Correction of GSK3β at young age prevents muscle pathology in mice with myotonic dystrophy type 1

Abstract

Myotonic dystrophy type 1 (DM1) is a progressive neuromuscular disease caused by expanded CUG repeats, which misregulate RNA metabolism through several RNA-binding proteins, including CUG-binding protein/CUGBP1 elav-like factor 1 (CUGBP1/CELF1) and muscleblind 1 protein. Mutant CUG repeats elevate CUGBP1 and alter CUGBP1 activity via a glycogen synthase kinase 3β (GSK3β)–cyclin D3–cyclin D-dependent kinase 4 (CDK4) signaling pathway. Inhibition of GSK3β corrects abnormal activity of CUGBP1 in DM1 mice [human skeletal actin mRNA, containing long repeats (HSA^LR) model]. Here, we show that the inhibition of GSK3β in young HSA^LR mice prevents development of DM1 muscle pathology. Skeletal muscle in 1-yr-old HSA^LR mice, treated at 1.5 mo for 6 wk with the inhibitors of GSK3, exhibits high fiber density, corrected atrophy, normal fiber size, with reduced central nuclei and normalized grip strength. Because CUG–GSK3β–cyclin D3–CDK4 converts the active form of CUGBP1 into a form of translational repressor, we examined the contribution of CUGBP1 in myogenesis using Celf1 knockout mice. We found that a loss of CUGBP1 disrupts myogenesis, affecting genes that regulate differentiation and the extracellular matrix. Proteins of those pathways are also misregulated in young HSA^LR mice and in muscle biopsies of patients with congenital DM1. These findings suggest that the correction of GSK3β-CUGBP1 pathway in young HSA^LR mice might have a positive effect on the myogenesis over time.—Wei, C., Stock, L., Valanejad, L., Zalewski, Z. A., Karns, R., Puymirat, J., Nelson, D., Witte, D., Woodgett, J., Timchenko, N. A., Timchenko, L. Correction of GSK3β at young age prevents muscle pathology in mice with myotonic dystrophy type 1.

Myotonic dystrophy type 1 (DM1) is a multisystemic disease that affects skeletal muscle, heart, brain, eye, and the endocrine system (1). Patients with DM1 have CTG triplet-repeat expansions in the 3′ UTR of the dystrophia myotonica protein kinase (DMPK) gene (2). The length of the CTG repeats correlates with the severity of the DM1, with the longest expansions (>1000 CTG repeats) found in the most severe forms of the disease, which affects infants [congenital myotonic dystrophy type 1 (CDM1)]. Skeletal muscle in patients having the adult form of DM1 is characterized by progressive weakness, wasting, and myotonia. In severe CDM1, skeletal muscle is underdeveloped and extremely weak. There is no cure for CDM1 or DM1.

It has been established that the DM1 mutation leads to the accumulation of the mutant DMPK transcripts, which misregulate RNA homeostasis in patients’ tissues by interfering with RNA-binding proteins (3–7). Two of the most-studied proteins, associated with the disruption of RNA processing in DM1, are muscleblind 1 (MBNL1) and CUGBP1 [(also known as CUGBP1 elav-like factor (CELF1)] (3–7). Other RNA-binding proteins are also affected in DM1, including Staufen1 and RNA helicases DDX5 and DDX6 (8–11).

Degradation of the mutant DMPK mRNA using antisense oligonucleotides or by increase of RNA helicase p68 corrects muscle pathology in HSA^LR mice (10, 12). Normalization of the activities of RNA-binding proteins, misregulated by CUG repeats (CUGBP1 and MBNL1), was also shown to be beneficial for the reduction of DM1 muscle pathology in HSA^LR mice (12). Previous studies have shown that CUG repeats increase CUGBP1 stability (13, 14). In addition, CUG repeats misregulate CUGBP1 activity through reduced phosphorylation at S302 (12, 15–17). In healthy muscle, CUGBP1 phosphorylated at S302 binds to the active eIF2α and increases protein translation of CUGBP1 targets (12, 15, 16). However, CUGBP1, unphosphorylated at S302, binds to the inactive form of eIF2α and inhibits protein translation (16). Levels of unphosphorylated (unp)-S302–CUGBP1 are very low in healthy muscle; however, they are significantly increased in DM1. The increase of unp-S302–CUGBP1 in DM1 is caused by increased levels of active GSK3β, which reduces cyclin D3–CDK4 kinase phosphorylation of CUGBP1 at S302 (17). Inhibition of GSK3β in adult HSA^LR mice with inhibitors of GSK3, such as lithium and TDZD-8, corrects CUGBP1 activity and significantly reduces DM1 muscle pathology. The degree of beneficial effect of inhibitors of GSK3 depends on the age of the HSA^LR mice. Although lithium effectively reduced muscle pathology in adult HSA^LR mice, the grip weakness reappeared once treatment stopped. In contrast, inhibition of GSK3β in young HSA^LR mice for 1 wk was sufficient to maintain improved grip strength for an additional 3 wk. These findings prompted us to examine whether correction of myogenesis via the GSK3β-CUGBP1 pathway in young HSA^LR mice before the occurrence of pathologic changes might have long-term benefits for muscle health in those mice.

MATERIALS AND METHODS

Antibodies and reagents

Antibodies to actin (I-19), α-actinin (H-300), β-actin (C-4), calreticulin (H-170), cyclin D3 (C-16), collagen 4A (E-14), CUGBP1 (3B1), desmin (H-76), Pax7 (EE-8), phosphorylated (p-)S51-eIF2α (sc-101670), and troponin T-FS (H-8) (all from Santa Cruz Biotechnology, Dallas, TX, USA); GSK3β (3891) and DCX (4604) (Cell Signaling Technology, Danvers, MA, USA); LEF1 (Thermo Fisher Scientific, Waltham, MA, USA); and myosin skeletal fast (MY-32; MilliporeSigma, Billerica, MA, USA) were used. Two affinity-purified antibodies to RBM45 from MilliporeSigma (HPA020448 and AV41154) were also used. Antibody 3B1 recognizes both active (p-S302) and inactive (unp-S302) CUGBP1. To distinguish those forms of CUGBP1, activity of CUGBP1 was tested by its interaction with inactive p-S51–eIF2α using a combination of the immunoprecipitation (IP) of CUGBP1 with antibody 3B1 and Western blot analysis of the CUGBP1-IPs with antibodies to p-S51–eIF2α. 6-Bromoindirubin-3′-oxime (BIO) and indirubin were from MilliporeSigma.

