BACKGROUND: Pathological response to neoadjuvant chemotherapy (NAC) is critical in prognosis and selection of systemic treatments for patients with triple-negative breast cancer (TNBC). The aim of this study is to identify gene expression-based markers to predict response to NAC.

PATIENTS AND METHODS: A survey of 43 publicly available gene expression datasets was performed. We identified a cohort of TNBC patients treated with NAC (n = 708). Gene expression data from different studies were renormalized, and the differences between pretreatment (pre-NAC), on-treatment (post-C1), and surgical (Sx) specimens were evaluated. Euclidean statistical distances were calculated to estimate changes in gene expression patterns induced by NAC. Hierarchical clustering and pathway enrichment analyses were used to characterize relationships between differentially expressed genes and affected gene pathways. Machine learning was employed to refine a gene expression signature with the potential to predict response to NAC.

RESULTS: Forty nine genes consistently affected by NAC were involved in enhanced regulation of wound response, chemokine release, cell division, and decreased programmed cell death in residual invasive disease. The statistical distances between pre-NAC and post-C1 significantly predicted pathological complete response [area under the curve (AUC) = 0.75; p = 0.003; 95% confidence interval (CI) 0.58-0.92]. Finally, the expression of CCND1, a cyclin that forms complexes with CDK4/6 to promote the cell cycle, was the most informative feature in pre-NAC biopsies to predict response to NAC.

CONCLUSIONS: The results of this study reveal significant transcriptomic changes induced by NAC and suggest that chemotherapy-induced gene expression changes observed early in therapy may be good predictors of response to NAC.