ALBERT

Description: Genome wide association studies (GWAS) in multiple myeloma (MM), and other neoplasias, have provided important insights into candidate germline variations, which may influence the risk of an individual to develop a given cancer, experience adverse clinical outcomes or develop side effects after treatment. However, it has been typically challenging to pursue further mechanistic evaluation of all these germline variants. Consequently, the role of many such variants in myeloma biology and clinical behavior often remains to be elucidated. We reasoned that recently developed functional genomics platforms, such as the CRISPR/Cas9 gene editing methodology, could provide insight into the role of GWAS-derived germline variants in MM. Specifically, we hypothesized that at least some germline variants previously proposed to correlate with higher risk for development of MM or its adverse clinical outcome could reside in or be proximal to genes which influence the proliferation and survival of MM cells. To address this hypothesis, we examined the results from our genome-wide CRISPR/Cas9-based gene editing screens in 2 MM cell lines (MM1.S and RPMI-8226; using the GeCKOv2 library of single guide RNAs [sgRNAs]), as well as additional results from other in-house or publicly available genome-wide CRISPR/Cas9 gene editing studies in 50 cell lines from other hematologic malignancies and 8 different types of solid tumors. In this functional genomic dataset, we examined the performance of 50 genes located in close proximity to over 60 risk loci identified in 6 different previously published GWAS studies for MM. Many of these genes had plausible potential involvement in MM/tumor biology, given their participation in transcriptional control or epigenetic regulation (e.g. CBX7, ASXL2, LCOR, MED24, SMARCD3, POU5F1); immunoglobulin secretion in plasma cells (ELL2), cell-to-cell adhesion (e.g. CDH12, CDH13); DNA repair (e.g. POLQ). We determined whether sgRNAs against these genes exhibited statistically significant (for 3 or more sgRNAs/gene, FDR=0.05) depletion or enrichment among the MM or non-MM cell lines of our study. While 14 and 6 of these genes exhibited statistically significant depletion (FDR=0.05) of their sgRNAs (3 or more per gene) in RPMI8226 and MM.1S cells, respectively, almost all of these genes were not ranked within the top 2000 genes with the most pronounced sgRNA depletion (in terms of log2 fold change and number of depleted sgRNAs/gene) in either MM or non-MM cell lines. Similarly, statistically significant sgRNA enrichment was not observed for the overwhelming majority of the genes in question in MM or non-MM cell lines. Interestingly, however, several genes showed a statistically significant association with clinical outcome in at least one clinically annotated gene expression profiling dataset in MM (e.g. correlation of ELL2, CDH13 transcript levels with clinical outcome of bortezomib-treated MM patients). These results taken together suggest that the majority of genes identified through prior GWAS studies for MM risk or adverse clinical outcome in this disease may have modest, if any, impact of the proliferation or survival of MM cells, as well as many other types of non-MM tumor cells, in CRISPR/Cas9-based screens conducted in cell-autonomous assay systems. In turn, these observations imply that, if these candidate genes are validated to play important roles in the pathophysiology of MM cells in vivo, this may likely involve cell-nonautonomous roles of these genes, e.g. in regulating tumor cell interaction with non-malignant cells in the local microenvironment or immune evasion. Our results highlight the value of a previously underappreciated approach in integrating genome-wide CRISPR/Cas9 in vitro genomic results with GWAS studies, in order to more comprehensively examine the putative roles of candidate germline variants and their proximal genes in the pathophysiology of myeloma and other neoplasias. Disclosures No relevant conflicts of interest to declare.

