mTOR-dependent proliferation defect in human ES-derived neural stem cells affected by myotonic dystrophy type 1

Owing to their ability to differentiate into a large spectrum of cell types, human disease-specific pluripotent stem cells of embryonic origin (hESCs), or more recently cells derived from the reprogramming of somatic cells (hiPSCs), appear to be a unique material in which to correlate molecular and functional pathological mechanisms within difficult-to-obtain cell populations (Maury et al., 2011). In the context of human neurological disorders, most of our current knowledge about molecular and cellular phenotypes has been gathered from studies in postmortem brain tissues, which are typically the end-stage of the disease and consequently may not be a fair illustration of how the disease developed. Human disease-specific pluripotent stem cells provide a unique opportunity to complement findings in human biopsies and animal models by analyzing the first stages of disease progression and deciphering molecular mechanisms that could be at the origin of a disease-related functional phenotype.

Over the past few years, a growing number of studies have demonstrated the usefulness of this approach to study both neurodevelopmental and neurodegenerative disorders. For example, pre-implantation genetically diagnosed embryos have been used to analyze molecular and functional defects in hESCs-derived neurons carrying a mutant gene responsible for fragile X (Eiges et al., 2007) and in motoneurons with myotonic dystrophy type 1 (DM1) (Marteyn et al., 2011). Other early childhood-onset monogenic disorders, such as Rett syndrome, familial dysautonomia disorder and spinal muscular atrophy, have been studied following the generation of hiPSCs from patients, and a defective neuronal phenotype associated with the expression of the mutant gene (Ebert et al., 2009; Kim et al., 2011; Lee et al., 2009). Human iPSCs have also been instrumental in deciphering a familial case of the late-onset neurodegenerative Parkinson's disease, associating the expression of a mutant gene with an increased sensitivity of hiPSCs-derived dopamine neurons to oxidative stress (Seibler et al., 2011).

Myotonic dystrophy type 1 is an autosomal dominant neuromuscular disorder with late clinical onset, caused by the expansion of trinucleotide CTG repeats in the 3′-untranslated region of the DMPK gene (Mahadevan et al., 1992). At the skeletal muscle level, three main molecular events can be described: (1) formation of nuclear foci that are composed at least of mutant DMPK mRNA and recruited RNA-binding proteins, such as splicing regulators and transcription factors (Ebralidze et al., 2004; Mankodi et al., 2001; Timchenko and Caskey, 1996); (2) disturbance of specific gene expression (Mankodi et al., 2002; Savkur et al., 2001; Yadava et al., 2008); and (3) impairment of cell proliferation and differentiation (Bigot et al., 2009; Furling et al., 2001; Timchenko et al., 2001a). Although DM1 has long been considered mainly as a muscle disorder, there is extensive evidence for the involvement of the central nervous system. Psychological dysfunction, mental retardation, excessive daytime sleepiness, and neuropathological abnormalities have been described in DM1 patients (Abe et al., 1994; Delaporte, 1998; Modoni et al., 2004; Modoni et al., 2008; Perini et al., 1999; Turnpenny et al., 1994). Nevertheless, in contrast to the substantial advances in understanding DM1 muscle pathology, the molecular and functional bases of DM1 in the central nervous system are still largely unknown.

In looking for molecular and functional mechanisms that could be implicated in neural development in DM1, we based our approach upon DM1-specific human pluripotent stem cells and their ability to differentiate into neural cells. We recently demonstrated that human embryonic stem cells (hESCs) obtained during pre-implantation genetic diagnosis for DM1 (Mateizel et al., 2006) represent a relevant cellular model for DM1. In particular, these human DM1-specific pluripotent stem cells have allowed us to discover new molecular mechanisms impeding the connectivity between DM1 hESCs-derived motor neurons and their muscle targets (Marteyn et al., 2011). To more broadly analyze the molecular and functional effects of the DM1 mutation in neural cells, we have taken advantage of progress made in neural differentiation protocols to obtain robust and homogenous neural precursor cell populations (Chambers et al., 2009). We show here that these neural cells exhibit a proliferation deficiency linked to a DM1-related impairment in mTOR signaling pathway activity, associated with increased autophagy.

