Determinants Of Drug Sensitivity Research Articles

Functional genomic screens with death rate analyses reveal mechanisms of drug action.

A common approach for understanding how drugs induce their therapeutic effects is to identify the genetic determinants of drug sensitivity. Because 'chemo-genetic profiles' are performed in a pooled format, inference of gene function is subject to several confounding influences related to variation in growth rates between clones. In this study, we developed Method for Evaluating Death Using a Simulation-assisted Approach (MEDUSA), which uses time-resolved measurements, along with model-driven constraints, to reveal the combination of growth and death rates that generated the observed drug response. MEDUSA is uniquely effective at identifying death regulatory genes. We apply MEDUSA to characterize DNA damage-induced lethality in the presence and absence of p53. Loss of p53 switches the mechanism of DNA damage-induced death from apoptosis to a non-apoptotic death that requires high respiration. These findings demonstrate the utility of MEDUSA both for determining the genetic dependencies of lethality and for revealing opportunities to potentiate chemo-efficacy in a cancer-specific manner.

Abstract B010: A novel systemic methodology for mapping the clonal architectures of human prostate tumors that develop resistance to androgen receptor (AR)-targeted therapy

Abstract Developing resistance to therapeutics is a major factor leading to the high mortality rate of metastatic prostate cancer. One critical factor that mediates therapeutic resistance is the diversity of genetic mechanisms used by numerous cancer subpopulations that may be geographically co-mingled in the same physical tumor. Nonetheless, our fundamental knowledge about heterogenous mechanisms that globally contribute to drug resistance is lacking. Preliminary studies conducted by our group have shown that preexisting subclonal heterogeneity in patient tumors prior to treatment can impact clinical and pathologic outcomes. Therefore, we aimed to reconstruct tumor phylogenies and model subclonal evolution using genomic alterations identified in human prostate tumors using samples from a recent clinical trial (NCT02430480), in which patients received six months of intense androgen deprivation (ADT plus enzalutamide) prior to surgery. In this study, MRI/ultrasound-targeted biopsies were acquired at baseline, and whole mount prostatectomies were completely mapped to identify residual disease. To achieve our goal, we are performing laser capture microdissection to obtain samples of histologically distinct tumor from every tumor focus, both at baseline and post-treatment. Whole-exome sequencing (WES) is being used to identify genomic variations and then computationally map tumor subclones to the physical tissue. In addition, bulk whole-genome sequencing is being performed to resolve global subclonal architecture, to be further informed by the ground truth clonal exclusivity determined by WES of geographically distinct tumor foci. From these two complementary datasets, clonal relationships between each patient’s primary disease and its treatment-resistant counterpart will be determined and will reveal the genomic origins of drug resistance and their contribution to disease evolution. Further analyses will distinguish the evolutionary determinants of drug sensitivity (n=15 in this cohort) versus those patients with significant residual disease (n=22). To the best of our knowledge, this is the most comprehensive study mapping tumor subclones both temporally and spatially to understand the genetic origins of therapy resistance in human prostate cancer. Uncovering evolutionary tracks that each tumor utilizes for developing advanced diseases holds an unprecedented potential for effectively preventing disease from progressing. We expect that our analysis will uncover both known (AR variants and TMPRSS2:ERG fusion) and novel resistance mechanisms. Histopathological studies are being conducted in parallel for validating the spatial distribution of uncovered genetic drivers in each case. Establishment of this research framework will be essential to determining bona fide causes of various diseases from an evolutionary standpoint, which will be necessary for future precision medicine that is centered on timely and accurate clinical decision-making. This presentation will primarily be about the proposed research methodology and preliminary data will be discussed. Citation Format: Chennan Li, Scott Wilkinson, Shana Y. Trostel, William L. Dahut, Fatima Karzai, Adam G. Sowalsky. A novel systemic methodology for mapping the clonal architectures of human prostate tumors that develop resistance to androgen receptor (AR)-targeted therapy [abstract]. In: Proceedings of the AACR Special Conference: Advances in Prostate Cancer Research; 2023 Mar 15-18; Denver, Colorado. Philadelphia (PA): AACR; Cancer Res 2023;83(11 Suppl):Abstract nr B010.

DeepDSC: A Deep Learning Method to Predict Drug Sensitivity of Cancer Cell Lines.

High-throughput screening technologies have provided a large amount of drug sensitivity data for a panel of cancer cell lines and hundreds of compounds. Computational approaches to analyzing these data can benefit anticancer therapeutics by identifying molecular genomic determinants of drug sensitivity and developing new anticancer drugs. In this study, we have developed a deep learning architecture to improve the performance of drug sensitivity prediction based on these data. We integrated both genomic features of cell lines and chemical information of compounds to predict the half maximal inhibitory concentrations [Formula: see text] on the Cancer Cell Line Encyclopedia (CCLE) and the Genomics of Drug Sensitivity in Cancer (GDSC) datasets using a deep neural network, which we called DeepDSC. Specifically, we first applied a stacked deep autoencoder to extract genomic features of cell lines from gene expression data, and then combined the compounds' chemical features to these genomic features to produce final response data. We conducted 10-fold cross-validation to demonstrate the performance of our deep model in terms of root-mean-square error (RMSE) and coefficient of determination [Formula: see text]. We show that our model outperforms the previous approaches with RMSE of 0.23 and [Formula: see text] of 0.78 on CCLE dataset, and RMSE of 0.52 and [Formula: see text] of 0.78 on GDSC dataset, respectively. Moreover, to demonstrate the prediction ability of our models on novel cell lines or novel compounds, we left cell lines originating from the same tissue and each compound out as the test sets, respectively, and the rest as training sets. The performance was comparable to other methods.

