APPLICATION NOTE10x GenomicsA new way of exploring immunity: linking highlymultiplexed antigen recognition to immune repertoireand phenotypeIntroductionT cells, key components of the adaptive immune system,are involved in the initiation of the immune responsemediated by specific antigen recognition. T-cell receptors(TCRs), located on the surface of T cells, recognize and bindspecific antigenic peptides presented by a major histocompatibility complex (MHC) on the surface of other cells (1).TCRs are heterodimeric proteins, commonly comprisingan alpha and a beta chain. The heterodimers recognizepeptide–MHC (pMHC) through their complementarity-determining region (CDR) loops: CDR1, CDR2, andCDR3, such that each paired TCR can bind to a particularset of pMHCs. Since each T cell typically expresses a single paired TCR, sufficient TCR sequence diversity isessential across an individual’s entire repertoire to bindthe broad spectrum of antigens that may be encountered.Somatic V(D)J recombination of the TCR loci during T-celldevelopment gives rise to an enormous potential space ofTCR sequences, resulting in TCR diversity (2).The amino acid sequence of the paired TCR directlydetermines its binding specificity. However, we do not yethave a complete understanding of the factors underlyingthe recognition of pMHC complexes by their cognateTCRs. A generalized and predictive understanding of thisinteraction would enable entirely new approaches in basicimmunological research and in clinical practice; thiswould encompass areas such as TCR-based diagnosticmethods and the rational design of immunotherapies.Recent work has shown that it is possible to identifyshared motifs within TCRs that are specific to particularpMHCs (3, 4). These studies identified TCR sequences thatbind a limited number of pMHCs and use the sharedmotifs to classify previously unseen TCRs according totheir predicted ability to bind one of the pMHCs in thetraining data. These initial studies are extremely promising but are limited by the amount and diversity of10xgenomics.comHighlightsWe demonstrate the power of the Single Cell ImmuneProfiling Solution with Feature Barcode technology to: Assess the binding specificities of over 150,000 CD8 T cells from 4 human donors across a highly multiplexedpanel of 44 distinct, specific peptide–MHC (pMHC)multimers Uncover data at single cell resolution for gene and cellsurface protein expression, TCR–pMHC binding specificity, and paired aß T-cell receptor (TCR) sequences Generate highly multiplexed binding data from manydifferent pMHC multimers in a single experiment Enable the potential to develop novel experimental andanalytical approaches Help facilitate a greater understanding of how the adaptive immune system responds to immune insults andcontribute to the development of therapeuticstraining data. A generalizable understanding of TCR recognition will require more extensive data with orders ofmagnitude more pMHC diversity.The 10x Genomics Chromium Single Cell Immune ProfilingSolution with Feature Barcode technology is particularlywell suited to generating the data required to address thisquestion. Single cell–based methods are the only way todirectly observe antigen binding to T cells, while simultaneously sequencing the paired and recombined alpha andbeta TCR genes. This is achieved by generating Single Cell5' libraries and V(D)J enriched libraries, which provideTCR sequences, in combination with Feature Barcodetechnology, which uses highly multiplexed pMHC multimer reagents, each carrying a pMHC and a distinctmolecular barcode, to identify binding specifities.

