Splicing misregulation of SCN5A contributes to cardiac-conduction delay and heart arrhythmia in myotonic dystrop

ARTICLE Received 2 Jun 2015 |Accepted 16 Feb 2016 |Published 11 Apr 2016 Splicing misregulation of SCN5A contributes to cardiac-conduction delay and heart arrhythmia in myotonic dystrophy Fernande Freyermuth1,*,w,Fre ´de ´rique Rau2,*, Yosuke Kokunai3, Thomas Linke4, Chantal Sellier1, Masayuki Nakamori3, Yoshihiro Kino5, Ludovic Arandel2, Arnaud Jollet2, Christelle Thibault1, Muriel Philipps1, Serge Vicaire1, Bernard Jost1, Bjarne Udd6,7,8, John W. Day9, Denis Duboc10, Karim Wahbi10, Tsuyoshi Matsumura11, Harutoshi Fujimura11, Hideki Mochizuki3, Franc¸ois Deryckere12, Takashi Kimura13, Nobuyuki Nukina14, Shoichi Ishiura15, Vincent Lacroix16, Amandine Campan-Fournier17, Vincent Navratil18, Emilie Chautard19, Didier Auboeuf19, Minoru Horie20, Keiji Imoto21, Kuang-Yung Lee22, Maurice S. Swanson23, Adolfo Lopez de Munain24, Shin Inada25, Hideki Itoh20, Kazuo Nakazawa25, Takashi Ashihara20, Eric Wang23, Thomas Zimmer4, Denis Furling2, Masanori P. Takahashi3& Nicolas Charlet-Berguerand1 Myotonic dystrophy (DM) is caused by the expression of mutant RNAs containing expanded CUG repeats that sequester muscleblind-like (MBNL) proteins, leading to alternative splicing changes. Cardiac alterations, characterized by conduction delays and arrhythmia, are the second most common cause of death in DM. Using RNA sequencing, here we identify novel splicing alterations in DM heart samples, including a switch from adult exon 6B towards fetal exon 6A in the cardiac sodium channel, SCN5A. We find that MBNL1 regulates alternative splicing of SCN5A mRNA and that the splicing variant of SCN5A produced in DM presents a reduced excitability compared with the control adult isoform. Importantly, reproducing splicing alteration of Scn5a in mice is sufficient to promote heart arrhythmia and cardiac-conduction delay, two predominant features of myotonic dystrophy. In conclusion, misregulation of the alternative splicing of SCN5A may contribute to a subset of the cardiac dysfunctions observed in myotonic dystrophy. DOI: 10.1038/ncomms11067 OPEN 1Department of Translational medicine and neurogenetics, IGBMC, CNRS UMR7104, INSERM U964, Universite ´de Strasbourg, Illkirch 67400, France. 2Sorbonne Universite ´s UPMC Univ Paris 06, Inserm, CNRS, Centre de Recherche en Myologie UMRS974/FRE3617, Institut de Myologie, GH Pitie ´-Salpe ˆtrie `re, Paris 75013, France. 3Department of Neurology, Osaka University Graduate School of Medicine, Osaka 565-0871, Japan. 4Department of Physiology, Friedrich Schiller University Hospital, Jena 07743, Germany. 5Department of Bioinformatics and Molecular Neuropathology, Meiji Pharmaceutical University, Kiyose 205-8588, Japan. 6Neuromuscular Research Center, Tampere University and University Hospital, Tampere 33520, Finland. 7Department of Medical Genetics, Folkha ¨lsan Institute of Genetics, Helsinki University, Helsinki 00250, Finland. 8Department of Neurology, Vaasa Central Hospital, Vaasa 65130, Finland. 9Department of Neurology, Stanford University, Stanford, California 94304, USA. 10 Service de Cardiologie, Universite ´Paris-Descartes, Ho ˆpital Cochin, AP-HP, Paris 75014, France. 11 Department of Neurology, Toneyama National Hospital, Toyonaka 560-8552, Japan. 12 CNRS UMR7175, Ecole Supe ´rieure de Biotechnologies de Strasbourg, Illkirch 67400, France. 13 Division of Neurology, Hyogo Medical College, Nishinomiya 663-8501, Japan. 14 Laboratory of Structural Neuropathology, Doshisha University Graduate School of Brain Science, Kyoto 610-0394, Japan. 15 Graduate School of Arts and Sciences, University of Tokyo, To kyo 15 3 - 8 9 0 2, Japan . 16 Universite ´Lyon 1, CNRS, UMR5558 LBBE, Villeurbanne 69622, France. 17 Hospices civils de Lyon, Laboratoire de cytoge ´ne ´tique constitutionelle, Bron 69500, France. 18 Po ˆle Rho ˆne Alpes de Bioinformatique, Universite ´Lyon 1, Ba ˆtiment Gregor Mendel, Villeurbanne 69100, France. 19 Centre de Recherche en Cance ´rologiedeLyon,Lyon69373,France.20 Department of Cardiovascular and Respiratory Medicine, Shiga Medical University, Otsu 520-2192, Japan. 21 Department of Information Physiology, National Institute for Physiological Sciences, Okazaki 444-8585, Japan. 22 Department of Neurology, Chang Gung Memorial Hospital, Keelung 20401, Taiwan. 23 Department of Molecular Genetics and Microbiology, Center for NeuroGenetics and the Genetics Institute, University of Florida, College of Medicine, Gainesville, Florida 32610, USA. 24 Department of Neurology, Hospital Universitario DONOSTIA, Neuroscience Area, Institute Biodonostia CIBERNED and University of Basque Country UPV-EHU, San Sebastia ´n 20014, Spain. 25 Laboratory of Biomedical Sciences and Information Management, National Cerebral and Cardiovascular Center Research Institute, Osaka 565-8565, Japan. * These authors contributed equally to the work. wPresent address: Massachusetts General Hospital, MassGeneral Institute for Neurodegenerative Diseases, Charlestown, Massachusetts 02129, USA. Correspondence and requests for materials should be addressed to D.F. (email: [email protected]) or to M.P.T. (email: [email protected]d.osaka-u.ac.jp) or to N.C-B. (email: [email protected]r). NATURE COMMUNICATIONS | 7:11067 | DOI: 10.1038/ncomms11067 | www.nature.com/naturecommunications 1 Myotonic dystrophy (DM), the most common adult-onset muscular dystrophy, includes two genetically distinct forms. DM of type 1 (DM1) and its severe congenital form (CDM1) are caused by an expansion of CTG repeats in the 30-untranslated region (UTR) of the DMPK gene1–3. In contrast, DM of type 2 (DM2) is caused by an expansion of CCTG repeats within the