Abstract 381: Combined inhibition of Bcl-2 and MCL-1 in small cell lung cancer (SCLC) is most effective in tumors with low Bcl-xL expression

Abstract Introduction: SCLC is an aggressive high-grade neuroendocrine malignancy in which targeting anti-apoptotic regulators such as Bcl-2 and Bcl-xL has shown efficacy in pre-clinical models but has not resulted in successful clinical trials (Rudin et al., Clin Cancer Res. 2012). Although SCLC cell-lines do not reflect the clinical impact of these inhibitors, patient-derived xenograft (PDX) models may more accurately recapitulate Bcl-2 family expression profiles and BH3 mimetic efficacy. One promising hypothesis is that the fellow anti-apoptotic protein MCL-1 rescues viability in the presence of Bcl-2/Bcl-xL antagonists. Here we evaluate the efficacy of the MCL-1 inhibitor S63845 in combination with a novel specific inhibitor of Bcl-2, BCL201/S55746, in SCLC patient-derived xenografts. Methods: BH3 mimetic compounds were tested for synergy in vitro in SCLC cell lines. A set of ten cell lines was chosen based on relative expression of BCL2, MCL1, and BCL2L1 (Bcl-xL) mRNA. Single agent and and pair-wise combinations of Bcl2 family inhibitors were compared in three-day growth inhibition assays. Loewe synergy scores were plotted versus Bcl2 family mRNA expression to identify the determinants of drug sensitivity. Based on the cell line synergy assays, a combination of BCL201/S55746 and S63845 was selected to test in PDX models of SCLC. Bcl-2 family expression was profiled across a panel of 37 SCLC PDX models generated at MGH by quantitative western blot, and standardized to the most sensitive SCLC cell line, NCI-H211. Ten models were selected based on absolute expression of Bcl-2, Bcl-xL and MCL-1. Mice were treated when subcutaneous tumors reached a volume of 400-800 cc, enabling precise measurement of tumor regression and time to tumor regrowth. Findings: Bcl-2 family dependency in SCLC cell lines was profiled with selective inhibitors as single agents or combinations. Maximum synergy was found between BCL201/S55746 and S63845 in cell lines with the highest Bcl-2:Bcl-xL expression ratio. Bcl-2 family expression was profiled across a panel of 37 PDX models of SCLC, and a representative set of 10 models was selected for in vivo testing. Consistent with cell line results, the two most sensitive models to BCL201/S55746+S63845 demonstrated the highest Bcl-2:Bcl-xL ratios, with moderate to high expression of MCL-1. In these models BCL201/S55746+S63845 resulted in a 44-70% tumor regression that was stable throughout 4 weeks of treatment. Efficacy was not dependent on MCL-1 expression, and was not strongly correlated with PDX sensitivity to platinum-etoposide. Conclusions: Combined inhibition of Bcl-2 with BCL201 and MCL-1 with S63845 is effective in SCLC tumors with relatively low Bcl-xL expression. This combination overcomes MCL-1 mediated resistance to Bcl-2 inhibitors, and represents a promising strategy to target anti-apoptotic dependency in SCLC. Citation Format: Benjamin J. Drapkin, Sneha Sanghavi, David T. Myers, Jun Zhong, Sarah Phat, Youzhen Wang, Ensar Halilovic, Javad Golji, Anna Farago, Erick Morris, Nicholas J. Dyson. Combined inhibition of Bcl-2 and MCL-1 in small cell lung cancer (SCLC) is most effective in tumors with low Bcl-xL expression [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 381.

Abstract 702: Kinetic analysis identifies determinants of sensitivity to MEK inhibitor-induced cell death

Abstract Successful identification of the genetic determinants of drug sensitivity has been a long-standing goal of cancer research and, more recently, of precision medicine. In cancer cells, drug treatment can induce both growth arrest and death. The distinction between these outcomes is important but can be overlooked in large-scale analyses that rely on traditional metabolism-based methods for quantifying cell viability. We recently demonstrated that Scalable Time-Lapse Analysis of Cell Death Kinetics (STACK) effectively quantifies cell death in response to diverse lethal stimuli. We investigated cases in which this approach captured drug-induced lethality more effectively than traditional assays. We found that inhibitors of mitogen activated protein kinase 1 and 2 (MEK1/2) arrested proliferation in diverse cancer cell lines, but induced cell death in only a subset of these lines. We sought to identify the factors that determine whether a cancer cell will undergo growth arrest or death in response to MEK inhibition. MAP kinase pathway activation has previously been reported as a determinant of MEK inhibitor sensitivity; we determined that the dynamics of Bcl-2 family proteins in response to MEK inhibitor treatment correlate with death sensitivity independent of pathway activation. In a genome-wide CRISPR screen, we identified a lipid metabolic enzyme that is a novel modulator of MEKi sensitivity. Our results suggest new approaches for predicting and enhancing sensitivity to cancer therapeutics in tumors and demonstrate that direct measurement of cell death provides clear advantages over proliferation-based metrics. Citation Format: Zintis Inde, Kyuho Han, Michael C. Bassik, Scott J. Dixon. Kinetic analysis identifies determinants of sensitivity to MEK inhibitor-induced cell death [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 702.

