Rare mutation-dominant compound EGFR-positive NSCLC is associated with enriched kinase domain-resided variants of uncertain significance and poor clinical outcomes

Abstract

Background: Compound epidermal growth factor receptor (EGFR) mutations are less responsive to tyrosine kinase inhibitors (TKIs) than single EGFR mutations in non-small cell lung cancer (NSCLC). However, the detailed clinical characteristics and prognosis of various compound EGFR mutations remain to be elucidated. Methods: We retrospectively studied the next-generation sequencing (NGS) data of treatment-naïve tumors from 1025 NSCLC patients with compound EGFR mutations, which were sub-categorized into different combinations of common mutations (19-Del and EGFR exon 21 p.L858R), rare mutations, and variants of uncertain significance (VUSs). Prognosis and drug resistance to first-line TKIs were analyzed in 174 and 95 patients, respectively. Results: Compound EGFR mutations were enriched with EGFR exon 21 p.L858R and rare mutations, but not 19-Del (P < 0.001). The common + rare and rare + rare subtypes had fewer concurrent mutations in the PI3K pathway (P = 0.032), while the rare + rare and common + VUSs subtypes showed increased association with smoking- and temozolomide-related mutational signatures, respectively (P < 0.001). The rare mutation-dominant subtypes (rare + VUSs and rare + rare) had the worst clinical outcomes to first-line TKIs (P < 0.001), which was further confirmed using an external cohort (P = 0.0066). VUSs in the rare + VUSs subtype selectively reside in the EGFR kinase domain (P < 0.001), implying these tumors might select additional mutations to disrupt the regulation/function of the kinase domain. Conclusions: Different subtypes of compound EGFR mutations displayed distinct clinical features and genetic architectures, and rare mutation-dominant compound EGFR mutations were associated with enriched kinase domain-resided VUSs and poor clinical outcomes. Our findings help better understand the oncogenesis of compound EGFR mutations and forecast prognostic outcomes of personalized treatments.