A first-in-human germline-targeting HIV nanoparticle vaccine induced broad and publicly targeted helper T cell responses

Abstract

The engineered outer domain germline targeting version 8 (eOD-GT8) 60-mer nanoparticle was designed to prime VRC01-class HIV-specific B cells that would need to be matured, through additional heterologous immunizations, into B cells that are able to produce broadly neutralizing antibodies. CD4 T cell help will be critical for the development of such high-affinity neutralizing antibody responses. Thus, we assessed the induction and epitope specificities of the vaccine-specific T cells from the IAVI G001 phase 1 clinical trial that tested immunization with eOD-GT8 60-mer adjuvanted with AS01_B. Robust polyfunctional CD4 T cells specific for eOD-GT8 and the lumazine synthase (LumSyn) component of eOD-GT8 60-mer were induced after two vaccinations with either the 20- or 100-microgram dose. Antigen-specific CD4 T helper responses to eOD-GT8 and LumSyn were observed in 84 and 93% of vaccine recipients, respectively. CD4 helper T cell epitope “hotspots” preferentially targeted across participants were identified within both the eOD-GT8 and LumSyn proteins. CD4 T cell responses specific to one of these three LumSyn epitope hotspots were observed in 85% of vaccine recipients. Last, we found that induction of vaccine-specific peripheral CD4 T cells correlated with expansion of eOD-GT8–specific memory B cells. Our findings demonstrate strong human CD4 T cell responses to an HIV vaccine candidate priming immunogen and identify immunodominant CD4 T cell epitopes that might improve human immune responses either to heterologous boost immunogens after this prime vaccination or to other human vaccine immunogens.

INTRODUCTION

Although numerous modalities to prevent HIV infection have been developed and disseminated, an effective preventative vaccine is still needed to alleviate the HIV/AIDS epidemic in high-risk communities globally (1). HIV-specific broadly neutralizing antibodies (bnAbs) isolated from people living with HIV-1 have been shown to provide sterilizing protection in nonhuman primate (NHP) challenge models (2–5) and protection against circulating neutralization-sensitive HIV-1 isolates in clinical trials (6, 7). However, induction of HIV bnAbs by vaccination has been a substantial challenge for the field.

The engineered outer domain germline targeting version 8 (eOD-GT8) immunogen was designed from the HIV-1 gp120 protein to have high affinity specifically for the inferred germline versions of diverse CD4 binding site (CD4bs)–specific VRC01-class bnAbs (8–10). The IAVI G001 phase 1 clinical trial demonstrated that the eOD-GT8 60-mer nanoparticle with AS01_B adjuvant induced VRC01-class CD4bs-specific immunoglobulin G (IgG) B cells in 35 of 36 vaccine recipients (11). However, as anticipated, these B cells had limited B cell receptor mutations and no detectable neutralizing activity (11). Heterologous boost immunogens will be needed to selectively affinity mature these B cell responses toward broad and potent neutralization (12–16).

The eOD-GT8 60-mer is composed of the eOD connected by a 15–amino acid glycine-serine linker to a lumazine synthase (LumSyn) protein, which forms the self-assembling icosahedral nanoparticle backbone. Self-assembling nanoparticles are a relatively new category of vaccine modalities and provide multimeric antigen display, improved in vivo trafficking, and enhanced immunogenicity compared with recombinant subunit vaccines, as demonstrated in preclinical studies for HIV and severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) (9, 17–21). Conceptually similar self-assembling virus-like particle vaccines have been licensed for prevention of hepatitis B virus (HBV) (22) and human papillomavirus (HPV) (23). These licensed efficacious antiviral vaccines induce potent antibody responses that correlate with vaccine-specific CD4 T cell help (24, 25). In addition, the use of the potent AS01_B-adjuvant has been shown to further enhance immunogenicity in the AS01_B-adjuvanted HBV vaccine, with increased antibody and CD4 T cell responses compared with alternative adjuvants (26, 27). Together, these data suggest that multimeric nanoparticle design combined with a potent adjuvant can enhance immunogenicity of antiviral subunit vaccines.

CD4 T cells regulate the survival, expansion, and maturation of germinal center (GC) B cells and are thereby critical for achieving high-affinity neutralizing antibodies (28, 29). In rhesus macaques, vaccine-induced autologous tier 2 HIV-1 neutralizing antibody responses were associated with the presence of envelope (Env)–specific circulating CD4 T cells with up-regulated T follicular helper (Tfh) cell–associated genes (30). The eOD-GT8 60-mer has also been shown to induce robust CD4 T cell responses in rhesus macaques (31); however, it is still unclear how this will translate to humans. Vaccine-induced Env-specific T cell responses in humans tend to be limited in breadth and variant recognition, with a median of one to four T cell epitopes detected per participant (32–35). Therefore, we assessed the ability of the eOD-GT8 60-mer to induce broad T helper cell activity across diverse human leukocyte antigen (HLA) backgrounds in the IAVI G001 clinical trial.

Whereas CD4 T cells have the potential to enhance the antibody response, CD8 T cells have direct antiviral cytotoxic activity that can be complementary to neutralizing antibodies for achieving vaccine-mediated protection. For instance, Gag-specific CD8 T cell responses have been shown to extend the durability and reduce the neutralization titer required for vaccine-mediated protection in NHPs (36). In addition to eliciting bnAbs, an ideal HIV-1 vaccine will likely need to elicit CD8 T cell responses against conserved HIV-1 epitopes to allow for control of any breakthrough virions (37, 38). Whereas adjuvanted protein vaccines are generally very poor at inducing CD8 T cell responses, nanoparticle vaccines seem to have potential to access cross-presentation pathways and induce antigen-specific CD8 T cells (39). However, the rules for the induction of CD8 T cells using nanoparticles are not well understood.

Here, we detail the magnitude, phenotype, and immunodominant epitopes of peripheral blood vaccine-specific CD4 and CD8 T cells and frequency of GC Tfh cells after vaccination with the eOD-GT8 60-mer/AS01_B nanoparticle in healthy adults. These analyses can help inform rational vaccine strategies for reliably eliciting CD4 T helper responses to support humoral immunity both in the context of germline-targeting HIV-1 vaccines and more broadly for other human vaccine regimens.

RESULTS

Vaccine-specific CD4 T cells were induced in almost all vaccine recipients

Vaccines at either of two doses, 20 or 100 μg, were administered at weeks 0 and 8. We measured vaccine-specific peripheral blood T cell responses 2 weeks after the second vaccination (Fig. 1A). To enable evaluation of T cell responses, we synthesized 15-mer peptides overlapping by 11 amino acids across the length of the eOD-GT8 and LumSyn amino acid sequence and pooled the peptides by protein. T cell responses were evaluated using a validated intracellular cytokine staining (ICS) assay, in which antigen-specific responses were assessed using up-regulation of cytokines and other functional markers after peptide stimulation. The specificity of the ICS assay and peptide pools was confirmed by testing 40 HIV-1–negative unvaccinated control samples (fig. S1).

