rbFOX1/MBNL1 competition for CCUG RNA repeats binding contributes to myotonic dystrophy type 1/type 2 differences

ARTICLE rbFOX1/MBNL1 competition for CCUG RNA repeats binding contributes to myotonic dystrophy type 1/type 2 differences Chantal Sellier1, Estefanía Cerro-Herreros2,3, Markus Blatter4, Fernande Freyermuth1, Angeline Gaucherot1, Frank Ruffenach1, Partha Sarkar5, Jack Puymirat6, Bjarne Udd7,8,9, John W. Day10, Giovanni Meola11,12, Guillaume Bassez13, Harutoshi Fujimura14, Masanori P. Takahashi15, Benedikt Schoser16, Denis Furling13, Ruben Artero 2,3, Frédéric H.T. Allain4, Beatriz Llamusi2,3 & Nicolas Charlet-Berguerand1,17,18,19 Myotonic dystrophy type 1 and type 2 (DM1, DM2) are caused by expansions of CTG and CCTG repeats, respectively. RNAs containing expanded CUG or CCUG repeats interfere with the metabolism of other RNAs through titration of the Muscleblind-like (MBNL) RNA binding proteins. DM2 follows a more favorable clinical course than DM1, suggesting that specific modifiers may modulate DM severity. Here, we report that the rbFOX1 RNA binding protein binds to expanded CCUG RNA repeats, but not to expanded CUG RNA repeats. Interestingly, rbFOX1 competes with MBNL1 for binding to CCUG expanded repeats and overexpression of rbFOX1 partly releases MBNL1 from sequestration within CCUG RNA foci in DM2 muscle cells. Furthermore, expression of rbFOX1 corrects alternative splicing alterations and rescues muscle atrophy, climbing and flying defects caused by expression of expanded CCUG repeats in a Drosophila model of DM2. DOI: 10.1038/s41467-018-04370-x OPEN 1IGBMC, INSERM U964, CNRS UMR7104, University of Strasbourg, 67404 Illkirch, France. 2Translational Genomics Group, Interdisciplinary Research Structure for Biotechnology and Biomedicine BIOTECMED, University of Valencia, 46010 Valencia, Spain. 3INCLIVA Health Research Institute, 46010 Valencia, Spain. 4Institute for Molecular Biology and Biophysics, Swiss Federal Institute of Technology (ETH) Zurich, 8092 Zurich, Switzerland. 5Department of Neurology, University of Texas Medical Branch, Galveston, TX 77555, USA. 6Human Genetics Research Unit, Laval University, CHUQ, Ste-Foy, Quebec QC G1V 4G2, Canada. 7Neuromuscular Research Center, Tampere University Hospital, 33521 Tampere, Finland. 8Department of Medical Genetics, Folkhälsan Institute of Genetics, Helsinki University, 00290 Helsinki, Finland. 9Department of Neurology, Vasa Central Hospital, 65130 Vaasa, Finland. 10 Department of Neurology, Stanford University, San Francisco, CA 94305, USA. 11 Department of Biomedical Sciences for Health, University of Milan, 20097 Milan, Italy. 12 Neurology Unit, IRCCS Policlinico San Donato, San Donato Milanese, 20097 Milan, Italy. 13 Sorbonne Université, Inserm, Association Institut de Myologie, Center of Research in Myology, 75013 Paris, France. 14 Department of Neurology, Toneyama National Hospital, Toyonaka 560-0045, Japan. 15 Department of Neurology, Osaka University Graduate School of Medicine, Suita 565-0871, Japan. 16 Friedrich-Baur-Institute, Department of Neurology, Ludwig Maximilian University, 80539 Munich, Germany. 17 UMR7104, Centre National de la Recherche Scientifique, 67404 Illkirch, France. 18 Institut National de la Santé et de la Recherche Médicale, U964, 67404 Illkirch, France. 19 Université de Strasbourg, 67404 Illkirch, France. Correspondence and requests for materials should be addressed to B.L. (email: [email protected]) or to N.C.-B. (email: [email protected]) NATURE COMMUNICATIONS | (2018) 9:2009 |DOI: 10.1038/s41467-018-04370-x |www.nature.com/naturecommunications 1 1234567890():,; Myotonic dystrophy is the most common muscular dystrophy in adults and comprises two genetically distinct forms. Myotonic dystrophy type 1 (DM1) and its severe congenital form (CDM1) are caused by an expansion of CTG repeats in the 3′-untranslated region (UTR) of the DMPK gene1–3. In contrast, myotonic dystrophy type 2 (DM2) is caused by an expansion of CCTG repeats within the first intron of the CNBP (also known as ZNF9) gene4. Expression of mutant RNAs containing hundreds to thousands of CUG or CCUG repeats interferes with the metabolism of other RNAs through dysfunctions of mainly two classes of RNA binding proteins. First, expression and phosphorylation of the CUG-binding protein 1 (CUGBP1, encoded by the CELF1 gene) are increased in DM1 heart samples, especially in the most severely affected individuals5. Second, Muscleblind-like proteins (MBNL1, MBNL2 and MBNL3), which are RNA binding proteins specifically recognizing YGC RNA motifs, are titrated away from their normal mRNA targets as a result of their binding to expanded CUG and CCUG RNA repeats. MBNL proteins titration is illustrated by their mislocalization within nuclear RNA foci formed by expanded CUG and CCUG repeats