Description: Multiple Myeloma (MM) remains an incurable malignancy in part because of an incomplete understanding of which genes are critically responsible for MM cell survival and proliferation. To address this unmet need, and building on our recent functional genomics studies with the CRISPR/Cas9 gene editing platform (ASH 2015; Int. MM Workshop, Rome 2015), we reasoned that quantification of sgRNA depletion in the absence of any treatment could identify genes essential for the survival or proliferation of MM cells and better define their role as candidate therapeutic targets. To this end, we transduced Cas9-expressing RPMI-8226 and MM.1S cells with the lentiviral genome-scale GeCKO pooled library of sgRNAs. After culture of these cell lines for 2, 6, 8 or 12 weeks without any treatment, we identified, based on next generation sequencing for the sgRNA sequences, genes with significantly depleted sgRNAs (4-6 sgRNAs/gene, 〉2-fold average depletion, FDR=0.05, based on MAGECK algorithm) in Cas9+ cells compared to their initial sgRNA plasmid pools, baseline cultures, or isogenic parental Cas9-negative cells. These results were confirmed for each cell line with a 2nd independent genome-wide analysis and with a focused sgRNA library containing a subset of candidates defined by the genome-wide analyses. We compared these results with data from our in-house or publicly available CRISPR/Cas9 gene editing studies, involving a total of 50 cell lines from other hematologic malignancies (leukemia, lymphoma) and from 8 different types of solid tumors. We identified 3 broad categories of essential genes in MM cells: a) core essential genes, with sgRNA depletion across the majority of MM and non-MM lines of our study, representing cellular processes critical for practically all lineages (e.g. genes involved in regulation of basic transcription factor complexes, ribosomal function, proteasome, spliceosome, structural proteins for mitochondria and other key organelles, et.c.); b) genes selectively essential for MM cell lines, but not for the overwhelming majority of leukemia, lymphoma or solid tumor cell lines; c) genes with a role in small subset(s) of cell lines, across diseases, which harbor defined genetic features correlating with this dependency. We integrated our CRISPR/Cas9-based data on MM-selective essential genes with a reanalysis of the Achilles Heel shRNA screen in MM and non-MM cell lines (10 and 493, respectively) of the Cell Line Encyclopedia Program (CCLE) program. We applied a series of statistical tests (e.g. Wilcoxon rank test or marker selection feature of GENE-E algorithm with 1000 permutation tests) to identify genes with a significantly lower rank in sgRNA or shRNA depletion in MM vs. non-MM cell lines, across different specific thresholds for fold change and statistical significance. We identified more than 50 high-value candidate target genes with preferential essentiality in MM, compared to non-MM cell lines of diverse lineages. Prominent examples of such MM-selective, essential genes included: transcription factors (e.g. IRF4, CCND2, MAF, NFKB1, NFKB2, RELA, RELB); otherNF-kB-related genes (e.g. IKBKB); PIM2 (but not PIM1 or PIM3 in this cell line panel); regulators of protein homeostasis, including diverse E2 and E3 ubiquitin ligases; and several other known or biologically-plausible dependencies which are under further evaluation. Many of these MM-selective dependencies exhibited significantly higher expression in MM, compared to non-MM cells, both in cell lines (based on the CCLE dataset) and patient-derived samples (comparison of Broad/MMRF vs. TCGA datasets, respectively). Notable observations of context-dependent essential genes include ARID1A in MM.1S cells (plausibly due to deficiency in its paralog ARID1B); and cases of both MM and non-MM cells with RAS mutations but lack of dependency on that gene. Targeting of lineage-specific dependencies (e.g. ER or AR in breast or prostate Ca, respectively) has provided major clinical benefit in some tumors; while context-specific dependencies are a cornerstone of molecularly-guided individualized treatments. Therefore, by identifying lineage- and context-dependent essential genes for MM, our integrated genome-wide CRISPR/Cas9 and shRNA analyses in molecularly annotated panels of MM vs. non-MM cell lines provide an attractive framework towards designing novel therapies for MM. Disclosures No relevant conflicts of interest to declare.

