Comparative Evaluation of Drug Response Metrics for Predicting Cancer Sensitivity Using Transcriptomic Profiles

Comparative Analysis of Independent Ex Vivo functional Drug Screens Identifies Predictive Biomarkers of BCL-2 Inhibitor Response in AML

Identifying AML patients most likely to benefit from venetoclax using apoptotic gene ratios

SummaryLung cancer is an extremely heterogeneous disease, and its treatment remains one of the most challenging tasks in medicine. Few existing laboratory lung cancer models can faithfully recapitulate the diversity of the disease and predict therapy response. Here, we establish 12 patient-derived organoids from the most common lung cancer subtype, lung adenocarcinoma (LADC). Extensive gene and histopathology profiling show that the tumor organoids retain the histological architectures, genomic landscapes, and gene expression profiles of their parental tumors. Patient-derived lung cancer organoids are amenable for biomarker identification and high-throughput drug screening in vitro. This study should enable the generation of patient-derived lung cancer organoid lines, which can be used to further the understanding of lung cancer pathophysiology and to assess drug response in personalized medicine.

SummaryPhenotypic heterogeneity in cancer is often caused by different patterns of genetic alterations. Understanding such phenotype-genotype relationships is fundamental for the advance of personalized medicine. We develop a computational method, named NETPHIX (NETwork-to-PHenotype association with eXclusivity) to identify subnetworks of genes whose genetic alterations are associated with drug response or other continuous cancer phenotypes. Leveraging interaction information among genes and properties of cancer mutations such as mutual exclusivity, we formulate the problem as an integer linear program and solve it optimally to obtain a subnetwork of associated genes. Applied to a large-scale drug screening dataset, NETPHIX uncovered gene modules significantly associated with drug responses. Utilizing interaction information, NETPHIX modules are functionally coherent and can thus provide important insights into drug action. In addition, we show that modules identified by NETPHIX together with their association patterns can be leveraged to suggest drug combinations.

Genomic Subtypes of AML Define Sensitivity to NK Cell Cytotoxicity

High-Throughput Functional Drug Screening and Genomic Analysis to Guide Individualized Therapy for Relapsed/Refractory Multiple Myeloma

Introduction: In pancreatic ductal adenocarcinoma (PDAC), mutations in KRAS and deletion or promoter methylation of the CDKN2A gene - resulting in deregulation of the p16/CDK4/CDK6/RB axis - each occur in over 90% of cases. Therefore, combination therapy approaches targeting KRAS effectors and CDK4/6 may be a promising treatment strategy in a majority of PDAC patients. Lerociclib (G1T38) is an oral and selective CDK4/6 inhibitor in clinical development as a potential backbone therapy for multiple combination regimens in cancer. Here, we used a powerful and reliable ex vivo drug screening platform, the live tissue sensitivity assay (LTSA), to evaluate the ability of lerociclib to synergize with inhibitors of KRAS effectors across a panel of 24 well-characterized PDAC patient-derived xenografts (PDXs). Methods: PDX tumors were sectioned into uniform tissue slices at 200 µm thickness and arrayed in 96-well plates. For each PDX model, tumor tissue slices were treated with 0.3 µM, 1 µM, and 3 µM of lerociclib, alone and in combination with identical concentrations of a PI3K (pictilisib), mTOR (AZD2014), MEK (trametinib), ERK (ulixertinib), BRAF (vemurafenib) or EGFR (erlotinib) inhibitor for 72 hours. The viabilities of individual tissue slices were measured with PrestoBlue® reagent. A quantitative drug response analysis approach was implemented in which “responder” PDXs were defined as those exhibiting a statistically significant reduction in area under the dose-response curve (AUC) for the combination treatment compared to the most efficacious single agent. Statistical significance was defined as p < 0.05 using Student’s t-test. Results: Treatment of 24 PDAC PDXs using the LTSA revealed a range of sensitivities to each inhibitor as a single agent. When KRAS effectors and CDK4/6 were inhibited in tandem, statistically significant (p < 0.05) enhancements of efficacy were observed when lerociclib was added to pictilisib (21% of PDX models tested), ulixertinib (15%), vemurafenib (10%), erlotinib (5%), trametinib (4%), or AZD2014 (4%). These ex vivo data narrowed the focus of follow-up in vivo work; studies of PDX models in mice treated with lerociclib +/- pictilisib or ulixertinib are ongoing to validate these results. Moreover, differential gene expression and mutational analysis for responders versus non-responders identified through the PDAC PDX LTSA platform are underway, to elucidate molecular factors that predict tumor response or enrich for populations that are sensitive to CDK4/6 and PI3K/ERK inhibitors. Conclusion: The LTSA platform was used to evaluate drug combinations in viable, PDX tumor tissue. We found that lerociclib significantly enhances the response to PI3K or ERK inhibition in a substantial proportion of the tested PDAC PDXs as measured by AUC analysis in the LTSA platform, supporting future clinical evaluation of these combinations. Citation Format: Bingbing Dai, Daniel M. Freed, Jessica A. Sorrentino, Jithesh Jose Augustine, Tara G. Hughes, Ya'an Kang, Patrick J. Roberts, Jason B. Fleming, Michael P. Kim. CDK4/6 inhibition with lerociclib (G1T38) enhances response to PI3K or ERK inhibitors in high-throughput, ex vivo pancreatic PDX 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 4732.

