HER kinase inhibition in patients with HER2- and HER3-mutant cancers

Authors

David M. Hyman

Sarina A. Piha-Paul

Helen Won

Jordi Rodon

Cristina Saura

Geoffrey I. Shapiro

Dejan Juric

David I. Quinn

Victor Moreno

Bernard Doger

Ingrid A. Mayer

Valentina Boni

Emiliano Calvo

Sherene Loi

Albert C. Lockhart

Funda Meric-Bernstam

José Baselga

David B. Solit

Doi

10.1038/nature25475

PMID: 29420467 · DOI: 10.1038/nature25475 · Journal: Nature (2018)

TL;DR

SUMMIT (NCT01953926) is a multi-histology, genomically selected basket trial of the irreversible pan-HER tyrosine kinase inhibitor neratinib in 141 patients with advanced solid tumours harbouring somatic ERBB2 (HER2; n=125) or ERBB3 (HER3; n=16) mutations across 21 cancer types. Response to neratinib varied as a joint function of tumour lineage and mutant allele: the strongest activity was in breast cancer (ORR8 32%, 95% CI 15–54%), with additional responses in cervical, biliary and salivary cancers, while bladder and colorectal cohorts showed lineage-based resistance and no ERBB3-mutant patient responded. Co-mutations in TP53 and ERBB3 and cell-cycle pathway alterations were enriched in non-responders.

Cohort & data

  • 141 patients (125 ERBB2-mutant, 16 ERBB3-mutant) across 21 cancer types; most common were breast, lung, bladder and colorectal (61% of treated patients).
  • Pre-specified disease cohorts: endometrial, gastroesophageal, ovarian, colorectal, bladder/urinary tract, plus a “Solid tumour (NOS)” cohort. Breast, cervical, biliary and lung accrued enough patients in NOS to permit independent efficacy analysis. All ERBB3-mutant patients enrolled in a single histology-agnostic cohort.
  • Treatment: neratinib 240 mg PO daily continuous, with mandatory loperamide prophylaxis during cycle 1.
  • Primary endpoint: ORR at week 8 (ORR8) by RECIST v1.1; PET Response Criteria (PRC, modified PERCIST v1.0 with FDG PET/CT) used for patients not RECIST-evaluable.
  • Local mutation testing accepted from 30 unique sequencing assays in 25 laboratories; 85% (120/141) by NGS, of which 81% had full ERBB2/ERBB3 exon coverage.
  • Central sequencing on archival FFPE tumour (n=91) and plasma cfDNA (n=15) with matched germline (n=102) using the MSK-IMPACT panel — either IMPACT341 (n=18) or IMPACT410 (n=88) — at mean 738× coverage (range 253–1383). Custom single-gene ERBB2 capture cfDNA assay used in a subset.
  • Copy number, purity and ploidy estimated with FACETS v0.3.9; clonality estimated with the ABSOLUTE v1.0.6 algorithm; MSI assessed with MSIsensor; pathway-level oncogenicity calls from OncoKB (Sept 2017); mutational signatures decomposed into 30 constituent processes.
  • Data interim cut: 10 Mar 2017 (enrollment through 16 Dec 2016). All clinical and sequencing data deposited in cBioPortal as study summit_2018.

