|Home | About | Journals | Submit | Contact Us | Français|
Human APOBEC3F (hA3F) and APOBEC3G (hA3G) are antiretroviral cytidine deaminases that can be encapsidated during virus assembly to catalyze C→U deamination of the viral reverse transcripts in the next round of infection. Lentiviruses such as human immunodeficiency virus (HIV) and simian immunodeficiency virus (SIV) have evolved the accessory protein Vif to induce their degradation before packaging. HIV type 1 (HIV-1) Vif counteracts hA3G but not rhesus macaque APOBEC3G (rhA3G) or African green monkey (AGM) APOBEC3G (agmA3G) because of a failure to bind the nonhuman primate proteins. The species specificity of the interaction is controlled by amino acid 128, which is aspartate in hA3G and lysine in rhA3G. With the objective of overcoming this species restriction, mutations were introduced into HIV-1 Vif at amino acid positions that differed in charge between HIV-1 Vif and SIV Vif. The mutant proteins were tested for the ability to counteract hA3G, rhA3G, and agmA3G. Alteration of the conserved sequence at positions 14 to 17 from DRMR to SERQ, which is the sequence in AGM Vif, caused HIV-1 Vif to functionally interact with rhA3G and agmA3G. Mutation of three residues to the sequence SEMQ allowed interaction with rhA3G. SEMQ Vif also counteracted D128K mutant hA3G and wild-type hA3G. Introduction of the sequence into an infectious molecular HIV-1 clone allowed the virus to replicate productively in human cells that expressed rhA3G or hA3G. These findings provide insight into the interaction of Vif with A3G and are a step toward the development of a novel primate model for AIDS.
Several nonhuman primate models for AIDS have been used to provide insight into human immunodeficiency virus (HIV) pathogenesis in humans. In particular, simian immunodeficiency virus of macaques (SIVmac) and chimeric SIV-HIV induce AIDS-like symptoms in rhesus macaques through pathogenic mechanisms that mimic those of HIV type 1 (HIV-1) (7). HIV-1, however, does not infect rhesus macaques or nonhuman primates such as the African green monkey (AGM) (1, 10-12, 19, 23). This species restriction is caused by at least two blocks to HIV-1 replication in nonhuman primate T cells and macrophages: a defect in the particle infectivity of virions that are released which is mediated by host APOBEC3 proteins and a postentry block that is mediated by Trim5α (25).
Lentiviruses such as HIV-1 and SIV encode Vif, an accessory protein that counteracts the antiviral activity of the APOBEC3 family of cytidine deaminases (9). Human APOBEC3G (hA3G) was identified initially (22), and subsequently additional family members, including human APOBEC3B, human APOBEC3C, and human APOBEC3F (hA3F), were found to be active (14, 27, 34). In cells infected with HIV-1 with vif deleted (Δvif), hA3G and hA3F are packaged into assembling virions. In the next round of infection, the packaged enzymes deaminate the newly synthesized minus-strand cDNA, resulting in plus-strand G→A mutations (8, 15, 17, 30, 33). In cells infected with wild-type HIV-1, Vif binds to hA3G and hA3F, forming a complex with a Cul5-based E3 ligase. The A3G proteins are then ubiquitinated and degraded by proteasomes (18, 31). Human APOBEC3B and human APOBEC3C are weakly active against HIV-1 but are potent inhibitors of SIV (2, 29). Other elements targeted by APOBEC3 family members include hepatitis B, endogenous murine long terminal repeat and non-long terminal repeat retroelements, and yeast retrotransposons (4-6, 26).
The interaction of Vif with hA3G is species specific (17). HIV-1 Vif binds to hA3G but not to rhesus macaque APOBEC3G (rhA3G) or AGM APOBEC3G (agmA3G). Conversely, SIV of AGMs (SIVagm) Vif binds to agmA3G but not to hA3G. SIVmac Vif is more permissive, binding to rhA3G, agmA3G, and hA3G. The determinant on hA3G that controls the species specificity of its interaction with Vif was mapped on a panel of hA3G-agmA3G chimeras. The specificity mapped to amino acid 128, which is Lys in hA3G but Asp in agmA3G (3, 16, 21, 28). Exchange of the Lys and Asp residues reverses the species specificity for binding to Vif. SIVagm Vif binds to D128K mutant hA3G but not to K128D mutant agmA3G. Conversely, HIV-1 Vif binds to K128D mutant agmA3G but not to D128K mutant hA3G. Modeling on the Escherichia coli cytidine deaminase predicts that amino acid 128 lies on an exposed loop that is positioned away from the enzyme's active site. The exchange does not affect the catalytic activity of the enzyme (21).
