Pluripotent Stem Cell-Based Drug Screening Reveals Cardiac Glycosides as Modulators of Myotonic Dystrophy Type 1.

There is currently no treatment for myotonic dystrophy type 1 (DM1), the most frequent myopathy of genetic origin. This progressive neuromuscular disease is caused by nuclear-retained RNAs containing expanded CUG repeats. These toxic RNAs alter the activities of RNA splicing factors, resulting in alternative splicing misregulation. By combining human mutated pluripotent stem cells and phenotypic drug screening, we revealed that cardiac glycosides act as modulators for both upstream nuclear aggregations of DMPK mRNAs and several downstream alternative mRNA splicing defects. However, these occurred at different drug concentration ranges. Similar biological effects were recorded in a DM1 mouse model. At the mechanistic level, we demonstrated that this effect was calcium dependent and was synergic with inhibition of the ERK pathway. These results further underscore the value of stem-cell-based assays for drug discovery in monogenic diseases.

Introduction

Myotonic dystrophy type 1 (DM1) is the most common form of adult muscular dystrophy of genetic origin, with a prevalence of 1/8,000 worldwide, and remains a disease for which there is no treatment. Its multisystemic symptoms, which include myotonia, muscle wasting, cardiac conduction defects, insulin resistance, cataracts, and cognitive dysfunction, are linked to dysregulation and alternative splicing of RNA that affects hundreds of genes (Du et al., 2010, Nakamori et al., 2013, Turner and Hilton-Jones, 2014). The origin of this dysregulation lies in changes to the bioavailability of RNA-binding proteins, in particular MBNL1, which is sequestered in intranuclear ribonucleoprotein aggregates (called foci) and triggered by a CTG repeat expansion in the 3′ UTR of the DMPK (dystrophic myotonia protein kinase) gene (Kanadia et al., 2003, Mahadevan et al., 1992). Accordingly, a major ongoing effort aims to generate therapeutic tools that are specifically engineered to target the upstream molecular substrate of the disease by modifying the pathological binding between the mutant mRNA and proteins (Thornton et al., 2017).

A complementary approach to drug discovery has recently been made possible by demonstrating that cells differentiated from pluripotent stem cell lines carrying the mutant DMPK gene—obtained as embryonic stem (ES) cell lines from preimplantation genetically diagnosed (PGD) embryos or via reprogramming induced pluripotent stem cell lines from patients—recapitulate major cellular and molecular hallmarks of the disease (Denis et al., 2013, Du et al., 2013, Gauthier et al., 2013, Marteyn et al., 2011, Ueki et al., 2017). These cell lines provide an endless supply of well-characterized biological resources that are amenable to unbiased phenotypic high-throughput drug screening. Over recent years, such an approach has been revealed to be instrumental in identifying candidate therapeutic compounds for other monogenic diseases, such as familial dysautonomia, Huntington disease, Phelan-McDermid syndrome, and progeria (Charbord et al., 2013, Darville et al., 2016, Lee et al., 2012a).

In the current study, we used cells differentiated from an ES cell line, derived from a human embryo carrying a DM1 mutation, to seek compounds that would affect foci and alternative splicing changes. This approach revealed the beneficial effects of US Food and Drug Administration (FDA)-approved cardiac glycosides both on DM1 biological markers and in an in vitro myogenic defect as a functional validation of our findings.

Results

High-Throughput Drug Screening on DM1 Cells

The drug screening workflow involved two sequential steps: (1) a primary assay in which foci were quantified by automated cell imaging, (2) a dose-response exploration of hit compounds for their effects on both foci and selected DM1-specific alternative splicing defects using qRT-PCR. The entire screening process was conducted on mesodermal stem cells (DM1_MSCs), differentiated from an ES cell line derived from a PGD embryo carrying over 1,000 CUG repeats. Screening values were compared with non-treated DM1_MSCs and wild-type (WT)_MSCs that were differentiated from an unaffected ES cell line (Marteyn et al., 2011). Results were secondarily validated in other cell and animal models for the disease.

