Mutational landscape of metastatic cancer revealed from prospective clinical sequencing of 10,000 patients

Authors

Ahmet Zehir

Ryma Benayed

Ronak H. Shah

Aijazuddin Syed

Sumit Middha

Hyunjae R. Kim

Preethi Srinivasan

Jianjiong Gao

Debyani Chakravarty

Sean M. Devlin

Matthew D. Hellmann

David A. Barron

Alison M. Schram

David M. Hyman

Marc Ladanyi

Nikolaus Schultz

David B. Solit

Michael F. Berger

Doi

10.1038/nm.4333

PMID: 28481359 · DOI: 10.1038/nm.4333 · Journal: Nature Medicine (2017)

TL;DR

Zehir et al. report the first-wave results of MSK-IMPACT, a prospective, CLIA-certified, hybrid-capture targeted-sequencing initiative at Memorial Sloan Kettering. Between January 2014 and May 2016, they sequenced matched tumor + patient-matched normal DNA from 10,945 advanced/metastatic tumors (10,336 patients; 62 principal tumor types, >300 detailed subtypes) using 341- or 410-gene panels. They cataloged 78,066 non-synonymous mutations, 22,989 CNAs, and 1,875 rearrangements; identified mutation signatures (MMR, POLE, UV, TMZ, smoking, APOBEC, BRCA1/2, aging) directly from targeted data; and matched 11% of the first 5,009 patients to a genomically-targeted clinical trial. The full dataset was released through cBioPortal as study msk_impact_2017.

Cohort & data

  • Patients / samples: 10,336 patients, 10,945 successfully sequenced tumors (rescue rate 75% on technical failures; final assay success rate 86%) (PMID:28481359).
  • Cancer types: 62 principal tumor types and >300 detailed OncoTree subtypes; 43% of specimens were from metastatic sites (most commonly liver, lymph node, bone) (PMID:28481359).
  • Dataset: msk_impact_2017 — released through cBioPortal at cbioportal.org/msk-impact.
  • Assay / panels: msk-impact-panel using two versions: IMPACT341 (341 genes; 2,809 tumors, 26%) and IMPACT410 (410 genes; 8,136 tumors, 74%; superset of IMPACT341). Hybrid-capture, mean depth 718X, 98% with patient-matched normal blood DNA.
  • Bioinformatics: bwa MEM alignment to hg19; ABRA realignment; GATK base-quality recalibration; union of mutect, pindel, and gatk-somatic-indel-detector for SNVs/indels; in-house caller for CNAs; delly v0.6.1 for structural variants; manual review of every alteration in integrative-genomics-viewer (PMID:28481359).
  • Orthogonal RNA fusion confirmation: msk-fusion (MSK-Solid Fusion) targeted RNA panel using archer-fusionplex Anchored Multiplex PCR over 35 rearranged genes (PMID:28481359).
  • WES comparison: 106 tumors re-captured with the Agilent Exome Kit v3 (whole-exome-seq, mean 240X) for cross-platform validation; TMB correlation R²=0.76 between MSK-IMPACT and WES (PMID:28481359).
  • Clinical annotation: Tumor-type-specific actionability tiers via oncokb; mutation signatures decomposed against Alexandrov 30-signature catalog (mutational-signatures); MSI calls cross-checked with msisensor.
  • Turnaround: Median <21 days; throughput 563 cases/month over the final 12 months of the study (PMID:28481359).
  • Consent: >85% enrolled on IRB-approved research protocol NCT01775072 with return of results.

