Mutational landscape determines sensitivity to PD-1 blockade in non–small cell lung cancer

Authors

Naiyer A. Rizvi

Matthew D. Hellmann

Alexandra Snyder

Pia Kvistborg

Vladimir Makarov

Jonathan J. Havel

William Lee

Jianda Yuan

Phillip Wong

Teresa S. Ho

Martin L. Miller

Natasha Rekhtman

Andre L. Moreira

Fawzia Ibrahim

Cameron Bruggeman

Billel Gasmi

Roberta Zappasodi

Yuka Maeda

Chris Sander

Edward B. Garon

Taha Merghoub

Jedd D. Wolchok

Ton N. Schumacher

Timothy A. Chan

Doi

10.1126/science.aaa1348

PMID: 25765070 · DOI: 10.1126/science.aaa1348 · Journal: Science (2015)

TL;DR

Rizvi et al. performed whole-exome sequencing on tumors from two independent cohorts of NSCLC patients (discovery n=16, validation n=18) treated with the anti–PD-1 antibody pembrolizumab. Higher nonsynonymous somatic mutation burden was significantly associated with objective response, durable clinical benefit (DCB), and progression-free survival (PFS). A molecular smoking signature, neoantigen burden, and deleterious mutations in DNA repair/replication genes (notably POLD1, POLE, MSH2) also correlated with efficacy. In one exceptional responder, anti–PD-1 therapy expanded a CD8+ T cell response specific for a mutant HERC1 P3278S neoantigen that paralleled radiographic tumor regression PMID:25765070.

Cohort & data

  • Two independent cohorts of advanced NSCLC patients treated with pembrolizumab: discovery cohort n=16 and validation cohort n=18 (34 paired tumor/normal exomes total), cBioPortal study nsclc_mskcc_2015 PMID:25765070.
  • Assay: tumor/normal whole-exome sequencing with mean target coverage 164× (94.5% of target covered ≥10×); orthogonal Ampliseq resequencing of 376 variants confirmed 357 (95%) PMID:25765070.
  • Cancer type: NSCLC (advanced, anti–PD-1–treated) PMID:25765070.
  • PD-L1 expression measured by PD-L1 IHC 22C3 (clone 22C3, Merck) on 30/34 patients PMID:25765070.
  • HLA class I–restricted neoantigen prediction (mutant nonamers with ≤500 nM binding affinity) and MHC multimer screening on serially collected peripheral blood lymphocytes PMID:25765070.
  • Data deposited at cBioPortal (study “Rizvi lung cancer”) and dbGaP accession phs000980.v1.p1 PMID:25765070.

