328. MOLECULAR SUBTYPING OF ESOPHAGEAL ADENOCARCINOMA BY NON-NEGATIVE MATRIX FACTORIZATION OF LASER CAPTURE MICRODISSECTED RNA-SEQ SAMPLES. (24th September 2022)

• Record Type: Journal Article
• Authors: Wilson, Gavin; Mathew, Elizabath; Shibahara, Yukiko; Allison, Frances; Bach, Yvonne; Allen, Jonathan; Kalimuthu, Sangeetha N; Elimova, Elena; Darling, Gail; Yeung, Jonathan
• Is Part Of: Diseases of the esophagus. Volume 35(2022)Supplement 2
• Publication Date: 2022-09-24
• Subjects: Esophagus -- Diseases -- Periodicals; 616.32
• DOI: 10.1093/dote/doac051.328
• Languages: English
• ISSNs: 1120-8694

Abstract

Abstract: Previous attempts to construct prognostic molecular subtypes for esophageal adenocarcinoma (EAC) using gene expression profiling has met with little success when compared to adenocarcinomas from other disease sites. We hypothesized that this is in part due to the low tumour cellularity obtained from EAC biopsy and resection specimens, which obscures the tumour-intrinsic gene expression signals due to normal cell contamination. We systematically collected biopsies from pre-treatment, post-induction, and metastatic EAC tumors. Laser capture microdissection (LCM) was utilized to enrich for tumour cells for whole transcriptome sequencing (RNA-seq) (N = 54). We utilized non-negative matrix factorization (NMF) with 7 components to identify common gene expression programs across our samples. After applying NMF to our samples, we identified 4 tumour-intrinsic components and 3 tumour-extrinsic components (including immune infiltrate and normal esophagus). These components validated on TCGA EAC and esophageal squamous cell carcinoma RNA-seq, single cell RNA-seq from EAC organoids, and the cancer cell line encyclopedia. Moreover, we observed that these tumour-intrinsic gene expression programs were preserved in EAC samples metastatic to other organ sites such as liver and lymph nodes. We observed that most tumours expressed more than one tumour component and that approximately 25% of our samples had low tumour signals despite LCM enrichment. We used LCM to enrich for tumour signal in EAC gene expression profiling data and identified 4 tumour-intrinsic gene expression programs that validate across multiple datasets. We have currently conducting survival analysis on our data and TCGA data to further validate the prognostic potential of our tumour signatures.