Comprehensive Molecular Characterization of Salivary Duct Carcinoma Reveals Actionable Targets and Similarity to Apocrine Breast Cancer

Abstract

Purpose: Salivary duct carcinoma (SDC) is an aggressive salivary malignancy, which is resistant to chemotherapy and has high mortality rates. We investigated the molecular landscape of SDC, focusing on genetic alterations and gene expression profiles.

Experimental design: We performed whole-exome sequencing, RNA sequencing, and immunohistochemical analyses in 16 SDC tumors and examined selected alterations via targeted sequencing of 410 genes in a second cohort of 15 SDCs.

Results: SDCs harbored a higher mutational burden than many other salivary carcinomas (1.7 mutations/Mb). The most frequent genetic alterations were mutations in TP53 (55%), HRAS (23%), PIK3CA (23%), and amplification of ERBB2 (35%). Most (74%) tumors had alterations in either MAPK (BRAF/HRAS/NF1) genes or ERBB2 Potentially targetable alterations based on supportive clinical evidence were present in 61% of tumors. Androgen receptor (AR) was overexpressed in 75%; several potential resistance mechanisms to androgen deprivation therapy (ADT) were identified, including the AR-V7 splice variant (present in 50%, often at low ratios compared with full-length AR) and FOXA1 mutations (10%). Consensus clustering and pathway analyses in transcriptome data revealed striking similarities between SDC and molecular apocrine breast cancer.

Conclusions: This study illuminates the landscape of genetic alterations and gene expression programs in SDC, identifying numerous molecular targets and potential determinants of response to AR antagonism. This has relevance for emerging clinical studies of ADT and other targeted therapies in SDC. The similarities between SDC and apocrine breast cancer indicate that clinical data in breast cancer may generate useful hypotheses for SDC. Clin Cancer Res; 22(18); 4623-33. ©2016 AACR.

Figure 1

Genetic landscape of salivary duct carcinoma. (A) Number of somatic mutations (top), clinical data (middle) and selected genetic alterations (bottom) in 16 patients with SDC. SDC ex. PA, Salivary duct carcinoma ex pleomorphic adenoma; AR, Androgen receptor; IHC, Immunohistochemistry; CNA, copy number alteration; MAPK, mitogen-activated protein kinase pathway; RTKs, receptor tyrosine kinases. *Hotspot mutation: At least 20 cases of the mutation reported in the COSMIC database (cancer.sanger.ac.uk/cosmic). (B) Most commonly altered genes in cohorts 1 and 2 combined. Mutations and significant copy number alterations are included. (C) Percent of patients with potentially targetable alterations according to the Memorial Sloan Kettering Cancer Center (MSKCC) levels of evidence, and the specific alterations listed for each category.

Figure 2

Alterations related to androgen signaling. (A) Illustration of the androgen receptor (AR) full length (FL) gene, including exons 1–8, and AR transcript variant 7 (V7), including exons 1–3 and cryptic exon 3 (CE3). (B) Sashimi plots showing RNA sequencing reads that span more than one AR exon. Representative cases with (top) or without (bottom) reads spanning exon 3 and CE3 are shown. (C) Quantification of AR-FL and AR-V7 expression based on RNA sequencing (top), and reverse transcriptase-PCR using cDNA extracted from tumor RNA (bottom). (D) Correlation between AR and FOXA1 expression. P<0.0001, 2-tailed Pearson correlation test. (E) Location of SDC mutations within the FOXA1 gene. Previously reported mutations were acquired from the cBioPortal database (cbioportal.org), and represents cases from any cancer. TA, transactivation domain; Forkhead N, Forkhead N-terminal domain. (F) Correlation between AR immunohistochemistry (red, positive; blue, negative), AR expression (based on z-score) and AR signaling index (see methods) with FOXA1 mutation status. (G) Correlation between AR expression and AR signaling index. P, one-tailed Spearman correlation.

Figure 3

SDC shows a similar gene expression pattern as estrogen-receptor negative breast cancer. (A) Unsupervised clustering of 16 cases of SDC (blue) and 591 cases of breast cancer from TCGA (x-axis), based on expression of the 100 most variable genes (y-axis). Breast cancer subtype was determined using the PAM50 gene signature. (B) AR and ERBB2 gene expression (FPKM) in SDC tumors based on RNA sequencing. (C) Enlargement of cluster 1 and 5 from figure 3A, with SDC cases colored according to AR and ERBB2 expression.

Figure 4

SDC resembles molecular apocrine breast cancer. (A) Consensus clustering of 16 SDC tumors (blue) and 109 basal-like breast cancers from TCGA (red), defined using the PAM50 gene signature. (B) Gene set enrichment analysis showing an overlap between differentially expressed genes in cluster 1 (from Figure 4A), and gene signature 7 as determined by Farmer et al. FDR, false discovery rate. (C) Gene set enrichment analysis between clusters (from Figure 4A) and gene signatures according to Farmer et al. Ellipses represent significant enrichment of gene sets between two clusters. Ellipse area corresponds to level of significance.