Definition of the viral targets of protective HIV-1-specific T cell responses

HIV-1 infection induces strong and broadly directed HLA class I restricted T cell responses for which specific epitopes and restricting HLA class I alleles have been associated with relative in vivo viral control [1]. The bulk of the anti-viral CTL response appears to be disproportionately HLA-B restricted, but the relative contribution of targeted viral regions and restricting HLA molecules on the effectiveness of these responses remains unclear [2‐5]. In addition, the impact of HIV-1 sequence diversity on the effectiveness of virus-specific T cell immunity in vivo is unclear, as functional constraints of escape variants, codon-usage at individual protein positions, T cell receptor (TCR) plasticity and functional avidity and cross-reactivity potential may all contribute to the overall antiviral activity of a specific T cell response [6‐13]. Of note, T cell responses to Gag have most consistently been associated with reduced viral loads in both clade B and clade C infected cohorts [14‐16]; however, the specific regions in Gag responsible for this effective control remain poorly defined. In addition, it is unclear whether the relative benefit of Gag is due to any other specific characteristic of this protein, such as rapid antigen-representation upon infection, protein expression levels, amino acid composition and/or inherently greater processability and immunogenicity, particularly in the context of selected HLA class I alleles [17, 18]. Thus, concerns remain that a purely Gag-based vaccine might mainly benefit those people with a particular HLA genotype and will not take advantage of potentially beneficial targets outside of Gag [4, 16, 17, 19]. In addition, CTL escape and viral fitness studies have focused largely on Gag-derived epitopes presented in the context of protective HLA class I alleles such as HLA-B27 and -B57 [7, 20, 21], yielding results that may not be generalizable to the genetically diverse majority of the human population. Furthermore, many studies have focused on immunodominant targets only, despite some studies in HIV-1 and SIV infection demonstrating a crucial contribution of sub-dominant responses to targets outside of Gag to the effective in-vivo viral control [4, 22]. Thus, the current view on what may constitute a protective cellular immune response to HIV-1 is likely biased towards a immunodominant responses and those restricted by frequent HLA class I alleles and HLA alleles associated with superior disease outcome.

To overcome these potential limitations, the design of an effective and broadly applicable HIV-1 vaccine should to be based on information gained through comprehensive analyses that extend across large portions of the population's HLA class I heterogeneity. Here we focus on three cohorts totaling more than 950 untreated, chronically HIV-1 infected individuals with clade B and C infections, from which responses to certain regions of the viral genome and specific T cell response patterns emerge as correlates of viral control. Importantly, the analyses identify functional properties unique to these responses and control for the impact of HLA class I alleles known to be associated with superior control of HIV-1 infection, thus providing vaccine immunogen sequence candidates with potential usefulness in a broadly applicable HIV-1 vaccine.

Defining functional correlates of HIV-1 immune control is critical to the design of effective immunogens. T cell responses to specific HIV-1 proteins and protein-subunits have been associated before with relatively superior viral control in vivo [14, 16, 49], but evidence from recent clinical trials suggests that including maximal immunogen content into various vectors does not necessarily induce more effective CTL responses [50, 51]. In fact, it has been argued that the existence of potential "decoy" epitopes may divert an effective CTL response towards variable and possibly less effective targets in the viral genome [52]. Thus, the definition of a minimal yet sufficient immunogen sequence that can elicit CTL responses in a broad HLA context is urgently needed. Thereby, focusing vaccine responses on conserved regions could help induce responses towards mutationally constrained targets and provide the basis for protection from heterologous viral challenge.

We present here the results of an extensive analysis that included more than 950 HIV-1 infected individuals with diverse HLA genotypes, from three different continents and including clade B and C infections. In both, the analysis in clade B in Lima and clade C in Durban, individual OLP were identified that are predominantly targeted by individuals with reduced or elevated viral loads, although the different size of the cohorts required different statistical approaches for their identification. In general, most of these OLP were among the more frequent targets in the HIV proteome, possibly due to both, the need for sizable responder groups to achieve statistical significance in the viral loads comparison as well as the high epitope density in these OLP. The identified OLP were frequently located in HIV-1 Gag and Pol, but rarely in the more variable proteins such as Env and Nef. With one exception, Nef and Env featured only non-beneficial OLP, thus arguing against their inclusion, at least as full proteins, in a CTL immunogen sequence [16]. In addition, in both cohorts, the Vif protein yielded few, yet exclusively beneficial OLP, which may warrant a renewed look at the inclusion of regulatory proteins in vaccine design [53, 54]. Also common to both clades, (and despite the wide scatter possibly due to the inclusion of less-frequently targeted OLP), an negative correlation between sequence entropy and PR was observed providing strong rationale for vaccine approaches that focus on conserved viral regions where T cell escape may be complicated by structural constrains [55]. This was particularly evident in the clade C cohort, where even within the relatively conserved Gag protein, a lower entropy was seen for the beneficial OLP compared to the remainder of the OLP spanning the protein. On the other hand, while beneficial and non-beneficial OLP showed a significant difference in their median entropy in the clade B cohort, this comparison was not significant in the clade C cohort. It is possible that the immunogen sequence, designed in 2001, did not optimally cover the circulating viral population in Durban throughout the enrollment period (until 2006), leading to missed responses particularly in the more variable segments of the virus [32, 56]. The study may have thus failed to identify beneficial as well as non-beneficial OLP in the more variable genes of HIV. This should have preferentially affected highly variable OLP due to a more frequent mismatch between autologous viral sequence and in vitro test set in these regions. However, even if scoring as beneficial OLP, such high-entropy OLP may from an immunogen-design point of view be of less interest as they would possible contribute only little to protection from heterologous viral challenge. It needs however also to be considered that the OLP-specific entropy values are based on variable numbers of sequences in the Los Alamos HIV database covering the different OLP, introducing potential further bias into these analyses, particularly for less covered proteins such as Vpu and other viral protein products. Such differences between autologous viral sequences and in vitro test sets may also have impacted the assessment of functional avidities. These determinations included responses in the same individual towards epitopes located in beneficial and non-beneficial OLP; with the former overall being more conserved. Thus, the higher functional avidity towards epitopes located in beneficial OLP could be biased by the higher chance that these epitopes matched the autologous viral sequence compared to epitopes located in non-beneficial OLP and which may thus have induced a more robust, avid response. Apart from covering autologous sequences, future studies will ideally also include comparable analyses in individuals identified and tested in acute infection that go on to control the infection at undetectable levels of viral replication (i.e. elite-controllers) so that the selective early emergence of responses to beneficial OLP could be linked to relative control of viral replication in chronic infection. As is, the identified beneficial responses may be particularly important to maintain low viral replication in chronic stages of infection, which in theory could be different (for instance due to more accelerated intra-individual viral evolution in variable genes) from responses determining viral set point during acute infection. However, the existing HLA bias in such cohorts and the small number of responses identified during earliest stages of infection may make such analyses a formidable undertaking that will require large numbers of individuals to be tested longitudinally.