Fig. 1. Vaccine-specific CD4 and CD8 T cell responses target eOD-GT8 and LumSyn.

(A) Shown is a schematic of the T cell analyses in the IAVI G001 clinical trial. Cryopreserved PBMCs obtained by leukapheresis 2 weeks after second immunization (week 10) were analyzed by ICS assay. Tfh cells in LN fine needle aspirates (FNA) 3 weeks after first and second vaccination were measured by flow cytometry (weeks 3 and 11, respectively; fig. S2). (B) Boxplots show the interquartile range of background-adjusted frequencies of cytokine-positive eOD-GT8 and LumSyn-specific CD4 T cells at week 10. Statistically positive responses were determined by MIMOSA and are indicated by colored closed circles. Nonresponders are indicated by open gray triangles. The number of positive responders of the participants tested and the percentage of positive responders are indicated above each graph. (C) Frequencies of vaccine-specific cTfh-like cells were measured in the PBMCs at week 10 as the background-adjusted percentage of IL-2⁺ or CD40L⁺ CXCR5⁺ CD4 T cells of total circulating CD4 T cells after peptide stimulation. No positivity call was applied to the vaccine-specific cTfh cells. Individual data points are indicated as colored open circles by dose group. (D) Shown are background-adjusted frequencies of cytokine-positive CD8 T cells in response to eOD-GT8 and LumSyn in the PBMC samples collected at week 10. Positive responses (by MIMOSA) are indicated by colored closed circles, and nonresponders are indicated by open gray triangles. Response rates are indicated above each graph. Significant differences between the placebo and 20- and 100-μg dose groups were determined by Barnard’s test (two-sided, α = 0.05).

At 2 weeks after second vaccination, both vaccine dose groups had significantly higher frequencies of eOD-GT8– and LumSyn-specific CD4 T cells expressing the combined expression of interferon-γ (IFN-γ) or interleukin-2 (IL-2) compared with the placebo recipients (P < 0.0001 for low or high dose compared with placebo; Fig. 1B). The frequencies of eOD-GT8– and LumSyn-specific CD4 T cells expressing any of the individual cytokines: IFN-γ (P = 0.0002), IL-2 (P < 0.0001), tumor necrosis factor–α (TNF-α) (P < 0.0001), or CD40L (P < 0.0001) were also higher compared with the placebo recipients (Fig. 1B). CD4 T cells producing IFN-γ, IL-2, or TNF-α in response to eOD-GT8 or LumSyn were not detected among any placebo recipients. Vaccine-induced eOD-GT8–specific CD4 T cells expressing the combined IFN-γ or IL-2 were detected in 88 and 81% of vaccine recipients, with median frequencies of 0.26 and 0.21% of CD4 T cells across all participants in the 20- and 100-μg dose groups, respectively (Fig. 1B). Compared with recent clinical trials evaluating adenovirus or canarypox viral vector HIV vaccines expressing either the full gp120 or gp140 HIV Env sequence with the same assay, Env-specific CD4 T cell responses were detected in higher proportion of vaccine recipients (>81% versus 40 to 60% among comparators) and at a similar magnitude (>0.2% versus about 0.2% among positive responders) (33, 40–42). In addition, 94 and 88% of eOD-GT8 60-mer vaccine recipients also had a significant [mixture models for single-cell assays (MIMOSA) probability > 0.999] frequency of CD4 T cells that produced IFN-γ or IL-2 singly or in combination in response to LumSyn, with a median of 0.26 and 0.3% of CD4 T cells responding in the 20- and 100-μg dose groups, respectively (Fig. 1B). Thus, the combined vaccine-specific CD4 T cell response was about 0.5% of the total circulating CD4 T cells in most vaccine recipients. There were no differences in the magnitudes or response rates of eOD-GT8– or LumSyn-specific CD4 T cell responses observed between the low (20 μg) and higher (100 μg) dose vaccine groups.

Not only were vaccine-specific CD4 T cell responses prevalent and of high magnitude, but they were also highly polyfunctional. We measured similar frequencies of CD40L (CD154), IL-2, TNF-α, and, to a lesser extent, IFN-γ, expressed among the eOD-GT8– responding CD4 T cells, with no detection of IL-4 or IL-17a expression. Although the LumSyn-specific CD4 T cells were also highly polyfunctional, they had slightly higher IFN-γ expression (Fig. 1B). We also determined the frequency of circulating Tfh (cTfh) cell–like CD4 T cells that express CXCR5 that were vaccine specific, indicated by expression of CD40L⁺ or IL-2⁺ after stimulation with eOD-GT8 or LumSyn (Fig. 1C). The median frequencies of vaccine-specific CD40L⁺ cTfh cells of CD4 T cells were low, 0.0292 and 0.0229% in the 20- and 100-μg groups, respectively, but still significantly expanded when compared with the placebo recipients (median of 0.0006% of CD4 T cells; P < 0.001; Fig. 1C). However, the vaccine-specific expansion of cTfh cells was even more pronounced when considering the proportion of cTfh cells that were eOD-GT8 and LumSyn specific compared with placebo (fig. S2). The expansion of vaccine-specific cTfh cells was comparable to previous reports assessing vaccine-induced cTfh cells after one of the SARS-CoV-2 vaccines that received emergency use authorization (43). However, differences in assay methodology made it difficult to make quantitative comparisons to the SARS-CoV-2 vaccine study.

Lymph node GC T follicular helper cells increased after vaccination compared with placebo

We monitored GC activity 3 weeks after each vaccination by evaluating mononuclear cells (MNCs) from the fine needle aspirates (FNAs) of axillary lymph nodes (LNs) ipsilateral to the site of immunization as previously described (11). GC Tfh cells were identified by the very high coexpression of the phenotypic markers CXCR5 and programmed cell death protein 1 (PD-1) on CD4 T cells in the fresh LN MNCs directly ex vivo, without any exogenous stimulation. Because we only evaluated the total GC Tfh cells, we were unable to quantify the subset of activated GC Tfh cells that were vaccine specific. However, the percentage of GC Tfh cells of total LN CD4 T cells was higher after vaccination compared with the placebo recipients. After a single vaccination, the frequency of GC Tfh cells was expanded about threefold among the vaccine recipients compared with placebos (2.136% in 20-μg dose group and 3.063% in 100-μg dose group versus 0.789% in placebos; fig. S3).

