The antiviral factor APOBEC3G improves CTL recognition of cultured HIV-infected T cells

We first analyzed how T cells infected with WT or Δvif HIV stimulate HS-CTLs. To this end, primary CD4⁺ T cells harboring endogenous A3G were infected with HIV particles produced in the absence of A3G (Fig. 1 A). From the literature, we anticipated that after CD4⁺ T infection, A3G will be incorporated in the progeny virions, thus exerting its editing activity starting from the second cycle of replication. Viral propagation was monitored by flow cytometry (intracellular Gag staining; Fig. 1 B). As expected, replication of HIV_Δvif was markedly reduced compared with HIV, with 3 and 20% Gag⁺ cells at day 5 after infection, respectively. We compared the ability of HIV and HIV_Δvif-infected CD4⁺ T cells to activate EM40-F21, an HS CD8⁺ CTL clone (Fig. 1 C). EM40-F21 was derived from an HIV-infected patient and recognizes a well-characterized immunodominant epitope of Gag p17 (SL9) presented by HLA-A*0201 (Moris et al., 2004). Surprisingly, HIV_Δvif-infected CD4⁺ T cells activated EM40-F21 to a greater extent than cells infected with the WT virus (Fig. 1 C). Similar results were obtained with cells from three different donors infected with various HIV inocula (Fig. 1 D). When expressed as the number of IFN-γ–producing CTL/Gag⁺ CD4 cells, CD4⁺ T cells infected with HIV_Δvif were found to be two to three times more efficient at activating HS-CTLs than HIV-infected cells (Wilcoxon rank-sum test: P = 0.032; Fig. 1 D). These data strongly suggest that there is dissociation between the capacity of HIV_Δvif to infect CD4⁺ T cells and to activate HS-CTLs.

To escape CTL recognition, the HIV Nef protein down-regulates the surface expression of HLA class I molecules (Schwartz et al., 1996). We asked whether Nef might affect the immunogenicity of HIV_Δvif-infected T cells. To this end, we constructed a ΔnefΔvif double mutant (HIV_Δnefvif). HIV_Δnef isolates replicate less efficiently than WT virus. To obtain sufficient numbers of infected cells, CD4⁺ T lymphocytes were exposed to VSV-G–pseudotyped HIV_Δnef and HIV_Δnefvif (Fig. 1 E). HIV_Δnefvif replication was reduced (three- to sixfold) compared with HIV_Δnef at 48 and 72 h after infection. Remarkably, at each time point, even if very few Gag⁺ cells were detected, HIV_Δnefvif induced a robust activation of HS-CTLs (Fig. 1 F). As expected, Nef-deficient isolates were more efficient than WT virus in activating CTLs (Fig. 1, compare C and F). Similar results were obtained with cells from eight different donors (Fig. 1 G), demonstrating that cells infected with HIV_Δnefvif were two to three times more efficient than HIV_Δnef-infected cells in activating HS-CTLs (Wilcoxon rank-sum test: P = 0.0005; Fig. 1 G). Therefore, the enhancing effect of Δvif on CTL activation is observed both in the context of WT and Δnef HIV.

