretroviruses with very high sequence identity but with strikingly different sensitivities to restriction and deamination by mA3. Here we performed infection assays with the two viruses being produced in cells expressing human APOBEC3G (hA3G), its catalytically inactive mutant (hA3G [E259Q]), mA3, and its catalytically inactive mutant (mA3 [E73A]). hA2 was used as a negative control because it neither deaminates nor restricts these viruses (22). Viruses were produced by cotransfection in 293T cells, harvested, and normalized for p30 content by ELISA (Fig. 1A). The levels of cell expression and efficiencies of virion packaging of all APOBEC proteins were monitored and comparable (Fig. 2A). Target NIH 3T3 cells, which do not express detectable levels of mA3, were infected at an MOI of 1 in respect to the amount of control viruses produced in the presence of hA2, and infection was assessed by measuring eGFP reporter gene expression over 72 h (Fig. 2B). By using NIH 3T3 cells as targets, viruses that survive the initial restriction by the virion-packaged A3 proteins are then free to replicate and spread over the course of the experiment. Our results showed that all A3 proteins tested, including the catalytically inactive mutants, reduced the level of infection of both viruses at the 24-h time point compared to the level of infection achieved with hA2 (Fig. 2B). This is the result of the well-documented deamination-independent restriction that is especially prominent in tissue culture assays. After 48 h postinfection and likely after a second cycle of viral infection, both viruses remained completely restricted by hA3G but continued to rapidly infect new cells when produced with catalytically inactive hA3G (E259Q) and mA3 (E73A). In contrast, mA3 delayed the ability of AKV to spread, as can be observed by a downward inflection in the growth curve (Fig. 2B) and a nearly 45% reduction in the level of infection compared to that achieved for M-MLV at 48 h (Fig. 2C).

FIG 2.

Sensitivity of M-MLV and AKV to deamination and restriction by A3 proteins. (A) (Top) Expression of the various APOBEC proteins in lysates of cotransfected 293T cells used to produce M-MLV and AKV was analyzed; (bottom) virions harvested from the culture supernatant of 293T cells were analyzed for the efficiency of packaging of the various APOBEC proteins. (B) Analysis of the spread of M-MLV and AKV infection produced in the presence of APOBEC proteins, represented as the percentage of cells expressing the eGFP reporter at the various time points after infection. Infection results are from at least three independent transfections with triplicate infection values for each. Results are presented as the mean level of infection ± standard deviation. (C) Comparison of the relative infection of M-MLV and AKV produced in the presence of mA3 or mA3 (E73A) at 48 h postinfection. Values were normalized to those for infection with virus produced in the presence of hA2 at 48 h. Error bars represent standard deviations. Statistical significance was determined using a two-tailed unpaired Student t test; P values are indicated on the graph. n.s., not significant. (D) 3D-PCR analysis performed on genomic DNA extracted from infected NIH 3T3 cells from the 48-h time point for M-MLV and AKV. Representative gels from 3 independent assays are shown. Td, denaturing temperature; −ve, negative.

We next looked at the intensity of cytidine deamination in integrated provirus DNA. Here we used 3D-PCR, which is a method used to selectively amplify hypermutated DNA sequences on the basis of the premise of reduced PCR amplicon melting temperatures as a consequence of A/T base enrichment due to the deamination of cytosines into uracils (54). 3D-PCR, however, does not provide information on the proportion of sequences mutated in a population. It is a method that makes a qualitative assessment of how intensely proviral sequences are hypermutated. We optimized our assays so that unmutated proviral DNA (virus alone) amplified only at the two highest melting temperatures selected from a gradient of seven possible temperatures (Fig. 2D). Positive amplification at lower melting temperatures is indicative of hypermutated proviral DNA sequences. Under our conditions, the threshold for reliable mutation detection by 3D-PCR is 3 or more G-to-A/C-to-T mutations per 279-bp sequence (53). For amplicons indicative of hypermutation, the presence of transition mutations in an A3 deamination context (5′-CC or 5′-TC, where the boldface C represents the deaminated C residue) was confirmed by DNA sequencing (Table 1). 3D-PCR assays were repeated at least 3 times from independent infection assays, and representative gel images are presented. Our results show that hA3G deaminates both M-MLV and AKV at the same intensity, meaning that the most intensely mutated sequences have similar numbers of G-to-A mutations (Fig. 2D). Catalytically inactive A3G (E259Q) was used as a negative control, and virus produced with A3G (E259Q) displayed amplification only under the first two melting conditions, similar to the findings for virus produced alone or with hA2 (data not shown). On the other hand, mA3 weakly hypermutated AKV, as judged by the presence of amplification in the 3rd lane in Fig. 2D, and did not at all hypermutate M-MLV.

