Maturation of hiPSC-derived cardiomyocytes promotes adult alternative splicing of SCN5A and reveals changes in sodium current associated with cardiac arrhythmia

Abstract Aims Human-induced pluripotent stem cell-cardiomyocytes (hiPSC-CMs) are widely used to study arrhythmia-associated mutations in ion channels. Among these, the cardiac sodium channel SCN5A undergoes foetal-to-adult isoform switching around birth. Conventional hiPSC-CM cultures, which are phenotypically foetal, have thus far been unable to capture mutations in adult gene isoforms. Here, we investigated whether tri-cellular cross-talk in a three-dimensional (3D) cardiac microtissue (MT) promoted post-natal SCN5A maturation in hiPSC-CMs. Methods and results We derived patient hiPSC-CMs carrying compound mutations in the adult SCN5A exon 6B and exon 4. Electrophysiological properties of patient hiPSC-CMs in monolayer were not altered by the exon 6B mutation compared with isogenic controls since it is not expressed; further, CRISPR/Cas9-mediated excision of the foetal exon 6A did not promote adult SCN5A expression. However, when hiPSC-CMs were matured in 3D cardiac MTs, SCN5A underwent isoform switch and the functional consequences of the mutation located in exon 6B were revealed. Up-regulation of the splicing factor muscleblind-like protein 1 (MBNL1) drove SCN5A post-natal maturation in microtissues since its overexpression in hiPSC-CMs was sufficient to promote exon 6B inclusion, whilst knocking-out MBNL1 failed to foster isoform switch. Conclusions Our study shows that (i) the tri-cellular cardiac microtissues promote post-natal SCN5A isoform switch in hiPSC-CMs, (ii) adult splicing of SCN5A is driven by MBNL1 in these tissues, and (iii) this model can be used for examining post-natal cardiac arrhythmias due to mutations in the exon 6B. Translational perspective The cardiac sodium channel is essential for conducting the electrical impulse in the heart. Postnatal alternative splicing regulation causes mutual exclusive inclusion of fetal or adult exons of the corresponding gene, SCN5A. Typically, immature hiPSCCMs fall short in studying the effect of mutations located in the adult exon. We describe here that an innovative tri-cellular three-dimensional cardiac microtissue culture promotes hiPSC-CMs maturation through upregulation of MBNL1, thus revealing the effect of a pathogenic genetic variant located in the SCN5A adult exon. These results help advancing the use of hiPSC-CMs in studying adult heart disease and for developing personalized medicine applications.


Introduction
Human-induced pluripotent stem cells (hiPSCs) are widely used to derive cardiomyocytes (hiPSC-CMs) and study cardiac diseases at the cellular and molecular level. [1][2][3] Patient-specific hiPSC-CMs capture the genetic landscape of the affected individual, allowing precise dissection of pathogenic mechanisms and testing drug responses that may ultimately lead to (personalized) treatments. Among the cardiac diseases that have been modelled using hiPSC-CMs are inherited cardiac arrhythmias caused by mutations in cardiac ion channels. 1,4 The SCN5A gene encodes the pore-forming α-subunit of the cardiac sodium (Na + ) voltage-gated channel Nav1. 5. This channel conducts the depolarizing Na + current (I Na ), which is responsible for the fast upstroke of the action potential (AP) in working cardiomyocytes; it is essential for conducting the electrical impulse in the heart. Mutations in SCN5A are associated with different cardiac diseases, including Brugada syndrome, long QT syndrome, isolated conduction defects, and dilated cardiomyopathy. 5 Nav1.5 expression is developmentally regulated, with foetal -or adult isoforms generated by alternative splicing of the SCN5A exon 6 (6A foetal and 6B adult), which encodes the voltage sensor of the first channel subunit. 6 The mechanism of SCN5A alternative splicing regulation and other cardiac genes during development has been studied in animal models, where the muscleblind-like RNA-binding protein 1 (Mbnl1) has been identified as key for inducing a post-natal switch. [7][8][9] In humans, MBNL1 is sequestered in the heart of myotonic dystrophy patients causing re-expression of the foetal SCN5A exon 6A 10 and the importance of MBNL proteins for late myogenic maturation has been recently demonstrated using hiPSCs. 11 hiPSC-CMs can recapitulate many electrical alterations due to mutations in SCN5A or in genes of Nav1.5 beta-subunits [12][13][14][15][16][17][18][19][20][21][22] and these are in some cases notably different from those observed in commonly used heterologous expression systems, 23 most likely because hiPSC-CMs carry the full complement of cardiac ion channel and accessory protein genes. Moreover, gene editing in hiPSCs allows effects of specific mutations to be distinguished from potential confounding elements resulting from line-to-line variability; patient mutations can be introduced into wild-type (WT) hiPSCs or mutations in patient hiPSC lines can be corrected. 15,24 However, cardiomyocytes differentiated from hiPSCs have an immature (foetal-like) phenotype with their electrical properties, structure, and gene expression profiles resembling first or second gestational trimester foetal cardiac cells 25,26 rather than adult cardiomyocytes. Although I Na in hiPSC-CMs display half-maximal potential (V 1/2 ) of activation and inactivation consistent with values reported for native human ventricular myocytes, 27 the maximal AP upstroke velocity (V max ) is slower than in adult cardiomyocytes. 28 In addition, hiPSC-CMs mostly express the Nav1.5 foetal isoform, masking effects due to mutations located in the adult exon 6B. 29 Previously, we and others addressed the issue of SCN5A in vitro maturation by culturing hiPSC-CMs for long periods to promote spontaneous maturation 29 and by using a medium inducing a metabolic switch from glycolysis to fatty acid oxidation. 20 We recently discovered that hiPSC-CM maturation is overall enhanced in a three-dimensional (3D) microtissue (MT) environment containing cardiac fibroblasts (CFs) and endothelial cells (ECs). 30 hiPSC-CMs isolated from MTs have more hyperpolarized RMP and faster AP upstroke velocity, suggesting an electrical maturation involving Nav1. 5.
Here, we showed that maturation of hiPSC-CMs in MTs promoted expression of the SCN5A adult isoform and allowed (i) evaluation of functional effects of the p.R225W mutation, located in adult exon 6B and (ii) detection of the more severe disease phenotype associated with the compound heterozygosity for p.W156X mutation, 13 as observed in patients. 31 Using gene-corrected isogenic hiPSCs was crucial in identifying mutation-specific effects. We also found that MBNL1 upregulation in MTs was necessary and sufficient to induce the expression of SCN5A exon 6B.

