Intrinsic Subtype Switching and Acquired ERBB2/HER2 Amplifications and Mutations in Breast Cancer Brain Metastases

Abstract

Importance: Patients with breast cancer (BrCa) brain metastases (BrM) have limited therapeutic options. A better understanding of molecular alterations acquired in BrM could identify clinically actionable metastatic dependencies.

Objective: To determine whether there are intrinsic subtype differences between primary tumors and matched BrM and to uncover BrM-acquired alterations that are clinically actionable.

Design, setting, and participants: In total, 20 cases of primary breast cancer tissue and resected BrM (10 estrogen receptor [ER]-negative and 10 ER-positive) from 2 academic institutions were included. Eligible cases in the discovery cohort harbored patient-matched primary breast cancer tissue and resected BrM. Given the rarity of patient-matched samples, no exclusion criteria were enacted. Two validation sequencing cohorts were used-a published data set of 17 patient-matched cases of BrM and a cohort of 7884 BrCa tumors enriched for metastatic samples.

Main outcomes and measures: Brain metastases expression changes in 127 genes within BrCa signatures, PAM50 assignments, and ERBB2/HER2 DNA-level gains.

Results: Overall, 17 of 20 BrM retained the PAM50 subtype of the primary BrCa. Despite this concordance, 17 of 20 BrM harbored expression changes (<2-fold or >2-fold) in clinically actionable genes including gains of FGFR4 (n = 6 [30%]), FLT1 (n = 4 [20%]), AURKA (n = 2 [10%]) and loss of ESR1 expression (n = 9 [45%]). The most recurrent expression gain was ERBB2/HER2, which showed a greater than 2-fold expression increase in 7 of 20 BrM (35%). Three of these 7 cases were ERBB2/HER2-negative out of 13 ERBB2/HER2-negative in the primary BrCa cohort and became immunohistochemical positive (3+) in the paired BrM with metastasis-specific amplification of the ERBB2/HER2 locus. In an independent data set, 2 of 9 (22.2%) ERBB2/HER2-negative BrCa switched to ERBB2/HER2-positive with 1 BrM acquiring ERBB2/HER2 amplification and the other showing metastatic enrichment of the activating V777L ERBB2/HER2 mutation. An expanded cohort revealed that ERBB2/HER2 amplification and/or mutation frequency was unchanged between local disease and metastases across all sites; however, a significant enrichment was appreciated for BrM (13% local vs 24% BrM; P < .001).

Conclusions and relevance: Breast cancer BrM commonly acquire alterations in clinically actionable genes, with metastasis-acquired ERBB2/HER2 alterations in approximately 20% of ERBB2/HER2-negative cases. These observations have immediate clinical implications for patients with ERBB2/HER2-negative breast cancer and support comprehensive profiling of metastases to inform clinical care.

Figure 1. Expression Profile Similarity in Patient-Matched Breast Cancer Metastases to the Brain

(A) Unsupervised hierarchical clustering heatmap of 20 patient-matched cases with hormone status (green indicates positive; black, negative), tissue site source or institution (yellow indicates Royal College of Surgeons [RCS], Ireland; purple, University of Pittsburgh [Pitt], United States), and tumor site (blue indicates primary; red, metastasis) of each sample indicated. The asterisks below the plots indicate patient-matched pairs that clustered in the same doublet of a clade in the dendrogram. (B) PAM50 intrinsic molecular subtype calls in patient-matched cases (red indicates basal; green, ERBB2/HER2; blue, LumA; yellow, LumB). Discordant pairs are marked with a delta symbol. BP indicates breast primary; BrM, brain metastasis.

Figure 2. Recurrent Expression Alterations in Breast Cancer Brain Metastases

(A) OncoPrint plot of highly recurrent (>25% of cases) expression alterations in 20 cases, ranked by frequency of alteration by gene. Blue tile represents a greater than 2-fold decrease in the patient-matched brain metastasis relative to the primary, while a red tile represents a greater than 2-fold increase. (B) Tile plot visualizing expression alterations in clinically actionable genes. Top panel consists of recurrent increases in expression (light red indicates a >2-fold increase; dark red, >4-fold increase), bottom panel are recurrent decreases in expression (light blue indicates a >2-fold decrease; dark blue, >4-fold decrease) between patient-matched pairs. Pitt indicates University of Pittsburgh; RSC, Royal College of Surgeons.

Figure 3. ERBB2/HER2 Gains in Breast Cancer Brain Metastases

(A) Paired ladder plot of ERBB2/HER2 expression in patient-matched cases. Gray dots represent samples with suspected hormone status switching; P values are from Wilcoxon signed-rank tests (primaries vs metastases). (B) The top panel shows ERBB2/HER2 alterations (amplification or mutation) in 3135 local tumors and 4130 metastatic tumors, and the bottom panel shows ERBB2/HER2 alterations in local tumors and 167 brain metastases (mets). NS indicates not significant.