High-Content Analysis of Breast Cancer Using Single-Cell Deep Transfer Learning

High-content analysis has revolutionized cancer drug discovery by identifying substances that alter the phenotype of a cell, which prevents tumor growth and metastasis. The high-resolution biofluorescence images from assays allow precise quantitative measures enabling the distinction of small molecules of a host cell from a tumor. In this work, we are particularly interested in the application of deep neural networks (DNNs), a cutting-edge machine learning method, to the classification of compounds in chemical mechanisms of action (MOAs). Compound classification has been performed using image-based profiling methods sometimes combined with feature reduction methods such as principal component analysis or factor analysis. In this article, we map the input features of each cell to a particular MOA class without using any treatment-level profiles or feature reduction methods. To the best of our knowledge, this is the first application of DNN in this domain, leveraging single-cell information. Furthermore, we use deep transfer learning (DTL) to alleviate the intensive and computational demanding effort of searching the huge parameter's space of a DNN. Results show that using this approach, we obtain a 30% speedup and a 2% accuracy improvement.

Response of high-risk of recurrence/progression bladder tumours expressing sialyl-Tn and sialyl-6-T to BCG immunotherapy

High risk of recurrence/progression bladder tumours is treated with Bacillus Calmette-Guérin (BCG) immunotherapy after complete resection of the tumour. Approximately 75% of these tumours express the uncommon carbohydrate antigen sialyl-Tn (Tn), a surrogate biomarker of tumour aggressiveness. Such changes in the glycosylation of cell-surface proteins influence tumour microenvironment and immune responses that may modulate treatment outcome and the course of disease. The aim of this work is to determine the efficiency of BCG immunotherapy against tumours expressing sTn and sTn-related antigen sialyl-6-T (s6T). METHODS: In a retrospective design, 94 tumours from patients treated with BCG were screened for sTn and s6T expression. In vitro studies were conducted to determine the interaction of BCG with high-grade bladder cancer cell line overexpressing sTn. RESULTS: From the 94 cases evaluated, 36 had recurrence after BCG treatment (38.3%). Treatment outcome was influenced by age over 65 years (HR=2.668; (1.344-5.254); P=0.005), maintenance schedule (HR=0.480; (0.246-0.936); P=0.031) and multifocality (HR=2.065; (1.033-4.126); P=0.040). sTn or s6T expression was associated with BCG response (P=0.024; P<0.0001) and with increased recurrence-free survival (P=0.001). Multivariate analyses showed that sTn and/or s6T were independent predictive markers of recurrence after BCG immunotherapy (HR=0.296; (0.148-0.594); P=0.001). In vitro studies demonstrated higher adhesion and internalisation of the bacillus to cells expressing sTn, promoting cell death. CONCLUSION: s6T is described for the first time in bladder tumours. Our data strongly suggest that BCG immunotherapy is efficient against sTn- and s6T-positive tumours. Furthermore, sTn and s6T expression are independent predictive markers of BCG treatment response and may be useful in the identification of patients who could benefit more from this immunotherapy.

The predominance of M2-polarized macrophages in the stroma of low-hypoxic bladder tumors is associated with BCG immunotherapy failure

OBJECTIVE: Bacillus Calmette-Guérin (BCG) immunotherapy is the gold standard treatment for superficial bladder tumors with intermediate/high risk of recurrence or progression. However, approximately 30% of patients fail to respond to the treatment. Effective BCG therapy needs precise activation of the type 1 helper cells immune pathway. Tumor-associated macrophages (TAMs) often assume an immunoregulatory M2 phenotype and may directly interfere with the BCG-induced antitumor immune response. Thus, we aim to clarify the influence of TAMs, in particular of the M2 phenotype in stroma and tumor areas, in BCG treatment outcome. PATIENTS AND METHODS: The study included 99 patients with bladder cancer treated with BCG. Tumors resected before treatment were evaluated using immunohistochemistry for CD68 and CD163 antigens, which identify a lineage macrophage marker and a M2-polarized specific cell surface receptor, respectively. CD68+ and CD163+ macrophages were evaluated within the stroma and tumor areas, and high density of infiltrating cells spots were selected for counting. Hypoxia, an event known to modulate macrophage phenotype, was also assessed through hypoxia induced factor (HIF)-1α expression. RESULTS: Patients in whom BCG failed had high stroma-predominant CD163+ macrophage counts (high stroma but low tumor CD163+ macrophages counts) when compared with the ones with a successful treatment (71% vs. 47%, P = 0.017). Furthermore, patients presenting this phenotype showed decreased recurrence-free survival (log rank, P = 0.008) and a clear 2-fold increased risk of BCG treatment failure was observed in univariate analysis (hazard ratio = 2.343; 95% CI: 1.197-4.587; P = 0.013). Even when adjusted for potential confounders, such as age and therapeutic scheme, multivariate analysis revealed 2.6-fold increased risk of recurrence (hazard ratio = 2.627; 95% CI: 1.340-5.150; P = 0.005). High stroma-predominant CD163+ macrophage counts were also associated with low expression of HIF-1α in tumor areas, whereas high counts of CD163+ in the tumor presented high expression of HIF-1α in tumor nests. CONCLUSIONS: TAMs evaluation using CD163 is a good indicator of BCG treatment failure. Moreover, elevated infiltration of CD163+ macrophages, predominantly in stroma areas but not in the tumor, may be a useful indicator of BCG treatment outcome, possibly owing to its immunosuppressive phenotype.