MAGEC3 is a prognostic biomarker in ovarian and kidney cancers

Rare variants in MAGEC3, members of the melanoma antigen gene family, are associated with BRCA-independent early onset ovarian cancers, while somatic mutations of this gene have been associated with kidney cancers. In this report, we quantified normal and tumor protein expression of MAGEC3 via immunohistochemistry in N=394 ovarian cancers and N=220 renal cell carcinomas. MAGEC3 protein levels fell into two categories - normal MAGEC3 and MAGEC3 loss - characterized by expression equivalent to normal tissue or significantly lower than normal tissue, respectively. Interestingly, cases with MAGEC3 loss demonstrated better overall survival in both ovarian cancers and renal cell carcinomas, which resembles patient outcomes with BRCA2 loss. MAGEC3 protein expression was associated with upregulation of pathways regulating G2/M checkpoint (NES: 4.13, FDR<0.001) and mitotic spindle formation (NES: 2.84, FDR<0.001). Increased CD8+ cell infiltration, coordinate expression of other cancer testis antigens, and tumor mutational burden were also associated with MAGEC3 expression. To emphasize the impact of these results, we built a prognostic RNA-based model using N=180 cancers of an independent cohort with matching transcriptomic data and tested its performance in two large public cohorts (N=282 ovary and N=606 kidney). Results based on predicted protein scores within these patients validated those discovered in patients with directly measured MAGEC3 protein. The RNA model was reproduced in independent cohorts implying a broader potential for MAGEC3-driven disease etiology and relevance to potential treatment selection. STATEMENT OF TRANSLATIONAL RELEVANCEMAGEC3 protein is expressed in multiple tissues and is dysregulated in cancer. In this work, we show that ovarian and kidney cancer patients with loss of MAGEC3 protein have favorable prognosis, indicating that MAGEC3 protein level may be used as a prognostic biomarker. Integrative genomic analysis of patients spanning more than nine cancer types showed an association between MAGEC3 protein and genes affecting stress response, including those involved in cell cycle and DNA damage repair. Additionally, it is correlated with tumor mutational burden in patients with mutated oncogenes. These associations suggest that MAGEC3 protein levels may be used to identify patients with deficient DNA damage repair mechanisms that can be targeted by PARP inhibitors. To operationalize this idea, we use machine learning to predict MAGEC3 protein levels from RNA sequencing data which can facilitate the identification of patients for treatment stratification according to their MAGEC3 status.