Human-based approaches to pharmacology and cardiology: an interdisciplinary and intersectorial workshop

REVIEW Human-based approaches to pharmacology and cardiology: an interdisciplinary and intersectorial workshop Blanca Rodriguez1*, Annamaria Carusi2, Najah Abi-Gerges3, Rina Ariga4, Oliver Britton1, Gil Bub5, Alfonso Bueno-Orovio1, Rebecca A.B. Burton5, Valentina Carapella6, Louie Cardone-Noott1, Matthew J. Daniels4, Mark R. Davies7, Sara Dutta1, Andre Ghetti3, Vicente Grau6, Stephen Harmer8, Ivan Kopljar9, Pier Lambiase10, Hua Rong Lu9, Aurore Lyon1, Ana Minchole1, Anna Muszkiewicz1, Julien Oster6, Michelangelo Paci11, Elisa Passini1, Stefano Severi12, Peter Taggart10, Andy Tinker8, Jean-Pierre Valentin13, Andras Varro14, Mikael Wallman15, and Xin Zhou1 1 Department of Computer Science, University of Oxford, Oxford, UK; 2 Medical Humanities, University of Sheffield, Sheffield, UK; 3 AnaBios Corporation, San Diego Science Center, San Diego, CA 92109, USA; 4 Radcliffe Department of Medicine, University of Oxford, John Radcliffe Hospital, Oxford, UK; 5 Department of Physiology, Anatomy and Genetics, University of Oxford, Oxford, UK; 6 Department of Engineering Science, University of Oxford, Oxford, UK; 7 QT-Informatics Limited, Macclesfield, UK; 8 William Harvey Heart Centre, Barts and The London School of Medicine and Dentistry, Queen Mary University of London, Charterhouse Square, London, UK; 9 Discovery Sciences, Dis&Dev Research, Janssen Pharmaceutical NV, Beerse, Belgium; 10 Institute of Cardiovascular Science, University College London, Bars Heart Centre, London, UK; 11 Department of Electronics and Communications Engineering, Tampere University of Technology, BioMediTech, Tampere, Finland; 12 Department of Electrical, Electronic and Information Engineering, University of Bologna, Cesena 47521, Italy; 13 Non-Clinical Development, UCB-Biopharma, Braine L’Alleud B-1420, Belgium; 14 University of Szeged, Szeged, Hungary; and 15 Fraunhofer-Chambers Centre, Gothenburg, Sweden Received 11 June 2015; accepted after revision 20 August 2015; online publish-ahead-of-print 29 November 2015 Both biomedical research and clinical practice rely on complex datasets for the physiological and genetic characterization of human hearts in health and disease. Given the complexity and variety of approaches and recordings, there is now growing recognition of the need to embed computational methods in cardiovascular medicine and science for analysis, integration and prediction. This paper describes a Workshop on Computational Cardiovascular Science that created an international, interdisciplinary and inter-sectorial forum to define the next steps for a human-based approach to disease supported by computational methodologies. The main ideas highlighted were (i) a shift towards humanbased methodologies, spurred by advances in new in silico,in vivo,in vitro, and ex vivo techniques and the increasing acknowledgement of the limitations of animal models. (ii) Computational approaches complement, expand, bridge, and integrate in vitro,in vivo, and ex vivo experimental and clinical data and methods, and as such they are an integral part of human-based methodologies in pharmacology and medicine. (iii) The effective implementation of multiand interdisciplinary approaches, teams, and training combining and integrating computational methods with experimental and clinical approaches across academia, industry, and healthcare settings is a priority. (iv) The human-based cross-disciplinary approach requires experts in specific methodologies and domains, who also have the capacity to communicate and collaborate across disciplines and cross-sector environments. (v) This new translational domain for human-based cardiology and pharmacology requires new partnerships supported financially and institutionally across sectors. Institutional, organizational, and social barriers must be identified, understood and overcome in each specific setting. ----------------------------------------------------------------------------------------------------------------------------------------------------------- Keywords Human-based methods †Computational approaches †Human electrophysiology †Stem-cell-derived cardiomyocytes †Computer modelling and simulations †Arrhythmias †Biomarkers *Corresponding author. Tel: +44 0 1865610806. E-mail address: [email protected] &The Author 2015. Published by Oxford University Press on behalf of the European Society of Cardiology. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. Europace (2016) 18, 1287–1298 doi:10.1093/europace/euv320 Downloaded from https://academic.oup.com/europace/article/18/9/1287/2802553 by Universidad de Zaragoza user on 26 February 2021 Motivation Both biomedical research and clinical practice rely on complex datasets for the physiological and genetic characterization of human hearts in health and disease. The information shaping our knowledge of human hearts is obtained from a variety of techniques and models, including recordings obtained in vivo invasively and noninvasively, in ex vivo tissue and isolated human adult cardiomyocytes recordings, and more recently in vitro using human stem-cell-derived cardiomyocytes. Increasing evidence suggests that non-human animal models may have limited ability to predict human in vivo effects due to important species differences between humans, dogs, guinea pigs, and rabbits. 