The objective of this study was to generate a genetically altered HIV-1 that could functionally interact with nonhuman primate APOBEC3 to counteract its antiviral activity. The approach taken was to identify amino acids in HIV-1 Vif that contribute to it species specificity and to alter these to resemble the sequence in SIV Vif. We found that a four-amino-acid sequence near the amino terminus of Vif which differs in charge between HIV-1 and SIV is important in determining the species specificity of the interaction with A3G. Alteration of the sequence resulted in a Vif that was able to counteract rhA3G and hA3G and allowed productive replication in human cells that stably expressed the deaminases. The mutation also allowed interaction with D128K mutant hA3G. These findings provide an important step toward the development of a primate model for HIV-1 replication.
A codon-optimized HIV-1 vif gene was assembled from synthetic oligonucleotides and amplified with a forward primer containing an EcoRI site and a reverse primer encoding a V5 epitope tag and an XhoI site (see Fig. S1 in the supplemental material). The PCR product was cloned into pcDNA3.1 at the EcoRI and XhoI sites to generate pcVif-Co. Mutations in Vif were generated in pcVif-Co by overlapping PCR and confirmed by nucleotide sequencing. Expression vector pcVifmac was constructed by amplification of the Vif open reading frame (ORF) from SIVmac239 with primers containing EcoRI and XhoI sites and encoding a carboxy-terminal V5 epitope tag. The PCR product was cleaved and cloned into pcDNA3.1 at the EcoRI and XhoI sites.
Single-cycle luciferase reporter viruses pNL-Luc-E−R−, pNL-Luc-E−R−Δvif, and pSIVmac-Luc-E−R− and replication-competent viruses NL-R− and NL-R−V− have been previously described (17). The SEMQ mutation was introduced into pNL-Luc-E−R− and NL43-R− by overlapping PCR in the fragment spanning the region between sites AgeI and PflmI (nucleotides 3485 to 5303) (data not shown). pNL4-3(Vifmac) is a Δvif NL4-3 Vpr− plasmid into which the SIVmac239 Vif ORF has been inserted in the nef gene. The plasmid was constructed by amplifying the SIVmac239 Vif ORF with primers containing NotI and XhoI restriction sites for cloning into nef at the same position as in NL-Luc.
Vif function was analyzed as previously described (17, 21). Briefly, vesicular stomatitis virus G (VSV-G)-pseudotyped luciferase reporter viruses were produced in 293T cells by cotransfection of 2 μg of pNL-Luc-E−R−Δvif, 1 μg of pcVif-Co.V5, 0.5 μg of an hA3G or rhA3G expression vector, and 0.5 μg of pcVSV-G. Virus was harvested 48 h posttransfection, and 1 ng of p24 was used to infect HOS.T4 cells in triplicate. Three days later, intracellular luciferase activity was measured with Luc-Lite Plus reagent (Packard) in a Topcount luminometer (Perkin-Elmer). The data are presented as mean counts per second ± the standard deviation. Vif expression in transfected cells was verified by immunoblot assay and probing with anti-V5 monoclonal antibody (MAb; Invitrogen). For titration experiments, 2 μg of pNL-Luc-E−R−, pNL-Luc-E−R−Δvif, pNL-Luc-E−R− Vif SEMQ, or pSIVmac-Luc-E−R− was cotransfected with indicated amounts of rhA3F, hA3G, rhA3G, or D128K mutant hA3G and pcVSV-G. Viruses were normalized for p24 or p27, and 1 ng was used to measure their infectivity as described above.
Virions were produced by cotransfection of 293T cells with 2 μg of pNL-Luc-E−R−, pNL-Luc-E−R−Δvif, pNL-Luc-E−R− Vif SEMQ with or without 1 μg of pcSIVmac Vif.V5, and 1.0 μg of hA3G.HA (where HA is hemagglutinin), rhA3G.HA, or D128K mutant hA3G.HA. The cells and supernatant were harvested 2 days posttransfection, and virus was concentrated by ultracentrifugation through 20% sucrose at 40,000 rpm for 1 h. A3G was detected in 10 ng of p24 and 20 μg of cell lysates by an immunoblot assay and probing with an anti-HA MAb. The blots were stripped and reprobed with an anti-Vif MAb, an anti-V5 MAb, or AIDS patient serum. The blots were also probed with an antitubulin MAb to control for equivalent loading.