Nuclear foci were identified in DM1_MSCs using an automated RNA fluorescence in situ hybridization assay that was developed to detect expanded CUG repeats (Figure S1A). DM1_MSCs contained an average of 2.5 foci, whereas none were observed in control cells (Figure 1A). Results were confirmed as robust based on a high Z-factor, of 0.71, calculated for the assay (Figure 1B).

Figure 1

Development of a High-Content Assay for Mutant DMPK Foci

(A) Mean number of detected foci per cell in untreated DM1-MSCs (red dots) and WT_MSCs (blue dots). Each dot represents one well in 90 screening plates.

(B) Determination of the Z′ factor by using the number of detected foci per cell between DM1 and WT_MSCs for each plate, with a mean value of around 0.7 for the whole screening campaign.

(C) Scatterplot representation of the primary screening analysis based on the number of detected foci per cell and average area of foci after normalization in non-treated DM1_MSCs. Validated compounds are shown with colored dots.

(D) Dose-response analysis for number of foci, area of foci, and cell density in DM1_MSCs after 48 hr treatment with seven cardiac glycosides. Mean values ± SD are given (10 points in dose-response curve, each in triplicate).

(E) Representative images of mutant DMPK mRNA foci detected by RNA-fluorescence in situ hybridization in DM1_MSCs after treatment with three different cardiac glycosides for 48 hr. Scale bar indicates 10 μm.

The primary drug screening was carried out by considering the number of intranuclear foci per cell using 12,089 compounds assayed at a concentration of 10 μM, including a set of 1,120 FDA-approved drugs. Compounds were selected as candidates when the number of foci reached 1.5 standard deviations below the mean value for untreated DM1_MSCs, in the absence of unacceptable cytotoxicity (cutoff for cell survival set at 20% of control). This target was then revealed as non-informative, because none of the compounds that were tested demonstrated a capacity for decreasing the number of foci in the absence of unacceptable cell toxicity. Four hit compounds only emerged once these results were secondarily filtered by taking into account the average area of individual intranuclear foci. Their effects were characterized by an increased number of foci per nuclei in conjunction with their size reduction, which suggested a partial disaggregation of ribonucleoprotein inclusions. These compounds were cycloheximide (CHX) and three members of the cardiac glycoside family, namely, strophantidin, ouabain, and digoxigenin (Figures 1C and S1B). The first compound, CHX, was excluded from further analysis owing to its well-known activity as a protein synthesis inhibitor. In support of the hypothesis that the apparent reduction in size of the foci was due to a general decrease in cell protein content, we observed a decreased expression of MBNL1 and CUGBP1, as well as an exacerbation of defective alternative splicing associated with DM1 after treatment with CHX (Figure S2). The identification of three cardiac glycosides in the screening campaign suggested a common mechanism of action, and four additional members of this chemical family were therefore included in dose-response experiments, revealing similar dose-dependent effects: an increased number of foci per nucleus associated with a size reduction of each focus (Figures 1D and 1E).

In parallel, a concomitant rescue of dysregulated alternative splicing was sought using the defective inclusion of exon 11 in the insulin receptor (IR) gene that was described as a representative in DM1 (Savkur et al., 2001) (Figure 2A). The ratio of IRB (containing exon 11) to IRA (excluding exon 11) was used as a readout. Although the action of cardiac glycosides on the foci parameters was efficient at the micromolar range, it was associated with decreased cell viability, as well as an exacerbation of the DM1-related IR splicing defect, decreasing the ratio of IR transcripts. Conversely, all cardiac glycosides that were tested normalized the DM1-altered ratio of IR transcripts at concentrations ranging from 10 nM to 500 nM, depending on the chemical structure, in the absence of overt cytotoxicity and of any effect on foci (Figures 2B and 2C).

