Abstract
It has been established that a high degree of tumor-infiltrating T cells is associated with ovarian cancer prognosis. We hypothesized that tumors display an immune-related program of transcription that can act in a stimulatory or a regulatory manner. We analyzed transcriptome-wide gene expression data from 503 ovarian tumors from the Cancer Genome Atlas to identify genes that show differential prognoses when stratified by CD3 expression. Genes with immunological functions and tumor antigen genes were selected for analysis. We repeated our analysis in an independent validation study. Five genes showed stimulatory/regulatory patterns at a high level of confidence (Bonferroni p < 0.05). Three of these (MAGEA8, MPL, AMHR2) were validated and one (WT1) could not be evaluated. These patterns show specific prognostic effect only in conjunction with CD3 expression. When patients express multiple transcripts in poor prognosis directions, there is a dose response: increasingly regulatory type tumors are associated with higher stage, lower treatment response and shorter overall survival and progression free survival. The high-confidence set of transcripts (MAGEA8, MPL, AMHR2, WT1) and selected low-confidence hits (EPOR, TLR7) alone or in combination represent candidate prognosis markers for further investigation.