HLA-II-Associated HIV-1 Adaptation Decreases CD4⁺ T-Cell Responses in HIV-1 Vaccine Recipients

ABSTRACT

Epitopes with evidence of HLA-II-associated adaptation induce poorly immunogenic CD4⁺ T-cell responses in HIV-positive (HIV⁺) individuals. Many such escaped CD4⁺ T-cell epitopes are encoded by HIV-1 vaccines being evaluated in clinical trials. Here, we assessed whether this viral adaptation adversely impacts CD4⁺ T-cell responses following HIV-1 vaccination, thereby representing escaped epitopes. When evaluated in separate peptide pools, vaccine-encoded adapted epitopes (AE) induced CD4⁺ T-cell responses less frequently than nonadapted epitopes (NAE). We also demonstrated that in a polyvalent vaccine, where both forms of the same epitope were encoded, AE were less immunogenic. NAE-specific CD4⁺ T cells had increased, albeit low, levels of interferon gamma (IFN-γ) cytokine production. Single-cell transcriptomic analyses showed that NAE-specific CD4⁺ T cells expressed interferon-related genes, while AE-specific CD4⁺ T cells resembled a Th2 phenotype. Importantly, the magnitude of NAE-specific CD4⁺ T-cell responses, but not that of AE-specific responses, was found to positively correlate with Env-specific antibodies in a vaccine efficacy trial. Together, these findings show that HLA-II-associated viral adaptation reduces CD4⁺ T-cell responses in HIV-1 vaccine recipients and suggest that vaccines encoding a significant number of AE may not provide optimal B-cell help for HIV-specific antibody production.

IMPORTANCE Despite decades of research, an effective HIV-1 vaccine remains elusive. Vaccine strategies leading to the generation of broadly neutralizing antibodies are likely needed to provide the best opportunity of generating a protective immune response against HIV-1. Numerous studies have demonstrated that T-cell help is necessary for effective antibody generation. However, immunogen sequences from recent HIV-1 vaccine efficacy trials include CD4⁺ T-cell epitopes that have evidence of immune escape. Our study shows that these epitopes, termed adapted epitopes, elicit lower frequencies of CD4⁺ T-cell responses in recipients from multiple HIV-1 vaccine trials. Additionally, the counterparts to these epitopes, termed nonadapted epitopes, elicited CD4⁺ T-cell responses that correlated with Env-specific antibodies in one efficacy trial. These results suggest that vaccine-encoded adapted epitopes dampen CD4⁺ T-cell responses, potentially impacting both HIV-specific antibody production and efficacious vaccine efforts.

INTRODUCTION

Despite effective therapy to prevent and treat HIV-1 infection, a protective HIV-1 vaccine is still needed to optimally decrease the number of new infections, which remains unacceptably high. Despite numerous prior HIV-1 trials evaluating active vaccine strategies, no vaccine candidate has consistently demonstrated efficacy in protecting against infection or reducing viral load in those with breakthrough infections (1–6). A recent study (AMP study) examined antibody-mediated protection from two efficacy trials focusing on passive vaccination with the broadly neutralizing antibody (bNAb) VRC01 (7). While this strategy did not decrease overall HIV-1 infection, analysis of AMP study recipients showed that those with high titers of Ab exhibited protection against VRC01-susceptible HIV-1 strains (7). This study demonstrated that protection may be achieved through an antibody-focused vaccine response, although the generation of high concentrations of multiple complementary bNAbs may be required. CD4⁺ T cells, especially T follicular helper (Tfh) cells, are important in enhancing antibody production, and optimization of CD4⁺ T-cell responses can assist in the improvement of HIV-1 vaccine-induced antibodies. Indeed, a study showed that HIV-1 vaccine recipients with specific HLA class II (HLA-II) alleles exhibited enhanced HIV-specific antibody responses, suggesting that vaccine-generated HLA-II-restricted CD4⁺ T cells were involved in antibody production (8). More specifically, another HIV-1 vaccine trial found that the frequency of CXCR5⁺ circulating T follicular helper (cTfh) cells coexpressing PD1 and ICOS correlated with Env-specific antibody titers (9).

HIV-1 can adapt rapidly under selective pressures like antiretroviral therapies or host immune responses (10–12). CD8⁺ T-cell escape against HIV-1 through the process of HLA-I-associated adaptation has been well described (13, 14); HLA-I-associated adaptation also impacts CD8⁺ T-cell responses in HIV-1 vaccine recipients (15). Although not as well described, HIV-1 adaptation from CD4⁺ T cells also occurs, and we have previously demonstrated that HLA class II-associated HIV-1 polymorphisms or adaptations can be used to predict escape against CD4⁺ T cells in the context of natural infection (16). Epitopes predicted to have evidence of CD4⁺ T-cell adaptation (termed adapted epitopes [AE]) are encoded within all HIV-1 vaccines currently being assessed; however, the impact of these escaped or adapted epitopes on vaccine-induced CD4⁺ T-cell responses has not yet been investigated. In the current study, analyses of samples from three previous HIV-1 vaccine trials show that vaccine-encoded adapted epitopes poorly induce CD4⁺ T-cell responses. Furthermore, epitopes with no such adaptation (termed nonadapted epitopes [NAE]) induced more-immunogenic CD4⁺ T-cell responses and were associated with enhanced HIV-specific antibodies. These findings provide evidence that HLA-II-associated adaptation negatively impacts CD4⁺ T-cell responses and, thus, exclusion of adapted epitopes should be considered when designing future HIV-1 vaccine immunogens.