Description: The functional characterization of kinases in multiple myeloma (MM) cells has mainly involved the use of RNAi, a mechanistic approach inherently dissimilar to small molecule inhibitors that are applied in the clinic. Furthermore, in the bone marrow (BM) microenvironment, where MM cells primarily reside, BM stromal cells (BMSCs) and other nonmalignant cell populations can function as "accessory" cells and alter the vulnerability of MM cells to diverse targeted therapies, including kinase inhibitors. In order to characterize how the kinase dependencies of MM cells differ in the presence vs. absence of BMSCs, we evaluated 16 MM cell lines and their response, in monoculture vs. co-culture conditions, to a panel of 273 kinase inhibitors (100nM, 24-72 h exposure), which target a total of 43 known primary oncogenic targets. We observed that virtually all tested cell lines showed no response, in either monocultures or co-cultures, to inhibitors of c-met, ALK, Abl, EGFR superfamily members (EGFR, HER2), c-kit, PDGFR, Flt3, FAK, VEGFR and Syk, including cell lines with detectable transcript levels for the respective kinases at baseline or stroma-induced increase in co-cultures. These results suggest that the respective kinases do not represent major dependencies for MM cell lines in either the cell-autonomous state or in the context of interaction with stromal cells. Dual PI3K/mTOR inhibitors exhibited similarly potent anti-tumor activity in both monocultures and stromal co-cultures of essentially all MM cell lines tested; while mTOR inhibitors devoid of PI3K activity had examples (e.g. in U266 and KMM2 cells) of stroma-induced resistance, likely reflecting a role of PI3K signaling in mediating, at least in some MM cells, stroma-induced decrease in dependence to mTOR function. Inhibitors of IKK or JAK (e.g. ruxolitinib, baricitinib) exhibited higher activity against several MM cell lines in the presence compared to absence of BMSCs. This stroma-induced sensitization to these kinase inhibitors is consistent with the role of BMSC-derived cytokines or cell adhesion molecules in inducing in MM cells NF-kappaB transcriptional activation or JAK/STAT3 signaling. Interestingly, several other classes of kinase inhibitors, e.g. against Aurora, PLK, MEK, Akt, exhibited more heterogeneous stroma-induced effects on their activity (decrease against some cell lines vs. sensitization against others; without an obvious common pattern of these effects for each of these kinase inhibitors across different cell lines). These results suggest that the dependence of MM cells on the corresponding kinases and the signaling networks they regulate are not only subject to stroma-induced changes, but also exhibit a higher degree of cell type-dependent plasticity than previously appreciated. In further support of the complexity of these stoma-induced events, we also observed that BRAF inhibitors (e.g. vemurafenib, dabrafenib, PLX-4720, AZ-628, et.c.) induced at least modest increase in proliferation of several BRAF-wild-type MM cell lines cultured in the absence of stromal cells, but co-culture with BMSC blunted this effect in some cell lines (JJN3, AMO1, OPM2), and modestly enhanced it in others (e.g. KMS34). BRAF inhibitor-induced proliferation of BRAF-wild-type cell lines has been previously reported in different tumor types and has been attributed to activation and signaling through C-RAF: our current observations suggest that, in MM patients harboring both V600E-BRAF mutant and wild-type clones, the impact of treatment with BRAF inhibitor on decreasing the burden of the former clone(s) vs. selecting for outgrowth of the latter may actually be subject to complex, genotype-dependent, influence of the BMSCs of the local microenvironment. In summary, our studies demonstrate the feasibility of functionally annotating the kinase dependencies in MM and potentially other neoplasias by using libraries of small-molecule kinase inhibitors in phenotypic assays against panels of tumor cell lines. Furthermore, our studies suggest that future efforts to individualize the administration of kinase inhibitors in MM should take into account not only the genotype of MM cells in the respective patient, but also the heterogeneous impact that nonmalignant "accessory" cells such as BMSCs can have on the MM cell dependence on each respective kinase and its downstream targets. Disclosures Aftab: Cleave Biosciences, Inc.: Research Funding; Onyx Pharmaceuticals, Inc.: Research Funding; Atara Biotherapeutics, Inc.: Employment, Equity Ownership; Omniox, Inc.: Research Funding. Mitsiades:TEVA: Research Funding; Janssen/Johnson & Johnson: Research Funding; Novartis: Research Funding.