CTLs secrete a panel of chemokine and cytokine upon activation (Price et al., 1998; Wagner et al., 1998; Appay et al., 2008). We performed a peptide titration experiment to identify cytokines other than IFN-γ that are secreted in an antigen-specific manner by EM40-F21. Primary CD4⁺ T cells were loaded with the SL9 peptide, co-cultured with HS-CTLs, and lymphokine production was analyzed in culture supernatants using the luminex technology. As previously reported (Price et al., 1998; Wagner et al., 1998), we observed a dose-dependent secretion of MIP-1α and MIP-1β (Fig. 2, F and I). Thereafter, we examined the effect of Δvif viruses on MIP-1β and MIP-1α secretions by EM40-F21. Primary T cells were infected with VSV-G-HIV_Δnef and -HIV_Δnefvif, monitored for Gag expression (12 and 1% of Gag⁺ cells, respectively), and used to activate EM40-F21 (Fig. 2, A, B, D, and G). Using IFN-γ Elispot, we first confirmed the enhancing effect of Δvif on IFN-γ secretion by EM40-F21 (Fig. 2, B and C). We then measured in co-culture supernatants the release of MIP-1α and MIP-1β. Upon co-culture with HIV_Δnef- and HIV_Δnefvif-infected targets (at a ratio of 1 CTL to 50 CD4⁺ target cells), EM40-F21 secreted similar levels of MIP-1α (130 and 106 pg/ml, respectively; Fig. 2 D). HIV_Δnef- and HIV_Δnefvif-infected cells also allowed the secretion of MIP-1β (292 and 167 pg/ml, respectively; Fig. 2 G). These levels were at least three times higher than background secretion induced by uninfected control cells (Fig. 2, D and G). HIV infection did not increase the background lymphokine production by target CD4⁺ T cells alone (unpublished data). We noticed that the secretion of MIP-1β with Δnefvif viruses was reduced as compared with Δnef (Fig. 2 G), suggesting that the effect of Δvif viruses might vary depending on the lymphokine. Nonetheless, even if very few Gag⁺ cells were detected, HIV_Δnefvif induced a potent secretion of both MIP-1α and MIP-1β by HS-CTLs (Fig. 2, E and H). To gain information on the magnitude of A3G-mediated effect, we used MIP-1α and MIP-1β SL9 peptide titration curves (Fig. 2, F and I) as standards to calculate the relative antigen presentation levels on the surface of infected (Gag⁺) target cells. Based on MIP-1α secretion, HIV_Δnef- and HIV_Δnefvif-infected cells presented 6.6 × 10⁻¹⁵ and 4 × 10⁻¹⁴ exogenous peptide equivalent, respectively. Based on MIP-1β secretion, the relative antigen presentation levels were 1.98 × 10⁻¹⁵ and 7.9 × 10⁻¹⁵ exogenous peptide equivalent for HIV_Δnef- and HIV_Δnefvif Gag⁺ cells, respectively. Thereafter, cells infected with HIV_Δnefvif viruses likely present four to six times more peptide antigen than cells infected with HIV_Δnef. This calculation was done using populations harboring different percentages of Gag⁺ cells. It would have also been informative to confirm these findings by calculating exogenous peptide equivalents in the presence of equal number of infected (Gag⁺) target cells. The low replicative capacity of HIV_Δnefvif in primary CD4⁺ T cells precluded this possibility. As expected, titrating down the amount of infected cells in the co-culture (1 CTL for 10 CD4⁺ targets) reduced the overall secretion levels and confirmed that in the absence of Vif, infected cells enhance cytokine production by HS-CTLs (Fig. S1). It is noteworthy that Nevirapine, a reverse transcription inhibitor of viral replication, blocked antigen presentation, indicating that activation of EM40-F21 was not a result of the presentation of incoming HIV antigens (Fig. 2, A and B). Collectively, our results demonstrate that Vif-deficient HIV-1 replicates less efficiently but is a potent activator of HS-CTLs, inducing the secretion of various lymphokines such as IFN-γ, MIP-1α, and MIP-1β. More specifically, Vif impacts the immunogenicity of HIV-infected T cells. In infected cells, Vif targets A3G for proteasomal degradation (Mariani et al., 2003; Yu et al., 2003); therefore, A3G may render HIV-infected cells more able to activate CTLs.

We further analyzed the contribution of A3G restriction in HS-CTL activation. We examined the effect of A3G, incorporated into incoming virions, on viral replication. Viral particles were produced in the presence or absence of A3G (Fig. 3 A; and Fig. S2, B and C; Gaddis et al., 2003). A3G expression was confirmed by Western blotting (Fig. S2 C). As expected, the amount of A3G was markedly reduced with Vif⁺ viruses (Fig. S2 C). In an assay of a single cycle of replication, A3G moderately reduced the infectivity of WT or HIV_Δnef, whereas HIV_Δvif isolates were highly sensitive to A3G inhibition (Fig. S2 B).