Two human embryonic stem cell (hES) lines carrying large CTG expansions (VUB03_DM and VUB24_DM) obtained from independent couples, and one wild-type control cell line (VUB01_WT) were successfully differentiated toward a homogenous population of neural stem cells (NSCs). No differences in morphology and in the expression of NSC-specific markers such as Nestin and Sox2 were observed during the differentiation process (Fig. 1A–C). Moreover, no difference in neuronal differentiation was observed between control and mutant cells at day 20 for most of the analyzed gene markers, including SNAP25, DLG4 and MAPT, although we cannot completely exclude a difference between DM1 and WT over the time course of neuronal differentiation (supplementary material Fig. S1). To validate their pathological relevance, we performed in situ hybridization combined with immunostaining which showed that DM1-NSCs and derived neurons harbored intranuclear ribonucleoprotein aggregates (known as ‘foci’), formed by the mutant RNA and the splice factor MBNL1, which are typical of the disease (Fig. 1D and supplementary material Fig. S1). The nuclear retention of MBNL1 was also confirmed by western blotting (Fig. 1E).

Quantitative analysis of doubling time by measurement of the ATP concentration in cell cultures indicated that the proliferation rate of DM1-NSCs was reduced compared to controls (Fig. 2A,B). Consistent with this observation, a decreased number of Ki67-positive cells was observed in DM1-NSC cultures (Fig. 2C), in association with decreased PCNA gene expression and decreased phosphorylation of the Rb protein (Fig. 2D,E). In addition, expression of other cell cycle regulators was deregulated in DM1 cells in comparison with control, with increased p27^kip1 (Fig. 2F) and P15 proteins and decreased cyclin D1 and P21 (Fig. 2G).

Measurement of necrosis or apoptosis by annexinV and propidium iodide staining, PARP cleavage and caspase-3 expression revealed no differences between DM1- and WT-NSCs. Similarly, staining for β-galactosidase did not reveal any differences in the proportion of senescent cells in the DM1- and WT-NSC cultures (supplementary material Fig. S2). Altogether, these results indicated that the reduced proliferative capacity of DM1-NSCs was not associated with an increase in either cell mortality or senescence.

Large cytoplasmic vacuoles were observed in DM1-NSCs, suggesting the induction of autophagy (Fig. 3A). Confirming this functional phenotype, increased expression of the autophagic markers LC3B-II, p62 and ATG12 protein was observed in DM1-NSCs (Fig. 3B–F), as well as significantly increased expression of the lysosomal cathepsin B marker both at the mRNA and the protein level (supplementary material Fig. S3). In addition, the nucleofection of a plasmid encoding the LC3–GFP fusion protein, used as a marker for autophagic vesicles and pre-autophagic compartments, resulted in more LC3–GFP aggregates in DM1-NSCs relative to controls (Fig. 3C,D). As a positive control, treatment of WT-NSCs with either EBSS medium or chloroquine resulted in an increased number of cells containing LC3–GFP aggregates. No effect of chloroquine treatment was observed on DM1-NSCs suggesting that these cells could not accumulate additional autophagic vacuoles.

To assess the central role of AKT/mTOR signaling pathways in both cell-cycle control and autophagy, expression of their main components, namely Akt, AMPK, GSK3α/β and rpS6 (Ser235/236) (ribosomal protein S6) was systematically analyzed. Gene expression and protein levels were unaltered for all of these components, when analyzed by quantitative PCR and western blotting, respectively (Fig. 4; supplementary material Fig. S4). In contrast, analysis of the post-translational activation of these proteins through phosphorylation revealed major differences between DM1-NSCs and controls. Whereas phosphorylation of the upstream components Akt (Ser473) and AMPK (Thr172) appeared unaffected, the downstream GSK3α/β (Ser21/9) and rpS6 (Ser235/236 and Ser240/245) were decreased in DM1-NSCs (Fig. 4A,B) through a mechanism independent of a stress-induced activation of P53 (supplementary material Fig. S4).

