Electrophysiological and calcium-handling development during long-term culture of human-induced pluripotent stem cell-derived cardiomyocytes

hiPSC cell line UMGi014-C clone 14 (isWT1.14) was derived from the dermal fibroblasts of a healthy male donor (31 years). They were cultured in feeder-free conditions using the integration-free CytoTune iPS 2.0 Sendai Reprogramming Kit (Thermo Fisher Scientific) with reprogramming factors OCT4, KLF4, SOX2, c-MYC. Previously published pluripotency and karyotype analysis of this line revealed no abnormalities or chromosomal instability [59]. Experimental protocols were approved by the ethics committee of the University Medical Center Göttingen (10/9/15). Directed feeder-free cardiac differentiation was achieved via canonical WNT modulation with small-molecules CHIR and IWP2, followed by metabolic selection with lactate as previously described [19, 39]. Day-3 (d-3) indicates final passaging whilst day 0 (d0) marks the onset of differentiation with WNT stimulation.

Between d27 and d30, purified hiPSC-CMs were digested with TrypLE (Thermo Fisher Scientific) and sparsely plated on 1:60 Matrigel-coated borosilicate glass 10 mm #0 round coverslips at a density of 15,000 cells/cm². Cells were incubated at 37 °C in 5% CO₂ and maintained every 2–3 days with a culture medium containing RPMI 1640 supplemented with B-27 (both Thermo Fisher Scientific). Cellular beating rate was routinely measured by photometric capture at × 40 magnification with a Retiga R6™ CCD camera mounted on an inverted microscope. Recordings were taken at 13 frames per second and analysed offline using the MUSCLEMOTION™ plugin on ImageJ [62]. For experimentation, coverslips were removed from their media and inserted directly into a heated bath chamber mounted on the stage of an inverted epifluorescence microscope. The differentiation and preparation process is outlined in Fig. 1a.

Whole-cell ruptured-patch techniques were employed to measure membrane currents in single, isolated early (d30–d46) and late (d47–d80) stage hiPSC-CMs. I_Ca,L and intracellular Ca²⁺ were measured simultaneously at 0.5 Hz with a voltage-clamp protocol consisting of a 100-ms ramp to − 40 mV (inactivating the fast Na⁺ current; I_Na) followed by a 100-ms depolarising test-pulse to + 10 mV, as previously described [15, 77]. For current voltage (I–V) curves, the test pulse was altered from − 40 to + 60 mV with 5 mV steps. Bath solution contained (in mmol/L): CaCl₂ 2, glucose 10, HEPES 10, KCl 4, MgCl₂ 1, NaCl 140, probenecid 2; pH = 7.35. 4-aminopyridine (5 mmol/L) and BaCl₂ (0.1 mmol/L) were added to block K⁺ currents [8, 73, 74, 78]. Pipette solution contained (in mmol/L): Fluo-3 pentapotassium salt 0.1, EGTA 0.02, GTP-Tris 0.1, HEPES 10, K-aspartate 92, KCl 48, Mg-ATP 1, Na₂-ATP 4; pH = 7.2. Sarcoplasmic reticulum (SR) Ca²⁺ content was assessed through integration of the Na⁺/Ca²⁺ exchanger-mediated current (I_NCX) during perfusion with 10 mmol/L caffeine.

Peak I_Na was measured in a bath solution containing (in mmol/L): NaCl 5, HEPES 10, MgCl₂ 1, CsCl 10, glucose 10, CaCl₂ 0.5, and TEA-Cl 120 (pH = 7.4, adjusted with CsOH). A voltage-clamp protocol was applied consisting of a holding potential at − 80 mV, followed by a 1000-ms pre-pulse step at − 110 mV (to increase availability of Na⁺ channels), and then 30-ms steps from − 80 to + 20 mV for I–V curves. Late Na⁺ current (I_Na,L) measurements were conducted with a bath solution containing (in mmol/L): NaCl 120, HEPES 10, MgCl₂ 1, CsCl 10, glucose 10 and CaCl₂ 0.5 (pH = 7.4, adjusted with CsOH). A voltage-clamp protocol was applied consisting of a holding potential of − 120 mV, followed by a 5-ms activating step to + 50 mV and a 300-ms step to − 30 mV to assess I_Na,L. Pipette solution for both I_Na and I_Na,L measurements contained (in mmol/L): NaCl 5, EGTA 10, GTP-Tris 0.4, HEPES 10, Mg-ATP 4, CsCl 20, CaCl₂ 3, Cs-Methansulfonate 90, pH = 7.2.

