Cancer Cell Line Screens Research Articles

The Anti-Tumor Activity of Igm-8444, an Agonistic Death Receptor 5 (DR5) IgM Antibody, Is Sensitized in Combination with Chemotherapy and Bcl-2 Inhibitors in NHL and AML

Background: Death receptor 5 (DR5) is a member of the tumor necrosis factor (TNF) receptor superfamily that multimerizes when bound to its ligand, TNF-related apoptosis inducing ligand (TRAIL), to activate the extrinsic apoptotic pathway. DR5 is broadly expressed on solid and hematologic cancers and has been targeted with both recombinant TRAIL and agonistic antibodies in the clinic. However, these therapeutics have been unsuccessful due to lack of efficacy or hepatotoxicity. We have developed IGM-8444, a pentameric IgM with 10 binding sites specific for DR5, that multimerizes DR5 to selectively and potently induce tumor cell apoptosis while maintaining tolerability. We have previously presented the in vitro and in vivo efficacy of IGM-8444 in solid tumor models, demonstrating low picomolar potency across multiple tumor cell lines, strong tumor regressions in cell line and patient derived xenograft mouse tumor models, and dose-dependent increases in apoptotic biomarkers. Here, we evaluate the activity of IGM-8444 in hematologic malignancies in combination with chemotherapy or targeted agents including Bcl-2 inhibitors targeting the intrinsic apoptotic pathway. Methods: Human hematologic cancer cell lines and primary human hepatocytes were evaluated in vitro for dose-dependent IGM-8444-induced cytotoxicity. Cell lines were further evaluated using IGM-8444 in combination with chemotherapy or targeted agents including Bcl-2 inhibitor ABT-199. In vivo efficacy was evaluated using IGM-8444 in combination with ABT-199 in cell line-derived xenograft mouse tumor models. Results: In a previous cancer cell line screen profiling single agent IGM-8444 cytotoxicity across 190 solid and hematologic cell lines, 25 (13%) were classified as highly responsive and 75 (39%) as moderately responsive to IGM-8444 induced cell death. Here the in vitro activity of IGM-8444 was evaluated across a subset of 32 NHL and AML cell lines. 5/21 (24%) of NHL cell lines and 5/11 (45%) of AML cell lines tested were classified as highly responsive or moderately responsive to IGM-8444-induced cytotoxicity. The DOHH-2 and JEKO1 NHL cell lines were amongst the most sensitive, with growth-normalized EC50 values as low as 0.03 ng/mL (0.03 pM) for JEKO1. Combinations with chemotherapy including cytarabine and doxorubicin or targeted agents such as Bcl-2 inhibitor ABT-199 resulted in synergistic in vitro cytotoxicity in multiple cell lines, as determined by Bliss synergy scores. IGM-8444 demonstrated minimal to no in vitro cytotoxicity to primary human hepatocytes at doses several log-fold higher than efficacious doses, and this favorable in vitro safety profile was maintained in combination with chemotherapeutic agents and ABT-199. Combination of IGM-8444 with ABT-199 also resulted in synergistic in vivo efficacy. In a DOHH-2 NHL model, IGM-8444 and ABT-199 showed modest tumor growth inhibition as single agents. However the combined treatment regimen led to tumor regressions during the first 2 weeks of treatment, with 3 of 10 animals showing a partial response and 2 of 10 animals achieving a complete response. The combined treatment also extended median overall survival compared to the control group, which was a significant improvement compared to either agent alone. Collectively, these results provide a strong rationale for simultaneously targeting the extrinsic and intrinsic apoptotic pathways to achieve enhanced efficacy. Conclusions: These data support the clinical development of IGM-8444 in hematological malignancies as a single agent, in combination with standard of care chemotherapy, and in combination with targeted agents that impact the intrinsic signaling pathway such as Bcl-2 inhibitor ABT-199. Initiation of a Phase I clinical study evaluating the safety of IGM-8444 is anticipated in 2020. Disclosures Wang: IGM Biosciences Inc: Current Employment, Current equity holder in publicly-traded company. Matthew:IGM Biosciences Inc: Current Employment, Current equity holder in publicly-traded company. Wang:IGM Biosciences Inc: Current Employment, Current equity holder in publicly-traded company. Kothambawala:IGM Biosciences Inc: Current Employment, Current equity holder in publicly-traded company. Calhoun:IGM Biosciences Inc: Current Employment, Current equity holder in publicly-traded company. Humke:IGM Biosciences Inc: Current Employment, Current equity holder in publicly-traded company. Sinclair:IGM Biosciences Inc: Current Employment, Current equity holder in publicly-traded company. Keyt:IGM Biosciences Inc: Current Employment, Current equity holder in publicly-traded company.

MEK is a promising target in the basal subtype of bladder cancer.

While many resources exist for the drug screening of bladder cancer cell lines in 2D culture, it is widely recognized that screening in 3D culture is more representative of in vivo response. Importantly, signaling changes between 2D and 3D culture can result in changes to drug response. To address the need for 3D drug screening of bladder cancer cell lines, we screened 17 bladder cancer cell lines using a library of 652 investigational small-molecules and 3 clinically relevant drug combinations in 3D cell culture. Our goal was to identify compounds and classes of compounds with efficacy in bladder cancer. Utilizing established genomic and transcriptomic data for these bladder cancer cell lines, we correlated the genomic molecular parameters with drug response, to identify potentially novel groups of tumors that are vulnerable to specific drugs or classes of drugs. Importantly, we demonstrate that MEK inhibitors are a promising targeted therapy for the basal subtype of bladder cancer, and our data indicate that drug screening of 3D cultures provides an important resource for hypothesis generation.