Patients' clinical history and phenotype
The clinical characteristics of the family were previously reported. 31 The proband patient ( Figure 1A, II-3), born with a severe tachycardia and conduction disorder, was treated with beta-blockers and later fitted with an ICD implant. Despite the ICD, he died from intractable ventricular arrhythmias at age 21. The parents and a sister ( Figure 1A, II-1) of the proband were asymptomatic, whereas the other sister ( Figure 1A, II-2) carrying the same mutations of the proband was affected by intractable arrhythmia and died at age of 1.

hiPSCs generation and culture
A skin punch biopsy was obtained following written informed consent from the patient at the age of 18 and approval by the medical ethics committee of the Amsterdam Medical Center (AMC), in accordance with the Declaration of Helsinki. Dermal fibroblasts were reprogrammed using the Sendai virus by the Leiden University Medical Center (LUMC) hiPSC core facility, as described previously, 13 following the protocol approved by the LUMC and AMC ethical committees. Details for hiPSC culture and differentiation are reported in the Supplementary material online.

hiPSC editing using CRISPR/Cas9
For the correction of c.468G.A mutation in SCN5A exon 4 (W156X) and excision of SCN5A exon 6A specific sgRNA sequences ( Figures 1E  and 3A, Supplementary material online, Table S1) were cloned into the pSpCas9(BB)-2A-Puro (PX459) v2 plasmid encoding Cas9 nuclease and puromycin resistance (Addgene #62988). 32 For the correction of c.673C . T mutation in exon 6B (p.R225W) and the excision of exon 4 of MBNL1, the Alt-R® CRISPR-Cas9 System (IDT Technologies) was used. Specific single-stranded DNA oligonucleotide (ssODN) repair template carrying the WT base and silent mutations were used for each mutation correction (see Supplementary material online, Table S1). Additional information is provided in the Supplementary material online.

Electrophysiology
Beating monolayers of hiPSC-CMs were dissociated at day 21 of differentiation using Tryple Select 0.5× (ThermoFisher Scientific) and plated onto Matrigel-coated glass coverslips at low density. Electrophysiological properties of hiPSC-CMs were analysed 10 days after dissociation. MTs were dissociated at day 21 of culture using collagenase II (Worthington Industries), as previously described. 33 Cells were plated on Matrigel-coated coverslips in LI-BPEL supplemented with 50 ng/mL VEGF and 5 ng/mL bFGF, refreshed with LI-BPEL medium the next day, and analysed for electrophysiological properties 7-10 days after dissociation.
Patch-clamp experiments were performed using an Axopatch 200B amplifier (Molecular Devices, Sunnyvale, CA, USA). Clampex 10.7 was used for data acquisition, voltage control, and analysis. APs were recorded in isolated single cells using the perforated patch-clamp technique in current-clamp mode and sodium current (I Na ) was recorded with the ruptured patch-clamp technique in voltage-clamp mode. Patch-clamp solutions and protocols are reported in the Supplementary material online.

