key: cord-0299956-4sn49j85 authors: Ma, Ke-Yue; Schonnesen, Alexandra A.; He, Chenfeng; Xia, Amanda Y.; Sun, Eric; Chen, Eunise; Sebastian, Katherine R; Balderas, Robert; Kulkarni-Date, Mrinalini; Jiang, Ning title: High-Throughput and High-Dimensional Single Cell Analysis of Antigen-Specific CD8+ T cells date: 2021-03-05 journal: bioRxiv DOI: 10.1101/2021.03.04.433914 sha: 053474cf57cdfe169e8c81e76d0df5ca5334a1d5 doc_id: 299956 cord_uid: 4sn49j85 Although critical to T cell function, antigen specificity is often omitted in high-throughput multi-omics based T cell profiling due to technical challenges. We describe a high-dimensional, tetramer-associated T cell receptor sequencing (TetTCR-SeqHD) method to simultaneously profile TCR sequences, cognate antigen specificities, targeted gene-expression, and surface-protein expression from tens of thousands of single cells. Using polyclonal CD8+ T cells with known antigen specificity and TCR sequences, we demonstrated over 98% precision for detecting the correct antigen specificity. We also evaluated gene-expression and phenotypic differences among antigen-specific CD8+ T cells and characterized phenotype signatures of influenza- and EBV-specific CD8+ T cells that are unique to their pathogen targets. Moreover, with the high-throughput capacity of profiling hundreds of antigens simultaneously, we applied TetTCR-SeqHD to identify antigens that preferentially enrich cognate CD8+ T cells in type 1 diabetes patients compared to healthy controls, and discovered a TCR that cross reacts between diabetic and microbiome antigens. TetTCR-SeqHD is a powerful approach for profiling T cell responses. Due to their multi-facet role in controlling infection, fighting against cancer, and responding to vaccines, 38 the T cell has been subjected to extensive analysis 1, 2 . Recently developed multi-omic single cell 39 profiling methods have enabled multi-dimensional analysis in single T cells, such as combining ATAC-40 seq with single cell RNA-seq (scRNA-seq) 3 , DNA-labeled antibody based phenotyping with scRNA-41 seq (CITE-seq 4 and REAP-seq 5 ), and DNA-labeled antibody based phenotyping with targeted single 42 cell gene expression 6 . They have greatly advanced our understanding of T cell immune responses in 43 multiple disease settings [7] [8] [9] . 44 T cell antigen-specificity, although critical to T cell function and T cell based immunotherapy 45 development, has been challenging to analyze in a high-throughput manner until recently. Using T cell 46 trogocytosis 10 or reporter genes 11-14 , a suite of technologies have been developed in this areas 47 enabling the high-throughput screening for T cell antigens, such as SABR 11 , MCR-TCRs 12 , T-scan 13 , 48 granzyme B based target cell tag 14 . These methods have provided much needed T cell epitope 49 information in the context of cancer 11-14 and SARS-CoV-2 15 . However, because these methods use 50 expanded T cells or TCR-transduced cell lines, they do not support the profiling of phenotype or gene 51 expression in primary T cells. Peptide-MHC tetramer (pMHC) based methods can be applied to 52 primary T cells. In combination with mass cytometry, it has been shown that over 100 antigens can be 53 screened in parallel along with phenotype 16, 17 . Yet, the destructive nature of mass cytometry prevents 54 the acquisition of TCR sequences, which is critical for T cell antigen validation. 55 We previously developed TetTCR-Seq 18 to link the T cell receptor (TCR) sequence information to its 56 cognate antigens by sequencing DNA-barcoded pMHC tetramers bound on individual single T cells 18 . 57 TetTCR-Seq enables the screening of hundreds of antigens on primary T cells. To better understand 58 functional profiles of antigen-specific CD8 + T cells, a method to simultaneously profile two other 59 "dimensions" of parameters, gene expression and surface-protein expression, is imperative. In this 60 study, we describe a high-dimensional TetTCR-Seq (TetTCR-SeqHD) method that enables us to 61 simultaneously profile paired TCR sequences, cognate antigen specificities, targeted gene expression, 62 and selected surface-protein expression in tens of thousands of single cells from multiple biological 63 samples. Using a mixture of T cell clones, we demonstrated high precision and recall rates with 64 TetTCR-SeqHD. We then developed a panel of 215 endogenous antigens, majority of which are type 65 1 diabetes (T1D) related peptides, and 65 foreign antigens. Using these antigens on a set of primary 66 CD8 + T cells from a cohort of healthy individuals and T1D patients, we showed that foreign pathogen-4 specific T cells exhibited infection dependent states. Analyzing the 209 T1D related peptides, we 68 identified three peptides that have elevated antigen-specific CD8 + T cell frequencies in T1D patients 69 compared with healthy controls. Transducing TCR sequences identified into human CD8 + T cells 70 allowed us to functionally validate these TCRs including one that cross-reacts between a T1D related 71 peptides and a peptide derived from microbiome. TetTCR-SeqHD together with the flexibility and the 72 speed of generating high-throughput antigen libraries through in vitro transcription and translation 73 (IVTT), created a powerful technology to characterize the function and phenotype and track clonal 74 lineage of antigen-specific T cells at single cell level in one assay. 