Genetic Characterization of ABT-199 Sensitivity in Human AML

Acute myeloid leukemias (AML) are aggressive blood cancers characterized by an overall survival of 27% at 5 years. The main challenge in AML treatment originates from the genetic heterogeneity of the disease that contributes to the wide range of clinical outcomes observed. A large proportion (~35%) of AML patients exhibit no distinguishable chromosomal abnormalities that can be used to guide treatment selection and are therefore classified in the poorly characterized intermediate risk category. Approximately 50-60% of intermediate risk AML patients carry mutations in the NPM1 gene. These mutations are associated with a favorable outcome unless a concomitant mutation in the FLT3 gene is detected, which accounts for 39% of cases. The survival rate is further worsened when a third mutation is detected in the DNMT3A gene, dropping from 40% to 20% 5 years post-treatment for double and triple mutants, respectively.This study aimed to identify drugs selectively affecting the viability of leukemic cells from AML patients with NPM1 mutations. To achieve this, we took advantage of culture conditions developed by our group that prevent differentiation of leukemic cells and preserve leukemia stem cell activity from primary AML specimens (Pabst et al., Nature methods, 2014), enabling chemical screening of primary AML specimens. We conducted a chemical screen using a collection of ~300 clinical grade drugs on a cohort of 38 primary human AML specimens containing NPM1 mutated (NPM1c+) and NPM1 wild-type (NPM1wt) samples. These specimens belonged to the Leucegene collection of sequenced and clinically annotated samples. The screen identified ABT-199 as the compound with the most discriminatory effect toward NPM1c+ AML. ABT-199 is a specific BH3-mimetic that prevents anti-apoptotic BCL2 from binding pro-apoptotic BAX and BAK1 proteins, leading to apoptosis. ABT-199 demonstrated encouraging results for AML treatment, but the determinants of drug sensitivity have not been well defined. Analysis of the enrichment of clinical variables in ABT-199 sensitive and resistant groups of AML specimens to define characteristics/biomarkers associated with ABT-199 sensitivity in AML revealed that mutations in NPM1, RAD21, IDH1, IDH2, DNMT3A and FLT3 (ITD), as well as normal karyotype and the FAB M1 class all significantly associate with increased ABT-199 sensitivity. At the other side of the spectrum, mutations in TP53 and FAB class M5B were significantly enriched in the resistant group. Additional analyses revealed that NPM1c+/DNMT3Amut/FLT3-ITD specimens are sensitive to ABT-199, which may provide a rationale to prioritize patients from this adverse risk AML subgroup for explorative ABT-199 based regimens. Specimens with RAD21 mutations were the most sensitive to ABT-199 treatment and further analyses demonstrated a clear association between mutation of cohesin genes (RAD21, SMC1A, SMC3, STAG2) and increased ABT-199 sensitivity. In line with this, we demonstrated that RAD21 knockdown alone is able to sensitise AML cell lines to BCL2 inhibition. Comparative transcriptome analysis of ABT-199 sensitive and resistant specimens also revealed an apoptotic gene signature linked to ABT-199 resistance with BCL2A1, an anti-apoptotic BCL2 homolog, being the most differentially expressed apoptotic gene between these response groups and showing increased expression in the resistant subset. Expression correlation analysis over the 415 specimens of the Leucegene cohort showed that BCL2A1 is one of the top genes anti-correlated to BCL2, and accordingly, high BCL2 and BCL2A1 expressors were enriched among ABT-199 sensitive and resistant specimens, respectively.In conclusion, using an unbiased pharmacogenomic approach, we identified ABT-199, a compound with the potential to eradicate NPM1c+ AML, which has already been tested in a phase 2 clinical trial for AML. Our results shed light on determinants of ABT-199 sensitivity which could readily impact AML therapy by providing a rationale for prioritizing patients with NPM1, RAD21, IDH1 and/or IDH2 mutations for ABT-199 AML trials. Our results also uncover potential mechanisms of resistance to ABT-199, providing grounds to design combination therapies to overcome ABT-199 chemoresistance. DisclosuresSauvageau:ExCellThera: Employment, Equity Ownership.

Senescence Markers from a DLBCL-Reminiscent Mouse Lymphoma Model Predict Patient Outcome

