Science for Life Laboratory, Center for Infectious Medicine (CIM), Department of Medicine, Karolinska Insitutet, Karolinska University Hospital Huddinge, Stockholm, Sweden
Full correspondence Dr. Adnane Achour, Science for Life Laboratory, Center for Infectious Medicine, Department of Medicine, Karolinska University Hospital Huddinge, Karolinska Institutet, Stockholm, Sweden
See accompanying commentary by Hickman and Yewdell
The immunogenicity of H-2Db (Db) restricted epitopes can be significantly increased by substituting peptide position 3 to a proline (p3P). The p3P modification enhances MHC stability without altering the conformation of the modified epitope allowing for T-cell cross-reactivity with the native peptide. The present study reveals how specific interactions between p3P and the highly conserved MHC heavy chain residue Y159 increase the stability of Db in complex with an optimized version of the melanoma-associated epitope gp10025–33. Furthermore, the p3P modification directly increased the affinity of the Db/gp10025–33-specific T-cell receptor (TCR) pMel. Surprisingly, the enhanced TCR binding was independent from the observed increased stability of the optimized Db/gp10025–33 complex and from the interactions formed between p3P and Y159, indicating a direct effect of the p3P modification on TCR recognition.
The identification of major histocompatibility complex (MHC) restricted tumor-associated antigens (TAAs) as targets for CD8+ cytotoxic T-cell lymphocytes (CTLs) has opened new possibilities for immunotherapy and led to intense efforts to harness and increase immune responses against tumors . T-cell responses against MHC/TAA complexes are constrained by immune tolerance and most of the identified TAAs are poorly immunogenic . In general, cancer-associated epitopes display weak MHC-binding affinity and low complex stabilization capacity, and the avidity of TAA-reactive CTLs is moderate at best . Therefore, the use of altered peptide ligands (APLs) with enhanced stabilization capacity and immunogenicity may allow overcoming tolerance and the intrinsic low immunogenicity of WT TAAs. Nevertheless, the appropriate design and possible clinical application of APLs has proven difficult and still represents one of the major hurdles in the design of cancer vaccines.
The most commonly used approach to improve the immunogenicity of TAAs is to modify peptide anchor residues in order to optimize the interactions of the targeted epitope with the MHC peptide binding groove [3-5]. However, it has been demonstrated that even subtle anchor-residue modifications may compromise molecular mimicry, therefore limiting the capacity of these APLs to elicit effective CTL responses against native TAAs [6-9]. Therefore, successful design of structurally conservative peptide modifications yielding enhanced MHC stabilization capacity and immunogenicity would represent a key advance in T-cell epitope immunogen design.
Expression of the differentiation antigen gp100, normally restricted to melanocytes, is significantly increased in malignant melanoma , rendering it an attractive vaccine candidate [11, 12]. Although the mouse gp100-derived epitope EGS (gp10025–33 EGSRNQDWL) contains the theoretically optimal motif required for binding to H-2Db (Db), with an asparagine at position 5 (p5N) and a leucine at the C-terminus (p9L), it binds with very weak affinity to Db and is poorly immunogenic [11, 12]. Peptide position 3 has been described as an auxiliary anchor position for Db-restricted epitopes [13-16]. Nevertheless, the dramatically increased MHC-I affinity, binding stability, and immunogenicity resulting from the substitution of serine at position 3 of gp100 to a proline at peptide position 3 (p3P) in the optimized epitope EGP (EGPRNQDWL) was unexpected . Vaccination of C57BL/6 mice with EGP elicited high frequencies of endogenous CTLs that cross-reacted with Db/EGS complexes on the surface of melanoma cells, leading to tumor cell lysis and impairing tumor progression [11, 17].