To eliminate the possibility of a non-specific effect of the growth factors present in the medium [i.e. epidermal growth factor (EGF), fibroblast growth factor (FGF2) and brain-derived neurotrophic factor (BDNF)] on mTOR signaling pathway activity, NSCs were starved for 48 hours and then treated for 1 hour with each factor alone or in combination. No effect of EGF and/or BDNF was observed on phosphorylated rpS6 (Ser235/236) levels. Treatment with FGF2 resulted in a dramatic increase in phosphorylated rpS6 (Ser235/236) through the activation of phospho-Erk1/2, equally in DM1- and WT-NSCs (Fig. 4C,E). The effect of FGF2-mediated ERK1/2 activation on the phosphorylation of rpS6 was transient, as it disappeared 48 hours after treatment. As all our experiments were performed 48 hours after treatment with growth factors, these results suggest that there might be a differential impact of growth factors in the medium on mTOR signaling pathways in DM1- and WT-NSCs.

A potential link between the functional defects in proliferation and autophagy observed in DM1-NSCs and the decreased activation of mTOR signaling was explored by analyzing the functional consequences of a partial pharmacological blockade of the pathway in WT-NSCs. Cells were treated with rapamycin, a specific mTOR inhibitor. Rapamycin provoked a dramatic decrease in phosphorylated rpS6 (Ser235/236) (P-rpS6) without substantial modulation of either phosphorylated Akt (Ser473), GSK3α(Ser21) or GSK3β (Ser9) (Fig. 5A,B; supplementary material Fig. S4). Functional defects associated with this molecular alteration were reminiscent of the phenotype described above for DM1-NSCs, as rapamycin-treated WT-NSCs exhibited impaired proliferative capacity (Fig. 5C,D; supplementary material Fig. S4) and induction of autophagy, as shown by LC3–GFP aggregation after transfection with the fusion protein plasmid (Fig. 5E,F).

Association of mTOR signaling alterations with expression of the mutant DMPK gene was controlled by transient transfection of WT-NSCs with a plasmid containing an extended stretch of 960 CTG repeats. Treated cells exhibited a profound decrease in P-rpS6 as well as P-GSK3β, which was consistent with our working hypothesis (Fig. 6A,B; supplementary material Fig. S5). The main consequence of the presence of an extended stretch of CUG repeats in DM1-NSCs is the decreased bioavailability of the MBNL1 protein due to its sequestration in intranuclear foci. To reproduce this phenomenon in WT-NSCs, MBNL1 gene expression was knocked down using specific siRNAs. MBNL1-depleted cells exhibited significantly decreased levels of P-rpS6 and, to a lesser extent, decreased levels of P-GSK3α/β (Ser21/9) (Fig. 6C). Conversely, overexpression of MBNL1 in DM1-NSCs elicited a partial restoration of the phosphorylation of rpS6 (Ser232/236) (Fig. 6D). These changes in MBNL1 expression in DM1- and WT-NSCs led to altered proliferation and autophagy (Fig. 6E,F).

The primary finding of the current study was the identification of an altered mTOR signaling pathway, which results in reduced proliferative capacity and induction of autophagy in neural cells derived from DM1 gene-carrying human embryonic stem cells. This alteration in the mTOR signaling pathway could be reproduced by reducing the bioavailability of the RNA-binding protein MBNL1 in WT cells, mimicking the defect associated with the expression of the DM1 mutation. The functional abnormalities associated with the altered mTOR pathway might contribute to the neurological aspects of this disease. This study further highlights the value of mutant gene-carrying human stem cell lines obtained from pre-implantation genetically diagnosed embryos in helping to decipher the molecular mechanisms and the functional consequences related to these mutations.

Molecular mechanisms involved in DM1 pathophysiology have been mainly identified by using animal models. Few human cell culture models have been developed; most are based on the use of muscle precursor cells derived from congenital forms of DM1, which are characterized by very large stretches of CTG repeats (>2000). Myoblasts derived from congenital forms of DM1 have alteration in various cell-cycle modulators, including p21 and cyclin D1 (Timchenko et al., 2001b; Timchenko et al., 2004). They also exhibit premature senescence through a p16-dependent mechanism and defect in p38MAPK and ERK MEK (Bigot et al., 2009), leading to defective proliferation and differentiation (Beffy et al., 2010; Bigot et al., 2009; Timchenko et al., 2001b; Timchenko et al., 2004). In addition, the induction of autophagy has been recently observed in myoblasts derived from congenital DM1 patients; a link to a p53-dependent inhibition of mTOR pathway in response to metabolic stress has been hypothesized (Beffy et al., 2010).