HSA^LR mice and treatments

All experiments with mice were approved by Institutional Animal Care and Use Committee at Baylor College of Medicine and Cincinnati Children’s Hospital Medical Center. HSA^LR mice expressing 250 CTG repeats under human skeletal muscle promoter (strain 20LRb), provided by Dr. Charles Thornton (University of Rochester, Rochester, NY, USA) were described previously (6). HSA^LR mice were bred as a homozygous (hom) genotype, and the expression of the mutant RNA CUG repeats was confirmed by FISH analysis (10). As controls, wild-type (WT) mice of FVB (Friend leukemia virus B) background from The Jackson Laboratory (Bar Harbor, ME, USA) were used. All mice were housed under standard conditions with free access to food and water.

Optimization of BIO and indirubin doses was performed in pilot experiments based on normalization of the levels of active GSK3β, monitored by Western blot analysis. Our previous work showed a positive correlation between normalization of the levels of active GSK3β and improvement of grip strength in HSA^LR mice; therefore, the reduction of the grip weakness in the treated HSA^LR mice was used as an indicator of the effectiveness of inhibitors (17).

Initially, the effectiveness of 2 doses of induribin (2.0 and 5.0 μg/g of body weight) administered intraperitoneally was tested. Adult (5–6-mo-old) HSA^LR mice of both genders (n = 5–6) were treated for 4 wk with indirubin every other day. Grip strength of the front paws in the gender- and age-matched, WT, treated and untreated HSA^LR mice was measured before and after the treatment. Grip strength steadily improved in all adult HSA^LR mice treated with indirubin at both doses, indicating that those doses of indirubin or its derivatives (such as BIO) were efficient for the normalization of GSK3β in HSA^LR mice.

Two groups (n = 6) of 1.5-mo-old HSA^LR mice were treated i.p. with indirubin (2.0 μg/g) or BIO (3.6 μg/g) for 6 wk every 48 h. Grip strength of the front paws in the gender- and age-matched, WT, treated and untreated HSA^LR mice was measured before the treatments, every other day during first week of the treatment, and then almost every month up to 12 mo old. WT, untreated and treated HSA^LR mice were euthanized at 12 mo old, and muscles of several groups, including gastrocnemius (gastroc) were collected for histologic and biochemical analyses.

Histologic analysis

Hematoxylin and eosin staining was performed using paraffin-imbedded muscle (gastroc) sections from the matched WT mice and untreated and treated HSA^LR mice. The total number of fibers was counted in transverse sections of entire gastroc in each mouse group, and the average numbers are presented. Average fiber size was calculated based on the analysis of 400 fibers for each mouse group. The numbers of centralized nuclei and bundles were counted in ×20 magnification views, with a total 100 views for each mouse group using MetaMorph software (Molecular Devices, Sunnyvale, CA, USA), and average data per ×20 view are presented.

Generation of Celf1 knockout mice

Celf1 knockout (KO) mice were generated using BayGenomics embryonic stem cells (ES) from the Mutant Mice Regional Resource Center at the University of California (Davis, CA, USA). The ES contain the insertion of the trap vector, in front of AUG codon of the Celf1 gene. The disruption of the Celf1 gene was confirmed by sequencing of ES. Those ES were used for the microinjection of C57BL/6 blastocysts at the Transgenic Mouse Core facility at Baylor College of Medicine. Genotyping of Celf1 KO mice was performed by quantitative PCR using primers specific for the trap vector of the following sequences: forward primer, 5′-GTTGCAGTGCACGGCAGATACACTTGCTGA-3′; reverse primer, 5′-GCCACTGGTGTGGGCCATAATTCAATTCGC-3′. The size of the PCR product was 389 nt.

Electron transmission microscopy

Electron transmission microscopy was performed as previously described (10). Fresh mouse tissues were fixed overnight in 4% formaldehyde and for 24 h in 3% glutaraldehyde, containing 0.4% of tannic acid. Muscle samples were postfixed in 1% osmium tetroxide, counterstained with 1% uranyl acetate for 2 h, and infused with epoxy resin. Ultrathin sections, contrasted with uranyl acetate and Sato lead stain, were examined in the Electron transmission microscopy Pathology Core of Cincinnati Children’s Hospital Medical Center at magnifications of ×7000–20,000; 300–600 myofibrils from the hind limb muscles of matched WT and hom Celf1 KO mice were analyzed at magnification ×7000, and the numbers of myofibrils and sarcomeres per view were counted. The average length of the sarcomeres (pixels) was examined by analysis of 3700 sarcomeres in WT muscle and 1850 sarcomeres in KO muscle in 40 images taken at magnification ×7000. The number of sarcomeres in WT and KO muscle was counted in 40 views at ×7000, and average number per view was deduced.

Microarray analysis

Gene expression analysis of mouse tissue samples was performed at Miltenyi Biotec (Bergisch Gladbach, Germany) using Agilent Technologies (Santa Clara, CA, USA) whole mouse genome oligo microarrays. Three-replicate (hind limb muscle) and 4-replicate (whole brain) samples from the 1-d-old WT and hom Celf1 KO mice were used. Student’s t test (2-tailed, equal variance) was applied for normalization of log₂-intensity data. Differences of expression were considered significant if the uncorrected value was P ≤ 0.05. The genes showing ≥1.5-fold average expression difference were identified. The functional grouping was determined using Gene Ontology Consortium data. A statistical test (Fisher’s exact test with Benjamin-Hochberg correction of multiple testing) was also performed. Values of less than 0.05 showed a significant enrichment relatively to the background. Expression data may be accessed through the National Center for Biotechnology Information (Bethesda, MD, USA) Gene Expression Omnibus, under accession number GSE85396.

