Structure-based design of altered MHC II-restricted peptide ligands with heterogeneous immunogenicity

Abstract

Insights gained from characterizing MHC-peptide-TCR interactions have held the promise that directed structural modifications can have predictable functional consequences. The ability to manipulate T cell reactivity synthetically or through genetic engineering might thus be translated into new therapies for common diseases such as cancer and autoimmune disorders. In the current study, we determined the crystal structure of HLA-DR4 in complex with the non-mutated dominant gp100 epitope, gp100_44–59, associated with many melanomas. Altered peptide ligands (APLs) were designed to enhance MHC binding and hence T cell recognition of gp100 in HLA-DR4+ melanoma patients. Increased MHC binding of several APLs was observed, validating this approach biochemically. Nevertheless, heterogeneous preferences of CD4+ T cells from several HLA-DR4+ melanoma patients for different gp100 APLs suggested highly variable TCR usage, even among six patients who had been vaccinated against the wild type gp100 peptide. This heterogeneity prevented the selection of an APL candidate for developing an improved generic gp100 vaccine in melanoma. Our results are consistent with the idea that even conservative changes in MHC anchor residues may result in subtle, yet crucial, effects on peptide contacts with the TCR or on peptide dynamics, such that alterations intended to enhance immunogenicity may be unpredictable or counterproductive. They also underscore a critical knowledge gap that needs to be filled, before structural and in vitro observations can be used reliably to devise new immunotherapies for cancer and other disorders.

INTRODUCTION

Melanoma is an aggressive form of skin cancer which is curable in its early stages but carries a poor prognosis following distant organ metastasis (1). It is also highly immunogenic, as evidenced by endogenous anti-melanoma T and B cell responses and the susceptibility of melanoma to drugs with a purely immunological mode of action, such as interleukin-2 (2), anti-CTLA-4 (3), anti-PD-1 (4) and anti-PD-L1 (5). Efficient vaccination with tumor specific antigens can re-direct the anti-tumor immune response and provide synergistic treatment effects when combined with systemic immune-enhancing agents (6–8). Thus, there is a need to develop optimal cancer vaccines and tumor antigen-specific detection methods for monitoring treatment outcomes in vitro. Rational chemical modification of tumor specific peptide antigens to increase their immunogenicity, based on structural models, may facilitate this approach.

Gp100, a melanocyte lineage-specific transmembrane glycoprotein, is expressed in most melanomas and is involved in a multiple-step process of pigment production (9). Gp100 has been a widely-used target for melanoma immunotherapy since the demonstration that tumor infiltrating lymphocytes and circulating T cells from melanoma patients commonly recognize this antigen (10, 11). Despite the fact that the most gp100-directed melanoma therapies have focused on stimulating CD8⁺ T cell responses, CD4⁺ T cells play a central role in inducing and maintaining tumor specific CD8⁺ T cells (12). Devising immunotherapies which can efficiently raise specific CD4⁺ T cell responses is therefore an important goal.

A gp100 MHC II restricted peptide, gp100_44–59, was identified from HLA-DRB1*0401 (hereafter HLA-DR4) positive melanoma cell lines (13) and was subsequently validated as a dominant epitope in a transgenic animal model (14). This peptide can generate melanoma-specific CD4⁺ T cells from the peripheral blood of melanoma patients following repetitive in vitro stimulation (14, 15). Nevertheless, in a clinical trial using gp100_44–59 as a vaccine, no enhancement of gp100-specific reactivity was detected in the peripheral blood of patients following vaccination, dampening enthusiasm for its therapeutic potential (16).

Because gp100_44–59 is a non-mutated self antigen with intermediate binding affinity for HLA-DR4 (15), we hypothesized that altered peptide ligands (APLs) with single amino acid substitutions could be designed to confer higher MHC binding affinity and hence improved immunogenicity. Such APLs derived from gp100 MHC I-restricted epitopes have been employed as melanoma vaccines (17–20). Whereas unmodified HLA-A2-restricted gp100_209–217 and gp100_280–288 peptides induced melanoma-reactive CTLs from limited numbers of melanoma patients in vitro, and numerous re-stimulations were required, the APLs gp100_209–217(210M) and gp100_280–288(288V) with enhanced MHC affinity showed superior immunogenicity in vitro and in vivo (17). Similarly, in mice, a variant of gp100 that bound H-2D^b with increased affinity induced high frequencies of melanoma-specific CTLs in the endogenous CD8⁺ repertoire (21).

APLs based on MHC II-restricted epitopes have rarely been explored, since these peptides are heterogeneous in length and more degenerate in MHC binding specificity than class I-restricted peptides (22), making it difficult to precisely define MHC II-specific peptide binding motifs. However, combined information from MHC II-peptide crystal structures, ligand sequencing and binding affinity determinations has enriched our knowledge of the general chemical properties permitting optimal peptide binding to HLA-DR4. A dominant large hydrophobic or aromatic residue in the P1 binding position, a hydroxylated residue at P6, and a hydrophobic or polar residue at P9 appear to be favored (22–25).

In order to obtain precise information about the binding characteristics of gp100_44–59 to HLA-DR4, as a basis for designing optimal melanoma vaccines and immunomonitoring tools, we determined the crystal structure of this MHC-peptide complex. APLs based on structural data were compared to the wild type peptide for their ability to detect gp100-specific reactivity in melanoma patients vaccinated against the wild type peptide, or to raise melanoma-specific T cells from pre-vaccination peripheral blood mononuclear cells (PBMCs).

