Zebrafish deficient for Muscleblind-like 2 exhibit features of myotonic dystrophy

Myotonic dystrophy (DM; also known as dystrophia myotonica), the most common form of muscular dystrophy in adults, is a progressively disabling disease that is inherited in an autosomal dominant manner, with anticipation (more severe phenotype and earlier onset from one generation to the next) and variable expressivity (symptoms vary among individuals) (Harper, 2001). The predominant symptoms are neuromuscular, including myotonia, muscular weakness and wasting. Other features of DM are cardiac conduction abnormalities, cognitive impairment, posterior cataract, frontal balding, testicular atrophy and insulin resistance. There are two forms of DM (DM1 and DM2), and these share several clinical symptoms and a similar molecular basis. DM1 and DM2 are both caused by unstable repeat expansions and their severity varies in proportion to the length of the repeat. The mutation underlying DM1 is a CTG repeat expansion in the 3′ untranslated region (UTR) of the DMPK gene (Brook et al., 1992; Buxton et al., 1992; Fu et al., 1992; Mahadevan et al., 1992), whereas DM2 is linked to a CCTG repeat in the first intron of the ZNF9 (CNBP) gene (Liquori et al., 2001). Both mutations are in UTRs; however, these regions are transcribed, resulting in mutant RNAs that are thought to exert an RNA trans-dominant effect (Michalowski et al., 1999). Previous studies have shown that mutant mRNAs containing the transcribed repeats are retained within the nuclei of DM1 and DM2 cells, where they bind and sequester specific RNA-binding proteins, including Muscleblind-like (MBNL) (Miller et al., 2000). It has been proposed that the sequestration of these proteins plays a key role in determining the classic features of DM (Michalowski et al., 1999), including alternative splicing abnormalities of specific transcripts (Philips et al., 1998; Savkur et al., 2001; Charlet et al., 2002; Mankodi et al., 2002). However, the functions of the MBNL genes are not completely understood.

MBNL proteins contain four zinc-finger domains of the CCCH type, and their homologue is required for muscle and photoreceptor development in Drosophila (Begemann et al., 1997; Artero et al., 1998). In humans, there are three members of the MBNL family, MBNL1, MBNL2 and MBNL3, all of which have been shown to colocalize with the nuclear foci of expanded repeat transcripts in DM1 and DM2 cells and tissues (Fardaei et al., 2001; Mankodi et al., 2001; Fardaei et al., 2002). It is known that each of the MBNL proteins is expressed at different levels in different tissues, which points towards functional specialization. One of the proposed functions for MBNL proteins is that of regulating the alternative splicing patterns of particular transcripts at specific points during development (Ho et al., 2004; Ladd et al., 2005). In the case of MBNL1, overexpression studies in a mouse model for DM support its role in DM pathogenesis, as it does in Mbnl1 knockout mice (Kanadia et al., 2003; Kanadia et al., 2006). However, the specific functions of MBNL2 and its relevance to DM are not entirely elucidated.

The zebrafish (Danio rerio) is a powerful tool in which to study developmental biology, organogenesis and diseases of vertebrates, including, in recent years, humans (Dodd et al., 2000; Bassett and Currie, 2003). We employed a reverse genetics approach using antisense technology in order to determine the relevance of MBNL2 to DM and learn more about the function of this protein. As a result, we have produced and characterized an mbnl2 knockdown model in zebrafish, which exhibits features of DM. We showed that loss of zebrafish mbnl2 function produces splicing abnormalities and muscle defects, similar to those observed in DM. Moreover, mbnl2 zebrafish morphants showed morphological abnormalities at the eye, heart and brain level, as well as defective somite patterning, suggesting that mbnl2 plays an essential role during embryonic development.

The annotation of the zebrafish gene family MBNL was carried out by sequencing cDNA clones and analysing the whole zebrafish genomic and expressed sequences available from public databases. Sequence comparison between human and zebrafish MBNL homologues, and the phylogenetic tree based on full-length alternatively spliced forms of the three members of this family, can be seen in supplementary material Fig. S1A,B.

To establish the spatial expression pattern of mbnl2, whole-mount in situ hybridization studies were conducted on zebrafish embryos at different stages of development, spanning from cleavage [3–4 hours post-fertilization (hpf)] to the hatching period (48–72 hpf). It was observed that the expression of mbnl2 peaks during the segmentation period (10–24 hpf), with a strong bilateral signal detected at the mesencephalon and hindbrain level, in spinal cord neurons and in the caudal portion of the neural tube (Fig. 1A–D). Later in development, during the pharyngula and hatching periods, the mbnl2 signal was present in the pectoral fin bud, lens and telencephalon (Fig. 1E). The use of a control sense probe allowed us to differentiate between the true positive signal and the background noise. A probe directed to pax2a was used with the purpose of highlighting the surrounding structures and delimiting the expression of mbnl2 (Fig. 1G–K). Pax2a is a well-known marker for the midbrain-hindbrain boundary, hindbrain, otic capsule and optic nerve (Fig. 1I,J). Parallel and simultaneous detection of mbnl2 and pax2a transcripts was carried out in a set of wild-type zebrafish at different stages of development. No overlap between the two probes was detected.

