|Home | About | Journals | Submit | Contact Us | Français|
As part of the series on Controversies in Cardiovascular Research, the article reviews the strengths and limitations of induced pluripotent stem cell-derived cardiomyocytes (iPSC-CM) as models of cardiac arrhythmias. Specifically, the article attempts to answer the following questions: Which clinical arrhythmias can be modeled by iPSC-CM? How well can iPSC-CM model adult ventricular myocytes? What are the strengths and limitations of published iPSC-CM arrhythmia models? What new mechanistic insight has been gained? What is the evidence that would support using iPSC-CM to personalize anti-arrhythmic drug therapy? The review also discusses the pros and cons of using the iPSC-CM technology for modeling specific genetic arrhythmia disorders such as long QT syndrome, Brugada Syndrome or Catecholaminergic Polymorphic Ventricular Tachycardia.
As part of the series on Controversies in Cardiovascular Research, I have been asked to discuss the limitations of induced pluripotent stem cell-derived cardiomyocytes (iPSC-CM) for modeling cardiac arrhythmias. A daunting task, given that the 2012 Nobel Prize for Physiology or Medicine has been awarded jointly to Shinya Yamanaka and Sir John B. Gurdon for the invention of the iPSC technology, which allows the generation of pluripotent stem cells by reprogramming mature somatic cells harvested from patients.1 This technology theoretically enables an unlimited supply of patient-specific stem cells, which can be differentiated into different cell types for disease modeling, molecular diagnosis, cell-based therapy and testing of personalized treatments.2 Among the first reports of iPSC disease models have been iPSC-CM models of genetic arrhythmia disorders (i.e., long QT syndrome),3 generating substantial enthusiasm for adopting the iPSC technology in the field. Potential applications of iPSC-CM for arrhythmia disorders can be either patient-specific (i.e., personalized medicine) or independent of the individual patient (Table 1). Among the patient-independent applications, drug safety testing, which utilizes iPSC-CM derived from control subjects, has received a lot of attention recently and is frequently mentioned as promising application of the iPSC technology.4-9 However, normal myocytes can also be generated from human embryonic stem cell (ESC), a technology that has been available for over a decade,10 but interestingly has not been widely adopted for drug safety screening. In this review, I will focus primarily on the role of iPSC-CM for patient-specific applications, which are the unique feature of the iPSC technology and theoretically could be used clinically to help diagnosis and personalized treatment for patients with arrhythmia disorders.
To begin assessing the relevance of iPSC-CM models for clinical arrhythmia diseases, let us consider the following hypothetical case: A man collapses during a basketball game. He is resuscitated and brought to the hospital for further evaluation and treatment. After obtaining a skin biopsy in the emergency room, several lines of iPSC derived CM are generated and phenotyped in the lab. A diagnosis of the underlying arrhythmia disorder is made, and drug therapy selected based on efficacy testing using the patient’s iPSC-CM. To address how realistic such a scenario is, I will review the published iPSC-CM arrhythmia models and attempt to answer the following questions: Which clinical arrhythmias can be modeled by iPSC-CM? How well can iPSC-CM model adult ventricular myocytes? What are the strengths and limitations of published iPSC-CM arrhythmia models? What new mechanistic insight has been gained? What is the evidence that would support using iPSC-CM to personalize anti-arrhythmic drug therapy?
Table 2 lists clinical arrhythmia disorders that could be responsible for our hypothetical case presentation of what is commonly referred to as sudden cardiac death (SCD).11 The vast majority of SCD cases (~80%) in adults are due to ventricular tachyarrhythmias caused by myocardial ischemia.11 It seems implausible that iPSC-CMs generated from a skin biopsy of a patient with SCD caused by myocardial ischemia will provide any relevant insight into the underlying arrhythmia mechanism or will help tailor anti-arrhythmic therapy for this patient. Moreover, of the remaining 20% of non-ischemic forms of SCD, most are due to underlying structural heart disease,12 and other acquired conditions that develop of over decades (i.e., atrial fibrillation from valvular heart disease, bradycardia from degenerative disease of the specialized conduction system). Of the subset of arrhythmia with a strong heritable component,13 many are also associated with structural heart disease either in the atria, the specialized conduction system and/or the ventricle, which will be challenging to model in single cells provided by the iPSC-CM technology. Hence, even in the best case scenario, the iPSC-CM technology will only be applicable to very small minority of patients with arrhythmia disorders. Accordingly, 14 of the 15 published iPSC-CM models of arrhythmia disorders have investigated rare genetic arrhythmia syndromes that can cause ventricular tachyarrhythmia in structurally-normal hearts (Table 3).
