To identify HLA-I–associated HIV-1 subtype C polymorphisms, we performed HLA-I typing on 701 untreated South African patients with chronic HIV-1 infection and sequenced the gag
genes of their circulating virus (unpublished data) (12
). Associations were corrected for viral phylogeny, as this approach was shown to significantly improve the specificity of HLA-I associations (16
). We limited our initial analysis to HLA-I viral polymorphisms that occurred within or flanked (within five amino acids of the N or C termini of the epitope) known HLA-I–restricted CTL epitopes. Such HIV-1 polymorphisms are more likely to represent CTL escape and less likely to represent confounding secondary compensatory mutations (11
). These analyses resulted in the identification of 20 mutations within Gag (18 within p24) and 11 within Nef (). The majority of the mutations within Gag (17 out of 20; 85%) and Nef (8 out of 11; 73%) were HLA-B restricted. Of those mutations that have been fully evaluated, all (five in Gag and two in Nef) were formally shown to be CTL escape variants (12
HLA class I allele–associated HIV-1 polymorphisms occurring within or flanking known CTL epitopes in chronically clade C–infected South Africans
We then applied these HLA-I–associated polymorphisms, found in chronic infection, to the analyses of early HIV-1 transmission within a cohort of 114 epidemiologically linked Zambian transmission pairs. These couples were initially identified as HIV-1 discordant (i.e., one partner was HIV-1 infected and their spouse was uninfected). Despite counseling and condom provision, transmission from the chronically infected partner to their spouse continues, but at a significantly reduced rate (~8% per year) (21
). Epidemiological linkage of the transmission pairs was established through sequence analysis of the regions encoding gp41 and Gag (22
). Population-based sequencing of gag
genes was performed on DNA RT-PCR amplified from plasma virus RNA obtained ~6 mo after the estimated date of infection. A phylogenetic tree, constructed to determine the relationships of HIV-1 gag
gene sequences between the two cohorts studied in this paper and other clade C sequences derived from the Los Alamos National Laboratory HIV Molecular Immunology Database (), demonstrated that Southern African sequences tend to cluster together, distinct from those prevalent in representative countries in East Africa, South America, and Asia. These data suggest that sequence analyses derived from one country may be closely applicable to viruses found in neighboring countries in this region. This is supported by previous experiments that have reported the identical escape mutations selected in geographically distinct but otherwise similar clade B–infected cohorts (12
Figure 1. Southern African clade C Gag sequences tend to cluster together. Neighbor-joining phylogenetic tree of Gag (p17 and p24) clade C taxa is shown. These taxa (20 from each cohort) were randomly selected from the total pool of sequencing data available from (more ...)
After HLA-I typing of both partners from each of the 114 epidemiologically linked Zambian transmission pairs, plasma VL was determined from the same samples that viral gag and nef gene sequences were derived. We examined the relationship of amino acid polymorphisms in Gag and Nef sequences to early VL in the newly infected partners of this Zambian transmission pair cohort. We reasoned that early VL (after seroconversion) would be more reflective of the characteristics of the virus transmitted from the donor than VL seen later during chronic infection, when reversion and compensatory mutations might occur. Indeed, the majority of HIV-1 polymorphisms (579 out of 610; 94.9%) were identical in the donor and recipient at 6 mo after transmission. Gag mutations were more likely to be unchanged in the recipient when compared with Nef mutations, because the latter had a greater tendency to exist as the clade C consensus amino acid in the recipient (97.1 vs. 91.3%, respectively; P = 0.002).
Next, we analyzed the relationship between the numbers of transmitted HLA-I–linked Gag and Nef mutations within or flanking known CTL epitopes () to early VL in recipient partners. Accumulation of HLA-I–associated polymorphisms in Nef was not linked to either a higher or lower VL in recipients (ρ = 0.01; NS; ). In contrast, a higher number of HLA-I–associated mutations within Gag was significantly associated with lower VL (ρ = −0.27; P = 0.01; ).
Figure 2. HLA-B–associated mutations in donor Gag affect HIV-1 plasma VL in linked recipients. (A and B) In linked recipients in the Zambian cohort (n = 114), the association between the numbers of mutations within or flanking known CTL epitopes (more ...)