Resistance to deamination maps to the gPr80 accessory protein.

In order to identify the regions of AKV that are responsible for rendering the virus sensitive to deamination by mA3, we generated three hybrid viruses by progressively replacing the proviral DNA of M-MLV with that of AKV (Fig. 3A). We then performed an infection assay under the same conditions described above. We found that A3 protein encapsidation was similar for all hybrid viruses (data not shown). Viral infection assays showed that hybrid 1, containing an AKV segment spanning from the R region to the N terminus of Gag, spread similarly to wild-type M-MLV (Fig. 3B). Hybrids 2 and 3, however, displayed a delay in spreading at the 48-h time point, similar to that seen for AKV, with, respectively, 18% and 24% reductions in the levels of infection relative to that for M-MLV being seen (Fig. 3C). Catalytically inactive mutants A3G (E259Q) and mA3 (E73A) did not have a significant effect on hybrid virus spread (Fig. 3B and C). 3D-PCR analysis performed on infected target cells revealed that hA3G hypermutated all three hybrids with the same intensity, while mA3 was able to hypermutate only hybrids 2 and 3 to levels similar to those for AKV (Fig. 3D).

The common proviral DNA segment of hybrids 2 and 3 maps to the gag gene of AKV, which includes sequences coding for both the gPr80 and Pr65 polyproteins. Because the spread curves and deamination intensities were similar between the two hybrids, this led us to conclude that all determinants responsible for the contrasting phenotypes between M-MLV and AKV are located within this region. Two recent studies have highlighted the involvement of the gPr80 Gag protein in resisting restriction by mA3 (15, 18). To distinguish if resistance to deamination was conferred by elements within the Pr65 or gPr80 polyproteins, we generated viral mutants, termed CTA mutants, that do not express gPr80 because their CTG initiation codon was replaced by a CTA trinucleotide (Fig. 4). Packaging of A3 proteins into these mutant viruses was similar to that for their WT counterparts (Fig. 4A). Restriction assays and growth curves of both viruses showed that they had similar profiles that closely resembled the profile for AKV (Fig. 4B and C). However, there was a small but noticeable decrease in infectivity of these mutants compared to that of the WT parental viruses. M-MLV and AKV produced with A2 infected 65 to 70% of the cell population after 24 h (Fig. 2B). With the CTA mutant viruses, this infection was reduced to 40% when the same amount of input virus (as measured by p30 ELISA) was used (Fig. 4B). Although both the CTA and CTG codons code for leucine, the position of this trinucleotide is located in the stem of one of the major stem-loops involved in packaging the viral RNA dimer, which could explain this decrease in infectivity (55). Hypermutation analysis of the M-MLV CTA and AKV CTA mutants revealed an increase in sensitivity to hypermutation by both mA3 and hA3G (Fig. 4D). This therefore allowed us to exclude the possibility of any potential roles for elements within Pr65 to be the cause for the resistance to deamination and to focus our attention on gPr80.

FIG 4.

Analysis of M-MLV and AKV mutants that do not express gPr80. (A) (Top) Western blots showing the expression of the various APOBEC proteins in lysates of cotransfected 293T cells used to produce M-MLV (CTA) and AKV (CTA) mutant viruses are shown; (bottom) viruses (virions) harvested from the culture supernatant of 293T cells were analyzed for the efficiency of packaging of the various APOBEC proteins. (B) Analysis of the spread of infection of mutant viruses produced in the presence of the various A3 proteins. Infection results are from at least three independent transfections with triplicate infection values for each. Results are presented as the mean ± standard deviation. (C) Comparison of the level of infection with the CTA mutant viruses relative to that with M-MLV and AKV produced in the presence of mA3 or mA3 (E73A) at 48 h postinfection. Values were normalized individually to the values for infection with virus produced in the presence of hA2 at 48 h. Error bars represent standard deviations. Statistical significance was determined using a two-tailed unpaired Student t test; P values are indicated on the graph. (D) 3D-PCR analysis performed on genomic DNA extracted from infected NIH 3T3 cells from the 48-h time point for the various viruses. Representative gels from 3 independent assays are shown.

