Novel, in-natural-infection subdominant HIV-1 CD8+ T-cell epitopes revealed in human recipients of conserved-region T-cell vaccines

Background Fine definition of targeted CD8+ T-cell epitopes and their human leucocyte antigen (HLA) class I restriction informs iterative improvements of HIV-1 T-cell vaccine designs and may predict early vaccine success or failure. Here, lymphocytes from volunteers, who had received candidate HIVconsv vaccines expressing conserved sub-protein regions of HIV-1, were used to define the optimum-length target epitopes and their HLA restriction. In HIV-1-positive patients, CD8+ T-cell responses predominantly recognize immunodominant, but hypervariable and therefore less protective epitopes. The less variable, more protective epitopes in conserved regions are typically subdominant. Therefore, induction of strong responses to conserved regions by vaccination provides an opportunity to discover novel important epitopes. Methods Cryopreserved lymphocytes from vaccine recipients were expanded by stimulation with 15-mer responder peptides for 10 days to establish short term-cell-line (STCL) effector cells. These were subjected to intracellular cytokine staining using serially truncated peptides and peptide-pulsed 721.221 cells expressing individual HLA class I alleles to define minimal epitope length and HLA restriction by stimulation of IFN-γ and TNF-α production and surface expression of CD107a. Results Using lymphocyte samples of 12 vaccine recipients, we defined 14 previously unreported optimal CD8+ T-cell HIV-1 epitopes and their four-digit HLA allele restriction (6 HLA-A, 7 HLA-B and 1 HLA-C alleles). Further 13 novel targets with incomplete information were revealed. Conclusions The high rate of discovery of novel CD8+ T-cell effector epitopes warrants further epitope mining in recipients of the conserved-region vaccines in other populations and informs development of HIV-1/AIDS vaccines. Trial registration ClinicalTrials.gov NCT01151319


Introduction
The protective role of CD8 + T cells against HIV-1 infection has been implicated by combined data from genome-wide association studies, viral sequence polymorphisms and replicative fitness analyses, and longitudinal maps of epitope escape [1][2][3]. Therefore, induction of effective CD8 + T cells by vaccines will likely be needed to complement induction of binding or broadly neutralizing antibodies for prevention of HIV-1 infection as well as assist HIV-1 cure.
Evidence is emerging that the CD8 + T-cell specificity, cognitive breadth and human lymphocyte antigen (HLA) restriction are crucial determinants of the T-cell response protective efficacy [1][2][3]. It follows that careful definition of optimal epitopes can critically inform T-cell vaccine design and increase the power of early prediction of vaccine success or failure. For this reason since 1998, the Los Alamos National Laboratory HIV Sequence Database (LANL-HSD; www.hiv.lanl.gov) has been collecting and posting now well over 300 of the best-defined, finemapped epitopes in the HIV-1 proteome restricted by over 80 HLA class I alleles as the 'A list' of HIV-1 CD8 + T-cell epitopes, while the less defined specific T-cell targets are gathered as the 'B list' [4].
One of the major challenges in vaccine development is the enormous capacity of HIV-1 for adaptation and diversification [5,6]. This is because over the course of natural infection, immune responses through the actions of antibodies, CTL and perhaps innate responses drive rapid HIV-1 evolution known as immune escape [7][8][9]. Under this strong selective pressure, HIV-1 evolved to have immunodominant epitopes in the most variable regions of its proteins as decoys, while the more functionally/structurally conserved, harder-to-change and therefore more protective determinants remain subdominant [10,11]. As a result, many of the potentially protective epitopes are left underutilized and/or completely ignored by the immune system. During natural infection, the initially strong CD8 + T-cell responses are thought to be swayed towards rapidly changing immunodominant epitopes and by the time the more protective subdominant epitopes may be targeted, damage to the immune system is already irreparable [3,11,12].
Harnessing the protective potential of and (re)focusing CD8 + T-cells on the HIV-1 conserved regions (HIVconsv) by active immunization is the central theorem of our T-cell vaccine strategy [11]. An additional advantage of this approach is that the conserved protein segments are common to most M group HIV-1 variants, and thus if successful, the vaccines could provide broad cross-clade protection. We constructed two generations of conserved-region vaccines HIVconsv and tHIVconsvX [13,14], and showed by elution studies that the vaccine epitopes were presented by HLA class I molecules on the surface of vaccine-and HIV-1-infected cells [15,16]. We also demonstrated in human volunteers that naturally subdominant regions when taken out of the context of the full-length viral proteins and delivered by a potent adenovirus prime-poxvirus boost regimens induced robust CD8 + T-cell responses capable of broad HIV-1 inhibition in vitro [17][18][19]. In the work presented here, we defined the optimal epitopes and their restricting HLA alleles induced by the first generation HIVconsv vaccines in healthy HIV-1-negative volunteers. Many of these epitopes have not been described before and a number of them are candidates for the Los Alamos HIV immunology database 'A list' [4].

