Large Drug Screening Research Articles

Stromal cells engineered to express T cell factors induce robust CLL cell proliferation in vitro and in PDX co-transplantations allowing the identification of RAF inhibitors as anti-proliferative drugs

Several in vitro models have been developed to mimic chronic lymphocytic leukemia (CLL) proliferation in immune niches; however, they typically do not induce robust proliferation. We prepared a novel model based on mimicking T-cell signals in vitro and in patient-derived xenografts (PDXs). Six supportive cell lines were prepared by engineering HS5 stromal cells with stable expression of human CD40L, IL4, IL21, and their combinations. Co-culture with HS5 expressing CD40L and IL4 in combination led to mild CLL cell proliferation (median 7% at day 7), while the HS5 expressing CD40L, IL4, and IL21 led to unprecedented proliferation rate (median 44%). The co-cultures mimicked the gene expression fingerprint of lymph node CLL cells (MYC, NFκB, and E2F signatures) and revealed novel vulnerabilities in CLL-T-cell-induced proliferation. Drug testing in co-cultures revealed for the first time that pan-RAF inhibitors fully block CLL proliferation. The co-culture model can be downscaled to five microliter volume for large drug screening purposes or upscaled to CLL PDXs by HS5-CD40L-IL4 ± IL21 co-transplantation. Co-transplanting NSG mice with purified CLL cells and HS5-CD40L-IL4 or HS5-CD40L-IL4-IL21 cells on collagen-based scaffold led to 47% or 82% engraftment efficacy, respectively, with ~20% of PDXs being clonally related to CLL, potentially overcoming the need to co-transplant autologous T-cells in PDXs.

Integration of Computational Docking into Anti-Cancer Drug Response Prediction Models.

Cancer is a heterogeneous disease in that tumors of the same histology type can respond differently to a treatment. Anti-cancer drug response prediction is of paramount importance for both drug development and patient treatment design. Although various computational methods and data have been used to develop drug response prediction models, it remains a challenging problem due to the complexities of cancer mechanisms and cancer-drug interactions. To better characterize the interaction between cancer and drugs, we investigate the feasibility of integrating computationally derived features of molecular mechanisms of action into prediction models. Specifically, we add docking scores of drug molecules and target proteins in combination with cancer gene expressions and molecular drug descriptors for building response models. The results demonstrate a marginal improvement in drug response prediction performance when adding docking scores as additional features, through tests on large drug screening data. We discuss the limitations of the current approach and provide the research community with a baseline dataset of the large-scale computational docking for anti-cancer drugs.

Mechanisms of Primaquine Induced Hemolysis in a Novel Humanized Murine Model of Mediterranean G6PD Deficiency