Targeting N-myristoylation for therapy of B-cell lymphomas

Myristoylation, the N-terminal modification of proteins with the fatty acid myristate, is critical for membrane targeting and cell signaling. Because cancer cells often have increased N-myristoyltransferase (NMT) expression, NMTs were proposed as anti-cancer targets. To systematically investigate this, we performed robotic cancer cell line screens and discovered a marked sensitivity of hematological cancer cell lines, including B-cell lymphomas, to the potent pan-NMT inhibitor PCLX-001. PCLX-001 treatment impacts the global myristoylation of lymphoma cell proteins and inhibits early B-cell receptor (BCR) signaling events critical for survival. In addition to abrogating myristoylation of Src family kinases, PCLX-001 also promotes their degradation and, unexpectedly, that of numerous non-myristoylated BCR effectors including c-Myc, NFκB and P-ERK, leading to cancer cell death in vitro and in xenograft models. Because some treated lymphoma patients experience relapse and die, targeting B-cell lymphomas with a NMT inhibitor potentially provides an additional much needed treatment option for lymphoma.

Abstract A73: BCL-XL dependency in a subset of ovarian clear-cell carcinoma cell lines

Abstract Clear-cell ovarian carcinoma is a rare histologic subtype of epithelial ovarian cancers. Clear-cell ovarian cancers tend to be less responsive to platinum-based chemotherapy compared to high-grade serous ovarian cancers. In a dataset of large-scale CRISPR knockout and RNAi screens of cancer cell lines (depmap.org), a subset of ovarian clear-cell carcinoma cell lines were sensitive to knockout or knockdown of BCL2L1, encoding antiapoptotic protein BCL-XL. Most clear-cell ovarian cancer cell lines also express BCL2L1 mRNA and protein. We therefore tested a panel of ovarian cancer cell lines for response to BH3 mimetic drugs, which inhibit different members of the BCL-2 family of antiapoptotic proteins. Initial data showed that a subset of ovarian clear-cell carcinoma cell lines was sensitive to single-agent BCL-XL inhibition with A1331852 and dual BCL-2/BCL-XL inhibition with ABT-263. In contrast, these cell lines were relatively insensitive to single-agent BCL-2 inhibition or MCL1 inhibition. We are currently applying BH3 profiling, a functional assay for apoptotic priming, to test the response of ovarian clear-cell carcinoma cell lines to BCL-XL inhibition using a BCL-XL specific peptide, HRK. Collectively, these studies suggest that ovarian clear-cell carcinomas may be sensitive to BCL-XL inhibition, either alone or in combination with chemotherapy. Citation Format: Elizabeth H. Stover, Maya K. Gelbard, Ursula A. Matulonis, Matthew Meyerson. BCL-XL dependency in a subset of ovarian clear-cell carcinoma cell lines [abstract]. In: Proceedings of the AACR Special Conference on Advances in Ovarian Cancer Research; 2019 Sep 13-16, 2019; Atlanta, GA. Philadelphia (PA): AACR; Clin Cancer Res 2020;26(13_Suppl):Abstract nr A73.

Expression of Annexin A2 Promotes Cancer Progression in Estrogen Receptor Negative Breast Cancers.

When breast cancer progresses to a metastatic stage, survival rates decline rapidly and it is considered incurable. Thus, deciphering the critical mechanisms of metastasis is of vital importance to develop new treatment options. We hypothesize that studying the proteins that are newly synthesized during the metastatic processes of migration and invasion will greatly enhance our understanding of breast cancer progression. We conducted a mass spectrometry screen following bioorthogonal noncanonical amino acid tagging to elucidate changes in the nascent proteome that occur during epidermal growth factor stimulation in migrating and invading cells. Annexin A2 was identified in this screen and subsequent examination of breast cancer cell lines revealed that Annexin A2 is specifically upregulated in estrogen receptor negative (ER-) cell lines. Furthermore, siRNA knockdown showed that Annexin A2 expression promotes the proliferation, wound healing and directional migration of breast cancer cells. In patients, Annexin A2 expression is increased in ER- breast cancer subtypes. Additionally, high Annexin A2 expression confers a higher probability of distant metastasis specifically for ER- patients. This work establishes a pivotal role of Annexin A2 in breast cancer progression and identifies Annexin A2 as a potential therapeutic target for the more aggressive and harder to treat ER- subtype.