1–3 Thus, methods firmly rooted in understanding physiology and pathophysiology in humans are clearly needed. Advances in imaging technologies such as the multiple modalities of cardiac magnetic resonance (CMR) are combined with body surface or intra-cardiac electrophysiological recordings to evaluate in specific patients the in vivo structural and functional implications of cardiac disease. Recent progress in research using human cardiomyocytes derived from induced pluripotent stem cells promises exciting new developments as it allows the in vitro characterization of the phenotype of cardiomyocytes of specific patients, and therefore has the potential of introducing the flexibility of in vitro methodologies in personalized medicine. Furthermore, ex vivo tissue from biopsies or from donor human hearts provides measurements of tissue, cellular, and ionic properties of human adult cardiomyocytes in nondiseased and diseased conditions. Each of these types of human-based assays and measurements provides us a single snapshot from one perspective of a complex set of variables through both time and spatial dimensions. In turn, this complex set of variables is able to define and explain the myriad of dynamic mechanisms and properties that underlie the activity of human hearts in health and disease. Given the complexity and variety of approaches and recordings, there is now growing recognition of the need to embed computational methods in cardiovascular medicine and science for analysis, integration, and prediction. Computational approaches in biomedicine range over a variety of techniques for signal, data, and image analysis, but also importantly multiscale modelling and simulation. Together, they provide a synergistic approach to organize and augment the information obtained from experimental and clinical recordings. The benefits gained include the quantitative analysis and organized reassembly of multiscale and multimodality datasets to probe, challenge and expand our knowledge of the complex and dynamic interactions in cardiac electrophysiology. Advances in computational cardiac electrophysiology were recently illustrated in two dedicated special issues of Europace 2014. 4,5 Furthermore, new initiatives such as the Comprehensive in vitro Proarrhythmia Assay (CiPA) launched by the United States Food and Drug Administration (FDA) recognize the potential of human-based in silico and in vitro approaches as a new paradigm for drug safety assessment. 6 On 17 September 2014, a Workshop on Computational Cardiovascular Science was hosted at the University of Oxford with the aim of creating an international, interdisciplinary, and inter-sectorial forum to discuss current trends in computational technologies to augment cardiovascular physiology, pharmacology, and medicine, and to propose solutions for the replacement, refinement and reduction of animal experimentation. 7 Invited participants included experts in cardiology, computer science, physiology, pharmacology, philosophy, and biomedical engineering from academia and industry, from 11 Universities and 12 companies, from UK, several countries in Europe, USA, and Japan. In this paper, we aim at describing the main ideas discussed during the meeting, rather than providing a thorough review of the literature. Table 1summarizes the forms of human-based in vivo,ex vivo,in vitro,andin silico experiments and techniques discussed, which are further illustrated in Figure 1. The main ideas highlighted through the workshop were the following: †A shift towards human-based methodologies in pharmacology and medicine is occurring, spurred by advances in new in silico, in vivo,in vitro, and ex vivo techniques and the increasing acknowledgement of the limitations of animal models. †Computational approaches complement, expand, bridge, and integrate in vitro,in vivo, and ex vivo experimental and clinical data and methods, and as such they are an integral part of humanbased methodologies in pharmacology and medicine. †The effective implementation of multiand interdisciplinary approaches, teams, and training combining and integrating computational methods with experimental and clinical approaches across academia, industry, and healthcare settings is a priority. †The human-based cross-disciplinary approach requires experts in specific methodologies and domains, who also have the capacity to communicate and collaborate across disciplines and to work productively in interdisciplinary and cross-sector environments. †This new translational domain for human-based cardiology and pharmacology requires new partnerships supported financially and institutionally across pharma and biotech industry, contract research organizations (CRO), academia and research institutes, technology and service providers, non-profit and governmental organizations, and regulatory agencies. Institutional, organizational, and social barriers must be identified, understood, and overcome in each specific setting. Description of the workshop The workshop consisted of four sessions aiming at exploring different aspects of human-based cardiovascular science, and specifically the synergies with in silico approaches in the three main experimental settings, i.e. in vivo,in vitro,andex vivo. Table 1summarizes the forms of experiments and techniques discussed. Human in vivo cardiovascular science Prof. Peter Taggart examined the challenges involved in obtaining basic electrophysiological data from in vivo human subjects, during routine clinical procedures. The example discussed was the application of percutaneous transluminal coronary angioplasty (PTCA) to study in vivo the effect of myocardial ischaemia (a major cause of mortality) on human electrophysiology, by recording either monophasic action potentials or unipolar electrograms on the ventricular endocardium in the region served by the artery undergoing PTCA. In vivo electrophysiological recording techniques include the use of the multi-electrode sock (264 electrodes) over the human ventricles to study global patterns of activation such as ventricular B. Rodriguez et al.1288 Downloaded from https://academic.oup.com/europace/article/18/9/1287/2802553 by Universidad de Zaragoza user on 26 February 2021 fibrillation, action potential duration (APD) changes, and postrepolarization refractoriness during early ischaemia and the demonstration of mechano-electric feedback in humans. 