Background: Approximately half a billion individuals are genetically deficient in glucose 6-phosphate dehydrogenase (G6PD) due to amino acid variants generally thought to decrease enzyme half-life. Decreased G6PD activity limits the antioxidant capacity of red blood cells (RBCs). As such, certain drugs that induce oxidative stress can cause life-threatening hemolysis in G6PD deficient (G6DPd) individuals. Standard regiments of the only approved drugs that can cure liver phase Plasmodium vivax ( p. vivax) (i.e. primaquine and tafenoquine) are contraindicated in G6PDd patients due to risks of hemolysis. Effective treatment of G6PDd individuals infected with p. vivax requires a prolonged course of primaquine (8 weeks), leading to non-adherence, and subsequent reduced efficacy and increased risk of transmission. Despite decades of research and large drug screening programs, no non-hemolytic compound has been identified that can eradicate p. vivax. Lack of progress in this area has likely been hampered by an incomplete mechanistic understanding of primaquine induced hemolysis, in part due to the lack of an animal model that recapitulates G6PD instability. Methods: Recombinant non-deficient human hG6PD(ND) or the deficient Mediterranean variant (hG6PD(MED)) were expressed, purified, and subjected to thermal proteome profiling in combination with TMT10-assisted quantitative cross-linking proteomics. Mice were generated in which murine genomic G6PD is replaced with either genomic human hG6PD(ND) or hG6PD(MED). Hemolysis in humans is caused by primaquine metabolites (e.g. primaquine-5,6-orthoquinone (5,6-POQ)) and not primaquine itself. As such, RBCs from both strains were exposed to 5,6-POQ and then analyzed by 1) high resolution metabolomics and proteomics, 2) electron paramagnetic resonance (EPR) to measure superoxide, 3) methemoglobin (MetHb) formation, 4) light and electron microscopy, 5) in vitro hemolysis, and 6) in vivo hemolysis (clearance) after infusion into animals. In vivo pulse chase biotinylation allowed visualization and identification of younger (1-6 days) RBCs vs. older (7-55 day) RBCs. Results: Recombinant hG6PD(MED) enzyme had a 2.6-fold lower specific activity compared to hG6PD(ND). Thermal gradient crosslinking proteomics identified specific amino acid spacing (less than 26 angstroms) that was higher in hG6PD(ND) [82-95 and 91-432] vs. higher in hG6PD(MED) [89-205, 97-429, and 429-497]. Analysis of mice demonstrated that hG6PD(MED) RBCs had 5% of G6PD activity as compared to hG6PD(ND) RBCs, with an RBC age dependent decrease in G6PD activity in hG6PD(MED) RBCs. Exposure to 5,6-POQ caused increased superoxide generation, increased MetHb formation, alterations to morphology, in vitro hemolysis, and in vivo hemolysis of hG6PD(MED) RBCs compared to hG6PD(ND) RBCs. However, younger hG6PD(MED) RBCs with higher G6PD activity were highly resistant to 5,6-POQ induced hemolysis (both in vivo and in vitro). All experiments were carried out a minimum of 3 times and all stated findings were statistically significant (p<0.05). Proteomic and metabolomic analysis (with combined samples from multiple experiments) demonstrated oxidation of glyceraldehyde 3-phosphate dehydrogenase (GAPDH), resulting in a block in glycolysis that increases dihydroxyacetone phosphate (DHAP) leading to cytotoxic methylglyoxal that broadly oxidizes protein lysines to carboxymethyllysine (CML) - the first step in advanced glycation end products (AGEs) known to facilitate RBC clearance. Discussion: We identify specific conformational changes caused by the hG6PD(MED) variant resulting in decreased specific activity and half-life, and confirm that G6PD activity decreases with RBC age, in a humanized mouse model. The resistance of young hG6PD(MED) RBCs to 5,6-POQ has immediate practical ramifications, in that it supports the rationale of clinical studies currently proposing new dose escalation primaquine regimens to induce low level early hemolysis, leading to reticulocytosis, and decreasing the mean age of circulating RBCs such that increased parricidal doses of primaquine can then be tolerated in G6PDd patients. The elucidation of the metabolic and proteomic lesion (under conditions confirmed to cause 5,6-POQ induced hemolysis of hG6PD(MED) but not hG6PD(ND) RBCs) provides a mechanistic landscape to guide development of non-hemolytic drugs to treat p. vivax.

Efficacy of convection enhanced delivery of MTX110 (soluble panobinostat) in preclinical Diffuse Intrinsic Pontine Glioma models using metabolic hyperpolarized 13C imaging

BackgroundDiffuse intrinsic pontine glioma (DIPG) usually occurs in children and has poor outcomes despite treatment. Large drug screens have identified the pan-histone deacetylate inhibitor panobinostat as a promising agent, however clinical response might be hampered by limited blood brain barrier penetration. Hyperpolarized 13C magnetic resonance (MR) metabolic imaging has successfully been applied to non-invasively assess metabolic activity of cancer therapies. Here, we use in vitro and in vivo DIPG models to validate the therapeutic efficacy of MTX110, an aqueous form of panobinostat delivered by convection enhanced delivery (CED) and apply metabolic imaging. MethodsMTX110 inhibitory effect was assessed in 11 DIPG cell lines. Caspase 3/7 levels were measured to assess mode of cell death. FACS analysis was utilized to determine impact on cell cycle. Hyperpolarized 13C imaging determined changes in pyruvate and lactate levels after treatment. In vivo activity of CED of MTX110 was assessed in a patient-derived xenograft rat model and tissue half-life of MTX110 was determined using mass spectrometry. ResultsMTX110 showed similar IC50 to panobinostat ranging from 5.34 nM and 47.96 nM. Anti-proliferative effects of MTX110 are mediated by G1 cell cycle arrest and subsequent apoptosis. Drug treatment led to reduced pyruvate to lactate conversion. CED of MTX110 significantly prolonged survival of tumor-bearing rats (p = 0.0372) with no signs of systemic toxicity. Tissue half-life after single CED of MTX110 is 2 h. ConclusionsOur results demonstrate that CED of MTX110, has potent antitumor activity with limited systemic toxicity and that hyperpolarized 13C imaging is able to assess metabolic impact.