Figure 2

Digoxin Treatment in the Nanomolar Range Rescues Alternative IR and SERCA1 Splicing in DM1 Cells

(A) Dose-response analysis for alternative IR splicing in DM1_MSCs treated with digoxin for 48 hr. Relative expression level of the two isoforms of IR (IRA [-exon 11] and IRB [+exon 11]) were measured by quantitative real-time RT-PCR. Data represent the ratio of IRB/IRA and are indicated as mean ± SD (n = 3). Data were analyzed with one way ANOVA, followed by a Dunnett post-hoc test. ∗∗∗: p-value < 0.001.

(B) Superposition of the concentration-dependent effect of digoxin on the number of foci per cell (solid line, left axis) and alternative IR splicing (black bars, right axis) on DM1_MSCs treated for 48 hr.

(C) Dose-response analysis of six different cardiac glycosides on mutant DMPK foci (solid lines; n = 2; mean ± SEM) and alternative IR splicing (bars; n = 2, mean ± SD) in DM1_MSCs treated for 48 hr.

Digoxin Effect on Alternative mRNA Splicing

Digoxin, the best documented and most commonly prescribed medication in the cardiac glycoside family, was used as a representative to further characterize the therapeutic potential in DM1.

The long-term effect of this drug was demonstrated by chronic treatment of DM1_MSCs with 50 nM digoxin for 4 weeks in standard tissue culture T75 flasks, with analysis of one sample at each time point. This treatment induced a stable splicing switch of the IR exon 11 and SERCA1 exon 22 from a DM1 to a WT pattern (Figure 3A). This effect was observed in the absence of a concomitant effect of the treatment on the number and size of the foci. Other splice defects associated with DM1 were also shown to be normalized, including cTNT exon 5, CLCN-1 exon 7a in DM1_hES-derived MSCs, and NMDAR1 exon 5 in DM1_hES-derived neurons after a 48-hr treatment period (Figure S3).

Figure 3

Digoxin Treatment Rescues Alternative IR and SERCA1 Splicing in DM1 Cells

(A) Effect of chronic digoxin treatment on mutant DMPK foci and alternative splicing of IR and SERCA1. DM1_MSCs were continuously treated with 50 nM of digoxin over 4 weeks. Mutant DMPK foci and alternative splicing of IR and SERCA1 were analyzed every 7 days. Untreated WT_MSCs were used as the control. For alternative IR and SERCA1 splicing, one sample was analyzed at each time point. Relative expression levels of the two isoforms of IR (IRA [-exon 11] and IRB [+exon 11]) were measured by real-time quantitative RT-PCR. Data represent the ratio of IRB/IRA (n = 1 at each time point). Alternative IR splicing was analyzed using real-time quantitative RT-PCR. The percentage for inclusion of SERCA1 exon 22 was determined by RT-PCR, followed by analysis with a 2100 Bioanalyzer (Agilent). Quantification of each band was performed using 2100 Expert software. For mutant DMPK foci, data are presented as mean ± SD (n = 4).

(B) Concentration-dependent effect of digoxin on alternative IR splicing in WT and DM1_MSCs. Relative expression levels of the two isoforms of IRA [-exon 11] and IRB [+exon 11]) were measured by real-time quantitative RT-PCR. Data represent the ratio of IRB/IRA.

(C) Effect of digoxin treatment at 50 nM for 48 hr on the inclusion of exon 22 of SERCA1 in WT and DM1_MSCs, determined by RT-PCR, followed by analysis with a 2100 Bioanalyzer (Agilent).

(D) Effect of digoxin treatment on alternative IR and SERCA1 splicing in primary myoblasts isolated from healthy individuals and patients with DM1. For all graphs, data are presented as mean ± SD (n = 3) and analyzed with one-way ANOVA, followed by a Bonferroni post-hoc test. *p < 0.05, **p < 0.01, ***p < 0.001.

