Human iPSC-derived cardiomyocytes cultured in 3D engineered heart tissue show physiological upstroke velocity and sodium current density

1Animal-heart tissue is commonly used as a model for human-heart tissue, but exhibits a significantly different action potential (AP) duration and shape, due to different ion channel contributions, interactions and regulation. Human induced pluripotent stem cell-derived cardiomyocytes (hiPSC-CM) have the great advantage of generating human-like AP-duration and shape. In addition, hiPSC-CM represent a theoretically unlimited source of CM, lacking the ethical concerns that come along with sacrificing animals. The recent progress in the development and generation of hiPSC provides great opportunities to study individualised cardiac electrophysiology, focusing on genetic disorders and individualised drug treatment. However, there are concerns about the immaturity of hiPSC-CM. One important difference relates to AP upstroke-velocity which, in initial publications, was found to be markedly lower (~2–50%) in hiPSC-CM– than in adult CM. These findings suggest lower sodium current (INa) density in hiPSC-CM, which is of great physiological importance as INa determines excitability, conductance velocity, refractoriness and triggered activity. Furthermore, INa is an established drug target for antiarrhythmic therapy (flecainide, propafenone, amiodarone, vernakalant and ranolazine). Recent improvements have been brought about by co-culture with non-cardiac cells, long-term culture, hormone stimulation, continuous field stimulation and variation of substrate stiffness,, revealing upstroke velocities of up to 147 V/s. While these values approach the expected range (200–300 V/s) for human adult ventricular tissue, differences remain and a head-to-head comparison under same conditions is lacking.

2An alternative approach to increase the maturation of hiPSC-CM is cardiac tissue engineering. CM in hydrogel-based engineered heart tissue (EHT) form a synchronously beating syncytium, which generates contractile force by rhythmically deflecting the two elastic silicone posts it is attached to and thereby performs auxotonic contractile work,. Morphological and functional evidence suggest that hiPSC-CM reach a higher degree of maturity in EHT, but electrophysiological data are lacking. Here we directly compared upstroke velocity in hiPSC-CM cultured in 3D (EHT) and in human heart tissue biopsies obtained during the implantation of left ventricular assist devices (LVAD) or heart transplantation, and studied INa properties in hiPSC-CM from 2D monolayers (ML) and EHT under the conditions published for human adult CM.

3We compared the INa of EHT with standard 2D ML using the whole-cell patch clamp technique. Cell size as measured by the cell capacitance showed no statistically significant difference between EHT and ML (Fig. 1A): EHT 28.2 ± 2.0 pF (n = 37) vs. ML 23.3 ± 1.9 pF (n = 38). We studied INa at a reduced extracellular Na concentration, in order to ensure good voltage control and comparability to previous studies on human adult CM,,. As expected, current amplitude showed a proportional relation to cell size (Fig. 1D). Mean INa density was remarkably higher (~80%) in EHT (−18.5 ± 1.9 pA/pF; n = 17) than in ML (−10.3 ± 1.2 pA/pF; n = 23; p < 0.001; Fig. 1C). I–V curves (Fig. 2A) show that the INa was activated around −55 mV and peaked at −30 mV in EHT and ML. EHT hiPSC-CM showed higher INa density than ML over the entire activation range (p < 0.05, Fig. 2A). Thus, EHT showed increased INa density in comparison to ML. To test for a late sodium current, we measured currents at the end of our test-pulse. Currents amounted to −68.9 ± 13.9 pA under control condition and to −70.1 ± 13.9 pA after the application of 30 µM tetrodotoxin (TTX; n = 10, ns, paired t-test). Therefore, we did not find evidence for a persistent/late INa.

