Landscape mapping of shared antigenic epitopes and their cognate TCRs of tumor-infiltrating T lymphocytes in melanoma.

HLA-restricted T cell responses can induce antitumor effects in cancer patients. Previous human T cell research has largely focused on the few HLA alleles prevalent in a subset of ethnic groups. Here, using a panel of newly developed peptide-exchangeable peptide/HLA multimers and artificial antigen-presenting cells for 25 different class I alleles and greater than 800 peptides, we systematically and comprehensively mapped shared antigenic epitopes recognized by tumor-infiltrating T lymphocytes (TILs) from eight melanoma patients for all their class I alleles. We were able to determine the specificity, on average, of 12.2% of the TILs recognizing a mean of 3.1 shared antigen-derived epitopes across HLA-A, B, and C. Furthermore, we isolated a number of cognate T cell receptor genes with tumor reactivity. Our novel strategy allows for a more complete examination of the immune response and development of novel cancer immunotherapy not limited by HLA allele prevalence or tumor mutation burden.


Introduction
Cancer immunotherapy is a cancer treatment that is designed to harness the power of the body's immune system to fight cancer (Finn, 2018a). The magnitude and breadth of T cell responses can serve as a critical biomarker in cancer immunotherapy and conventional cancer treatments, such as chemotherapy and radiation (Schumacher et al., 2015). Immunotherapy is now widely considered a game-changer, as it is rapidly becoming the 4th pillar of cancer treatment, following surgery, radiotherapy, and chemotherapy.
The adoptive transfer of antitumor T cells can induce clinical responses in cancer patients Hu et al., 2018;Karpanen and Olweus, 2015;Rapoport et al., 2015;Robbins et al., 2015;Sahin and Türeci, 2018;Tran et al., 2016;Tran et al., 2017). However, comprehensive analysis of the specificity of antitumor T cell responses against mutated or non-mutated tumor antigens is lacking. This deficit is exacerbated by the fact that non-mutated antigens are greater in number than mutated antigens by multiple orders of magnitude (Finn and Rammensee, 2018) and the high polymorphism of HLA genes (Shao et al., 2018). The clinical importance of identifying antitumor T cell responses is highlighted by the results of recent clinical trials which attest that a shared antigen, such as NY-ESO-1, can be targeted with potent on-target efficacy and minimal off-target toxicity in T cell receptor (TCR) gene therapy (Rapoport et al., 2015;Robbins et al., 2015;Thomas et al., 2018). Unlike shared antigens, the vast majority of neoantigens are unique to each patient (Coulie et al., 2014;Finn, 2018b). The elucidation of T cell epitopes derived from shared antigens may facilitate the robust development of an efficacious and safe adoptive T cell therapy that is readily available to a larger cohort of cancer patients.
Analysis of antigen-specific T cells using peptide/HLA (pHLA) multimers has been utilized as a standard technique in immunology over the past two decades (Bentzen and Hadrup, 2017;Dey et al., 2019;Dolton et al., 2015). Recent progress in multimer technology has enabled the high-throughput analysis of antigen-specific T cells during immune responses (Bentzen and Hadrup, 2017). However, the production of each multimer is still cumbersome and expensive because in vitro peptide exchange of generated complete pHLA proteins requires multiple complicated steps (Toebes et al., 2006) and some class I alleles are not easily produced in bacteria (Migueles et al., 2000). Therefore, further advancements in multimer technology are needed to investigate the multitude of potential tumor-associated antigen (TAA) peptides presented by any given class I allele.
The combination of pHLA multimer-based analysis and functional assays are utilized to measure antigen-specific T cell responses. We have developed a novel peptide-exchangeable pHLA class I eLife digest The immune system is the body's way of defending itself, offering protection against diseases such as cancer. But to remove the cancer cells, the immune system must be able to identify them as different from the rest of the body. All cells break down proteins into shorter fragments, known as peptides, that are displayed on the cell surface by a protein called human leukocyte antigen, HLA for short. Cancer cells display distinctive peptides on their surface as they generate different proteins to those of healthy cells. Immune cells called T cells use these abnormal peptides to identify the cancer so that it can be destroyed.
Sometimes T cells can lack the right equipment to detect abnormal peptides, allowing cancer cells to hide from the immune system. However, T cells can be trained through a treatment called immunotherapy, which provides T cells with new tools so that they can spot the peptides displayed by HLA on the previously 'hidden' cancer cells.
There are many different forms of HLA, each of which can display different peptides. Current research in immunotherapy commonly targets only a subset of HLA forms, and not all cancer patients have these types. This means that immunotherapy research is only likely to be of most benefit to a limited number of patients. Immunotherapy could be made effective for more people if new cancer peptides that are displayed by the other 'under-represented' forms of HLA were identified.
Murata, Nakatsugawa et al. have now used T cells that were taken from tumors in eight patients with melanoma, which is a type of skin cancer. A library of fluorescent HLA-peptides was generated -using a new, simplified methodology -with 25 forms of HLA that displayed over 800 peptides. T cells were then mixed with the library to identify which HLA-peptides they can target. As a result, Murata, Nakatsugawa et al. found the cancer targets of around 12% of the tumor-infiltrating T cells tested, including those from under-represented forms of HLA. Consequently, these findings could be used to develop new immunotherapies that can treat more patients. multimer technology that can be applied for any given class I allele and bound peptide. Unlike comparable methods (Andersen et al., 2012;Migueles et al., 2000), our novel pHLA multimer technology enables a one-step peptide exchange in one tube and is more effective in class I alleles that are difficult to produce in bacteria. These technical advantages allow for a rapid, versatile, and less labor-intensive examination of the immune response. In addition, to functionally detect HLArestricted antigen-specific T cell responses, we previously demonstrated the versatility of gene-engineered artificial antigen-presenting cells (APCs) individually expressing various class I alleles as a single HLA allele (Butler and Hirano, 2014). Our strategy using a paired library of the novel pHLA multimer and human cell-based artificial APC system enables identification of novel class I epitopes and detection of cognate T cells in a comprehensive and systematic way.