Key findings

  • Discovery cohort (n=16): median nonsynonymous mutations 302 in DCB vs 148 in NDB (Mann-Whitney P=0.02); 73% of high-burden patients (above median 209) achieved DCB vs 13% of low-burden patients (Fisher P=0.04). ORR 63% vs 0% (Fisher P=0.03); median PFS 14.5 vs 3.7 months (HR 0.19, 95% CI 0.05–0.70, log-rank P=0.01) PMID:25765070.
  • Validation cohort (n=18): median nonsynonymous mutations 244 in DCB vs 125 in NDB (Mann-Whitney P=0.04). DCB 83% vs 22% above median 200 (Fisher P=0.04); median PFS not reached vs 3.4 months (HR 0.15, 95% CI 0.04–0.59, log-rank P=0.006) PMID:25765070.
  • Cutoff ≥178 nonsynonymous mutations (chosen for maximal sensitivity/specificity in discovery): discovery sensitivity 100%, specificity 67%; validation sensitivity 86%, specificity 75%; only one patient with <178 mutations achieved an objective response (transient, 8 months) PMID:25765070.
  • Pooled (n=34): median nonsynonymous mutations 299 (DCB) vs 127 (NDB), Mann-Whitney P=0.0008; PFS HR 0.19 (95% CI 0.08–0.47, log-rank P=0.0004) PMID:25765070.
  • Mutation spectrum: C→A transversions enriched and C→T transitions depleted in DCB vs NDB (both Mann-Whitney P=0.01); transversion/transition ratio 0.74 overall PMID:25765070.
  • Molecular smoking signature (TH vs TL classifier): ORR 56% TH vs 17% TL (Fisher P=0.03); DCB 77% vs 22% (Fisher P=0.004); PFS not reached vs 3.5 months (HR 0.15, 95% CI 0.06–0.39, log-rank P=0.0001). Self-reported smoking history did not significantly stratify DCB or PFS (Fisher P=0.66, log-rank P=0.29) PMID:25765070.
  • Neoantigen burden: median 112 candidate neoantigens per tumor (range 8–610); strongly correlated with mutation burden (Spearman ρ=0.91, P<0.0001). DCB tumors had median 203 vs 83 in NDB (Mann-Whitney P=0.001); PFS 14.5 vs 3.5 months by neoantigen burden (log-rank P=0.002). Specific HLA alleles did not correlate with efficacy PMID:25765070.
  • Exceptional responder (Study ID 9): anti–PD-1 induced a CD8+ T cell response against a HERC1 P3278S neoantigen (ASNASSAAK, HLA-A–restricted). Baseline frequency <0.005%, rising to 0.040% of CD8+ cells by day 21 and maintained at day 44, paralleling tumor regression. Reactive T cells were CD45RA−CCR7−HLA-DR+LAG-3+, consistent with activated effector phenotype. Polyfunctional response (IFNγ, TNFα, CD107a, CCL4) to mutant but not wild-type peptide confirmed specificity PMID:25765070.
  • PD-L1 + mutation burden: among PD-L1–expressing tumors (weak or strong), DCB was 91% (10/11) with high mutation burden (>200) vs 10% (1/10) with low burden; in weak PD-L1 specifically, high burden gave 75% DCB vs 11% for low PMID:25765070.

Genes & alterations

  • POLD1 — exonuclease-domain mutations E374K (in a never-smoker DCB patient with the highest never-smoker burden, 507 nonsynonymous mutations) and C284Y (in the highest-burden tumor in the series). Both lie in the proofreading domain and likely produce a hypermutator phenotype with C→T predominance distinct from the smoking signature PMID:25765070.
  • POLE — deleterious mutation identified in one of three responders with the highest mutation burden PMID:25765070.
  • MSH2 — deleterious mutation identified in one of three top-burden responders, consistent with mismatch-repair–driven hypermutation PMID:25765070.
  • PRKDC — nonsense mutations observed; required for DNA-dependent protein kinase (DNA-PK) activity and genomic integrity PMID:25765070.
  • RAD17 — nonsense mutations observed; required for proper DNA damage response PMID:25765070.
  • POLR2A, KEAP1, PAPPA2, PXDNL, RYR1, SCN8A, SLIT3 — each harbored deleterious mutations in ≥4 DCB patients and 0 NDB patients (associations not corrected for mutation burden) PMID:25765070.
  • KRAS — mutated in 7/14 DCB tumors vs 1/17 NDB tumors; authors attribute the imbalance to the established association between smoking and KRAS mutations in NSCLC rather than a direct causal effect on PD-1 response PMID:25765070.
  • CD274 (PD-L1) — no mutations or copy-number alterations in CD274 or in antigen-presentation pathway genes were associated with response or resistance PMID:25765070.
  • HERC1 — P3278S missense mutation generated the validated HLA-A–restricted neoantigen ASNASSAAK that drove a measurable, treatment-induced CD8+ T cell response in an exceptional responder PMID:25765070.

Clinical implications

  • Biomarker: Nonsynonymous tumor mutation burden (with a candidate cutoff of ≥178 mutations in this exome-derived series) is a candidate predictive biomarker for durable clinical benefit and prolonged PFS on pembrolizumab in NSCLC PMID:25765070.
  • Mechanism: The data support a neoantigen-mediated mechanism for anti–PD-1 efficacy — total nonsynonymous burden correlated more tightly with benefit than total exonic burden, and a single neoantigen-specific T cell expansion paralleled tumor regression in vivo PMID:25765070.
  • Combinatorial biomarker: Mutation burden stratified outcomes even within PD-L1–positive tumors, suggesting mutation burden and PD-L1 IHC (pd-l1-ihc-22c3) capture complementary information PMID:25765070.
  • DNA repair as a precision-immuno-oncology axis: Deleterious mutations in POLD1, POLE, and MSH2 identify a never-smoker subset of NSCLC that can still reach the high-burden state and benefit from PD-1 blockade PMID:25765070.
  • Monitoring: Detection of neoantigen-specific CD8+ T cell expansions in peripheral blood raises the possibility of blood-based pharmacodynamic assays for anti–PD-1 therapy PMID:25765070.