Our results have in part confirmed these data, by extending them to DM1 neural precursor cells, in which altered proliferation and induction of autophagy were observed. However, molecular mechanisms responsible for those effects were not similar as neither persistent activation of the FGF2-dependent MEK–ERK pathway nor abnormal p53 activation was observed in DM1-NSCs. Therefore, the current results do not support the hypothesis of a cellular stress response in DM1 cells, which would be secondarily responsible for the mTOR signaling pathway alterations. Nevertheless, we cannot exclude the possibility that these discrepancies are due to the differences between either the cell types analyzed or the number of CTG repeats in the two cellular systems. Indeed, mTOR signaling was altered in the current study in the presence of only 1000 CTG repeats in DM1-NSCs or after transfection of a plasmid with 960 repeats in WT-NSCs, i.e. in conditions that are unlikely to be linked with congenital forms of the disease.

Conversely, our data point to a specific defect in the activity of the glycogen synthase kinase 3 (GSK3) being involved in the disease. GSK3 activity is inhibited through phosphorylation of serine 21 in GSK-3α and serine 9 in GSK-3β, which have been identified as targets of Akt and AMPK (Fang et al., 2000). The inhibition of GSK3 may participate in the activation of mTOR, which results in the increased phosphorylation of rpS6 (Jastrzebski et al., 2007). According to this cascade of events and in the absence of any alteration in the activation of Akt or AMPK in DM1 neural cells, the decreased mTOR-dependent phosphorylation of rpS6 can be directly associated with this decreased inhibition of GSK3. Alternatively, as GSK3 acts as an integrator of Akt and Wnt signals, the observed alteration of the GSK3/mTOR pathway is related to a Wnt defect (Ma et al., 2010; Vigneron et al., 2011). Consistent with the results of the current study, alteration of GSK3 phosphorylation was also observed in a rat PC12 cell line expressing 90 CUG repeats, which resulted in altered expression and phosphorylation of Tau (Hernández-Hernández et al., 2006). Given the crucial role of Tau in neurodegenerative diseases, further investigations into whether the decreased GSK3 phosphorylation observed in DM1-NSCs could lead to defects in Tau expression would be of interest.

Mechanisms by which the mutation causing DM1 could modulate the phosphorylation of specific molecular targets of the mTOR signaling pathways remain to be determined. Three main hypotheses may be proposed: (1) haploinsufficiency of the DMPK gene (Reddy et al., 1996); (2) alteration of the expression of neighboring genes such as SIX5 and DMDW (Klesert et al., 2000; Sarkar et al., 2000); (3) toxic gain-of-function of the mutant mRNA because of the sequestration of proteins involved in RNA processing, such as the MBNL1 protein (Ranum and Cooper, 2006). Although several lines of evidence suggest that the toxic gain-of-function of the mutant RNA corresponds to the main mechanism by which the mutation resulting in DM1 causes the complex pattern of the disease, the contributions of each of these mechanisms are not mutually exclusive. As an example, insulin-resistance associated with DM1 has been correlated with both DMPK haploinsufficiency and aberrant regulation of alternate splicing of the insulin receptor because of the loss of bioavailability of MBNL1 (Llagostera et al., 2007; Savkur et al., 2001). The current results following gain- or loss-of-function of MBNL1 support the hypothesis of a potential role for a toxic RNA mechanism in mTOR signaling alterations that involves loss of MBNL1 bioavailability. Further studies will be required to determine the mechanism by which MBNL1 may modulate the GSK3/mTOR signaling pathway. However, it is worth mentioning that the effect of gain- and loss-of-function of MBNL1 is less pronounced on proliferation than on autophagy, which suggests the involvement of other modulators in the proliferation phenotype. Accordingly, CUGBP1, another RNA-binding protein whose expression is affected by the presence of the DM1 mutation, has been shown to affect the expression of cell-cycle inhibitor p21 expression in DM1.