Recently, a number of different approaches for improving cancer treatment have been developed. One scientific approach is to understand the biochemical pathways and coding genes involved in cancer causation, while other approaches look to distinguish the biochemical pathways involved in drug response. Many chemotherapeutic drugs have been developed targeting different cellular properties, and even more targeted therapies impacting specific oncogenic pathways are on the horizon. Pathways implicated in drug response have been studied to determine why some tumors might respond to a given chemotherapy while others do not. Understanding these pathways may provide potential for improving therapeutic efficacy. Pathway analysis methods have been proposed to discover pathways of interests using gene expression profiling. In this study, the biological pathways associated with response to Epirubicin (E), Cyclophosphamide (C) or the combination of the two drugs (combo EC) were evaluated in both breast cancer cell lines and patients. Thirty immortalized breast cancer cell lines were evaluated for response to these drugs using ChemoFx®. Area under the dose-response curve (AUC) was calculated to measure the chemosensitivity. We applied pathway enrichment to detect specific biochemical pathways associated with these drugs based on biological pathway databases (Kyoto Encyclopedia of Genes and Biocarta), breast cancer cell lines’ drug responses and public microarray data. Comparison of pathways associated with single drug E, C, and combo EC revealed that there are pathways associated with drug combination but not significantly associated with single drugs. These drug response-associated pathways offer great potential for new biomarker identification and drug targeting discovery. In addition, EC associated pathways in breast cancer patients were identified using public microarray data and the associated patient clinical outcomes. Of importance, four biological pathways related to adherens junction, leukocyte transendothelial migration, regulation of actin cytoskeleton and ribosomal protein were identified as EC response-associated pathways in both breast cancer cell lines and patients. These common EC response-associated pathways indicate the potential possibility to use cell line in vitro drug responses to predict patient chemotherapy outcomes. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 101st Annual Meeting of the American Association for Cancer Research; 2010 Apr 17-21; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2010;70(8 Suppl):Abstract nr 118.

Cancer cell spheroids present a biologically relevant model of avascular tumors and a unique tool for anti-cancer drug discovery. Despite being used in research laboratories for several decades, spheroids are not routinely used in the mainstream drug discovery pipeline primarily due to the difficulty of mass-producing uniformly sized spheroids and intense labor involved in handling, drug treatment, and analyzing of spheroids. We overcome this barrier using a novel technology to robotically microprint spheroids in standard 384-microwell plates. An aqueous drop containing cancer cells is dispensed into a bath of a second, immiscible aqueous phase to allow generation of a single spheroid. Using several cancer cells, we establish that this approach results in spheroids of well-defined size with less than 8% deviation from the mean diameter in 384-microwell plates. We demonstrate the feasibility of robotic, high throughput compound screening against tumor spheroids using of a collection of 25 standard chemotherapeutics and molecular inhibitors against three different cancer cells, HT-29 colon cancer cells, U-87 MG brain cancer cells, and MDA-MB-157 breast cancer cells. Dose-dependent experiments are done with each drug at a wide range of concentrations. After six days of incubation, viability of cancer cells in drug-treated spheroids is measured using a PrestoBlue assay that we have optimized for three-dimensional cell cultures. Post-treatment morphological changes are used as a secondary measure for analysis. This screening identifies specific MEK inhibitors including Trametinib and Selumetinib that potently inhibit growth of HT-29 spheroids at nanomolar concentrations, whereas several PI3K inhibitors such as pictilisib significantly block proliferation of U-87 MG spheroids at micromolar concentrations. Spheroids of MDA-MB-157 cells show resistance to most of the drugs. We identify few compounds such as a Survivin inhibitor, YM155, and doxorubicin that disintegrate spheroids and induce significant cell death in all three cells. To generate a scoring system, we use half-maximum inhibitory concentration (IC50), maximum inhibition (Emax), and area under the dose-response curve (AUC) to present a multi-parametric approach that takes into account both potency and efficacy parameters for evaluating drug responses in tumor spheroids. In conclusion, our robotic technology offers a low cost and convenient platform for high throughput screening of large collections of chemical compounds against tumor spheroids. Incorporation of this technology in drug discovery applications will make drug testing and screening with realistic tumor models a routine laboratory technique prior to expensive and tedious in vivo analyses, dramatically improving testing throughput and efficiency and reducing costs. Citation Format: Pradip Thakuri, Gary Luker, Hossein Tavana. High throughput robotic drug screening against microprinted tumor spheroids. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 616.

