Modification of MHC anchor residues generates heteroclitic peptides that alter TCR binding and T-cell recognition

Abstract

Improving T-cell antigens by altering MHC anchor residues is a common strategy used to enhance peptide vaccines but there has been little assessment of how such modifications affect TCR binding and T-cell recognition. Here, we use surface plasmon resonance and peptide-MHC tetramer binding at the cell surface to demonstrate that changes in primary peptide anchor residues can substantially and unpredictably alter TCR binding. We also demonstrate that the ability of TCRs to differentiate between natural and anchor-modified heteroclitic peptides distinguishes T-cells that exhibit a strong preference for either type of antigen. Furthermore, we show that anchor-modified heteroclitic peptides prime T-cells with different TCRs compared to those primed with natural antigen. Thus, vaccination with heteroclitic peptides may elicit T-cells that exhibit suboptimal recognition of the intended natural antigen and, consequently, impaired functional attributes in vivo. Heteroclitic peptide-based immune interventions therefore require careful evaluation to ensure efficacy in the clinic.

INTRODUCTION

αβ T-cells orchestrate adaptive immune responses and provide protection against pathogens and cellular malignancies by recognizing short peptide fragments complexed with MHC molecules on the cell surface (1, 2). At the molecular level, the specificity of T-cell recognition is governed by the tripartite interaction between the TCR and peptide-MHC (pMHC). Enhanced binding of TCR to pMHC, and mutations that stabilize the fastening of the antigenic peptide to its restricting MHC molecule, are both known to increase immunogenicity (3-11). Thus, a number of sequence-modified, or heteroclitic, peptides have been developed for this purpose. However, peptide alterations that improve TCR binding are very difficult to predict in the absence of detailed structural data as they are likely to be different for each TCR/pMHC pairing. In addition, there is a danger that substitution of TCR contact residues may induce the expansion of T-cells with unwanted or irrelevant specificities. In contrast, preferred MHC binding motifs are now widely known (12-15) and make the improvement of peptide binding to MHC molecules relatively straightforward. This approach is widely used in the field of cancer vaccination and could also have appeal in strategies designed to induce autoantigen-specific regulatory T-cell responses in the context of autoimmune diseases such as type 1 diabetes (T1D) and multiple sclerosis (16).

The generation of effective cancer vaccines is proving to be difficult as the majority of cancer antigens are derived from self-molecules (17). Thus, T-cells that respond strongly to these antigens are likely to be culled during thymic selection in order to maintain self-tolerance. In support of this concept, we have found that tumor-specific TCRs bind to their cognate pMHC antigens with affinities that are approximately 5-fold lower than those derived from pathogen-specific T-cells (18). The development of cancer vaccines has generally focused on MHC class I (MHCI)-restricted epitopes with the aim of generating CD8⁺ effector T-cells that directly kill tumor cells. To date, most efforts have targeted peptides restricted by HLA A*0201, the most common human MHCI molecule. However, the preferred peptide binding motif of HLA A*0201, xLxxxxxxV/L (14, 15), is not found in the majority of the most promising HLA A*0201-restricted cancer-derived epitopes (Table I). As a result, vaccine-oriented approaches, including clinical trials, have utilized peptides with altered MHCI anchor residues that enhance the stability of the pMHCI complex (3, 19, 20). A key assumption in this strategy is that peptide anchor residue modifications do not transmit major structural changes to the outward facing TCR recognition platform or otherwise alter T-cell recognition.

Table I.

List of current natural human cancer antigens and their related heteroclitic peptides. Anchor residue modifications are highlighted in bold and underlined

Tumor-associated protein antigen	HLA allele	Natural peptide	Heteroclitic peptide
Melan-A26-35	HLA A*0201	EAAGIGILTV	ELAGIGILTV(9)
NY-ESO-1157-165	HLA A*0201	SLLMWITQC	SLLMWITQL(3)
gp100209-217	HLA A*0201	ITDQVPFSV	ILDQVPFSV(6)
gp100280-288	HLA A*0201	YLEPGPVA	YLEPGPVV(6)
gp100154-162	HLA A*0201	KTWGQYWQV	KLWGQYWQV(6)
HER-1/neu369-377	HLA A*0201	KIFGSLAFL	KVFGSLAFV(4) KLFGSLAFV(4)
PSA178-187	HLA A*0201	VISNDVCAQV	VLSNDVCAQV(8)

In the present study, we show that substitution of amino acid side chains at primary anchor positions in the antigenic peptide can substantially affect TCR binding in ways that are difficult to predict. These data are pertinent in the light of recent findings by Speiser and colleagues, which show that anchor residue-modified heteroclitic peptides can be less effective that natural peptides for the purposes of vaccination (21). Our results have important implications for the use of heteroclitic peptides and demonstrate the need for careful evaluation of this strategy on an individual basis prior to clinical application.

MATERIALS AND METHODS

Generation of CD8⁺ T-cell clones and primed lines

CD8⁺ T-cell clones were generated as described previously (5). The HLA A*0201-restricted Melan-A/MART-1_26-35 ELAGIGILTV-specific MEL5 clone and the HLA A*0201-restricted preproinsulin_15-24 ALWGPDPAAA-specific 1E6 clone were described previously (22-24). PBMCs from an HLA A*0201⁺ donor were primed with either ELAGIGILTV or EAAGIGILTV as described previously (25).

