4-gene score to predict response to neoadjuvant chemotherapy, metastasis, and survival.

Abstract

e12592 Background: At this point, there is no treatment to cure metastatic breast cancer. On the other hand, improvement of treatment of metastatic breast cancer is the key to further improve breast cancer survival. One possible opportunity is to detect metastatic breast cancer in it’s “early” phase when it many respond to treat such as immune therapy. We hypothesized that a scoring system generated from metastatic clone predict metastasis and survival. Methods: A gene signature score was generated using differentially expressed genes between MDA-MB-231 human breast cancer cell-line and its highly metastatic variant, LM2-4, and TCGA breast cancer dataset. METABRIC cohort as well as six neoadjuvant therapeutic cohorts and four metastatic cohorts were used as validation cohort. The totals of 4,199 samples from a large number of cohorts were analyzed in the study. Results: Among the 297 genes that were up-regulated in LM2-4, tumor expression of four genes had the strongest association with disease-free survival in TCGA breast cancer dataset (hazard ratio [HR] > 1.2 with p < 0.02). A 4-gene score calculated from the combined expression of these genes correlated with overall survival in TCGA data, with HR of 1.44 (95% confidence interval [CI] = 1.52-2.98; p < 0.001). The association of higher scores with worse survival was also observed for disease-specific survival in the METABRIC breast cancer cohort (HR = 1.57, 95% CI = 1.23-1.72, p < 0.001). Tumors with high score enriched gene sets related with cancer aggressiveness such as E2F, G2F, MYC v1 and v2 (FDR = 0.23, 0.19, 0.20, 0.21, respectively), whereas those with lower score enriched estrogen response pathway related set (FDR = 0.13) in TCGA and it was validated with METABRIC cohorts. We also found that the score associated with distant metastasis to brain or lung (p < 0.001). Kaplan-Meier analyses of site-specific metastasis-free interval demonstrated that for all three cohorts, patients with high scores had significantly increased risk for development of metastasis to the lung (p≤0.02). This association of the score with metastasis risk was also seen for brain for two cohorts, and it was seen for bone in one cohort. Importantly, we found that the high score group had a higher pCR rate than the low score group for especially ER+/HER2- breast cancer in two cohorts (p = 0.003 and p = 0.004). And the score significantly decreases with good response to neoadjuvant chemotherapy (p = 0.022) or hormonal therapy (p = 0.013). Conclusions: This novel gene score may be clinically useful with both prognostic and predictive values for breast cancer.