Genomic correlates of response to CTLA-4 blockade in metastatic melanoma

Authors

Eliezer M. Van Allen

Diana Miao

Bastian Schilling

Sachet A. Shukla

Christian Blank

Lisa Zimmer

Antje Sucker

Uwe Hillen

Marnix H. Geukes Foppen

Simone M. Goldinger

Jochen Utikal

Jessica C. Hassel

Benjamin Weide

Katharina C. Kaehler

Carmen Loquai

Peter Mohr

Ralf Gutzmer

Reinhard Dummer

Stacey Gabriel

Catherine J. Wu

Dirk Schadendorf

Levi A. Garraway

Doi

10.1126/science.aad0095

PMID: 26359337 · DOI: 10.1126/science.aad0095 · Journal: Science (2015)

TL;DR

Van Allen et al. performed whole-exome sequencing on pretreatment tumors and matched germline DNA from 110 patients with metastatic melanoma treated with single-agent ipilimumab (anti-CTLA-4), with matched tumor RNA-seq for 40 of them. Overall nonsynonymous mutational load (P = 0.0076), predicted neoantigen load (P = 0.027), and a cytolytic tumor microenvironment signature (GZMA/PRF1 expression, P = 0.042) were each significantly associated with clinical benefit from ipilimumab. Tumor expression of CTLA4 itself (P = 0.033) and PD-L2 / PDCD1LG2 (P = 0.041) was also elevated in responders. However, no recurrent neoantigen peptide sequence predicted response — putative response-associated neoantigens were overwhelmingly private events — and a previously reported tetrapeptide signature failed to validate in this cohort.

Cohort & data

  • 110 metastatic melanoma patients treated with ipilimumab monotherapy (skcm_dfci_2015); pretreatment tumor + matched germline whole-exome sequencing for all 110, tumor RNA-seq for 42 (40 with matched WES).
  • Histology breakdown: 92 cutaneous (SKCM), 4 mucosal, 14 occult melanomas; OncoTree parent MEL.
  • Average exome-wide target coverage: 183.7× (tumor) and 157.2× (germline).
  • Patients stratified by RECIST + overall survival into clinical benefit (CR/PR or SD with OS >1 yr, n = 27), no clinical benefit (PD or SD with OS <1 yr, n = 73), and a separate long-term-survival/early-progression subgroup (PFS <6 mo but OS >2 yr, n = 10).
  • All sequencing data deposited in dbGaP accession phs000452.v2.p1.
  • Assays: whole-exome-seq, rna-seq. Drug: ipilimumab.

Key findings

  • Median nonsynonymous mutational load was 197 per sample (range 7–5854), consistent with the high mutational burden of cutaneous melanoma (PMID:26359337).
  • Nonsynonymous mutational load was significantly higher in patients with clinical benefit from ipilimumab than in those without (P = 0.0076; Mann-Whitney) (PMID:26359337).
  • Predicted neoantigen load (9–10 aa peptides, HLA class I binding ≤500 nM) was significantly associated with clinical benefit (P = 0.027); median predicted neoantigen load was 369 per sample (range 9–16,300). Correlation with mutational load was very high (Spearman ρ = 0.97, P < 0.0001) (PMID:26359337).
  • The association remained significant with randomly redistributed HLA types (P = 0.0096), and in multivariate models controlling for prior RAF inhibitor and M class (mutational load >100: P = 0.0169; neoantigen load >100: P = 0.0371) (PMID:26359337).
  • Increasing the HLA-binding affinity stringency did not improve the predictive signal (P = 0.034, 0.038, 0.042 for affinity <250, <100, <50 nM, respectively) (PMID:26359337).
  • Of 77,803 unique neoantigens, only 28 (~0.04%) were observed in more than one patient with clinical benefit and absent from non-responders; none of these shared obvious sequence features, and a previously reported tetrapeptide signature (Snyder et al. 2014) did not validate in this cohort (PMID:26359337).
  • In the 40-patient RNA-matched subset, filtering DNA-predicted neoantigens by patient-matched tumor RNA expression cut median neoantigen load from 395 to 198 per patient; TCGA bulk-melanoma expression filtering gave only partial overlap with patient-matched RNA filtering (PMID:26359337).
  • The cytolytic-activity signature defined as the geometric mean of GZMA and PRF1 expression was significantly elevated in clinical-benefit tumors vs. no-benefit (P = 0.042; Mann-Whitney) (PMID:26359337).
  • Tumor CTLA4 expression (P = 0.033) and PD-L2 / PDCD1LG2 expression (P = 0.041) were each significantly higher in clinical-benefit tumors (PMID:26359337).
  • HLA class I genes were expressed in all patients with no significant difference across response groups, and somatic mutations in HLA class I were rare and did not segregate by response (PMID:26359337).
  • No genes — including BRAF and NRAS — were enriched for nonsynonymous mutations in either response subgroup (figs. S1, S2) (PMID:26359337).
  • Clinical covariates (age, gender, histology, primary site, lines of prior therapy, baseline LDH) did not correlate with response (all P > 0.05) (PMID:26359337).