10–20 Findings addressed the debate as to whether rotors or multiple wavelets drive ventricular fibrillation (VF) and showed that both coexist in human VF, and the very rapid time course of the early electrophysiological changes during early ischaemia in humans. In vivo electrophysiological recordings are critical for translational research (from basic to clinical) but they are limited due to considerations for patient comfort and safety. Therefore, they need to be complemented by and combined with alternative ways of probing the human heart, including non-invasive in vivo imaging as well as ex vivo investigations, as described below. Dr Rina Ariga provided an overview of multimodality magnetic resonance imaging and its application in patients with hypertrophic cardiomyopathy (HCM). HCM is the most common genetic heart disease (affects 1 in 500) 21 and the commonest cause of sudden cardiac death in the young. 22 Transthoracic echocardiography (TTE) is routinely used to assess HCM, but is limited in patients with poor acoustic windows or poor visualization of some LV regions. CMR ............................................................................................................................................................................... Table 1 Summary of human-based in vivo,ex vivo, and in vitro techniques and in silico approaches Human-based Data acquisition technique Limitations of the data Progress of in silico approaches In vivo Electrophysiological recordings during clinical procedures [Taggart, Zhou] – Limited data sets due to ethical and practical obstacles; – Datasets derived from diseased hearts; – Inter-patient variability; – Limited experimentation – Signal analysis and integration; – Multiscale electrophysiological models and simulations; – Population of models to mimic action potential variability Multimodality imaging including magnetic resonance [Ariga, Grau] – Limited data sets; – No experimentation – Image analysis (tissue characterization); – Ventricular shape analysis; – Structural models and simulations – Computer models for link between structure and diffusion; – And link between micro structure and function Body surface potentials, electrocardiogram [Minchole, Lu] – Captures global patterns of heart behaviour; – Variability – Automated quantification of ECG features for clinical diagnosis and identification of new bioamarkers (morphological QRS or T-wave based, iCEB); – Electrocardiographic imaging; – Multiscale human heart simulations from ion channel and microstructure to the electrocardiogram mHealth recordings through mobile devices [Oster] – Very large quantities of data not amenable to manual analysis; – Noisy and patchy data; – Social, ethical and legal challenges – Automated and semi-automated techniques for analysis, such as machine learning, implementable on mobile phones Isolated human primary cells and non-clinical, real-world data from biopsies and medical histories [Ghetti] – Limited data sets; – Social, ethical and legal challenges – Computational models to integrate experimental data and to investigate multiscale mechanisms of disease and pharmacological interventions Ex vivo Microelectrode, optical mapping, patch clamp, protein, and mRNA expression [Varro, Britton, Dutta] – Limited availability, and mostly from diseased hearts; – Variability; – Difficulty of technique implementation (cell isolation; current separation); – Change of properties due to cell isolation – Data analysis and integration; – Multiscale models for greater contextualization; – Investigation of variability through approaches such as population of models In vitro Human cardiomyocytes derived from induced pluripotent stem cells (hiPSC-CMs) [Daniels, Severi, Kopljar, Harmer] – Inconsistent immaturity; – Variability and associated difficulty of comparison – Models to investigate variability, assist interpretation, and facilitate translation to in vivo cells – Models to investigate gene mutations; – Models for drug safety assessment Cell cultures and high-speed optical imaging [Burton] – Limited cross-institution and cross-sector access to experiments – Multiscale modelling to explain dynamics in heterogeneous preparations Human-based approaches to pharmacology and cardiology 1289 Downloaded from https://academic.oup.com/europace/article/18/9/1287/2802553 by Universidad de Zaragoza user on 26 February 2021 is now the gold standard in assessing mass, hypertrophy, volume, and function in HCM due to high spatial and temporal resolution. 23 Unlike TTE, CMR also provides tissue characterization using late gadolinium enhancement (detects focal fibrosis which has been associated with ventricular arrhythmias and SCD) 24 and T1 mapping (detects diffuse and focal fibrosis). CMR also offers insights into several hallmark features of HCM that are potential contributors of disease progression using novel techniques such as stress perfusion imaging (impaired perfusion), 25 blood oxygen level dependent (BOLD) imaging at stress (deoxygenation at stress), 26 phosphorus MR spectroscopy (abnormal myocardial energetics at rest with further deficit in exercise), 27 and most recently, diffusion tensor imaging (to assess fibre disarray). 28 CMR is not only a useful imaging adjunct in cases with limited TTE views, but also provides accurate disease characterization of subtle differences. Research using CMR is improving our understanding of this complex heterogeneous disease and is helping to guide risk stratification and treatment strategies. 29 Prof. Vicente Grau spoke about methods to investigate myocardial microstructure, including quantification using imaging as well as determination of functional repercussions of microstructural changes, investigated using computational modelling and simulation. Recent developments in MRI technology, in particular using diffusion sensitive sequences, allow the quantification of microstructure, initially in fixed hearts 8,30,31 or in hearts at different stages of contraction. 