The ability of cardiac glycosides to modulate RNA processing independently of an effect on foci was confirmed by treating WT_MSCs with nanomolar digoxin concentrations. This also led to an increased inclusion of exon 11 in IR and exon 22 in SERCA1 in these cells, although above control values in these cases (Figures 3B and 3C). Similar results were obtained with other cellular systems, namely, primary cultures of myoblasts and human embryonic stem cells (hES)-derived neurons (Figures 3D and S3C). In addition, there was no effect of digoxin, either on the level of expression or on the localization of MBNL1 and CUGBP1 in DM1 cells, the mislocalization and change in bioavailability of which are deemed responsible for the impairment of alternative mRNA splicing (Figures S4A–S4C and S4E). The consequences of digoxin treatment on alternative mRNA splicing were thus explored more comprehensively using deep RNA sequencing, comparing DM1_MSCs with or without treatment with 50 nM of the drug over a period of 2 days. Analysis of annotated splice junctions identified 44 modified splices with Delta ψ ≥ 20 and 258 with 10 ≤ Delta ψ ≤ 20 (Tables S1 and S2). The comparison of this dataset with a recently published list of genes, which exhibited dysregulated alternative RNA splicing in muscle biopsies from patients with DM1 (Nakamori et al., 2013), revealed a beneficial effect of digoxin treatment on 10 different DM1-specific splicing defects as determined by qRT-PCR analysis (Figure 4).

Figure 4

Beneficial Effect of Digoxin Treatment in the Nanomolar Range on Other Alternative Splicing Associated with DM1

RT-PCR analysis of 10 forms of alternative splicing reported to be affected in patients with DM1 and rescued by digoxin treatment (50 nM for 48 hr) in WT and DM1_MSCs. The percentage for inclusion of isoform was determined by RT-PCR, followed by analysis with a 2100 Bioanalyzer (Agilent). Quantification of each band was performed using 2100 Expert software. Data are presented as mean ± SD (n = 3). Data were analyzed with one-way ANOVA, followed by a Dunnett post-hoc test. *p < 0.05, **p < 0.01, ***p < 0.001.

Finally, the effect of digoxin was evaluated in vivo, in the HSALR transgenic DM1 mouse model, which expresses 220 CUG repeats in skeletal muscle and exhibits dysregulated alternative splicing (Mankodi et al., 2002). Digoxin was administrated daily by intraperitoneal injection for 7 days at two dosage regimens: 0.02mg/kg and 2 mg/kg/ d. Rescue of Mbnl1-, Ldb3-, and Clcn1-dysregulated RNA splicing was observed after digoxin treatment at the lower dosage (Figures 5A and S5), independent of any changes in expression of the HSA transgene (Figure 5B). Mbnl1 exon 5 inclusion, for instance, was recorded at 17.44% ± 2.7% after digoxin treatment at 0.02mg/kg/d, down from 35.13% ± 2.2% in untreated HSALR mice (FigureS5).

Figure 5

Digoxin Treatment Partially Restores In Vivo Splicing Defects in a DM1 Mouse Model

(A) Analysis of Mbnl1, Ldb3, Clcn1 alternative splicing by RT-PCR in HSALR mice treated with intraperitoneal injection of digoxin for 7 days (n = 3 for wild-type, n = 4 for mock-treated, and n = 5 for treated). RT-PCR was followed by analysis with a 2100 Bioanalyzer (Agilent). Quantification of each band was performed using 2100 Expert software. Data are presented as mean ± SEM. The p value was determined using the Mann-Whitney U test; *p < 0.05.

(B) Quantitative real-time qRT-PCR analysis of HSA transgene as a function of digoxin treatment. Data were normalized with the housekeeping gene 18S.

Molecular Mechanisms that Underlie the Effects of Cardiac Glycosides on Alternative RNA Splicing

A series of hypotheses were sequentially explored in a search for the molecular mechanisms underlying the potential therapeutic effects of cardiac glycosides.

