Wild-type hiPSCs were derived from fibroblasts of pooled fetal livers originally isolated in 2006 (a gift by Dr. Tobias Cantz, Hannover Medical School). These were expanded in 10% serum/Knockout™ DMEM (Thermo)/1 x penicillin/streptomycin/glutamine (PSG, Thermo), replated at 100,000 cells per well of a 12-well plate, and transfected next day using 4 μl Ribojuice (Merck) and 1 μg of self-replicating RNA containing reading frames for OCT4, SOX2, KLF4, and GLIS1 (Addgene plasmid #58974). Reprogramming RNA was generated as previously described (Yoshioka et al., 2013), using a RiboMAX T7 in vitro transcription kit (Promega) on linearized plasmid DNA, and Cellscript kits #C-SCCE0610, #C-SCMT0610, and #CPAP5104H for RNA capping, 2′-O-Methylation, and PolyA tailing, respectively. Reprogramming medium contained 100 ng/ml B18R protein (eBioscience), 0.5 μg/ml puromycin (for 3 days, then 0.1 μg/ml), 0.5 mM valproic acid, and 5 μM EPZ004777 (Tocris). hPSC-like colonies emerging several weeks later were picked and expanded under hiPSC maintenance conditions (see below). A selected, newly generated hiPSC line termed F1 was characterized using a panel of assays which have been described elsewhere (Greber et al., 2011). Karyotyping was performed according to standard procedures (n = 10). Primers for RT-qPCR and bisulfite sequencing are given in Table S2.

For KCNA5 mutagenesis, a “4n” CRISPR approach targeting the 5′ end of the KCNA5 coding sequence was employed, using 4 CRISPR/Cas9n nickase vectors as illustrated in Figure 1I. Fwd oligonucleotide sequences defining the gRNAs targeting this genomic region are given in Table S2. These were phosphorylated, annealed with their corresponding reverse oligos and cloned into a modified version of pX335 (Addgene plasmid #42335) carrying a GFP-puromycin selection cassette. F1 hiPSCs were co-transfected with all 4 CRISPR vectors using Fugene HD (Roche) and transiently selected for 24 h using 0.5 μg/ml puromycin, to enrich for transfected cells. Following replating of the cells at clonal dilution several days after transfection, clonal half-colonies were picked another ~2 weeks later, to be analyzed for successful mutagenesis. This was done using conventional DNA isolation and PCR spanning the target region (Table S2). CRISPR-induced genomic deletions were indicated by smaller PCR products which were subsequently cloned and sequenced (~10 TOPO clones per mutant hiPSC line, to statistically obtain sequence information on both alleles).

F1 hiPSCs, the KCNA5^fs/fs derivative line, and HuES6 human embryonic stem cells were routinely cultured in 6-well plates on 1:75 diluted Matrigel™ HC (Corning #354263), in FTDA medium [10]. FTDA consisted of DMEM/F12, 1 × ITS (Corning #354350), 0.1% human serum albumin (Biological Industries #05-720-1B), 1 × defined lipids (all Thermo), 1 × PSG, 10 ng/ml FGF2 (PeproTech #100-18B), 0.2 ng/ml TGFβ1 (eBioscience #34-8348-82), 50 nM Dorsomorphin (Santa Cruz #sc-200689), and 5 ng/ml Activin A (eBioscience #34-8993-85). Cells were routinely passaged as single cells or, initially, as clumps of cells. For single cell splitting, cells were grown to full confluence (until cultures seemingly appeared syncytial), digested for 15 min using Accutase™ containing 10 μM Y-27632 (R&D # 1254/50), and replated in the presence of Y-27632 at 600,000 cells per well of a 6-well plate. hiPSCs reached full confluence after about 4 days under these conditions and were subsequently harvested as above, for continuous maintenance or for the induction of differentiation. hiPSCs were kept in culture for a maximum of 30 passages. Cell lines were tested negative for mycoplasma.

