To examine the susceptibility of primary HIV-2 to hTRIM5α, we constructed chimeric clones expressing plasma-derived HIV-2 CA coding sequences in the context of an HIV-1 backbone. The backbone was pNL4-3-ΔENV-LucR 43 , which permits optimal virus production and highly sensitive quantification of HIV-1 infectivity on a single-cycle basis. As shown on Figure 1, HIV-2 or SIV CA coding sequences (aa 17 to 224) were inserted between the HIV-1 MA/CA and CA/p2 cleavage site sequences, thereby preserving the HIV-1 identity of these sequences, and allowing optimal proteolytic cleavage of the HIV-2 CA by the HIV-1 protease expressed by the pNL4-3-derived vector. The HIV-2 and SIV CA sequences were categorized as follows: i) CA from laboratory adapted HIV-2 strains GL-AN and ROD10; ii) CA from primary plasma-derived HIV-2 sequences, including 8 subtype A and 7 subtype B viruses, iii) CA from HIV-2 CRF01_ AB NMC (Nagoya Medical Center) 307, NMC716 and NMC842 viruses, iv) synthetic CA sequences from 6 non-epidemic HIV-2 (subtypes C to H), 4 SIVsmm strains, and SIVmac239, and v) CA from 9 primary HIV-1 belonging to subtypes B and CRF02_AG.

CA maturation in the chimeric viruses was analyzed by western blot after purification by ultracentrifugation of virions over a 20% sucrose cushion, normalization for RT content by ELISA and reaction with a mixture of human sera from HIV-2 positive patients. As shown on Figure 2A, chimeric virions contained high amounts of fully mature p26 CA protein, together with small amounts of higher molecular weight Gag cleavage intermediates. The ratio of fully processed CA to cleavage intermediates was comparable to that observed for pNL4-3 particles, although the pattern of incompletely processed Gag proteins was different in the two virion preparations. The profile of incompletely cleaved Gag proteins was similar among the different HIV-2 CA-expressing chimeric virions studied.

To further validate that the mature capsids of these chimeric viruses had biological properties expected from HIV-2, we also examined their susceptibility to the cyclophilin A (CypA) antagonist Debio-025. Indeed, while most HIV-1 strains have been described as being susceptible to CypA antagonists, the replication of HIV-2 has been consistently reported to be unaffected by these antagonists, a property that is related to specific genetic traits of its CA protein 44 45 46 . As expected, and in contrast with the HIV-1 control virus, no detectable inhibition of viral infectivity by Debio-025 was observed for the two chimeric viruses carrying HIV-2 CA sequences tested (Figure 2B). This observation indicates that these viruses had acquired biological properties consistent with that of expression of a functional HIV-2 CA protein, making them suitable for hTRIM5α susceptibility testing.

Infectivity of recombinant viruses carrying HIV-2 and SIV CA sequences was tested in a luciferase-based single-cycle assay in U373-X4-hTRIM5γ cells, in which the antiviral activity of hTRIM5α is offset by stable overexpression of hTRIM5γ. Stocks of viruses were produced by transfection of 293T cells and normalized for RT content by ELISA. To ensure that no viral propagation would occur in the target cell cultures, virus particles were pseudotyped with VSV-G. The results of viral infectivity obtained using U373-X4-hTRIM5α cells, which were pretreated with IFNα before infection, are presented on Figure 3A. The viruses were grouped according to the origin of their CA proteins: laboratory-adapted HIV-2 strains (n=2), HIV-2 clinical isolates from subtype A (n=8) or from subtype B (n=7), CRF01_AB (n=3), synthetic sequences from HIV-2 viruses from non-epidemic subtypes C to H (n=6), synthetic sequences from SIVsmm (n=4) and from SIVmac239. The results were compared to those obtained for NL4-3 and to those from a panel of recombinant viruses expressing CA from a panel of primary HIV-1 plasma sequences. This panel included recombinant virus NRC10-5, a previously described NL4-3 derived virus carrying the CA sequences from a clinical isolate that expresses high susceptibility to hTRIM5α 12 13 . For the HIV-2 CA recombinants, infectivity values ranged from 1×10⁴ to 8×10⁶ relative light units (RLU)/0.2 ng of RT (Figure 3A), but no virus group displayed infectivity levels that differed significantly from those of any of the other groups (p > 0.05). The mean infectivity values of viruses carrying primary HIV-1 CA and NL4-3 CA were 6×10⁴ and 3×10⁸ RLU/0.2 ng of RT, respectively. Among the panel of our chimeric viruses, one of the seven viruses carrying CA from a primary HIV-2 subtype B and one of the four viruses carrying CA from a SIVsmm strain could not be tested phenotypically for their susceptibility to hTRIM5α since their infectivities were less than 4×10⁴ RLU/0.2 ng of RT (Figure 3A).

