Abstract
Abstract
Gastric Cancer (GC) is the second leading cause of cancer-related mortality (9.7% of the total) and most patients with advanced disease will die within one year of diagnosis. GC is histologically classified into intestinal, diffuse and the mixed types, and into four molecular subtypes based on genetic profiling (i.e. microsatellite instable (MSI), EBV positive, chromosomal instable, and genomically stable). Although the molecular annotation is meaningful, the tumor immune microenvironment (TIME) has largely not been evaluated. Using an integrated collection of 11 public cohorts (n=2,270), we identified 5 unique GC TIME phenotypes by unsupervised clustering of gene expression signatures of RNA that describe the composition and biology of TIME and properties of malignant cells: two stromal-enriched subtypes - 1) Mesenchymal-wnt activated, high stroma activation and WNT signaling and 2) Fibrotic, only high stroma activation; two immune enriched subtypes 3) Inflamed non-fibrotic, high immune infiltration and low stromal compartment and 4) B-cell inflamed, high B cell activation; and 5) Immune Depleted, lowest immune and the highest malignant cell properties. The clusters were robust, being identified across all 11 datasets and across all stages of disease. Intestinal, diffuse, and mixed histologies were identified in each cluster. These clusters will be orthogonally validated using immunohistochemistry.
The current molecular subtypes were represented in each of our TIME clusters, with some enrichment. Specifically, we found enrichment of MSI molecular subtype, characterized by hypermutation and microsatellite instability, in “Inflamed, non-fibrotic” cluster whereas the MSS/EMT subtype, associated with poor overall survival, was enriched in “Mesenchymal, wnt activated” cluster. We also saw specific enrichment of EBV positive tumors, known to have good prognosis, in both immune-enriched clusters. These results show a high concordance with the current TCGA/ACRG molecular subtypes of GC.
These TIME clusters are prognostic in GC. The “Inflamed, non-fibrotic” cluster demonstrated a better overall and relapse free survival whereas stromal enriched clusters exhibited the worst (p<0.001, HR=2.27). Interestingly, the most aggressive “Mesenchymal, wnt activated” subtype was also enriched of metastatic samples. These results were confirmed using an independent validation cohort (n=231) from four other datasets. Additionally, comparison of matching on-treatment vs baseline biopsies from 7 patients treated with cabazitaxel further suggested that TIME changed upon treatment and, in some cases, was indicative of poor response to taxanes.
We have defined and characterized the TIME for GC. The GC microenvironment is both prognostic for patient outcome and predictive of response to cytotoxic therapy.
Citation Format: Prashant V. Thakkar, Olga Kudryashova, Daria Melikhova, Naira Samarina, Sandrine Degryse, Alexander Bagaev, Felix Frenkel, Svetlana Podsvirova, Dmitry Tychinin, Sandipto Sarkar, Rhonda K. Yantiss, Nathan Fowler, Manish A. Shah. Tumor immune microenvironment based molecular functional clustering reveals a prognostic signature that predicts overall survival in patients with gastric cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 3182.