Recent analyses of the STEP trial suggest that T-cell vaccination strategies can be successful, but only if effective CD8+
T-cell responses are induced, and in this case, they were seen only in subjects expressing the already protective alleles HLA-B*27/57/5801 (1
). This highlights the importance of fully understanding which specific responses mediate improved immune control of HIV to guide future HIV vaccine design. The study presented here is the first to comprehensively define what epitopes are targeted by the HLA class I alleles most strongly associated with an outcome of immune control in adult chronic C-clade HIV infection. Despite HLA-B*57/5801 alleles having been well studied, it has remained unclear which epitopes are presented and to what extent they are targeted. Here, we report a detailed characterization of all the epitopes within the HIV proteome that are restricted by any of three closely related, protective HLA alleles, HLA-B*5702, HLA-B*5703, and HLA-B*5801, that are significantly targeted in a large cohort (n
> 1,000) of chronically infected study subjects. Having undertaken this analysis, we conclude that HLA restriction has a distinct impact on both the immunogenicity and the selection pressure imposed on the virus by each response. In addition, the differences in control of disease progression observed for these closely related HLA alleles are most clearly correlated with the number of p24 Gag-specific responses that drive significant selection pressure on the virus.
The amino acid differences between the 3 closely related HLA-B*57/5801 alleles represent micropolymorphisms, a term previously coined in relation to three closely related HLA-B*44 alleles, HLA-B*4402, HLA-B*4403, and HLA-B*4405, that differ at two positions, 116 and 156 () (2
). This HLA-B*44 study showed that the amino acid changes do not alter the T-cell receptor (TCR) contact sites on the MHC but alter the confirmation observed for peptide residues 4 to 7 for the three HLA-B*44 alleles. The same principle may apply to peptides presented by HLA-B*5702, HLA-B*5703, and HLA-B*5801, where they bind the same peptides but may present these peptides in very different peptide-MHC landscapes. For example, HLA-B*4402 and HLA-B*4403 differ by only 1 amino acid, and in the same position (Asp and Leu, respectively, at residue 156) as HLA-B*5702 and HLA-B*5703 (Arg and Leu, respectively, at residue 156). The crystal structures of the same peptide presented by HLA-B*4402 and HLA-B*4403 showed a wider peptide-binding cleft for HLA-B*4403 than for HLA-B*4402, and while the majority of peptides presented by HLA-B*4403 are also presented by HLA-B*4402, a broader repertoire of unique peptides is presented by HLA-B*4403 (36
). The hydrophobic side chain in residue 156 for HLA-B*5703 (156-Leu) may similarly disrupt water-mediated hydrogen bonds serving to keep the peptide-binding cleft relatively narrow in HLA-B*5702 (156-Arg) compared to HLA-B*5703. This may alter the precise orientation of the same peptide in the groove of B*5702 versus B*5703, as well as increasing the number of possible epitopes presented by HLA-B*5703 compared to HLA-B*5702, as was observed here (). Other studies of the HLA-B*41 family show how 6 HLA-B*41 subtypes, differing in 6 MHC I residues, present different peptides of various lengths and thereby create structurally distinct peptide-HLA complexes (4
). These studies underline the consequences of HLA micropolymorphism changes in residues lining the peptide-binding groove.
In this study, the impact of only 1 residue change between HLA-B*5703 and B*5702 on peptide presentation was highly evident, as these alleles share only 50% of the HIV epitopes, with 7 more epitopes targeted by HLA-B*5703 than by HLA-B*5702. The larger repertoire of unique epitopes presented by HLA-B*5703 is of particular interest in relation to the KF11 Gag epitope, targeting of which has been associated with a lower viral load than HLA-matched nonresponders (30
) and in which the accumulation of escape mutations increasingly cripples the virus, thereby facilitating immune control (12
). HLA-B*5703 is the only one of these closely related HLA-B*57/5801 molecules capable of a broad p24 Gag-specific response. The differential targeting of these well-studied epitopes in p24 Gag, ISW9, KF11, TW10, and QW9, is related to the viral-load set point in that HLA-B*5703 targets all four, while HLA-B*5702 and B*5801 target three and two, respectively. This result is consistent with our previous findings (40
), which showed that the viral-load set point associated with each HLA-B allele is correlated with the number of Gag responses restricted by that HLA-B allele (r
= −0.49; P
= 0.013). Our findings here also support an additional result of that study, namely, that the ability of a particular HLA-B-restricted Gag response to impose selection pressure on the virus is more strongly correlated with the viral-load set point (r
= −0.62; P
= 0.0009) (40
). Here, we have shown that the only Gag epitopes restricted by the HLA-B*5702/5703/5801 alleles where there is evidence of escape are ISW9, KF11, TW10, and QW9. Furthermore, if one compares the number of p24 Gag epitopes restricted by these 3 HLA-B*57/5801 alleles that show evidence of escape (ISW9 and TW10 for B*5702; TW10 and QW9 for B*5801; and ISW9, KF11, and TW10 for B*5703) with the number of Gag responses driving escape via the other 22 HLA-B alleles (7 responses divided among 22 alleles; mean, 0.32 responses/allele), one arrives at a remarkably strong correlation with the viral-load set point (F). Although it is clear that many other factors contribute to the viral-load set point, in addition to the number of CD8+
T-cell Gag responses that can impose selection pressure on the virus, nonetheless, it is a surprisingly strong correlation, given the small number of data points available. Of note, no such correlation of type was observed when we considered epitopes in Pol, in the accessory/regulatory proteins, or in Env. However, other mechanisms may also be involved, such as the different linkage disequilibrium (LD) to HLA-A and Cw alleles. In particular, the LD of HLA-B*5703 with HLA-A*7401 has recently been shown to have an additive effect for immune control over HLA-A*7401 (34
), whereas HLA-B*5801 in LD with HLA-A*0205, for example, may not add any beneficial effect.