Results
High-throughput staining of polyclonally expanded TILs with a panel of pHLA multimers The adoptive transfer of TILs can induce sustained antitumor responses in patients with metastatic melanoma in combination with lymphodepletion and IL-2 administration (Yang and Rosenberg, 2016). It is well established that melanoma TILs contain antitumor T cells that are specific for both non-mutated and mutated antigens (Andersen et al., 2012;Bentzen et al., 2016;Duhen et al., 2018;Ye et al., 2014). Here, TILs were isolated from eight metastatic melanoma patients, polyclonally expanded in vitro (Nguyen et al., 2019), and then examined for their shared antigen specificities recognizing epitopes in the context of all 25 different class I alleles present. The combination of structure-based analysis using pHLA multimers and functional analysis has been widely used to measure antigen-specific T cell responses (Bentzen and Hadrup, 2017;Bentzen et al., 2016;Newell et al., 2013). We initially stained the T cells using our novel peptide-exchangeable pHLA multimers with bound peptides that were previously known or predicted by publicly available algorithms (Andreatta and Nielsen, 2016). The previously known peptides were chosen from the Cancer Antigenic Peptide Database (https://caped.icp.ucl.ac.be) for all the class I alleles of eight melanoma patients, and we also synthesized several high-affinity binding peptides derived from TAAs based on prediction algorithms for the class I alleles that were positive in multiple patients (Figure 1 Table 1, all eight samples showed positivity for at least one of the chosen multimers. In addition to the previously known epitopes, we found that the newly predicted epitopes A*02:01/ABCB5 700-708 and C*05:01/tyrosinase 460-468 were immunogenic ( Figure 1A,B, Figure 2A-C).
T cell epitope determination of in vitro-expanded TILs using overlapping peptides Since pHLA multimer production requires the use of a peptide with a known exact sequence, it is not straightforward or practical to conduct high-throughput screening for new epitope peptides using a pHLA multimer-based strategy. To identify new epitope peptides, we conducted functional assays using artificial APCs, which can take up and process longer peptides and present epitope peptides via class I molecules, as stimulator cells (Butler and Hirano, 2014). The class I-matched artificial APCs were pulsed with overlapping peptides (20-mers with an overlap of 15 amino acids) to cover the whole proteins of six shared antigens (MART1, NY-ESO-1, SSX2, gp100, MAGE-A1, and survivin) that are frequently expressed by melanomas (Finn, 2018b; Supplementary file 2) and used as stimulators in cytokine ELISPOT assays. When stimulated with B*18:01-artificial APCs pulsed with MART1-derived overlapping peptides, B*18:01 + M87 TILs showed positive responses to two adjacent peptides with the shared sequence 21 YTTAEEAAGIGILTV 35 ( Figure 4A, Supplementary file 2a). Using a series of deletion mutant peptides, we determined the minimally required epitope peptide, 25 EEAAGIGIL 33 presented by B*18:01 molecules. Notably, this epitope partially overlaps with but is distinct from one of the most immunogenic epitopes, A*02:01/MART1 27-35 , suggesting that this region of the MART1 protein is an immunological hotspot (Cole et al., 2010;Kawakami et al., 1994). Importantly, the B*18:01/MART1 25-33 multimer successfully stained up to 9.2% of the polyclonally expanded M87 TILs, suggesting that the B*18:01/MART1 25-33 T cells were a dominant population of TILs ( Figure 4B,C). Similarly, we detected C*03:04/NY-ESO-1 92-100 T cells, the frequency of which was 18.2% of polyclonally expanded M31 TILs, and they were also a dominant population of TILs (  pulsed with the NY-ESO-1 125-133 peptide, the frequency of B*40:01/NY-ESO-1 125-133 T cells increased to 6.1% ( Figure 4E). Using a similar strategy, we identified HLA-B*40:01/gp100 448-458 , C*06:02/gp100 [190][191][192][193][194][195][196][197][198] , and C*07:01/gp100 479-487 T cells, which accounted for 0.11%, 1.2%, and 0.14% of CD8 + T cells among the polyclonally expanded TILs, respectively (Figure 4-figure supplement 1A-C,G-L, Supplementary file 2d). Following peptide-specific stimulation with class I-matched artificial APCs, the frequency of B*40:01/gp100 448-458 and C*07:01/gp100 479-487 T cells increased to 1.8% and 1.2%, respectively, excluding the possibility that the low percentages of staining represented false positives (Figure 4-figure supplement 1B,K). Although we performed similar experiments using overlapping peptides for SSX2, MAGE-A1, and survivin, no T cell response was observed. It should be noted that in this study, we did not investigate the expression of shared  antigens in the patient's own tumor tissues. Therefore, it is possible that the absence of a T cell response toward some shared antigens is due to the lack of their expression.

