Identifying the genomic landscape of EGFR-mutant lung cancers with central nervous system metastases

Abstract

Background: Despite the intracranial efficacy of osimertinib, central nervous system (CNS) metastases remain a major cause of morbidity and mortality in EGFR-mutant non-small-cell lung cancer (NSCLC). The genomic drivers of CNS dissemination are poorly understood.

Patients and methods: We analyzed the clinicogenomic features of patients with EGFR-mutant NSCLC receiving first-line osimertinib with extracranial next generation sequencing (NGS) (n = 262) and individuals with intracranial NGS (n = 81). Paired extra- and intracranial NGS was available for 14 patients. Time-to-event analyses were conducted from time of metastatic diagnosis, except for time-to-treatment discontinuation (TTD), which began at treatment initiation.

Results: Among 262 patients receiving first-line osimertinib, 53% developed CNS metastases (36% de novo, 16% acquired on treatment). The cumulative incidence of brain (BrM) and leptomeningeal metastases (LM) was 39% and 2% at 1 year, 49% and 6% at 3 years, and 54% and 12% at 5 years, respectively. CNS metastases correlated with a higher frequency of CARD11 amplifications (14% versus 3%, P = 0.031) and a lower frequency of MDM2 amplifications (1% versus 13%, P = 0.008) in extracranial NGS specimens, with otherwise similar genomic profiles. Patients who developed CNS metastases on treatment had worse overall survival (OS) [hazard ratio (HR) = 3.67, 95% confidence interval (CI) 2.41 to 5.59], followed by those with de novo (HR = 1.61, 95% CI 1.15 to 2.26), compared with those who never developed CNS metastases (P < 0.001). In multivariable Cox regression, atypical EGFR mutations were associated with shorter OS. Cell cycle pathway alterations were more frequent in BrM than LM samples (93% versus 47%, P = 0.003, q = 0.03). No other significant genomic differences were found between BrM and LM, or between paired CNS and systemic samples.

Conclusions: Patients with atypical EGFR mutations or acquired CNS metastases on osimertinib have worse outcomes. Comparative NGS profiling of intra- and extracranial tumors suggest that CNS dissemination is driven by mechanisms beyond single-gene alterations.

Keywords: EGFR-mutant; brain metastases; genomics; leptomeningeal metastases; non-small-cell lung cancer; osimertinib.

Figure 1.. Incidence of developing brain or leptomeningeal metastases in patients with EGFR-mutant non-small-cell lung cancer on first-line osimertinib.

Cumulative incidence of (A) brain (BrM) and (B) leptomeningeal metastases (LM) in patients on first-line osimertinib. Cumulative incidence of BrM by (C) EGFR alteration and (D) EGFR co-alteration in patients without baseline central nervous system (CNS) metastases. Cumulative incidence of LM by (E) EGFR alteration and (F) EGFR co-alteration in patients without baseline CNS metastases.

Figure 2.. Genomic analysis of extracranial tumors from patients with and without central nervous system (CNS) metastases.

(A) Genomic landscape of pretreatment samples from patients without (‘Never’) or with CNS metastases (de novo or acquired, ‘Ever’). The type of genetic alteration (structural variant, deletion, amplification, truncating mutation, splice variant, missense mutation, in-frame mutation, promoter mutation), mutational signature category [reactive oxygen species (ROS), homologous recombination deficiency (HRD), smoking, APOBEC, Clock-like, unknown], and whole-genome doubling (WGD) is described in the key. The frequency of mutations is noted in the middle. Tumor mutational burden (TMB) frequency is displayed on top. (B) Scatterplot of gene alterations with frequencies ≥5% in ‘Never’ versus ‘Ever’ tumor samples. TP53, RB1, EGFR, and MET are shown regardless of frequency. Genes highlighted in red indicate a ≥10% frequency difference between the cohorts. (C) Comparison of oncogenic signaling pathway alterations from ‘Never and ‘Ever’ patient samples.

Figure 3.. Clinical outcomes among patients with EGFR-mutant non-small-cell lung cancer on first-line osimertinib.

(A) Deaths per person-months and (B) time-to-treatment discontinuation by central nervous system (CNS) type (Never CNS mets, De novo, or CNS mets acquired on treatment) in patients with EGFR-mutant NSCLC on first-line osimertinib. mets, metastases; Tx, treatment.

Figure 4.. Genomic analysis comparing brain versus leptomeningeal metastases and paired central nervous system (CNS) versus systemic metastases.

(A) Scatterplot of gene alterations with frequencies ≥5% in brain (BrM) and leptomeningeal (LM) metastases, or (D) paired CNS (intracranial) and systemic (extracranial) metastases. TP53, RB1, EGFR, and MET are shown regardless of frequency. Genes highlighted in red indicate a ≥10% frequency difference between the cohorts. Oncogenic signaling pathway alterations from (B) BrM and LM, or (E) paired CNS and systemic metastases. Bold text indicates statistical significance (two-sided Fisher’s exact test, P < 0.05; Multiple testing correction by Benjamini-Hochberg, q < 0.05). (C) Genomic landscape of paired EGFR-mutant CNS (intracranial) and systemic (extracranial) samples. The type of genetic alteration (structural variant, deletion, amplification, truncating mutation, splice variant, missense mutation, inframe mutation, promoter mutation), mutational signature category [reactive oxygen species (ROS), homologous recombination deficiency (HRD), smoking, APOBEC, Clock-like, unknown], and whole-genome doubling (WGD) is described in the key. Time interval (in months) between paired samples is displayed under sample type. The frequency of mutations is noted on the left. Tumor mutational burden (TMB) frequency is displayed on top. CSF, cerebrospinal fluid.