SynBa: improved estimation of drug combination synergies with uncertainty quantification.

There exists a range of different quantification frameworks to estimate the synergistic effect of drug combinations. The diversity and disagreement in estimates make it challenging to determine which combinations from a large drug screening should be proceeded with. Furthermore, the lack of accurate uncertainty quantification for those estimates precludes the choice of optimal drug combinations based on the most favourable synergistic effect. In this work, we propose SynBa, a flexible Bayesian approach to estimate the uncertainty of the synergistic efficacy and potency of drug combinations, so that actionable decisions can be derived from the model outputs. The actionability is enabled by incorporating the Hill equation into SynBa, so that the parameters representing the potency and the efficacy can be preserved. Existing knowledge may be conveniently inserted due to the flexibility of the prior, as shown by the empirical Beta prior defined for the normalized maximal inhibition. Through experiments on large combination screenings and comparison against benchmark methods, we show that SynBa provides improved accuracy of dose-response predictions and better-calibrated uncertainty estimation for the parameters and the predictions. The code for SynBa is available at https://github.com/HaotingZhang1/SynBa. The datasets are publicly available (DOI of DREAM: 10.7303/syn4231880; DOI of the NCI-ALMANAC subset: 10.5281/zenodo.4135059).

DIPG-22. TRIPTOLIDE AS A POTENTIALLY PROMISING THERAPY FOR DIFFUSE MIDLINE GLIOMA

Abstract Diffuse Midline Gliomas (DMG) are amongst the most lethal pediatric brain tumors. Children diagnosed with DMG face a grim prognosis of 9-11 months from diagnosis. Over 85% of DMG patients harbor mutation affecting histone-encoding genes (H3.1K27M, H3.3K27M) leading to genome hypomethylation and transcriptional disruption. Despite decades of clinical studies, the development of an effective therapy remains a challenge. Triptolide is a diterpenoid triepoxide compound inhibiting cell growth and exhibiting preclinical antitumor activity. A large drug screen conducted by the NCI using DMG cell lines identified triptolide as a candidate for reducing DMG cell viability. To further investigate triptolide and its potential for clinical intervention, we used DMG preclinical in vitro and in vivo models. In vitro viability assays with triptolide alone and in combination with other therapeutic drugs (ONC201 and ONC206), were carried out using seven DMG cell lines. Single treatment with triptolide resulted in a significant decrease in DMG cell viability and combination with both ONC201 and ONC206 showed additivity. Triptolide’s toxicity was assessed in vivo through direct injections into the yolk sac of zebrafish. Preliminary data demonstrated how triptolide has higher toxicity compared to un-injected and DMSO control groups. However, a major limitation of triptolide’s clinical utility is its low solubility. Therefore, for in vivo studies, we used minnelide, a water-soluble analog of triptolide, with potent antitumor properties. Studies were performed in eleven RAG2 female mice. PDX tumor-bearing mice were treated with minnelide and ONC201 for three weeks as single or in combination. Survival studies showed no significance in minnelide monotherapy. However, minnelide + ONC201 combination therapy resulted in significant (p=0.01) survival of PDX models. Our current studies investigate the efficacy of a higher dose of minnelide in vivo. Our goal is to establish minnelide potential as a combination strategy for the treatment of children diagnosed with DMG.