We used antisense technology in a loss-of-function approach to examine the in vivo role of mbnl2. Two translation-blocking morpholinos, MO-1 and MO-2, were designed to specifically target mbnl2. Their sequences were compared with those of mbnl1 and mbnl3 in order to assess potential cross-hybridization. Both morpholinos showed more than five mismatches with both transcripts (Fig. 2A). Introducing morpholino MO-1 in zebrafish embryos resulted in a dose-dependent specific phenotype. The main abnormalities observed in the mbnl2 morphants consisted of eye, brain and muscle defects, as well as abnormalities of cardiac structure and function (Fig. 2B–M). Additionally, the movements of mbnl2 morphants were restricted and uncoordinated. A second morpholino (MO-2), designed upstream to MO-1, mimicked this phenotype, but the effect was milder (data not shown). By contrast, injection of a standard control morpholino, with no target nor significant biological activity (Gene Tools), did not have any effect. Because of the lack of a suitable antibody it was not possible to demonstrate a reduction in the level of Mbnl2 protein following morpholino knockdown. Thus, we generated two further morpholinos, MO-3 and MO-4, to block the mbnl2 splice sites flanking exon 1B, as shown in supplementary material Fig. S2. Embryos treated with MO-3 or MO-4 demonstrated the same phenotype as those treated with morpholino MO-1. Reverse transcriptase PCR (RT-PCR) analysis revealed that MO-4 treatment induced a mis-splicing event using an illegitimate splice site within exon 1B, resulting in the production of an anomalous isoform lacking the 3′ 42 bp of the exon, as shown in supplementary material Fig. S3. Translation of this isoform would result in an Mbnl2 protein lacking most of the first zinc finger. supplementary material Fig. S4 shows the sequence of the deleted fragment.

The caudal defects consisted of shortened and twisted tails. The severity of these malformations varied greatly, from a slight bending to a severe shortening. In moderately and severely affected embryos, the somite shape was altered, changing the classic chevron shape for a ‘U’ shape. Withdrawal movements in the mbnl2 morphants ceased after a few seconds of gentle stimulation with a pipette tip, whereas, in the controls, there was a continuous avoidance response. The movements of the embryos were restricted and embryos showed an abnormal trembling pattern (see supplementary material Movies 1 and 2). At the eye level, the morphants presented asymmetric defects, which varied from mild to severe unilateral hypoplasia and hypopigmentation. The normal morphology was lost at the midbrain level in moderate-to-severe morphants. In mild morphants, the hindbrain ventricle seemed swollen or enlarged, giving the head a convex appearance. The hearts of the morphants were hypoplastic and became elongated and/or dilated after 48 hpf. Pericardial oedema was of common occurrence, and in some cases pools of static blood formed in the surrounding area. The cardiac deformities caused progressive malfunctioning of this organ, with heart beating slowing down and finally stopping.

In order to determine whether there is compensatory expression of the paralogues following mbnl2 knockdown, we performed quantitative RT-PCR analysis to determine the levels of expression of mbnl1 and mbnl3. supplementary material Fig. S5 shows that there is no significant change in the levels of mbnl1 or mbnl3 following knockdown of mbnl2.

Morpholinos sometimes induce off-target effects, which are caused by upregulation of p53 and subsequent apoptosis. Injection of a morpholino against p53 can alleviate these effects and facilitate the observation of true phenotypes. When MO-p53 was co-injected with the mbnl2 MO-1 morpholino, the level of neural death was reduced, although a mild brain phenotype was still observed. The other defects – shortened twisted tails, hypoplastic hearts and pericardial oedema – all remained unchanged.

We assessed the effect of mbnl2 knockdown on muscle development and fibre type specification by light microscopy, transmission electron microscopy (TEM) and immunohistochemistry assays. Histological analysis of sections stained with toluidine blue was carried out on specimens at 52 hpf (Fig. 3A,B). The somite chevron shape expected at this stage of development was lost in the entire set of mbnl2 morphant embryos examined, which showed an abnormally irregular or ‘U’ somite shape. A decreased number of muscle fibres, which made the nuclei appear more prominent, was also noticed in mbnl2 morphants when compared with controls. Moreover, the diminished muscle fibres of the morphants were abnormally organized, appearing short and irregular. These fibres lacked the parallel tight alignment seen in the muscle of controls and were scattered in the somitic compartment.

