Authors: Diogo Teles (1Department of Biomedical Engineering, Columbia University, New York, NY 10027, USA.), Barry M. Fine (2Department of Medicine, Columbia University Irving Medical Center, New York, NY 10032, USA.)
Categories: Article, Arrythmia, Cardiomyocytes, Disease modeling, Drug development, Human induced pluripotent stem cells
Source: Expert opinion on drug discovery
Authors: Diogo Teles, Barry M. Fine
Arrhythmias are disturbances in the normal rhythm of the heart and account for significant cardiovascular morbidity and mortality worldwide. Historically, preclinical research has been anchored in animal models, though physiological differences between these models and humans have limited their clinical translation. The discovery of human induced pluripotent stem cells (iPSC) and subsequent differentiation into cardiomyocyte has led to the development of new in vitro models of arrhythmias with the hope of a new pathway for both exploration of pathogenic variants and novel therapeutic discovery.
The authors describe the latest two-dimensional in vitro models of arrhythmias, several examples of the use of these models in drug development, and the role of gene editing when modeling diseases. They conclude by discussing the use of three-dimensional models in the study of arrythmias and the integration of computational technologies and machine learning with experimental technologies.
Human iPSC-derived cardiomyocytes models have significant potential to augment disease modeling, drug discovery, and toxicity studies in preclinical development. While there is initial success with modeling arrhythmias, the field is still in its nascency and requires advances in maturation, cellular diversity, and readouts to emulate arrhythmias more accurately.
Proper mechanical function of the heart is dependent on the organized and coordinated electrical depolarization of atrial and ventricular muscle to move blood directionally forward into the pulmonary and systemic beds. Arrythmias are abnormal electrical depolarizations that can originate from any structure within the heart and lead to a loss of organized chamber depolarization such that mechanical function of the heart is impacted. These abnormalities can be triggered by a vast array of factors including fibrosis, ischemia, structural disease, and inherited mutations in ion channels. Arrythmias are a large clinical burden and can lead to sudden cardiac death, thrombosis, and heart failure. These disorders account for significant morbidity and mortality and carry a large financial burden to healthcare systems.
Model systems are key tools in cardiovascular research for investigating mechanisms of disease, genotype-phenotype relationships, and therapeutic development. Because of a historical lack of true cardiomyocyte culture systems, heterologous expression systems in cell lines like HEK293 cells have been successful not only in elucidating ion channel physiology but also drug mechanisms of action [1–3]. Genetically modified animal models, particularly the mouse, have been primarily utilized as a surrogate model system for both investigating disease pathogenesis and exploring new therapeutic agents. However, there are important and clinically relevant phenotypic differences between hearts from small animals and those from humans [4]. Ion channel currents, arrhythmia mechanisms, and responses to cardioactive drugs are species dependent, and extrapolating responses in one animal to a different species can prove to be an inaccurate assumption [5–9]. These differences are extremely relevant as we study mechanisms of disease pathophysiology and test the efficacy and safety of new therapeutic agents. There is a need for a physiologically relevant human model which accurately mimics in vitro the complexity of human pathophysiology and recapitulates the effects of known therapeutic agents to aid in the discovery of new cardioactive drugs.
The teams led by Shinya Yamanaka and James A. Thomson revolutionized the field of stem cell technology in 2007 with the generation of human induced pluripotent stem cells (iPSC) [10,11]. The ability to reprogram mature cells collected from peripheral blood or a skin biopsy into iPSC with the transduction of several transcription factors provided the field of cardiovascular research with a renewable resource for the in vitro study of human diseases. These cells have similar characteristics to embryonic stem cells (ESC), especially their potential to be differentiated into the three primary germ layers and respective derivative cells.