Description: In multiple myeloma (MM) and other neoplasias, several kinases have been extensively evaluated as potential therapeutic targets using RNAi-based approaches or pharmacological inhibitors. Attempts to map the functional dependence of MM cells on individual kinases have primarily utilized RNAi, a mechanistic approach inherently dissimilar to small molecule inhibitors that are applied in the clinic. For many of these oncogenic kinases, large numbers of such inhibitors have been designed: these inhibitors often exhibit very similar effect on their primary designated target(s), but also perturb other secondary kinases, which may vary for different inhibitors within the same class. Using large sets of such inhibitors can enable comparative analyses to reveal the functional roles of both the respective primary target(s), as well as non-overlapping secondary targets. We therefore pursued the functional mapping of the kinome dependencies of 16 MM cell lines, using a panel of 273 kinase inhibitors (100nM, 24-72 h exposure), which target a total of 43 known primary oncogenic targets. In this study, we observed universally potent activity of Aurora (n=18 compounds), PLK (n=5), and mTORC1/2 (n=20) inhibitors; this observation is consistent with the high proliferative rate of MM cell lines in vitro. In contrast, we observed modest to minimal cell-autonomous susceptibility of MM cells to selective inhibitors of PDK1, PI3K (excluding those that also inhibit mTOR), and Akt: this suggests that PDK1- and Akt-independent mechanisms mediate the effect of PI3K signaling on the survival of most of these cell lines. In addition, we observed lack of response in virtually all tested cell lines to inhibitors of c-met (n=17 inhibitors), ALK (n=2), EGFR superfamily members (EGFR, HER2; n=25 inhibitors), c-kit (n=3), PDGFR(n=5), VEGFR (n=21), Flt3 (n=7), FAK (n=2), Syk (n=5), Src (n=5) and BTK: this result was observed even in those cell lines with detectable transcript against the respective kinases. Notable exceptions to this pattern were inhibitors that, in addition to their primary target, also possess activity to other kinases with known roles in MM (e.g. potent activity of FAK or ALK inhibitors that also target IGF1R, such as TAE226 and GSK1838705A, respectively). Consistent with prior experience, several FGFR3 inhibitors showed modest activity against FGFR3- expressing cell lines (e.g. KMS11, KMS18, OPM2, KMS34). Our screen also revealed several previously underappreciated classes of inhibitors with "non-consensus", heterogeneous, activity across the tested MM cell lines. For instance, we identified 3 clusters of cell lines with high (e.g. AMO1, Karpas-620); intermediate (e.g. KMS20, MM1S), and low responsiveness, to 8 different MEK1/2 inhibitors. Notably, both Karpas-620 and AMO1 cells are KRAS-mutant, BRAF-wild-type and have inherently high levels of p-ERK; while AMO1 cells also harbor a MEK2-Q60P mutation, previously reported to positively regulate the kinase domain activity of MEK2 and induce resistance of BRAF-V600E mutant melanoma cells to MEK1/2 inhibitors. These results raise the possibility that the response to MEK1/2 inhibitors and the role of specific mutations, such as MEK2-Q60P, are tumor-type dependent and/or influenced by concurrent BRAF mutation status. Notably, BRAF inhibitors (n=7) were inactive as cytoreductive agents against our cell line panel of BRAF wild-type cells; while several MM cell lines exhibited significantly increased proliferation upon treatment with these inhibitors. This stimulation has been previously noted in melanoma and has been attributed to activation and signaling through C-RAF; it also suggests that treatment of MM patients harboring both V600E-BRAF mutant and wild-type clones with BRAF inhibitor may decrease the burden of the former clone(s), but select for outgrowth of the latter. In summary, our studies establish the value of using large libraries of small-molecule kinase inhibitors in phenotypic assays against panels of tumor cell lines, as an approach to functionally annotate the kinome dependencies across a given neoplasia, such as MM. Furthermore, our studies provide insight into the possible clinical implications that specific molecular lesions (e.g. mutation status of MEK2 or BRAF) can have on the individualized administration of kinase inhibitors targeting the respective pathway. Disclosures Mitsiades: Millennium Pharmaceuticals: Consultancy, Honoraria; Celgene: Consultancy, Honoraria; Amgen: Research Funding; Johnson & Johnson: Research Funding.