The application of artificial intelligence to drug sensitivity prediction

The development of computational methods for the prediction of effective therapeutic strategies based on the genomic information of patients is the main challenge of precision medicine. Since the 21st century, next-generation sequencing (NGS) has opened up new possibilities for personalized medicine. Extensive characterization at the molecular level for hundreds of cancer cell lines has been brought to the public eye by many organizations and agencies around the world. For example, the National Cancer Institute 60 Human Cancer Cell Line Screen (NCI-60), Cancer Cell Line Encyclopedia (CCLE) and Genomics of Drug Sensitivity in Cancer (GDSC) have provided large-scale omics data such as genomic, transcriptomic and epigenomic data characterizing cancer cell lines, and The Cancer Genome Atlas (TCGA) has molecularly characterized over 20000 primary cancers of patients. Combined with the drug response data of cancer cell lines, multiomics data could be used to analyse the mechanisms of action of anticancer drugs, which could be incorporated into precision medicine strategies. Over several decades, artificial intelligence (AI) technologies based on big data have revolutionized bioinformatics. AI has built a bridge between genomics and drug sensitivity by promoting the development of predictive models for the drug response of cancer cell lines. The 2012 NCI-DREAM drug prediction challenge has been particularly influential, as the innovative applications of machine learning that emerged from it have laid the groundwork for future studies. However, classic machine learning models are still challenging in terms of predictability because they limit the systematic integration of high-dimensional multiomics data. Therefore, network-based approaches, including link prediction and network representation, have become mainstream methods for drug response prediction. On the one hand, network-based approaches have not faced the “small n, large p” problem since the multiomics features are either represented in a gene/protein network or embedded in similarity networks between cell lines. On the other hand, the introduction of gene regulatory networks (GRNs) and protein-protein interactions (PPIs) into the predictive model can provide a functional background for the integration of genomic data and thereby improve the predictive performance of drug response. In addition to network-based approaches, multimodal deep learning models can systematically integrate multiomic data by considering them as different modalities. Generally, there are three feature fusion methods in deep neural networks: Input-level feature fusion (early fusion), intermediate feature fusion and decision-level fusion (late fusion). Intermediate feature fusion is predominant in drug response prediction studies, by which features are learned separately for each type of omics data and then integrated into one unified representation to be used as the input for a classifier or a regressor. Moreover, the features of drug structures can be used as a model to improve the performance. In brief, we summarize the characteristics of publicly accessible genomic databases and discuss the trends of artificial intelligence applications in drug sensitivity prediction for cancer cell lines, including machine learning, networks and multimodal deep neural networks.

Abstract 36: Virtual screening with deep learning using cancer cell line dose-response data

Abstract Current high-throughput ligand and structural virtual screening methods typically rely on the identification of a specific functional target for the screen; however, the relationship between cancer genetic phenotype and the specific functional target is not fully understood at the onset of drug development. Cancer cell line (CCL) screening panels are continually used to study the relationship between cellular genetic and ligand structure. We propose combining the speed of machine learning-based virtual screening techniques with the greater than four million in vitro cancer cell line data points generated from the NCI-60, GDSC, and CTRP cancer cell line screens. Using deep neural networks, we train a regression neural network on an integrated dataset of CCL to predict area under the dose-response curve (AUC). Molecules are characterized using the Mordred descriptor package and with a message-passing network on the structure of the molecule, and cell lines are represented by LINCS100 RNA-seq and SNP quantification data. We present two validation styles using our model on random unseen training drugs and by unseen scaffolds. On unseen drugs, the enrichment factor is 42% for the top 1% predictions and BEDROC score is 0.25, based on binning active compounds as the bottom 2% of the overall AUC distribution from the NCI-60. Using the model as a screening tool in inference mode, we scored 1 million compounds from Enamine REAL lead-like compounds on all NCI60 and CCLE cell lines, though the entire screening results will be forthcoming (>450M compounds screened). Initial results show compounds screened by the model have a higher rate of bioactivity just based on the PubChem listings. We verify the robustness of the virtual screen by presenting common functional groups across highly scored drugs and finding stability across small perturbations both to the featurization and molecular structure. For molecules listed in PubChem, the top 10% of predicted compounds had 10x the bioactivity ranking than the bottom 10% of predictions from our model. We present a new high-throughput method for cancer drug virtual screening using CCL dose response. Our method utilizes aggregated data from in vitro CCL panels to train the dose-response model and is a multiclass model across all common CCLs. While we believe this technique is valuable for the community as a tool for integrating experimental cancer cell line data into existing virtual screening infrastructure, we acknowledge verification with in vitro testing outside the model validation data needs to be addressed in future work. Citation Format: Austin Clyde, Arvind Ramanathan, Rick Stevens. Virtual screening with deep learning using cancer cell line dose-response data [abstract]. In: Proceedings of the AACR Special Conference on Advancing Precision Medicine Drug Development: Incorporation of Real-World Data and Other Novel Strategies; Jan 9-12, 2020; San Diego, CA. Philadelphia (PA): AACR; Clin Cancer Res 2020;26(12_Suppl_1):Abstract nr 36.

Multilayered VBC score predicts sgRNAs that efficiently generate loss-of-function alleles.

CRISPR-Cas9 screens have emerged as a transformative approach to systematically probe gene functions. The quality and success of these screens depends on the frequencies of loss-of-function alleles, particularly in negative-selection screens widely applied for probing essential genes. Using optimized screening workflows, we performed essentialome screens in cancer cell lines and embryonic stem cells and achieved dropout efficiencies that could not be explained by common frameshift frequencies. We find that these superior effect sizes are mainly determined by the impact of in-frame mutations on protein function, which can be predicted based on amino acid composition and conservation. We integrate protein features into a 'Bioscore' and fuse it with improved predictors of single-guide RNA activity and indel formation to establish a score that captures all relevant processes in CRISPR-Cas9 mutagenesis. This Vienna Bioactivity CRISPR score (www.vbc-score.org) outperforms previous prediction tools and enables the selection of sgRNAs that effectively produce loss-of-function alleles.