in cell and animal models of myotonic dystrophy6–8. MBNL and CUGBP1 are RNA binding proteins that regulate pre-mRNA alternative splicing9–14. Thus, alterations of MBNL and CUGBP1 functional levels result in reversion to embryonic splicing patterns for various mRNAs, which are associated with several symptoms of myotonic dystrophy15–22. Importantly, knockout of Mbnl1 and/ or Mbnl2 in mouse reproduces splicing alterations and keys features of myotonic dystrophy10,13,14,20,23. Conversely, overexpression of MBNL1 corrects splicing alterations and myotonia in mice expressing expanded CUG repeats24. Finally, a higher number of pathogenic CUG repeats leads to a greater titration of MBNL proteins, resulting in increased RNA metabolism alterations that correlate with increased disease severity in DM1 individuals23,25. These results highlight the importance of MBNL titration in myotonic dystrophy type 1. Myotonic dystrophy type 2 resembles adult-onset DM1 with autosomal dominant inheritance pattern and similar clinical multi-organ features, including progressive skeletal muscle atrophy and weakness, myotonia, cardiac arrhythmia and conduction defects, posterior subcapsular iridescent cataract, insulin resistance, hypogammaglobulinemia, as well as cognitive and personality changes26–28. Despite these similarities, there are significant differences between DM1 and DM2, including a usually more favorable clinical course in DM2 compared to DM1. Indeed, symptoms are generally milder in DM2 compared to DM1 and include slower and less severe progression of the disease, reduced severity of the cardiac involvement, later and less prominent weakness of the respiratory, facial and bulbar muscles, less evocable myotonia and preserved social and cognitive abilities29–31. A milder involvement in DM2 compared to DM1 is also found at the cellular level as in vitro cultures of muscle cells originating from individuals with DM1 reveal reduced fusion capacity and alternative splicing alterations compared to control cells. However, only subtle or even no detectable fusion defects or splicing alterations are observed in cultures of DM2 myoblasts32–34. Paradoxically, these milder clinical and cellular features are in contradiction with the 3 to 5 folds higher expression of the CNBP mRNA, which first intron hosts the expanded CCUG repeat, compared to the DMPK mRNA, which 3’UTR hosts the expanded CUG repeats7,35,36. Furthermore, the average size of expanded repeats is generally higher in DM2 (up to ~11,000 CCTG repeats in blood) compared to adult-onset DM1 (up to ~1000 CTG repeats in blood)4. Thus, expanded CCUG repeats appear inherently less pathogenic than expanded CUG repeats. A likely contribution to this difference of toxicity is their genomic localization as CCUG repeats are embedded in the first intron of the CNBP pre-mRNA, which is presumably less stable compared to the 3′UTR of the DMPK mRNA that hosts the CUG repeats. In favor of this hypothesis, transgenic Drosophila expressing either expanded CUG or CCUG repeats, embedded in a comparable genomic context deprived of any DMPK or CNBP sequences, show similar DM-like phenotypes37,38. Nonetheless, it is not excluded that other mechanisms may further contribute to the lesser toxicity of expanded CCUG repeats in DM2. Searching for novel RNA binding proteins that interact specifically with either CUG or CCUG expanded repeats, we identified that the rbFOX RNA binding proteins bind to expanded CCUG repeats, but not to expanded CUG repeats. The rbFOX family comprises three members, rbFOX1 (also named Fox-1, A2BP1 or HRNBP1), rbFOX2 (also known as Fox-2, RBM9, Fxh or HRNBP2) and rbFOX3 (also called Fox-3, NeuN or HRNBP3), which are involved in the regulation of various aspects of mRNA metabolism39–47. While rbFOX2 is widely expressed, rbFOX1 is enriched in skeletal muscle, heart and brain, and rbFOX3 expression is restricted to neuronal cells48. All three rbFOX proteins possess a near-identical RNA-recognition motif (RRM) that recognizes the UGCAUGY RNA sequence39–41,49. Here, we find that rbFOX1 also binds to expanded CCUG repeats, albeit with a lesser affinity compared to its favorite UGCAUGY sequence. Interestingly, rbFOX1 co-localizes with CCUG RNA foci in muscle cells and skeletal muscle tissues of individuals with DM2. In contrast, rbFOX1 does not bind to expanded CUG repeats and does not co-localize with CUG RNA foci in DM1 samples. Furthermore, we find that the binding of MBNL1 and rbFOX1 to expanded CCUG repeats are mutually exclusive. Addition of rbFOX1 competes with MBNL1 binding to CCUG repeats and partially releases MBNL1 from CCUG RNA foci in DM2 muscle cells. Importantly, expression of rbFOX1 corrects splicing alterations and reduces muscle atrophy, as well as climbing and flying defects caused by expression