Human myoblast cell culture

Normal and CDM1 myoblasts were generated from biopsied muscle from patients with healthy muscle histopathology and healthy metabolism and from patients with CDM1 containing 2000 CTG repeats. Human myoblasts were maintained in an F10 medium (Thermo Fisher Scientific) containing 15% fetal bovine serum (GE Healthcare Life Sciences, Marlborough, MA, USA), 1% sodium bicarbonate (Thermo Fisher Scientific), 5% defined supplemental calf serum (GE Healthcare Life Sciences), 1% l-glutamine (Thermo Fisher Scientific), and 1% penicillin/streptomycin (GE Healthcare Life Sciences). Differentiation was induced by switching growth medium to the fusion medium containing DMEM, supplemented with horse serum and insulin for 3–5 d. Growth medium was changed every other day, and fusion medium was changed every day. Where indicated, BIO was added at a dose of 1 μg/ml for 24 h.

Western blot and co-IP analyses

Human myoblasts and myotubes were pelleted, and total protein extracts were purified with RIPA buffer. Mouse muscle tissues were homogenized in RIPA buffer, and total proteins were collected; 50 µg of proteins were separated by the SDS-gel electrophoresis, transferred onto membrane, and probed with antibodies, according to the manufacturer protocols. The protein signals were quantified by the scanning densitometry relatively to β-actin or calreticulin. Mean deviations were determined based on 3 repeats.

In the co-IP–Western blot assay, CUGBP1 was immunoprecipated from gastroc of 12-mo-old WT, untreated HSA^LR mice and of HSA^LR mice treated at a young age with BIO using 3B1 antibodies and the protein A resin. The CUGBP1-IPs were carefully washed with PBS, resuspended in the loading buffer, and examined by Western blot with antibodies to p-S51-eIF2α.

Immunofluorescence analysis

Paraffin sections from hind-limb muscles of 1-d-old WT and hom Celf1 KO mice were analyzed by immunofluorescence (IF) using antibodies to actin, α-actinin, fast myosin, troponin T, and desmin, according to the manufacturer recommendations. Frozen human muscle sections (quad) from 2 patients with CDM1 (1 and 3 yr old) and age- (1 and 3 yr) and gender-matched, healthy controls (with healthy muscle histology and without diagnostic abnormalities) were examined with antibodies to DCX, LEF1, RBM45, and Col4A. Control staining without first antibodies confirmed the specificity of the signals. The images were analyzed under the same exposure time and brightness, allowing quantification of the IF signals, using Nikon (Tokyo, Japan) microscope.

Quantitative RT-PCR

Total RNA (2 μg), isolated from the hind-limb muscle of 1-d-old mice, was subjected to reverse transcription, using a high-capacity cDNA Reverse Transcription kit (Thermo Fisher Scientific). RT-PCR assay was performed with the TaqMan gene expression system (Thermo Fisher Scientific) under conditions recommended by the manufacturer. The RT-PCR assay (a volume of 10 μl) contained 3 μl of cDNA (diluted 1:5 with diethyl pyrocarbonate–treated water), 5 µl of the TaqMan Master mix, and 0.5 µl of the gene-specific TaqMan Assay probe purchased from Thermo Fisher Scientific. The following mouse probes were used: LEF1, Mm00550265_m1; RBM45, Mm00463002_m1; DCX, Mm00438400_m1; and β-actin, Mm02619580_g1. The sequences of the primers for Col4A were as follows: 5′-ATTTCCAGGTGTGGATGGTG-3′ (forward) and 5′-GCCACCAGTAATGTACACAG-3′ (reverse).

Statistical analysis

Wherever possible, data were analyzed in a blinded manner. Western blot images were analyzed by scanning densitometry using 3 repeats of each experiment. Signals of proteins were normalized to signals of control protein actin. Data are presented as means ± sem. IF images were obtained under quantitative conditions using the same time of exposure and brightness. Statistical analysis was performed using 2-way ANOVA and 2-tailed Student’s t test. A value of P < 0.05 was considered statistically significant. To determine differences between expected and observed ratios of WT, heterozygous (het), and hom Celf1 KO mice, Fisher’s exact test was used.

RESULTS

Inhibition of GSK3β in young HSA^LR mice prevents DM1 muscle pathology

To examine the effect of the correction of GSK3β on DM1 muscle pathology over time, two groups of 1.5-mo-old HSA^LR mice (6 mice/group) were treated for 6 wk with BIO or indirubin, and grip strength was monitored in treated mice for 9 mo after completion of treatment. WT and untreated HSA^LR mice (n = 6) were analyzed in parallel with treated mice. We found that the grip strength in the untreated 3–12-mo-old HSA^LR mice was significantly less than that of matched WT mice (Fig. 1A). In contrast, HSA^LR mice treated at 1.5 mo of age for 6 wk with BIO had almost normal grip strength shortly after treatment (at 3 mo of age), and this near-normal grip strength was maintained for 9 mo after completion of the treatment. Similar results were obtained when HSA^LR mice were treated at 1.5 mo of age for 6 wk with indirubin (data not shown). Histologic analysis of the whole gastroc showed that the gastroc in the 12-mo-old HSA^LR mice in which GSK3β was inhibited at 1.5 mo of age with BIO had improved appearance, with increased fiber bundling (Supplemental Fig. 1). Microscopic analysis at the higher magnification (×20) showed a reduction of hypertrophic fibers and a reduction of fibers with central nuclei (Fig. 1B).

Figure 1.

Inhibition of GSK3 in young HSA^LR mice prevents grip weakness and reduces muscle histopathology over time. A) Longitudinal analysis of grip strength (calculated as a ratio to the mouse body weight) of WT, untreated HSA^LR mice, and HSA^LR mice treated at 1.5 mo for 6 wk with BIO. The means ± sd are shown. *P < 0.05, **P < 0.01, ***P < 0.001 for the matched untreated HSA^LR vs. the treated HSA^LR mice and for the untreated HSA^LR mice vs. the WT mice. B) Hematoxylin and eosin staining for images of the gastroc from 12-mo-old WT and HSA^LR mice, untreated, and treated briefly at a young age with BIO. Arrows point to myofibers in the gastroc from untreated HSA^LR mice, containing central nuclei. As an example of fiber-size variability, enlarged and small fibers in the gastroc of untreated HSA^LR mice are shown with red and yellow stars, respectively. Original magnification ×20; Scale bars, 20 μm.

