RESULTS AND DISCUSSION Dose response analysis In this study, we used HCA to examine the effects of 6 compounds that have reported effects on cardiomyocyte function, mitochondrial status, or viability. Based on published literature, some of the compounds were anticipated to have acute cytotoxic effects, while others have been implicated in cardiac pathologies associated with longer-term exposure. Doxorubicin, a chemotherapeutic agent, is known to induce acute cardiotoxicity, and has been issued a Black Box Warning for cardiovascular toxicity by the FDA. Like doxorubicin, amiodarone has an associated Black Box Warning from the FDA for cardiovascular toxicity. Inspection of representative images of cells treated with these compounds (Figure 3) and also antimycin A, confirms that there are distinct differences in the cellular phenotypes induced by these drugs.

Figure 3. Multiplexed cytotoxicty assay reveals phenotypic differences. Cardiomyocytes were treated with A) 0.5% DMSO only, B) amiodarone, C) doxorubicin or D) antimycin A then stained for mitochondrial status (TMRM, red), calcium mobilization (Fluo-4 AM, green), DNA/ nuclear status (Hoechst 33342, blue) and cell viability (TOTO-3 iodide, not shown). Cells were imaged live with IN Cell Analyzer 2000.

As the dose-response plots in Figure 4 show, doxorubicin and amiodarone toxicity were readily detected by the assay, which reported a dose-dependent decrease in mitochondrial shape (1/ form factor) and plasma membrane integrity (viability), concomitant with changes in nuclear area and intracellular calcium concentration (not shown). The toxic dose inducing half maximal plasma membrane integrity of the population (TD50) was 1.6µM and 4.5µM for doxorubicin and amiodarone respectively, on the same order of magnitude as values reported in the literature with other model systems (3). The effects of these compounds are observed for similar parameters, but the images and dose response profiles suggest different pharmacodynamics. Evidence suggests that acute doxorubicin cardiotoxicity involves cardiomyocyte apoptosis via a Bax-mediated pathway (4) whereas the effect of amiodarone is mediated by a Bax-independent process.

. Figure 4. Multi-parametric analysis reveals differential responses to test drugs. Spotfire DecisionSite Trellis plots of dose response curves for A) mitochondrial shape (1/ form factor) and B) plasma membrane integrity (% viability) for 6 compounds. Cardiomyocytes were exposed to increasing concentrations of compounds for 24 – 72h. Data were analyzed by non-linear regression (n=3 replicate wells).

Profiling tools As this example illustrates, cell imaging has the potential not only to identify compounds with potential cardiotoxic liabilities, but also to differentiate between different toxic compounds and group them based on similarities in their responses. While high content analysis of the 4-color toxicity assay captured a wealth of cell-by-cell data from the study, conventional dose-response plots do not take full advantage of all the information collected. Compound profiles that take into account a larger number of cell measurements from every treatment condition are likely to be more robust, as well as more sensitive in discriminating differential effects (5). The HCA image analysis protocol applied in this study extracted 54 different measurements from every cell in every image. Powerful pattern detection methods were then employed and profiles visualised in a variety of ways for more in-depth analysis of the whole data set. Data for each compound were analyzed by hierarchical cluster analysis, and visualized as a heatmap (Figure 5A), where each row comprises the normalized response for a particular treatment condition. The heatmap reveals that replicate samples and different doses of the same compound tend to cluster together and produce similar profiles. The combination of multiple parameters comprises a reproducible “signature” that reflects the various cell pathways and processes impacted by a particular drug.

Cath Hather*, Nick Thomas, Liz Roquemore and Alla Zaltsman GE Healthcare, Amersham Place, Little Chalfont, Buckinghamshire, England, UK HP7 9NA. Tel: +44 (0)29 2052 6499; Fax: +44 (0)29 2052 6230

e-mail: cath.hather@ge.com

Conclusions

• GE Healthcare Cardiomyocytes were employed in a live cell, 6 compound toxicity study and toxic compounds readily identified by HCA profiling

• High-throughput automated imaging with IN Cell Analyzer 2000 and high content analysis of the resulting images readily distinguished the more acutely cardiotoxic compounds from those with more chronic effects

• HCA is shown to be a powerful means of distinguishing and grouping toxic compounds based on multi-parametric phenotypic signatures

• The pattern recognition tools accessible through IN Cell Investigator software allowed identification of compounds with different mechanisms of action

