- Research article
- Open Access
- Open Peer Review
This article has Open Peer Review reports available.
Assessing the translatability of In vivo cardiotoxicity mechanisms to In vitro models using causal reasoning
© Enayetallah et al.; licensee BioMed Central Ltd. 2013
Received: 18 April 2013
Accepted: 3 September 2013
Published: 6 September 2013
Drug-induced cardiac toxicity has been implicated in 31% of drug withdrawals in the USA. The fact that the risk for cardiac-related adverse events goes undetected in preclinical studies for so many drugs underscores the need for better, more predictive in vitro safety screens to be deployed early in the drug discovery process. Unfortunately, many questions remain about the ability to accurately translate findings from simple cellular systems to the mechanisms that drive toxicity in the complex in vivo environment. In this study, we analyzed translatability of cardiotoxic effects for a diverse set of drugs from rodents to two different cell systems (rat heart tissue-derived cells (H9C2) and primary rat cardiomyocytes (RCM)) based on their transcriptional response. To unravel the altered pathway, we applied a novel computational systems biology approach, the Causal Reasoning Engine (CRE), to infer upstream molecular events causing the observed gene expression changes. By cross-referencing the cardiotoxicity annotations with the pathway analysis, we found evidence of mechanistic convergence towards common molecular mechanisms regardless of the cardiotoxic phenotype. We also experimentally verified two specific molecular hypotheses that translated well from in vivo to in vitro (Kruppel-like factor 4, KLF4 and Transforming growth factor beta 1, TGFB1) supporting the validity of the predictions of the computational pathway analysis. In conclusion, this work demonstrates the use of a novel systems biology approach to predict mechanisms of toxicity such as KLF4 and TGFB1 that translate from in vivo to in vitro. We also show that more complex in vitro models such as primary rat cardiomyocytes may not offer any advantage over simpler models such as immortalized H9C2 cells in terms of translatability to in vivo effects if we consider the right endpoints for the model. Further assessment and validation of the generated molecular hypotheses would greatly enhance our ability to design predictive in vitro cardiotoxicity assays.
In 2007, the leading cause for drug withdrawal from the market was attributed to cardiotoxicity (31%) . The voluntary withdrawal of the COX-2 selective inhibitor Rofecoxib in 2004 due to increased risk of myocardial infarction and stroke is one of the more prominent examples . Addressing the safety issues early would significantly reduce such costly surprises in the drug discovery process and would also improve the survival of pharmaceutical drugs to the market. Although using animal models to predict late stage safety issues has been the norm in the industry for years, there is increased expectation that progress in utilization of computational toxicology predictive models, specialized in vitro models and a combination of both these models will enhance early de-risking, reduce animal use and enhance compound survival. In addition, the US National Academy of Sciences recently released a toxicity testing framework emphasizing the utilization of high throughput in vitro toxicity assays and computational models to assess the risk and underlying mechanism of toxicities triggered by pharmaceutical chemicals and environmental contaminants. This is envisioned to include model systems based on stem cell biology, functional genomics and physiologically based pharmacokinetic (PBPK) modeling .
There have been several reports wherein computational models have been utilized for predicting the early safety risks based on potassium voltage-gated channel, subfamily H (HERG) binding [4, 5], Absorption, Distribution, Metabolism, Excretion and Toxicity (ADMET) properties , Adenosine tri-phosphate Binding Cassette (ABC) transporter substrates  and Cytochrome P450 (CYP450) inductions . However, the successful utilization of mechanism-based screening assays has been a challenge despite the plethora of published studies on the known mechanisms of drug-induced cardiac toxicity. These include well studied mechanisms of cardiotoxicity such as oxidative stress, calcium dysregulation, energy metabolism disruption, cell cycle/proliferation and tissue remodeling [9–11].
It is believed that a major factor contributing to the limited success of predicting clinical outcome using preclinical models or predicting in vivo outcome using in vitro models is due to limited understanding of the translatability across model systems and species. Hence, the recent increase of models believed to better reflect the physiological and functional roles of cardiomyocytes such as progenitor cardiomyocytes, human embryonic stem cells (ESC) and inducible pluripotent stem cell (iPS) derived cardiomyocytes [12, 13]. Recently, Force and Kolaja reviewed the most commonly used models of cardiomyocytes summarizing their advantages and disadvantages . It should be noted, of course, that this methodology will only reveal mechanisms that result from direct action of a compound on a cardiomyocyte. This in vitro system is inadequate for predicting secondary effects mediated by the interaction of multiple complex organ systems, such a rise in heart rate due to increased epinephrine release.
The primary goal of this study is to evaluate the translatability of cardiotoxicity mechanisms from in vitro to in vivo and to compare the elicited mechanisms in different in vitro models. To achieve this we utilized gene expression microarray experiments from rat toxicity studies (Drugmatrix, Iconix [14, 15]) and in vitro experiments in H9C2 (embryonic BD1X rat heart tissue derived cells) and neonatal rat ventricular cardiomyocytes (RCM) using nine known pharmaceutical compounds known to induce cardiotoxicity in vivo.