Abstract 786: STAG2 acts as a transcriptional repressor and promotes invasion of muscle invasive bladder cancer cells

Abstract STAG2 (Stromal Antigen 2) functions in chromatid cohesion, DNA damage repair and genome organization, but its role in muscle invasive bladder cancer (MIBC) remains unknown. We have previously found that in MIBC, loss of STAG2 protein expression is associated with better overall survival (n=169; p=0.049) and progression free survival (n=169; p=0.016). Based on these retrospective analyses, we hypothesized that STAG2 promotes an aggressive cell phenotype through its genomic interactions and transcriptomic regulation in MIBC. First, to study the effects of STAG2 on gene expression, we stably knocked down (KD) STAG2 in T24 MIBC cells using two short hairpin RNAs. Scrambled shRNA served as a control in all experiments. Second, using KD and control T24 cell lines, we performed RNA and chromatin-immunoprecipitation sequencing, then integrated these results utilizing the Cistrome analysis algorithm (Wang, S. et al., 2013) to determine how STAG2 regulates gene expression at its genomic binding sites. STAG2 KD led to the differential expression of 2158 genes, with 648 overlapping between the two shRNA cell lines. Through Cistrome analysis, we discovered that genes with increased expression after STAG2 KD were enriched for STAG2 binding sites (p=0.000503), yet genes with decreased expression did not show significant enrichment (p=0.873). This suggests that STAG2 functions as a transcriptional repressor, and subsequently we focused on genes with increased expression after STAG2 KD. We identified the significantly upregulated gene Reelin (RELN) (log2FC=2.89, 1.86; p<1*10-42, 1*10-17 for each shRNA, respectively), which has a well-established role in cell migration and invasion. To investigate this in vitro in MIBC cells, we performed time lapse microscopy and invasion assays to quantitatively determine cell movement over time. Compared to controls, T24 cells with STAG2 KD had reduced displacement (78 vs 114 µm, p<0.05), speed (0.30 vs 0.41 µm/min, p<0.05) and invasion (137 vs 190 cells/field, p<0.001) in vitro. Altogether, our results indicate that STAG2 functions as a transcriptional repressor and promotes movement and invasion of MIBC cells. This may explain why STAG2 protein loss leads to better outcomes for MIBC patients, and points to STAG2 protein expression as a potential prognostic biomarker. In our current work, we are utilizing a large drug screen of FDA-approved agents to identify vulnerabilities of tumor cells lacking STAG2. These studies will help to identify drugs that augment current standard of care and benefit the large proportion of patients with STAG2-mutant tumors. Citation Format: Sarah R. Athans, Nithya Krishnan, Swathi Ramakrishnan, Eduardo Cortes Gomez, Sofia Lage-Vickers, Monika Rak, Zara Kazmierczak, Aimee Stablewski, Kristopher Attwood, Jianmin Wang, Anna Woloszynska. STAG2 acts as a transcriptional repressor and promotes invasion of muscle invasive bladder cancer cells [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 786.

Abstract 3264: Ex vivo treatment in high grade serous ovarian cancer demonstrates the benefit of EZH2 inhibition in combination with standard therapy

