G. Center-Driven Research Project: Genomic signature-based screening in robust HTSformatsG.1. IntroductionWe propose to develop general and robust HTS methods that enable the discovery of small molecules thatswitch, for example, disease states to healthy states without requiring knowledge of the relevant cellulartarget(s) in advance of the screen. These methods require the ability to perform multiple measurements perwell that together define signatures of the relevant states.Small-molecule, high-throughput screens often entail a single measurement per well, for example, the activityof a singular, purified enzyme. In principle and in practice, the ability to perform multiple measurements perwell provides novel scientific insights, as demonstrated through high-content expression- and image-basedscreening. To illustrate: 1) cells treated with small molecules have been probed with Luminex beads todetermine the relative amounts of RNAs (mRNA, miRNA)13'43, 2) multiple cellular features have beenmeasured and selected as state classifiers using image-analysis software developed by researchers at theBroad Institute38'39, and 3) protein microarray and Luminex bead-based methods have been adapted to theanalyses of phosphoproteins in cells65'66. Such capabilities underlie signature-based state-switching screens,which permit the probing of biological and disease circuitries in order to discover small molecules able toswitch one state to the other. (Methods for making multiple measurements can also permit extremely efficientscreens where each of the individual measurements is of particular, but not necessarily related, importance,e.g., where the many RNAs being measured each represent a singular biological or therapeutic probe.)Important understandings in biology have already been gained from the two most advanced and informativemethods for multiplexed measurements in screens at the Broad Institute: gene expression-based screeningand automatic scoring of complex cellular phenotypes by machine learning from multiple features of cells. Inthe former case, for example, one clinical trial was initiated based on the discovery that sirolimus converts adexamethasone-resistant state of leukemia (ALL)- HRG-bl +HRG-D1 ce||Sj jdentjfjec| by its gene-expression signature, to adexamethasone-sensitive state, and a second clinicaltrial involving patients with relapsed or refractoryAMI was initiated based on the discovery thatgefitinib and erlotonib induced differentiation of acutemyeloid leukemia cells13. In the case of imagedfeatures of cells, for example, the discovery ofmodulators of heregulin (HRG)-mediated, ErbB2-dependent filopodia extension, an example of acomplex phenotype, was achieved based on supervised selection of imaged cellular features of the non and heregulin-teated cell states (Figure 13).While the capability to make two such multiplexed measurements already exists at the Broad Institute, it is notyet possible to do so in a fully automated high-throughput format. Until that challenge is met, the full potentialto exploit the novel technologies and to share them with the MLPCN and other labs cannot be realized. Inaddition, the continuing evolution of technologies like single molecule-based measurements provide theopportunity for continued process improvement through greater sensitivity and resolution. Finally, othermethods are being developed that have not yet reached a stage of throughput sufficient for use in smallmoleculescreens, but that offer considerable promise in terms of the unique insights they provide into cellcircuitry and the relationship of small-molecule structure to protein binding. Two examples of such technologiesunder development currently at the Broad Institute are: 1) chromatin immunoprecipitation followed byexhaustive (Solexa-based) DNA sequencing, and 2) the use of stable isotope labeling of cells (SILAC) andMS-based proteomics to determine rank-ordered small-molecule/protein interactions in cells treated with smallmolecules (unpublished results, Steve Carr, Stuart Schreiber and colleagues). While we propose here initiallyto undertake two specific signature-based screening techniques and to convert them to robust HTS formats,we will also explore other promising new areas and develop a process, in consultation with MLPCN leadership,for determining whether they merit effort as future Center-driven research projects.Two biological systems illustrate how signature-based screening might be used in the future. As describedearlier (Section B), we have developed a 384-well format for culturing human primary pancreatic islets havingfunctional endocrine cells. We have also cultured individual cell types from these islets. If signatures of theindividual endocrine cell types could be identified, small molecules, including ones targeted to chromatinmodifyingenzymes, would be screened for their ability to convert non-beta cells to beta cells. Earlier (SectionB), we also described our ability to co-culture primary hematopoietic or leukemic stem cells with stromal (e.g.,osteoblast) cells. Signature-based screens would be of great value for discovering small molecules thatselectively alter the developmental states of these individual cell types.The objective of the Broad Center-Driven Research Project is to convert promising or existing (pilot stage)methods into robust, fully automated HTS methods.
The specific aims are: To advance benchmark systems in gene expression- and image-based signature screening from theircurrent pilot stage to a mature, HTS stage defined by the ability to screen the complete MLPCN smallmoleculecollection. To develop tools for the comparative analysis of signature-based, multidimensional HTS datasets.To incubate nascent yet promising methods for multiplexed measurements yielding signatures of cellstates that are well suited for HTS and complementary to the first two systems. To make these capabilities available to the MLPCN community, initially by advancing them to theProduction Facility of the BCSC.We envision that this Center-Driven Research Project will impact several facets of the MLPCN. Used inprimary screens, these methods offer powerful new discovery capabilities, giving researchers the opportunityto probe expression and image-based biological outcomes, neither of which can be used routinely in HTS. Theability to compare directly the information obtained from such methods will enable evaluation of multiplex assaymethods and help determine priorities for future development. They can also facilitate probe development bypermitting richer and more informative structure/activity relationships to be obtained (mechanism-associatedsignatures), ones based on multiple measurements that provide greater assurance that structural variants arestaying 'on mechanism' while showing improved cell-based-selectivity and potency. Potential off-target effectsthat would otherwise be 'silent' in single-measurement short-term cell-based systems could then be identifiedat the earliest stages of probe development. On-mechanism signatures provide a systematic means to identifyand to validate biomarkers of value in small-molecule experiments involving organ cultures and even animals.Looking ahead, the capacity and capability to perform multi-dimensional data analysis will help elucidatecomplex pathway inter-relationships, provide a means for comprehensively probing the basis for clinicallyobserved drug resistance, and enable the design of screens for generating disease-relevant cell states thatcan be used in an HTS mode to systematically discover novel, multiple-drug therapies.
Showing the most recent 10 out of 46 publications