The gene expression microarray data was analyzed using a novel computational tool called the Causal Reasoning Engine (CRE) [16, 17]. CRE interrogates prior biological knowledge to generate testable hypotheses about the molecular upstream causes of the observed gene expression changes. Each such hypothesis summarizes (“explains”) a certain number of gene expression changes (see KLF4+ example below). Notably, hypotheses usually make statements about predicted protein abundance or activity changes, e.g. increased or decreased TGFB1 activity. In our experience, CRE hypotheses tend to robustly identify biological phenomena driving gene expression changes and provide several advantages over other gene expression analysis methods . In particular, for the purpose of this study, CRE provided the advantage of better abstracting biological information from gene expression data obtained across different experimental settings (see Causal Reasoning Convergence below).
Following the CRE analysis of all individual compound treatments in vitro and in vivo, we compared the hypotheses and the biological processes they compose to assess the translatability of mechanisms from one model system to the other. Subsequently, we experimentally tested KLF4 and TGFB1 activities, two of the central molecular hypotheses predicted by CRE, in response to the cardiotoxic compounds used in the CRE analysis using qPCR and reporter assay. Finally, we discuss the implications of our analysis and suggest potential future experiments.
H9C2 cells (derived from embryonic BD1X rat heart tissue) were purchased from ATCC. H9C2 cells were grown DMEM (Gibco# 11965) with 10% FBS as per manufacturer’s protocol. Neonatal, ventricular Clonetics® Rat Cardiac Myocytes (P1-3) (RCM)(Catalog # R-CM-561) were purchased from Lonza and were grown in RCGM media with supplements as per manufacturer’s protocol.
For ATP depletion assays, H9C2 and RCM’s cells were plated in 96 well plates per the manufacturer’s protocol for 24 hr prior to treatments. For gene expression experiments, H9C2 and RCM cells were plated in 24 well plates per the manufacturer’s protocol for 24 hr prior to adding of treatments.
In vitro cytotoxicity phenotype (ATP depletion) and known in vivo cardiac safety liabilities of the test compounds
ATP depletion IC50 at 48 hrs (μM ± SE)
In vivo 5-day treatment (mg/kg)
Reported ECG abnormalities & arrhythmia
Reported structural cardiotoxicity
Primary pharmacology & indication
12 ± 1.17
5.7 ± 0.67
2.71 ± 0.92
22.9 ± 0.83
β1 agonist, inotropic
4.23 ± 0.52
39.5 ± 2.11
1.16 ± 0.47
ATP depletion assays
ATP depletion measurements were done using The CellTiter-Glo® Luminescent Cell Viability Assay from Promega (Catalog # G7570) per the manufacturer’s protocol. 100 μl per well of reconstituted ATP depletion reagent was added directly to 96 well plate and incubated for 10 minutes on orbital shaker. Luminescence signal was measured using Envison plate reader.
Microarray gene expression data
RNA was extracted 24 hrs after compound treatment using Qiagen’s RNeasy Mini kit (Catalog # 74104) per the manufacturer’s protocol. Quality and quantity of RNA was assessed using Nanodrop 2000c (A260/280 ratio) from Thermo Fisher Scientific and Agilent RNA analyzer (RIN scores). RNA (n = 2) was submitted to Genelogic for Affymetrix Genechip profiling using Rat Expression Array 230 2.0 chip. The in vivo rat cardiac tissue gene expression comparisons in response to the same compounds (Table 1) used in the in vitro experiments were obtained from the Drugmatrix toxicogenomic database [14, 15]. The gene expression data for the effect of Isoprenaline on mouse cardiac tissue was obtained from the public domain (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE18801), from a study published by Galindo et al. .
For quality control, RNA degradation plots were generated for each CEL file. To assess potential RNA degradation, 3′/5′ ratios and their associated confidence intervals were evaluated . Two techniques were used to distill the probe results into a small number of representative variables; Multidimensional scaling (MDS)  and Principal component analysis (PCA). These two techniques were applied to the data before and after Robust Multi-Array Average (RMA)  signal processing. During this processing, only the perfect match (PM) probe data were used; the mismatch (MM) probes were not used. To assess differential expression of genes between groups of interest, a common statistical model was applied independently to each probeset. Gene expression for all sample types was analyzed on the log2 scale. Linear models were used to calculate t-statistics, which were subsequently adjusted using the moderated t-statistic procedure . The Benjamini and Hochberg adjustment procedure  based on controlling the False Discovery Rate (FDR) was used.
Causal reasoning engine algorithm
Gene expression changes are analyzed to detect potential upstream regulators as previously described [16, 17]. Briefly, the approach relies on a large collection of curated biological statements in the form:
A [increases or decreases] B, where A and B are measurable biological entities.