Total number of fibers was reduced in the gastroc of 12-mo-old, untreated HSA^LR mice relatively WT mice (Fig. 2A). However, the total number of fibers in gastroc from 12-mo-old HSA^LR mice that had been briefly treated with BIO at 1.5 mo was significantly increased. The fibers in HSA^LR muscle were varied in size, with the presence of hypertrophic, small, and split fibers (6). The presence of hypertrophic fibers led to an increase in the average fiber size in 12-mo-old, untreated HSA^LR muscle (Fig. 2B). We found that the abnormal myofiber size was significantly reduced in HSA^LR mice, in which GSK3β was inhibited at a young age for 6 wk with BIO (Figs. 1B and 2B). The number of fibers with central nuclei was also significantly reduced in the muscle of 12-mo-old, treated HSA^LR mice (Figs. 1B and 2C). The number of fiber bundles was decreased in untreated 12-mo-old HSA^LR mice relative to WT mice, but the bundling was near normal in the matched HSA^LR mice treated for 6 wk with BIO at a young age (Fig. 2D).

Figure 2.

Inhibition of GSK3 in young HSA^LR mice prevents muscle atrophy and myofiber size variability. A–D) Quantitative analysis of the total fiber number (A), the average cross-sectional fiber area (B), total central nuclei per ×20 view (C), and the number of bundles per ×20 view (D) in the gastroc of 12-mo-old WT, untreated HSA^LR, and BIO-treated (for 6 wk at 1.5 mo old) HSA^LR mice. See Materials and Methods for details. A) *P < 0.05 for untreated HSA^LR mice vs. WT mice; **P < 0.01 for treated vs. untreated HSA^LR mice. B) ***P < 0.001 for untreated HSA^LR mice vs. WT mice and for treated vs. untreated HSA^LR mice. C, D) ***P < 0.001 for untreated HSA^LR mice vs. WT mice (C, D) and treated vs. untreated HSA^LR mice (C); *P < 0.05 for untreated HSA^LR mice vs. treated HSA^LR mice (D). E) Western blot analysis of GSK3β, cyclin D3, and total CUGBP1 in gastroc from 12-mo-old WT and HSA^LR mice, untreated and treated (at 1 mo with BIO for 6 wk). β-actin was used as a control. F) Analysis of interaction between CUGBP1 and p-S51–eIF2α in the gastroc of 12-mo-old WT and untreated and treated HSA^LR mice. CUGBP1 was immunoprecipitated, and the levels of p-S51–eIF2α were determined by Western blot analysis. The total levels of p-S51–eIF2α (input) are shown on the bottom. The p-S51–eIF2α in the CUGBP1-IP is shown with an arrow. G) Western blot analysis of Pax7 in gastroc from 12-mo-old WT and HSA^LR mice, untreated and treated at 1.5 mo with BIO for 6 wk. Calreticulin is shown as a loading control.

Because the inhibitors of GSK3, lithium, and TDZD-8 reduce the activity of GSK3 and its downstream substrate, cyclin D3, in HSA^LR muscle (17), we examined the levels of GSK3β and cyclin D3 in 12-mo-old HSA^LR mice that had been briefly treated at 1.5 mo with BIO. Levels of GSK3β and cyclin D3 were near normal in 12-mo-old HSA^LR mice, which had been briefly treated at a young age with BIO (Fig. 2E). Moreover, increase of the total levels of CUGBP1 was also corrected in the 12-mo-old, treated HSA^LR mice.

To examine whether the translational activity of CUGBP1 was improved in old HSA^LR mice treated at a young age with BIO, the interaction of CUGBP1 with inactive, p-S51-eIF2α in skeletal muscle (gastroc) of 12-mo-old WT, untreated HSA^LR, and HSA^LR mice treated at a young age with BIO was measured by the IP-Western blot analysis. CUGBP1 was immunoprecipitated, and the amounts of p-S51-eIF2α in the CUGBP1-IPs were determined by Western blot analysis. Figure 2F shows that p-S51–eIF2α was undetectable in the CUGBP1-IP in skeletal muscle of 12-mo-old WT mice, whereas inactive p-S51–eIF2α interacts with CUGBP1 in muscle of old HSA^LR mice. In contrast to the untreated mice, p-S51–eIF2α was undetectable in the CUGBP1-IPs from skeletal muscle of 12-mo-old HSA^LR mice, treated at a young age with BIO (Fig. 2F). This result shows that the translational activity of CUGBP1 was corrected in HSA^LR mice briefly treated at a young age with BIO.

We found that the number of active myogenic satellite cells (MSCs) that were positive for the transcription factor Pax7 was reduced in 7-mo-old HSA^LR mice (17). We compared the levels of Pax7 in skeletal muscle of 12-mo-old, untreated HSA^LR mice and HSA^LR mice treated at 1.5 mo for 6 wk with BIO. As shown in Fig. 2G and Supplemental Fig. 2, levels of Pax7 were reduced in old, untreated HSA^LR mice relative to WT mice. However, Pax7 levels were normalized in HSA^LR mice that had been treated at a young age for 6 wk with BIO. Taken together, these findings indicate that correction of GSK3β from 1.5 to 3 mo of age is sufficient to maintain a corrected GSK3β–cyclin D3–CUGBP1 pathway and almost-normal muscle health with normal levels of Pax7 in HSA^LR mice over a long period.

Loss of CUGBP1 disrupts neonatal myogenesis

Despite the increase of the total levels of CUGBP1 in DM1, CUGBP1 activity is partially reduced in DM1 because of conversion of active, p-S302–CUGBP1 into inactive unp-S302–CUGBP1 caused by misregulation of the GSK3β–cyclin D3/CDK4 pathway (12, 15–17). To determine the contribution of the reduced CUGBP1 in the maintenance of myogenesis in vivo, we examined the effects of loss of CUGBP1 on skeletal muscle in Celf1 KO mice.