Susceptibility to hTRIM5α was expressed as the ratio of viral infectivity measured in U373-X4-hTRIM5γ cells to that measured in U373-X4-LacZ control cells. This method has been shown by our laboratory to be fully valid for hTRIM5α susceptibility testing of viruses carrying primary HIV-1 Gag sequences 12 13 . Susceptibility to hTRIM5α of viruses expressing HIV-2 CA is shown on Figure 3B. Overall, susceptibility of most HIV-2 CA-carrying viruses was greater than that measured for HIV-1 NL4-3. Highest susceptibility values were found with viruses expressing CA from laboratory-adapted HIV-2 strains GL-AN and ROD10, respectively 6.4-fold (p = 0.0086) and 8.9-fold (p = 0.0016) more susceptible to hTRIM5α than NL4-3. The lowest susceptibility values were measured in viruses CA from non-epidemic HIV-2 subtypes (mean = 3.3). Susceptibility values for this series of viruses were markedly heterogeneous, ranging from 7.5-fold for CA from HIV-2 subtype F to 1.4-fold with CA from HIV-2 subtype G. Such heterogeneity was not observed for viruses expressing SIVsmm CA. As expected, hTRIM5α susceptibility values obtained for viruses carrying primary HIV-1 CA was low (mean = 2.0-fold), including 8 viruses for which susceptibility values ranged 1.4- to 2.3-fold. Consistent with previous observations, the susceptibility of NRC-10, which expresses mutations conferring CTL resistance in the HLA-B*27-targeted epitope KK10, was 4.7-fold. Overall, susceptibility to hTRIM5α of primary HIV-1 CA-carrying viruses was significantly lower than that measured for recombinant viruses carrying CA from either HIV-2 subtype A (p = 0.001), HIV-2 subtype B (p = 0.008) or HIV-2 CRF01_ AB (p = 0.009), respectively.

Quantification of plasma HIV-2 viral load (VL) was performed by a method derived from that described by Damond et al. 34 , using a modified set of primers targeting the 5'-long terminal repeat (5'-LTR) region with a detection limit of 10 HIV-2 RNA copies/ml (Figure 4). The quantification of viral load was performed on plasma samples obtained from 8 HIV-2 subtype A-infected and 6 HIV-2 subtype B-infected patients for whom we were able to reconstruct chimeric viruses and measure their sensitivity to hTRIM5α. HIV-2 viral loads ranged from 1 to 5.15 Log₁₀ RNA copies/ml, and were higher for HIV-2 subtype A-infected patients than for HIV-2 subtype B-infected patients (mean values of 3.6 Log₁₀ and 2.1 Log10 RNA copies/ml, respectively, p < 0.02). However, no correlation was found between HIV-2 viral load and susceptibility to hTRIM5α of the chimeric viruses carrying the CA from these HIV-2 infected patients.

Song et al. 41 reported that HIV-2 propagation in human cells overexpressing hTRIM5α is less efficient for viruses carrying a proline at position 119 of CA protein, as compared to that of viruses expressing a glutamine or an alanine at this position. More recently, studies by Onyango et al., evaluating HIV-2-infected patients from Guinea Bissau, found that the concommittent presence of a proline at CA positions 119, 159 and 178 is more frequent in subjects with low VL while non-proline aminoacids at these three sites are more frequent in subjects with high VL 42 . The authors suggested that this relationship could be explained by differences in susceptibility to hTRIM5α.

To explore the potential impact of these variations on the susceptibility to hTRIM5α, we grouped our HIV-1 chimeric viruses according to whether the residues at CA positions 119, 159 and 178 were a proline (P) or non-proline (N) amino acid. For these studies, the panel of HIV-1 chimeric viruses studied was increased by including of variants harboring CA sequences derived from minority populations present in the plasma from some of the infected patients. As shown on Figure 5A, the mean hTRIM5α susceptibility ranged from 4.4 (NPP viruses) to 7.8 (NPN viruses). Considerable overlap between the different groups was observed, however, and these differences were not statistically significant. In addition, no correlation was observed between the number of proline residues present at these three positions and hTRIM5α susceptibility.

We also analyzed the impact on hTRIM5α susceptibility of amino acid substitutions involving one or more of these proline residues. Mutagenesis was performed on the recombinant virus carrying CA from the GL-AN strain, which, like its parental strain GH123 47 , harbored proline residues at CA positions 119, 159 and 178. Modifying individually these proline residues (P119A, P159S, P178A) did not significantly change viral susceptibility to hTRIM5α, although the mutation P178A induced a 1.3-fold increase compared with that of CA-GL-AN (Figure 5B). Similarly, the double mutant P159S-P178A and the triple mutant P119A-P159S-P178A had comparable hTRIM5α susceptibilities (mean values of 7.0 and 6.4, respectively) that were not significantly different from that of the CA-GL-AN virus (mean value of 6.4).