With exposure to the correct stimuli, these cells demonstrated the ability to be differentiated into functional cardiomyocytes [12,13]. Some of this work initially had been pursued with embryonic stem cells, which displayed the ability to form cardiomyocytes spontaneously in embryoid body aggregates stimulated with FBS [14]. Methods have expanded significantly in terms of complexity by incorporating a variety of mesodermal morphogens to direct differentiation in both two and three dimensions. Stimulation of specific pathways such as MAPK and WNT led eventually led to chemically defined media in which iPSC differentiation could predictably proceed through mesodermal development to generate cardiac progenitors and then cardiomyocytes [15–17]. For a overview of this evolution and all current methods for cardiac differentiation, we guide readers to this review [18]. Cardiomyocytes differentiated from iPSC resembled in vivo cardiomyocytes in terms of morphology, gene expression, and contraction behavior. However, there are significant differences between in vitro and in vivo cardiomyocytes, including mature gene expression, sarcomeric organization, and automaticity. One of the immediate advantages of iPSC-derived cardiomyocytes is their genetic recapitulation of the patient from whom they were derived, and thus their potential to model inherited disease states. With the simultaneous improvements in gene editing, specific genetic variations could also be introduced in apparently healthy wild-type cell lines or repaired in cells obtained from donors with genetic variants to create isogenic matched pairs (Figure 1) [19–21].
Two-dimensional (2D) culture is straightforward for investigating molecular mechanisms, signal transduction pathways, and high-throughput drug screening. However, 2D cardiomyocyte systems alone lack the maturity and physiological complexity of the heart as an electrical-mechanical organ [22]. Tissue engineering has made considerable advances in three-dimensional (3D) systems through the inclusion of scaffold biomaterials, integration of different stromal populations, and application of external stimuli (electrical, mechanical, or biochemical) [23]. These developments have led to the generation of more nuanced and physiologically relevant models to study tissue function and mimic disease [24–27]. IPSC technology combined with tissue engineering now allows for establishing multiple isogenic tissue types from the same cell line which can be connected to emulate organ systems. This has proved particularly advantageous for studying drug metabolism and potential off-target toxicities [28–31].
In this review, we describe the use of iPSC-derived cardiomyocytes in the study of human arrhythmias and their application in drug development. We discuss several in vitro arrhythmia disease models, highlight examples of drug discovery and gene editing when modeling diseases, and conclude with our opinion on their current role and future applications.
Shortly after the establishment of protocols for the differentiation of human cardiomyocytes from iPSC, somatic cells donated from patients with arrhythmias were reprogramed into iPSC and differentiated into cardiomyocytes to study if these cells still exhibited measurable cellular traits that mirrored electrophysiologic properties of clinical arrhythmias. These early studies of human arrhythmias in vitro using patient iPSC-derived cardiomyocytes confirmed the utility of this model system in studying a cardiovascular disease phenotype in vitro. Inherited arrhythmias are rare diseases, and the majority of disease-causing variants are near unique in terms of their individual prevalence. Thus, a critical question asked by both iPSC-based models and in vivo models more broadly, is whether using a patient-derived variant with inherited arrhythmia can generate physiologically relevant phenotype in vitro that more broadly represents the disease in question. This is a critical scientific assumption that is made frequently in animal models of disease and is a critical benchmark for iPS models broadly in order to demonstrate their utility for clinically relevant therapeutic discovery. Below, we summarize several in vitro models of arrhythmias (important details can be found in Table 1), discuss their applications, and examine how they have been used successfully to further expand our knowledge of these disorders by demonstrating key electrophysiological qualities linked to disease pathogenesis.
There are several subtypes of long QT (LQT) syndrome, each associated with variants in specific genes encoding ion channels or channel-associated proteins. These hereditary channelopathies are characterized by repolarization abnormalities, prolongation of the QT interval on the electrocardiogram, and the propensity to develop life-threatening cardiac arrhythmias and sudden cardiac death [32]. LQT syndrome 1 (LQT1) is a disorder of cardiac repolarization due to an abnormal potassium current. It is the most common LQT syndrome and is caused by mutations in KCNQ1, a gene that encodes a voltage-gated potassium channel [32,33]. In a seminal paper, somatic cells from a family affected by LQT1 were obtained, reprogrammed into iPSC, and differentiated into cardiomyocytes, which successfully demonstrated several canonical features of prolonged repolarization [34].