Cardiac induction was performed under serum and serum albumin-free conditions (Zhang et al., 2015). Fully confluent hiPSC cultures—which then appeared as a flat, seemingly syncytial monolayer—were harvested using Accutase™, centrifuged twice at 300 g, resuspended in day 0 differentiation medium and seeded out at 550,000 cells per well of a Matrigel-coated 24-well plate, in a total volume of 2 ml. Compared to maintenance plates, differentiation plates were coated with a further 1:3 dilution of Matrigel (a total dilution of approximately 1:250). Day 0 differentiation medium consisted of KO-DMEM, 1 × ITS, 10 μM Y-27632, PSG, 5 ng/ml Activin A, 10 ng/ml FGF2, 0.5–1 ng/ml BMP4 (R&D #314-BP-050), and 1 μM CHIR99021 (AxonMedchem #Axon 1386). Before placing freshly seeded differentiation cultures back into the incubator, plates were tapped several times under a stereo microscope and then left standing at room temperature for 30 min, to ensure an even distribution and initial attachment of cells. Medium in differentiation plates was exchanged on a daily basis. From day 1 onwards, the basal differentiation medium consisted of KO-DMEM, 1 × TS (transferrin/selenium), 250 μM 2-phospho-ascorbate (Sigma-Aldrich #49752), and PSG. 100 × TS stock was prepared in advance by dissolving 55 mg transferrin (Sigma #T8158) in 100 ml PBS containing 0.067 mg sodium selenite (Sigma #S5261). WNT inhibitor C-59 (Tocris #5148/10) was added to the cultures from 48to 96 h of differentiation at 0.2 μM, to promote cardiac specification. Optionally, for promoting an atrial fate, all-trans-retinoic acid (Sigma #R2625-50MG) was supplemented from 72 to 120 h of differentiation.

Differentiating cultures typically started to display spontaneous beating after 6 (no RA) or 8 days (+RA). To allow maturation of the cultures until the preferred time-point of analysis (typically at ~2.5 weeks after the initiation of differentiation), beating monolayers were dissociated using TrypLE Select (Thermo) and replated at a ratio of ~1:4 using CM splitting medium consisting of RPMI 1640 (Thermo), 1 × ITS, 0.2% (w/v) HSA, 250 μM phospho-ascorbate, 0.008% thioglycerol, PSG, and 10 μM Y-27632. Next day, medium was replaced by CM maintenance medium consisting of KO-DMEM, 1 × ITS, 0.2% (w/v) HSA, 250 μM phospho-ascorbate, 0.008% thioglycerol, and PSG. Following CM maturation, cells were replated once again, at an appropriate ratio, onto MEA or glass surfaces for downstream analyses several days later (~3 week total).

For electrophysiological analysis on microelectrode arrays (USB-MEA256 system, Multichannel Systems), the electrode areas of plasma-cleaned 9-well MEAs were coated with 3 μl of a 1:150 diluted Matrigel/0.1% gelatin solution in KO-DMEM for ~2 h at 37°C in a humidified cell culture incubator. hiPSC-CMs were dissociated from maintenance cultures using a 10 × TrypLE Select digestion to obtain a single-cell/small aggregate suspension. Coating solution was removed from the electrode arrays to be replaced by 25,000–50,000 cells resuspended in a ~3 μl droplet of CM replating medium. CMs were allowed to attach for ~30 min. Subsequently, MEA chambers with attached cells were filled with 150 μl of CM replating medium. Next day, medium was replaced by CM maintenance medium once more. From the following day onwards, cell preparations were used for electrophysiological recordings at 37°C. Recordings of drug-treated cells were initiated after a wash-in time of about 5 min (4-AP: 1 mM, propranolol: 10 μM, isoprenaline: 10 μM, AF-DX-116: 10 μM, cisapride: 100 nM, carbachol: 0.2–0.5 mM). Wash-out recordings were performed after five media changes. T_max and peak-to-peak finding algorithms were implemented in MC Rack software v4.5.7. Field potential durations (FPDs, QT_max intervals) and peak-to-peak (RR) intervals were averaged from five consecutive measurements of a given sample, if appropriate, and averaged between independent replicates. Data were processed in MS Excel where QT_max-like intervals were frequency-corrected using Bazett's formula: QTc_max = QT_max [ms]/(RR [s])^0.5. Poincaré plots were generated from representative recordings.

For action potential (AP) measurements, culture tissues were enzymatically dissociated into single cells as described previously (Meijer van Putten et al., 2015) and plated at a low density on Matrigel-coated coverslips. AP measurements were performed using the amphotericin-B perforated patch-clamp technique and an Axopatch 200B amplifier (Molecular Devices). Data acquisition and analysis were realized with custom software. Signals were low-pass-filtered with a cutoff of 5 kHz and digitized at 40 kHz. The potentials were corrected for the calculated liquid junction potential of 15 mV (Barry and Lynch, 1991). Cell membrane capacitance (C_m) was determined with -5 mV voltage step from −40 mV by dividing the time constant of the decay of the capacitive transient by the series resistance.