CELLector: Genomics-Guided Selection of Cancer In Vitro Models

Selecting appropriate cancer models is a key prerequisite for maximizing translational potential and clinical relevance of invitro oncology studies. We developed CELLector: an R package and R Shiny application allowing researchers to select the most relevant cancer cell lines in a patient-genomic-guided fashion. CELLector leverages tumor genomics to identify recurrent subtypes with associated genomic signatures. It then evaluates these signatures in cancer cell lines to prioritize their selection. This enables users to choose appropriate invitro models for inclusion or exclusion in retrospective analyses and future studies. Moreover, this allows bridging outcomes from cancer cell line screens to precisely defined sub-cohorts of primary tumors. Here, we demonstrate the usefulness and applicability of CELLector, showing how it can aid prioritization of invitro models for future development and unveil patient-derived multivariate prognostic and therapeutic markers. CELLector is freely available at https://ot-cellector.shinyapps.io/CELLector_App/ (code at https://github.com/francescojm/CELLector and https://github.com/francescojm/CELLector_App).

Identification of Intrinsic Drug Resistance and Their Biomarkers in High-Throughput Pharmacogenomic and CRISPR Screens

Precision medicine is empowered by biomarker discovery. High-throughput drug screens in cancer cell lines enable the identification of drug sensitivity biomarkers, whilst resistance biomarkers remain underexplored: dissecting meaningful drug responses at high concentrations is impossible due to cytotoxicity, making non-responders and resistant cell lines indistinguishable, thus limiting resistance biomarker discovery to frequently mutated cancer genes. In order to address this, we developed a statistical method to indirectly identify resistance biomarkers. Here, we interrogate subpopulations carrying sensitivity biomarkers, and consecutively investigate resistant outliers for unique genetic alterations that putatively drive this resistance. Out of 78 cases with resistant outliers, our method predicts infrequent resistance biomarkers for 43 of them, e.g. the clinically relevant EGFRT790M mutation in lung adenocarcinoma treated with EGFR inhibitors. Interrogating the underpinnings of drug resistance with publicly available CRISPR phenotypic assays assists in prioritising resistance drivers, offering hypotheses for potentially effective drug combinations, and thereby paving the way for the next generation of precision medicine.

Abstract P1-05-13: Dissecting the interplay of the long non-coding RNA NORAD and PARP inhibitors in triple negative breast cancer

Abstract Background and Aims: Triple-negative breast cancer (TNBC) is the most aggressive subtype, and is often associated with mutations in the breast cancer susceptibility protein (BRCA) genes, which are known tumor-suppressor genes involved in DNA damage repair. Chromosomal instability (CIN) has been added to the hallmarks of cancer and is characterized by a frequent gain or loss of chromosomes during mitosis (i.e. chromosomal aberrations). BRCA1 and BRCA2 breast tumors develop by specific and distinct evolutionary paths, as their gene profiles and genome aberration spectra differ from each other and from those in sporadic BC. The lncRNA Non-Coding RNA Activated by DNA Damage (NORAD) has been demonstrated to contribute to the progression of several cancer entities by regulating genomic stability. Therefore, we wanted to explore a potential additive anti-cancerous effect by treating TNBC cells with a combination of NORAD silencing and PARP inhibition. Material and Methods: To clarify the potential clinical significance of NORAD in human TNBC, patient data from publicly available databases were analyzed. In order to further characterize NORAD, we performed phenotypic experiments after establishing an siRNA-mediated knock-down approach. Expression levels of NORAD as well as DNA-damage repair proteins were analyzed by qRT-PCR and Western Blot after combined cell treatment with NORAD silencing and PARP inhibition. To further unravel the molecular mechanism behind, we performed immunofluorescent experiments against markers for DNA-double strand breaks and homologous recombination efficiency (i.e. Rad51 and yH2AX) after TNBC treatment with NORAD siRNAs and PARP inhibitors. Results: High NORAD levels were identified to be a negative prognostic factor for disease-free survival of TNBC patients (HR with 95% CI: 2.4 (1.35-4.25); n=161; p=0.002). Furthermore, the NORAD gene is amplified in up to 3% of BC cases and mRNA levels are higher in tumor tissue when compared to healthy tissue. A qPCR based screening of several breast cancer cell lines showed an at least 4-times increased expression of NORAD when compared to normal mammary cells (i.e. MCF 12A). Knock-down of NORAD in TNBC cells resulted in a slightly reduced proliferation as well as to a decreased expression of several DNA repair proteins. Immunofluorescent experiments of TNBC cells treated with the PARP inhibitor olaparib and/or after NORAD silencing revealed increased yH2AX-dependent nuclear foci formation when compared to control conditions, indicating augmented DNA double-strand breaks after treatment. Conclusion: Our first data let us hypothesize that a combination of NORAD knock-down and PARPi treatment increases inhibition of DNA repair mechanisms and leads to a more pronounced “anti-tumorigenic” phenotype in TNBC cells. This approach could lead to the optimization of current treatment concepts. Furthermore, as increased NORAD expression is associated with an unfavorable patient survival, this lncRNA could serve as potential prognostic biomarker. Citation Format: Christiane Klec, Anita Kapeller, Felix Prinz, Christoph Trenk, Herbert Stöger, Martin Pichler. Dissecting the interplay of the long non-coding RNA NORAD and PARP inhibitors in triple negative breast cancer [abstract]. In: Proceedings of the 2019 San Antonio Breast Cancer Symposium; 2019 Dec 10-14; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2020;80(4 Suppl):Abstract nr P1-05-13.