first intron of the CNBP (also known as ZNF9) gene4. The pathogenesis of DM involves a RNA gain-of-function mechanism caused by expression of mutant RNAs containing hundred to thousands of CUG or CCUG repeats that interfere with the splicing of other pre-mRNAs through dysfunction of two classes of RNA-binding proteins. MBNL proteins (MBNL1, MBNL2 and MBNL3) are sequestered within nuclear RNA foci formed by expanded CUG and CCUG repeats5,6, whereas expression and phosphorylation of CUG-binding protein 1 (CUGBP1, encoded by the CELF1 gene) are increased in DM1 heart samples7. MBNL and CUGBP1 proteins regulate alternative splicing, and alterations of their functional levels in myotonic dystrophic tissues results in reversion to fetal splicing patterns for several mRNAs, such as the insulin receptor (INSR) (ref. 8), the muscle chloride channel (CLCN1) (refs 9,10), dystrophin (DMD) (refs 11,12) and key components of the skeletal muscle excitation–contraction coupling process, including amphiphysin2 (BIN1) (ref. 13), ryanodine receptor 1 (RYR1) (ref. 14), sarcoplasmic/endoplasmic reticulum Ca2þ-ATPase SERCA1 (ATP2A1) (ref. 14) and the muscle calcium channel Ca V 1.1 (CACNA1S) (ref. 15). Misregulation of the alternative splicing of the insulin receptor INSR,CLCN1 and DMD mRNAs are associated with the insulin resistance8, myotonia9,10,16 and dystrophic process12, respectively, while alterations of the alternative splicing of BIN1,RYR1,ATP2A1 and CACNA1S may contribute to the skeletal muscle weakness observed in DM13–15. In contrast, the molecular mechanisms underlying the cardiac defects, which affect 80% of individuals with DM and represent the second most common cause of death in this disease17,18, are yet to be defined. Cardiac involvements in DM are characterized by cardiac-conduction delay that may result in fatal atrioventricular block, and by atrial or ventricular tachycardia17,18. Electrocardiography (ECG) analyses in DM patients indicate prolonged conduction time from the sinoatrial node to the ventricles (PR interval) and elongated ventricular depolarization (QRS duration). Interestingly, cardiac dysfunctions in DM are reminiscent in some aspect to an alteration of the cardiac sodium current. The a-subunit of the cardiac voltage-gated Na þ channel, Na v 1.5, is encoded by the SCN5A gene and plays a key role in the excitability of cardiomyocytes and for rapid propagation of the impulse through the cardiac-conduction system. Mutations in SCN5A lead to a variety of arrhythmic disorders, including long QT3, progressive and non-progressive cardiac-conduction disease (also known as Lev-Lene `gre disease), atrial fibrillation, sick sinus syndrome, Brugada syndrome and numerous overlapping syndromes19–21. Using transcriptomic approaches, we identified various novel splicing changes in heart samples of DM1 individuals. Analysis of the RNA motifs enriched in the vicinity of these misregulated exons indicates that sequestration of the MBNL proteins is probably the main cause of splicing misregulation in heart of individuals with DM. Among these novel splicing alterations, we focused on misregulation of alternative splicing of the SCN5A pre-mRNA. This splicing alteration results in expression of a fetal isoform of SCN5A with altered electrophysiological properties. Of importance, we demonstrate that reproducing the splicing alteration of Scn5a in mouse is sufficient to cause heart arrhythmia and cardiac-conduction delay with elevated PR interval, which are key characteristics of the heart alterations observed in DM. These results suggest that altered splicing of SCN5A mRNA may participate to the electrical cardiac abnormalities observed in DM. Results Identification of splicing changes in DM heart samples.To determine novel splicing abnormalities in DM heart samples, we first used whole-genome microarrays (GeneChip Human Exon 1.0 ST array) on polyadenylated RNA extracted from left ventricle samples of three adult DM1 patients compared with three agematched control individuals. Bioinformatic analyses predicted significant (Fold Change Z2, Sudent t-test, Pvalue r0.01) changes in the splicing of 24 exons between control and DM1 samples (Supplementary Table 1), including a misregulation of the alternative splicing of the SCN5A pre-mRNA. To extend this analysis, we performed paired-end RNA sequencing (RNA-seq) on the same DM1 and control heart samples, yielding 1,611 million of mapped 100 bp reads. DESeq and Cuffdiff were then applied to estimate differential gene expression and over or under-expressed mRNAs were selected by using the Benjamini and Hochberg adjusted Pvalues (false discovery rate (FDR) r0.1). A total of 9 and 19 upregulated genes were predicted differentially expressed with DESeq and Cuffdiff, respectively, but none were confirmed by quantitative real-time RT-qPCR analyses. This low number of differentially expressed mRNAs suggests that cardiac pathology in DM is not associated with drastic modifications of gene expression levels. In contrast, DEXSeq (ref. 22), which tests differential exon usage between two conditions, predicted 134 significant (Log2 Fold Change Z1.2, FDR r0.1) alternative splicing changes between control and DM1 heart samples (Supplementary Data 1). Similarly, MISO (ref. 23) analysis, which computes the fraction of mRNA that includes a given cassette alternative exon, predicted 259 significant (DPSI Z0.3; Z-score Z1.2) alternative splicing changes between control and DM1 heart samples (Fig. 1a and Supplementary Data 2), including a robust misregulation of the alternative splicing of SCN5A (Fig. 1b). MISO and DEXSeq predictions overlapped, but with some exceptions, such as the skipping of the consecutive