The biological entities can be of different types (e.g. phosphorylated proteins, transcript levels, biological process and compound exposure) and each statement is tied to accessible, peer-reviewed articles. For this work, we licensed approximately 450,000 causal statements from commercial sources (Ingenuity Systems and Selventa).
Network modeling of the CRE hypotheses
The analysis results are visualized using the Causal Reasoning Browser, a Java application based on the open-source biological network viewer Cytoscape  as previously described . Briefly, in the CRE browser an overview graph allows users to visualize hypotheses and examine their network relationships in the context of the causal relationships obtained from the literature based knowledgebase. To facilitate the construction of biological networks from the generated hypotheses, several analytical tools were developed e.g. a clustering tool uses cosine similarity metric and an average linkage method to group related hypotheses together .
HEK293 TGFβ reporter assay methods
HEK-293 cell line was obtained from American Type Culture Collection (ATCC; Manassas, VA). HEK-293 cells were grown in Eagles Minimum Essential Medium (ATCC) containing 10% fetal bovine serum and 1% penicillin-streptomycin. Cells were maintained at 37°C, 5% CO2, 95% humidity.
TGFβ (SMAD2/SMAD3/SMAD4) Cignal lentiviral construct and transducing reagents were purchased from SABiosciences (Frederick, MD). Cells were plated in 12-well plates at 2.5×105 cells per well. Transductions were performed according to manufacturer’s directions, using 20 μL of lentiviral particles and 8 μM concentration of Sureentry (SABiosciences) transfection reagent. Stable cell lines were selected using 1 μg/mL puromycin (Sigma, St. Louis, MO). Single cells were isolated from Polyclonal cell lines using a FACS Vantage Cell Sorter (BD, Franklin Lakes, NJ), and expanded.
Transduced cells were plated in 384-well plates at 2000 cells/well. After overnight incubation, cells were induced using 25 ng/ml hTGFβ1 protein (Sigma # T7039) for 1 hour. Cells were then dosed with varying concentrations of test compound at a final 1% DMSO concentration and incubated for 24 hours in a 37° incubator with 5% CO2. Luciferase activity was determined using Steady-Glo Luciferase Assay Reagent to cells. (Promega, Madison, WI). Luminescence was measured on an EnVision 2103 Multilabel Reader (Perkin-Elmer, Waltham, MA). To evaluate inhibitory effects of the test compounds on the TGFB1 reporter, it was necessary to first stimulate TGFB1 expression. The in vitro reporter cell lines express low basal levels of TGFB1 by design for the original purpose of agonist evaluation. In addition, the Envision plate reader used for detection of the reporter assay luciferase readout is unable detect values lower that zero. Induction of TGFB1 expression with a stimulant allowed us to induce TGFB1 luciferase readout such that we were able run the assay in antagonist mode. This differs from in vivo TGFB1 expression levels, which allow for evaluation of a decrease or increase in expression.
Quantitative real time polymerase chain reaction assays were performed in triplicates (n = 3 per treatment group) in rat heart tissue derived immortalized H9C2 cells treated with cardiotoxic and reference compounds using a 384 well format on the ABI 7900HT. Relative quantification values (ΔΔCt) for Klf4 (p/n 4331182) message were calculated using the ABI SDS 2.3 software comparing compound treatment to DMSO vehicles after normalization to β-actin (p/n 4352340E.) The ABI 2X Master Mix (p/n 4370074) was used with standard cycling protocols.
Causal reasoning convergence
One of the proposed advantages in this study is the ability of the causal reasoning approach to abstract similar molecular events from microarray experiments from different sources, models and chips, thus overcoming technical and biological variability that otherwise make the comparison at the gene level challenging. Therefore, we investigated the convergence capability of CRE in detecting expected similar biological events from data generated in different species, gene-chips and different experimental settings (Iconix Drugmatrix database and GEO public data, see Methods). Isoprenaline is a widely studied prototypic compound for hypertrophic cardiomyopathy with documented molecular mechanisms  and its effect in rats and mice is compared here. Indeed, comparison of two independently generated gene expression datasets, for Isoprenaline treated mouse heart tissue and from rat heart tissue, reveals very similar causal reasoning biological networks (Figure 1).
The major molecular events (Figure 1) were constructed by selecting the highest ranking hypotheses and their closest significant neighbors followed by elimination of redundant and surrogate hypotheses as previously described . The molecular networks from both rats and mice largely support similar biological events such as increased hypoxia/ischemia, angiotensin signaling, oxidative stress and inflammation, all of which are known mechanisms of cardiac stress response [25–29].