of expanded CCUG repeats in Drosophila models of DM2. Thus, we propose that the rbFOX proteins, by competing with and reducing the titration of MBNL1 within CCUG RNA foci, may participate to the lesser toxicity of the CCTG repeat expansion in myotonic dystrophy type 2. Results Identification of proteins associated with expanded CCUG repeats. To identify novel RNA binding proteins involved in myotonic dystrophy, we incubated radioactively labeled CUG or CCUG RNA repeats with nuclear extract of differentiated mouse C2C12 muscle cells and analyzed the RNA-bound proteins by UV-light crosslink followed by SDS page gel electrophoresis (Fig. 1a). Consistent with the pioneering work of the Swanson group6, a UV-crosslink signal around 42 to 45 kDa, which corresponds to Mbnl1 molecular weight, was evident for both CUG and CCUG RNAs. In contrast, a signal at 35 to 40 kDa was observed only with CCUG RNA repeats (Fig. 1a). To identify this factor, proteins from differentiated C2C12 muscle cells were captured on streptavidin resin coupled to biotinylated RNA composed of thirty CUG or CCUG repeats, eluted, separated on SDS–PAGE gels, silver stained and each protein band was cut, gel extracted and proteins were identified by nanoLC/MS-MS analysis (Supplementary Table 1). Mass spectrometry revealed that the band of ~35 to 40 kDa contains various proteins, including Tra2a (33 kDa), Pcbp1 and Pcbp2 (37 and 38 kDa), Hnrnpa3 (40 kDa), Tiar (43 kDa) and rbFox1 (38 kDa). Importantly, only rbFox1 was preferentially associated with CCUG repeats and not with CUG repeats (Supplementary Table 1). We repeated these capture experiments using mouse brain nuclear extract. Silver ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-04370-x 2NATURE COMMUNICATIONS | (2018) 9:2009 |DOI: 10.1038/s41467-018-04370-x |www.nature.com/naturecommunications staining of the captured proteins confirmed that proteins of ~35 to 40 kDa were specifically pulled-down by the CCUG RNA repeats (Fig. 1b). Gel extraction followed by mass spectrometry analysis identified these proteins as rbFox1, rbFox2 and rbFox3 (Supplementary Table 2). Western blotting analysis of the proteins eluted from the RNA affinity columns confirmed that rbFox1 binds to CCUG repeats, but not to CUG repeats (Fig. 1c). As a positive control, Mbnl1 binds to both CUG and CCUG repeats (Fig. 1c). To confirm these results, candidate proteins identified by mass spectrometry were tested for co-localization with RNA foci of either expanded CUG or CCUG RNA repeats in transfected muscle C2C12 cells. As previously reported, Mbnl1 co-localizes with RNA foci of both expanded CUG and CCUG repeats (Fig. 1d). In contrast, rbFox1, rbFox2, and rbFox3 co-localize only with RNA foci of expanded CCUG repeats, but not with foci of expanded CUG repeats (Fig. 1e and Supplementary Fig. 1). All other tested candidate proteins do not co-localize significantly with CUG or CCUG RNA foci and thus were not investigated further (Fig. 1f and Supplementary Fig. 1). rbFOX1 directly binds to expanded CCUG repeats RNA. Presence of rbFOX1 within a large protein complex46 questions whether rbFOX1 contacts directly CCUG RNA repeats or requires indirect protein-protein interactions. Both gel-shift and UV-crosslinking experiments demonstrated that purified recombinant GST-tagged rbFOX1 directly binds to CCUG repeats, but not to CUG repeats (Fig. 2a and Supplementary Fig. 2A, B). Similarly, rbFOX2, which contains a RNA-recognition motif identical to rbFOX1, also binds to CCUG repeats, but not to CUG repeats (Supplementary Fig. 2C). As a positive control, MBNL1 binds to both CUG and CCUG repeats (Fig. 2b and supplementary Fig. 2D). Of interest, MBNL1 binds to CUG or CCUG repeats with higher affinity compared to its natural BIN1 or INSR pre-mRNA targets (Supplementary Fig. 2E). In contrast, rbFOX1 CUG 30x CCUG 30x CUG 30x CCUG 30x 250 130 95 72 55 35 Mbnl1 CCUG 1000x CUG 960x ? Mbnl1 rbFox1/2/3 25 100 50 % Co-localization 0 Zcchc11 Zc3h10 Pcbp1 Pcbp2 Tardbp Matr3 Mbnl1 rbFox1 rbFox2 rbFox3 Eif2c1 Eif2c2 Hnrnp A2B1 Hnrnp K Hnrnp L Hnrnp M Hnrnp UI1 Hnrnp UI2 Ddx3y Ddx6 Dhx57 Srsf2 Srsf3 Srsf5 Prpf31 Prpf38b Tial1 Khsrp Fmrp Ncl Raly Celf2 Sfpq MW a d f b 250 130 95 72 55 35 25 MW CUG 30x CCUG 30x Mbnl1 rbFox1 c 55 35 55 35 MW Mbnl1 RNA Merged eRNA MergedrbFox1 CUG 960xCCUG 1000x CUG 960xCCUG 1000x Fig. 1 Identification of proteins specifically associated with expanded CCUG repeats. aUV-crosslinking binding assays of 20 µg of nuclear extract from C2C12 muscle cells differentiated four days incubated with 30,000 CPM of uniformly [αP32] internally labeled in vitro transcribed RNAs containing 30 CUG or CCUG repeats. bSilver staining of proteins extracted from 1 mg of mouse brain and captured on streptavidin resin coupled to biotinylated RNA containing 30 CUG or CCUG