Primary T cells transduced with TIL-derived TCR genes recognize tumor cells
The results of TCR gene therapy clinical trials demonstrate that the adoptive transfer of T cells transduced with high-affinity TCR genes can induce sustained clinical responses in cancer patients Fesnak et al., 2016;Karpanen and Olweus, 2015;Morgan et al., 2006;Rapoport et al., 2015;Robbins et al., 2015). Several tumor-reactive TCR genes have been cloned from melanoma TILs Scheper et al., 2019). Indeed, some of these TCRs have been tested in TCR gene therapy clinical trials and shown to induce clinically relevant responses Karpanen and Olweus, 2015). For the majority of tumor reactivities that we identified, multimer-positive antitumor T cells were collected and their TCR genes were molecularly cloned. All the HLA-B-and HLA-C-restricted T cell populations contained one pair of TCR genes  Peptide-specific stimulation (+) Peptide-specific stimulation (-)  each. These results suggested that all the HLA-B-and HLA-C-restricted antigen-specific T cells found in this study were monoclonal (Table 1, Figure 5-figure supplements 1 and 2, Supplementary file  3). The antigen specificity and functional reactivity of the cloned TCRs were verified by multimer staining and ELISPOT assays of TCR-reconstituted T cells. For example, when reconstituted on primary T cells, B*18:01/MART1 25-33 TCR-transduced T cells were successfully stained with the cognate multimer ( Figure 5A) and strongly reacted with the MART1 25-33 peptide presented by surface B*18:01 molecules ( Figure 5B). Importantly, these cells were able to recognize B*18:01-matched and peptide-unpulsed tumor cells naturally expressing the MART1 gene ( Figure 5C).