HLA typing of volunteers
Molecular typing for HLA-A, HLA-B, HLA-C & HLA-DRB was performed by The DNA Sequencing MISEQ HLA Laboratory, Weatherall Institute of Molecular Medicine, The John Radcliffe, on EDTA blood samples taken from all volunteers prior to vaccination. The DNA was purified using an ArchivePure TM DNA purification kit (5 PRIME GmbH) according to the manufacturer's instructions and typing was performed using specific oligonucleotide probes to amplify specific HLA genes by SS-PCR.

Prediction of epitopes and HLA restriction
Candidate peptides obtained from the ICS analysis together with volunteer's HLA data were entered into the ELF algorithm (epitope location finder; www.hiv.lanl.gov) and compared to previously reported epitopes retrieved from the CTL databases. Potential epitopes were identified based on either anchor residues that matched motifs associated with the submitted HLA's or using the Immune Epitope Database (IEDB: www.immuneepitope.org), which predicts epitopes based on the affinity of given HLA/peptide interactions. The same data were also run on HLArestrictor with NetMHCpan version 2.4 (www.cbs.dtu.dk).
re-suspended in 1 ml of medium and placed with loose lids at 37˚C, 5% CO 2 for 48 hours to rest prior to the ICS assay.
The following gating strategy was employed for data analysis Intracellular cytokine assay (ICS) for HLA restriction 721.221 Epstein-Barr virus-transformed B-cell lines transfected with a single HLA gene were maintained in RPMI 1640 medium supplemented with 20% FC, and 40 μg/ml puromycin. For peptide pulsing, 0.6 x 10 6 721.221 cells were pelleted in a conical test tube, incubated in 25 μl of 200-μg/ml peptide solution for 1 hour at 37˚C, washed by centrifugation three times to remove excess peptide. Effector cells were STCL generated as described above and rested for 48 hours prior to use. Targets and effectors were mixed at 1:1 ratio, minimum 2 x 10 5 of each, in medium supplemented with anti-CD28 and anti-CD49d (Becton-Dickinson) both at 1 μg/ml. Controls were unpulsed targets plus effectors and effectors with peptide only. After incubating for 1 hour at 37˚C, 10 μg/ml brefeldin A (Becton-Dickinson) was added and the cells were incubated at 37˚C for a further 4-5 hours, stored overnight at +4˚C, stained on the following day and analyzed as described above for the ICS assay.

