Myotonic dystrophy [Dystrophia Myotonica (DM); Steinert’s disease] is the most common muscular dystrophy in adults. The prevalence rates are high in Europe and countries with European descendants (Thornton, 2014; Wenninger et al., 2018). In addition to myotonia, a delayed relaxation after forceful muscle contraction, muscle weakness/atrophy and features involving brain, heart, eyes and other organs usually accompany and worsen the course of disease. Although trials applying a variety of strategies are underway, currently no effective therapy is available (Ashizawa and Sarkar, 2011; Udd and Krahe, 2012; Thornton, 2014).

There are two types of DM: type 1 (DM1) and type 2 (DM2). Both DM1 and DM2 are microsatellite disorders that carry extended short tandem repeats within non-coding regions of different genes (Fu et al., 1992; Liquori et al., 2001). The mutations do not cause loss of function of the genes where they are located. Instead, extended (CTG)_n repeats in the 3′-UTR of DMPK gene (DM1) or (CCTG)_n repeats in the intron 1 of CNBP gene (DM2) generate pathological effects through RNA toxicity, such as stabilizing CELF1 protein and generation of repeat-associated non-ATG (RAN) proteins (Ranum and Day, 2004; Thornton, 2014; Gourdon and Meola, 2017). One of the main mechanisms is Muscleblind-like (MBNL) loss of function- When the DMPK gene is transcribed, the extended RNAs retain in the nuclei and formed “RNA foci,” colocalized and sequestered MBNL proteins. MBNLs are RNA-binding proteins (RBPs) that normally regulate splicing of their downstream targets during developmental transition. Once one of the targets Clcn1 is mis-spliced, myotonia occurs due to loss of CLCN1 protein production (Miller et al., 2000; Mankodi et al., 2003; Pascual et al., 2006; Wheeler et al., 2007). It has been shown that RNA foci also exist in neuronal populations, and mis-spliced targets (e.g., GRIN1, MAPT and APP) were found in DM brains (Jiang et al., 2004). In fact, the consequence of MBNL loss of function may alter several key features of RNA processing, including polyadenylation and transportation. Therefore, it involves a very intricate underlying mechanism that is in need of further investigation (Gourdon and Meola, 2017).

Adult DM patients develop various neuropsychological symptoms including excessive daytime sleepiness, cognitive decline, depression, apathy and avoidant personality (Thornton, 2014; Minier et al., 2018). Similar to DM1, the CNS symptoms in DM2 are common but less severe. In congenital DM (CDM) with several hundreds to more than a thousand of (CTG)_n repeats, intellectual disability is as frequent as respiratory distress that patients may face. Autism spectrum disorder (ASD), attention deficit hyperactivity disorder (ADHD) and social difficulties are accompanied in one-third of the patients (Douniol et al., 2012; Gourdon and Meola, 2017). Brain MRI studies showed diffuse brain atrophy, predominant white matter changes in DM1 brains involving mainly frontal and temporal lobes and milder in DM2 patients (Minnerop et al., 2011, 2018; Wozniak et al., 2014). Neuropathological approaches revealed cell loss, presence of neurofibrillary tangles with tauopathy as the major findings (Caillet-Boudin et al., 2014).

