Molecular Determinants of Response to Anti–Programmed Cell Death (PD)-1 and Anti–Programmed Death-Ligand 1 (PD-L1) Blockade in Patients With Non–Small-Cell Lung Cancer Profiled With Targeted Next-Generation Sequencing

Authors

Hira Rizvi

Francisco Sanchez-Vega

Konnor La

Walid Chatila

Philip Jonsson

Darragh Halpenny

Andrew Plodkowski

Niamh Long

Jennifer L. Sauter

Natasha Rekhtman

Travis Hollmann

Kurt A. Schalper

Justin F. Gainor

Ronglai Shen

Ai Ni

Kathryn C. Arbour

Taha Merghoub

Jedd Wolchok

Alexandra Snyder

Jamie E. Chaft

Mark G. Kris

Charles M. Rudin

Nicholas D. Socci

Michael F. Berger

Barry S. Taylor

Ahmet Zehir

David B. Solit

Maria E. Arcila

Marc Ladanyi

Gregory J. Riely

Nikolaus Schultz

Matthew D. Hellmann

Doi

10.1200/JCO.2017.75.3384

PMID: 29337640 · DOI: 10.1200/JCO.2017.75.3384 · Journal: Journal of Clinical Oncology (2018)

TL;DR

Rizvi et al. retrospectively analyzed 240 patients with advanced NSCLC treated with anti–PD-(L)1 therapy at MSKCC and profiled by the targeted MSK-IMPACT panel. Tumor mutation burden (TMB) measured on the targeted panel correlated tightly with whole-exome estimates (Spearman r = 0.86, P < .001) and was significantly higher in patients with durable clinical benefit (DCB) than in those with no durable benefit (NDB). High TMB and PD-L1 expression were independent predictors of benefit; combining them further enriched for response (50% DCB when both high, vs 18% when both low). Mutations in EGFR and STK11 associated with lack of benefit, and high fraction of copy-number–altered genome (FGA) was enriched in non-responders. The work established that targeted NGS panels routinely deployed in the clinic can reliably estimate TMB and serve as a practical immunotherapy biomarker.

Cohort & data

  • 240 patients with advanced NSCLC treated with anti–PD-1 or anti–PD-L1 therapy (alone or with anti–CTLA-4) at Memorial Sloan Kettering between April 2011 and January 2017, all profiled by MSK-IMPACT targeted NGS (PMID:29337640).
  • Histology: 78% adenocarcinoma (LUAD), 14% squamous (LUSC), 8% other.
  • Treatment: 86% PD-(L)1 monotherapy, 14% PD-(L)1 + CTLA-4 combination (nivolumab, pembrolizumab, atezolizumab, durvalumab, ipilimumab).
  • Panel versions: IMPACT341 (56 patients), IMPACT410 (164), IMPACT468 (20). Mean coverage 744×.
  • Validation subset: 49 patients had paired whole-exome sequencing.
  • PD-L1 IHC available for 84 patients (22C3, 28-8, E1L3N antibodies).
  • Comparison cohort: 1,836 non–ICI-treated NSCLC patients sequenced on MSK-IMPACT 2014–2017 (dataset: nsclc_pd1_msk_2018).
  • Endpoints: RECIST 1.1; durable clinical benefit (DCB) = CR/PR/SD lasting > 6 months; no durable benefit (NDB) = PD or SD ≤ 6 months. 49 (20%) had CR/PR; 69 (29%) had DCB; 158 (66%) NDB; 13 (5%) not evaluable.

Key findings

  • Targeted NGS reliably quantifies TMB. TMB by MSK-IMPACT correlated with TMB by WES at Spearman r = 0.86 (P < .001, n = 49) (PMID:29337640).
  • TMB associates with benefit. Median TMB 8.5 SNVs/Mb in DCB vs 6.6 in NDB (P = .006); 8.5 vs 6.6 vs 6.6 across CR/PR vs SD vs PD (P = .049 / .015).
  • Dose–response with TMB. DCB rate above the 50th percentile was 38.6% vs 25.1% below (P = .009). PFS HR for above- vs below-median TMB = 1.38, P = .024. Odds ratios of DCB rose stepwise across TMB percentiles: 25th OR 1.75, 50th OR 2.02, 75th OR 2.06, 90th OR 3.24.
  • TMB is predictive, not prognostic. In the non–ICI-treated comparison cohort, higher TMB was associated with worse survival, ruling out a generic prognostic effect.
  • FGA is highest in non-responders. Fraction of copy-number–altered genome was significantly higher in NDB than in non-ICI NSCLC (median 0.16 vs 0.11; P = .007). FGA and TMB were modestly positively correlated despite their opposite associations with response.
  • TMB and PD-L1 are independent. No correlation between TMB and PD-L1% (Spearman r = 0.19, P = .08). AUC for predicting DCB: TMB 0.601, PD-L1 0.646. A composite of TMB-high + PD-L1 ≥ 1% gave 50% DCB rate, vs 18% when both were low (n = 22).
  • EGFR mutations underrepresented in DCB (7% of EGFR-mutant patients had DCB; FDR-adjusted P = .013 vs non-ICI cohort).
  • STK11 mutations enriched in NDB (FDR-adjusted P = .007 vs non-ICI cohort).
  • B2M / JAK pathway hits rare in this cohort. Likely-deleterious B2M mutation: 1 patient (S40* in trans with Q28L, with loss of B2M expression by IHC) — paradoxically had ongoing PR at 8.9 months. Homozygous deleterious JAK2 mutation: 1 patient with PD.
  • MDM2/MDM4 amplification (n = 8) did not show distinct hyperprogression in this series (HR 1.4, P = .44), in contrast to prior reports.

