Community Research and Development Information Service - CORDIS


HECATOS Report Summary

Project ID: 602156
Funded under: FP7-HEALTH
Country: Netherlands

Periodic Report Summary 3 - HECATOS (Hepatic and Cardiac Toxicity Systems modelling)

Project Context and Objectives:
Liver toxicity and cardiovascular toxicity have proven to be the major causes for two out of three market withdrawals of drugs and drug development project terminations in clinical phases I-III from 1992 to 2002 in the respective time period. Between 2007 and 2010, as late as at Phase III, 21% of the failures across all therapeutic areas were still attributable to safety issues. Human toxicity of novel drugs may even not be recognized for years after a drug has been introduced to the market. Overall, unexpected liver and cardiovascular toxicity are major reasons for drug candidate failure.
Within this context, HeCaToS aims at developing integrative in silico tools for predicting human liver and heart toxicity. The objective is to develop an integrated modeling framework, by combining advances in computational chemistry and systems toxicology, for modelling toxic perturbations in liver and heart across multiple scales. This framework includes vertical integrations of representations from drug(metabolite)-target interactions, through macromolecules/proteins, to (sub-)cellular functionalities and organ physiologies, and even the human whole-body level. In view of the importance of mitochondrial deregulations and of immunological dysfunctions associated with hepatic and cardiac drug-induced injuries, focus is on these particular Adverse Outcome Pathways. Models are populated with data from innovative in vitro 3D liver and heart assays challenged with prototypical hepato- or cardiotoxicants; data are generated by advanced molecular and functional analytical techniques retrieving information on key (sub-)cellular toxic evens. For validating perturbed AOPs in vitro in appropriate human investigations, case studies on patients with liver injuries or cardiomyopathies due to adverse drug effects, have been developed, and biopsies are subjected to similar analyses. Existing ChEMBL and diXa data infrastructures are advanced for data gathering, storing and integrated statistical analysis.

Model performance in toxicity prediction is assessed by comparing in silico predictions with experimental results across a multitude of read-out parameters, which in turn will suggest additional experiments for further validating predictions. HeCaToS, organized as a private-public partnership, will generate major socioeconomic impact because it will develop better chemical safety tests leading to safer drugs, but also industrial chemicals, and cosmetics, thereby improving patient and consumer health, and sustaining EU’s industrial competitiveness.

Project Results:
WP1: Computational chemistry
As reported in D1.6, an integrated system has been created that enumerates metabolites for an input compound and generates predictions of the activities of these species against a panel of toxicity-associated biomolecular targets, activity against these targets representing possible interactions with Adverse Outcome Pathways.
In addition, small molecule passive membrane transfer is simulated using MD with an improved lipid bilayer force field. As an example high level ab initio DFT calculations have been conducted of activation barriers to arene oxidation. The results correlate well with previous studies undertaken in the development of the SMARTCyp application, both for the full system and a surrogate methoxy (MeO) radical that will make routine calculations more tractable. AMBER for field parameterization has been completed for the entire reference compound library.
A web-based system has been created that takes the predicted set of Cytochrome P450 metabolites and produces predictions of activities against the targets of toxicological interest with reference to Adverse Outcome Pathways.
Furthermore, test cases of DNMT-1 and cardiolipin have been elaborated for the purpose of developing homology models for membrane transporters downstream of protein receptors and initial MD simulations have been conducted of holoenzyme and inhibitor-bound structures.

WP2: Molecular pathway modelling
With respect to mitochondrial signalling evidence was found for adrenergic signalling as an important pathway in the context of cardiac toxicity resulting from anthracyclines. Modelling of immune response has been deferred to a later stage of the project when more data from liver micro tissue will be available. Changes in adrenergic signalling are prominent in our data on cardiotoxicity. Combining our observations with the findings on changes in mitochondrial function, there is now growing evidence that a combination of oxidative stress and changes in adrenergic signalling may lead to the induction of apoptotic processes. This in turn leads to progressive degeneration of cardiac tissue under the effect of a toxic dose of anthracyclines. Our analysis suggests that this induction may be mediated via the regulation of the BCL-2 gene. BCL-2 is an integral outer mitochondrial membrane protein that regulates cell death by controlling the mitochondrial membrane permeability. Expression changes were much weaker and more difficult to interpret in case of the lower dose (therapeutic dose). In the further analysis we want to use our modelling approach to gain better understanding of competing effects in molecular pathways inducing apoptotic processes.

WP3: Physiological modelling
A multi-scale cardiac model, covering the cellular, organ, and wider systemic levels, has been created. Parameters at all levels can be set using experimental and clinical data. We have successfully used the changes in protein expression reported from mass spectrometry protein concentration from the cardiac microtissues exposed to a series of anthracyclins recorded over a two week dynamic exposure protocol to alter the contractile function of the simulated human heart and predict how these changes in protein concentration alter the ejection fraction of the heart.
With respect to liver toxicity, a physiologically-based model of drug-induced cholestasis is currently developed. To this end, bile-acid metabolism is described at a large level of physiological detail.