Alpha interferon (IFNα) (Sigma-Aldrich) was dissolved in deionized water (final concentration, 2 × 10⁵ U/ml), and stored in aliquots at −80°C. Debio-025 (kindly provided by Debiopharma, Lausanne, Switzerland) was dissolved in dimethyl sulfoxide (DMSO) at a final concentration of 10 mM, and stored in aliquots at −80°C.

293T and U373-X4 cell lines were cultured in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal calf serum, 100 U/ml penicillin G and 100 μg/ml streptomycin. The two U373-X4 cell lines used as target cells, have previously been described 12 : (i) U373-X4 cells over-expressing human TRIM5γ (hTRIM5α cells), which has a dominant-negative effect against endogenous human TRIM5α; (ii) U373-X4 cells over-expressing LacZ (LacZ cells), which were used as a control. Human TRIM5γ cells and LacZ cells were maintained in the presence of 10 μg/ml puromycin, 100 μg/ml hygromcin B, and 8 μg/ml blasticidin.

To generate recombinant viruses expressing CA sequences from HIV-2 and SIV viruses, we used the HIV-1 pNL4-3-ΔENV-LucR vector as background, which contains Renilla luciferase reporter gene in place of the HIV-1 nef locus and carries a large deletion in env 43 . To facilitate the construction of CA chimeric proviruses, SacII and XmaI restriction sites were introduced at positions 49 and 670 of the CA domain of pNL4-3-ΔENV-LucR, respectively, using the QuickChange site-directed mutageneis kit (Stratagene), thereby creating pNL4-3-ΔENV-LucR-SX (Figure 1). The HIV-2 and SIV capsid sequences tested in this study were derived from different origins.

The HIV-2 viral load measurements were performed at Hôpital Saint-Louis using the real-time quantitative PCR method for measuring the HIV-2 RNA load as previously described 34 , with the exception that a new set of primers overlapping the 5'-long terminal repeat (5'-LTR) region allowing to distinguish between HIV-2 subtypes A and B. Six of these patients were clinically considered as aviremic with viral loads below 200 RNA copies/ml.

The pNL4-3-ΔENV-LucR-SX vector or the corresponding chimeric proviruses (1 μg) and a VSV-G expression plasmid (phCMV-G, 0.1 μg) were cotransfected into 5×10⁴ 293T cells seeded into 6 well plates using jetPEI reagent (PolyPlus Transfection). Virus-containing culture supernatants were harvested 40 h after transfection, clarified by centrifugation to remove cell debris, assayed for their reverse transcriptase (RT) content using the HS-Lenti RT activity kit (Cavidi AB), and used directly for western blot analysis and infectivity assays.

The VSV-G-pseudotyped HIV-1 NL4-3-derived particles were harvested 40 h post-transfection, purified and concentrated by centrifugation through a 20% sucrose cushion at 110,000 × g for 90 min using a TLA 100.4 rotor and an Optima TLX Ultracentrifuge (Beckman Coulter) and then resuspended in 50 μl Laemmli sample buffer (Bio-Rad). Particle-associated proteins corresponding to 0.3 ng of RT were separated by gel electrophoresis into a 12% SDS-PAGE gel, transferred to a PVDF membrane (Whatman), and blocked in Odyssey western blot blocker for 3 h at room temperature. To analyze the Gag processing pattern an to detect HIV-2 CA proteins, the membranes were probed with a mixture of sera derived from HIV-2 subtype A- and B-infected patients at a 1:2,000 dilution for 2 h at room temperature, washed, incubated with a HRP-conjugated goat anti-human IgG - H&L (ab6858, Abcam) at a 1:20,000 dilution for 1 h at room temperature, and detected using a ECL Advance western blotting detection Kit (RPN2135, GE healthcare).

Each viral infectivity was measured by determining Renilla luciferase activity in target cells 40 h after infection, as previously described 12 . Briefly, the U373-X4-hTRIM5γ and/or the U373-X4-LacZ target cells (1 × 10⁴ cells) were initially plated in black-walled, clear bottom 96-well plates, cultured in the presence of 100 U/ml IFNα for 20 h to increase the expression of TRIM5α, and infected with 2 ng RT/ml of virus supernatant. After 40 hours, the supernatant of adherent cells was completely removed and 50 μl of 1X lysis buffer (Renilla Luciferase kit, Promega) were added to each well. Plates were maintained at room temperature for 30 min, after which wells were sequentially injected with 100 μl of luciferase substrate (Promega), and three seconds later, light emission (relative light units, RLU) was measured over a two sec interval using a luminometer (Varioskan Flash, Thermo scientific). Each sample was evaluated in triplicate and three independent experiments were performed. Results were expressed as mean for three experiments.

All results are expressed as mean ± standard error of the mean (SEM). Comparisons between two groups were performed using the Mann–Whitney test; comparisons between multiple groups within a single series of experiments were performed using the Kruskal-Wallis test; post-test comparisons, performed only if p < 0.05, were performed using Dunn’s multiple comparison test. In all cases, GraphPad Prism 5 software was used.