RNA isolation, cDNA synthesis, and gene expression analysis
Total RNA was isolated either from hiPSC-CMs on day 21 of differentiation and from hiPSCs using Nucleospin kit (Machery and Nagel), or from MTs on day 21 of culture using RNeasy Micro Kit (Qiagen), according to the manufacturers' instructions. cDNA was synthesized using iScript™ cDNA Synthesis Kit and gene expression was analysed by qPCR with IQ Syber Green Supermix (Bio-Rad), according to the manufacturer's protocol. All qPCR reactions were performed in duplicate. Gene expression levels were normalized to the RPL37A housekeeping gene.
For quantification of mRNA-splicing isoforms containing 6A or 6B exons, BioRad QX200™ Droplet Digital PCR (ddPCR) system was used following the manufacturer's instructions. Primer and probe sequences used for qPCR and ddPCR are reported in Supplementary material online, Table S2 and S3.

RNA-seq analysis
The RNA was extracted with the RNeasy micro kit (QIAGEN) as from the manufacturer's instructions, followed by paired-end sequencing at 150 bp, 40 million reads for each sample (Novogene, Cambridge, UK). Residual Illumina adapters were trimmed from the RNA sequences using cutadapt 34 with parameter -m 20. Reads were aligned to the human genome (build 38, Ensembl version 96) using STAR. 35 BAM files were used by rMATS 36 to detect differentially spliced exons and sashimi plots were made using the function rmats2sashimiplot (https://github.com/Xinglab/ rmats2sashimiplot). Exon counts for the SCN5A and MBNL1 canonical transcripts, ENST00000413689 and ENST00000282486, respectively, were retrieved using featureCounts 37 with parameters -t exon and -g exon_id. Transcripts per kilobase million (TPM) was calculated considering all exons counts as library size and by using the R script convertCounts. The coverage of constitutive and alternative exons for SCN5A or MBNL1 was calculated as percentage of TPM assigned to each exon bin over the TPM mean of the whole transcript.
RNA-seq data are available in the GEO repository with accession number GSE180290. Publicly available foetal heart (FH) and adult heart (AH) RNA-seq data (GEO, accession number GSE62913 38 ) were used for Supplementary material online, Figure S7A.

Statistical analysis
Data were obtained from at least three independent differentiation-or MT formation experiments and were expressed as mean + standard error of the mean (SEM). Data were compared using t-test or one-way ANOVA followed by Fisher LSD post hoc test. Statistical analysis was performed using OriginPro 2016 (Origin Lab), RStudio, or GraphPad Prism (GraphPad Software, Inc.).

Generation and genetic correction of hiPSCs from a patient with compound heterozygous mutations in SCN5A
A patient ( Figure 1A, II-3) presented with severe cardiac conduction abnormalities and ventricular arrhythmias early after birth and was genetically characterized as carrying bi-allelic mutations in the SCN5A gene. 31 These were a c.468G.A nucleotide change in exon 4 of SCN5A (NM_198056.2) inherited from the father, leading to the stop codon mutation p.W156X (NP_932173.1), and a c.673C.T nucleotide change in exon 6B of SCN5A inherited from the mother. This led to the substitution of a positively charged arginine with an aromatic tryptophan (p.R225W) in the voltage-sensor segment of the first domain of the channel ( Figure 1A and B). When the patient turned 18, we reprogrammed dermal skin fibroblasts using the Sendai virus to generate the hiPSC W156X/R225W line. Sanger sequencing confirmed the presence of both SCN5A mutations ( Figure 1C). After several passages, hiPSCs were negative for the Sendai virus (see Supplementary material online, Figure S1A), had a normal karyotype (see Supplementary material online, Figure S1B), expressed the pluripotency markers NANOG, SSEA4, POU5F1 ( Figure 1D), and genome-wide gene expression analysis indicated high 'pluripotency score' and low 'novelty score' with the PluriTest algorithm 39 (see Supplementary material online, Figure S1C, left).
While the p.W156X Nav1.5 mutation was previously characterized in a heterologous system 31 and using hiPSC-CMs derived from the proband's father ( Figure 1A, I-1), 13 the phenotypic effects of the p.R255W Nav1.5 mutation was only studied in heterologous system 31 since hiPSCs from the proband's mother were not available. To examine the effects of this mutation specifically, we corrected the c.468G.A SCN5A mutation in the hiPSC W156X/R225W line using CRISPR-Cas9 and generated the hiPSC corr/R225W line ( Figure 1E and F). One hiPSC clone was selected, in which Sanger sequencing confirmed the correction of the c.468G.A SCN5A mutation and the presence of the silent mutations in heterozygosity ( Figure 1F). No modifications were present in the in silico-predicted two intragenic Cas9 off-target sites by Sanger sequencing (data not shown). hiPSC corr/R225W had a normal karyotype (see Supplementary material online, Figure S1D) and high pluripotency score, as confirmed by PluriTest (see Supplementary material online, Figure S1C, right). The original hiPSC W156X/R225W line was heterozygous for the rs6781731 SNP located in the intron downstream exon 4. The corrected hiPSC corr/R225W was also heterozygous for this SNP, confirming that the recombination involved only the paternal allele as intended (see Supplementary material online, Figure S1E).