The kinetics of viral replication was determined with HOS.T4.X4 cells that stably expressed hA3G.HA or rhA3G.HA. HOS.T4.X4 cells were infected with a VSV-G-pseudotyped pBABE-neo retroviral vector, and cells were selected with 0.5 mg/ml G418. Resistant clones were isolated and expanded, and rhA3G expression was confirmed by Western blot analysis. Cell clones were infected with 10 ng of p24 from wild-type, Δvif, or Vif SEMQ replication-competent virus, and virus growth was measured over 15 days by p24 enzyme-linked immunosorbent assay.
A functional interaction of HIV-1 Vif with hA3G requires amino acid 128 of hA3G to be negative or neutral. Conversely, interaction of SIVagm Vif with agmA3G requires a positive charge at amino acid 128. We hypothesized that these charge requirements may result from the interaction of complementary charged amino acids in Vif and A3G. D128 in hA3G could interact with a positive charge in HIV-1 Vif, and K128 in agmA3G could interact with a negative charge in SIV Vif. To search for potential complementary amino acids in Vif that might interact with A3G amino acid 128, we searched known viral sequences to identify amino acid residues of Vif that were positively charged in HIV-1 and negative in SIV. In the SIVagm Vif sequences of the Los Alamos National Laboratory sequence database, there are seven conserved, negatively charged positions (E4, E17, E102, E104, D109, D/E122, and E215). Only two of the corresponding amino acids in HIV-1 Vif are positively charged (R4 and R15, corresponding to SIVagm Vif E4 and E17, respectively). HIV-1 Vif R15 is flanked by a nearby positive charge at R17, and these amino acids are conserved in HIV-1 Vif (R15 in >99% of isolates, and position 17 is either R or K in all sequenced viruses [82% R and 18% K]). We therefore focused on amino acids 14 to 17, which in HIV-1 Vif are DRMR. In SIVagm, positions 15 and 17 of Vif are neutral or negatively charged (Fig. (Fig.1A1A).
To test the role of amino acids 4 and 14 to 17 of HIV-1 Vif in determining the species specificity of the interaction with A3G, mutant Vif expression vectors were generated and their function was tested in a single-cycle infection assay. Single-cycle viruses were generated by cotransfection of 293T cells with a Δvif HIV-1 luciferase reporter virus plasmid and a mutant or wild-type pcVif-HA expression vector. The pcVif plasmids expressed a Vif ORF that was codon optimized over its entire length to increase expression in the triple transfection (data not shown). A VSV-G expression vector was included in the transfection to generate infectious pseudotyped virus. The results of this analysis showed that a swap that included the R4 sequence (MEEEKR) maintained function against hA3G but did not counteract rhA3G or agmA3G. In contrast, the swap of the four-amino-acid sequence DRMR at positions 14 to 17 in the SERQ mutant (changes are in bold) was active against hA3G and rhA3G and partially active against agmA3G (Fig. (Fig.1B).1B). Further analysis within positions 14 to 17 showed that changing only three of the four amino acids in the SEMQ mutant allowed interaction of HIV-1 Vif with rhA3G but not with agmA3G. Mutation of R14E and R17E to DEME was not sufficient to alter the interaction with A3G. Insertion of an E at position 17 in DRMER was also insufficient. Immunoblot analysis of the cell lysates showed that the mutant proteins were expressed comparably to wild-type HIV-1 Vif (Fig. (Fig.1B,1B, bottom). These findings suggest that positions 14 to 17 of HIV-1 Vif can influence the species specificity of the interaction with hA3G.
The role of this region in the interaction with rhA3G was tested with additional HIV-1 Vif mutants with changes at positions 14 to 17. Point mutations at each of the three positions (D14S, R15E, and R17Q) were tested (Fig. (Fig.2A).2A). R15E failed to function on hA3G or rhA3G, while D14S and R15E were active against hA3G but not against rhA3G. Thus, single mutations were not sufficient to change the species specificity of the interaction. SEMR, DEMQ, and SRMQ double mutants were also tested (Fig. (Fig.2B).2B). All double mutants remained at least in part active against hA3G. Only DEMQ was partially active on rhA3G. Immunoblot analysis confirmed the equivalent expression of each of the mutant Vif proteins (Fig. 2A and B, bottom). Taken together, these findings mapped the minimal required change to amino acids 14, 15, and 17. R15 and R17 appear to be the most important.