Initial in vitro study on stable transfectans of PC12 cell line expressing extended (CTG)₉₀ has shown inhibition of differentiation but not short repeats (CTG)₅ and (CTG)₆₀, suggesting the pathological (CTG)_n expansion is sufficient to affect terminal differentiation in neuronal cells (Quintero-Mora et al., 2002). In terms of in vivo models, several mouse models have been created for investigating DM mechanisms (Braz et al., 2018), but only limited transgenic lines showed brain phenotypes: one carrying DM human mutation (DM300-328 and DMSXL lines thereafter; Hernández-Hernández et al., 2013); the other is a Tamoxifen-inducible EpA960/CaMKII-Cre line which expresses 960 copies of interrupted (CTG) repeats within human DMPK 3′-UTR (Wang et al., 2017). Our group proved the “MBNL loss of function” hypothesis by creating knockout (KO) mouse lines of three MBNL family members (Mbnl1, Mbnl2 and Mbnl3): While Mbnl1 KO mice (Mbnl1^ΔE3/ΔE3) recapitulate myotonia and typical DM muscle pathology (Kanadia et al., 2003), Mbnl2 constitutive KO mice (Mbnl2 KO, Mbnl2^ΔE2/ΔE2) exhibit REM sleep misregulation and cognitive dysfunction (Charizanis et al., 2012). The deprivation of Mbnl1/2 and all three members in the skeletal muscle recapitulated symptoms of muscle atrophy and respiratory distress, the end-stage DM1 or CDM features (Lee et al., 2013; Thomas et al., 2017). Based on these results, MBNL2 is considered the most critical player among the three for DM brain pathology. To investigate if loss of Mbnl1/2 may cause deleterious effects to the brain and to overcome the embryonic lethality of constitutive KO of Mbnl1/2, we chose Nestin-Cre mouse, a transgenic line which specifically expresses Cre recombinase in neuronal and glial cell precursors, to conditionally KO Mbnl2 specifically in the nervous system. We found combined loss of Mbnl1 and Mbnl2 in Nestin-Cre double KO (DKO, Mbnl1^ΔE3/ΔE3; Mbnl2^cond/cond; Nestin-Cre^+/–) mice showed pronounced missplicing in the hippocampi similar to that of DM patients (e.g., Tau exon 2/3 and exon 10), indicating Mbnl1 may also play a compensatory role in DM brains. Since many of the downstream targets regulated by Mbnl2 were associated with the function of brain wiring (e.g., NMDAR1 or GRIN1, MAPT and DCLK1; Shin et al., 2013; Jadhav et al., 2015; Perez-Rando et al., 2017), we wondered if structural changes of cortical neurons may also be present in our Mbnl KO lines.

MBNL family proteins are RBPs critical for post-transcriptional regulations (Batra et al., 2015; Taylor et al., 2018). Aside from the RNA toxicity derived from microsatellite repeat expansions including RAN translation and secondary CELF1 upregulation (Wang et al., 2007, 2009; Zu et al., 2011, 2017), MBNL loss of function is essential for causing DM phenotypes(Thomas et al., 2017; Chen et al., 2018; Ramon-Duaso et al., 2019). The 2KO and DKO models have shown their potential and suitability for the investigations of DM brain pathogenesis due to extensive and distinct splicing alterations similar to DM patients (Charizanis et al., 2012; Goodwin et al., 2015). However, detailed morphological evaluation has been lacking. Therefore, this project was aiming to look for structural evidence at the cellular level in these mouse models.

The macroscopic studies on DM1 adult patients were generally normal, without characterizing significant brain atrophy. However, in the longitudinal studies of air encephalography, progressive atrophy did occur in adult DM1 patients (Refsum et al., 1967). On the other hand, the severity of brain atrophy was higher in CDM patients and the observations were further confirmed by most series of CT or MRI image studies showing ventricular dilatation and marked brain atrophy (Regev et al., 1987). In our results, we were unable to see major changes in gray and white matter thickness. Although a trend of reduced EC thickness in the mutants was observed, it was unable to reach statistical significance. This may be a result of the relatively young mice selected (2–4 months, due to the reduced life expectancy of the DKOs) of our experimental animals or limited numbers of brain checked. Nevertheless, our results did not notably deviate from the previous findings in adult DM patients.

