Frequencies of Gag-restricted T-cell escape “footprints” differ across HIV-1 clades A1 and D chronically infected Ugandans irrespective of host HLA B alleles

Highlights • A and D infected subjects even though they bear the same presenting HLA alleles, and live in the same environment. Escape mutations that are known to confer survival advantage were more frequent in clade A-infected subjects irrespective of host HLA alleles.• There was no evidence to link this difference in outcome to the evaluated adaptive T-Cell responses (IFN-γ responses and polyfunctional responses) to those key structurally constrained Gag epitopes.• However, we have demonstrated that there was significantly greater selective pressure on the Gag protein of clade A than that of clade D.• The data are in line with the known faster disease progression in clade D than clade A infected individuals.• The data also highlight that the current difficulties in formulating a global HIV vaccine design will be further challenged by clade associated differences in outcome.

HLA-restricted imprints in structurally compromised epitopes would be expected to follow predicted patterns in subjects with the same presenting alleles; however, this has not always been the case [26,27]. It is not clear how concurrent T-cell responses are attributable to this outcome. Here, we combined adaptive Tcell responses, host HLA alleles and the KF11, ISW9 and TW10 epitope sequences to evaluate how frequencies of critically relevant epitope escape correlate with concurrent T-cell responses across clades A and D infection among subjects living in the same environment.

Study population and evaluation of immune responses
HIV-1 infected, therapy-naïve subjects were recruited for a cross sectional evaluation. Participant plasma viral loads (HIV RNA copies per ml), CD4+ T-cell counts (cells/l) and HLA alleles were determined as previously described [10]. Infecting clades, estimation of selection pressure and KF11, ISW9 and TW10 epitope diversities were determined from the gag sequences. Cryopreserved peripheral blood mononuclear cells (PBMC) from 44 subjects were initially evaluated for IFN-␥ response to KF11, ISW9 and TW10, Table 1. Sixteen subjects were further assessed for simultaneous secretion of IFN-␥, IL-2, TNF-␣ and Perforin in response to KF11, ISW9 and TW10, using intracellular cytokine staining assay. Selection for flowcytometry evaluations was based on cell availability, presence of A163G mutations in the KF11epitope and/or possession of HLA B*57 or B*5801 alleles. Uganda Virus Research Institute Ethics Review Board and the Uganda National Council of Science and Technology reviewed and approved this study. All subjects provided written informed consent for collection and subsequent evaluation of their specimens.

Estimation of synonymous (dS) and non-synonymous (dN) rates
Selective pressure was computed from the rates of nonsynonymous (dN) and synonymous (dS) substitutions. The contributions of dN and dS rates to the overall substitution rate were estimated based on the posterior substitution rates estimated by Bayesian analysis. Their estimation was performed using a local codon model, as implemented in HYPHY [28]. We used the MG94xHKY85 codon model [29], an extension of the classical MG94 model with estimation of equilibrium codon frequencies, using nucleotide frequencies specific to each codon position.

The ELISpot assay
Virus-specific IFN-␥ responses were quantified using ELISpot assay, as previously described [30]. Responses were enumerated as spot forming units per million PBMCs (SFU/10 6 PBMCs). The test acceptance criteria were: ≥300 SFU per PHA well; ≤100 cumulative spots in all the six background wells; and ≤5 cumulative SFU in the two wells that contained media only. Test wells with ≥100 net SFU/10 6 PBMCs after subtracting three times the background were considered positive.

Functional avidity of the KF11 peptides
The IFN-␥ ELISpot assay was used to evaluate five HLA B57 subjects for binding affinities to the 11-mer KF11 peptide; the study subjects were selected based on PBMC availability. Binding affinities were determined using duplicate 2-fold serial peptide dilutions ranging from 2 g/ml to 4 pg/ml), as described elsewhere [31]. Peptide concentrations that yielded half the maximum number of sport forming units were determined from a sigmoidal doseresponse curve fit using Graph Pad Prism 5. Functional avidities were expressed as half-maximal stimulatory peptide doses (SD 50%).

Intracellular cytokine staining assay
HIV-specific T-cell polyfunctional responses were assessed using intracellular cytokine staining assay, as previously described [32], but with slight modifications. Briefly, thawed and rested PBMCs were incubated with 1 g/ml peptide for 6 h at 37 • C in a 5% CO 2 -in air, humidified environment, in the presence of Golgi Plug TM and CD28/CD49d co-stimulatory antibodies. Negative controls were PBMCs incubated as above but without peptides. Positive controls were PBMCs incubated with Staphyloccoccus enterotoxin B (SEB) instead of peptides. Stimulated PBMCs were then washed in 10% FBS in PBS, and incubated with Aqua viability dye before surface staining for T-cell lineage markers. The PBMCs were subsequently intracellularly stained, for the simultaneous expression of Interferon (IFN)-␥ IL and Perforin. Flowcytometry data was analyzed using FlowJo (version 9.5.3, TreeStar), Pestle (version 1.6.2) and SPICE (Version 5.3101) [33] software. The gating strategy for definition of IFN-␥, IL-2, TNF ␣ and Perforin is illustrated in Supplementary Fig. 1.
Supplementary material related to this article can be found, in the online version, at http://dx.doi.org/10.1016/j.vaccine. 2015.02.037.