Several groups have generated patient-specific iPSC from members of families affected by LQT1 and recapitulated in vitro cardiac arrhythmias and drug responses as observed in patients. Cardiomyocytes with these mutations displayed prolonged action potential duration (APD) and corrected field potential duration (FPD), an analog to the QT interval on an electrocardiogram [34,35]. The slow delayed rectifier potassium current (IKs) was reduced with decreased peak and tail current densities as well as altered channel activation and deactivation properties [34,35]. Patients with LQT1 have an increased susceptibility to catecholamine-induced tachyarrhythmia, which was emulated in vitro when cardiomyocytes were exposed to the β1-adrenergic agonist isoproterenol. Conversely, this effect was successfully attenuated with β-adrenergic blockade [34,35]. Compounds with known ion channel activity demonstrated the utility of human iPSC-derived cardiomyocytes to study their impact on arrhythmias. ML277, an activator of KCNQ1, rescued the disease phenotype by increasing the amplitude and enhancing the activation of IKs [36]. Nifedipine, a calcium channel blocker, and pinacidil, an activator of ATP-sensitive potassium channels, also improved the phenotype of these cells through shortening the APD and calcium transient duration.
The use of iPSC-derived cardiomyocytes also allows for mechanistic studies of how mutations lead to disease beyond mere loss of function. For example, KCNQ1, which is normally found at the cell membrane as an ion channel, was instead observed in the cytoplasm and perinuclear areas of cells derived from several donors with several in-frame and missense heterozygous mutations [35,36]. Similar to defects observed in cystic fibrosis for chloride channels, these observations suggest a potassium channel trafficking abnormality plays a role in the pathogenesis of LQT1 due to this mutation and have opened the door to potential therapies that correct ion channel trafficking for patients with these mutations.
LQT syndrome 2 (LQT2) involves mutations in KCNH2, which encodes a pore-forming subunit of the voltage-gated inward rectifying potassium channel. This subunit mediates the rapidly activating component of the delayed rectifying potassium current (IKr) [37]. Several iPSC lines have been generated from patients and the resulting differentiated cardiomyocytes emulated several arrhythmogenic features in vitro [38–41]. These cardiomyocytes displayed reduced IKr, prolonged APD and FPD, early after depolarizations (EAD or abnormal depolarization during phase 2 or 3 repolarization) and delayed after depolarizations (DAD or abnormal depolarizations during the phase 4 repolarization).
These electrical perturbations were then used to explore drug effects in vitro. EADs were exacerbated by β-adrenergic stimulation with isoproterenol, which could be reversed by β-blockers. The selective hERG blocker E-4031, as well as cisapride, prolonged APD and FDP and increasing arrhythmogenesis, while nifedipine and pinacidil led to significant APD and FPD attenutation and eliminated all arrhythmogenic events [38,39,41]. Nicorandil and PD-118057, experimental potassium channel enhancers, caused shortening of APD and reduced EAD, though this could be reversed by isoproterenol [38,41]. These findings demonstrated the power of human iPSC-derived cardiomyocytes to reconstitute multiple different ion channel and signaling pathways that impact both therapeutic utility and toxicity.
LQT syndrome 3 (LQT3) is associated with gain of function mutations in the voltage gated sodium channel SCN5A [32]. These mutations lead to delayed sodium channel inactivation and consequently delayed repolarization of cardiomyocytes. The phenotype of several mutations observed in patients with LQT3 was successfully modeled in vitro with human iPSC-derived cardiomyocytes [42,43]. These cells exhibited prolonged APD, both when spontaneously beating or electrically paced, and a higher time to peak of sodium current. Exposing these cells to blockers of the sodium channels led to the reduction of APD and late sodium current.
LQT syndrome 8 (LQT8), otherwise known as Timothy Syndrome, is a disease caused by mutations in the CACNA1C gene, which encodes a subunit of the voltage-gated calcium channel [44]. Cardiomyocytes from a patient-derived iPSC line demonstrated lower beat frequency, irregular contractions, increased APD, and slow and irregular calcium transients [45]. Roscovitine, a cyclin-dependent kinases (CDK) inhibitor, successfully rescued the calcium mishandling observed in the cells derived from patients [45,46]. To study the molecular mechanism underlying the positive effects of CDK inhibition in LQT8, iPSC-derived cardiomyocytes were used to test several roscovitine analogues and other CDK inhibitors [47]. Four compounds demonstrated beneficial effects increasing the spontaneous beating rate of these cardiomyocytes and decreasing the irregularity of their contractility. Three of these compounds were CDK5 inhibitors, supporting the hypothesis that inhibition of CDK5 had a positive effect on the phenotype of LQT8 and may be a potential therapeutic avenue in patients with this genetic mutation. In further work, the non-opioid intracellular receptor SIGMAR1 was shown to inhibit CDK5 activity. FDA-approved SIGMAR1 agonists fluvoxamine and dextromethorphan rescued the diseased phenotype, reverting the electrophysiological changes observed in these cardiomyocytes [48]. The effective dosages were shown to be safe in isogenic controls, and both in vitro efficacy and safety of dextromethorphan were confirmed in an animal model of LQT8.
