对患者细胞进行高内涵成像以识别可预测药物反应的CLL和AML患者队列

Biomarkers that predict therapy response greatly facilitate applying precision medicine to patient treatment decisions. However, within populations of Chronic Lymphocytic Leukemia (CLL) and Acute Myelogenous Leukemia (AML) patients there is heterogeneity that is inherent to the disease and also between patients. This heterogeneity, obscures conventional potential biomarkers. As an alternative, we are applying confocal microscopy of live primary patient samples in 2D and 3D microenvironment models that mimic the bone marrow niche to identify cellular phenotypes that can be used as alternative biomarkers. For CLL, live cell painting using non-toxic dyes of cells from 133 patient samples that were grown in 2D niche mimetic cultures enabled machine learning from images. Feature reduction followed by unbiased image clustering revealed five cohorts of patients each with unique drug responses. The results of these studies suggest that high content imaging combined with machine learning and automated image analysis can be used to predict drug responses in a patient specific manner. For AML live cell painting revealed that a novel 3D microenvironmental model was required to inhibit differentiation and enable monitoring the growth of cells from bone marrow aspirates. We are currently applying machine learning to micrographs of AML patient samples stained for CD34 and CD45 to identify putative cancer stem cells within the 3D cultures and for image analysis of staining with a nuclear dye and Annexin V to assess cell-type specific responses to drugs targeting apoptosis proteins. Our results suggest that cellular phenotyping by high content imaging of patient cells grown in microenvironmental models may provide the biomarkers needed to enable precision patient treatment.