75 Results 76 In Tetramer associated TCR Sequencing High Dimension (TetTCR-SeqHD) technology, each peptide 78 encoding DNA oligo was individually in vitro transcribed/translated (IVTT) to generate corresponding 79 peptide, which was later loaded onto MHC molecules. Then peptide-MHC (pMHC) tetramer was 80 tagged with its corresponding peptide oligo bearing a 3' polyA overhang, which serves as the DNA 81 barcode for that antigen specificity (Fig. 1a) . This enables the tetramer barcodes to be captured by 82 BD Rhapsody™ beads and reverse transcribed together with other mRNAs captured, including TCR 83 transcripts ( Fig. 1b) . At the same time, 59 DNA-labeled antibodies 19 were used to stain cells. Similar 84 to the tetramer, the DNA barcodes labeled to the antibodies were captured by the same beads. Thus, 85 TetTCR-SeqHD integrates TCR sequencing with TCR antigen-specificity, gene expression, and 86 phenotyping in tens of thousands of single cells for hundreds of antigens simultaneously. 87 We first assessed the precision of TetTCR-SeqHD to detect correct antigen specificities using 88 polyclonal CD8 + T cells sorted and stimulated with seven known antigens including potentially cross-89 reactive epitopes (Supplementary Table 1) . PE-labeled, DNA-barcoded tetramers were used to stain 90 cultured T cells. Tetramer + CD8 + T cells were sorted (Supplementary Fig. S1 ) and loaded to BD 91 Rhapsody™ to perform reverse transcription and PCRs. A total of 4,533 single cells were recovered 92 after sequencing (Supplementary Table 2 ). Further filtering of low-quality cells and putative multiplets 93 led to 4,462 cells retained, among which, a median of 140 genes were detected and TCRα and TCRβ 94 capture efficiencies were 89% and 94% respectively. For each of these six polyclonal CD8 + T cell 95 cultures, our previously developed MIDCIRS technology 20 was used to assess TCRβ sequence 96 diversity and distribution. These TCRβ sequences were then set as internal references for identifying 97 true antigen specificities (Supplementary Table 3 ). Although the tetramer negative cells had a lower 98 level of target gene expression, a similar level of gene expression were observed among different 99 antigen-specific T cell clones (Supplementary Fig. S2a, b ). An average of 17,249 reads per cell were 100 sequenced for tetramer DNA-barcodes. 101 We detected antigen binding events based on MID count distribution of tetramer DNA-barcodes in 102 each single cell, which helped us to define antigen specificity and possible cross-reactive binding 103 antigens for individual T cells (See methods). Using the known TCR sequences from T cell clones, 104 their known antigen specificities, and detected antigen-specificity by pMHC DNA-barcode, we showed 105 that the precision, which is antigen-matched TCRs divided by antigen-specific TCRs identified by 106 pMHC DNA-barcodes, is over 98% and the recall, which is antigen-matched TCRs divided by all TCRs 107 determined by TCRβ clonality, is over 80%, except for GAD specific clones (Fig 2a, b) . Additional 108 analysis revealed the lower recall rate for GAD specific clones was due to one non-GAD binding clone 109 (TCRβ: CASRFLGTEAFF) that accounted for 26% of all GAD specific T cells, which is likely to be a 110 non-specific contaminant in the polyclonal culture ( Fig. 2c-h) . 111 112 To further demonstrate the advantages of TetTCR-SeqHD in characterizing antigen-specific CD8 + T 114 cells, we curated a panel of 215 endogenous and 65 foreign antigens from the IEDB database and 115 based on peptide-MHC class I binding prediction (Supplementary Table 4 , see methods) covering 116 HLA-A01:01, HLA-A02:01, and HLA-B08:01 alleles and applied TetTCR-SeqHD in ten non-type 1 117 diabetes (T1D) healthy donors and eight T1D patients (Supplementary Table 5 ). Endogenous and 118 foreign peptides were UV-exchanged 18 onto PE and APC-labeled tetramers respectively. CD8 + T cells 119 were stained and sorted similar to the polyclonal T cell cultures (Supplementary Fig. S3 ). An HCV 120 antigen-specific CD8 + T cell clone was spiked in to primary CD8 + T cells for all HLA-A02:01 donors. A 121 total of 35,168 cells were recovered across four experiments. An average of 50,000 reads per cell 122 were sequenced covering all six groups of attributes (Supplementary Table 2 had TCRα and TCRβ captured, with pairing efficiency of 34%. Since the primary CD8 + T cells were 126 recovered