exons 18, 19 and 20 of CAMK2B predicted by DEXSeq but not by MISO; or the shift of SCN5A exon 6B towards exon 6A identified by MISO but not by DEXSeq. These differences are inherent to their computation models, since MISO does not detect alterations of successive exons and DEXSeq does not identify mutually exclusive exons, highlighting that MISO and DEXSeq are complementary bioinformatics approaches. Next, we tested by PCR with reverse transcription (RT–PCR) forty candidate mRNAs having the highest probability of misregulation in DEXSeq and/or MISO analyses. We validated splicing alterations for 32 of them, including some that have been identified in previous studies (TNNT2, TNNT3,ABLIM1, LDB3, MBNL1,CAMK2B,MAPT and so on)24–26, and 20 others that represent, to the best of our knowledge, novel alterations of alternative splicing (ADD3,GOLGA4,CRTC2,ARHGEF10L, ANK3,DCLK2,EPN2,UNC13B,TECR,ARVCF,SOCS7,CELF1 and so on) in DM1 heart samples (Fig. 1c). Of interest, some of these splicing alterations may be of pathological consequence in DM. For example, knockout of the Socs7 gene in mouse results in insulin resistance27. Whether the splicing misregulation of SOCS7 in DM contributes to insulin resistance remains to be tested. Also, RNA sequencing predicts an increased retention of the penultimate intron of FCGRT, which encodes the Fc fragment of the IgG receptor transporter a(FCRN) protein, involved in IGG recycling28. Whether splicing alteration of FCGRT in DM is responsible to the decreased level of IGG in blood of these patients is an attractive hypothesis that remains to be tested. ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11067 2NATURE COMMUNICATIONS | 7:11067 | DOI: 10.1038/ncomms11067 | www.nature.com/naturecommunications Finally, RNA sequencing predicts misregulation of the alternative splicing of a cardiac-specific exon located in the 50-UTR of CELF1, which encodes CUGBP1. Whether this alternative splicing may contribute to the increase levels of CUGBP1 protein observed in DM1 hearts remains also to be evaluated. Splicing altered in DM are enriched for MBNL-binding sites. Mutants RNAs containing expanded CUG or CCUG repeats interfere with the functional levels of CUGBP1 and MBNL proteins. Earlier studies determined that MBNL proteins bind to YGC RNA motifs (where Y is a pyrimidine)29–32, while CUGBP1 binds to UGU-enriched sequences33,34. To determine whether these RNA motifs are indeed present in the vicinity of exons misregulated in DM, we determined all 4-mer RNA motifs enriched within, upstream or downstream of the exons predicted as misregulated by MISO in DM1 heart samples, compared with 2,000 control exons (Fig. 2). Most RNA motifs significantly enriched (binomial test, Pvalue o1.0 107) contained YGC sequences, while none were found to contain UGU sequences. Furthermore, YGC sequences were enriched upstream of exons abnormally included in DM1, while YGC motifs were enriched downstream of exons repressed in DM1. These results matched the MBNL splicing regulatory map determined by CLIP experiments, where binding of MBNL upstream of an exon tends to inhibit exon inclusion whereas binding of MBNL downstream Exons excluded in DM Exons included in DM Z score 1 –1 –0.5 0 Δ PSI 0.5 1 2 356A 6B 7 SCN5A SCN5A CTL DM1 6A 6B CTL DM1 SCN5A – 16 + 16 ADD3 CTL DM1 – 6 + 6 MYH11 CTL DM1 - 47 + 47 NCOR2 CTL DM1 – 6A + 6B TPM2 CTL DM1 – 2 + 2 TECR CTL DM1 100 200 100 200 100 200 300 200 200 300 200 100 CTL DM1 – 14 + 14 ABLIM1 CLTB – 6 + 6 CTL DM1 CRTC2 – 13 + 13 CTL DM1 NUMA1 – 20 + 20 CTL DM1 EPN2 CTL DM1 – 5 + 5 – 6 + 6 ZFYVE21 CTL DM1 100 200 200 300 300 200 200 100 300 200 400 200 100 – 8 + 8 CTL DM1 COPZ2 GOLGA4 CTL DM1 – 24 + 24 MXRA7 CTL DM1 – 4 + 4 ANK3 CTL DM1 – 40 + 40 UNC13B CTL DM1 – 38 + 38 ARVCF CTL DM1 – 19 + 19 100 200 200 300 200 100 200 100 200 100 100 200 ARHGEF10L DCLK2 CAMK2B SOCS7 CELF1 SUN1 – 18.19.20 – 10 + 10 – 8 + 8 – 5 + 5 – 5 + 5 – 2 + 2 + 18.19.20 + 19.20 CTL DM1 CTL DM1 CTL DM1 CTL DM1 CTL DM1 CTL DM1 100 400 200 300 100 200 100 200 100 200 100 200 300 200 300 100 100 300 300 500 300 300 300 100 300 400 30 20 10 30 20 10 30 20 10 30 20 10 30 20 10 30 20 10 38662462 RPKM RPKM RPKM RPKM RPKM RPKM 38658666 38654924 38651227 Genomic coordinate (chr3), “-” strand ab c Figure 1 | Identification of novel splicing misregulations in DM1 heart samples. (a)D-PSI versus Z-score plot of exon cassettes misregulations predicted by MISO analysis. (b) Exons structure and coverage of RNA-seq reads across SCN5A exons 5–7 show increased inclusion of exon 6A and decreased inclusion of exon 6B in heart samples of three DM1 patients (bottom, blue) versus three control samples (top, red). (c) Validation by RT–PCR of RNA-seq predictions in human heart samples of normal adult individuals (CTL, black) versus adult DM1 patients (DM1, red). Molecular size markers in bps are reported to the left of each RT–PCR gels. bp, base pair. NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11067 ARTICLE NATURE COMMUNICATIONS | 7:11067 | DOI: 10.1038/ncomms11067 | www.nature.com/naturecommunications 3 of an exon generally stimulates exon inclusion35,36. In contrast, we found no enriched motifs for other RNA-binding proteins, including CUGBP1, rbFOX1, hnRNP H or Staufen. These results, as well as previous data36–38, support a model in which titration of MBNL proteins is the main cause of splicing change in DM1 heart, while misregulation of other RNA-binding proteins may contribute to a subset of splicing alterations. Splicing of SCN5A is misregulated in DM heart samples. Both microarray and RNA-seq predicted misregulation of alternative splicing of SCN5A pre-mRNA in DM1 heart samples. Splicing of SCN5A is developmentally