RESULTS

Identification of HLA-II-associated adapted epitopes encoded in prior HIV-1 vaccines used in trials.

We have previously described 29 polymorphisms across the HIV-1 proteins Gag, Nef, and Pol that associated with specific HLA-II alleles in a chronic HIV-1 cohort (16). Utilizing a similar prediction strategy, 18 novel HLA-II-associated polymorphisms were identified in Env. These Env polymorphisms are summarized in Table S1 in the supplemental material. Using these 47 HLA-II-associated polymorphisms and NetMHCII (17), we identified epitopes encompassing each polymorphic site with the highest binding affinity to the respective HLA-II allele. Two different variants of each epitope were identified: those containing the HLA-II-associated polymorphism and therefore potentially having evidence of CD4⁺ T-cell escape were termed adapted epitopes (AE), and those that did not contain evidence of HLA-II-associated adaptation were termed nonadapted epitopes (NAE). Of note, the predicted in silico binding affinities showed no significant difference between NAE and AE (Fig. S1A). We next investigated the prevalence of HLA-II-associated polymorphisms encoded in vaccine immunogens across three prior HIV-1 vaccine trials: (i) an MRKAd5 vaccine trial (Step Study/HIV Vaccine Trials Network [HVTN] 502), (ii) a DNA/rAd5 vaccine trial (HVTN 505), and (iii) a DNA/MVA vaccine trial (HVTN 106). An overview of these vaccine trials is given in Table 1. While the MRKAd5 and DNA/rAd5 trials were both phase IIb trials that ultimately showed no protection against HIV-1 infection (3, 5), the DNA/MVA trial was a phase I trial that investigated multiple immunogen strategies, including a mosaic vaccine approach. Immunogen amino acid sequences were examined to identify the presence of HLA-II-associated polymorphisms. A representative example of this is shown in Fig. S1B. After analysis of sequences from all three vaccine trials, we found that between 30 and 40% of our HLA-II-associated polymorphic sites encoded AE (Fig. S1C). This finding emphasizes the need to determine the impact of these vaccine-encoded adapted epitopes.

TABLE 1.

Overview of HIV-1 vaccine trials

Vaccine	Vaccine strategy	HVTN no.	Phase
MRKAd5	rAd5 (gag/pol/nef)	502	IIb
DNA/rAd5	DNA prime (gag/pol/nef/env) + rAd5 (gag/pol/env)	505	IIb
DNA/MVA	DNA prime (env) + MVA (env)*a	106	I

^a*, multiple types.

HLA-II-associated adaptation diminishes CD4⁺ T-cell immunogenicity.

To determine if adapted epitopes negatively impacted CD4⁺ T-cell immunogenicity in HIV-1 vaccine recipients, thereby representing HIV-1 immune escape, we synthesized vaccine-encoded sequences and combined all vaccine-matched NAE and AE for each trial into separate pools. CD4⁺ T-cell immunogenicity was assessed using two methods: (i) a CD8-depleted interferon gamma (IFN-γ) enzyme-linked immunosorbent spot assay (ELISpot) assay as done in previous studies by our group (16) and (ii) a more-sensitive flow cytometric activation-induced marker (AIM) assay measuring the antigen-specific upregulation of the activation markers OX40 and PDL1, as shown by our group and others (18–21). The ELISpot results showed that NAE elicited a nonsignificant increase in IFN-γ production in MRKAd5 recipients but a significant increase of IFN-γ production in DNA/rAd5 recipients (Fig. 1A, P = 0.10 and P = 0.004, respectively). As expected, NAE was seen to induce a significant increase in the magnitude of IFN-γ-positive (IFN-γ⁺) T cells when the results for samples from both studies were combined (Fig. 1A, P = 0.001). We found a significantly increased number of positive responses to NAE, with 20% (8/40) showing detectable IFN-γ production compared to 2.5% (1/40) positive responses to AE (Fig. 1B, P = 0.03). The results from our AIM assay (a representative example is shown in Fig. 1C, and a representative gating strategy in Fig. S2) showed that CD4⁺ T cells elicited higher coexpression of OX40 and PDL1 in response to NAE across both the MRKAd5 and DNA/rAd5 trials (Fig. 1D, P = 0.03 and P = 0.01, respectively; P = 0.001 when combined). However, there was no significant difference in terms of responder frequency (Fig. 1E). In total, we observed 32 positive responses to NAE (21 definitive positive and 11 positive with uncertainty) and 26 positive responses to AE (14 definitive positive and 8 positive with uncertainty). Notably, we observed an increased number of responses with the AIM assay, highlighting this assay’s increased sensitivity compared to that of assays detecting cytokine production (18, 19). These results suggest a decreased CD4⁺ T-cell immunogenicity directed toward adapted epitopes.

FIG 1.

