Search

Image-based profiling has emerged as a powerful technology for various steps in basic biological and pharmaceutical discovery, but the community has lacked a large, public reference set of data from chemical and genetic perturbations. Here we present data generated by the Joint Undertaking for Morphological Profiling (JUMP)-Cell Painting Consortium, a collaboration between 10 pharmaceutical companies, six supporting technology companies, and two non-profit partners. When completed, the dataset will contain images and profiles from the Cell Painting assay for over 116,750 unique compounds, over-expression of 12,602 genes, and knockout of 7,975 genes using CRISPR-Cas9, all in human osteosarcoma cells (U2OS). The dataset is estimated to be 115 TB in size and capturing 1.6 billion cells and their single-cell profiles. File quality control and upload is underway and will be completed over the coming months at the Cell Painting Gallery:https://registry.opendata.aws/cellpainting-gallery. A portal to visualize a subset of the data is available athttps://phenaid.ardigen.com/jumpcpexplorer/.

Fluorescent-based microscopy screens carry a broad range of phenotypic information about how compounds affect cellular biology. From changes in cellular morphology observed in these screens, one key area of medicinal interest is determining a compound’s mechanism of action. However, much of this phenotypic information is subtle and difficult to quantify. Hence, creating quantitative embeddings that can measure cellular response to compound perturbation has been a key area of research. Here we present a deep learning enabled encoder called MOAProfiler that captures phenotypic features for determining mechanism of action from Cell Painting images. We compared our method with both a traditional computer vision means of feature encoding via CellProfiler and a deep learning encoder called DeepProfiler. The results, on two independent and biologically different datasets, indicated that MOAProfiler encoded MOA-specific features that allowed for more accurate clustering and classification of compounds over hundreds of different MOAs.

Metabolic alterations in cancer represent convergent effects of oncogenic mutations. We hypothesized that a metabolism-restricted genetic screen, comparing normal primary mouse hematopoietic cells and their malignant counterparts in an ex vivo system mimicking the bone marrow microenvironment, would define distinctive vulnerabilities in acute myeloid leukemia (AML). Leukemic cells, but not their normal myeloid counterparts, depended on the aldehyde dehydrogenase 3a2 (Aldh3a2) enzyme that oxidizes long-chain aliphatic aldehydes to prevent cellular oxidative damage. Aldehydes are by-products of increased oxidative phosphorylation and nucleotide synthesis in cancer and are generated from lipid peroxides underlying the non–caspase-dependent form of cell death, ferroptosis. Leukemic cell dependence on Aldh3a2 was seen across multiple mouse and human myeloid leukemias. Aldh3a2 inhibition was synthetically lethal with glutathione peroxidase-4 (GPX4) inhibition; GPX4 inhibition is a known trigger of ferroptosis that by itself minimally affects AML cells. Inhibiting Aldh3a2 provides a therapeutic opportunity and a unique synthetic lethality to exploit the distinctive metabolic state of malignant cells.

Deep profiling of cell states can provide a broad picture of biological changes that occur in disease, mutation, or in response to drug or chemical treatments. Morphological and gene expression profiling, for example, can cost-effectively capture thousands of features in thousands of samples across perturbations, but it is unclear to what extent the two modalities capture overlapping versus complementary mechanistic information. Here, using both the L1000 and Cell Painting assays to profile gene expression and cell morphology, respectively, we perturb A549 lung cancer cells with 1,327 small molecules from the Drug Repurposing Hub across six doses. We determine that the two assays capture some shared and some complementary information in mapping cell state. We find that as compared to L1000, Cell Painting captures a higher proportion of reproducible compounds and has more diverse samples, but measures fewer distinct groups of features. In an unsupervised analysis, Cell Painting grouped more compound mechanisms of action (MOA) whereas in a supervised deep learning analysis, L1000 predicted more MOAs. In general, the two assays together provide a complementary view of drug mechanisms for follow up analyses. Our analyses answer fundamental biological questions comparing the two biological modalities and, given the numerous applications of profiling in biology, provide guidance for planning experiments that profile cells for detecting distinct cell types, disease phenotypes, and response to chemical or genetic perturbations.