Catecholaminergic polymorphic ventricular tachycardia (CPVT) is an arrhythmic disorder that leads to syncope or sudden cardiac death in states of high catecholamine surges such as physical activity or emotional stress [49]. This disease arises from abnormal calcium homeostasis caused by either mostly missense mutations in the ryanodine receptor (RYR2) or recessive mutations in calsequestrin (CASQ). In these situations, when the levels of catecholamines increase in the body, they can trigger abnormal calcium release from intracellular stores leading to the development of ventricular tachycardia.
Several iPSC lines have been derived from patients with CPVT and cardiomyocytes differentiated therefrom displayed several relevant molecular high arrhythmic activity, higher calcium amplitude, and longer durations of spontaneous calcium release events [50,51]. As expected, stimulation with either isoproterenol or adrenaline induced abnormal calcium handling, increased EAD’s and increased DAD’s [52,53]. Studies with different compounds were tested for managing arrhythmic activity in iPSC models of CPVT. KN-93 a Ca2+/calmodulin-dependent protein kinase II (CaMK-II) inhibitor with promising results in animal models, suppressed isoproterenol-induced DAD’s and stabilized calcium activation, while dantrolene was shown to improve RYR2 channel hyperactivity and reduce DAD’s [54,55].
Arrhythmogenic right ventricular cardiomyopathy (ARVC) is characterized by the replacement of normal heart muscle tissue with fatty or fibrous tissues, primarily in the right ventricle [56]. Mutations in desmosomal proteins such as plakophilin-2 (PKP2), desmoglein-2 (DSG2) and desmoplakin (DSP) account for most cases that lead to ARVC. As desmosomes are integral to the structural integrity of the myocardium, these mutations lead to disruption of cell-cell interactions followed by cell death and fibrofatty replacement. Cardiomyocytes differentiated from patient-specific iPSC lines with mutations in PKP2 display several features of ARVC and, interestingly, implicate abnormal metabolism as a contributing mechanism to the pathophysiology of this disease [57]. Several known compounds were applied to these models in order to validate iPSC cardiomyocyte physiological response[58,59]. For example, contraction amplitude was reduced after treatment with the calcium channel blocker nifedipine, while treatment with isoproterenol increased beat frequency. Caffeine, a potentiator of cAMP-mediated signal transduction pathways, evoked an instantaneous, rapid, and large release of calcium indicating that PKA modulation of calcium flux was intact in these cells.
iPSC obtained from donors with pathogenic variants in other genes associated with ARVC (DSP and DSG2) have demonstrated arrhythmogenicity and are potential models to identify novel therapeutics for treating arrhythmias in patients with these mutations [60,61]. As an example, cardiomyocytes derived from a patient with ARVC secondary to a mutation in the DSG2 gene displayed higher pacemaker activity, more arrhythmic events, and highly expressed NDPK-B and SK4 channels [62]. Treatment with PHP-1, an NDPK-B antagonist, suppressed cell beating, reduced SK4 channel activity, and diminished arrhythmic events, indicating a mechanistic role of NDPK-B and SK4 channels in the pathogenesis of ARVC. Future studies with other genotypes have similar potential to promote this precision medicine approach for mechanistic studies followed by therapeutic development tailored to specific genotypes.
The discovery of precise and facile gene editing techniques as zinc finger nuclease (ZFN) [63,64], transcription activator-like effector nuclease (TALEN) [65–68], and clustered regularly interspaced short palindromic repeats (CRISPR) [69–73] has accelerated iPSC disease modeling through both the introduction of variants in wild-type cell lines, as well as the correction of a mutation in a patient-derived cell line. These isogenic pairs are critical for studying genotype-phenotype relationships and making comparisons in readouts that can be potentially impacted by other genomic variations between cell lines. Isogenic controls for several diseases like LQT1 [74], LQT2 [75,74], LQT8 [47,48], and ARVC [57] have been generated using gene editing. By using these cells as controls for functional, electrophysiological, and calcium handling studies, the influence of the mutation in their phenotype can be investigated, confirming their validity for mechanistic and drug testing studies in the future.