Crystal structure analysis revealed that the pyrrolidine ring of p3P closely interacts with the side chain of the highly conserved MHC-I heavy chain tyrosine residue Y159 in the peptide-binding groove of Db/EGP . The juxtaposition of p3P on Y159 was well suited for CH–π interactions between the side chain of p3P and the electron ring of the hydroxyphenyl group of Y159. The importance of CH–π interactions formed between the aromatic side chains of tyrosine, phenylalanine, or tryptophane residues and aliphatic CH donors, including in particular prolines, has been previously demonstrated in model systems and protein–protein interactions [18, 19]. Indeed, proline–aromatic interactions, which are generally considered as relatively weak, appear frequently in 3D protein structures and model systems, suggesting that they provide significant and seemingly underappreciated contributions to protein structural stability, interactions, and functions [18-24]. Despite an initial study by Umezawa et al. , the importance of CH–π interactions for interactions between peptides and MHC heavy chains has to our knowledge not yet been systematically assessed.
In this study, the molecular mechanisms underlying the observed impact of the p3P modification on epitope immunogenicity were investigated. Our results demonstrate that CH–π interactions formed between p3P and Y159 significantly enhanced Db/peptide complex stability. Additionally, p3P-substituted Db/EGP complexes appear to be recognized with higher affinity by a cognate TCR independent of the increased MHC/peptide complex stability. The presented results indicate that CH–π interactions may indeed play an important but underappreciated role for MHC-I-binding affinity, stability, and immunogenicity and suggest a novel and unexpected direct impact of proline substitution on TCR recognition.
The aromatic ring of Y159 is essential for the increased MHC stabilization capacity of EGP
Previous analysis of the crystal structures of Db in complex with the melanoma-associated EGS epitope and the highly immunogenic APL EGP suggested that specific interactions between p3P and Y159 underlie the significantly increased stabilization capacity conferred by this substitution to several Db-restricted TAA epitopes . Comparative analysis of a large amount of crystal structures of Db/peptide complexes revealed the juxtaposition of the aromatic ring of Y159 and the side chains of peptide residues p3 as an absolutely conserved feature (Fig. 1). This juxtaposition is defined by the geometry of the peptide-binding groove and in particular by the conserved hydrogen bond formed between the hydroxyl group of Y159 and the carbonyl of the first peptide residue, independent of the specific sequence of the Db-restricted epitopes (Fig. 1) [15, 26].
The Db residue Y159 was mutated to phenylalanine (Y159F), leucine (Y159L) or alanine (Y159A), and soluble complexes of each mutated MHC were generated with either EGS or EGP. While the Y159F substitution abrogated the hydrogen bond interaction between the Y159 hydroxyl group and p1, it conserved the aromatic phenyl ring that formed van der Waals and CH–π interactions with p3. In contrast, the Y159L and Y159A mutations removed the phenyl ring and progressively reduced the size of the side chain. Thermostability measurements using circular dichroism (CD) demonstrated that the p3P modification substantially increased the stabilization capacity of Db-restricted TAAs, resulting in a 12°C increase in melting temperature (Tm) of Db/EGP over Db/EGS (Fig. 2A and Supporting Information Fig. 1). This striking shift in stabilization capacity was well in line with previously measured differences in complex stability between Db/EGS and Db/EGP using cell-based assays . The Y159F mutation, which abrogated the conserved hydrogen bond formed between p1 and Y159, reduced the Tm values of Db/EGP and Db/EGS by 6 and 3°C, respectively (Fig. 2B and Supporting Information Fig. 1). The Tm values of both Db/Y159L and Db/Y159A in complex with either EGP or EGS were between 35 and 37°C with no statistically significant differences between the two peptides. Together this demonstrates that, while the Y159F modification significantly reduces the stabilization advantage provided by EGP over EGS, the observed effects are abolished only upon removal of the aromatic ring in Db-Y159L and Db-Y159A (Fig. 2C).