Introduction: Treatment decisions based on patient-specific molecular features are central to personalized cancer precision medicine. Oftentimes treatment response of an individual patient remains an issue of trial and error. Expression profiling has been utilized to study molecular subtypes of tumor entities and their impact on treatment outcome. Biological effector programs, such as cellular senescence, however, remain largely understudied. Syngeneic mouse models of cancer that can reproduce critical molecular features of human malignancies could serve as useful models to explore genetic determinants of drug sensitivity, and, likewise, to unveil molecular mechanisms of treatment resistance. Here, we focus on the involvement of therapy-induced senescence on treatment outcome in mouse models and patients diagnosed with diffuse large B-cell lymphoma (DLBCL).Methods: We present and characterize here the utilization of Eµ-myc transgenic lymphomas as a faithful model of chemoresistance and demonstrate its cross-species validity for DLBCL patients. Specifically, primary Eµ-myc lymphomas, of which we generated gene expression profiles (GEP) at diagnosis, were exposed to genotoxic therapy in vivo, and subsequently monitored regarding long-term outcome in a clinical trial-like design. Lymphoma senescence capability, a central drug effector principle, was studied in mice by unbiased approaches as well as loss- and gain-of-function genetics.Results: Investigation of DLBCL-established gene expression based subtypes related to cell-of-origin (COO - i.e. GCB/ABC subtypes) and distinct DLBCL biologies (e.g. comprehensive consensus clusters [CCC]) using machine-learning methods demonstrated their relevance in the murine platform. Moreover, our findings show an important role of histone H3 lysine 9-trimethylation (H3K9me3) for senescence induction and treatment outcome as demonstrated by shorter time to death and time to relapse of mice bearing lymphomas with engineered loss of the H3K9me3-critical methyltransferase Suv39h1 on one hand and lymphomas with genetically transferred or endogenous overexpression of H3K9-active demethylases on the other hand. Furthermore, expression levels of H3K9me3-specific demethylases stratified unmodified Eµ-myc lymphomas and DLBCL patients into two groups with superior outcome for those with lower levels. In line with these findings, DLBCL patients with high levels of the senescence-associated H3K9me3 mark in their lymphomas presented with significantly longer survival times. Further transcriptomics-based investigations of our clinical-trial like mouse model and DLBCL patients suggests the presence of a molecular network distinguishing lymphomas into a clinically superior senescence responder from an inferior non-responder group.Conclusions: Our results conclude that Eµ-myc transgenic lymphomas serve as faithful model for human DLBCL and the importance of therapy-induced senescence for treatment outcome of DLBCL patients. Our data suggest the integration of tractable, transgenic mouse models to the repertoire of functional test platforms to better implement lesion- and state-based decisions in personalized cancer precision medicine. Ongoing mouse model-based work aims at specific targeting of aberrant demethylase activities and synthetic lethal approaches to selectively eliminate potentially detrimental senescent lymphoma cells after chemotherapy to assess therapeutic long-term effects and to determine the conditions for future early-phase DLBCL clinical testing. DisclosuresNo relevant conflicts of interest to declare.

Abstract 5115: Establishing tumoroid and mouse models for functional validation of progression markers in DCIS

Abstract Ductal Carcinoma in Situ (DCIS) is a starting lesion in the milk duct of the breast, which accounts for 25% of all ‘breast cancers' detected since the introduction of breast screening. DCIS is usually treated by surgery combined with radiotherapy, which can have a large impact on the life of patients. However, there is little to no evidence that treatment of low and intermediate grade DCIS reduces mortality, while women diagnosed with DCIS do perceive their risk of dying the same as patients with invasive disease. To reduce the negative perception and the overtreatment of DCIS, but assure proper treatment for high risk DCIS, it is critical to identify which factors can predict whether DCIS stays indolent or becomes invasive. We have successfully shown that it is possible to propagate tumor cells and organoids in vivo using intraductal injections. Adult female NSG and nude mice were injected intraductally into the mammary gland with suspensions of breast cancer cell and organoid lines as well as DCIS cell lines. Moreover, we were able to instigate a protocol for deriving fresh patient DCIS material to create in vitro primary tumoroid cultures, which we transplanted as suspended cells into NSG and nude mice. Tumor formation was examined via palpation and bioluminescence imaging. The resulting tumors will be compared to the original patient material to show that we are able to create a suitable in vivo model, retaining the morphologic and genomic features of the patient. In addition to these transplantation models we sought to take advantage of existing non-germline models for establishing a genetic DCIS model. To establish these, genes were selected from literature which are suspected to initiate DCIS formation. These genes were then incorporated in lentiviral vectors and injected in the mammary gland of immunocompetent mice. The mice will be sacrificed at set time points to follow DCIS initiation and progression to Invasive Ductal Carcinoma (IDC). For characterization, all these models will be subjected to in-depth histopathologic, transcriptomic, mutational, proteomic, immunologic and methylomic characterization, such as copy number sequencing, RNA sequencing and reverse-phase protein assays. Together these models will aid in finding factors initiating DCIS and finding the key characteristics driving DCIS to IDC switch, and ultimately making it possible to distinguish between high risk and low risk DCIS. On top of this, the in vitro and in vivo human DCIS models will place a unique opportunity for high-throughput drug screening to obtain sensitivity profiles. These will be combined with -omics data to identify (epi)genetic determinants of drug sensitivity. Citation Format: Stefan J. Hutten, Catrin Lutz, Stefano Annunziato, Ellen Tanger, Jelle Wesseling, Jos Jonkers. Establishing tumoroid and mouse models for functional validation of progression markers in DCIS [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 5115.