A new way of exploring immunity: linking highly multiplexed antigen recognition to immune repertoire and phenotypeHere, we demonstrate the use of a highly multiplexedpanel of 44 distinct, specific pMHC multimers (dCODEDextramer reagents) to assess the binding specificitiesof over 150,000 CD8 T cells from four human donors.This large dataset contains data at single cell resolutionon gene expression, expression of eleven surface proteins(using TotalSeq -C antibodies with Feature Barcodetechnology), pMHC binding, and paired aß TCRsequences (Figure 1).From an initial analysis of the dataset, we haveidentified 64,755 cells and 15,214 distinct paired TCRsequences with apparent specificity for at least onepMHC within the panel. This single dataset has thepotential to account for a significantly larger amountof knowledge regarding paired human TCR specificitiesthan has been generated to date.Within our data, we observed TCRs with cognate antigensthat had been reported previously, while also identifyingentirely new TCR–pMHC interactions. In addition, weobserved specific expanded non-naïve T-cell clones alongwith more diverse binding in the naïve compartment.We obtained CD8 T cells from four healthy donors fromAllCells and STEMCELL Technologies (Reference Table1). Donors were chosen to ensure each HLA allele in thepMHC panel was present in at least one donor. Cytomegalovirus (CMV) and Epstein Barr Virus (EBV) serostatuswas known for Donors 3 and 4.Peptide–MHC multimers with Feature Barcode technologyWe used a panel of 44 dCODE Dextramer reagents(Immudex, Reference Table 2) with antigenic peptidesderived from different viruses (CMV, EBV, influenza,HTLV, HPV and HIV) and known cancer antigens.cDNA Cleanup & QC7-AAD Dead CellMarkerFluorescentAntibodiesdCODEDextramer tramer T CellsDextramerCD4 T-CellB-Cell (CD19)7AADCD8 T CellsPan T CellsLive CellsSSC-ASSC-ALeukocytesRTSPRI Supernatant10x UMIBarcodeTSOFeature Read 2N Sample P7BarcodingIndexSequencePoolRemove OilSequencing LibrariesCell Rangerfor Feature BarcodeEnzyme,Labeled CellsOilSingle CellGEMscDNA Amplification10x BarcodedcDNATCRP5Read 110x UMIBarcodeTSOInsertRead 2Sample P7IndexLibrary ConstructionP5Read 1 10x UMIBarcodeTSOVD JCCD8 T CellCell Ranger for VDJ10x BarcodedGel Beads5’ Gene ExpressionLibrary ConstructionLibrary ConstructionRead 1TCR Target EnrichmentSPRI PelletSingle Cell 5’ Cell Surface ProteinP5CD8 T CellGEM Generation & BarcodingCollectLibrary Construction (cont’d)TotalSeq-CAntibodiesSequencingCD8 T cellsCellsCell SortingCell samplesSample PrepFACSLibrary ConstructionMethods10x BarcodedcDNASequencing LibrariesSequencing LibrariesCell RangerCell Rangerfor Gene Expressionfor V(D)JAnalysisCell LabelingThis rich and large dataset illustrates the power and scalability of the 10x Genomics Chromium Single CellImmune Profiling Solution with Feature Barcode technology and presents an exciting opportunity for researchersto explore and draw further conclusions about the mechanisms of TCR–pMHC interaction. Furthermore, thisexperiment serves as the next step on the path toward theeven larger-scale experiments that will be necessary tofully comprehend the rules of antigen recognition in theadaptive immune system. In addition, the experimentallowed us to enhance our understanding of experimental design and computational analysis, both essential forsingle cell–based immunology research.Loupe Browser and Custom AnalysisFigure 1. Experimental approach for generating high-throughput, highly multiplexed TCR–antigen binding data. CD8 T cells from healthy human donorswere labeled with fluorescent antibodies, TotalSeq -C antibodies, and dCODE Dextramer reagents. Dextramer positive CD8 T cells were sorted byflow cytometry and used as input for the 10x Genomics Chromium Single Cell Immune Profiling Solution with Feature Barcode technology. Libraries wereprepared to characterize gene expression, cell surface protein expression, paired TCR sequences, and TCR–pMHC binding in each single