Key findings

  • Mutation yield: 78,066 non-synonymous mutations, 22,989 CNAs, 1,875 rearrangements across 10,945 tumors; median variant allele fraction 0.21; mutation count and CNA count tended to be inversely proportional (PMID:28481359).
  • Why broad panels matter: 81% (n=63,184) of all mutations fell outside the combined target regions of common amplicon “hotspot” panels; downsampling showed ≥9% of mutations would have been missed by 150X WES, including targetable BRAF, EGFR, and MET alterations (PMID:28481359).
  • Matched-normal value: 69% of somatic mutations detected by MSK-IMPACT were not in COSMIC v78 — without matched-normal blood, these would be hard to distinguish from rare germline variants (PMID:28481359).
  • TCGA concordance & differences: Mutation frequencies in MSK-IMPACT were broadly concordant with TCGA across 16 common tumor types but several genes were significantly enriched in the metastatic cohort. TP53 was enriched in prostate cancer (29% vs 7%, >4-fold), kidney chromophobe (CHRCC), glioblastoma, and gastric cancer; AR was mutated in 18% of PRAD (vs 1% in TCGA) and ESR1 in 11% of BRCA (vs 4%) — both consistent with hormone-therapy resistance (PMID:28481359).
  • AR resistance hotspots: L702H and H875Y were each observed in 10 PRAD patients — both previously described as acquired AR-inhibitor-resistance mutations (PMID:28481359).
  • ESR1 mutations: Recurrent hotspots in both BRCA and UCEC, almost exclusively in metastatic post-hormone-therapy tumors (PMID:28481359).
  • Most-altered gene overall: TP53 — 41% of samples, altered in >10% of cases for 43/62 principal tumor types; highest rates in HGSOC (98%), esophageal adenocarcinoma (89%), and SCLC (85%) (PMID:28481359).
  • KRAS: 2nd most altered (15%); 90% of PAAD and 44% of COAD; G12 codon = 80% of all KRAS mutations and 12% of all patients (PMID:28481359).
  • Other recurrent codons: PIK3CA H1047, PIK3CA E545, and BRAF V600 each observed in >20 principal tumor types (PMID:28481359).
  • Recurrent fusions: TMPRSS2-ERG (n=151) in prostate cancer, EGFRvIII (exon 2–7 deletion; n=65), EML4-ALK (n=38), EWSR1-FLI1 (n=25); cryptic TMPRSS2 rearrangements in 23 prostate cancers consistent with chromoplexy (PMID:28481359).
  • TERT promoter: Largest pan-cancer TERT promoter analysis to date — G>A at −124/−146 dominated (96.3%), but 10 additional recurrent sites identified including position −138 (21 patients, presumptive ETS binding site). All novel recurrent TERT mutations within 100 bp of TSS. Highest rates: BLCA 70%, glioma 67%, thyroid 60%, melanoma 49% (predominantly cutaneous). Trend toward shorter overall survival in TERT-promoter-mutant cases (PMID:28481359).
  • Kinase fusions: 35% (n=268) of fusions involved kinase genes; ALK/RET/ROS1 fusions found in 11 additional tumor types beyond LUAD; 51 kinase fusions involved novel partners (mostly in tumors lacking other driver mutations) (PMID:28481359).
  • BRAF rearrangements: 33 BRAF fusions across 11 principal tumor types involving 18 distinct partner genes (10 novel), including a novel recurrent CDK5RAP2-BRAF fusion in two melanomas. 7 cases harbored intragenic multi-exon BRAF deletions predicted to be in-frame and analogous to splicing isoforms previously reported in V600E-mutant melanomas with acquired BRAF-inhibitor resistance — but only 3/7 had prior BRAF-inhibitor exposure, suggesting a de novo driver role for the others (PMID:28481359).
  • Mutation signatures from targeted data: Among 994 cases (9%) with TMB > 13.8 mut/Mb, signatures were assigned by KL-divergence decomposition against the Alexandrov 30-signature catalog, with a “dominant” call at >40%. Eight signatures were tracked: aging (1), APOBEC (2,13), smoking (4), BRCA1/2 (3), MMR (6,15,20,26), UV (7), POLE (10), and TMZ (11). UV/TMZ/smoking signatures predominated in melanoma, glioma, and lung respectively; POLE and MMR predominated in colorectal and endometrial (PMID:28481359).
  • MSI detection: 102 patients across 11 tumor types had a dominant MMR signature and MSI by msisensor; 45% had not been previously tested for MMR deficiency. Indel:SNV ratio in MMR samples was 0.46 vs 0.06 in POLE/other (p<0.001). Responses to immune checkpoint blockade observed in MSI COAD, UCEC, gastric, PRAD, and BLCA, including a 55-year-old prostate cancer patient with an unanticipated MMR signature who responded to anti-PD-L1 therapy (PMID:28481359).
  • Actionability: 36.7% of patients (n=3,792) harbored ≥1 oncokb-actionable alteration. Top tumor types by proportion: GIST 76%, thyroid 60%, BRCA 57%, melanoma 56% (PMID:28481359).
  • Trial enrollment: Of 5,009 patients sequenced ≥1 year before analysis, 1,894 (38%) enrolled on any clinical trial, 811 (16%) on a trial with a targeted agent, and 527 (11%) were genomically matched to ≥1 of 197 trials targeting an alteration in their tumor. 72% of matches occurred after MSK-IMPACT reporting (PMID:28481359).
  • BRAF V600 in non-melanoma: Detected in 211 non-melanoma patients; 75/211 (36%) received BRAF-targeted therapy on- or off-trial. Radiographic clinical-benefit rate was identical (71%) between melanoma and non-melanoma BRAF V600 patients (PMID:28481359).