HLA-II-associated HIV-1 adaptation diminishes CD4⁺ T-cell responses. (A) Response magnitudes as detected from CD8-depleted IFN-γ ELISpot assays in MRKAd5, DNA/rAd5, and both trials (combined data). Dotted line represents positivity threshold (50 SFU/10^6 cells). (B) Frequencies of positive responses by ELISpot in all individuals from both MRKAd5 and DNA/rAd5 trials. (C) Representative example of upregulation of OX40 and PDL1 in AIM assay. (D) Response magnitudes as detected by AIM assay, showing the net frequencies of OX40⁺ PDL1⁺ CD4⁺ T cells in MRKAd5, DNA/rAd5, and combined analyses. (E) Response frequencies by AIM in all individuals from both MRKAd5 and DNA/rAd5 trials. MRKAd5 recipients (n = 20) are represented by circles; DNA/rAd5 recipients (n = 20) are represented by squares. P values were determined by paired Wilcoxon signed-rank test (A and D) and Fisher’s exact test (B and E). ns, not significant.

We additionally looked to verify these results in the DNA/MVA vaccine trial samples. DNA/MVA recipients (n = 35) received natural, consensus, or mosaic-based HIV-1 Env immunogens. Overall, our findings showed that adapted epitopes elicited no IFN-γ responses by ELISpot, in agreement with our findings from the MRKAd5 and DNA/rAd5 vaccine trial samples (Fig. S3A). In contrast, we observed two positive IFN-γ ELISpot responses to NAE, although there was no significant difference by magnitude (Fig. S3A). There were no differences between the three vaccine arms (Fig. S3B). With the AIM assay, we saw no increase in the magnitude of responses or differences between the three arms (Fig. S3C and D). The overall response rate in DNA/MVA recipients was lower by the AIM assay.

Polyvalent vaccine recipients show decreased immunogenicity to single HLA-II-associated adapted epitopes.

Many prior vaccine studies have used a single immunogen sequence for each HIV-1 protein. As a result, either an NAE or AE form of each HLA-II-associated epitope was included, precluding us from assessing CD4⁺ T-cell immunogenicity to both an NAE and its variant AE in individual vaccine recipients. However, recent HIV-1 vaccine trials have used multiclade and mosaic vaccine immunogens encoding multiple HIV-1 sequences in an effort to increase the breadth of both T-cell and antibody responses (22, 23). These polyvalent vaccine approaches allowed us to directly study the impact of HLA-II-associated adaptation in cases where recipients received both NAE and AE forms of the same epitope. One such example is the epitopes encoded in the multiclade DNA/rAd5 trial at the Env316 site (Fig. 2A). At this site, the clade A and clade B sequences had no evidence of HLA-II-associated adaptation and were therefore classified as NAE; in contrast, the clade C sequence had an alanine-to-threonine polymorphic change, therefore classifying this epitope as an AE. When determining CD4⁺ T-cell responses to this site, one of the DNA/rAd5 vaccine recipients showed AIM responses to both single NAE, but not to the AE (Fig. 2B). No responses were observed by IFN-γ ELISpot. We found similar sequence variability at the same site in the DNA/MVA immunogen sequence, with mosaic strain 1 (MosI) encoding an adapted epitope and strains 2 and 3 (MosII and Mos3, respectively) encoding a nonadapted epitope (Fig. 2C). In one DNA/MVA recipient, we found IFN-γ production in response to both NAE strains based on our ELISpot assay, while the response to AE was just below our positivity threshold (Fig. 2D). We observed similar data in the AIM assay, with the Env316_Mos3-NAE being the only epitope to elicit a positive response (Fig. 2E). Importantly, both of these vaccine recipients carry HLA-DR*0301, the HLA-II allele that was originally identified with this adaptation based on our prediction methods. Taken together, for immunization with both NAE and AE simultaneously, adapted epitopes are still less immunogenic.

FIG 2.

Polyvalent vaccine recipients show decreased immunogenicity to single adapted epitopes. (A) Env316 epitopes from each sequence in the DNA/rAd5 vaccine. (B) AIM responses in one DNA/rAd5 vaccine recipient to each Env316 epitope. (C) Env316 epitopes from each mosaic sequence in DNA/MVA vaccine. (D and E) CD4⁺ T-cell responses in one DNA/MVA vaccine recipient as detected by CD8-depleted IFN-γ ELISpot (D) and AIM (E) assays. (D) Dotted line represents positivity threshold (50 SFU/10^6 cells). (B and E) Statistical analyses were performed with Fisher’s exact test.

HLA-II-associated adaptation impacts frequencies of CD4⁺ T-cell cytokine production.