Previous work from Dr. L. Paillard’s laboratory (18) with the Celf1 KO mouse model showed that CUGBP1 was necessary for normal development, growth, and spermatogenesis. Recent work with the same model showed a reduction of the cardiac function in neonatal hom Celf1 KO mice (19). Although a significant portion of hom Celf1 KO mice, described in the previous studies died, skeletal muscle in those mice was not examined. Surviving hom Celf1 KO mice were characterized by impaired fertility and cataracts but did not show defects in skeletal muscle (20).

Recently, we generated Celf1 KO mice in which the Celf1 gene was disrupted by an insertion of a trap vector into the intronic sequence upstream of exon 3, containing ATG codon. In agreement with the previous reports (18–20), we found that CUGBP1 has a crucial role in development. Hom Celf1 KO mice are smaller than WT and het littermates (Fig. 3A). The number of hom mice was reduced probably because of embryonic underdevelopment (Supplemental Fig. 3A). Most hom mice die shortly after birth. The results of the quantitative RT-PCR, Western blot, and IF analyses confirmed a complete loss of Celf1 RNA and protein in the skeletal muscle of hom Celf1 KO mice (Fig. 3B, C). Het Celf1 KO mice survive and age normally; however, they have a slightly reduced body weight (Supplemental Fig. 3B) and develop grip weakness (Fig. 3D, E).

Figure 3.

Critical role of CUGBP1 in myofibrillar organization. A) Growth retardation of hom Celf1 KO mice. Representative pictures of 6-d-old WT and hom Celf1 KO littermates are shown. B, C) Confirmation of deletion of CUGBP1 in skeletal muscle of hom Celf1 KO mice by immunostaining, Western blot, and quantitative RT-PCR analyses. Longitudinal and cross-sections of hind-limb muscle of 1-d old WT and hom Celf1 KO littermates were analyzed with antibodies to CUGBP1 under the same exposure time and brightness (B). Original magnification, ×60. Scale bar, 5 μm. Western blot (left) and quantitative RT-PCR (right) of CUGBP1 in hind-limb muscle of 1-d-old WT and Celf1 KO littermates (C). Actin was used as a control for the protein and RNA analyses. D, E) Grip strength analysis of WT and het Celf1 KO mice (females and males) at different ages. The number of analyzed mice in each mouse group is shown. *P < 0.05, **P < 0.01, ***P < 0.001 for WT vs. het Celf1 mice. F) Hematoxylin and eosin staining of cross-sections of the hind-limb muscle from 1-d-old WT and hom Celf1 KO littermates. Original magnification, ×40. Scale bars, 10 μm. G) Electron microscopy images of the hind-limb skeletal muscle from 1 d-old WT and hom Celf1 mice. Original magnification, ×10,000. Arrows point to the sarcomeres in WT and KO myofibrils. Scale bar, 2 μm. H–J) Comparison is shown of sarcomere length, sarcomere number, and myofibril number in 1-d-old WT and hom Celf1 KO muscle. For experimental details, see Materials and Methods. ***P < 0.001 for WT vs. hom Celf1 KO muscle.

Analysis of skeletal muscle histology in 1-d-old hom Celf1 KO mice revealed severe disruption of the muscle structure (Fig. 3F). Myofibers in hom Celf1 KO mice are thinner than those in WT mice, with reduced density, and they contain central nuclei. The most striking feature of hematoxylin and eosin staining of skeletal muscle in neonatal hom Celf1 KO mice is the almost complete lack of muscle striation. Fine analysis of skeletal muscle by electron microscopy confirmed disruption of myofibrillar structures in 1-d-old hom Celf1 KO mice (Fig. 3G) and showed a reduced number of myofibrils, reduced myofibril density, disorganization of myofibrils, and reduction of sarcomeres, which had reduced lengths relative to WT muscle (Fig. 3H–J). IF analysis with antibodies to the major contractile proteins, actin and myosin, showed that those sarcomeric proteins were not organized in sarcomeres and had a diffuse appearance in 1-d-old hom Celf1 KO fibers (Fig. 4A). Disorganization of myofibrils in hom Celf1 KO muscle is also supported by the diffuse appearance of the contractile muscle protein troponin T (Fig. 4B). A major muscle filament protein, desmin, and a structural protein of the Z-line, α-actinin, were also disorganized in hom Celf1 KO fibers. Those findings indicate that CUGBP1 has critical role in the formation of functional myofibers.

Figure 4.

IF analysis of actin and myosin (A) and troponin T (TnT), desmin, and α-actinin (B) in the hind-limb muscle of 1 d-old WT and hom Celf1 KO mice. The merge of actin and DAPI staining is also shown. Original magnification, ×60. Scale bars, 2 μm.

Misregulation of pathways controlling differentiation and extracellular matrix in the under-developed muscle of hom Celf1 KO mice

To determine the molecular pathways disrupted in the skeletal muscles of underdeveloped hom Celf1 KO mice, we performed microarray analysis of the global gene expression in the skeletal muscle of 1-d-old WT and hom Celf1 KO mice. In addition to skeletal muscle, gene expression in brains of 1-d-old WT and Celf1 KO mice was examined. That analysis showed that the loss of CUGBP1 in vivo caused significant changes in expression of genes associated with the control of development, cell differentiation, nucleotide metabolism, receptor signaling, cellular import and export, and protein degradation—in both skeletal muscle and in brain (Fig. 5A). Several enriched gene categories related to differentiation/transdifferentiation, extracellular matrix (ECM), and other functions of muscle cells were identified among the most highly up-regulated genes in hom Celf1 KO muscle. Those genes included matrix metallopeptidase 13; an inhibitor of protein phosphatase 1 associated with contraction of smooth muscle Ppp1r14d; the vitamin D receptor, which regulates muscle homeostasis and muscle strength (21); and midline 1 protein, which controls neuronal axon extension (22).

Figure 5.