Spheroids present a biologically relevant three-dimensional model of avascular tumors and a unique tool for discovery of anticancer drugs. Despite being used in research laboratories for several decades, spheroids are not routinely used in the mainstream drug discovery pipeline primarily due to the difficulty of mass-producing uniformly sized spheroids and intense labor involved in handling, drug treatment, and analyzing spheroids. We overcome this barrier using a polymeric aqueous two-phase microtechnology to robotically microprint spheroids of well-defined size in standard 384-microwell plates. We use different cancer cells and show that resulting spheroids grow over time and display characteristic features of solid tumors. We demonstrate the feasibility of robotic, high-throughput screening of 25 standard chemotherapeutics and molecular inhibitors against tumor spheroids of three different cancer cell lines. This screening uses over 7000 spheroids to elicit high quality dose-dependent drug responses from spheroids. To quantitatively compare performance of different drugs, we employ a multiparametric scoring system using half-maximum inhibitory concentration (IC50), maximum inhibition (Emax), and area under the dose-response curve (AUC) to take into account both potency and efficacy parameters. This approach allows us to identify several compounds that effectively inhibit growth of spheroids and compromise cellular viability, and distinguish them from moderately effective and ineffective drugs. Using protein expression analysis, we demonstrate that spheroids generated with the aqueous two-phase microtechnology reliably resolve molecular targets of drug compounds. Incorporating this low-cost and convenient-to-use tumor spheroid technology in preclinical drug discovery will make compound screening with realistic tumor models a routine laboratory technique prior to expensive and tedious animal tests to dramatically improve testing throughput and efficiency and reduce costs of drug discovery.

In vitro chemosensitivity assays are invaluable for assessing chemotherapeutic agents' effects on cancer cells. Yet the dose-response curves generated by those assays, usually approximated by four-parameter logistic (4PL) models, are oftentimes difficult to interpret, with no clear indication of which metric should be used to compare them. Here, five commonly used metrics, absolute and relative half-maximal inhibitory concentration (IC(50)), area under the dose-response curve (AUC) based on trapezoidal rule and a parametric approach, and the effect at the maximal concentrations (E(max)), were compared in both simulations and real-life scenarios to evaluate their use with 4PL curves. Despite the fact that IC(50) is the most widely used metric to analyze dose-response curves, this study demonstrated that it was not the most reliable of the metrics tested. Fitted AUC showed the best overall performance in both the simulation and real-life scenarios; trapezoidal AUC showed similar performance to fitted AUC in most cases.

