Divergent T-cell receptor recognition modes of a HLA-I restricted extended tumour-associated peptide

Human leukocyte antigen (HLA)-I molecules generally bind short peptides (8–10 amino acids), although extended HLA-I restricted peptides (>10 amino acids) can be presented to T cells. However, the function of such extended HLA-I epitopes in tumour immunity, and how they would be recognised by T-cell receptors (TCR) remains unclear. Here we show that the structures of two distinct TCRs (TRAV4+TRAJ21+-TRBV28+TRBJ2-3+ and TRAV4+TRAJ8+-TRBV9+TRBJ2-1+), originating from a polyclonal T-cell repertoire, bind to HLA-B*07:02, presenting a 13-amino-acid-long tumour-associated peptide, NY-ESO-160–72. Comparison of the structures reveals that the two TCRs differentially binds NY-ESO-160–72–HLA-B*07:02 complex, and induces differing extent of conformational change of the NY-ESO-160–72 epitope. Accordingly, polyclonal TCR usage towards an extended HLA-I restricted tumour epitope translates to differing TCR recognition modes, whereby extensive flexibility at the TCR–pHLA-I interface engenders recognition.

H uman leukocyte antigen (HLA)-I molecules are of central importance in the presentation of antigenic peptides, enabling CD8 + T cells to eliminate cancerous and virally infected cells. Typically, HLA-I molecules present peptides between 8 and 10 amino acids in length 1 , where the N-and Ctermini of the peptide are fixed within the P2 and PΩ binding pockets of the antigen (Ag)-binding cleft, respectively. However, HLA-I molecules can present extended peptides (>10 amino acids), whereby the N-and C-termini are similarly constrained within the HLA-I molecule, forcing the central region of the peptide to bulge from the Ag-binding cleft 2 . In addition, the HLA-associated peptide repertoire may be further expanded via N-terminal extensions on presented peptides, as observed for HLA-B*57:01 3 . C-terminal protrusions have also been observed 4 and were shown to extend out of the F pocket of the HLA-I binding groove 5 . Collectively, it has been estimated that extended peptides could comprise as much as 10% of the total HLA-I peptide repertoire [6][7][8][9][10] . Accordingly, it is difficult to predict a priori how extended peptides will be accommodated within the HLA-I molecule, and subsequently recognised by the T-cell receptor (TCR) 6,8,11 . Nevertheless, a number of studies have reported the importance of extended peptides in CD8 + T-cellmediated immunity-mostly in the axis of viral immunity 6,[12][13][14][15][16][17] . Extended peptides presented by HLA-I molecules can adopt highly dynamic conformations, thereby presenting differing energetic barriers for TCR ligation 1,18 . Extended peptides are generally considered to be challenging targets for TCR recognition owing to the dynamic nature of the central bulge [19][20][21][22][23][24][25] , in contrast to peptides of canonical length [26][27][28] . Extended peptides were frequently associated with highly biased T-cell repertoires 1,7,9,10,18,29 , thought to be resultant from conserved and HLA-I-centric TCR docking topologies [12][13][14][15]18,20,30 . Notwithstanding recent exceptions 31,32 , the majority of TCR-pHLA-I structural data available exhibits a fixed polarity, whereby the TCR αand β-chains are positioned over the HLA-I α2 and α1helices, respectively 1 , although how this is related to extended epitopes remained unclear. Relatively little is known regarding TCR engagement of extended peptides despite their apparent importance in tumour immunosurveillance 6,33 , with extended epitopes identified for the tumorigenic antigens CAMEL 34 , MAGE-A1 35,36 , and NY-ESO-1 17 . Presently, TCR recognition of extended epitopes has demonstrated starkly contrasting docking mechanisms. For example, one TCR docked atop the superbulged LPEP (BZLF1 [52][53][54][55][56][57][58][59][60][61][62][63][64] ) peptide, making limited contact with the HLA-I molecule itself 12 . A subsequent study described how another TCR docked towards the N-terminal end of this bulged peptide, making more extensive contacts with the HLA-I although the peptide conformation remained unchanged 16 . Conversely, another crystal structure described TCR recognition of an 11-amino-acid peptide, where the TCR flattened the bulged peptide upon ligation 13 .