It would seem that the first step in establishing the utility of a new model system is to compare it with the real thing, which is the freshly-isolated adult cardiomyocyte. Yet almost all published reports of iPSC-CM arrhythmia models have skipped that step. Rather, researchers went straight to comparing the electrophysiologic phenotype of mutant iPSC-CM derived from diseased individuals to that of iPSC-CM derived from healthy individuals. Possibly as a result of this lack of control studies, there is a large range of values for basic electrophysiogic parameters such as action potential duration, resting membrane potential or cell beating rates among the “normal” iPSC-CM in the published literature.14 Further contributing to the wide variability of the “normal” electrophysiology of iPSC-CM is that most studies published to date have used the embroid body technology to differentiate iPSC into CM, which is hampered by low yield of CM (5-10%) exhibiting a relatively immature phenotype. Furthermore, culture conditions vary widely between labs. Significant progress was made with the introduction of the matrix sandwich method of generating CM from pluripotent stem cells.15 Here, yields upward of 95% of cells expressing myocyte markers can be achieved, and the CM appear to have a more mature phenotype. The first comprehensive electrophysiological characterization of CM generated by the matrix sandwich method has been recently been reported by the January lab.16 In this article, the results obtained from iPSC-CM were compared to published values of adult human ventricular myocytes. As summarized in table 4, the iPSC-CM exhibit action potential parameters, resting membrane potential and membrane ionic currents that largely resemble those found in human ventricular myocytes. The main differences are a smaller cell size, reduced IK1 currents and the presence of a prominent If current. As a result, the iPSC-CM exhibit spontaneous membrane depolarizations, which are not observed in healthy human ventricular myocytes. Ca handling and SR function has been investigated to date only in iPSC-CM generated by the embryoid body method.17 Similar to adult human CM, Ca release and contraction of iPSC-CM was strongly dependent on Ca influx via L-type Ca channels. Furthermore, iPSC-CM exhibit intracellular Ca stores that can be released by activating RyR2-Ca release channels with caffeine,17 suggesting that some functional sarcoplasmic reticulum (SR) is present. However, approximately 50% of intracellular Ca stores contributing to the action potential induced Ca transients are sensitive to IP3,17 which is not the case in normal adult ventricular myocytes.18 Small IP3-sensitive Ca stores may contribute to E-C coupling of normal adult atrial myocytes,19 and upregulation of IP3-releasable stores has been reported in failing ventricular myocytes. Hence, the dependence of Ca release on IP3-sensitive Ca stores in iPSC-CM would be an important caveat when investigating disease states that may involve altered regulation of PLC-IP3 signaling. Moreover, CM generated from human embryonic stem cells using this approach often do not even have a functional SR.20
Thus far, no study has investigated in detail the excitation-contraction (E-C) coupling process of iPSC-CM. But given the lack of t-tubules and the underdeveloped and relatively disorganized sarcomeres observed in the iPSC-CM,21, 22 it is likely that these cells have E-C coupling properties that are quite different from those of adult ventricular myocytes. For example, one report indicates that Ca release was slow and spatially inhomogeneous in iPSC-CM, which was attributed in part to the lack of t-tubules and relatively low expression levels of SR Ca handling proteins,23 as also reported in ESC-derived CM.24
Another limitation of current iPSC-CM technology is that the resulting CM exhibit a mix of action potential morphologies, which has led investigators to classifying CM based on their action potential morphology as “atrial-like”, “nodal-like” and “ventricular-like”.3 An obvious disadvantage of this method is that action potentials have to be measured first before any experiments on a particular myocyte type can be conducted. One approach to circumvent this problem is to transfect cells with fluorescent reporter constructs that are cell type specific, so that cells can be purified by flow cytometry before the experiments. Time will tell how well this works. To improve the contractile performance and enhance the structural organization of the cells, 2D or 3D plating approaches have been proposed. An inherent limitation of using cultured CM of any kind is that even acute-isolated terminally-differentiated ventricular myocytes tend to remodel and de-differentiate in culture within a few days,25 a problem that investigators in the field have struggled with for decades, with no good solution in sight. Thus, the goal of generating iPSC-CM that completely resemble acutely-isolated myocytes may be unattainable.