CEM-A2, a CD4⁺ T cell line which does not express endogenous A3G (Gaddis et al., 2003), was infected with HIV_Δnef, produced in the presence (HIV_Δnef + A3G) or absence of A3G (Fig. 3 A). The use of HIV_Δnef enables better CTL activation (Fig. 1). We did not use a Δvif HIV in this setting because this mutant was not viable when produced with A3G (Fig. S2 B). Importantly, in this experimental system, A3G exerts its editing activity exclusively on the first cycle of infection (Fig. 3 A). An infection peak was reached 2 d after HIV_Δnef- and HIV_Δnef + A3G infection with 78 and 26% Gag⁺ cells, respectively, thus confirming that A3G generates partly defective viruses (Fig. 3 B). Remarkably, HIV_Δnef + A3G–infected cells induced a robust activation of the CTL clone EM40-F21 (Fig. 3 C). Similar data were obtained with two different HIV isolates (NL4-3 and SF2) used at various inocula (Fig. 3 D). Notably, when expressed as the number of IFN-γ–producing CTL/Gag⁺ CD4 T cells, HIV_Δnef + A3G–infected cells were two to three times more efficient in activating HS-CTLs than HIV_Δnef-infected cells (Wilcoxon rank-sum test: P = 0.0005; Fig. 3 D). To further highlight the role of A3G in enhancing CTL activation, we compared the capacity of CEM-A2 with similar levels of Gag⁺ cells to activate HS-CTLs. Data from six independent experiments are presented in Fig. 3 (E and F). In the presence of A3G, there was a stronger activation of EM40-F21 (Wilcoxon rank-sum test: P = 0.015; Fig. 3 F).

The CTL clone EM40-F21 is specific for a Gag-derived epitope presented by HLA-A2. We examined whether A3G also activates CTLs that recognize other viral epitopes. To this end, we generated a Nef-specific CTL line, restricted by HLA-B*07. As a negative control, we used CTLs raised against a CMV epitope. An HLA-B*07⁺ cell line, T1-B7, was used as target (Cardinaud et al., 2004). T1-B7 cells were infected with HIV or HIV + A3G. At 24 h after infection, cells were monitored for viral replication and used as targets in Elispot assays. As expected, replication of HIV + A3G was reduced compared with HIV alone (unpublished data). CMV-specific control T cells did not react with HIV-infected cells (Fig. 3 H). T1-B7 cells infected with HIV + A3G induced a strong activation of Nef-specific CTLs (Fig. 3 G). Activation with HIV + A3G–infected cells was threefold higher than with HIV-infected cells (Fig. 3 G). When normalized to the amount of infected (Gag⁺) cells, HIV + A3G–infected cells activated Nef-specific CTLs ninefold better than HIV-infected cells (unpublished data). Overall, using two HIV isolates (NL4-3 and SF2) and CTLs specific for two different HIV proteins (Gag and Nef) presented by distinct HLA molecules (HLA-A2 and -B7), we demonstrated that A3G enhances recognition of HIV-infected cells by HS-CTLs.

We then questioned whether the editing activity of A3G is necessary to promote HS-CTL stimulation. A3G possesses two consensus deaminase motifs, which is a hallmark of APOBEC protein family members. Mutations in the C-terminal deaminase motif (such as H257R and C288S) abrogate editing activity (Haché et al., 2005; Newman et al., 2005).

HIV_Δnef particles were produced in the presence of WT or mutant (H257R and C288S) A3G. The amounts of WT and mutant A3G proteins were comparable (Fig. S2 C). CEM-A2⁺ cells were infected with two different doses of virus (Fig. 4 A). At 48 h after infection, the percentage of infected cells was reduced using HIV_Δnef + A3G compared with HIV_Δnef, whereas HIV_Δnef produced in the presence of A3G mutants H257R and C288S replicated to levels similar to those of HIV_Δnef. Infected cells were then used to stimulate HS-CTL EM40-F21 (Fig. 4 B). As already observed, cells infected with HIV_Δnef + A3G activated HS-CTLs to a great extent, which was comparable to activation levels obtained with HIV_Δnef-infected cells. Cells infected with HIV_Δnef containing A3G mutants also induced strong IFN-γ secretion (Fig. 4 B). When expressed as IFN-γ⁺ CTLs/Gag⁺ CD4 cells, HIV_Δnef + A3G was twofold more efficient in activating HS-CTL than HIV_Δnef. Mutating the deaminase motif of A3G abrogated the capacity of A3G to enhance CTL activation (Fig. 4 C). Similar results were also obtained using another A3G mutant lacking both N- and C-terminal cytosine deaminase motifs (H65R/H257R; unpublished data). Thus, the editing activity of A3G is required to mediate its effect on HS-CTL-recognition.

