Human pathology

A 9-protein biomarker molecular signature for predicting histologic type in endometrial carcinoma by immunohistochemistry.

PMID 25282036


Histologic typing may be difficult in a subset of endometrial carcinoma (EC) cases. In these cases, interobserver agreement improves when immunohistochemistry (IHC) is used. A series of endometrioid type (EEC) grades 1, 2, and 3 and serous type (SC) were immunostained for p53, p16, estrogen receptor, PTEN, IMP2, IMP3, HER2, cyclin B2 and E1, HMGA2, FolR1, MSLN, Claudins 3 and 4, and NRF2. Nine biomarkers showed significant differences with thresholds in IHC value scale between both types (p53 ≥ 20, IMP2 ≥ 115, IMP3 ≥ 2, cyclin E1 ≥ 220, HMGA2 ≥ 30, FolR1 ≥ 50, p16 ≥ 170, nuclear PTEN ≥ 2 and estrogen receptor ≤ 50; P < .005). This combination led to increased discrimination when considering cases satisfying 0 to 5 conditions predicted as EEC and those satisfying 6 to 9 conditions predicted as SC. This signature correctly predicted all 48 EEC grade 1-2 cases and 18 SC cases, but 3 SC cases were wrongly predicted as EEC. Sensitivity was 86% (95% confidence interval [CI], 64%-97%), and specificity was 100% (95% CI, 89%-100%). The classifier correctly predicted all 28 EEC grade 3 cases but only identified the EEC and SC components in 4 of 9 mixed EEC-SC. An independent validation series (29 EEC grades 1-2, 28 EEC grade 3, and 31 SC) showed 100% sensitivity (95% CI, 84%-100%) and 83% specificity (95% CI, 64%-94%). We propose an internally and externally validated 9-protein biomarker signature to predict the histologic type of EC (EEC or SC) by IHC. The results also suggest that mixed EEC-SC is molecularly ambiguous.