Three N-linked glycosylated sites in M-MLV gPr80 are required for complete resistance to deamination.

Because the amino acid sequences of the gPr80 proteins of M-MLV and AKV are highly similar, we focused our attention on differences in the N-linked glycosylation patterns of the proteins (Fig. 3). We carried out an in silico analysis of the gPr80 proteins of both viruses using the NetNGlyc server to identify putative glycosylated sites within an Asn-XAA-Ser/Thr (where XAA is not Pro) sequon. We identified three likely glycosylated sites in the gPr80 of M-MLV (N113, N480, and N505) but only two in that of AKV (N113 and N482). N113 is located in the matrix (MA), while N480/N482 and N505 are located in the capsid (CA).

Using site-directed mutagenesis, we generated point mutants for each of the predicted glycosylated amino acids of gPr80 of M-MLV and AKV. We also introduced an additional N-linked glycosylated site at position S507 of AKV (AKV [S507N]), which is in a sequon favorable for glycosylation. The asparagine-to-aspartic acid substitutions chosen remove the glycosylated site, while they introduce a chemically conservative mutation. It should also be noted that changes in the gPr80 sequence downstream of amino acid 88 are also reflected in the Pr65 Gag protein. We also generated double and triple glycosylation mutants when appropriate. Western blot analyses of transfected cell extracts clearly show that the apparent molecular mass of gPr80 shifts according to the number of putative glycosylated sites mutated (Fig. 5A to C). Glycosidase treatment of the extracts reveals a gPr80 band lower than that of the gPr80 band of the M-MLV (N113Q/N505Q) double mutant, which indicates that amino acid N480 is also likely glycosylated, as predicted (Fig. 5A and B). The M-MLV (N113Q/N480Q/N505Q) triple glycosylation mutant demonstrated a band for gPr80 at the same height as that in the glycosidase-treated sample (Fig. 5C). Western blot analysis of AKV gPr80 showed that the S507N mutation increases the size of the band, suggesting that N-linked glycosylation at this site was indeed restored (Fig. 5B).

FIG 5.

Identification of N-linked glycosylated sites in M-MLV and AKV. (A to C) Western blot analysis of transfected 293T cell extracts showing the altered migration patterns caused by N-linked glycosylation. The last lane of each blot shows either M-MLV- or AKV-infected cell extracts treated with glycosidase. The gPr80 protein band for each virus is indicated by red arrows. Numbers in red indicate the number of glycosylated residues.

Restriction assays with the M-MLV single point mutants and the N113Q/N505Q double mutant clearly showed a delay in infection spread at 48 h with mA3 (Fig. 6A). The N113Q/N505Q double mutant was about twice as sensitive to mA3 restriction as the single point mutants (Fig. 6C). Mutant M-MLVs produced with mA3 (E73A) showed little or no difference in infectivity compared to that of the wild-type virus (Fig. 6C). Restoring glycosylation at position S507 of AKV increased the spreading kinetics to resemble that of mA3-resistant M-MLV (Fig. 6B); however, the N113D substitution did not render the virus more sensitive to mA3 restriction (Fig. 6D). Viral mutants M-MLV (N480Q) and AKV (N482D) displayed very poor infectivity (less than 2%), despite being efficiently released from the cells, as judged by p30 ELISA (data not shown). For this reason, these mutants were not analyzed further.

FIG 6.

Intensity of gPr80 glycosylation correlates with sensitivity to deamination by mA3. (A and B) Analysis of spread of infection of glycosylation mutants of M-MLV (A) and AKV (B) coproduced with various A3 proteins. (C and D) Comparison of the infection with the mutant viruses relative to that with M-MLV (C) and AKV (D) produced in the presence of mA3 or mA3 (E73A) at 48 h postinfection. Values were normalized individually to those for infection with virus produced in the presence of hA2 at 48 h. Error bars represent standard deviations. Statistical significance was determined using a two-tailed unpaired Student t test; P values are indicated on the graphs. n.s., not significant. (E and F) 3D-PCR analysis performed on genomic DNA extracted from infected NIH 3T3 cells from the 48-h time point for glycosylation mutants of M-MLV (E) and AKV (F).