Stem cells respond to the physicochemical parameters of the substrate on which they grow. Quantitative material activity relationships - the relationships between substrate parameters and the phenotypes they induce - have so far poorly predicted the success of bioactive implant surfaces. In this report, we screened a library of randomly selected designed surface topographies for those inducing osteogenic differentiation of bone marrow-derived mesenchymal stem cells. Cell shape features, surface design parameters, and osteogenic marker expression were strongly correlated in vitro. Furthermore, the surfaces with the highest osteogenic potential in vitro also demonstrated their osteogenic effect in vivo: these indeed strongly enhanced bone bonding in a rabbit femur model. Our work shows that by giving stem cells specific physicochemical parameters through designed surface topographies, differentiation of these cells can be dictated. (C) 2017 Elsevier Ltd. All rights reserved.

Liver disease is a leading cause of morbidity worldwide and treatment options are limited, with organ transplantation being the only form of definitive management. Cell-based therapies have long held promise as alternatives to whole-organ transplantation but have been hindered by the rapid loss of liver-specific functions over a period of days in cultured hepatocytes. Hypothesis-driven studies have identified a handful of factors that modulate hepatocyte functions in vitro, but our understanding of the mechanisms involved remains incomplete. We thus report here the development of a high-throughput platform to enable systematic interrogation of liver biology in vitro. The platform is currently configured to enable genetic knockdown screens and includes an enzyme-linked immunosorbent assay-based functional assay to quantify albumin output as a surrogate marker for hepatocyte synthetic functions as well as an image-based viability assay that counts hepatocyte nuclei. Using this platform, we identified 12 gene products that may be important for hepatocyte viability and/or liver identity in vitro. These results represent important first steps in the elucidation of mechanisms instrumental to the phenotypic maintenance of hepatocytes in vitro, and we hope that the tools reported here will empower additional studies in various fields of liver research.

Cell-based therapies hold the potential to alleviate the growing burden of liver diseases. Such therapies require human hepatocytes, which, within the stromal context of the liver, are capable of many rounds of replication. However, this ability is lost ex vivo, and human hepatocyte sourcing has limited many fields of research for decades. Here we developed a high-throughput screening platform for primary human hepatocytes to identify small molecules in two different classes that can be used to generate renewable sources of functional human hepatocytes. The first class induced functional proliferation of primary human hepatocytes in vitro. The second class enhanced hepatocyte functions and promoted the differentiation of induced pluripotent stem cell-derived hepatocytes toward a more mature phenotype than what was previously obtainable. The identification of these small molecules can help address a major challenge affecting many facets of liver research and may lead to the development of new therapeutics for liver diseases.

A fully automated and robust method was developed to quantify β-III-tubulin-stained retinal ganglion cells, combining computational recognition of individual cells by CellProfiler and a machine-learning tool to teach phenotypic classification of the retinal ganglion cells by CellProfiler Analyst. In animal models of glaucoma, quantification of immunolabeled retinal ganglion cells is currently performed manually and remains time-consuming. Using this automated method, quantifications of retinal ganglion cell images were accelerated tenfold: 1800 images were counted in 3 hours using our automated method, while manual counting of the same images took 72 hours. This new method was validated in an established murine model of microbead-induced optic neuropathy. The use of the publicly available software and the method’s user-friendly design allows this technique to be easily implemented in any laboratory.

Postmortem, genetic, brain imaging, and peripheral cell studies all support decreased mitochondrial activity as a factor in the manifestation of Bipolar Disorder (BD). Because abnormal mitochondrial morphology is often linked to altered energy metabolism, we investigated whether changes in mitochondrial structure were present in brain and peripheral cells of patients with BD. Mitochondria from patients with BD exhibited size and distributional abnormalities compared with psychiatrically-healthy age-matched controls. Specifically, in brain, individual mitochondria profiles had significantly smaller areas, on average, in BD samples (P = 0.03). In peripheral cells, mitochondria in BD samples were concentrated proportionately more within the perinuclear region than in distal processes (P = 0.0008). These mitochondrial changes did not appear to be correlated with exposure to lithium. Also, these abnormalities in brain and peripheral cells were independent of substantial changes in the actin or tubulin cytoskeleton with which mitochondria interact. The observed changes in mitochondrial size and distribution may be linked to energy deficits and, therefore, may have consequences for cell plasticity, resilience, and survival in patients with BD, especially in brain, which has a high-energy requirement. The findings may have implications for diagnosis, if they are specific to BD, and for treatment, if they provide clues as to the underlying pathophysiology of BD.