Abstract B167: KRAS gene mutations are associated with tumor resistance toward the LN1222/AD736 p38 inhibitor

Abstract The p38 MAPK pathway, also called stress activated protein kinase pathway, is involved in basic cellular processes of inflammation and acts on cell growth, proliferation, differentiation, migration, and apoptosis. p38 MAP kinase activation occurs by environmental stress and proinflammatory cytokines signaling and is frequently deregulated in cancers. p38 MAP kinase inhibition has been suggested for treatment of the cancer; a number of inhibitors have been developed and some are in an early phase of clinical evaluation. The present study aimed to investigate the antitumor activity and discover potential predictive biomarkers for the novel p38 inhibitor LN1222/AD736 in a large and diverse series of tumor models. Cancer cell inhibition was tested in vitro by means of a propidium-based monolayer assay in a total of 222 human tumor cell lines, covering 29 different types of solid tumors including 70 derived from PDX. To illuminate molecular determinants of drug sensitivity, we conducted a biomarker discovery study using the molecular profiles of cell lines, which included whole exome mutations, gene copy number variations, and gene expression profiles as determined by next-generation sequencing, Affymetrix SNP6.0 and Affymetrix U133 2plus. Throughout the 222 cell lines, the IC50 and IC70 geometric means were 0.33 μM and 1.3 μM, respectively. High antitumor potency (IC70 <500 nΜ) was found in 121 out of 222 cell lines (54%), and the cell line sensitivity toward LN1222/AD736 was well dichotomized into very sensitive (<500 nM) and very resistant (>30 μM) models. Very sensitive cell lines were observed in leukemia and lymphoma as well as a wide variety of solid tumors including sarcoma (soft tissues and osteosarcoma), prostate, bladder, breast, melanoma, colon, NSCLC, and kidney tumor models, irrespective of the cancer type. Molecular analyses showed that tumor models had comparable expression patterns of the four different isoforms of p38 mRNA (MAPK11, 12, 13, and 14) with those of the patient’s tumors (TCGA data set). None of the four isoforms was found significantly predictive for tumor sensitivity to LN1222/AD736 (Lima test and Spearman correlation analyses). Interestingly, the biomarker discovery screening associated KRAS gene mutations (24 of 127 model tested (20%)) as a strong factor for resistance toward LN1222/AD736, only wild type KRAS models were sensitive (Wilcoxon test p<0.0001). In parallel, we identified a subset of 80 differentially expressed genes that discriminates groups of tumors with a high response rate by using unsupervised hierarchical clustering (p<0.0001). The ingenuity pathway analysis indicates that part of these genes are related to inflammation and cell survival. These biomarker data will be evaluated in an independent data set of tumors treated with p38 inhibitors. Our results demonstrate that the p38 inhibitor LN1222/AD736 had antitumor activity in a wide range of tumors and may represent a relevant treatment option for tumors with wild type KRAS. Extending the analysis in vitro and in vivo in tumor types in which activity was seen and testing combinations of LN1222/AD736 and other compounds targeting the MAPK pathway are needed to refine the evaluation. The biomarker findings should be tested and validated in larger cohort. Citation Format: Vincent Vuaroqueaux, Gerhard Kelter, Hans R Hendriks, Stefan Laufer, Heinz-Herbert Fiebig. KRAS gene mutations are associated with tumor resistance toward the LN1222/AD736 p38 inhibitor [abstract]. In: Proceedings of the AACR-NCI-EORTC International Conference: Molecular Targets and Cancer Therapeutics; 2017 Oct 26-30; Philadelphia, PA. Philadelphia (PA): AACR; Mol Cancer Ther 2018;17(1 Suppl):Abstract nr B167.

Abstract 1083: Synergistic drug combination prediction through drug differential dependency network analysis

Abstract In an effort to discover strategies which identify effective drug combinations, we analyzed 39 of the 480 compounds screened in the Cancer Therapeutics Response Portal (CTRP) where combinations of two compounds were tested against 860 cancer cell lines; this enabled a comparison of the drug sensitivity of the combinations versus that of the individual compounds. More than half of the drug combinations (n=21) did not significantly improve the drug sensitivity, compared to individual compounds alone. In fact, some of the combinations showed reduced drug sensitivity. In EDDY-CTRP* analysis, the Cancer Cell Line Encyclopedia (CCLE) RNAseq data and CTRP compound response measurements were analyzed to discover both 1) pathways enriched with differential dependencies between sensitive and non-sensitive cell lines for each compound and 2) the mediators of cell line response to a drug. A mediator is a gene in a pathway that plays a significantly different role between sensitive and non-sensitive conditions. The significance is assessed for either essentiality, measured as a node’s centrality change, or specificity, measured as the difference in condition specific edges. These drug-pathway-mediator connections are predicted to reveal crucial molecular determinants of drug sensitivity that otherwise are hidden in the complexities of the molecular networks of the cell (Speyer et al., PSB 22:497-508, 2017). We further investigated whether mediators identified for single compounds could predict sensitivity to drug combinations. This analysis revealed that if two single compounds share the same specificity mediators, i.e. the genes with the most significant re-wiring of gene dependencies between sensitive and non-sensitive cell lines, combination of these two compounds correlate with improved sensitivity. The converse was also found: compounds that do not share mediators rarely show synergy. Further analysis and empirical testing of predicted combinations promises to prioritize synergistic drug combinations. We believe that this methodology may predict synergistic drug combinations from cancer cell line drug screening data. Supported by NIH U01CA168397. *available at http://biocomputing.tgen.org/software/EDDY/CTRP Citation Format: Seungchan Kim, Gil Speyer, Harshil Dhruv, Jeff Kiefer, Michael Berens. Synergistic drug combination prediction through drug differential dependency network analysis [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 1083. doi:10.1158/1538-7445.AM2017-1083

Chromatin-Remodeling Complex SWI/SNF Controls Multidrug Resistance by Transcriptionally Regulating the Drug Efflux Pump ABCB1.