Previous work has revealed that DM1 skeletal muscle shows a characteristic type 1 (slow) fibre hypoplasia, and DM2 a type 2 (fast) fibre hypoplasia (Brooke and Engel, 1969; Tohgi et al., 1994; Vihola et al., 2003; Schoser et al., 2004). To investigate whether the hypoplasia in the mbnl2 morphants occurred at the expense of a particular fibre type, slow and fast sarcomeric myosins were targeted with antibodies S58 and MF20. S58 monoclonal IgA is specific for slow muscle fibre myosin heavy chain (MHC) isoforms (Miller et al., 1985). MF20 is an IgG2b monoclonal antibody, which reacts with all sarcomeric myosins (Bader et al., 1982). It was found that the amount of both fast and slow fibres was significantly reduced, and their distribution pattern was altered (Fig. 3C,D). As shown in Fig. 3, control embryos at 52 hpf showed the expected normal distribution of muscle fibres, with the slow fibres forming an even layer in the periphery (yellowish green), and the rest of the fibres (red) organized towards the centre. In transverse sections of 52-hpf mbnl2 morphants, both slow and fast muscle fibres were disorganized and scattered, and the outer layer of slow muscle was discontinuous, alternating with fast fibres and non-muscle DAPI-positive tissue. We performed fibre counting in confocal micrographs from three mbnl2 morphant embryos and found no difference in the proportion of fast and slow type fibres when compared with three age-paired controls. Additionally, we confirmed an overall reduction of 37.7% on average in the number of skeletal muscle fibres in the morphants (Table 1). Analysis of muscle ultrastructure by TEM was performed on the morphants and controls at the skeletal muscle and heart level (Fig. 3E–L). Sections of skeletal muscle of normal embryos showed the expected ultrastructure with well-aligned myofibrils and their Z-bands being regularly stacked. All the sarcomeric elements and triads were easily identifiable. By contrast, in age-paired mbnl2 morphants, there were fewer fibres compared with in the controls, and the myofibrils were scattered and disorganized, running in different directions. Some of the Z-bands were condensed, and the Z-bands of adjacent myofibrils were not in register. Often, it was observed that the H-zone and the I-band were absent from morphant sarcomeres, giving the impression of muscle hypercontraction. Less frequently, in some of the morphants the sarcoplasmic reticulum was disorganized, often lacking the well-ordered triads evident in the control embryos. At the heart level, light microscopy analysis of the mbnl2 morphant revealed dilated heart cavities, which were smaller in size than those of sibling controls (Fig. 3G,H). TEM analysis of heart tissue from the morphants showed streaming and condensation of the Z-bands in the sarcomere. Additionally, the H-zone and I-band were not recognizable in most of the sarcomeres, and myofibril disorganization and disintegration were also seen (Fig. 3I–L). Although there was obvious evidence of muscle disruption, the underlying cause was not clear. To determine whether this disruption is due to a failure of muscle differentiation, we examined the expression of the muscle marker MyoD. Preliminary analysis (supplementary material Fig. S6) revealed no gross level change to the distribution of MyoD expression, suggesting that the early pattern of muscle formation is maintained in mbnl2 morphants. However, MyoD expression was slightly reduced in the more anterior somites of the morphants at 18 hpf compared with wild-type embryos (supplementary material Fig. S6C,D). This difference was more pronounced in 24-hpf embryos (supplementary material Fig. S6E-H). Further work will be required to assess whether these differences represent primary changes or whether they are secondary effects due to other morphological constraints.

Dysregulation of alternative splicing of specific transcripts, including cardiac troponin T (TNNT2) and the muscle chloride channel (CLCN1), are associated with DM (Philips et al., 1998; Charlet et al., 2002; Mankodi et al., 2002). We examined the effect of knocking down the function of mbnl2 on the alternative splicing pattern of transcripts of orthologous zebrafish genes. The genomic organization and alternative splicing pattern of zebrafish clcn1 and scn4ab (voltage-gated sodium channel type IV, alpha subunit) were established by sequence analysis, as described for the MBNL genes. The exon layout and splicing events in zebrafish tnnt2 have been described elsewhere (Hsiao et al., 2003).