Genes & alterations

  • TP53 — most frequently altered gene overall (41% of samples; >10% in 43/62 principal tumor types). Largely truncating/splice-disrupting and inactivating. Significantly enriched vs TCGA in PRAD, CHRCC, glioblastoma, and gastric cancer.
  • KRAS — 2nd most altered (15%); G12 codon dominates (80% of KRAS mutations). 90% in PAAD, 44% in COAD. Authors call out limited clinical actionability with then-available drugs.
  • PIK3CA — H1047 and E545 each observed in >20 principal tumor types (cross-lineage selection).
  • BRAF — V600 in >20 tumor types; non-V600 alterations include 33 fusions across 11 tumor types (18 partners; 10 novel), a novel recurrent CDK5RAP2-BRAF fusion in melanoma, and 7 intragenic multi-exon deletions phenocopying acquired-resistance splice isoforms (some without prior BRAF-inhibitor exposure → de novo drivers).
  • EGFR — tumor-type-specific localization: extracellular N-terminal in glioma, kinase domain in lung cancer. EGFRvIII (exon 2–7 deletion) detected in 65 cases.
  • TERT — promoter mutations at canonical −124/−146 hotspots (96.3% of TERT promoter events) plus 10 novel recurrent positions including −138 (n=21). Highest in BLCA, glioma, thyroid, melanoma. Trend toward shorter overall survival.
  • AR — 18% in PRAD (vs 1% TCGA); recurrent acquired-resistance hotspots L702H and H875Y (10 patients each).
  • ESR1 — recurrent hotspots in BRCA and UCEC, almost exclusively in post-hormone-therapy metastases.
  • VHL, APC, IDH1 — examples of lineage-restricted drivers (renal-clear, colorectal, glioma) contrasted with pan-cancer drivers like TP53/PIK3CA.
  • ALK, RET, ROS1 — kinase fusions enriched in LUAD but also detected across 11 additional tumor types.
  • MET — included in the targetable-alteration tally that broad-panel sequencing recovered vs amplicon panels and shallow WES.
  • TMPRSS2-ERG — most common rearrangement overall (n=151) in PRAD; cryptic TMPRSS2 rearrangements in 23 PRAD cases consistent with chromoplexy.
  • EML4-ALK — n=38, predominantly LUAD.
  • EWSR1-FLI1 — n=25 in ES.
  • DNAJB1-PRKACA — pathognomonic for fibrolamellar hepatocellular carcinoma (FLC).
  • POLE — loss-of-function somatic mutations underlie POLE-signature hypermutation, predominantly in colorectal and endometrial.

Clinical implications

  • Direct trial matching: 11% (527/5,009) of patients with ≥12-month follow-up were enrolled on a genomically-matched targeted-therapy trial; 72% of matches were enabled by MSK-IMPACT reporting itself rather than prior testing — evidence that broad prospective profiling expands trial eligibility above the prior NGS-driven baseline (PMID:28481359).
  • Off-label / cross-tumor BRAF targeting: 71% radiographic clinical-benefit rate to BRAF-directed therapy among 75 non-melanoma BRAF V600 patients — same as the melanoma rate, supporting basket-trial designs and tumor-type-agnostic BRAF inhibition where mutation status is the criterion.
  • MSI as immunotherapy biomarker: Comprehensive panel-based mutation-signature + msisensor calling identified 102 MSI tumors across 11 lineages; 45% had not been previously tested for MMR deficiency, including a PRAD case who responded to anti-PD-L1 — argues that broad genomic profiling expands the immunotherapy-eligible pool beyond conventional MSI testing’s current footprint (PMID:28481359).
  • Actionable-alteration prevalence as a planning metric: 36.7% of patients harbored an oncokb-actionable alteration; tumor-type-by-tumor-type proportions (GIST 76%, thyroid 60%, BRCA 57%, melanoma 56%) inform expected match rates.
  • Resistance biomarkers picked up by routine sequencing: AR L702H/H875Y in PRAD and ESR1 hotspots in BRCA/UCEC were enriched in the metastatic cohort vs TCGA — clinically relevant for hormone-therapy management.
  • Novel actionable-resistance class: BRAF intragenic in-frame multi-exon deletions, predicted to drive RAS-independent dimerization, may benefit from RAF-dimer inhibitors — applicable both as acquired-resistance lesions in V600E-mutant disease and as de novo drivers in patients without prior BRAF-inhibitor exposure (PMID:28481359).
  • Germline by-product: Patient-matched normal sequencing enables IRB-approved return of inherited pathogenic variants to consenting patients, with implications for PARP-inhibitor response and family-member cancer susceptibility (PMID:28481359).
  • Data sharing: Full dataset deposited to cBioPortal as msk_impact_2017 at cbioportal.org/msk-impact to enable downstream biomarker discovery.