Cardiac liabilities and cytotoxicity of test compounds
We selected a set of test compounds with reported ECG-type abnormalities and/or structural cardiac toxicities and of diverse pharmacology (Table 1). The ATP depletion IC50 concentration at 48 hours in H9C2 cell line was used to determine the microarray experimental concentrations. However, we harvested the cells at 24 hours for RNA extraction and microarray analysis with the rationale of investigating earlier molecular events preceding cell death. All compounds exhibited IC50 in the low micromolar range with the exception of Dexamethasone and Terbutaline.
Examples of in vivo to in vitro causal networks
Outlined in Figure 2 are the major signaling networks differentiating the Amiodarone effect on rat heart (Figure 2A) and primary rat cardiomyocytes (Figure 2B). In vivo, we found a number of hypotheses related to Amiodarone’s suggested mechanisms of action through cellular Ca++ and potassium modulation [30, 31], and reported side effects such as binding to thyroid antagonism and hypothyroidism [32, 33]. None of the mechanism related hypotheses were found in vitro. Moreover, all major causal reasoning supported biological networks were significantly different. Inflammation is one of the major signaling networks predicted, albeit with opposite directionality being predicted decreased in vivo and predicted increased in vitro. Suggested downstream effects varied significantly as well, decreased cell cycle in vivo versus apoptosis in vitro and a larger tissue remodeling/structural signal primarily driven by decreased TGFB in vitro. At the hypothesis level very few similarities were found between in vivo cardiac tissue and in vitro primary rat cardiomyoctes, e.g. Hypoxia- and SRF- hypotheses.
Contrary to Amiodarone, Dexamethasone shows high degree of in vivo to in vitro translatability at both the process and individual hypothesis levels. Figure 3 shows the causal reasoning inferred molecular response to Dexamethasone in rat cardiac tissue (Figure 3A) and Primary rat cardiomyocytes (Figure 3B). Causal reasoning generated a number of individual hypotheses reflective of dexamethasone action such as Dexamethasone+, NR3C1+ and glucocorticoid+. Known dexamethasone effect is also reflected by supported biological processes such as the anti-inflammatory sub-network both in vivo and in vitro. Dexamethasone is also highly translatable to H9C2 cells as well with a causal network that is highly similar to that of primary rat cardiomyocytes (not shown).
In vivo to in vitro translatability of the major biological processes
Identifying KLF4 as a potential common hub in cardiotoxicity
KLF4 mRNA fold changes in H9C2 from the gene microarrays and subsequent confirmation using qRT-PCR
Potential role of TGFB1 in cardiotoxicity and TGFB1 reporter assay
Predicted TGFB signaling in vivo and in cell lines
Predicted TGFB signaling
Gene expression changes of nine compounds known to induce cardiotoxicity were profiled in rat cardiomyocytes, rat embryonic heart tissue-derived H9C2 cells, and heart tissue from treated rats. There was, as expected, significant variation between drugs and test systems at the individual gene level. In this work we applied a recently developed method [16, 17] to understand convergence of gene expression changes based on their potential upstream regulators. As described the CRE analysis revealed a convergence of the explained changes around a set of biological pathways. Specifically, pathways associated with tissue remodeling, cell cycle, oxidative stress, and DNA damage were particularly well conserved across cardiotoxic drugs and between in vivo and in vitro test systems. This level of concordance between the in vivo and in vitro systems was encouraging but there were some clear points of disagreement between the experimental systems providing a stark reminder of the limitations of in vitro systems. An example of this difference is the greater diversity of signaling in H9C2 cells compared to rat cardiomyocytes. This may be explained by the immortalized nature of H9C2 cells with active cell cycle compared to the primary rat cardiomyocytes. Another possibility is that H9C2 cells are less similar to cardiomyocytes thus more likely to exhibit non-cardiomyocyte phenotype. Although, the wholesale differences between the Amiodarone in vitro and in vivo transcriptional changes highlights that the overall predictivity of cellular systems can vary from compound to compound depending on specific expression of drug targets, the opposing TGFB signals observed across the majority of tested drugs points to a more fundamental inability of the in vitro systems to replicate in vivo signaling networks. By better understanding these limitations though, we might still be able to address those instances of successful translations of pathway-level signals of toxicity between in vivo and in vitro systems to quickly and efficiently triage potential therapeutics for their potential to induce adverse events.
The CRE method provided interesting insights in this case and summarized the observed expression changes efficiently for further analysis. However, it is important to note its potential shortcomings. The approach is only as powerful as its underlying knowledgebase of prior biological knowledge. Even a knowledgebase that encompasses all currently known biomedical relationships would not be able to summarize changes that have never been observed before. In our experience  the approach usually provides helpful insights as many molecular regulatory processes have been well researched over time. Given a comprehensive knowledgebase results often turn up combinations of upstream regulators that have been observed in a different biological context previously but are novel for the biological problem under study.