Anthracyclines are among the most effective yet most toxic drugs used in the oncology clinic. The nucleosome-remodeling SWI/SNF complex, a potent tumor suppressor, is thought to promote sensitivity to anthracyclines by recruiting topoisomerase IIa (TOP2A) to DNA and increasing double-strand breaks. In this study, we discovered a novel mechanism through which SWI/SNF influences resistance to the widely used anthracycline doxorubicin based on the use of a forward genetic screen in haploid human cells, followed by a rigorous single and double-mutant epistasis analysis using CRISPR/Cas9-mediated engineering. Doxorubicin resistance conferred by loss of the SMARCB1 subunit of the SWI/SNF complex was caused by transcriptional upregulation of a single gene, encoding the multidrug resistance pump ABCB1. Remarkably, both ABCB1 upregulation and doxorubicin resistance caused by SMARCB1 loss were dependent on the function of SMARCA4, a catalytic subunit of the SWI/SNF complex. We propose that residual SWI/SNF complexes lacking SMARCB1 are vital determinants of drug sensitivity, not just to TOP2A-targeted agents, but to the much broader range of cancer drugs effluxed by ABCB1. Cancer Res; 76(19); 5810-21. ©2016 AACR.

Constitutive BAK activation as a determinant of drug sensitivity in malignant lymphohematopoietic cells.

Mitochondrial outer membrane permeabilization (MOMP), a key step in the intrinsic apoptotic pathway, is incompletely understood. Current models emphasize the role of BH3-only BCL2 family members in BAX and BAK activation. Here we demonstrate concentration-dependent BAK autoactivation under cell-free conditions and provide evidence that this autoactivation plays a key role in regulating the intrinsic apoptotic pathway in intact cells. In particular, we show that up to 80% of BAK (but not BAX) in lymphohematopoietic cell lines is oligomerized and bound to anti-apoptotic BCL2 family members in the absence of exogenous death stimuli. The extent of this constitutive BAK oligomerization is diminished by BAK knockdown and unaffected by BIM or PUMA down-regulation. Further analysis indicates that sensitivity of cells to BH3 mimetics reflects the identity of the anti-apoptotic proteins to which BAK is constitutively bound, with extensive BCLXL•BAK complexes predicting navitoclax sensitivity, and extensive MCL1•BAK complexes predicting A1210477 sensitivity. Moreover, high BAK expression correlates with sensitivity of clinical acute myelogenous leukemia to chemotherapy, whereas low BAK levels correlate with resistance and relapse. Collectively, these results inform current understanding of MOMP and provide new insight into the ability of BH3 mimetics to induce apoptosis without directly activating BAX or BAK.

Thymosin-β4 is a determinant of drug sensitivity for Fenretinide and Vorinostat combination therapy in neuroblastoma

Retinoids are an important component of neuroblastoma therapy at the stage of minimal residual disease, yet 40–50% of patients treated with 13-cis-retinoic acid (13-cis-RA) still relapse, indicating the need for more effective retinoid therapy. Vorinostat, or Suberoylanilide hydroxamic acid (SAHA), is a potent inhibitor of histone deacetylase (HDAC) classes I & II and has antitumor activity in vitro and in vivo. Fenretinide (4-HPR) is a synthetic retinoid which acts on cancer cells through both nuclear retinoid receptor and non-receptor mechanisms. In this study, we found that the combination of 4-HPR + SAHA exhibited potent cytotoxic effects on neuroblastoma cells, much more effective than 13-cis-RA + SAHA. The 4-HPR + SAHA combination induced caspase-dependent apoptosis through activation of caspase 3, reduced colony formation and cell migration in vitro, and tumorigenicity in vivo. The 4-HPR and SAHA combination significantly increased mRNA expression of thymosin-beta-4 (Tβ4) and decreased mRNA expression of retinoic acid receptor α (RARα). Importantly, the up-regulation of Tβ4 and down-regulation of RARα were both necessary for the 4-HPR + SAHA cytotoxic effect on neuroblastoma cells. Moreover, Tβ4 knockdown in neuroblastoma cells increased cell migration and blocked the effect of 4-HPR + SAHA on cell migration and focal adhesion formation. In primary human neuroblastoma tumor tissues, low expression of Tβ4 was associated with metastatic disease and predicted poor patient prognosis. Our findings demonstrate that Tβ4 is a novel therapeutic target in neuroblastoma, and that 4-HPR + SAHA is a potential therapy for the disease.

Differential modulation of p‐4EBP1, p‐ERK and cyclin D1 by amuvatinib/erlotinib treatment in prostate cancer cells (1055.4)

Cyclin D1 is a key regulator of cell cycle progression and a therapeutic endpoint for cancer treatment. The human prostate cancer cell lines, LNCaP (PTEN‐) and DU145 (PTEN+), were used to examine the differential regulation of cyclin D1 as a determinant of drug sensitivity to amuvatinib, a receptor tyrosine kinase inhibitor, and erlotinib, an epidermal growth factor inhibitor. 1D and 2D Western blot analysis revealed that combination drug treatment of LNCaP cells caused a decrease in 4EBP1 pSer65, pThr70, and pThr37/46, with concomitant decreases in cyclin D1 levels. Such combination treatment, however, did not modulate pERK status. Moreover, single drug treatment with amuvatinib, but not erlotinib, decreased p‐4EBP1 and cyclin D1 levels. In DU‐145 cells, combination drug treatment had no effect on p‐4EBP1 status, but decreased p‐ERK and cyclin D1 levels. Additionally, single drug treatment with erlotinib, but not amuvatinib, caused a loss of p‐ERK and cyclin D1. The data reveal the differential modulation of cyclin D1, and its upstream regulators, in prostate cancer cell lines by amuvatinib and erlotinib. Such differential responses likely contribute to specific anti‐tumor agent efficacy. An appreciation of factors that determine the cell‐dependent modulation of cyclin D1 will assist in the selection of appropriate biomarkers for directing patient‐targeted therapy.Grant Funding Source: Supported by Ventana/Roche, P30 ES006694