4The contribution of INa to the electrical activity of CM depends critically on the voltage and time-dependent activation and inactivation. Activation curves were calculated from individual I–V curves by normalising peak current amplitudes for their actual driving force and a Boltzmann function was fitted to the data set (Fig. 2C). Voltages for the half-maximum activation (V0.5act) of INa and curve steepness (kact) did not differ between EHT and ML (Table 1). In order to characterise the inactivation kinetics of INa, we fitted a single exponential function to current traces at different test pulse potentials (Fig. 2B). Time constants were shorter at increasing voltages of the test pulse without differences between EHT and ML.

5In general, resting membrane potential (RMP) in CM is less negative than V0.5inact. Therefore, only a minority of cardiac Na channels can be activated. Even small changes in RMP have strong effects on Na channel availability. We simulated different RMPs by applying variable conditioning pre-pulses from −120 mV to −30 mV for 1000 ms to determine steady-state inactivation. The mean data revealed no differences between EHT and ML in V0.5inact or kinact (Table 1). A Boltzmann curve (Fig. 2C) showed that activation and steady-state inactivation curves overlap between ~−77 mV and ~−45 mV (inset Fig. 2C). The maximal overlap was reached at −61.9 mV (ML) and −61.3 mV (EHT), where normalised INa availability and conductance were 1.4% (EHT) and 1.8% (ML) of the maximum. EHT and ML did not differ in window current amplitude or in voltage-dependency.

6The refractoriness of heart muscle depends critically on the fast recovery of INa from inactivation. The proportion of second INa to first INa was decreased by shorter interpulse duration as shown by the mean data and the two-phase exponential fit (Fig. 2D). Fitting a two-phase exponential function to the data set of each individual cell revealed no difference between EHT and ML (Table 1).

7TTX is a Na channel blocker with high affinity for neural isoforms and low affinity for cardiac isoforms of the Na channel. We found a concentration-dependent inhibition of INa in hiPSC-CM by TTX (Fig. 3A) with a sigmoidal concentration-response relationship. A single-site binding model could be fitted to the data points (Fig. 3C). The IC50 was calculated at 1.3 µmol/L (95% CI 1.1 and 1.6 µmol/L) with a slope factor of 1.08. A two-site binding model did not show a better fit. Applying the extra sum-of-square-test a single-site binding model was identified as the preferred fit. Our results argue against a relevant contribution of a high affinity binding site. Accordingly, transcript levels of the low-TTX-sensitive cardiac isoform NaV1.5 (SCN5A) were predominant without differences between EHT, ML and non-failing human left ventricular tissue (Fig. 3B), and the TTX-resistant neuronal isoform NaV1.8 (SCN10A) had 8-fold (LV), 50-fold (EHT) and 350-fold (ML) lower mRNA concentrations than NaV1.5 (SCN5A). Transcript levels of the neuronal isoform NaV 1.8 (SCN10A) were significantly lower in EHT- and ML-hiPSC-CM than in LV. Transcript levels of the highly-TTX sensitive brain-type isoforms NaV1.1–1.3, 1.6 (SCN1A, SCN2A, SCN3A, SCN6A) fell mainly below the cycle threshold cut-off level of 30. At the beginning of the differentiation from stem cells to cardiomyocytes SCN2A was the dominant isoform (Supplementary Figure 2). Expression levels of all isoforms described a U-shaped curve in development between day 0 and 20 of differentiation. However, at the late phase SCN5A clearly became the dominant isoform.

8EHT beat spontaneously (EHT: 0.83 ± 0.12 Hz, N = 5), whereas LV tissue was quiescent. We paced EHT slightly above its intrinsic rate at 1 Hz pacing in order to compare data to LV tissue (Fig. 4) and to the literature (Table 2). AP upstroke velocity in EHT did not differ significantly from that recorded in LV tissue (EHT: 219 ± 15 V/s, n = 13, N = 6 vs. LV: 253 ± 16 V/s, n = 25, N = 5; ns; Fig. 4A and B). Maximum diastolic potential (MDP), RMP immediately before the upstroke (= take-off potential) and AP amplitude (Fig. 4B) also did not differ significantly between EHT (MDP: −78.4 ± 2.9 mV, RMP: –73.5 ± 1.6 mV, 102.7 ± 2.8 mV, n = 13, N = 6) and LV (MDP: −74.8 ± 1.1 mV, RMP: −74.8 ± 1.1 mV, APA: 104.8 ± 1.4 mV, n = 25, N = 5).