The typical “design” approach to image-based assay development involves choosing measurements that are likely to correlate with the phenotype of interest, based on the researcher’s intuition and knowledge of image analysis. An alternate “screening” approach is to measure a large number of cellular features and systematically test each feature to identify those that are best able to distinguish positive and negative controls while taking precautions to avoid overfitting the available data. The cell measurement software the authors previously developed, CellProfiler, makes both approaches straightforward, easing the process of assay development. Here, they demonstrate the use of the screening approach to image assay development to select the best measures for scoring publicly available image sets of 2 cytoplasm-to-nucleus translocation assays and 2 Transfluor assays. The authors present the resulting assay quality measures as a baseline for future algorithm comparisons, and all software, methods, and images they present are freely available.

Biologists increasingly use co-culture systems in which two or more cell types are grown in cell culture together in order to better model cells’ native microenvironments. Co-cultures are often required for cell survival or proliferation, or to maintain physiological functioning in vitro. Having two cell types co-exist in culture, however, poses several challenges, including difficulties distinguishing the two populations during analysis using automated image analysis algorithms. We previously analyzed co-cultured primary human hepatocytes and mouse fibroblasts in a high-throughput image-based chemical screen, using a combination of segmentation, measurement, and subsequent machine learning to score each cell as hepatocyte or fibroblast. While this approach was successful in counting hepatocytes for primary screening, segmentation of the fibroblast nuclei was less accurate. Here, we present an improved approach that more accurately identifies both cell types. Pixel-based machine learning (using the software ilastik) is used to seed segmentation of each cell type individually (using the software CellProfiler). This streamlined and accurate workflow can be carried out using freely available and open source software., National Science Foundation (U.S.) (NSF CAREER DBI 1148823), National Institutes of Health (U.S.) (NIH UH3 EB017103)

The perception and control of self-movement relies on visual cues derived from the radial patterns of optic flow and from the relative motion of objects within view. Optic flow and object motion processing impairments might limit independent self-movement in a manner like that seen in ageing and in Alzheimer's disease. We used optic flow and object motion stimuli to simulate aspects of the self-movement scene. Stimulus salience was individualized to present comparable stimuli to young [n = 18; mean age = 25.6, standard error of measurement (SEM) = 1.4], middle-aged (n = 17; mean age = 53.9, SEM = 0.9), older adult (n = 30; mean age = 72.4, SEM = 1.4) and Alzheimer's disease (n = 15; mean age 75.2, SEM = 1.6) subjects. All groups were tested in two tasks: pointing towards the simulated direction of self-movement and steering the simulated self-movement towards a straight-ahead direction. We found that young and middle-aged subjects show similar pointing accuracy using either optic flow or object motion, but steer better with object motion than with optic flow. Older adult subjects show better performance with optic flow than object cues for pointing (P < 0.001), but their performance improves when both cues are combined in the pointing (P = 0.012) and steering (P = 0.02) tasks. Alzheimer's disease patients show poorer performance with optic flow and object motion than all other groups and do not benefit from the combined presentation of cues for either pointing or steering. We conclude that ageing and Alzheimer's disease are associated with distinct profiles of visual processing deficits that limit the ability to use optic flow and object motion to perceive and control self-movement.

As arboreal primates move through the jungle, they are immersed in visual motion that they must distinguish from the movement of predators and prey. We recorded dorsal medial superior temporal (MSTd) cortical neuronal responses to visual motion stimuli simulating self-movement and object motion. MSTd neurons encode the heading of simulated self-movement in three-dimensional (3-D) space. 3-D heading responses can be evoked either by the large patterns of visual motion in optic flow or by the visual object motion seen when an observer passes an earth-fixed landmark. Responses to naturalistically combined optic flow and object motion depend on their relative directions: an object moving as part of the optic flow field has little effect on neuronal responses. In contrast, an object moving separately from the optic flow field has large effects, decreasing the amplitude of the population response and shifting the population's heading estimate to match the direction of object motion as the object moves toward central vision. These effects parallel those seen in human heading perception with minimal effects of objects moving with the optic flow and substantial effects of objects violating the optic flow. We conclude that MSTd can contribute to navigation by supporting 3-D heading estimation, potentially switching from optic flow to object cues when a moving object passes in front of the observer.