10x GenomicsEach Dextramer reagent included a distinct nucleic acidbarcode, along with a phycoerythrin (PE) fluorophore.The panel also contained 6 dCODE Dextramer reagentswith irrelevant negative control peptides to assist in thedetection of non-specific binding events.Just prior to labeling, 2 µl of each Dextramer reagent wascombined for a total volume of 100 µl.Surface marker antibodies with Feature BarcodetechnologyWe used a panel of eleven TotalSeq -C antibodies (BioLegend, Reference Table 3) that were chosen to enablediscrimination between CD8 T cell (CD8A /CD3 ) subpopulations, such as naïve (CD45RA /CD45RO /CCR7 ),effector (TEF, CD45RA /CD45RO /CCR7 ), effector memory(TEM, CD45RA /CD45RO /CCR7 ), exhausted (PD-1 ),central memory (TCM, CD45RA /CD45RO /CCR7 ), andactivated (HLA-DR /CD127 ), and exclude any CD4 T cells(CD4 ), B cells (CD19 ), and monocytes (CD14 , CD15 ,CD16 ). Prior to staining, the TotalSeq -C antibodies werepooled by combining 0.5 µg of each antibody per reaction.Fluorescently labeled surface marker antibodiesWe used a panel of fluorescently labeled antibodies (BioLegend, Reference Table 4) to enable sorting of pure CD8 T cells. Fluorescent antibodies were used in staining reactions at a dilution of 1:100 and were different clones thanthe TotalSeq -C antibodies to avoid binding competition.Cell labelingFor each donor, 20 million CD8 T cells were thawed in a37ºC water bath. 1 ml warm medium (RPMI 10% FBS)was added to the cells in a 15 ml conical tube containing9 ml warm medium. The cells were washed once in 5 mlmedium and finally resuspended in 4–6 ml PBS 2% FBSand then counted. For Dextramer and antibody labeling,20 million cells were centrifuged and resuspended in 105µl PBS 2% FBS. 12.5 µl Human TruStain FcX (Fc receptor blocking solution, BioLegend, Cat# 422301), 5 µldextran sulfate (10 mg/ml), and 100 µl pooled Dextramer panel were added to the cells and incubated for 10 minuteson ice. 7 µl TotalSeq -C antibody panel and 15 µl fluorescent antibody panel were then added and the cells wereincubated for 30 minutes on ice. After incubation, cellswere washed by adding 1.2 mL PBS 2% FBS, followedby centrifugation for 5 minutes at 350 Xg (swing bucket).Cells were washed 2 more times in 1.5 mL PBS 2% FBS.After the final centrifugation, the cells were resuspendedin 3–4 ml PBS 2% FBS. 100 µl cell suspension wasreserved for a non-sorted cell population. Before cellsorting, 7-AAD viability staining solution (BioLegend,Cat# 420403) was added at 1:200 dilution. See our Demonstrated Protocols CG000149 and CG000203 for full details.Cell sortingCells were sorted using a MA900 Multi-Application CellSorter (Sony Biotechnology) with 100 µm flow cell chip.Cells were sorted into 55.9 µl Reaction Mix containing RTReagent Mix and Poly dT RT primers in 8 wells of a chilled96-well plate. We targeted a cell yield of 9000 cells per well.Cells were gated using the Sony MA900 system software toobtain single (FSH/BSC, FSW/BSC), live (7-AAD-), CD8 cellswhilst excluding CD14 /CD15 /CD16 /CD19 /CD4 cells.We then selected all Dextramer -positive cells within theCD8 T-cell population. After sorting, 5 µl water and a total12.4 µl of Additive A (2.4 µl) and RT Enzyme Mix B (10 µl)were added to each well to complete the Reaction Mix,which was then directly loaded onto a Chromium chip.Generation of the single cell immune profiling librariesThe Chromium Single Cell Immune Profiling workflowwith Feature Barcode technology generates Single Cell 5’Gene Expression, V(D)J, and Cell Surface Protein libraries.The Cell Surface Protein libraries are generated fromamplified DNA derived from antibody or Dextramer -conjugated Feature Barcode oligonucleotide, which are boundto the cell surface proteins and TCRs, respectively. A total180 libraries were generated (60 libraries each for GeneExpression, Cell Surface Protein, and V(D)J). The librarieswere prepared following the User Guide for ChromiumSingle Cell V(D)J Reagent kits with Feature Barcode technology for Cell Surface Protein (CG000186).Sequencing librariesChromium Single Cell V(D)J enriched libraries, 5’ GeneExpression libraries, and Cell Surface Protein librarieswere quantified, normalized, and sequenced accordingto the User Guide for Chromium Single Cell V(D)J Reagentkits with Feature Barcode technology for Cell Surface Protein (CG000186). The 5’ Gene Expression libraries and CellSurface Protein libraries from Donors 1 and 3 and fromDonors 2 and 4 were combined into 2 library pools andsequenced on an Illumina NovaSeq sequencer with aNovaSeq 6000 S4 Reagent Kit (200 cycles) (Illumina).The V(D)J enriched libraries from all 4 donors were pooledand sequenced with an Illumina NovaSeq 6000 S2Reagent Kit (300 cycles) (Illumina). The Gene Expression,Cell Surface Protein, and V(D)J libraries were targeted,respectively, for sequencing depths of at least 20,000,20,000 and 10,000 read pairs per cell (rppc).10x Genomics3