To date, 15 articles describe the use iPSC-CM to model arrhythmia disorders (Table 3). Fourteen of those modeled genetic arrhythmia diseases that occur in structurally normal hearts.26 Of those, 9 have modeled the long QT syndrome caused either by mutations in genes responsible for delayed rectifier K currents (LQT1, LQT2), cardiac Na currents (LQT3) or L-type Ca currents (LQT8). Five articles reported iPSC-CM models of Catecholaminergic Polymorphic Ventricular Tachycardia (CPVT), which is caused either by mutations in the cardiac ryanodine receptor Ca release channel (RyR2), cardiac calsequestrin (Casq2) or triadin.13 Only one study investigated iPSC-CM derived from a patient with an arrhythmia disorder associated with structural heart disease – Arrhythmogenic Right Ventricular Cardiomyopathy (ARVC) – a disease caused by mutations in genes encoding desmosomal proteins.13 The reader is referred to the companion piece to this article, which describes in more detail the major findings of the published iPSC-CM arrhythmia models.
To assess the relative strengths and limitations of the different iPSC-CM arrhythmia models, I have rated each iPSC-CM model using the following score: 0 = iPSC-CM phenotype exhibits significant discrepancy with the published literature. 1 = iPSC-CM reproduces findings from heterologous expression systems or isolated myocytes from mouse models. 2 = iPSC-CM replicates key disease features observed in patients. 3 = validated iPSC-CM model that replicates disease phenotype of adult myocytes isolated from diseased hearts. The results of this analysis are shown in Table 3. Only 2 models received a score of 3, both of which are studies of Na channel mutations. In each study, the investigators validated the model by comparing the results obtained in iPSC-CM to the Na channel phenotype of acutely-isolated adult ventricular myocytes. Due to the lack of human tissue, this was done in ventricular myocytes obtained from mouse models which reproduce key features of the disease in vivo. A score of 2 was given to six models, all of them models of LQT due to either K channel or L-type Ca channel mutation. Here, the clinical patient data of long QT corroborate the finding of action potential prolongation in patient-specific iPSC-CM. Models of CPVT and ARVC received a score of 1, because there are no clinical data available to corroborate the validity of iPSC-CM phenotype. One study describing a Casq2 linked CPVT-iPSC-CM model received a score of 0 because the iPSC-CM phenotype was at odds both with in vitro findings from mouse models and clinical patient data. Next, I will discuss in more detail the latter study, which illustrates important limitations of the iPSC-CM approach.