Delayed rectifier (rapid component; I_Kr) tail currents were measured using a high-performance automated patch clamp system (SyncroPatch 384; Nanion Technologies GmbH) with a voltage-clamp protocol consisting of a holding potential of − 80 mV followed by a 2-s step to 60 mV with steps of 10 mV for I–V acquisition. Bath solution contained (in mmol/L): CsCl 144, CaCl₂ 2, MgCl₂ 2, glucose 5, HEPES 10 (pH = 7.4 adjusted with CsOH). Pipette solution contained (in mmol/L) CsCl 20, EGTA 10, HEPES 10, CsF 110 (pH = 7.2 adjusted with CsOH) in accordance with a recently published protocol [26]. Cs⁺ was used as a charge carrier due to its selectivity for the hERG channel [65].

Basal inward-rectifier K⁺ current (I_K1) was measured at 0.5 Hz with a ramp pulse from − 100 to + 40 mV at 0.5 Hz while superfusing a modified Tyrode’s bath solution containing (mmol/L): NaCl 120, KCl 20, MgCl₂ 1, CaCl₂ 2, glucose 10, HEPES 10, pH = 7.4. Pipette solution contained (in mmol/L): EGTA 0.02, GTP-Tris 0.1, HEPES 10, K-aspartate 92, KCl 48, Mg-ATP 1, Na₂-ATP 4; pH = 7.2. I_K1 was identified as Ba²⁺ (1 mmol/L)-sensitive current as previously described [76].

APs were measured in current-clamp configuration at 0.5 Hz in bath solution containing (in mmol/L) the following: CaCl₂ 2, glucose 10, HEPES 10, KCl 4, MgCl₂ 1, NaCl 140; pH = 7.35. Pipette solution contained (in mmol/L): EGTA 0.02, GTP-Tris 0.1, HEPES 10, K-aspartate 92, KCl 48, Mg-ATP 1, Na₂-ATP 4; pH = 7.2. Mean holding currents were − 0.86 ± 0.13 pA/pF for early hiPSC-CM and − 1.05 ± 0.16 pA/pF for late hiPSC-CM (P = 0.3651).

All electrophysiological experiments were carried out at 37 °C, except for I_Na, I_Na,L and I_Kr which were measured at room temperature. Seal resistances were 3–6 GΩ. Borosilicate glass pipettes with tip resistances of 2–7 MΩ were used for voltage clamp experiments. High resistance borosilicate glass pipettes (5–10 MΩ) were used for current clamp experiments. All current recordings (except for I_Kr) were acquired using an Axopatch 200B microelectrode amplifier and analysed using pClamp-Software V 10.7 (both from Axon Instruments Inc., Foster City, USA). Membrane currents were corrected for membrane capacitance and expressed in pA/pF. Action potentials were acquired using a HEKA amplifier and HEKA patchmaster software and analysed using Lab Chart 7 (AD instruments, Otago, New Zealand).

[Ca²⁺]_i of single, isolated early- and late-stage hiPSC-CM was measured using the fluorescent Ca²⁺ indicator fluo-3-acetoxymethyl ester (Fluo-3-AM, 10 µmol/L, 10 min loading, 30 min de-esterification, λ_Ex = 488 nm, λ_Em = 535 nm) during simultaneous I_Ca,L measurement at 37 °C as previously described [77]. Fluorescence emission was collected with a photomultiplier optimised for high-speed signal capture (10 kHz). Emission was correlated to [Ca²⁺]_i with the formula [Ca²⁺]_i = K_d [F/(F_max − F)]. Here, K_d represents the dissociation constant of Fluo-3 (864 nmol/L), F denotes Fluo-3 fluorescence, and F_max describes Ca²⁺-saturated fluorescence obtained through cellular laceration at the end of each experiment [15].