Integrating the NCI‐60 Data with “Omics” for Drug Discovery

Abstract Preclinical Research The National Cancer Institute (NCI)‐60 screening panel represents one of the largest chemical cell line screening endeavors ever undertaken. This screen, starting in the early 1990s, has evaluated the effect of ∼400,000 compounds on the growth of a panel of over 60 human cancer cell lines. The screen and the extensive characterization of the cell lines provide a wealth of data on determinants of drug sensitivity, much of it publicly available. Here, an overview of the data is provided including basic characteristics of cell lines, the variance of 50% growth inhibitory concentration of over 40,000 chemicals, and expression of 17,775 genes across these cell lines. This analysis revealed that only a fraction of compounds have differential sensitivities across lines. For the majority of compounds, the genomic underpinnings of drug sensitivity cannot possibly be discerned from the data as there is no significant variance between cell lines. This limitation applies, surprisingly, to platinum agents and even to tamoxifen, demonstrating the limitation of this screening methodology to capture all differences that predict drug sensitivity in the context of human cancer. In contrast, methotrexate has remarkably disparate effects across cell lines, suggesting that the origin of its sensitivity can be discerned. Correlations of drug sensitivity and gene expression reveal some remarkably robust correlations across the NCI‐60 cell lines, suggesting their value as predictive biomarkers, or perhaps even drug targets. These results demonstrate both the power and limitations of NCI‐60 type screens to identify the determinants of drug sensitivity.

N-methylpurine DNA glycosylase inhibits p53-mediated cell cycle arrest and coordinates with p53 to determine sensitivity to alkylating agents

Alkylating agents induce genome-wide base damage, which is repaired mainly by N-methylpurine DNA glycosylase (MPG). An elevated expression of MPG in certain types of tumor cells confers higher sensitivity to alkylation agents because MPG-induced apurinic/apyrimidic (AP) sites trigger more strand breaks. However, the determinant of drug sensitivity or insensitivity still remains unclear. Here, we report that the p53 status coordinates with MPG to play a pivotal role in such process. MPG expression is positive in breast, lung and colon cancers (38.7%, 43.4% and 25.3%, respectively) but negative in all adjacent normal tissues. MPG directly binds to the tumor suppressor p53 and represses p53 activity in unstressed cells. The overexpression of MPG reduced, whereas depletion of MPG increased, the expression levels of pro-arrest gene downstream of p53 including p21, 14-3-3σ and Gadd45 but not proapoptotic ones. The N-terminal region of MPG was specifically required for the interaction with the DNA binding domain of p53. Upon DNA alkylation stress, in p53 wild-type tumor cells, p53 dissociated from MPG and induced cell growth arrest. Then, AP sites were repaired efficiently, which led to insensitivity to alkylating agents. By contrast, in p53-mutated cells, the AP sites were repaired with low efficacy. To our knowledge, this is the first direct evidence to show that a DNA repair enzyme functions as a selective regulator of p53, and these findings provide new insights into the functional linkage between MPG and p53 in cancer therapy.

Abstract IA15: Predicting drug sensitivity from cancer cell lines

Abstract High-throughput molecular characterization technologies have enabled detailed genetic and genomic characterization of large panels of cancer cell lines, including profiling of gene expression, copy number variation (CNV), SNP, mutation, and RNA-seq. The same cell lines have also been profiled for sensitivity to potential anti-cancer compounds, allowing development of computational models used to predict genotype-specific drug treatments linked to tumor molecular subtypes. Inferring such models is challenging because the model inputs contain far more features than observations, known as the p>>n problem, precluding the use of classical statistical models such as least squares (a.k.a. multiple linear regression). Recently, generalized linear models or penalized linear models have been proposed and extended for feature selection and overcoming the p>>n. Bayesian extension of such methods have also been developed, using algorithms such as Metropolis-Hasting and Markov Chain Monte Carlo, and demonstrated to improve prediction accuracy at the expense of increased computing cost. A comprehensive evaluation of predictive modeling techniques is essential to leverage cell line studies to infer the most accurate genetic predictors of drug sensitivity, which can then be used to inform patient selection strategies in clinical trials and to identify functional genetic determinants of drug sensitivity or resistance. In this study, we evaluate state-of-the-art predictive models which utilize dimension reduction methods (e.g., partial least square, support vector machine, principal component regression), feature selection (e.g., LASSO, RIDGE, and elastic-net), and Bayesian feature selection (e.g., Bayesian LASSO and Bayesian RIDGE). We assess each method based on 1) the accuracy of sensitivity predictions in a cross validation setting and in an independent dataset, and 2) the biological coherence of inferred predictive features by comparison to publicly available pathway databases. Citation Format: Adam Arne Margolin, In Sock Jang, Stephen Friend. Predicting drug sensitivity from cancer cell lines. [abstract]. In: Proceedings of the AACR Special Conference on Chemical Systems Biology: Assembling and Interrogating Computational Models of the Cancer Cell by Chemical Perturbations; 2012 Jun 27-30; Boston, MA. Philadelphia (PA): AACR; Cancer Res 2012;72(13 Suppl):Abstract nr IA15.