MATERIALS AND METHODS

Isolation and sequencing of native gp100 peptides complexed to HLA-DR4

Peptide-HLA-DR complexes were isolated from cultured 1102-mel melanoma cells on an anti-HLA-DR affinity column and peptides were eluted as described (26). Peptides were then analyzed by nanoflow HPLC-microelectrospray ionization coupled to either a hybrid linear quadrupole ion trap-Fourier-transform ion cyclotron resonance (LTQ-FT) mass spectrometer or an LTQ-Orbitrap mass spectrometer (Thermo Fisher Scientific) modified to perform electron transfer dissociation (ETD). Data were acquired as previously described (27). In brief, a precolumn loaded with 1e7 cell equivalents of MHC peptides was connected with polytetrafluoroethylene tubing (0.06-inch o.d. and 0.012-inch i.d.; Zeus Industrial Products) to the end of an analytical HPLC column (360-μm o.d. and 50 μm i.d.) containing 7 cm of C18 reverse-phase packing material (5-μm particles; YMC). Peptides were eluted into the mass spectrometer at a flow rate of 60 nL/min with a gradient: A = 0.1 M acetic acid (Sigma-Aldrich) in H₂O; B = 70% acetonitrile (Malinckrodt) and 0.1M acetic acid in H₂O; 0–60% B in 40 min, 60–100% B in 5 min. Parameters used to acquire ETD/MS/MS spectra in a data-dependent mode on the modified LTQ instrument have been described previously (28).

Sequence analysis was performed by searching MS/MS spectral data against a database consisting only of the gp100 protein using the Open Mass Spectrometry Search Algorithm (OMSSA) software to generate a list of candidate spectra (29). A precursor mass tolerance of ± .01 Da was used, MS/MS spectra were searched using a monoisotopic fragment ion mass tolerance of ± 0.5 Da. Data were searched allowing variable modifications for phosphorylation of serine, threonine and tyrosine residues and oxidation of methionine, with a total of 1024 variable modifications per peptide being allowed. Other parameters utilized were; peptide charge range from +2 to +5, +2 charge state products allowed for peptides of charge +3 and above, peptide size range 4–25 a.a. with no enzyme restriction. Peptide sequence assignments were validated by manual interpretation of the corresponding ETD and/or CAD MS/MS spectra. Approximate copy numbers per melanoma cell for each peptide were determined by comparing peak areas of the observed parent ions to that of angiotensin I and vasoactive intestinal peptide (DRVYIHPFHL HSDAVFTDNYTR, 100 fmol; Sigma-Aldrich) spiked into the sample mixture.

Synthesis of HLA-DR4-restricted peptides

Synthetic peptides used in this study were made with Fmoc chemistry, isolated by HPLC to ≥90% purity, and validated with mass spectrometry (Global Peptides or Pi Proteomics). Peptides included: gp100 overlapping peptides (Table I); wild type (WT) gp100_44–59 (WTWNRQLYPEWTEAQRLD); gp100_44–59 APLs (Table II); influenza hemagglutinin (HA)_307–319 (PKYVKQNTLKLAT); and CDC27_758–772 (MNFSWAMDLDPKGAN) (30).

Table I.

Naturally processed HLA-DR4-restricted overlapping gp100 peptides.

Residues	Copy number/cell	Amino acid sequence
44–59	4500	WNRQLYPEWTEAQRLD
44–57	2000	W------------R
45–57	900	N-----------R
45–59	500	N-------------D
46–57	300	R----------R
46–59	800	R------------D
40–57	100	RTKAW------------R

All peptides have the same core binding region (in bold) as wild type gp100_44–59.

Table II.

gp100_44–59 APLs and their affinity for HLA-DR4

gp10044–59 and APLs	Amino acid sequence	Substituted position	IC50 (nM)
WT	WNRQLYPEWTEAQRLD	NA	627
L48F	----F-----------	P1*	280
Y49M	-----M----------	P2	521
P50A	------A---------	P3	189
E51A	-------A--------	P4*	199
E51Q	-------Q--------	P4*	200
W52A	--------A-------	P5	372
T53V	---------V------	P6*	489
E54L	----------L-----	P7*	65
E54T	----------T-----	P7*	104
A55G	-----------G----	P8	553
Q56A	------------A---	P9*	276
Q56I	------------I---	P9*	926

^*Indicates MHC anchor residues according to crystal structure. IC50, 50% inhibitory concentration. WT, wild type gp100 peptide sequence. NA, not applicable.

MHC-peptide binding affinity assay

Competition assays to quantitatively measure the binding affinities of native and altered gp100 peptides for purified HLA-DR4 were performed essentially as detailed elsewhere (31–33). Briefly, purified HLA-DR4 molecules (5–500 nM) were incubated with various concentrations of unlabeled peptide inhibitors and 0.1–1 nM ¹²⁵I-radiolabeled probe peptide (human MBP_85–101, sequence PVVHFFKNIVTPRTPPY) for 48h in PBS containing 0.05–0.15% Nonidet P-40 (NP40) in the presence of a protease inhibitor cocktail. MHC binding of the radiolabeled peptide was determined by capturing MHC/peptide complexes on L243 (anti-HLA-DRA) antibody-coated Lumitrac 600 plates (Greiner Bio-one, Frickenhausen, Germany), and measuring bound cpm using the TopCount (Packard Instrument Co., Meriden, CT) microscintillation counter. Peptides were typically tested at six different concentrations covering a 100,000-fold dose range, in three or more independent assays. Under the conditions utilized, where [label]<[MHC] and IC₅₀ ≥ [MHC], the measured IC₅₀ values are reasonable approximations of the K_D values (34, 35).