We next sought to determine differences in cytokine production between NAE-specific and AE-specific CD4⁺ T-cell responses by using an intracellular cytokine staining (ICS) assay to measure the expression of IFN-γ, tumor necrosis factor alpha (TNF-α), interleukin-2 (IL-2), IL-4, and CD154. In line with findings from other groups, we added IL-13 in combination with IL-4 since both these cytokines are associated with a Th2 subset (24). Due to the potential of cells with a cTfh phenotype, we also included IL-21, the canonical Tfh cytokine. The gating strategy with negative and positive controls can be seen in Fig. S4. Only individuals from the MRKAd5 and DNA/rAd5 trials with a previous positive NAE or AE CD4⁺ T-cell response (n = 31 and n = 24, respectively) were tested. Overall, we observed low-frequency IFN-γ production in the CD154⁺ cells (Fig. 3A). Despite these low frequencies of response, we did observe more CD154⁺ IFN-γ⁺ CD4⁺ T cells in response to NAE (Fig. 3B, P = 0.003), supporting our earlier findings from CD8-depleted IFN-γ ELISpot assays. We observed no differences in the magnitudes of TNF-α or IL-2 production (Fig. 3B). We also found very little evidence of either IL-4/IL-13 or IL-21 production in any samples besides the positive controls.

FIG 3.

HLA-II-associated HIV-1 adaptation impacts CD4⁺ T-cell cytokine production. Detection of cell-associated and soluble cytokine levels from ICS assay and ELISA is shown. (A) Representative example of CD154⁺ IFN-γ⁺ responses with ICS assay. (B) Overall CD4⁺ T-cell production of IFN-γ, TNF-α, and IL-2 with CD154 expression in response to NAE and AE stimulation as determined by ICS. MRKAd5 and DNA/rAd5 samples with previously positive NAE (n = 31) or AE (n = 24) responses were tested. (C) Cytokine production of IFN-γ, TNF-α, and IL-2 in response to NAE and AE stimulation detected by multiplex ELISA. A subset of positive NAE (n = 12) and AE (n = 11) responses were tested. P values were determined by the unpaired Wilcoxon rank sum test.

Due to the low frequencies of cytokine detection with the ICS assay and to capture any other cytokines/effector/helper molecules produced, we next utilized a multiplex enzyme-linked immunosorbent assay (ELISA). Supernatants were collected following AIM assay stimulation with either NAE or AE in a small subset of samples; only supernatants from samples with a positive AIM response were included (NAE, n = 12, and AE, n = 11). We detected an increase in IFN-γ production in samples stimulated with NAE, supporting our ICS and ELISpot results (Fig. 3C, P = 0.007). In addition, there were increased levels of TNF-α and IL-2 in response to NAE, albeit to a lesser extent than for IFN-γ (Fig. 3C, P = 0.04 and P = 0.06, respectively). We did not observe an increase in any other CD4-related cytokines. Taken together, these data show that HIV-1 vaccine recipients elicit Th1 cytokines in response to NAE stimulation.

Transcriptomic analyses show that HLA-II-associated adaptation leads to an altered CD4⁺ T-cell phenotype.

To gain comprehensive insights into the transcriptional landscape of these CD4⁺ T-cell responses, we performed single-cell RNA sequencing to fully capture the heterogeneity of NAE-specific and AE-specific CD4⁺ T-cell responses. These analyses were performed on samples from 4 vaccine recipients (2 from MRKAd5 and 2 from DNA/rAd5) who had both an NAE and an AE response, as shown by the data in Fig. 1. Single cells with OX40⁺ PDL1⁺ coexpression after stimulation with both NAE and AE were sorted into 96-well plates before undergoing PCR and cDNA conversion. In total, 518 cells were analyzed (293 from the NAE condition and 225 from the AE condition), with a total of 7,688 genes detected. When examining the full transcriptional profile, we observed differences between NAE-specific and AE-specific CD4⁺ T cells as seen in the distinct clusters obtained with uniform manifold approximation and projection for dimension reduction (UMAP) (Fig. 4A). In total, we found that 36 genes were differentially upregulated after NAE stimulation, with only 14 genes upregulated in response to AE. A list of all differentially expressed genes is given in Table S3. In support of our previous findings from our IFN-γ ELISpot assay, we observed that NAE-specific CD4⁺ T cells had an increased fold change of IFNG transcripts, but not to a level that was statistically significant (Fig. 4B). We did, however, observe upregulation of the interferon-related genes IRF3 and IFI6 (Fig. 4B and C, P_adj = 0.017 and P_adj = 0.046, respectively). Interestingly, AE-specific CD4⁺ T cells exhibited a demonstrable differential upregulation of CCR4 and IL-13 (Fig. 4B and D, P_adj = 0.016 and P_adj = 0.001, respectively), both of which are considered Th2-associated genes.

FIG 4.

HLA-II-associated HIV-1 adaptation impacts immune gene expression of CD4⁺ T-cell responses at single-cell level. (A) UMAP showing different clusters of NAE-stimulated (n = 293) and AE-stimulated CD4⁺ T cells (n = 225) across 4 vaccine recipients. (B) Volcano plot showing differentially expressed genes in response to NAE and AE. (C) Upregulation of interferon-related IRF3 and IFI6 transcripts in response to NAE. (D) Upregulation of Th2-related CCR4 and IL-13 transcripts in response to AE. Statistical analysis performed using MAST.

HLA-II-associated adaptation impacts Env-specific antibody production and cTfh phenotype of antigen-specific CD4⁺ T cells.