NY-ESO-1 is an immunogenic cancer-testis antigen that is spontaneously expressed on a range of melanomas and other cancers including myelomas 17,37 . A key mediator of NY-ESO-1 immunity is CD8 + T cells with observations of CD8 + T-cell infiltration correlating with NY-ESO-1 expression and inversely correlating with tumour progression in vivo 38 . NY-ESO-1 restricted T cells therefore are of great interest due to their potential use for targeted immunotherapeutic treatment of tumours. Indeed, NY-ESO-1-specific engineered T cells have been studied for therapeutic use in multiple myeloma treatment 39 . Here T cells raised against NY-ESO-1 157-165 presented by HLA-A*02:01 were clonotyped 40 , structurally characterised 41 , and used for phage display to generate TCRs with picomolar affinity for the NY-ESO-1 157-165 antigen 42 . The engineered T cell then formed the framework (FW) for engineered T-cell therapy, with the NY-ESO-1 restricted T cells showing targeted antitumour activity in clinical trials 39 .
In addition to the HLA-A*02:01-directed response, an immunodominant extended peptide was identified, which was presented by HLA-B*07:02 17 . To identify the principles underpinning extended peptide recognition, here we investigated TCR binding of an immunodominant NY-ESO-1 13-amino-acid peptide (APRGPHGGAASGL) derived from positions 60-72 of the cancer-testis antigen, NY-ESO-1. We examined the HLA-B*07:02-NY-ESO-1 restricted CD8 + T-cell repertoire, previously shown to exhibit a diverse TRBV gene repertoire in vaccinated HLA-B*07:02 + melanoma patients 17 . Further, we isolated and characterised four individual T-cell clones that were representative of the many TRBV families. The binding of two distinct TCRs to NY-ESO-1 60-72 -HLA-B*07:02 was via either flattening or stabilisation of the extended peptide. This represents the first example of how a HLA-restricted peptide adopts markedly differing conformations upon engaging differing TCRs, thereby providing key insights into TCR recognition of extended peptides in the context of polyclonal TCR recognition.
The NY-ESO-1 60-72 peptide adopts a bulged conformation. To evaluate presentation of the NY-ESO-1 60-72 peptide by HLA-B*07:02, we determined the structure of HLA-B*07:02 presenting the NY-ESO-1 60-72 peptide to a resolution of 1.5 Å, which was significantly higher than the 2.1 Å resolution structure reported previously 17 ( Table 2). Superimposition of the HLA-B*07:02 molecule onto the structure of HLA-B*07:02 bound to a nonameric RFL9 peptide 48 illustrated little movement in the HLA-I molecule (root-mean-squared deviation (r.m.s.d.) of 0.2 Å) ( Fig. 2e), despite presenting distinct peptides. Given the anchored restraints at the peptide extremities, the central residues, encompassing P4-Gly→P11-Ser of the NY-ESO-1 60-72 peptide, bulged extensively out of the peptide-binding groove. Here, the solvent accessible surface area of the NY-ESO-1 60-72 peptide was double that of the RFL9 peptide (620 Å 2 compared to 290 Å 2 ), thereby providing an extended surface for TCR-mediated recognition. Notably, although solved to high resolution, P6-His-P8-Gly side chains were not clearly observed in the electron density ( Fig. 2d), which suggested conformational plasticity in this region of the peptide.