MHC-peptide protein preparation, crystallization, data collection, and structure determination

The gp100_44–59–HLA-DR4 complex was assembled in two steps in order to maximize the yield of recombinant protein. First, HLA-DR4 bearing CLIP_87–101 peptide (PVSKMRMATPLIMQA) was prepared by in vitro folding from bacterial inclusion bodies. Second, gp100_44–59 was loaded into CLIP–HLA-DR4 using the peptide-exchange catalyst HLA-DM. Briefly, the extracellular portions of the HLA-DR4 α and β chains (residues 1–181 and 1–192, respectively) were expressed separately as inclusion bodies in Escherichia coli BL21 (DE3) cells (Novagen). Inclusion bodies were dissolved in 8 M urea, 50 mM Tris-HCl (pH 8.0), and 10 mM DTT, followed by purification on a Poros HQ20 anion exchange column (PerSeptive Biosystems) in 50 mM Tris-HCl, 8 M urea, and 1 mM DTT at pH 8.0 (DRα) or pH 8.5 (DRβ), using a linear NaCl gradient (36). For in vitro folding, the purified subunits were diluted to a final concentration of 40 mg/L each in a folding solution containing 50 mM Tris-HCl, 30% (w/v) glycerol, 0.5 mM EDTA, 3 mM reduced glutathione, and 0.9 mM oxidized glutathione (pH 8.0). CLIP peptide (GenScript) was added to a final concentration of 5 μM, and the folding mixture was kept for two weeks at 4 °C. The final folding solution was concentrated and dialyzed against 50 mM Mes (pH 6.0). Purification was carried out with sequential Superdex S-200 and Mono Q FPLC columns (GE Healthcare). The CLIP–HLA-DR4 complex was concentrated to 0.8 mg/ml and loaded with gp100_44–59 by overnight incubation at 37 °C in 100 mM sodium citrate-HCl (pH 5.8) containing 200 μM gp100_44–59 and 0.2 mg/ml soluble HLA-DM.

The gp100_44–59–HLA-DR4 complex was crystallized at room temperature in hanging drops by mixing equal volumes of the protein solution at 5 mg/ml and a reservoir solution of 14% (w/v) polyethylene glycol (PEG) 8000, 0.2 M magnesium acetate, and 0.1 M HEPES (pH 7.0). For data collection, crystals were transferred to a cryoprotectant solution [mother liquor containing 30% (w/v) PEG 8000], prior to flash-cooling in a nitrogen stream. X-ray diffraction data to 2.5 Å resolution were recorded at beamline X29 of the Brookhaven National Synchrotron Light Source with an ADSC Quantum-315 CCD detector. All data were indexed, integrated, and scaled with the program HKL 2000 (37). Data collection statistics are shown in Table III.

Table III.

Data collection and refinement statistics

	gp10044–59–HLA-DR4
Data collection statistics
Space group	P21212
Unit cell (Å, °)	a = 90.3, b = 117.6, c = 41.9
Resolution (Å)	30.0–2.50
Observations	144,536
Unique reflections	14,952
Completeness (%)a	93.0 (71.2)
Mean I/σ(I)a	37.6 (6.2)
Rsym (%)a,b	6.1 (24.4)
Refinement statistics
Resolution range (Å)	30.0–2.50
Rwork (%)c	26.9
Rfree (%)c	29.8
Protein atoms	3,133
R.m.s. deviations from ideality
Bond lengths (Å)	0.009
Bond angles (°)	1.47
Ramachandran plot statistics
Most favored (%)	86.2
Additionally allowed (%)	12.6
Generously allowed (%)	0.9
Disallowed	0.3

^aValues in parentheses are statistics for the highest resolution shell (2.50–2.54 Å).

^bR_sym = Σ|I_j – <I>|/ΣI_j, where I_j is the intensity of an individual reflection and <I> is the average intensity of that reflection.

^cR_work = Σ||F_o| – |F_c||/Σ|F_o|, where F_c is the calculated structure factor. R_free is as for R_work but calculated for a randomly selected 10.0% of reflections not included in the refinement.

The structure was solved by molecular replacement with the program Phaser (38) using HLA-DR4 (Protein Data Bank accession code 1D5Z) (39) as a search model. Refinement was carried out using CNS1.2 (40) including iterative cycles of simulated annealing, positional refinement and B factor refinement, interspersed with model rebuilding into δ_A-weighted F_o– F_c and 2F_o– F_c electron density maps using XtalView (41). Refinement statistics are summarized in Table III. Stereochemical parameters were evaluated with PROCHECK (42). Atomic coordinates and structure factors have been deposited in the Protein Data Bank (accession code 4IS6 http://www.rcsb.org/pdb/home/home.do).

Patients

Patients with unresectable stage IV melanoma who expressed HLA-DRB1*0401 were treated with a synthetic gp100 peptide vaccine. Patients were vaccinated subcutaneously with 5 mg of gp100_44–59 peptide emulsified in incomplete Freund’s adjuvant every 3 weeks for a series of 4 inoculations (1 treatment cycle), as described (16). PBMCs were collected by leukapheresis before treatment and 3 weeks after every two vaccinations for in vitro immunologic monitoring studies. Patients underwent radiologic restaging after each treatment cycle. All patients were treated in the Surgery Branch of the National Cancer Institute, National Institutes of Health in Bethesda, Maryland on a clinical trial approved by the Institutional Review Board of the National Cancer Institute, after signing informed consent (16). In addition to DRB1*0401, HLA-DR alleles expressed by the six patients reported here included the following: DRB1*0404, patient #6; DRB1*0701, patients #1, 3 and 5; DRB1*1104, patient #2; DRB1*1501, patient #4; DRB3*02, patient #2; DRB4*01, all patients; DRB5*01, patient #4.

gp100-specific CD4⁺ T cell clone 100.1.G7

The gp100_44–59–specific CD4⁺ T cell clone 100.1.G7 (G7 clone hereafter) was raised by repetitive in vitro peptide stimulation of peripheral blood lymphocytes (PBLs) from a HLA-DR4⁺ melanoma patient whose tumor expressed gp100 protein. Briefly, T cell cultures were initiated in 24-well plates in the presence of IL-2. GM-CSF (200 U/ml) and IL-4 (100 U/ml) were added on day 0 to generate dendritic cells as APCs, along with 25 μM peptide. T cells were restimulated every 10–14 days with irradiated peptide-pulsed autologous PBMCs or HLA-DR4⁺ EBV-transformed B cells and were cloned after 28 days by limiting dilution culture, as previously described (26). Long-term CD4⁺ T cell clones were cultured in RPMI 1640 medium supplemented with 10% heat inactivated human AB serum, IL-2, −7, and −15, and 20% conditioned medium from lymphokine-activated killer cell cultures.