from frozen samples, lower gene and TCR capture rates were seen compared with cultured 127 We started by performing joint modeling of RNA expression and surface-protein expression using 129 totalVI 21 , followed by dimensionality reduction using uniform manifold approximation and projection 130 (UMAP) 22 and single cell clustering with the Leiden algorithm ( Fig. 3a) 23 . Minimum batch effects among 131 chips were detected (Fig. 3b) . A total of 13 clusters were identified, consisting of major conventional 132 CD8 + T cell phenotypes including naïve T cells (Tnaïve, clusters 1-4), central memory T cells (Tcm, cluster 133 6), effector memory T cells (Tem, clusters 8-10), effector T cells (Teff, clusters 11-12) and transitional T 134 cells between effector and memory populations (Ttrans, cluster 7) based on CCR7 and 135 CD45RA/CD45RO protein expression, spike-in HCV specific clone (cluster 13) and CD56 + T cells, 136 which are likely to be NK-like T cells 24 (cluster 5) (Fig. 3c, d) . The large number of primary CD8 + T 137 cells processed and the combined analysis of target gene and surface-protein expression provided a 138 superior resolution to identify sub-populations. While clusters 8, 9, and 11 represent early stages of 139 Tem and Teff, clusters 10 and 12 represent late stage Tem and Teff based on the graduate changes of 140 gene/protein expression. Similarly, Tnaive was also further separately into four clusters (1-4) (Fig. 3d , 141 Supplementary Fig. S4 ). Of note, we found that cluster 5 (CD56 + T cells) is characterized by a low 142 tetramer DNA-barcode signal fraction (Supplementary Fig. S5) , and no enrichment of antigen-143 specific CD8 + T cells was identified. Among the 12 clusters of primary CD8 + T cells identified from all 144 donors, the four Tnaïve clusters and the CD56 + T cell cluster have the lowest TCR clonality, which is 145 ubiquitous in all donors. However, different activated T cell subpopulations display various degrees of 146 clonal expansion and clusters 8, 9, 10, and 12 (Tem and Teff) have a relatively high TCR clonality in 147 majority of donors (Fig 3e) . Table 6 ). Almost all of the clonally 155 expanded TCRs had unique antigen specificities identified, confirming the precision of TetTCR-SeqHD 156 in primary CD8 + T cells from human PBMCs (Fig. 3f) . We further used the HCV clone to characterize 157 the precision and recall of TetTCR-SeqHD in primary CD8 + T cell experiments. Of the cluster 13 158 identified to harbor the HCV specific spike-in clone, there were total of 623 cells, 536 (86%) of which 159 were accurately identified as binding to at least one HCV wildtype (WT) and associated variant 160 antigens (Fig. 3g) . Of these cells, a total of 421 cells were identified to have the same paired TCRα/β 161 sequences as the HCV specific clone in this experiment. 91% of them bind to at least one HCV wildtype 162 (WT) and associated variant antigens (Fig. 3h) . 40 viral antigens that were detected in greater than 5 163 cells across all donors were selected for further analysis (see Methods). As expected, different T cell 164 phenotypic clusters are comprised of distinct antigen specificities, with endogenous antigens 165 occupying Tnaïve, while foreign antigens populating non-naïve T cell clusters (Fig. 4a) . In general, 166 different donors, regardless of their T1D status, presented varying frequency and phenotypic profiles 167 of viral antigen-specific CD8 + T cells, possibly due to different infection or vaccination history ( Fig. 4b-168 e). However, we also found that some viral antigens induced distinct T cell phenotypes. Influenza 169 antigen experienced T cells are mostly within cluster 7, where T cells display Tim3 + , CD25 + and CD26 + 170 phenotype 25-27 (Fig. 4c, Supplementary Fig. S7 ). Epstein-Barr Virus (EBV) antigens showed 171 distinguishable phenotypes compared with influenza antigens (Fig. 4c, Supplementary Fig. S8 ). Two 172 different categories of EBV antigens originated from lytic and latent viral proteins also present distinct 173 phenotypes. Antigens from latent viral proteins, such as LMP1 and LMP2, preferentially induced T 174 cells in central memory states (cluster 6), while lytic viral proteins, such as BRLF1 and BLMF1, display 175 effector and effector memory phenotypes (clusters 8, 9, 10, and 12) (Fig. 4c) , consistent with previous 176 findings using CyTOF 16 . We also found M1 specific CD8 + T cells display a more uniform phenotype 177 distribution among donors, compared with other antigens (Fig. 4d, e) . 178 Another advantage of TetTCR-SeqHD is its capacity to identify putative cross-reactive CD8 + T cells. 179 Similar to TetTCR-Seq 18 , 85% of HCV specific clone display binding to all five HCV antigens 18 (Fig. 180 4f). We also examined the cross-reactivity detection in primary CD8 + T cells using MART1 antigens. 