repeats. cWestern blotting against either rbFox1 or Mbnl1 on mouse brain proteins captured by RNA-column containing either 30 CUG or 30 CCUG repeats. dRNA FISH against CCUG repeats coupled to immunofluorescence against Mbnl1 on differentiated C2C12 cells transfected with a plasmid expressing either 960 CUG or 1000 CCUG repeats. eRNA FISH against CCUG repeats coupled to immunofluorescence against rbFox1 on differentiated C2C12 cells transfected with a plasmid expressing either 960 CUG or 1000 CCUG repeats. Scale bars, 10 µm. Nuclei were counterstained with DAPI. fQuantification of the co-localization of CUG or CCUG RNA foci with candidate proteins in transfected C2C12 cells. Error bars indicate s.e.m. of three independent experiments. Representative images are presented in Supplementary Fig. 1 NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-04370-x ARTICLE NATURE COMMUNICATIONS | (2018) 9:2009 |DOI: 10.1038/s41467-018-04370-x |www.nature.com/naturecommunications 3 binds to ten CCUG repeats with a 2 to 5 folds lesser affinity compared to its known UGCAUGC RNA motif or to its natural BIN1 or INSR pre-mRNA targets (Supplementary Fig. 2F). Competition experiments confirmed that rbFOX1 binds preferentially to the UGCAUGC RNA motif compared to CCUG repeats (Supplementary Fig. 2G). Of interest, binding of rbFOX1 to two CCUG repeats (UGCCUGCC RNA sequence) is negligible compared to the UGCAUGC sequence (Supplementary Fig. 2F), indicating that significant binding of rbFOX1 to the CCUG RNA motif required repetition of that motif. Apparent K D are indicated in the supplementary Fig. 2H. Next, we investigated how rbFOX1 can interact, even with a weak affinity, with expanded CCUG RNA repeats. Previously determined NMR structure of the RNA-recognition motif (RRM) of rbFOX1 in complex with the UGCAUGU sequence shows that the central adenosine (underlined) is not involved in any direct RNA-protein interaction49. Strikingly, replacement of this adenine by a cytosine yields a motif matching perfectly DM2 mutation (Fig. 2c), questioning whether rbFOX1 can accommodate a cytosine instead of the adenine in the UGCAUGC RNA motif. Indeed, energy minimization and computer modeling of rbFox1 RRM in complex with UGCCUGC RNA shows that the substitution to a cytosine in position four allows the base to adopt a similar base pairing with guanine two compared to the previously shown adenine-guanine interactions (Fig. 2, e). Importantly, we confirmed this model by NMR spectroscopy, and comparison of the amide backbone chemical shifts of both complexes (Supplementary Fig. 2I, 2J) indicates that the number of intra RNA hydrogen bonds is preserved and supports the overall similar binding topology between UGCCUGC and UGCAUGC RNAs. These results confirm that rbFOX1 directly binds to CCUG RNA repeats. rbFOX1 co-localizes with CCUG RNA foci in DM2 patients. We next assessed the localization of endogenous rbFOX proteins in muscle cells and tissues of individuals with DM2. Importantly, GST-rbFOX1 a b c de GST-MBNL1Δ101 GST-MBNL1Δ101 CUG 10x CCUG 10x CUG 10x 1234567 CCUG 10x G6 U5 Val188 lle124 Gly157 Arg184 Arg127 Asn151 Val188 Gly157 Asn151 lle124 Arg184 Arg127 Phe126 Phe126 C4 C3 U1 G2 C7 C4 C3 G2 U1 3′ 3′ 5′ 5′ β2β4 α1 β1β1α1-loop β1α1-loop β1 β4 β3′β3′′ β3′′ α2 α1 β3 rbFOX binding site DM2 mutation CCUG 10x CUG 10x CCUG 10x CUG 10x Free Bound Free Bound % Bound 100 80 60 40 20 0 % Bound 100 80 60 40 20 0 0.01 0.1 1 10 100 0.0 0.1 1 10 100 nM GST-rbFOX1 nM GST-MBNL1Δ101 GST-rbFOX1 Fig. 2 rbFOX1 binds to expanded CCUG RNA repeats. aLeft panel, gel-shift assays of 0, 0.1, 0.3, 1, 3, 10, 30, and 100 nM of purified recombinant GSTrbFOX1 with 10 pM (3000 CPM) of uniformly [αP32] internally labeled in vitro transcribed RNAs containing either 10 CUG or CCUG repeats. Right panel, gel-shift quantification. bGel-shift as in abut with purified recombinant GST-MBNL1Δ101.cAlignment of the UGCAUGC consensus RNA binding site for rbFOX1 with expanded CCUG repeats that constitute the DM2 mutation. dModel of rbFOX1 RRM bound to UGCCUGC. RNA is shown in stick representation (yellow) and potential hydrogen bonds in dashed lines (purple). eMagnification of dshowing that guanine 2 and cytosine 4 form a nonWatson-Crick base pair, in an analogous way to guanine 2 and adenine 4 in rbFOX1 RRM bound to UGCAUGU described previously (pdb 2ERR)49. Error bars indicate s.e.m. of five independent experiments ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-04370-x 4NATURE COMMUNICATIONS | (2018) 9:2009 |DOI: 10.1038/s41467-018-04370-x |www.nature.com/naturecommunications RNA FISH coupled to immunofluorescence analysis revealed that endogenous rbFOX1 co-localizes with endogenous CCUG RNA foci in primary cultures of differentiated muscle cells originating from skeletal muscle biopsies of individuals with