Although the sequential mutation of glycosylated residues in gPr80 had an overall modest effect on the sensitivity to restriction by mA3, a more pronounced impact on hypermutation could clearly be observed. Both mA3 and hA3G mutated the M-MLV and AKV mutants with an intensity higher than that for their WT counterparts (Fig. 6E and F). All mutants with N-to-Q point mutations of M-MLV gPr80 became more sensitive to deamination; this effect was slightly increased in the M-MLV (N113Q/N505Q) double mutant and the AKV (N113D) mutant. AKV (S507N), which had three glycosylated sites, became resistant to deamination (Fig. 6F).

To further characterize the intensity and frequency of deamination in proviral sequences, we performed mutation analyses on individual clones isolated from infected cells. We used HyperHRM analysis, which is a high-throughput method that we developed to quantify the number of A3-induced mutations in a PCR amplicon (53). Confirming the 3D-PCR data, we found that there was an inverse correlation between the number of glycosylated residues in gPr80 and the proportion of hypermutated sequences in a specific virus (Fig. 7A and B and Table 2). Our data also show that the intensity of hypermutation increases when fewer glycosylated sites are present in either virus.

FIG 7.

N-linked glycosylation inhibits deamination and deamination-independent restriction. (A and B) Analysis of mutation intensities in proviral DNA by HyperHRM analysis. The histograms depict the proportion of total sequences containing the indicated number of mutations. The results of clone analysis are presented in Table 2. (C and D) Relative infection depicting the effect of gPr80 mutants on virus sensitivity to restriction at 48 h postinfection. Infection levels from experiments whose results are presented in Fig. 2B, 4B, and 6A and B were normalized to those for hA2 for each virus. Values presented are the means ± SEMs from three independent transfection experiments with triplicate infection samples. The Student unpaired t test was performed to assess statistical significance: *, P = 0.05; **, P = 0.19; ***, P = 0.02.

gPr80 antagonizes both arms of mA3 restriction.

Having identified M-MLV mutants that are sensitive to deamination, we next wanted to evaluate how gPr80 glycosylation affects deamination-independent restriction by mA3. Here we normalized the infection data at the 48-h time point to those for the hA2 control independently for each virus set: M-MLV (Fig. 7C) and AKV (Fig. 7D). We found that viruses were more restricted by catalytically inactive mA3 (E73A) when the gPr80 protein had fewer glycosylated sites or was not expressed altogether. The increased sensitivity to restriction of the CTA mutant viruses by mA3 (E73A) could reflect the protective effect of N480 glycosylation. Altogether, these results clearly show that N-linked glycosylation of gPr80 prevents restriction by deamination and also by deamination-independent mechanisms.

Endogenous mA3 inhibits MLV infection but is also a source of genetic diversification.

Two reports have already established that the antiretroviral activity of mA3 is inhibited by gPr80 in vivo (15, 18). These studies were carried out by comparing the infectivity of deamination-resistant MLV strains with coisogenic strains unable to either express or produce a full-length gPr80 protein. Here we asked whether endogenous levels of mA3 are sufficient to hypermutate deamination-sensitive MLVs to affect viral fitness and the replication of released viruses. We performed our experiments only with MLVs that expressed gPr80 in order to measure the direct contribution of N-linked glycosylation to the resistance.

We isolated splenocytes from neonatal C57BL/6 mice with WT mA3 or mA3-deficient (mA3-KO) neonatal C57BL/6 mice. The splenocytes were cultured, activated, and then infected with a similar MOI with viruses that were produced in the absence of mA3 in NIH 3T3 cells (Fig. 1B). The infectious titer of the viruses released by the splenocytes, measured as the number of transducing units per ml of supernatant harvested, was assessed 96 h later (Table 3). M-MLV was generally more efficient at replicating in these splenocytes than AKV. Differences in titers were also presented as relative infectivity to better illustrate that endogenous mA3 had a similarly potent effect at reducing the infectious titers of all viruses, including those resistant to deamination (Fig. 8A). This indicates that deamination by endogenous mA3 has little effect on viral restriction under these conditions.

TABLE 3.

Titers of gammaretroviruses released from splenocytes^a

Virus	Source	Virus titer (no. of TU/ml [105])a
M-MLV	WT	0.047 ± 0.006
	KO	5.099 ± 1.964
M-MLV (N113Q/N505Q)	WT	0.037 ± 0.007
	KO	3.896 ± 1.740
AKV	WT	0.025 ± 0.005
	KO	1.796 ± 0.615
AKV (S507N)	WT	0.012 ± 0.002
	KO	0.302 ± 0.053

^aTU, transducing units; WT, wild type C57BL/6 mouse splenocytes; KO, mA3^−/− C57BL/6 mouse splenocytes. Values represent the means of triplicate infection values ± standard deviation (for KO mice, n = 5; for WT mice, n = 6).