A new way of exploring immunity: linking highly multiplexed antigen recognition to immune repertoire and phenotypeAnalysisThe sequencing data combined with the description ofthe sequences identifying TotalSeq -C antibodies anddCODE Dextramer reagents were analyzed with theCell Ranger analysis pipeline (see What is Cell Ranger?).Results and Preliminary ObservationsGenerating high-throughput, highly multiplexedTCR binding dataWe acquired CD8 T cells from four healthy humandonors (Reference Table 1) and used FACS to isolate thoseT cells that appeared to bind at least one pMHC multimerfrom a panel of 50 dCODE Dextramer reagents. TheseT cells were used to generate gene expression, TCRsequence, cell surface protein expression, and pMHCbinding data (44 specific pMHCs; Reference Table 2) in189,512 single cells where at least one productive TCRchain was detected.We used the annotated TCR sequences produced by CellRanger to identify putative mucosal associated invariant T (MAIT) cells, based on whether they expressed aTCR a chain, comprising the typical V and J segments(5). We identified 3,914 MAIT cells and excluded themfrom further analysis. No iNKT cells were identified bya similar approach.Within the remaining cells, we were able to identify aproductive TCR a and ß chain in 136,447 (72%) and weassigned these sequences to 55,221 distinct clonotypes,where each clonotype had a unique set of recombinedTCR nucleotide sequences.We used graph-based clustering to cluster the cells intosubpopulations based on their gene expression profiles.We classified the clusters as either naïve (CD45RA CD45ROCCR7 ) or non-naïve (the remaining cells) by analyzing geneand cell surface protein expression in Loupe Cell Browser.As expected, the non-naïve T-cell population was highlyenriched for expanded clonotypes (Figure 2).IdentityNaïveNon-NaïveClonotype Size103102101100Donor 1Donor 2Donor 3Donor 4DonorFigure 2. Expanded clonotypes are found in the non-naïve compartment. CD8 T cells were classified as either naïve or non-naïve, based on cell surfacemarker and gene expression. Cells were assigned to clonotypes based on TCR sequences. Box plots indicate the distribution of the number of cells in eachclonotype within the two

10x GenomicsPreliminary analysis of T cells that bind to specificpMHC multimersDextramer reagents analyzed for specificity, the oneswithout a specific binding T cell were the following:We classified the T cells according to whether theyexhibited specific binding to at least one of thedCODE Dextramer reagents used in the experiment.We focused on cells that had strong evidence of specificbinding by setting a threshold such that a specificbinding event required a UMI count greater than 10 thatwas also greater than five times the highest negativecontrol UMI count for that cell. In cases where a cell wasassigned more than one specificity, we considered it tobe specific only for the pMHC with the highest UMIcount. We also excluded any cells with apparentspecificities for more than four Dextramers . This gave62,858 cells with pMHC specificity distributed across15,214 unique clonotypes. Amongst the 44 dCODE GGGAMpMHC-specific cells were not distributed evenly betweenthe donors (Supplemental Figure 1). This was expectedgiven that only Donors 1 and 2 possessed the HLAA*02:01 allele that was the most common MHC in theantigen panel (Reference Tables 1 and 2). Additionally,combining the cell type classifications with binding dataallowed us to determine the specific binding eventsbetween the naïve and memory T-cell compartments foreach donor (Figure 3 and Supplemental Figure 1).IdentityNaïveNumber of specific