Despite the in vitro nature of iPSC models, there have been reports that the degree of phenotypes observed in patient-derived cell lines can correlate to the severity of the disease of that patient. This has potential utility in classifying pathogenicity of variants and personalized drug response as an overall effort in precision medicine approaches. As an example, cells derived from LQT2 patients with milder phenotypes presented less frequent EAD and lower susceptibility to drugs known to increase arrhythmogenicity [40]. In another example, iPSC-derived cardiomyocytes were able to disentangle multiple mutations in a single patient with variants in both KCNH2 and SCN5A, genes associated with LQT2 and 3, respectively [76]. In vitro studies demonstrated arrhythmias due to aberrant sodium flux, but not influenced by the potassium flux from the polymorphism observed in the KCNH2 gene. These cardiomyocytes displayed an increased inhibition of their late sodium current when treated with mexiletine, a sodium channel blocker indicating potential drug response in that patient. These types of studies demonstrate promise for iPSC in both annotating genomic variants as well as in predicting possible therapeutic/toxicity interactions with specific variants.
Other groups have used gene editing techniques with iPSC to evaluate variant pathogenicity by interrogating serial mutations in a wild-type line [77,78]. Several mutations in KCNH2 were introduced into a single wild-type iPSC line, and mechanistic studies revealed that mutation proximity to the pore-forming domains was directly related to higher susceptibility to drug-induced arrhythmias [77]. These types of experiments demonstrate how this approach can reveal mechanistic insight into genotype-phenotype relationships, leading to enhanced risk stratification for patients.
These technologies also offer the potential to explore gene therapy modalities that can either suppress, correct, or replace mutant alleles with the ultimate goal of curative therapy. Animal alleles contain polymorphisms and critical sequence differences that can interfere with gene targeting beyond proof of principle studies. Approaches that specifically target human genes can ultimately require genomic humanization in animals to demonstrate safety and efficacy in preclinical studies. Human iPSC-derived cardiomyocytes, however, provide a model system in which the human gene therapy approaches can be safely optimized and validated in vitro. Several groups have tested this approach by delivering suppression and replacement gene therapy to rescue cells derived from donors with LQT syndromes [79,80]. For example, a short hairpin RNA (shRNA) (KCNQ1 for LQT1 or KCNH2 for LQT2) was used to suppress both alleles of the gene of interest, followed by gene replacement expression of an shRNA-resistant wild-type allele. Human iPSC-derived cardiomyocytes from several donor cell lines with genetic mutations were successfully rescued with the shortening of their APD. This approach is mutation agnostic and can be implemented in a range of pathogenic variants of different genetic diseases. Previous approval of similar genetic replacement therapies by the US Food and Drug Administration (FDA) in cases of retinal dystrophy and spinal muscular atrophy support this promising strategy to potentially cure genetic arrhythmogenic diseases by acting directly on the genetic level [81,82].
Drug toxicity is one of the most common reasons for drug development failure. Prolongation of the QT interval and inhibition of the hERG potassium channel are two major safety concerns as these can lead to a fatal ventricular tachycardia called Torsades de Pointes in humans. However, predicting toxicity in humans using in vitro models has been very difficult. The Comprehensive in vitro Proarrhythmia Assay (CiPA) collaboration, an international team comprised of regulatory officials, non-profits, academics, and drug developers, was assembled to evaluate the proarrhythmic risk of novel therapeutic agents [83]. This collaboration is focused on improving the prediction by integrating ion current studies, EKG analysis, in silico predictive modeling and, importantly, studies using iPSC-derived cardiomyocyte models. This was predicated on several published studies demonstrating iPSC cardiomyocytes display APD and FPD prolongation as well as changes in ion channel currents in response to known QT-prolonging drugs [84–86]. The goal of this initiative is to identify the risk of malignant arrhythmia prior to human testing in order to avoid costly late-stage drug failure [87], and the incorporation of iPSC-derived cardiomyocytes provides a scalable tool to quickly screen for new potential toxicities early in hit-to-lead campaigns (Figure 2).