Three weeks after the second vaccination, the difference in the frequency of GC Tfh cells among the vaccine and placebo recipients was even more pronounced (4.226% in 20-μg dose group and 5.669% in 100-μg dose group versus 0.108% in placebos; fig. S3). Thus, at either dose, the vaccine induced GC activity, as indicated by sustained expansion of GC Tfh cells detected up to 3 weeks after the boost immunizations.

LumSyn-specific CD8 T cells were detected among about half of vaccine but not placebo recipients

Although the eOD-GT8 vaccine was designed to induce humoral immunity, we evaluated the induction of CD8 T cells to gain information about protein nanoparticle induction of such responses, because cytotoxic T cells can potentially contribute to vaccine-induced protection. No eOD-GT8–specific CD8 T cell responses were induced, consistent with other clinical studies of adjuvanted recombinant protein vaccines (Fig. 1D) (44–47). In contrast, we observed induction of LumSyn-specific CD8 T cells among vaccine, but not placebo, recipients. Among the 20- and 100-μg dose groups, 52.9 and 56.2% of vaccine recipients had CD8 T cells that produced IFN-γ, IL-2, or both in response to LumSyn peptides, with median frequencies of 0.157 and of 0.172% of CD8 T cells, respectively. The LumSyn-specific CD8 T cells were highly polyfunctional, as indicated by similar frequencies of IFN-γ–, IL-2–, and TNF-αγexpressing CD8 T cells. The lack of CD8 T cell responses among the placebo groups and in the original validation of the peptide pools confirmed that these responses were specifically induced by the vaccine (fig. S1).

Vaccine-specific CD4 T cells were polyfunctional and had diverse phenotypes

To further characterize the detailed phenotypic profiles of the vaccine-specific CD4 T cell responses, we used K-means clustering to simultaneously evaluate the expression of all of the functional and phenotypic markers in the flow cytometry panel. To be as inclusive as possible of antigen-specific CD4 T cells while maintaining specificity, we included all CD4 T cells that expressed any combination of at least two cytokines or activation markers. Using these criteria, we included all reactive CD4 T cells from all of the stimulation conditions, including control stimulations, vaccine-matched peptide pools, and cytomegalovirus (CMV) peptides. Ten clusters were then defined (Fig. 2A and table S1), and these clusters were quantified as frequencies of CD4 T cells per sample and stimulation (Fig. 2B).

Fig. 2. Antigen-specific CD4 T cells have diverse phenotypes by K-means clustering analysis.

K-means clustering was conducted on CD4 T cells that were positive for at least two cytokines or activation markers in the eOD-GT8, LumSyn, negative control, and CMV stimulations by ICS assay. (A) Heatmap indicates proportion of cells that were positive for each marker within the specified cluster. GzB, granzyme B. (B) Shown are the background-adjusted percentages of positive CD4 T cells for each cluster of the total CD4 T cells stimulated with CMV pp65 (top), eOD-GT8 (middle), or LumSyn (bottom). Box plots indicate the median and interquartile range.

Several distinct clusters were the most prevalent among the vaccine-specific CD4 T cell responses, including clusters 2, 5, 7, 9, 10, and, to a lesser extent, 8 (Fig. 2B and fig. S4). These clusters represented polyfunctional T cells expressing all four of the key functional markers (CD40L, IL-2, TNF-α, and IFN-γ; cluster 5), three such markers (clusters 2, 8, and 10), or two markers (clusters 7 and 9). The cells were either central memory (CM) or effector memory, with some clusters predominantly one or the other. Although not visualized well on the heatmap because of the low frequency of these cells, cluster 10 included the inducible T cell costimulator (ICOS)⁺ CXCR5⁺ CM CD4 T cells, which would include the cTfh cell responses. In total, these multiple clusters likely represent T cells with distinct functions, indicating that the vaccine induced a broad array of CD4 cell types.

Overall, the LumSyn-specific CD4 T cells were quite similar in phenotype to the eOD-GT8–specific CD4 T cells, with the exception that cluster 1 was also enriched among the LumSyn-specific CD4 T cells and not among the eOD-GT8–specific CD4 T cells (Fig. 2B and fig. S4). Cluster 1 was T helper 1–biased with high expression of IFN-γ, highly polyfunctional with coexpression of CD40L, IL-2, and TNF-α, and increased cytolytic potential, as indicated by increased granzyme B expression. Nonvaccine-induced CMV-specific CD4 T cells had less diverse phenotypes and were dominated by IFN-γ and granzyme B expression (clusters 1 and 4) as well as a more effector phenotype consistent with a T cell response induced by chronic infection in contrast to vaccination.

LumSyn-specific CD8 T cells were highly polyfunctional and had a predominantly effector memory phenotype

We also used K-means clustering to characterize the phenotypes of the antigen-specific CD8 T cell responses. Although we initially identified 10 clusters, four of the clusters had too few cells for analysis (median < 10 cells per sample) and were therefore excluded, leaving six antigen-specific CD8 T cell clusters (fig. S5A). The LumSyn-specific CD8 T cells were predominantly found in clusters 1, 2, 3, and 8 (figs. S5B and S6). Clusters 1, 2, 3, and 8 shared an effector memory phenotype. Cluster 1 was the most polyfunctional, with high expression of IFN-γ, IL-2, TNF-α, and, to a lesser extent, granzyme B. Cluster 2 expressed IFN-γ and TNF-α but lacked IL-2. Cluster 3 expressed IL-2 and TNF-α but lacked IFN-γ and granzyme B and included some CM T cells. Cluster 8 was mostly IFN-γ and granzyme B, with a lower frequency of IL-2 expression. In contrast, CMV-specific CD8 T cells were dominated by clusters 1 and 2, which were effector memory and polyfunctional with (cluster 1) or without (cluster 2) IL-2 expression. Thus, the LumSyn-specific CD8 T cells included more diversity in phenotypes, representing early rather than late differentiated cells. For both the vaccine-specific CD4 and CD8 T cells, the early time point after vaccination may contribute to the predominance of early differentiated cells. In addition, although the functional importance of the various clusters cannot be determined in this phase 1 study, this type of analysis can provide insights in later-phase testing such as for identifying immune correlates of protection.

CD4 T cell responses were driven by immunodominant epitopes with diverse and promiscuous HLA restriction

The high response rates to eOD-GT8 and LumSyn peptide pools were likely composed of a polyclonal T cell response targeting multiple epitopes. To determine whether there were any broadly immunogenic epitopes that should be retained in heterologous boost immunogens to follow this priming immunogen, we identified the specific epitope regions that were targeted. We started by measuring CD4 T cell responses to subpools (13 to 14 peptides each) covering the eOD-GT8 portion of the immunogen: eOD-GT8-1 subpool (165 to 235 amino acids), eOD-GT8-2 subpool (225 to 287 amino acids), and eOD-GT8-3 subpool (277 to 341 amino acids), where amino acid numbering is relative to the eOD-GT8–LumSyn linear sequence. We found that, although all three subpools were recognized by some vaccine recipients, the C-terminal subpool eOD-GT8-3 was targeted by 70% of vaccine recipients (Fig. 3A).