T cell functional assays: ELISAs

Peptide-specific CD4⁺ T cells at 1e5–3e5 cells per well were co-cultured overnight in flat-bottom 96-well plates with 1e5 peptide-pulsed HLA-DR4⁺ allogeneic EBV-B cells or autologous PBMCs. Culture supernatants were harvested for measurement of GM-CSF and IFN-γ secretion by T cells, using ELISA with commercially available kits (R&D Systems).

In vitro stimulation (IVS) and ELISPOT assay

Cryopreserved PBMCs from vaccinated patients were thawed and suspended in complete medium (CM, RPMI 1640 plus 10% heat-inactivated human AB serum, 2 mM glutamine, 10 mM HEPES buffer, and antibiotics) at 1e6 cells/ml in 24-well plates with 20 μM peptide, 200 U/ml GM-CSF and 100 U/ml IL-4. Parallel cultures were grown with IL-7 and IL-15 (25 ng/ml each) instead of GM-CSF and IL-4. Cells were incubated at 37°C. After 3 days, IL-2 was added at a final concentration of 10 IU/ml. At 9–12 days, some cells were harvested for ELISPOT assay and the remaining cells were restimulated with peptide-pulsed irradiated autologous PBMCs and cultured in CM containing 150 IU/ml IL-2. At day 20, cells were harvested again for ELISPOT assay. One day prior to ELISPOT assays, approximately half of the culture volume was replaced with fresh CM (without IL-2).

ELISPOT assays were conducted in MultiScreen-IP Filter Plates coated overnight at 4°C with 50 μl/well of 20 μg/ml mouse anti-human IFN-γ antibody (clone 1D1K, Mabtech). On the following day, the plates were washed and blocked with AB medium for 2 hours at 37 °C. For some assays, fresh cryopreserved PBMCs were thawed into ELISPOT media (EM, RPMI with 10% heat-inactivated human AB serum, 2 mM glutamine, 10 mM HEPES buffer) at 4e6 cells/ml. After 2 hours incubation at 37°C and 5% CO₂, 2e5 PBMCs were plated directly to each IFN-γ antibody-coated well. For ELISPOT assays using peptide-stimulated T cell cultures, 1e5 T cells were plated into each IFN-γ antibody-coated well. Then, 1e5 irradiated (5000 rad) autologous PBMCs were added to T cells in each well as APCs. Gp100 peptides were added at 20 μM in 100 μl total volume. An unrelated HLA-DR4-restricted peptide, CDC27_758–772, was used as a negative control. PMA-ionomycin stimulation provided a positive control. After overnight incubation at 37°C, cells were discarded and the plates were washed with PBS/0.05% Tween 20 followed by PBS. 100 μl biotinylated mouse anti-human IFN-γ mAb (clone 7-B6-1, Mabtech) diluted at 2 μg/ml in PBS with 0.5% BSA was added to each well and incubated at 37°C for 2 hours. The plates were washed and developed with avidin-peroxidase-complex (Vector Laboratories) and stained with AEC substrate (Sigma). Spots were counted by KS ELISPOT automated reader system (Carl Zeiss). The number of spots was averaged from triplicate wells. Peptide-specific T cells were defined as showing ≥ 20 spots per 1e6 fresh PBMCs or cultured T cells, and ≥ twice the numbers of spots observed in the negative control wells.

RESULTS

T cell recognition of native DR4-restricted gp100 peptides

The DR4-restricted peptide gp100_44–59 was originally described by Halder et al. (13) as a dominant peptide displayed on the surface of melanoma cells. The immunogenicity of this peptide was confirmed in DR4-transgenic mice and in human in vitro studies (14, 15), leading to its clinical testing as a melanoma vaccine. Although many MHC II-restricted peptides occur as nested peptide sets, naturally occurring sequences overlapping gp100_44–59 have not been previously identified. In order to identify native nested gp100 peptides with potentially enhanced immunogenicity, we searched peptides eluted from DR4 molecules displayed on cultured melanoma cells. A set of seven nested gp100 peptides spanning residues 40–59 was identified. Peptide abundance ranged from 100 to 4500 copies/cell (Table I). We assessed the recognition of these peptides by measuring specific cytokine secretion from the CD4⁺ gp100-specific DR4-restricted G7 clone (Figure 1). As shown in Figure 2, recognition of gp100_44–59 exceeded all other peptides. Truncated peptides lacking the N-terminal residues Trp44 and Asn45, outside the peptide’s MHC core binding sequence, were not well recognized by T cells; this is reminiscent of our findings with another MHC II-restricted melanoma-associated peptide, phospho-MART-1, and highlights the important effects that N-terminal residues outside the binding groove may have on non-mutated (self) epitope recognition (43). Thus, we were unable to identify a naturally processed overlapping peptide that was better recognized by the G7 clone than gp100_44–59. Furthermore, this peptide was the most abundantly expressed member of the nested peptide set. Therefore, all subsequent experiments were based on the gp100_44–59 native peptide sequence.

Figure 1.