Neuronal migration is a critical step during cortical development for neurons to accomplish proper positioning and form brain circuitry. During the early embryonic stage, newborn cortical neurons starting from the ventricular zone move radially alongside with radial glial fibers to the cortical plate and form six cortical layers in an inside-out manner (Pilz et al., 2002; Marin et al., 2010). Early cortical plate neurons that settle in layer VI become the corticothalamic projection neurons while layer V pyramidal neurons are mainly corticospinal projection neurons. Layers IV, III and II neurons receive thalamic inputs and projections from other cortical regions and layers (Petreanu et al., 2009). Migration defects could cause detrimental effects and may potentially be linked to neuropsychiatric disorders such as autism (Reiner et al., 2016). Recently, neuroscientists have started to realize that RBPs may potentially regulate migration. For example Nova2, a splicing factor, has shown its function in modulating migration for the upper layer cortical neurons in mice, suggesting the importance of RBPs during this process (Yano et al., 2010; Vuong et al., 2016). Previous reports on gyral architecture in CDM patients and if migratory defects exist in DM were under debate. However, a report showing multiple cases with poor lamination of cortex and architectural arrangement of neurons was published more than half a century ago (Rosman and Kakulas, 1966; Hageman et al., 1993).

In our study, a considerable amount of cortical neurons, such as Cux1-positive upper cortical neurons, failed to reach their final destination in the mutants, suggesting a defect in cortical neuron migration. Unlike glutamatergic cortical neurons, GABAergic inhibitory interneurons derive from another proliferative region, the ganglionic eminence, in the ventral telencephalon. PV-positive interneurons, which play an important role in the gating of sensory-motor circuits (Sachidhanandam et al., 2016), are mostly derived from the medial ganglionic eminence; whereas CR-positive interneurons are largely derived from the caudal ganglionic eminence (Cauli et al., 2014). During development, inhibitory cortical interneurons first migrate dorsally and then tangentially toward the cortex (Marin, 2013). The patterns of PV- and CR-positive cells in the upper cortical regions of our KO mice suggested a relatively mild defect in the distribution of cortical interneurons as well. Our results demonstrated for the first time that loss of function of Mbnl proteins may cause abnormalities in neuronal distribution, which may be caused by migratory alteration. To further confirm if these changes were directly caused by migration deficits rather than a shift in the neuronal lineage, alteration of progenitor proliferation or a result of apoptotic cell death, BrdU pulse-chase experiments would be necessary in the future to directly answer this question.

Positioned cortical neurons initiate dendrite and axon identities postnatally for their specific input/output functions after migration. The growth and formation of dendrites is a fundamental process for proper synaptic connections, equally important as axonal maturation. Dendritic structures are constantly changing, under the influence of experience-dependent plasticity (Maravall et al., 2004). Alterations of dendritic morphology have also been reported in neuropsychiatric diseases, such as Fragile X syndrome (Irwin et al., 2001). In our morphological studies on dendrites, we chose parameters that were functionally relevant. For example, the number of branch orders (bifurcation nodes) may influence signal propagations (Brown et al., 2008). Surprisingly, the impact of lacking Mnbl proteins on apical and basal dendrites was different. We observed greater structural changes in the basal dendrites than the apical dendrites in layers II/III pyramidal neurons of both 2KO and DKO mice. Layers II/III basal dendrites receive glutamatergic excitatory inputs from the thalamus as well as layer VI neurons and the branching patterns are sensitive to sensory experience (Bose et al., 2010). Our results suggest Mbnl loss of function may alter this feedforward signals in the sensory-motor circuitry (Petreanu et al., 2009). Interestingly, the defective basal dendrites in layers II/III cortical neurons were evident and comparable between 2KO and DKO lines, indicating the predominant role of Mbnl2 in determining dendritic branching.

