A Signature Predicting Poor Prognosis in Gastric and Ovarian Cancer Represents a Coordinated Macrophage and Stromal Response

Purpose: Gene-expression profiling has revolutionized the way we think about cancer and confers the ability to observe the synchronous expression of thousands of genes. The use of putative genome-level expression profiles has allowed biologists to observe the complex interactions of genes that constitute recognized biologic pathways. We used gastric and ovarian datasets to identify gene-expression signatures and determine any functional significance. Experimental Design: Microarray data of 94-tumor and 45-benign samples derived from patients with gastric cancer were interrogated using Hierarchical Ordered Partitioning and Collapsing Hybrid analysis identifying clusters of coexpressed genes. Clusters were further characterized with respect to biologic significance, gene ontology, and ability to discriminate between normal and tumor tissue. Tumor tissues were separated into epithelial and stromal compartments and immunohistochemical analysis performed to further elucidate specific cell lineages expressing genes contained in the signature. Results: We identified a stromal-response expression signature, highly enriched for inflammatory, extracellular matrix, cytokine, and growth factor proteins. The majority of genes in the signature are expressed in the tumor-associated stroma but were absent in associated premalignant conditions. In gastric cancer, this module almost perfectly differentiates tumor from nonmalignant gastric tissue and hence can be regarded as a highly tumor-specific gene-expression signature. Conclusions: We show that these genes are consistently coexpressed across a range of independent gastric datasets as well as other cancer types suggesting a conserved functional role in cancer. In addition, we show that this signature can be a surrogate marker for M2 macrophage activity and has significant prognostic implications in gastric and ovarian high-grade serous cancer. (C) 2014 AACR.