The CD4⁺ T cell clone 100.1.G7 (“G7”) is gp100_44–59-specific and HLA-DR4-restricted. (A) G7 T cells were cultured overnight with allogeneic HLA-DR4⁺ EBV-B cells pre-pulsed with gp100_44–59 or HA_307–319 peptides (25 μM). GM-CSF secreted by T cells into culture supernatants was measured by ELISA. Results representative of 4 experiments. (B) Recognition of gp100_44–59 peptide-pulsed EBV-B cells by G7 is specifically blocked by an anti-HLA-DR mAb. G7 cells are homozygous for HLA-DR4. (C) Specific recognition of HLA-DR4+ melanoma cells by G7 T cells, which were co-incubated with HLA-DR4⁺ (1359 and 1102) or HLA-DR4⁻ (1363) EBV-B or melanoma cells overnight. HA, influenza hemagglutinin_307–319, an HLA-DR4-restricted peptide that was used as a negative control. EBV, Epstein-Barr Virus. mAb, monoclonal antibody. Mel, melanoma cell lines.

Figure 2.

G7 clone recognition of gp100_44–59 exceeds recognition of other members of a nested set of naturally processed gp100 peptides. Antigen presenting cells were pulsed with gp100_44–59 or nested peptides at the indicated concentrations overnight. G7 cells were added and cultured overnight. Supernatants were harvested and tested for IFN-γ secretion by ELISA. Similar results were obtained for GM-CSF secretion (not shown). HA, influenza hemagglutinin_307–319, an HLA-DR4-restricted peptide that was used as a negative control.

Structure of gp100_44–59 bound to HLA-DR4

The crystal structure of gp100_44–59 complexed to HLA-DR4 was determined to 2.5 Å resolution (Table III). Continuous and unambiguous electron density was observed for gp100_44–59 residues Arg46 to Leu58; however, N-terminal residues Trp44 and Asn45 and C-terminal residue Asp59 were not defined in the electron density map (Figure 3A), despite their importance for recognition by clone G7 (Figure 2).

Figure 3.

Structure of the gp100_44–59–HLA-DR4 complex. (A) Electron density map for the bound gp100_44–59 peptide. The 2F_o – F_c map at 2.5 Å resolution is contoured at 1δ. The peptide is drawn in stick representation with carbon atoms in yellow, oxygen atoms in red, and nitrogen atoms in blue. (B) Top view of the gp100_44–59–HLA-DR4 complex, looking down on the peptide-binding groove. The molecular surface of HLA-DR4 is cyan (MHC α-chain) and green (MHC β-chain). (C) Conformation of high- and low-affinity peptides bound to HLA-DR4. The conformation of gp100_44–59 (yellow) was compared with those of CII_1168–1180 (green), HA_307–319 (blue), and MBP_111–129 (pink) by superposing the α1β1 domains of HLA-DR4 in the gp100_44–59–HLA-DR4, CII_1168–1180–HLA-DR4 (Protein Data Bank accession code 2SEB) (47), HA_307–319–HLA-DR4 (1J8H) (45), and MBP_111–129 –HLA-DR4 (3O6F) (48) complexes. The peptides are viewed from the side of the β1 helix; gp100_44–59, CII_1168–1180, and HA_307–319 bind HLA-DR4 with higher affinity than MBP_111–129.

In the structure, the primary anchor residues for gp100_44–59 bound to HLA-DR4 are Leu48 (P1) and Thr53 (P6); the secondary anchor residues are Glu51 (P4), Glu54 (P7) and Gln56 (P9). This result is consistent with the sequence motif for HLA-DR4 deduced from phage display and synthetic peptide libraries (24, 33, 44). Leu48 fulfills the requirement for a large nonpolar residue at P1 for efficient binding to HLA-DR4. Thr53 at P6 is a suitable primary anchor as well. By contrast, none of the secondary anchors (P4 Glu51, P7 Glu54 and P9 Gln56) conform to the optimal binding motif for HLA-DR4. Thus, the P4 pocket of HLA-DR4, which is lined by β-chain residues Phe26, Lys71, Ala74 and Tyr78, is hydrophobic and prefers nonpolar residues (24, 33, 44). In the gp100_44–59–HLA-DR4 complex, the charged carboxy group of the P4 Glu51 side chain points toward the surface of HLA-DR4 instead of into the P4 pocket (Figure 3B). The P7 pocket of HLA-DR4, formed by β-chain residues Try47, Trp61, Leu67 and Lys71, is likewise hydrophobic, such that the side chain of P7 Glu54 adopts a conformation very similar to that of P4 Glu51. The side chain of P9 Gln56 projects deep into the primarily nonpolar P9 pocket of HLA-DR4 (Figure 4 left panel), which is formed by α-chain residues Asn69, Ile72 and Met73 and β-chain residues Tyr37, Asp57 and Trp61. P9 Gln56 makes extensive van der Waals contacts with these residues, in addition to two side-chain–side-chain hydrogen bonds: P9 Glu56 Oε1–Oη DR4 Tyr37β and P9 Glu56 Nε2–Oδ1 DR4 Asp57β (Figure 4 right panel).

Figure 4.

Interaction of P9 Gln56 with HLA-DR4. Left panel, molecular surface of HLA-DR4 (MHC α-chain, cyan; MHC β-chain, green) showing the P9 pocket that accommodates the side chain of P6 Gln56. Right panel, contact residues of HLA-DR4 are drawn and labeled. Hydrogen bonds are indicated by broken black lines.