• Automated profile classification could be used to generate libraries of “liability” profiles to help predict drug effects during new compound screening

References [1] Product sheet: Cardiomyocytes. Relevant. Reliable. Confident. A new era in safety screening with human cell models, GE Healthcare, 28-9801-12, Edition AB (2010). [2] Stem-Cell Derived Human Cardiomyocytes: utility for cardiac Safety Pharmacology and Detection of Complex Drug Effects. Bruening-Wright, A. et.al. (Poster presented at the 9th Annual World Pharmaceutical Congress, June 15-17, 2010, Philadelphia, Pennsylvania [3] Filigheddu, N. et al. Hexarelin protects H9c2 cardiomyocytes from doxorubicin-induced cell death. Endocrine (2001) 14:113-119. [4] An, J. et al., ARC is a critical cardiomyocyte survival switch in doxorubicin cardiotoxicity. J. Mol. Med. 87, 401–410 (2009). [5] Freeley, M. et. al. A high-content analysis toolbox permits dissection of diverse signaling pathways for T-lymphocyte polarization. J. Biomol. Screening (2010) 15:541-555.

Figure 5. Pattern detection using data clustering and profiling tools. A) Hierarchical cluster analysis of data from the top 4 doses for 6 compounds based on 13 cellular parameters derived from the fluorescent reporters Fluo-4 AM (calcium mobilization), Hoechst 33342 (nuclear phenotype), TMRM (mitochondrial status) and TOTO-3 iodide (membrane integrity). B) Profile plot based on nuclear and mitochondrial parameters at 0 µM (green) and 33 µM (red) are compared for amiodarone (top), doxorubicin (middle) and nifedipine (bottom). All data shown are n=3 replicate wells per treatment condition.

In the parallel axis profile plot (Figure 5B), each vertical line in a given profile represents an axis upon which is plotted the relative response for a single parameter. Superimposing the profiles generated at concentrations of 0µM and 33µM for several compounds on the same profile plot makes it easier to visualize differences in phenotypic profiles. The phenotypic changes induced by amiodarone become distinguishable from those of doxorubicin at the 33µM dose. By contrast, cell phenotype is not affected by nifedipine treatment. It is readily apparent that 3 of the test compounds—doxorubicin, amiodarone and antimycin A—elicit markedly different cellular phenotypes, particularly at the higher doses. However, diclofenac, zidovudine and nifedipine have phenotypic profiles more similar to the DMSO controls. While all of these compounds have been associated in some way with adverse cardiovascular effects, literature suggests that their damaging effects are likely to be cumulative and less acutely cytotoxic than those of doxorubicin, amiodarone and antimycin A, which trigger cell death pathways. Automated profiling The profiling tools used so far have enabled the key parameters important in distinguishing compound toxicity to be identified. These parameters were then used in automated profiling of the entire set of compounds at all concentrations for phenotypic classification (Figure 6). The phenotypic changes induced by amiodarone, doxorubicin and antimycin A become distinguishable at certain doses and are highlighted in red. Cell phenotype is not affected by the remaining compounds over the dose-range tested.

Figure 6. Automated profile classification. Phenotypic profiles for each compound at each dose were generated based on a subset of key parameters (n=3 replicate wells per treatment condition).

ABSTRACT During the drug development process, late-stage detection of cardiotoxic side effects can significantly increase program costs and time to market. Failure to detect cardiotoxicity prior to launch can present a serious health risk for patients. Following costly withdrawals of a number of drugs from the market due to unexpected adverse cardiovascular effects, there has been an increased demand for more relevant and readily available cell models for in vitro cardiotoxicity testing. To address this need, GE Healthcare provides differentiated cardiomyocytes derived from human embryonic stem cells. To explore the utility of GE Healthcare Cardiomyocytes in image-based assays for toxicity screening, we challenged the cells with a panel of test compounds, including those with known or suspected cardiotoxic liabilities. High-content analysis (HCA) of the results allowed us to identify cardiotoxic compounds and distinguish them from each other based on their phenotypic profiles.

INTRODUCTION GE Healthcare Cardiomyocytes are an industrial-scale source of physiologically relevant cells for use during drug development and toxicity testing. Differentiated from karyotypically normal human embryonic stem cells, the cardiomyocytes have been characterized by electrophysiology, flow cytometry and sub-cellular imaging (1). Upon recovery from cryo-preserved stocks, GE Healthcare Cardiomyocytes form contractile monolayers of cells expressing cardiac-related transcription factors, structural proteins (Figure 1) and functional ion channels (2).