The conformations of F159 and p3P are conserved in Db-Y159F/EGP compared to Db/EGP
Given the significant loss of stability observed for Db-Y159F/EGP, the crystal structure of this MHC complex was determined (Supporting Information Table I) and compared to the previously determined structure of Db/EGP  in order to assess potential structural alterations following the loss of the hydrogen bond between p1 and Y159 (Fig. 3A). The two structures are highly similar with an overall root mean square deviation (RMSD) of 0.9 Å and EGP takes nearly identical conformations in both complexes with an RMSD of 0.55 Å (Fig. 3B). Most importantly, the juxtaposition of the side chain of residue F159 and p3P compared to Db/EGP remained conserved (Fig. 3C). Thus the observed reduction in stabilization capacity in Db-Y159F/EGP is not caused by a different conformation of the peptide EGP nor of the side chain of F159.
The p3P substitution independently enhances Db stabilization and TCR affinity
T cells carrying the murine TCR pMel recognize target cells presenting the Db/EGS complex [11, 12, 27]. Vaccination of C57BL/6 mice with the p3P-modified EGP peptide induced strong cross-reactive endogenous CTL responses against Db/EGS+ target cells, based on structural mimicry between Db/EGP and Db/EGS . The affinity of a soluble version of pMel to Db, Db-Y159F, and Db/Y159A in complex with both EGS and EGP was assessed using surface plasmon resonance (SPR) (Fig. 4). The binding affinities for the interactions of pMel with Db/EGP and Db/EGS were within the range generally observed for MHC-restricted agonist/weak agonist peptides , with KD values between 5 and 30 μM (Fig. 4A). Several independent measurements consistently resulted in higher affinities for the interaction of pMel with Db/EGP compared to Db/EGS demonstrating that the p3P modification directly enhances TCR binding.
Surprisingly, the enhanced pMel binding to Db/EGP compared to Db/EGS remained conserved when Y159 was modified to either phenylalanine or alanine, as the binding affinity of this TCR to Db-Y159F/EGP and Db-Y159A/EGP remained higher compared to Db-Y159F/EGS and Db-Y159A/EGS, respectively (Fig. 4B). These results suggest that, while the MHC stabilization capacity of the peptides plays an important role in TCR affinity, the p3P substitution in EGP by itself provides a TCR recognition advantage over EGS, by a mechanism independent of the interactions formed between the modified peptide and Y159 as well as the peptide-binding groove.
The observed increase in pMel TCR binding was confirmed by staining CD8+ pMel T-cells with tetrameric complexes of Db/EGP, Db/EGS and Db-Y159F/EGP (Fig. 4C). Indeed, a significant binding difference was observed between Db/EGP and Db/EGS, resulting in a threefold higher avidity. Furthermore, CD8+ pMel T cells were stained with higher intensity by Db-Y159F/EGP compared to Db/EGS, well in line with both the thermostability (Fig. 2) and SPR measurements (Fig. 4A and B). In conclusion, these results indicate that the affinity increases measured with SPR could have important implication for TCR recognition at low MHC/peptide complex density. In order to verify that the lower tetramer staining intensity was not caused by early denaturation of Db/EGS complexes due to their low thermal stability, the integrity of Db/EGS was followed with CD for 60 min at the same temperature used for the tetramer staining (25°C), demonstrating only minimal denaturation (Supporting Information Fig. 2).
The p3P substitution stabilizes the TCR-interacting MHC interface
To further assess the molecular impact of the p3P modification on MHC/peptide dynamics, molecular dynamics simulations were performed for 40 ns using the entire Db/EGS and Db/EGP complexes as well as coordinated water molecules at 350 K. Analysis of average backbone root mean square fluctuations (RMSFs) indicated an increased overall stability of the peptide-binding groove of Db/EGP compared to Db/EGS (Fig. 5, A and B, Supporting Information Video 1). The higher flexibility of Db/EGS resulted in the partial loss of conserved hydrogen bond interactions between p1 and Y159, and between the main peptide anchor residue p5 and Q97 localized at the bottom of the peptide-binding groove (Fig. 5C and Supporting Information Video 1). However, in contrast to the previously published analysis of a variant of the human melanoma antigen MART-1 that displays decreased TCR recognition , no significant differences were found between Db/EGS and Db/EGP in our analysis of the distance fluctuations between main TCR contact residues.