The conformation of gp100_44–59, which binds HLA-DR4 with intermediate affinity (15), was directly compared with those of three other peptides that bind HLA-DR4: 1) a self peptide from collagen II (CII_1168–1180) which binds HLA-DR4 with relatively high affinity (24, 44); 2) a foreign peptide from influenza virus hemagglutinin (HA_307–319) that also binds HLA-DR4 with high affinity (45); and 3) a self peptide from myelin basic protein (MBP_111–129), which binds weakly to HLA-DR4 (46). The main chain of gp100_44–59 superposes very closely onto those of CII_1168–1180 and HA_307–319 in complex with HLA-DR4 (45, 47), from residues P2 to P9 (Figure 3C). The C-terminus of gp100_44–59 (residues P10 and P11), which is positioned ~1 Å higher than that of CII_1168–1180 or HA_307–319, is located outside the peptide-binding groove. However, all three peptides sit deeply in the binding groove. By contrast, the low-affinity MBP_111–129 peptide diverges from gp100_44–59, CII_1168–1180, and HA_307–319 at anchor residues P6 and P7 (Figure 3C), due to a longer side chain at P6 (MBP_111–129: Gln; gp100_44–59: Thr; CII_1168–1180: Ala; HA_307–319: Thr) (48). The consequent elevation of MBP_111–129 results in fewer contacts to HLA-DR4 compared to gp100_44–59, CII_1168–1180, or HA_307–319.

APLs and their affinity for DR4

According to the crystal structure of the DR4-gp100_44–59 complex as well as published HLA-DR4 peptide binding motifs (22–24), we designed several altered peptide ligands (APLs) with substituted MHC anchor residues (Table II). The amino acid substitutions L48F (P1), E51Q and E51A (P4), E54L and E54T (P7), and Q56A (P9) are located at MHC anchor residues revealed by crystal structure and were designed to enhance MHC-peptide affinity. Indeed, these anchor-modified APLs were found to have 2–10 fold higher affinities for DR4 than the WT peptide. Y49M (P2), P50A (P3), W52A (P5) and A55G (P8), in which the substituted amino acids are positioned at potential TCR contact residues, had higher or similar MHC-binding affinities compared to WT peptide. As a negative control, Q56I was designed to reduce binding affinity to DR4 due to the large isoleucine residue in P9, incompatible with the corresponding shallow binding groove in HLA-DR4; the MHC binding affinity of this modified peptide was approximately 50% lower than WT (Table II).

Recognition of gp100_44–59 APLs by gp100_44–59 -specific CD4⁺ T cell clone G7

T cell recognition of gp100_44–59 APLs was first characterized by cytokine secretion from the G7 clone. Peptides modified at putative TCR contact residues -- Y49M (P2), P50A (P3) and W52A (P5) -- were not recognized by the G7 clone, confirming the importance of these residues for T cell specificity (Figure 5). In addition, a replacement of threonine by valine in anchor position 6 also abrogated T cell recognition (not shown), suggesting that P6 has high stringency for a hydroxylated residue. However, G7 secreted 2 to 3 times more IFN-γ and GM-CSF in response to Q56A than to WT peptide (Figure 5), nominating the Q56A APL for further study. Other APLs showed equivalent or lower recognition by G7, compared to WT peptide (Figure 5). The ability of G7 to cross-react with gp100_44–59 and Q56A APL is easily understood in structural terms, since these peptides differ by only a single residue, at anchor position P9. The Gln56 side chain of gp100_44–59 (and presumably the Ala56 side chain of Q56A APL) is sequestered in the P9 pocket of HLA-DR4 and is therefore not exposed to TCR G7 (Figure 3B).

Figure 5.

Q56A substitution at the P9 anchor residue of gp100_44–59 confers enhanced recognition by the G7 clone. HLA-DR4⁺ 1102-EBV cells were pulsed with gp100 peptides at the indicated concentrations for 24 hrs. Then G7 cells were added and cultured overnight. Supernatants were harvested and tested for IFN-γ (left) and GM-CSF (right) secretion by ELISA. T cell recognition of the Q56A was enhanced approximately 5-fold compared to WT gp100 peptide. APLs not shown in this figure did not stimulate detectable cytokine secretion, including Y49M (P2 substitution), P50A (P3), W52A (P5) and T53V (P6). Results representative of 3 experiments. HA, influenza hemagglutinin_307–319, an HLA-DR4-restricted peptide that was used as a negative control. WT, wild type gp100_44–59.

Comparison of gp100 WT peptide versus the Q56A APL in detecting gp100-specific immunity in PBLs from vaccinated patients

In a prior study of melanoma patients receiving the gp100 WT vaccine, no detectable immunity against the vaccinating peptide was detected in post-treatment PBMCs, using a standard IFN-γ ELISPOT assay. Since the Q56A APL had higher affinity for DR4 and provoked greater recognition by the G7 clone, we assessed Q56A for its ability to detect gp100-specific reactivity in pre- and post-treatment PBLs from six patients previously reported in this clinical trial (16) (Figure 6). Uncultured PBLs were assessed for reactivity against WT peptide or the Q56A APL, using an IFN-γ ELISPOT assay. In addition, PBLs were stimulated in vitro (IVS) for 10–20 days with WT or Q56A peptide to amplify gp100_44–59 specific T cells, prior to ELISPOT. With IVS, the CD4⁺ T cell population increased to 60%-90% of cultured T cells (data not shown). We found reproducible evidence that new specific anti-gp100 responses developed after vaccination in 2 of 6 patients, using uncultured or 10–20 day IVS T cells (Patients #1 and 2). In Patient #3, IVS cultures showed greater gp100-specific activity after 4 compared to 2 vaccines; pre-treatment T cells did not proliferate in IVS culture and thus could not be assessed. There was pre-existing immunity against gp100_44–59 which persisted after vaccination in uncultured PBMCs and IVS cultures, in Patient #4. Finally, two patients did not manifest evidence of a pre-existing or vaccination-induced anti-gp100 response (data not shown). Among the 4 patients with anti-gp100 immunity, only Patient #3 had improved detection of responses using the Q56A APL, while also recognizing WT peptide. IVS cultures raised against WT or Q56A peptides showed equivalent recognition of the reciprocal peptide in patients #1, 2 and 4 (Figure 6). In summary, IVS-enhanced ELISPOT techniques enabled the detection of gp100-specific immunity in 4 of 6 melanoma patients (in 3 of 4 patients, immunity increased following vaccination). Unlike results with the G7 T cell clone, the Q56A APL was not consistently superior to WT peptide in revealing these responses.