The total BSA of the KFJ5 TCR-NY-ESO-1 60-72 -HLA-B*07:02 interface was notably smaller at 1560 Å 2 . Here, the KFJ5 TCR ternary interface was dominated by TCR α-chain contacts, which contributed~75% of the BSA. The CDR1α, CDR2α, CDR3α and FWα loops contributed 30%, 4%, 38% and 2%, respectively, to the BSA of the interface, whereas only the CDR3β of the β-chain contributed (25% BSA) to the interface ( Fig. 3h and Supplementary Table 2). The KFJ5 β-chain formed limited contacts with the HLA-I surface and comprised contacts only from the CDR3β loop, including Arg109β, which contacted Thr73 and the main chain of Ala69 of the α1-helix, and Gln110β, which contacted Glu152 of the α2-helix ( Fig. 4e and Supplementary Table 2). In the KFJ5 TCR ternary complex, contact to the HLA-I surface was facilitated via an extension of the F-strand of the KFJ5 Vα domain over the N-terminal extremity of the α1-helix (Fig. 4d). Here the CDR3α loop is distended through the incorporation of four additional amino acids whereby the non-germline residues served to cap the peptide-binding groove of the HLA-I along with the germline Asn112α (CDR3α), which formed a hydrogen bond with Gln65 of the α1-helix (Fig. 4d and Supplementary Table 2). Given the commonality of the TRAV4*01 chains between the KFJ37 and KFJ5 TCRs, it was of interest that there were some conserved TRAV4*01-mediated contacts. Namely, Thr30α of the CDR1α made conserved van der Waals contacts to Gln163, Glu166 and Trp167 (Fig. 5f). Further, the neighbouring Asn31α made a conserved hydrogen bond to Trp167 and contacts with Arg62 (Fig. 5f). Interestingly, Tyr33α made conserved van der Waals contact to Glu163, as well as conserved coordination of P6-His (Fig. 4c, f), In addition, the TRAV4*01-encoded Thr30α and Gln55α made conserved contacts to the P6-His of the peptide (Fig. 5f). Further, non-germline commonalities included a saltbridge contact from Asp/Glu108α to Arg62 of the HLA-B*07:02 α1helix (Fig. 5f). Thus, while the KFJ37 and KFJ5 TCRs share the same TRAV-encoded region and adopt a similar positioning atop the pHLA-I, the extent and details within the respective interfaces differ markedly, despite recognising the same HLA-peptide complex.
Differential TCR recognition of the NY-ESO-1 60-72 peptide. The position of the KFJ37 TCR CDR loops on the HLA-B*07:02 surface allows the TCR to completely encapsulate the protruding NY-ESO-1 60-72 peptide (Fig. 4c). Indeed, the central part of the peptide was sandwiched between both CDR3 loops and three-tyrosine residues from CDR1α, FWβ and CDR2β (Supplementary Table 1). As observed in the binary structure, the central portion of the NY-ESO-1 60-72 peptide arched out of the peptide-binding groove, yet here was stabilised upon KFJ37 TCR ligation (Fig. 4c). Namely, the central region (P6-His→P9-Ala) of the NY-ESO-1 60-72 peptide formed extensive contacts with the KFJ37 TCR. Here, P6-His made hydrogen bonding contacts with Tyr33α of the CDR1α loop and Gln55α of the FWα (Fig. 4c and Supplementary Table 1) and also made a water-mediated contact to Ser113β of the CDR3β. The bulging peptide was further stabilised by Ser107β and Ser113β of the CDR3β loop, which formed hydrogen bonds to the peptide main chain at positions P9-Ala, and P8-Gly, respectively, in addition to numerous van der Waals contacts to the peptide (Fig. 4c and Supplementary Table 1). Indeed, the bulging peptide comprised 36% of the interaction surface (BSA) for KFJ37 TCR-mediated HLA-B*07:02 recognition, comparable to the~41% contribution to peptide-mediated recognition in SB27 TCR 12 .