The formation of synapses, the specialized points of chemical communication between neurons, is a highly regulated developmental process fundamental to establishing normal brain circuitry. Perturbations of synapse formation and function causally contribute to human developmental and degenerative neuropsychiatric disorders, such as Alzheimer's disease, intellectual disability, and autism spectrum disorders. Many genes controlling synaptogenesis have been identified, but lack of facile experimental systems has made systematic discovery of regulators of synaptogenesis challenging. Thus, we created a high-throughput platform to study excitatory and inhibitory synapse development in primary neuronal cultures and used a lentiviral RNA interference library to identify novel regulators of synapse formation. This methodology is broadly applicable for high-throughput screening of genes and drugs that may rescue or improve synaptic dysfunction associated with cognitive function and neurological disorders., National Institutes of Health (U.S.) (MH095096), National Institutes of Health (U.S.) (R01 GM089652)

Efforts to develop more effective therapies for acute leukemia may benefit from high-throughput screening systems that reflect the complex physiology of the disease, including leukemia stem cells (LSCs) and supportive interactions with the bone marrow microenvironment. The therapeutic targeting of LSCs is challenging because LSCs are highly similar to normal hematopoietic stem and progenitor cells (HSPCs) and are protected by stromal cells in vivo. We screened 14,718 compounds in a leukemia-stroma co-culture system for inhibition of cobblestone formation, a cellular behavior associated with stem-cell function. Among those compounds that inhibited malignant cells but spared HSPCs was the cholesterol-lowering drug lovastatin. Lovastatin showed anti-LSC activity in vitro and in an in vivo bone marrow transplantation model. Mechanistic studies demonstrated that the effect was on target, via inhibition of HMG-CoA reductase. These results illustrate the power of merging physiologically relevant models with high-throughput screening.

The Plasmodium liver stage is an attractive target for the development of antimalarial drugs and vaccines, as it provides an opportunity to interrupt the life cycle of the parasite at a critical early stage. However, targeting the liver stage has been difficult. Undoubtedly, a major barrier has been the lack of robust, reliable, and reproducible in vitro liver-stage cultures. Here, we establish the liver stages for both Plasmodium falciparum and Plasmodium vivax in a microscale human liver platform composed of cryopreserved, micropatterned human primary hepatocytes surrounded by supportive stromal cells. Using this system, we have successfully recapitulated the full liver stage of P. falciparum, including the release of infected merozoites and infection of overlaid erythrocytes, as well as the establishment of small forms in late liver stages of P. vivax. Finally, we validate the potential of this platform as a tool for medium-throughput antimalarial drug screening and vaccine development., Bill & Melinda Gates Foundation (51066)

The cellular content of mitochondria changes dynamically during development and in response to external stimuli, but the underlying mechanisms remain obscure. To systematically identify molecular probes and pathways that control mitochondrial abundance, we developed a high-throughput imaging assay that tracks both the per cell mitochondrial content and the cell size in confluent human umbilical vein endothelial cells. We screened 28,786 small molecules and observed that hundreds of small molecules are capable of increasing or decreasing the cellular content of mitochondria in a manner proportionate to cell size, revealing stereotyped control of these parameters. However, only a handful of compounds dissociate this relationship. We focus on one such compound, BRD6897, and demonstrate through secondary assays that it increases the cellular content of mitochondria as evidenced by fluorescence microscopy, mitochondrial protein content, and respiration, even after rigorous correction for cell size, cell volume, or total protein content. BRD6897 increases uncoupled respiration 1.6-fold in two different, non-dividing cell types. Based on electron microscopy, BRD6897 does not alter the percent of cytoplasmic area occupied by mitochondria, but instead, induces a striking increase in the electron density of existing mitochondria. The mechanism is independent of known transcriptional programs and is likely to be related to a blockade in the turnover of mitochondrial proteins. At present the molecular target of BRD6897 remains to be elucidated, but if identified, could reveal an important additional mechanism that governs mitochondrial biogenesis and turnover.

We assessed visual processing related to navigational impairment in Alzheimer's disease hypothesizing that visual motion evoked responses to optic flow simulating observer self-movement would be linked to navigational performance. Mild Alzheimer's disease and older adult control subjects underwent open-field navigational testing, visual motion perceptual threshold determination and a battery of neuropsychological examinations. We recorded visual motion evoked potentials (EPs) at occipital and parietal sites during centred visual fixation. Randomly moving or stationary pattern pre-stimuli preceded horizontal motion and radial optic flow stimuli to separate motion N200s from pattern onset responses. Radial optic flow evoked N200 responses comparable with those obtained with uniform horizontal motion, despite the variety of motion directions in radial optic flow. Alzheimer's disease patients showed smaller radial optic flow N200s than older adult subjects, and these were greatly diminished when preceded by stationary dots. Combining N200 amplitudes with optic flow perceptual thresholds and contrast sensitivities yielded a strong correlation with navigational impairment in Alzheimer's disease (R2 = 0.95). We conclude that navigational impairment in Alzheimer's disease is linked to a disorder of extrastriate visual cortical motion processing reflected in specific perceptual and neurophysiological measures.