To further probe the role of positions 14 to 17, we tested the dependence of charge on Vif function. The SDMQ and SHMQ mutants tested the requirement for a negative charge at position 15. The AEMQ mutant tested the importance of amino acid 14; the SEMA and SEMN mutants tested position 17; and the AAMA mutant tested positions 14, 15, and 17. All of the mutants interacted at least partially with hA3G (Fig. (Fig.2C,2C, upper part). SDMQ but not SHMQ interacted with rhA3G, suggesting a requirement for a negative charge at position 15 (Fig. (Fig.2C,2C, lower part). AEMQ, SEMA, and SEMN also interacted with rhA3G, suggesting that the identity of amino acids 14 and 17 is not critical but that they cannot be negative. AAMA was inactive against rhA3G, further demonstrating the importance of a negative charge at position 15. Taken together, these results suggest that the charge at amino acids 14 to 17 influences the interaction with rhA3G. Amino acid 15 must be negative, and amino acid 17 must be neutral.
The requirement for positive charges at positions 14 and 17 in HIV-1 Vif suggested an interaction with a negative charge on hA3G. Because of its role in determining the interaction with Vif, amino acid 128 was considered a likely candidate interaction site on hA3G. To determine whether this might be the case, we tested whether SEMQ HIV-1 Vif would interact with D128K hA3G. As expected, DRMR did not. Interestingly, SEMQ interacted with D128K mutant hA3G (Fig. (Fig.3A,3A, right part). These results are consistent with a direct interaction of hA3G D128 with DRMR in Vif. In this analysis, the Vif mutations were introduced directly into the HIV-1 reporter construct. Because of the overlap with pol, the SEMQ mutation introduced three amino acid changes in IN (see Fig. S1 in the supplemental material). However, these did not interfere with the production of infectious reporter virus. In addition, the analysis showed that the Vif mutations were functional as expressed in cis in the context of the virus, arguing that the previous results were not simply the result of overexpression from the codon-optimized Vif expression vector.
To determine the relative efficiency with which the SEMQ mutation altered the Vif phenotype, the reporter viruses were produced from 293T cells cotransfected with hA3G, rhA3G, and D128K mutant hA3G expression vectors over a range of plasmid ratios. Reporter viruses that expressed wild-type HIV-1 Vif, SEMQ Vif, and SIVmac Vif were all similarly resistant to hA3G and differed from Δvif mutant virus, which was sensitive (Fig. (Fig.3B,3B, left part). A reporter virus that expressed HIV-1 Vif was inhibited by rhA3G (Fig. (Fig.3B,3B, middle part), in contrast to viruses that expressed SEMQ Vif and SIVmac Vif, which were resistant. Viruses that expressed HIV-1 Vif, SEMQ mutant HIV-1 Vif, and SIVmac Vif were all similarly resistant to D128K mutant hA3G (Fig. (Fig.3B,3B, right part). Taken together, the results show that the SEMQ mutation allows HIV-1 Vif to counteract rhA3G and D128K mutant hA3G with an efficiency comparable to that of SIVmac Vif. These results show that the SEMQ gene is functional as expressed by the virus and that the alteration to IN was not deleterious.
To biochemically assess SEMQ mutant Vif function, its effect on A3G virion encapsidation and steady-state intracellular level was measured. 293T cells were transfected with Δvif mutant HIV-1, wild-type HIV-1, or Δvif mutant HIV-1 plus pcSIVmac-vif or SEMQ mutant HIV-1 Vif and an A3G expression vector. The virions were pelleted, and their A3G content was quantitated by immunoblot analysis. hA3G was present in Δvif virions but not in controls that lacked viral DNA or that expressed SIVmac Vif or SEMQ mutant Vif (Fig. (Fig.4A,4A, upper part). The cell lysates showed that SIVmac Vif and SEMQ mutant Vif reduced the steady-state level of hA3G (Fig. (Fig.4A,4A, lower part). HIV-1 Vif and SIVmac Vif were detected with anti-Vif and anti-V5 epitope tag MAbs, respectively. SIVmac Vif appeared to be present at higher levels, probably because it was supplemented in trans. HIV-1 Vif did not exclude rhA3G from virions (Fig. (Fig.4B,4B, upper part). In contrast, SIVmac Vif and SEMQ mutant Vif effectively prevented the encapsidation of rhA3G. SIVmac Vif and SEMQ mutant Vif correspondingly reduced steady-state rhA3G levels in the cell lysates (Fig. (Fig.4B,4B, lower part). Similarly, HIV-1 Vif did not exclude D128K mutant hA3G from virions but SIVmac and SEMQ mutant Vif were effective and reduced the steady-state levels of D128K mutant hA3G in the cell lysates (Fig. (Fig.4B,4B, upper and lower parts). SEMQ mutant Vif was expressed at amounts comparable to wild-type Vif in each panel, suggesting that the effect was not caused by differences in levels of Vif.