A B Figure 1. GE Healthcare Cardiomyocytes express cardiac markers. Indirect immunofluorescence staining for cardiac Troponin I (A) and α-actinin (B). In both images, nuclei are counterstained with Hoechst 33342 (blue).

Cardiotoxins can impact cell behavior and phenotype in a variety of ways, including interfering with mitochondrial respiration, compromising plasma membrane integrity, and triggering cell death pathways. By simultaneously monitoring multiple toxicity indicators on a cell-by-cell basis, HCA can provide a sensitive and robust means of capturing subtle changes in cell phenotype that may be indicative of toxicity, but would not easily be detected by electrophysiological assays or more conventional in vitro tests.

In this study, the effects of test compounds on cardiomyocytes were examined using a multiplexed assay that employs a cocktail of four fluorescent probes to monitor plasma membrane integrity, calcium mobilization, nuclear phenotype and mitochondrial status. By rapidly collecting over 50 measurements from every cell across a range of doses and replicate sample wells, we have generated multi-parametric phenotypic profiles for the test compounds, and demonstrate that HCA analysis using GE Healthcare Cardiomyocytes can be highly informative in detecting and also differentiating cardiotoxic compounds.

METHODS The live assay, image acquisition and analysis were performed as described in the workflow (Figure 2). Cardiomyocytes were seeded into Matrigel-coated plates following the methods for plate coating and cell preparation as described in the Cardiomyocytes User Guide (28980586AA). After compound exposure, images of live cardiomyocytes were acquired with an IN Cell Analyzer 2000 high-content analysis system using the following excitation (x) and emission (m) filter combinations: 350/50x, 455/50m for Hoechst 33342; 490/20x, 525/20m for Fluo-4 AM; 579/34x, 624/40m for TMRM; and 645/30x, 705/72m for TOTO-3. Automated image analysis was performed with IN Cell Investigator software. For hierarchical data clustering, profile plots and multi-dimensional scatter plots, we used Spotfire DecisionSite® tools (Tibco Software Inc., Palo Alto, CA), accessed through links integrated into the IN Cell Investigator software interface.

Figure 2. Live multiplexed assay workflow. Description of workflow for assay of live GE Healthcare Cardiomyocytes with 6 compounds and subsequent image acquisition, analysis and data visualization.

D

H

P

High content cell profiling as a tool for early drug safety testing using GE Healthcare Cardiomyocytes and IN Cell Analyzer 2000

GE, imagination at work, and GE Monogram are trademarks of General Electric Company. All third party trademarks are the property of their respective owners. The IN Cell Analyzer 2000 and associated analysis modules are sold under use licenses from Cellomics Inc. under US patent numbers US 5989835, 6416959, 6573039, 6620591, 6671624, 6716588, 6727071, 6759206, 6875578, 6902883, 6917884, 6970789, 6986993, 7060445, 7085765, 7117098; Canadian patent numbers CA 2282658, 2328194, 2362117, 2381334; Australian patent number AU 730100; European patent numbers EP 0983498, 1095277, 1155304, 1203214, 1348124, 1368689; Japanese patent numbers JP 3466568, 3576491, 3683591 and equivalent patents and patent applications in other countries. GE Healthcare Cardiomyocytes are sold under licence from Geron Corporation and Wisconsin Alumni Research Foundation under US patent and publication numbers: US 7,425,448, US 2009/0017465, US 6,800,480, US 5,843,780, US 6,200,806, US 7,029,913, US 7,582,479, US 7,413,902; US 7,297,539, US 2009/0047739 and US 2007/0010012 and equivalent patent and patent applications in other countries. © 2011 General Electric Company – All rights reserved. All goods and services are sold subject to terms and conditions of sale of the GE Healthcare Company which supplies them. A copy of these terms and conditions are available on request. Contact your GE Healthcare representative for the most current information and a copy of the terms and conditions. GE Healthcare Bio-Sciences AB Bjorkgatan 30 SE-751 84 Uppsala Sweden. This poster was presented at the Cambridge Healthtech Institute’s 8th Annual High Content Analysis conference, 11th – 14th January 2011, San Francisco, California. *To whom all correspondence should be addressed.