PS915 HIGH‐THROUGHPUT IMAGING ANALYSIS FOR DRUG COMBINATION EFFICACY AND SPECIFICITY IN CHILDHOOD ACUTE LYMPHOBLASTIC LEUKAEMIA

Abstract

Background:Childhood acute lymphoblastic leukaemia (ALL) responds well to conventional treatment, however there is a need to find more targeted therapies without the associated toxicities. There is great inconsistency of drug responses in cell lines, and often these do not represent the outcomes observed in the clinic. Patient‐derived xenografts (PDXs) have been shown to more closely represent patient cancers, and increasingly better demonstrate the drug response clinically observed. However, in vivo assessment of multiple drug combinations is costly and time consuming. The development of an ex vivo ALL patient‐derived xenograft/mesenchymal stromal cell (MSC) co‐culture platform has allowed for screening of novel drug combinations in multiple childhood ALL cytogenetic subtypes.Aims:The aim of this study was to develop a high‐throughput whole‐well image‐based analysis of PDX and MSC cell numbers after exposure to double combinations of 4 drugs. A panel of PDXs were assessed to take into account heterogeneity in pharmacological response.Methods:A previously reported PDX/MSC platform was utilised to assess double drug combinations of dexamethasone, palbociclib, idasanutlin and ABT‐199 in a panel of childhood ALL PDXs. Thawed PDXs were seeded on MSCs and treated with drug combinations in a 7x7 matrix format for 96 h. These were subsequently stained with a fluorescent DNA dye, and high‐throughput microscopy was conducted to obtain whole‐well images of PDX and MSC nuclei. Images were segmented and measurements of nuclear intensity, size and shape were produced with CellProfiler software. Segmented images were used to train a random forest classifier, within CellProfiler Analyst software, to distinguish PDX and MSC nuclei, utilising a multitude of image features. Combination treatments were assessed for synergy with the zero interaction potency model (ZIP).Results:Whole‐well cell numbers were achieved for PDXs and MSCs in the context of drug combinations across a range of ALL cytogenetic subgroups. Comparing fluorescence intensity readings with imaging and cell scoring revealed a background fluorescence level that previously could not be removed, leading to the underestimation of drug response. Combination of dexamethasone and ABT‐199, previously shown to be additive or synergistic was also confirmed across our panel. Cell numbers in control wells were consistently equal to or higher than the original seeding densities, demonstrating the capacity of the platform to support PDX proliferation. Interestingly, the CDK4/6 inhibitor palbociclib scored strongly across many of the PDXs, and even more so in combination with ABT‐199: something which has not been previously observed. Particularly strong synergy of ABT‐199 and idasanutlin was observed in a PDX with a t(9;22) translocation (max ZIP score 69). None of the double combinations tested had an effect on MSC cell number.Summary/Conclusion:This platform allows accurate quantification of absolute PDX cell number and relative MSC cell number. DNA stains have previously been shown to be superior to many standard viability assays such as fluorescent metabolic readouts. This methodology goes further to enable stratification of different cell types. The effective combination of dexamethasone and ABT‐199 in childhood ALL, previously seen in mice, was confirmed with this platform, and several novel combinations were identified which appeared to show synergy, and will be taken forward to in vivo testing. This image‐based platform is now being expanded to increase the complexity of the co‐culture model with other haematopoietic niche cells.