A 107 Gene Nanostring Assay Effectively Translates the Cancer Genome Atlas, and Tumour Microenvironment Gastric...

There is a need to improve the translation of gastric cancer molecular classification schemes, such as those proposed by the Cancer Genome Atlas (TCGA) and Tumour Microenvironment score (TME), to clinical specimens and three-dimensional organoid culture models. In this study, we validate a 107-gene Nanostring assay informed by previously established machine learning models using a prospective cohort of gastric adenocarcinoma tumours and tumour-organoid pairs. Thirty-eight gastric adenocarcinoma specimens and twelve parent tumour-tumour organoid pairs were assigned TCGA and TME subtypes using gene expression measured by our custom Nanostring gene set. Subtypes were validated using gold-standard tests for Epstein-Barr virus (EBV) and microsatellite instability (MSI). Molecular subtype scores were compared to known clinicopathologic characteristics. The correlation between dose-response and molecular subtypes using an organoid drug assay and the Cancer Cell Line Encyclopedia (CCLE) was investigated. TCGA and TME subtypes were successfully applied to all specimens. The relationship of molecular subtype scores in our population compared to public cohorts was statistically identical for Lauren Class and Signet Ring status. Our method achieved 100% accuracy in labeling EBV and MSI subtypes. We identified 81.8% and 63.6% concordance between parent tumour-tumour organoid pairs for TME and TCGA subtypes, respectively. No significant correlation was identified between dose response to chemotherapy and molecular subtype scores. Analysis of the CCLE identified promising personalized therapy candidates for each molecular subtype. Our 107-gene Nanostring test successfully assigns TCGA and TME molecular subtypes to clinical tumour and tumour organoid samples for use in future study.