181 MART1 wildtype antigens (MART127-35 nonamer and MART126-35 decamer) and its variants have been 182 widely used as a model system of human cancer antigens. By changing one or two amino acids, such 183 as MART126-35 A27L and MART126-35 E26A/A27L, it was noted previously that the resulting variant 184 peptides greatly improved the binding/stability of peptide/HLA-A*0201 complexes and enabled the 185 otherwise weak wildtype antigens to potent immunogens 28, 29 . We thus used these set of peptides and 186 studied the robustness of TetTCR-SeqHD in detecting both strong and weak pMHC ligands. Among 187 cells with same tetramer binding frequency greater than 10, a total of 2,308 cells were identified to 188 bind MART1 WT or variant antigens. 84% of cells bind to more than one MART1 WT or variant 189 antigens (Fig. 4g) . Interestingly, our method also detected previously noted cross-reactivities among 190 the PGT-178 (LLAGIGTVPI) peptides and a MART1 variant antigen (ELAGIGILTV) 30 and an 191 additionalMART1 variant cross-reactive antigen (ALAGIGILTV), despite five or more amino acid 192 differences in these peptides (Fig. 4h) . 193 194 Among 209 T1D-related autoantigens included in the antigen pool, 106 and 102 different autoantigens 196 were detected more than 3 times in 1,109 and 814 T1D antigen-specific cells from T1D and non-T1D 197 donors, respectively. The total T1D autoantigen tetramer + CD8 T cell frequency was comparable 198 between T1D and healthy donors (Supplementary Fig. S9 ). However, comparing the frequency of 199 T1D autoantigen-specific CD8 + T cells individually, we found INS-WMR-10, PPI-29-38 and PTPRN-200 805-813 specific cells exhibit a significantly higher cell frequency in T1D patients compared to healthy 201 control donors within this donor cohort (Fig. 5a) . Among them, PTPRN-805-813 was reported before 202 as a potential marker in PBMC of T1D patients 31 and PPI-29-38 was identified as an HLA-A02:01 low 203 binder but present in T1D patients 32 . To ensure the sensitivity of our analysis, we increased the 204 tetramer MID negative threshold to 15 and compared the frequency difference between T1D and 205 healthy donors again. Five antigens were identified, including previously identified INS-WMR-10 and 206 PTPRN-805-813, further validating the potential of these two antigens to distinguish between T1D and 207 healthy donors (Supplementary Fig. S10a) . We also noticed varying degree of clonal expansion in 208 T1D autoantigen-specific T cells isolated from different T1D patients, revealing the complexity of 209 antigen landscape in T1D (Supplementary Fig. S10b ). This could also be caused by limited sampling 210 from PBMC. 211 In addition, we identified an expanded T cell clone cross-recognizing three different antigens, 212 INSDRIP-1-9, DUF5119-124-133 and PTPRN-797-805 in a type 2 diabetes patient. This led us to test 213 the plasma banked from the same blood draw and showed the patient was positive for GAD (Glutamic 214 Acid Decarboxylase) reactive auto-antibody. Further review of the medical record showed that the 215 patient was later diagnosed with Latent Autoimmune Diabetes in Adults (LADA) after the sample was 216 collected for this study. Interestingly, INSDRIP-1-9 is derived from an alternative open reading frame 217 within human insulin mRNA and a significantly higher levels of INSDRIP-1-9 + specific CD8 + T cells 218 were reported to be detected in T1D patients 33 . DUF5119-124-133 is derived from Bacteroides fragilis 219 /thetaiotaomicron, a common bacteria found in human gut microbiota 34 and PTPRN-797-805 is derived 220 from IA2 protein, a previously known T1D autoantigen 35 . This is likely due to cross-reactivity of the 221 three antigens by the same TCR. To confirm the result by TetTCR-SeqHD, we thus transduced this 222 TCR together with some TCRs identified among T1D and healthy subjects to further validate the 223 accuracy of TetTCR-SeqHD (Supplementary Table 7) . Tetramer staining (Fig. 5b) and antigen 224 stimulation experiments (Fig. 5c) both confirmed that cognate TCRs identified from TetTCR-SeqHD 225 can bind and be stimulated by respective antigens. 226 In this study, we developed a method to simultaneously profile TCR sequences, cognate antigen 228 specificity, gene expression, and surface-protein expression, for single primary CD8 + T cells in a high 229 throughput manner. We addressed the precision of TetTCR-SeqHD, ability to profile TCR cross-230 reactivity, as well as its application to study diverse phenotypes of foreign-and self-antigen specific 231 CD8 + T cells. By using in vitro cultured polyclonal T cells with known antigen specificities and TCR 232 sequences, TetTCR-SeqHD established over 98% precision for detecting the correct antigen 233 Recently, DNA-barcoded dextramer, dCODE ™ dextramer ® was adapted by 10x Genomic platform to 235 enable profiling of antigen-specific CD8+ T cells. However, the dCODE ™ dextramer ® suffers from the 236 high cost of generation of dextramers, thus lacks the flexibility to screen large antigen panels. This 237 prevents it from profiling antigens in a high-throughput manner. By combining in vitro transcription and 238 translation (IVTT) with UV exchange technique, TetTCR-SeqHD enables the creation of a panel of 239 antigens (in hundreds) affordably and quickly (within a week). Therefore, we created a large panel of 240 antigens consisting of foreign-specific antigens derived from various virus and self-specific antigens 241 derived from known T1D autoantigens, and profiled CD8 + T cells recognize these antigens from 242 healthy subjects and T1D patients. 