Abstract Introduction Gene expression microarrays, artificial neural network (ANN), tissue microarray and immunohistochemistry (IHC) techniques allow for the analysis of huge numbers of gene transcripts and their corresponding proteins and have been widely applied in predicting clinical outcome. Methods 1- In this study, we analysed 48,000 gene transcripts of 171 unselected series of BC using ANN and pathways analysis to identify genes that can be used to predict clinical outcome of BC. 2- The clinic-pathological outcome of candidate genes were validated by using IHC in 4 independent primary BC data sets: a) a series of 379 consecutive high risk BC (NPI>3.4) who treated with surgery (S)+ radiotherapy (RT) and did not received neither endocrine (ET) nor chemo-therapies (CT), b) A series of 1650 consecutive cases of BC who treated with S + RT and received adjuvant CMF and/or ET according to Nottingham prognostic index (NPI), menopausal and ER status, c) 250 locally advanced BC treated with anthracycline-based combination with or without Taxane followed by S + RT and d) 145 BC overexpressing HER-2 treated with S + RT followed by sequential adjuvant anthracycline combination CT + trastuzumab. Results Gene expression analysis ANN analysis revealed that KIF2C gene was the highest ranked gene that predicted clinical outcome and accurately differentiated between low and high grade BC based on a 10-fold external cross-validation analysis with an average classification accuracy of >98%. High KIF2C gene expression level was associated with shortest BC specific survival (BCSS), disease free (DFS) and distal metastasis free survivals (DM-FS); p<0.0001. In univariate analysis, high level of KIF2C gene expression was associated with large tumour size, higher lymph node stage, negative ER, positive p53 expression and HER2 overexpression. However in multivariate analysis, KIF2C gene expression level was only statistically associated with histological grade (p<00001) and mitosis (p<0.0001). Pathways analysis revealed that KIF2C is likely to play a significant role in cytokinesis, cell division and cell cycle regulations. Immunohistochemistry 75% of BC showed overexpression of KIF2C protein. KIF2C protein overexpression was associated with unfavourable clinic-pathological features including high grade, high mitotic index, basal like phenotype, triple negative phenotype, HER2 overexpression, TOP2A overexpression, p53 mutation, and loss of BRCA1 (adjusted p<0.0001). In univariate analysis, KIF2C protein overexpression was associated with patient's BCSS in both ER+/high risk patients (NPI > 3.4) who did not received ET (HR: 3.3, 95% CI: 1.2−9.3, p=0.02) and ER-/high risk patients who did not received CT (HR: 3.2, 95% CI: 1.1−8.8, p=0.025). In 1650 BC series, multivariate Cox regression model including validated prognostic factors, confirmed that KIF2C overexpression is an independent prognostic factor. KIF2C overexpression showed increase in the risk of death (HR: 1.5, 95% CI: 1.1−2.0, p=0.009), recurrence (HR: 1.4, 95% CI: 1.1−1.8, p=0.017) and DM (HR: 1.6, 95% CI: 1.2−2.3, p=0.005). In conclusion, our findings provide a new insight to a better understanding of mammary carcinogenesis and that KIF2C is a promising molecular prognostic factor and a potential therapeutic target. Citation Information: Cancer Res 2011;71(24 Suppl):Abstract nr P4-09-11.
Read full abstract