No functional effects of exon 6B mutation in monolayer hiPSC-CM cultures expressing the foetal but not adult SCN5A isoform
We differentiated hiPSC W156X/R225W and hiPSC corr/R225W lines into cardiomyocytes (CMs) in monolayer ( Figure 1G). Differentiation efficiencies were comparable between the two lines and consistently yielded .90% cardiac troponin T (TNNT2)-positive cells (see Supplementary material online, Figure S2A). The cardiac ion channel genes SCN5A, NCX1, KCND2, KCNH2, KCNQ1, and KCNJ12 were expressed at similar levels in CMs from hiPSC W156X/R225W and hiPSC corr/R225W (see Supplementary material online, Figure S2B).  We next examined the electrophysiological properties of CMs from hiPSC W156X/R225W and hiPSC corr/R225W lines. Figure 2A shows representative I Na traces from hiPSC W156X/R225W -and hiPSC corr/R225W -CMs. While the voltage dependence of I Na activation and inactivation was similar between the two lines (see Supplementary material online, Figure 2B), I Na density was lower in hiPSC W156X/R225W -CMs over the whole physiological voltage range ( Figure 2C), with the mean peak current density in hiPSC W156X/R225W -CMs (− 31 + 4.5 pA/pF) significantly reduced compared with hiPSC corr/R225W -CMs (− 94.2 + 11.4 pA/pF) ( Figure 2D). Notably, the current density recorded in hiPSC corr/ R225W -CMs was comparable to that previously reported (≏ − 100 pA/ pF) for CMs differentiated with the same method from a WT hiPSC line (hiPSC WT/WT ), whereas I Na density in hiPSC W156X/R225W -CMs was similar to that recorded in hiPSC-CMs carrying only the p.W156X mutation (≏ − 30 pA/pF) derived from the proband's father ( Figure 1A I-1). 13 Altered I Na impacts the electrical activity of cardiomyocytes, particularly the AP upstroke velocity. We, therefore, recorded APs in hiPSC W156X/R225W -and hiPSC corr/R225W -CMs paced at 1 Hz. To compensate for the typical depolarized RMP of foetal-like hiPSC-CMs, we used dynamic clamp to set the RMP to − 86 mV. 40 Figure 2E shows representative AP traces recorded from hiPSC W156X/R225W -and hiPSC corr/ R225W -CMs, and their respective first derivative trace. In agreement with the differences in I Na density, the maximal first derivative of AP upstroke, corresponding to the maximal upstroke velocity (V max ), was significantly lower in hiPSC W156X/R225W -compared with hiPSC corr/R225W -CMs ( Figure 2F). As observed for I Na , V max values from hiPSC corr/ R225W -CMs were similar to those previously reported for WT hiPSC-CMs using the same dynamic clamp technique. 41 Other AP parameters, namely RMP, AP amplitude (APA), and AP duration (APD), were not different between the two lines (see Supplementary material online, Figure S2C).
To check whether both mutations contributed to I Na and V max , we measured exon 6A and 6B expression in CMs from hiPSC W156X/ R225W and hiPSC corr/R225W lines. Using digital droplet PCR (ddPCR) we quantified exon 6A-and exon 6B-containing transcripts (see Supplementary material online, Figure S2D and E). Figure 2G illustrates representative ddPCR results for hiPSC W156X/R225W -and hiPSC corr/ R225W -CMs. Exon 6B-containing SCN5A transcripts accounted for approximately 5% and were much lower than exon 6A-containing transcripts ( Figure 2G and H ). In FH and AH tissues, exon 6B-containing transcripts accounted for 50% and 94% of all transcripts, respectively, in line with previous reports, 29,42 and thus were expressed at higher levels than in hiPSC-CMs ( Figure 2H).
These results indicated that hiPSC W156X/R225W and hiPSC corr/ R225W -CMs expressed the foetal SCN5A isoform almost exclusively and I Na density and V max reduction in the compound mutant was due to the p.W156X Nav1.5 mutation. The effect of p.R225W mutation could not be properly determined.