APs were recorded at 36 ± 0.2°C from single and spontaneously contracting hiPSC-CMs. Cells were superfused with solution containing (in mmol/L): 140 NaCl, 5.4 KCl, 1.8 CaCl₂, 1.0 MgCl₂, 5.5 glucose, 5.0 HEPES; pH 7.4 (NaOH). Patch pipettes (borosilicate glass; resistance ≈2.5 MΩ) contained (in mmol/L): 125 K-gluconate, 20 KCl, 5 NaCl, 0.44 amphotericin-B, 10 HEPES; pH 7.2 (KOH). hiPSC-CMs typically lack the inward rectifier K⁺ current, I_K1, that limits the functional availability of Na⁺ current (I_Na) and transient outward K⁺ current (I_To) (Hoekstra et al., 2012; Giles and Noble, 2016). To overcome this limitation, we injected an in silico I_K1 with kinetics of Kir_2.1 channels through dynamic clamp, as we previously described in detail (Meijer van Putten et al., 2015). An amount of 2 pA/pF peak outward current was applied, resulting in quiescent hiPSC-CMs with a maximal membrane depolarization (MDP) of −80 mV or more negative. APs were elicited at 0.2–3 Hz by 3 ms, ~1.2 × threshold current pulses through the patch pipette. We analyzed the MDP, maximum AP amplitude (APA_max), AP duration at 20, 50, and 90% of repolarization (APD₂₀, APD₅₀, and APD₉₀ respectively), maximal upstroke velocity (V_max) and plateau amplitude (APA_plat) measured 20 ms after the AP upstroke. Averages were taken from 10 consecutive APs.

Immunofluorescence analysis was carried out according to standard procedures following paraformaldehyde fixation and permeabilization/blocking with 0.2% Triton X-100/5% FCS/2% BSA/2% glycine in PBS for 45 min. Antibodies used were anti-AFP (1:250, Sigma #A8452), ANP (1:50, R&D # AF3366), K_v1.5 (1:100, Santa Cruz #sc-377110), MLC2v (1:200, ProteinTech Group #10906-1-AP), NKX2.5 (1:200, R&D #AF2444), OCT4 (1:100, Santa Cruz #sc-5279), SMA (1:200, Sigma #C6198), SOX2 (1:200, R&D #AF2018), Troponin C (1:100, Thermo #MS-295-P), Troponin I (1:250, Santa Cruz #sc15368), βIII-Tubulin (1:2000, Covance #PRB-435P), as well as appropriate Alexa Fluor 488 or 568-conjugated secondary antibodies (Thermo). Images shown are full or cropped frames taken with a 10 or 20 × objective mounted to a Zeiss Axiovert inverted microscope.

Intracellular staining prior to FACS analysis was performed as described (Zhang et al., 2015). Briefly, cells were dissociated using TrypLE Select, fixed, and stained in 0.5% saponin/5% FCS in PBS using α-CTNT (1:200, Labvision #MS-295-P) and Alexa-488-conjugated α-mouse (Life Technologies).

RNA was isolated using NucleoSpin RNA kits with on-column DNA digestion (Machery Nagel). Reverse transcription was performed using M-MLV reverse transcriptase (Affymetrix #78306) with oligo-dT₁₅ priming at 42°C. Real-time PCR was carried out using validated primers in Table S2 and BioRad iTaq™ Universal SYBR Green Supermix (#172-5124) on an ABI 7500 cycler. Efficiency-validated primers used are given in Table S2. RPL37A served as housekeeping control. Data are expressed relative to an indicated control sample or, alternatively, as percentage of RPL37A abundance (100^*2^−ΔCt).

Unless otherwise stated, data are presented as means between biological replicates ± SEM. Statistical analysis was carried out with SigmaStat 3.5 software. Two groups were compared using unpaired t-tests or, in case of a failed normality and/or equal variance test, Mann-Whitney Rank Sum Test. In action potential measurements, normality and equal variance assumptions were tested with the Kolmogorov-Smirnov and the Levene median test, respectively. More than 2 groups were compared using One-Way ANOVA or Two-Way repeated ANOVA followed by a Student-Newman-Keuls Method post-hoc test. In case of non-normally distributed parameters, Kruskal-Wallis test followed by pairwise comparisons with Dunn's Method was performed. P < 0.05 defines statistical significance.