DM2 (Fig. 3a and Supplementary Fig. 3A, 3B). In contrast, rbFOX1 is not present within CUG RNA foci in DM1 muscle cells (Fig. 3a). Concomitant labeling of endogenous MBNL1 demonstrated colocalization of MBNL1 with both CUG and CCUG RNA foci (Fig. 3a and Supplementary Fig. 3B). We confirmed these results in skeletal muscle sections of individuals with myotonic dystrophy. RNA FISH coupled to immunofluorescence indicated that both rbFOX1 and rbFOX2 co-localize with CCUG RNA foci in DM2 muscle sections, but not with CUG RNA foci in DM1 muscle sections (Fig. 3b, c). Concomitant labeling of endogenous MBNL1 demonstrated co-localization of MBNL1 with both CUG and CCUG RNA foci (Fig. 3b, c). As controls, RNA FISH coupled to immunofluorescence analysis indicated that neither endogenous rbFox1 nor rbFox2 co-localize with RNA foci of overexpressed expanded CUG, CGG or AUUCU repeats, which are involved in DM1, Fragile X Tremor and Ataxia Syndrome (FXTAS) and spinocerebellar ataxia of type 10 (SCA10), respectively (Supplementary Fig. 3C and 3D). Overall, these data indicate that rbFOX1 and rbFOX2 are specific components of CCUG RNA foci in myotonic dystrophy type 2. rbFOX splicing regulatory functions are not altered in DM2. To test a potential titration of rbFOX1 within CCUG RNA foci, we first assessed its mobility. Dendra2-tagged rbFOX1 was cotransfected with a vector expressing either no repeats or expanded CUG or CCUG repeats, then a nuclear spot of Dendra2-rbFOX1 was photoconverted from green to red and imaged every minute to follow the nuclear diffusion of rbFOX1. Interestingly, photoconverted Dendra2-rbFOX1 present in nuclear foci of CCUG expressing cells moved away from the CCUG foci and diffused freely in the nucleoplasm, albeit with a slightly slower kinetic compared to rbFOX1 in control condition or with expanded CUG repeats (Fig. 4a). As a control, photoconverted Dendra2MBNL1 diffuse freely in control cells but was immobilized in foci of expanded CUG or CCUG repeats (Fig. 4b). These data suggest that in contrast to MBNL1, rbFOX1 is not immobilized within CCUG RNA foci. In a second approach, we tested whether expanded CCUG repeats would modify rbFOX1 splicing regulatory functions. Alternative splicing of the human mitochondrial ATP synthase gamma-subunit (ATP5C1 also named F1gamma) exon 9 minigene is regulated by the rbFOX proteins39. However, overexpression of expanded CCUG repeats does not modify ATP5C1 minigene splicing (Fig. 4c). As a control, overexpression of expanded CCUG repeats alters the splicing regulation of the insulin RNA a b c DM1DM2DM1DM2DM1DM2 rbFOX1 MBNL1 Merged 100 *** *** *** * 50 % Co-localization 0 100 50 % Co-localization 0 100 50 % Co-localization 0 MBNL1 rbFOX1 MBNL1 rbFOX1 MBNL1 rbFOX2 CCUG CUG CCUG CUG CCUG CUG RNA rbFOX1 MBNL1 Merged RNA rbFOX2 MBNL1 Merged Fig. 3 rbFOX1 co-localizes with CCUG RNA foci in DM2 patients. aLeft panel, representative confocal images of RNA FISH against CCUG repeats coupled to immunofluorescence against MBNL1 and rbFOX1 on four days differentiated muscle cells originating from muscle biopsies of individuals with either DM1 or DM2. Right panel, quantification of the co-localization of MBNL1 and rbFOX1 with CUG or CCUG RNA foci. bLeft panel, CUG or CCUG repeats RNA FISH combined with immunofluorescence against MBNL1 and rbFOX1 on skeletal muscle sections of adult individuals with DM1 or DM2. Right panel, quantification of the co-localization of rbFOX1 within a hundred CUG or CCUG RNA foci. cRNA FISH/ immunofluorescence as in bbut with an antibody directed against rbFOX2. Scale bars, 10 µm. Nuclei were counterstained with DAPI. Error bars indicate s.e.m. of three independent experiments. Student’sttest, asterisk (*) indicates p< 0.5, asterisk (***) indicates p< 0.001 NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-04370-x ARTICLE NATURE COMMUNICATIONS | (2018) 9:2009 |DOI: 10.1038/s41467-018-04370-x |www.nature.com/naturecommunications 5 receptor (INSR) exon 11 minigene (Supplementary Fig. 4A). Of interest, MBNL1 regulates INSR minigene splicing, but does not modulate ATP5C1 minigene splicing (Fig. 4c and Supplementary Fig. 4A). These results suggest that expression of expanded CCUG repeats alters MBNL1 splicing regulatory function, but does not modify rbFOX1 splicing activity. Importantly, identical results were observed with endogenous alternative splicing events regulated by the rbFOX proteins42. Namely, overexpression of expanded CCUG repeats has no significant effect on endogenous Fmnl3 exon 26, Tbx3 exon 25, and Enah exon 11 A alternative splicing (Fig. 4d and Supplementary Fig. 4B, 4C). We confirmed that overexpression or siRNA-mediated depletion of rbFox1 regulates Fmnl3,Tbx3 and Enah alternative splicing, while expression or depletion of Mbnl1 has no effect (Fig. 4d and