Decoding the antigen specificity of polyclonally expanded TILs in melanoma
Using a library of pHLA class I multimers and artificial APCs, we were able to determine the specificity of 12.2 ± 7.3% (mean ± SD, max 25.9%, min 4.6%) of the CD8 + TILs from eight melanoma patients toward 3.1 ± 2.0 (mean ± SD, max 7, min 1) previously known and novel peptides derived from shared antigens across HLA-A, B, and C. Notably, the maximum accumulative total percentage of multimer-positive CD8 + T cells was 25.9% (M31 TILs) (Figure 6). At least one shared antigenderived epitope was identified in all the TILs that were studied. One TIL sample (M87) possessed reactivity to as many as seven different antigens in the context of four different class I alleles. Interestingly, we observed that the M66 TIL sample reacted with three different A*02:01-restricted gp100-derived epitopes. Furthermore, M31 TILs exhibited reactivity to three different NY-ESO-1 peptides via three different class I alleles, and M87 TILs possessed reactivity to two different HLArestricted epitopes each from gp100, tyrosinase, and MART1 (Table 1). These results underscore the established strong immunogenicity of these shared antigens.

Discussion
HLA-restricted T cell responses toward immunogenic peptides, mutated or non-mutated, can induce antitumor effects in cancer patients Rapoport et al., 2015;Robbins et al., 2015;Tran et al., 2016). Since the HLA gene is the most polymorphic gene in the human genome (Shao et al., 2018), previous human T cell research has largely been limited to the few HLA alleles that are prevalent in a subset of ethnic groups. Therefore, a precise and comprehensive understanding of the antigen specificity for antitumor T cell responses remains lacking, including against nonmutated tumor antigens and infrequent HLA alleles. The sheer number of potential non-mutated antigens and the high polymorphism of HLA genes may have hampered comprehensive analysis of the specificity of antitumor T cell responses.
The adoptive transfer of tumor-reactive T cells, such as TILs or T cells transduced with high-affinity TCR genes, can induce sustained tumor regression in some cancer patients  Rapoport et al., 2015;Robbins et al., 2015;Tran et al., 2016). In this study, using a library of the paired pHLA multimers and artificial APCs for 25 different class I alleles and greater than 800 peptides, we systematically and comprehensively mapped shared antigenic epitopes recognized by TILs from eight melanoma patients for all their class I alleles. Furthermore, we isolated multiple TCR genes highly tumor-reactive to shared antigens from the TILs. In particular, NY-ESO-1 is one of the shared antigens that have been most promising and extensively studied, and many clinical trials using A*02:01/NY-ESO-1 157-165 TCR genes are ongoing. The use of these newly cloned tumor-reactive B*07:02, B*40:01, and C*03:04-restricted NY-ESO-1 TCR genes may widen the applicability of anti-NY-ESO-1 TCR gene therapy and immune mobilizing monoclonal TCRs against cancer (ImmTAC) therapy beyond HLA-A*02:01-positive cancer patients. ImmTAC therapy trials targeting A*02:01/gp100 are also ongoing in patients with metastatic uveal melanoma (Liddy et al., 2012). The newly cloned tumor-reactive gp100 TCRs restricted by four different class I alleles (A*24:02, B*40:01, C*06:02, and C*07:01) could substantially extend the applicability of ImmTAC therapy targeting gp100.
The strategy employed in this study has enabled us to decipher the antigen specificity of tumorspecific T cells for any given HLA class I allele, regardless of allele frequency (Supplementary file 4), and for a very large number of peptides. Although we studied T cell responses to greater than 800 peptides derived from more than 90 proteins across all the 25 class I alleles expressed by eight melanoma patients, our study is still limited. Additional comprehensive studies are required to elucidate the full spectrum of antitumor T cell response in TILs. Since in vitro expanded TILs were studied in the current study, further assessment of TIL directly ex vivo is needed in future studies. Our strategy has also allowed us to build a large database of class I-restricted peptides and cognate tumor-reactive TCR genes at an unprecedented scale. This database will facilitate individual examination of the personal immune response with precision as well as the identification and validation of biomarkers to aid in patient-based selection of a cancer immunotherapy regimen. Furthermore, this database will help the robust development of novel cancer vaccines, TCR gene therapies, and ImmTAC therapies for patients without being limited by HLA allele prevalence or tumor mutation burden.  . All cell lines were routinely checked for the presence of mycoplasma contamination using PCR-based technology. TILs isolated from eight metastatic melanoma patients were grown in vitro as reported previously (Nguyen et al., 2019). High-resolution HLA DNA typing (American Red Cross) was performed for all TIL samples. Melanoma specimens were obtained from UHN Biospecimen Program. This study was conducted in accordance with the Helsinki Declaration and approved by the Research Ethics Board of the University Health Network, Toronto, Canada. Written informed consent was obtained from all healthy donors who provided peripheral blood samples.