We first checked whether these effects were associated with a change in the intracellular concentration of calcium, which is the common outcome of treatments with cardiac glycosides. To this end, we used the calcium ionophore A23187, which increases intracellular calcium content (Dedkova et al., 2000). Treatment with the calcium ionophore A23187 had a similar effect on the IR splicing ratio as that produced by digoxin. A combination of digoxin and A23187 led to neither an additive nor a synergistic effect (Figure 6A), supporting the hypothesis of a calcium-dependent mechanism for digoxin effects on alternative mRNA splicing. This was further validated by the abolition of the drug effect on alternative IR and SERCA1 splicing when calcium was withdrawn from the cell culture medium (Figures 6B and 6C).

Figure 6

Calcium-Dependent Effect of Digoxin on Alternative Splicing of IR and SERCA1

(A) Pharmacological approaches to analyze the combinatorial effect of AKT, GSK3, ERK, and calcium pathways with digoxin on alternative IR splicing. DM1_MSCs were treated for 48 hr, with or without 50 nM of digoxin, in combination with different concentrations of AKT inhibitor (LY29002), GSK3 inhibitor (CHIR99021), ERK inhibitor (PD0325901), or calcium ionophore (A23187). WT_MSCs were used as a positive control. Alternative IR splicing was quantified using qRT-PCR. Data are presented as mean ± SD (n = 3).

(B and C) Influence of calcium on the digoxin-mediated effect on alternative splicing of IR (B) and SERCA1 (C) in DM1_MSCs. Cells were maintained in medium with or without calcium and treated with digoxin (50 or 100 nM) for 48 hr. Medium without calcium was added 30 min before the digoxin (n = 3). Data were analyzed with one-way ANOVA followed by a Bonferroni post-hoc test. *p < 0.05, **p < 0.01, ***p < 0.001.

(D) Effect of different concentrations of digoxin on mitogen-activated protein kinase and mammalian target of rapamycin (mTOR) signaling pathways. Expression level of key modulators of Akt/mTOR signaling in DM1_MSCs treated with different concentrations of digoxin for 48 hr, including Phospho(Ser473)-Akt, Phospho(Ser21/9)-GSK3α/β, Phospho(Ser235-236)-ribosomal protein S6 (rpS6), Phospho (Thr202/Tyr204) ERK, and the corresponding total forms. Antibodies used are liste in Table S3. Western blot analysis was quantified by using ImageJ software. Data are presented as mean ± SD (n = 4) and p value were determined using paired t test. *p < 0.05, **p < 0.01.

The next step in the analysis aimed at determining intracellular signaling pathways that may be triggered by the drug and that may affect alternative splicing. Cardiac glycosides are sodium-potassium pump inhibitors that activate multiple cell signaling pathways, representing many other possible mechanisms by which these compounds might affect alternative splicing. Following treatment of DM1_MSCs with 50 or 100 nM digoxin for 48 hr, western blot analyses indicated a moderately increased phosphorylation of AKT and GSK3-β and a decreased activation of ERK (Figure 6D). A combination of digoxin with either LY294002 or CHIR99021, which specifically inhibit AKT and GSK3, respectively, attenuated the beneficial effect of digoxin on IR exon 11 inclusion (Figure 6A), indicating that these pathways contribute toward the effects of the drug. Conversely, inhibition of ERK phosphorylation by PD0325901 in DM1_MSC-normalized IR exon 11 inclusion in the absence of digoxin, and co-treatment with digoxin, resulted in an additive or synergistic effect, suggesting a cooperative mode of action (Figure 6A). Finally, two RNA-binding proteins that were previously shown to be modulated by cardiac glycosides (Anderson et al., 2012), namely, SRSF3 (SRp20) and TRA2B, were investigated. Western blot analysis revealed that their expression levels were not modified in treated DM1_MSCs (Figures S4D and S4E).