Recognition screen and combinatorial peptide library (CPL) scan

For the recognition screen, 2×10³ 1E6 CD8⁺ T-cells were cultured in triplicate for 16 h with position 2 (p2) variants of the wildtype preproinsulin_15-24 peptide (ALWGPDPAAA) at a concentration of 1μg/ml. Supernatant was harvested and analyzed for TNFα production by ELISA according to the manufacturer’s instructions (Invitrogen, Carlsbad, CA, USA). The combinatorial peptide library (CPL) comprised a total of 9.36×10¹² ((10+19)×19⁹) different decamer peptides divided into 200 different peptide mixtures. For the CPL scan, 1E6 CD8⁺ T-cells were washed and rested overnight in RPMI 1640 containing 100 U/mL penicillin, 100 μg/mL streptomycin, 2 mM L-glutamine and 2% heat inactivated FCS (Gibco, Invitrogen, Carlsbad, CA, USA) (R2 medium). In 96-well U-bottom plates, 6×10⁴ C1R cells stably transfected with HLA A*0201 were pulsed in duplicate with various peptide library mixtures (100 μg/mL) at 37°C for 2 h. Following peptide pulsing, 3×10⁴ 1E6 CD8⁺ T-cells were added and the assay was incubated overnight at 37°C. Subsequently, the supernatant was harvested and assayed for MIP1β by ELISA according to the manufacturer’s instructions (R&D Systems, Minneapolis, MN, USA).

Generation of expression plasmids

The MEL5, MEL187.c5 and 1E6 TCRs were generated from the corresponding CD8⁺ T-cell clones. All sequences were confirmed by automated DNA sequencing (Lark Technologies, Essex, UK). A disulphide-linked construct was used to produce the soluble domains (variable and constant) for both the TCR α and TCR β chains (26, 27). The HLA A*0201 α chain, tagged with a biotinylation sequence, and β2m were also cloned and used to produce the pMHCI molecules. The MEL5, MEL187.c5 and 1E6 TCR α and TCR β chains, the HLA A*0201 α chain and β2m sequences were inserted into separate pGMT7 expression plasmids under the control of the T7 promoter (26).

Protein expression, refolding and purification

Competent Rosetta DE3 E.coli cells were used to produce the MEL5, MEL187.c5 and 1E6 TCR α and TCR β chains, the HLA A*0201 α chain and β2m in the form of inclusion bodies using 0.5 mM IPTG to induce expression as described previously (26). For a 1 L TCR refold, 30 mg of TCR α chain was incubated at 37°C for 15 mins with 10 mM DTT and added to cold refold buffer (50 mM TRIS pH 8.1, 2 mM EDTA, 2.5 M urea, 6 mM cysteamine hydrochloride and 4 mM cystamine). After 15 mins, 30 mg of TCR β chain, also incubated at 37°C for 15 mins with 10 mM DTT, was added. For a 1 L pMHCI refold, 30 mg of HLA A*0201 α chain was mixed with 30 mg of β2m and 4 mg of peptide at 37°C for 15 mins with 10 mM DTT. The following peptides were used in separate refolds: EAAGIGILTV, ELAGIGILTV, ALWGPDPAAA and AQWGPDPAAA (heteroclitic changes are denoted in bold and underlined). This mixture was then added to cold refold buffer (50 mM TRIS pH 8, 2 mM EDTA, 400 mM L-arginine, 6 mM cysteamine hydrochloride and 4 mM cystamine). Refolds were mixed at 4°C for >1 hour. Dialysis was carried out against 10 mM TRIS pH 8.1 until the conductivity of the refolds was <2 mS/cm. The refolds were then filtered, ready for purification steps. Refolded proteins were purified initially by ion exchange using a Poros50HQ™ column (GE Healthcare, UK) and then gel filtered into BIAcore buffer (10 mM HEPES pH 7.4, 150 mM NaCl, 3 mM EDTA and 0.05% (v/v) Surfactant P20) using a Superdex200HR™ column (GE Healthcare, UK). Protein quality was analyzed by Coomassie-stained SDS-PAGE. Biotinylated pMHCI was prepared as described previously (28).

Surface plasmon resonance (SPR) analysis

Binding analysis was performed using a BIAcore T3000™ (GE Healthcare, UK) equipped with a CM5 sensor chip as reported previously (28). Between 200 and 400 response units (RUs) of biotinylated pMHCI was immobilized on streptavidin, which was chemically linked to the chip surface; pMHCI was injected at a slow flow rate (10 μL/min) to ensure uniform distribution on the chip surface. Combined with the small amount of pMHCI bound to the chip surface, this reduced the likelihood of off-rate limiting mass transfer effects. The MEL5, MEL187.c5 and 1E6 TCRs were purified and concentrated to ~100 μM on the day of SPR analysis to minimize TCR aggregation. For equilibrium analysis, eight serial dilutions were carefully prepared in triplicate for each sample and injected over the relevant sensor chip at a flow rate of 45 μL/min at 25°C. Results were analyzed using BIAevaluation 3.1™, Microsoft Excel™ and Origin 6.1™. The equilibrium binding constant (K_D) values were calculated using a nonlinear curve fit (y = (P₁x)/(P₂ + x)).