9HiPSC-CM in EHT were oriented in parallel and showed a rod-shape morphology with sarcomere alignment comparable to LV tissue (Fig. 5). In contrast, hiPSC-CM in ML format showed an increased circularity and the sarcomeres were oriented in different directions even within the cell. NaV1.5 was distributed in the Z-disks and in the intercalated disks of adult cardiac tissue (Fig. 5). The subcellular distribution of NaV1.5 in ML hiPSC-CM was markedly different. It shows perinuclear enhancement with less pronounced signalling at the cell periphery and without co-localisation with α-actinin or enrichment at cell-cell contacts similar to what has been found in previous publications–. In contrast, hiPSC-CM in EHT showed a more pronounced expression of NaV1.5 in the periphery of the CM. Few CM in EHT showed co-localisation of NaV1.5 to Z-disks (arrowheads in Fig. 5B) and enhanced expression of NaV1.5 at cell-cell contacts orthogonal to the CM orientation (arrow in Fig. 5B), comparable to the intercalated disk of adult cardiac tissue. Proper impulse propagation does not only depend on the sodium channels, but also on polarized connexin-43 expression. We found pronounced connexin-43 staining at the cell membranes of hiPSC-CM in EHT, but no clear enhancement at end-to-end over lateral cell-cell contacts typical of adult human LV (Supplementary Figure 3).

10In the present study, we investigated whether culture of hiPSC-CM in the EHT format leads to a higher resemblance with adult human CM in terms of INa density, upstroke velocity, CM morphology and subcellular distribution of NaV1.5.

11INa densities are difficult to compare among different studies, as experimental conditions differ widely with respect to temperature and Na concentrations (as listed in Table 2). Here we used the same experimental conditions used previously to measure INa in human adult ventricular CM, allowing for reliable comparisons with hiPSC-CM. Due to the limited availability of human adult ventricular tissue, studies analysing its electrophysiological properties are rare. To our knowledge, only one publication has studied INa in human adult ventricular CM reporting an INa density of −20.2 pA/pF. Human atrial CM INa density measured under the same conditions amounted to −17.8 pA/pF, others reported values between −4 and -30 pA/pF,,,. Thus, the INa densities in EHT hiPSC-CM (-18.5 pA/pF) fit nicely with ventricular and atrial adult CM. The observation that INa density in EHT CM was nearly two fold higher than CM cultured in conventional ML (−10.7 pA/pF) provides evidence for the hypothesis that EHT culture improves the maturity of CM. The isometric mode of contraction of hiPSC-CM cultured on rigid surfaces (ML) in comparison to the auxotonic mode of contraction with defined load (EHT) might cause the difference in INa density. This supports recent data from hiPSC-CM cultured on soft substrate,, suggesting that auxotonic work is essential for the development of proper INa density. INa density in mammalian CM increases constantly during cardiac development from embryonic to neonatal stages until the adult state–, which occurs in parallel with increasing SCN5A expression during differentiation and culture of hiPSC-CM (Supplementary Figure 2 and Fig. 3B). Thus, the increase in INa density might be part of the maturation process.