Key findings

  • Tumour-type efficacy (PMID:29420467): Breast cancer met its pre-specified efficacy endpoint with ORR8 32% (8/25, 95% CI 15–54%); responses observed across extracellular and kinase domain missense mutations and kinase domain insertions, in both ER+ (30%, 6/20) and ER− (40%, 2/5) tumours, all centrally HER2-non-amplified. Biliary, cervical and salivary gland cancers also produced responses; bladder (n=16) and colorectal (n=12) cohorts produced no RECIST responses, consistent with lineage-based resistance to single-agent pan-HER kinase inhibition.
  • Lung cancer outcome (PMID:29420467): Of 26 non-small-cell lung patients (predominantly LUAD with HER2 exon 20 insertions), only 1 RECIST response — in a patient with kinase domain missense L755S. Median PFS in recurrent lung cancer was 5.5 months with 6 patients on therapy >1 year, comparing favourably to second-line chemotherapy and checkpoint inhibitors despite the low ORR. HER2 exon 20 insertions are paralogous to EGFR exon 20 insertions resistant to first/second-generation EGFR TKIs.
  • HER3-mutant cohort (PMID:29420467): No RECIST responses in 16 patients with ERBB3-mutant tumours despite preclinical data suggesting some HER3 mutations are oncogenic drivers via wild-type HER2.
  • Allele-specific responses (PMID:29420467): Responses observed in HER2 S310, L755, V777, G778_P780dup, Y772_A775dup mutants. Among 42 patients with kinase domain hotspot missense mutations, responders included L755S (n=4), V777L (n=4), L869R (n=1) across breast, biliary, lung, salivary. In 30 patients with HER2 S310 (extracellular hotspot), responses occurred in breast, cervical and biliary but not bladder, where S310 predominates. Glycine preservation at position 770 in exon 20 insertions did NOT predict response, contrary to preclinical predictions.
  • Hotspot vs non-hotspot (PMID:29420467): 87% (109/125) of HER2 mutations and 75% (12/16) of HER3 mutations occurred at known hotspots. Among 15 patients with non-hotspot HER2 mutations, only 1 responded — a breast cancer patient with the previously unobserved insertion/substitution L755_E757delinsS at a recurrent insertion locus.
  • HER2 amplification co-occurrence (PMID:29420467): 17% (15/86) of patients with broader profiling had concurrent ERBB2 mutation + amplification; 86% (12/14 evaluable) of amplifications targeted the mutant allele locus. Concurrent amplification did NOT correlate with clinical benefit (p=0.50), suggesting amplification adds no extra sensitivity beyond mutation alone for irreversible HER kinase inhibitors.
  • Clonality (PMID:29420467): 95% (70/74 with adequate material) of HER2 mutations were clonal (CCF >0.85 by ABSOLUTE integrated with FACETS segmentation). None of 4 patients with subclonal HER2 mutations achieved clinical benefit.
  • Tumour mutational burden (PMID:29420467): Using a ≥13.8 nonsyn mut/Mb cutoff, 20% (17/86) had high TMB. High TMB was 24% (16/66) in patients without clinical benefit vs 5% (1/20) in those with benefit (p=0.10, n.s.).
  • Co-mutation modifiers (PMID:29420467): Coincident TP53 (nominal p=0.018) and ERBB3 (nominal p=0.064) mutations were enriched in non-responders; no patient with concurrent ERBB2+ERBB3 mutation achieved benefit (n=8 unique HER2 + n=9 unique HER3 alleles, across breast n=3, bladder n=2, gastroesophageal n=2, colorectal n=1, pancreatic n=1).
  • Pathway modifiers (PMID:29420467): Cell cycle checkpoint aberrations associated with lack of clinical benefit (p=0.043, but driven by TP53; p=0.769 after TP53 removal); RTK/RAS/RAF activation trended worse (p=0.060); PIK3CA/AKT/mTOR activation did NOT adversely affect benefit (p=0.753) — diverging from the established negative-predictor role of PI3K pathway alterations in HER2-amplified breast cancer.
  • Central confirmation (PMID:29420467): Concordance between local and central HER2 mutation calls was 95% (69/73) by tissue and/or plasma cfDNA. One locally reported HER2 V773M was reclassified as germline polymorphism by central testing (patient had renal cell carcinoma, PD at first scan). Discordant cases had median PFS of 43 days (range 5–58); none responded. HER3 concordance was 75% (6/8).
  • Safety (PMID:29420467): Grade 3 diarrhoea rate 22% with mandatory anti-diarrhoeal prophylaxis; median onset day 10, median episode duration 2 days; only 2.8% permanent discontinuation due to diarrhoea.

Genes & alterations

  • ERBB2 (HER2): 125 patients with 31 unique somatic mutations. Domain distribution: kinase 66%, extracellular 26%, transmembrane/juxtamembrane 8%. Class distribution: missense 74%, in-frame insertions 22% (kinase domain only), 2 indels, 1 in-frame kinase-retaining fusion (GRB7-ERBB2). Most frequent alleles: S310, L755, Y772_A775dup, V777. V777L (13.6% vs 5.3%) and Y772_A755dup (12.0% vs 2.7%) were over-represented vs TCGA/ICGC. Responders identified across S310, L755S, V777L, G778_P780dup, Y772_A775dup, L869R, and the novel insertion/substitution L755_E757delinsS. HER2 mutation was clonal in 95% of evaluable patients.
  • ERBB3 (HER3): 16 patients with 11 unique missense mutations clustered in the extracellular furin-like and receptor domains. No RECIST responses to neratinib monotherapy. Frequently co-occurred with ERBB2 mutations in non-responders.
  • GRB7: One patient harboured a GRB7-ERBB2 in-frame kinase-retaining fusion as the qualifying ERBB2 alteration.
  • TP53: Co-mutation enriched in patients without clinical benefit (nominal p=0.018); drives the cell-cycle pathway association.
  • PIK3CA: Activation of PI3K/AKT/mTOR pathway did NOT adversely affect benefit (p=0.753), in contrast to its established negative-predictor role in HER2-amplified disease.
  • EGFR: Cited as the paralogous receptor whose exon 20 insertions resist first/second-generation TKIs, motivating the lung cancer findings; not itself a study qualifier.
  • BRAF, KRAS, ALK, ROS1: Cited as comparators (other oncogenic drivers with approved targeted therapies achieving higher response rates than neratinib in HER2-mutant disease).