The vaccines and subjects
The first generation HIVconsv immunogen is derived from 14 highly conserved regions of the HIV-1 proteome. It uses clade A, B, C and D consensus amino acid sequences and alternates the clade of origin for each adjacent region [13] (Fig 1A). The HIVconsv immunogen was delivered by three vaccine modalities: plasmid DNA, non-replicating simian (chimpanzee) adenovirus ChAdV-63 and non-replicating poxvirus modified vaccinia virus Ankara (MVA) in heterologous prime-boost regimens ( Fig 1B). All PBMC samples used in this study were drawn from HIV-1-negative adult vaccine recipients in phase 1/2a trial HIV-CORE 002, which took place in Oxford, UK between March 2011 and April 2015 [18,20]. All volunteers were tissue typed and each HLA-A and B allele was assigned an HLA supertype [21] (S1 Table).
Definition of optimal CD8 + T-cell epitopes A panel of 15-mer overlapping peptides (HC001-HC199) across the entire HIVconsv protein sequence was originally used to determine the breadth of HIVconsv responses in an IFN-γ  [18] that the volunteers analyzed in this study received. ChAdV-recombinant non-replicating simian (chimpanzee) adenovirus 63 ChAdV63.HIVconsv; MVA-recombinant non-replicating poxvirus modified vaccinia virus Ankara MVA.HIVconsv; and DNA-'naked' plasmid pSG2.HIVconsv DNA. (C and D) Optimal epitope mapping was performed using thawed vaccine-recipients' PBMCs, which were in vitro expanded with the parental 15-mer peptides for 10 days to establish STCL. (C) Frequencies of vaccine-elicited, HIV-1-specific, in vitro 15-mer peptide (x-axis)expanded sequential PBMC samples of several volunteers from 5 (black) and 12 (grey) months after the last vaccine administration. Specific cells were enumerated in an IFN-γ ELISPOT assay. Panel (D) overviews the frequencies of peptide-specific CD8 + STCLs over the volunteers and peptides used. Frequencies of specific cells from all tested volunteers expanded by the same 15-mer peptide are shown next to each other above each peptide given on the x-axis. Top and bottom graphs show frequencies of CD8 + T cells producing IFN-γ and TNF-α, respectively. S2 Table for  ELISPOT assay [18]. Here, the PBMCs used for epitope mapping were collected between 10 weeks and 1 year after the last vaccine administration and were expanded in vitro using a 10-day peptide culture to establish an STCL. For selected responders and peptides, STCLs were generated from two different time points and the recognition patterns corresponded well, although individual peptide responses were with time and magnitude varied (Fig 1C and  1D). For all 12 studied individuals, optimal lengths of epitopes contained within the 15-mer peptides stimulating the up to 5 highest frequencies of HIVconsv-specific STCLs in each individual (S2 Table) were first predicted in silico based on the subjects' HLAs (S3 Table). To define optimal peptides experimentally, progressively truncated peptides were used to stimulate 15-mer-expanded STCL in an ICS assay including antibodies to CD3, CD4 and CD8 markers (CD4 + T-cell determinants will be published separately). For most volunteers, a complete set of their HLA class I allele gene transfectants of lymphoblastoid 721.221 or C1R cells was generated to determine the restricting HLA class I allele. The results are summarized in Table 1 and their narrowing and HLA-restriction are discussed in the order of the 'parental' 15-mer peptide numbers and shown in S1-S19 Figs. Some parental peptides had additional lysine(s) (K or KK) added to the HIV-1-derived amino acid sequences for solubility.

Discussion
Our T-cell vaccines against HIV-1 express immunogen HIVconsv and aim to focus CD8 + effector T cells on the functionally conserved, typically subdominant epitopes [44]. This approach has now been tested in 8 clinical trials in Europe and Africa (refs. [18,19,45] and unpublished) generating a wealth of cryopreserved PBMC samples with vaccine-elicited CD8 + T-cell responses. In the work presented here, these samples were utilized for definition of novel T-cell epitopes that have the potential to serve as important targets of an effective HIV-1-specific T-cell response. Overall, we identified in the HIV-1 proteome 14 previously unreported, and therefore considered novel, 'A-list' epitope candidates (6 HLA-A, 7 HLA-B and 1 HLA-C) and further 13 novel CD8 + T-cell targets that add to the epitope 'B list' (Table 1). These epitopes were derived mostly from HIV-1 Pol, but also Gag and Vif proteins. While discovery of novel epitopes is not unexpected given the subdominance of these epitopes in naturally-infected HIV-1 patients, the novelty of a small proportion of these could also come from the fact that our vaccines employ 'artificial' clade consensus amino acid sequences, which are processed into peptides that may not all be present in the natural HIV-1 isolates. Nevertheless, the vaccine-elicited CD8 + T cells in vitro proliferated and were plurifunctional producing IFN-γ and TNF-α and expressing cell surface CD107a, a degranulation marker, upon a specific peptide stimulus. The IFN-γ, TNF-α and CD107a responses concurred well. While it is reassuring that epitopes AIFQSSMTK (Pol) and RTWKSLVK (Vif) were eluted from HLA class I molecules on HIVconsv vaccine-infected human cells [36], the relevance of many of the other subdominant epitopes in terms of their presentation on HIV-1-infected cells is being currently studied in in vitro HIV-1 inhibition assays [17][18][19]46] and in HIV-1-potisitive patients [47]. However, the latter will be likely less useful given the epitope subdominance in natural infection.
In conclusion, in vivo induction of CD8 + T-cell effectors, their antiviral activity and assessment of their role in HIV-1 control rely strongly on unequivocal definition of the targeted epitopes and their HLA restriction collected as the 'A-list' epitopes [4]. As more vaccine candidates enter clinical trials, immunogen selection and in vitro monitoring of vaccine success become critical for go/no go developmental decisions. Through identification of new conserved targets on HIV-1 proteins, the information gathered in the present study contributes to the carefully collected list of regions of HIV-1 vulnerability and will help inform iterative improvements towards truly effective T-cell vaccines and their target product profile definition.