A previously described approach based on a self-replicating RNA vector expressing four reprogramming factors was employed to generate integration-free wild-type hiPSCs from fetal human fibroblasts (Figure 1A) (Yoshioka et al., 2013). Following RNA vector transfection and antibiotic selection to maintain robust transgene expression (Figure S1A), cell line F1 was obtained and characterized in more detail. F1 hiPSCs showed typical human pluripotent stem cell morphology, stained positive for alkaline phosphatase under hPSC maintenance conditions (Frank et al., 2012), and displayed a normal male karyotype (Figures 1B,C). Following withdrawal of puromycin selection pressure, transgene expression was no longer detectable indicating a transgene-free reprogrammed state, as expected (Figure 1D). Instead, F1 cells showed reset DNA methylation of the OCT4 and NANOG promoters as evidenced by bisulfite sequencing (Figure 1E). Expression of endogenous pluripotency markers was indistinguishable from that in human embryonic stem cells and robustly detectable at protein level (Figures 1F,G).

Immunocytochemical analysis of spontaneously differentiated cultures indicated competency to form derivatives of all three germ layers suggesting acquired pluripotent characteristics (Figure S1B). Finally, a previously established protocol for directed differentiation along the cardiac lineage was applied revealing that CMs could be generated at high efficiency (Figure S1C, Video S1). These data suggest that F1 cells are fully reprogrammed to a pluripotent state.

The KCNA5 gene encoding K_v1.5 (Figure 1H) is composed of a single exon. Four CRISPR/Cas9 nickase vectors were designed to target the 5′ end of the open reading frame, to potentially induce frame shift-causing deletion mutations in a controlled manner. Indeed, we were able to isolate one clone showing small homozygous frame shift-causing deletions leading to a premature stop codon on both alleles (Figure 1I). This engineered cell line designated KCNA5^fs/fs hence resembles the genotype of a well-documented patient case of lone AF caused by a heterozygous but dominant KCNA5 nonsense mutation, to lead to a loss of K_v1.5 function (Olson et al., 2006). The homozygous 76 bp deletion in our model was confirmed using conventional PCR and sequencing at the genomic and transcriptional levels (Figure 1J).

Next, we sought to assess whether both the mutant and the wild-type (WT) control lines were capable of undergoing selective differentiation into ventricular vs. atrial-like CMs. The Zhang et al. protocol (Zhang et al., 2015) served as a baseline for ventricular-like differentiation, as this presents the default fate for most cells in virtually any directed differentiation procedure. Optionally, transient stimulation with RA may be administered for conferring an atrial CM identity (Devalla et al., 2015). Optimization experiments suggested that a 2-day supplementation with RA between 72 and 120 h of differentiation presented a sufficient time period (Figure 2A and data not shown). Panels of ventricular and atrial-specific marker genes were defined based on our previous analysis of primary human heart tissue samples in comparison to hPSC-CMs (Piccini et al., 2015). These were used for analyzing RA titration experiments aiming at identifying an optimal dose. Results revealed that a stimulation with 0.5 μM RA during CM induction was sufficient for robustly inducing atrial markers including KCNA5 while strongly suppressing ventricular ones like IRX4 and MYL2 (encoding myosin light chain 2, Figure S2A).

Importantly, this variant of the protocol was applicable both to WT and KCNA5^fs/fs hiPSCs suggesting that the model as well as its isogenic control may efficiently be converted into the desired cardiac subtype (atrial-enriched CMs). Default ventricular-like CMs could instead serve as an important specificity control in subtype-specific disease modeling (Figure 2B). After ~1.5 weeks of maturation under CM maintenance conditions, RA untreated CMs showed MYL2 (MLC2v) expression and low atrial natriuretic peptide (ANP) abundance at the protein level, whereas the reverse pattern was obtained after RA treatment (Figure 2C, bottom). By contrast, K_v1.5 was detectable only in atrial-like WT cells and absent in KCNA5^fs/fs CMs, which confirmed the homozygosity of the underlying mutation and the consistency of the differentiation methodology (Figure 2C, Figure S2B).

Electrophysiological analyses on multielectrode arrays (MEAs) revealed that spontaneous beating frequencies were increased in atrial CMs as compared to ventricular ones, as expected given shorter AP durations (APDs) in the former (Figure 3A). In ventricular CMs, beat-to-beat (RR) intervals were indistinguishable between WT and KCNA5^fs/fs cells. In atrial-enriched cell preparations, however, KCNA5^fs/fs CMs showed slower beating frequencies than isogenic WT controls implying that K_v1.5 functionally influences atrial CM beat rates (Figure 3A, Videos S2, S3). Analysis of frequency-corrected field potential durations (QTc_max intervals, FPDs) likewise showed no difference between the cell lines under default differentiation conditions. Atrial-enriched WT CMs displayed greatly reduced (< 100 ms) QTc_max intervals as compared to ventricular WT CMs, which is expected given the lack of a plateau phase in underlying atrial APs. Compared to atrial WT cells, atrial KCNA5^fs/fs CMs however showed a substantial FPD prolongation (~200 ms, Figure 3B).