12In order to elucidate potential mechanisms causing the greater INa density in EHT vs. ML, we reanalysed our previously published transcriptome of hiPSC-CM. We focused on the expression of genes thought to influence INa. On the one hand, we found lower mRNA levels in EHT for the transforming growth factor beta 1 (TGFB1l1) which is a multifunctional cytokine and may reduce NaV1.5 expression. On the other hand, we found higher mRNA levels in EHT of proteins believed to enhance INa function: epidermal growth factor (EGF), promoting ubiquitously growth, proliferation and differentiation; anchoring adaptor ankyrin-G (ANK3), which is involved in ion channel trafficking to the cell membrane, and plakophilin-2 (PKP2), which is part of cytoskeleton and cell-cell contact. Interestingly, missense mutations in plakophilin-2 are known to cause arrhythmogenic right ventricular cardiomyopathy and INa deficit.

13The contribution of the short lasting INa to the AP also depends on inactivation kinetics, which can be characterised by fitting a single exponential non-linear curve to the negative downslope of the INa. Inactivation kinetics revealed similar time constants and voltage-dependency in ML and EHT (Fig. 2B) fitting nicely to data reported for human adult CM. The voltage-dependency of activation and steady-state inactivation were indistinguishable between EHT and ML (Fig. 2C), implying that the higher INa density in EHT may be explained by a greater number of functional Na channels. As experimental conditions like time after rupture and temperature are known to shift V0.5act and V0.5inact to approximately the same extent, we calculated the actual difference (delta) between V0.5act and V0.5inact. We found values of 55.1 mV and 55.3 mV for EHT and ML, respectively (Table 2), which are comparable to published data,, for adult CM (54–58 mV). This finding is in contrast to previous measurements in hiPSC-CM cultured in ML showing markedly smaller values,,, including a recent study that describes the culturing of hiPSC-CM on a soft substrate. It is unclear if these discrepancies relate to technical differences or to intrinsic properties of different cell lines. Overall, we conclude that EHT format does not affect the voltage-dependent steady-state inactivation and activation in hiPSC-CM and these parameters are similar in our hiPSC-CM to that in human adult CM.

14The difference (delta) between V0.5act and V0.5inact mentioned above is not trivial: more overlap should generate more window current. Previously published studies on hiPSC-CM,, have shown reduced differences (delta) between V0.5act and V0.5inact and as a consequence a higher calculated window current (4–12%, Table 2). Larger window currents may partly explain persistent/late INa in CM and might have a critical influence on AP duration. In our experiments with hiPSC-CM, we found only a small amount of window current as in adult CM and no persistent/late INa. At the maximum degree of overlap (inset Fig. 2C), INa in hiPSC-CM showed similar fractions of maximal availability and conductance (1.4% and 1.8% for EHT and ML, respectively) as human adult CM (~2% and ~1% for atrial and ventricular CM, respectively). It should be noted that the amount of window current might be an additional marker in the maturation process, since window currents decrease during the development of chick embryonic CM.

15Interestingly, not only does the RMP influence INa density, but also vice versa, INa may influence RMP. Window INa may contribute significantly to RMP, especially when inward rectifier currents are small, as in human atrial trabeculae. Imanishi et al. have shown that high concentrations of TTX hyperpolarized the membrane potential of quiescent human atrial trabecular by about 7 mV. Consequently, window INa may be of particular relevance in hiPSC-CM, as less negative RMP and small inwardly rectifying potassium currents are consistently reported for hiPSC-CM.

16Sodium currents are known to recover quickly from voltage-dependent inactivation. Recovery from inactivation critically determines refractoriness and influences the susceptibility to tachyarrhythmia. In human adult atrial and ventricular CM recovery from inactivation could be fitted by a two-time constant function when plotting peak currents against different recovery time intervals,. Earlier studies on hiPSC-CM from ML reported similar characteristics of recovery from inactivation as here. It should be noted that both this study and Ma et al. found a somewhat faster recovery from inactivation in hiPSC-CM than in human adult CM. Whether this small difference has physiological relevance needs to be elucidated by computer modelling and functional studies.