Abstract 208 Altered metabolism has long been recognized as a signature finding of malignant cells with changes in glucose handling termed the ‘Warburg effect’. Now recognized as an adaptation of metabolic pathways to meet the distinctive demands of malignant cells, metabolism represents a potential area of opportunity for developing novel therapeutics. We established a system by which a wide spectrum of metabolic pathways could be examined for differential activity in tumor cells and their normal counterparts in the context of acute Myeloid Leukemia (AML). We studied metabolism using a biased shRNA library comprising 1000 shRNAs on a well-defined population of leukemia stem or initiating cells (LICs) in a model of AML driven by the MLL-AF9 oncogene (Krivtsov, 2006) and performed identical assays on the normal counterparts of LICs called normal granulocyte monocyte precursors (N-GMPs). The shRNAs chosen, targeted canonical ‘rate limiting enzymes’ as well as those enzymes which were found to be differentially expressed in LICs and N-GMPs by gene expression profiling. Both LICs and N-GMPs were maintained in the presence of murine bone marrow mesenchymal cells in order to mimic the native micro-environment in which these cells grow. To maximize physiologic relevance we used primary mouse LICs, N-GMPs and bone marrow mesenchymal cells, instead of cell lines, which may have distinct metabolic dependencies secondary to their adaption to culture conditions. Our screen has identified Aldehyde Dehydrogenase 3a2 (Aldh3a2 or Fatty Aldehyde Dehydrogenase, FALDH) as a key target gene. Aldh3a2 depletion in LICs with three independent validated shRNAs resulted in improvement in survival of transplanted mice (p value of 0.0004 to 0.0018) and significantly decreased growth of cells in methylcellulose soft agar assays and in co-culture with primary bone marrow stroma. In a knock-out mouse model of Aldh3a2, normal hematopoiesis as assessed by immunophenotype seems unperturbed. Using Tumorscape, the Broad Institute's database on gene copy number variation in tumors, we showed that Alh3a2 is significantly focally amplified (Q value =1.39−5) across the entire dataset of 3,131 tumors, including hematologic malignancies and is located within a focal peak region containing 11 other genes (chromosome 17:19117656–19932585). We have also probed a gene expression dataset of 436 samples obtained from two different multicenter clinical trials: the AMLSG HD98a trial (patients < 60 years of age) and the AMLSG HD98b trial (patients > 60 years of age). Analysis revealed significant overexpression of Aldh3a2 in MLL-AF9 samples compared to all other AML subtypes (p< 0.0006 by ANOVA). Patients in the highest quartile of Aldh3a2 expression had a significantly prolonged overall survival compared to those in all other quartiles (p=0.038). Whether this survival association is due to increased dependency on Aldh3a2 and sensitivity to chemotherapy is unclear, but an issue for future experiments. We have observed increased apoptosis, decreased cycling and altered ROS in L-GMPs with knock-down of Aldh3a2. Metabolomic experiments to identify the precise metabolites which are responsible for the differential sensitivity of LICs and normal hematopoietic cells to Aldh3a2 depletion are ongoing. Our work is aimed at confirming the metabolic dependency of leukemia on Aldh3a2 and dissecting the biology of this enzyme in the context of this disease. We hope this will lead to the development of novel therapeutic strategies and provide the groundwork for the further study of leukemia metabolism. Disclosures: No relevant conflicts of interest to declare.