Description: Myeloablative high-dose melphalan (HDM) followed by autologous stem cell transplantation (ASCT) remains one of the cornerstones of multiple myeloma (MM) treatment. Conceptually, HDM is based on the notion that a single melphalan dose (typically 200 mg/m2) exceeding the myelotoxicity threshold is associated with a steep increase in the dose response curve and higher degree of MM cell killing compared to fractionated delivery (without the need for hematopoietic stem cell support) of the same cumulative dose. Despite its importance in the therapeutic management of MM, HDM-ASCT is not considered a curative procedure, presumably because chemoresistant subpopulations of MM cells survive HDM and lead to eventual relapse. Elucidating the molecular mechanisms responsible for resistance to HDM in MM could have major implications for the identification of patients with the highest probability of major clinical benefit from this procedure. However, these resistance mechanisms remain incompletely understood. To address this void in the field, we examined whether an open-ended unbiased genome-wide functional characterization of MM cells could uncover genes associated with resistance to HDM and quantify the degree to which other, non-genetically determined, mechanisms of resistance contribute to this process. Specifically, we used the human myeloma cell line MM1.S, transduced with lentiviral construct for the Cas9 nuclease and with pooled lentiviral particles of the GeCKO library (Shalem et al., 2014), which consists of 2 pooled single guide RNA (sgRNA) sub-libraries (~120,000 sgRNAs; targeting ~19,000 genes and ~1800 miRNAs). Using this CRISPR/Cas9-based approach to mutagenize and cause loss of function of the genes recognized by the respective sgRNAs, we sought to facilitate the prospective isolation of MM cells resistant to HDM. To better simulate the exposure of MM cells to HDM in the autologous transplant setting, we adapted our in vitro treatment to include an initial dose of 25 µM of melphalan and sequential one-hour interval partial wash-outs, to achieve gradual reduction of melphalan concentrations, consistent with the pharmacokinetic profile observed in the clinical setting (Nath et al., 2010). We repeated twice the genome-wide knock-out screens, with biological replicates on both the transduction and cell culture level (maintaining a coverage of at least 1000 cells per individual sgRNA). While the in vitro simulation of clinical HDM with our wash-out regimen achieved 〉99% estimated reduction in MM cell viability in both screens, viable cells re-emerged in cultures from all biological and technical replicates. HDM-resistant cells were processed to quantify their sgRNA enrichment or depletion, using next generation sequencing, and also characterize the sensitivity of these MM cells to repeat exposure to HDM concentrations consistent with the genome-wide CRISPR screens. We observed that, despite having survived a previous round of exposure to HDM, the majority (〉90%) of these sgRNA-transduced cells were again highly responsive to repeat treatment with HDM. A similar result was also obtained with MM1.S cells which had not been previously transduced with the sgRNA sub-libraries. While ongoing investigation in our lab addresses these findings on additional MM cell lines, these observations on a well-established MM cell line model of melphalan responsiveness raises the intriguing possibility that genetically-determined forms of HDM resistance, conferred after an unbiased genome scale evaluation through the CRISPR/Cas9-editing methodology, may only represent a small fraction of the MM cells which survive a round of HDM treatment, and that other non-genetically determined mechanisms may mediate the principal mode of resistance, even in the cell autonomous context of our experiments. In turn, this suggests that future efforts to individualize the administration of HDM-ASCT in patients and their post-ASCT monitoring should not rely exclusively on genetic markers identified and validated from preclinical experimentation with repeated rounds of MM cell exposure to the equivalent of myelotoxic melphalan concentrations. Instead, renewed emphasis is warranted on identification of functional markers correlating with the ability of MM cells to develop transient resistance to HDM, even in the absence of constitutive genetically-determined mechanisms of resistance. Disclosures Mitsiades: Novartis: Research Funding; TEVA: Research Funding; Janssen/Johnson & Johnson: Research Funding.

