T-Scan: A Genome-wide Method for the Systematic Discovery of T Cell Epitopes

Abstract

T cell recognition of specific antigens mediates protection from pathogens and controls neoplasias, but can also cause autoimmunity. Our knowledge of T cell antigens and their implications for human health is limited by the technical limitations of T cell profiling technologies. Here, we present T-Scan, a high-throughput platform for identification of antigens productively recognized by T cells. T-Scan uses lentiviral delivery of antigen libraries into cells for endogenous processing and presentation on major histocompatibility complex (MHC) molecules. Target cells functionally recognized by T cells are isolated using a reporter for granzyme B activity, and the antigens mediating recognition are identified by next-generation sequencing. We show T-Scan correctly identifies cognate antigens of T cell receptors (TCRs) from viral and human genome-wide libraries. We apply T-Scan to discover new viral antigens, perform high-resolution mapping of TCR specificity, and characterize the reactivity of a tumor-derived TCR. T-Scan is a powerful approach for studying T cell responses.

Keywords: T cell epitope; T cell screening; TCR epitope; TCR-specificity; antigen discovery; epitope discovery; epitope mutagenesis; epitope screening; immunological screening; off-target screening; peptide MHC recognition.

Figure 1.. Design of the T-Scan Platform

(A) Schematic of T-Scan workflow. T cells of interest are co-cultured with target cells expressing a library of candidate antigens. Target cells recognized by the T cells are isolated by FACS. The antigens these cells had displayed are identified by PCR and next generation sequencing of the antigen cassette from the isolated cells. (B) Schematic of the fluorogenic GzB reporter. GzB cleaves a constraining linker sequence between two domains of IFP to enable protein maturation and fluorescence. (C) Representative flow cytometry plots of IFP fluorescent signals in target cells co-cultured with cytotoxic T cells in the absence (top) or presence (bottom) of cognate antigen. (D) Efficiency of GzB reporter activation in target cells co-cultured with pp65-specific T cells. Target cells were pulsed with control peptide (HIV IV9), the cognate pp65 peptide, or express one of two 56-aa fragments or the full-length pp65 ORF that contain the cognate pp65 peptide. Error bars indicate the SD across three replicates. See also Figure S1.

Figure 2.. CMV Genome-wide T Cell Antigen Discovery

(A) Overview of CMV screens. Peptides tiling across the CMV genome in 56-aa steps with 28-aa overlap are expressed via lentiviral transduction of target cells expressing HLA-A2, IFP^GZB, and ICAD^CR. For NLV2 TCR T cells, donor CD8 T cells are transduced with lentivirus expressing the NLV2 TCR. For NLV-expanded T cells, CD8 T cells from a CMV-positive donor are expanded in the presence of the NLV peptide. (B) T-Scan screen of NLV2 TCR cells against CMV genome-wide library. Each dot represents one peptide with the y axis plotting the geometric mean of the enrichment of the peptide across six total replicates (three screen replicates with two internal barcode replicates each). Fold enrichment is defined as the ratio of the abundance of the peptide in the sorted population relative to the input library. Peptides highlighted in red contain the known cognate antigen of the NLV2 TCR. Data presented in Table S1. (C) Reproducibility of internal replicates of NLV2 TCR T cell screen. Each dot represents one peptide, with the X and Y axes plotting the geometric mean of the fold-change of each barcode across three replicates. (D–F) Performance of T-Scan assays in various conditions. Bars show the average fold-enrichment of the four NLV-containing peptides in screens performed with NLV3 TCR T cells with: (D) a 2:1 (4× T cells), 1:2 (standard), and 1:8 (0.25× T cells) ratio of NLV3 TCR T cells. (E) antigens introduced into target cells at an MOI of 1 or 5, and (F) a 1:2 ratio of NLV3 TCR T cells (standard), a 1:8 ratio of NLV3 TCR T cells (0.25× T cells), and a 1:8 ratio of NLV3 TCR T cells mixed with a 3-fold excess of non-specific T cells (0.25× T cells in 1:4 mix). Data for (D)–(F) are shown in Table S2. Error bars for (D)–(F) indicate SD across four target peptides. (G) T-Scan screen of NLV-expanded cells against CMV genome-wide library, plotted as in (B). Peptides highlighted in red contain the NLV peptide and the blue peptides highlight additional enriching peptides. Data presented in Table S3. (H) Reproducibility of internal replicates of NLV-expanded T cell screen, plotted as in (C). See also Figures S2 and S3.