FIG 8.

gPr80 glycosylation levels influence the intensity of deamination by endogenous mA3 expressed in mouse splenocytes. (A) Relative infectivity of viruses released from infected splenocytes. The viral titers used to generate the figure are presented in Table 3. (B) 3D-PCR analysis performed on genomic DNA extracted from infected KO (left) or WT (right) C57BL/6 splenocytes. Representative gels from 3 independent assays are shown. (C) Analysis by HyperHRM of mutation intensities in proviral DNA isolated from infected C57BL/6 splenocytes. The histograms depict the proportion of total sequences containing the indicated number of mutations. The results of clone analysis are presented in Table 2. (D) Infection spread in NIH 3T3 cells. Cells were infected with an MOI of 0.2 with viruses released from the splenocytes for which the results are depicted in panel A. Infection was measured every 24 h for 72 h. The graphs represent nonlinear regression curves of the infection. Results are presented as the mean level of infection ± standard deviation. (E) 3D-PCR analysis performed on genomic DNA extracted from infected NIH 3T3 cells. Representative gels from 3 independent assays are shown. (F) Analysis by HyperHRM of mutation intensities in proviral DNA isolated from infected NIH 3T3 cells.

To ensure that hypermutation of proviral DNA did occur in the WT splenocytes, 3D-PCR was carried out on proviral DNA from splenocytes at 96 h postinfection (Fig. 8B). Hypermutated viral sequences were detected only in WT splenocytes at various intensities for all viruses. Surprisingly, M-MLV and AKV (S507N) also showed evidence of hypermutation caused by mA3, but the intensity and frequency of the mutations were very low (Table 1). These data were also supported by HyperHRM analysis of individual clones, showing that they contained several transition mutations (Fig. 8C and Table 4).

TABLE 4.

HyperHRM analysis of editing of M-MLV, AKV, and glycosylation mutants by endogenous mA3 expressed in C57BL/6 splenocytes^a

Infected cell and virus	No. of clones analyzed	No. of mutated clones	% clones mutated	Total no. of mutations	Predicted G-to-A mutation frequency (no. of mutations/kb)
WT splenocytes
M-MLV	81	12	15	38	0.73
M-MLV (N113Q/N505Q)	53	15	28	59	1.72
AKV	40	19	48	52	2.00
AKV (S507N)	63	8	13	21	0.51
NIH 3T3 cells
M-MLV	42	1	3	4	0.15
M-MLV (N113Q/N505Q)	60	6	10	16	0.41
AKV	39	10	26	30	1.19
AKV (S507N)	63	5	8	21	0.51

^aHyperHRM analysis was performed on gDNA from infected WT splenocytes 96 h after infection. NIH 3T3 cells were infected with viruses released from WT splenocytes. HyperHRM analysis was performed 48 h later.

To address whether the replicative fitness of MLVs sensitive to deamination was affected by being produced in splenocytes expressing mA3, we then used these viruses to infect NIH 3T3 cells with an MOI of 0.025. We monitored their spread in culture every 24 h for 72 h. These assays revealed that the pool of infectious viruses released from WT splenocytes did not exhibit a proliferation efficiency different from that of viruses released from mA3-KO cells (Fig. 8D). These results were confirmed by statistical analyses indicating that the slopes of the viral spreading curves were similar for viruses produced in WT or KO splenocytes (data not shown). We also analyzed proviral DNA sequences amplified from infected NIH 3T3 cells. We found by 3D-PCR clear evidence of hypermutation in viruses originating from WT splenocytes (Fig. 8E and Table 1). HyperHRM analysis revealed that the intensity and frequency of the transition mutations were, however, lower (Fig. 8F and Table 4). The mutation frequency for AKV dropped from 2 mutations per kb in splenocytes to 1.19 mutations per kb in NIH 3T3 cells (Table 4). This indicates that more heavily mutated viruses were selected against and were not represented in NIH 3T3 cells but a significant proportion of sublethally mutated viruses still continued to replicate.