Although limiting the HLA-I–associated HIV-1 polymorphisms to those occurring within or flanking known CTL epitopes strengthens the case that these changes represent CTL escape mutations, it potentially introduces selection bias into our dataset. We therefore included all HLA-I–associated polymorphisms (P < 0.001; q < 0.2) whether or not they were coupled with known CTL epitopes (; and Table S1, available at http://www.jem.org/cgi/content/full/jem.20072457/DC1
). This provided an additional 16 mutations in Gag and 12 mutations in Nef. Unlike what was observed for mutations associated with known CTL epitopes, we saw a more even distribution of the HLA-I restriction of these mutations for both Gag (seven HLA-A, four HLA-B, and four HLA-C restricted) and Nef (four restricted by each HLA-I allele). No associations were seen when comparing the numbers of Nef mutations with VL (ρ = 0.01; NS; ), consistent with the previous analysis, and there was no longer a significant association with the total number of Gag mutations and VL (ρ = −0.11; NS; ). In contrast, when we focused the analysis on HLA-B–associated Gag polymorphisms, these remained significantly associated with diminished VL (ρ = −0.24; P = 0.02; ). HLA-A– and HLA-C–associated Gag polymorphisms were not associated with a lower recipient VL (ρ = 0.08; NS; unpublished data). Similarly, HLA-B–associated Nef mutations were not associated with either positive or negative VL changes in the recipients (ρ = 0.02; NS; ).
These results () support previous studies demonstrating that the HIV-1–specific CTL responses with the greatest impact on VL are HLA-B restricted (23
) and are directed against Gag epitopes associated with CTL escape mutations (6
), perhaps reflecting the selection pressure imposed by these effective responses. Thus, a confounding factor in this analysis of the in vivo fitness cost of escape mutations may be the new HIV-1–specific CTL responses, certainly present a 6 mo after infection, which could by themselves reduce VL. This would imply that newly infected individuals lacking the HLA-B molecules that drive CTL escape mutations in Gag () would benefit most from the receipt of virus with detrimental escape mutations in Gag. When only these latter recipients were included in the analysis, the association between the increasing number of Gag mutations and reduced VL was much stronger (ρ = −0.57; P = 0.0003; ).
Figure 3. Recipients lacking HLA-B alleles associated with Gag mutations benefit most from these mutations. (A and B) The relationship between VL and HLA-B–associated polymorphisms in Gag is shown for recipients who lack (A) or carry (B) the HLA-B alleles (more ...)
This finding also suggests that the recipients who are able to effectively target Gag themselves would derive less absolute benefit from CTL-induced Gag mutations because of the equipoise between CTL-derived virus suppression and the fitness costs associated with escape mutations. Indeed, we did not see any association between the number of Gag mutations and VL in recipient carriers of HLA-B alleles that select for the Gag CTL escape mutations described in (ρ = 0.07; NS; ). To determine whether newly induced Gag-specific CTL responses could be contributing to VL control in this latter recipient group, the number of HLA-B–restricted Gag epitopes that could potentially be targeted in each individual was quantified and compared with VL. In line with previous findings (4
), the greater the number of Gag CTL epitopes potentially targeted by these recipients, the greater the decrease in VL (ρ = −0.28; P = 0.047; ). This latter finding in recipients with Gag-specific CTLs capable of inducing mutations is also similar to a recently published paper in which an inverse relationship was demonstrated between VL and HLA-I–associated polymorphisms in Gag, Pol, and Nef in chronically infected hosts (24
). Our analysis expands on this study because CTL escape mutations induced by the recipient's immune system were almost nonexistent at the time point analyzed (unpublished data). Therefore, it is likely that the potency of CTLs associated with mutations in Gag plays an important role in early viral control, even if these mutations do not occur.
Because certain HLA-B alleles are associated with improved disease outcomes (7
), it was important to consider the possibility that the significant associations with VL were driven by Gag mutations restricted only by those HLA-I alleles previously associated with a lack of disease progression (e.g., HLA-B*57 or -B*5801). We therefore stratified the data according to which HLA-I–associated Gag mutation was present in recipients who lacked HLA-B alleles that drive escape (). Although none of the VL associations with any individual HLA-I allele were statistically significant, certain patterns could be seen. All 12 HLA-B–associated Gag mutations were linked to lower VLs in HLA-mismatched recipients (compared with only one out of five HLA-A– or HLA-C–associated mutations; P = 0.002). Of note, this trend was not unique to HLA-B alleles associated with good clinical markers, but instead was also observed with HLA-B*07, -B*35, -B*41, and -B*44. These data suggest that it is number of CTL epitopes targeted by any individual HLA allele (as assessed by the frequency of Gag mutations induced) that drives its association with lower VL. Notably, 13 out of 20 (65%) of the mutations in Gag () were restricted by HLA molecules (HLA-B*42, -B*57, -B*5801, and -B*8101) previously associated with lower VLs and a better clinical outcome (7
Figure 4. Lower VL in recipients is not restricted to a select class of donor HLA-B alleles. VLs based on transmission of viruses with (closed squares) or without (open squares) Gag mutations restricted by individual HLA-I alleles are shown. The data represent (more ...)