Figure 3.. Virome-wide T Cell Antigen Discovery

(A) Schematic of virome-wide T-Scan screen. Peptides tiling across the proteomes of 206 viral species in 56-aa steps with 28-aa overlap are expressed via lentiviral transduction of target cells expressing HLA-A2, IFP^GZB, and ICAD^CR. (B) T-Scan screen of NLV-expanded T cells against CMV genome-wide library. Each dot represents one peptide, with the y axis plotting the geometric mean of the fold-change of each peptide across four replicates. Peptides highlighted in red contain the NLV peptide and the blue peptides highlight additional enriching peptides. Data presented in Table S4. (C) Table of enriching peptides highlighted in (B) and Figure 2E. Predicted HLA-A2 binding peptides, including the known NLV epitope are bolded. (D) GzB reporter activation in target cells pulsed with the pp65 peptide (NLVPMVATV) or IE1 peptide (VLEETSVML) in the presence of NLV3 TCR T cells (top panel) or NLV-expanded T cells (bottom panel). Error bars indicate SD across three replicates. (E) Tetramer staining of the NLV-expanded T cells with the pp65 peptide and the IE1 peptide. Gated CD8⁺ cells are shown.

Figure 4.. CMV Genome-wide Screen with Bulk Memory CD8 T Cells

(A) Generation of memory CD8 T cells. Bulk memory CD8 T cells are purified from donor blood and undergo polyclonal expansion in the presence of feeder cells and anti-CD3. (B) T-Scan screen of bulk memory CD8 cells against CMV genome-wide library. Each dot represents one peptide, with the X and Y axes plotting the performance of two unique barcodes for the peptide. The X and Y values indicate the modified geometric mean (see STAR Methods) of the fold-change of each barcode across four replicates. Colored dots highlight pairs of overlapping peptides. Data presented in Table S5. (C) List of enriching peptides highlighted in (B). Predicted high-affinity HLA-A2 binding epitopes in the peptide overlap regions are bolded and underlined. (D) Tetramer staining of the memory CD8 T cells used in the screen with the predicted peptides bolded in (C).

Figure 5.. Comprehensive Mutagenesis Analysis of NLV-Specific T Cells

(A) Design of the NLV epitope comprehensive mutagenesis library. Every position in the NLV epitope (NLVPMVATV) is mutated to each of the 19 alternative amino acids. The mutant epitopes are expressed in the context of a 56-aa fragment with an N-terminal tag or as short peptides with an N-terminal tag. The two amino acids directly N-terminal and C-terminal to the epitope are also mutated in the 56-aa versions. (B–D) Recognition of mutant peptides in the context of 56-aa fragments by NLV-specific T cells. Each box in the heatmap represents one mutant, where the amino acid along the x axis is mutated to the amino acid indicated along the y axis. The value in the heatmap represents the enrichment of this mutant compared to the WT NLV epitope. Heatmaps are plotted for (B) NLV-expanded T cells, (C) T cells expressing the NLV2 TCR, and (D) T cells expressing the NLV3 TCR. Data for (B)–(D) are presented in Table S6. (E) Relative activation of the GzB reporter in cells pulsed with identified NLV2 TCR-specific and NLV3 TCR-specific mutant peptides and co-cultured with NLV2 or NLV3 TCR T cells. Error bars indicate SD across three replicates. p values were determined by a two-tailed t test and are shown with asterisks, *<0.05, **<0.01, ***<0.001. (F) Predicted MHC binding affinity of each mutant. The mutants are displayed as in (B)–(D) and the values plotted are nM affinities based on the NetMHC algorithm. Boxes are drawn around all mutants that enriched to at least 50% of WT levels in one or more of the T cell experiments from (B)–(D). (G) GzB reporter activation following NLV3 T cell co-culture with target cells that were pulsed with dilutions of the WT NLV epitope or two predicted low-affinity mutants of the NLV epitope. Error bars indicate SD across three replicates. See also Figures S4 and S5.

Figure 6.. Human Genome-wide T Cell Antigen Discovery

(A) Schematic of human genome-wide T-Scan screen. Peptides tiling across the human proteome in 90-aa steps with 45-aa overlap are expressed via lentiviral transduction of target cells expressing HLA-A1, IFP^GZB, and ICAD^CR. MAGE-A3 TCR T cells are generated by lentiviral transduction of donor T cells with a construct encoding the MAGE-A3 TCR. (B) T-Scan screen of MAGE-A3 TCR expressing T cells against human genome-wide library. Each dot represents one peptide, with the y axis plotting the geometric mean of the fold-change of each peptide across eight replicates. Data presented in Table S7. (C) List of the predicted HLA-A1 binding epitopes from the antigens identified in (B) from MAGE-A3 (NM_005362.4), MAGE-A6 (NM_005363.3), PLD5 (NM_152666.1), and FAT2 (NM_001447.2). (D) Activation of GzB reporter by MAGE-A3 TCR T cells in the presence of the candidate antigens. Untransduced EDCs (parental), EDCs pulsed with the HIV IV9 peptide (peptide), and EDCs transduced with the GFP ORF (ORF) were used as controls. Peptides were added at 1 μM. Fragments of 90 amino acids and full-length ORFs containing each antigen were stably expressed by lentiviral transduction. The FAT2 synthetic ORF fragments include the region surrounding the FAT2 epitope fused to the CD8 or FAT2 signal peptide and transmembrane domain. Error bars indicate SD across three replicates.