Description: Acquired or de novo resistance to established and investigational therapies represents a major clinical challenge for multiple myeloma (MM) and other neoplasias. Despite extensive efforts, clinically-validated molecular markers that predict for proteasome inhibitor (PSI) resistance in most MM patients remain elusive. This challenge is partly due to limited availability so far of molecular data on MM patients before the start of PSI treatment vs. immediately after resistance to it develops; this challenge may also reflect the heterogeneity of the complex molecular mechanisms regulating MM cell response to PSIs. We hypothesized that resistance to PSIs can be mediated by disruption of several functionally overlapping genes, and that the prevalence of any of these lesions may be too low to detect in datasets available thus far. To examine this latter hypothesis, we performed a genome-wide screen for genes whose loss confers to MM cells resistance against bortezomib, through the use of the CRISPR (clustered regularly interspaced short palindromic repeats)–associated nuclease Cas9 system. Specifically RPMI-8226 MM cells were transduced with lentiviral construct for Cas9 nuclease, followed by lentiviral delivery of a genome-scale pooled library of 123,411 single-guide RNAs (sgRNAs), which selectively align to target sequences at the 5′ constitutive exons of 18,080 genes and direct the Cas9 nuclease to cause double-stranded cleavage and loss of function of the respective gene. From the pool of MM cells transduced with the sgRNA library and treated with bortezomib, treatment-resistant cells were processed for deep sequencing, to identify enriched sgRNAs and their corresponding genes. We identified that loss-of-function of 33 candidate genes is associated with bortezomib resistance. We observed a high level of consistency between independent sgRNAs targeting the same gene, as well as a high rate of hit confirmation across different biological replicates. Notably, this set of candidate bortezomib-resistance genes was distinct from the "hits" we identified through a parallel CRISPR screen on the same cell line for resistance to a different targeted therapy (namely the bromodomain inhibitor JQ1), supporting the ability of this approach to identify treatment-specific resistance genes. These candidate bortezomib-resistance genes have documented or presumed roles in the regulation of extrinsic and intrinsic apoptotic cascades, autophagy, Toll-like receptor and NF-kappaB signaling, aggresome function, heat shock protein expression, chromatin remodeling, nutrient sensing, and tumor suppressor gene networks. Importantly, information from several publically available molecular profiling datasets converge to support the putative clinical relevance of these genes. For instance, gene expression data from tumor cells of bortezomib-naive patients with advanced MM revealed several transcriptional signatures of these candidate genes (defined by low transcript levels for any of the genes in the signature) which correlated with shorter time to disease progression after treatment with bortezomib (p0.426). Congruent with these findings, the highly bortezomib-responsive clinical setting of newly-diagnosed MM is associated with low cumulative frequency of mutations of these bortezomib-resistance genes (e.g. cumulative mutation rate of 3.9%, 95% confidence interval [CI] 1.25-6.55%). Notably, in other malignancies that are typically PSI-resistant, a higher cumulative frequency of such lesions is observed (average of ~28%, range 0-76%, 95% CI 22.46-32.70%; 57 datasets from 20+ neoplasias examined). In summary, this first application of the CRISPR/Cas9-based technology in MM illustrates its power to interrogate gene function on a genome-wide scale. This approach identifies bortezomib-resistance genes that are associated with pathways linked with the regulation of proteasome inhibitor response. Results from molecularly-annotated clinical samples converge to support a possible role for these genes in bortezomib resistance. This experience supports the value of CRISPR/Cas9-based studies to dissect the molecular mechanisms of treatment resistance in MM and other hematologic neoplasias (* equal contribution of M.S. and Y.H.). Disclosures Shalem: Broad Institute: Patent application for CRISPR technology Patents & Royalties. Sanjana:Broad Institute: Patent application for CRISPR technology Patents & Royalties. Zhang:Broad Institute: Patent application for CRISPR technology Patents & Royalties. Mitsiades:Johnson & Johnson: Research Funding; Amgen: Research Funding; Celgene: Consultancy, Honoraria; Millennium Pharmaceuticals: Consultancy, Honoraria.

