Somatic Mutations and Neoepitope Homology in Melanomas Treated
Total Page:16
File Type:pdf, Size:1020Kb
Published OnlineFirst December 12, 2016; DOI: 10.1158/2326-6066.CIR-16-0019 Research Article Cancer Immunology Research Somatic Mutations and Neoepitope Homology in Melanomas Treated with CTLA-4 Blockade Tavi Nathanson1, Arun Ahuja1, Alexander Rubinsteyn1, Bulent Arman Aksoy1, Matthew D. Hellmann2,3, Diana Miao4,5, Eliezer Van Allen4,5, Taha Merghoub2,3,6, Jedd D. Wolchok2,3,6, Alexandra Snyder2,3, and Jeff Hammerbacher1 Abstract Immune checkpoint inhibitors are promising treatments for narrowed from somatic mutation burden, to inclusion of only patients with a variety of malignancies. Toward understanding the those mutations predicted to be MHC class I neoantigens, to determinants of response to immune checkpoint inhibitors, it was only including those neoantigens that were expressed or that previously demonstrated that the presence of somatic mutations had homology to pathogens. The only association between is associated with benefit from checkpoint inhibition. A hypoth- somatic mutation burden and response was found when exam- esis was posited that neoantigen homology to pathogens may in ining samples obtained prior to treatment. Neoantigen and part explain the link between somatic mutations and response. To expressed neoantigen burden were also associated with further examine this hypothesis, we reanalyzed cancer exome data response, but neither was more predictive than somatic muta- obtained from our previously published study of 64 melanoma tion burden. Neither the previously described tetrapeptide patients treated with CTLA-4 blockade and a new dataset of RNA- signature nor an updated method to evaluate neoepitope Seq data from 24 of these patients. We found that the ability to homology to pathogens was more predictive than mutation accurately predict patient benefit did not increase as the analysis burden. Cancer Immunol Res; 5(1); 84–91. Ó2016 AACR. Introduction In our study of melanomas treated with checkpoint blockade agents targeting cytotoxic T-lymphocyte associated protein 4 Checkpoint blockade therapies are improving outcomes for (CTLA-4; ref. 8), we present the hypothesis that responding patients with metastatic solid tumors (1–4). As only a subset of tumors may share features with each other or with infectious patients responds, there is a critical need to identify determi- agents and that such resemblance may predict response. In this nants of response. Expression ofprogrammeddeathligandone report, we reanalyzed the data in that study using updated (PD-L1) is the lead companion diagnostic for PD-1/PD-L1 methods and integrating new RNA sequencing (RNA-Seq) data blockade therapies, but sensitivity and specificity are limited from a subset of 24 samples. (5–7). An association between elevated tumor mutation bur- We found that in this small dataset, nonsynonymous mutation den and benefit from checkpoint blockade therapies has been burden was associated with clinical benefit from therapy in demonstrated (8–11). samples collected before, but not after, treatment with CTLA-4 blockade. Predicted neoantigen burden and percentage of C!T 1Genetics and Genomic Sciences, Icahn School of Medicine at Mount Sinai, New transitions characteristic of ultraviolet damage were associated York, New York. 2Department of Medicine, Memorial Sloan Kettering Cancer with, but did not outperform, mutation burden. We develop- Center, New York, New York. 3Department of Medicine, Weill Cornell Medical ed a publicly available tool, Topeology (https://github.com/ College, New York, New York. 4Department of Medical Oncology, Dana-Farber hammerlab/topeology), to compare neoantigens to known Cancer Institute, Broad Institute of MIT and Harvard, Boston, Massachusetts. pathogens. Neither the resemblance of tumor neoantigens to 5Center for Cancer Precision Medicine, Dana-Farber Cancer Institute, Boston, known antigens nor the previously published tetrapeptide signa- Massachusetts. 6Swim Across America–Ludwig Collaborative Research Labo- ratory, Immunology Program, Ludwig Center for Cancer Immunotherapy, New ture outperformed mutation burden as a predictor of response. York, New York. Note: Supplementary data for this article are available at Cancer Immunology Materials and Methods Research Online (http://cancerimmunolres.aacrjournals.org/). Patient samples T. Nathanson, A. Ahuja, A. Snyder, and J. Hammerbacher contributed equally to All analyzed samples were collected in accordance with local this article. Internal Review Board policies as described in ref. 8 and summa- Online supplementary data files can be found at http://www.hammerlab.org/ rized in Table 1. Thirty-four patients had tumor samples collected melanoma-reanalysis. Epitope comparison tool can be found at https://github. prior to initiating CTLA-4 blockade, and 30 patients had samples com/hammerlab/topeology collected after initiating CTLA-4 blockade. Clinical benefit Corresponding Author: Alexandra Snyder, Memorial Sloan Kettering Cancer was defined as progression-free survival lasting for greater than Center, 300 East 66th Street, New York, NY 10065. Phone: 646-888-5122; Fax: 24 weeks after initiation of therapy (Online Data File 1). Nine 646-888-4265; E-mail: [email protected] discordant lesions were present, where overall patient benefit did doi: 10.1158/2326-6066.CIR-16-0019 not match individual tumor progression. See Table 1 for details Ó2016 American Association for Cancer Research. about this patient cohort. 