Abstract While high-grade serous ovarian cancers (HGSOC) are initially sensitive to platinum-based therapy, resistance to chemotherapy remains a major challenge. Recently, the role of EZH2 in establishing a chemo-resistant state has been described in small cell lung cancer with a potential mechanism thought to involve the silencing of SLFN111. Preclinical studies showed that EZH2 inhibition results in re-expression of SLFN11 protein and restores sensitivity to DNA damaging agents. A study in ovarian cancer2 emphasized the need to investigate the effect further. Here we describe the response of platinum-resistant HGSOC patient-derived xenograft (PDX) tumors to CPI-0209, an investigational new drug that is currently being explored for its potential as a reversible second generation EZH2 inhibitor as monotherapy and in combination with carboplatin or PARP inhibition. Four luciferized PDXs (DF101, DF149, DF181, and DF216) were selected based on known platinum sensitivity and BRCA mutation status. PDX tumor ascites were harvested from mice at the IVIS signal intensity of 1011. Xenograft-derived organotypic tumor spheroids (xDOTS3) were loaded into 3D devices (AIM Biotech), treated with either CPI-0209, carboplatin, or olaparib alone, or with CPI-0209 for 7 days prior to adding carboplatin or olaparib for an additional 5-11 days. H3K27me3 was assessed by immunofluorescence. Expression of SLFN11 following treatment was assessed by western blot. Activity was measured at endpoint by dual-labeling fluorescence microscopy. EZH2-dependent target engagement via demethylation of H3K27me3 was seen in all models. While DF181 (BRIP1m) showed resistance to all treatments, DF216 (BRCAwt) showed a consistent trend in reduction in live area in the combination of carboplatin or olaparib with CPI-0209 vs single agent, and significant reduction (>20%; P<0.05) vs control. DF101 (BRCAm) and DF149 (BRCAwt), which showed no evidence of sensitivity to carboplatin and CPI-0209 monotherapy, showed sensitivity to olaparib which was not enhanced by the addition of CPI-0209. Further assessment of DF216 via western blot analysis showed elevated expression of SLFN11 after 7 days of EZH2i treatment. Since PDX in vivo studies are time-consuming and costly, it is impractical to use them for large drug screens. Here we show that our short-term and efficient xDOTS platform allows for rapid evaluation of multiple targeted therapies or therapeutic combinations. The results suggested that CPI-0209 directed combination therapy could enhance activity over standard chemotherapy alone in a platinum-resistant model of ovarian cancer ex vivo. Increased SLFN11 protein level supported the importance of EZH2-SLFN11 axis. Further exploration of CPI-0209 combination in ovarian cancer in known genomic contexts is warranted. Citation Format: Ha V. Vo, Qing Zeng, David A. Barbie, Prafulla C. Gokhale, Elizabeth Adams, Cloud P. Paweletz, Elena Ivanova. Ex vivo treatment in high grade serous ovarian cancer demonstrates the benefit of EZH2 inhibition in combination with standard therapy [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 3264.

Drug response prediction in patient-derived xenografts with data augmentation and multimodal deep learning.

e13572 Background: Prediction of drug response is a critical research area in precision oncology and has been previously explored with large drug screening studies of cancer cell lines (CCLs). Patient-derived xenografts (PDXs) are an appealing platform for preclinical drug studies because the in vivo environment of PDXs helps preserve tumor heterogeneity and usually better mimics drug response of patients with cancer compared to CCLs. Methods: We investigate multimodal neural network (NN) and data augmentation for drug response prediction in PDXs. The multimodal NN learns to predict response using drug descriptors, gene expressions (GE), and histology whole-slide images (WSIs) where the multi-modality refers to tumor features only. The NN uses late integration where separate subnetworks are used to encode the input feature types before concatenation and prediction layers. Median tumor volume per treatment group is assessed relative to the control group to create a binary variable representing response. The data include twelve single-drug and 36 drug-pair treatments resulting in 2,556 single-drug and 2,203 drug-pair response values. Pathology and omics data from 487 PDXs from NCI's Patient Derived Models Repository are used as tumor feature model inputs. We explore whether the integration of WSIs with GE improves predictions as compared with models that use GE alone. We use two methods to address the limited number of response values in the dataset: 1) homogenize drug representations which allows to combine single-drug and drug-pairs into a single dataset, 2) augment drug-pair samples by switching the order of drug features which doubles the sample size of all drug-pair samples. These methods enable us to combine single-drug and drug-pair treatments which results in 6,962 responses, allowing us to train multimodal and unimodal NNs without changing architectures or the dataset. Results: Prediction performance of three unimodal NNs which use GE (um1, um2, and um3) are compared to assess the contribution of data augmentation methods. NN um1 that uses the full dataset which includes the original and the augmented drug-pair treatments as well as single-drug treatments significantly outperforms NNs (p-values < 0.01) that ignore either the augmented drug-pairs (um2) or the single-drug treatments (um3). In assessing the contribution of multimodal learning, results show that the multimodal NN (mm) outperforms both unimodal NNs that ignore either the GE (um4) or the WSIs (um1). However, the improvement of mm over um1 is not statistically significant (p-value < 0.26). Conclusions: Our results show that data augmentation and integration of histology images and GE can help improve prediction performance of drug response in PDXs.[Table: see text]