Abstract PL03-01: Discovery of candidate biomarkers of anticancer drug sensitivity by high-throughput cell line screening.

Abstract PL03-01: Discovery of candidate biomarkers of anticancer drug sensitivity by high-throughput cell line screening.

Abstract 3804: The pathobiology of perturbed SEPT9 expression in ovarian neoplasia

Abstract We have made observations of the deregulated expression of certain SEPT9 transcripts and we wish to place these in a pathobiological context. Septins are an evolutionarily conserved family of GTP binding proteins which have been implicated in a wide range of cellular processes. SEPT9 was initially identified through mapping of a region of chromosome 17q which demonstrated allelic imbalance in sporadic ovarian tumours. The large SEPT9 locus is complex since up to 18 splice variants are possible through alternate splicing of six 5’ ends and three 3’ ends. Of note are two transcripts SEPT9_v4 and v4* differing only in their 5’UTRs but encoding the same polypeptide SEPT9 isoform 4 (SEPT9_i4). Ovarian neoplasia is accompanied by down regulation of SEPT9_v4 and up regulation of SEPT9_v4*. We have also shown that translation of the former is by cap-dependant mechanisms while translation of the later is cap-independant and stress responsive (Hum Mol Gen 2007;6:742-52). We therefore sought to understand the mechanism of control of transcription of these two transcripts and how this is perturbed in neoplasia. Using bioinformatic analysis we have identified multiple putative p53 responsive elements upstream of the transcription start site of SEPT9_v4*. Such sites are rare upstream of all other SEPT9 transcripts. In wild type p53 cells but not an isogenic p53 null cell line, the SEPT9_v4* transcript and isoform 4 are induced under conditions in which p53 is induced. No other SEPT9 transcript responded in this way. We used chromatin immunoprecipitation assays to demonstrate the validity of at least some of these putative p53 sites. Using a luciferase assay we show that these same sites bind and respond to wild type but not mutant p53. Previous studies from our lab have shown that over expression of isoform 4 perturbs microtubule dynamics in cells such that depolymerisation of microtubules (in response to cold) and their subsequent repolymerisation is delayed. We therefore sought to determine if the p53 dependant upregulation of SEPT9_v4* might also be associated with altered microtubule dynamics and resistance to microtubule-interacting drugs. In a tissue culture model we observation that induction of p53 and SEPT9_v4* using a DNA damaging agent reduces the cellular response to paclitaxel. Further, the transfection of SEPT9_v4* specific siRNA (with SEPT9_i4 knock down) leads to dramatically enhanced cell kill after taxol treatment compared with scrambled siRNAs. The clinical relevance of this is emphasised by the observation that SEPT9_v4* expression is associated with disease stage in ovarian cancer. Our data provide novel insight into septin function and the role of stress pathways including p53 in its regulation and has clinical significance as a determinant of drug sensitivity. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 102nd Annual Meeting of the American Association for Cancer Research; 2011 Apr 2-6; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2011;71(8 Suppl):Abstract nr 3804. doi:10.1158/1538-7445.AM2011-3804

Farnesyl transferase expression determines clinical response to the docetaxel‐lonafarnib combination in patients with advanced malignancies

Lonafarnib (LNF) is a protein farnesyl transferase (FTase) inhibitor that has shown synergistic activity with taxanes in preclinical models and early stage clinical trials. Preclinical findings suggested tubulin acetylation and FTase expression levels may be important determinants of drug sensitivity that would help identify patient populations more likely to benefit from this regimen. This pilot study evaluated the biological effects of LNF and docetaxel (DTX) combination therapy in refractory solid tumors by comparing pretreatment and post-treatment tumor biopsies. Patients with histologically confirmed locally advanced or metastatic solid malignancies refractory to standard therapies or with no effective therapies available were eligible. Patients were randomized to 1 of 4 dosing cohorts: 1) 30 mg/m², 100 mg; 2) 36 mg/m², 100 mg; 3) 30 mg/m², 150 mg; or 4) 36 mg/m², 150 mg of DTX intravenously weekly, LNF orally twice daily, respectively. Of the 38 patients enrolled, 36 were treated, and 29 were evaluable for toxicity and response assessment. The combination of LNF and DTX was tolerated in all cohorts with the exception of a 28% incidence of grade 3/4 diarrhea, which was manageable with aggressive antidiarrheal regimens. Seven patients derived clinically meaningful benefit from this combination treatment; these patients had significantly lower basal FTase-beta mRNA expression levels than the mean study population level (P < .05). Correlation of clinical benefit with tubulin acetylation content as well as basal acetyl-tubulin content were evaluated. However, no significant correlation was found. Despite the small number of patients, these findings support our preclinical mechanistic studies and warrant further clinical investigations using FTase-beta mRNA expression as a potential predictive biomarker to select for an enriched patient population to study the effects of taxane and FTase inhibitor combination therapies.