Early and late hiPSC-CM were trypsinised and cellular membranes were isolated by differential centrifugation and then solubilized at 1 mg/ml of total protein in solubilisation buffer. Ca²⁺-handling proteins SERCA2a and NCX1, as well as expression of Kir2.1 were analysed with immunoblotting techniques. (LI-COR Biotechnology, US). Antibodies are outlined in Online Table S1. Immunofluorescent screening stained for nuclear and sarcomeric proteins using hiPSC-CM (d29) fixed in 4% PFA and permeabilised in 0.1% Triton X-100 with an AxioObserver A1 fluorescence microscope (Carl Zeiss, Jena, Germany). Flow cytometry utilised trypsinised, fixed and permeabilised hiPSC-CM (d29). Cells were screened using the LSRII flow cytometer (BD Biosciences, US).

The state-of-the-art in silico hiPSC-CM model by Kernik et al. [36] formed the basis for our simulations. The model was implemented in Myokit [9] and model parameters were adjusted to reproduce experimental data from early- or late-stage hiPSC-CM obtained in the present study (Online Table S2). The model age was set to 30 days for early-stage hiPSC-CM and 50 days for late-stage hiPSC-CM. For optimisation of I_K1, 40 days and 60 days were used, based on the maturity levels of the experimental data. Finally, we interpolated the parameters obtained for the early- and late-stage hiPSC-CM models to obtain parameter values as a function of age (Online Table S2, right columns). Linear functions were used for interpolation whenever possible. Alternatively, Hill functions were employed to prevent unphysiological values (e.g., negative membrane capacitance) at advanced age. AP simulations were performed and the steady-state AP following 1000 beats of prepacing was used for analysis in the presence of 0.2 pA/pF hyperpolarising current injected to suppress hiPSC-CM automaticity. The model code is freely accessible at www.​github.​com/​jordiheijman. The installation guide for the induced pluripotent stem cell-derived maturity evaluator (iMATURE) is available in the Online Supplement.

Data are reported as mean ± SEM and n-numbers as n/N, where n indicates number of hiPSC-CM studied from N differentiations, unless otherwise stated. Analyses were carried out with Prism 8 software (Graphpad, San Diego, USA). Normality of the data distribution was assessed using the Shapiro–Wilk normality test. Normally distributed data were compared using unpaired two-tailed Student’s t-test unless otherwise indicated. Data with unequal variance were compared using Welch’s t-test. Non-normally distributed data were compared using the Mann–Whitney U test. P < 0.05 was taken as statistically significant.

Next, we assessed I_Ca,L-triggered Ca²⁺ transients (CaTs) at different stages of hiPSC-CM development. I_Ca,L was initiated by a voltage-clamp protocol and was measured simultaneously with CaT (Fluo-3) in hiPSC-CM (Fig. 2a). Peak I_Ca,L was significantly larger in late hiPSC-CM versus early culture (− 29.24 ± 3.98 vs. − 10.52 ± 1.27 pA/pF, n/N = 28/3 vs. 19/3, P = 0.0005; Fig. 2b). Current–voltage relationships showed a positive shift in maximal current density in late cultures (Fig. 2c). Diastolic [Ca²⁺]_i was similar between both groups; however, systolic [Ca²⁺]_i was higher in late cultures. This was matched by a significant increase in CaT amplitude in late cells versus early cells (191.6 ± 26.81 vs. 134.3 ± 15.49 nmol/L, n/N = 28/3 vs. 19/3, P = 0.0358; Fig. 2d). The Ca_v1.2 blocker nifedipine (1 µmol/L) produced a reduction in I_Ca,L in every instance, suggesting presence of functional Ca_V1.2 from an early stage in cellular differentiation (Online Fig. S1).

SR Ca²⁺ content was assessed through caffeine application (10 mmol/L) after cessation of the I_Ca,L-activating protocol (Fig. 3a). The resulting caffeine-induced CaT (cCaT) amplitude and integral of corresponding membrane current (reflecting NCX-mediated Ca²⁺ extrusion; charge) were comparable between late and early hiPSC-CM cultures, indicating that higher CaT amplitude in late cells is mainly due to increased trigger I_Ca,L (Fig. 3b).