Abstract C083: Profiling of PDK1 inhibitor SNS-510 shows potent activity in hematologic and solid tumor cancer models and identifies CDKN2A-mutated cancers as potential targets with enhanced sensitivity

Abstract Background: Phosphatidylinositol (PI) dependent kinase 1, PDK1, is a master kinase that activates kinases important in cell growth and survival including members of the AKT, PKC, RSK and SGK families. PDK1 can interact with its substrates through PI-dependent or PI-independent (PIF-mediated) mechanisms. SNS-510 is a potent inhibitor of active and inactive conformations of PDK1 and binds deep in the adaptive pocket. This interaction perturbs the PIF-pocket, thereby affecting both PI-dependent and -independent substrates (Hansen Mol Cancer Ther 2015). SNS-510 is also a potent FLT3 inhibitor. Methods and Results: SNS-510 activity was evaluated in 320 OncoPanel™ cell lines derived from 20 different cancer types and exposed for 72 hours to a range of drug concentrations (10 µM-1 nM). SNS-510 showed strong and consistent anti-proliferative activity (overall median IC50 537 nM; range 14 nM-10 µM). Cell lines with IC50 <250 nM were classified as sensitive to SNS-510, while those with IC50 >2 μM were classified as resistant. Among the 59 cell lines classified as sensitive, the most prevalent tumor origin was lymphoma followed by leukemia, kidney and lung cancer. The AML FLT3-ITD positive cell line MV4-11 was the most sensitive cell line across the entire OncoPanel™ (IC50 14 nM). Univariate analysis comparing drug response data and public genomics data for the OncoPanel™ cell lines revealed potential biomarkers associated with sensitivity and resistance to SNS-510. The presence of mutations or deletions in the CDKN2A (Cyclin Dependent Kinase Inhibitor 2A) gene was the genomic feature most significantly associated with sensitivity to SNS-510 (p=4.2E-05). Deletions in C9orf53 (CDKN2A divergent transcript), CDKN2B (Cyclin Dependent Kinase Inhibitor 2B) and CDKN2BAS (CDKN2B Antisense RNA 1) genes were also associated with significant sensitivity to SNS-510. Twenty-six (44%) sensitive cell lines were carriers of a CDKN2A mutation or deletion while only 7 (12%) resistant cell lines had a genetic defect in CDKN2A. We previously demonstrated (Hansen Mol Cancer Ther 2015) PDK1/RSK2/P70S6K/AKT pathway modulation and sustained antitumor effects in a human AML MV4-11 FLT3-ITD xenograft mouse model after SNS-510 administration PO. We have now evaluated SNS-510 antitumor activity in a human AML MOLM-16 FLT3 wild-type xenograft mouse model. Potent and dose-dependent tumor growth inhibition (range 52-97%) was observed in the MOLM16 model at SNS-510 PO dose levels of 10, 25 and 50 mg/kg qdx21. The antitumor efficacy was correlated with a profound drug-induced decrease in phospho-PDK1 (Ser241), phospho-AKT (Thr308) and phospho-RSK2 (Ser227) in MOLM16 tumor lysates. Conclusion: SNS-510 is a potent, orally bioavailable PDK1 inhibitor with broad activity in both hematologic and solid tumor cancers. SNS-510 inhibition of PDK1-signaling correlates with antitumor activity at tolerated doses and schedules in both FLT3-ITD and FLT3 wild-type AML xenograft mouse models. Profiling (anti-proliferative and molecular) in the OncoPanel cancer cell line screen identified CDKN2A mutated tumors as particularly sensitive to SNS-510, supporting a potential role by PDK1 inhibitors in combination with CDK4/6 inhibitors to reverse resistance and improve activity (Jansen Cancer Res 2017). Citation Format: Pietro Taverna, Stig Hansen, Jeff Iwig, Gene Jamieson, Judy A Fox. Profiling of PDK1 inhibitor SNS-510 shows potent activity in hematologic and solid tumor cancer models and identifies CDKN2A-mutated cancers as potential targets with enhanced sensitivity [abstract]. In: Proceedings of the AACR-NCI-EORTC International Conference on Molecular Targets and Cancer Therapeutics; 2019 Oct 26-30; Boston, MA. Philadelphia (PA): AACR; Mol Cancer Ther 2019;18(12 Suppl):Abstract nr C083. doi:10.1158/1535-7163.TARG-19-C083

Genome-Scale CRISPR-Cas9 Synthetic Lethal Screening of AML Cell Line Identified Functional Modulators of Etoposide Resistance Predictive of Clinical Outcome in AML Patients