Introduction: In vitro experimentation has been suggested as a rapid tool for identifying and evaluating biomarkers associated with drug response. Excision repair cross-complementation 1 (ERCC1) enzyme plays a rate-limiting role in the nucleotide excision repair pathway and its up regulation has been associated with resistance to platinum agents. Recent studies have suggested that low expression levels of ERCC1 are related to a better survival benefit from cisplatin-based chemotherapy among patients with advanced non-small cell lung carcinoma (NSCLC). This study evaluates the relationship between ERCC1 expression and in vitro drug response to cisplatin using an in vitro chemoresponse assay, ChemoFx® Drug Response Marker (DRM), in human NSCLC cell lines and primary cultures of human lung cancers. Methods: The ChemoFx® DRM was performed on 13 immortalized human NSCLC cell lines (NCI-H520, HOP_92, HOP-62, A549, Calu-3, HCC827, OK, NCI-H460, NCI-H596, NCI-H1666, EKVX, NCI-H358, and NCI-H1975) and 30 primary cultures established from de-identified lung cancer surgical specimens. Cells were treated with a 10-dose range of cisplatin for 72 hours before DAPI-nuclear staining and counting. Response Index scores (RI scores), which are derived from an area under the dose-response curve (AUC), were calculated on the resulting dose-response curves. Protein expression of ERCC1 was evaluated with In-Cell Western analysis. Pearson correlation coefficient was used to show the association between ERCC1 expression and in vitro drug response. Result: Increased ERCC1 expression was significantly associated with cisplatin resistance in the 13 immortalized NSCLC cell lines (r=−0.72, p=0.004). A similar trend was observed across 30 primary cultures of human lung cancers, but the association did not reach the level of significance (r=−0.22, p=0.24). Conclusion: Clinical observation of a negative correlation between ERCC1 expression level and improved survival benefit from cisplatin-based chemotherapy was mirrored by in vitro ChemoFx® DRM analysis with strong correlation in immortalized human NSCLC cell lines and in primary cultures from human lung cancers with a correlation trend. This result suggests the value of in vitro chemoresponse assay in identification and evaluation of biomarkers and their association with chemotherapeutic response. The less significant correlation in human lung cancer primary cultures indicated that ERCC1 alone may not be sufficient for clinical prediction of response in human lung cancers across different pathological subtypes and stages, and additional biomarkers may be involved. These results suggest the value of in vitro chemo response assay, in helping to determine cisplatin response in human lung cancer. 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 4114. doi:10.1158/1538-7445.AM2011-4114

Spheroids present a biologically relevant model of avascular tumors and a unique tool for discovery of anti-cancer drugs. Despite being used in research laboratories for several decades, spheroids are not routinely used for drug discovery primarily due to the difficulty of mass-producing uniformly-sized spheroids and intense labor involved in handling, drug treatment, and analyzing them. We overcome this barrier using a novel technology to robotically microprint spheroids in standard 384-well plates. An aqueous drop containing cancer cells is dispensed into a bath of a second, immiscible aqueous phase. The drop maintains cells in close proximity to aggregate into a single spheroid. Using U-87 MG brain cancer cells, we show that this approach produces spheroids of well-defined size with ~10% deviation from their mean diameter. We demonstrate the feasibility of robotic, high throughput compound screening against tumor spheroids using a collection of 25 standard chemotherapeutics and molecular inhibitors against U-87 MG spheroids. Each drug is used in a wide range of concentrations. Viability of cancer cells in drug-treated spheroids is measured using a PrestoBlue assay. Morphological changes are used as a secondary measure for analysis of drug effect. We identify several compounds that effectively inhibit growth of spheroids. To generate a scoring system for effectiveness of drugs, we use half-maximum inhibitory concentration (IC50), maximum inhibition (Emax), and area under the dose-response curve (AUC) to present a multi-parametric approach that takes into account both potency and efficacy of drugs. Our robotic technology offers a low cost and convenient platform for screening large collections of chemical compounds against realistic tumor models prior to expensive and tedious in vivo tests, dramatically improving testing throughput and efficiency, and reducing costs.