243 With the ability to profile targeted gene expression and surface-protein expression simultaneously 244 using BD Rhapsody™ platform, we resolved 12 clusters for primary CD8 + T cells, plus 1 cluster for in 245 vitro cultured HCV-specific CD8 + T cell clone. Most importantly, T cell phenotypic and functional sub-246 classes represented by gradual changes of gene expression were revealed among these 12 clusters, 247 from naïve to early stage of effector and memory populations to transitional state between effector and 248 memory, and to late stage of effector and memory populations. By investigating the composition of 249 phenotypic clusters for each antigen, phenotype signatures of distinct antigens were assessed. We 250 found that viral antigens from influenza and EBV display distinct phenotypes. Influenza-specific CD8 + 251 T cells were mostly enriched in cluster 7, displaying a transitional phenotype between effector and 252 memory populations, while EBV-specific CD8 + T cells were largely memory and effector populations. 253 Similar phenotypic differences between EBV latent and lytic antigens were observed previously using 254 mass cytometry 16 . This example further validates the robustness of TetTCR-SeqHD to capture the 255 phenotypic profiles of antigen-specific CD8 + T cells. Moreover, studied subjects also showed diverse 256 phenotype signatures of influenza-and EBV-specific CD8 + T cells, due to different viral infection (or 257 vaccination) history. 258 In addition to its high precision and high-throughput capacity, TetTCR-SeqHD also enables detection 259 of cross-reactivity of CD8 + T cells. We examined cross-reactivity in both in vitro cultured HCV-specific 260 CD8 + T cell clones and primary CD8 + T cells. We not only detected cross-reactivity among HCV and Peptide-encoding DNA oligonucleotides were purchased from Sigma Aldrich. 50nM DNA templates 298 were first amplified by PCR as described previously with modifications (Zhang et al.) . 1µM IVTT_r and 299 Table 8 ) were used following below reaction conditions: 95°C 3min; 300 then 22 cycles of 95°C for 20s, 59°C for 30s, 72°C for 30s; then 72°C for 5min. The PCR product was 301 then diluted with 50µl of nuclease free water and proceeded to IVTT reaction. 302 IVTT generated peptides were mixed with biotinylated pMHC containing a UV-labile peptide. The final 304 concentration of biotinylated pMHC is 0.2mg/ml. Individual pMHC was formed through UV exchange 305 as described previously 38 . Individual pMHC tetramer and tetramer library pool were generated and 306 tested as described previously 18 . Tetramer pool can be stored in 4°C temporarily. 307 Anti-CD2 antibody was purchased from Biolegend (Clone RPA-2.10, Biolegend). Amine modified 309 oligonucleotide was purchased from Sigma Aldrich (Supplementary Table 8) . The conjugation 310 between oligonucleotide and CD2 antibody followed CITE-Seq protocol 4 . 311 Corresponding antibodies and used oligonucleotides were listed in supplementary table 312 (Supplementary Table 9 ). 313 12 CD50 antibody SampleTags 39 were customized by BD Biosciences using the commercial 314 SampleTag oligos. 315 Seven types of tetramers with peptides chemically synthesized and UV-exchanged to MHC were used 317 to raise antigen-specific polyclonal T cells (Supplementary Table 1) . For each tetramer, 20 Tetramer + 318 CD8 + single T cells were sorted into each well of the 96 well plate for culture for three weeks. 319 Human whole blood from diagnosed T1D and T2D patients were obtained at Seton Family of Hospitals 321 at Austin with informed consent. The use of whole blood from these patients was approved by the 322 all relevant ethical regulations. Human peripheral blood mononuclear cell (PBMC) from healthy donors 324 were purchased from ePBMC. PBMC from T1D whole blood was isolated using Ficoll-paque density-325 gradient centrifugation (GE Healthcare). CD8 + T cells were then enriched from PBMC of T1DM and 326 healthy donors using EasySep TM Human CD8 + T cell isolation kit (STEMCELL). 327 CD8 + T cells were resuspended in FACS buffer containing 0.05% sodium azide and 50nM of Dasatinib. 