A previous study showed a correlation between antigen-specific CD4⁺ T-cell responses and antibody (Ab) production in the setting of HIV-1 vaccines (9). To this end, we wanted to investigate whether NAE-specific CD4⁺ T cells correlated with Env-specific antibody production. To do this, we focused our investigation on individuals from the DNA/rAd5 vaccine trial that were known to produce Env-specific antibodies (25). We restricted the analyses to only those individuals with a positive NAE or AE response by ELISpot or AIM as shown by the data in Fig. 1 (NAE, n = 15, and AE, n = 13). We used the IgG Env Ab score, which is a summary of antibody binding to all Env sites by ELISA (25). We found that the AIM response magnitude (net OX40/PDL1 coexpression) in response to NAE correlated strongly with Env-specific antibody production (Fig. 5A, P = 0.008 and R = 0.65). In contrast, the magnitudes of AE-specific responses did not correlate with Env-specific Ab levels (Fig. 5A, P = 0.59 and R = 0.16). To investigate this further, we next analyzed whether DNA/rAd5 NAE-specific and AE-specific CD4⁺ T-cell responses exhibited a circulating T follicular helper (cTfh) phenotype by measuring the OX40⁺ PDL1⁺ coexpression on CXCR5⁺ CD4⁺ T cells, as our group has done previously (a representative example is shown in Fig. 5B) (20). Although we did not observe a significant difference between the two conditions by cTfh response magnitude, there was a trend toward cTfh upregulation of OX40 and PDL1 in response to NAE, with 87% (13/15) of NAE-specific CD4⁺ responses having a matching cTfh response, while only 53% (7/13) of AE-specific responses had cTfh responses (Fig. 5C, P = 0.09). There was no evidence that the cTfh response magnitude itself correlated with Env-specific IgG.

FIG 5.

HLA-II-associated HIV-1 adaptation impacts generation of Env-specific antibodies and cTfh phenotype. (A) Correlations between magnitudes of AIM responses to NAE/AE and Env-specific IgG scores, as previously described (20), in DNA/rAd5 vaccine recipients; R and P values were determined by Spearman’s rank correlation test. (B) Representative gating strategy of cTfh AIM responses. SSC, side scatter. (C) cTfh response magnitudes and responder frequencies in DNA/rAd5 vaccine responses; P values were determined by the unpaired Wilcoxon rank sum test (left) and Fisher’s exact test (right). ns, not significant. Only DNA/rAd5 samples with positive NAE (n = 15) or AE (n = 13) responses as shown by the results in Fig. 1 were included in these analyses.

DISCUSSION

Although our group previously showed that HLA-II-associated HIV-1 adaptation negatively impacts CD4⁺ T-cell responses in HIV-1 infection (16), this study provides the first evidence that predicted adaptation dampens CD4⁺ T-cell responses following HIV-1 vaccination. We show that HLA-II-associated AE are less immunogenic than their NAE counterparts. Although most of these assays were done at the pool level, CD4⁺ T cells from two participants (one from the multiclade DNA/rAd5 trial and the other from the mosaic DNA/MVA trial) preferentially responded to NAE at the single-epitope level when vaccinated with both forms of the same epitope.

Our finding that only the NAE-specific CD4⁺ T-cell response magnitude correlated with Env-specific IgG antibody production suggests that AE-specific CD4⁺ T cells do not contribute to effective antibody production. In part, this result can be explained by a trend toward increased frequency of cTfh responses in NAE-specific cells, although confirmation of this in a larger number of recipients is needed. Taken together, these findings suggest that NAE-specific CD4⁺ T cells provide support to B cells, allowing improved antibody production. Recent studies have shown the promise of bNAbs leading to protection against sensitive HIV-1 strains (7). Utilization of sequential immunogen strategies has been shown to enrich precursor B cells capable of generating HIV-specific bNAbs (26), although further strategies are needed to form bNAb-producing B cells from these promising B-cell precursors. While many other factors are needed to induce neutralizing antibodies, immunization with a fully nonadapted vaccine could generate optimized CD4⁺ T-cell responses that assist in the formation of these bNAb-producing B cells.

Investigation at the cytokine level suggested that NAE-specific CD4⁺ T cells produced low magnitudes of Th1-level cytokines. The presence of polyfunctional CD4⁺ T cells expressing IFN-γ, TNF-α, IL-2, IL-4, and CD154 has been found to decrease the risk of infection in a prior HIV-1 vaccine efficacy trial (27). ICS assay results showed that CD4⁺ T cells responding to NAE had an increased CD154⁺ IFN-γ⁺ phenotype. When utilizing a more sensitive ELISA, we detected increases in IFN-γ, TNF-α, and IL-2 in the supernatants of samples following stimulation with NAE, confirming the detection of a low-magnitude Th1 phenotype. Although the results from the ELISA appear to be more sensitive than those from the ICS, it cannot be determined whether these cytokines were produced by CD4⁺ T cells or CD8⁺ T cells. At the transcriptional level, we observed an increase in the interferon-related genes IRF3 and IFI6 in NAE-specific CD4⁺ T cells. While there was also an increase in the fold change of IFNG, our analyses were not adequately powered to detect a significant difference between the two conditions for this gene. In contrast, CD4⁺ T cells responding to AE had increased expression of the genes for the surface marker CCR4 and the cytokine IL-13, suggestive of a Th2 phenotype. While we were unable to detect the Th2 cytokines IL-4 or IL-13 in our ICS assays or ELISAs, these are difficult to measure at the protein level, and future investigations should aim to confirm this phenotype in AE-specific CD4⁺ T-cell responses.