Drug sensitivity and resistance are conventionally quantified by IC50 or Emax values, but these metrics suffer from a fundamental flaw when applied to growing cells: they are highly sensitive to the number of divisions that take place over the course of a response assay. Division rate varies with cell line, experimental conditions, and genetic alterations. The dependency of IC50 and Emax on division rate creates artefactual correlations between genotype and drug sensitivity while obscuring important biological insights and interfering with biomarker discovery. In this work, we derive alternative drug response metrics that are insensitive to number of divisions occurring during the assay. These are based on estimating growth rate inhibition (GR) in the presence of a drug using endpoint or time-course assays. The latter provides a direct measure of phenomena such as adaptive drug resistance. Using a simple model of drug response, we first show how GR50 and GRmax are superior to IC50 and Emax for assessing the effects of drugs in dividing cells. By expressing an oncogene in a transformed cell line, we illustrate how conventional metrics can lead to artefactual connections between mutations and drug sensitivity. We further validate the superiority of GR50 over IC50 values by reanalyzing a recently published large dataset of drug sensitivity and showing cases where difference in division rates is the only reason why IC50 values correlate with tissue type or genetic alterations. Using GR50 values prevents these artificial correlations and restores known connections between drug resistance and genomic markers. Finally, we show how GRmax values, which reflect efficacy, quantify differences in the phenotypic response and thus can be used to identify new biomarkers of sensitivity. Adopting GR metrics requires only modest changes in experimental protocols. GR values and metrics can be evaluated using scripts are available on github (www.github.com/sorgerlab/gr50_tools) or using an interactive website: www.grcalculator.org. We expect GR metrics to improve the use of drugs to identify response biomarkers, study mechanisms of cell signaling and growth, and identify drugs effective on specific patient-derived tumor cells.Drug sensitivity and resistance are conventionally quantified by IC50 or Emax values, but these metrics suffer from a fundamental flaw when applied to growing cells: they are highly sensitive to the number of divisions that take place over the course of a response assay. Division rate varies with cell line, experimental conditions, and genetic alterations. The dependency of IC50 and Emax on division rate creates artefactual correlations between genotype and drug sensitivity while obscuring important biological insights and interfering with biomarker discovery. In this work, we derive alternative drug response metrics that are insensitive to number of divisions occurring during the assay. These are based on estimating growth rate inhibition (GR) in the presence of a drug using endpoint or time-course assays. The latter provides a direct measure of phenomena such as adaptive drug resistance. Using a simple model of drug response, we first show how GR50 and GRmax are superior to IC50 and Emax for assessing the effects of drugs in dividing cells. By expressing an oncogene in a transformed cell line, we illustrate how conventional metrics can lead to artefactual connections between mutations and drug sensitivity. We further validate the superiority of GR50 over IC50 values by reanalyzing a recently published large dataset of drug sensitivity and showing cases where difference in division rates is the only reason why IC50 values correlate with tissue type or genetic alterations. Using GR50 values prevents these artificial correlations and restores known connections between drug resistance and genomic markers. Finally, we show how GRmax values, which reflect efficacy, quantify differences in the phenotypic response and thus can be used to identify new biomarkers of sensitivity. Adopting GR metrics requires only modest changes in experimental protocols. GR values and metrics can be evaluated using scripts are available on github (www.github.com/sorgerlab/gr50_tools) or using an interactive website: www.grcalculator.org. We expect GR metrics to improve the use of drugs to identify response biomarkers, study mechanisms of cell signaling and growth, and identify drugs effective on specific patient-derived tumor cells. Citation Format: Hafner M, Niepel M, Sorger PK. Metrics of drug sensitivity based on growth rate inhibition correct for the confounding effects of variable division rates [abstract]. In: Proceedings of the 2016 San Antonio Breast Cancer Symposium; 2016 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2017;77(4 Suppl):Abstract nr P6-07-33.

BackgroundOne of the challenges in personalized medicine is to determine specific drugs and their dosages for patient individuals who are undergoing a common disease. The technique of cell lines provides a safe approach to capture the drug responses of patient individuals when given specific drugs with varied dosages. However, it is still costly to determine drug responses in cells w.r.t dosages by biological assays. Computational methods provide a promising screening to infer possible drug responses in the cells of patient individuals on a large scale. Nevertheless, existing computational approaches are insufficient to interpret the underlying reason for drug responses.MethodsIn this work, we propose an interpretable model for analyzing and predicting drug responses across cell lines. The proposed model bridges drug features ( e.g., chemical structure fingerprints), cell features ( e.g., gene expression profiles), and drug responses across cells (measured by IC50) by a triple matrix factorization (TMF), such that the underlying reason for drug responses in specific cells is possibly interpreted.ResultsThe comparison with state‐of‐the‐art computational approaches demonstrates the superiority of our TMF. More importantly, a case study of drug responses in lung‐related cell lines shows its interpretable ability to find out highly occurring drug substructures, crucial mutated genes, as well as significant pairs between substructures and mutated genes in terms of drug sensitivity and resistance.ConclusionTMF is an effective and interpretable approach for predicting cell lines responses to drugs, and can dig out crucial pairs of chemical substructures and genes, which uncovers the underlying reason for drug responses in specific cells.