Abstract P1-10-26: Gene expression levels of DTX3, CACNA1G, IL11, ETV4 and TSPAN7 selected by LASSO penalty regression could predict pCR after neoadjuvant chemotherapy in breast cancer tumors

Abstract

Introduction: While pCR has been amply associated with improved treatment outcome in patients treated with NAC and surgery, there are only few studies that explore the molecular determinants of pCR. Given the discrete percentage of patients who achieve pCR after NAC, it is of upmost need to identify biomarkers that can discern between responder and non-responder patients before NAC and therefore determine the best therapeutic approach in advance. Our aim is to find the best combination of variables to predict pCR in our cohort. Patients: To evaluate if there is a gene expression profile that could predict pathological complete response (pCR) to NAC, we selected a cohort of 69 patients treated with neoadjuvant chemotherapy consisting of anthracyclines + taxanes; HER2+ patients were further treated with trastuzumab. The median age was 48 years, and 71% of patients were premenopausal. The median of mammographic tumor size was 3,7. Most patients (89%) had invasive ductal carcinoma, and the histological grade was mainly 2 (55,7% Grade 1+2; 44,3% Grade 3). ER and PR positivity were seen in 63% and 53% of cases respectively. HER2 overexpression was seen in 20% of tumors. Methods: Gene expression was evaluated in 770 genes involved in 13 canonical cancer pathways as screened with the nCounter® PanCancer Pathways panel in FFPE pre-treatment tumor biopsies and analyzed with the nSolverTM software. The collected clinical and pathologic variables were age, mammographic tumor size, nodal status, tumor grade, HR status, HER2-overexpression, Ki67 percentage, IHQ subtype and PAM50 subtype. Finally, LASSO regression was used to predict pCR based on normalized gene expression levels and clinical and pathological data. Results: Out of the 69 enrolled patients, 27 achieved pCR (39%) and 42 did not (determined by Miller and Payne criteria). High-quality RNA was isolated from each patient for doing Nanostring nCounter experiment. Based on the gene expression levels of the PanCancer Pathways panel, Lasso regression identified a subset of genes as predictive of pCR. Therefore, the combination of expression levels of DTX3, CACNA1G, IL11, ETV4 and TSPAN7 genes, predicted pCR with an average 70% of Area Under the Curve ROC (AUC) on the test set in a cross-validation scheme. DTX3 is a member of Notch pathway, IL11 from JAK-STAT pathway, CACNA1G from MAPK pathway, ETV4 and TSPAN7 from Transcriptional misregulation pathway. DTX3, CACNA1G and TSPAN7 are down-regulated and IL11 and ETV4 are up-regulated in patients who achieved pCR. Regarding the clinical and pathological data, the best model for predict pCR obtained an 60% of AUC and the selected variables were age, mammographic tumor size, nodal status, tumor grade and Ki67 percentage. Conclusions: In our cohort, the combination of gene expression levels of DTX3, CACNA1G, IL11, ETV4 and TSPAN7 can predict tumor response to NAC even better than any clinical or pathological variable. Although validation in a separate cohort is necessary, our results indicate that this combination could constitute a predictive gene signature to discriminate resistant patients to NAC before the treatment in breast cancer tumors. Citation Format: Maria Rosario Chica-Parrado, Julio Montes-Torres, Cynthia Robles-Podadera, Martina Alvarez, Jose Jerez, Luis Vicioso, Lidia Pérez-Villa, Begoña Jimenez, Nuria Ribelles, Ana Godoy, Isabel Barragán-Mallofret, Emilio Alba. Gene expression levels of DTX3, CACNA1G, IL11, ETV4 and TSPAN7 selected by LASSO penalty regression could predict pCR after neoadjuvant chemotherapy in breast cancer tumors [abstract]. In: Proceedings of the 2019 San Antonio Breast Cancer Symposium; 2019 Dec 10-14; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2020;80(4 Suppl):Abstract nr P1-10-26.