Interestingly, CD4⁺ T-cell responses to AE also showed increased expression of ACOT8, encoding a transcription factor that has been shown to bind to the HIV-1 protein Nef and play a role in the transcription of viral proteins within host cells (28, 29). This may suggest that HLA-II viral adaptation promotes HIV-1 infection; our group has previously shown a similar phenomenon with vaccine-generated CD8⁺ T-cell responses and HLA-I adaptation, where viral adaptation increased dendritic cell maturation and increased transinfection in MRKAd5 vaccine recipients (30). However, more investigation is needed to determine whether HLA-II-associated adaptation may promote HIV-1 infection or whether it simply leads to lower immunogenicity of CD4⁺ T-cell responses as shown here.

Our overall findings also have implications for mosaic vaccine design strategies. Mosaic vaccines have been shown to lead to an increased breadth of T-cell responses (22, 23), which led to two recent efficacy trials. Results from the Imbokodo study (HVTN 705, ClinicalTrials.gov identifier NCT03060629) in sub-Saharan Africa were recently released, and the mosaic vaccine candidate tested was ultimately shown to not be effective at preventing HIV-1 infection. A trial of a second mosaic vaccine called Mosaico (HVTN 706, ClinicalTrials.gov identifier NCT03964415) is under way and currently pending efficacy results. We show here that the mosaic sequences encoding adapted epitopes were not immunogenic in two polyvalent vaccine recipients, in contrast to sequences encoding nonadapted epitopes. Future vaccine trials should confirm this impact of HLA-II-associated viral adaptation on T-cell immunogenicity when vaccine recipients are given mosaic immunogens encoding both adapted and nonadapted epitopes. Our findings here suggest that vaccination with fully nonadapted, mosaic sequences would provide the best framework for increased CD4⁺ T-cell response breadth and overall immunogenicity.

Our study has a few limitations. As highlighted earlier, samples from both MRKAd5 and DNA/rAd5 recipients were obtained 4 weeks following the final vaccination time point, while samples from DNA/MVA recipients were collected 2 weeks following the final DNA vaccination. Many studies have found that CD4⁺ T-cell responses, and cTfh responses in particular, peak approximately 1 to 2 weeks following the final vaccination (9, 31). Despite this, our findings that HLA-II adaptation impacts CD4⁺ T-cell responses, and potentially cTfh responses, despite not using samples from the most immunogenic time point, underscore the role that HLA-II-associated HIV-1 adaptation may have in vaccine immunogenicity. While we used relatively low exogenous peptide concentrations, peptide pulse experiments may not completely replicate in vivo results. Finally, analysis of functional antibody responses, such as neutralizing capability, was not performed for many of the samples used in this project, and the impact of HLA-II-associated adaptation on functional antibody production remains an important unanswered question.

In conclusion, our findings demonstrate that HLA-II-associated AE are included in HIV-1 vaccine immunogens and elicit weakened CD4⁺ T-cell responses. Adapted epitopes continue to be included in current HIV-1 vaccines, resulting in decreased CD4⁺ T-cell immunogenicity that could be negatively affecting both HIV-specific antibody production and efficacious vaccine efforts. Future studies may need to determine if CD4⁺ T-cell responses to AE or NAE have any impact on infection risk and investigate whether immunogenicity and clinical outcomes are improved following immunization with a fully nonadapted HIV-1 vaccine.

MATERIALS AND METHODS

Samples.

Vaccine recipient peripheral blood mononuclear cell (PBMC) samples from three different vaccine trials, HVTN 502 (MRKAd5; ClinicalTrials.gov identifier NCT00095576), HVTN 505 (DNA/rAd5; ClinicalTrials.gov identifier NCT00865566), and HVTN 106 (DNA/MVA; ClinicalTrials.gov identifier NCT02296541), were obtained through the HIV Vaccine Trials Network (HVTN). Specifically, MRKAd5 recipients (n = 20) received three Ad5-based vaccines with HIV-1 gene inserts (gag, pol, and nef); the samples we obtained were collected at week 30 (4 weeks following the third/final vaccination). DNA/rAd5 recipients (n = 20) were primed with 3 DNA vaccines before receiving an rAd5 vaccine with HIV-1 gene inserts (gag, pol, env A/B/C, and nef [DNA only]); the samples we obtained were collected at week 28 (4 weeks following the fourth/final vaccination). Participants in this trial were randomized into one of three arms that each received unique DNA Env immunogens: (i) a natural vaccine (Nat-B), encoding a singular clade B HIV-1 strain; (ii) a consensus vaccine (CON-S), encoding the single amino acid which appears most frequently in HIV-1; and (iii) a mosaic vaccine (Mos), encoding three different strains of HIV-1, designed to account for the viral diversity of different HIV-1 strains. Importantly, the samples we obtained were collected at visit 7, 2 weeks following the third DNA prime vaccination time point. Placebo vaccine recipients from all three vaccine trials (MRKAd5, n = 4; DNA/rAd5, n = 4; and DNA/MVA, n = 10) were included, with immunogenicity results discussed below. All vaccine recipients had given informed consent for the HVTN studies.

