Prognostic Significance of ARID1A Expression Patterns Varies with Molecular Subtype in Advanced Gastric Cancer

BACKGROUND/AIMS: AT-rich interactive domain 1A (ARID1A) is frequently mutated in gastric cancer (GC), especially Epstein-Barr virus (EBV)-associated and microsatellite instability high GC. The loss of ARID1A expression has been reported as a poor prognostic marker in GC. However, the relationships between ARID1A alteration and EBV-associated and microsatellite instability high GC, which are known to have a favorable prognosis, has hampered proper evaluation of the prognostic significance of ARID1A expression in GC. We aimed to analyze the true prognostic significance of ARID1A expression by correcting confounding variables. METHODS: We evaluated the ARID1A expression in a large series (n=1,032) of advanced GC and analyzed the relationships between expression pattern and variable parameters, including clinicopathologic factors, key molecular features such as EBV-positivity, mismatch repair protein deficiency, and expression of p53 and several receptor tyrosine kinases including human epidermal growth factor receptor 2, epidermal growth factor receptor, and mesenchymal-epithelial transition factor. Survival analysis of the molecular subtypes was done according to the ARID1A expression patterns. RESULTS: Loss of ARID1A expression was found in 52.5% (53/101) of mutL homolog 1 (MLH1)-deficient and 35.8% (24/67) of EBV-positive GCs, compared with only 9.6% (82/864) of the MLH1-proficient and EBV-negative group (p<0.001). The loss of ARID1A expression was associated only with MLH1 deficiency and EBV positivity. On survival analysis, the loss of ARID1A expression was associated with worse prognosis only in MLH1-proficient and EBV-negative GC. Multivariate analysis revealed that both loss of ARID1A and decreased ARID1A expression were independent worse prognostic factors in patients with advanced GC. CONCLUSIONS: Only in MLH1-proficient and EBV-negative GC, the loss of ARID1A expression is related to poorer prognosis.