DISCUSSION

The aim of this study was to identify what makes AKV different from M-MLV in respect to its sensitivity to deamination by mA3. Our results have clearly revealed that the number of N-linked glycosylated sites in the gammaretrovirus-specific gPr80 accessory protein inversely correlates with the intensity of mA3-induced deamination. This therefore supports the concept that gammaretroviruses employ the host's protein glycosylation machinery to protect themselves against innate restriction by mA3. Considering that the gPr80 protein of MLVs also reduces the potency of the mutator activity of human A3G, N-linked glycosylation may therefore be part of a broader strategy allowing gammaretroviruses to increase the success of zoonotic transmission between hosts of different species.

M-MLV is not the only murine retrovirus that is resistant to deamination by mA3. In fact, most mouse retroviruses, including the prevalent Friend MLV, which has the same number of putative gPr80 glycosylated sites as M-MLV, are resistant (Fig. 9A) (34). The xenotropic MLV-related virus (XMRV) is perhaps the only other currently known exception, along with AKV, of a murine retrovirus that is sensitive to deamination by mA3 (56). However, XMRV, contrary to AKV, is not an endogenous murine retrovirus with a long history of coevolution with its host but, rather, a recently emerged laboratory virus that was inadvertently created through provirus (PreXMRV-1 and PreXMRV-2) recombination during human xenograft passages in mice (57). What is further interesting about XMRV is that it has a deletion in the leader sequence of the pr80 gene, as do some other murine endogenous retroviruses (ERVs), and therefore does not express gPr80 or any alternate form of a glycosylated Gag-like protein (58). The mouse mammary tumor virus (MMTV) is a betaretrovirus that is restricted by mA3 but is resistant to deamination (35). Although the MMTV genome does not encode a glycosylated Gag protein, it does encode a superantigen (SAg) with multiple N-linked glycosylated residues that could be packaged into virions and act as a functional surrogate to gPr80 (59). It would also be interesting to investigate whether N-linked glycosylation of virion-packaged proteins other than gPr80 can also inhibit A3 deaminase activity.

FIG 9.

Phylogenetic analysis of the gPr80 amino acid sequence of various MLVs. (A) Detailed alignment of MLVs able to produce gPr80. Residues at positions homologous to putative glycosylated sites for M-MLV are highlighted in orange. Amino acids different from asparagine at these positions are highlighted in yellow. N60 is a putative glycosylated site for several MLVs of the Friend-Moloney-Rauscher (FMR) group. (B) The amino acid sequences of exogenous and endogenous MLVs were aligned by use of the DNAStar Lasergene (version 8) MegAlign program (Clustal W method), and a phylogenetic tree was generated. Green shading, exogenous retroviruses belonging to the Friend-Moloney-Rauscher group; yellow shading, endogenous MLVs that produce a full-length gPr80; purple shading, endogenous MLVs unable to produce gPr80 because of deletions in the sequence. Endogenous MLV subgroups were xenotropic (Xmv), ecotropic (Emv), polytropic (Pmv), and modified polytropic (Mpmv). FrMLV and Fr-MLV, Friend MLV.

The exact role of the gPr80 protein of MLVs has long remained enigmatic. Early studies investigated various functional aspects of the gPr80 protein by comparing viral release, infectivity, and virulence in the presence or absence of the protein or by exchanging gag gene segments between different MLV strains (40, 49, 50, 58, 60, 61). The general conclusion of these studies was that the absence of a functional and full-length gPr80 decreases virus release and infectivity in vitro, but especially in vivo. However, recent studies have revealed new and important roles of this accessory protein in helping the virus evade innate immune defenses. Two studies have shown that gPr80 inhibits the antiviral activity of mA3 (15, 18). Stavrou et al. took these findings further by showing that gPr80 acts by preventing mA3 from accessing the reverse transcription complex in viral cores and also by hiding replication intermediates from cytosolic sensors (15). Our work here adds to current knowledge by showing that the N-linked glycans attached to gPr80 are, in fact, essential for its activity against host innate defenses. MLVs that express few N-linked glycosylated sites in gPr80 are more susceptible to being hypermutated and restricted by mA3.