Inhibiting the Activity of NADPH Oxidase in Cancer

Significance: The primary function of NADPH oxidases (NOX1-5 and dual oxidases DUOX1/2) is to produce reactive oxygen species (ROS). If inadequately regulated, NOX-associated ROS can promote oxidative stress, aberrant signaling, and genomic instability. Correspondingly, NOX isoforms are known to be overexpressed in multiple malignancies, thus constituting potential therapeutic targets in cancer. Recent Advances: Multiple genetic studies aimed at suppressing the expression of NOX proteins in cellular and animal models of cancer have provided support for the notion that NOXs play a pro-tumorigenic role. Further, large drug screens and rational design efforts have yielded inhibitor compounds, such as the diphenylene iodonium (DPI) analog series developed by our group, with increased selectivity and potency over "first generation" NOX inhibitors such as apocynin and DPI. Critical Issues: The precise role of NOX enzymes in tumor biology remains poorly defined. The tumorigenic properties of NOXs vary with cancer type, and precise tools, such as selective inhibitors, are needed to deconvolute NOX contribution to cancer development. Most NOX inhibitors developed to date are unspecific, and/or their mechanistic and pharmacological characteristics are not well defined. A lack of high-resolution crystal structures for NOX functional domains has hindered the development of potent and selective inhibitors. Future Directions: In-depth studies of NOX interactions with the tumor microenvironment (e.g., cytokines, cell-surface antigens) will help identify new approaches for NOX inhibition in cancer.

Identifying potential targets for prevention and treatment of amyotrophic lateral sclerosis based on a screen of medicare prescription drugs

Background: Few well-established factors are associated with risk of amyotrophic lateral sclerosis (ALS). We comprehensively evaluate prescription drugs use in administrative health claims from U.S. Medicare beneficiaries in relation to ALS risk to generate hypotheses for further research. Methods: This is a population-based case–control study of 10,450 U.S. Medicare participants (ages 66–89 years) diagnosed with ALS, based on Medicare Parts A and B fee-for-service claims, between 1 January 2008, and 31 December 2014, and 104,500 controls (1:10 ratio) frequency-matched on age, sex, and selection year. Odds ratios (ORs) for the ALS association with 685 prescription drugs were estimated using logistic regression models for both a one- and three-year lag period. Covariates included demographic characteristics and key comorbidities, among other factors. Prescription drug use was based on Medicare Part D claims. We adjusted for multiple comparisons using a Bonferroni correction. Additional a priori analyses of sex hormone drugs were also undertaken. Results: In the large drug screen, we found 10 drugs significantly associated with lower ALS risk after the multiple-testing correction in a one-year and three-year lag analysis. These included several drugs for hypertension, diabetes, and cardiovascular disease. In a separate a priori inquiry of sex hormone drugs, tamoxifen was related to lower ALS risk, and testosterone to a higher risk in women. Conclusions: These associations warrant replication in databases that include information on the severity and duration of medical conditions underlying drug use, and drug use over a longer portion of individuals’ lifespans, to further help evaluate confounding by indication.

Abstract 3883: Gene expression predicts ex vivo drug sensitivity in acute myeloid leukemia