We have previously demonstrated that substitution of position 3 in the Db-restricted melanoma-associated epitope EGS to a proline dramatically increased MHC stabilization and enhanced the immunogenicity of this TAA  to levels usually displayed by immunodominant viral epitopes (Fig. 2) . While several prediction servers including IEDB (http://www.immuneepitope.org), SYFPEITHI (http://www.syfpeithi.de/index.html), and netMHC (http://www.cbs.dtu.dk/services/NetMHC/) [31-33] predicted such an increase in binding affinity for EGP compared to EGS, proline at peptide position 3 is not always listed as favorable and, to our knowledge, a molecular mechanism has so far not been provided for this prediction. Based on our previous structural analysis of interactions between p3P and Y159 , mutated Db versions were generated. Thermostability measurements revealed that the stabilization advantage conferred by p3P was mediated by CH–π interactions between the pyrrolidine side chain of p3P and the aromatic ring of Y159 or F159 (Fig. 2). This is emphasized by the reduced stabilization advantage of EGP observed with Db-Y159F, which is consistent with the dependence of CH–π interactions on the electrostatic potential of the aromatic ring and mirrors previous observations in which tyrosine–proline interactions were found to be more energetically favorable compared to phenylalanine–proline interactions [19, 34, 35]. The importance of CH–π interactions for the stability and immunogenicity of MHC/peptide complexes as well their potential for the enhancement of MHC-restricted epitopes has to our knowledge not been discussed previously. It thus seems plausible that CH–π and aromatic ring-stacking interactions play an underappreciated role in the binding stabilization capacity and thus the immunogenicity of MHC-I- and MHC-II-restricted epitopes.
The data presented here suggest that CH–π interactions and the conserved juxtaposition between the third peptide residue and Y159 found in all known Db/peptide complex structures (Fig. 1) should shape the amino acid distribution at position 3 of the repertoire of Db-restricted peptides. Accordingly, we found that published sequences of large sets of peptides eluted from Db+ cells demonstrated an overrepresentation of proline as well as other aliphatic residues that also should be favorable for CH–π interactions at peptide position 3 [13, 14, 36]. This strong representation of proline is also notable considering the fact that the murine transporter associated with antigen processing seems to disfavor p3P-containing peptides [37, 38]. Finally, it should also be noted that Y159 is conserved across MHC-I alleles and that proline is for example described as a main anchor at position 3 for H-2Dd-restricted epitopes [39-41] and is regularly observed in H-2Kb-restricted epitopes [13, 42].
The remarkable similarity between the crystal structures of Db/EGS and Db/EGP explains the observed pMel TCR cross-reactivity between these two complexes  but not how interactions between Y159 and p3P result in higher Tm values. MD simulations suggested that the increased thermostability of Db/EGP was due to a pronounced reduction in flexibility of large parts of both the peptide-binding groove and the presented peptide. (Fig. 5A and B and Supporting Information Video 1). Furthermore, the simulations also suggested that conserved interactions were lost between Db residues and peptide positions 1 and 5 in Db/EGS compared to Db/EGP (Fig. 5C and Supporting Information Video 1). However, it should be noted that MD simulations provide only approximate analyses of the dynamic behavior of proteins and are in here limited by the relatively short time-scales so that detailed interpretations may require further experimental validation.
Additionally, SPR analysis revealed a subtle but consistent increase in pMel recognition of EGP complexes compared to each respective EGS complex, independent from the modifications introduced at the heavy chain position 159. Furthermore, tetramer staining of pMel+ T cells was significantly increased with Db/EGP compared to Db/EGS (Fig. 4C). The fact that the affinity increase conferred by p3P in Db/EGP remained conserved in Db-Y159A/EGP was unexpected (Fig. 4B), as no difference in thermostability between the complexes was observed (Fig. 1C). This could indicate that the advantage in pMel binding was independent from CH–π interactions between p3P and Y159 and may represent an intrinsic effect of the proline residue. While the underlying mechanisms remain to be explored, it seems likely that the introduction of a proline rigidifies the peptide thereby reducing its flexibility, which may facilitate TCR binding.