regulated, such that exon 6A is included in fetal heart but rapidly replaced by exon 6B after birth39. Consequently, SCN5A exon 6A is named as embryonic or fetal, while exon 6B is known as adult. Exons 6A and 6B are mutually exclusive exons encoding part of the voltage sensor, segments 3 and 4 located in the domain I of the sodium channel (Fig. 3a,b). These are key segments for the electrical activity of the sodium channel, and inclusion of either fetal exon 6A or adult exon 6B results in channel isoforms, named, respectively, hNa v 1.5e and hNa v 1.5, with different electrophysiological properties39–41. We confirmed our microarray and RNA-seq predictions by RT–PCR and found that adult SCN5A exon 6B is partly replaced by its fetal exon 6A in heart samples of individuals with DM, including adult DM1 and adult DM2 cases (Fig. 3c). Note that to differentiate exon 6B from exon 6A that have the exact same length of 92 bp, we took advantage of a BstbI restriction site present only in exon 6A, which thus appears as a BstbI-digested doublet band in Fig. 3c. These results are consistent with the recent report of a splicing misregulation of SCN5A in one DM1 heart sample42. Although splicing of SCN5A is misregulated in DM1, we observed no correlation between the percentage of SCN5A exon 6A inclusion and the increased duration of the PR interval and only a very limited, if any, correlation between misregulation of SCN5A exon 6A splicing and alteration of the QRS duration in individuals with DM1 (R2of 0.2 with six DM1 samples; Supplementary Fig. 1). Misregulation of SCN5A splicing was specific to DM, as we did not observe inclusion of exon 6A in heart samples from individual affected with Duchenne muscular dystrophy (DMD), amyotrophic lateral sclerosis (ALS) or dilated cardiomyopathy (DCM) (Fig. 3d). Moreover, misregulation of splicing in DM1 was specific and not global, as we observed no splicing changes of SCN5A alternative exon 18, of CACNA1C mutually exclusive exons 8A and 8B, of KCNAB1 alternative exons 2 and 11, or of KCNQ1 alternative exons 2 and 5 (Supplementary Fig. 2). Finally, we observed no significant alteration of the expression level of SCN5A mRNA by quantitative real-time RT-qPCR (Fig. 3e). Overall, these results indicate a specific misregulation of alternative splicing of SCN5A resulting in expression of a fetal form of this channel in adult DM heart. These results are consistent with previous studies where alternative splicing changes in DM resume a MBNL-dependent fetal splicing pattern that persist in adult tissues26,37. Alternative splicing of SCN5A is regulated by MBNL1.To determine the mechanisms underlying misregulation of SCN5A splicing, we first determined its splicing pattern in cell models of DM. Since SCN5A is expressed at low level in culture of immature skeletal muscle cells, we investigated its splicing in primary cultures of differentiated skeletal muscle cells originating from muscle biopsies of control and DM1 individuals. RT–PCR experiments determined a switch of exon 6B towards exon 6A in DM1 muscle cells compared with control, reproducing the splicing alteration observed in cardiac tissue (Fig. 4a). Of technical interest, the basal level of exon 6A inclusion was higher in muscle cell cultures than in adult heart samples (compare Fig. 4a to Fig. 3c), which probably reflect the immature aspect of cell cultures. Since mutant RNAs containing expanded CUG or CCUG repeats interfere with alternative splicing through titration of MBNL proteins, we tested whether MBNL1 regulates SCN5A splicing. Reduction of MBNL1 expression through a siRNA-mediated approach in human control primary muscle cells mimicked the effect of CUG repeats and promoted a switch from adult exon 6B towards fetal exon 6A (Fig. 4b). Western blotting analysis confirmed the successful depletion of MBNL1 expression (Supplementary Fig. 3A). Next, we assessed alternative splicing of Scn5A in heart samples of Mbnl knockout mice43. RT–PCR analysis shows that inclusion of the exon 6A of Scn5a is increased in heart samples of mice with no Mbnl1 and reduced level of Mbnl2 (Mbnl1/,Mbnl2 þ/)(Fig. 4c). The increased inclusion of Scn5a exon 6A in Mbnl knockout mice is significant (Student t-test, Pvalue r0.01) but rather mild, probably reflecting difference in regulation of alternative splicing between human and mouse or the compensatory effect of residual Mbnl2 expression43. This hypothesis is consistent with the mild splicing alteration of Scn5A observed in the sole Mbnl1 knockout mice44. Overall, these results suggest that MBNL proteins regulate the alternative splicing of SCN5A exons 6A and 6B. To determine whether this regulation is direct or indirect, we constructed a minigene containing exons 6A and 6B of SCN5A bordered by their intronic regions. Expression of this construct in mouse C2C12 myoblasts reproduced a fetal pattern with mainly inclusion of exon 6A (Fig. 4d). Since inclusion of exon 6B was repressed, reduction of Mbnl1 activity through siRNA or expression of expanded CUG repeats had no further repressive effect on exon 6B. In contrast, expression of MBNL1 promoted a switch from fetal exon 6A towards adult exon 6B, while CCCC (2×10–18) CUGC (9×10–15) UGCU (2×10–14) CUAA (5×10–10) UUGC (5×10–8) CCUG (4×10–7) UGCC (7×10–7) UGCU (2×10–26) GCUU (2×10–19) CUGC (2×10–15) UUGC (2×10–9) CGCU (4×10–9) GCUC (7×10–7) GAAG (5.1 10–7) CCGC (2×10–22) CGCC (8×10–17) CGCU (9×10–10) UCGC (4×10–7) GCGC (7×10–7) CCCG (8×10–7) CUGC (2×10–11) UGCU (3×10–9) GCUG (6×10–8) Exons excluded in DM Exons included in DM Figure 2 | MBNL-binding motifs are enriched in vicinity of exons misregulated in DM1. Sequence and binomial test Pvalues of 4-mer RNA motifs enriched downstream, within and upstream of exons misregulated in DM1 heart samples. Sequences enriched in exons excluded in DM are indicated in red, while sequences enriched in exons included in DM are indicated in blue. ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11067 4NATURE COMMUNICATIONS | 7:11067 | DOI: 10.1038/ncomms11067 | www.nature.com/naturecommunications expression or siRNA-mediated depletion of CUGBP1 had no effect (Fig. 4d). Western blotting analysis confirmed that siRNA transfection efficiently reduced endogenous Mbnl1 or Cugbp1 expression (Supplementary Fig. 3B and C). Next, gel-shift assays determined that recombinant purified GST-tagged MBNL1 bound to UGC RNA motifs located upstream of exon 6A (Fig. 4e). Of interest, this UGC sequence is absent from the mouse genome, which may explain the mild splicing alteration of Scn5A observed in mice knockout for Mbnl proteins. Mutation of these UGC motifs abolished MBNL1 binding (Fig. 4f), as well as the regulatory effect of MBNL1 on a mutant SCN5A minigene (Fig. 4g). Overall, these results establish that MBNL1 regulates directly alternative splicing of SCN5A exons 6A/6B. SCN5A splicing forms present different electrical properties. SCN5A encodes Na v 1.5, the main cardiac voltage-gated sodium channel, and loss-of-function mutations in SCN5A lead to a variety of arrhythmic disorders, which share some common pathological features with DM. Furthermore, exons 6A and 6B differ at seven amino acid positions, resulting in channel variants with different electrophysiological properties39–41. To investigate the consequences of the switch from SCN5A exon 6B towards exon 6A observed in DM, we first examined in Xenopus oocytes the sodium currents generated by either hNa v 1.5e, the splice variant of SCN5A containing the fetal exon 6A, or hNa v 1.5, encoded by SCN5A containing the adult control exon 6B (Fig. 5a and Supplementary Table 2). Injection of RNA encoding hNav1.5e, which is the splicing isoform of SCN5A found in DM, indicated a significant reduction of the sodium current amplitude of 45%, compared with hNa v 1.5, the normal adult SCN5A exon 6B form (Fig. 5b,c). Since, the extent of splicing misregulation varies among DM individuals, which typically express a mix of SCN5A splicing forms containing either exon 6A or exon 6B (cf. Fig. 3c), we analysed sodium currents generated by a mix of both SCN5A isoforms (Fig. 5a). Injecting Xenopus oocytes with an equimolar mix of RNA encoding each channel, namely 50% of hNa v 1.5e (SCN5A containing fetal exon 6A) and 50% of hNa v 1.5 (SCN5A expressing adult exon 6B), resulted in a reduction of 30% of the current amplitude compared with the control hNa v 1.5 (Fig. 5b,c). Next, two-electrode voltage clamp recording experiments revealed that the steady-state activation of the fetal hNa v 1.5e was shifted by 7 mV towards depolarized potential compared with the control adult hNa v 1.5 form (Fig. 5d and Supplementary Table 2). This shift is consistent with the shift observed previously in transfected mammalian cells39–41, thus validating our approach in Xenopus oocytes. To better reproduce the situation observed in DM, we injected in Xenopus oocytes an equimolar mix of DM (hNa v 1.5e, fetal exon 6A) and control (hNa v 1.5, adult exon 6B) RNA isoforms of SCN5A. Importantly, this mix of splicing forms also presented a significant shift of steady-state activation towards depolarized potentials by 3.8 mV, compared with the control hNa v 1.5 form (Fig. 5d, Supplementary Table 2). Correspondingly, a similar shift Adult Fetal 5 76A 6B COOH NH2 IIIIIIIV Adult CTL Adult DM1 Adult DM2 Adult ALS Congenital DM1 Adult DM1 Adult DM2 Congenital DM1 Fetal CTL Exon 6B Exon 6A 0 Fetal CTL Adult CTL ALS DCM DMD 20 40 60 80 100 0.5 1 CTL DM1 0 100 200 300 bp BstBI 12345 6 12345 6 12345 6 12345 6 % Exon 6A inclusion SCN5A mRNA expression ab c de Figure 3 | Splicing of SCN5A exon 6A is altered in DM heart samples. (a) Schematic representation of mutually exclusive exons 6A and 6B of SCN5A. SCN5A mRNA includes exon 6A (red) in fetal heart, while SCN5A mRNA expresses exon 6B (blue) in adult heart. (b) Schematic representation of SCN5A topology expressing exon 6A (red). Exons 6A or 6B encodes part of segment 3, connecting loop between S3 and S4 and most part of the voltage-sensitive segment 4 of domain 1 of the sodium channel SCN5A. (c). Representative BstBI-digested RT–PCR analysis of endogenous SCN5A mRNA from human heart samples of normal adult (CTL), adult ALS, non-DM fetuses (20, 24 and 35 weeks), congenital DM1 fetuses (CDM1 of 22, 25 and 28 weeks), adults DM1 and DM2 individuals. Molecular size marker is indicated in bp. (d) Graphical representation of RT–PCR analysis depicting the percentage of SCN5A mRNA including exon 6A in left ventricular heart samples from fetal and adult control, ALS, DCM, DMD, CDM1 and adult DM1 and DM2 individuals. (e) Graphical representation of quantitative real-time RT-qPCR depicting the mRNA expression of SCN5A relative to RPLP0 in control normal adults (n¼5) versus adult DM1 (n¼5) heart samples. Bars indicate s.e.m. bp, base pairs. NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11067 ARTICLE NATURE COMMUNICATIONS | 7:11067 | DOI: 10.1038/ncomms11067 | www.nature.com/naturecommunications 5 was observed for the time constant of inactivation (Fig. 5e). Consistent with previous electrophysiological studies39–41,no significant differences were observed between hNa v 1.5 and hNa v 1.5e regarding steady-state inactivation and recovery from inactivation (Fig. 5f,g). Overall, our results are consistent with previous studies39–41, and demonstrate that hNa v 1.5e, the splicing form of SCN5A expressed in DM and containing the fetal