Histo- and immunohistochemistry-based estimation of the TCGA and ACRG molecular subtypes for gastric carcinoma and their prognostic significance: A single-institution study.

Abstract

Gastric cancers comprise molecularly heterogeneous diseases; four molecular subtypes were identified in the cancer genome atlas (TCGA) study, with implications in patient management. In our efforts to devise a clinically feasible means of subtyping, we devised an algorithm based on histology and five stains available in most academic pathology laboratories. This algorithm was used to subtype our cohort of 107 gastric cancer patients from a single institution (St. Michael's Hospital, Toronto, Canada), which was divided into 3 cases of EBV-positive, 23 of MSI, 27 of GS and 54 of CIN tumours. 87% of the tumours with diffuse histology were classified as GS subtype, which was notable for younger age. Examining for characteristic molecular features, aberrant p53 immunostaining was seen most frequently in the CIN subtype (43% in CIN vs. 6% in others), whereas ARID1A loss was rarely seen (6% vs. 35% in others). HER2 overexpression was seen exclusively in CIN tumours (17% of CIN tumours). PD-L1 positivity was seen predominantly in the EBV and MSI tumours. As with the TCGA study, no survival differences were seen between the subtypes. A similar strategy was employed to approximate the Asian Cancer Research Group (ACRG) molecular subtyping, with the addition of p53 IHC to the algorithm. We observed rates of ARID1A loss and HER2 overexpression that were comparable to the ACRG study. In summary, our algorithm allowed for clinically feasible means of subtyping gastric carcinoma that recapitulated the key molecular features reported in the large scale studies.