Using X-ray Crystallography, Biophysics, and Functional Assays to Determine the Mechanisms Governing T-cell Receptor Recognition of Cancer Antigens

Human CD8+ cytotoxic T lymphocytes (CTLs) are known to play an important role in tumor control. In order to carry out this function, the cell surface-expressed T-cell receptor (TCR) must functionally recognize human leukocyte antigen (HLA)-restricted tumor-derived peptides (pHLA). However, we and others have shown that most TCRs bind sub-optimally to tumor antigens. Uncovering the molecular mechanisms that define this poor recognition could aid in the development of new targeted therapies that circumnavigate these shortcomings. Indeed, present therapies that lack this molecular understanding have not been universally effective. Here, we describe methods that we commonly employ in the laboratory to determine how the nature of the interaction between TCRs and pHLA governs T-cell functionality. These methods include the generation of soluble TCRs and pHLA and the use of these reagents for X-ray crystallography, biophysical analysis, and antigen-specific T-cell staining with pHLA multimers. Using these approaches and guided by structural analysis, it is possible to modify the interaction between TCRs and pHLA and to then test how these modifications impact T-cell antigen recognition. These findings have already helped to clarify the mechanism of T-cell recognition of a number of cancer antigens and could direct the development of altered peptides and modified TCRs for new cancer therapies.


Introduction
X-ray crystallography has been, and will continue to be, an extremely powerful technique to understand the nature of ligand-receptor interactions. By visualizing these interactions in atomic detail, not only is it possible to divulge the molecular mechanisms governing many biological processes, but it is also possible to directly alter contact interfaces for therapeutic benefit. Coupled with techniques such as surface plasmon resonance and isothermal titration calorimetry (to name just a couple), such modifications can then be analyzed biophysically to assess the direct impact on binding affinity, interaction kinetics, and thermodynamics. Finally, by performing functional experiments on relevant cell types, a detailed picture of the molecular and functional impact of modifications to receptor-ligand interactions can be gleaned, providing very specific mechanistic information. Overall, these types of methods provide an atomic resolution picture enabling the determination of how biological systems work, with attendant implications for diagnostic and therapeutic advances.
Our laboratory routinely uses these techniques to study the receptors that mediate human T-cell immunity to pathogens and cancer, in autoimmunity, and during transplantation. Here, we focus on the human CD8 + T-cell response to cancer, mediated by an interaction between the T-cell receptor (TCR) and human leukocyte antigen (HLA)-restricted tumor-derived peptides (pHLA). This is important because, although CD8 + T cells are able to target cancer cells, we and others have previously shown that anti-cancer TCRs suboptimally bind to their cognate pHLA 1,2 . Thus, many laboratories have attempted to alter either the TCR 3,4,5 or the peptide ligand 6,7,8 in order to increase immunogenicity and to better target cancer cells. However, these approaches are not always effective and can have severe side effects, including off-target toxicities 4,9,10 . Further research exploring the molecular mechanisms that govern T-cell recognition of cancer antigens will be vital to overcome these shortfalls.
In the present study, we focused on the responses against autologous melanoma cells by CD8 + T cells specific for a fragment of the differentiation melanocyte antigen glycoprotein 100 (gp100), gp100 280-288 , presented by HLA-A*0201 (the most commonly-expressed human pHLA class I). This antigen has been a widely studied target for melanoma immunotherapy and has been developed as a so-called "heteroclitic" peptide in which a valine replaces alanine at anchor position 9 to improve pHLA stability 12 . However, modifications to peptide residues can have unpredictable effects on T-cell specificity, demonstrated by the poor efficacy of most heterolitic peptides in the clinic 6,13 . Indeed, another heteroclitic form of gp100 280-288 , in which peptide residue Glu3 was substituted for Ala, abrogated recognition by a polyclonal population of gp100 280-288 -specific T cells 14,15 . We have previously demonstrated that even minor changes in peptide anchor residues can substantially alter Tcell recognition in unpredictable ways 6,16 . Thus, the study focused on building a more detailed picture of how CD8 + T cells recognize gp100 and how modifications of the interaction between TCRs and pHLA could impact this function.
Here, we generated highly pure, soluble forms of two TCRs specific for gp100 280-288 presented by HLA-A*0201 (A2-YLE), as well as the natural and altered forms of pHLA. These reagents were used to generate protein crystals to solve the ternary atomic structure of a human TCR in complex with the heteroclitic form of A2-YLE, as well as two of the mutant pHLAs in unligated form. We then used a peptide scanning approach to demonstrate the impact of peptide substitutions on TCRs by performing in-depth biophysical experiments. Finally, we generated a genetically modified CD8+ T-cell line, re-programmed to express one of the A2-YLE-specific TCRs, in order to perform functional experiments to test the biological impact of the various peptide modifications. These data demonstrate that even modifications to peptide residues that are outside of the TCR binding motif can have unpredictable knock-on effects on adjacent peptide residues that abrogate TCR binding and T-cell recognition. Our findings represent the first example of the structural mechanisms underlying T-cell recognition of this important therapeutic target for melanoma.