Diastolic Ca²⁺ removal from the cytosol was faster in late cultures versus early cultures as indicated by the rate constant of systolic CaT decay (inverse of CaT τ, k_syst, Fig. 3c). Decay of cCaT mainly reflects NCX-mediated Ca²⁺ removal and was also faster in more mature cells. This is consistent with higher peak I_NCX density (2.03 ± 0.38 vs. 0.97 ± 0.16 pA/pF, n/N = 22/3 vs. 13/3, P = 0.0455; Fig. 3d), pointing to increased NCX activity in late hiPSC-CM cultures. In accordance, average protein expression of NCX1 was numerically larger in late hiPSC-CM (Fig. 3e, f). Absolute levels of housekeeping protein calsequestrin (CSQ2) showed no difference between early and late cultures (1.00 ± 0.33 vs. 1.03 ± 0.28 a.u relative to early, n/N = 3/3 vs. 3/3 [not shown]). The rate constant k_SERCA represents the difference between the rate constant of cCaT decay and that of systolic CaT decay [15]. k_SERCA was significantly larger in late cultures versus early cultures (1.82 ± 0.38 vs. 0.91 ± 0.21 s⁻¹, n/N = 22/3 vs. 13/3, P = 0.0453; Fig. 3c). This was supplemented by the western blot findings (Fig. 3e, f).

Confocal line-scan analysis of Ca²⁺ sparks revealed a tendency towards decreased Ca²⁺ spark frequency and significantly decreased Ca²⁺ leak from the SR in late hiPSC-CM cultures versus early (12.73 ± 5.06 vs. 30.35 ± 9.04 100 µm⁻¹ s⁻¹, n/N = 41/7 vs. 39/2, P = 0.0416; Online Fig. S2). In addition, spontaneous beating rate, a marker of automaticity, was decreased in late-stage hiPSC-CM (0.45 ± 0.18 vs. 0.79 ± 0.4 Hz, n/N = 11/2 vs. 12/2, P = 0.0067; Online Fig. S2).

Peak I_Na was significantly larger in late-stage hiPSC-CM (− 71.12 ± 15.77 vs. − 26.63 ± 4.89 pA/pF, n/N = 29/3 vs. 21/3; P = 0.0237) with current–voltage relationships showing a slight negative shift in peak current density (Fig. 4a, b). I_Na,L was subsequently measured as current responsive to tetrodotoxin (TTX, 10 µmol/L) in both late and early hiPSC-CM cultures (Fig. 4c). In contrast to peak I_Na, integrated I_Na,L was not different between early and late hiPSC-CM (Fig. 4d).

K⁺ currents are responsible for repolarisation and stabilisation of resting membrane potential (RMP). Assessment of the rapid component of the delayed rectifier K⁺ current (I_Kr) following complete block with 25 µmol/L E-4031 revealed no age-dependent difference in tail current between late and early hiPSC-CM (− 2.94.12 ± 0.55 vs. − 3.84 ± 0.82 pA/pF, n/N = 77/4 vs. 60/4, P = 0.6733; Fig. 5a, b). Comparable I–V curves and pharmacology were also observed across early and late hiPSC-CM (Online Fig. S3).

The basal inward-rectifier K⁺ current I_K1 was measured in both early- and late-stage hiPSC-CMs using a modified ramp protocol with high extracellular [K⁺] (20 mmol/L), which produces a positive shift in reversal potential and allows for precise current detection, as previously described [76] (Fig. 5c). Ba²⁺-sensitive I_K1 was markedly increased in late-stage cells compared with early cells in both inward direction (− 100 mV: − 48 ± 12.3 vs. − 12.63 ± 2.28 pA/pF, n/N = 17/4 vs. 43/6, P = 0.0004; Fig. 5c, d) and outward direction (− 10 mV: 4.10 ± 0.62 vs. 2.75 ± 0.47 pA/pF, n/N = 17/4 vs. 43/6, P = 0.009), without changes in rectification (Online Fig. S4). In accordance, expression of Kir2.1 protein was significantly increased in late hiPSC-CM cultures (Fig. 5e, f).