Description: Multiple Myeloma (MM) is a prototypical neoplasm for the study of tumor-microenvironment interactions and influences on drug response. These interactions within the bone marrow (BM) alter the signaling state of MM cells and their relative dependence on pharmacological targets. Conversely, many efforts to identify and validate drug targets in MM are conducted outside of this context. This raises the possibility that systematic re-evaluation of the current pharmacopeia may identify drugs with previously unappreciated capacity for targeting MM cells within the marrow environment. To this end, we utilized the compartment-specific bioluminescence platform (CS-BLI) to characterize the activity of 2,684 FDA-approved drugs from The Johns Hopkins Drug Library (JHDL) in three distinct MM subtypes, in the presence or absence of patient-derived CD138-negative bone marrow stromal cells (BMSCs). Anti-MM activity was evaluated at 100 nM concentrations over 72 h in MM1S (t(14;16), KRASG12A, TRAF3LOF), L363 (t(20;22), NRASQ61H, p53S261T), and OPM2 (t(4;14), FGFRK560E, p53R175H) lines. These lines demonstrate phenotypes of strong, medium, and low BMSC-induced growth enhancement, respectively. Active drugs were placed into 4 categories: type 1 - having potent anti-MM activity independent of BMSC interactions (no stromal effect), type 2 - having anti-MM activity only in the presence of BMSCs (stroma-dependent "synthetic lethality"), type 3 - having anti-MM activity that is decreased in the context of BMSCs (stroma-dependent resistance), and type 4 - otherwise inactive agents that demonstrate pro-MM activity in context of BMSCs (stroma-dependent stimulants). In this study, for MM1S, L363, and OPM2, respectively, we identified 103, 118, and 108 type 1 drugs, 217, 105, and 76 type 2 drugs, 128, 75, and 16 type 3 drugs, and 124, 33, and 38 type 4 drugs. For each category of drug phenotype, we assessed overlap across the three MM cell lines. We observed high degree of overlap for type 1 drugs (67 drugs active in all three models), while more diversity between lines was evident across the 3 lines for type 2-4 drugs, whose activity is altered by interaction with BMSCs (Figure 1). Specifically, focusing on agents demonstrating BMSC-associated stimulation, adrenergic drugs consistently stimulated MM growth in context of BMSCs, while glucocorticoids consistently grouped as type 3 agents (demonstrating BMSC-associated resistance). Interestingly, carfilzomib was also subject to BMSC-associated resistance. Despite differences in drugs demonstrating stroma-induced lethality across the MM cell lines, salicylates were commonly represented in this category. In addition to the salicylates, tofacitinib, a Janus kinase (JAK) inhibitor, demonstrated a strong capacity to elicit a stroma-dependent synthetic lethal phenotype and ruxolitinib, another inhibitor in the same class, showed a similar, yet distinct pattern of stroma-mediated sensitization. In conjunction with our screen, we performed an RNA-seq analysis to assess differential gene expression between MM in monoculture vs. in co-culture with BMSCs. Expression analysis revealed 4.0 fold increase in JAK3 expression induced by co-culture with primary BMSCs, as well as induction of a STAT3 transcription factor fingerprint by ChIP-seq enrichment analysis. A detailed dose-response analysis of tofacitinib revealed no anti-MM activity against MM cells in isolation at physiological concentrations, but showed typical sigmoidal log-dose response dynamics in the presence of stroma and a dynamic range that completely abrogated the growth advantage attributable to stromal stimulation. This phenomenon of BMSC-dependent pharmacology identifies tofacitinib as an intriguing candidate for repurposing as an agent demonstrating stroma-induced synthetic lethality against MM. Further evaluation of this agent in combination with other anti-MM agents, like bortezomib, is also warranted. Taken together, this study demonstrates specific anti-MM activity for a wide array of clinically relevant drugs and drug classes in the context of BM microenvironment interactions and provides context for further validation and potential suitability for repurposing to treat MM within the medullary compartment. Figure 1. Figure 1. Disclosures Aftab: Cleave Biosciences, Inc.: Research Funding; Omniox, Inc.: Research Funding; Atara Biotherapeutics, Inc.: Employment, Equity Ownership; Onyx Pharmaceuticals, Inc.: Research Funding. Off Label Use: The use of tofacitinib citrate and ruxolitinib will be discussed in preclinical contexts for treatment of multiple myeloma. Other approved drugs and drug classes will be generally presented in similar off-label context..