WP4: Vertical model integration
A generic workflow for model-based contextualisation of ‘omics data (PICD: PBPK-based in vivo contextualisation of human in vitro toxicity data) has been developed and published (Thiel et al., Arch. Toxicol. 2016). Following an initial use case for acute and chronic cases of azathioprine-induced liver injury, PICD was next applied in a comparative analysis of 15 hepatoxicants from the HeCaToS list of compounds and in a case study analysing stimulatory and inhibitory effects during acetaminophen and caffeine co-administration.
In a complementary analysis a genotype specific PBPK/PD model for isoniazid-based therapies of mycobacterial infections has been developed and published (Cordes et al., Antimicrob Agents Chemother. 2016). The model was used to analyse the trade-off between antibacterial activity and liver tissue exposure of a toxic metabolite in specific patient subgroups. In particular, the model was used to assess different treatment schedules.

WP5: in vitro assays
Work has been focused on:
- Constitution of a panel of dose response curves for the cardiotoxic and hepatotoxic compounds to help determine the toxic dose in the assay design;
- Completed assay run for cardiotoxic compound Danuroubicin as well as the cardiac un-treated control and vehicle fluctuation control;
- Completed assay run for hepatotoxic compounds Azathioprine and Acetaminophen as well as the hepatic vehicle run (0.1% DMSO) and un-treated hepatic control;
- Completed controls that fully characterise microtissues and vehicle effect.

WP6: Patients
During year 3, a substantial number of patients with the suspicion diagnosis of DILI, has been identified, investigated and finally unequivocally classified (ca. 60 patients). A large number of serum samples along the onset of the DILI event has been gathered and constitutes a valuable collection for the metabolomic studies being currently initiated. Liver samples of almost 20 patients are available for further studies. Metabolomic studies are scheduled for the 4th year. The causality and occurrence of acute events recorded along this period, has resulted in the consideration of clavulanic acid as a model drug for cholestasis which is now in the list of assayed compounds. Thus it will be possible to compare experimental data obtained in vitro and in vivo experiments.

About 100 patients demonstrating acute cardiotoxicity upon anthracycline treatment have been included by now. From 1 patient also cardiac biopsies have been generated. In addition, 16 patients with chronic cardiomyopathy as well as 16 controls have been included. From all these subjects cardiac biopsies have been successfully taken.
Measurement of high sensitive cardiac troponin T from serum samples of acute cardiotoxicity patients pre- and post-chemotherapy took place at the Maastricht hospital. A database was built for the analysis of these patients. However, patient numbers are currently too low to reach statistical significance. Clinical data from about 100 patients with serial chemotherapy has been completed.

WP7: ‘Omics analyses
The analysis of DNA, RNA, proteins, and metabolites on the omics level has been continued and datasets for a first series of compounds have been generated in the reporting period and made available for quality control, data warehousing, and data integration WPs. An additional comprehensive series of control experiments on 3D tissues has been carried out to further solidify the basis for the interpretation of treatment-data of spheroid cultures. Further work has been invested in the optimization of minimal input analytical protocols for the analysis of clinical samples. While work with the minute amounts of tissues and media from the culture experiments remain a challenge, the analysis of (also very small) biopsy samples from clinical patients has been greatly optimized and started to generate important and comprehensive datasets at all molecular levels. All data have been made available to the downstream WPs and data processing and interpretation workflows and methods have been coordinated between the different WPs.

WP8: Functional analyses
The following tasks have been performed:
- Application of high-sensitivity single spheroid assay measurement of ETC activity to all available complex dosing schemes (WP4) for the cardiac spheroid model, generating data 7 days post initial treatment for Doxorubicin & Epirubicin compounds.
- Data generated on the impact of short term (2h) treatment of the 3 anthracyclines (Idarubucin, Doxorubicin and Epirubicin) on the mitochondrial function of cardiomyocyte-derived spheroids using the dosing scheme developed within WP4.
- Application of high-sensitivity single spheroid assay measurement of ETC activity to all available complex dosing schemes (WP4) for the liver spheroid model, generating data 7 days post initial treatment for Acetaminophen and Azathioprine compounds.
- Further demonstrated capacity to measure O2 consumption, ECA and intracellular O2 on xCELLigence E-plates thereby facilitating sequential analysis of 2D cardiomyocyte mitochondrial function / metabolism and beating with a view to providing a more holistic picture of the impact of drug treatment on cardiomyocyte function.
- Caspase activity data generated for Doxorubicin & Epirubicin treatment on cardiac spheroids and Acetaminophen & Azathioprine treated liver spheroids, 7 days post initial treatment using complex dosing schemes from WP4.
- Caspase activity data generated at short time-point exposure (2h) with the 3 Anthracyclines (Idarubucin, Doxorubicin & Epirubicin) treatment of cardiac spheroids, using the Therapeutic and Toxic concentrations from WP4 dosing schemes.
- xCELLigence data generated on a 2D iPS cardiomyocyte model treated with classical mitochondrial compounds, and Cardio-active drugs to demonstrate specific altered beating, such as Isoproterenol, Nifedipine and E-4031, facilitating the delineation of inter-relationships between cardiomyocyte beating and cell metabolism when sequential assays measurements are performed.