Fig. 3. Immunodominant T cell responses to eOD-GT8 and LumSyn identified by epitope mapping by ICS.

(A) CD4 T cell responses to eOD-GT8 and corresponding subpools (13 to 14 peptides each), eOD-GT8-1, eOD-GT8-2, and eOD-GT8-3, are shown. Frequencies of cytokine positive T cells are background-adjusted. The numbers of positive responders out of participants tested and percent of positive responders are indicated above each graph. Positive responses were determined by Fisher’s exact test and are indicated by colored closed circles. Negative responses are indicated by open gray triangles. Box plots indicate the median and interquartile range. Significant differences between the placebo and 20- and 100-μg dose groups were determined by Barnard’s test (two-sided, α = 0.05). ns, not significant. (B) Shown are individual peptide-specific CD4 and CD8 T cell responses to LumSyn (41 peptides; 1 to 175 amino acids) and eOD-GT8-3 by ICS assay. Amino acid numbering is in relative to the eOD-GT8–LumSyn linear sequence. Each row (y-axis tick) indicates a participant. Statistically significant responses are plotted using one horizontal line representing the amino acids covered by the given peptide. Nonresponders were included as blank rows. Because of the overlapping design of the peptides, recognition of a single epitope may be indicated by overlapping peptide responses. (C and D) Shown is the percentage of vaccine recipients, by dose group, with positive CD4 T cell responses to the eOD-GT8-3 (C) and LumSyn (D) individual peptides (numbered sequentially in order from N terminus to C terminus relative to the eOD-GT8–LumSyn linear sequence), hotspots, and combinations of hotspots.

For more granular resolution of immunodominant T helper responses, we mapped the responses to individual overlapping peptides from the eOD-GT8-3 subpool and LumSyn pool using the ICS assay (Fig. 3, B to D). In parallel, we conducted HLA genotyping to infer potential genetic restrictions influencing the T cell responses (table S2). We found two epitope “hotspots” across vaccine recipients that accounted for most of the CD4 T cell response to the eOD-GT8-3 subpool. The first eOD-GT8 hotspot (HXB2 Env numbering 338 to 356 amino acids), eOD-GT8-3 peptide 2 (WNNTLKQIASKLREQ), and eOD-GT8-3 peptide 3 (LKQIASKLREQYGNK) had a 36% combined response rate (Fig. 3, B and C). Responses to this first eOD-GT8 hotspot were enriched among participants with diverse HLA class II alleles: HLA-DQA1*01 (P = 0.03), HLA-DRB1*15 (P = 0.03), HLA-DRB5*01(P = 0.03), and HLA-DQB1*06 (P = 0.03) (table S3). HLA-DRB5*01 was predicted to bind strongly to the NNTLKQIASKLREQ peptide included in the first hotspot; the NetMHCIIpan-predicted affinity was in the top 4% of predicted binding affinities of 5 × 10⁶ randomly generated peptides (48, 49). The second eOD-GT8 hotspot (HXB2 Env numbering 379 to 397 amino acids), eOD-GT8-3 peptide 12 (GGEFFYCDSTQLFNS), and eOD-GT8-3 peptide 13 (FYCDSTQLFNSTWFN) had a 48% combined response rate (Fig. 3C). The responses to peptides 12 and 13 were enriched among participants with the HLA class II allele HLA-DPB1*04 (unadjusted P < 0.05), and the 15-mer peptide, GEFFYCDSTQLFNST, is within the top 5% of peptides predicted to bind DPB1*04, according to NetMHCIIpan (48, 49).

Given the high proportion of vaccine recipients with responses to the LumSyn master pool, we decided to map the entire length of LumSyn using the individual overlapping peptides in the ICS assay (Fig. 3, B and D). We also considered that identification of any broadly immunogenic CD4 epitopes might allow for their use as heterologous T helper epitopes in other human vaccines. We found three CD4 T cell hotspots within LumSyn: peptide 5 (IVASRFNHALVDRLV) and peptide 6 (RFNHALVDRLVEGAI); peptide 22 (ATPHFDYIASEVSKG) and peptide 23 (FDYIASEVSKGLADL); and peptide 28 (KPITFGVITADTLEQ) and peptide 29 (FGVITADTLEQAIER), with response rates of 32, 68, and 36%, respectively. A large majority of vaccine recipients (77.4%) responded to either LumSyn peptides 22 and 23 or peptides 28 and 29, and 85% responded to at least one of these three LumSyn hotspots (Fig. 3D). HLA-DQA1*01 was enriched among participants responding to LumSyn peptides 5 and 6 (unadjusted P = 0.02; table S3) and was a predicted binder of the peptide ASRFNHALVDRLVE (top 5.4% of predicted binders). HLA-DRB4*01 was enriched among participants responding to LumSyn peptides 22 and 23 (unadjusted P < 0.05; table S3). Multiple HLA alleles were enriched among responders to each of the LumSyn epitope hotspots without clear dominance of specific alleles, indicating that multiple HLA alleles may be able to present these epitopes (table S3).

CD8 T cell responses to LumSyn were driven by HLA-A*02–restricted immunodominant epitopes

Because LumSyn also elicited CD8 T cell responses, we determined the specific peptides with CD8 T cell reactivity using the ICS assay. There was a single immunodominant epitope within peptides 24 and 25 (ASEVSKGLADLSLELRKPI), which was also adjacent to the most immunodominant CD4 T cell epitope (Fig. 3B and D). The epitope appeared to be restricted by HLA-A*02, based on enrichment among responders (unadjusted P < 0.0001; adjusted P = 0.003; table S3). The GLADLSLEL 9-mer peptide was also among the top 0.02% of peptides predicted to bind HLA-A*02 (NetMHCpan) (48, 49).

B and T cell responses correlated within but not between LN and peripheral blood compartments

We investigated the relationship between vaccine-specific T and B cells within and between anatomical compartments. We compared the frequencies of vaccine-specific CD4 T and B cell responses from the peripheral blood at 2 weeks after second vaccination with the frequencies of GC Tfh cells, eOD-GT8⁺ GC B cells, and total GC B cells found in FNAs from the draining LNs at 3 weeks after second vaccination (Fig. 4). We found that patterns were very similar between the individual and combined dose groups, although statistical significance varied. In the periphery, vaccine-specific CD4 T cells were positively associated with vaccine-specific cTfh cells (combined dose groups: r = 0.49, P < 0.05) and eOD-GT8⁺ IgG memory B cells (combined dose groups: r = 0.39, P < 0.05). Similarly, within the LNs, GC Tfh cells correlated with GC B cells (combined dose groups: r = 0.69, P < 0.0001) and eOD-GT8–specific IgG GC B cells (combined dose groups: r = 0.71, P < 0.0001). However, between anatomical compartments, there was only one significant correlation, whereby GC Tfh cell frequencies were inversely associated with the magnitude of eOD-GT8⁺ IgG memory B cells (combined dose groups: r = −0.54, P < 0.01). Together, we concluded that vaccine-specific CD4 T cell responses were correlated with the induction of memory B cells. However, the relationship between anatomical compartments was not so simple and may require more qualitative or longitudinal analyses for further interpretation.

