Most canine ameloblastomas harbor HRAS mutations, providing a novel large-animal model of RAS-driven cancer

a Mandibular CAA case prior to resection. b Histologic architecture (hematoxylin–eosin (H&E) stain) of typical CAA case; note tumor epithelium (violet) interdigitates with stroma (pink). Inset shows tumor region at higher magnification. CAA formalin-fixed paraffin-embedded (FFPE) tissue blocks (dated 2007–2015) were retrieved from the clinical archives of the Department of Pathology, UC Davis School of Veterinary Medicine, and H&E-stained sections reviewed by a trained veterinary pathologist (N.V.). c Integrated Genome Viewer display of mapped reads from WES of CAA case harboring HRAS-Q61R mutation. Red and blue reads map to plus and minus strands, respectively; only a subset of mapped reads is shown. WES was done on 16 CAA samples; while this was an exploratory study, sample sizes of 10–15 should provide 80% power to identify driver mutations if present at ≥20–30% frequency. Genomic DNA was extracted from CAA FFPE tissue scrolls using the Qiagen (Germantown, MD, USA) DNA FFPE Tissue Kit. WES was done using the Agilent (Santa Clara, CA, USA) SureSelect Canine All Exon Kit, following modifications recommended for FFPE-derived DNA samples. Barcoded WES libraries were sequenced (101？bp？×？2) on an Illumina HiSeq2500 or 4000 instrument (Stanford Genome Sequencing Service Center) to an average 116× mean base pair coverage. Raw reads were aligned to the dog genome (CanFam3.1) using BWA21. Single-nucleotide variants (SNVs) were called using SAMtools22 mpileup and, in the absence of matched normal, restricted to 597 canine gene orthologs of known human cancer genes (the union of Cancer Gene Census and FoundationOne gene lists) (Table S2). SNVs were annotated using the Ensembl Variant Effect Predictor23. Subsequently, SNVs were filtered to exclude known germline variants (SNPs) and to retain only those SNVs with High evidence (read depth ≥20; minor allele frequency 20–50%) and High consequence (missense, stop-gain, or splice donor/acceptor variants), yielding 171 SNVs (in 91 genes) across 16 tumors (Table S4). To further distinguish likely somatically acquired SNVs from personal germline SNPs, we focused only on those SNVs occurring at the orthologous position of known human cancer hotspot mutations24 (Table S3), determined from the Catalogue of Somatic Mutations in Cancer (COSMIC)25. Finally, we performed manual inspection of reads spanning HRAS-61, HRAS-13, and BRAF-595, identifying one additional HRAS-Q61R case (CAA-20) with mutant allele frequency 11%, missed by the automated SNV caller. All WES data are available from NCBI SRA (accession PRJNA516699). d