Despite the use of intensive chemotherapeutic regimens containing ara-C, daunorubicin and etoposide (ADE), in combination with stem cell transplantation for high-risk group patients, approximately significant patients experience relapse. Resistant and relapsed disease remains the most prevalent forms of clinical failure in treating this disease. The current scientific challenge is to identify candidate genes that are truly essential to drug resistance and thus may be pharmacologically targeted in a clinically effective manner. The development of the CRISPR/cas9 genome editing tool has dramatically improved the capabilities for functional screening in multiple systems including AML where genome-wide drop-out screens in AML cancer cell lines identified AML-essential genes. None of the studies have unfortunately integrated the CRISPR/cas9 screening tool with patient outcome data. In this study, we have successfully performed Genome scale Crispr knock out GeCKO -CRISPR/cas9 screening using Brunello library (targeting 19,114 genes) in K562 cell lines followed by treatments with etoposide, dauno and cytarabine. We performed a large-scale transduction into ~30 x 106 cells at a low multiplicity of infection (0.3). After puromycin selection the mutant cell library was exposed in triplicate to IC30 concentration of chemotherapeutic agents for up to 18 days with collection of samples at day 4 (ara-C, dauno, etoposide) days 12 and 18 (dauno and etoposide). A vehicle-only exposure was similarly carried out for 18 days. Direct sequencing of PCR amplified sgRNA guides from the pooled cells by short read sequencing (Illumina) was used to quantify the representation of each knockout clone in the pooled population (schema shown in Fig 1). The relative enrichment/depletion of each knockout clone was determined by ratio of the abundance of each clone between two samples. The FASTX-Toolkit, was used to extract the unique sgRNA sequences which were assembled into a Burrows-Wheeler index using the Bowtie build-index function and number of uniquely aligned reads for each sgRNA were calculated. The MaGeCK algorithms was used to analyze the read count data and perform statistical comparisons. The results for the etoposide screening are shown in Fig 2, overall 17 genes showed significant association with responsiveness to etoposide at all the three different time points (4, 12 and 18 days). Further investigation of these candidate genes in AML patients treated on AML02 trial demonstrated consistent and significant association between expression levels of several genes identified in CRISPR screening and clinical outcome. Fig 3 shows a representative result for 2 of the genes identified in etoposide screening, ABCC1and RAD54L2 both were associated with etoposide resistance in CRISPR screening and concordantly high expression was associated with greater induction 1 MRD, inferior event free survival (EFS, Fig 3) and overall survival (OS) in the AML02 cohort. ABCC1 is a drug efflux transporter that has been implicated in resistance to etoposide and daunorubicin as well as has been associated with poor prognosis in AML consistent with our results. RAD54L2, is a helicase involved in DNA damage repair response pathway. Other genes of interest identified in etoposide screening with significant association with clinical endpoints as MRD, EFS and OS included TKT, involved in pentose phosphate pathway and implicated in regulation of metabolic switch in cancer. A nucleotide excision repair gene, RAD23B, implicated in breast cancer progression as well as in CML and ALL. Similarly, for daunorubicin and cytarabine CRISPR/cas9 synthetic lethal screening identified target genes of clinical relevance that will be discussed in the presentation. Our approach shows that using CRISPR/cas9 targeted synthetic lethal screening as a reliable approach to not only identify and establish genes predictive of drug resistance and poor outcome but also potential targets for novel drug development that could be used in combination with currently approved agents. Disclosures No relevant conflicts of interest to declare.

Discovering long noncoding RNA predictors of anticancer drug sensitivity beyond protein-coding genes

Large-scale cancer cell line screens have identified thousands of protein-coding genes (PCGs) as biomarkers of anticancer drug response. However, systematic evaluation of long noncoding RNAs (lncRNAs) as pharmacogenomic biomarkers has so far proven challenging. Here, we study the contribution of lncRNAs as drug response predictors beyond spurious associations driven by correlations with proximal PCGs, tissue lineage, or established biomarkers. We show that, as a whole, the lncRNA transcriptome is equally potent as the PCG transcriptome at predicting response to hundreds of anticancer drugs. Analysis of individual lncRNAs transcripts associated with drug response reveals nearly half of the significant associations are in fact attributable to proximal cis-PCGs. However, adjusting for effects of cis-PCGs revealed significant lncRNAs that augment drug response predictions for most drugs, including those with well-established clinical biomarkers. In addition, we identify lncRNA-specific somatic alterations associated with drug response by adopting a statistical approach to determine lncRNAs carrying somatic mutations that undergo positive selection in cancer cells. Lastly, we experimentally demonstrate that 2 lncRNAs, EGFR-AS1 and MIR205HG, are functionally relevant predictors of anti-epidermal growth factor receptor (EGFR) drug response.

Systematic Prediction of Cancer Essential Genes Using Multi-Omics Data

Background: Genome-wide phenotype screens in cancer cell lines have identified putative drug targets. However, these experiments are time-consuming and costly, and essential genes differ significantly across cell lines. Methods: Here we aimed to predict cancer essential genes using only sequencing data. We trained our machine-learning (ML) models on 470 cell lines using DNA mutation, methylation and gene expression data as input and RNAi screening results as output, and obtained models for 177 genes with predictive performance of AUC>=0.8. Findings: We applied these models to 549 cell lines, and validated the predicted essential genes using public CRISPR-Cas9 screening data in 42 of these cell lines. With the top ten genes ranked by our models, we could provide >=3 valid targets for three-quarter of the cell lines, and >=5 targets for half the cell lines. Interpretation: Although most of these 177 genes were not previously implicated in cancers, their associated genes were enriched in various cancers and diseases, indicated that the 177 genes could be ideal drug targets. Funding Statement: This work was partially supported by National Key Research and Development Program of China 2018YFC0910502 (to W.H.C.), and Shanghai Municipal Science and Technology Major Project (No.2018SHZDZX01), LCNBI and ZJLab. Declaration of Interests: All authors declare that they have no competing interests. Ethics Approval Statement: Not required.

Abstract 5097: Clinical trial outcomes for drug combination therapies can be predicted by modeling independent action using cancer cell line screens