32 Diffusion MRI uses an indirect measurement to estimate cardiac structure, and its relationship to cardiac microstructure is not fully understood. Two methods to improve this understanding were discussed. Histology offers direct insights into microstructure, but three-dimensional reconstruction from histological slices is Body Body surface potentials ECG recordings mHealth data Arrhythmic-risk biomarkers High-resolution arrhythmia mechanisms Atrial and ventricular electrophysiology models in health and disease 1 0.5 0 −0.50 200 400 Time (ms) 600 Potential 2.00 1.38 0.75 0.12 −0.50 Organ Clinical EP Studies Multi-modality clinical MRI Tissue High-resolution MRI Histology Optical mapping Cell cultures Single cell Microelectrode recordings Protein/mRNA expression Optical mapping Ionic current Voltage/patch clamp Isolated cells/hiPSC-CMs Figure 1 Sources of experimental data integrated in computational models of human cardiac electrophysiology, and applications in physiology. Ionic current models are constructed mostly based on voltage/patch clamp data from ex vivo and in vitro preparations. The integration of ionic current models in single cell models, accounting for variability in protein expression and disease remodelling, allows for the simulation of the action potential and electrolyte concentrations in healthy and disease. Additionally, cardiac simulations at the whole organ and body levels require the construction of image-based anatomical models. When coupled to mathematical descriptions of electrical excitation through cardiac tissue, they allow for the high-resolution investigation of arrhythmia mechanisms based on clinical electrophysiology studies, as for the interpretation and identification of arrhythmic-risk biomarkers at the surface potential level. Transmural visualization of ventricular myofibre orientation, adapted from reference 8with permission. ECG/whole body simulation, adapted from reference 9with permission. B. Rodriguez et al.1290 Downloaded from https://academic.oup.com/europace/article/18/9/1287/2802553 by Universidad de Zaragoza user on 26 February 2021 challenging. 33 Computational models, simulating water diffusion and MRI acquisition sequences, can be used to provide a direct link between structure and diffusion. 34 The relationship between microstructural and functional changes is not fully understood, and here computational models can again provide a unique tool as shown for example in references. 35–37 In addition to imaging, clinical in vivo recordings to evaluate the human heart include the surface electrocardiograms (ECGs), a rapid and cost-effective method to acquire non-invasive recordings. A large body of research has been devoted to the quantification of ECG features for patient stratification in terms of arrhythmic risk and disease. 38 Dr Ana Minchole ´described computational approaches applied to the ECG for the detection of electrophysiological abnormalities and patient stratification in HCM. Using a database of Holter recordings obtained in HCM patients and volunteers, she described the quantification of standard ECG-based biomarkers such as QT T peak to T end, as well as new mathematical model-based morphological features of the QRS complex and T wave morphology. 39,40 Dr Minchole also described how information about the electrophysiological and structural signature of disease from the ionic to the whole organ level can be integrated in multiscale human heart models and used in the development and understanding of ECG-based biomarkers. 41 Multiscale simulations as illustrated in Figure 2allow the identification of key structural and functional factors that determine each of the ECG biomarkers and provide a deeper and more precise understanding of the information each of the biomarkers conveys. 42 The new knowledge also aids in the identification of more selective and specific biomarkers for specific disease conditions with complex functional and structural signatures such as HCM or myocardial infarction. Body surface ECGs combined with imaging data using mathematical algorithms to non-invasively reconstruct the electrical activity on the epicardial surface of the heart in vivo as demonstrated for example in references. 43–45 Human ex vivo and in vitro cardiovascular science Prof. Andras Varro began by discussing the knowledge we have of human electrophysiology largely through classic techniques such as conventional microelectrode and more modern patch clamp techniques, protein and RNA expression approaches using human cardiomyocytes from biopsies or donor hearts. 46–48 Obtaining good representative experimental human data is restricted by a number of practical problems. Sources of tissue particularly ventricular cells are difficult to obtain and isolation of cells is complicated by disease. The source of undiseased donor hearts is particularly limited. The process of cell isolation may also lead to changes from properties studied in multicellular tissue preparations. Furthermore, the separation of each of the ionic currents illustrated in Figure 3is problematic since the available pharmacological inhibitors are not totally selective. Finally, Prof. Varro highlighted the importance of human-based studies rather than highlighted the use of animal models. It is often underappreciated that there are substantial species differences between human and even large mammals, such as the dog, considered to be a representative model. 