Apical and basal dendrites are functionally specified structures. The molecular mechanisms underlying the formation and arborization of apical and basal dendrites are just about to be revealed (Chow et al., 2009). For example, TAOK2 and EPAC2 are ASD—related genes. EPAC2 is a guanine nucleotide exchange factor (GEF) for the Ras-like small GTPase Rap, downstream of sema3A-Npn1-PlexinA4 signaling cascade, which is activated by TAOK2 (Srivastava et al., 2012; de Anda et al., 2012). Both Epac2 knockdown and Taok2 KO mice exhibit poor basal dendrite complexity yet relatively normal apical dendrites in layers II/III neurons (Srivastava et al., 2012; de Anda et al., 2012), similar to our findings. These phenotypes of basal dendrites may be a shared feature in disease showing autism-associated symptoms, including DM1/CDM (Angeard et al., 2018). Dendritic spines are micron-scale protrusions from dendritic shafts and the sites where most excitatory synapses are located. These dendritic spines functionally connect nerve fibers, which are fundamental for brain circuits. The structure of dendritic spines is dynamic and influenced by synaptic activity. Neurotrophic factors and extracellular matrix glycoproteins are key players that regulate their formation. They are crucial for learning and memory but may be severely damaged under pathological conditions (Arikkath, 2012; Adrian et al., 2014). During spinogenesis, the number of spines were quite few in fetal or newborn animals. The immature appendage/filopodia/filopodia-like spines emerged first and were followed by mature spines later postnatally (García-López et al., 2010). Compared to the length that is more specifically linked to calcium compartmentation and calcium-dependent learning, the width of the spine has a stronger correlation with the synaptic strength (Arellano et al., 2007). Therefore, the primary width alterations suggested the indispensable roles of Mbnl1/2 in maintaining synaptic strength.

Impaired hippocampal electrophysiological activity and neuronal circuits have been identified in the Mbnl2 constitutive KO mice (Charizanis et al., 2012; Chen et al., 2018; Ramon-Duaso et al., 2019). Since dendritic spine morphology is correlated with synaptic function, we applied electron microscopy to evaluate PSDs that were most likely affected by impaired dendritic compositions (Arellano et al., 2007). The PSDs are composed of a variety of proteins including glutamatergic receptors and adhesion molecules, located at the tip of dendritic spines and critical for maintaining structural and physiological integrity of the synapses (Sheng and Hoogenraad, 2007). As expected, the thickness and width of PSDs were reduced in the DKO mice, which is compatible with the results found in dendritic spine analysis, suggesting defects in synaptic transmission and plasticity may occur in these mutants.

Based on previously published data, Mbnl1 and Mbnl2 are highly expressed in the neurons. The distribution of Mbnl1 and Mbnl2 are both nuclear and cytoplasmic (Jiang et al., 2004; Charizanis et al., 2012; Wang E. T. et al., 2012). The distribution of Mbnl proteins may differ in different cell types. For example, the Mbnl1 expression in the cerebellar Purkinje neurons is exclusively cytoplasmic, while in the cortical neurons, cerebellar molecular interneurons and deep cerebellar nuclei neurons, the Mbnl1 expression is both nuclear and cytoplasmic. On the other hand, Mbnl2 is highly expressed in the nucleus of neurons in the frontal cortex (Jiang et al., 2004; Daughters et al., 2009; Charizanis et al., 2012; Wang et al., 2017).

KO mouse models have demonstrated the functions of Mbnl proteins. The overall splicing changes were very mild in Mbnl1 KO but readily detectable through RT-PCR, microarray and RNAseq experiments in Mbnl2 KO. The behavioral evaluation of Mbnl1 KO was limited due to its constitutive KO and marked myotonia that may affect locomotion. On the other hand, we saw spatial learning deficits as well as eletrophysiological changes in Mbnl2 KO mice. Therefore, we thought MBNL2 is the key player in DM brains, in contrast to the relative predominant role of MBNL1 in the skeletal muscle. However, this does not mean an exclusive tissue-specific role for individual MBNL. Evidence from Myo-Cre DKO that KO of Mbnl1/2 in muscle shows severe muscular atrophy and Nestin-Cre DKO that KO of Mbnl1/2 in the nervous system shows complete Tau spliceopathy suggest compensatory roles Mbnl1 and Mbnl2 in different organs (Lee et al., 2013; Goodwin et al., 2015).