Abstract Background: In many cancer settings, it is evident that combination drug therapies show increased efficacy compared to monotherapies. However, the development of new combination therapies is slowed by the fact that it is infeasible to experimentally evaluate the vast number of possible drug combinations. In an effort to overcome this problem, we developed a computational method which uses monotherapy screening data in cell lines to estimate the efficacy of combinations of two or more chemotherapy drugs. When validated against published phase 3 clinical cancer trial results, our method is capable of predicting clinical trial success or failure with a high degree of sensitivity and specificity. Methods and Results: We processed existing high-throughput in vitro drug screening datasets (i.e. GDSC and CTRPv2) to obtain fitted dose response curves for more than 600,000 drug-cell line pairs, including data for 695 unique compounds and 1347 unique cell lines. We then used this dataset to develop a method of estimating the efficacy of combinations of two or more drugs. While previous efforts to predict drug combination efficacy have largely focused on dose-agnostic estimates of drug synergy, we modeled dose-dependent estimates of independent drug action, where each patient treated with a drug combination only derives benefit from the single drug in the combination to which they are most sensitive. For method validation, we systematically searched clinicaltrials.gov to identify completed, phase 3 cancer clinical trials that evaluated drug combinations for which we could estimate efficacy and which had at least 50 patients per trial arm. Of the >9,000 clinical trials considered, 37 trials (11 successful and 26 unsuccessful) met our inclusion criteria. Using clinically determined plasma concentrations for each drug, we then predicted relative efficacy for each of the experimental drug combinations versus the control group therapies in these trials. These predictions were clustered into two groups (predicted high benefit vs low benefit) using k-means clustering, and these clusters were compared to whether or not the clinical trials were considered a success by the original study investigators. Our method correctly classified 84% of the trials, with a sensitivity of 73% and a specificity of 88%. Furthermore, efficacy predictions for combinations containing erlotinib were dependent on EGFR mutation status, suggesting that this method may be useful for identifying molecular subtypes which are sensitive to a given combination therapy. Conclusion: Clinically meaningful predictions of drug combination efficacy can be generated from single-agent, in vitro drug screens under the assumption of independent drug action. This enables efficacy predictions to be made for >150,000 2-drug combinations and millions of 3- and 4-drug combinations using the existing GDSC and CTRPv2 datasets. Citation Format: Alexander Ling, R. Stephanie Huang. Clinical trial outcomes for drug combination therapies can be predicted by modeling independent action using cancer cell line screens [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 5097.

Abstract 810: Differential use of lactate for mitochondria respiration by NSCLC cells

Abstract Metabolic liabilities in cancer cells provide the opportunity for novel therapeutic approaches. This is especially true as the dogma that all cancer cells are glycolytic is being challenged. Phenotypic screening of cancer cell lines can demonstrate what these liabilities are and suggest approaches for drugs that target these vulnerabilities. As an example, non-small cell lung cancer (NSCLC) is typically driven by oncogenic mutations in either KRAS or EGFR. However, the impact of these mutations on cellular metabolic phenotype is not well-studied. In this study, we observed that the two KRAS-mutated NSCLC cell lines (A549 and H460) primarily relied on glycolysis for ATP production while EGFR-mutated cell lines (H1975 and PC9) were more reliant on mitochondrial respiration to meet energy demands. There is emerging evidence to suggest that lactate is a major fuel for cancer cell energy metabolism especially in the glucose-limited tumor microenvironment. Thus, we next compared the use of lactate with other carbon sources for ATP production in NSCLC cells. We found there was differential usage of lactate for mitochondrial respiration between KRAS and EGFR-mutated NSCLC cells. EGFR-mutated NSCLC cells used lactate and increased mitochondrial respiration when lactate was acutely administrated regardless of glucose availability. In contrast, the use of lactate for the mitochondrial respiration by KRAS-mutated cells was limited. In A549 cells, no significant change in mitochondrial respiration rate was observed at all by acute injection of lactate. Together, these results imply that NSCLC variants have adopted different metabolic phenotypes depending on the oncogenic background and more oxidative phenotype may be correlated to higher lactate use for TCA cycle in production of ATP. Citation Format: Pamela Swain, Natalia Romero, Yoonseok Kam, Brian P. Dranka. Differential use of lactate for mitochondria respiration by NSCLC cells [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 810.

Abstract 2470: A molecular oncology almanac for integrative clinical interpretation of molecular profiles to guide precision cancer medicine

Abstract Background: Tumor molecular profiling is increasingly used to detect first-order genomic alterations associated with therapeutic actions (e.g. BRAF V600E & RAF/MEK inhibition). Simultaneously, more complex molecular features are being discovered and applied to clinical scenarios (e.g. mutational signatures, somatic-germline interactions). As patients receive expanded profiling, such as clinical whole-exome and RNA sequencing, novel algorithms are needed to integrate interpretation of multiple data modalities. Furthermore, the clinical-preclinical gap continues to widen as data from high-throughput screens of cancer cell lines are generated without accessibility at the point of care. Here, we introduce a paired interpretation algorithm and knowledge system for cancer genomic data, the Molecular Oncology Almanac, to inform treatment decisions through rapid assessment of tumor actionability. Methods: We implemented a cloud-based interpretation algorithm that annotates and evaluates variants from WES and RNA-seq (SNVs from WES and RNA-seq, InDels, CNAs, and fusions) and infers additional features such as mutational burden, mutational signatures, MSI, somatic-germline interactions, and aneuploidy. Predictive implication levels were assigned to reflect confidence in the database's catalogued relationships to therapeutic response and prognosis for each molecular feature. We also developed a patient-preclinical matchmaker function to expand the theoretical therapeutic modalities for any given patient. Towards timeliness of updates and knowledge system accessibility, we developed API endpoints, a browser extension for suggesting citations, and workflows in the FireCloud framework. Results: A total of 260 patients with metastatic castration-resistant prostate cancer (n=150) and metastatic melanoma (n=110) were evaluated with 569 alteration-action relationships catalogued in the Molecular Oncology Almanac. Overall 80% of patients had at least one alteration suggesting therapeutic sensitivity based on FDA approval, clinical trials, or studies in humans; which increased to 95.8% by also considering preclinical and inferential associations. Per patient, the matchmaker function on average highlighted 1.56 additional therapies that would not have otherwise been nominated. At least one feature associated with resistance or prognosis was observed in 85% and 90% of patients, respectively. Conclusion: Clinical actionability of sequence data was increased by including integrative molecular profiling of DNA and RNA, global molecular features, and preclinical alteration-action relationships. Increased accessibility of clinical interpretation through our cloud-based web portals and API endpoints may aid in sample contextualization. Source code and a web portal for this project are available at moalmanac.org. Citation Format: Brendan Reardon, Nicholas Moore, Nathanael Moore, Eric Kofman, Eliezer Van Allen. A molecular oncology almanac for integrative clinical interpretation of molecular profiles to guide precision cancer medicine [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 2470.