84 Cancer Immunol Res; 5(1) January 2017 Downloaded from cancerimmunolres.aacrjournals.org on September 24, 2021. © 2017 American Association for Cancer Research. Published OnlineFirst December 12, 2016; DOI: 10.1158/2326-6066.CIR-16-0019 Mutations and Neoepitopes in Melanoma Table 1. Cohort summary 18, 2015, from the MSigDB website (http://www.broadinstitute. Group Benefit No benefit Discordant org/gsea/msigdb/index.jsp). The Hallmark gene set collection was N 27 28 9 extended by adding gene symbols corresponding to well-known % Cutaneous 20/27 19/28 5/9 peptides that are (i) tumor specific; (ii) associated with differen- OS 3.7 (1.6–7.3) 0.8 (0.2–2.7) 4 (1.7–7.9) Age 65 (33–81) 58.5 (18–79) 68 (40–90) tiation; and (iii) overexpressed in cancer cells. To do this, gene Mutations 611 (165–3,394) 321 (6–1,816) 549 (93–1,336) symbols were imported from the Cancer Immunity Peptide Neoantigens 1,388 (209–6,502) 714.5 (3–4,510) 1,048 (197–2,584) Database as gene sets that are compatible with the GSEA software. NOTE: Features of tumors from patients with clinical benefit, no benefit, or in Before running the GSEA, the gene expression data (FPKM) were which a discordant lesion was resected. collapsed using official gene symbol identifiers and the median Abbreviation: OS, overall survival. expression value used when multiple transcripts mapped to the same gene symbol. To normalize the data further, noninformative genes with no variation (SD of 0) across all samples were Mutation calls removed. Three GSEA analyses were conducted comparing: (i) Single-nucleotide variants (SNV) were called with an ensemble pretreatment benefiting versus pretreatment nonbenefiting; (ii) of four variant callers: Mutect, Strelka, SomaticSniper, and Varscan pretreatment benefiting versus posttreatment benefiting; and (iii) as described previously (9). Insertions and deletions (indels) were pretreatment nonbenefiting versus posttreatment nonbenefiting. called using Strelka with default settings In all these comparisons, the normalized gene expression values were used as the input matrix. The number of permutations was HLA typing set to 1,000, restricting these permutations to gene set labels rather HLA types were determined by ATHLATES for all samples using than the sample phenotype labels due to our sample size, and we fi exome sequence data and con rmed with seq2HLA for samples kept the rest of the default options (see http://www.hammerlab. that had RNA-Seq available (24 samples; Online Data File 2). org/melanoma-reanalysis/gsea-results/ for complete reports and instructions to replicate them). Neoepitope prediction Somatic SNVs that occurred a single base away from other Neoantigen homology somatic SNVs were combined into larger variants containing both We developed a tool, Topeology, to compare tumor neoepitopes SNVs. For each somatic variant, we used Topiary (https://github. with entries in the Immune Epitope Database (IEDB; Online Data com/hammerlab/topiary) to generate the predicted 8–11mer File 7; ref. 13), accounting for position, amino acid gaps, and amino acid product resulting from somatic alterations (SNV or biochemical similarity between amino acids. Epitopes were com- indel), including predicted neoepitopes generated from com- pared, and the comparison scored, using the Smith–Waterman bined SNVs (Online Data Files 3–4). Each variant was linked to alignment algorithm (14) supplied with a substitution matrix its corresponding coding DNA sequence (CDS) from Ensembl consisting of PMBEC correlation values derived from the PMBEC based on its B37 coordinates. The CDS sequence was retranslated covariance matrix (15). We compared amino acids from position 3 with the mutated DNA residue producing the mutated peptide to the penultimate amino acid of the peptide, assuming that the product. NetMHCcons v1.1 generated a predicted binding affi- anchor residues would be necessary for MHC class I presentation nity for all 8-mers to 11-mers containing the mutated amino and would therefore not be "visible" to a T cell. A gap penalty equal acid and all peptides with an IC50 score below 500 nmol/L were to the lowest PMBEC correlation