Identification of HLA-II-associated polymorphisms in HIV-1 Env.

HIV-1 Gag, Pol, and Nef polymorphisms associated with specific HLA-II alleles have been previously described by our group (16). A similar strategy was utilized to identify HLA-II-associated polymorphisms in the HIV-1 Env protein. In short, using a large cohort of clade B and clade C chronically infected, antiretroviral-naive individuals (n = 350), HIV-1 polymorphisms that correlated with specific HLA-II alleles at the population level were identified. Utilizing the PhyloD software program (32), logistical regression was performed to determine which polymorphisms met our false discovery rate threshold (q > 0.2) when accounting for other variables, including verification that our identified adaptations did not overlap known HLA-I-associated polymorphisms. This led to the identification of 18 novel polymorphisms within Env. These Env polymorphisms are listed in Table S1. Including the 29 original predictions in Gag, Pol, and Nef, this study investigated a total of 47 HLA-II-associated HIV-1 polymorphic sites.

Peptide synthesis and peptide pool design.

Peptides (20 amino acids in length) were designed based upon optimal HLA-II allele binding predictions from NetMHCII (17) and the clade B consensus sequence. Each peptide contained a predicted HLA-II-associated adaptation site with a clade B consensus backbone. AE and NAE peptides were designed with all AE, including the HLA-II-associated polymorphism, and all NAE showing no evidence of escape or adaptation. Analysis of the binding affinities of these 47 NAE and their matching AE showed no significant differences (Fig. S1A). Predicted binding scores were calculated by the following equation: predicted score = {1 − log[50,000(affinity)]}. We next examined HIV-1 vaccine immunogen sequences from each of the three vaccine trials discussed above. Vaccine-matched variant sequences that encompassed the predicted HLA-II polymorphic sites were synthesized by New England Peptides (www.newenglandpeptide.com) in a 96-well array format. Using the HLA-II-associated polymorphisms as described above, we classified our designed peptides into two groups: (i) epitopes containing the amino acid that matched our HLA-II-associated HIV-1 polymorphisms were termed adapted epitopes (AE), and (ii) epitopes that did not contain any evidence of HLA-II-associated adaptation were termed nonadapted epitopes (NAE). Due to limited sample availability, all designed epitopes were then combined into either an AE pool or an NAE pool. A representative example showing the identification process of AE and NAE within the DNA/rAd5 vaccine-based Gag immunogen is shown in Fig. S1B. In this example, two adapted epitopes containing an HLA-II-associated polymorphism were identified at Gag112 and Gag247, while two nonadapted epitopes were identified at Gag147 and Gag339.

CD8-depleted IFN-γ ELISpot assay.

The ELISpot assay protocol was similar to that described previously by our group and others (16, 33). In brief, vaccine recipient PBMCs were thawed and rested overnight at 37°C in RPMI medium supplemented with 10% human AB serum (R10 medium). ELISpot plates were coated with anti-IFN-γ antibody overnight and then blocked for 2 h with R10 medium at 37°C the following day. PBMC samples underwent CD8 depletion (Dynabeads CD8; Invitrogen) to enrich for CD4⁺ T-cells. CD8-depleted PBMCs were then added to the ELISpot plate at 100,000 cells/well in duplicate and were stimulated with peptide pools at 2 μM for 40 h at 37°C. Negative and positive controls of unstimulated conditions and phytohemagglutinin treatment, respectively, were included for each sample. The plates were then washed, removing all cells, and developed by adding biotinylated anti-IFN-γ antibody (2 h), streptavidin (45 min), and 5-bromo-4-chloro-3-indolyl phosphate (BCIP)/nitroblue tetrazolium (NBT) substrate solution (10 min) sequentially. Washes between each staining step used phosphate-buffered saline (PBS) with 0.01% Tween 20 by volume. ELISpot plate spots were counted and analyzed using the ImmunoSpot analyzer and software (version 5.0; Cell Technology Limited), and the results were normalized to mean spot-forming units per 1 million cells (SFU/10⁶ cells). In agreement with previous publications investigating T-cell responses in HIV-1 vaccine samples, the positivity criteria were responses greater than or equal to 50 SFU/10⁶ cells in magnitude and at least three times higher than the unstimulated negative control as done by previous groups (34–38). Net SFU values were used in all analyses and were calculated by subtracting each response from the value for the respective unstimulated negative control. Placebo recipients did not elicit any positive immune responses to pooled AE or NAE based on ELISpot.

Flow cytometric-based AIM assay.