In HIV samples derived from patients, editing patterns are dominated by GG-to-AG hypermutations (Vartanian et al., 1991). On the protein level, this preferential editing leads to a high proportion of tryptophan (TGG) to STOP codon (e.g., TAG) substitutions.

We asked whether the expression of aberrant HIV proteins with premature STOP codons might be involved in the enhanced immunogenicity mediated by A3G. Downstream of the Gagp17 epitope SL9 recognized by EM40-F21, we identified two tryptophan codons (W₂₁₂ and W₂₄₉) that are potential targets for A3G editing. HIV proviruses with STOP codons at these positions have been previously isolated from PBMCs of infected individuals (Kijak et al., 2008). In addition, A3G edits these tryptophans into STOP codons in vitro (Armitage et al., 2008). We thus mimicked A3G editing by introducing STOP codons at these two positions, yielding HIV_ΔnefP2-Stop and HIV_ΔnefP3-Stop, respectively. Upon transfection, HIV_ΔnefP2 and HIV_ΔnefP3 led to the expression of truncated Gag proteins of 24 and 34 kD, respectively (Fig. 5 B). Viral particles produced were not infectious (not depicted) and lacked Gag p24 expression (Fig. 5 C). Transfected HeLa-HLA-A2⁺ cells expressing HIV_ΔnefP2-Stop and HIV_ΔnefP3-Stop were then used as targets to activate EM40-F21 (Fig. 5 D). Activation levels obtained with truncated P2 and P3 Gags were similar to or higher than those induced by cells producing infectious HIV (Fig. 5 D). CTL activation by infectious HIV, or by HIV-P2-Stop or -P3-Stop, was blocked by the proteasome inhibitor epoxomicin (Fig. 5 E), strongly suggesting that processing of full-length and truncated Gag-derived epitopes occurs through the classical MHC-I presentation pathway. Therefore, defective HIV strains bearing premature STOP codons are efficiently recognized by HS-CTLs.

AID/APOBEC family members are cytidine deaminases that exert different functions in the host defense against pathogens. In B cells, AID is necessary for antibody somatic hypermutation and class-switch recombination. AID provides the molecular flexibility that is crucial to establishing efficient humoral immune responses. APOBEC3 proteins offer a strong innate antiviral defense. Recent studies suggested that the adaptive and innate immune functions of APOBEC3/AID family members might overlap. AID protects the host against the oncogenic viruses Abelson murine leukemia virus, Epstein-Barr virus, and hepatitis B virus (Rosenberg and Papavasiliou, 2007). AID seems to act on the host genome through a mechanism that is not yet fully understood (Gourzi et al., 2007). Conversely, mouse APOBEC3 (mA3) is required to establish an efficient neutralizing antibody response against Friend virus (Santiago et al., 2008). Again the mechanism remains to be determined, however it has been proposed that mA3 might be involved in shaping the antibody repertoire. In the present study, we reveal a new role for A3G. In addition to restricting viral replication, A3G enhances the capacity of infected T cells to activate virus-specific CTLs. This phenomenon requires the editing activity of A3G.

Misfolded or truncated proteins represent a major source of peptides for the loading of MHC-I molecules (Yewdell and Nicchitta, 2006). This was demonstrated using several model antigens, such as influenza NP, OVA, or HIV-1 proteins. For instance, targeting HIV-1 Gag/Nef and Env to rapid proteasomal degradation by the addition of N-terminal degron signals (i.e., Arginine residue and/or ubiquitin fusion sequence) resulted in enhanced MHC-I presentation and increased activation of mouse CTLs (Tobery and Siliciano, 1997; Goldwich et al., 2008). We extend these observations by showing that premature STOP codons in HIV genome, a hallmark of A3G editing, enhance CTL activation. We suggest that truncated/misfolded HIV polypeptides, generated by A3G, supply the pool of MHC-I–restricted HIV antigens. In addition to A3G, other mechanisms (e.g., viral recombination or errors during reverse transcription) generate crippled viruses. These defective viruses might also be an underestimated source of antigens.