exon 6A, presents a reduced excitability compared with hNa v 1.5, which is the adult control SCN5A isoform containing exon 6B. Alteration of SCN5A splicing leads to heart conduction defects. Misregulation of the alternative splicing of SCN5A in DM is one alteration identified among many others, thus questioning the CMV 6A 6B 5 CTL DM1 siCTL siMBNL1 % Exon 6B 0 20 40 60 *** % Exon 6B 0 20 40 60 6B *** UUUGCUAUGCUGUGCUAUGCCUUGCAG MBNL1 Free Bound SCN5A minigene WT UUU_CUAU_CUGU_CUAU_CCUU_CAG MBNL1 Free Bound SCN5A minigene MUT XXX % Exon 6B 0 20 40 60 6B CMV 6A 6B 5 % Exon 6A ** 0 10 20 30 6B 6A 6B 6A % Exon 6B 0 20 40 60 *** 6B 6A CTL #6A 6A 100 200 300 bp 100 bp 200 100 200 300 bp 100 200 bp Control Control CUG 960x CUG 960x MBNL1 MBNL1 CUGBP1 siMbnl1 siMbnl1 siCelf1 100 200 bp Mbnl1–/– Mbnl2+/– PolyA PolyA abc de fg Figure 4 | MBNL1 regulates alternative splicing of SCN5A.(a) Upper panel, RT–PCR analysis of endogenous SCN5A mRNA from differentiated primary muscle cell cultures derived from biopsies of control or DM1 individuals. (lower) Quantification of the percentage of SCN5A mRNA including exon 6B. (b, upper) RT–PCR analysis of endogenous SCN5A mRNA from human differentiated cultures of control primary muscle cells transfected with a scrambled siRNA (siCTL) or a siRNA targeting MBNL1 mRNA (siMBNL1). (lower) Percentage of SCN5A mRNA including exon 6B. (c, upper) RT–PCR analysis of endogenous Scn5a mRNA in heart samples of wild-type and compound Mbnl1/,Mbnl2þ/double knockout mice. (lower) Percentage of Scn5a mRNA including exon 6A. (d, upper) RT–PCR analysis of exogenous SCN5A mRNA from differentiated C2C12 muscle cells co-transfected with a SCN5A minigene containing exons 6A and 6B bordered by their introns and with either a plasmid expressing 960 CTG repeats, MBNL1, CUGBP1 or with a siRNA directed against Mbnl1 (siMbnl1)orCelf1 (encoding Cugbp1; siCelf1). # Indicates usage of a cryptic splice site inherent to the minigene. (lower) Percentage of SCN5A mRNA including exon 6B. (e, upper) Schematic representation of SCN5A minigene, including the UGC-rich sequence used for binding assays. (lower) Gelshift assays were performed using 5–1,000 nM of purified bacterial recombinant GST-MBNL1D101 and a uniformly 32P-CTP labelled RNA. (f, upper) Schematic representation of mutant SCN5A minigene, including the mutant sequence, used for binding assays. (lower) Gel-shift assay performed as in e.(g, upper) RT–PCR analysis of exogenous SCN5A mRNA from differentiated C2C12 muscle cells co-transfected with mutant SCN5A minigene and with a plasmid expressing 960 CTG repeats or MBNL1 or with a siRNA directed against Mbnl1 (siMbnl1). (lower) Percentage of SCN5A mRNA including exon 6B. All transfection and gel-shift experiments were repeated three to five times. Molecular size markers are indicated in bp. Bars indicate s.e.m. Student test, ** indicates Po0.01, *** indicates Po0.001. bp, base pairs. ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11067 6NATURE COMMUNICATIONS | 7:11067 | DOI: 10.1038/ncomms11067 | www.nature.com/naturecommunications contribution of SCN5A misregulation to the cardiac symptoms observed in DM. To test the physiological importance of SCN5A splicing misregulation, we artificially forced the switch from adult exon 6B towards fetal exon 6A into wild-type adult mouse heart using an exon-skipping strategy (Fig. 6a). To insure efficient transduction of the cardiac muscle and continuous expression of nuclear antisense oligonucleotides, we engineered and produced adeno-associated virus (AAV2/9) expressing optimized U7-snRNA fused to Scn5a antisense sequences (U7-ASScn5a). Splicing analysis revealed that combination of two U7-AS constructs, spanning intron 6/exon 6B junction and exon 6B of Scn5A, promoted a switch from inclusion of adult exon 6B towards inclusion of the fetal exon 6A (Supplementary Fig. 4A). Thus, AAV2/9 expressing both U7-AS constructs 2 ms 1 μA 0 1 2 3 4 *** *** Inactivation time constant (ms) Voltage (mV) –40 –30 –20 –10 0 10 0 2 4 6 8 10 12 14 Steady-state activation Voltage (mV) –60 –50 –40 –30 –20 –10 0 10 20 0.0 0.2 0.4 0.6 0.8 1.0 hNav1.5 hNav1.5e 50% hNav1.5 + 50% hNav1.5e Volta g e (mV) 0.0 0.2 0.4 0.6 0.8 1.0 –130 –110 –90 –70 –50 –30 Steady-state inactivation 0.0 0.2 0.4 0.6 0.8 1.0 Fractional recovery –60 –40 –20 0 20 Voltage (mV) –4 –3 –1 –2 0 –5 hNav1.5 hNav1.5e 50% hNav1.5 + 50% hNav1.5e hNav1.5 hNav1.5e 50% hNav1.5 + 50% hNav1.5e hNav1.5 hNav1.5e 50% hNav1.5 50% hNav1.5e 40 60 Recovery interval Δt (ms) 0 20406080100 50% hNav1.5 + 50% hNav1.5e hNav1.5e (SCN5A exon 6A)hNav1.5 (SCN5A exon 6B) Peak current (μA) Current (μA) hNav1.5 hNav1.5e 50% hNav1.5 + 50% hNav1.5e hNav1.5 hNav1.5e 50% hNav1.5 + 50% hNav1.5e fg a bc de Figure 5 | Electrophysiological properties of hNa v 1.5 and hNa v 1.5e channels. (a) Representative Naþcurrents generated in Xenopus oocytes by hNa v 1.5 (encoded by SCN5A containing the adult exon 6B), hNa v 1.5e (encoded by SCN5A including the fetal exon 6A), and simultaneously expressed Na v 1.5 and Na v 1.5e channels at a 1:1 ratio. (b) Peak current amplitudes at the test potential of 10 mV in Xenopus oocytes injected with equimolar amount of cRNA encoding hNa v 1.5, hNa v 1.5e or 1:1 combination of Na v 1.5 and Na v 1.5e channels. (c) Current–voltage relationships. (d) Steady-state activation curves. (e) Inactivation time constants th (ms) at different test pulses. (f) Steady-state inactivation curves. (g) Fractional recovery curves. Data were obtained from 11 different batches of oocytes. To illustrate steady-state activation, steady-state inactivation and recovery from inactivation, we used 3–5 representative measurements. For total number of measurements (n¼25–27) and