WP9: Data Warehousing
The HeCaToS data warehouse became operational and data upload procedures were defined and refined:
- “ToxDB: pathway-level interpretation of drug-treatment data” paper was published;
- Tables specifically designed for storing routes of metabolism data were added to the ChEMBL data model in ChEMBL 21;
- Established a process for annotating ChEMBL assay descriptions using the Medical Subject Headings (MeSH) controlled vocabulary and processes all 1.2 million ChEMBL assay descriptions;
- Integrated the latest data from the Tox21 Phase II data set into ChEMBL version 22;
- Incorporated data gathered on the L-type calcium channel, a relevant gap target, into the ChEMBL database;
- Performed literature searches for routes of metabolism for all HeCaToS compounds and were available extracted these pathways and enzymes into the new metabolism tables.

WP10: Data quality control
Work has focused on:
- Profiler 10.0 release, incl. new workflows and data management functionalities;
- Seamless integration between Genedata and EBI data infrastructures;
- First analysis of RNA-seq data from cardiac microtissues treated with anthracyclines;
- First analyses of RNA-seq data from biopsies;
- Characterization of DMSO-effect on microtissue mRNA expression leading to correction of experimental design;
- First analysis of ChIP-Seq data from microtissues;
- Establishment of an efficient and automated pipeline for analyzing proteomics data;
- Primary analysis of proteomics data.

WP11: Integrated statistics
The following output has been achieved:
- Five scientific publications;
- Developed new workflow for MeDIP-seq data analysis;
- Developed a statistical model for the quantification of absolute methylation levels from MeDIP-seq enrichment data;
- Explored public repositories for additional data sets on liver and heart toxicity in vivo and in vitro;
- Developed frameworks for integrative data analysis;
- Identified potential biomarkers that extrapolate from in vitro to in vivo cardiac drug toxicity;
- Conducted integrative analysis in cooperation with WPs 2, 5, 6, 7, 10 and 12.

WP12: Predictive comparisons
The following has been generated:
- Two scientific publications;
- Computed pathway responses for cardiotoxicity from a large number of drugs and identified molecular themes related to cardiotoxicity (D12.1), in cooperation with WP2;
- Started the anthracycline use case;
- Further developed a repository for the molecular interpretation of benchmark omics data, ToxDB;
- Integrated this repository to the central project data warehouse (in cooperation with WP9).

WP13: Training and dissemination
Publication of the first HeCaToS newsletter, and updating the HeCaToS website: At the 3rd HeCaToS Annual Consortium Meeting a ‘Workshop on Data infrastructure, quality control, data warehousing and integrated statistics’ has been organised.

Potential Impact:
The ultimate results of the HeCaToS project are to be in silico models for liver and for heart toxicity model, which have been developed by combining recent advances in computational chemistry and systems biology and systems toxicology. These models will predict toxic perturbations in the liver and in the heart across multiple scales, from molecular interactions, through macromolecules, to (sub-)cellular functionalities and organ physiologies, and even the intact human being. Our in silico models will achieve high predictive accuracy because we will be able to iteratively inform these models with novel data on perturbed AOPs from the best existing in vitro heart and liver assays, obtained by cutting edge methodologies, in association with chemical structure and validated against relevant clinical hepatic and cardiac data.

The potential socioeconomic impact of these in silico models for predicting heart and liver toxicity, relates to the fact that:
- Our models, once successfully developed, will outperform current animal-based models for predicting human drug safety, thus decreasing current attrition rates of new candidate-drugs;
- Our models, though developed for the purpose of predicting drug safety, may also be applied in the context of REACH concerning the safety assessment of industrial chemicals, and for assessing safety of cosmetic products since they are non-animal based and therefore comply with the EU ban to test cosmetic ingredients and products on animals;
- Our models will ultimately lead to lower numbers of animal test for evaluating toxicity, and are therefore in compliance with the 3R Principle.

In Year 3, the HeCaToS project has successfully exploited the workflow established in Year 2 , from challenging organotypical in vitro models to prototypical toxicants, through benchmarking this data against data obtained from biopsies taken from patients treated with the same drugs and demonstrating signs of target organ toxicity, to in silico modelling of molecular, functional and physiological responses to such toxicants. Cross-omics data and functional data have been generated from cardiac microtissues treated with 4 anthracyclins and appropriate controls. Also, the first set of cardiac biopsies from anthracycline-treated patients has been analysed. Data have been used to populate molecular models which in particular yielded a list of differentially expressed proteins which are used for validating the initial physiological model for cardiac toxicity.
In addition, work on the liver microtissues has started: molecular data from 2 test compounds have been generated.
In Year 4 it is foreseen that all experimental work on the cardiac and hepatic microtissues will be completed, and all patient samples analysed.

List of Websites:

Related information

Reported by



Life Sciences
Follow us on: RSS Facebook Twitter YouTube Managed by the EU Publications Office Top