49,50 For example, marked differences in ventricular repolarization reserve have been reported between human and dog, 49 with larger rapid component of the delayed rectifying current (I Kr )inhumanbut stronger slow component (I Ks ) and inward rectifying current (I K1 ) 0.4–0.8 mV I 0.226 0.3 0.4 gks (nS/pF) 0.5 0.586 –0.4 –1 1 0 0.8 Figure 2 Computer simulation of the human heart electrophysiology from ion channel to body surface potentials and the electrocardiogram. Simulations are conducted using human biophysically detailed models considering heterogeneity in specific ionic properties (left, colour scale correspond to the maximum conductance of the slow component of the delayed rectifying current) to determine their effect on the spatiotemporal evolution of electrical potentials across the whole torso (middle, extracellular potentials throughout the torso) and on the ECG (right, main leads displayed). Human-based approaches to pharmacology and cardiology 1291 Downloaded from https://academic.oup.com/europace/article/18/9/1287/2802553 by Universidad de Zaragoza user on 26 February 2021 in the dog. Consequently, APD prolongation caused by selective I Kr block is three-fold larger in human than in dog, which suggests caution in translation of animal findings to human. Species-specific differences in pharmacological action between animal models have been also described, 50 with rabbit exhibiting larger APD prolongation and proneness to repolarization abnormalities upon selective I Kr block compared to others. Computational modelling and simulation can facilitate interspecies comparison by identifying and addressing differences and maximizing the re-use of experimental data from human hearts. Dr Oliver Britton described how the complex electrophysiological datasets obtained from ex vivo human hearts have been integrated in multiscale computer models, specifically focusing on human data over the past decade 51 –54 as examples. He described the recent construction of a population of human ventricular cell computational models that captures the inter-subject variability seen in action potential recordings from human ventricular tissue preparations from Prof Varro’s laboratory. The models in the population have a wide range of different configurations of ionic current strengths to mimic variability in a population as in reference 55. The team investigated whether there were configurations that were particularly vulnerable to developing repolarization abnormalities such as alternans and early afterdepolarizations (EADs), in response to blockade of different combinations of currents known to be important in repolarization—rapid delayed rectifying (I Kr ), slow delayed rectifying (I Ks ), inward rectifying (I K1 ), and late calcium (I CaL ) current. The computational approach therefore integrates and extends experimental recordings, generating predictions and refining hypotheses about proarrhythmic mechanisms that can then be tested experimentally. Importantly, the in silico human models provide a multiscale framework to investigate with high spatiotemporal resolution key ionic mechanisms in drug safety and efficacy in human, with high degree of flexibility in the possible interventions (such as heart rate, concentrations, and adrenergic challenge) compared to experiments. Prof. Matt Daniels reviewed the status of human cardiomyocytes derived from induced pluripotent stem cells (hiPSC-CMs). He saw three uses for these cells: †screening for cardiotoxicity of novel therapeutics, †modelling and drug discovery for inherited cardiac conditions and †in the medium to long term, regenerative cell therapy. While there is progress in scaling up production of these cells, 56 functional maturity is still a major impediment. Stem cell derivatives in many ways (force generation, gene expression, sarcomeric maturity, electrophysiological properties, etc.) resemble immature neonatal cardiac substrates. 57 However, there is one key difference that is rather unique to these cells and best illustrated by comparison to the existing alternatives. Adult cardiomyocytes are consistently mature; neonatal cardiomyocytes are consistently immature. By contrast, stem cell derivatives display an inconsistent immaturity (Figure 4) such that in any experimental test to date, variability within the sample is typically one order of magnitude of measures such as APD and cycle length as shown in reference 58. This will complicate comparisons between samples, which are typically made on small numbers of cells. Dr Stefano Severi showed how a computational approach can be supportive and complementary to the functional in vitro study of hiPSC-CMs. In exploiting hiPSC-CMs as in vitro models for the electrophysiological effects of evolving or new drugs, a detailed understanding of the electrophysiological properties of hiPSC-CMs is necessary. In silico models of hiPSC-CMs action potential have been constructed for control 60 and some genetic mutations, such as those causing long QT (LQT) syndrome type 1 (LQT1), 61 type 2, and type 3, 62 based on recent electrophysiological measurements 63 –66 and validated against drug administration. Computer simulations showed that in principle hiPSC-CMs are qualitatively consistent with adult CMs in response to many current blockers, 67 but differences also emerged. Moreover, hiPSC-CMs show a highly variable electrophysiological behaviour, namely a variable and depolarized resting potential and diverse rates of spontaneous action 50 0 –50 V (mV) INa (A/F) ICaL (A/F) IK1 IKr IKs INaCa INaK Ito (A/F) I (A/F) I (A/F) –100 –100 –200 –300 –1 –2 –0.2 1 0.5 0 1 0.5 0.2 0 0 0 0 0 0.5 1 0 0.5 1 0 0.5 1 0 0.5 1 0 0.5 1 0 0.5 Time (s) 1 Figure 3 Simulation of the human ventricular action potential and the underlying ionic currents. From top to bottom, time course of the action potential, sodium current (I Na ), L-type calcium current, the rapid and slow component of the delayed rectifying current (I Kr ,I Ks ) and the inward rectifying current (I K1 ), the transient outward current (I to ) and the sodium potassium pump (I NaK ) and the sodium calcium exchanger (I NaCa ). B. Rodriguez et al.1292 Downloaded from https://academic.oup.com/europace/article/18/9/1287/2802553 by Universidad de Zaragoza user on 26 February 2021 potentials. Such high variability can be perceived as a