A) The percentage of genes for the functional pathways altered in skeletal muscle and in the brain of hom Celf1 KO mice. B) Confirmation of alterations of selected mRNAs identified by microarray analysis in skeletal muscle of Celf1 KO mice. RNA expression analysis of LEF1, DCX, RBM45, and Col4A in the hind-limb muscle of 1–4-d-old WT and Celf1 KO mice. LEF1, DCX, and RBM45 were analyzed in hom Celf1 KO muscle. Col4A mRNA was examined in het Celf1 KO muscle. C) Western blot analysis of LEF1, RBM45, Col 4A, and DCX in the gastroc from WT and Celf1^+/− littermates at 4 d old. β-actin was used as the control for protein loading. D) Western blot analysis of skeletal muscle from WT and Celf1^+/− littermates at 5 d and 12 mo old with antibodies to Pax7 and β-actin as control.

Ion signaling genes, particularly calcium signaling, such as parvalbumin (23), were among the most down-regulated genes in skeletal muscle of hom Celf1 KO mice. Global microarray analysis also showed many misregulated long intergenic noncoding RNAs in hom Celf1 KO muscle and brain.

A list of mRNAs regulating cell differentiation includes mRNAs encoding the transcription regulators LEF1, TCF7, RUNX2, and forkhead box N4 factor, which were increased in skeletal muscle of hom Celf1 KO mice (Table 1). Loss of CUGBP1 led to elevation of mRNAs encoding doublecortin (DCX), type 13 collagen (Col13), and RNA-binding protein, RBM45. DCX regulates myoblast migration and synaptogenesis (24, 25). Col13 is a protein of ECM associated with the maturation of neuromuscular junction (26). RBM45 is a developmentally regulated protein controlling splicing, RNA transport, translation, and RNA stability (27).

TABLE 1.

Selected genes linked to cell differentiation altered in skeletal muscle of hom Celf1 KO mice

Name	Fold change	Probability
Lymphoid enhancer binding factor 1 (LEF1)a	1.720	1.84 × 10−2
Collagen, type XIII, α1 (Col13A1)	3.025	4.19 × 10−2
Vitamin D receptor (VDR)	9.283	1.80 × 10−2
Doublecortin, variant 4 (DCX)a	2.526	3.00 × 10−2
Hedgehog-interacting protein (HHIP)	1.515	1.75 × 10−2
RNA binding motif protein 45 (RBM45)a (BC3H1 cDNA encoding RNA-binding region RNP-1)	4.621	1.53 × 10−3
Transcription factor 7, T cell specific (TCF7)	2.087	9.60 × 10−3
CD44 antigen (CD44)	2.251	6.19 × 10−3
Ubiquitination factor E4B (UBE4A)	2.375	1.20 × 10−3
Runt related transcription factor 2 (RUNX2)	4.022	8.77 × 10−3
Forkhead box N4 (FOXN4)	2.468	7.25 × 10−3
Desmoplakin	−1.688	2.64 × 10−2
Gasdermin A3	0.900	4.49 × 10−2
IL-1 receptor accessory protein-like 1	−2.336	1.65 × 10−2

^aGenes examined in this study.

The microarray analysis also revealed that ECM proteins required for myogenesis and muscle function were altered in hom Celf1 KO muscle (Table 2). In addition to the above-mentioned Col13, loss of CUGBP1 was associated with up-regulation of type 4A (Col4A), a major component of the basement membrane, which protects myofibers and transmits forces produced within the muscle to the tendon and to other muscles (28). Type 9 collagen was also increased. Loss of CUGBP1 likely affects ECM turnover because matrix metalloproteinase 13 expression was significantly increased in skeletal muscle of hom Celf1 KO mice.

TABLE 2.

Selected genes linked to ECM altered in skeletal muscle of hom Celf1 KO mice

Name	Fold change	Probability
Matrix metallopeptidase 13 (MMP13)	47.284	1.72 × 10−2
Collagen, type IX, α2 (Col9A1)	1.649	3.93 × 10−2
Alkaline phosphatase, liver/bone/kidney (ALPL)	4.479	3.89 × 10−2
Collagen, type 13, α1 (Col13A1)	3.025	4.19 × 10−2
Brevican (BCAN)	4.991	1.72 × 10−2
Collagen, type 4, α4 (Col4A1)a	2.091	2.92 × 10−2
Sialophorin (SPN)	−1.782	2.33 × 10−3
ADAM metallopeptidase thrombospondin type 1 motif, 8 (ADAMTS8)	−1.677	4.13 × 10−2

^aGene examined in this study.

We confirmed the alterations of DCX, RBM45, LEF1, and Col4A in Celf1 KO muscle using quantitative RT-PCR and Western blot assays (Fig. 5B, C). Taken together, RNA expression analysis showed that the loss of CUGBP1 disrupted myogenesis and changed pathways, including ECM, ECM remodeling, transcriptional regulation, myoblast migration, and formation of neuromuscular junctions.

We found that Pax7 levels were increased in the muscles of young HSA^LR mice before development of muscle pathology; however, the levels of Pax7 were reduced in adult and old HSA^LR mice [(17) and Fig. 2G]. Because Pax7 is a marker of MSCs, these findings suggest that the elevation of Pax7 in young muscle of HSA^LR mice might occur in response to the expression of toxic CUG repeats. However, in adult HSA^LR mice, the activation of MSCs might be exhausted, and as the result, the levels of Pax7 are reduced [(17) and Fig. 2G]. To determine whether Pax7 expression was altered in Celf1 KO muscle, we examined the levels of Pax7 in muscles of 5-d-old and 12-mo-old het Celf1 KO mice. The immunoanalysis showed that Pax7 was elevated in the skeletal muscle of 5-d-old Celf1 KO mice (Fig. 5D). However, in the muscle of 12-mo-old Celf1 KO, the levels of Pax7 were reduced. Those findings suggest that the reduction of CUGBP1 function affects muscle, causing alterations of Pax7 levels. In 5-d-old muscle, the levels of Pax7 were increased, likely in response to the reduction of CUGBP1, which disrupts myogenesis. The reduction of Pax7 in old muscle of het Celf1 KO mice might occur because of exhaustion of MSCs.