cells 1104Non-Naïve103102101B3501 pp65 CMV IPSINVHHYB0801 IE-1 CMV ELRRKMMYMB0801 BZLF1 EBV RAKFKQLLB0801 EBNA-3A EBV FLRGRAYGLB0702 pp65 CMV TPRVTGGGAMB0702 pp65 CMV RPHERNGFTVLB0702 EBNA-6 EBV QPRAPIRPIB0702 EBNA-3A EBV RPPIFIRRLA2402 WT1-(235-243)236M Y CYTWNQMNLA2402 IE-1 CMV AYAQKIFKIA2402 pp65 CMV QYDPVAALFA1101 EBNA-3B EBV IVTDFSVIKA1101 EBNA-3B EBV AVFDRKSDAKA0301 IE-1 CMV KLGGALQAKA0301 BCL-2L1 Cancer RIAAWMATYA0301 EMNA-3A EBV RLRAEAQVKA0201 WT-1 RMFPNAPYLA0201 BCL-X Cancer YLNDHLEPWIA0201 PSA146-154 KLQCVDLHVA0201 Ca2-indepen-Plip-A2 FLASKIGRLVA0201 MAGE-A1 Cancer KVLEYVIKVA0201 Kanamycin-B-dioxygenase CLLGTYTQDVA0201 MAGE-A3 Cancer KVAELVHFLA0201 NY-ESO-1 Cancer SLLMWITQVA0201 Tyrosinase Cancer CLLWSFQTSAA0201 gp100 Cancer IMDQVPFSVA0201 MART-1 Cancer ELAGIGILTVA0201 gp100 Cancer KTWGQYWQVA0201 HPV-16E7 82-91 LLMGTLGIVCA0201 HTLV-1 LLFGYPVYVA0201 16E7 HPV MLDLQPETTA0201 Gag-protein HIV SLFNTVATLYA0201 Gag-protein HIV SLFNTVATLA0201 Gag-protein HIV SLYNTVATLYA0201 RT HIV ILKEPVHGVA0201 Gag-protein HIV RTLNAWVKVA0201 pp65 CMV NLVPMVATVA0201 Flu-MP Influenza GILGFVFTLA0201 BMLF1 EBV GLCTLVAMLA0201 LMP2A EBV FLYALALLLA0201 EBNA-3B EBV LLDFVRFMGVA0201 LMP-2A EBV CLGGLLTMVA0101 IE-1 CMV VTEHDTLLYA0201 LMP1 EBV YLLEMLWRL100Dextramer reagentFigure 3. pMHC binding landscape for Donor 1. Bars indicate the number of cells classified as having specific binding to each pMHC multimer in the naïveand non-naïve compartments.10x Genomics5

A new way of exploring immunity: linking highly multiplexed antigen recognition to immune repertoire and phenotypeWe observed multiple pMHC-specific naïve T cells withdiverse TCR sequences. These naïve binders were oftenspecific to endogenous or tumor-associated antigens (e.g.MART-1) or to antigens derived from viruses for which thedonor was seronegative (e.g. HIV).We looked for evidence of pMHC-specific expandedclonotypes within the CD8 memory compartment.We observed binding of multiple clonally related cells tothe same pMHC, implying a specific clonal expansiondriven by response to that particular antigen (Figure 4).These apparently specific expansions were typicallyassociated with pMHCs that were HLA-matched to aparticular donor; in addition, expansions were consistentwith donors being seropositive for particular common viralinfections (e.g. EBV and Influenza in Donor 1 and Donor 2).To further investigate binding within expanded clonotypes, we determined the binding concordance withineach clonotype. For each expanded clonotype with a binding specificity, we calculated the proportion of the entireclonotype within the donor that had the particular binding specificity (the 'binding concordance'; Figure 4 andSupplementary Figure 2). A concordance of 1 indicatesthat all the cells within the clonotype have the same binding specificity. Clonotypes with high binding concordancewere observed in instances of expected specific expansions, while lower binding concordances were more typicalfor pMHCs that were not HLA-matched.610xgenomics.comIntriguingly, we observed clonotypes with apparentlycross-reactive binding to A0301 EMNA-3A EBVRLRAEAQVK, A0301 IE-1 CMV KLGGALQAK,A1101 EBNA-3B EBV IVTDFSVIK and A1101 EBNA-3BEBV AVFDRKSDAK. Binding to these multimers wasobserved in all four donors irrespective of their HLA haplotype and serostatus. While some of the expandedclonotypes exhibited high levels of binding concordance,others did not. This suggests an avenue for further studyto investigate the causes of these binding events andwhether they represent specific TCR-mediated binding oranother, less specific binding mode.Comparison of pMHC-binding TCR sequences withpreviously reported observationsWe compared the pMHC-specific TCRs that we observedin these data with those that have been previouslyreported to have the same binding specificities in VDJdb(6). VDJdb contains paired, human TCR sequences for 11of the pMHCs for which we observed binding events. Wecalculated the TCRdist metric (3) between each TCR pairwe observed and the closest entry in VDJdb. We foundexamples of exactly matching sequences (Table 1) alongside sequences with varying degrees of similarity to thosealready reported (Figure 5).