Almost as important as the overlap between the in vivo and in vitro outcomes of drug treatment is the notion that the critical biological processes that seem to underlie the drug toxicity can be visualized across various cell types. Much work has been devoted to trying to build an in vitro system that accurately replicates intact organ systems in a dish. These technologies have tended to be expensive, laborious, and low-throughput, thus limiting their utility in any type of routine predictive screening strategy. In addition, these complex culture systems still fail to fully recapitulate the in vivo organ system they seek to model, particularly for long term dosing studies. What this work suggests though is that these types of convoluted cell models might not be necessary for understanding the safety risk of a segment of compounds. When the underlying mechanism of the toxicity is a basic pathway associated with cell health and viability, the specific cell system is of minimal importance. Moving from a primary cardiomyocyte, which recapitulates many important activities of an in vivo cardiac cell; to an immortalized rat heart tissue derived cell line such as H9C2 did not result in the loss of translational power. Likewise, the primary cardiomyocytes were just as likely to show discordance from the in vivo as the immortalized cell line was.
The traditional thinking has been that the reason for the organ specificity of drug toxicity was due to unique innate traits of the particular organ being affected. This thinking has largely driven a desire to have more organ-like in vitro culture systems. The notion that very generic, non-organ specific mechanisms of toxicity might explain a large portion of organ-specific toxicity runs counter to this thinking and leads to questions of why compounds with these types of liabilities do not show gross, multi-organ toxicities in vivo. It has long been appreciated that differences in distribution and accumulation of medications directly affect their efficacy . The same can be said about toxicity. Cardiotoxicity is not entirely due to the unique “cardiac-ness” of the cells but due to the fact that the heart is the organ that sees the greatest concentration of the compound as a result of a combination of intrinsic and extrinsic expression of transporters and clearance mechanisms. Therefore, in an in vitro system, where one can ensure exposure of the compound to the cell, reproducing an intact organ system is not necessary for visualizing the toxicity risk.
This is not to say that all types of toxicity can be modeled in a generic cell line. There are several types of specific drug-induced toxicities were specific functionalities must be present in a cell system in order to visualize that toxicity. For example, induced pluripotent stem cell derived cardiomyocytes have been extensively characterized (including comparative gene expression profiling) and evaluated to study cardiac specific end points (such as Beating and Contractility) [35, 36]. Utilization of these types of advanced test systems that take advantage of ‘cardiac-ness’ of these cells might be helpful for certain evaluations. This may be the case for Amiodarone in this study. For instance, drug-induced arrhythmias could be attributed to a very unique feature of cardiomyocytes. Ideally, an in vitro system that predicts this outcome would incorporate a cell that beats so that any alteration in pace or occurrence of rhythmic cell contraction could be directly measured. But even with this example, distilling this very organ-specific toxicity down to the basic molecular mechanism that drives it enables a simple, cell-neutral assay for predicting it, hERG binding and dofetilide competition. As we gain a better appreciation of the mechanisms of toxicity, there will be a reduction in the need for costly primary cell cultures in predictive toxicology.
The mechanisms of toxicity uncovered in this work are not entirely novel. Disregulating cell cycle, inducing DNA damage, and producing oxidative stress has long been appreciated as having a negative effect on cellular health, often leading to obvious cytotoxicity. It is not surprising then that a basic cytotoxicity assay has been shown to have high predictive power for in vivo toxicity regardless of the organ-specific nature of that toxicity [37, 38]. This similarity in toxicity across cell lines of different tissue origins can also be seen in our data. Both the primary cardiomyocytes and immortalized skeletal muscle cells showed a clear down regulation of TGFB signaling upon application of cardiotoxicants. We were able to reproduce this data utilizing a reporter system cloned in cell line derived from kidney. Although this response was in opposition to what was observed in vivo, upon moving to the in vitro system, there was a complete conservation of signaling at the pathway level regardless of the tissue type the cell line was meant to model. Similarly, both primary cardiomyoctes and immortalized H9C2 cells showed predicted increase in KLF4 which we were able to reproduce by measuring KLF4 expression levels using RT-PCR. KLF4 is a hub that mediates the effect of different cell stress signals such as oxidative stress and DNA damage on critical cell functions such as cell proliferation and differentiation . In particular, KLF4 is known to play a role in cardiac function. For example, KLF4 has been shown to mediate cardiac myofibroblast differentiation in response to Angiotensin II stimulation partly through regulating TGFB1 . KLF4 has been also shown to be involved in regulating the cardiac hypertrophic response .
The finding concerning TGFB signaling has implications beyond this work. In recognition of the need for more and better in vitro tools for toxicity prediction, many different reporter assays and screening systems have been built and are being marketed for this purpose. The choice of signaling pathways and cellular endpoints used for these products are, for the most part, based not on detailed validation of the tools for their designed purpose. Instead the significance of these endpoints is taken exclusively from literature without fully understanding the impact of moving them to an in vitro detection system. The link between aberrant TGFB signaling and potential adverse events is well established [42–44]. Using a reporter system to measure the potential of a compound to induce that signaling network in vivo is clearly not that straightforward though, based on the finding of this work. Until the translatability of tools like the TGFB reporter system can be validated, caution must be taken in utilizing it and tools like it for predictive screening.