The study by Novak et al aimed to establish an iPSC-CM model of CPVT caused by the CASQ2-D307H mutation.22 CPVT is a disease caused by mutations in the RyR2 Ca release channel or its protein partners triadin and Casq2.27 Clinically, patients have normal hearts, normal contractility and normal ECG parameters including a normal QT duration.28 Polymorphic or bi-directional ventricular arrhythmia develops with exercise or emotional stress. Based on extensive work in heterologous expression systems and mouse models, the underlying arrhythmia mechanism has been well established.27 The molecular defect in the RyR2 Ca release channel complex renders the SR prone to premature spontaneous Ca release during periods of catecholaminergic stimulation. The resulting spontaneous Ca elevation during diastole activates the electrogenic NaCa exchanger, which in turn depolarizes the cell membrane and generates a delayed afterdepolarization (DAD). DADs of sufficient amplitude then trigger action potentials, which give rise to the typical polymorphic ventricular tachycardia in CPVT. Novak et al indeed find that CPVT-iPSC-CM derived from patients carrying the CASQ2-D307H mutation develop DADs after catecholaminergic stimulation with isoproterenol. But Novak et al also report that CPVT-iPSC-CM exhibit dramatic action potential prolongation (Fig. 1), which is completely at odds with the normal QT interval observed clinically.28 Furthermore, catecholaminergic stimulation causes a paradoxical slowing in the beating rate of CPVT-iPSC-CM, with DADs observed during those bradycardic periods, which is at odds with the ventricular arrhythmia observed clinically only during periods of fast heart rates.28 Moreover, Novak et al find that CPVT-iPSC-CM have an immature ultrastructure with disorganized contractile apparatus and impaired contractility, again at odds with the normal heart morphology and contractile function observed in CPVT mouse models27 and in CPVT patients.28 The reasons for these discrepancies are not clear. One possibility is that the presence of mutant CASQ2-D307H with reduced Ca buffering properties29 alters the maturation of iPSC-CM in culture, analogous to how the level of Casq2 protein expression influences the maturation of ESC derived CM.30 Regardless of the underlying reason, the CASQ2-D307H example illustrates the pitfalls of attempting to use iPSC-CM to define a functional phenotype for diseases of known or unknown genetics. Here, the functional phenotype of the CPVT-iPSC-CM derived from a skin biopsy of a patient carrying the D307H mutation clearly does not represent the phenotype of the patients’ own ventricular myocytes.
Of the 15 reports, 9 studies merely reproduced disease mechanisms that had previously been discovered using other approaches. Only 6 studies proposed new arrhythmia mechanisms (Table 3). Of those, 4 studies used patient-derived iPSC-CM to establish the pathogenicity of new mutations found in patients with LQT1,3, 31, LQT332 and CPVT,33 respectively. Both LQT1 mutations cause a dominant-negative protein trafficking defect.3, 31 Results of iPSC-CM were confirmed using a heterologous expression system, which makes one wonder whether the use of iPSC-CM was warranted in the first place. Terrenoire et al32 used iPSC-CM to diagnose the pathogenicity of a de novo SCN5A LQT3 mutations and rule out the influence of KCNH2 polymorphism also present in the patient. Based on the results of iPSC-CM, Yazawa et al postulated that delayed afterdepolarizations (DADs) contribute to the pathogenesis of ventricular arrhythmia in a LQT8 patient.34 Since iPSC-CM are particularly likely to exhibit DADs due to their high levels on If, weak IK1 and their proclivity to develop spontaneous Ca release from intracellular stores (Table 4), the relevance of these findings to the pathogenesis of human arrhythmia in LQT8 remains unclear. Finally, Kujala et al report that iPSC-CM derived from a CPVT patient exhibit early afterdepolarizations (EAD).35 The authors then go on and confirm the presence of EADs in monophasic action potential recordings in the same CPVT patient. Hence, only one out of 15 iPSC-CM studies appears to have generated novel mechanistic insight that was validated in the index patient.
Using iPSC-CM models to personalize drug therapy has been advertised as one of the major advantages of this technology. Yet based on the published reports, evidence supporting this strategy is essentially lacking. Not counting reports where β-blockers were used to prevent the pro-arrhythmic effects of pharmacological β-adrenergic stimulation with isoproterenol, only 5 out of 15 studies tested drugs using iPSC-CM (Table 3). Yet, despite their beneficial effects in patient-derived iPSC-CM, none of the drugs identified as being effective in iPSC-CM were actually given to the patient (Table 3). Regardless of the reasons the investigators had for not given the drug to their patients, there are several inherent limitations of using iPSC-CM to personalize anti-arrhythmic drug therapy.
Testing investigational compounds is not warranted, because such compounds could never be used without extensive safety testing, which is not an option for treating an individual patient. Furthermore, FDA-approved can be tested for efficacy in the patient without even waiting for results from in vitro testing.
Even if a drug is found to be beneficial in the patient-derived iPSC-CM, it is far from clear if this will translate into a clinical benefit. For example, drugs such as nifedipine identified as being effective in an iPSC-CM model of LQT236 have not been associated with clinical benefit in LQT patients in general.37 Furthermore, the predictive value of the in vitro testing is unlikely to ever be validated in a clinical trial, given the small number of individuals involved.