Genetic excision of exon 6A does not increase adult SCN5A isoform expression
To investigate whether the removal of exon 6A might force SCN5A exon 6B expression and thus reveal the effect of the p.R225W Nav1.5 mutation, we used CRISPR-Cas9 to excise this exon from both SCN5A alleles in the hiPSC W156X/R225W line. As illustrated in Figure 3A, we designed 2 gRNAs targeted to the flanking regions of exon 6A and identified a clone in which exon 6A had been excised in both alleles. The actual excision spanned a larger intronic region than intended by the strategy design ( Figure 3A and Supplementary material online, S3A). In the hiPSC lines with complete exon 6A excision (referred to as 6A-KO hiPSC W156X/ R225W ), we corrected the p.W156X mutation in exon 4 (generating the 6A-KO hiPSC corr/R225W line), using CRISPR-Cas9 with the same strategy as earlier ( Figure 1E). The excised hiPSC lines differentiated into CMs with similar efficiencies and beating properties as the parental hiPSC line (data not shown). As expected, exon 6A was not expressed at all in excised CMs at 21 or 60 days of culture, whereas expression of exon 6B was detected (see Supplementary material online, Figure S3B). To examine SCN5A transcripts expressed by CMs from the excised hiPSC lines, we amplified by PCR a region between exon 3 and exon 8 of SCN5A ( Figure 3B). Here, we only used the 6A-KO hiPSC corr/R225W line, since in the 6A-KO hiPSC W156X/R225W line, the stop codon p.W156X mutation in exon 4 leads to nonsense-mediated mRNA decay 13 and could confound the results. Gel electrophoresis of PCR products revealed three distinct bands ( Figure 3C). Sanger sequencing demonstrated these corresponded to three different transcript species: the highest band represented the adult isoform of SCN5A transcript including exon 6B; the middle band, a transcript in which exon 6B was skipped, and in the lowest band, both exons 5 and 6B were skipped ( Figure 3C).
To investigate whether the overall expression of the adult SCN5A isoform was increased compared with the non-excised line, we performed ddPCR analysis, normalizing the expression of exon 6B with that of the TATA-binding protein (TBP). Exon 6B expression was similar in the excised hiPSC-CMs and in the parental non-excised hiPSC-CMs ( Figure 3D). Since exon 6B-containing transcripts were likely the only functional transcript species produced in the exon 6A-excised lines, we investigated whether this would unmask the functional phenotype of the p.R225W Nav1.5 with electrophysiology. We attempted to record I Na in 6A-KO hiPSC W156X/R225W -and 6A-KO hiPSC corr/R225W -CMs but I Na was virtually absent in all (n = 9) 6A-KO hiPSC W156X/R225W -CMs ( Figure 3E), while 50% of the 6A-KO hiPSC corr/R225W cells (n = 10) showed very small I Na ( Figure 3E) with mean peak current density of − 12.9 + 5.5 pA/pF. Analysis of SCN5A transcript expression by RNA-seq showed that upon excision of exon 6A, not only was there spurious mRNA produced but also the intronic region between exon 5 and exon 6B was aberrantly included at a similar level to the exons ( Figure 3F).  Taken together, these results demonstrated that excision of SCN5A exon 6A impairs the splicing mechanism, leading to abnormally spliced transcripts of unknown significance and the inclusion of intronic regions in the transcript population. Importantly, 6B-containing transcript levels were not sufficient to generate detectable I Na for comparison of the two mutant lines.

Electrical maturation of hiPSC-CMs in 3D cardiac MTs with hiPSC-derived cardiac non-myocytes reveals functional effects of p.R225W mutant Nav1.5
We next tested whether hiPSC-CM maturation in cardiac MTs containing hiPSC-derived CFs and cardiac ECs, 30,33 could promote SCN5A exon 6B expression and thus reveal the effect of the p.R225W mutation. The cardiac MTs were generated using defined ratios of the three cell types as shown in Figure 4A and dissociated into single cells after 21 days of culture to analyse I Na by patch clamp ( Figure 4B). Since I Na had not been characterized in hiPSC-CMs derived in this way before, we used hiPSC-CMs from hiPSC WT/WT as a process control. 13,43 In addition, we corrected the p.R225W mutation in hiPSC corr/R225W , generating the isogenic control hiPSC corr/corr (see Supplementary material online, Figure S4A). Correction of the mutation in hiPSC corr/corr was confirmed by Sanger sequencing (see Supplementary material online, Figure S4B). The line had a normal karyotype and differentiated efficiently into cardiomyocytes (see Supplementary material online, Figure S4C and D). Four groups of MTs were analysed, containing CMs from hiPSC W156X/R225W , hiPSC corr/R225W , hiPSC corr/corr (Supplementary material online, videos S1-S3), and hiPSC WT/WT . Representative traces of normalized I Na recorded in CMs dissociated from MTs are shown in Figure 4C. The inset shows respective current trace at − 40 mV normalized on peak current, showing smaller activation for hiPSC W156X/R225W compared with the other lines, indicating that a smaller fraction of I Na channels opened at the given voltage. Indeed, the activation curve of hiPSC W156X/R225W showed a significant rightward shift compared with the other lines ( Figure 4D). I Na current density showed instead a genotype-specific reduction, with I Na significantly lower along the physiological voltage range in hiPSC W156X/R225W -CMs compared with all the other lines, and significantly lower in hiPSC corr/R225W -compared with hiPSC corr/ corr and hiPSC WT/WT -CMs ( Figure 4E).
Finally, we examined the effect of the altered I Na on the AP. Figure 4F shows representative AP traces and the derivative of the AP upstroke recorded in hiPSC-CMs dissociated from MTs and paced at 1 Hz. V max was significantly smaller in hiPSC W156X/R225W -CMs compared with CMs from the other lines ( Figure 4G), showing that the compound mutations have an additive effect on I Na density, thus impacting on CM electrical activity. Other AP parameters such as spontaneous resting membrane potential, APA and APD were instead similar in the different lines (see Supplementary material online, Figure S5).