From the comprehensive sequence analysis of escape mutants within the 17 epitopes studied here ( and ), we note that sequence variability from the wild-type consensus is present in >50% of individuals not expressing HLA-B*5702, HLA-B*5703, and HLA-B*5801 for 5 of these epitopes (LF11-p17, FF9-Pol, IAW9-Pol, KF9-Nef, and HW9-Nef). The accumulation of the A83G mutant within KF9-Nef and H119N within HW9-Nef, selected by HLA-B*57/5801, and M377L within IAW9, selected by HLA-B*5801, may represent epitopes being driven toward extinction (33
), precipitating new “consensus” sequences. In the cases of LF11-p17 and FF9-Pol, the high sequence variability may not be driven by HLA-B*57/5801 but may result from epitope clustering and selection pressure imposed by other alleles on overlapping epitopes. This finding raises the possibility that protection from HLA-B*5702, HLA-B*5703, and HLA- B*5801 may alter over time. However, whether this protection increases or decreases as a result of such changes is unknown. Taken together, these data support the hypothesis that a broad Gag-specific response is protective against HIV disease progression, that Gag-specific responses capable of driving selection pressure on the virus have the strongest protective effect, and that the outstanding difference between HLA-B*5703 and the other closely related HLA alleles, HLA-B*5702 and HLA-B*5801, is the immunodominant response to KF11.
As described above, we have illustrated an approach to characterizing the significant CD8+
T-cell responses restricted by different HLA class I alleles that is not biased by use of peptide-binding motifs and in which confirmation by a peptide-MHC tetramer demonstrates both the optimal epitope and the restriction element. There are relatively common examples in the literature of epitopes that have proven to be incorrect in sequence and/or by HLA restriction (14
In this paper, we also took advantage of the opportunity to assess the part played in immunogenicity by peptide-MHC binding avidity and the peptide-MHC stability half-life and found significant differences between immunogenic and nonimmunogenic optimal peptides, which is in agreement with other findings (M. Harndahl, unpublished data). We also found significant correlations between the percentage of chronically infected subjects recognizing the peptide and peptide-MHC binding avidity and peptide-MHC stability. However, the Spearman rank correlation (r
) values found here were moderate, which is not unexpected and indicates that other factors also influence immunodominance, as comprehensively reviewed by others (3
). Presentation of an MHC class I-restricted epitope to a specific CD8+
T cell is the culmination of a number of individual upstream events, including proteasome cleavage of viral proteins, transportation to the ER lumen by transporters associated with antigen (TAP), and further N-terminal trimming by ERAAP1/2 before peptide loading onto empty MHC class I molecules by the peptide-loading complex (PLC) and subsequent trafficking to the cell surface via the Golgi apparatus for recognition of CD8+
T cells (56
). In addition to these processing events, the availability of CD8+
T cells expressing TCRs specific for the processed peptide applies another level of complexity influencing immunodominance, as TCR selection from a vast naïve T-cell repertoire containing millions of unique TCRs is not a stochastic process but is routinely ordered and biased (43
). Furthermore, protein abundances differ substantially, in the case of HIV proteins representing an immunogenicity advantage for the highly abundant Gag epitopes (7
), and some protein epitopes are presented earlier in the viral life cycle than epitopes from other viral proteins (49
). In addition, the presence of viral-sequence escape mutations further affects immunogenicity. Therefore, the measurement of peptide binding and peptide stability of the HLA molecule is unlikely to precisely predict the outcome of this very complex set of individual events that are important for immunogenicity. However, it is evident that peptide-MHC binding is a necessary but not sufficient prerequisite for the initiation of a CD8+
T-cell response (3
). This is exemplified by the low LW9-B*5702 stability half-life (<0.5 h) (D), which results in a complete lack of detectable LW9-specific CD8+
T-cell responses in subjects expressing HLA-B*5702 (C).
In conclusion, this study defines the differences between three closely related protective HLA class I alleles in terms of the epitopes presented, which are targeted by a large cohort of HIV-infected study subjects, and in terms of the selection pressure imposed by these responses on the virus. These data support earlier findings (40
) that the critical differences distinguishing HLA alleles are the breadth of the Gag-specific CD8+
T-cell responses, in particular the p24 Gag-specific response, and the abilities of those responses to drive selection pressure on the virus.