pMHCI stability assays

For SPR-based pMHCI stability assays, ~1000 RUs of biotinylated pMHCI was immobilized on streptavidin, which was chemically linked to the chip surface. The RUs on each flow cell were monitored in real time over 4,000 sec. Reductions in mass were analyzed using BIAevaluation 3.1™ to determine the stability of each pMHCI complex. For T2 cell-based pMHCI binding assays, the LCLxT-lymphoblastoid hybrid cell line .174×CEM.T2 was used. These cells, referred to as T2 cells (29), lack the transporter associated with antigen processing (TAP); thus, addition of exogenous binding peptide is required for stable expression of HLA A*0201 molecules on the cell surface. 10⁶ T2 cells per test were incubated in AIM V serum-free medium (Invitrogen, Carlsbad, CA, USA) with various concentrations (0.01, 0.1, 1, 10, and 100 μM) of peptide at 26°C for 14–16 h, then at 37°C for 2 h prior to staining for HLA A*0201 surface expression with the PE-labeled monoclonal antibody (mAb) BB7.2 (BD Biosciences, San Jose, CA, USA). Duplicate samples for each condition were acquired using a BD FACSCantoII flow cytometer (BD Biosciences Immunocytometry Systems, San Jose, CA, USA). Data were analyzed with BD FACSDiva software.

Cellular assays

The production of IFNγ, TNFα and IL2 by MEL5, MEL187.c5 and 1E6 CD8⁺ T-cells was measured using a Cytometric Bead Array Th1/Th2 Assay Kit (BD Biosciences, San Jose, CA, USA); degranulation was quantified using a flow cytometric assay to detect CD107a mobilization (30). In all experiments, 3×10⁴ HLA A*0201-transfected C1R cells, pulsed with peptide at concentrations in the range of 10⁻⁴ to 10⁻¹¹ M for 60 mins at 37°C in 50μl R2 medium, were mixed with 3×10⁴ clonal CD8⁺ T-cells in 50μl R2 medium. The cells were incubated for 18 hrs at 37°C before analysis. Tetramer binding analysis was conducted using 5×10⁴ clonal CD8⁺ T-cells stained in minimal volume with 0.1 μg of the corresponding pMHCI component at 37°C for 30 min (reviewed in (31)); cells were then washed with PBS and analyzed by flow cytometry. Tetramer association and dissociation assays were performed as described previously (32).

Flow cytometric sorting and molecular analysis of CD8⁺ T-cell clonotypes

PBMCs primed in vitro with either ELAGIGILTV or EAAGIGILTV peptide were stained first at room temperature for 20 min with live-dead fixable Aqua (Invitrogen, Carlsbad, CA, USA), then at 37°C for 15 min with 1 μg of the corresponding HLA A*0201/ELAGIGILTV or HLA A*0201/EAAGIGILTV tetramer conjugated to allophycocyanin. Cells were then stained with anti-CD3-PECy7 and anti-CD8-Alexa Fluor 700 mAbs (both BD Biosciences Pharmingen, San Diego, CA, USA) on ice for 20 min. Viable CD3⁺CD8⁺tetramer⁺ cells (median: 5,000; range: 2,781-5,000) were sorted at >98% purity into 100 μL Ambion® RNAlater (Applied Biosystems Inc., Foster City, CA, USA) using a custom-modified BD FACSAriaII flow cytometer (BD Biosciences Immunocytometry Systems, San Jose, CA, USA). Molecular analysis of all expressed TRB gene rearrangements was then performed as described previously (33). The IMGT nomenclature is used throughout this report (34).

RESULTS

Mutation at a primary MHCI anchor residue can substantially alter T-cell function

In this study, we set out to examine the effect of MHC anchor residue changes on T-cell recognition. We reasoned that the examination of such outcomes would be best explored initially in the context of a low affinity TCR/pMHCI interaction, which might facilitate the measurement of subtle effects. Autoimmune disease provides such a context and we therefore focused on the 1E6 CD8⁺ T-cell clone, which is specific for the HLA A*0201-restricted human preproinsulin_15-24 peptide ALWGPDPAAA. This clone was recently derived from the peripheral blood of a patient with T1D; a clone with an identical TCR was subsequently isolated from the same patient one year later, thereby demonstrating the long-term persistence in vivo of T-cells with this specificity (22). In a recognition screen using peptides with all possible proteogenic amino acids at position 2 (p2), which is a primary anchor residue for peptides that bind HLA A*0201, glutamine (Q) was found to be the optimal residue for the induction of IE6 T-cell activation (Figure 1A). This result was unexpected because Q is known to be a suboptimal anchor for HLA A*0201 at this position (14, 15). We confirmed that Q was the preferred residue for IE6 T-cell recognition at this position using a p2 CPL scan (Figure 1B). Furthermore, similar effects were observed in assays with different functional readouts. Thus, the AQWGPDPAAA variant peptide induced greater secretion of MIP1β, IFNγ, TNFα and IL2 (Figure 2A-D) and enhanced T-cell degranulation (Figure 2E) at low peptide concentrations. The difference in activation was most marked with IL2 production (5); AQWGPDPAAA induced IL2 production from 1E6 T-cells at concentrations of exogenously applied peptide that were 100-fold lower compared to the 2L wildtype peptide (Figure 2D). This unexpected finding, that substitution with a suboptimal anchor residue at a primary MHCI binding site generated a substantially improved T-cell agonist peptide, warranted further investigation.

Figure 1.