Alternative adult SCN5A splicing is promoted in hiPSC-CMs exposed to the MT microenvironment
To obtain a more general overview of gene expression, we performed bulk RNA sequencing in hiPSC W156X/R225W -and hiPSC corr/R225W -2D CMs and 3D MTs after 21 days of culture. Gene ontology (GO) analysis showed that genes related to GO terms of heart development and contraction had a similar expression in the two hiPSC lines, both in 2D CMs and in MTs (see Supplementary material online, Figure S6A), indicating that the cells had reached similar levels of differentiation and maturation. Our single-cell (sc) RNA-seq datasets from MTs and hiPSC-CMs previously published from the same WT line used here confirmed that SCN5A is only expressed in the hiPSC-CM population (Supplementary material online, Figure S6B). 30 Analysis by ddPCR showed that MTs from hiPSC W156X/R225W , hiPSC corr/R225W , and hiPSC WT/WT expressed similar fractions of SCN5A exon 6B ( Figure 5A), which accounted for around 27% of total SCN5A transcripts ( Figure 5B). Notably, the fraction of exon 6B-containing transcripts was 5-6 times higher in MTs from hiPSC W156X/R225W and hiPSC corr/R225W compared with their respective hiPSC-CMs derived in monolayer ( Figure 5C). RNA-seq analysis confirmed that exon 6B expression was increased in MTs compared with monolayer hiPSC-CMs in both lines ( Figure 5D). Overall SCN5A expression was also higher in MTs compared with monolayer hiPSC-CMs in both lines ( Figure 5E), in agreement with scRNA-seq in hiPSC WT/WT (see Supplementary material online, Figure S6C) and confirmed by qPCR (see Supplementary material online, Figure S6D, left). The total expression of exon 6 (6A+6B) was similar to upstream and downstream exons; however, the fraction of exon 6B in MTs was higher than in monolayer hiPSC-CMs in both lines ( Figure 5E). Exon 6B fraction was nonetheless lower in MTs compared with FH and AH (see Supplementary material online, Figure S7A). Interestingly, SCN5A expression was not increased in MTs containing 6A-KO hiPSC-CMs and the expression of exons downstream of exon 6B was drastically reduced, confirming that 6A excision impaired splicing in the entire gene (see Supplementary material online, Figure S7B).

The alternative splicing regulator MBNL1 promotes SCN5A exon 6B inclusion during hiPSC-CM maturation
The RNA-binding protein Mbnl1 has been previously shown to promote exon 6B-and inhibit exon 6A expression in the mouse heart. 7, 9 We, therefore, investigated whether MBNL1 is involved in hiPSC-CM alternative splicing.
RNA-seq analysis showed that MBNL1 was up-regulated in MTs assembled using hiPSC W156X/R225W -and hiPSC corr/R225W -CMs compared with 2D hiPSC-CMs cultures ( Figure 6A and B). Although we found MBNL2 was also up-regulated in MTs (see Supplementary material online, Figure S6D), scRNA-seq from our dataset previously published showed that MBNL1 but not MBNL2 was up-regulated in the CM population of MTs compared with monolayer hiPSC-CMs (see Supplementary material online, Figure S6C), suggesting a major role of MBNL1 specifically in CMs. In addition, MTs displayed skipping of MBNL1 exon 5, which is promoted by MBNL1 itself in adult CMs 44 ( Figure 6B) and thus suggesting that MBNL1 is functional in MTs.
To test whether MBNL1 is necessary and sufficient to promote exon 6B inclusion, we overexpressed or knocked out MBNL1 in hiPSC-CMs.
First, we transiently transfected hiPSC-CMs with an mRNA construct encoding MBNL1 and eGFP. Transfection efficiency reached around 50%, as shown by FACS analysis of eGFP+ cells ( Figure 6C) and overexpression of MBNL1 in transfected hiPSC-CMs compared with untransfected cells was confirmed by qPCR ( Figure 6D). ddPCR showed that the exon 6B fraction was increased in MBNL1-transfected hiPSC-CMs to levels comparable with those in MTs ( Figure 6E). We then knocked out MBNL1 in hiPSCs (MBNL1-KO hiPSCs) (see Supplementary material online, Figure S8). Cardiac in vitro differentiation was not affected in MBNL1-KO hiPSCs, but MBNL1 expression in 2D hiPSC-CMs was negligible and did not differ between unexcised and MBNL1-KO lines ( Figure 6D). When included in MTs, MBNL1-KO hiPSC-CMs did not up-regulate MBNL1, differently from the unexcised line ( Figure 6D). Moreover, exon 6B fraction in MBNL1-KO MTs was comparable to the 2D cultures by ddPCR analysis, and significantly lower than in MTs composed by unexcised hiPSC-CMs ( Figure 6E).
Taken together these results demonstrated that 3D cardiac MT constructs composed of hiPSC-cardiac ECs and -CFs promoted alternative splicing maturation in hiPSC-CMs such that they displayed increased expression of SCN5A exon 6B and its regulator MBNL1. This was sufficient to reveal the functional defects caused by the p.R225W mutation on I Na and to dissect the different contributions of p.W156X and p.R225W Nav1.5 mutations to the disease in the patient.