Protein Expression
1. Make gene constructs for the generation of soluble TCRs and pHLAs, as described in detail previously 17,18 . Design each construct with a 5' BamH1 and a 3' EcoR1 restriction site for insertion into the pGMT7 vector. 10. Use the K D values determined by SPR at different temperatures to calculate ΔG° using the standard thermodynamic equation (ΔG° = -RTlnK D ) 19 . NOTE: R = gas constant, T = temperature in K, ln = natural log. 11. Calculate the thermodynamic parameters according to the Gibbs-Helmholtz equation (ΔG° = ΔH − TΔS°) 19 .

Isothermal Titration Calorimetry (ITC)
1. Perform ITC experiments using an isothermal titration calorimeter. Inject 30 µM pHLA into the calorimeter cell and load 210 µM soluble TCR into the syringe. Use the following buffer conditions: 20 mM Hepes (pH 7.4) containing 150 mM NaCl. 2. Perform 20 TCR injections, each of a 2 µL volume. Calculate ΔH and K D using analytical software.

Crystallization, Diffraction Data Collection, and Model Refinement
1. Perform crystallization trials using a crystallization robot. 2. Grow crystals by vapor diffusion at 18 °C via the sitting drop technique in a 96-well plate with a reservoir containing 60 µL of crystallization buffer (mother liquor) 21 . 3. Concentrate the soluble pHLA to approximately 10 mg/mL (0.2 mM) in crystal buffer by spinning at 3,000 x g in a 10-kda molecular weight cut-off centrifugal concentrator. 4. For the co-complex structures, mix the TCR and pHLA at a 1:1 molar ratio to obtain a protein solution at approximately 10 mg/mL (0.1 mM). 5. Add 200 nL of pHLA alone, or the 1:1 molar ratio mix of TCR and pHLA, to 200 nL of each reservoir solution from the crystallization screen using a crystallization robot and score for crystals under a microscope after 24 h, 48 h, 72 h, and then once a week. 6. Harvest single crystals by manually mounting them in cryo-loops under a microscope and cryo-cool them by submerging and storing them in liquid nitrogen (100 K). NOTE: Loading crystals takes a bit of practice, and deciding which crystals are good enough for data collection comes with experience. As a rule of thumb, the larger and more regular the crystal, the better. 7. Collect data in a stream of nitrogen gas at 100 K.
NOTE: This data was acquired at the Diamond Light Source (DLS) national synchrotron science facility in the UK. 8. Analyze the data by estimating the reflection intensities with xia2 using both MOSFLM 22 and XDS packages 23 , and then scale the data with SCALA or AIMLESS 24 and the CCP4 package 25 . 9. Solve the structures with molecular replacement using PHASER 26 .
10. Adjust the model with COOT 27 and refine the model with REFMAC5 28 . 11. Prepare graphical representations with PYMOL 29 .
12. Calculate the contacts by using the "contact" program in the CCP4 package. Use a 4 Å cut-off for van der Waals contacts and a 3.4 Å cut-off for hydrogen bonds and salt bridges. 13. Calculate surface complementarity using the "SC" program in the CCP4 package. 14. Calculate the crossing angle of the TCR-pHLA complex, as described 30 . NOTE: For this study, the reflection data and final model coordinates were deposited with the PDB database (PMEL17 TCR-A2-YLE-9V PDB: 5EU6, A2-YLE PDB: 5EU3, A2-YLE-3A PDB: 5EU4, and A2-YLE-5A PDB: 5EU5).