limiting factor to the application of the computational approach to hiPSC-CMs. Indeed, computational models of cell electrophysiology are usually developed on the basis of the average values of quantities measured in in vitro experiments. The underlying idea is that the predictions obtained with the model of the ‘average cell’ can apply to all the cells with a tolerance that is of the same order of the variability of the data on which the model is based. Therefore, high variability in the data results in predictions with lower reliability. One way to overcome this limitation is to include the variability within the model itself (which is no longer a model of the ‘average cell’) in order to take it into account both in the investigation of physiological mechanisms and in model-based predictions. The aforementioned population of models approach is a relevant and promising example. Furthermore, when experimental data show high variability, computational models can help to identify the causes of such heterogeneity and potentially help to reduce it. As a relevant example, computational analysis can be used to assess to what extent the variability is due to (i) different experimental systems (e.g. cell lines) that could then be described by different, specific, computational models (ii) differences in the expression of ionic channels from cell to cell, which is eventually much larger than in adult cardiomyocytes or (iii) lack of robustness in the electrical activity of incompletely mature hiPSC-CMs. An example of the latter case are the effects of small changes in depolarizing currents, which can lead to dramatic changes in the rate of spontaneous beating in hiPSC-CMs, whereas they lead to only minor changes in resting potential in adult cardiomyocytes. In this sense, the variability can be reduced if observed in the model parameter’s space since cells showing very different APs could be quite close in terms of their ionic current maximal conductances. Dr Ivan Kopljar continued the theme looking at experimental studies on hiPS-CMs in safety assessments in the drug development setting, with an emphasis on the investigation of long-term drug Variable expression MERGEActininTroponin I Unorganized Rudimentary Organized Variable sarcomere maturation Figure 4 Stem-cell-derived cardiomyocytes have variable phenotypes: current methods of stem cell differentiation produce mixed populations at two distinct levels—gene expression, and post-transcription. This is demonstrated for sarcomeric morphology here, with the panel on the left showing two cells in the field of view positive for the z-disc marker alpha-actinin (white), and the thin filament protein troponin I (magenta). However, further heterogeneity exists even within the cell populations expressing both markers, as only some cells demonstrate ordered sarcomeric units with clear cell polarity (panel on the right). Methods to eliminate (or compensate for) this will be needed to enable small differences between samples to be identified above the noise of the difference within the sample. Scale bar 10 mm. The human ES line OXF2 was grown to confluency on Matrigel and differentiated as described in reference 59. Cells were dissociated by incubation with trypsin/EDTA (0.05%, Lifetech) for 15 min at room temperature prior to seeding onto 0.1% gelatin coated glass coverslips. Ten days after seeding, cells were fixed in 4% PFA (10 min, room temperature), permeabilized (0.1% Triton X-100 in Tris-buffered saline), and blocked with 2% BSA plus 0.001% sodium azide in TBS-T (1 h RT) and incubated with Primary antibodies (mouse monoclonal anti alpha-actinin, (sigma), and rabbit polyclonal anti-troponin T, prior to washing and incubation with Fab fragment anti-mouse 488, and anti-rabbit 568 (molecular probes). Images were acquired on an upright Leica SP5 confocal with a63×lens. Human-based approaches to pharmacology and cardiology 1293 Downloaded from https://academic.oup.com/europace/article/18/9/1287/2802553 by Universidad de Zaragoza user on 26 February 2021 effects. Currently, various technologies such as multi-electrode array (MEA), Ca 2+ transient, impedance, and optical action potential measurements are applied on hiPSC-CMs for their characterization. Their potential has been highlighted by the CiPA initiative launched by the FDA. On the other hand, drug-induced delayed and chronic cardiotoxicity is one of the main risk for drug withdrawal from the market. Therefore, Dr Kopljar described investigations of the acute and delayed (5 days) effects of various oncological compounds on hiPS-CMs using an impedance-based functional assay. Different functional parameters such as beat rate, cell index and incidence of arrhythmia-like events were evaluated, and indicate that the hiPS-CMs can be used to detect different levels of acute and chronic cardiotoxicity and could be valuable in drug safety. In another industrial perspective, Dr Najah Abi-Gerges discussed currently employed strategies in the pharmaceutical industry for cardiotoxicity screening. He gave an overview of the issue pointing out the high attrition rate of new molecules during the iterative drug discovery process: a significant number of these are caused by cardiotoxicity. 68 He highlighted the extent to which some safe new drugs are not developed because of these potential concerns. The new CiPA initiative attempts to address this issue. 