Flow cytometric assays were performed to analyze the activation status and determine the phenotype of CD4⁺ T cells by activation-induced marker (AIM) assay as described by other groups (18–21). Frozen PBMC samples were thawed and rested in R10 medium at 37°C for approximately 3 h. Between 500,000 and 1 million cells were then aliquoted into flow cytometry tubes (Falcon), and the respective antigen stimulations (NAE pool or AE pool) were added at 2 μM per peptide. Unstimulated (with equimolar dimethyl sulfoxide [DMSO] added) and superantigen Staphylococcus enterotoxin B (SEB)-stimulated (0.25 μg/mL) tubes were used for negative and positive controls, respectively, with each sample. Anti-CD28 and anti-CD49d antibodies were added for optimal T-cell responses. After 18 h of stimulation, supernatants from certain samples were collected, centrifuged to remove any residual cells, and frozen at −80°C for use in multiplex ELISAs (see below). Cells were stained with fluorescent antibodies as detailed in panel 1 of Table S2. Staining was done for 30 min at 4°C, except for staining with CCR7 (peridinin chlorophyll protein [PerCP]-Cy5.5), which was completed for 20 min at 37°C. Cells were fixed in 5% formalin solution immediately after staining and analyzed within 24 h on an LSR II or FACSymphony flow cytometer (BD Biosciences). Samples were then analyzed using FlowJo software (version 10). Figure S2 shows the overall AIM gating strategy. Positivity criteria for NAE- and AE-stimulated conditions were set at two hierarchical thresholds as done previously by others (39): definite positive responses exhibited an OX40⁺ PDL1⁺ frequency at least three times higher than under unstimulated conditions and a Fisher’s exact P value of <0.0001, while responses that were only two times higher than under unstimulated conditions and with a Fisher’s exact P value of <0.01 were deemed positive with uncertainty. A few placebo AIM responses met our positivity criteria (1 definite positive response and 2 positive with uncertainty). Net values (background subtracted) were used in all analyses.

Intracellular cytokine staining (ICS) assay.

PBMC samples from HIV-1 vaccine recipients were thawed and rested in a manner similar to that described above for the AIM assay. Samples were then stimulated with their respective antigens, as well as a cocktail of CD154-BV785 (BD Biosciences), anti-CD28, and anti-CD49d antibodies. After 1 h of stimulation, GolgiStop containing monensin was added at 0.5 μL per tube and the mixture stimulated for an additional 12 h. GolgiPlug containing brefeldin A was not used per previous protocols emphasizing decreased CD154 staining (40). Cells were then stained with surface markers for 30 min at 4°C. Cells were then permeabilized for 20 min with the Cytofix/Cytoperm buffer (BD Biosciences) before staining with intracellular cytokine markers. For all details on flow cytometry panels, see Table S2. Samples were then fixed in 5% formalin solution and analyzed within 24 h on a FACSymphony flow cytometer (BD Biosciences).

Multiplex ELISA.

Supernatants collected from the AIM assay were quantitatively analyzed for multiple human cytokines. In brief, 200 μL of each sample was analyzed based upon the Quantibody human cytokine array 4000 kit (RayBiotech, Peachtree Corners, GA, USA) to probe for 200 human proteins. This kit utilizes a primary protein-specific antibody and a secondary biotin-labeled detection antibody, allowing quantification of the cytokine-antibody-biotin complexes. Values below the limit of detection and above the maximum detection level were normalized to these values. Net values were used in all analyses and were obtained by subtracting the protein concentrations under the NAE and AE conditions from the respective protein concentration under the unstimulated condition in each vaccine recipient.

Single-cell-based RNA-sequencing assay.

PBMC samples were thawed and stimulated in a manner similar to the AIM assay protocol as described above with a simplified staining protocol as detailed in panel 2 of Table S2. Unfixed samples were analyzed, and CD3⁺ CD4⁺ OX40⁺ PDL1⁺ single cells were sorted into each well of a 96-well plate using a FACSAria flow cytometer (BD Biosciences). Each well of the 96-well plate was preloaded with 3 μL of lysis buffer (5 U/μL RNAseOUT, 25 μM deoxynucleoside triphosphates [dNTPs], and 0.01% Triton X-100 by volume) and kept on ice. After sorting, cells were immediately frozen at −70°C until sequencing. Single-cell RNA sequencing was performed at the Institute for Immunology and Infectious Diseases in Perth, Australia, as previously described (41–43). In short, oligo(dT)-primed reverse transcription was done on a single-cell level, utilizing uniquely tagged primers and a preamplification step to increase the yield and transcript length of the single-cell cDNA library. Next, cDNA libraries were analyzed by calculating gene-specific read counts using HTSeq-count. Quality control steps removed cells with fewer than 200 genes and more than 5% mitochondrial DNA, while genes were excluded that did not have positive counts in at least three cells. Normalization and scaling of the remaining genes were performed using the R package Seurat (version 4.1) (44). Differential expression analysis was also done using Seurat, with significance determined by using a fold change of 0.2 and an adjusted P value of <0.05 as determined by the model-based analysis of single-cell transcriptomics (MAST) statistical test.

Statistical analysis.

All statistical analyses and figure creation were done in R with supporting R packages. Statistical significance when testing for immunogenicity between NAE and AE conditions across all 40 donors was determined by using the paired Wilcoxon signed-rank test (Fig. 1A and C). When testing for differences in response frequencies, Fisher’s exact test was utilized (Fig. 1D and 5C). The unpaired Wilcoxon rank sum test was utilized to determine phenotypic and functional differences in samples that exhibited positive responses (Fig. 3B and C and Fig. 5C). Differential gene expression significance was determined using Seurat as discussed above. All graphs were made with the R package ggplot2 (45), with help from ggrepel (version 0.9) and ggbreak (24).