In stark contrast, the NY-ESO-1 60-72 peptide underwent dramatic remodelling upon recognition by the KFJ5 TCR, due in part to the steric occlusion from the distended CDR3α loop and the coordination of the CDR3β loop. Here, the super-bulged NY-ESO-1 60-72 topology was deformed upon KFJ5 TCR ligation and adopted a 'helical' peptide conformation (P5-P8) (Fig. 4f), the impact of which resulted in a relatively flattened epitope upon KFJ5 TCR engagement, which in turn resulted in a markedly reduced (~20%) contribution to the interaction surface. Peptide recognition in the KFJ5 TCR ternary complex was almost entirely modulated via van der Waals interactions, notwithstanding the solitary hydrogen bond to the peptide from Asp113β of the CDR3β to P6-His, which flipped 180°into an upright conformation, compared to the KFJ37 TCR complex, while retaining contacts with Tyr33α of the CDR1α (Fig. 4f and Supplementary Table 1). The large deformation of the bulged peptide is coordinated via water-mediated interactions from the KFJ5 TCR CDR3 loops with Glu108α and Gln110β serving to stabilise the helical peptide main chain at P6-His and P5-Pro, respectively ( Fig. 4f and Supplementary Table 2). Van der Waals interactions included Glu108α, Ile109α and Leu110α of the CDR3α, to P4-Gly and P5-Pro of the peptide, which further contributed to the peptide interface with the α-chain of the KFJ5 TCR ( Fig. 4f and Supplementary Table 2). Conversely, the β-chain made limited contribution to the TCR-peptide interface, with only Arg109β and Gln110β of the CDR3β loop making van der Waals contacts with P10-Ala and P11-Ser peptide side chains (Supplementary Table 2).
Peptide plasticity is complemented by TCR malleability. To further understand the remodelling at the TCR-NY-ESO-1 60-72 -HLA-B*07:02 interface, we determined the structure of the KFJ5 TCR in an unligated form (Table 2). When comparing the KFJ5 TCR in a ligated and unligated form ( Supplementary  Fig. 5a), the CDR regions underwent significant remodelling. A large movement of the CDR3α was observed in which residues Glu108α to Phe116α ( Supplementary Fig. 5b & d) moved~9 Å upon recognition of the NY-ESO-1 60-72 -HLA-B*07:02 complex. A more modest remodelling was observed within the CDR regions of the β-chain, whereby the CDR3β displayed the largest shift of~2 Å upon ligation ( Supplementary Fig. 5c & d). The pronounced movement of the KFJ5 TCR CDR3α loop is indeed prerequisite for HLA recognition, due to major steric clashes with P5-Pro and P6-His of the NY-ESO-1 60-72 peptide. The remodelling suggests that the structural plasticity of extended epitopes can be accommodated to some extent by complementary remodelling of the TCR, but at an energetic cost, as illustrated by the reduced K D of the KFJ5 TCR.

Discussion
A generic tenet of T-cell-mediated immunity is the ability of the TCR to mould around a given HLA-I molecule when presenting a peptide of canonical length 1,46,49 . However, our understanding of TCR recognition towards extended HLA-I restricted peptides is unclear despite their growing importance in T-cell-mediated immunity 6 . Two contrasting TCR recognition modes have been described for such extended peptides. Firstly, a TCR was observed to perch atop a rigid and super-bulged 13-amino-acid peptide, thereby making limited contacts with the HLA-I molecule 12,16 . This same epitope was recognised differently by a distinct TCR, but the epitope retained its rigid conformation 16 . Secondly, a TCR flattened a bulged peptide upon recognition to enable a more conventional HLA-I centric footprint 13 . We describe a polyclonal CD8 + T cell response to HLA-B*07:02 presenting the immunodominant NY-ESO-1 60-72 peptide, which was detected in patients with melanomas expressing NY-ESO-1 17 , an antigen of importance in onco-immunology given its tumour-dependent expression and growing importance as a therapeutic target 39,43,50,51 .