.4TRA:CAVGGGGGSQGNLIF;TRB:CASSIRASYEQYFTotal clonotype EQYFTRA:CAVSAASGGSYIPTF;TRB:CASSPRDRERGEQYF10x GenomicsBinding concordance355000.81.0A0201 EBNA-3B EBV LLDFVRFMGVA0201 LMP2A EBV FLYALALLLA0201 Gag-protein HIV SLFNTVATLYA0201 Flu-MP Influenza GILGFVFTLA0201 gp100 Cancer KTWGQYWQVA0301 EMNA-3A EBV RLRAEAQVKA1101 EBNA-3B EBV IVTDFSVIKA0301 IE-1 CMV KLGGALQAKA1101 EBNA-3B EBV AVFDRKSDAKA2402 pp65 CMV QYDPVAALFB0801 EBNA-3A EBV FLRGRAYGLB0801 BZLF1 EBV RAKFKQLLFigure 4. Binding specificities of expanded clonotypes from Donor 1. The 50 largest clonotypes are plotted along with their binding specificities andconcordance. A circle indicates that at least one member of the clonotype was classified as specific for a particular pMHC. Circle size indicates the totalwithin-donor clonotype size. Circle color indicates the proportion of cells within the clonotype that bind the Dextramer (the ‘binding concordance’).10x Genomics7

A new way of exploring immunity: linking highly multiplexed antigen recognition to immune repertoire and 201 ELAGIGILTV MART-1 Cancer181440donor1A0201 SLLMWITQV NY-ESO-1 Cancer310donor1A0201 GILGFVFTL Flu-MP Influenza34957744donor1A0201 GLCTLVAML BMLF1 EBV151485donor1A0201 LLFGYPVYV HTLV-1920donor1A0201 RMFPNAPYL WT-1350donor1A1101 IVTDFSVIK EBNA-3B EBV15180donor1A1101 AVFDRKSDAK EBNA-3B EBV28170donor1B3501 IPSINVHHY pp65 CMV320donor1B0801 FLRGRAYGL EBNA-3A EBV350donor2A0201 ELAGIGILTV MART-1 Cancer27440donor2A0201 GILGFVFTL Flu-MP Influenza68157754donor2A0201 GLCTLVAML BMLF1 EBV691482donor2A0201 LLFGYPVYV HTLV-1620donor2A1101 IVTDFSVIK EBNA-3B EBV6580donor2A1101 AVFDRKSDAK EBNA-3B EBV24370donor2B3501 IPSINVHHY pp65 CMV120donor2B0702 RPHERNGFTVL pp65 CMV110donor2B0801 FLRGRAYGL EBNA-3A EBV1750donor3A0201 ELAGIGILTV MART-1 Cancer33440donor3A0201 SLLMWITQV NY-ESO-1 Cancer110donor3A0201 GILGFVFTL Flu-MP Influenza105770donor3A0201 LLFGYPVYV HTLV-1820donor3A0201 RMFPNAPYL WT-1150donor3A1101 IVTDFSVIK EBNA-3B EBV9880donor3A1101 AVFDRKSDAK EBNA-3B EBV71970donor3B3501 IPSINVHHY pp65 CMV220donor3B0801 ELRRKMMYM IE-1 CMV120donor4A0201 ELAGIGILTV MART-1 Cancer19440donor4A0201 GILGFVFTL Flu-MP Influenza25770donor4A0201 GLCTLVAML BMLF1 EBV11480donor4A0201 LLFGYPVYV HTLV-1220donor4A1101 IVTDFSVIK EBNA-3B EBV9180donor4A1101 AVFDRKSDAK EBNA-3B EBV7670donor4B0801 ELRRKMMYM IE-1 CMV120Table 1. Comparison between pMHC-specific paired TCR sequences identified in this experiment and those reported in VDJdb. Overlapping sequencesare those with a TCRdist separation of

10x GenomicsA0201 Flu-MP Influenza 0200250300TCRdist to nearest VDJdb entryA1101 EBNA-3B EBV 50200250300350400TCRdist to nearest VDJdb entryFigure 5. Similarities between paired TCR sequences binding to two pMHC-specific Dextramers with a peptide previously reported in VDJdb.Histograms show the distribution of TCRdist values between pMHC-specific paired TCRs and the closest TCR paired chain entry in VDJdb with the samereported specificity.DiscussionThis experiment, the first of its kind at this scale, demonstrates that Feature Barcode technology enables thegeneration of highly multiplexed binding data from manydifferent pMHC multimers in a single experiment. Thedataset also illustrates the complexity inherent in