Fig. 4. Associations of the frequencies of T and B cell responses within and between peripheral and LN compartments.

Frequencies of eOD-GT8⁺ IgG memory B cells and vaccine-specific CD4 T and cTfh (CD40L⁺ CXCR5⁺) CD4 T cells were evaluated in PBMCs collected at 2 weeks after second vaccination. Frequencies of eOD-GT8⁺ IgG GC B cells, total GC B cells, and total GC Tfh cells were analyzed in draining LN FNA samples collected 3 weeks after second vaccination. The direction and degree of correlation by Spearman r is indicated by color and size of the interior square. *P < 0.05, **P < 0.01, and ***P < 0.001, by active vaccine treatment arm: 20-μg, 100-μg, and combined dose groups.

DISCUSSION

Germline-targeting priming and subsequent shepherding of B cell responses to induce HIV-specific bnAbs by vaccination is a promising approach to HIV vaccine design. The IAVI G001 phase 1 clinical trial of the eOD-GT8 60-mer/AS01_B vaccine demonstrated that VRC01-class memory B cells could be consistently induced by vaccination, validating the germline-targeting approach (11). However, maturing those responses to become bnAbs will require heterologous boosting strategies and robust CD4 T cell help.

Here, we showed that the eOD-GT8 60-mer/AS01_B vaccine induced robust CD4 T cell responses to eOD-GT8 and LumSyn in nearly all vaccine recipients receiving either the 20- or 100-μg dose. Total vaccine-specific CD4 T cells accounted for approximately 0.5% of circulating CD4 T cells in most vaccine recipients. This is similar or higher in magnitude and response rate than the approved SARS-CoV-2 mRNA and adenoviral vaccines (43, 50, 51). No differences were observed between vaccine dose groups for T cell magnitudes, response rates, or phenotypes, suggesting that the low dose was sufficiently potent for reproducible induction of CD4 T helper cells in participants of diverse HLA haplotypes. When we examined the epitope specificities of the CD4 T cell responses, we found that 70% of vaccine recipients had CD4 T cell responses to the subpool 3 of eOD-GT8 driven by two immunodominant regions in the C terminus of eOD-GT8 (subpool 3). Ninety-one percent of vaccine recipients had CD4 T cell responses to LumSyn, and 85% recognized at least one of three immunodominant regions, with the single highest epitope region being recognized by 68% of vaccine recipients.

Helper CD4 T cells are critical for the recruitment and maturation of vaccine-elicited antigen-specific B cells, especially in the context of rare or low-affinity antigen-specific B cells (28, 52). We identified two immunodominant CD4 T cell epitope regions within the HIV Env eOD-GT8 sequence: eOD-GT8-3 peptides 2 and 3 (WNNTLKQIASKLREQYGNK) and peptides 12 and 13 (GGEFFYCDSTQLFNSTWFN). These regions have high percent identity among HIV-1 subtype B sequences, and previous reports have described T helper responses overlapping with these eOD-GT8 peptide 2 and 3 or peptide 12 and 13 epitope regions in untreated people living with HIV-1 (53, 54). It is not clear whether other HIV-1 vaccine candidates have induced responses to these epitope regions, because CD4 T cell response data at the individual peptide level is very limited, but it is likely that they would be present in other Env-based immunogens and therefore likely have the potential to provide T cell help to heterologous HIV vaccines with sufficient sequence similarity.

Vaccine induction of HIV-specific CD4 T cells is not an unambiguously positive outcome. Multiple studies have indicated that increases in activated total or vaccine-specific CD4 T cells after vaccination can enhance acquisition of HIV in humans or of simian immunodeficiency virus or simian-human immunodeficiency virus (SHIV) in rhesus macaques (55–59). Increased frequencies of total activated CD4 T cells could, in principle, be caused by any HIV or non-HIV vaccine inducing T cell responses (57). Increased frequencies of activated adenovirus-specific CD4 T cells in the mucosa have been implicated as a possible explanation for the increased risk of HIV infection among adenovirus-5–seropositive individuals in the Step trial that tested a human adenovirus serotype 5 vector delivering HIV gag, pol, and nef (56, 60). However, there is evidence that HIV preferentially infects HIV-specific CD4 T cells compared with non-HIV–specific CD4 memory T cells, suggesting that vaccine induction of HIV-specific CD4 T cells, in particular, may increase susceptibility to infection (55, 61). Furthermore, the presence of Env-specific CD4 T cell responses was associated with faster disease progression in untreated chronic HIV-infected participants (62). Overall, the benefits of vaccine-induced CD4 T cells in supporting the development of protective B cell or CD8 T cell responses will need to outweigh the potential risks of increased target cells for an efficacious protective HIV vaccine (57).

LumSyn is a bacterial enzyme derived from Aquifex aeolicus and forms the scaffold of the nanoparticle for the display of the 60 copies of eOD-GT8 monomer. A LumSyn T helper epitope termed LS-3 (LRFGIVASRANHALV), which partially overlaps with the region covered by our LumSyn peptides 5 and 6 (IVASRFNHALVDRLVEGAI), was previously found to be immunogenic in mice immunized with eOD-GT8 60-mer expressed by synthetic DNA delivered by electroporation (63). In that study, mice immunized with a variant of eOD-GT8 60-mer lacking the LS-3 epitope produced a weaker humoral response. Furthermore, when this epitope was fused to a minimal domain immunogen, as DNA or protein, humoral responses to the minimal domain were increased (63). Those data support the premise that the LumSyn T helper hotspots found here to be immunodominant in humans might be sufficient to provide T cell help and support humoral responses in minimal-epitope–designed immunogens for human vaccination. We are not aware of other human data demonstrating such broad CD4 T cell immunogenicity across vaccine recipients for any one peptide (e.g., ATPHFDYIASEVSKGLADL with a 68% response rate) or three peptides (with a 85% response rate). Whether such peptides will be similarly immunogenic in other LumSyn-based immunogens or when transferred to non-LumSyn immunogens is not known. Hence, the potential utility of these LumSyn peptides providing strong and broad CD4 T help to other LumSyn- and non-LumSyn–based immunogens merits further investigation. LumSyn has the potential to provide continuity of T helper responses between heterologous prime and boost immunogens that are built on the same backbone but with different HIV-specific components. That strategy is being tested with an eOD-GT8 60-mer prime and a core-gp120 60-mer heterologous boost also based on LumSyn, both delivered by mRNA, in the IAVI G002 phase 1 trial (NCT05001373).