Previous studies have examined VL differences based on CTL escape that occurs during the course of chronic infection. These analyses are complicated by the fact that the occurrence of a CTL escape mutation in an individual abolishes immunological suppression of viral replication in that individual. Indeed, escape from an HLA-B*27–restricted Gag response results in a higher VL (3
) despite the fact that it is associated with reduced replicative capacity in vitro (14
). It is therefore difficult to assess whether any particular escape mutation results in a viral fitness cost in chronically infected patients. This current study is unique in that the identity of the transmitted virus was known in each of the individuals early after infection. This allowed us to observe the impact of HLA-I–induced mutations on virus replication after it entered a new immunogenetic environment where the fitness cost of the mutations would be most pronounced. Importantly, we demonstrate two different mechanisms of viral control in newly infected recipients. Similar to previous experiments (6
), those recipients expressing certain HLA-B alleles likely control VL by effective CTL Gag targeting. In contrast, newly infected recipients who lack HLA-B alleles associated with Gag polymorphisms can derive significant benefit from the transmission of viruses containing Gag mutations induced in partners with potent HLA-B–restricted responses.
One possible explanation for the finding that mutations in Gag, but not Nef, affect viral fitness is that the former protein (and in particular p24) must make multiple interactions with other Gag molecules during the assembly of both immature viral capsids and mature viral cores, as well as with cellular components such as cyclophilin A (25
). This relative structural inflexibility may allow the immune system, through its targeting of Gag, to significantly affect viral replication, and would provide a strong mechanistic rationale for the repeated observation that Gag-specific CTLs correlate with biological markers of improved clinical outcome (4
It is not clear why HLA-B–restricted responses directed against HIV-1 play such an important role in immunopathogenesis, although several groups have demonstrated this association (6
). It is possible the HLA-B alleles just happen to target important regions in p24 with a higher binding affinity (26
). Furthermore HLA-A, -B, and -C are differentially expressed depending on the cell type (27
), and it is plausible that certain antigen-presenting cells preferentially express HLA-B molecules. In addition to CTL function, HLA-B Bw4 interacts with NK cells, and this was recently demonstrated to play an important role in HIV-1 control (28
The ~10-fold VL reduction associated with increasing numbers of escape mutations (fewer than two versus more than six Gag mutations; ) observed in this report may be clinically significant, because even a 2.5-fold lower VL among recent HIV-1 seroconverters was associated with a benefit in AIDS-free survival (29
). Compensatory mutations (11
) or reversions (30
) occurring after our analyzed time point could dampen the impact of CTL escape on viral fitness constraints. Follow-up VL data were available for 39 recipients at ~1 yr after seroconversion or 6 mo after the first evaluation. Although the VL significantly increased at the second time point compared with the first (70,928 vs. 21,961 copies/ml, respectively; P < 0.0001 using the Wilcoxon signed-rank test), it was higher in all patients irrespective of the number of Gag polymorphisms or whether or not the patients carried the “protective” HLA-I alleles. Therefore, it is difficult to know whether the increase in plasma VL seen after the first year of infection is caused by the development of compensatory mutations, the loss of immune control, or other factors. Nevertheless, these HLA-I–induced mutations are certainly present at the earliest point of infection and would be expected to also influence peak VL. Because much of the CD4 depletion occurs before seroconversion (31
), these viral fitness mutations may contribute to long-lasting clinical benefit even if their biological impact is lost at some point after seroconversion. It would have been interesting to evaluate the impact of viral fitness mutations on CD4 T cells; however, this clinical test was not available at the time of the study.
Although these findings need to be replicated in other cohorts, they strengthen the argument for CTL targeting of Gag antigens as an important mechanism for sustained viral control. Moreover, if a virus encoding escape mutations in gag were to be transmitted to a new host in the absence of fully compensatory mutations (as observed in this report), these new recipients would likely benefit as well. These data imply that for CTL-based HIV vaccines to effectively control VL, they must simultaneously target multiple Gag epitopes, thereby ensuring that fitness constraints prevent the virus from easily mutating. Thus, efforts to identify other CTL-induced HIV-1 mutations in structurally important regions of Gag (or other proteins), and to define mechanisms by which these mutations influence VL set point and disease progression will ultimately benefit the design of an AIDS vaccine.