328 CD8 + T cells were then incubated at 37°C for 30min-60min. About 10,000 cells from an HCV peptide 329 binding clone used previously 18 were pre-stained with BV510 anti-CD8a antibody (clone: RPA-T8, 330 Biolegend) and spiked into the primary CD8 + T cells. Following the Dasatinib treatment, tetramer pool 331 together with anti-CD8a antibody (clone: RPA-T8, Biolegend) was directly added into the cells. Cells 332 were incubated at 4°C for 1hr with continuous rotation. After washing, cells were further stained at 4°C 333 for 20min with the presence of 5 µg/ml mouse anti-PE (clone: PE001, Biolegend) and/or mouse anti-334 APC (clone: APC003, Biolegend). AbSeq staining mastermix was prepared by pooling 1µl of each 335 AbSeq together (Supplementary Table 9 ). Cells were washed in FACS buffer once and stained with 336 the AbSeq mastermix. Additional dump-channel antibodies (AF488-anti-CD4, AF488-anti-CD14 and 337 AF488-anti-CD19), 7-AAD and 2µl of anti-CD50 SampleTag were mixed in cells. Cells were incubated 338 at 4°C for 40mins, prior to washing in FACS buffer twice and proceeded for sorting. 339 During cell sorting, about 50,000 Tet -CD8 + T cells were also sorted, which will later be spiked into Tet + 340 T cells. 341 Prior to BD Rhapsody™ processing, Tetramer -CD8 + T cells were first stained with 2µl of CD2 343 SampleTag at 4°C for 30mins. Cells were washed in FACS buffer for three times and resuspended in 344 100µl BD Sample Buffer. Sorted Tet + CD8 + T cells and Tetramer -CD8 + T cells were counted using BD 345 Rhapsody™. Tetramer + and Tetramer -CD8 + T cells were pooled and processed on BD Rhapsody™ 346 cartridge following user's manual. Single cell cDNA synthesis and library amplification were performed 347 following manufacturer's protocol with some modifications. Briefly, in PCR1, 1.2µl of tetramer PCR1 348 primer was added to the PCR reaction in addition to primers for gene expression panel, AbSeq, 349 SampleTag, and universal oligo (Supplementary Table 9 ). 9 and 10 PCR cycles were used for 5000-350 10,000 and 10,001-20,000 cells respectively. Double-sided AMPure beads purification was processed 351 to purify short amplicons (AbSeq, SampleTag and tetramer DNA-barcodes) and long amplicons (target 352 genes and TCRα/β) separately. In PCR2, five separate PCR reactions with 15 reaction cycles were 353 carried out to amplify gene panel, SampleTag, TCRα, TCRβ, and tetramer DNA-barcodes. AbSeq, 354 tetramer and TCRα/β libraries were gel extracted for the desired band before proceeding to PCR3. 355 Finally, 8 cycles of PCR reactions were performed for all six elements following manufacturer's 356 instruction. All PCR libraries were quantified using Bioanalyzer 2100 and pooled. 15% PhiX was used 357 in all sequencing runs. Pooled libraries were sequenced on HiSeq X with PE150. 358 Sequencing reads from target gene expression, AbSeq, SampleTag, TCRα/β and tetramer DNA-360 barcodes were processed as below (Supplementary Fig. S11) . 361 For the SampleTag DNA-barcodes, reads were mapped to SampleTag DNA-barcode reference using 375 bowtie2 with --norc and --local mode 41 . Aligned reads were then processed using umi_tools to count 376 the number of MIDs for each SampleTag DNA-barcode in each cell. Distribution of MID counts for 377 each SampleTag was fitted by a bimodal distribution and the cutoff between two distributions were set 378 as the negative threshold for the corresponding SampleTag. In addition, to recover false negative SampleTag signals, SampleTag, whose MID counts account for >50% total SampleTag MID counts, 380 was also classified as positive event. Cells containing CD2 SampleTag were Tetcells, while cells with 381 more than two regular SampleTags were multiplets and were removed from further analysis. 382 For the TCR sequencing reads, we adapted sub-clustering algorithm as previously described 42 to 383 remove PCR and/or sequencing errors and identify VDJ and CDR3 with some changes. Reads were 384 first aligned to TCR J and C region reference. Only reads that are >62.5% identical were retained. 385 Reads with same cellular barcodes and MID were grouped together. Under each group, reads within 386 a Levenshtein distance of 15% were further clustered into a subgroup. For each subgroup, a 387 consensus sequence was built based on the average nucleotide at each position, weighted by quality 388 score. After ranking the consensus sequences by their abundance, the most abundant consensus 389 sequence is selected and other sequences with edited distance less than three were removed. In case 390 the most abundant consensus sequence is non-productive, the next most abundant productive 391 sequence, if exists, was selected as the unique consensus sequence for that cell. The 2 nd TCR chain 392 was retained when its MID count accounts for more than 20% of total TCRα or TCRβ MID counts. 