Discussion
The mammalian heart undergoes extensive functional modification after birth, regulated by changes in gene expression at both transcriptional and post-transcriptional level. 45 Alternative splicing is an important mechanism in the post-natal transition and applies to many cardiac genes, 8,46 including the cardiac sodium channel Nav1.5. 6,29 The switch between the foetal SCN5A isoform containing exon 6A and the adult SCN5A isoform containing exon 6B is completed postnatally in mice and humans. 42,47 Immature hiPSC-CMs predominantly express the foetal SCN5A transcript. Although this does not preclude evaluating the functional effects of SCN5A mutations using hiPSC-CMs in general, it does represent a hurdle for those mutations located in exon 6B which is expressed only postnatally. Here, we studied a patient who developed severe cardiac conduction defects early after birth and carried compound mutations in Nav1.5 (p.W156X and p.R225W), one of which (p.R225W) located in the adult exon 6B. Clinical characterization of the family indicated that the cardiac disorder was present only in individuals with compound heterozygosity for the two mutations, with individuals carrying only one of the two Nav1.5 mutations being asymptomatic. 31 This suggests an additive effect of the two mutations. To dissect the contribution of each mutation, we generated a hiPSC line from the patient and an isogenic corrected line carrying only the mutation in SCN5A exon 6B (hiPSC corr/R225W ). However, when we recorded I Na in hiPSC corr/R225W -CMs, current density was similar to that measured in WT hiPSC-CMs, whereas the compound mutant showed a reduction of I Na density similar to that previously observed in hiPSC-CMs with only p.W156X mutation. 13 The lack of effect of p.R225W contrasts with the strong reduction in current density shown in Xenopus laevis oocytes overexpressing the same p.R225W Nav1.5 mutation 31 ; this could be partly due to differences between heterologous expression systems and human cells. Interestingly, the single mutation effect in patients was negligible, suggesting that there might be differences between in vitro systems and the patient phenotype. Analysis of SCN5A exon expression revealed that only a small fraction of exon 6B-containing transcripts was expressed in hiPSC-CMs, in line with previous reports in independent hiPSC lines. 29 This explained the functional results and confirmed that the immature hiPSC-CM phenotype precludes analysis of mutations located in the adult SCN5A isoform.
Remarkably, while genetic excision of Scn5a exon 6B resulted in selective re-expression of foetal Nav1.5 in mice, 48 our complementary approach of SCN5A exon 6A excision in hiPSCs did not lead to selective expression of the adult Nav1.5. Indeed, 6A-KO hiPSC-CMs expressed similar (low) amounts of SCN5A exon 6B compared with the original non-excised lines and presented negligible I Na . This demonstrates that the exon 6B expression levels in standard monolayer hiPSC-CM cultures are not sufficient to generate I Na and that I Na measured in non-excised hiPSC-CMs is mainly conducted by the foetal channel. The extremely low (or negligible) I Na in 6A-KO hiPSC-CMs is likely a consequence of very few functional channels formed, because 6A excision interferes with the splicing machinery, as shown by the inclusion of intronic regions and exclusion of exon 5. This indicates that some regulatory sequences important for correct splicing might be present either within the intronic region removed by the excision, or in exon 6B itself, driving its own exclusion during splicing. Further studies with consecutive removal of different sequences within the 6A region are needed to elucidate the exact mechanisms underlying the regulation of SCN5A splicing. However, the switch between exon 6A and exon 6B is likely to need not only dedicated regulatory sequences, but also specific factors which may be lacking in immature CMs.
We and others developed different systems to improve the maturation of hiPSC-CMs using more physiological environments involving 3D engineering and/or co-culture with other (cardiac) cell types. 49,50 We previously showed that hiPSC-CMs in our cardiac tri-cell-type MTs undergo maturation at the functional, structural, metabolic, and gene expression levels 30 ; however, we did not examine whether it involved alternative splicing transition. Both ddPCR and RNA-seq revealed that in cardiac MTs there is a significant increase in the fraction of SCN5A exon 6B-vs. exon 6A-containing transcripts. A similar switch between SCN5A exon 6A and exon 6B was also previously promoted by culturing hiPSC-CMs in monolayer for extended periods. 29 Of note, cardiac MT maturation is less time-consuming than prolonged monolayer culture (40 days vs. more than 60), and the number and ratio of cells in MTs remain unchanged during the 21 days of culture required for maturation. 30 In our experience, long-term culture of hiPSC-CMs was accompanied by some cell loss, the proliferation of non-cardiomyocyte cells and the risk of detachment of the beating monolayer from the dish. Importantly, cardiac MTs showed higher expression of the exon 6B fraction (five-to sixfold, MT vs. 20 days 2D hiPSC-CMs) compared with long-term monolayer hiPSC-CM culture (two-to three-fold, 60 days vs. 20 days). 29 Nevertheless, levels of exon 6B expression comparable to adult human heart were not reached, suggesting that splicing regulation may still differ between in vitro and in vivo conditions. It remains to be investigated whether the effects of exon 6B mutation are revealed in other 3D or monolayer maturation systems or media.