Abstract 2560: Werner syndrome helicase is a selective vulnerability of microsatellite instability-high tumor cells

Abstract Targeted cancer therapy is based on exploiting selective dependencies of tumor cells. By leveraging recent large-scale genomic profiling and functional screening of cancer cell lines we identified Werner syndrome helicase (WRN) as a novel specific vulnerability of microsatellite instability-high (MSI-H) cancer cells. MSI, caused by defective mismatch repair is frequently detected in human malignancies, in particular in colorectal, endometrial and gastric cancers. We demonstrate that WRN inactivation selectively impairs the viability of MSI-H but not microsatellite stable (MSS) colorectal and endometrial cancer cell lines. In MSI-H cells, WRN loss results in the emergence of chromosome breaks, chromatin bridges and micronuclei highlighting defective genome integrity. WRN variants harboring mutations abrogating the ATPase function of WRN helicase fail to rescue the viability phenotype of WRN-depleted MSI-H colorectal cells. Our study suggests that pharmacological inhibition of WRN helicase function might represent a novel opportunity to develop a targeted therapy for MSI-H cancers. Citation Format: Simone Lieb, Silvia Blaha-Ostermann, Elisabeth Kamper, Katharina Ehrenhöfer-Wölfer, Andreas Schlattl, Andreas Wernitznig, Jesse Lipp, Kota Nagasaka, Gerd Bader, Ralph Neumueller, Norbert Kraut, Mark Pearson, Simon Woehrle, Mark Petronczki. Werner syndrome helicase is a selective vulnerability of microsatellite instability-high tumor cells [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 2560.

Abstract 2883: Mithramycin inhibits KDM6A (UTX) and SWI/SNF to induce apoptosis in malignant rhabdoid tumor

Abstract BACKGROUND: Rhabdoid tumor (RT) is an aggressive pediatric malignancy defined by the genetic deletion of SMARCB1 (BAF47/SNF5/INI1), a subunit of the SWI/SNF chromatin remodeling complex. SMARCB1 loss redistributes the SWI/SNF complex to favor a proliferative non-differentiated cellular state. Nevertheless, while EZH2, the catalytic subunit of PRC2, is an established, synthetically lethal therapeutic target, other vulnerabilities that result from this imbalance between the SWI/SNF and PRC2 complexes are relatively unexplored. METHODS: In a screen of multiple pediatric cancer cell lines, we identified a heightened sensitivity of RT cells to mithramycin (MMA), a small molecule transcription inhibitor. We hypothesized the hypersensitivity of rhabdoid tumor to mithramycin was due to the disruption of SWI/SNF and PRC2 dynamics. Here, we describe an epigenetic mechanism of action for mithramycin in rhabdoid tumor. We characterize KDM6A, a H3K27me3 histone demethylase, as a novel therapeutic vulnerability in rhabdoid tumor with live cell imaging, western blot, small molecule inhibition, and chromatin immunoprecipitation (ChIP) sequencing. We identify SWI/SNF eviction from chromatin as a novel mechanism of action for mithramycin with biochemical fractionation, ChIP-seq, and ATAC-seq. Finally, we demonstrate these findings in vivo using RT xenograft models. RESULTS: Mithramycin induced apoptosis and a striking increase in H3K27me3, the catalytic mark of PRC2, in rhabdoid tumor cell lines in a dose and time-dependent manner. This amplification of H3K27me3 preceded suppression of cellular proliferation and induction of apoptosis suggesting H3K27me3 amplification may drive the apoptotic phenotype. In addition, knockdown of KDM6A, a H3K27me3 histone demethylase, potentiates the effects of MMA providing further evidence that KDM6A loss and H3K27me3 amplification lead to apoptosis. We have identified SP1 as the transcription factor driving KDM6A and mithramycin interferes with SP1-dependent activation of KDM6A by evicting SWI/SNF from chromatin. We are currently working on the mechanism of action for mithramycin in rhabdoid tumor cells, including H3K27me3 ChIP-seq and ATAC-seq to elucidate global transcription and chromatin changes. We have recapitulated these results in vivo with xenograft mouse models. CONCLUSIONS: Mithramycin inhibits KDM6A, a histone demethylase, and SWI/SNF leading to apoptosis through the amplification of H3K27me3. Eviction of SWI/SNF from chromatin is a novel mechanism of action for mithramycin. Overall, this study indicates KDM6A and SWI/SNF loss are novel therapeutic vulnerabilities in rhabdoid tumor and mithramycin may be a promising clinical candidate for the treatment of rhabdoid tumor. Citation Format: Maggie H. Chasse, Elissa Boguslawski, Katie Sorensen, Ian Beddows, Kristin Rybski, Zachary Madaj, Susan M. Kitchen-Goosen, Patrick J. Grohar. Mithramycin inhibits KDM6A (UTX) and SWI/SNF to induce apoptosis in malignant rhabdoid tumor [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 2883.