Alternatively spliced variants of tnnt2 were amplified from mbnl2 morphants and age-paired control embryos at 51 hpf by RT-PCR. When using primers located in exons 1 and 8, two main bands of 318 bp and 282 bp were revealed (Fig. 4A). Both products were detected in the controls, but only the 318 bp product was present in the morphants. Direct sequencing of the purified products showed that they contained variants of tnnt2 in which exons 4 and 6 are alternatively spliced. It was confirmed that the 318 bp product includes alternatively spliced exon 6, whereas the 282 bp product excludes it. To learn more about the splicing of tnnt2 transcripts during zebrafish development, we analyzed pools of embryos at different stages. We found that, during the earliest stages (18 somites and 24 hpf), the embryos express the 318 bp transcript. By contrast, later in development (51 hpf, 72 hpf and adults) both of the isoforms were clearly expressed (Fig. 4B).

The expression of clcn1 isoforms was also analyzed by RT-PCR using a set of primers designed to cover different exon splicing combinations. Fig. 4C shows that primers located in exons 15 and 21 produced two bands of 588 bp and 555 bp. These products were expressed at a ratio of 1:1.2 in control embryos. However, in the mbnl2 morphants there was a shift in the ratio of the two forms to 1:1.98 due to increased expression of the 555 bp isoform. After sequencing three independently amplified RT-PCR fragments, from both controls and morphants, it was found that the two products differ in the inclusion or exclusion of a newly identified 33 bp exon (18a), which was predicted to encode 11 amino acids (FTKLSPAGGVK) located in between the first and second cytoplasmic CBS domains, at the C-terminal end of the Clcn1 protein. A similar shift in the ratio of splice forms was observed in embryos treated with the other mbnl2-directed morpholinos. A further example of the skewed ratio of clcn1 isoforms is shown in supplementary material Fig. S7 following treatment with MO-4.

To determine whether the alternative splicing changes observed in the mbnl2 morphants represented specific events or global changes to alternative splicing, we analyzed the expression of zebrafish scn4ab in control and morphant embryos (supplementary material Fig. S8A,B). No difference in size or exon content for scn4ab transcripts was detected when using multiple primer combinations, restriction enzyme digests and sequencing (supplementary material Fig. S8C and Table S1). These results are comparable to those in DM patients, in whom the alternative splicing of the homologous gene is also unaltered (Charlet et al., 2002).

To further assess the specificity of the mbnl2 morpholino effects, we performed rescue experiments with full-length capped mbnl2 mRNA co-injected with morpholino MO-1 (Fig. 5 and Table 2). The amount of MO-1 injected was the optimum for producing a moderate-to-severe phenotype with a low mortality rate (4.28 ng). The dosage of mbnl2 mRNA was titrated to obtain the highest numbers of rescued embryos without the synthetic mRNA toxic effects. It was found that 150 pg mRNA was the most effective rescue dose. As a further control we also co-injected embryos with MO-1 and unrelated RNA, Xenopus elongation factor 1α, to ensure that the injected RNA did not affect morpholino activity non-specifically.

A complete rescue of the external phenotype was achieved in 64% of treated embryos, and a partial rescue was obtained in 32% of embryos (Fig. 5A–D and Table 2). Furthermore, recovery of the wild-type splicing pattern of tnnt2 and clcn1 transcripts was observed in rescued embryos (Fig. 5E,F). In addition, we tested the ability of the human mbnl2 homologue, MBNL2, to rescue the mbnl2 morphant phenotype: when using 75 pg of MBNL2 mRNA plus 4.28 ng of MO-1, up to 55% of embryos were completely rescued plus 24% were partially rescued (Table 2). The unrelated Xenopus RNA had no effect on the mutant phenotype. These results suggest that the mechanisms involved in MBNL2-dependent alternative splicing have been conserved throughout vertebrate evolution.

At present, there is substantial evidence to support an RNA-mediated pathogenic mechanism for DM1 and DM2, in which pathogenic RNAs accumulate within the nuclei of DM cells, where they sequester specific RNA-binding proteins such as MBNL2, causing disruptions to different cellular processes. Currently, the functions of the three MBNL genes are not completely understood. It has been suggested that MBNL proteins participate in regulating the alternative splicing of specific transcripts such as insulin receptor (IR) and TNNT2 (Ho et al., 2004). Interestingly, an additional function has been suggested involving MBNL2 in the localization of the mRNA of transmembrane protein integrin α3 to focal adhesion sites (Adereth et al., 2005).