6 CiPA proposes a paradigm based on screening new molecules against specific cardiac ion channels combined with integrative computer modelling to predict the proarrhythmic potential of new drugs. This is subsequently combined with non-rodent and early human clinical studies to assess drug effects on QTc measurements. Dr Abi-Gerges highlighted the need for further developing robust and predictive in silico models that represent native myocyte physiology and the heart of healthy volunteers and patients. Such models will predict acute and chronic drug effects with high predictive value on ECG abnormalities other than QT/proarrhythmia, heart rate, contractility, blood pressure, and cardiac structure. Dr Abi-Gerges concluded by advocating that scientists, modellers, regulators, and pharmaceutical industry are currently well positioned to shape how future cardiac modelling would positively impact drug development. Arrhythmia mechanisms and biomarkers A fourth session presented the use of experimental and computational methods for investigations into arrhythmia mechanisms and biomarkers. Dr Stephen Harmer continued the theme of iPSC technologies, in this case for investigations on mechanisms of disease pathogenesis in LQT1 using iPSC technology to model the hereditary cardiac arrhythmia syndrome. Both heterologous (HEK293/ CHO-K1 cell expression systems) and iPSC-based cellular models are used to investigate the underlying disease mechanisms, and a comparison is conducted to evaluate whether the disease mechanisms are similar in both cell types. Results show differences on I Ks channel function and trafficking in heterologous systems for five LQT1 patient mutations. The results are being modelled computationally to investigate the implications of I Ks channel mutations on action potential duration and morphology, and the penetrance of the mutation in human cell populations such as those described in the previous session. Dr Sara Dutta presented a state-of-the-art multiscale human whole heart computational framework to investigate arrhythmic effects of drugs and disease, and specifically hERG block in acute myocardial ischaemia. The framework presented here consists of a human anatomically based model with biophysically detailed representation of membrane kinetics including ionic current and concentration dynamics, as well as fibre orientation and ventricular heterogeneity. The human multiscale model is parameterized using experimental data, from voltage clamp at the ionic level to MRI scans at the whole heart level. They provide a multiscale platform to dissect and analyse specific ischaemic and arrhythmic processes with high spatiotemporal resolution, which is not possible to obtain through experiments alone, especially in human. The new insights provided by the human model could help in the design of further experimental and clinical studies to improve patient risk stratification, as well as decisions about drug dose during management of antiarrhythmic therapy. This study 69 can be extended to explore the mechanisms under different conditions, such as varying sizes and locations of the ischemic region, different drug compounds and their multichannel effects, and inter-subject variability in ionic currents and repolarization patterns. The presentation therefore highlighted the power of human multiscale simulations using anatomically based heart models to investigate safety and efficacy of pharmacological action in lethal disease conditions. Using a different computational approach, Dr Julien Oster presented advances on the computational detection of arrhythmia episodes in ECGs. The development of mobile technologies (mobile phones, tablets, etc.) for health services (mHealth) is currently rapidly growing for two main reasons: (i) cost reduction and (ii) access to resource-scarce communities. 70 The importance for the automatic or semi-automatic detection of arrhythmias on ECG recordings was highlighted in the presentation, given the simplicity of data acquisitions and therefore the multiplication of such data. Manual expert analysis would be more a burden for the clinicians than help for the diagnosis of cardiovascular diseases. Machine-learning approaches have been demonstrated to be a powerful tool for such an analysis for several applications, such as atrial fibrillation (AF) episodes or ventricular ectopic beat detection. Many Holter softwares already require electrophysiologists or laboratory technicians to annotate beats clustered together automatically. Llamedo and Martinez 71 recently suggested a technique requesting expert labelling of the clusters, outputted based on both morphologicaland rhythm-based features. Other techniques for rhythm classification or AF episodes detection, based uniquely on the heart rhythm, have also been presented, and are implemented in implantable devices. 72 Dr Oster’s presentation focused on two major arrhythmic types: AF and premature ventricular contraction (PVC). A machinelearning approach has recently been proposed, through the implementation of a Support Vector Machine 73 on a mobile phone. 74 PVC is another type of ventricular arrhythmia, identified as a predictor for mortality after myocardial ischaemia. 75 The application of an ECG morphology model-based Bayesian filtering 76,77 was shown to be effective for PVC detection. 78 Ms Xin Zhou presented computational investigations into the ionic mechanisms underlying repolarization alternans in a population of human ventricular models calibrated using in vivo recordings such as those presented by Prof. Taggart. Previous research into cardiac alternans has mainly been carried out in animals rather than in human. Ms Zhou presented the construction and calibration of a population of over 2000 human ventricular cell electrophysiology models B. Rodriguez et al.1294 Downloaded from https://academic.oup.com/europace/article/18/9/1287/2802553 by Universidad de Zaragoza user on 26 February 2021 to mimic the action potential variability exhibited in in vivo electrophysiological recordings from 41 patients. 