17We used TTX-inhibition of INa to evaluate whether non-cardiac isoforms contribute significantly to INa in hiPSC-CM, which might be a sign of immaturity (Supplementary Figure 2). In hiPSC-CM, the effect size for low nanomolar concentrations of TTX fits perfectly to a single-site binding model with low sensitivity (IC50 1.3 µmol/L). Therefore, we would not assume relevant contribution from highly sensitive isoforms of Na channels to peak currents. Published TTX-sensitivities in atrial and ventricular CM amount to 1.1 µmol/L17 and 1.7 µmol/L16 showing that the TTX-sensitivity of INa in hiPSC-CM was similar to that in adult CM. Quantitative evaluation of the transcript levels of different Na channel isoforms confirmed that expression was dominated by the expression of the low-TTX sensitive isoform NaV 1.5 (SCN5A) compared to the highly-TTX sensitive neuronal isoforms NaV 1.1–1.3, 1.6 (SCN1–3A, SCN6A). The TTX-resistant neuronal isoform NaV1.8 (SCN10A) was expressed at intermediate levels and showed lower absolute transcription levels in hiPSC-CM than in LV.

18To the best of our knowledge, we show here for the first time AP in hiPSC-CM with an upstroke velocity similar to human adult ventricular tissue (200–300 V/s). Previous studies in hiPSC-CM have reported heterogeneous and overall lower upstroke velocities (Table 2) at ~40 V/s for ventricular-like AP in isolated cells–, or in embryoid bodies. Recent approaches to culture hiPSC-CM on a soft substrate of extracellular matrix as single CM or ML revealed higher INa density, and higher upstroke velocity of 147 V/s12 in comparison to cultures on a stiff substrate (65 V/s). Collectively, the data indicate that culture conditions allowing auxotonic contractions of hiPSC-CM against a flexible resistance increases the resemblance in INa density known for adult CM. This will be important for the further use of hiPSC-CM in drug testing and modelling of genetically determined cardiac diseases.

19We also found that AP in hiPSC-CM in EHTs showed a steeper early repolarisation and were considerably shorter than APs derived from LV tissue. A possible explanation is that APs from human tissue were all recorded from the subendocardial myocardium, which is known to express much less transient outward potassium current (Ito) and therefore exhibits longer APs than subepicardial regions. Future work is warranted to evaluate whether APs in EHT are in fact “subepicardial-like” or whether the observed difference indicate a specific hiPSC-CM phenotype or a peculiarity of the cell line under investigation.

20Immunohistochemistry revealed a parallel orientation of CM with a more rod-shaped morphology and sarcomere alignment of hiPSC-CM than in ML (Fig. 5). Interestingly, the subcellular distribution of NaV1.5 in EHT hiPSC-CM showed an enhancement of NaV1.5 at cell-cell contacts and, in some CM, pronounced signals, similar to those in the intercalated disks of the LV. Since NaV1.5 relocates from lateral to intercalated disks during cardiac development, the enhancement of NaV1.5 in the direction of sarcomere orientation might be another hint for structural maturation of hiPSC-CM by the EHT format. Additionally, some hiPSC-CM in EHT showed co-localisation of NaV1.5 with α-actinin at Z-disks, while hiPSC-CM in ML did not, as shown previously.

21Although the EHT format facilitated structural maturation, the cell size of hiPSC-CM was similar to those in ML format (~25 pF) and much smaller in comparison to adult LV CM (~100–200 pF),. As cell size increases during the embryonic development of cardiomyocytes, the small cell size may indicate an early stage of development. Smaller cells have a greater membrane area to volume ratio, but the physiological relevance of this remains unclear. From a technical point of view, smaller cells should conduct smaller absolute membrane currents making patch-clamp studies more technically demanding.

22A limitation of this study is that all LV tissue was obtained from patients with advanced heart failure due to dilated cardiomyopathy, since access to living non-failing heart tissue is virtually impossible. We therefore cannot fully exclude electrophysiological differences to healthy tissue. However, previous publications have shown no differences in sodium current properties from failing or non-failing hearts and upstroke velocity in our hands were similar to values reported for non-failing human heart.