We employed computational modelling to assess (1) whether the experimentally identified changes in I_Ca,L, I_NCX and SERCA are sufficient for the observed changes in hiPSC-CM Ca²⁺-handling and (2) to predict the maturation-dependent changes in AP characteristics resulting from the remodelling of all ionic currents. The recent Kernik et al. hiPSC-CM model was fit to our experimental data from early- or late-stage hiPSC-CM [36]. Besides the changes in I_Ca,L, I_NCX and SERCA function, adjustments in Ca²⁺ buffering and background Ca²⁺ influx were needed to reproduce the experimentally observed Ca²⁺-handling properties (Online Table S2). Nevertheless, with this limited number of changes, the model was able to reproduce all experimentally observed properties of early- and late-stage hiPSC-CM (Online Figs. S4–S8). Of note, interpolation of these Ca²⁺-handling parameters produced non-linear maturation-dependent changes in CaT properties that reflected the non-linear patterns observed in the experimental data (Fig. 6a). These modelling data suggest that gradual increases or decreases in expression levels of Ca²⁺-handling proteins may produce complex temporal changes at the cellular level.

Under current-clamp conditions at 0.5 Hz, the in silico hiPSC-CM model predicted a maturation-dependent decrease in action potential duration (APD) from 365 ms at d30 to 174 ms at d70. (Fig. 6b). In order to corroborate and validate the maturation-dependent AP shortening predicted by the modelling data, APs were measured at multiple time points during long-term hiPSC-CM monolayer culture (Fig. 6c). Under experimental current-clamp conditions, hiPSC-CM indeed displayed maturation-dependent changes, with late cells showing APD shortening at 50% and 90% repolarisation (APD_50, APD₉₀) compared with early-stage cells (APD₅₀: 163 ± 35.76 vs. 205.34 ± 26.87 ms [not shown]; APD₉₀: 259.1 ± 42.13 vs. 393.8 ± 38.65 ms, n/N = 13/3 vs. 23/3, P = 0.0358; Fig. 6d). Repolarisation fraction ([APD₉₀-APD₅₀]/APD₉₀), a representation of repolarisation profile and, therefore, an index of cardiomyocyte subtype, remained unchanged throughout hiPSC-CM culture suggesting the absence of a transient subtype shift during long culture periods (Online Fig. S9). Upstroke velocity and AP amplitude (APA) were increased in late-stage hiPSC-CM (upstroke velocity: 66.98 ± 20.9 vs. 30.22 ± 10.24 mV/ms, n = 10/3 vs. 16/3, P = 0.0309; APA: 123.7 ± 4.06 vs. 110.6 ± 3.67 mV, n = 13/3 vs. 23/3, P = 0.0298; Online Fig. S9), consistent with the increase in I_Na. No change was observed between early- or late-stage RMP both in the presence (Fig. 6d) and absence of injected current (Online Fig. S9), which is in line with our in silico simulations. In silico, the unchanged RMP could be attributed to a parallel increase in depolarizing NCX during maturation, counterbalancing the effects of increased I_K1 on RMP.

We then went back to the model to establish the major ionic determinant of the AP shortening by making use of the perfect control offered by in silico models. Preventing the maturation-dependent increase in I_K1 abolished the APD reduction and eventually elicited spontaneous activity (Fig. 7a). A progressive depolarisation of RMP was also observed due to increased NCX (Fig. 7a, b), highlighting the importance of maturation-dependent changes in I_K1 for cellular electrophysiology of hiPSC-CM. To further highlight the electrophysiological consequence of increased I_K1 in hiPSC-CM experimentally, we evaluated the effects of partial I_K1 inhibition. Older cells showed a 50% increase in AP duration following I_K1 blockade with BaCl₂ (1 mmol/L) compared to a non-significant 5% change in early cells (Online Fig. S10). In addition, application of BaCl₂ produced more instability and a high occurrence of arrhythmogenic events in late cells compared to early hiPSC-CM (Online Fig. S10).