S1 Fig. HC031 CTERQANFLGKIWPS (Gag)-Definition of CD8 + T-cell determinants.
(A) The box. 15-mer peptide HC031 was recognized by volunteer 421 of the indicated HLA type, and the optimal peptides and determined HLA restriction are summarized. Cryopreserved lymphocytes from the vaccine recipient were expanded by stimulation with 'parental' 15-mer responder peptide for 10 days to establish STCL effector cells. These were subjected to ICS using serially truncated (B), and overlapping 9-mer (C) peptides. In (B), IFN-γ (green) and TNF-α (orange) production and surface expression of CD107a (pink) served as the readout. Arrows next to an amino acid indicate the peptide-terminal amino acid residue required for efficient peptide recognition. (PDF)

S2 Fig. HC049 KNFPISPIETVPVKLK / HC050 SPIETVPVKLKPGMD (Pol)-Definition of CD8 + T-cell determinants. (A)
The box. Overlapping 'parental' peptides HC049 and HC050 were recognized by volunteer 406 of the indicated HLA type. Optimal peptide and its HLA restriction are shown. (B) The subject's cryopreserved lymphocytes were expanded by stimulation with the 'parental' responder peptide for 10 days to establish STCL effector cells, which were tested in ICS using serially truncated peptides, whereby IFN-γ (green) and TNF-α (orange) production and surface expression of CD107a (pink) were detected. Arrows with an amino acid on the horizontal-bar graphs indicate the peptide-terminal amino acid residue required for efficient peptide recognition. Optimal peptides are listed below. Cryopreserved 411 lymphocytes were expanded by stimulation with 'parental' 15-mer peptide for 10 days to establish STCL effector cells. These were subjected to ICS using serially truncated (B), and overlapping 9-mer or 8-mer (C) peptides monitoring IFN-γ (green) and TNF-α (orange) production and surface expression of CD107a (pink). In (B), arrows next to an amino acid indicate the peptide-terminal amino acid residue required for efficient peptide recognition. (D) Three strong stimulatory peptides were subjected to titration against the HC080-expanded SCTL. (PDF)

S6 Fig. HC081 YTAFTIPSINNETPG (Pol)-Definition of CD8 + T-cell determinants.
(A) The box. 15-mer peptide HC081 was recognized by volunteer 415 of the indicated HLA type. While restricting HLAs were not determined for the three stimulatory peptides, the HLA-A Ã 02:01 restriction for the 'parental'-derived peptide was confirmed. Volunteer's lymphocytes were expanded by stimulation with peptide HC081 for 10 days to establish an STCL, which was subjected to ICS using serially truncated (B), and overlapping 9-mer (C) peptides. IFN-γ (green) and TNF-α (orange) production and surface expression of CD107a (pink) served as the read-out. Arrows next to an amino acid indicate the peptide-terminal amino acid residue required for efficient peptide recognition. (D) 721.221 and C1R cells expressing individual HLA class I alleles of volunteer 415 were utilized to determine the HLA restriction of 'parental' peptide HC081. (PDF)

S7 Fig. HC088 GSPAIFQSSMTKILE (Pol)-Definition of CD8 + T-cell determinants. (A)
The box. 15-mer peptide HC088 was recognized by volunteer 409 of the shown HLA type. Optimal peptides and determined HLA restriction are summarized below. Volunteer's lymphocytes were expanded by stimulation with peptide HC088 for 10 days to establish an STCL, which was subjected to ICS using serially truncated (B), and overlapping 9-mer (C) peptides monitoring IFN-γ (green) and TNF-α (orange) production and surface expression of CD107a (pink). In (B), arrows next to an amino acid indicate the peptide-terminal amino acid residue required for efficient peptide recognition. (D) 721.221 cells expressing the HLA-B Ã 07:02 and HLA-B Ã 08:01 alleles were used to determine the HLA restriction of peptide SPAIFQSSMTK. (PDF)