To test whether introduction of the SEMQ mutation would allow HIV-1 to productively replicate in cells that express rhA3G, the mutation was introduced into replication-competent NL4-3. The SEMQ mutant NL4-3 virus, along with control Δvif mutant and wild-type NL4-3, was used to infect HOS.T4.X4 cells that stably expressed CD4/CXCR4 and hA3G or rhA3G or that lacked A3G. Clones of HOS cells that expressed moderate (clone 8) or high (clone 5) levels of rhA3G were used (Fig. (Fig.5A).5A). Virus replication kinetics was measured over 2 weeks. The three viruses replicated similarly in cells that lacked A3G (Fig. (Fig.5B).5B). On cells that express hA3G, the wild-type and SEMQ mutant viruses replicated but the Δvif mutant virus replicated poorly (Fig. (Fig.5C).5C). On clone 8 and clone 5 rhA3G cells, the SEMQ mutant virus replicated efficiently, unlike the Δvif mutant and wild-type viruses, which failed to replicate (Fig. (Fig.5D).5D). These results show that the SEMQ mutation allows efficient replication of HIV-1 on cells that expressed relatively high levels of rhA3G.
hA3F has previously been shown to block Δvif mutant HIV-1 infectivity in single-round reporter virus assays (14, 27, 34). To test whether A3F restricts HIV-1 in a species-specific manner similar to that of A3G, we cloned rhA3F (see Fig. S2 in the supplemental material) and tested its activity against wild-type, Δvif mutant, and SEMQ mutant HIV-1 reporter viruses. As expected, hA3F inhibited Δvif mutant but not wild-type HIV-1 (14, 27, 34). In contrast, hA3F inhibited the SEMQ mutant virus. This suggested that the SEMQ mutation caused Vif to lose its ability to counteract hA3F (Fig. (Fig.6A).6A). rhA3F inhibited wild-type HIV-1 and the SEMQ mutant. hA3F and rhA3F inhibited Δvif mutant SIVmac, and this was partially overcome by SIVmac Vif, consistent with a recent report by Zennou and Bieniasz (32) which showed that SIVmac Vif was partially active against rhA3F. Our data suggested that the interaction between Vif and A3F is species specific and that SEMQ mutant Vif fails to interact with rhA3F.
To determine whether rhA3F blocks the replication of HIV-1, we infected HOS cells that stably express CD4/CXCR4 and rhA3F with replication-competent wild-type, Δvif mutant, or SEMQ mutant NL4-3 or NL4-3(Vifmac), an HIV-1 strain that contains an engineered SIVmac vif gene in the nef position. The HOS cells expressed moderate (C1) or high (C4) levels of rhA3F (Fig. (Fig.6B).6B). All viruses replicated well in the parental HOS.CD4.CXCR4 cell line (Fig. (Fig.6B).6B). Surprisingly, intermediate and high levels of rhA3F had no effect on the replication kinetics. Some differences in peak p24 production were evident among the HOS cell clones, but these were caused by differences in coreceptor expression (data not shown). Taken together, these data demonstrate that although rhA3F is active against HIV-1 in the single-cycle reporter virus assay, it does not block virus replication under more physiological conditions.