Abstract Introduction: Ex vivo drug sensitivity studies of samples derived from acute myeloid leukemia (AML) patients have been shown to be predictive of in vivo response. These findings are based on a limited number of well-characterized agents for which in vivo patient response data and ex vivo drug sensitivity data—on that same patient—are available. To show the feasibility of scaling such ex vivo studies to large drug screens, we characterized the reproducibility of expression-based models of drug response across two independent data sets—one generated at the Oregon Health and Science University (OHSU) and the second at the Institute for Molecular Medicine Finland (FIMM). Methods: We harmonized two large-scale AML ex vivo studies screened for drug response and profiled transcriptomically—OHSU (303 AML patient samples and 160 drugs) and FIMM (48 AML samples and 480 drugs). The two panels have 94 drugs in common. Log-logistic curves were fit to the dose-response data and area under the dose-response curves (AUCs) were calculated. Predictive modeling using Ridge regression or an integrative Bayesian approach was performed for each drug AUC independently using 202 highly-variable and/or cancer-associated genes as features. Results: For each of the 94 drugs in common between the two data sets, we trained a Ridge regression model on the OHSU data set, used the model to predict response in the FIMM data set, and calculated the Pearson correlation between the predicted and observed FIMM responses. 41 of the 94 drug models had a positive and statistically significant correlation [false discovery rate (FDR) < 20%; mean ρ = 0.43; 95% CI = 0.29 – 0.77]. Drugs corresponding to the top decile of these significant models (mean ρ = 0.54; 95% CI = 0.48 – 0.77) clustered into four primary classes: MEK inhibitors (PD184352, Selumetinib, and Trametinib), EGFR/VEGFR inhibitors (Cabozantinib, Erlotinib, Foretinib, and Sorafenib), and singletons Venetoclax and Sirolimus. To confirm these results, we applied a second modeling approach—an integrative Bayesian machine learning method—that allows systematic combination of both data sets. Training and evaluation of this approach using 10-fold cross validation yielded 82 positive and statistically significant correlations (FDR < 20%; mean ρ = 0.35; 95% CI = 0.13 – 0.58). Five of nine drugs (Cabozantinib, Selumetinib, Sirolimus, Sorafenib, and Trametinib) corresponding to the top decile of these significant models (mean ρ = 0.54; 95% CI = 0.49 – 0.60) overlapped with drugs from the top decile of Ridge results (one-sided Fisher p = 2.5 x 10-4) Conclusions: Our results using independent data sets and two statistical approaches suggest that certain drugs (including MEK and EGFR/VEGFR inhibitors) are amenable to expression-based predictive modeling in AML. Future work will focus on inferring individual biomarkers of response. Citation Format: Brian S. White, Suleiman A. Khan, Muhammad Ammad-ud-din, Swapnil Potdar, Mike J. Mason, Cristina E. Tognon, Brian J. Druker, Caroline A. Heckman, Olli P. Kallioniemi, Stephen E. Kurtz, Kimmo Porkka, Jeffrey W. Tyner, Tero Aittokallio, Krister Wennerberg, Justin Guinney. Gene expression predicts ex vivo drug sensitivity in acute myeloid leukemia [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 3883.

Deep Learning for Drug Discovery and Cancer Research: Automated Analysis of Vascularization Images.

Likely drug candidates which are identified in traditional pre-clinical drug screens often fail in patient trials, increasing the societal burden of drug discovery. A major contributing factor to this phenomenon is the failure of traditional in vitro models of drug response to accurately mimic many of the more complex properties of human biology. We have recently introduced a new microphysiological system for growing vascularized, perfused microtissues that more accurately models human physiology and is suitable for large drug screens. In this work, we develop a machine learning model that can quickly and accurately flag compounds which effectively disrupt vascular networks from images taken before and after drug application in vitro. The system is based on a convolutional neural network and achieves near perfect accuracy while committing potentially no expensive false negatives.

Exploitation of the Apoptosis-Primed State of MYCN-Amplified Neuroblastoma to Develop a Potent and Specific Targeted Therapy Combination

SummaryFewer than half of children with high-risk neuroblastoma survive. Many of these tumors harbor high-level amplification of MYCN, which correlates with poor disease outcome. Using data from our large drug screen we predicted, and subsequently demonstrated, that MYCN-amplified neuroblastomas are sensitive to the BCL-2 inhibitor ABT-199. This sensitivity occurs in part through low anti-apoptotic BCL-xL expression, high pro-apoptotic NOXA expression, and paradoxical, MYCN-driven upregulation of NOXA. Screening for enhancers of ABT-199 sensitivity in MYCN-amplified neuroblastomas, we demonstrate that the Aurora Kinase A inhibitor MLN8237 combines with ABT-199 to induce widespread apoptosis. In diverse models of MYCN-amplified neuroblastoma, including a patient-derived xenograft model, this combination uniformly induced tumor shrinkage, and in multiple instances led to complete tumor regression.

A Semi-Supervised Approach for Refining Transcriptional Signatures of Drug Response and Repositioning Predictions.