Peptides
Synthetic peptides were purchased from Genscript (Piscataway, NJ) and dissolved at 50 mg/ml in DMSO. The purity of the vast majority of the peptides exceeded 85%. The peptide sequences are shown in Supplementary files 1 and 2.

Genes
All the HLA class I genes except for HLA-A*02:01 were fused with a truncated version of the human nerve growth factor receptor (DNGFR) via the internal ribosome entry site (Renaud-Gabardos et al., 2015). DNGFR-transduced cells were isolated using anti-NGFR mAb. The full-length MART1, NY-ESO-1, SSX2, and tyrosinase genes were cloned from Malme-3M, Me275, SK-MEL-37, and SK-MEL-28 cells via RT-PCR according to their published sequences, respectively. The full-length gp100, MAGE-A1, and MAGE-A3 genes were purchased from Dharmacon (Lafayette, CO). Genomic DNA of gp100 was isolated using PureLink Genomic DNA Mini Kit (Thermo Fisher Scientific). TCR genes were cloned via 5'-rapid amplification of cDNA ends (RACE) PCR as previously described (Nakatsugawa et al., 2016). The 5'-RACE PCR products were cloned into a retrovirus vector and sequenced. All genes were cloned into the pMX retrovirus vector and transduced using the 293GPG cell-based retrovirus system (Ory et al., 1996).

Cytokine ELISPOT analysis
IFN-g ELISPOT analysis was conducted as described previously (Kagoya et al., 2018). PVDF plates (Millipore, Bedford, MA) were coated with the capture mAb (clone 1-D1K; MABTECH, Mariemont, OH), and T cells were incubated with 2 Â 10 4 target cells per well in the presence or absence of a peptide for 20-24 hr at 37˚C. The plates were subsequently washed and incubated with a biotin-conjugated detection mAb (clone 7-B6-1; MABTECH). HRP-conjugated SA (Jackson ImmunoResearch) was then added, and IFN-g spots were developed. The reaction was stopped by rinsing thoroughly with cold tap water. ELISPOT plates were scanned and counted using an ImmunoSpot plate reader and ImmunoSpot version 5.0 software (Cellular Technology Limited, Shaker Heights, OH).