393 All single cells were first filtered to exclude low quality cells whose total gene and AbSeq expression 395 MID counts were in the last 1% quantile. Then cells identified as multiplets with SampleTag and cells 396 with two productive TCRβ chains were also removed. Additionally, genes or AbSeqs whose expression 397 were detected in less than 50 cells were filtered. Gene expression and AbSeq data from different 398 Rhapsody chips were pooled together and performed joint probabilistic modeling of RNA expression 399 and surface protein measurement with totalVI 21 . Each donor was treated as an independent batch 400 factor and 200 epochs were used to train the model. Other parameters were set as default in totalVI. 401 Posterior dataset was then used for dimensionality reduction (UMAP algorithm) and clustering (Leiden 402 algorithm), both with Scanpy 43 . 403 First, for each tetramer fluorescent color, distribution of total tetramer DNA-barcode counts per cell 405 was fitted to a bimodal distribution. The cutoff counts were set as negative threshold to capture positive 406 tetramer binding events. Tetramer DNA-barcode counts were then ranked for each single cell and the 407 knee point on the count-rank plot was selected. Antigens rank higher than the inflection point are 408 putative binding antigens. Besides, antigens that rank below inflection point, but with <=3 amino acid 409 difference compared with higher ranking antigens, were also included as putative cross-reactive 410 binding antigens. For each cell, tetramer MID signal fraction was defined as the fraction of cumulative 411 MID count from putative binding antigens over cumulative MID count from all bound antigens: 412 Further, cells with the same TCRα/β were pooled together. The correlation coefficient of antigen 414 binding for each single cell in the pool were calculated between detected tetramer DNA-barcode 415 counts and corresponding median tetramer DNA-barcode counts within the pool. This correlation 416 coefficient for each single cell is used as the tetramer binding noise. The knee point of the distribution 417 of correlation coefficients were set as the threshold, below which cells were removed due to high 418 tetramer binding noise. 419 For analysis of viral antigens, we select antigens detected in more than 5 cells to ensure capturing low 420 frequent antigen specific CD8 + T cells while limiting non-specific binding. 421 For sensitivity analysis to demonstrate the robustness of TetTCR-SeqHD, we set the negative 422 threshold of tetramer MID to 15 to capture positive binding events. This threshold was then used for 423 all experiments (Supplementary Fig. S19) . 424 In the TetTCR-SeqHD clone experiment, true positive is defined as antigen-matched TCRs between 426 MIDCIRS and TetTCR-SeqHD. Predicted condition positive is defined as antigen-specific TCRs 427 identified by pMHC DNA-barcodes. The condition positive is defined as antigen-specific TCRs 428 identified by MIDCIRS. Precision and Recall is then calculated as below. 429 HLA-A02:01 bound T1D autoantigens were curated from the IEDB (www.iedb.org) database, while 434 HLA-A01:01 and HLA-B08:01 bound T1D autoantigens were predicted using NetMHCpan 4.0 44 . The 435 IC50 cutoff for HLA-A01:01 and HLA-B08:01 was 950nM and 500nM respectively. 436 TCRs that have productive paired α and β chains were used to calculate TCR clonality, which is a 438 score to characterize T cell expansion. Higher TCR clonality indicates that corresponding TCR are 439 more clonally expanded. If there is singleton TCR, we define the TCR clonality being 0, while single 440 TCR species with multiple copies have TCR clonality being 1. For all other situations, the TCR clonality 441 is defined using following formula. 442 We generated TCR constructs as previously described 18 and cloned them into an empty pCDH 445 (System Biosciences) vector driven by the MSCV promoter. Lentivirus was generated using the 446 Virapower (ThermoFisher Scientific) system and concentrated 10 times using an Amicon Ultra column. 447 Freshly thawed CD8 + T cells from an HLA-A2/B8/A1 negative donor were stimulated with Immunocult 448 (StemCell Technologies) and incubated with the concentrated virus for 2-3 days. The cells were 449 expanded for a minimum of ten days and then assessed for murine TCRβ chain expression. The presence of anti-GAD, -IA2 and -Znt8 antibodies were determined via ELISA assay obtained from 466 Kronus and performed according to the manufacturer's instructions. Whole, undiluted plasma was 467 used in this assay. Absorbance was measured using a SpectraMax M3 plate reader and analysis of the standard curve was performed in R using a cubic-spline fit. The antibody concentration for each 469 sample was then interpolated, with all positive controls falling within the reported concentrations. 470 Patients were reported as positive if the detectable antibody levels were in excess of 5 IU/mL, 7.5 and 471 15 U/mL for the anti-GAD, -IA2 and -Znt8 antibodies, respectively according to the manufacturer's 472 instructions. 