Autophagy In Breast Cancer Patients With Residual Tumors After Neoadjuvant Chemotherapy And Its Correlation With Patients Outcome

Objective:To evaluate autophagy in breast cancer patients with residual tumors after neoadjuvant chemotherapy through immunohistochemical assays for LC3B and Beclin-1protein, and determine whether the level of autophagy had correlation to patients outcome.Methods:We carried out an immunohistochemical assay for Ki-67, LC3B, Beclin-1markers on formalin-fixed invasive carcinoma samples from patients who were diagnosed between Jan.2003and Dec.2006in Shanghai Cancer Center, all of these samples were residual tumors after neoadjuvant chemotherapy for breast cancer. Variables at baseline and at surgery were investigated.Results:120patients considered as non-pCR responders were included in this study.35.8%were considered as LC3B positive among all these patients. Beclin-1and Ki-67were all positively correlated to LC3B (P<0.001,P=0.033). The LC3B positivity was observed in25.0%,18.2%,58.8%, and47.2%of patients in luminal-A, luminal-B,HER-2and triple-negative patients, respectively. The expression of LC3B was siginificantly higher in HER-2and TN subtypes (P=0.017). Combining clinical and pathological features of all patients, expression of LC3B was not correlated to clincal response of neoadjuvant chemotherapy (P=0.624). However, LC3B-positive patient had a relatively worse prognosis than LC3B-negative patients. The5-year DFS were47.0%and75.4%, respectively (P<0.001). And the5-year OS were62.3%and85.3%, respectively (P=0.001). In multivariate Cox regression analysis, lymph node metastasis (P<0.001), HER-2(P=0.028), Ki-67(P=0.015), LC3B (P<0.001) were proved to be independent predictors of DFS, whereas the independent predictors of OS were LC3B (P=0.001; and lymph node metastasis (P=0.033). Based on these variables, we built a prognostic nomogram to predict5-year DFS of non-pCR patients with a C-index of0.768. Conclusion:Autophagy in residual breast cancer after neoadjuvant chemotherpay was correlated to patients prognosis. Our prognostic model could easily discriminate patients with different risks of experiencing an event or death, which could allow physicians to tailor treatment strategies specifically and individually.