Genes & alterations

  • EGFR — Activating mutations in 17 patients (7%). Significantly underrepresented in the DCB group; only 7% of EGFR-mutant patients achieved DCB. Authors attribute this to the link between EGFR mutations, never-smoker status, and consequent low TMB (PMID:29337640).
  • STK11 — Significantly enriched in NDB vs non-ICI NSCLC (FDR-adjusted P = .007). Consistent with STK11/LKB1-loss murine and human reports of low tumor inflammation.
  • KRAS — Mutated in 83 patients; DCB rate (36%) similar to overall cohort. KRAS mutation alone did not stratify response.
  • TP53 — Altered in 67% of DCB and 54% of NDB tumors (descriptive, no enrichment claim).
  • B2M — Truncating mutations rare; one patient with biallelic deleterious B2M (S40* + Q28L) and confirmed loss of B2M IHC nonetheless had a PR ongoing at 8.9 months and TMB of 48 SNVs/Mb.
  • JAK1 / JAK2 — Few events overall; one patient with homozygous loss-of-function JAK2 splice mutation + LOH had primary resistance (PD), consistent with interferon-gamma signaling defects described in melanoma.
  • MDM2 / MDM4 — Amplifications identified in 8 patients; PFS not significantly different from the overall cohort (HR 1.4, P = .44). Hyperprogression signal previously reported elsewhere was not reproduced here.
  • Other actionable drivers reported with low frequency: ALK (1%), BRAF (2%), ROS1 (3%), RET (1%), MET (3%), ERBB2 — too few events for response analysis.
  • Antigen-presentation / immunology genes profiled in the OncoPrint: HLA-A, POLE, JAK3, CD274, PTEN, ATR.

Clinical implications

  • Targeted NGS panels (≥ ~1 Mb of coding sequence) can replace WES for TMB estimation in routine practice — directly supports clinical adoption of MSK-IMPACT and similar assays as immunotherapy biomarkers (PMID:29337640).
  • TMB and PD-L1 should be combined, not substituted. Their AUCs for DCB are comparable but the variables are independent; a composite (TMB > median + PD-L1 ≥ 1%) selects a subgroup with 50% DCB.
  • EGFR-mutant and STK11-mutant NSCLC are enriched in non-responders to anti–PD-(L)1 therapy — supporting use of these alterations as negative selectors when considering single-agent ICI.
  • No universal TMB cut point is recommended; appropriate thresholds are panel- and assay-specific. The authors caution against transferring numerical cutoffs across platforms.
  • Hyperprogression with MDM2/MDM4 amplification was not reproduced, leaving the prior signal an open question rather than a confirmed contraindication.

Limitations & open questions

  • Moderate sample size (n = 240) limits power, particularly for subgroup and multivariable analyses.
  • Retrospective clinical outcome adjudication for the non-trial subset, though authors argue the mixed real-world + trial composition improves generalizability.
  • Single-institution, single-panel design; the absolute TMB cutpoints derived here are not portable across panels of different sizes or informatics pipelines (caution noted for panels < 0.5 Mb).
  • Only 85% of MSK-IMPACT samples were collected pre-immunotherapy; the analysis was not designed to study acquired resistance or selection pressure on B2M/JAK genes.
  • PD-L1 testing used three different antibodies (22C3, 28-8, E1L3N) — possible inter-assay variability despite reported concordance.
  • Targeted panels lack the immunogenomic genes (HLA, antigen-processing machinery beyond a handful) that may be biologically central; authors call for panel expansion and continued WES/WGS discovery work.
  • Mechanism of MDM2/MDM4 hyperprogression and the apparent discordance with prior reports remains unresolved.

Citations from this paper used in the wiki

  • “Estimates of TMB by targeted NGS correlated well with WES (r = 0.86; P < .001). TMB was greater in patients with DCB than with NDB (P = .006).” — Abstract.
  • “DCB was more common, and progression-free survival was longer in patients at increasing thresholds above versus below the 50th percentile of TMB (38.6% v 25.1%; P < .001; hazard ratio, 1.38; P = .024).” — Abstract.
  • “Variants in EGFR and STK11 associated with a lack of benefit. TMB and PD-L1 expression were independent variables, and a composite of TMB plus PD-L1 further enriched for benefit to ICIs.” — Abstract / Results.
  • “TMB in patients with DCB was similar to those with CR/PR (P = .85) and greater in those with non-ICI NSCLC (P < .001).” — Fig 1B legend.
  • “STK11 was significantly enriched in the NDB group compared with the non-ICI NSCLC group (FDR-adjusted P = .007).” — Results, Gene Alterations.
  • “MDM2/MDM4 amplifications were identified in eight patients … PFS was not substantially different in this group compared with the overall patient cohort (HR, 1.4; P = .44).” — Results.
  • “Patients with high TMB (greater than the group median) and PD-L1 positivity (≥ 1% expression) had a 50% rate of DCB, whereas the presence of only one or neither variable was associated with a lower rate of DCB.” — Results, PD-L1 + TMB composite.

This page was processed by crosslinker on 2026-05-15.