A variety of animal models have been produced to gain insight into the molecular basis of DM. However, none of them have recapitulated all of the signs and symptoms of this disease, reinforcing the idea that DM is a disease with an unconventional complex pathophysiology. Nonetheless, different lessons can be learned from each model. In particular, Mbnl1 knockout mice show DM features including myotonia, abnormal central nuclei in myofibres, ocular cataracts and splicing alterations of Clcn1, Tnnt2 and fast skeletal muscle troponin T (Tnnt3) (Kanadia et al., 2003). Some of these defects, namely myotonia and chloride channel splicing abnormalities, are shared by a recently published knockout model in mice for Mbnl2 (Hao et al., 2008), produced by disruption of the protein at the beginning of the second zinc finger. However, a similar knockout model, in which mice were generated by a gene trap vector targeted to intron 4, in which the first two zinc fingers of Mbnl2 are conserved, shows no specific phenotype (Lin et al., 2006). We observe a consistent phenotype with both translation-blocking and splice-blocking morpholinos, which have a different mechanism of action. Our results suggest that Mbnl2 deficiency in zebrafish produces features more similar to those reported in the mouse by Hao et al. (Hao et al., 2008). At the muscle level, apart from the obvious differences in muscle organization between the two species, increased heterogeneity and an overall reduction in skeletal muscle fibres was reported in both. Similarly, these authors also report a modest increase in the presence of an alternative RNA splice form of Clcn1, and normal splicing of other transcripts tested, whereas clcn1 splicing defects are more prominent in our zebrafish morphants and are accompanied by notable alterations of tnnt2 (not examined in the mouse model). The absence of reported developmental defects in the Mbnl2-deficient mice might be due to evolutionary divergence. Analysis of muscleblind (Mbl)-deficient flies reveals developmental defects of the eye and muscle, with disorganization of sarcomeres, muscle contraction and failure to survive beyond the first instar larva stage. This demonstrates that a higher deficiency of Muscleblind function can affect the development of specific organs in other species.

We have identified and characterized three MBNL genes in zebrafish as a counterpart of those in humans. Our data is consistent with the recently published findings of another research group (Liu et al., 2008). In order to further dissect the function of mbnl2, we generated a zebrafish model using antisense technology in a loss-of-function approach. The antisense morpholino molecule was designed to block the translation of the entire mbnl2 mRNA. The phenotype obtained in mbnl2 morphants was demonstrated to be specific and dose dependent, and included morphological abnormalities at the eye and skeletal muscle levels. Our results were consistent with those previously reported in Drosophila, in which a role in photoreceptor and muscle development was established (Begemann et al., 1997; Artero et al., 1998). Moreover, brain and cardiac defects were also part of the phenotype of mbnl2 morphants. Whether these were due to unknown activity of Mbnl2 during development in vertebrates, or to a secondary effect on downstream genes, remains to be elucidated.

Dysregulation of alternative splicing of specific transcripts, including TNNT2 and CLCN1, is a key feature of DM and is thought to contribute to the production of cardiac abnormalities (TNNT2) and myotonia (CLCN1) in DM patients (Philips et al., 1998; Charlet et al., 2002; Mankodi et al., 2002). Likewise, we found that tnnt2 and clcn1 alternative splicing patterns are disrupted in mbnl2 morphants, which show cardiac and skeletal muscle abnormalities. Interestingly, the analysis of scn4ab, another ion channel linked to myotonia, showed no splicing alterations, consistent with the absence of such changes in DM patients (Mankodi et al., 2005). In the case of clcn1, we showed the presence of two distinctive isoforms expressed in almost equal proportions in the wild type, whereas mbnl2 morphants were found to preferentially express the 555 bp form. Additionally, we established that the splicing pattern of tnnt2 shown by mbnl2 morphants was similar to that observed in fish during earlier developmental stages. In a similar manner, DM adult patients express a TNNT2 isoform that includes an exon preferentially expressed in embryonic tissue but not in adult controls (Philips et al., 1998). These observations support a role for MBNL2 as a regulator of developmentally programmed alternative splicing of specific transcripts. Our results demonstrate that the loss of mbnl2 function is sufficient to produce alternative splicing defects in transcripts of particular genes that are involved in DM.

Originally, myopathic changes in DM muscle, including fibre size variation, seemed to be similar in both DM1 and DM2 (Schoser et al., 2004). However, some distinction emerged when using immunohistochemistry to assess fibre type differentiation. DM1 muscle exhibited a characteristic type 1 (slow) fibre atrophy, in contrast with type 2 (fast) fibre atrophy in DM2 (Brooke and Engel, 1969; Vihola et al., 2003). We set out to investigate whether the mbnl2-deficient fish showed any muscle alteration parallel to those observed in DM. First, it was observed that the mbnl2 morphants displayed an abnormal somite shape and overall reduction of the number of muscle fibres. Immunohistological studies were used to assess fibre type proportion, which confirmed that the number of both fast and slow fibres was significantly reduced, and their distribution pattern was altered, in mbnl2 morphants. In addition, electron microscopic studies on skeletal and cardiac muscle showed impaired alignment of myofibrils, abnormal Z-bands and disorganization of the sarcoplasmic reticulum, all of which are features shared with DM muscle (Klinkerfuss, 1967; Ludatscher et al., 1978).