23In conclusion, we have characterized INa in hiPSC-CM in 3D EHT and conventional 2D ML culture and compared AP characteristics in hiPSC-EHTs and human ventricular tissue. The main findings are 1) a higher INa density and a similar upstroke velocity in EHT as in human adult ventricular tissue, 2) similar voltage-dependent inactivation and activation of INa in EHT and ML, 3) no evidence for relevant non-cardiac isoforms contributing to INa in hiPSC-CM and 4) a higher resemblance of hiPSC-CM in EHT to LV concerning structure and subcellular NaV1.5 distribution than ML. Thus, our data suggest that EHT culture of hiPSC-CM may improve the validity of in-vitro experiments studying electrophysiological questions.

24This investigation conforms to all principles outlined by the Declaration of Helsinki and the Medical Associaton of Hamburg. According to the guidelines of the ethical review committee of the Medical Association of Hamburg, Germany, there is no need for a specific approval in this case since patient data were used anonymized. All materials from patients were taken with informed consent of the donors. Left ventricular free wall samples were obtained from patients undergoing implantation of left ventricular assist device (LVAD) or heart transplantation.

25As previously described, single cell suspensions of hiPSC-CM were either subjected to EHT generation in a 24-well format (1x106 hiPSC-CM/EHT in a fibrin matrix (total volume 100 µl) consisting of 10 µl/100 µl Matrigel [BD Bioscience, 256235], 5 mg/ml bovine fibrinogen (200 mg/ml in NaCl 0.9% [Sigma, F4753] plus 0.5 µg/mg aprotinin [Sigma, A1153]), 2x DMEM, 10 µM Y-27632 and 3 U/ml thrombin [Biopur, BP11101104]) or cultured conventionally in ML gelatin-coated 24-well plates (4x105 hiPSC-CM per well, 2 cm2). Culture media and duration were kept identical. For patch clamp measurements, hiPSC-CM in EHT and ML were isolated with collagenase II (200 U/ml, Worthington, LS004176) after a 24–29 day culturing period, and re-plated on gelatin-coated coverslips for 24–48 h in order to maintain adherence under perfusion.

26INa recordings were performed as described previously. In brief, borosilicate glass microelectrode pipettes (tip resistances 1.5–3.0 MOhm) were used to record INa in whole-cell configuration at room temperature (21 ± 1 °C) with an Axopatch-200B amplifier (Axon Instruments, Foster City, CA). Pipette and bath solution contained 5 mmol/L NaCl.

27APs were recorded as described previously with standard sharp microelectrodes in intact EHTs (25–60 days old) or LV tissue superfused with Tyrode´s solution at 36.5 ± 0.5 °C field-stimulated at 1 Hz (n = number of total impalements, N = number of EHT/LV tissue).

28Immunofluorescence was performed as described previously. Briefly, EHT or LV tissue were fixed in formaldehyde overnight at 4 °C, blocked for 6 h and incubated with primary antibodies (monoclonal mouse anti-α-actinin; monoclonal rabbit anti-NaV1.5) and secondary antibodies and nuclear staining (Alexa Fluor® 488 goat-anti-rabbit; Alexa Fluor® 546 goat-anti-mouse; DRAQ5TM). 2D cultures were cultivated on 96-well plates and were fixed for 20 minutes at 4 °C and stained accordingly with the exception of using Hoechst 33342 for nuclei staining.

29GraphPad Prism 5 (GraphPad Software, San Diego, CA, USA) was used for data analyses. Curves were fitted to data points from individual experiments and all data were compared using unpaired t-tests and for groups greater than 2 One-way ANOVA followed by Tukey corrections. Two-way ANOVA was used to assess repeated measurements of current-voltage relationship (Fig. 2A). All analyses were two-tailed and a p < 0.05 was considered to be statistically significant. Group data are presented as mean ± SEM.