Statistical analysis
Statistical analyses were performed using Stata V 10.0 software (Stata Corp, TX, USA) and Graph Pad 5.0 (GraphPad Software, Inc., San Diego, CA, USA). Graphical presentations were performed using SPICE and Graph Pad software. Continuous data is presented as medians, with their interquartile ranges (IQR). Medians were evaluated using Mann Whitney test if there were two groups; or Kruskal-Wallis Rank Sum test if there were more than two groups. Proportions were compared using the Fisher's Exact test. Means were compared using the Student's t-test. p values ≥0.05 were considered significant.
3.6. In HLAB*57 subjects with the A163X mutation, virus-specific CD8+ T-cell polyfunctionality were at similar frequencies across clades In HLA B*57/*5801 subjects, measurable CD4+ T-cell responses ( Fig. 2A) and CD8+ T-cell responses (Fig. 2B) were detected at varying frequencies against at least one KF11 variant. Assessment of the relationship between host HLA B alleles, possession of A163X, and the associated KF11-specific CD8+ T-cell response revealed no evidence for superior T-cell response(s) in clade A1-than clade D-infected subjects. Overall, frequencies of the detected virusspecific Perforin and CD8 + T-cell polyfunctionality were equivalent or higher in some clade D infected individuals ( Fig. 3A and B, respectively) compared to A1 subjects (Fig. 3C).
The CD8+ T-cell response to wild type KF11 (KF11-1) were of comparatively lower frequencies in both clade A infected (Fig. 4A and B) and clade D infected subjects ( Fig. 4C and D). Secretion of TNF-␣ dominated the CD8+ T-cell response, while IFN-␥ responses were marginal or absent. Wild type KF11 responses were largely monofunctional and lacked Perforin. On the other hand, variant responses were more polyfunctional and were detectable at comparable frequencies in both clades A1 and D subjects. Taken together, these data do not suggest any evidence for superiority of T-cell responses to KF11 in clade A1 infected patients compared to clade D infected patients.

Clade A sequences were subjected to greater selective pressure than clade D sequences
We then evaluated the rates of non-synonymous (dN) and synonymous (dS) substitutions, as a measure of the selection pressure on the clades A and D Gag protein sequences, Fig. 5. The dN/dS ratio provides a measure of the selection pressure to the reference sequence. We found four positively selected sites, and 79 negatively selected sites in clade A sequences; while clade D had eight and 67 selected sites, respectively, 0.1 significance level. There was significantly more negative selection in clade A [mean dN/dS ratio 0.328214 (95% CI = [0.29962,0.358633]) than in clade D [mean 0.438708 (95% CI = [0.407466,0.471585]) sequences, continuous extension of the binomial distribution test, p = 5 × 10 −7 (<0.005). We also observed an overall low synonymous substitution rate for HIV-1 clade A, which is known to be less pathogenic than HIV-1 clade D.  Fig. 2A and B) and D (Supplementary Fig. 2C-E). Some low-grade ELISpot titres were observed in both clades. Wild type KF11 responses were of higher avidity than mutant responses, Supplementary Fig. 3A and D, clades A and D respectively. Sigmoid curves from two datasets that showed more robust IFN-␥ responses ( Fig. 2B and E) revealed no difference in functional avidities across clades A and D, Supplementary Fig. 3.
Supplementary material related to this article can be found, in the online version, at http://dx.doi.org/10.1016/j.vaccine. 2015.02.037.