473 474 We thank T1D patients for donating blood samples to our study. We also thank anonymous blood 476 donors and staff members at We Are Blood for sample collection. We thank P. Parker for assistance 477 with blood sample purification. This work was supported by NIH grants S10OD020072 (N.J.) and antigen identification among HCV-specific T cell-cluster (cluster13) (f) and HCV-specific TCR bearing cells (g). Cells were classified into "filter" category based on following criteria: 1) more than one antigens bind to single cell, and these antigens are more than 3 amino acid distance away from each other; 2) correlation of tetramer MID between single cell and median of all cells with same TCR sequence is below 0.9, identified as described in Methods. PTPRN-FGD-9: FGDHPGHSY. *, p ≤ 0.05; **, p ≤ 0.01; ***, p ≤ 0.001; ****, p ≤ 0.0001. Recent progress in the analysis of alphabetaT cell and B cell receptor 507 repertoires The science and medicine of human immunology Transcript-indexed ATAC-seq for precision immune profiling Simultaneous epitope and transcriptome measurement in single cells Multiplexed quantification of proteins and transcripts in single cells A Targeted Multi-omic Analysis Approach Measures Protein Expression and 517 Single-cell multiomic analysis identifies regulatory programs in mixed-519 phenotype acute leukemia Single-cell immune landscape of human atherosclerotic plaques Multi-Omics Resolves a Sharp Disease-State Shift between Mild and Moderate 523 COVID-19 T cell antigen discovery via trogocytosis T cell antigen discovery via signaling and antigen-presenting bifunctional 526 receptors Deciphering CD4(+) T cell specificity using novel MHC-528 TCR chimeric receptors T-Scan: A Genome-wide Method for the Systematic Discovery of T Cell Epitopes Rapid selection and identification of functional CD8(+) T 532 cell epitopes from large peptide-coding libraries Unbiased Screens Show CD8(+) T Cells of COVID-19 Patients Recognize 534 Shared Epitopes in SARS-CoV-2 that Largely Reside outside the Spike Protein Combinatorial tetramer staining and mass cytometry analysis facilitate T-537 cell epitope mapping and characterization Bystander CD8(+) T cells are abundant and phenotypically distinct in human 539 tumour infiltrates High-throughput determination of the antigen specificities of T cell receptors 541 in single cells Ultrahigh-throughput 543 single cell protein profiling with droplet microfluidic barcoding Immune Repertoire Sequencing Using Molecular Identifiers Enables Accurate 545 A Joint Model of RNA Expression and Surface Protein Abundance in Single 547 Cells Uniform manifold approximation and projection for 549 dimension reduction Data-Driven Phenotypic Dissection of AML Reveals Progenitor-like Cells that 551 Correlate with Prognosis Adaptive Memory of Human NK-like CD8(+) T-Cells to 553 Potential killers exposed: tracking endogenous influenza-specific CD8(+) T 555 cells T cell immunoglobulin and mucin protein-3 (Tim-3)/Galectin-9 interaction 557 regulates influenza A virus-specific humoral and CD8 T-cell responses Differential expression of CD26 on virus-specific CD8(+) T cells during active, 560 latent and resolved infection Assessment of immunogenicity of human Melan-A peptide analogues in HLA-562 A*0201/Kb transgenic mice A novel population of human melanoma-specific CD8 T cells recognizes Melan-564 AMART-1 immunodominant nonapeptide but not the corresponding decapeptide Degeneracy of antigen recognition as the molecular basis for the high frequency 567 of naive A2/Melan-a peptide multimer(+) CD8(+) T cells in humans Immunization of HLA class I transgenic mice identifies autoantigenic epitopes 570 eliciting dominant responses in type 1 diabetes patients CD8 T cell autoreactivity to preproinsulin epitopes with very low human 572 leucocyte antigen class I binding affinity Autoimmunity against a defective ribosomal insulin gene product in type 1 574 diabetes Hotspot autoimmune T cell receptor binding underlies pathogen and insulin 576 peptide cross-reactivity Simultaneous detection of circulating autoreactive CD8+ T-cells specific for 578 different islet cell-associated epitopes using combinatorial MHC multimers Autoreactive CD8+ T cell exhaustion distinguishes subjects with slow 581 type 1 diabetes progression Islet-reactive CD8(+) T cell frequencies in the pancreas, but not in blood, 583 distinguish type 1 diabetic patients from healthy donors Generation of peptide-MHC class I complexes through UV-mediated ligand 585 exchange Single Molecule and Single Cell UMI-tools: modeling sequencing errors in Unique Molecular 589 Identifiers to improve quantification accuracy Fast gapped-read alignment with Bowtie 2 Accurate immune repertoire sequencing reveals malaria infection driven 593 antibody lineage diversification in young children SCANPY: large-scale single-cell gene expression data 595 analysis NetMHCpan-4.0: Improved Peptide-MHC Class I Interaction Predictions 597 Integrating Eluted Ligand and Peptide Binding Affinity Data