Functional Effects of Digoxin on Myogenic Differentiation

In an attempt to identify a functional correlate for these biological results, the digoxin effect was challenged on DM1 primary human myoblast cells carrying 2000 CTG repeats and displaying defective myotube differentiation in vitro (Bigot et al., 2009, Thornell et al., 2009). Digoxin treatment at 50 nM for 7 days increased the number of myotubes that were formed, as quantified by both myosin heavy chain immunostaining (number of MF20-positive cells), and the number of nuclei per MF20-positive cell (Figures 7A and 7B and S6). Calcium deprivation in the culture medium during myotube differentiation fully abolished the functional benefit of digoxin treatment (Figure 7C). Similar positive results on the myogenic process were obtained by applying a 50 nM digoxin pulse for the first two days of myoblast differentiation. This observation was further confirmed using other cardiac glycosides (Figure S7). In addition, qRT-PCR analysis of alternate splicing variants revealed a normalization of MBNL1 exon 7, INSR exon 11, and DMD exon 78 in myotubes treated with 50 nM digoxin.

Figure 7

Digoxin Treatment Restores In Vitro Myogenic Defect in DM1

(A) Representative immunostaining for the myogenic marker Mf20 in WT and DM1 myoblasts differentiated for 6 days. DM1 myoblasts were treated with 50 nM of digoxin at day 1 of differentiation.

(B) Automated quantification of myogenic differentiation as determined by the measurement of the total area stained for Mf20 per field in WT-differentiated myoblasts and in DM1-differentiated myoblasts, treated or untreated with 50 nM digoxin, as well as the quantification of nucleus area per Mf20-positive myotube. All data were analyzed with one way ANOVA followed by a “Bonferroni's Multiple Comparison Test” ∗p<0.01, ∗∗∗∗p<0.0001.

(C) Calcium-dependent effect of digoxin on the in vitro myogenic impairment observed in DM1 myoblasts. Cells were differentiated in the presence or absence of calcium. The effect of digoxin treatment (50 nM for 7 days) on myogenesis was evaluated by automated quantification according to the total area stained for Mf20 per field and as the percentage of the total area of nuclei identified within Mf20-positive myotubes. Data are presented as mean ± SD (n = 3 independent experiments, 3 replicates) and were analyzed with one-way ANOVA, followed by a “Bonferroni's multiple comparison test”; **p < 0.05, ***p < 0.001.

(D) Analysis of splice variants in WT and DM1 myoblasts differentiated for 8 days. Digoxin 50nM was applied as a chronic (7 days) or pulse (2 days) treatment. Percentage of exon inclusion was quantified by RT-PCR and Capillary Electrophoresis using the Agilent bioanalyzer 2100. Data represent mean ± SD (n=1 experiment – triplicates). All data were analyzed with one way ANOVA followed by a “Bonferroni's Multiple Comparison Test” ∗∗p<0.05, ∗∗∗p<0.001.

Discussion

The main outcome of this study is the identification of the cardiac glycoside chemical family as constituting a potential therapeutic strategy for DM1. These compounds had the capacity to promote changes toward normalization in alternative mRNA splicing of genes affected by the DM1 mutation and myogenic fusion of affected myoblasts. These results were obtained using high-throughput drug screening, based on human pluripotent stem cell line derivatives that carried the mutant gene and recapitulated the cardinal cellular features of the disease. Similar experimental paradigms may be proposed for a vast array of monogenic diseases as these cellular models give access to relevant readouts.

DM1 is characterized by the accumulation of the mutant DMPK mRNA in cell nuclei, due to the presence of an abnormal tract of CTG triplet repeats in the 3′ UTR of the DMPK gene. This mutation provokes a major alteration in the bioavailability or activity of RNA-binding proteins, most particularly MBNL1 and CUGBP1, in part through the formation of intranuclearribonucleoprotein aggregates (Mankodi et al., 2001). This, in turn, induces changes in alternative RNA splicing in a number of genes that are physiological targets of those proteins. In the present study, cardiac glycosides were shown to affect these pathological mechanisms in different ways, depending upon their concentration. At high cytotoxic concentrations—in the micromolar range—they provoked a partial disaggregation of the inclusions. This, however, did not correlate with a normalization of the alternative RNA splicing activity, which rather deteriorated further as a consequence of cellular toxicity. Such a phenomenon was reminiscent of what was observed in a previous study in which the expression of MBNL1 was decreased in primary fibroblasts derived from patients with DM1 using a specific small interfering RNA (Childs-Disney et al., 2013).