Supplementary Fig. 4B, 4C). Western blotting indicated correct siRNA-mediated depletion of rbFox1 and Mbnl1 proteins 1.0 ab c e f g d 0.8 0.6 Relative fluorescence 0.4 0.2 0.0 1.0 0.8 0.6 Relative fluorescence 0.4 0.2 0.0 0 100 *** *** *** ** 50 % Inclusion 0 Controls DM2 Controls DM2 Controls DM2 100 50 % Inclusion 0 100 50 % Inclusion 0CTL DM2 100 50 % Inclusion 0CTL DM2 100 50 % Inclusion 0CTL DM2 1234 Time (min) CTL GFP-rbFOX1 siRNA rbFox1 siRNA Mbnl1 CUG 960x CCUG 1000x GFP-MBNL1 CTL GFP-rbFOX1 siRNA rbFox1 siRNA Mbnl1 CUG 960x CCUG 1000x GFP-MBNL1 rbFOX1 + CCUG 1000x rbFOX1 + CUG 960x MBNL1 + CCUG 1000x MBNL1 + CUG 960x MBNL1 rbFOX1 * ** ** *** *** *** *** *** 567 ATP5C1 + Ex 9 – Ex 9 Fmnl3 + Ex 26 – Ex 26 FMNL3 + Ex 26 – Ex 26 ENAH + Ex 11A – Ex 11A ECT2 + Ex 25 – Ex 25 01234 Time (min) 567 Fig. 4 rbFOX1 is not sequestered within CCUG RNA foci. aTime course quantification of photoconverted spot of dendra2-rbFOX1 in COS7 cells cotransfected with a plasmid expressing dendra2-rbFOX1 and a plasmid expressing either no repeats (CTL), 960 CUG or 1000 CCUG repeats. Each data point is the average of 7 spot. bAs in abut with dendra2-MBNL1. cUpper panel, RT-PCR analysis of RNA extracted from two days differentiated C2C12 cells co-transfected with a minigene expressing the exon 9 of the mitochondrial ATP synthase gamma-subunit gene and either with a plasmid expressing rbFOX1, MBNL1, 960 CUG repeats or 1000 CCUG repeats or with a siRNA directed against rbFox1 or Mbnl1. Lower panel, quantification of exon 9 inclusion of transfected ATP5C1 minigene. dUpper panel, RT-PCR analysis of endogenous Fmnl3 exon 26 alternative splicing from GFP-FACS sorted C2C12 cells differentiated two days and co-transfected with a plasmid expressing eGFP and either with a plasmid expressing rbFOX1, MBNL1, 960 CUG repeats or 1000 CCUG repeats or with a siRNA directed against rbFox1 or Mbnl1. Lower panel, quantification of Fmnl3 exon 26 inclusion. e–gRT-PCR analysis (left panel) and quantification (right panel) of alternative splicing of FMNL3, ENAH, and ECT2 performed on total RNA extracted from adult skeletal muscle of control or DM2 individuals. Error bars indicate s.e.m. of three independent experiments. Student’st-test, asterisk (*) indicates p< 0.5, asterisk (**) indicates p< 0.01, asterisk (***) indicates p< 0.001 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-04370-x 6NATURE COMMUNICATIONS | (2018) 9:2009 |DOI: 10.1038/s41467-018-04370-x |www.nature.com/naturecommunications (Supplementary Fig. 4D). As a further control, expression of endogenous Mbnl1 is not altered upon overexpression or depletion of rbFOX1 (Supplementary Fig. 4E). Finally, we tested whether rbFOX-dependent alternative splicing events were modified in skeletal muscle samples of DM2 patients. RT-PCR analysis demonstrated that exon 26 of FMNL3, exon 11 A of ENAH, and exon 25 of ECT2 do not display any splicing abnormalities in muscle samples of individuals with DM2 compared to control skeletal muscle samples (Fig. 4e, g). As controls, alternative splicing of exons regulated by MBNL1, namely exon 29 of CACNA1S (CAV1.1), exon AS1 of RYR1, exon 22 of ATP2A1 (SERCA), exon 7 of MBNL1 and exon 11 of LDB3 (CYPHER/ZASP), are all altered in DM2 muscle samples (Supplementary Fig. 4F, 4G, 4H, 4I, 4J). Of interest, the extend of splicing alterations in distal skeletal muscle samples is greater in DM1 compared to DM2 (Supplementary Fig. 4F, 4G, 4H, 4I, 4J), which correlates with the greater impairment of distal muscle in DM1 compared to DM2. These results indicate that, in contrast to MBNL1, the splicing regulatory functions of the rbFOX proteins are not altered by expression of expanded CCUG repeats. rbFOX compete the binding of MBNL1 to expanded CCUG repeats. The binding of both MBNL1 and rbFOX1 to expanded CCUG repeats questions whether these interactions are mutually exclusive. In vitro, an excess of either rbFOX1 or rbFOX2 chases MBNL1 from binding to CCUG repeats (Fig. 5a and supplementary Fig. 5A). Conversely, an excess of MBNL1 competes rbFOX1 binding to CCUG repeats (Fig. 5b). In DM2 muscle cell cultures, overexpression of rbFOX1 reduces the quantity of MBNL1 present within RNA foci of expanded CCUG repeats and concomitantly increases the labeling of free diffuse MBNL1 (Fig. 5c). Inversely, shRNA-mediated depletion of rbFOX1 expression increases the localization of endogenous MBNL1 within CCUG RNA foci (Supplementary Fig. 5B). Consistent with a mutually exclusive competition mechanism, overexpression of MBNL1 partially displaces rbFOX1 from CCUG RNA foci in DM2 muscle cells (Fig. 5d). As controls, we observed no displacement of MBNL1 localization upon modulation of rbFOX1 expression in DM1 muscle cell cultures (Fig. 5c and Supplementary Fig. 5C). These results suggest that rbFOX1 competes with