Misregulation of proteins—altered in undeveloped muscle of hom Celf1 KO mice and in skeletal muscle of pediatric patients with CDM1

Because skeletal muscle in severe CDM1 is underdeveloped, we analyzed several proteins important for myogenesis and healthy muscle function that were identified via assessment of hom Celf1 KO muscle in patients with CDM1 using quantitative IF analysis. The underdevelopment of the CDM1 muscle was confirmed by hematoxylin and eosin staining, which showed reduction of fiber size in CDM1 muscle, with an increase of central nuclei and reduced fiber density (Supplemental Fig. 4). Quantitative IF analysis (with images at the same exposure time and brightness) showed that DCX, LEF1, and Col4A were increased in CDM1 fibers relative to healthy fibers (Fig. 6A, B). In addition to elevation of expression, intracellular distribution of DCX and Col4A was altered in CDM1 myofibers. In healthy myofibers DCX and Col4A were predominately membrane proteins. However, in CDM1 myofibers, DCX and Col4A were mainly cytoplasmic.

Figure 6.

Proteins linked to myogenesis are disrupted in the underdeveloped muscle of patients with CDM1. A) Quantitative IF analysis of DCX and Col4A and LEF1 (B) in skeletal muscle sections from the matched pediatric healthy (Normal) control patients and patients with CDM1. DAPI staining is shown under each IF image. Note that the size of the myofibers is larger in healthy muscle than in CDM1 muscle. Original magnification, ×40. BG, background staining without first antibodies. Scale bars, 2 μm.

We suggest that alteration and mislocalization of proteins important for myogenesis (such as LEF1, DCX, and Col4A and possibly other genes misregulated by CUGBP1) might disturb the balance among proliferation, differentiation, myoblast migration, synaptogenesis, and ECM remodeling and, hence, causing delays in CDM1 myogenesis.

One of the important proteins identified by global gene analysis in hom Celf1 KO muscle is RBM45 (27). Similar to CUGBP1, RBM45 belongs to a superfamily of elav proteins, which normally regulate cell development, including myogenesis through control of RNA processing. RBM45 contains 4 RNA recognition motifs (RRMs) with 1 pseudo-RRM (Fig. 7A). RRMs of RBM45 are ∼30% homologous to the RRMs of human CUGBP1. Analysis of RBM45 expression in normal and CDM1 myogenesis revealed that RBM45 was increased during myoblasts differentiation in both normal and CDM1 myotubes; however, levels of RBM45 were greater in CDM1 myotubes relative to healthy myotubes (Fig. 7B). This result shows that RBM45 is a new regulator of muscle differentiation and that it is linked to abnormal myogenesis in CDM1. IF analysis identified large RBM45⁺ inclusions in sarcolemma and in the nuclei in human fibers from healthy control patients (Fig. 7C). Some RBM45 inclusions were located outside the nuclei. However, in CDM1 myofibers, RBM45 signal was mainly diffuse and was located in the cytoplasm with little or no RBM45 inclusions. We suggest that the altered function of RBM45 in CDM1 contributes to a delay in CDM1 myogenesis.

Figure 7.

The inhibitor of GSK3 BIO corrects proteins linked to cell differentiation and ECM, assessed through analysis of the underdeveloped muscle of hom Celf1 KO mice, in human CDM1 myotubes, and in muscle of young HSA^LR mice. A, top) Structures and locations of RRMs of CUGBP1 and RBM45. Pseudodomain in RBM45 is shown in blue. A, bottom) Partial homology of RBM45 and RRM3 of CUGBP1. B) Western blot analysis of RBM45 in healthy and CDM1 myoblasts and myotubes, differentiated for 5 d. Protein levels of RBM45, adjusted to β-actin are shown. C) Quantitative IF analysis of RBM45 in transverse muscle sections from the matched healthy (N) control patient and from a patient with severe CDM1. Merged signals of nuclei, stained by DAPI and RBM45, are shown. Arrows point to nuclear RBM45 aggregates in healthy fibers. Original magnification, ×40. Scale bar, 5 μm. D) Western blot analysis of RBM45, LEF1, DCX, and Col4A in healthy (N) and CDM1 myotubes, untreated or treated with BIO. E) Levels of RBM45, LEF1, Col4A, and DCX shown in D were calculated as ratios to β-actin. F) Western blot analysis of RBM45, LEF1, DCX, and Col4A in the gastroc from the 2.5-mo-old WT, untreated, and BIO-treated at 1 mo for 6 wk HSA^LR mice. G) The protein levels, shown in F were calculated as ratios to β-actin.

We next examined whether inhibition of GSK3 with BIO corrects RBM45 levels in CDM1 myotubes. In addition to RBM45, the levels of LEF1, DCX, and Col4A were also measured in CDM1 myotubes, with and without treatments with BIO. Immunoanalysis showed that levels of RBM45, LEF1, DCX, and Col4A were increased in CDM1 myotubes relative to control healthy myotubes (Fig. 7D, E). However, treatment of CDM1 myotubes with BIO reduced levels of those proteins.

We found that RBM45, LEF1, DCX, and Col4A are misregulated in skeletal muscle of young HSA^LR mice. Western blot analysis showed that levels of RBM45, LEF1, DCX, and Col 4A are increased in 1-mo-old, untreated HSA^LR mice (Fig. 7F, G). However, the treatment of 1-mo-old HSA^LR mice with BIO for 6 wk reduced LEF1, DCX, RBM45, and Col 4A. These findings show that treatment of HSA^LR mice with the inhibitors of GSK3 corrects expression of proteins associated with myogenesis in young, actively growing HSA^LR muscle.

DISCUSSION

The most important result of this study is that temporal correction of GSK3β in muscle of young HSA^LR mice is crucial for the maintenance of skeletal muscle health over time. Remarkable reduction in muscle histopathology and almost-normal grip strength in 12-mo-old HSA^LR mice treated at a young age with BIO showed that the correction of the GSK3β levels in HSA^LR mice at 1.5–3 mo old was critical for the long-term reduction of DM1 muscle pathology. We applied 5 tests that included the analysis of the total fiber number, the average fiber area, the number of the fiber bundles, the number of the centralized nuclei, and the grip strength. We found that 4 of 5 tests were normal or near normal in the 12-mo-old HSA^LR mice briefly treated at a young age with BIO (Figs. 1 and 2A, B, D). Although central nuclei were detected in skeletal muscle of the treated mice, their numbers were significantly reduced (Figs. 1C and 2C). The molecular analysis of the GSK3β–cyclin D3–CUGBP1 pathway showed that the levels of GSK3β and cyclin D3 were normal in HSA^LR mice treated at a young age with BIO (Fig. 2E). CUGBP1 activity was also normalized (Fig. 2F).