Our ternary structures highlighted the extent to which TCR plasticity can be harnessed by the adaptive immune system to evoke an immune response. Our findings provide novel insight into the recognition of extended peptide epitopes via distinct Tcell populations, and highlight their utilisation of widely differing docking modes to recognise an oncogenic epitope of significant clinical interest.  Table 3. Cells were washed once with 100 μL PBS and were resuspended in 100 μL 1% paraformaldehyde (PFA; EMS) for 20 min at room temperature. Cells were washed once with PBS and stained with 50 μL of PBS-diluted (1:100) antihuman IFN-γ (BD Biosciences) in 0.2% saponin (Sigma-Aldrich) for 30 min at 4°C . Cells were washed once with PBS, resuspended in 100 μL PBS and analysed using FACSCanto II flow cytometer (BD Biosciences). Data were processed with FlowJo (Tree Star Inc.) software. NY-ESO-1 60-72 -HLA-B*07:02 tetramer staining was performed by staining cells with 100 μL of PBS-diluted (1:150) NY-ESO-1 60-72 -HLA-B*07:02 tetramer (0.44 mg mL −1 ) for 30 min at room temperature. Tetramers were synthesised at the Tetramer Production Facility of the Ludwig Institute for Cancer Research (LICR). Tetramer-stained cells (~2 × 10 6 ) were further stained with antibodies specific to CD3 and CD8 (BD Biosciences). Tetramer and antibodies were used at an empirically determined dilution factor with an optimum signal to noise ratio. Tetramer hi , CD3 hi and CD8 hi cells were individually sorted using FACSAria III flow cytometer (BD Biosciences) into a round-bottom 96-well plate containing 200 µL of complete RPMI-1640 medium (as described above but without FCS) supplemented with 10% (v/v) heat-inactivated human AB sera (LICR), 600 U recombinant human IL-2, phytohaemagglutinin (1 μg mL −1 , Sigma-Aldrich) and 3 × 10 5 healthy donors' PBMCs (Australian Red Cross Blood Service). PBMCs were mixed from three different donors in 1:1:1 ratio before being irradiated at 3000 rad using Gammacell 1000 Elite Irradiator 137 Cs source (Nordion). The gating strategy used to sort these cells is presented in Supplementary Fig. 6a. Tetramer + CD8 + clonal cells were further expanded on day 15 post cell-sorting for another 10 to 12 days in the same setting as described above, with~5000 clonal cells transferred into each well on a round-bottom 96-well plate initially.
Characterising the CD8 + T-cell clone TCRβ repertoire. All NY-ESO-1 60-72 -HLA-B*07:02-specific CD8 + T-cell clones were stained with a panel of 24 antibodies specific to TCR Vβ chains (Beckman Coulter) and analysed using FACS-Canto II flow cytometer (BD Biosciences), and data were processed with FlowJo (Tree Star Inc.) software. The gating strategy used to analyse these cells is presented in Supplementary Fig. 6b. CD8 + T-cell clones were co-incubated with these melanoma lines in 1:2 ratio in 10 μg mL −1 of Brefeldin A (BFA, Sigma-Aldrich) for 5 h followed by antibody staining for CD3, CD4 and CD8 (BD Biosciences), and ICS. The gating strategy used to analyse these cells is presented in Supplementary Fig. 6c. All cell lines were tested and/or treated for mycoplasma and were mycoplasma free.
TCR clonotyping and αβ TCR cell line generation. RNA was prepared from CD8 + T-cell clones with Trizol (Invitrogen) and was reverse transcribed with 5′ rapid amplification of cDNA end (RACE, Invitrogen) using gene-specific primers, Cα-GSP1 (GGGAAGAAGGTGTCTTCTGGAAT) and Cβ-GSP1 (GGCTGCTCAGGGCGTA). DNA fragments containing the sequence encoding the CD8 + T-cell clone's αor β-chains were obtained by 5′ RACE PCR amplification of cDNA with combinations of the 5′ RACE abridged anchor (Invitrogen) and gene-specific primers, Cα-GSP2 (GCTGTTGTTGAAGGCGTTTGC) or Cβ-GSP2 (GTGGCCAGGCATACCAGTGT). Each PCR-derived α or β gene was cloned into pGEM-T