(A) Recognition of all p2 variants in the preproinsulin_15-24 peptide sequence by the 1E6 CD8⁺ T-cell clone. 2×10³ IE6 T-cells were cultured in triplicate for 16 h with p2 variants of the preproinsulin_15-24 peptide at a concentration of 1μg/mL. Supernatants were then harvested and assayed for TNFα production by ELISA. Results are representative of three independent assays. (B) A p2 combinatorial peptide library screen. Each 10-mer peptide pool was defined by a fixed amino acid at position 2 and a random equimolar mixture of all 20 proteogenic L-amino acids except cysteine at the remaining 9 positions (O₁X₉). Cysteine was included at the fixed position (O) but omitted from the degenerate positions (X) in order to limit the problem of oxidation. In each 18 h assay, 3×10⁴ 1E6 CD8⁺ T-cells were exposed in the presence of individual peptide mixtures to 6×10⁴ C1R target cells stably transfected with HLA A*0201. Supernatants were then harvested and assayed for MIP1β production by ELISA. In each case, the star indicates the wildtype p2 leucine residue.

Figure 2.

Activation of the 1E6 CD8⁺ T-cell clone in response to wildtype ALWGPDPAAA (ALW) and heteroclitic AQWGPDPAAA (AQW) peptides. In all experiments, 3×10⁴ HLA A*0201-transfected C1R cells, pulsed with peptide at concentrations in the range of 10⁻⁴ to 10⁻¹¹ M for 60 mins at 37°C in 50μl R2 medium, were mixed with 3×10⁴ 1E6 CD8⁺ T-cells in 50μl R2 medium. The cells were incubated for 18 hrs at 37°C before analysis. (A) MIP1β ELISA assay. Error bars represent one standard deviation and, in most cases, are smaller than the plot symbols. (B-D) Cytometric bead array showing IFNγ, TNFα and IL2 production. (E) Degranulation (CD107a) assay.

Mutation at a primary MHCI anchor residue can substantially alter TCR binding

Analysis of pHLA A*0201 stability on the surface of T2 cells showed that the AQWGPDPAAA (2Q) variant exhibited suboptimal binding compared to ALWGPDPAAA (Figure 3, inset). In order to examine this effect in more detail, we developed a sensitive SPR-based assay to examine pMHCI stability in real time. Biotinylated HLA A*0201/ALWGPDPAAA and HLA A*0201/AQWGPDPAAA were immobilized on the surface of separate BIAcore™ CM5 chips using streptavidin-biotin coupling. A reduction in RUs over time was observed, corresponding to a reduction in mass at the chip surface (Figure 3). The loss of peptide from pMHCI complexes cannot wholly account for this RU reduction due to the relatively small mass of the peptide (~1 kDa). Instead, we reasoned that the majority of this mass reduction was due to the dissociation of β2m from the MHCI heavy chain caused by destabilization of the pMHCI complex upon loss of the antigenic peptide; thus, this decay process represents the half-life of peptide binding. Consistent with this hypothesis, the ALWGPDPAAA peptide, which contains a preferred HLA A*0201 binding residue at p2, displayed a slower dissociation rate compared to the AQWGPDPAAA peptide, which incorporates a suboptimal p2 anchor residue (Figure 3). This new assay showed that the half-life of the wildtype HLA A*0201/ALWGPDPAAA complex was 4.8x longer than the corresponding HLA A*0201/AQWGPDPAAA complex (Figure 3). Thus, the 2Q peptide elicited substantially better activation of IE6 T-cells despite reduced MHCI binding.

Figure 3.

Real-time pMHCI stability assay comparing the binding of ALWGPDPAAA (ALW) and AQWGPDPAAA (AQW) peptides to HLA A*0201 using SPR. Approximately 1000 RUs of each pMHCI was attached on to each flow cell via a biotin-streptavidin linker. The assay was monitored over 4,000 seconds at 25°C. The reduction in RUs over time reflects dissociation of β2m from the pMHCI heavy chain caused by de-stabilization of the complex upon loss of the antigenic peptide. HLA A*0201/ALWGPDPAAA (half-life ≈ 87 h) dissociated more slowly than HLA A*0201/AQWGPDPAAA (half-life ≈ 18 h). Inset: cellular pMHCI stability assay. Live T2 cells were pulsed with 10 μM peptide and pHLA A*0201 was quantified on the cell surface by flow cytometry using a specific PE-labeled mAb. Exogenous application of ALWGPDPAAA (ALW) exhibited superior pMHCI surface stabilization compared to AQWGPDPAAA (AQW). These data show that the wildtype preproinsulin_15-24 peptide with leucine at p2 forms a more stable pHLA-A*0201 complex than the corresponding 2Q peptide.

We next examined TCR/pMHCI affinity to investigate the molecular basis for the enhanced agonist properties of the AQWGPDPAAA peptide. Soluble 1E6 TCR bound to HLA A*0201/AQWGPDPAAA with a substantially greater affinity (K_D ≈ 8.4×10⁻⁵ M) compared to HLA A*0201/ALWGPDPAAA (K_D ≈ 2.7×10⁻⁴ M) in SPR experiments (Figure 4A&B). These data clearly show that anchor residue modification can have a profound effect on TCR/pMHCI binding, even though such buried amino acids do not usually contact the TCR during antigen recognition. In addition, we examined ligand binding at the T-cell surface using pMHCI tetramers (31, 32). The HLA A*0201/AQWGPDPAAA tetramer stained 1E6 T-cells with greater intensity than the HLA A*0201/ALWGPDPAAA tetramer (Figure 5A). Furthermore, kinetic analysis showed that HLA A*0201/AQWGPDPAAA tetramer binding was characterized by slower dissociation and faster association rates compared to the HLA A*0201/ALWGPDPAAA tetramer (Figure 4C&D; Table II). Thus, mutation at a primary MHCI anchor residue can substantially alter TCR binding.

Figure 4.