Immunoblotting
Equal amounts of proteins were separated on 8% gels by SDS-PAGE and transferred to Immobilon-P PVDF membranes (Millipore). The membranes were probed with the primary antibodies at 4˚C overnight. The membranes were then washed and incubated with HRP-conjugated anti-mouse IgG (Promega) or anti-rabbit IgG (Santa Cruz Biotechnology) secondary antibody at room temperature for 1 hr. The following antibodies were used: anti-SSX2 pAb (Thermo Fisher Scientific), anti-MAGE-A3 pAb (LifeSpan Biosciences), and anti-b-actin antibody (Santa Cruz Biotechnology). The signal was detected by Amersham ECL Prime Western Blotting Detection Reagent (GE Healthcare).

TIL culturing and rapid expansion protocol (REP)
The TIL expansion procedure was performed as previously published (Nguyen et al., 2019). Briefly, melanoma tissue was processed by cutting into~1 mm 3 fragments. Tissue fragments were either plated directly into 24-well plates or enzymatically dissociated in IMDM containing collagenase (Sigma) and Pulmozyme (Roche) and then plated in 24-well plates. Cells were cultured in complete medium (as previously described) and 6,000 IU/mL IL-2 and expanded for approximately 4 weeks prior to cryopreservation.
Expansion of CD8 + TILs in an HLA-restricted peptide-specific manner CD8 + TILs were purified through negative magnetic selection using a CD8 + T Cell Isolation Kit (Miltenyi Biotec). HLA class I-matched artificial APCs were pulsed with 10 mg/ml class I-restricted peptides of interest for 6 hr. The artificial APCs were then irradiated at 200 Gy, washed, and added to the TILs at an effector to target (E:T) ratio of 20:1. After forty-eight hours, 10 IU/ml IL-2 (Novartis), 10 ng/ml IL-15 (Peprotech), and 30 ng/ml IL-21 (Peprotech) were added to the cultures every three days.

Expansion of primary CD8 + T cells transduced with cloned TCRs
CD3 + T cells were purified through negative magnetic selection using a Pan T Cell Isolation Kit (Miltenyi Biotec). Purified T cells were stimulated with artificial APC/mOKT3 irradiated with 200 Gy at an E:T ratio of 20:1. Starting on the next day, activated T cells were retrovirally transduced with the cloned TCR genes via centrifugation for 1 hr at 1,000 g at 32˚C for three consecutive days. On the following day, 100 IU/ml IL-2 and 10 ng/ml IL-15 were added to the TCR-transduced T cells. The culture medium was replenished every 2-3 days.

Production of human cell-based pHLA multimers
The affinity-matured HLA class I gene was engineered to carry a Glu (E) residue in lieu of the Gln (Q) residue at position 115 of the a2 domain and a mouse K b gene-derived a3 domain instead of the HLA class I a3 domain (Wooldridge et al., 2007). By fusing the extracellular domain of the affinitymatured HLA class I gene with a Gly-Ser (GS) flexible linker followed by a 6x His tag, we generated the soluble HLA class I Q115E -K b gene. HEK293T cells were individually transduced with various soluble HLA class I Q115E -K b genes using the 293GPG cell-based retrovirus system (Ory et al., 1996). Stable HEK293T cells expressing soluble affinity-matured class I Q115E -K b gene were grown until confluent, and the medium was then changed. Forty-eight hours later, the conditioned medium was harvested and immediately used or frozen until use. The soluble HLA class I Q115E -K b -containing supernatant produced by the HEK293T transfectants was mixed with 100-1000 mg/ml of class I-restricted peptide of interest overnight at 37˚C for in vitro peptide exchange. Soluble monomeric class I Q115E -K b loaded with the peptide was multimerized using an anti-His mAb (clone AD1.1.10; Abcam) conjugated to a fluorochrome such as phycoerythrin (PE) or allophycocyanin (APC) at a 2:1 molar ratio for 2 hr at room temperature or overnight at 4˚C. The concentration of functional soluble HLA class I Q115E -K b molecules was measured by specific enzyme-linked immunosorbent assay (ELISA) using an anti-pan class I mAb (clone W6/32; in-house) and an anti-His tag biotinylated mAb (clone AD1.1.10; R and D systems) as capture and detection Abs, respectively.