In summary, our study provides broader insight into the function of MBNL2 as a splicing regulator and as a major participant during muscle, heart, eye and brain development. We showed that key features of DM are replicated in our zebrafish mbnl2 knockdown model, with ultrastructural and molecular parallels. Taken together, the results presented herein support the idea that the sequestration of MBNL2 is an important determinant of muscle pathophysiology and splicing disruption in DM, which must be considered when developing therapeutic approaches for this condition.

In view of the accessibility of the zebrafish as a model for developmental studies, our data provides the starting point for experiments to dissect the molecular pathway producing muscle weakness and wasting in DM.

Human MBNL nucleotide and protein sequences were obtained from public databases (NCBI, ENSEMBL, ZFIN) and used to retrieve ESTs and mRNA sequences from zebrafish. ESTs containing a full open reading frame were computationally predicted based on homology with known genes. The corresponding cDNA clones were obtained from MRC geneservices and RZPD, and were fully sequenced by primer walking. Alignments and phylogenetic analyses were carried out using BLAST and ClustalW tools, and BioEdit and TreeView software. The phylogenetic tree was constructed by the neighbour-joining method (Pearson et al., 1999). The zebrafish genomic sequences used to annotate the MBNL genes correspond to Zv4 and Zv5 assemblies from the Danio rerio Sequencing Project.

Zebrafish stocks were maintained as described (Westerfield, 2000). Two translation blocking morpholino oligonucleotides were designed for mbnl2 and manufactured by Gene Tools: MO-1: 5′-TAAAGCCATAGTTGTGTTGTGAATG-3′ and MO-2: 5′-GTTTGTGTGTTGCTGTACTTTTGAG-3′, which was designed upstream of MO-1. Wild-type zebrafish embryos at the one- to four-cell stage were microinjected with 1 nl of titrated doses of morpholino to determine the efficiency/toxicity window. Siblings from the same pool of morpholino-injected embryos served as non-injected controls. Additionally, a standard control morpholino, 5′-CCTCTTACCTCAGTTACAATTTATA-3′ (Gene Tools), was injected into a group of control embryos. Full-length capped mbnl2 and MBNL2 mRNAs were synthesized using the mMessage mMachine kit (Ambion) according to the manufacturer’s instructions, and cleaned up using Ambion’s MEGAclear kit. Titrated doses of the resulting capped mRNA were injected alone and in conjunction with either control or MO-1 morpholinos. All experimental procedures complied with institutional and national animal welfare laws, guidelines and policies.

Zebrafish mbnl2 plasmid DNA was linearized with EcoRI-XhoI and then purified by phenol/chloroform/isoamyl alcohol extraction and subsequent ethanol precipitation. A pax2a antisense probe was produced after linearizing the corresponding plasmid cDNA with BamHI as described elsewhere (Brand et al., 1996; Majumdar et al., 2000). Sense and antisense riboprobes were obtained by in vitro transcription of 0.2 pmol of linear plasmid DNA. Reactions were set up with 10–40 U of the corresponding (SP6, T7 or T3) RNA polymerase (Roche), 10× concentrated transcription buffer [400 mM Tris-HCl (pH 8.0; 20°C), 60 mM MgCl₂, 100 mM dithiothreitol (DTT), 20 mM spermidine], 20 U RNase inhibitor (RNasin, Promega), 10 mM ATP, 10 mM CTP, 10 mM GTP, 6.5 mM UTP, 3.5 mM DIG-UTP and sterile distilled water to make a total volume of 20 μl. The reaction was incubated overnight at 37°C. The resulting DIG-labelled RNA was precipitated with 0.3M NH₄Ac and ethanol, and was resuspended in DEPC-treated water. The in situ hybridization process was carried out as described (Thisse et al., 2004). The embryos were mounted in glycerol and the results registered using a Nikon digital camera attached to a SMZ1000 Nikon microscope with a cold light source (Photonic). The images were processed with ACT-2U (Nikon) imaging software.

