P3-05-03: Transcriptomic Validation of Molecular Classification of Invasive Ductal Carcinoma Based on Immunohistochemical Markers and Grade

Transcriptomic analyses identified four major groups among invasive ductal carcinomas, associated with different clinical outcomes. Their definitions in practice is still matter of debate. Our aims were 1) to validate definitions based on immunohistochemical markers and grade: luminal A= ER and or PR+ve, grade I HER2−ve, luminal B= ER+ve and/or PR+ve, HER2+ve or grade III, HER2 enriched carcinomas= HER2+ve and ER-ve, basal-like/triple negative carcinomas= grade III, ER-ve, PR-ve, HER2−ve and expression of at least one of the basal-like markers (CK5/6, CK14, EGFR); 2) to refine this immunohistochemical definition. 142 consecutive tumors were selected in our tumour bank (42 triple-negative and grade III; 31 HER2+ve and ER-ve; 35 luminal B ER or PR+ve and grade III (7 cases) or HER2+ve (28 cases); luminal A (34 ER+ve or PR+ve, grade I)). Transcriptomic analyses were performed using Affymetrix U133+2 arrays (good quality RNAs obtained for all frozen specimens). The molecular classes determined according to proposed definitions were compared to those obtained with unsupervised clustering analyses using datas after GC-RMA normalisation (with the intrinsic gene list genes and with the highest variance genes). Marker patterns of expression (CK 8/18, 5/6, 14, EGFR, BCL2 and Ki67) were analysed within each molecular class. Based on the phenotypical definition and grade, 10 and 9% of cases were misclassified respectively using unsupervised clustering either with the intrinsic gene list or the highest variance genes. The misclassified tumors were luminal B HER2+ve case with low ER level of expression, or HER2+ve and ER-ve cases classified among the triple-negative group. 39 out of 42 (93%) triple negative expressed at least one of the basal markers. RP expression pattern differed between luminal A and B carcinomas. All luminal A showed at least > 15% of positive cells and 65% of them harboured > 50% of positive cells. In contrast, luminal B showed < 15% of positive cells in 70% of the cases. Bcl2 was negative in 40% of the luminal B cases and positive (> 20% stained cells) in more than 90% of the luminal A cases. CK14 was positive (i.e. > 1% positive cell) in 65% of the triple negative cases, compared to CK5/6 positive in 58% of the cases. Ki67 was > 20% in 90% of the triple negative and in 55% of luminal B cases compared to less than 5% of the luminal A cases. 20 and 30% of HER2+ve ER-ve carcinomas expressed CK5/6, CK14 and EGFR associated in more than 90% of the cases to CK8/18 positivity (> 20% positive cells). Identification of molecular classes of breast carcinomas was accurately determined by immunohistochemistry and grade. Low level of RP expression and BCL2 negativity were part of the luminal B phenotype. CK14 was more sensitive in this population of triple negative carcinomas to identify basal-like carcinomas. KI67 was highly expressed in the vast majority if not all breast triple negative carcinomas. HER2+ve ER-ve carcinomas can express basal markers but in contrast to basal-like carcinomas, associated in the large majority of the cases to CK8/18 expression. The accurate determination of molecular groups of breast carcinomas should be a key parameter for development of targeted therapies within each group. Citation Information: Cancer Res 2011;71(24 Suppl):Abstract nr P3-05-03.