Limitations & open questions

  • Small cohorts (n=16 discovery, n=18 validation) limit precision of cutoff selection and subgroup analyses PMID:25765070.
  • The trial enrolled patients enriched for PD-L1 expression (24/30 with detectable expression), limiting independent assessment of mutation burden in PD-L1–negative tumors PMID:25765070.
  • Gene-level associations (e.g. POLR2A, KEAP1, PAPPA2, PXDNL, RYR1, SCN8A, SLIT3) were not corrected for the elevated overall mutation burden in DCB tumors and require independent validation PMID:25765070.
  • The 178-mutation cutoff was derived from whole-exome data and does not directly transfer to targeted panels without recalibration PMID:25765070.
  • Five of 18 high-burden tumors (≥178 mutations) had no durable benefit, indicating mutation burden is necessary but not sufficient for response — additional immune resistance mechanisms remain to be characterized PMID:25765070.
  • Neoantigen-specific T cell tracking was demonstrated in only one exceptional responder; the generalizability of peripheral neoantigen reactivity as a response biomarker is unestablished PMID:25765070.
  • The cohort and mutation spectrum are consistent with prior NSCLC genomic series including the TCGA LUAD landscape PMID:25079552; the conceptual framework extends Snyder et al.’s prior neoantigen findings in melanoma–ipilimumab PMID:25409260 into the anti–PD-1 NSCLC setting PMID:25765070.

Citations from this paper used in the wiki

  • “In the discovery cohort (n = 16), the median number of nonsynonymous mutations was 302 in patients with durable clinical benefit (DCB)…versus 148 with no durable benefit (NDB) (Mann-Whitney P = 0.02)” (p. 2).
  • “Seventy-three percent of patients with high nonsynonymous burden…experienced DCB, compared with 13% of those with low mutation burden (below median) (Fisher’s exact P = 0.04)” (p. 2).
  • “median PFS 14.5 versus 3.7 months, log-rank P = 0.01; hazard ratio (HR) 0.19, 95% confidence interval (CI) 0.05 to 0.70” (p. 3, discovery cohort).
  • “HR 0.15, 95% CI 0.04 to 0.59” (p. 3, validation cohort PFS).
  • “Patients with nonsynonymous mutation burden ≥178, the cut point that combined maximal sensitivity with best specificity, had a likelihood ratio for DCB of 3.0” (p. 3).
  • “The ORR in TH tumors was 56% versus 17% in TL tumors (Fisher’s exact P = 0.03); the rate of DCB was 77% versus 22% (Fisher’s exact P = 0.004)” (p. 3).
  • “in three responders with the highest mutation burden, we identified deleterious mutations in POLD1, POLE, and MSH2” (p. 4).
  • “a POLD1 E374K mutation was identified in a never-smoker with DCB whose tumor harbored the greatest nonsynonymous mutation burden (n = 507) of all never-smokers in our series” (p. 4).
  • “Mutations in KRAS were found in 7 of 14 tumors from patients with DCB compared to 1 of 17 in the NDB group” (p. 4).
  • “There were no mutations or copy-number alterations in antigen-presentation pathway–associated genes or CD274 [encoding programmed cell death ligand-1 (PD-L1)] that were associated with response or resistance” (p. 4).
  • “This analysis revealed a CD8+ T cell response against a neoantigen resulting from a HERC1 P3278S mutation (ASNASSAAK)” (p. 5).
  • “Among those with high nonsynonymous mutation burden (>200, above median of overall cohort) and some degree of PD-L1 expression (weak/strong), the rate of DCB was 91% (10 of 11…)” (p. 6).

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