Figure 6.

Detection of gp100 specific immunity in 4 vaccinated melanoma patients. Reactivity against gp100_44–59 in uncultured PBMCs or in 10–20 day IVS cultures was determined by IFN-γ ELISPOT assays. 20 μM WT or Q56A peptide was used to stimulate CD4⁺ T cells in both IVS cultures and ELISPOT assays. Data from uncultured PBMCs from patient #3 are not shown, because all values were < 20 spots/10⁶ cells. CDC27, an unrelated HLA-DR4-restricted control peptide. IVS, in vitro stimulation. Pre, pre-vaccine. P2, P4, P8, PBMCs collected from patients post 2, 4 or 8 vaccinations. SFC, spot forming colonies, mean± SEM. WT, wild type gp100_44–59. * P<0.02, one-tailed t test compared to CDC27.

T cells raised against WT peptide prefer a variety of APLs

We proceeded to evaluate a panel of gp100 APLs for their ability to detect gp100-specific CD4⁺ T cell responses in vitro, by testing their recognition by T cells from vaccinated melanoma patients that were stimulated in vitro with WT peptide. Figure 7A shows results from 3 patients, demonstrating that IVS T cells from each patient had a unique preference for recognizing various APLs. These results suggest heterogeneous TCR usage by gp100 WT-specific T cells from individual patients. Several APLs that were preferred by individual patients (L48F, E54L and E54T) were assessed for their ability to generate gp100-specific CD4⁺ T cells from the pre-vaccination PBLs of the 6 melanoma patients. None of these APLs was superior to WT or Q56A peptide in stimulating gp100-specific CD4⁺ T cells, and in some cases APLs provoked APL-specific responses that did not cross-react with WT peptide (not shown). Finally, a peptide pool comprised of WT peptide plus the APLs L48F, E54L, E54T, and Q56A was evaluated for its ability to raise gp100-specific immunity in pre-vaccination T cells or to detect post-vaccination gp100 specific immunity, in melanoma patients #1, 2 and 4, similar to experiments shown in Figure 6. The peptide pool was not superior to the WT peptide in raising or detecting gp100-specific immunity (Figure 7B).

Figure 7.

T cells stimulated in vitro with gp100_44–59 WT peptide show selective recognition of a variety of APLs. (A) T cells were derived from melanoma patients receiving a gp100 WT peptide vaccine. Results of IFN-γ ELISAs are shown. Background values for IFN-γ secretion by T cells stimulated with APC + HA_307–319 have been subtracted from data shown, as follows: Patient #2, 213 pg/ml (1 μM peptide stimulation) and 208 pg/ml (10 μM peptide stimulation); Patient #4, 187 and 191 pg/ml; Patient #5, 346 and 388 pg/ml. (B) A peptide pool including APLs was not superior to the WT peptide in raising or detecting gp100-specific immunity. Reactivity against gp100_44–59 in 10-day IVS cultures was determined by IFN-γ ELISPOT assays. CD4⁺ T cells were stimulated with 20 μM WT gp100 or peptide pool (WT, L48F, E54L E54T and Q56A) in both IVS cultures and ELISPOT assays. CDC27, an unrelated HLA-DR4-restricted peptide, was used as a negative control. P4, P6, P8, PBMCs collected from patients post 4, 6 or 8 vaccinations. SFC, spot forming colonies, mean± SEM. WT, wild type gp100_44–59. * P<0.02, ^× P<0.05, one-tailed t test compared to HA or CDC27 control peptides.

DISCUSSION

Insights gained from the characterization of MHC-peptide-TCR interactions hold the promise that directed structural modifications can have predictable functional consequences. The ability to manipulate T cell reactivity synthetically or through genetic engineering might thus be translated into new therapies for common diseases such as cancer and autoimmune disorders. In an early example with the HLA-A2-restricted gp100 epitope 209–217, an MHC anchor residue substitution improved MHC-peptide binding affinity, which translated into enhanced CD8 T cell recognition in vitro and increased immunogenicity in the clinic as a melanoma vaccine (8, 17). However, subsequent investigations on the same theme using MHC I-restricted melanoma-associated peptides and melanoma-specific TCRs have yielded variable results, demonstrating that structural modifications of MHC-peptide-TCR interactions may have unpredictable or undesirable functional consequences (49, 50), which may involve collateral damage to normal tissues when tolerance to non-mutated tumor antigens such as gp100 is alleviated. Furthermore, in vitro functional testing may not always be predictive of in vivo effects (51). These difficulties are amplified in the context of MHC II interactions, for which peptide binding motifs are permissive and not rigid, enabling binding of a single peptide to multiple MHC II alleles (52). An APL of an HLA-DR2-restricted myelin basic protein (MBP) epitope containing modified TCR contact residues, designed to antagonize autoreactive T cells in patients with multiple sclerosis, caused disease exacerbation rather than alleviation in some vaccinated patients (53).