The large variety of alternate splicing defects observed in DM1 are thought to be responsible for the multisystemic nature of this disease, and some reports have already linked the alteration of single mRNA variants to specific disease phenotypes. For instance, the increased inclusion of CLC1 exon 7a is involved in the onset of myotonia, and repression of this exon inclusion using morpholino antisense oligonucleotide greatly reduced the myotonic grade observed in genetic DM1 mouse models (Wheeler et al., 2007). Therefore normalization of abnormal ratios of these transcripts, if obtained under treatment, is expected to be therapeutically relevant. The effects of cardiac glycosides do not, however, specifically target the causative molecular defect of the disease and the extent of the success of treatment for different disease phenotypes can only be partial. Whether the partial recovery of ratios might be therapeutically relevant will need to be explored in a clinical setting. Nevertheless, the positive effect of digoxin recorded in the present study on a myogenesis assay indicates that cardiac glycosides might have functional consequences. These compounds are still regularly prescribed for the management of atrial fibrillation and heart failure in the general population, although their clinical use has been steadily decreasing because of their narrow therapeutic window, which makes them a toxicity risk (Kanji and MacLean, 2012). Given that cardiac conduction abnormalities are critical symptoms in DM1, careful consideration of the indication and cautious observation will be required for administration, as well as close monitoring of the side effects of cardiac glycosides in patients with DM1. However, there are already reports on several patients with DM1 exhibiting good tolerance to digoxin or digitoxin (Finsterer and Stollberger, 2012, Lagoeiro Jorge et al., 2012, Phillips and Harper, 1997, Schmitt and Schmidt, 1975). It is also notable that the narrow concentration range in which digoxin exerts a beneficial effect on DM1 phenotypes in this study is similar to tissue concentrations found in patients treated for heart failure (Hothi et al., 2014, Schoner and Scheiner-Bobis, 2007).