79 The in silico human ventricular cell population is illustrated in Figure 5and it was shown to reproduce two types of alternans restitution curves also observed in human in vivo recordings. By analysing the population of human in silico models, she dissected the mechanisms underlying cardiac alternans and how the complex interaction between sarcolemmal currents and calcium dynamics contribute to the initiation and maintenance of each alternans type. Therefore, in this presentation, multiscale human in silico models were used to capture key repolarization properties of in vivo recordings, to investigate the ionic mechanisms underlying the occurrence of proarrhythmic repolarization alternans and to identify potential anti-arrhythmic targets. Dr Hua Rong Lu described a new, non-invasive and translational biomarker—the index of cardiac electrophysiological balance (iCEB, the ratio between QT and QRS)–in drug and ischaemiainduced cardiac arrhythmias and in genetic LQT syndrome and Brugada syndrome. Currently used biomarkers may not be adequate to detect all types of drug-induced cardiac arrhythmias. Furthermore, there is also a need for a new biomarker to detect cardiac risks in patients with gene-defects in the heart such as LQT syndrome and Brugada syndrome, in patients with heart failure and to detect cardiac risks in sportsmen and women. iCEB was successfully developed and introduced in 2013, 80 it may detect potential risks for drug-induced cardiac arrhythmias beyond LQT and Torsade de Pointes. It may also a be better than currently used biomarkers derived from animal models, such as the isolated rabbit leftventricular wedge model, because iCEB is also detecting additional drug-induced potential cardiac arrhythmias by slowing conduction and QT-shortening. 80 iCEB was also found to be significantly increased in patients in genetic LQT syndrome and significantly decreased in patients with Brugada syndrome. New perspectives/beyond the human heart Dr Jean-Pierre Valentin introduced the background of compound testing strategies in response to ICH guidelines and how the CiPA initiative 6,81 –83 will develop a new non-clinical paradigm for cardiac safety evaluation of new drugs by shifting the focus away from QT prolongation to an assessment of proarrhythmia to mitigate against the thorough QT study. The proposals would be to consider using ion channel effects in tandem with utilizing the emerging technology of in silico assessment and stem-cell-derived cardiomyocyte effects, but not at the expense of clinical ECGs or an understanding of pharmacokinetics/pharmacodynamics. 6 The initiative aims to deliver an implementation of recommendations initially by mid-2016 onwards; however, this is dependent on community support in understanding more about which experimental inputs, which models and which levels of predictive capacity are required. 6 Finally, Valentin reminded the audience of a need to keep QT in perspective and to consider cardiac effectsbeyondQTasitonlyaccountsfor4%ofcardiacsafety-related drug attrition. 84,85 Among the adverse events that are observed are arrhythmia, tachycardia and changes in blood pressure. 84,85 Dr Andre Ghetti set out the problem facing the development of pharmaceutical drugs in the translation between in vitro and animal studies to the clinical setting and the critical need to have stronger models at the non-clinical phase for improving success and understanding of drug action. The approach that Dr Ghetti’s company AnaBios will be taking is to use primary human tissue to derive cells that can be used more reliably in predicting later clinical effects than in vitro or animal studies can alone. This work is potentially to be used in parallel to the development of in silico approaches that are better informed (parameterized) by the data being generated from these isolated human primary cells. Taken together with a much improved characterization of donors, including medical history, allows us to consider the concept of incorporating real-world-type data into models. Dr Rebecca Burton presented her recent work on developing a cell culture model of neurally mediated arrhythmogenesis and non-invasive optical imaging methods being pursued at Oxford. 86,87 Biological models with varying degrees of complexity have been developed to shed light on re-entrant arrhythmias and cardiac monolayers are one of the simplest models. The next stages of this research is to pursue remote monitoring of in vitro cell cultures that would increase experimental access, reduce the need to sacrifice additional animals, and spur the adoption by other laboratories working in allied research areas. There has been a proliferation of remote access platforms that offer promising functionality. Remote access imaging offers advantages to both ‘wet-lab’ and ‘dry’ experimentation (computational and mathematical modelling), and there is a need for platforms which offer secure and reliable implementation. Key challenges moving forward In order to succeed in a programme of research and implementation for drug discovery and testing that takes full advantage of the state of the art in cardiovascular science, the workshop discussions identified five key challenges to be met (1) Each of the human-based methodologies and techniques presented at the workshop has advantages and disadvantages, as 80 Accepted models Rejected models Original O’Hara–Rudy model 40 0 –40 Membrane voltage (mV) –80 0 600 Time (ms) Figure 5 Population of human ventricular action potential models calibrated using in vivo electrogram recordings. Each simulated action potential in the population is generated using the O’Hara– Rudy model with ionic conductances sampled in a wide range of possible values. Calibration is then conducted using the in vivo electrograms by selecting the models that yield action potentials with properties such as action potential duration consistent with the electrograms (red traces, accepted models), and rejecting those that are outside range. In this figure, the action potential for the original O’Hara–Rudy model is shown in black. Human-based approaches to pharmacology and cardiology 1295 Downloaded from https://academic.oup.com/europace/article/18/9/1287/2802553 by Universidad de Zaragoza user on 26 February 2021