Proteogenomic Analysis of Salivary Adenoid Cystic Carcinomas
Total Page:16
File Type:pdf, Size:1020Kb
Published OnlineFirst November 10, 2020; DOI: 10.1158/1078-0432.CCR-20-1192 CLINICAL CANCER RESEARCH | TRANSLATIONAL CANCER MECHANISMS AND THERAPY Proteogenomic Analysis of Salivary Adenoid Cystic Carcinomas Defines Molecular Subtypes and Identifies Therapeutic Targets A C Renata Ferrarotto1, Yoshitsugu Mitani2, Daniel J. McGrail3, Kaiyi Li1, Tatiana V. Karpinets4, Diana Bell2, Steven J. Frank5, Xingzhi Song4, Michael E. Kupferman6, Bin Liu7, J Jack Lee8, Bonnie S. Glisson1, Jianhua Zhang4, Jon C. Aster9, Shiaw-Yih Lin3, P. Andrew Futreal4, John V. Heymach1, and Adel K. El-Naggar2 ABSTRACT ◥ Purpose: Salivary gland adenoid cystic carcinoma (ACC) has of TP63 and receptor tyrosine kinases (AXL, MET, and EGFR) and heterogeneous clinical behavior. Currently, all patients are treated less aggressive clinical course. TP63 and MYC were sufficient to uniformly, and no standard-of-care systemic therapy exists for assign tumors to ACC subtypes, which was validated in one metastatic ACC. We conducted an integrated proteogenomic anal- independent cohort by IHC and two additional independent yses of ACC tumors to identify dysregulated pathways and propose cohorts by RNA-seq. Furthermore, IHC staining for MYC and a classification with therapeutic implications. P63 protein levels can be used to identify ACC subtypes, enabling Experimental Design: RNA/DNA sequencing of 54 flash-frozen rapid clinical deployment to guide therapeutic decisions. Our data salivary ACCs and reverse phase protein array (RPPA) in 38 speci- suggest a model in which ACC-I is driven by MYC signaling mens were performed, with validation by Western blotting and/or through either NOTCH mutations or direct amplification, which IHC. Three independent ACC cohorts were used for validation. in turn suppress P63 signaling observed in ACC-II, producing Results: Both unbiased RNA sequencing (RNA-seq) and RPPA unique therapeutic vulnerabilities for each subtype. analysis revealed two molecular subtypes: ACC-I (37%) and ACC-II Conclusions: Cooccurrence of multiple actionable protein/path- (63%). ACC-I had strong upregulation of MYC, MYC target genes, ways alterations in each subtype indicates unique therapeutic and mRNA splicing, enrichment of NOTCH-activating mutations, vulnerabilities and opportunities for optimal combination therapy and dramatically worse prognosis. ACC-II exhibited upregulation for this understudied and heterogeneous disease. Introduction part restrained malignant effect on tumor. Usually, loss of myoepithe- lial cells within a tumor is associated with emergence of a solid Adenoid cystic carcinoma (ACC) is a biphasic tumor that typically epithelial form and aggressive behavior, although solid subtype may originates in the salivary glands of the head and neck. A unique be composed of myoepithelial cells (1). Treatment of progressive ACC characteristic of this entity is the concurrent participation of myoe- remains a major challenge largely due to their high biological vari- pithelial and epithelial neoplastic cells in the formation of this ability and the lack of biomarkers for targeted therapy (2, 3). tumor (1). Although the underlying events associated with the dual Recently, DNA sequencing has led to major advances in under- cellular formation are unknown, the presence of myoepithelial cells in standing the genomic landscape of ACC. However, only a few ther- apeutic targets have been identified with this approach (4–6). RNA sequencing (RNA-seq) of paraffin-embedded ACC samples has 1Department of Thoracic/Head and Neck Medical Oncology, The University of 2 revealed differential expression patterns based on predominant his- Texas MD Anderson Cancer Center, Houston, Texas. Department of Pathology, MYB-MYBL1 3 tologic subtype (epithelial vs. myoepithelial) and expres- The University of Texas MD Anderson Cancer Center, Houston, Texas. Depart- fi ment of Systems Biology, The University of Texas MD Anderson Cancer Center, sion (1, 7). While these ndings underscore the need for more Houston, Texas. 4Department of Genomic Medicine, The University of Texas MD personalized therapy for patients with ACC, the direct therapeutic Anderson Cancer Center, Houston, Texas. 5Department of Radiation Oncology, implications are unclear. The University of Texas MD Anderson Cancer Center, Houston, Texas. 6Depart- Our group has previously utilized reverse phase protein array (RPPA) ment of Head and Neck Surgery, The University of Texas MD Anderson Cancer to identify altered pathways and new therapeutic targets in non–small Center, Houston, Texas. 7Department of Epigenetics and Molecular Carcino- cell lung cancer (8). Similar studies in ACC are lacking. Proteomic genesis, The University of Texas MD Anderson Cancer Center, Houston, Texas. 