HIV has developed multiple mechanisms for escaping CTLs. Nef, by modulating MHC-I expression, decreases the recognition and killing of infected cells by CTLs. The efficacy of Nef-mediated CTL escape is partial and depends on the kinetic of antigen expression and on the quantity of MHC-I–HIV epitope complexes (Fig. 1; Tomiyama et al., 2005). We show in this paper that A3G enhances the generation of HIV-derived epitopes, thus partially counteracting the action of Nef. Moreover, in primary CD4⁺ T cells, Vif reduced the capacity of infected cells to activate CTLs by degrading A3G. Recognition of HIV-infected cells by CTLs results from a balance between the efficacy of the generation of antigens enhanced by A3G editing and the immune-escape mechanisms mediated by Nef and Vif.

A3G has several genetic variants that might influence the progression to AIDS (An et al., 2004). It will be worth examining the contribution of A3G polymorphism in the induction of CTL responses. Simian immunodeficiency virus (SIV) might also be used to further characterize the role of A3G in CTL activation because Vif-deficient SIV replicates poorly but elicits an antibody response in macaques (Desrosiers et al., 1998).

In conclusion, cellular and viral factors determine the efficiency of antigen presentation that, in turn, influences the quality of T cell responses and thus the control of HIV infection (Appay et al., 2008). Our study demonstrates that A3G-mediated viral restriction contributes to the immunogenicity of HIV-infected cells and to CTL activation, thus linking innate and adaptive immunity. It is tempting to speculate that the function of A3G in enhancing CTL recognition occurs with other APOBEC3/AID family members and with other viruses sensitive to cytidine deamination.

PBMCs were isolated by ficoll density gradient from the blood of healthy donors and screened by FACS for the expression of HLA-A*02 using BB7.2 antibody. CD4⁺ T lymphocytes were then isolated from PBMCs using negative depletion (Miltenyi Biotec). HeLa-CIITA-HLA-A*0201⁺ and HUT-HLA-A*0201⁺ cell lines were generated by electroporation and selection using G418. CEM-HLA-A*0201⁺ cells were provided by K.L. Collins (University of Michigan, Ann Arbor, MI). T1 (174xCEM, CCR5⁺ LTR-GFP⁺) cells stably transfected with the HLA-B7^mα3 construct were used as antigen-presenting cells (Cardinaud et al., 2004).

The HS EM40-F21 CTL clone is specific for Gag p17 (aa 77–85) presented by HLA-A*0201 and has been previously described (Moris et al., 2004). Human protocols were approved by the ethics committee of the Pasteur Institute. CTL lines specific for the HIV Nef (F10LR, Nef ₆₈FPVTPQVPLR₇₇) and CMV (T10AM, pp65 ₄₁₇TPRVTGGGAM₄₂₆) were derived from splenocytes of peptide-immunized HLA-B7^mα3 transgenic mice (Cardinaud et al., 2004). Splenocytes were cultured in RPMI 1640 supplemented with 10% fetal bovine serum and stimulated every 10 d with peptide-pulsed syngeneic LPS lymphoblasts.

The HIV strains used in the study were HIV_NL4-3, HIV_SF2, and their Δnef counterparts (Fackler et al., 2006). HIV and HIV(VSV) virions were produced and titrated as previously described (Moris et al., 2004). HIV virions containing A3G WT and H257R or C288S mutants were produced in 293T cells by cotransfection (1:1 DNA ratio) of HIV provirus with a pcDNA3-V5 plasmid encoding for WT or mutant A3G (A3G and mutant constructs provided by A.J. Hance [Hopital Bichat, Paris, France] and M.H. Malim [King’s College, London, England, UK]). pNL4-3 XCS and pNL4-3_Δvif XCS (provided by A.J. Hance) were used to generate HIV_NLΔnef and HIV_NLΔnefvif, respectively. In brief within nef, a frame shift mutation of four bases was inserted at a unique XhoI site (Schwartz et al., 1996). HIV_NLΔnefP2 and HIV_NLΔnefP3 proviruses were generated by introducing in HIV_NLΔnef a STOP codon at positions 212 and 249 (amino acid sequence), respectively, using the primers for HIV_NLΔnefP2 (5′-GAGGAAGCTGCAGAATAGGATAGATTGCATCCAG-3′) and HIV_NLΔnefP3 (5′-CAGGAACAAATAGGATAGATGACACATAATCCACCTATCC-3′; mutagenized nucleotides in bold) with QuikChange XL Site-Directed Mutagenesis kit (Stratagene). 48 h after transfection, viral supernatants were collected and p24 content was measured by ELISA (PerkinElmer). When stated, the infectivity was tested using HeLa-CD4⁺ cells (P4 cells), which carry the integrated HIV LTR-lacZ reporter cassette. P4 was infected with 5 ng p24 of the indicated viruses. 48 h later, β-galactosidase activity was measured in cell extracts.