We evaluated the HLA genotypes among the vaccine recipients and found numerous HLA class II alleles that were associated with publicly targeted CD4 T cell responses. Many of these interactions were further supported by computationally predicted HLA binding affinities (48, 49). For multiple epitopes, such as LumSyn peptides 22 and 23, the individuals targeting the epitope were enriched for more than one HLA class II allele. This suggests that these regions could include multiple distinct optimal epitopes or promiscuous epitopes presented by multiple HLA allelic variants, and this may contribute to their wide recognition.

In a study of HLA allele distribution in South Africa (n = 402), HLA-DQA*01, associated with the LumSyn peptide 5 and 6 hotspot, and HLA-DRB4*01, enriched among LumSyn peptide 22 and 23 responders, were commonly observed in 74 and 76% of the study population, respectively (64). HLA-DRB3*01, enriched among LumSyn peptide 28 responders, was less common, observed in 38% of participants (64). In contrast, the other HLA alleles predicted to bind to the eOD-GT8 hotspots were less commonly observed in the South African cohort. HLA-DRB5*01, a predicted binder for the eOD-GT8 peptide 2 and 3 region, and HLA-DPB1*04, associated with the eOD-GT8 peptide 12 and 13 region, were only observed in 16 and 15% of the South African cohort, respectively. Together, these data suggest that the three epitope regions in LumSyn are restricted by sufficiently prevalent alleles to have at least one corresponding restricting allele to be expressed in the majority of South African population. This also raises the possibility that these LumSyn hotspots could be included in other vaccine immunogens to enhance breadth and magnitude of CD4 T helper responses.

The induction of vaccine-specific CD8 T cells was unexpected. There has been some suggestion that nanoparticles could more efficiently target dendritic cells for antigen cross-presentation to prime CD8 T cell responses, but this has not been demonstrated in humans. If the mechanism underlying this response could be determined, then it could possibly be exploited to induce HIV-specific CD8 T cells in parallel. However, the CD8 T cell response against LumSyn seemed to be driven by a specific HLA-A*02–restricted 9-mer epitope. We contemplated whether this response was truly induced by the vaccine. In Leggat et al. (11), we reported that vaccine and placebo recipients had detectable binding antibody responses to LumSyn at baseline. We hypothesized that these preexisting LumSyn-specific antibody responses were primed after exposure to homologs expressed by environmental, pathogenic, or microbiome bacteria or yeasts (11), and we found here that the predicted 9-mer epitope can be found in homologous LumSyn sequences from multiple microbial sequences. However, the LumSyn peptides used in this study were prescreened for false-positive responses using peripheral blood mononuclear cell (PBMC) samples from 48 healthy unvaccinated participants and 16 contemporaneous blinded placebo controls, all of which were negative. Therefore, one hypothesis is that these responses were possibly primed by previous exposure but specifically expanded by the vaccine. This would also explain why CD8 T cell responses were detected to LumSyn, but not to the HIV-derived, eOD-GT8, domain. It will be critical to distinguish between these hypotheses in future studies to determine whether the nanoparticle design can be exploited for induction of HIV-specific CD8 T cells.

In addition, we used FNAs from the draining LNs to longitudinally monitor the induction of GC activity after the prime and boost immunizations (65, 66). GC Tfh cells have been associated with induction of effective neutralizing antibodies in NHPs (66) and recruitment of rare VRC01-class B cells into GCs in quantitative studies of T cell help in a murine model (52). However, there remain limited detailed immunogenicity assessments after vaccination in clinical studies. Here, we observed strong induction of GC activation as indicated by increased Tfh cells after a single nanoparticle immunization similar to preclinical studies (31). We also found that the frequencies eOD-GT8–specific GC B cells, total GC B cells, and total Tfh cells correlated with each other, similar to previous reports of the SARS-CoV-2 mRNA and influenza vaccines (67). We also observed correlations between the vaccine-specific cTfh cells, vaccine-specific CD4 T cells, and eOD-GT8–specific IgG B cells in the periphery (68–70).

Unexpectedly, there were no positive associations of vaccine-induced B and T cells between the secondary lymphoid and peripheral blood compartments. Frequencies of LN GC Tfh cells tended to be inversely associated with peripheral eOD-GT8–specific IgG B cells. These results were counterintuitive given the overall positive relationship between vaccine-elicited B and CD4 T cells in both LN and peripheral blood compartments, as seen in our study. Presumably, LN FNAs have increased variability due to the physical selection of an LN to sample and the rapid changes in GC activity over time. Therefore, one hypothesis is that this increased variability and the fact that LN FNA and peripheral blood collections were conducted 1 week apart could have resulted in discordant frequencies between compartments. A second hypothesis is that, after the potent nanoparticle vaccine, the overall Tfh cells and GC B cells are so elevated that the range observed is insufficient for correlations to vaccine-elicited memory B cells. It is also possible that there are qualitative differences in vaccine-elicited memory B cells that are not captured in these analyses, such as affinity maturation or the eventual generation of high-affinity neutralizing antibodies.

Here, we have reported a comprehensive analysis of T cell responses induced by the eOD-GT8 nanoparticle vaccine. However, there were some limitations to our study that are worth noting. Vaccine-specific T cell responses were evaluated by a validated ICS assay in cryopreserved leukapheresis samples obtained at week 10. In contrast, bulk GC Tfh cells were evaluated by phenotypic marker expression at week 11 from freshly isolated LN FNAs. The temporal and technical differences in these analyses could have affected our ability to assess correlations between the physiological compartments. In addition, although the rates of vaccine-specific CD4 T cell responses were impressive, three participants did not have readily detectable vaccine-specific CD4 T cell responses in the peripheral blood. We confirmed that these participants did have detectable eOD-GT8⁺ IgG memory B cells at the same time point, and two of the three participants had available data from the LN aspirates where eOD-GT8⁺ IgG GC B cells and GC Tfh cells were also observed. Thus, it is unlikely that these B cell responses were driven by T cell–independent extrafollicular B cells. Instead, we hypothesize that the vaccine-specific CD4 T cells in these participants were either sequestered in lymphoid compartments or below the limits of detection of the highly specific validated ICS assay (71). In future studies, we would like to dissect the relationship between GC Tfh cells, peripheral vaccine–specific CD4 T cells, and induction of effective humoral responses in response to vaccination. Now that we have identified the common vaccine-specific CD4 T cell epitopes for this vaccine, minimal epitopes could be designed for optimal stimulation or tetramer staining to identify vaccine-specific GC Tfh cells and peripheral CD4 T cells for repertoire and phenotype comparisons longitudinally and across compartments.