Although the extent to which the enhanced pMel recognition of Db/EGS applies to the general TCR repertoire as well as its functional relevance remain to be fully addressed, our study indicates that the increased thermostability of Db/EGP (Fig. 1) reduces the flexibility of the MHC/peptide interface (Fig. 5), which might be favorable for enhanced TCR binding. Indeed, several studies using a range of different TCRs and MHC/peptide complexes have previously indicated that TCR recognition can be tuned by the dynamic properties of MHC/peptide complexes [29, 43-45]. However, the relative contribution and universal nature of such dynamic tuning remain to be investigated further, requiring additional studies of several different TCR/MHC/peptide complexes.
In conclusion, we consider that the demonstrated ability of the p3P peptide modification significantly enhances MHC/peptide stability through CH–π interactions and increases TCR binding of class-I restricted MHC epitopes has important implications for the future design of enhanced APLs.
Materials and methods
Peptides and antibodies
Peptides EGP (EGPRNQDWL) and EGS (EGSRNQDWL) were purchased at >95% purity from GenScript (Piscataway, NJ, USA). Uncoupled and APC-conjugated anti-Cβ antibodies (H57-597) were purchased from eBioscience (San Diego, CA, USA).
Expression, refolding, and purification of WT and mutated Db variants
The Db/peptide complexes were produced as described previously [46, 47]. Refolded complexes were purified on Superdex 200 gel filtration columns (GE Healthcare, Uppsala, Sweden).
Thermostability measurements using circular dichroism (CD)
CD measurements were performed as described before , using protein concentrations between 0.15 and 0.25 mg/mL and a 2-mm quartz cuvette. Melting temperatures are an average of at least three measurements from at least two independent refolding assays per complex. Spectra were analyzed using GraphPad Prism 5 (GraphPad Software, La Jolla, USA).
Crystallization of Db-Y159F/EGP
Refolded Db-Y159F/EGP was concentrated to 4 mg/mL in 20 mM Tris pH 8.0. Crystallization was performed at 25°C using the hanging drop vapor diffusion method with a 4:2 protein:reservoir mixture. The best diffracting crystals were obtained in 1.6–1.8 M ammonium sulfate, 0.1 M Tris HCl pH 7.5–9.0.
Data collection, processing, and structure determination
Data collection was performed under cryogenic conditions (T = 100 K) using beamline ID14-4 at ESRF (Grenoble, France) . Crystals were soaked in crystallization solution supplemented with 25% glycerol. Diffraction data was processed and scaled using MOSFLM 7.0.3 , XDS , and SCALA . The crystal structure of Db-Y159F/EGP was determined by molecular replacement using PHASER  and the structure of Db/EGP (PDB ID: 3CH1) , with the peptide omitted, as a search model. Random 5% of reflections were set aside for Rfree cross-validation . Refinement was performed using Refmac5  and Coot . All figures were created using PyMol (PyMOL Molecular Graphics System, Version 1.2r1, Schrödinger, LLC).
Expression, refolding, and purification of soluble pMel TCR
Codon-optimized sequences encoding the soluble domains of the α and β chains (kindly provided by Professor T.N. Schumacher, NKI, Amsterdam, Netherlands) were cloned into pET-22b. A C-terminal His-tag was added to the alpha chain and stabilizing mutations were introduced as previously described . Expression, refolding, and purification of pMel were based on published protocols [30, 57]. Insoluble proteins were expressed in Rosetta pLysS cells (Merck) at 20°C. The inclusion bodies were washed and proteins extracted using 8 M Guanidine-HCl, 10 mM DTT, 5 mM EDTA. Thirty milligrams per liter of each TCR chain was refolded for 3–4 days at 4°C in a refolding buffer (50 mM Tris pH 8.0, 200 mM l-Arginine, 5 mM reduced glutathione, 0.5 mM oxidized glutathione, 1 mM EDTA, 4 M urea, and 0.5 mM PMSF) and thereafter dialyzed against 10 L of 10 mM Tris pH 8.0, 1M urea, 100 mM l-Arginine followed by dialysis against 10 mM Tris pH 8.0 for a total of 10 days. Refolded TCR was captured on a MonoQ ion exchange column (GE Healthcare) and further purified using both HisTrap and Superdex 200 chromatography columns (GE Healthcare).