Discussion
The protocols outlined here provide a framework for the molecular and cellular dissection of T-cell responses in the context of any human disease. Although cancer was the main focus of this study, we have used very similar approaches to investigate T-cell responses to viruses 32,33,34,35,36,37 and during autoimmunity 38,39,40 . Furthermore, we have used these techniques more broadly to understand the molecular principles that govern T-cell antigen recognition 2,19,41,42 . Indeed, the unpredictable nature of modifications to peptide residues, even those outside of the of the TCR contact residues, impacts T-cell recognition has important implications for the design of heteroclitic peptides. These findings have directly contributed to the development of novel T-cell therapies, including peptide vaccines 6,43 and artificial high-affinity TCRs 3,4,5,20,44 , as well as of enhanced diagnostics 45,46,47 .
Critical steps within the protocol The generation of a highly pure, functional protein is essential for all of the methods outlined in this paper.
Modifications and troubleshooting Difficulties in generating highly pure protein often relate to the expression of highly-pure, insoluble IBs from the E. coli expression system. Usually, modifying the expression protocol (e.g., inducing at different optical densities, using different E. coli strains, or using different media formations) resolves these issues.
Limitations of the technique These techniques use soluble protein molecules (TCR and pHLA) that are normally expressed at the cell surface. Thus, it is important to ensure that structural/biophysical findings are consistent with cellular approaches to confirm biological significance.
Significance of the technique with respect to existing/alternative methods Through the use of X-ray crystallography and biophysics substantiated through functional analysis, we and others have demonstrated that TCRs specific for cancer epitopes are generally characterized by low binding affinities 48 . This low TCR affinity may help explain why T cells are not naturally effective at clearing cancer. High-resolution atomic structures of complexes between anti-cancer TCRs and cognate tumor antigens are starting to reveal the molecular basis for this weak affinity. Furthermore, these studies are helpful for determining the mechanisms that underlie the therapeutic interventions designed to overcome this issue, seeding future improvements 16 . In this study, we examined the first structure of a naturally-occurring αβTCR in complex with a gp100 HLA-A*0201-restricted melanoma epitope. The structure, combined with an in-depth biophysical examination, revealed the overall binding mode of the interaction. We also uncovered an unexpected molecular switch, which occurred in a mutated form of the peptide, that abrogated TCR binding (assessed using surface plasmon resonance) and CD8 + T-cell recognition (functional experiments). It was only possible to demonstrate this new mechanism of HLA antigen presentation using the highresolution methods described.
Future applications or directions after mastering this technique Overall, our results demonstrate the power of X-ray crystallography and biophysical methods when combined with robust functional analyses. Using these approaches, it is possible to dissect out precise molecular mechanisms that govern T-cell antigen recognition. Indeed, it is also possible to use this approach to solve the structure of unligated TCRs, demonstrating how conformational changes can play a role during antigen discrimination 49,50,51 . A better understanding of the highly complex and dynamic nature that underpins TCR-pHLA interactions also has obvious implications for therapy design. Being able to directly "see" the molecules that are being therapeutically targeted, as well as the effect that modifications have on antigen recognition, will clearly improve the development of these medicines going forward. In this study, we show that even changes in a single peptide residue that is not heavily engaged by a TCR can unpredictably transmit structural changes to other residues in the HLA-bound peptide, which, in turn, dramatically alters T-cell recognition. A more complete understanding of the molecular mechanisms employed during T-cell antigen recognition will be hugely beneficial when designing future therapies for a wide range of human diseases.

Disclosures
The authors have no conflicts of interest or competing financial interests.