Combined Analysis of Antigen Presentation and T-cell Recognition Reveals Restricted Immune Responses in Melanoma

Abstract

The quest for tumor-associated antigens (TAA) and neoantigens is a major focus of cancer immunotherapy. Here, we combine a neoantigen prediction pipeline and human leukocyte antigen (HLA) peptidomics to identify TAAs and neoantigens in 16 tumors derived from seven patients with melanoma and characterize their interactions with their tumor-infiltrating lymphocytes (TIL). Our investigation of the antigenic and T-cell landscapes encompassing the TAA and neoantigen signatures, their immune reactivity, and their corresponding T-cell identities provides the first comprehensive analysis of cancer cell T-cell cosignatures, allowing us to discover remarkable antigenic and TIL similarities between metastases from the same patient. Furthermore, we reveal that two neoantigen-specific clonotypes killed 90% of autologous melanoma cells, both in vitro and in vivo, showing that a limited set of neoantigen-specific T cells may play a central role in melanoma tumor rejection. Our findings indicate that combining HLA peptidomics with neoantigen predictions allows robust identification of targetable neoantigens, which could successfully guide personalized cancer immunotherapies.Significance: As neoantigen targeting is becoming more established as a powerful therapeutic approach, investigating these molecules has taken center stage. Here, we show that a limited set of neoantigen-specific T cells mediates tumor rejection, suggesting that identifying just a few antigens and their corresponding T-cell clones could guide personalized immunotherapy. Cancer Discov; 8(11); 1366-75. ©2018 AACR. This article is highlighted in the In This Issue feature, p. 1333.

Figure 1.

Tumor antigen discovery pipeline. Whole-exome sequencing (WES) of 15 melanoma tumor samples derived from 6 patients and one melanoma cell line was performed in parallel to HLA-peptidome analysis of the cells’ HLA-I and HLA-II repertoires. Integrating the WES data with the human proteome database in the mass spectrometry analysis allowed us to reveal the neoantigens and TAAs presented by the patients’ tumor cells. In parallel, we applied a neoantigen prediction pipeline followed by a long peptide screen. NT, not tested. TILs isolated from the tumor were sequenced to identify their TCR sequences and reveal any high similarity between the different metastases from the same patient. Neoantigens and TAAs were tested for their reactivity to TILs and the ability of TILs to target the melanoma cells was characterized using in vitro and in vivo imaging. Neoantigen-specific clones were isolated using tetramers and sequenced to identify their TCR sequence, and TIL reactivity was also derived from their prevalence in the bulk TILs.

Figure 2.

High similarity in the HLA peptides and TCR repertoires of metastases from the same patient. A and B, Similarity in the heat maps of HLA-I (A) and HLA-II (B) peptides from the different tumor metastases. Color code indicates the Jaccard index. The highest similarity was observed between metastases from the same patient and between patients with shared HLA alleles. C and D, The similarity between the presented peptides is observed not only in terms of their identity but also in terms of their intensity. The log₂ of the peptide intensities were plotted for HLA-I (C) and HLA-II (D) peptides. Unique peptides for each sample were given a constant value of 15. Peptides derived from TAAs are marked in red. Pearson correlations are indicated in red. E, Similarity heat maps of the TCR amino acid sequences identified in the various tumor metastases. Color code indicates the Jaccard index. F, The frequencies of the various TCRs in the different metastases from the same patient. Because at lower frequencies we found many different TCRs with the same frequency in both metastases, we color-coded the number of TCRs represented by each dot. Pearson correlation is indicated in red. G-H, The frequency of the nucleotide sequences of the two most convergent amino acid TCRβ sequences. In patients 51 (G) and 92 (H), one of the most convergent sequences was detected as most convergent in both metastases. Each nucleotide sequence is represented by layer, with the overlapping sequences presented in the same color in both metastases.

Figure 3.

Neoantigen-specific T cells show more reactivity, killing ability, and proliferation in vitro and in vivo. A, IFNγ release measured after overnight coculture of the TILs with EBV-transformed B cells that were pulsed with 10 μM of the peptides that were identified in the HLA peptidomics analysis. B, IFNγ release measured after overnight coculture of the TILs with EBV-transformed B cells that were pulsed with 10 μM of the 25-mer peptides. For the reactive 25-mers, we also checked the reactivity of each minimal epitope that was predicted to bind the patients’ HLA alleles. C, A fluorescence-based in vitro assay comparing the killing of 12T melanoma cells by autologous bulk 12TILs, enriched neoantigen TIL population, or the rest of the non-neoantigenic TIL population with an increasing effector:target (E:T) ratio. D, Bulk TILs were stained with NCAPH2, MED15, GPNMB, and CDK4 tetramers to evaluate the percentage of the different populations in the bulk TILs. E, 12T melanoma cells were cocultured with 12TILs for 24 hours, and later were stained with anti-4–1BB antibody and the two tetramers against the neoantigens. The percentage of reactive and nonreactive T cells for each neoantigen, neoantigen-negative, and bulk TIL are indicated. The mean fluorescence intensity (MFI) of the 4–1BB staining was calculated for the reactive T cells in each population. F, Flow cytometry analysis of the different antigen populations in the 12TILs before and after injection to NSG mice with 12T melanoma cells tumor. The first panel shows the percentage of the 12TILs and irrelevant 108TILs in the T-cell mixture. The last three panels are gated only to the violet positive cells (only 12TILs) and show the percentage of T cells in each neoantigen or tetramer-negative population that proliferate. The images are representative for three replicates.

Figure 4.

Ten most frequent TCRs in the bulk 12TILs comprise >99% of the TIL population, and the neoantigens are shown to be the most dominant clones. Bulk TILs and five different populations isolated from two different sorting experiments were analyzed using TCR sequencing to identify neoantigen-specific clones and reactive and nonreactive T-cell populations. A, Frequency of the top-10 abundant amino acid sequences found in the bulk TILs across the different samples and the second most abundant sequence from NCAPH2. As shown in the bars, these 11 TCRs comprise 90.75%−99.94% of the productive rearrangements in all the samples. B, The table indicates the 11 TCR sequences and their frequencies in the different samples. Clones 1 and 11 are against NCAPH2 and clones 3 and 4 are against MED15. Clone 2 is found in the tetramer-negative population and mostly negative to 4–1BB, and therefore probably represents a clone that expended in the tumor but its antigen was downpresented in the tumor cells as part of the immune-editing process.