Limitations & open questions

  • Treatment-decision attribution is observational. Clinical-utility assessment relied on retrospective trial-enrollment counts, not randomized comparison; many patients may have been too debilitated to enroll, and others receiving conventional therapy may later benefit from targeted agents (PMID:28481359).
  • Incomplete follow-up. Only patients tested ≥12 months before analysis (5,009/10,336) were considered for enrollment metrics; the long-term match rate could be higher with maturing follow-up.
  • MSI prevalence underestimated. Signature-based MSI calling was restricted to tumors with TMB > 13.8 mut/Mb (the top 9%); lower-burden MSI tumors may be missed by this approach (PMID:28481359).
  • TERT-promoter survival association is preliminary. A consistent trend toward shorter survival was observed but the prognostic significance “remains incompletely understood” pending longer longitudinal data.
  • Targeted-panel signature inference is bounded. Mutation-signature decomposition required ≥13.8 mut/Mb and 994 samples; while WES correlation was R²=0.76, signature precision in low-TMB tumors is not addressed.
  • KRAS handling is conservative. Authors classify KRAS as non-actionable under their stricter OncoKB-tiered criteria, acknowledging that this will change as KRAS-targeting agents mature (PMID:28481359) — predates clinical approval of KRAS G12C inhibitors.
  • Single-institution cohort. MSKCC referral patterns, pathology workflow, and trial portfolio may not generalize to community settings; broader inter-institution data sharing is explicitly called for.
  • Cohort skews to advanced disease. By design, the cohort is metastatic and often pretreated, which makes comparison to TCGA primary-tumor cohorts informative for treatment-emergent biology but limits inferences about untreated disease.

Citations from this paper used in the wiki

  • “Altogether, we successfully sequenced 10,945 tumor samples from 10,336 patients (91%).” (Results, Description of the Sequencing Cohort.)
  • “Two panels were used throughout this study, encompassing 341 genes (2,809 tumors, 26%) and 410 genes (8,136 tumors, 74%), with all 341 genes included in the latter expanded panel.” (Results.)
  • “81% (n=63,184) of all mutations fell outside the combined target regions of commercially available amplicon-based ‘hotspot’ panels … at least 9% of all mutations would have been missed by WES to a mean target depth of 150X, including therapeutically targetable alterations in BRAF, EGFR, and MET.” (Results.)
  • “69% of somatic mutations detected by MSK-IMPACT were not previously reported in the COSMIC database (v78)…” (Results.)
  • “In prostate cancer alone, the frequency of TP53 mutations was >4-fold greater in MSK-IMPACT than TCGA (29% versus 7%)…” (Landscape of Somatically Altered Genes.)
  • “AR (androgen receptor) in prostate cancer (18% versus 1%) and ESR1 (estrogen receptor) in breast cancer (11% versus 4%)…” (Landscape.)
  • “The most common AR mutations in our cohort were L702H and H875Y (10 patients each)…” (Landscape.)
  • “TP53 mutations occurred most often in high-grade serous ovarian cancer (98%), esophageal adenocarcinoma (89%), and small cell lung cancer (85%)…” (Landscape.)
  • “KRAS mutations were most prevalent in pancreatic adenocarcinoma (90%) and colon adenocarcinoma (44%). KRAS also harbored the most frequently altered codon among tumors sequenced (G12), accounting for 80% of all KRAS mutations and 12% of all patients.” (Landscape.)
  • “The most commonly observed rearrangements were TMPRSS2-ERG (n=151), EGFRvIII (deletion of exons 2–7; n=65), EML4-ALK (n=38), and EWSR1-FLI1 (n=25).” (Landscape.)
  • “TERT promoter mutations were most commonly observed in bladder cancer (70%), glioma (67%), thyroid cancer (60%), and melanoma (49%)…” (TERT Promoter Mutations.)
  • “Altogether we detected 33 BRAF fusions across 11 principal tumor types involving 18 distinct partner genes (10 novel), including a novel recurrent fusion gene, CDK5RAP2-BRAF…” (Kinase Fusions and Rearrangements.)
  • “Altogether we identified 102 patients across 11 tumor types harboring both a dominant MMR signature and MSI classification by MSIsensor (Fig. 5e), 45% of whom were not previously tested for MMR deficiency.” (Mutation Signatures and Somatic Hypermutation.)
  • “Altogether, 36.7% of patients (n=3,792) harbored at least one actionable alteration… The tumor types with the highest proportion of actionable mutations were gastrointestinal stromal tumor (76%), thyroid cancer (60%), breast cancer (57%), and melanoma (56%).” (Clinical Actionability and Utility.)
  • “527 (11%) patients were enrolled on at least one of 197 trials involving molecularly targeted agents at our institution, based on a target aberration in their tumor.” (Clinical Actionability.)
  • “BRAF V600 mutations were detected in 211 patients with non-melanoma histologies, of which 75/211 (36%) received BRAF targeted therapy on- or off-trial. While the proportion of non-melanoma patients receiving BRAF inhibitors was smaller, the rate of clinical benefit assessed by radiographic measurement among non-melanoma and melanoma patients was identical (71%)…” (Clinical Actionability.)

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