A large number of neurological abnormalities have been reported in patients both at the clinical and histological levels (de León and Cisneros, 2008). Central nervous system involvement in DM1 implicates cognitive impairment, hypersomnolence, and personality and behavioral disturbances (Abe et al., 1994; Delaporte, 1998; Meola et al., 2003; Modoni et al., 2008; Turnpenny et al., 1994). Grey matter reduction has also been described in various cortical regions, the hippocampus and the thalamus of DM1 patients (Minnerop et al., 2008; Minnerop et al., 2011; Weber et al., 2010). At the molecular level, nuclear aggregation of mutant mRNAs in combination with MBNL1 has been described in cortical and subcortical neuronal cells of DM1 patients post-mortem (Jiang et al., 2004). Alternative splicing of N-methyl-D-aspartate receptor1 (NMDAR1), amyloid beta precursor protein (APP) and microtubule-associated protein tau (MAPT) is abnormally regulated in DM1 brain tissue samples, even though direct involvement of MBNL1 in these splice defects has not been demonstrated (Jiang et al., 2004). However, the functional consequences as well as the precise neuropathological contributions of these molecular abnormalities are still unclear. MBNL1 knockout mice exhibit cognitive impairments, suggesting that MBNL1 loss-of-function might be implicated in the neuropathology of DM1 (Matynia et al., 2010). The MBNL1 knockout mouse model has also identified novel splicing defects in the brain, including Sorbs1, Dclk1 and Camk2d (Suenaga et al., 2012), but these genes have not been associated with clinical symptoms, yet. Lastly, a very recent study using a MBNL2 knockout mouse model suggests an essential role for this gene in the DM1 developing brain (Charizanis et al., 2012). However, this conclusion needs to be qualified in the light of redundant functions of the different types of MBNL1 and MBNL2 proteins (Wang et al., 2012).

The results of the current study shed some light on the neuropathology of DM1, as we have identified an MBNL1-dependent mTOR signaling defect, leading to functional abnormalities, in neural precursors that have been shown by previous authors to exhibit a telencephalic phenotype (Chambers et al., 2009). Tentatively, and subject to further validation, one may relate the proliferation defects observed in DM1-NSCs to the reduced volume of the grey matter in DM1 patients. mTOR-dependent induction of autophagy may also participate in the neuropathology, as a similar process has been suggested to play a role in other neurological disorders involving abnormal protein aggregations, such as Parkinson's, Alzheimer's and Huntington's diseases (Garelick and Kennedy, 2011).

hES cells were propagated as previously described (Marteyn et al., 2011). The differentiation of hES into NSCs was performed using a SMAD inhibitor protocol (Chambers et al., 2009). Briefly, hES colonies were mechanically dissociated and incubated in suspension with N2B27 medium containing DMEM/F12 medium with Glutamax I mixed with neurobasal medium; N2 supplement; B27 supplement without vitamin A; β-mercaptoethanol (all from Invitrogen) (1:1 vol). The following were also added: human recombinant noggin, a BMP pathway inhibitor, used at 300 ng/ml (PeproTech); SB-431542, a TGFβ pathway inhibitor, used at 20 µM (Tocris); and Y-27632, a ROCK inhibitor, used at 20 µM (Calbiochem). After 6 hours, aggregates were transferred to a dish pre-coated with 0.01% polyornithine (Sigma) and laminin 1 ng/ml (Sigma) and maintained with medium (without Y-27632). After the appearance of neural rosettes following 8–10 days of differentiation, the medium was replaced with the N2B27 medium supplemented with EGF at 10 ng/ml (R&D Systems), FGF2 at 10 ng/ml (PeproTech) and human BDNF (hBDNF) at 20 ng/ml (R&D Systems). For differentiation into neurons, NSCs were seeded at a density of 50,000 cells per cm² on polyornithine/laminin-coated dishes in N2B27 medium containing hBDNF but without EGF and FGF2. Medium was changed every other day for 20 days. Treatment with chemical compounds such as rapamycin (Cell Signaling), chloroquine and bafilomycin (Sigma) was performed according to the manufacturer's instructions.

A senescence assay was performed using a senescence detection kit based on β-galactosidase activity, as per the manufacturer's instructions (Abcam). Apoptosis assay was performed using Vybrant Apoptosis Assay Kit (Molecular Probes) according to manufacturer's recommendations and proliferation was measured using CTG Cell Titer Glo® (Promega). Briefly, at day 1 NSCs were plated in 96-well plates in 100 µl medium at 2000 cells/cm². CellTiter-Glo® reagent (Promega) was added to the cells and the bioluminescence signal was read 40 minutes later on an AnalystGT microplate reader (Molecular devices).