Clinical implications

  • Tumour-agnostic targeting is incomplete (PMID:29420467): Pan-HER kinase inhibition with neratinib is NOT tumour-agnostic for HER2 mutations. Sensitivity is jointly determined by tissue lineage and specific mutant allele — bladder and colorectal HER2-mutant tumours show intrinsic lineage-based resistance regardless of allele.
  • Practice signal (PMID:29420467): HER2-mutant, non-amplified breast cancer met its pre-specified efficacy bar with neratinib monotherapy (ORR8 32%) across ER subgroups and across mutation domains. This supports HER2 mutation as an actionable biomarker in BRCA independent of HER2 amplification status.
  • HER2 exon 20 insertion lung cancer is poorly served by single-agent neratinib (PMID:29420467): ORR was minimal in HER2-exon-20-insertion-driven LUAD, but median PFS of 5.5 months with a long-tail of >1-year responders suggests neratinib still has activity worth pursuing in combination regimens.
  • HER3 mutations are not actionable with single-agent pan-HER TKI (PMID:29420467): Despite preclinical oncogenic data, no HER3-mutant patient responded — argues against single-agent HER kinase inhibition as a strategy in HER3-mutant disease.
  • Co-mutation context matters (PMID:29420467): Concurrent TP53 or ERBB3 mutation and cell-cycle pathway lesions associate with lack of benefit; these may represent stratification or combination-strategy biomarkers in future trials. SUMMIT was amended to evaluate neratinib combinations in multiple HER2-mutant tumour types.
  • Local-to-central NGS concordance is high (PMID:29420467): 95–96% concordance across 30 unique assays from 25 labs supports basket-trial enrollment based on local NGS testing without mandatory central confirmation.
  • Permissive enrolment validates novel variants (PMID:29420467): Patients with previously uncharacterized HER2 variants (V697, D769N, L869R) responded, providing first clinical proof-of-concept of gain-of-function before formal preclinical characterization. Supports allele prioritization by recurrence and paralogy as whole-exome/genome testing scales.

Limitations & open questions

  • Each tumour-specific cohort is small; subgroup p-values are nominal and several pathway associations did not survive Benjamini-Hochberg correction.
  • Patients with prior HER kinase inhibitor exposure and unstable brain metastases were excluded — generalizability to those populations unknown.
  • The interim data cut (10 Mar 2017) leaves several cohorts (cervical, biliary, salivary) without completed efficacy analyses.
  • Why does the PI3K/AKT/mTOR pathway, a strong negative predictor in HER2-amplified breast cancer, not predict resistance in HER2-mutant disease? Authors call for disease-specific follow-up.
  • Mechanism of lineage-based resistance in BLCA and COAD HER2-mutant tumours remains unexplained — preclinical work suggests combination HER-targeted therapy may overcome it in colorectal.
  • Master/umbrella protocols enrolling 30–40 patients per arm may be underpowered to discover the kind of lineage × allele interactions SUMMIT (n=141) revealed.
  • Whether HER3 mutations are oncogenic drivers at all — or merely co-mutations of HER2 — is open; the failure of single-agent neratinib in HER3-mutant disease does not resolve this.

Citations from this paper used in the wiki

  • “Efficacy in HER2-mutant cancers varied as a function of both tumour type and mutant allele to a degree not predicted by preclinical models, with the greatest activity seen in breast, cervical and biliary cancers and with tumours harbouring kinase domain missense mutations.” (Summary)
  • “In total, 141 patients (125 with HER2-mutant tumours, 16 with HER3-mutant tumours) received neratinib. These patients were diagnosed with one of 21 unique cancer types…” (Results, Patient and mutation characteristics)
  • “Neratinib exhibited the greatest degree of activity in patients with breast cancer (n=25 total, objective response rate at week 8 [ORR8] 32%, 95% confidence interval [CI] 15–54%)” (Results, Treatment outcomes)
  • “Despite preclinical data suggesting that HER3 mutations can be oncogenic drivers, no responses to neratinib were observed in patients with HER3-mutant tumours.” (Results, Treatment outcomes)
  • “ERBB2 amplification did not correlate with outcome (p=0.50)… in the presence of ERBB2 mutations, amplification may not confer additional sensitivity to irreversible HER kinase inhibitors.” (Results, Genomic modifiers of response)
  • “In patients with HER2-mutant disease, coincident mutations in TP53 and HER3 were enriched in patients with no clinical benefit (nominal p=0.018 and p=0.064, respectively)” (Results, Genomic modifiers of response)
  • “All datasets generated during and/or analysed during the current study, including patient-level clinical data as well as all sequencing data have been deposited and are publically available in the cBioPortal for Cancer Genomics under the accession code ‘SUMMIT, Nature, 2018’ (http://www.cbioportal.org/study?id=summit_2018).” (Methods, Data availability)
  • “concordance on retrospective central review was extremely high (96%)… 30 unique sequencing assays performed in 25 different laboratories.” (Discussion)

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