MBNL1 to bind to CCUG RNA repeats. Nevertheless, it is possible that other +++++ + 100 50 0 % Binding +++ + T T UU T U T U T U T U T U T U T U T U T U T U TUTUTUTU c c T U T U T U T U T U T U T U T U T U T U T U T U T U T U T U T U +++ +++ rbFOX1 a c d b MBNL1 rbFOX1 MBNL1 MBNL1 MergedFlag-rbFOX1RNA Flag-MBNL1 MergedrbFOX1RNA DM1DM2DM1DM2 rbFOX1 MBNL1 CCUG 30x rbFOX1 MBNL1 CCUG 30x GST-rbFOX1 GST-MBNL1Δ101 GST-rbFOX1 GST-MBNL1Δ101 GST-rbFOX1 GST-MBNL1Δ101 GST-rbFOX1 GST-MBNL1Δ101 100 50 0 % Binding 100 50 0DM1 CTL Flag-rbFOX1 CTL Flag-MBNL1 DM2 DM1 DM2 *** *** % MBNL1 in foci 100 50 0 % rbFOX1 in foci Fig. 5 rbFOX1 competes with MBNL1 for binding to expanded CCUG repeats. aUpper panel, UV-cross-linking binding of 1 µg of GST-MBNL1Δ101 to 10,000 CPM of uniformly [αP32] internally labeled in vitro transcribed RNA containing 30 CCUG repeats was competed by increasing amounts (0.25, 0.5, 1, and 2 µg) of GST-rbFOX1. Middle panel, loading of recombinant MBNL1 and of rbFOX1 proteins was verified by coomassie staining. Lower panel, quantification of the binding of MBNL1 and rbFOX1 to expanded CCUG repeats. bUV-cross-linking as in abut with binding of 0.5 µg of GST-rbFOX1 competed by 0.5, 1, and 2µg of GST-MBNL1Δ101.cLeft panel, RNA FISH of expanded CUG or CCUG repeats coupled to immunofluorescence against Flag-rbFOX1 and MBNL1 on primary cultures of myoblasts originating from muscle biopsies of individuals with either DM1 or DM2 and transfected with a plasmid expressing FLAGtagged rbFOX1. Each image shows at least two different cells, one transfected (T) and one not transfected (U), with insets showing higher magnification focusing on RNA foci. Right panel, quantification of the signal of endogenous MBNL1 localized within the CCUG RNA foci. dAs in cbut with transfection of FLAG-tagged MBNL1 and immunofluorescence against rbFOX1 and Flag-MBNL1. Scale bars, 10 µm. Nuclei were counterstained with DAPI. Error bars indicate s.e.m. of three independent experiments. Student’st-test, asterisk (***) indicates p< 0.001 NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-04370-x ARTICLE NATURE COMMUNICATIONS | (2018) 9:2009 |DOI: 10.1038/s41467-018-04370-x |www.nature.com/naturecommunications 7 mechanisms are at play. As noted previously44,50,51, overexpression of rbFOX1 inhibits inclusion of Mbnl1 exon 7 (Supplementary Fig. 5D). This 36 nts long exon is reported to modulate MBNL1 self-dimerization52. However, this splicing change is unlikely to contribute significantly to MBNL1 reduced localization within CCUG RNA foci upon rbFOX1 overexpression as in the same conditions of rbFOX1 overexpression we observed no changes in MBNL1 localization within CUG RNA foci in DM1 muscle cells (Fig. 5c and Supplementary Fig. 5C). Furthermore, rbFOX1 overexpression neither changes MBNL1 nuclear/ cytoplasmic localization (Fig. 5c), nor MBNL1 global expression (Supplementary Fig. 5E). Similarly, expressions of rbFOX1, rbFOX2,MBNL1, and MBNL2 mRNAs are not altered in muscle samples of individuals with DM2 compared to non-DM controls or individuals with DM1 (Supplementary Fig. 5F). A potential mechanism of competition requires that a sufficient quantity of endogenous rbFOX1 is expressed to prevail over MBNL1. In human skeletal muscle, MBNL1 mRNA is expressed at slightly higher level compared to rbFOX1 mRNA with 50 to 70 Transcripts Per Kilobase Million (TPM) for MBNL1 compared to 30 to 40 TPM for rbFOX1, while MBNL2 and rbFOX2 mRNAs are expressed at similar levels (20 to 30 TPM)35,36. However, western blotting analysis indicated that both rbFOX1 and MBNL1 are expressed at similar levels in control or in DM2 adult human skeletal muscle samples (Supplementary Fig. 5G). To avoid a bias due to antibody differences, recombinant purified proteins were used as standards (Supplementary Fig. 5H). Thus, a competition between MBNL1 and rbFOX1 is possible as these proteins are expressed at comparable levels in human adult skeletal muscle. Expression of rbFOX1 corrects splicing alterations of CCUG repeats. Titration of the MBNL proteins within CUG or CCUG RNA foci leads to specific alternative splicing changes. Hence, we tested whether the release of MBNL1 from CCUG RNA foci upon rbFOX1 overexpression may correct splicing alterations typical of myotonic dystrophy. Importantly, overexpression of rbFOX1 in C2C12 muscle cells partly corrects splicing alterations of the chloride channel (Clcn1) and of the cardiac troponin T (TNNT2) minigenes induced by overexpression of expanded CCUG repeats (Figs. 6a, b and Supplementary Fig. 6A). Inversely, siRNAmediated decreased expression of endogenous rbFox1 increases the pathogenic effect of expanded CCUG repeats on Clcn1 and TNNT2 minigene splicing (Supplementary Fig. 6B, 6C). In contrast, overexpression or decreased expression of rbFOX1 has no correcting