for statistical data evaluation (Vm, s) see the Supplementary Table 2. Bars indicate s.e.m. Student test, *** indicates Po0.001. NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11067 ARTICLE NATURE COMMUNICATIONS | 7:11067 | DOI: 10.1038/ncomms11067 | www.nature.com/naturecommunications 7 (AAV-U7-ASScn5a) were injected systemically into newborn wild-type mice and cardiac functions were investigated 4 and 6 months post injection. Control animals injected either with saline or empty AAV2/9 presented no splicing alterations of Scn5a and normal cardiac functions. In contrast, mice injected with AAVU7-ASScn5a presented a decreased inclusion of adult exon 6B with a concomitant 30–40% increase of the inclusion of exon 6A, thus reproducing the situation observed in DM (Fig. 6b). Quantitative RT–PCR demonstrated no changes in the expression of Scn5a mRNA or of its associated subunit Scn1b between controland AAV-U7-ASScn5a-injected mice (Fig. 6c). Importantly, AAV-U7-ASScn5a-injected mice reproduce some of the key AdultFetal 5 76A 6B 5 76A Forced inclusion of exon 6A in adult % Exon 6A CTL U7-ASScn5a 0 20 40 60 *** c mRNA Scn1b Gja1Scn5a CTL U7-ASScn5a 0 0.5 1 2 1.5 Control U7-ASScn5a Col3a1 mRNA 0 0.5 1 2 1.5 T g fb1Cola1a CTL U7-ASScn5a * ** QT 0 10 20 30 50 40 PR 0 10 20 30 50 40 *** ms CTL U7-ASScn5a mn 150 75 0 225 0.5 1 1.5 20 0.5 1 1.5 20 RR RR 150 75 0 225 CTL U7-ASScn5a 6B 6A P QRS P QRS 25 ms CTL U7-ASScn5a P QRS P QRS CTL U7-ASScn5a g 0 5 10 20 15 QRS 0.058 100 200 300 bp Scn5A antisense sequences a b de f hi Figure 6 | Alteration of Scn5a splicing causes heart conduction defects and arrhythmias. (a) Schematic representation of mutually exclusive exons 6A and 6B of Scn5a and of antisense sequences driven by optimized U7-snRNAs (U7-ASScn5a) to force fetal exon 6A inclusion in adult wild-type mouse heart. (b, upper) RT–PCR analysis of the alternative splicing of endogenous Scn5a mRNA from heart samples of mice injected with AAV2/9 expressing U7-ASScn5a compared with control injected mice. Molecular size marker is indicated in bp. (lower) Percentage of Scn5a mRNA including exon 6A. (c) Realtime RT-qPCR quantification of the expression of Scn5a, Scn1b and GJja1 (connexin 43) mRNAs in heart samples of mice expressing U7-ASScn5a (n¼6) compared with control injected mice (n¼6). (d) Representative ECG traces show prolongation of the PR interval in U7-ASScn5a-injected mice compared with control mice. (e) ECG measures of PR interval, QRS and QT intervals in 4-month-old mice injected with AAV2/9 expressing U7-ASScn5a (n¼25) compared with age-matched control mice (n¼17). (f) Representative ECG traces reveal atrial fibrillation in U7-ASScn5a-injected mice compared with control mice. (g) Variation of the RR interval indicates evidences of heart arrhythmias in U7-ASScn5a-injected mice (n¼25) compared with control mice (n¼17). (h) Representative image of six analysed heart samples showing mild fibrosis revealed by Red Sirius histology staining in AAV-U7-ASScn5a-injected mice. Scale bar, 100 mm. (i) Real-time RT-qPCR quantification of the expression of Cola1a, Col3a1 and Tgfb mRNAs in heart of control (n¼6) or AAV-U7ASScn5a-injected mice (n¼6). Bars indicate s.e.m. Student test, * indicates Po0.5, ** indicates Po0.01, *** indicates Po0.001. bp, base pair. ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11067 8NATURE COMMUNICATIONS | 7:11067 | DOI: 10.1038/ncomms11067 | www.nature.com/naturecommunications pathological features of DM, including conduction defects and heart arrhythmias. Indeed, ECG performed 4 months post injection revealed a significant prolongation of the PR intervals (Student t-test, Pvalue r0.001) in AAV-U7-ASScn5a-injected mice compared with control injected mice (Fig. 6d,e). In contrast, QT interval was not significantly altered, and we identified only a trend towards increased QRS duration (Student t-test, Pvalue of 0.058 with 8 AAV-U7-ASScn5a-injected mice on 25 presenting a QRS higher than 19 ms versus 16.5 ms in control mice) (Fig. 6e and Supplementary Fig. 4B). Similarly, analysis of heart functions in 6-month-old animals showed that AAV-U7-ASScn5a-injected mice present a consistent increase of the PR interval compared with control injected mice (40.5 ms versus 34.8 ms respectively; Student t-test, Pvalue r0.05), without significant changes of the QRS and QT intervals (Supplementary Fig. 4B). Of interest, a similar elongation of the PR interval was observed in Scn5aþ/ mice, which are hemizygote for Scn5a expression and represent an established model for cardiac-conduction disease45–47. Furthermore, ECG analyses also revealed that 44% of AAV-U7-ASScn5a-injected mice develop significant (Student t-test, Po0,001) heart arrhythmia at 4 months post injection with an average of five arrhythmic events, defined as variation of the RR interval, per minute whereas control injected animals showed no alterations (Fig. 6f,g). We did not detect ventricular fibrillations or second and third-degree heart blocks in any injected animals. In contrast, we observed supraventricular premature contractions and atrial fibrillation in AAV-U7-ASScn5a-injected mice (Fig. 6f), and five of these injected mice died suddenly between 4 and 6 months post injections (none of the control mice died). These electrical alterations were specific and not caused by global cardiac remodelling since we observed neither systolic nor diastolic alterations by doppler echocardiography (Supplementary Table 3) and no change in heart/body weight ratio (4.4±0,1 mg g1 in control, n¼9, versus 4,7±0,2 mg g1in AAV-U7-ASScn5ainjected animals, n¼14). As further control, H&E-staining revealed normal heart structures with no evident cardiomyopathy or dilation