After harvesting at different developmental stages, the fish embryos were immersed in 30% sucrose for 60 minutes and frozen down in OCT (RA Lamb) using liquid nitrogen cooled isopentane. 20-μm sections were cut on a cryostat (Microm HM505E) and collected on 3-aminopropyltriethoxysilane (APES)-coated glass slides. The sections were post fixed with 0.5% paraformaldehyde (PFA) for 3 minutes, washed three times in PBS, and permeabilized by incubation with 0.15% Triton X-100 (Promega) for 10 minutes. Blocking was carried out by incubation with 5% bovine serum albumin (BSA; Roche) for 30 minutes. Antigens of interest were labelled first by incubating with S58 anti-slow MHC primary antibody [DSHB (Developmental Studies Hybridoma Bank, University of Iowa, NICHD); diluted 1:50] and visualized by incubating with anti-mouse Alexa Fluor 488 secondary antibody (Invitrogen; diluted 1:300), then by incubating with MF20 anti-sarcomere myosin primary antibody (DSHB; diluted 1:50) and visualized by incubating with anti-mouse Alexa Fluor 555 secondary antibody (Invitrogen; diluted 1:300). For the fibre migration study, the F59 antibody (DSHB) was used at a dilution of 1:50 and visualized with Alexa Fluor 555. Sections were mounted and coverslips applied with VECTORSHIELD containing DAPI (Vector Labs). Sections were analyzed and photographed using fluorescent microscopy (Nikon Optiphot-2 microscope with Spot Insight imaging system). Further imaging was performed at room temperature on a confocal laser-scanning microscope (Zeiss LSMuv META Combi confocal on a Zeiss Axiovert 100 microscope). Stacks were acquired and all the images were processed and analyzed using Zeiss LSM software.

Selected zebrafish embryos were dechorionated manually and killed using an overdose of anaesthetic ethyl m-aminobenzoate tricane methane sulfonate (MS-222; Sigma). The embryos were then fixed in 3% glutaraldehyde in PBS for a minimum of 1 hour. Secondary fixation was achieved by incubation in 1% aqueous osmium tetroxide (OsO₄) for 1 hour. Subsequently, the fish were rinsed in water and dehydrated in an acetone series (10 minutes each in 50%, 70%, 80%, 90%, 95%, 100% and dried), infiltrated with 1:1 Araldite:acetone overnight, embedded in Araldite (Agar Scientific) and finally polymerized on oven at 50°C. In order to ensure adequate orientation of the specimens, most of the fish were cut into two regions, thorax and tail. These pieces and further sections cut from the tail were embedded separately. Semi-thin sections (0.5 μm) were first cut using an Ultracut E microtome (Reichert-Jung), stained with toluidine blue and examined by light microscopy to observe the morphology. Ultra-thin sections of about 100 nm were cut and then collected on a copper grid and stained with 2% uranyl acetate and Reynold’s lead citrate. The sections were examined using a JEOL 1200 electron microscope and images recorded at various magnifications between 5000× and 20,000× on Kodak SO-163 electron microscope film.

Sample homogenates were obtained by grinding 20–25 mg of zebrafish embryo samples in lysis buffer (QIAGEN) using a Polytron blender. Total RNA was extracted from pools of zebrafish embryos using the QIAGEN RNeasy kit, according to the manufacturer’s recommendations. The first-strand cDNA was synthesized from 1–2 μg of total RNA using anchored oligo-dT for tnnt2, and a mix of anchored oligo-dT 1:3 random hexamers in the case of clcn1 and scn4ab. A blend of AMV and MMuLV reverse transcriptases, 5× buffer (ABgene), 5 nm dNTPs and RNase inhibitor was then added, and the samples were incubated at 55.6–59°C for 90 minutes, followed by 10 minutes at 75°C for enzyme inactivation. Different sets of primers were designed according to the alternative splicing analysis of each transcript. Zebrafish tnnt2 was amplified by RT-PCR with primers published elsewhere (Hsiao et al., 2003). Zebrafish clcn1 and scn4ab were amplified using primers listed in supplementary material Tables S2 and S3. PCR amplification of reverse transcribed products was carried out in 25 μl reactions, containing 200 nM primers, 100–250 ng cDNA, 350 μM dNTPs, 2.25 mM MgCl₂ and 1.25U Taq Extensor Hi-fidelity DNA polymerase (ABgene). The amplification cycles were performed as follows: 94°C for 2 minutes, with 25–30 cycles of 94°C for 10 seconds, 50–60°C for 10 seconds and 68°C for 1 minute followed by an extension cycle of 68°C for 5 minutes. RT-PCR products were resolved by electrophoresis on 1.5–2.5% agarose gels and visualized with ethidium bromide. The gels were scanned and the density of the bands was quantified using ImageJ software (Wayne Rasband, NIH). Percentage exon inclusion was calculated as: [exon inclusion band/(exon inclusion band + exon exclusion band)] × 100. All the samples were analyzed in triplicate.