There is a paucity of systematic studies assessing passive maturation of electrophysiological and Ca²⁺-handling processes in isolated hiPSC-CM cultured for more than 30 days after differentiation onset. Here, we report evidence of robust Ca²⁺-handling and operational SR Ca²⁺ stores in early hiPSC-CM, similar to results reported by Hwang et al. [27]. However, as our cells aged further, they displayed increased I_Ca,L and CaT amplitude (Fig. 2) along with a functional increase in key Ca²⁺-removal mechanisms, NCX and SERCA (Fig. 3). The present work expands on that of Hwang et al. and indicates that further maturation of Ca²⁺-handling machinery is possible in hiPSC-CM under prolonged culture of more than 50 days. Previous studies using cells under 30 days of age have outlined a dominant role of NCX function and poor SR development in hiPSC-CM [44]. Increased NCX-mediated electrogenic activity coupled with a leaky SR leads to increased incidence of delayed after-depolarisations (DADs) and could play a role in the cellular automaticity typically displayed by hiPSC-CM. The interplay of low inward-rectifier K⁺ current density paired with increased funny current (I_f) is assumed to be the major cause of spontaneous activity in hiPSC-CMs [66]. However, previous studies have shown that I_f density alone is not sufficient to induce automaticity in hESC-CM. Instead, Ca²⁺ release from the SR has been hypothesised as a major driver of automaticity with RyR2 abolition leading to cellular quiescence [38]. I_K1 also influences cellular automaticity and previous work highlights a potential regulatory effect of increased cytosolic Ca²⁺ flux during diastole by increasing rectification in cardiomyocytes, effectively blocking I_K1 and facilitating spontaneous beating [16, 80]. We, among others, report high SR Ca²⁺ leak and increased automaticity in early hiPSC-CM. However, as the cells age, we noted a lower incidence of Ca²⁺ sparks and decreased SR Ca²⁺ leak coupled with a significant decrease in spontaneous beating rate (Online Fig. S2). Interestingly, SR Ca²⁺ load remained comparable between early and late cultures, suggesting coordinated development of Ca²⁺-handling machinery during maturation (Fig. 3). The increased NCX function identified in older hiPSC-CM was not strongly replicated at the protein level, possibly also implicating the concurrent development of intracellular signalling, trafficking and phosphorylation mechanisms. Indeed, increased cAMP and cGMP have been found to enhance forward mode NCX function through protein kinase activation in older, but not in younger, embryonic mouse ventricular cardiomyocytes [58].

Together, these data point to a strong age-dependence of SR function under standardised and prolonged culture conditions. When interpolated in silico, a non-linear behaviour with exponential consolidation of SERCA activity is observed as hiPSC-CM age over time (Fig. 6). Our in silico model is the first to incorporate a detailed analysis of CaT and Ca²⁺-reuptake mechanisms across various stages of hiPSC-CM maturation.

An overview of hiPSC-CM electrophysiology from 2D and 3D preparations compared with our data and with data from native ventricular cardiomyocytes is provided in Tables 1, 2 and 3. It is important to emphasize that ion currents are dependent on experimental ion concentrations, which differ between studies. Direct comparisons should, therefore, be made with caution, ideally between groups studied under identical conditions. Nevertheless, similar to previous studies, our hiPSC-CM show a relatively immature phenotype characterized by less negative RMP, slower AP upstroke velocity and automaticity [8, 18, 26]. Lack of I_K1 is a hallmark of hiPSC-CM [23, 50]. To allow for comparison between studies performed at different extracellular K⁺ concentrations and temperatures, we estimated the conductance values based on reported I_K1 densities and calculated the resulting reversal potentials (E_rev, Tables 1, 23). The reduced I_K1 conductance observed throughout all studies is a major contributor to the less negative RMP and thereby facilitates the occurrence of spontaneous activity in hiPSC-CM constructs [24, 31, 34]. In addition, the depolarized RMP also causes reduced availability of voltage-dependent Na⁺ channels due to incomplete recovery from inactivation. Therefore, the reduced I_K1 may also contribute to the typically lower AP upstroke velocity in hiPSC-CM [12, 43, 50]. This combination induces proarrhythmic traits in hiPSC-CM which presents a severe disadvantage for their use for drug safety screening initiatives such as CiPA. Targeting I_K1 is therefore an important aspect of enhancing hiPSC-CM maturity and several methods exist to increase inward rectifier density. The hybrid method of dynamic patch clamp aims to overcome this limitation using an in silico ion channel model to adjust hyperpolarising current injection in real time based on the measured membrane potential [18, 52]. This technology is promising; however, it cannot capture native regulation of I_K1, e.g., by intracellular Ca²⁺ or Na⁺ [22, 75, 80]. Our combination of standardised serum-free differentiation and prolonged monolayer culture of > 50 days is sufficient to produce a 70% increase in I_K1 density with increased I_Na and concomitant increases in AP upstroke velocity (Figs. 4, 5, Online Fig. S9). Similar maturation-dependent increases in I_K1 density have been reported previously in hESC-CMs during prolonged culture [37, 64]. Doss et al. reported a transient increase in I_K1 density in hiPSC-CMs after 2 months, followed by a decrease after 4 months [12]. Our in silico extrapolation of I_K1 maturation does not support this parabolic developmental pattern, which could be due to differences in cellular culture techniques or their harsher dissociation procedures prior to measurement.