Easy vector (Promega) and sequenced. Five full-length genes encoding TCR αand β-chains (Table 1) separated by a hydrolase element P2A linker 58 were synthesised (GenScript) and subcloned into the pMIG expression vector 58 . SKW3 cells, a TCR α/β −/− and derivative of KE-37 cells of acute lymphoblastic leukaemia (Leibniz Institute DSMZ-German Collection of Microorganisms and Cell Cultures) were transduced with vectors encoding the sequenced αβ TCRs (Table 1, Supplementary Fig. 4) or the control LC13 TCR 44 . Retrovirus transduction was performed by first plating~10 6 HEK 293 T cells in tissue culture plate in 10 mL of complete DMEM medium containing 10% (v/v) FCS (Thermo Trace), Glutamax (2 mM), benzylpenicillin (100 U mL −1 ), streptomycin (100 µg mL −1 ), HEPES buffer (10 mM), 2-mercaptoethanol (50 µM), sodium pyruvate (1 mM) and 1 × nonessential amino acids (all from Gibco), and were incubated overnight at 37°C. 4 µg TCR 2A-linked/pMIG, 2 µg pVSV-G 58 and 4 µg pPAME 58 plasmids were dropwise added into a microtube containing 470 μL OptiMEM medium (Invitrogen) and 30 μL Fugene 6 reagent (Roche), gently tapped to mix and incubated at room temperature for 15 min. The mixture was dropwise added to overnight-cultured HEK 293 T cells for plasmid co-transfection and further incubated overnight at 37°C. Cell culture medium of the overnight transfected HEK 293 T cells was replaced with 10 mL of complete DMEM medium. About 10 5 SKW3 cells were seeded in tissue culture flask in 10 mL of complete DMEM medium and incubated overnight at 37°C. After 12 h incubation, HEK 293 T-cell culture medium containing retroviruses was carefully harvested and filtered through a 0.45 μm filter (Pall) to resuspend SKW3 cells and 10 μL polybrene (  The conserved interaction codon of the TRAV4*01 variable domain with the KFJ37 TCR and KFJ5 TCR coloured blue and green, respectively. Hydrogen and van der Waals bonds are shown by black and yellow lines with the conserved non-germline encoded Asp/Glu108α salt bridge also shown, in red mL −1 , Sigma-Aldrich) was added to the resuspended cells. Ten millilitres of complete DMEM medium was added to the HEK 293 T cells and incubated at 37°C. This process was repeated for another six times in every 12 h interval. Finally, TCR-transduced SKW3 cell culture medium was carefully aspirated, cells were washed three times with PBS, resuspended in complete DMEM and further incubated at 37°C for 7-10 days. For T-cell activation assays, TCR-transduced SKW3 cells (~10 5 ) were co-incubated with peptide-pulsed P03-BLCL cells (prepulsed with 10 −5 M NY-ESO-1 60-72 peptide for 2.5 h at 37°C) in 1:1 ratio for 5 h at 37°C. P03-BLCL cells without peptide pulsing were kept as a control. Cells were then stained with antibodies specific to CD3 and CD69 (BD Biosciences), washed once with PBS and subsequently fixed in 1% (v/v) paraformaldehyde (EMS) before flow cytometric analysis. The gating strategy used to analyse these cells is presented in Supplementary Fig. 6d. SPR. SPR experiments were conducted at 20°C on a BIAcore 3000 instrument in 10 mM HEPES, pH 7.4, 150 mM NaCl supplemented with 0.005% (v/v) surfactant P20 and 1% BSA. The analyte was captured on a CM5 sensorchip using the conformational monoclonal antibody 12H8 59 , whereby amine-coupled 12H8 was used to immobilise the KFJ4, KFJ5, KFJ15 and KFJ37 TCRs to a surface density of 300-350 response units. Serial dilutions from 200 to 0 µM of NY-ESO-1 60-72 -HLA-B*07:02 were injected over the immobilised TCR. The antibody surface was regenerated with Actisep (Sterogene) between two injections. All experiments were performed in duplicate with the kinetic parameters calculated using the BIAevaluation program using 1:1 Langmuir binding models with the addition of a drifting baseline parameter.