Total RNA from cells was extracted using the ‘RNeasy Mini Kit Protocol’ (Qiagen). Reverse transcription was performed with SuperScript III reverse transcriptase (Invitrogen) and real-time PCR were performed with SyberGreen PCR Master Mix (Applied Biosystems) on a Chromo4 real-time system (Bio-Rad) as previously described (Marteyn et al., 2011). Primers are listed in supplementary material Table S1.

The sequence encoding the 43 kDa isoform of MBNL1 inserted in the pCDNA3.1 vector and the plasmid (CTG)₉₆₀ were kindly provided by Dr Nicolas Charlet (CERBM-GIE, Illkirch, France). The plasmid expressing the LC3–GFP fusion protein was kindly provided by Prof. Codogno (INSERM U984, Chatenay-Malabry, France). The siRNA sequence for MBNL1 knock-down was previously described (Dansithong et al., 2005).

These expression vectors and siRNAs were nucleofected into NSCs using the Rat Neural Stem Cell Nucleofector Kit (Lonza) according to the manufacturer's instructions.

Immunocytochemisty and fluorescent in situ hybridization were performed as previously described (Marteyn et al., 2011). Primary antibodies are listed in supplementary material Table S2. Immunostaining was analyzed by epifluorescence microscopy with the Zeiss Imager Z1 and the Zeiss Axiovert 40CFL, and images were captured with the Zeiss Axiocam mRM.

The number of nuclei with foci, the number of foci per nucleus and the number of cells with at least two LC3–eGFP-positive autophagosomes were counted using the ArrayScan VTI HCS Reader (Cellomics). Data were examined using the Cell Selecting software (Cellomics) as previously described (Marteyn et al., 2011).

Cells were lysed in RIPA lysis buffer (Sigma) supplemented with anti-protease cocktail (Sigma) and anti-phosphatase (Roche). Nuclear and cytoplasmic fractions were extracted using the NE-PER Kit (Thermoscientific). The protein concentration of the cell extracts was measured at 562 nm using the Pierce® BCA Protein Assay Kit (Perbio Thermo Scientific) according to the manufacturer's instructions and using a plate analyzer (Biotek). SDS-PAGE was performed using NuPAGE® Novex 4–12% Bis-Tris gels (Invitrogen); 10–30 µg of total protein was loaded per lane. iBlot® Gel Transfer Stack (Invitrogen) was used for the transfer onto nitrocellulose membranes. Membranes were blocked with 5% non-fat milk in phosphate-buffered saline containing 0.1% Tween 20 for 1 hour (except for MBNL1 antibody which requires 2% fetal calf Serum in PBST), then incubated overnight at 4°C in 5% non-fat milk in PBST (except for MBNL1 in PBST only) with primary antibodies at the appropriate dilution. The antibodies used are listed in supplementary material Table S2. Immunoreactive bands were revealed using the Amersham ECL Plus™ Western Blotting Detection Reagents (GE Healthcare). Equal protein loading was verified by the detection of actin. Quantification was performed using ImageJ software (NIH).

All the data are given as means ± standard error of the mean (s.e.m.) and were analyzed using GraphPad Prism 6.0 software. All the statistical tests were performed at least three times and one-way ANOVA with Dunnett's post-hoc test was used.

We thank Dr Genevieve Gourdon (INSERM U781, Paris, France) for performing CTG repeats analysis; Karen Sermon (AZ-VUB Brussels, Belgium) for providing DM1 and control hES cell lines; Dr Nicolas Charlet (INSERM, Strasbourg, France) and Dr Glen Morris for providing us with resources; Prof. Patrice Codogno and Isabelle Beau (INSERM U984) for the LC3BeGFP construction and their very helpful advice concerning autophagy; Dr Denis Furling (INSERM U787, Paris, France), Dr Nicolas Sergeant and Dr Marie-Laure Caillet-Boudin (INSERM U815, Lille, France) for their helpful discussions and their expertise on the field of myotonic dystrophy disease; and Dr Delphine Laustriat, Dr Fabrice Casagrande, Julien Come, Pauline Georges, Jacqueline Gide and Rémi Vernet (UMR861, CES, I-Stem, Evry, France) for technical assistance.