Surface plasmon resonance (SPR) binding analysis
All measurements were performed on a BIAcore 2000 (GE Healthcare) at 25°C in HBS-EP (10 mM HEPES pH 7.4, 150 mM NaCl, 0.005% Tween-20, 3 mM EDTA). Soluble pMel was noncovalently coupled to the anti-Cβ antibody H57-597, immobilized on a CM5-chip via standard amine coupling, and 4-6000 response units of H57–597 were coupled, immobilizing 1000–2000 response units of pMel. A control surface without antibody was used as reference flow cell, and 1–100 μM of freshly produced Db/peptide complexes were injected over the chip surfaces at 5 μL/mL. The sample rack was cooled to 4°C during the run and complex integrity was confirmed using analytical gel filtration. Chip surfaces were regenerated using 5 μL 0.1 M Glycine-HCl, 500 mM NaCl, pH 2.5, Tween 0.05% at 30 μL/min after each injection. Data was analyzed with BIAevaluation 3.0 (BIAcore AB, Uppsala, Sweden) and GraphPad Prism. KD values were obtained from steady-state fitting of equilibrium-binding curves.
Tetramer staining experiments
Tetramerization of biotinylated Db/peptide complexes  was performed freshly before each stain by mixing with streptavidin-PE (BD Biosciences). Transgenic pMel mice were bred in the animal facility at the Leiden University Medical Center (Leiden, The Netherlands) according to national ethical guidelines. CD8+ T lymphocytes were purified by negative selection from single-cell splenocyte suspensions (Miltenyi Biotec, Germany), and 1 × 105 cells were washed with MACS buffer (PBS, 2 mM EDTA, 1% fetal calf serum) and stained with 10–600 nM of each tetramer together with an APC-conjugated anti-TCR Cβ antibody (H57–597) for 30 min at 25°C. TCR+ single-cell lymphocytes were gated, levels of tetramer staining evaluated as the geometric mean fluorescence intensity (MFI) of the tetramer positive T-cell population, and normalized against the highest staining level of each tetramer. Data were analyzed using the program FlowJo (Tree Star, Inc., Ashland, OR, USA) and GraphPad Prism.
Molecular dynamic simulations
Two sets of free energy MD simulations were performed using the sander module of Amber 9.0  and the parm03 force field  on the entire MHC/peptide complexes Db/EGP (PDB 3CH1) and Db/EGS (PDB 3CCH)  for 40 ns at 350 K. Each individual measurement was performed in two repeats. Proteins were initially placed in a reference box with periodic boundary conditions together with six Na+ counter-ions, twelve Na+ and Cl− ions as well as 9000 TIP3 water molecules . Short-range nonbonded interactions were taken into account up to a cut-off value of 9 Å. Long-range electrostatic interactions were treated with the particle mesh Ewald option  using a grid spacing of 0.9 Å. The structures were energy minimized and heated from 100 to 300K or 350K within 0.1 ns using positional restraints on solute atoms of constant force 50 kcal/mol Å−2, followed by step-wise removal of the positional restraints on the solute within another 0.1 ns. Resulting structural ensembles were evaluated with respect to RMSDs compared to starting structures and average residue-specific root mean square fluctuations over the course of measurements. Visualization of trajectories and preparation of figures were performed using VMD (Visual Molecular Dynamics) .
We gratefully acknowledge excellent technical assistance and help by Kathleen Loschinski and Guillaume Stewart-Jones, as well as access to synchrotron beamline 14.4 at ESRF (Grenoble, France). This work was supported by grants from the Swedish Research Council and the Swedish Cancer Society (AA).
Conflict of interest
The authors declare no financial or commercial conflict of interest.