Beyond assessing for potential arrhythmia toxicity, iPSC-based assays can help identify therapeutic potential in compounds using in vitro readouts that have some physiological relevance to the studied disease state. For example, iPSC-derived cardiomyocytes derived from a patient with CPVT were treated with several drugs in clinically relevant concentrations, and these findings were compared with the same drugs’ response in the same patient [88]. Treatment with the sodium channel blocker flecainide reduced exercise-induced ventricular tachycardia in the patient while simultaneously reducin adrenergic-induced arrhythmia in vitro. While treatment with labetalol had no effect in both settings, propranolol prevented the development of arrhythmias in vitro and successfully predicted decreased exercise-induced VT in the patient clinically.
Lumacaftor is an FDA approved drug that works as a chaperone to rescue trafficking defects in mutant CFTR chloride channels in a subset of patients with cystic fibrosis. This drug was tested in two patients with LQT2 who had mutations that were associated with defective membrane trafficking of hERG [89]. Both patients had their QT interval shortened significantly after one day of therapy, which rebounded after a wash-out recording [90]. Although these studies were only performed in a small number of patients, it is encouraging to observe a positive correlation between in vitro drug effects and clinical observations in patients, supporting the further use of human iPSC-derived cardiomyocytes to screen new or already approved drugs before testing them in large clinical trial settings.
iPSC-derived cardiomyocytes also play a role in the development of new drugs, especially in hit-to-lead campaigns. Novel therapeutic agents can be tested in vitro and can undergo several cycles of chemical refinement based on preclinical data before being tested in patients. Mexiletine is an antiarrhythmic drug used to reduce the QT interval in patients with LQT3, even though the mechanism of its therapeutic properties is not fully understood. Cardiomyocytes were differentiated from cells obtained from patients with LQT3 and were treated with mexiletine analogues [91]. Based on the analogues’ effects on APD shortening, arrhythmic activity, and ion current, the compounds were further engineered for optimized APD shortening and anti-arrhythmic activity. The resulting lead compounds from this drug development approach presented greater on-target potency and selectivity than the original drug. The efficacy of one of the new analogues was further tested in cells with different mutations in the same gene (SCN5A) and in cells which a LQT3-like phenotype generated pharmacologically. The new analogues shortened the APD in all models, while other treatments showed mutation-specific efficacy, and suppressed the emergence of arrythmia in healthy cells.
Machine learning involves the development of algorithms that allow computers to make predictions and decisions based on pattern learning. Tools that have arisen from machine learning have enhanced our ability to decipher high-dimensional and complex data sets in biomedical research. Clinically, this has led to several advances in the interpretation of several areas including electrocardiograms, pathology sections, and radiographic images. This has also proved fruitful in translational research, assisting in the identification of disease phenotypes in vitro, the classification of cardioactive drugs, and predicting drug response [92].
The integration of machine learning algorithms with iPSC models allows for the facile analysis of high-dimensional data to identify phenotypes and drug response in vitro using novel classifiers in unsupervised or supervised manners. Several groups have reported the use of machine learning to classify iPSC-derived cardiomyocytes in healthy or diseased phenotypes based on different parameters of cardiac function [93–96]. Cells obtained from donors with different arrhythmias were successfully classified when comparing cells obtained from healthy donors. Using a noninvasive label-free approach, machine learning algorithms successfully discriminated iPSC-derived cardiomyocytes derived from patients with LQT8 as diseased cells when compared to healthy controls [97].
Machine learning has also been used to identify drug-induced arrhythmias in toxicity studies and to classify cardioactive drugs. Cardiomyocytes treated with drugs with different mechanisms were properly identified by algorithms based on changes in the cardiac functionality [98–100]. With elegant integration of in silico technologies and experimental methodologies, complex platforms can help identify the proarrhythmic risk of new drugs early in the development process [101].
In silico-based drug discovery has been proven particularly impactful, rapidly accelerating drug design campaigns by incorporating computational tools for drug binding optimization through molecular modeling and quantum mechanics [102]. Computational molecular dynamics and docking have been used to generate structure-activity relationships quickly. The integration of structural and ligand-based approaches for virtual screening allows for large lead compound screens with reduced labor costs, improved efficiency, and increased specificity of hits [103]. Moving forward, there is potential for synergy in drug development by combining computer-aided drug design with disease modeling with iPSC to rapidly accelerate validation of hits, identification of promising leads with high throughput functional assays, and early identification of toxicity.