We present a novel strategy to identify drug-repositioning opportunities. The starting point of our method is the generation of a signature summarising the consensual transcriptional response of multiple human cell lines to a compound of interest (namely the seed compound). This signature can be derived from data in existing databases, such as the connectivity-map, and it is used at first instance to query a network interlinking all the connectivity-map compounds, based on the similarity of their transcriptional responses. This provides a drug neighbourhood, composed of compounds predicted to share some effects with the seed one. The original signature is then refined by systematically reducing its overlap with the transcriptional responses induced by drugs in this neighbourhood that are known to share a secondary effect with the seed compound. Finally, the drug network is queried again with the resulting refined signatures and the whole process is carried on for a number of iterations. Drugs in the final refined neighbourhood are then predicted to exert the principal mode of action of the seed compound. We illustrate our approach using paclitaxel (a microtubule stabilising agent) as seed compound. Our method predicts that glipizide and splitomicin perturb microtubule function in human cells: a result that could not be obtained through standard signature matching methods. In agreement, we find that glipizide and splitomicin reduce interphase microtubule growth rates and transiently increase the percentage of mitotic cells–consistent with our prediction. Finally, we validated the refined signatures of paclitaxel response by mining a large drug screening dataset, showing that human cancer cell lines whose basal transcriptional profile is anti-correlated to them are significantly more sensitive to paclitaxel and docetaxel.

Right Ventricular Systolic Pressure Measurements in Combination with Harvest of Lung and Immune Tissue Samples in Mice

The function of the right heart is to pump blood through the lungs, thus linking right heart physiology and pulmonary vascular physiology. Inflammation is a common modifier of heart and lung function, by elaborating cellular infiltration, production of cytokines and growth factors, and by initiating remodeling processes. Compared to the left ventricle, the right ventricle is a low-pressure pump that operates in a relatively narrow zone of pressure changes. Increased pulmonary artery pressures are associated with increased pressure in the lung vascular bed and pulmonary hypertension. Pulmonary hypertension is often associated with inflammatory lung diseases, for example chronic obstructive pulmonary disease, or autoimmune diseases. Because pulmonary hypertension confers a bad prognosis for quality of life and life expectancy, much research is directed towards understanding the mechanisms that might be targets for pharmaceutical intervention. The main challenge for the development of effective management tools for pulmonary hypertension remains the complexity of the simultaneous understanding of molecular and cellular changes in the right heart, the lungs and the immune system. Here, we present a procedural workflow for the rapid and precise measurement of pressure changes in the right heart of mice and the simultaneous harvest of samples from heart, lungs and immune tissues. The method is based on the direct catheterization of the right ventricle via the jugular vein in close-chested mice, first developed in the late 1990s as surrogate measure of pressures in the pulmonary artery. The organized team-approach facilitates a very rapid right heart catheterization technique. This makes it possible to perform the measurements in mice that spontaneously breathe room air. The organization of the work-flow in distinct work-areas reduces time delay and opens the possibility to simultaneously perform physiology experiments and harvest immune, heart and lung tissues. The procedural workflow outlined here can be adapted for a wide variety of laboratory settings and study designs, from small, targeted experiments, to large drug screening assays. The simultaneous acquisition of cardiac physiology data that can be expanded to include echocardiography and harvest of heart, lung and immune tissues reduces the number of animals needed to obtain data that move the scientific knowledge basis forward. The procedural workflow presented here also provides an ideal basis for gaining knowledge of the networks that link immune, lung and heart function. The same principles outlined here can be adapted to study other or additional organs as needed.

Evaluation of PAF antagonists using human neutrophils in a microtiter plate assay

This paper describes a testing model for the detection and evaluation of PAF antagonists, based on the inhibition of PAF-elicited elastase release by human neutrophils. Incubations are performed in microtiter plates in the presence of a specific fluorogenic elastase substrate allowing direct measurement of the exocytosis response by means of a 96-well fluorescence reader. Determinations of the IC 50 values for five established PAF antagonists, Ro 19-3704, BN 52021, CV-3988, 48740 RP and kadsurenone, showed that the new model is comparable in sensitivity and discriminative capacity to other in vitro assays. From the effect of antagonists on the PAF concentration-response curve pA2 values could be calculated and information on the type of antagonism obtained. BN 52021 was found to behave as a competitive antagonist while Ro 19-3704 showed a more complex type of inhibition. As a one-plate system, the test is simple to handle and highly reproducible, and appears therefore particularly useful for large drug screening programs.