We report here progress in addressing the APOBEC3 block to HIV-1 replication in nonhuman primate cells. Changing the sequence DRMR at positions 14 to 17 to SEMQ, which resembles the SIV sequence, allowed it to counteract the antiviral activity of rhA3G and agmA3G. SEMQ mutant Vif induced rhA3G degradation and prevented its encapsidation. The SEMQ mutation allowed the production of infectious reporter virus in cells cotransfected with rhA3G, and in replication-competent HIV-1, it allowed productive replication in human cells that stably expressed relatively high levels of hA3G or rhA3G. The mutant Vif protein was inactive against rhA3F in reporter virus assays, but HIV-1 was able to productively replicate in cells that stably expressed rhA3F. The DRMR sequence is highly conserved in HIV Vif, consistent with a role as an interaction site for A3G. D14, R15, and M16 are 99% conserved in database HIV-1 sequences, while R17 is conserved in 82% of isolates and replaced with Lys in 18% of sequences.
The DRMR sequence in HIV-1 Vif is important for interaction with APOBEC3G but is not the only region that is required for a functional interaction. The SEMQ mutant gained the ability to interact with rhA3G but did not lose its ability to interact with hA3G. This suggests that the interaction of Vif with APOBEC3G is stabilized by additional binding sites. With the exception of Vif R15E, which lost its function against hA3G, mutations in this region in general had little effect on the interaction with hA3G. Thus, the DRMR region is critical for the strength of binding with A3G but is not the sole determinant of its interaction with hA3G. Consistent with this finding, HIV-1 Vif binds with low affinity to D128K hA3G in coimmunoprecipitation experiments (21). This interaction was not sufficient to counteract the antiviral function.
A prediction of this work is that position 128 of A3G interacts with Vif at amino acid 15 or 17. This is suggested by the finding that replacement of R15 and R17 in HIV-1 Vif with neutral or negatively charged residues allowed interaction with rhA3G, which is positively charged at position 128. In addition, mutations at R15 and R17 allowed HIV-1 Vif to interact with D128K mutant hA3G, which contains only a single amino acid difference from wild-type hA3G. The complementarity of the changes in Vif and APOBEC3G suggests a direct electrostatic interaction at these positions, although indirect effects on the conformation of the proteins cannot be ruled out.
An earlier report by Kar et al. (13) described findings that appear to be at odds with those presented here. In that work, an HIV-1 vif gene was inserted into the nef position of Δvif mutant SIVmac239. The chimeric virus replicated in rhesus T cells, suggesting that HIV Vif-1 can substitute for Vif in SIVmac239 and implying that HIV-1 Vif functionally interacts with rhA3G. In contrast, we found that low levels of rhA3G blocked HIV-1 replication. The explanation for this difference is not obvious but could be related to the expression of Vif in the nef position instead of its native context or could be related to the use of primary rhesus T cells, which may express relatively low levels of rhA3G.
SEMQ mutant Vif did not interact with rhA3F, probably because the binding site on Vif for A3F is not identical to that for A3G (24). This raised the question of whether rhA3F is an important restriction factor for HIV-1 in nonhuman primate cells. In humans, hA3F is expressed in T cells and is active against HIV-1 (14, 27, 34). rhA3F is likely to be similarly expressed in nonhuman primates. In single-cycle luciferase virus analyses, rhA3F inhibited HIV-1 and the inhibition was not reversed by Vif. SIVmac was also inhibited by rhA3F, but this was somewhat reversed by SIV Vif. In contrast, our analysis of replication-competent virus on cells that stably expressed rhA3F led to somewhat different conclusions. In this analysis, which more closely mimics physiological conditions, rhA3F did not inhibit HIV-1 replication. This suggests that rhA3F is not an impediment to HIV-1 replication in nonhuman primate cells. The SEMQ mutation in Vif may be sufficient to overcome the APOBEC3 barrier to HIV-1 replication in nonhuman primates without the need for additional mutational alteration.
Trim5α remains as a barrier to HIV-1 replication in nonhuman primate cells. As a result, the SEMQ mutant does not replicate in primary rhesus cells (data not shown). Owens et al. (20) showed that various mutations in the HIV-1 capsid, most in the vicinity of the cyclophilin binding loop, partially allowed the virus to escape rhTrim5α. The combination of these mutations could be helpful in the establishment of a novel primate model for AIDS.
We thank Hui Chen and Qin Yu for critical reading of the manuscript.
This work was funded by grants from the National Institutes of Health (AI51686 and DA14494) and the American Foundation for AIDS Research. N.R.L. is an Elizabeth Glaser Fellow of the Pediatric AIDS Foundation.
†Supplemental material for this article may be found at http://jvi.asm.org/.