Together, the eOD-GT8 60-mer with AS01_B induced robust and broad vaccine-specific CD4 T cell responses that were targeted by commonly expressed HLA alleles. We identified numerous human CD4 T cell epitopes among the nanoparticle backbone, LumSyn, which has the potential to provide continuity in T cell help across heterologous prime-boost vaccine regimens. Furthermore, we identified small peptide regions that were publicly recognized across vaccine recipients and have the potential to provide or enhance T cell help in minimal epitope designed vaccines to support induction of effective humoral immunity.

MATERIALS AND METHODS

Study design

IAVI G001 (ClinicalTrials.gov NCT03547245) was a phase 1, randomized, double-blind, placebo-controlled dose escalation study to evaluate the safety and immunogenicity of the eOD-GT8 60-mer vaccine adjuvanted with AS01_B in HIV-uninfected, healthy adult volunteers. eOD-GT8 60-mer is a glycosylated self-assembling nanoparticle composed of 60 copies of a fusion protein in which the eOD of HIV gp120, eOD-GT8, is fused to the C terminus of a modified version of the enzyme LumSyn from the bacteria A. aeolicus through a flexible glycine-serine linker (9, 10, 72). The protein molecular mass of one protomer is 36.6 kDa, and the protein molecular mass of the entire nanoparticle is about 2200 kDa. Details of the study design, sample size determination, randomization, inclusion/exclusion criteria, participant demographics, and safety have been previously reported (11). Two doses of 20- or 100-μg eOD-GT8 60-mer with AS01_B or two doses of placebo were given by deltoid intramuscular injection 8 weeks apart, with both immunizations given to the same arm. Placebo was the buffer used in the vaccine: Dulbecco’s phosphate-buffered saline (PBS) containing 10% sucrose at pH 7.5. The trial adhered to IAVI standard operating procedures in accordance with the guidelines formulated by the International Committee on Harmonization for Good Clinical Practice in clinical studies and complied with applicable local standards and regulatory requirements, including review and approval by the institutional review boards at Fred Hutchinson Cancer Center (FHCC) and George Washington University (GWU).

PBMC samples were isolated and cryopreserved within 8 hours of leukapheresis. There were 18 vaccine and six placebo recipients enrolled in groups 1 and 2, for a total of 48 expected participants. Of those, three participants had no ICS results at visit 8. One vaccine recipient missed their visit, and the other two had small blood draws and did not have specimens shipped to the laboratory for assays. Therefore, the maximum number of participants available for analysis was 45. An HIV-negative, CMV-positive PBMC sample from a leukapheresis was stained with every experiment and used as a negative control for the vaccine-matched peptide pools and as a positive batch control using CMV stimulation.

ICS and epitope mapping

The ICS assay was used to examine eOD-GT8–specific CD4 and CD8 T cell responses using previously cryopreserved PBMC specimens from leukaphereses obtained 2 weeks after the second vaccination (week 10). Individual peptides of 15 amino acids overlapping by 11 amino acids covering the eOD-GT8 and LumSyn linear amino acid sequences (Biosynthesis) were individually reconstituted in dimethyl sulfoxide (DMSO) at 50 mg/ml, pooled, and diluted with diethylpyrocarbonate-treated water (100 μg/ml). Peptide pools covered LumSyn (1 to 175 amino acids) and eOD-GT8 (165 to 341 amino acids), where amino acid numbering is relative to the eOD-GT8–LumSyn linear sequence. To establish baseline T cell reactivity to these antigens, we first evaluated PBMCs from unvaccinated HIV-negative (n = 40) and HIV-infected (n = 8) donors. No LumSyn-specific CD4 or CD8 T cell responses were detected among the 48 unvaccinated donors (fig. S1). Similarly, for eOD-GT8, no T cell responses were detected among the uninfected participants.

The peptides covering the eOD-GT8 sequence were further split into subpools of 13 to 15 peptides each from the N terminus to C terminus: eOD-GT8-1 (165 to 235 amino acids), eOD-GT8-2 (225 to 287 amino acids), and eOD-GT8-3 (277 to 341 amino acids), for mapping. Samples with positive responses to the LumSyn or eOD-GT8-3 pools were then tested using the corresponding individual overlapping peptides: 41 peptides for LumSyn and 14 peptides for eOD-GT8-3.

A validated 17-color ICS assay modified from a previously published panel was used (73). Briefly, PBMCs were rested overnight (16 to 18 hours). The next day, 0.5 × 10⁶ to 1 × 10⁶ PBMCs were plated in 200 μl per well in R10 medium [RPMI 1640 + 10% fetal bovine serum (FBS)] and stimulated for 6 hours with peptides or peptide pools. A final concentration of 1 μg/ml per peptide was used for peptide pool stimulations, and 2 μg/ml per peptide was used for testing individual peptides. Costimulatory molecules, anti-CD28 and CD49d antibodies (final concentration, 1 μg/ml; BD Biosciences), and brefeldin A (final concentration, 10 μg/ml; Sigma-Aldrich), to block cytokine secretion, were added to the stimulations for the 6-hour incubation. After the incubation, 20 μl of 20 mM EDTA to each well, and the plates were transferred to 4°C overnight. The next day, cells were washed and stained with viability dye (Thermo Fisher Scientific) in 1× PBS for 20 min in the dark. The cells were washed again and incubated with the surface staining antibodies in 1× PBS with 2% FBS (table S4) for 20 min in the dark. Subsequently, cells were fixed (BD FACS Lyse, BD Biosciences) for 10 min, centrifuged, and permeabilized (BD FACS Perm II, BD Biosciences) for 10 min. After the permeabilization, the cells were washed with 1× PBS with 2% FBS, centrifuged, decanted, resuspended with the intracellular staining cocktail (table S4) in 1× PBS with 2% FBS, and incubated in the dark for 30 min. Cells were washed twice with 1× PBS with 2% FBS, resuspended in 1× PBS with 1% paraformaldehyde, and acquired the same day on a custom LSRII (BD Biosciences).

As a negative control, two wells containing DMSO, the peptide diluent, were run per participant PBMC sample. As a positive control, cells were stimulated with a polyclonal stimulant, staphylococcal enterotoxin B (SEB; final concentration of 0.25 μg/ml; Millipore). An additional positive control well containing CMV pp65 peptide pool–stimulated cells was also run per each participant sample. An HIV-negative known CMV responder PBMC sample was also included for each batch as an assay control. Records were filtered for the following low T cell counts: CD4 T cell count fewer than 10,000 events or CD8 T cell count fewer than 5000 events. Results were reported as mock-subtracted percentages of CD4 or CD8 T cells.