In contrast to hA3G, which stringently restricts MLV infection and spread in human cells, mA3 does not inactivate mouse retroviruses with the same potency. All MLVs that we tested, including those with fewer glycosylated sites in gPr80, and that were produced in the presence of mA3 spread to nearly all target cells 72 h after infection. Differences in the intensity of the deamination cannot alone explain this striking difference. In a previous study, we reported that hA3G mutated AKV at an average rate of 8 mutations per kb of proviral DNA; here we have shown that AKV (N113D) is also mutated to that level by mA3 (Table 2) (33). It is therefore likely that a factor in addition to deamination intensity, such as the consensus DNA sequence being deaminated, is also required for viral genetic inactivation. While mA3 deaminates cytidines preferentially in a 5′-TC context, hA3G prefers 5′-CC (26, 33, 62). With the tryptophan codon being TGG (5′-CCA when it is read on the minus-strand DNA), hA3G is much better suited than mA3 at generating all of the three termination codons. Mutations caused by mA3 therefore appear to have subtler and less deleterious effects on viral replicative fitness than hA3G-induced mutations, which are almost always lethal for the virus.

An interesting result that emerged from the ex vivo assays is that infectious viruses with low levels of mutations have replicative fitness similar to that of nonmutated viruses (Fig. 8D). In a previous study, we showed how AKV released from WT splenocytes expressing endogenous mA3 was restricted when analyzed 48 h after infection in NIH 3T3 cells (33). Here we took these results further by looking at the impact of mutations on virus spread over time. By normalizing input infectious viral titers, we compared the infection efficiency of a pool of mA3-mutated viruses to that of nonmutated viruses. Viruses that infected target NIH 3T3 cells had, as would be expected, fewer mutations than those recovered from the splenocytes, but nevertheless, a relatively large proportion (26% for AKV) had between 1 and 3 mutations per sequence analyzed (Fig. 8C and F). It therefore appears that viruses containing low levels of mutations can be as infectious as unmutated viruses under these experimental conditions. In the context of a natural infection, such mutations could be beneficial and help the virus evade immune defenses.

These results now raise the question as to whether gammaretroviruses that are permissive to low levels of deamination, by way of less intensely glycosylated gPr80, have a greater success at infecting and persisting in vivo. In support of this concept, a prior study in which the authors performed a phylogenetic analysis of the nucleotide sequence of various MLV pr80 genes revealed that clade A, B, and C murine ERVs (xenotropic, polytropic, and modified polytropic viruses) are descendants of the same ancestral progenitor, while viruses of the Friend-Moloney-Rauscher (FMR) group come from a different progenitor (58). If we now look instead at the amino acid sequence of their respective gPr80 proteins, we find that clade B and C endogenous retroviruses (containing AKV) cluster together and that their genomes generally encode a gPr80 with fewer putative glycosylated sites than the gPr80 proteins of the FMR group (Fig. 9A and B). The only exceptions are Rauscher MLV and amphotropic MLV, which contain only two putative glycosylated sites. However, it remains to be experimentally determined whether these viruses are sensitive or resistant to deamination by mA3 and also whether they in fact contain only two glycosylated sites. Additionally, these two viruses have a serine at position 60, whereas Friend MLV has an asparagine in a perfect N-glycosylation sequon. It would therefore be interesting to investigate the role of this variable residue, especially if Rauscher MLV and amphotropic MLV were, in fact, shown to be sensitive to deamination. Overall, these observations indicate that moderate sensitivity to mA3 deamination may have provided a subset of exogenous gammaretroviruses with an additional strategy to increase their genetic diversity. As such, permitting sublethal mA3 mutations could help these gammaretroviruses persist in their murine host and thereby increase their chances of becoming endogenized in the mouse germ line. This concept that sublethal levels of A3-induced deamination could be a driving force behind retroviral/HIV evolution, drug resistance, and immune escape has been raised several times before (63–68).

How the N-linked glycans of gPr80 prevent A3-induced deamination is still unknown. Although gPr80 is cleaved by cellular proteases, with its N-terminal extremity (containing N113) being packaged into virions (46), it is intriguing that the secreted C-terminal fragment containing glycosylated residues N480 and N505 also prevents deamination. It remains to be established whether gPr80 and mA3 interact. If mA3 were to facilitate the packaging of the full-length gPr80 protein or even that of each of its cleaved polypeptides, then a case could be made that gPr80 glycans act collectively to physically block A3 proteins from accessing the viral cDNA substrate in the reverse transcription complex. Further studies are required to fully understand how N-linked glycans on gPr80 impede the deamination of gammaretroviruses by A3 proteins. Such efforts could potentially lead to the identification of a new class of A3 inhibitors.