A broadly applicable T cell immunogen sequence should include T cell targets restricted by a wide array of HLA class I alleles. Although broad representation of HLA-B alleles may be particularly important in this regard, emerging data on the effects HLA-C alleles in these cohorts may warrant a broad HLA-C representation as well [2, 47, 57]. In the present study, the 26 beneficial OLP from Lima and the 22 beneficial OLP from Durban covered 26 described, optimally defined CTL epitopes restricted by 20 different HLA alleles for the clade B cohort and 33 epitopes presented by 34 alleles for the clade C cohort, respectively [44]. As this is likely to be an underestimate of the true diversity in HLA restriction (Table 2 and ref [58]), it is reasonable to predict that the inclusion of identified beneficial OLP, or even a subset thereof, could evoke potential responses in a widely diverse HLA context. This could also provide the basis for the induction of poly-specific T cell responses with increased breath, which the present data clearly associates with progressively lower viral loads and which emerge as a potentially important parameter from several recent vaccine studies showing superior protection from SIV challenge in animals with a broad vaccine induced responses to Gag p17 [59, 60].

Recent studies have suggested a global adaptation of HIV-1 to its various host ethnicities [4, 46]. The consequence of such adaptation has led in some cases to the elimination of protective CTL targets, causing a profound absence of responses to these epitopes and detrimentally changing the association between HLA allele and HIV-1 disease outcome [4]. It is thus not surprising that the two main cohorts tested here yielded only partially overlapping sets of beneficial OLP as the impact of host genetics and viral evolution in the studied populations cannot readily be overcome. In fact, given studies by Frahm et al [4], the past and current adaptation of HIV-1 to common HLA class I alleles will likely still call for somewhat population tailored vaccine approaches, especially if the immunogen sequences should be kept short to avoid regions of potentially reduced immunological value [52]. Such approaches will also profit from more extensive structural analyses that may identify specific domains of viral proteins that are or are not enriched in valuable T cell targets; of which the latter could possibly be ignored for the design of T cell immunogen sequences. Additional analyses in other genetically unrelated cohorts of HIV-1 infected individuals and studies in SIV infection may further help to guide such selective immunogen design and to understand the factors defining the effectiveness of different epitopes in mediating relative HIV-1 control. Of note, the beneficial OLP identified here, 24 in clade B and 22 in clade C infection matched other immunogen design based on conserved elements in some parts as well, i.e. of the 14 conserved elements proposed by Hanke et al, eight (57%) overlapped at least partly with beneficial OLP identified here [61]. Similarly, among the highly conserved elements proposed by Rolland et al [52], 35% (5/14) were covered by our beneficial OLP in clade B infection. These differences possibly emerge because the present analysis is based on functional T cell data rather than viral sequence alignments, which may not take into consideration epitope density and processing preferences of certain regions. Nevertheless, the partial overlap with these other immunogen design support the focus on conserved regions and offers the opportunity for alternative or combined vaccine approach that elicit responses to regions where the virus is and possibly remains vulnerable [4, 46, 55, 62].

Finally, we used the extensive data set available to approach the question of relative effects of host genetics (i.e. HLA) and CTL specificity on HIV-1 control. While the two factors cannot be entirely disentangled, our data suggest that CTL specificity has an at least equal if not stronger effect on viral control than HLA class I allele expression. These findings are also in line with data by Mothe et al [63] showing that targeting key regions in p24 surrounding the dominant epitopes restricted by known protective alleles (KK10 for HLA-B27 and TW10 for HLA-B57/58) in HLA-B27, -57 or B58 negative individuals is associated with significantly reduced viral loads. In addition, the presence of individuals not expressing known beneficial alleles in HIV-1 elite controller cohorts [64], further indicates that HIV-1 control is not necessarily bound to a few specific HLA class I alleles. A detailed study of the total HIV-1-specific CTL response of subjects not expressing these alleles yet effectively controlling HIV-1 can be expected to provide further and crucially needed insight into the importance of targeting specific (conserved) regions of the viral genome for HIV-1 control. Similarly, the characterization of functional attributes of these responses, including functional avidity and the ability to suppress in vitro viral replication will need to be further assessed in such individuals. Building on experimentally derived and potentially promising immunogen sequences as defined here may thus provide a suitable basis for further immunogen design and iterative clinical trials in the human setting.