In the majority of studies, the maximal peak I_Na amplitude appears to be larger in hiPSC-CM compared with native cardiac tissue (Tables 1, 2, 3). This contradicts the lower expression levels of the underlying Nav1.5 subunit which are found in hiPSC-CM in comparison to human native ventricular cardiomyocytes [7]. Furthermore, direct comparison of peak I_Na currents is often hindered by different experimental conditions such as extracellular Na⁺ concentration and temperature. A direct comparison of hiPSC-CM and native human ventricular cardiomyocytes by Lemoine et al. suggests that peak I_Na currents are indeed lower in hiPSC-CM [46]. In addition, Lemoine et al. provide evidence for higher peak I_Na currents in more advanced 3D tissue culture models. This is in agreement with our study and a previous study in hiPSC-CM showing a tendency towards increased peak Na⁺ current in older hiPSC-CM [81]. Taken together, it appears that electrophysiological maturation of hiPSC-CM is associated with increased I_Na densities, although AP upstroke velocity remains low compared to adult ventricular cardiomyocytes.

The present work identifies maturation-dependent I_K1 augmentation as a key mediator of AP shortening in hiPSC-CM. Rapid delayed rectifier K⁺ currents remained unchanged throughout long term culture (Fig. 5) and, in the absence of maturation-dependent changes of I_K1 in silico, APD prolonged as the cells matured in the presence of increasing cytosolic Ca²⁺ activity (Fig. 7). This is in contrast to the Paci model [54], which attributes maturation-dependent AP shortening of hiPSC-CM to I_Ca,L and I_Kr dynamics. AP shortening upon I_K1 injection has indeed been shown previously in native cardiomyocytes [72], in hiPSC-CMs [4, 70], and in silico models [14]. To our knowledge, we provide the first evidence of maturation-induced I_K1 mediation of APD shortening in hiPSC-CM. This finding, along with our accompanying iMATURE maturation simulation software, could provide insight into the heterogeneous AP profiles which are regularly reported within and between hiPSC-CM cohorts. Future innovative studies promoting hiPSC-CM maturation should focus similarly on I_K1 development as this will be a key component in optimising hiPSC-CM for widespread screening initiatives [25].

Inherent variability in hiPSC-CM function hinders reliable quantification of average behaviour. In silico modelling has emerged as a powerful solution to link dispersed data sets and precisely define cellular parameters that contribute to experimentally observed heterogeneity. The recent state-of-the-art Kernik hiPSC-CM model integrates a wide range of experimental data by building a predictive array of cellular variability that allows for detailed investigation of cellular electrophysiology and underlying causal mechanisms of phenotypical heterogeneity [36]. Despite the availability of several useful hiPSC-CM models, none has comprehensively considered the ages of the hiPSC-CM used in the underlying experimental data. Using our own experimental findings and the framework of the Kernik model, we present the first hiPSC-CM model in which minimal parameter change can reproduce a wide range of electrophysiological properties of early- and late-stage hiPSC-CM (Online Figs. S4–S8). To facilitate analyses of maturation-dependent effects, our model has been integrated into an open-source interface, iMATURE, which allows the user to manually select the age of hiPSC-CM post differentiation and receive predictive readouts of AP morphology and ion channel dynamics over their specified age range. This platform also allows for manual modulation of individual ion channels and Ca²⁺ fluxes at any cellular age between 21 and 80 days post differentiation, enabling the investigation of age-dependent responses to hypothetical cardiotropic compounds (Fig. 8). Increased understanding of the impact of time-dependent phenotypic changes in hiPSC-CM is expected to contribute to standardisation of methodological techniques. Accordingly, this may enhance the quality of hiPSC-CM platforms, reduce variability in functional readouts, and promote efficient outlets for personalised medicine and streamline drug development [21].