5 × 10⁶ CEM-A2⁺, Hela-A2⁺, HUT-A2⁺, and T1-B7 cells were infected with various amount of HIV (ranging from 10 to 100 ng p24/ml) for 3 h at 37°C in 10 mM Hepes, pH 7.4, and 2 µg/ml DEAE-dextran containing culture medium. Primary CD4⁺ T cells were PHA activated (1 µg/ml for 48 h; Abbot Laboratories), washed, and grown in IL-2–containing culture medium (50 IU/ml; Proleukin [Novartis]) for at least 10 d. Cells were then infected for 3 h at 37°C with 10–100 ng/ml HIV in the presence of DEAE-dextran and grown with IL-2. Viral replication was then followed using either p24 intracellular staining. Infected cells were then used as antigen presenting cells in Elispot assays.

Cells were fixed, permeabilized, and stained using standard procedures with anti-Gagp24 (KC57; Beckman Coulter) or an IgG1 isotype mAb as a negative control. Samples were analyzed using FACSCalibur (BD).

Stimulator cells were co-cultured for at least 8 h with CD8⁺ T cell clones (ranging from 2,500 to 10⁴ EM40-F21/well and 2,000 to 10⁴ anti-Nef CTLs/well). IFN-γ production was then measured in an Elispot assay as previously described (Moris et al., 2004). As positive controls, stimulators were incubated with 1 µg/ml of cognate (SL9) peptides before addition of CTLs.

Infected or SL9 peptide–loaded stimulator cells were co-cultured in a 96-well plate with EM40-F21 at a 1/50 and 1/10 ratio (5,000 CTLs for 250,000 or 50,000 CD4⁺ T cells, respectively). Cell culture supernatants were collected after 24 h, and MIP-1α and MIP-1β secretion was measured according the manufacturer’s instructions (Chemokine Human 5-plex Panel LHC0005; Invitrogen).

24 or 48 h after transfection, cells were lyzed in PBS-1% Triton X-100 (Sigma-Aldrich) supplemented with protease inhibitors (Roche). Viral supernatants were ultracentrifuged (10⁵ rpm for 20 min) and resuspended in PBS containing 1% triton and protease inhibitors. Cell lysates and viral particles were analyzed by SDS-gel electrophoresis using 4–12% NuPAGE gels (Invitrogen). Anti-HIVp17, anti-V5 (both provided by M.H. Malim), and anti-A3G (National Institutes of Health reagent 9968) monoclonal antibodies were used.

Data were analyzed by a Wilcoxon rank-sum test with StatView statistics program (Abacus Concepts).

Fig. S1 shows that Vif-deficient HIV-1 stimulates chemokine secretion of HS-CTLs at a 10:1 ratio. Fig. S2 shows the infectivity and phenotype of HIV particles produced in the presence or absence of A3G.

We thank Hans-Georg Rammensee, Stefan Stevanovic, and Andreas Weinzierl for discussions, Matthew Albert and Georges Azar for critical reading of the manuscript, and Yoram Reiter, Maya Haus, Kathleen Collins, Alan Hance, Mike Malim, and the National Institutes of Health AIDS research and reference reagent program for kind gifts of reagents. We also thank Helen K.W. Law from the Centre d’Immunologie Humaine for assistance.

This work was supported by grants from the Agence Nationale de Recherche sur le SIDA (ANRS), SIDACTION, the Centre National de la Recherche Scientifique, the European Community, Institut Pasteur, and Janssen-Cilag. N. Casartelli is a fellow of SIDACTION funding. S. Brandler is a Fellow of the ANRS vaccine network (AHVN).

The authors have no conflicting financial interests.