In agreement with these data based on bulk cultures, perforated patch clamp measurements revealed an atrial-specific phenotype of our model at single-cell level. Compared with ventricular-like CMs, APs of RA-treated cells were in general characterized by fast phase-1 repolarization resulting in short APs with a negative plateau phase (Figure 3C). APs in atrial KCNA5^fs/fs CMs, however, were significantly prolonged at 20 and 50% of repolarization and had a more positive AP plateau phase, as compared to atrial WT CMs (Figure 3C). Notably, these marked differences in atrial APDs remained conserved at various stimulation frequencies (Figure 3D). By contrast, APDs in ventricular-enriched cultures did not differ significantly between the WT and KCNA5^fs/fs CMs. Furthermore, there was no marked difference in AP amplitudes, plateau, or upstroke velocities between the groups (Figure 3C). These data indicate the functional importance of K_v1.5 in shaping the atrial action potential in the hPSC system and reveal a pronounced cardiac subtype-specific phenotype in the disease model.

To affirm that this primary phenotype is indeed due to the loss of K_v1.5 in atrial KCNA5^fs/fs CMs, we next evaluated the effects of 4-aminopyridine (4-AP), an inhibitor of K_v1.5 (Devalla et al., 2015). 4-AP administration had little effect on beating frequencies (Figure S3A). Likewise, ventricular WT and KCNA5^fs/fs CMs did virtually not respond to this compound when assessing FPDs on MEAs or APDs using single-cell patch clamping (Figure 4A, and data not shown). By contrast, 4-AP provoked a significant FPD and APD prolongation specifically in atrial WT cardiomyocytes but exerted no effects in K_v1.5-deficient KCNA5^fs/fs mutants (Figures 4A,B, Figure S3B, Table S1). Taken together, these data imply that the atrial-specific FPD and APD prolongation phenotype in the KCNA5^fs/fs model is indeed due to the loss of K_v1.5 function and that this channel accounts for much of the repolarization reserve.

A patient-specific KCNA5 nonsense mutation has previously been shown to increase vulnerability for triggered activity under adrenergic stress conditions (Olson et al., 2006). Hence, to assess the relevance of the KCNA5^fs/fs model in the context of AF, we exposed atrial WT and mutant CMs to several distinct stimuli and channel blockers. Modulating β-adrenergic signaling using propranolol and isoprenaline promoted a marked decrease and increase in spontaneous beating frequencies, respectively, albeit without causing any irregularities (Figures S4A,B). Furthermore, to interfere with the function of K⁺ channels potentially collaborating with K_v1.5 in atrial CMs, we employed AF-DX-116, a muscarinic receptor antagonist (to indirectly compromise K_ir3.1/3.4) (Hammer et al., 1986), and cisapride, a potent hERG blocker (Rampe et al., 1997). These treatments did however not reveal any signs of irregular beating in atrial KCNA5^fs/fs CMs (Figures S4C,D), contrasting with our previous results based on ventricular-enriched K_v7.1-deficient hiPSC-CMs (Zhang et al., 2014).

To challenge the model in an alternative manner, we next used carbachol, a cholinergic agonist promoting the acethylcholine-sensitive K⁺ current (I_KACh). Although this type of treatment might seem counterintuitive in the context of a KCNA5 mutation, carbachol administration presents an accepted means of inducing AF in animal models (Kovoor et al., 2001). Supplementation of as much as 0.5 mM carbachol to bulk cultures only promoted an expected reduction in beat-to-beat intervals of atrial WT CMs, with no irregularities (Figure 5A, left). By contrast, atrial KCNA5^fs/fs CMs tended to show irregular baselines in-between contractions following carbachol administration (Figure 5A, middle), or chaotic beating (Figure 5A, right), which was never seen in WT or ventricular-enriched preparations. These effects were reversible, as normal beating was restored following drug washout, albeit these observations were only made in a subset of experiments (4 out of 8). Because irregular beating in atrial KCNA5^fs/fs CMs tended to occur at reduced beating rates (Figure 5A), we finally sought to identify possible triggers for this mutant-specific behavior using patch clamp methodology in single cells at low stimulus frequencies. Interestingly, at 0.2 but not at 1 Hz, pronounced EADs were obtained in atrial KCNA5^fs/fs CMs but not in WT ones (Figure 5B). Again, these events were observed in a subset of samples (5 out of 12 cells analyzed). Together, these data show that atrial KCNA5^fs/fs hiPSC-CMs are prone to develop EAD-like triggering events that may translate into macroscopic beating irregularities in a stochastic manner.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.