SPR equilibrium binding of soluble 1E6 TCR to (A) HLA A*0201/ALWGPDPAAA (A2-ALW) and (B) HLA A*0201/AQWGPDPAAA (A2-AQW). The mean response for each concentration is plotted (n = 2). The equilibrium dissociation constant (K_D) values were calculated assuming 1:1 Langmuir binding and were plotted using a nonlinear curve fit (y = (P₁x)/(P₂ + x)). These data show that TCR binding affinity is directly related to the sensitivity of a T-cell for cognate antigen. (C-D) HLA A*0201/ALWGPDPAAA (ALW) and HLA A*0201/AQWGPDPAAA (AQW) real-time tetramer binding kinetics. Tetramer association and dissociation rates were calculated as described previously (32).

Figure 5.

Wildtype and heteroclitic pMHCI tetramer staining of IE6, MEL5 and MEL187.c5 CD8⁺ T-cell clones. (A) HLA A*0201/ALWGPDPAAA (ALW) and HLA A*0201/AQWGPDPAAA (AQW) tetramer staining of the IE6 CD8⁺ T-cell clone. (B) HLA A*0201/EAAGIGILTV (EAA) and HLA A*0201/ELAGIGILTV (ELA) tetramer staining of the MEL5 CD8⁺ T-cell clone. (C) HLA A*0201/EAAGIGILTV (EAA) and HLA A*0201/ELAGIGILTV (ELA) tetramer staining of the MEL187.c5 CD8⁺ T-cell clone. These data show that tetramer staining avidity is directly related to monomeric TCR/pMHCI affinity.

Table II.

Tetramer dissociation kinetics

CTL clone	HLA A*0201 tetramer	Tetramer decay half-life (mins)
1E6	ALWGPDPAAA	0.83
	AQWGPDPAAA	3.07
MEL5	ELAGIGILTV	0.94
	EAAGIGILTV	1.78
MEL187.c5	ELAGIGILTV	2.16
	EAAGIGILTV	1.67

Stability and specificity investigations in the clinically relevant Melan-A/MART-1_26-35 system

To extend our findings to a clinically relevant system, we examined CD8⁺ T-cell responses to the HLA A*0201-restricted Melan-A/MART-1_26-35 peptide EAAGIGILTV, which is the most widely studied tumor-associated epitope to date (9, 35-37). The ELAGIGILTV (2L) heteroclitic variant of the Melan-A/MART-1_26-35 peptide, which contains the preferred HLA A*0201 motif and binds with greater stability compared to EAAGIGILTV, has been widely adopted in this system because it induces far greater CD8⁺ T-cell expansion than the natural peptide (21). The 2L heteroclitic peptide is also used exclusively in Melan-A/MART-1_26-35 pMHCI multimers due to their greater stability. We confirmed the enhanced stability of HLA A*0201/ELAGIGILTV in a T2 binding assay (Figure 6, inset). This result was further confirmed using our novel SPR pMHCI stability assay, which showed that the half-life of HLA A*0201/ELAGIGILTV was 5.3x longer than the corresponding HLA A*0201/EAAGIGILTV complex (Figure 6).

Figure 6.

Real-time pMHCI stability assay comparing the binding of EAAGIGILTV (EAA) and ELAGIGILTV (ELA) peptides to HLA A*0201 using SPR. The anchor residue-modified heteroclitic peptide complex HLA A*0201/ELAGIGILTV dissociated more slowly (half-life ≈ 32h) than the natural cancer peptide complex HLA A*0201/EAAGIGILTV (half-life ≈ 6h). Inset: cellular pMHCI stability assay. Assays were performed as described in the legend to Figure 3.

We next examined recognition of the wildtype and 2L heteroclitic variant Melan-A/MART-1_26-35 epitopes by two distinct CD8⁺ T-cell clones, MEL5 and MEL187.c5, using five distinct functional readouts (Figure 7). The MEL187.c5 clone responded more strongly to HLA A*0201/ELAGIGILTV and did not recognize the wildtype EAAGIGILTV sequence at exogenous peptide concentrations <1 μM. Conversely, the MEL5 clone recognized both peptides well with a slight preference for the natural sequence, even though this peptide exhibits substantially weaker binding to HLA A*0201. On this basis, we reasoned that the distinct TCRs from these CD8⁺ T-cell clones bound with different affinities to the two pHLA A*0201 antigens. To test this hypothesis, we produced soluble versions of the MEL5 and MEL187.c5 TCRs and examined their binding properties using SPR. As predicted, the two TCRs exhibited differential binding to the HLA A*0201/EAAGIGILTV and HLA A*0201/ELAGIGILTV antigens. The MEL5 TCR bound to the natural HLA A*0201/EAAGIGILTV (K_D ≈ 6.4×10⁻⁶ M) antigen with an affinity >2.5x that of the HLA A*0201/ELAGIGILTV (K_D ≈ 1.7×10⁻⁵ M) heteroclitic complex (Figure 8A&B). Conversely, the MEL187.c5 TCR bound HLA A*0201/ELAGIGILTV (K_D ≈ 1.8×10⁻⁵ M) with an affinity more than twice that of the corresponding HLA A*0201/EAAGIGILTV (K_D ≈ 4.2×10⁻⁵ M) antigen (Figure 9A&B). To confirm this preference at the cell surface of MEL5 and MEL187.c5 T-cells, we conducted pMHCI tetramer staining studies. The MEL5 T-cell clone stained more avidly with the natural HLA A*0201/EAAGIGILTV tetramer (Figure 5B). In contrast, the MEL187.c5 T-cell clone exhibited substantially better staining with the heteroclitic HLA A*0201/ELAGIGILTV tetramer (Figure 5C). These differences were reflected in pMHCI tetramer kinetic studies (Figures 8C-D&9C-D; Table II). Thus, TCRs specific for Melan-A/MART-1_26-35 can distinguish between the natural HLA A*0201/EAAGIGILTV antigen and the widely used heteroclitic HLA A*0201/ELAGIGILTV complex.