Measurement of peptide exchange efficiency using ELISA
The efficiency of peptide exchange in soluble class I Q115E -K b monomers was assessed using a competition binding assay and ELISA. The monomer was loaded with 100 mg/ml biotinylated peptide and incubated overnight at 37˚C. Biotinylated peptide-monomer was purified and exchanged with PBS using Amicon Ultra filters (molecular weight cut-off (MWCO) 10 kDa) (Millipore Sigma, Burlington, MA) and mixed with 1 mg/ml HLA-restricted peptide of interest or an equivalent volume of DMSO followed by overnight incubation at 37˚C. ELISA plates were coated with anti-pan class I mAb (clone W6/32) at 10 mg/mL in PBS overnight at 4˚C. The plates were washed and blocked with 10% nonfat dry milk in PBS for 30 min at room temperature. Peptide-monomer was added and incubated for 2 hr at room temperature. After washing, the plates were incubated with streptavidin-conjugated alkaline phosphatase for 30 min at room temperature. Finally, the plates were washed and incubated with p-nitrophenyl phosphate (PNPP) substrate (Pierce, Rockford, IL) at room temperature. The reaction was terminated by adding 1 mol/L NaOH. The optic density (OD) (405 nm) was read (Spectramax 190 Microplate Reader; Molecular Devices, Sunnyvale, CA). The OD values from the control wells containing nonbiotinylated peptide were subtracted from the OD values in test wells containing biotinylated peptide. The efficiency of peptide exchange for each monomer was calculated as follows: Peptide exchange efficiency = [1 -(OD value with peptide/OD value with DMSO)] x 100. Every sample was assayed in triplicate wells. The biotinylated peptides used were HLA-A*02:01restricted telomerase 540-548 ILAK(-biotin)FLHWL, B*07:02-restricted MiHAg SMCY 1041-1051 SPSVDK(biotin)ARAEL, and C*07:02-restricted adenovirus B 585-593 FRK(-biotin)DVNMVL. The efficiencies of peptide exchange in the A*02:01, B*07:02, and C*07:02 monomers were shown in Figure 6-figure supplement 1. pHLA multimer staining T cells (1 Â 10 5 ) were incubated for 30 min at 37˚C in the presence of 50 nM dasatinib (LC laboratories). The cells were then washed and incubated with 5-10 mg/ml of PE-conjugated multimer for 30 min at room temperature, and R-phycoerythrin-conjugated AffiniPure Fab fragment goat anti-mouse IgG1 antibody (Jackson ImmunoResearch Laboratories) was added for 15 min at 4˚C. Next, the cells were washed three times and costained with an anti-CD8 mAb for 15 min at 4˚C. Dead cells were finally discriminated using the LIVE/DEAD Fixable Dead Cell Stain kit. For multiplex staining, after incubation with PE-conjugated multimer, the cells were washed three times and then incubated with 5-10 mg/ml of APC-conjugated multimer for 30 min at room temperature followed by costaining with an anti-CD8 mAb. The stability and multiplexing ability of the reagents were confirmed (Figure 6-figure supplements 2 and 3).

Statistical analysis
Statistical analysis was performed using GraphPad Prism 5.0. To determine whether two groups were significantly different for a given variable, we conducted an analysis using Welch's t test (twosided). P values < 0.05 were considered significant. exchange efficiency. (e) Peptides used for initial B*08:01 multimer staining. (f) Peptides used for initial C*05:01 multimer staining. (g) Peptides used for initial C*07:01 multimer staining. (h) Peptides used for initial C*07:02 multimer staining and measurement of peptide exchange efficiency. (i) Other peptides used for initial pHLA multimer staining. . Supplementary file 4. Twenty-five class I alleles for which paired multimers and artificial APCs were generated.
. Transparent reporting form Data availability All data generated or analysed during this study are included in the manuscript and supporting files.