Both for LQT and CPVT, heterologous expression systems are readily available, cheaper, and have been used to identify drugs that later worked either in animal models and/or patients.26 Hence, it is not clear that iPSC-CM provide significant advantage of existing model system.
Given these formidable challenges, it remains to be seen what role iPSC-CM will play in the clinic for personalizing drug therapy of arrhythmia patients in the future.
A common question in the field of inherited arrhythmia syndromes is how a specific mutation identified in a patient causes arrhythmia susceptibility. Current approaches to answer this question have relied on heterologous expression systems, gene-transfer in cultured myocytes, gene-targeted mice, rats or rabbits, and computer modeling, either alone or in combination. In theory, the iPSC-CM technology would generate the diseased human myocyte in culture and enable the investigation of their cellular disease physiology without having to harvest myocytes from patients, which is essentially never possible. In practice, the iPSC-CM technology is hampered by inherent limitations of the model (i.e., single-cell model, lack of control for the effect of modifier genes) and technical limitations (i.e., incomplete reprogramming of mature cells by the Yamanaka factors resulting in residual epigenetic memory, high cell-to-cell and line-to-line variability, confounding effects of culturing, differences from adult CM physiology [Table 4]). Based on the analysis of published iPSC-CM arrhythmia models (Table 3), these limitations may have differential impact on the value of the iPSC-CM model depending on which disease is being studied. Table 5 summarizes the major pros and cons of using iPSC-CM as models of genetic arrhythmia syndromes, which may serve as a guide for investigators contemplating the use of this technology for investigating genetic arrhythmia disease. For example, atrial fibrillation caused by gain of function K-channel mutations may be an attractive disease to model using iPSC-CM, once the technical problem of generating high purity atrial-like iPSC-CM has been solved. On the other hand, diseases associated with either an increase or decrease in the SR Ca buffering protein Casq2 will be a major challenge, given that the level of Casq2 protein expression appears to strongly influence the maturation of Ca handling properties of ESC derived CM in culture.30 Moreover, in adult CM even a 25% reduction of Casq2 protein is sufficient to confer an increased risk for spontaneous Ca release in vitro and catecholaminergic arrhythmias in vivo.38 Hence, given that Casq2 protein is expressed at relatively low levels in iPSC-CM from healthy individuals,22 using iPSC-CM as model of CPVT appears to be problematic. Accordingly, the average journal impact factor of published iPSC-CM CPVT models of 7 is much lower than that of LQT iPSC-CM models, which is 20 (Fig. 2). At present time, it is too early to tell what role the iPSC technology will play in the study of genetic arrhythmia disorders associated with structural heart disease (i.e., hypertrophic cardiomyopathy, ARVC). Regardless of the disease in questions, it seems paramount to have a specific, testable hypothesis before embarking the costly and time-intensive journey of generating and studying iPSC-CM.
Based on the rapidly-declining journal impact factor of articles that report iPSC-arrhythmia models (Fig. 2), one might indeed consider iPSC-CM as boutique science that is flashy and fashionable but short lived. Furthermore, as shown in Table 3, iPSC-CM models thus far have not provided major new insights into arrhythmia mechanism, nor have they generated personalized therapy for patients. Moreover, as illustrated by the hypothetical case discussed here, iPSC-CM are unlikely to be relevant for the clinical care of the vast majority of patients with arrhythmia disorders. Hence, it appears unlikely that iPSC-CM approaches will revolutionize the diagnosis and treatment of arrhythmia disorders. On the other hand, compared to other arrhythmia models, iPSC-CM have the major advantage of being patient-specific human myocytes. Further improvement and standardization of the culturing methods, the use of nucleases (i.e., transcription activator-like effector nuclease [TALEN]) to correct the underlying genetic defects and thereby generate genotype-matched iPSC control cell lines,39 and 2D- or 3D-culturing methods may help to overcome many of the limitations mentioned in this article. Thus, it is likely that the iPSC-CM technology is here to stay and will be a valuable addition to the arsenal of existing arrhythmia models. Indeed, as I am writing about their limitations, I am using iPSC-CM approaches in the lab to investigate the biology of certain arrhythmia disorders.
Financial Support: NIH-R01 HL071670, HL088635 & HL108173
Disclosures: No conflict of interest declared
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.