Figure 7.

Activation of the MEL5 and MEL187.c5 CD8⁺ T-cell clones in response to wildtype EAAGIGILTV (EAA) and heteroclitic ELAGIGILTV (ELA) peptides. In all experiments, 3×10⁴ HLA A*0201-transfected C1R cells, pulsed with peptide at concentrations in the range of 10⁻⁴ to 10⁻¹¹ M for 60 mins at 37°C in 50μl R2 medium, were mixed with 3×10⁴ MEL5 or MEL187.c5 CD8⁺ T-cells in 50μl R2 medium. The cells were incubated for 18 hrs at 37°C before analysis. (A, F) MIP1β ELISA assay. Error bars represent one standard deviation and, in most cases, are smaller than the plot symbols. (B-D, G-I) Cytometric bead array showing IFNγ, TNFα and IL2 production. (E, J) Degranulation (CD107a) assay.

Figure 8.

SPR equilibrium binding of the MEL5 TCR to (A) HLA A*0201/ELAGIGILTV (A2-ELA) and (B) HLA A*0201/EAAGIGILTV (A2-EAA). The equilibrium dissociation constant (K_D) values were calculated as in Figure 4. (C-D) HLA A*0201/EAAGIGILTV (EAA) and HLA A*0201/ELAGIGILTV (ELA) real-time tetramer binding kinetics. Tetramer association and dissociation rates were calculated as described previously (32).

Figure 9.

SPR equilibrium binding of the MEL187.c5 TCR to (A) HLA A*0201/ELAGIGILTV (A2-ELA) and (B) HLA A*0201/EAAGIGILTV (A2-EAA). The equilibrium dissociation constant (K_D) values were calculated as in Figure 4. (C-D) HLA A*0201/EAAGIGILTV (EAA) and HLA A*0201/ELAGIGILTV (ELA) real-time tetramer binding kinetics. Tetramer association and dissociation rates were calculated as described previously (32).

Different CD8⁺ T-cell repertoires emerge following priming with wildtype and heteroclitic peptide antigens

The differential binding of TCRs to HLA A*0201/EAAGIGILTV and HLA A*0201/ELAGIGILTV raises questions about the validity of using Melan-A/MART-1_26-35 heteroclitic peptides in vaccine trials (35, 37-39). Specifically, this binding disparity suggests that the heteroclitic ELAGIGILTV peptide might prime T-cells with different constituent clonotypes that exhibit suboptimal recognition of the natural antigen presented on the tumor cell surface. To examine this possibility, we dissected the clonotypic composition of CD8⁺ T-cell populations specific for HLA A*0201/EAAGIGILTV and HLA A*0201/ELAGIGILTV after in vitro priming of PBMC with the respective peptides. In all cases, antigen-specific CD8⁺ T-cells primed with either EAAGIGILTV or ELAGIGILTV were identified using the corresponding pHLA A*0201 tetramers and sorted by flow cytometry to >98% purity; a quantitative molecular analysis of all expressed TRB gene products was then conducted using a template-switch anchored RT-PCR (33). To minimize the effect of stochastic processes during in vitro priming, we pooled all TRB gene transcripts specific for either HLA A*0201/EAAGIGILTV (260 sequences) or HLA A*0201/ELAGIGILTV (141 sequences) derived from one HLA A*0201⁺ donor at different time points and weighted each distinct sequence equally regardless of frequency (Figures 10&11). A common bias towards TRBV4-1, TRBV6-2/6-3, TRBV6-5, TRBV27 and TRBV28 was observed in the CD8⁺ T-cell repertoires primed by both HLA A*0201/EAAGIGILTV and HLA A*0201/ELAGIGILTV. In contrast, TRBJ gene usage was more diverse in both repertoires, although HLA A*0201/EAAGIGILTV-specific clonotypes exhibited preferential usage of TRBJ1-5. Glycine residues were well represented within the CDR3 sequences, but no prevalent motifs were apparent. However, HLA A*0201/EAAGIGILTV-specific clonotypes tended to use shorter CDR3s compared to HLA A*0201/ELAGIGILTV-specific clonotypes (Figures 10&11). Notably, only 15 clonotypes were shared between the two concatenated datasets; in contrast, there were 62 distinct clonotypes within the HLA A*0201/EAAGIGILTV-primed repertoire and 39 distinct clonotypes within the HLA A*0201/ELAGIGILTV-primed repertoire. These data indicate that differences in T-cell fine specificity affect clonotype selection, at least in vitro, and are consistent with the biophysical and functional data derived from different CD8⁺ T-cell clones specific for Melan-A/MART-1_26-35 (Figures 7-9).

Figure 10.

Concatenated molecular analysis of CD8⁺ T-cell clonotypes specific for the Melan-A/MART-1_26-35 antigen. PBMCs from an HLA A*0201⁺ donor were primed at two different time points with (A) EAAGIGILTV and (B) ELAGIGILTV peptides. Each distinct sequence for each specificity was counted once, irrespective of clonal frequency, to minimize bias arising from preferential clonotypic expansions in vitro. Panels show TRBV and TRBJ gene usage, and CDR3 amino acid sequence. Clonotypes that are common to both repertoires are color-coded.