Abstract 759 We hypothesized that metabolic differences between leukemia initiating cells and their normal counterparts represent a vulnerability in the leukemia initiating cell, which can be therapeutically exploited. To test this hypothesis, we used the MLL-AF9 acute myeloid leukemia (AML) model in mice. Actin-DsRed mouse bone marrow transduced with MLL-AF9 expressing retrovirus was used to produce serially transplantable leukemia. Leukemic granulocyte-monocyte precursors (L-GMPs), defined by others to be the leukemia initiating cells were flow sorted from secondary recipient mice and compared with normal GMPs (N-GMPs) from actin Ds-Red mice. Gene expression profiling, metabolomic profiling via liquid chromatography- mass spectrometry and an in vitro shRNA screen were used to identify metabolic pathways preferentially activated in leukemia initiating cells. Of 1574 defined metabolic enzymes, 44 were found to be differentially expressed between L-GMPs and their normal counterparts (N-GMPs). These together with 117 classic rate limiting enzymes were subjected to shRNA knockdown in vitro. Metabolomic profiling of both cell populations was used to corroborate findings from shRNA knockdowns. L-GMPs and N-GMPs were transduced with lentivirus expressing shRNAs of interest (5 shRNAs per gene) in a 384 well format, selected with puromycin and cultured for 72–96 hours in the presence of GFP-positive primary bone marrow stroma. The number of cells in each well at the end of this experiment was quantitated using an Image Xpress microscope. Genes, the knockdown of which by at least two independent shRNAs produced a two fold or more decrease in L-GMPs as compared to control wells and did not similarly decrease N-GMPs, were chosen for in vivo validation. Ten genes in the glycolysis pathway and TCA cycle, fatty acid metabolism and detoxification, and ketohexokinase were identified. Ketohexokinase, a rate-limiting enzyme in fructose metabolism was particularly potent and of interest given its potential to be exploited therapeutically. In vivo assessment of its relative ability to inhibit malignant versus normal hematopoietic cells is ongoing. These studies provide preliminary support for the hypothesis that specific metabolic circuits are differentially active in leukemia initiating cells in MLL-AF9 AML and may represent unique points of vulnerability that can be targeted therapeutically. Authors 1 and 2 contributed equally. Authors 3 and 4 contributed equally. Disclosures: No relevant conflicts of interest to declare.

Abstract 760 Recent insights into the molecular and cellular processes that drive leukemia have called attention to the limitations intrinsic to traditional drug discovery approaches. To date, the majority of cell-based functional screens have relied on probing cell lines in vitro in isolation to identify compounds that decrease cellular viability. The development of novel therapeutics with greater efficacy and decreased toxicity will require the identification of small molecules that selectively target leukemia stem cells (LSCs) within the context of their microenvironment, while sparing normal cells. We hypothesized that it would be possible to systematically identify LSC susceptibilities by modeling key elements of bone marrow niche interactions in high throughput format. We tested this hypothesis by creating and optimizing an assay in which primary murine stem cell-enriched leukemia cells are plated on bone marrow stromal cells in 384-well format, and examined by a high content image-based readout of cobblestoning, an in vitro morphological surrogate of cell health and self-renewal. AML cells cultured in this way maintained their ability to reinitiate disease in mice with as few as 100 cells. 14,720 small molecule probes across diverse chemical space were screened at 5uM in our assay. Retest screening was performed in the presence of two different bone marrow stromal types in parallel, OP9s and primary mesenchymal stem cells (MSCs). Greater than 60% of primary screen hits positively retested (dose response with IC50 at or below 5 μM) on both types of stroma. Compounds that inhibited leukemic cobblestoning merely by killing the stroma were identified by CellTiter-Glo viability analysis and excluded. Compounds that killed normal primary hematopoietic stem and progenitor cell inputs, as assessed by a related co-culture screen, were also excluded. Selectivity for leukemia over normal hematopoietic cells was additionally examined in vitro by comingling these cells on stroma within the same wells. Primary human CD34+ AML leukemia and normal CD34+ cord blood cells were also tested, by way of the 5 week cobblestone area forming cell (CAFC) assay. Additionally, preliminary studies of human AML cells pulse-treated with small molecules ex vivo, followed by in vivo transplantation, provided further evidence of potent leukemia kill across genotypes. A biologically complex functional approach to drug discovery, such as the novel method described here, has previously been thought impossible, due to presumed incompatibility with high throughput scale. We show that it is possible, and that it bears fruit in a first pilot screen. By these means, we discover small molecule perturbants that act selectively in the context of the microenvironment to kill LSCs while sparing stroma and normal hematopoietic cells. Some hits act cell autonomously, and some do not, as evidenced by observed leukemia kill when only the stromal support cells are treated prior to the plating of leukemia. Some hits are known, such as parthenolide and celastrol, and some are previously underappreciated, such as HMG-CoA reductase inhibition. Others are entirely new, and would not have been revealed by conventional approaches to therapeutic discovery. We therefore present a powerful new approach, and identify drug candidates with the potential to selectively target leukemia stem cells in clinical patients. Disclosures: No relevant conflicts of interest to declare.