S8 Fig. HC088 GSPAIFQSSMTKILE (Pol)-Definition of CD8 + T-cell determinants. (A)
The box. 15-mer peptide HC088 was recognized by volunteer 421 of the indicated HLA type. Optimal peptides and determined HLA restriction are shown below. Volunteer's lymphocytes were expanded by stimulation with peptide HC088 for 10 days to establish an STCL, which was subjected to ICS using serially truncated (B), and overlapping 9-mer (C) peptides monitoring IFN-γ (green) and TNF-α (orange) production and surface expression of CD107a (pink). In (B), arrows next to an amino acid indicate the peptide-terminal amino acid residue required for efficient peptide recognition. were recognized by volunteer 410 of the indicated HLA type. Optimal peptides and determined HLA restriction are shown. Volunteer's lymphocytes were expanded by stimulation with peptide 'parental peptides for 10 days to establish STCLs, which were subjected to ICS using serially truncated (B), and overlapping 9-mer (C) peptides monitoring IFN-γ (green) and TNF-α (orange) production and surface expression of CD107a (pink). In (B), arrows next to an amino acid indicate the peptide-terminal amino acid residue required for efficient S15 Fig. HC135 KLVSQGIRKVLFLDG (Pol)-Definition of CD8 + T-cell determinants. (A) The box. 15-mer peptide HC135 was recognized by volunteers 416 and 418 of the indicated HLA types. Optimal peptides and determined HLA restriction are shown. (B) Volunteers' lymphocytes were expanded by stimulation with 'parental' peptide for 10 days to establish STCLs, which were subjected to ICS using serially truncated peptides monitoring IFN-γ (green) and TNF-α (orange) production and surface expression of CD107a (pink). Arrows next to an amino acid indicate the peptide-terminal amino acid residue required for efficient peptide recognition. (C) The same SCTLs from volunteers 416 (left) and 418 (right) were tested for recognition of overlapping 9-mer peptides. 721.221 and C1R cells expressing HLA alleles of volunteers 416 (D) and 417 (E) were used to determine the HLA restriction of peptide KLVSQGIRKV. (PDF) S16 Fig. HC139 DKAQ-AKEIVASCDKC (Pol)-definition of CD8 + T-cell determinants.
(A) The box. Peptide HC139 was recognized by volunteer 404 of the indicated HLA type and the optimal peptides are shown. '-' indicates junction between two adjacent HIVconsv regions. (B) Cryopreserved lymphocytes from the vaccine recipient were expanded by stimulation with 'parental' peptide for 10 days to establish STCL, which was subjected to ICS using serially truncated peptides. IFN-γ (green) and TNF-α (orange) production and surface expression of CD107a (pink) served as the read-out. Arrows next to an amino acid indicate the peptide-terminal amino acid residue required for efficient peptide recognition. (C) The same SCTLs as in (B) were tested for recognition of overlapping 9-mer peptides. (PDF)

S17 Fig. HC145 GQVDCSPGIWQLDCTH (Pol)-Definition of CD8 + T-cell determinants.
(A) The box. Peptide HC145 was recognized by volunteer 404 of the indicated HLA type, and the optimal peptide is shown. (B) Cryopreserved lymphocytes from the vaccine recipient were expanded by stimulation with the 'parental' peptide for 10 days to establish STCL, which was subjected to ICS using serially truncated peptides. IFN-γ (green) and TNF-α (orange) production and surface expression of CD107a (pink) served as the read-out. Arrows next to an amino acid indicate the peptide-terminal amino acid residue required for efficient peptide recognition. (C) The same SCTLs were tested for recognition of overlapping 9-mer peptides. (PDF) S18 Fig. HC164 VQMAVFIHNFKRKGGI (Pol)-Definition of CD8 + T-cell determinants. (A) The box. Peptide HC164 was recognized by volunteers 404 and 410 of the indicated HLA types, and the optimal peptide is shown. (B) Cryopreserved lymphocytes from vaccine recipients 404 (left) and 410 (right) were expanded by stimulation with the 'parental' peptide for 10 days to establish STCLs, which were tested for recognition of overlapping 9-mer peptides. (PDF) S19 Fig. HC176 VVPRRKAKIIRDYGK (Pol)-Definition of CD8 + T-cell determinants. (A) The box. Peptide HC176 was recognized by volunteer 413 of the indicated HLA type. Optimal peptides and their HLA restriction are shown. (B) Cryopreserved lymphocytes from the vaccine recipient were expanded by stimulation with 'parental' peptide for 10 days to establish STCL, which was subjected to ICS using serially truncated peptides. IFN-γ (green) and TNF-α (orange) production and surface expression of CD107a (pink) served as the read-out. Arrows next to an amino acid indicate the peptide-terminal amino acid residue required for efficient peptide recognition. (C) The same SCTLs as in (B) were tested for recognition of overlapping 9-mer peptides. (D) 721.221 cells expressing HLA-B Ã 08:01 were used to confirm binding peptides. (PDF) S1