In addition to nuclear MBNL that regulate splicing, both cytoplasmic Mbnl1 and Mbnl2 may participate in RNA processing such as RNA localization and facilitate local translation. For example, cytoplasmic Mbnl1 binds to 3′-UTR of downstream gene targets and transports them to the membrane organelles, including the synapses (Adereth et al., 2005; Wang E. T. et al., 2012). Of note, the nuclear Mbnl1 splicing regulation may also be associated with 3′-UTR binding, since spliced transcripts regulated by Mbnl1 have twofold chances of 3′-UTR binding by Mbnl1, suggesting a collaborative work between nuclear and cytoplasmic Mbnl1. The potential cytoplasmic function of Mbnl1 is intriguing especially in the asymmetrical cells such as neurons, since focal structures far away from the nucleus, such as axon and synapse may be more sensitive to RNA localization. The enriched Mbnl1 binding sites in the synapse suggest its potential role in regulating synaptic function (Wang E. T. et al., 2012).

It has been known that nuclear Mbnl1 and Mbnl2 autoregulate splicing of themselves, including exon 7 (54 nt) of Mbnl1 and exon 6 (54 nt) of Mbnl2 with sequence of nuclear localizing signals. This translocation from cytoplasm into nucleus is especially important during developmental transition from fetal to adult (Lin et al., 2006). In the Mbnl KO mice, the splicing shifted these alternative-spliced cassettes towards exon-inclusion, therefore the nuclear portion of Mbnl1 and Mbnl2 was increased (Lee et al., 2013). Interestingly, it was reported that an alternative post-translational modification pathway may determine the translocation of Mbnl1 from cytoplasm to nucleus by deubiquitination (Wang et al., 2018). In the transgenic mice overexpressing (CTG)₉₆₀ specifically in the brain, they found reduced cytoplasmic Mbnl1 is an early event and this cytoplasmic isoform is necessary for maintaining neurite growth. In addition, loss of function of cytoplasmic Mbnl1 showed negative impact on the synaptic transmission, prior to the nuclear Mbnl2-mediated mis-splicing of downstream targets. Therefore, the mechanism and impact of loss of Mbnl1 or Mbnl2 on the dendrite growth may be different (Wang et al., 2017, 2018). In our in vivo system, we chose CKO strategy in order to get DKO mice in the nervous system. Initially, we did not include Mbnl1 KO due to its subtle brain phenotype as we reported previously (Suenaga et al., 2012). Although Mbnl1 plays a predominant role in the cytoplasm and Mbnl2 functions more significantly in the nuclei, our observation on Nestin-Cre DKO nevertheless revealed a phenotype that contributed by compound loss of MBNL1/2 in the nervous system. To compare and verify specific consequences of loss of Mbnl1 or Mbnl2 to the dendrite/dendritic spines, further work on morphological analysis will be worth doing by using individual Mbnl KO lines.

In conclusion, through the morphological studies on Mbnl KO models, we once again proved that Mbnl1/2 proteins play a critical role in the developing brain through affecting the distribution patterns of cortical projection neurons and inhibitory interneurons as well as dendritic complexity, spine morphology and PSD ultrastructures. The Mbnl1/2 loss of function caused the infant/immature structural phenotypes in mutant brains that were reminiscent of the fetal isoform retention in adults during splicing analysis (Charizanis et al., 2012). These deficits may alter the synaptic transmission and brain circuitry in KO mice and potentially also in DM patients. Our findings provided the first morphological evidence of compromised neurons in Mbnl KO brains that will be very helpful in understanding CNS pathogenesis in DM.

All datasets generated for this study are included in the manuscript.

Mice were generated in the animal center of Chang Gung Memorial Hospital, Keelung, Taiwan, an AAALAC accredited institute. To maintain the highest standard of animal welfare, the IACUC committee ensured the number of mice used was reduced and the experimental procedures were refined (IACUC No. 2015121903).

H-CC and CS conducted the experiments and analyzed the data. K-YL and L-JL designed the experiments, analyzed the data and prepared the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.