In the current study, APLs from a dominant HLA-DR4-restricted gp100 epitope were designed to enhance MHC binding and hence T cell recognition in melanoma patients. APL design was based on precise structural definition of the wild type peptide-MHC II complex, and increased MHC binding was validated by direct affinity measurements. Nevertheless, we observed heterogeneous preferences of CD4+ T cells from several HLA-DR4+ melanoma patients for different gp100 APLs, suggesting highly variable TCR usage even among patients who had been vaccinated against the wild type gp100 peptide. This could reflect heterogeneous responses of TCRs from different patients to APL-DRB1*0401 complexes, but might also reflect immune responses to APLs bound to alternative DR alleles expressed by individual patients. This heterogeneity prevented the selection of an APL candidate for developing an improved generic melanoma vaccine.

Although we did not determine crystal structure of Q56A APL bound to HLA-DR4, it is unlikely that the conformation of this anchor-modified peptide differs significantly from that of wild-type gp100_44–59, which could have explained loss of TCR reactivity. Q56A APL and gp100_44–59 differ by only a single residue, at anchor position P9. The possible effect on the peptide backbone of replacing P9 Gln by Ala may be assessed by examining two other peptide–HLA-DR4 structures, one involving CII_1168–1180, in which P9 is Gly (47), and the other involving HA_307–319, in which P9 is Leu (45). As shown in Figure 3C, the main chain of gp100_44–59 superposes very closely onto those of CII_1168–1180 and HA_307–319, even though these three unrelated peptides have different residues at P9 (Gln, Gly, and Leu, respectively). This suggests that the P9 anchor modification in Q56A APL exerts subtle effects on peptide contacts with TCR or on peptide dynamics, such that an alteration intended to enhance immunogenicity may be counterproductive. Indeed, there is growing evidence for an important role of peptide dynamics in modulating TCR recognition (54–56). In one especially relevant study (55), replacing suboptimal anchor residues of the HLA-A2-restricted MART-1_27–35 melanoma antigen unexpectedly abolished recognition by most MART-1-specific T cell clones, despite a 40-fold improvement in peptide binding to MHC. Moreover, crystal structures of wild-type and anchor-modified MART-1_27–35 peptides bound to HLA-A2 showed that the anchor modifications did little to alter the conformation of the peptide or the MHC binding groove (49). Instead, NMR and molecular dynamics simulations revealed that the anchor modifications increased the flexibility of both MART-1_27–35 and HLA-A2 (55). Although the resulting entropic effects improved binding of the peptide to MHC, they also increased the entropic penalty for TCR binding to prohibitive levels, resulting in loss of TCR affinity for peptide–MHC. A similar mechanism may explain the loss of T cell recognition that we observed for Q56A APL. In a related study, TCR recognition of an APL of lymphocytic choriomeningitis virus peptide p33 was found to be entropically driven, whereas recognition of wild-type p33 was enthalpically driven (57). If different peptides can alter peptide-MHC dynamics in ways that affect TCR recognition, the commonly used approach of optimizing anchor residues to enhance peptide binding to MHC may be insufficient to reliably generate improved vaccine candidates, as we found in the present study.

In the gp100_44–59–HLA-DR4 structure, N-terminal residues Trp44 and Asn45 lie outside the peptide-binding groove and are not defined in the electron density, suggesting flexibility. Nevertheless, these two N-terminal residues are required for efficient recognition by TCR G7, which is reminiscent of the way some autoreactive TCRs engage self or altered self peptides, including tumor antigens, presented by MHC II molecules (43, 58). For example, X-ray crystallographic studies of tumor-specific TCRs that recognize a somatically mutated human melanoma antigen have revealed substantial alterations in the topology of TCR binding to peptide–MHC compared to anti-foreign TCRs (59, 60). In these anti-tumor TCR–peptide–MHC complexes, the TCR is skewed toward the peptide N-terminus relative to its central position in anti-foreign TCR–peptide–MHC complexes, resulting in low affinity binding (K_D > 200 μM) that likely enabled escape from negative thymic selection. In another example, human autoimmune TCR Ob.1A12, which recognizes a self-peptide from myelin basic protein (MBP_85–99) bound to HLA-DR2b, was found to only contact the N-terminal half of MBP_85–99 (61). TCR Ob.1A12 cross-reacts with an E. coli peptide having limited sequence identity with MBP_85–99. Cross-reactivity is due to structural mimicry of a binding hotspot at the N-terminal portion of the bacterial and self-peptide (62).

We attempted to measure the binding of TCR G7 to gp100_44–59–HLA-DR4 by surface plasmon resonance (data not shown). However, we could not detect an interaction, even after injecting a high concentration (up to 200 μM) of recombinant TCR G7 over immobilized gp100_44–59–HLA-DR4, which precluded further analysis. This result is consistent with previous findings that autoreactive TCRs generally bind self-antigens, including tumor antigens, with very low affinity (58). For example, the human melanoma-specific TCR E8 binds mutTPI–HLA-DR1 with K_D > 200 μM (59). The low affinities of G7 and E8 for their self-peptide–MHC ligands likely enabled these autoreactive T cells to escape negative thymic selection (58).

Although an explosion of information in the field of molecular immunology over the past two decades has yielded extraordinarily precise and extensive data on molecular interactions and structure-function relationships in the immune system, there remains a critical gap in knowledge that hinders reliable clinical translation from structural and in vitro observations. As demonstrated in the present study, peptide modifications that improve binding to MHC class II molecules do not necessarily translate to increased antigenicity. Confounding this strategy is the heterogeneity of human CD4+ T cell responses, as well as possible dynamical effects of peptide modifications. Predicting these effects will require the use of molecular dynamics simulations in conjunction with structural information. Such structure-guided computational design may eventually allow successful re-orientation of immune responses against cancer with vaccines and other immunomodulatory therapies.

Abbreviations used in the article

APL

altered peptide ligand

HA

influenza hemagglutinin

IC50

50% inhibitory concentration

IVS

in vitro stimulation

MBP

myelin basic protein

WT

wild type