In humans it is difficult to predict which T-cell responses will be induced after vaccination because they are governed by the phenomenon of immunodominance [4]: most pathogen-specific CD8+ T cells are directed towards a few epitopes, despite the presence of thousands of potential epitopes. Factors governing immunodominance are complex and include 1) positive and negative selection of the T-cell receptor (TCR) repertoire in the thymus, 2) abundance of intracellular antigens, 3) efficiency of protease generation of class I epitopes [5], 4) efficiency of transporter associated with antigen processing (TAP)-dependent or -independent transport of peptides into the endoplasmic reticulum, 5) affinity of peptide binding to class I molecules, and 6) naïve T-cell precursor frequency [6]. It is important to determine which factors select the specificities and restrictions of vaccine-induced T-cell responses since some T-cell specificities seem more effective at controlling viral replication [7-8]. Here we investigated the epitopes and HLA molecules utilized to mount HIV-specific CD8+ T-cell responses in healthy HIV-seronegative individuals who participated in clinical trials of candidate recombinant adenoviral vaccines. These findings have profound implications for vaccine design as they show that immunodominance plays a substantial role in vaccine-induced responses.

Two peptide sets, potential T-cell epitopes (PTE) and Consensus B (ConB), were employed to detect HIV-specific T-cell responses. PTE peptides were 15-mers designed according to published methods [11] (synthesized by BioSyn, Lewisville, TX). ConB peptides were 15-mers overlapping by 11 a.a. based on HIV-1 clade B consensus sequence (provided by NIH AIDS Research and Reference Reagent Program, Bethesda, MD). Additionally, 8-11-mers corresponding to described class I HLA-restricted Cytotoxic T Lymphocyte (CTL) epitopes were tested according to an individual's HLA type (synthesized by BioSyn, New England Peptide (Gardner, MA), and Mimitopes (Minneapolis, MN)). When PBMC were limiting, only known CTL epitopes that corresponded to the individual's HLA type were tested.

IFN-γ ELISpot was performed as previously described [8]. Results were considered positive if the number of spot forming cells (SFC) was at least twice background and ≥50 SFC/10⁶ PBMC. Positive ELISpot results to a 15-mer from either peptide set were confirmed by testing positive for one of the following: IFN-γ ELISpot to a corresponding 15-mer from the other peptide set, IFN-γ ELISpot to an embedded optimal epitope (8-11-mers), or an Intracellular Cytokine Staining (ICS) assay detecting IFN-γ, TNF-α, IL-2, and/or CD107a. Additionally this ICS assay was used to discriminate T-cell responses as either CD4+ or CD8+; CD4+ T-cell responses were not included in this analysis. HLA restrictions for defined optimal epitopes were predicted based on published class I HLA-restricted CTL epitopes.

To measure T-cell avidity, IFN-γ ELISpot was performed using peptide concentrations titrated from 20 μg/ml to 0.0002 μg/ml. Functional avidity values (EC₅₀) were calculated based on regression curves drawn with the Sigmoidal Fit tool in Origin 6.0 software.

To measure HLA surface expression, CD4+ T-cells were enriched by negative selection (StemCell Technologies, Vancouver, BC) and incubated with mouse anti-human HLA monoclonal antibodies (One Lambda, Canoga Park, CA) for 30 minutes before washing and staining with goat anti-mouse IgG Fluorescein-conjugated antibody (Beckman Coulter, Fullerton, CA). Data acquisition was performed on an LSRII flow cytometer (BD) and analyzed with FlowJo (TreeStar, Ashland, OR). The mean fluorescent intensity (MFI) of samples was compared to QIFI beads (DakoCytomation, Glostrup, Denmark) coated with well-defined quantities of mouse monoclonal antibody stained with the same secondary antibody. Antibody-binding capacity (ABC) was calculated based on the QIFI bead MFI curve.

Quantitative assays to measure the binding of peptides to HLA class I molecules were based on the inhibition of binding of a radiolabeled standard peptide, as described previously [12].

Our data and others have shown in addition to specific HLA alleles preferentially utilized to mount vaccine-induced responses, specific HIV-1 epitopes were targeted (Table 1) [16]. All HLA-B27+ vaccinees mounted a B27-restricted Gag p24 response against KRWIILGLNK (KK10) in conjunction with a Pol response against FKRKGGIGGY (FY10). The latter epitope is likely KRKGGIGGY (KY9), which was a recently reported B27-restricted epitope found in chronic infection [17]. Similarly, all HLA-B57+ vaccinees mounted a B57-restricted Gag p24 response against TSTLQEQIGW (TW10) [12]. The preferential selection of certain epitopes may be because they have significantly higher affinity for their HLA alleles, as assessed by HLA binding assays (Figure 1A). The HLA-B35-restricted responses were split between the Px and PY alleles, which differ in their binding motif anchor residues. Both motifs bind peptides with a proline at position two, but peptides bound by PY alleles have a tyrosine at the C-terminus position, whereas peptides bound by Px alleles have other C-terminal residues. All B35Px vaccinees recognized the B35-restricted Env DPNPQEVVL (DL9) epitope in conjunction with the B35-restricted Pol IPLTEEAEL (IL9) epitope. Conversely, 71% of B35PY vaccinees recognized the B35-restricted Pol NPDIVIYQY (NY9) epitope (Table 1). Interestingly, B35+ participants receiving the VRC vaccine only mounted four responses to Gag and 28 responses to other HIV proteins regardless of HLA restriction. In addition, among 21 B35+ MRK-Gag vaccinees, only one mounted a B35-restricted response. These data suggest that in the absence of highly mutable Env and Pol epitopes (present within the VRC vaccine) HLA class I alleles other than B35 are utilized to mount a response.

Other data from our group suggest that epitopes targeted early in infection determine disease progression [8]. However, that study did not address why specific epitopes/HLA were immunodominant in both early infection and vaccination. It has been postulated that high-avidity T-cell responses dominate in early infection [18]. In this study we found vaccine-induced B27 KK10-specific T cells to be of low avidity. Of note, vaccine-induced B27 KK10-specific T cells were of lower avidity than KK10-specific T cells from HIV-1 infected individuals. In contrast B35-specific T cells were of higher avidity than those restricted by other alleles (p=0.0017; Figure 1B; see Supplemental Table 3 for complete list). This suggests that B35 may be preferentially utilized after vaccination, because T cells restricted by this allele are of higher avidity than T cells restricted by other alleles.

However, it remained unclear why T cells restricted by B27 dominate in early infection and after vaccination despite their low avidity. We hypothesized that this may be due to higher cell surface density of B27 molecules. To address this we quantitatively measured cell surface expression of HLA-B27 in relation to HLA-A24 and HLA-A11 alleles (Figure 1C) and found that HLA-B27 is expressed at a 45% higher level on the cell surface than other HLA alleles tested. Thus, T cells restricted by HLA alleles which are expressed at high levels at the cell surface can dominate even if they have low avidity TCRs. Unfortunately it was not possible to assess the surface expression of a larger set of HLA alleles including HLA-B57 due to the unavailability of specific IgG antibodies.

In addition to TCR avidity, epitope affinity, and HLA surface expression, factors such as intracellular abundance and stability of epitopes and proteosome/tapasin-dependence may also play a role in determining immunodominance [19-20], although these were outside the scope of this study.

1