There is a desperate need in modern drug discovery for high-throughput, cost effective assay technologies that are highly predictive of in vivo toxicity. One of the primary concerns in adapting these assays for triaging newly developed compounds is the ability to translate an in vitro signal to an in vivo outcome. This work adds to the growing literature that strongly suggests that an in vivo/in vitro connection can be drawn through the use of basic cellular mechanisms but there are limitations to these predictions that are independent of the relationship between the cell type and the target tissue.
- Wilke RA, Lin DW, Roden DM, Watkins PB, Flockhart D, Zineh I, Giacomini KM, Krauss RM: Identifying genetic risk factors for serious adverse drug reactions: current progress and challenges. Nature reviews Drug discovery. 2007, 6 (11): 904-916. 10.1038/nrd2423.View ArticlePubMedPubMed CentralGoogle Scholar
- Topol EJ: Failing the public health–rofecoxib, Merck, and the FDA. N Engl J Med. 2004, 351 (17): 1707-1709. 10.1056/NEJMp048286.View ArticlePubMedGoogle Scholar
- Andersen ME, Krewski D: Toxicity testing in the 21st century: bringing the vision to life. Toxicol Sci. 2009, 107 (2): 324-330.View ArticlePubMedGoogle Scholar
- Bhavani S, Nagargadde A, Thawani A, Sridhar V, Chandra N: Substructure-based support vector machine classifiers for prediction of adverse effects in diverse classes of drugs. J Chem Inf Model. 2006, 46 (6): 2478-2486. 10.1021/ci060128l.View ArticlePubMedGoogle Scholar
- Gepp MM, Hutter MC: Determination of hERG channel blockers using a decision tree. Bioorg Med Chem. 2006, 14 (15): 5325-5332. 10.1016/j.bmc.2006.03.043.View ArticlePubMedGoogle Scholar
- Bidault Y: A flexible approach for optimising in silico ADME/Tox characterisation of lead candidates. Expert Opin Drug Metab Toxicol. 2006, 2 (1): 157-168. 10.1517/17425255.2.1.157.View ArticlePubMedGoogle Scholar
- Demel MA, Schwaha R, Kramer O, Ettmayer P, Haaksma EE, Ecker GF: In silico prediction of substrate properties for ABC-multidrug transporters. Expert Opin Drug Metab Toxicol. 2008, 4 (9): 1167-1180. 10.1517/17425255.4.9.1167.View ArticlePubMedGoogle Scholar
- Hammann F, Gutmann H, Baumann U, Helma C, Drewe J: Classification of cytochrome p(450) activities using machine learning methods. Mol Pharm. 2009, 6 (6): 1920-1926. 10.1021/mp900217x.View ArticlePubMedGoogle Scholar
- Octavia Y, Tocchetti CG, Gabrielson KL, Janssens S, Crijns HJ, Moens AL: Doxorubicin-induced cardiomyopathy: from molecular mechanisms to therapeutic strategies. J Mol Cell Cardiol. 2012, 52 (6): 1213-1225. 10.1016/j.yjmcc.2012.03.006.View ArticlePubMedGoogle Scholar
- Greineder CF, Kohnstamm S, Ky B: Heart failure associated with sunitinib: lessons learned from animal models. Curr Hypertens Rep. 2011, 13 (6): 436-441. 10.1007/s11906-011-0225-8.View ArticlePubMedGoogle Scholar
- Force T, Kolaja KL: Cardiotoxicity of kinase inhibitors: the prediction and translation of preclinical models to clinical outcomes. Nature reviews Drug discovery. 2011, 10 (2): 111-126. 10.1038/nrd3252.View ArticlePubMedGoogle Scholar
- Braam SR, Tertoolen L, van de Stolpe A, Meyer T, Passier R, Mummery CL: Prediction of drug-induced cardiotoxicity using human embryonic stem cell-derived cardiomyocytes. Stem Cell Res. 2010, 4 (2): 107-116. 10.1016/j.scr.2009.11.004.View ArticlePubMedGoogle Scholar
- Braam SR, Passier R, Mummery CL: Cardiomyocytes from human pluripotent stem cells in regenerative medicine and drug discovery. Trends Pharmacol Sci. 2009, 30 (10): 536-545. 10.1016/j.tips.2009.07.001.View ArticlePubMedGoogle Scholar
- Fielden MR, Halbert DN: Iconix Biosciences, Inc. Pharmacogenomics. 2007, 8 (4): 401-405. 10.2217/146224184.108.40.2061.View ArticlePubMedGoogle Scholar