Previous findings from our laboratory and other laboratories, together with findings in this study, suggest that the improvement of muscle maintenance in HSA^LR mice after treatment with the inhibitors of GSK3 may be mediated by the correction of CUGBP1 activity. This suggestion is based on studies that showed that the inhibition of the GSK3β–cyclin D3 pathway is needed for normal CUGBP1 activity, and on studies that provided an understanding of the critical role of CUGBP1 in the regulation of muscle structure, myogenesis, and muscle function. The role of CUGBP1 in the regulation of myogenesis at the levels of proliferation and differentiation has been demonstrated in human muscle biopsies and in primary myoblasts from healthy control patients and patients with DM1, as well as in multiple cell culture models (12, 13, 15–17). Analysis of two mouse models in which CUGBP1 was overexpressed in the whole body demonstrated that the abnormal increase of CUGBP1 is toxic for muscle differentiation and causes variability of myofiber size with central nuclei and muscle degeneration, consistent with DM1 phenotype (29, 30). Conditional mice overexpressing CUGBP1 in skeletal muscle or in the heart developed muscle wasting and dilated cardiomyopathy (31, 32). Increased expression of CUGBP1 lacking the nuclear localization signal resulted in the alteration of skeletal muscle structure with reduced endomysial and perimysial spaces and with fiber-size variability (33). The findings in the current study show that the lack of CUGBP1 also severely affects myogenesis, disturbing sarcomeric structure (Fig. 3). Studies in mouse models, including the current study, show that different levels of CUGBP1 are likely required at different stages of muscle development, suggesting that either too much or too little CUGBP1 is equally deleterious for myogenesis. Similar findings were described for critical myogenic factors, such as myocyte enhancer factor type 2 (MEF2) and myogenin (34, 35).

Analysis of muscle in hom Celf1 KO mice showed that CUGBP1 function is critical for the formation of the functional myofibers. Despite the presence of other homologous members of the CUGBP1/CELF family of proteins, the lack of CUGBP1 disrupts myogenesis (Figs. 3 and 4). Multiple CUGBP1 targets associated with myogenesis have been identified previously with various models (36, 37). RNA array analysis in this study identified several additional pathways, critical for myogenesis, disrupted by the loss of CUGBP1. We found that the pathways regulating cell differentiation (transcription factors, LEF1 and Tcf7, which bind to β-catenin), DCX, and RBM45, and pathways regulating ECM (collagens) are misregulated by the loss of CUGBP1.

We are planning to determine which of those mRNAs are directly regulated by CUGBP1. However, alterations of LEF1, DCX, RBM45, and Col4A in the underdeveloped muscle of patients with CDM1 suggest that the reduced activity of CUGBP1 might be responsible for the misregulation of those proteins in CDM1 muscle (Figs. 6 and 7). Our data suggest that those proteins might be altered in underdeveloped muscle of CDM1 and in young HSA^LR mice via CUG–GSK3β–CUGBP1 pathway, which reduces the activity of CUGBP1 via a reduction of phosphorylation of CUGBP1 at S302 (15, 16). Correction of RBM45, LEF1, DCX, and Col4A in CDM1 myotubes and in young HSA^LR mice with BIO suggests that the inhibitors of GSK3 might improve ECM remodeling and cell adhesion, which will be beneficial for the migration of myoblasts and neuromuscular junctions increasing myogenesis over time. The analysis of satellite cell niches and the efficiency of MSC renewal, expansion, and differentiation in treated HSA muscle at different time points after the treatment should reveal the stage of myogenesis corrected by the inhibitors of GSK3 in young HSA mice, preventing DM1 muscle pathology.

Supplementary Material

This article includes supplemental data. Please visit http://www.fasebj.org to obtain this information.

Glossary

BIO

6-bromoindirubin-3′-oxime

CDK4

cyclin D–dependent kinase 4

CDM1

congenital myotonic dystrophy type 1

Col

collagen

CUGBP1/CELF1

CUG-binding protein/CUGBP1 elav-like factor 1

DCX

doublecortin

DDX

DEAD-box

DM1

myotonic dystrophy type 1

DMPK

dystrophia myotonica protein kinase

ECM

extracellular matrix

eIF2

eukaryotic initiation factor 2

ES

embryonic stem cell

GSK3β

glycogen synthase kinase 3β

het

heterozygous

hom

homozygous

HSA^LR

human skeletal actin mRNA, containing long repeats

IF

immunofluorescence

IP

immunoprecipitation

KO

knockout

LEF1

lymphoid enhancer binding factor 1

MBNL1

muscleblind 1 protein

MSC

myogenic satellite cell

p-

phosphorylated

Pax7

paired box 7 factor

Ppp1r14d

protein phosphatase 1 regulatory inhibitor subunit 14D

RBM45

RNA-binding motif 45 protein

RRM

RNA recognition motif

RUNX2

runt-related transcription factor 2

TCF7

T cell specific transcription factor 7

TDZD-8

4-benzyl-2-methyl-1,2,4-thiadiazolidine-3,5-dione

unp

unphosphorylated

WT

wild type

AUTHOR CONTRIBUTIONS

L. Timchenko generated hypotheses, designed the studies, and provided funds for the studies; C. Wei, L. Stock, L. Valanejad, and Z. A. Zalewski performed the experiments; R. Karns performed statistical analyses; J. Puymirat, D. Nelson, and D. Witte provided reagents; J. Woodgett provided conceptual advice; N. A. Timchenko and L. Timchenko provided interpretation of results; and C. Wei, J. Puymirat, D. Nelson, J. Woodgett, N. A. Timchenko, and L. Timchenko wrote and commented on the manuscript.