Restriction enzyme digests were performed on purified RT-PCR products produced with primers for zebrafish scn4ab. The following restriction enzymes were used: NdeI + BclI, BamHI + SspI, EcoRI + XbaI, SspI + BmrI (New England Biolabs). Reactions of 10 μl were set up with 1–3 U of restriction enzyme per μg DNA with 1μl 10× buffer and 0.1 μl 100× BSA (10 mg/ml), and were incubated overnight at the corresponding temperature according to the manufacturer’s instructions.

RT-PCR and restriction digestion products from tnnt2 and scn4ab were excised from agarose gels and purified using the QIAquick PCR purification kit (QIAGEN) for direct sequencing. Additionally, clcn1 RT-PCR products were cloned into T-easy vector (Promega). Sequencing reactions were set up using Big Dye Terminator (Applied Biosystems) and visualized on an ABI 3100 Genetic Analyser (Applied Biosystems). Gap4 and Pregap4 programs from the Staden Package (Staden, 1996) (MRC Laboratory of Molecular Biology, Cambridge, UK) were used for multiple DNA sequence assembly.

The corresponding human and murine transcripts and proteins were compared against the whole zebrafish genome and expressed sequences available from public databases. Zebrafish ESTs and mRNAs that exhibited high similarity with the query sequences were selected. The cDNA clones corresponding to ESTs with the highest scores were retrieved, cultured, sequenced by primer walking and analyzed. Three different populations of cDNA were characterized, corresponding to alternatively spliced isoforms derived from three orthologous MBNL genes, which showed high identity with their human counterparts. The four distinctive CCCH zinc-finger domains, organized in two pairs (CX₇CX₆CX₃H and CX₇CX₄CX₃H), as well as the regions enriched in alanine and proline, were identified in the three zebrafish MBNL proteins. The percentage of identity between human and zebrafish MBNL homologues was 79.41% for MBNL1, 80% for MBNL2 and 71.91% for MBNL3, at a protein level. The highest level of identity, near 100%, was found at the four zinc-finger domains level, whereas the highest divergence was found in the sequence located between and after the two sets of zinc fingers (supplementary material Fig. S1A). When other vertebrate and non-vertebrate species were included in the alignment, a similar conservation pattern was observed (data not shown). To examine the evolutionary relationship between zebrafish and human MBNL proteins, we constructed a phylogenetic tree based on full-length alternatively spliced forms of the three members of this family (supplementary material Fig. S1B). Zebrafish mbnl1, mbnl2 and mbnl3 clustered with their human counterparts into three distinct monophyletic groups. Orthologous relationships between zebrafish and humans in the three branches were supported.

To determine whether these observations were biased by sequence variation in the highly divergent regions, a phylogenetic tree based on partial sequences that omitted such regions was constructed. Nevertheless, there was no significant difference in the tree topology. During the preparation of this manuscript, an independent group reported an annotation for the zebrafish MBNL family of genes (Liu et al., 2008). Our annotation shows slight differences with those of Liu et al.

Total RNA was isolated from zebrafish embryos at 51 hpf in batches of 50–80 embryos, using a QIAGEN RNeasy Fibrous Tissue kit. First-strand cDNA was synthesized using 1 μg of total RNA with M-MuLV reverse transcriptase (Thermo Scientific) in a final volume of 25 μl including 1× RT buffer, 1 mM dNTPs, 20 units of RNase inhibitor and 1 μl of random hexamers (400 ng/μl) made up with DEPC water. The reaction mixture was incubated at 37°C for 2 hours.

Primers for the real-time assay were designed using the Universal Probe Library Assay design centre software on the Roche-Applied-Science website (see Table 3).

The real-time PCR assay was performed on an ABI 7500 Fast Real Time PCR system using a SensiMix Real-Time kit according to the manufacturers’ instructions. Amplification conditions were 95°C (10 minutes), followed by 40 cycles of 95°C (15 seconds) and 60°C (60 seconds). β-actin1 and β-actin2 were used as reference gene controls and a no-template control was also included. Each sample was amplified in triplicate and three biological replicates were used. A standard curve of four dilution points (in steps of fivefold) was generated for the wild-type cDNA and each test cDNA was diluted 1:5. The relative expression of each gene was calculated using the ABI 7500 software to produce plots illustrating relative fold changes in respect to sample.

Additional primer sequences used were: F3 + R7 Primers, F3 5′-CAAACGCGCAGACTATTTCA-3′ and R7 5′-TCTGCTGG-ATGAGGTTGTTG-3′.

This work was supported by the Muscular Dystrophy Association USA, the Muscular Dystrophy Campaign, the British Heart Foundation and the Wellcome Trust.