Discussion
The greatest impact on HIV disease may be mediated by HLA B-restricted induction of effective virus-specific CD8+ T-cell responses to certain Gag T-cell epitopes [18]. The resultant protection is achieved through selective pressure [10,11,18,21] on the structurally constrained ISW9, KF11 and TW10 Gag p24 CD8+ T-cell epitopes, generating variants with impaired replicative ability [11,12,34,35]. Ugandan is concurrently infected with nearly equal proportions of co-circulating clades A and D viruses [36]. We evaluated whether these virus variants that confer apparent advantage to hosts, accumulate at similar frequencies across clades A and D infection; and determined how frequencies of those variants correlate with the concomitant T-cell responses. We combined measurements of adaptive T-cell functionality with ISW9, KF11 and TW10 Gag p24 epitope sequence data to demonstrate that the A163X mutation in KF11 was more associated with clade A1-infection; the compensatory mutation I147X, known to restore immune recognition of the ISW9 epitope [37], was more frequent in clade A1 subjects; but frequencies of TW10 epitope polymorphisms did not significantly vary across clades.
Accumulation of HLA-restricted mutations in HLA matched hosts has been demonstrated before [38,39]. In matched hosts, HLA-B*57-restricted A163X escape mutation in the KF11 epitope has been shown to result in reduced plasma viral loads [11,40]; while I147X compensatory mutation has been linked to restoration of ISW9 epitope recognition [37]. Structurally constrained Gag p24 T-cell epitopes would be expected to yield similar escape pathways in allele presenting subjects. However, only a third of HLA matched hosts mount an epitope-specific response [41]; and the six major MHC determinants of HIV-1 protection in Caucasians [42] lacked protective associations in African cohorts [26]. It is also unclear why the widely protective HLA B*5801 [18,43,44] lacked protection in clade A-infected Rwandans [45]. Others also demonstrated higher A163X frequencies among clade A subjects [27]. However, it remained unclear whether this outcome depended on differences in concurrent adaptive responses. The higher rates of A163X mutation observed in HLA matched and mismatched clade A1 subjects in our population suggest better tolerance of those changes by clade A viruses. Differential clade-based outcome was also reported in clades B and C infected populations; where it was linked to functional constraints imposed by HIV clade on specific Gag residues [46]. Here, we found evidence for greater selective pressure on clade A Gag compared to clade D Gag; implying that the clade-specific outcome were linked to differences in immune pressure on the infecting virus clades. These findings are in line with those recently reported, which demonstrated clade-associated superior targeting of Gag among clade C-compared to clade B-infected, HLA B*0702 subjects [47].
Indisputably, T-cells exert strong selection pressure on HIV-1, generating HLA allele-restricted mutants that evade subsequent T-cell responses [48,49]. Emergence of HLA-B*5701-restricted variants has been linked to superior maintenance of IL-2 and Perforin, and persistence of polyfunctional CD8+ T-cell responses [50]. Superior virological control has been linked to greater functional avidity of the T-cell responses [31]. Here, we found no evidence for lower KF11 binding affinities to clade D peptides. However, avidity comparisons reported here needs to be interpreted with caution, as there were limitations of sample size. A larger sample size will be required to adequately evaluate relationships with functional avidity. In this study, plasma viral loads were eight-fold lower in presenting hosts bearing the A163X mutation suggesting the likelihood for potential T-cell involvement. The evaluated epitopespecific IFN-␥ and polyfunctional responses were similar across clade A1 and D suggesting that clade-specific variations observed were not merely due to differences in the adaptive T-cell responses assessed here. The detected preservation of wild-type KF11 despite existence of epitope-specific polyfunctional CD8+ T-cell responses suggests that either the responses to clade D infection was ineffective or the virus was less tolerant of the mutations. We observed that responses to mutant epitopes were polyfunctional, while CD8+ T-cell polyfunctionality to the wild type epitopes had diminished. Possibly the responses were exhausted; specimen limitations did not allow for evaluation of exhaustion status of the antigen-specific T-cells. It is not surprising that we detected few TW10 directed epitope escape. The evaluated subjects were all chronically infected; yet TW10 epitope is targeted by acute CD8+ T-cell responses while the B*57-restricted KF11 epitope is dominant in chronic infection [18].
Overall, association of clade A1 subjects with escape mutations that are known to confer protection supports the notion that clade A infection with slower disease progression than D [1][2][3]10]. Overall, the data also support others that linked Gag targeting with improved outcome [10]. While the associations we report here are significant, there were some limitations. First of all, a longitudinal follow up would have better enabled evaluations of relationships between the various factors and the trends in evolutionary pathways. Secondly, these subjects were randomly selected for HLA B*57/5801 expressions without bias; finding of such associations in relatively few subjects suggests prevailing circulation of KF11 polymorphisms in clade A1-infected populations. However, a larger cohort will possibly allow for evaluation of correlations with rarer determinants. Lastly, further studies will be necessary to evaluate how these persistent dominant antigenic stimulations translate into T-cell exhaustion.
Overall, these data imply that the existing obstacles to HIV Tcell based vaccines may be further complicated by clade-associated differences in selective pressure on known HLA associated Gag outcomes. Even if critical epitopes remained as targets, their continued accumulation and adaptation in infected populations may lower the cost to viral fitness possibly reducing the host benefit. If the wild epitopes remain preserved, T-cell responses will likely persist but may become exhausted and functionally impaired due to continued antigenic stimulation. The data also raises hope that increased selective pressure on Gag will contribute to protection irrespective of host HLA alleles. The work presented here highlights the need for improved understanding of how HIV-1 diversity influences population-specific correlates of protection. The data have implications for T-cell vaccine approaches targeting highly conserved virus sequences to attenuate the infecting virus, and underscore the need to understand the direction of virus evolution when designing HIV T-cell vaccines.