Crystallisation and structure determination. All crystals were grown by the hanging-drop, vapour-diffusion method at 20°C, crystals were flash-frozen in mother liquor supplemented with 30% (w/v) of the respective PEG, and data were collected at the Australian Synchrotron MX1 and MX2 beamlines, Melbourne. Data were processed with the program XDS 60 and were scaled with the SCALA program of the CCP4 suite 61 . Orthorhombic crystals of the KFJ5 TCR were grown in 0.1 M Tris-HCl pH 8.5 and 25% (w/v) PEG 3350 in 1:1 mixture with 8 mg mL −1 protein. KFJ5 TCR crystals belonged to the P2 1 2 1 2 1 space group and diffracted to 1.4 Å. Molecular replacement was conducted with the TRBV28 + TCR HA1.7 (PDB code 1J8H) 62 as the search ensemble. An initial run of rigid body refinement was performed using the Phenix REFINE software 63 , the CDR loops of the TCR were built in COOT 64 . Iterative rounds of refinement cycles lead to a final model with an R work and R free of 19.8% and 21.8%, respectively. Orthorhombic crystals of the NY-ESO-1 60-72 -HLA-B*07:02 complex were grown in 0.2 M NaCl and 20% (w/v) PEG 3350 at a 1:1 mixture with 5 mg mL −1 protein. The NY-ESO-1 60-72 -HLA-B*07:02 crystals belonged to the P2 1 2 1 2 1 space group and diffracted to 1.5 Å. Molecular replacement was conducted with RFL9-HLA-B*07:02 (PDB code 5eo0) 48 as the search ensemble. An initial run of rigid body refinement was performed using Phenix REFINE 63 to find optimal placement of the HLA-B*07:02 and β2m. Iterative rounds of refinement cycles lead to a final model with an R work and R free of 21.1% and 23.0%, respectively. Triclinic crystals of the KFJ37 TCR-NY-ESO-1 60-72 -HLA-B*07:02 and monoclinic crystals of the KFJ5 TCR-NY-ESO-1 60-72 -HLA-B*07:02 complexes were grown in 0.1 M MIB pH 7 and 25% (w/v) PEG 1500, and 0.1 M BTP pH 8.8, 0.2 M Na-Iodide, 14% (w/v) PEG 3350 and 2% (w/v) PEG 20,000, respectively, at a 1:1 mixture with 10 mg mL −1 protein. The KFJ37 TCR-NY-ESO-1 60-72 -HLA-B*07:02 and KFJ5 TCR-NY-ESO-1 60-72 -HLA-B*07:02 complexes belonged to the P1 and P2 1 spacegroups, and diffracted to 2.6 and 2.0 Å, respectively. Molecular replacement was conducted with the refined structures of the KFJ5 TCR and NY-ESO-1 60-72 -HLA-B*07:02 serving as separate search ensembles. The electron density at the interface was clear and unambiguous. Initial rounds of rigid body refinement were performed using Phenix REFINE 63 . The CDR loops of the respective TCRs were built in COOT 64 , where iterative rounds of refinement cycles lead to final models with an R work and R free of 19.3 and 24.3% for the KFJ37 TCR-NY-ESO-1 60-72 -HLA-B*07:02 complex and 20.4 and 25.6% for the KFJ5 TCR-NY-ESO-1 60-72 -HLA-B*07:02 complex. Atomic contacts were determined with the CONTACT program from the CCP4i suite 61 . The quality of all structures was confirmed at the Research Collaboratory for Structural Bioinformatics Protein Data Bank Data Validation and Deposition Services website. All presentations of molecular graphics were created with the PyMOL molecular visualisation system (The PyMOL Molecular Graphics System, Version 1.5.0.4 Schrödinger, LLC.
Data availability. The refined coordinate and structure factors files for the X-ray crystal structures of the NY-ESO-1 60-72 -HLA-B*07:02, KFJ5 TCR, KFJ37 TCR-NY-ESO-1 60-72 -HLA-B*07:02 and KFJ5 TCR-NY-ESO-1 60-72 -HLA-B*07:02 structures have been validated using the Protein Data Base validation site and the coordinates relating to the data reported in this study were deposited in the protein data bank with identification codes 6AT5, 6AT6, 6AVG, 6AVF, respectively. All remaining data are available within the article and its Supplementary Information files and from the corresponding authors on reasonable request.