Genes & alterations

  • BRAF — no enrichment of nonsynonymous mutations in clinical-benefit vs. no-benefit subgroups in this ipilimumab cohort (figs. S1, S2) (PMID:26359337).
  • NRAS — no enrichment of nonsynonymous mutations by response group, despite prior reports linking NRAS status to immune-therapy response (PMID:26359337).
  • CTLA4 — tumor RNA expression of CTLA4 itself was significantly higher in patients with clinical benefit from ipilimumab (P = 0.033; Mann-Whitney) (PMID:26359337).
  • PDCD1LG2 (PD-L2) — tumor RNA expression significantly higher in clinical-benefit patients (P = 0.041) (PMID:26359337).
  • GZMA — component of the cytolytic-activity signature (geometric mean of GZMA + PRF1); elevated in clinical-benefit tumors (P = 0.042) (PMID:26359337).
  • PRF1 — component of the cytolytic-activity signature; elevated in clinical-benefit tumors (P = 0.042) (PMID:26359337).

Clinical implications

  • Tumor mutational load and predicted neoantigen load measured from pretreatment whole-exome sequencing are statistically associated with clinical benefit from CTLA-4 blockade with ipilimumab in metastatic melanoma, supporting use of mutational/neoantigen burden as a candidate predictive biomarker — though responder/non-responder distributions overlap substantially (high-burden non-responders and low-burden responders exist) (PMID:26359337).
  • RNA-level features of the tumor microenvironment — cytolytic activity (GZMA + PRF1) and expression of immune-checkpoint molecules (CTLA4, PDCD1LG2) — add information complementary to DNA-level burden and are themselves elevated in responders (PMID:26359337).
  • Recurrent / shared neoantigen peptides are not a viable per-patient biomarker at the cohort size achievable in 2015: response-associated neoantigens are overwhelmingly private, and previously published candidate epitopes (including a tetrapeptide signature) did not validate (PMID:26359337).
  • HLA class I loss is unlikely to be a dominant resistance mechanism in this cohort: HLA class I genes were expressed across all response groups and somatic HLA class I mutations were rare (PMID:26359337).

Limitations & open questions

  • The cohort was powered to detect aggregate burden signals but underpowered to identify recurrent neoantigen sequences; the authors explicitly note that detecting recurrent response-mediating neoantigens will require cohorts much larger than 110 patients (PMID:26359337).
  • The RNA-matched subset is only 40 patients (13 clinical benefit, 22 no benefit, 5 long-term survival), too small to reach significance on RNA-filtered neoantigen load by response group (PMID:26359337).
  • The “long-term survival with early progression” subgroup (n = 10) had mutational-load distributions resembling non-responders, so the genomic basis for delayed/long-term benefit remains unresolved (PMID:26359337).
  • Neoantigen prediction was restricted to HLA class I 9–10mers; class II neoantigens and fusion-derived neoantigens were not assessed and are flagged as future directions (PMID:26359337).
  • Clinical benefit was defined using a composite endpoint mixing RECIST response and OS >1 yr; while clinically motivated, this binarization may obscure signal in patients with mixed/delayed responses (PMID:26359337).
  • Tumor samples were pretreatment biopsies from a single time point; on-treatment dynamics of mutational/neoantigen/cytolytic signals were not measured (PMID:26359337).

Citations from this paper used in the wiki

  • “Overall mutational load, neoantigen load, and expression of cytolytic markers in the immune microenvironment were significantly associated with clinical benefit. However, no recurrent neoantigen peptide sequences predicted responder patient populations.” (Abstract)
  • “The median nonsynonymous mutational load was 197 per sample (range: 7 to 5854).” (Results, p. 2)
  • “Overall, nonsynonymous mutational load was significantly associated with clinical benefit from ipilimumab (P = 0.0076; Mann-Whitney test).” (Results, p. 3)
  • “Overall, neoantigen load was significantly associated with clinical benefit (P = 0.027; Mann-Whitney).” (Results, p. 3)
  • “These analyses confirmed that patients with high neoantigen or mutational loads (>100) were significantly more likely to have clinical benefit from ipilimumab (P = 0.0371 and P = 0.0169, respectively).” (Results, p. 3)
  • “Of the 77,803 unique neoantigens in our cohort, 28 (~0.04%) were found in more than one patient having a clinical benefit but were absent in all patients having no clinical benefit or long-term survival.” (Results, p. 3)
  • “Both genes [GZMA and PRF1] were significantly enriched in the cohort treated with ipilimumab that had clinical benefit compared with the cohort that showed no clinical benefit (P = 0.042; Mann-Whitney).” (Results, p. 4)
  • “Expression of CTLA-4 itself and of the programmed cell death 1 ligand 2 (PD-L2)… was also significantly elevated in the cohort that showed clinical benefit compared with the one that showed no clinical benefit (P = 0.033 and P = 0.041; Mann-Whitney).” (Results, p. 4)
  • “All sequencing data are available in dbGap, accession number phs000452.v2.p1.” (Acknowledgements)

This page was processed by crosslinker on 2026-05-14.