performing and analyzing these experiments. Even thoughthe experimental design and the analyses presented hereare not fully optimized for understanding TCR binding atsuch a large scale, the data will be useful in extractingdetailed biological insights as well as developing futureexperimental and analytical approaches.Key areas for experimental improvement include optimizing the design of pMHC multimer panels and thesource of donor cells with respect to HLA types. For theanalysis, the most crucial step is the classification ofcells according to their binding specificities; there is avaluable opportunity to develop more sophisticatedapproaches that maximize sensitivity and specificity bymoving beyond simple count-based thresholds.Associating TCR sequences with pMHC specificity andcellular phenotypes at this scale and resolution willenable a new way of exploring adaptive immunity andunderstanding some of the most fundamental questionsin immunology.10x Genomics9

A new way of exploring immunity: linking highly multiplexed antigen recognition to immune repertoire and phenotypeReference tablesDonorProviderAllelesHLA-A 1HLA-A 2HLA-B 1HLA-B 2CMVstatusEBVstatusOther statusDonor 1StemCellTechnologies02:0111:0135:01NDNDNDHIV-1, HIV-2, HBV,HCV negativeDonor 2StemCellTechnologies02:0101:0108:01NDNDNDHIV-1, HIV-2, HBV,HCV negativeDonor 3AllCells24:0229:0235:0244:03 -HIV, HBV, HCV negativeDonor 4AllCells03:0103:0107:0257:01--HIV, HBV, HCV negativeReference Table 1. Information on the T cell donors used in this 201MAGE YVIKVWB5066A*0201Kanamycin B dioxygenaseCLLGTYTQDVWB2143A*0201EBNA 3B/EBVLLDFVRFMGVWB3307A*0201HPV 16E7, LALLLWB2161A*0201Flu KEPVHGVWB5335A*0201Ca2 -indepen. Plip A2FLASKIGRLVWF2639A*2402WT1 (235-243)236M- YCYTWNQMNLWB2646A*0201Gag protein/HIVRTLNAWVKVWB2157A*0201PSA 146-154KLQCVDLHVReference Table 2. List of the dCODE Dextramer reagents used in the

10x 1LLFGYPVYVWB3338A*0201Gag protein/HIVSLFNTVATLWB3339A*0201Gag protein/HIVSLYNTVATLYWB3340A*0201Gag 1EMNA YWD2175A*1101EBNA 3B/EBVIVTDFSVIKWD2149A*1101EBNA ALFWH2165B*0702EBNA B*0702EBNA MMYMWI2147B*0801EBNA GRGAALReference Table 2. List of the dCODE Dextramer reagents used in the study. (continued)10x Genomics11

A new way of exploring immunity: linking highly multiplexed antigen recognition to immune repertoire and phenotypeCat#CloneDescription300479UCHT1TotalSeq-C 0034 anti-human CD3 Antibody302265HIB19TotalSeq-C 0050 anti-human CD19 Antibody304163HI100TotalSeq-C 0062 anti-human CD45RA Antibody300567RPA-T4TotalSeq-C 0072 anti-human CD4 Antibody301071RPA-T8TotalSeq-C 0080 anti-human CD8A Antibody301859M5E2TotalSeq-C 0081 anti-human CD14 Antibody304259UCHL1TotalSeq-C 0087 anti-human CD45RO Antibody329963EH12.2H7TotalSeq-C 0088 anti-human CD279 (PD-1) Antibody351356A019D5TotalSeq-C 0390 anti-human CD127 (IL7Ra) Antibody353251G043H7TotalSeq-C 0148 anti-human CD197 (CCR7) Antibody3076

Cells were sorted using a MA900 Multi-Application Cell Sorter (Sony Biotechnology) with 100 µm flow cell chip. Cells were sorted into 55.9 µl Reaction Mix containing RT Reagent Mix and Poly dT RT primers in 8 wells of a chilled 96-well plate. We targeted a cell yield of 9000 cells per well. Cells were gated using the Sony MA900 system software to