- Ganter B, Snyder RD, Halbert DN, Lee MD: Toxicogenomics in drug discovery and development: mechanistic analysis of compound/class-dependent effects using the DrugMatrix database. Pharmacogenomics. 2006, 7 (7): 1025-1044. 10.2217/146224220.127.116.115.View ArticlePubMedGoogle Scholar
- Chindelevitch L, Ziemek D, Enayetallah A, Randhawa R, Sidders B, Brockel C, Huang ES: Causal reasoning on biological networks: interpreting transcriptional changes. Bioinformatics. 2012, 28 (8): 1114-1121. 10.1093/bioinformatics/bts090.View ArticlePubMedGoogle Scholar
- Enayetallah AE, Ziemek D, Leininger MT, Randhawa R, Yang J, Manion TB, Mather DE, Zavadoski WJ, Kuhn M, Treadway JL: Modeling the mechanism of action of a DGAT1 inhibitor using a causal reasoning platform. PloS one. 2011, 6 (11): e27009-10.1371/journal.pone.0027009.View ArticlePubMedPubMed CentralGoogle Scholar
- Galindo CL, Skinner MA, Errami M, Olson LD, Watson DA, Li J, McCormick JF, McIver LJ, Kumar NM, Pham TQ: Transcriptional profile of isoproterenol-induced cardiomyopathy and comparison to exercise-induced cardiac hypertrophy and human cardiac failure. BMC Physiol. 2009, 9: 23-10.1186/1472-6793-9-23.View ArticlePubMedPubMed CentralGoogle Scholar
- Archer KJ, Dumur CI, Joel SE, Ramakrishnan V: Assessing quality of hybridized RNA in Affymetrix GeneChip experiments using mixed-effects models. Biostatistics. 2006, 7 (2): 198-212.View ArticlePubMedGoogle Scholar
- Sammon JW: A nonlinear mapping for data structure analysis. IEEE Trans Comput. 1969, 18 (5): 401-409.View ArticleGoogle Scholar
- Smyth GK: Linear models and empirical bayes methods for assessing differential expression in microarray experiments. Stat Appl Genet Mol Biol. 2004, 3: Article3-PubMedGoogle Scholar
- Benjamini Y, Hochberg Y: Controlling the false discovery rate: a practical and powerful approach to multiple testing. Journal of the Royal Statistical Society Series B (Methodological). 1995, 57 (1): 289-300.Google Scholar
- Shannon P, Markiel A, Ozier O, Baliga NS, Wang JT, Ramage D, Amin N, Schwikowski B, Ideker T: Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res. 2003, 13 (11): 2498-2504. 10.1101/gr.1239303.View ArticlePubMedPubMed CentralGoogle Scholar
- Jain AK, Dubes RC: Algorithms for clustering data. 1988, Upper Saddle River: Prentice-Hall, Inc.Google Scholar
- Teekakirikul P, Eminaga S, Toka O, Alcalai R, Wang L, Wakimoto H, Nayor M, Konno T, Gorham JM, Wolf CM: Cardiac fibrosis in mice with hypertrophic cardiomyopathy is mediated by non-myocyte proliferation and requires Tgf-beta. J Clin Invest. 2010, 120 (10): 3520-3529. 10.1172/JCI42028.View ArticlePubMedPubMed CentralGoogle Scholar
- Van der Heiden K, Cuhlmann S, le Luong A, Zakkar M, Evans PC: Role of nuclear factor kappaB in cardiovascular health and disease. Clin Sci. 2010, 118 (10): 593-605. 10.1042/CS20090557.View ArticlePubMedGoogle Scholar
- Aragno M, Mastrocola R, Alloatti G, Vercellinatto I, Bardini P, Geuna S, Catalano MG, Danni O, Boccuzzi G: Oxidative stress triggers cardiac fibrosis in the heart of diabetic rats. Endocrinology. 2008, 149 (1): 380-388.View ArticlePubMedGoogle Scholar
- Wolkart G, Kaber G, Kojda G, Brunner F: Role of endogenous hydrogen peroxide in cardiovascular ischaemia/reperfusion function: studies in mouse hearts with catalase-overexpression in the vascular endothelium. Pharmacol Res. 2006, 54 (1): 50-56. 10.1016/j.phrs.2006.02.005.View ArticlePubMedGoogle Scholar
- Rona G: Catecholamine cardiotoxicity. J Mol Cell Cardiol. 1985, 17 (4): 291-306. 10.1016/S0022-2828(85)80130-9.View ArticlePubMedGoogle Scholar
- Kojima S, Wu ST, Wikman-Coffelt J, Parmley WW: Acute amiodarone terminates ventricular fibrillation by modifying cellular Ca++ homeostasis in isolated perfused rat hearts. J Pharmacol Exp Ther. 1995, 275 (1): 254-262.PubMedGoogle Scholar