8Department of Statistics, The University of Texas MD Anderson Cancer Center, analysis can yield unique and complementary data to that provided by Houston, Texas. 9Department of Pathology, Harvard Medical School and Brig- genomic analysis and can more directly guide therapeutic development. ham and Women's Hospital, Boston, Massachusetts. In this study, we performed RNA-seq, RPPA, and targeted DNA fl Note: Supplementary data for this article are available at Clinical Cancer deep sequencing on ash-frozen clinically annotated ACC samples Research Online (http://clincancerres.aacrjournals.org/). and conducted an integrative analysis to identify dysregulated path- Corresponding Author: Renata Ferrarotto, University of Texas MD Anderson ways and druggable therapeutic targets. Cancer Center, 1515 Holcombe Blvd, Box 432, Houston, TX 77030. Phone: 713- 792-4545; Fax: 713-792-1220; E-mail: [email protected] Materials and Methods – Clin Cancer Res 2020;XX:XX XX Patient selection doi: 10.1158/1078-0432.CCR-20-1192 The study population comprised of 54 patients with salivary ACC Ó2020 American Association for Cancer Research. and flash-frozen primary tumor samples available for RNA-seq, AACRJournals.org | OF1 Downloaded from clincancerres.aacrjournals.org on September 24, 2021. © 2020 American Association for Cancer Research. Published OnlineFirst November 10, 2020; DOI: 10.1158/1078-0432.CCR-20-1192 Ferrarotto et al. lysate was diluted with five dilutions and arrayed onto nitrocellulose- Translational Relevance coated slides by Aushon 2470 Arreyer (Aushon Biosystems). Each Our comprehensive unbiased proteogenomic analysis of ade- array was probed with extensively validated primary antibodies noid cystic carcinoma (ACC) revealed two molecular subtypes: by Dako Autostainer Plus (Dako North America, Inc.) and protein ACC-I (37%) and ACC-II (63%), with distinctive histologic char- expression signals were measured by Dako Catalyzed Signal Ampli- acteristics and prognosis. ACC-I was associated with worse prog- fication System according to the manufacturer's recommendations. nosis and exhibited enrichment of NOTCH-activating mutations, RPPA data processing was done using MicroVigene Software overexpression of MYC, and upregulation of mRNA splicing (VigeneTech) and an in-house R package. pathway. ACC-II was associated with better prognosis and over- expression of potential therapeutic targets, including AXL, phos- DNA extraction and targeted sequencing pho-MET, and phospho-EGFR. Notably, we established a two- Genomic DNAs from flash-frozen tissue samples were extracted gene (MYC/TP63) signature that accurately segregates tumors into using the Gentra Puregene Tissue Kit (Qiagen) according to the ACC-I and ACC-II subtypes. We validated the utility of this MYC/ manufacturer's protocols. Deep-targeted exome sequencing of TP63 signature in three additional cohorts. Furthermore, we show 263 genes (T200) was used to evaluate genomic alterations in 54 that assessment of MYC and p63 by IHC, which is routinely samples as described previously (14). Because matched normal tissue performed in many clinical laboratories, accurately stratifies was not available, we used a common normal control sample to tumors by subtype allowing for rapid implementation of ACC identify single-nucleotide variants by MuTect and small insertions subtyping for clinical trials and other therapeutic decisions. Col- and deletions by Pindel (15, 16). To reduce false positives, outputs lectively, our novel molecular classification can be highly valuable generated by MuTect and Pindel were filtered by the following for strategizing optimal targeted therapies for this orphan disease. approaches: (i) removal of novel mutations with variant allele fre- quency (VAF) > 0.4 and documented mutations from Catalogue of Somatic Mutations in Cancer with VAF > 0.6, (ii) removal of known SNPs and exon variants, as well as polymorphic genes with VAF < 0.4, RPPA, and targeted DNA deep sequencing. This is a subset of a cohort and (iii) sequencing depth of at least 100 reads for documented of 102 patients previously characterized by whole-exome sequencing mutations and 300 reads for novel mutations. The filtering criterion and various targeted sequencing panels from formalin-fixed, paraffin- for VAF was chosen after manual curation of mutations selected with embedded (FFPE) tissues as described previously (ref. 9; Supplemen- different VAF thresholds. Hits found across many samples were tary Fig. S1). This study was conducted in accordance with Declaration manually curated to reduce the rate of false-positive mutations. of Helsinki. Samples were obtained by either an institutional review board–approved waiver of informed consent (for deceased patients) or Subtype clustering informed consent (front-door consent) for molecular and clinicopath-