Figure 11.

The frequency of (A) TRBV gene expression, (B) TRBJ gene expression, and (C) CDR3 length within CD8⁺ T-cell populations specific for HLA A*0201/EAAGIGILTV and HLA A*0201/ELAGIGILTV amplified in vitro from an individual donor. Each distinct clonotype was counted once to exclude bias arising from preferential clonotype expansions.

DISCUSSION

There are currently a number of anchor residue-modified peptides undergoing clinical assessment in cancer vaccine trials; these heteroclitic peptides target Melan-A/MART-1 (9), NY-ESO-1 (3), gp100 (6), HER-1/neu (4) and PSA (8). The purpose of peptide modification at anchor residues is to enhance the stability of MHC binding and thereby increase immunogenicity, with the inherent assumption that responding T-cells will maintain specificity for the wildtype antigen. However, although there is evidence that such heteroclitic peptides can induce stronger T-cell responses compared to their wildtype counterparts, the induced T-cell expansions do not appear to mount robust tumor-specific responses (40-42). To shed light on this dilemma, we report here the first in-depth investigation into the effects of peptide anchor residue modification on TCR specificity.

Commencing our study with the HLA-A*0201/preproinsulin_15-24 epitope, one of very few defined TCR/pMHCI systems in autoimmune disease, we discovered that a cognate CD8⁺ T-cell clone from a patient with T1D activated more readily in response to an anchor residue-modified peptide containing a L2Q substitution (Figures 1&2). This result was unexpected, given that Q is not a preferred HLA A*0201 binding residue at p2 (consensus sequence: xLxxxxxxV/L) (14, 15). Indeed, using standard techniques, we confirmed that AQWGPDPAAA bound to HLA A*0201 with reduced affinity compared to the natural ALWGPDPAAA peptide (Figure 3). Furthermore, we developed a sensitive SPR-based assay to examine pMHCI complex stability in real time; using this approach, we found that the L2Q substitution reduced the pHLA A*0201 complex half-life from ~87 h to ~18 h. The finding that a cognate CD8⁺ T-cell clone preferred the modified HLA A*0201/AQWGPDPAAA antigen, despite the substantial reduction in peptide binding and complex stability, suggested that the clonotypic TCR might bind the heteroclitic form with greater affinity compared to the wildtype HLA A*0201/ALWGPDPAAA molecule. Indeed, soluble IE6 TCR bound HLA A*0201/AQWGPDPAAA with a substantially greater affinity (K_D ≈ 8.4×10⁻⁵ M) compared to HLA A*0201/ALWGPDPAAA (K_D ≈ 2.7×10⁻⁴ M) in SPR equilibrium binding experiments (Figure 4A&B). This difference in TCR binding was also evident at the cell surface in experiments with pMHCI tetramers (Figure 4C&D; Figure 5A). Thus, we concluded that amino acid substitution at a primary HLA A*0201 anchor residue (p2) (43, 44) could enhance T-cell recognition and TCR binding despite causing a substantial reduction in peptide affinity for MHCI. This striking observation led us to consider that the converse scenario might also occur, a proposition that could have important implications for the use of heteroclitic peptide-based vaccines in cancer immunotherapy.

The ELAGIGILTV heteroclitic variant of Melan-A/MART-1_26-35 is known to bind HLA A*0201 with higher affinity than the wildtype EAAGIGILTV peptide (Figure 6) (45). This enhanced stability is thought to explain why HLA A*0201/ELAGIGILTV is a more potent T-cell agonist (21, 45), due to the fact that antigen density on the cell surface and the duration of stimulus will be substantially increased. Two distinct CD8⁺ T-cell clones specific for Melan-A/MART-1_26-35, MEL5 and MEL187.c5, were able to distinguish between the EAAGIGILTV and ELAGIGILTV peptides in a wide range of functional assays (Figure 7). Biophysical analysis showed that the MEL5 TCR bound HLA A*0201/EAAGIGILTV with greater affinity (>2.5-fold) than HLA A*0201/ELAGIGILTV (Figure 8A&B). This binding preference was also evident at the cell surface (Figure 8C&D; Figure 5B). Conversely, the MEL187.c5 TCR bound HLA A*0201/EAAGIGILTV with lower affinity (>2-fold) than HLA A*0201/ELAGIGILTV (Figure 9A&B). Again, this difference was apparent at the cell surface (Figure 9C&D; Figure 5C). Thus, distinct TCRs can recognize an anchor residue-modified heteroclitic peptide differently in the Melan-A/MART-1_26-35 system.

To date, all of our experiments using altered peptide ligands and/or mutated TCRs with defined biophysical binding properties ((5, 46) and unpublished) have shown that the most functionally sensitive T-cells express TCRs with the highest affinity for cognate antigen. Furthermore, we have shown that TCRs raised against tumor-specific peptides bind cognate pMHC antigen with affinities that are substantially weaker than those raised against pathogen-derived peptides (18). In the present study, it is noteworthy that MEL187.c5 TCR binding to both forms of the Melan-A/MART-1_26-35 antigen and MEL5 TCR binding to HLA A*0201/ELAGIGILTV occur within or just below the range reported for other MHCI-restricted tumor-specific TCRs (K_D ≈ 1.1×10⁻⁵ M to 3.4×10⁻⁵ M) (18). In contrast, the MEL5 TCR bound HLA A*0201/EAAGIGILTV with a substantially higher affinity (K_D ≈ 6.4×10⁻⁶ M), which falls within the range typically observed for MHCI-restricted pathogen-specific TCRs (18). Therefore, it is likely that CD8⁺ T-cells with identical or similar specificity to MEL5 represent more desirable targets for optimal adoptive therapy and vaccination trials compared to MEL187.c5-like T-cells.