at 6 months post AAV injections (Supplementary Fig. 4C). Similarly, quantitative RT–PCR experiments show no alteration in the expression levels of Nppa,Nppb (encoding Anp and Bnp, respectively) and Myh7 mRNAs (Supplementary Fig. 4D), suggesting no overt cardiac remodelling in antisense AAV-U7-ASScn5a-injected mice. Moreover, Sirius Red staining confirmed normal heart structures but also revealed some mild fibrosis (Fig. 6h), which was confirmed by increased expression of collagen Cola1a and Tgfb1 mRNAs (Fig. 6i). Interestingly, mild fibrosis is also observed in DM cardiac samples17,18, as well as in individuals and mice models with loss-of-function mutations of the SCN5A gene20,21,46,47. Overall, heart arrhythmias and prolonged PR interval in AAV-U7-ASScn5a-injected animals demonstrate that inclusion of the fetal exon 6A of Scn5a is inappropriate to adult mouse heart physiology. However, while we found a clear elongation of the PR interval, we did not detect a significant alteration of the QRS duration as only a third of AAV-U7ASScn5a-injected mice present increased QRS duration (419 ms). Interestingly, similar findings have been described in Scn5aþ/ mice, which all show elongation of the PR interval, while only a subset of Scn5aþ/animals present a prolongation of the QRS interval45. Hence, elongation of the PR interval is not systematically associated with increased duration of the QRS in mouse model of Scn5a dysfunction. Thus, to strengthen our data, we mathematically tested whether human cardiac parameters would be altered by the electrophysiological differences caused by the switch from adult exon 6B towards fetal exon 6A of SCN5A. Simulation based on a modified O’Hara-Rudy model48,49 predicted a change of the QRS duration from 72 ms with control adult hNa v 1.5 to 88 ms with fetal hNa v 1.5e, hence a 22% increase (Fig. 7a and Supplementary Fig. 5). Furthermore, we also tested extent of atrio-ventricular change50. Mathematical simulation predicted a change of the atrium-His interval from 81 ms with control hNa v 1.5 to 143 ms with fetal hNa v 1.5e (Fig. 7b). Overall, these results support our mouse results and provide additional evidences that misregulation of SCN5A alternative splicing causes cardiac-conduction abnormalities, which is a key pathological feature of DM (Fig. 7c). Discussion Cardiac defects affect 80% of individuals with DM and represent the second most common cause of death in this disease17,18. However, the molecular mechanisms responsible for cardiac-conduction delay and ventricular tachycardia in DM are unclear. Using RNA sequencing we identified various novel splicing misregulation events in DM1 heart samples. Among these changes, the splicing switch from adult exon 6B to fetal exon 6A in SCN5A mRNA is of particular interest. Previous studies39–41 as well as ours indicate that hNa v 1.5e, the splicing variant of SCN5A found in DM and that contains the fetal exon 6A, possesses a reduced excitability compared with the normal adult splicing form of SCN5A containing the exon 6B. Consequently, the switch from the hNa v 1.5 to the hNa v 1.5e channel in DM may cause a slower upstroke velocity of the cardiac action potential, leading to conduction slowing. Importantly, this hypothesis is supported by mathematical simulation as well by animal model, since imposing a switch from inclusion of the control adult exon 6B towards using the fetal exon 6A of Scn5A in adult mouse heart led to cardiac-conduction delay and heart arrhythmias, two key features of DM. Moreover, clinical evidence also supports an alteration of the sodium current in DM. Indeed, the electrophysiological features39–41 of the fetal isoform of SCN5A expressed in DM are similar to the electrophysiological characteristics observed with loss-of-function mutations of SCN5A causing cardiac-conduction disease51–53. Also, there are some similarities of ECG recording, including prolongation of the PR interval and of the QRS duration, between individuals with DM and individuals affected by cardiac-conduction disease caused by loss-of-function mutations in SCN5A42,54. Finally, the induction of abnormal ECG pattern in DM patients treated with ajmaline55,56, a class Ia antiarrhythmic agent acting on the cardiac sodium channel and the abnormal sodium current observed in a mouse model of DM57, are also evocative of a dysfunction of the sodium channel in DM. Overall, our results suggest that misregulation of the splicing of SCN5A participates in a subset of electrical cardiac alterations observed in DM, namely the cardiac-conduction delay and the heart arrhythmias. However, it is likely that other alternative splicing alterations and/or mechanisms58–61 are participating to the full pattern of cardiac alterations in DM since knockout of Mbnl1 and Mbnl2 in mice leads to only mild alteration of Scn5A splicing, while these mice show severe conduction disease and cardiac dilatation43,44. In conclusion, this work may also have some clinical importance such as considering with caution the treatments of DM patients with pharmaceutical agents that reduce the activity of the cardiac sodium channel, including mexiletine, flecainide and other antiarrhythmic drugs of class I. In that aspect, this study may provide a molecular explanation to the adverse cardiac reaction of some patients with myotonic dystrophic to treatment with drugs reducing activity of SCN5A (refs 62,63). Involvement of the cardiac sodium channel in DM might also highlight the importance of considering polymorphism in the SCN5A gene, as NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11067 ARTICLE NATURE COMMUNICATIONS | 7:11067 | DOI: 10.1038/ncomms11067 | www.nature.com/naturecommunications 9