Description: Most conventional methods to sensitively quantify tumor cell proliferation and viability in vitro involve processing of cells in ways that preclude continuation of the respective experiment or prevent the longitudinal collection of data. This common technical feature of conventional assays limits their ability to provide detailed insight into the kinetics of tumor cell responses to treatment(s). Additionally, these limitations hinder the use of these assays to monitor how the kinetics of treatment response can be altered by nonmalignant "accessory" cells of the tumor microenvironment (e.g. bone marrow stromal cells [BMSCs] for hematologic malignancies or bone metastases of solid tumors). To address these obstacles, we modified our previously developed tumor cell compartment-specific bioluminescence imaging (CS-BLI) platform (McMillin et al. Nat Med. 2010), to enable longitudinal assessment of tumor cell response to diverse experimental conditions; we cultured luciferase-expressing tumor cells, with or without stromal cells, in the presence of bioluminescent substrates, using optimized conditions which provide detectable bioluminescent signal even after several days of culture, while having no adverse effect on the viability of tumor or non-malignant cells in this system. This modified approach (time-lapse CSBLI, [TL-CSBLI]) preserved the linear correlation of bioluminescent signal with tumor cell viability. Furthermore, results obtained at the end of the experiment and during interim time-points are consistent with those generated using either non-time-lapse applications of CS-BLI or conventional techniques. We applied TL-CSBLI to delineate, in high-throughput manner, the temporal dynamics of the responses of tumor cells (e.g. multiple myeloma (MM) and other hematologic malignancies) to diverse treatments (e.g. conventional chemotherapeutics, glucocorticoids; proteasome inhibitors (PIs, bortezomib or carfilzomib), and kinase inhibitors). Using the time-lapse capabilities of this assay, we evaluated tumor cell responses in the presence vs. absence of stromal cells. We observed that the kinetics of tumor cell response to diverse therapeutic classes are heterogeneous, even within the same tumor type: for instance, tumor cells with pronounced responses at the end of drug incubation (e.g. 24, 48, 72, hrs after initiation of treatment with PIs, DNA-damaging chemotherapeutics, or dexamethasone respectively), can have different magnitude of responses at intermediate time points. This suggests that TL-CSBLI data can further stratify treatment-responsive tumor cells into those with early vs. late kinetics of response. We also observed that the kinetics of the proliferative / anti-apoptotic effect conferred by stromal cells on tumor cells are highly variable between different cell lines, even within the same tumor type. For instance, the time between initiation of coculture and maximum stimulation of tumor cell viability by stromal cells was variable between cell lines and did not correlate with the magnitude of stimulation by stromal cells. Importantly, TL-CSBLI identified that the response of diverse types of tumor cells to treatments can be delayed in the presence of stromal cells, compared to conventional tumor cell monocultures: this initial delay in treatment response of tumor cells in stromal co-cultures may be observed even in cases where similar cytoreductive responses are eventually observed at later time-points in both the presence and absence of stromal cells. This observation suggests that a more expansive definition of stroma-induced resistance to a given treatment may be warranted, to specifically incorporate the ability of stromal cells to delay the tumor cell response to such treatment. In summary, TL-CSBLI enables detailed characterization of the kinetics of tumor cell responses to diverse experimental conditions. Its use can provide insight into the underappreciated impact that cell-autonomous variations or stroma-induced changes in the kinetics of tumor cell response to a given anti-tumor therapy can have on determining its efficacy. This is particularly consequential for agents (e.g. PIs) which have clinical pharmacokinetic profiles associated with transient peak exposure. Disclosures McMillin: Axios Biosciences: Equity Ownership; DFCI: patent submission on stromal co-culture technologies Patents & Royalties. Negri:DFCI: patent submission on stromal co-culture technologies Patents & Royalties. Mitsiades:Johnson & Johnson: Research Funding; Amgen: Research Funding; Celgene: Consultancy, Honoraria; Millennium Pharmaceuticals: Consultancy, Honoraria; DFCI: patent submission on stromal co-culture technologies Patents & Royalties.