Structural evidence suggests that substitution of alanine to leucine at the p2 anchor residue does not substantially alter Melan-A/MART-1_26-35 peptide conformation in the uncomplexed pHLA A*0201 molecule (47, 48). However a number of recent studies have shown that peptide conformation can be substantially altered during TCR docking (49-51). We have recently solved the crystal structure of the MEL5 TCR in complex with HLA A*0201/ELAGIGILTV (24). This structure shows that the MEL5 TCR α chain makes a number of electrostatic interactions with the N-terminus of the peptide, particularly residues 1E, 2L and 4G. The electrostatic interaction between the MEL5 TCR and the main chain of 2L is of particular note because it shows that TCR/anchor residue contacts can contribute to antigen specificity. It is likely that peptides with suboptimal anchor residues may be more flexible within the MHC binding groove, thereby enabling them to form subtly different conformational motifs that could have an effect on the fine specificity of different T-cell clones. In the case of the MEL5 TCR, increased flexibility in this region of the peptide may allow for stronger, or even new, TCR/peptide contacts that could explain the enhanced TCR affinity and preferential antigen sensitivity of MEL5 for HLA A*0201/EAAGIGILTV compared to HLA A*0201/ELAGIGILTV (Figures 7&8).

Current wisdom has resulted in the use of heteroclitic peptides for the majority of HLA A*0201-restricted tumor-associated epitopes (Table I). The differential binding of TCRs to such peptides, as demonstrated in this study, gives cause for concern as it may result in the priming of T-cell populations that recognize the natural antigen suboptimally. Indeed, we observed that different TCR repertoires were primed in vitro by the EAAGIGILTV and ELAGIGILTV peptides, even from the same PBMC sample under otherwise identical conditions (Figures 10&11). These differences likely reflect clonotypic preferences for either form of the Melan-A/MART-1_26-35 antigen. Our observations in this system are timely given recent elegant in vivo studies showing that unmodified self-antigen triggers human CD8⁺ T-cells with stronger tumor reactivity than the altered ELAGIGILTV antigen (21). These studies compared the CD8⁺ T-cell responses and clonotypic composition of Melan-A/MART-1_26-35–specific repertoires following peptide vaccination in melanoma patients (21, 52, 53). Thus, in these studies, inter-individual repertoires specific for either EAAGIGILTV or ELAGIGILTV were compared after in vivo priming. In contrast, we evaluated intra-individual clonotypic responses with the corresponding specificities after in vitro priming; this approach enables comparison based on samples derived from the same naïve T-cell pool. Overall, similar data emerge from all of these studies. Thus, irrespective of the priming peptide, we detected a clear bias towards TRBV4-1, TRBV6-2/6-3, TRBV6-5, TRBV27 and TRBV28 gene usage in the responding repertoires. Furthermore, glycine was commonly used within the CDR3 cores and the previously described “GXG” motif was observed in several of the responding clonotypes (52, 54). However, despite these preferences, the majority of clonotypes were non-overlapping between the two repertoires (Figure 10). Such subtle differences in the mobilized repertoires likely reflect differences in fine specificity at the level of TCR structure and function that could be influenced by two main factors (55). First, less stable pMHCI complexes will likely be present at lower concentrations on the cell surface, such that only CD8⁺ T-cell clones sensitive to low antigen doses will derive a sufficiently robust stimulus to undergo antigen-driven selection. Second, our data suggest that subtle structural variations between natural and heteroclitic peptides can result in two antigens that “look” very different to incoming T-cells. This is in contrast to the current assumption that anchor residues are “hidden” from view with respect to the TCR. Thus, pMHC stability and TCR affinity, which are both known to affect T-cell activation, can be influenced by altering peptide anchor residues (56, 57). In this respect, caution must be advised when vaccinating with structurally altered antigens because the amplified T-cell repertoire may be dissimilar to that induced with wildtype antigen; this could not only impair efficacy in vivo against the intended target, but could also lead to unknown cross-reactive specificities.

In summary, we show here that substitution of antigenic peptides to improve HLA A*0201 binding can alter TCR fine specificity with attendant functional consequences at the cellular level. The heteroclitic ELAGIGILTV peptide has been used in a number of clinical trials (37, 39), but these trials have been abandoned due to a low likelihood of success (58, 59). Our observation that TCRs can distinguish between the heteroclitic ELAGIGILTV peptide used in vaccination and the wildtype EAAGIGILTV sequence found on the surface of melanoma cells provides a biophysical reason for this failure. Indeed, our data synergize well with recent studies in which patients with melanoma were vaccinated using either the EAAGIGILTV or ELAGIGILTV peptides (21). These seminal studies revealed that, although analog vaccination was immunologically weaker than heteroclitic vaccination, the responding T-cell repertoires were therapeutically superior in terms of avidity and effector function (21). We therefore conclude that the use of anchor-modified, heteroclitic peptides requires careful re-evaluation to ensure that primed T-cells exhibit optimal specificity and sensitivity for the intended target.

Non-standard abbreviations

pMHC

peptide-major histocompatibility complex

RU

response unit

SPR

surface plasmon resonance

T1D

type 1 diabetes