Retained mismatch repair protein expression occurs in approximately 6% of microsatellite instability-high cancers and is associated with missense mutations in mismatch repair genes.

Immunohistochemistry for mismatch repair protein expression is widely used as a surrogate for microsatellite instability status-an important signature for immunotherapy and germline testing. There are no systematic analyses examining the sensitivity of immunohistochemistry for microsatellite instability-high status. Mismatch repair immunohistochemistry and microsatellite instability testing were performed routinely as clinically validated assays. We classified germline/somatic mutation types as truncating (nonsense, frameshift, and in/del) versus missense and predicted pathogenicity of the latter. Discordant cases were compared with concordant groups: microsatellite instability-high/mismatch repair-deficient for mutation comparison and microsatellite stable/mismatch repair-proficient for immunohistochemical comparison. 32 of 443 (7%) microsatellite instability-high cases had immunohistochemistry. Four additional microsatellite instability-high research cases had discordant immunohistochemistry. Of 36 microsatellite instability-high cases with discordant immunohistochemistry, 30 were mismatch repair-proficient, while six (five MLH1 and one MSH2) retained expression of the defective mismatch repair protein and lost its partner. In microsatellite instability-high tumors with discordant immunohistochemistry, we observed an enrichment in deleterious missense mutations over truncating mutations, with 69% (25/36) of cases having pathogenic germline or somatic missense mutations, as opposed to only 19% (7/36) in a matched microsatellite instability-high group with concordant immunohistochemistry (p = 0.0007).  In microsatellite instability-high cases with discordant immunohistochemistry and MLH1 or PMS2 abnormalities, less cells showed expression (p = 0.015 and p = 0.00095, respectively) compared with microsatellite stable/mismatch repair-proficient cases. Tumor mutation burden, MSIsensor score, and truncating mismatch repair gene mutations were similar between microsatellite instability-high cases with concordant versus discordant immunohistochemical expression. Approximately 6% of microsatellite instability-high cases have retained mismatch repair protein expression and would be missed by immunohistochemistry-based testing, hindering patient access to immunotherapy. Another 1% of microsatellite instability-high cases show isolated loss of the defective gene's dimerization partner, which may lead to germline testing of the wrong gene. These cases are enriched for pathogenic mismatch repair missense mutations.