The ultimate goal of a DM1 therapy is the correction of all affected genes. Targeting the causative mechanisms of the pathology, at the level of either the DMPK gene mutation or by introducing competition to its mRNA binding to splicing factors such as MBNL1, would obviously be the most elegant therapy. This is being actively pursued by a number of teams who have developed various approaches based on gene transfer, antisense oligonucleotides, or specific synthetic small molecules (Childs-Disney et al., 2012, Childs-Disney et al., 2013, Francois et al., 2011, Ketley et al., 2013, Laustriat et al., 2015, Lee et al., 2012b, Mulders et al., 2009, Nakamori et al., 2016, Pandey et al., 2015, Parkesh et al., 2012, Wheeler et al., 2012). Current results point to a parallel pharmacological approach, aiming at modifying alternative RNA splicing of as many DM1-affected genes as possible, without affecting the causative mechanisms of the disease. Such an approach would benefit from a combination of drugs. Indeed, cardiac glycosides are not an isolated case. Two small molecules have recently been clinically evaluated in DM1: the anti-myotonic mexiletine (Logigian et al., 2010) and, more recently, the anti-diabetic drug metformin (Laustriat et al., 2015). Interestingly, the mechanism of action for cardiac glyocosides differs from that for these previously mentioned compounds. Cardiac glycosides modulate the Na+/K+ ATPase protein complex upon binding and activate downstream signaling pathways via increased intracellular calcium concentrations (Fontana et al., 2013), which can independently induce changes in alternative mRNA splicing patterns (Razanau and Xie, 2013, Sharma and Lou, 2011, Sharma et al., 2014). In this case, the molecular mechanisms linking calcium concentration to the modulation of the splicing process may involve chromatin modifications, rather than RNA-binding proteins (Sharma et al., 2014). Thus the effects of metformin and cardiac glycosides do not rely on fully identical intracellular pathways. Therefore they may also modify various subsets of DM1-affected genes. It is thus tempting to speculate that a combination of several drugs may lead to additive or synergistic beneficial effects (Konieczny et al., 2017). From this perspective, it will also be interesting to analyze the effects of other drugs known to affect alternative RNA splicing, e.g., valproate, amyloride, or kinetin (Axelrod et al., 2011, Chang et al., 2011, Farrelly-Rosch et al., 2017), in the context of DM1 in vitro models. Calcium changes in treated cells may not, however, be the sole mechanism underlying the effects of cardiac glycosides. In the present study, use of PD0325901, an inhibitor of the ERK pathway, appeared as yet another path for regulating alternative mRNA splicing, which may normalize a further subset of DM1-affected genes. DM1 myoblasts have been described to sustain abnormal ERK activity that might be responsible for the defective myogenic differentiation (Beffy et al., 2010). It would therefore be interesting to determine whether the digoxin effect on the functional in vitro myogenesis was triggered by a splicing modification or through an ERK1/2 inhibitory effect. In addition, our results also demonstrated that digoxin treatment led to an increased level of the inactive form of GSK3-β (phosphorylated at S9) in DM1_MSCs. Interestingly, inhibition of GSK3-β signaling with tideglusib is actually under clinical investigation in patients with DM1 (ClinicalTrials.gov identifier: NCT02858908). This development was based on different studies demonstrating that the active GSK3-β is elevated in skeletal muscle biopsy samples from patients with DM1 and in skeletal muscle of mice. Inhibition of GSK3-β signaling in both DM1 cell culture and mouse models reduced muscle weakness and myotonia in DM1 mice (Jones et al., 2012, Wei et al., 2018). In our study, the use of CHIR99021, a well-characterized GSK3-β inhibitor, did not improve IR exon 11 inclusion and minimized the digoxin effect when co-treatment was applied. Even if a combination of a cardiac glycoside with a GSK3-β inhibitor might theoretically allow the modification of a larger variety of exons and phenotypes, a combinatorial therapeutic strategy needs to be carefully evaluated, as one strategy can abolish the effect of the other.

The present study further emphasizes the value of pluripotent stem cell lines derived from mutant-gene-carrying donors as in vitro cell models for monogenic pathologies and drug discovery. Access to a never-ending supply of cells expressing a physiologically and pathologically relevant human genome is a first advantage over other cell models. The physiological relevance of our cellular model was also of particular importance here, because ratios of gene transcripts are highly dependent upon cell phenotypes. It is interesting to note that a previous study has already demonstrated the ability of cardiac glycosides to modulate alternative mRNA splicing of a large number of genes (Stoilov et al., 2008). However, using HEK cells as a model, this list did not include the 15 DM1-affected genes identified in the present study using mesodermal and neural derivatives of ES cells. In addition, the use of stem cell derivatives allows a comparison of populations of cells displaying an exact same phenotype, which differ only in their genotype (e.g., DM1 versus WT).

In summary, stem cell derivatives provide a very interesting platform for in vitro drug discovery, but results need to be confirmed using other relevant in vitro and in vivo models, as performed here, before identifying compounds that can be moved forward toward clinical application.

Limitations of the Study

In this study, we identified the capacity of the FDA-approved cardiac glycosides to normalize several alternative mRNA splicing of genes affected by the causal mutation of DM1. Our results demonstrated that this effect is calcium dependent and can be synergic to the inhibition of the ERK signaling pathway. However, the exact molecular mechanisms by which cardiac glycosides can induce the normalization of these affected alternative splicing is still unclear. Our results suggest that the mechanisms of action of these compounds are not dependent on the two main splicing factors, known to be involved in the pathogenesis of DM1. Therefore additional experiments are needed to investigate the mechanisms by which cardiac glycosides normalize alternative splicing defects in DM1.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.