Moving beyond 2D culture systems is crucial for better model systems of the heart that appreciate the complexity of a 3D mechanical organ and can also consist of a large number of non-cardiomyocytes that play a role in disease, as well as a complex extracellular matrix. The development of organs-on-chip technologies have allowed the formation of 3D cardiac tissues that incorporate different stromal cell types, extracellular matrices, and even vasculature for infusing external stimuli [23].
The power of these models is bolstered by the plasticity of iPSC and their ability to generate multiple isogenic cell types. By genetically editing iPSC-derived cardiomyocytes and fibroblasts to present a point mutation in the KCNH2 gene and integrating both cell types in a collagen-based hydrogel, a team of researchers developed a LQT2 model [78]. These 3D tissues formed anisotropic sarcomeres as well as recapitulated key features of LQT2, including prolonged APD and calcium mishandling. Exposure to E-4031, a hERG channel blocker, rescued the phenotype, demonstrating the applicability of these models to test novel therapeutic agents for this arrhythmic disease. Additional models have been successfully generated and emulated the pathophysiology and drug response of LQT1 [104], ARVC [105], and CPVT [106]. In addition to the modeling of human diseases in vitro, 3D cardiac tissues can also been used to study the molecular mechanisms behind arrhythmias, identifying the role of TFAP2A in the fibro-fatty remodeling observed in ARVC [107], and the mechanism behind drug-induced tachyarrhythmias with characteristics similar to Torsade de Pointes [108].
There has been significant progress in the development of human iPSC-derived cardiomyocyte-based models of arrhythmias. Genetic studies through either derivation from patients with pathogenic variants or engineering mutations in healthy cell lines have captured several important features underlying arrhythmias. These have led to mechanistic insights into disease pathogenesis and clearly demonstrated clinically relevant drug responses. Advances in 3D tissue engineering have improved model fidelity and the application of machine learning has shown significant promise for high-dimensional data analysis and the discovery of new classifiers. All of this has set the stage for iPSC-based systems to provide both efficacy and safety data in preclinical drug development. This is timely as the FDA Modernization Act 2.0, passed in 2022, allows FDA regulators to consider preclinical data obtained from sources other than animal models when evaluating new therapeutic agents, empowering in silico computational-based models, artificial intelligence, and human iPSC culture systems. By integrating these models into the drug development pipeline, new leads can be studied in potentially more physiologically relevant models. These types of studies may give us more reliable safety and toxicity information before moving forward into clinical trials. With access to iPSC obtained from donors, in vitro clinical trials can be established to study the pharmacokinetic and pharmacodynamic profiles of new therapeutic agents and identify and predict risk for drug-induced toxicity. This will hopefully improve hit-to-lead/lead optimization campaigns and result in more effective and safe treatments for patients with arrhythmogenic disorders.
One of the biggest impediments in cardiovascular research has been a lack of available human disease models. This stems in part from both a paucity of available tissue and an inability to sustain or easily manipulate primary culture systems from human heart tissue. This has led to a near exclusive reliance on rodent models of heart disease throughout the literature. These models have been bulwarked by genomic studies that have allowed us to validate human genetic drivers of disease as well as larger animal studies such as those in the pig, which more closely resembles human physiology. However, the rodent model has remained a poor predictor of therapeutic translation and is a potential culprit of many early human trial failures for therapies that progressed optimistically through preclinical investigational new drug (IND) application-enabling studies. This has presented an opportunity for iPSC to fill in a crucial gap in cardiovascular research as a potential parallel pipeline of both validation as well as discovery. The promise of this system cannot be overstated and the ability of iPSC-derived cardiomyocytes to recapitulate genetic diversity, electromechanical coupling, and calcium handling is encouraging, opening new avenues of research into the pathophysiology and treatment of a variety of cardiomyopathies. However, several key issues remain to be addressed to fully realize the potential of this system to model cardiovascular disease, specifically when it comes to appreciating the pathophysiology of arrhythmias (Figure 3). Though there has been success in applying genetic variant models, iPSC-derived cardiomyocytes will require a significant leap forward to tackle the far more common arrhythmias of the atria and ventricles such as atrial fibrillation and scar-related ventricular tachycardia.