GC Tfh cell phenotyping

We conducted FNAs of ipsilateral axillary LNs as previously described (11). Briefly, with ultrasound assistance, a 22-gauge needle was used to aspirate from the largest and most accessible LN. The needle contents were expelled into a 50-ml conical tube containing R10 medium. The collection was repeated up to four times for each study participant, using a new sterile needle and syringe for each pass. After collection, the sample was transported to the laboratory on wet ice or cold packs as soon as possible. LN FNA samples ranged in recovery from very few cells up to 4 × 10⁷ cells (median recovery varied between sites from 2.4 × 10⁶ versus 4.7 × 10⁶).

The FNA sample in a 50-ml conical tube was spun at 325g for 10 min at 4°C (no brake). Supernatant was gently aspirated with a serological pipet down to about 500 μl. Cells were resuspended in 5 ml of cold 1× red blood cell lysis buffer (eBioscience) and incubated for 5 min at ambient temperature. Cells were washed with 40 ml of cold 1× PBS with 10% FBS and centrifuged at 325g for 10 min at 4°C. Supernatant was gently aspirated with a serological pipet, and the cell pellets were resuspended with 2 ml of cold 1× PBS with 10% FBS.

Cells were stained as described elsewhere (11) with a surface staining cocktail (table S5) in Brilliant Buffer (BD Biosciences) and incubated for 30 min at 4°C. Cells were washed in 10% FBS in 1× PBS, resuspended in a cocktail of viability dye, 7-aminoactinomycin D (7-AAD) in R10 medium, and acquired on an Aria flow cytometer (BD Biosciences). GC Tfh cells were identified as CD3⁺, CD4⁺, CXCR5⁺⁺, and PD-1⁺⁺ live lymphocytes.

Statistical analysis

Background-adjusted percentages of positive T cells were visualized using scatterplots by T cell subset, peptide pool, marker, and treatment. Positive responders were plotted as points colored by treatment, and nonresponders were plotted as open gray triangles. Positive responses were determined using the MIMOSA statistical test to compare the number of cytokine-positive CD4 or CD8 T cells under the antigen stimulated and unstimulated conditions (74). Briefly, responder status to each peptide pool during ICS was accessed using the MIMOSA package, which implements a two-component mixture model testing the null hypothesis that the proportion of cytokine-expressing cells is equal in the stimulated and unstimulated samples. Because counts in ICS assays are frequently small, the MIMOSA package pools information across individuals to increase sensitivity and specificity. However, the power to detect a “positive” antigen-specific response is affected by the degree of background cytokine secretion in the unstimulated control and the number of events observed. The lower the background in the unstimulated control and the higher the cell numbers observed, the higher power to determine a positive antigen-specific response. Responses with probability of response of >0.999 were considered positive responders.

Box plots were superimposed on the distribution of responders. Pairwise comparisons of response rates by treatment group were performed using Barnard’s test (two-sided, P ≤ 0.05). Placebo recipients were pooled. Comparisons of response rates between eOD-GT8 and LumSyn (vaccinees only) were performed within dose group using McNemar’s test for paired data (two-sided, P ≤ 0.05).

For the K-means clustering, individual antigen-reactive T cells were defined by the coexpression of at least two of six functional markers (IFN-γ, IL-2, TNF-α, CD154, IL-4, or IL17a). These two of six definitions of positive antigen-reactive T cells were chosen because of the strong response among the LumSyn and eOD-GT8 stimulations and low response for negative control stimulations. Cells stimulated with LumSyn, eOD-GT8, CMV, and negative control were included in the analysis. SEB-stimulated positive control cells were not considered for this analysis. Twelve ICS markers were included in the K-means clustering analysis: IFN-γ, IL-2, TNF-α, CD154, IL-4, IL-17a, granzyme B, ICOS, CXCR5, PD-1, CD45RA, and CCR7. For the marker IL-4, simultaneous coexpression of CD154 was required to exclude background fluorescence. The data were collected for two master peptide pools covering the eOD-GT8–LumSyn linear sequence: LumSyn and eOD-GT8 and to the three eOD-GT8 subpools, at week 10 (2 weeks after second vaccination). Cluster magnitude was defined as background-adjusted percentage of positive T cells for each combination of cellular markers. The percentage positive measurement was background-adjusted by subtracting the percentage of positive cells from the negative controls from the stimulated conditions. The marker positivity by cluster is the proportion of cells that are positive for a given marker of the total number of cells in a cluster.

K-means clustering (75) was performed on the full-dimensional dataset specified. The cells on the two-dimensional Uniform Manifold Approximation and Projection (UMAP) for dimension reduction space were then colored by cluster number. The number of clusters was set to 10, which resulted in well-defined clusters in the two-dimensional UMAP space. Marker positivity by cluster was visually displayed with a heatmap. The following measures were displayed with boxplots: background-adjusted percentage positive cells, based on 2⁺ of 6 function marker subset, in each cluster of CD4 T cells.

Within each cluster, response magnitude (background-adjusted percent positive cells) was compared between vaccine and placebo recipients using the Wilcoxon rank-sum test (two-sided, P ≤ 0.05). High- and low-dose vaccine recipients were pooled, as were placebo recipients. To reduce the false discovery rate (FDR) across the multiple Wilcoxon rank-sum tests, P values were adjusted using the Benjamini and Hochberg procedure (76). Paired response magnitudes for LumSyn versus eOD-GT8 (master pool) were compared within cluster between treatment group using the Wilcoxon signed-rank test (two-sided, P ≤ 0.05) with the FDR reduced as described above.

Statistical associations of HLA alleles corresponding to major histocompatibility complex (MHC) class I (HLA-A, HLA-B, and HLA-C) with CD8 T cell responses and MHC class II (HLA-DP, HLA-DM, HLA-DO, HLA-DQ, and HLA-DR) alleles with CD4 T cell responses were determined. To determine the relationship between allele presence and peptide response, Fisher’s exact test was computed for each peptide and allele combination, and results with a P < 0.05 were displayed. Adjusted P values using the FDR were also displayed. Comparisons were restricted to allele and peptide pool pairs with sufficient variability in allele frequency and peptide pool responses. Alleles with fewer than four participants having the allele present or fewer than four participants having the allele not present were excluded from comparisons. Similarly, peptide pools with fewer than four participants having a positive response and fewer than four participants having a negative response were excluded from comparisons.

Data and materials availability:

All data associated with this study are presented in the paper or the Supplementary Materials. All data files produced in this study are available in the Zenodo public data repository at https://doi.org/10.5281/zenodo.7563250 (77). The repository also contains code for data analysis. Requests for materials should be directed to the corresponding authors.