- Haworth RA, Goknur AB, Berkoff HA: Inhibition of ATP-sensitive potassium channels of adult rat heart cells by antiarrhythmic drugs. Circulation research. 1989, 65 (4): 1157-1160. 10.1161/01.RES.65.4.1157.View ArticlePubMedGoogle Scholar
- Goldfine ID, Maddux B, Woeber KA: Effect of amiodarone on L-triiodothyronine stimulation of [3H] thymidine incorporation into GH3 cells. J Endocrinol Invest. 1982, 5 (3): 165-168.View ArticlePubMedGoogle Scholar
- Jonckheer MH, Blockx P, Broeckaert I, Cornette C, Beckers C: ‘Low T3 syndrome’ in patients chronically treated with an iodine-containing drug, amiodarone. Clin Endocrinol (Oxf). 1978, 9 (1): 27-35. 10.1111/j.1365-2265.1978.tb03569.x.View ArticleGoogle Scholar
- Morgan P, Van Der Graaf PH, Arrowsmith J, Feltner DE, Drummond KS, Wegner CD, Street SD: Can the flow of medicines be improved? Fundamental pharmacokinetic and pharmacological principles toward improving Phase II survival. Drug Discov Today. 2012, 17 (9–10): 419-424.View ArticlePubMedGoogle Scholar
- Puppala D, Collis LP, Sun SZ, Bonato V, Chen X, Anson B, Pletcher M, Fermini B, Engle SJ: Comparative gene expression profiling in human-induced pluripotent stem cell–derived cardiocytes and human and cynomolgus heart tissue. Toxicol Sci. 2013, 131 (1): 292-301. 10.1093/toxsci/kfs282.View ArticlePubMedGoogle Scholar
- Guo L, Abrams RM, Babiarz JE, Cohen JD, Kameoka S, Sanders MJ, Chiao E, Kolaja KL: Estimating the risk of drug-induced proarrhythmia using human induced pluripotent stem cell-derived cardiomyocytes. Toxicol Sci. 2011, 123 (1): 281-289. 10.1093/toxsci/kfr158.View ArticlePubMedGoogle Scholar
- Lin Z, Will Y: Evaluation of drugs with specific organ toxicities in organ-specific cell lines. Toxicol Sci. 2012, 126 (1): 114-127. 10.1093/toxsci/kfr339.View ArticlePubMedGoogle Scholar
- Benbow JW, Aubrecht J, Banker MJ, Nettleton D, Aleo MD: Predicting safety toleration of pharmaceutical chemical leads: cytotoxicity correlations to exploratory toxicity studies. Toxicol Lett. 2010, 197 (3): 175-182. 10.1016/j.toxlet.2010.05.016.View ArticlePubMedGoogle Scholar
- Evans PM, Liu C: Roles of Krupel-like factor 4 in normal homeostasis, cancer and stem cells. Acta Biochim Biophys Sin (Shanghai). 2008, 40 (7): 554-564. 10.1111/j.1745-7270.2008.00439.x.View ArticleGoogle Scholar
- Zhang Y, Wang Y, Liu Y, Wang N, Qi Y, Du J: Kruppel-like factor 4 transcriptionally regulates TGF-beta1 and contributes to cardiac myofibroblast differentiation. PloS one. 2013, 8 (4): e63424-10.1371/journal.pone.0063424.View ArticlePubMedPubMed CentralGoogle Scholar
- Leenders JJ, Pinto YM, Creemers EE: Tapping the brake on cardiac growth-endogenous repressors of hypertrophic signaling. J Mol Cell Cardiol. 2011, 51 (2): 156-167. 10.1016/j.yjmcc.2011.04.017.View ArticlePubMedGoogle Scholar
- Al-Shabanah OA, Aleisa AM, Hafez MM, Al-Rejaie SS, Al-Yahya AA, Bakheet SA, Al-Harbi MM, Sayed-Ahmed MM: Desferrioxamine attenuates doxorubicin-induced acute cardiotoxicity through TFG-beta/Smad p53 pathway in rat model. Oxid Med Cell Longev. 2012, 2012: 619185-View ArticlePubMedPubMed CentralGoogle Scholar
- Wang Z, Yao T, Song Z: Extracellular signal-regulated kinases 1/2 suppression aggravates transforming growth factor-beta1 hepatotoxicity: a potential mechanism for liver injury in methionine-choline deficient-diet-fed mice. Exp Biol Med. 2010, 235 (11): 1347-1355. 10.1258/ebm.2010.010160.View ArticleGoogle Scholar
- Prud’homme GJ: Pathobiology of transforming growth factor beta in cancer, fibrosis and immunologic disease, and therapeutic considerations. Lab Invest. 2007, 87 (11): 1077-1091. 10.1038/labinvest.3700669.View ArticlePubMedGoogle Scholar
- The pre-publication history for this paper can be accessed here:http://www.biomedcentral.com/2050-6511/14/46/prepub
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.