Probably one of the most frequently raised concerns about cardiomyocytes derived from iPSC is their level of maturation. At the end of standard differentiation protocols, cardiomyocytes have been mostly characterized as neonatal [109]. This has significant impact on recapitulating both genetic variants and acquired pathological arrhythmias that depend on adrenergic signaling, mature ion channel constituency, cell to cell communication, and calcium homeostasis. Advances have been made to improve maturation of these cells after differentiation through a variety of methods such as age dependent cultures, specialized media, exposure to specific hormones, altering the metabolic source of energy, co-culture studies, and electromechanical stimulation in 3D culture [26,110–114]. This latter method has proved particularly impressive in terms of gene expression and force generation [26]. Maturity though remains a critical impediment and creating models that more closely resemble adult cardiomyocytes will be needed to boost the fidelity and translational potential of arrhythmia models.
Furthermore, not all cardiomyocytes are the same. There are significant differences between atria and ventricle and between left and right sided structures in terms of their developmental origins, gene expression profiles, electromechanical properties, and ion channel constitution. Arrhythmias can arise in nearly any part of the heart and to model these accurately, we will need to improve our ability to direct differentiation toward specific sublineages. Standard differentiation protocols produce a heterogeneous population that initially displays multiple foci of autologous depolarization with varied action potential morphologies. While there has been significant progress toward both improving discrimination of differentiation between subpopulations of cardiomyocytes and increasing population purity using commonly found expression markers, we are still far from having well-defined labels and in vitro characteristics to properly differentiate all the subpopulations in the native heart [115]. We believe initiatives to develop criteria that different subclasses of in vitro differentiated cardiomyocytes will improve the use of different cell subtypes in disease modeling [116] Whether these can enhance clinical translation significance remains to be tested. However, this improved specification will allow more nuanced models that can both investigate chamber-specific contributions to arrhythmogenesis as well as those in which chamber-specific cardiomyocytes are electrically connected to mimic actual human physiology.
Though cardiomyocytes by volume occupy the bulk of heart tissue, they numerically actually comprise a minority of cells in the heart. The contribution of other cell types to the development of cardiovascular disease is becoming increasingly appreciated and these cells need to be incorporated into model systems. Fibroblasts, for instance, play a key role in the maintenance of structural integrity after myocardial injury through a fibrotic response to the permanent loss of cardiomyocytes. The result, however, can be scarring heterogeneity with islands of viable cardiomyocytes creating short tracks of conduction that lead to scar-mediated ventricular tachycardia. Immune cells have also been recently implicated in exciting research demonstrating that recruited macrophages are key to the development of atrial fibrillation in inflammatory states [117]. Advances in iPSC differentiation now allow for directed differentiation of isogenic fibroblasts and immune cells including macrophages whose incorporation allows for more complex multicellular models. As a recent example of this power, improved atrial fibrillation models in 3D tissue organoids was recently demonstrated through incorporation of isogenic iPSC-derived atrial specific fibroblasts [118].
Modeling organotypic function in a cell or tissue-based culture systems requires a dimensional reduction of physiological phenotypes to laboratory-based readouts which we rationalize have a relationship to the original disease being studied. As an example, cardiomyocyte contractile properties by video microscopy can be correlated to sarcomeric dysfunction from cardiomyopathic iPSC lines. The danger of this approach is that most of these laboratory readouts can be confounded by small changes in differentiation efficiency, batch to batch variation, maturity level, and cell heterogeneity. Biological replicates and the use of several different cell lines can help control for both these phenotypic variations as well as potential unknown genetic interactions in the case of using just one genetic background. However, the fact that we utilize the same quality benchmarks for the generation of iPSC-derived cardiomyocytes and the surrogates for therapeutic response will remain a significant weakness of this system and modeling arrhythmogenesis is no exception. Cellular measurements such as patch clamp, action potentials, and calcium flux using various reporters are all subject to significant variation during the generation of iPSC-derived cardiomyocytes. Consistency and reduction of variability in the manufacturing of iPSC-derived cardiomyocytes are necessary to have confidence that model systems can measure a pathologically relevant change to a readout that is not the result of the variance of iPSC’s differentiation.
Overall, there have been significant progress in modeling arrhythmias using cardiomyocytes differentiated from iPSC. Advances in iPSC methodologies, gene editing, tissue engineering, and in silico technologies portend an optimistic future where these models will become established components of drug development pathways.