effect on Clcn1 and TNNT2 splicing changes induced by expanded CUG repeats (Fig. 6a, b and Supplementary Fig. 6B and 6C). As a positive control, expression of MBNL1 corrects splicing alterations caused by either CUG or CCUG expanded repeats (Fig. 6a, b). Note that MBNL1 regulates alternative splicing of the Clcn1 and TNNT2 minigenes, while rbFOX1 has no effect in absence of CCUG repeats (Fig. 6a, b). In support of a splicing regulation directly modulated by MBNL1 but not by rbFOX1, gel-shift experiments indicate that recombinant purified MBNL1, but not rbFOX1, binds to Clcn1 and TNNT2 minigenes (Supplementary Fig. 6D). As a further control, western blotting indicates that overexpression of rbFOX1 does not modify endogenous Mbnl1 levels (Supplementary Fig. 6E). Expression of rbFOX1 alleviates muscle atrophy. To test whether expression of rbFOX1 could rescue any deleterious phenotypes caused by expression of expanded CCUG repeats, we developed Drosophila models of DM1 and DM2 that overexpress rbFOX1. Expression of uninterrupted expanded CTG or CCTG repeats deleted of their natural DMPK or CNBP sequences were targeted to fly muscle through crossing UAS-CTG or UAS-CCTG flies with a Myosin heavy chain (Mhc)-GAL4 driver line. As noted previously37, expression of expanded CUG or CCUG repeats leads to muscle dysfunctions compared to a control GFP line or to Drosophila lines carrying control (20×) CTG or CCTG repeats GFP rbFOX1 CCUG 1000x a b 80 60 *** **** *** *** *** *** 40 20 0 % Inclusion CUG 960x MBNL1 GFP rbFOX1 MBNL1 Clcn1 + Ex 6B – Ex 6B rbFOX1 MBNL1 % Inclusion TNNT2 + Ex 5 – Ex 5 rbFOX1 rbFOX1 rbFOX1 MBNL1 MBNL1 MBNL1 GFP GFP siMBNL1 CCUG 1000x CUG 960x 100 50 0 *** *** *** *** *** Fig. 6 rbFOX1 corrects splicing alterations caused by CCUG repeats. aUpper panel, RT-PCR analysis of alternative splicing of the mouse chloride channel Clcn1 exon 6B minigene co-transfected in C2C12 mouse muscle cells with a plasmid expressing either 960 CUG repeats or 1000 CCUG repeats and a vector expressing either rbFOX1 or MBNL1. Lower panel, quantification of Clcn1 exon 6B inclusion. bAs in abut with TNNT2 (cTNT) exon 5 minigene. Error bars indicate s.e.m. of three independent experiments. Student’st-test, asterisk (*) indicates p< 0.5, asterisk (**) indicates p< 0.01, asterisk (***) indicates p< 0.001 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-04370-x 8NATURE COMMUNICATIONS | (2018) 9:2009 |DOI: 10.1038/s41467-018-04370-x |www.nature.com/naturecommunications (Supplementary Fig. 7A). We then generated recombinant Drosophila flies expressing GFP-tagged rbFOX1 under the UAS promoter and crossed these flies with either DM1 or DM2 lines. Importantly, overexpression of rbFOX1 fully rescues muscles atrophy caused by expression of expanded CCUG repeats (Fig. 7a). Two different independent Drosophila lines expressing rbFOX1 were tested and gave comparable results (Fig. 7a). Similarly, a second independent line expressing expanded CCUG repeats was tested and shows identical muscle correction upon overexpression of rbFOX1 (Supplementary Fig. 7B). Next, we tested the effect of rbFOX1 in DM1 flies. Interestingly, overexpression of rbFOX1 does not suppress muscle atrophy in Drosophila expressing expanded CUG repeats (Fig. 7b). As a positive control, expression of MBNL1 fully corrects muscle alterations in DM1 and DM2 flies (Figs. 7a, b). Consistent with correction of muscle atrophy, functional assays demonstrated that overexpression of rbFOX1 rescues both flying (Fig. 7c) and climbing (Fig. 7d) defects caused by expression of expanded CCUG repeats. This correction is specific, as overexpression of rbFOX1 has no rescue effect in CUG expressing flies (Figs. 7c, d). CCUG 1000x a b cd ef 100 50 % Muscle area 0 100 *** *** *** *** 50 % Muscle area 0 80 40 Average landing height 0 20 ** *** *** *** 10 V (mm/s) 0 1 0.5 Serca exon 13 Relative expression 0 GFP CCUG 1000x CCUG 1000x; rbFOX1 L1 CCUG 1000x; rbFOX1 L2 CCUG 1000x; MBNL1 GFP CUG 250x CUG 250x; rbFOX1 L1 CUG 250x; rbFOX1 L2 CUG 250x; MBNL1 GFP CUG 250x CUG 250x; rbFOX1 L1 CUG 250x; rbFOX1 L2 CCUG 1000x; rbFOX1 L1 CCUG 1000x; rbFOX1 L2 CCUG 1000x GFP CUG 250x CUG 250x; rbFOX1 L1 CUG 250x; rbFOX1 L2 CCUG 1000x; rbFOX1 L1 CCUG 1000x; rbFOX1 L2 CCUG 1000x GFP CUG 250x CUG 250x; rbFOX1 L1 CUG 250x; rbFOX1 L2 CCUG 1000x; rbFOX1 L1 CCUG 1000x; rbFOX1 L2 CCUG 1000x CCUG 1000x; rbFOX1 L1 CCUG 1000x; rbFOX1 L2 CCUG 1000x; MBNL1 CCUG 250x CUG 250x; rbFOX1 L1 CUG 250x; rbFOX1 L2 CUG 250x; MBNL1 *** *** GFP CUG 250x CUG 250x; rbFOX1 L1 CCUG 1000x; rbFOX1 L1 CCUG 1000x; rbFOX1 L2 CCUG 1000x Fhos + Ex 16 – Ex 16 *** *** 100 50 % Inclusion 0 NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-04370-x ARTICLE NATURE COMMUNICATIONS | (2018) 9:2009 |DOI: 10.1038/s41467-018-04370-x |www.nature.com/naturecommunications 9