The possibility to derive all cardiac MT cell components from hiPSCs allowed us to include either mutant, corrected isogenic, or WT hiPSC-CMs while the other cell types derived from the same (WT) stock, avoiding possible confounding factors. Since hiPSC-CM maturation achieved in MTs is maintained after dissociation, 30 we could analyse the electrical properties of single CMs. This revealed the contribution of p.R225W Nav1.5 mutation to the disease phenotype, uncovering the gene-dosage relationship and linking the presence of one or two mutated SCN5A alleles with a mild (≏45%) or severe (≏70%) I Na reduction, respectively. Reduced I Na density but no differences in the channel kinetics of activation or inactivation were measured in the p.R225W hiPSC-CMs compared with the corrected or WT hiPSC-CMs, much as has been observed for another mutation (p.I230T) in the same exon. 29 A positive shift in the voltage dependence of I Na activation was instead measured in the double mutant hiPSC-CMs, similar to the effects of the homozygous p.I230T Nav1.5 mutation. 29 This shift can be attributed to the p.R225W mutation carried on the maternal-and only functional -allele, since the p.W156X mutation likely generates a non-functional protein, due to the stop codon in the paternal allele. In hiPSC corr/ R225W , this change was instead masked by the corrected (and functional) allele. A shift in the voltage dependence of p.R225W Nav1.5 was also observed in transfected X. laevis oocytes 31 ; however, this was more pronounced and involved both activation and inactivation, underscoring the fact that ion channel properties are influenced by species-specific cellular context (for example ancillary proteins/subunits).
The comparison between immature hiPSC-CMs and more mature CMs derived from MTs allowed us to investigate the mechanism underlying the switch between foetal and adult SCN5A isoforms. Mbnl1 was previously identified as necessary for exon 6B expression in mice, as the Mbnl1-knockout mice expressed only the foetal isoform of SCN5A in the heart. 7,10 Here, we showed that MBNL1 expression is low in 2D hiPSC-CM monotypic cultures, while its expression increases significantly when CMs are included in MTs. Interestingly, only MBNL1 was specifically up-regulated in the CM population, and not MBNL2, another member of the same family. MBNL1 is also characterized by different developmentally regulated splicing isoforms and it is known to regulate its own splicing. 44,51 In particular, exon 5 is included in the embryonic isoform and is gradually excluded postnatally. 51,52 In MTs, we observed reduced expression of MBNL1 exon 5 compared with monolayer hiPSC-CMs, suggesting a functional change in MBNL1. Overall this indicates that MBNL1 is up-regulated and alternatively spliced in mature hiPSC-CMs and it might be specifically required to induce the switch in SCN5A isoform. Indeed, transient overexpression of MBNL1 increased SCN5A exon 6B fraction in hiPSC-CMs and lack of MBNL1 prevented splicing isoform switch in MTs. This confirms a key role for this factor in regulating SCN5A splicing also in human cells. Since we performed transient expression of MBNL1 in hiPSC-CMs, it remains to be elucidated whether the sole expression of the adult MBNL1 is sufficient to promote functional maturation of Nav1.5. Moreover, further studies are needed to understand how MBNL1 overexpression and knockout influence the expression of other target genes beside SCN5A. Indeed, Mbnl1 was shown to regulate the post-natal isoform switch of many other genes during mouse heart development 7-9 and to repress human stem cell pluripotency while promoting differentiation. 53,54 The expression of adult MBNL1 in hiPSC-CMs within the MT opens the possibility to study other developmentally/postnatally regulated genes using hiPSC-derived cells. The alternative splicing of the region of the voltage-sensing S4 segment is conserved also in voltage-gated sodium channels expressed in the brain, generating different isoforms underlying functional differences between foetal and adult tissues. 55 It is tempting to speculate that promoting hiPSC maturation in 3D organspecific tissues might facilitate the study of adult disease phenotypes in vitro in other organs.
In conclusion, our data demonstrate that a post-natal maturation of hiPSC-CMs is required to express the adult SCN5A isoform through MBNL1 regulation and thus reveal mutation contributions, allowing dissection of ionic current changes that cause adult arrhythmic disease phenotypes in humans.