In addition to prolonged cultivation times, strategies to enhance hiPSC-CM maturity include the adoption of appropriate cardiac differentiation protocols [6], culture substrates [23] and the application of mechanical, chemical or electrical stress to hiPSC-CM embedded in 3D hydrogels or fibrin blocks to more closely replicate the in vivo environment [35, 57, 68]. From an electrophysiological point of view, 2D cultivation strategies seem to be particularly limited with respect to the slow upstroke velocity of the AP, which is a common finding in all studies (Table 1). In contrast, 3D culture strategies seem to provide a promising improvement showing faster AP upstroke velocities (Table 2) [46]. It is important to note that the concept of hiPSC-CM maturation is a broad paradigm which also encompasses molecular, metabolic and structural properties [47]. Previous reports of hiPSC-CM maturation indeed show structural elongation and heightened sarcomere organisation, resembling the classical rod shape of cardiomyocytes [33, 49]. We observed highly heterogeneous morphological features in our developing cellular cultures with both rod-like and rounded cells. Measuring individual cells in sparsely seeded monolayers is essential for gathering true electrophysiological readouts; however, this environment is inherently artificial due to severely decreased cell-to-cell communication. Therefore, various external paracrine effects and electrical stimuli may not play a major role in the development of our cells [37]. Our differentiation technique enables a baseline assessment of cellular function under fully defined culture conditions. Further studies using targeted maturation-enhancing techniques can, therefore, build upon this foundation.

In addition to the remodelling characterised in the present study, other currents may contribute to hiPSC-CM electrical development and age-dependent action potential shortening. Cardiac currents such as t-type calcium current (I_Ca,T), I_f, transient-outward K⁺ current (I_to) or the slow component of the delayed-rectifier K⁺ current (I_Ks) were not experimentally quantified, although they are present within the Kernik hiPSC-CM computer model [36] that our work is based on. Previous studies have identified a large role for I_Ks and I_Kr in the maintenance of repolarization reserve in adult human cardiomyocytes [5, 32]. During our experimental examination of I_K1 block on AP duration, we cannot exclude the possibility that our high concentration of BaCl₂ could also cause unspecific inhibition of the delayed-rectifier K⁺ currents [32]. Following consolidation with our entire experimental data set, our iMATURE platform is able to provide appropriate readouts of all cardiac-related currents and does not indicate a strong contribution of the delayed-rectifier K⁺ currents in age-related AP shortening in hiPSC-CM.

hiPSC-CMs also show regional subtype-specific traits that allow them to be classified as ventricular-, atrial- or nodal-like cardiomyocytes [29]. In the absence of an intervention to direct subtype differentiation, for example by promoting an atrial phenotype with retinoic acid, hiPSC-CM generally show predominantly ventricular traits, with minimal mixing of other subtypes [10]. We cannot definitively state that our experimental data contains only ventricular-like cells. Differences in ionic makeup of atrial and nodal cells influence their AP morphology and repolarisation profile, which are, therefore, commonly utilised as functional markers of cellular subtype [13]. The repolarisation fraction provides an index of phase 3 AP kinetics and, in our hands, does not show evidence of atrial or nodal cell contamination in our experimental cohort. In addition, no significant change in repolarisation fraction was detected during maturation, indicating the absence of a subtype-shift during prolonged culture as has been recently proposed [3].

Additionally, some potential limitations should be considered related to the computational modelling performed in this study. (1) Although we employed a widely used and well-validated state-of-the-art in silico model of hiPSC-CM [36], the model dependence of our results cannot be excluded. (2) The intercellular heterogeneity of ion channel distribution and activity could potentially affect our observed AP properties. Here, we present a single deterministic hiPSC-CM model without intercellular variability. Although we do identify an overall trend of age-dependent AP shortening in hiPSC-CM, future research should employ high-throughput electrophysiological techniques and multiple populations of in silico hiPSC-CM models to properly capture and map this phenomenon.