Moving forward with the neuromuscular junction

Graphical abstract

The neuromuscular junction (NMJ) is indispensable for survival. This synapse between motoneurons and skeletal muscle fibers allows posture, movement and respiration. Therefore, its dysfunction creates pathologies than can be lethal. The molecular mechanisms of NMJ development and maintenance are the subject of intensive studies. This mini-review focuses on some of the most recent discoveries. An unexpected role for a protein, rapsyn, which has been known for 40 years to aggregate acetylcholine receptors (AChR) has emerged. A new cell partner at NMJ has been unmasked and is challenging our understanding of the functioning of this synapse. Toxins are now used as new tools to study degeneration/regeneration. The possibility of creating human NMJ in vitro is within reach with major consequences for drug screening. Wnts are secreted neurogenic factors that have been involved in vitro in AChR clustering, but their precise role in vivo remains to be clarified. All these data are raising new and exciting perspectives in the field and are discussed in this Review.

Graphical abstract

The best studied synapse for understanding mechanisms of synapse formation and maintenance is by far the neuromuscular junction (NMJ), a synapse between the motoneuron and the muscle. This is due to its large size and location outside of the brain, allowing a detailed analysis of cholinergic postsynaptic differentiation. During development, traits specific to this synapse are observed. First, the postsynaptic domain appears to differentiate before innervation in the center of the muscle, a process called prepatterning. The prepatterned domain is then recognized by the axon terminal, which upon contact with the muscle induces the pre- and postsynaptic elements to further differentiate to form a mature synapse. Signals originating first from the muscle then in addition from the nerve and the perisynaptic schwann cell orchestrate the formation of this synapse (Darabid et al., 2014). A number of organizing signals and their roles have been described in the past twenty years. However, a limited set of molecules play an instructive function and are indispensable for the formation of this synapse. Interestingly these factors play an anterograde as well as a retrograde role in this process though the mechanisms by which motoneuron secreted factors control the postsynaptic differentiation is by far the best characterized. Agrin is such a factor. The NMJ is absent in mice deficient for Agrin or its receptor complex (MuSK/LRP4) and intracellular adaptor proteins (Rapsyn, Dok-7). These molecules represent the core system for NMJ formation (Wu et al., 2010). A figure summarizes the essential steps and molecules in NMJ development. MuSK (Muscle-Specific Kinase) and its co-receptor LRP4, a member of the low-density lipoprotein receptor related protein family, are expressed by muscle cells and are responsible for the clustering of Acetylcholine receptors (AChR) even before innervation (DeChiara et al., 1996; Kim et al., 2008; Zhang et al., 2008). Beyond the clustering of AChR, they represent an essential hub from which successively the postsynaptic domain is built, the growth cone attracted, the presynaptic and postsynaptic elements instructed for some aspects of their differentiation and the synapse maintained (Shi et al., 2012). In mice, AChR aneural clusters induced by MuSK/LRP4 activation appear around E13.5 in a broad central region of the muscle during prepatterning (Lin et al., 2001; Yang et al., 2001;). Prepatterning has a transitory function and essentially serves to position the nerve on the synapse. AChR cluster size, number and shape are refined with innervation (E14) that disperses non-synaptic AChR clusters and induces new neural AChR synaptic clusters through the secretion of Agrin (Shi et al., 2012). Both prepatterned aneural and agrin-induced AChR cluster formation depend on the expression and activity of LRP4/MuSK. Agrin is a proteoglycan synthetized by the motoneuron and the muscle cell but only the nerve secreted isoform is active and induces AChR clustering. This molecule binds the N-terminal domain of LRP4 but agrin's postsynaptic effect is transduced through MuSK activation. Rapsyn is a membrane-bound cytoplasmic molecule that binds AChR and is necessary for their clustering. Dok-7 is a muscle-specific adapter protein that is recruited by MuSK once phosphorylated and induces a cascade of phosphorylation that leads to AChR clustering.

A remarkable trait of this developmental process is that the receptor MuSK/LRP4 is a multitask complex as described above. How this can be achieved by a single receptor complex? We believe that this is due to the existence of several ligands for the complex that are active at different stages of NMJ formation.

Very recently, unexpected results have challenged our knowledge of the NMJ, with the discovery of a new role for rapsyn, a new synaptic cell partner at this tripartite synapse, new approaches to decipher pathophysiological mechanisms in vitro in human neuromuscular junction, and new tools to study nerve regeneration. The complex world of Wnts and their roles during synapse formation are also being analyzed. Here we summarize the main points about these issues.

Rapsyn, a peripheral membrane protein, was first identified in postsynaptic membranes of Torpedo californica as a protein called 43kDa in reference to its molecular mass in the late 80's (Neubig et al., 1079; Sobel and Changeux, 1977; Frail et al., 1988). The co-localization of the 43kDa protein with acetylcholine receptors (AChR) in postsynaptic membrane and the interaction of this protein with the receptors lead to rename it rapsyn for receptor associated protein of the synapse. The function of rapsyn was first revealed in vitro when 43K, co-expressed with AChR in xenopus oocytes, was shown to induce the clustering of AChR (Froehner et al, 1990). The key role of rapsyn was further demonstrated in the mid 90's in vivo by a disruption of the gene in mice that resulted in a total absence of AChR clusters and abnormal nerve branching (Gautam et al, 1995). Rapsyn is then indispensable for postsynaptic differentiation, and the main role of rapsyn is thought to cluster AChR. What is the mechanism? The rapsyn-induced clustering results from the indirect activation of MuSK, a transmembrane receptor kinase. However, the mechanisms downstream of MuSK activation that induce rapsyn-AChR interaction are not known. AChR receptor interaction is mediated by the RING-H2 domain present in rapsyn and this domain also binds dystroglycan, a transmembrane protein connecting the extracellular matrix to the actin cytoskeleton (Ramarao and Cohen, 1998; Bartoli et al., 2001; Ramarao et al, 2001). In addition, rapsyn binds actin and actinin (Antolik et al., 2006; Dobbins et al., 2008). Interactions between these partners is regulated and suggest that rapsyn acts as a scaffold protein bridging AChR to the cytoskeleton and clustering them. Lin Mei's laboratory (Augusta University, USA) now reveals that rapsyn has an enzymatic activity and acts as a E3 ligase. This activity lies in the RING-H2 domain. A mutant that impairs the enzymatic activity of rapsyn inhibits AChR clustering in heterologous cells as well as in muscle cells. The ligase activity of rapsyn induces a posttranslational modification and neddylates AChR. In contrast, the activity mutation results in reduced neddylation. AChR neddylation is increased at NMJ and necessary for AChR clustering. Moreover, stimulation with agrin or overexpression of MuSK induces an increase in neddylated AChR in muscle cells. Mutation in the ring domain leads to the absence of AChR neddylation. Besides this new role of rapsyn, this is the first demonstration of this type of posttranslational AChR modification. The full story can be found in a recent issue of Neuron (Li et al., 2016).

A number of problems inherent to in vivo approaches and species differences are limiting the understanding of human synapse formation and functioning : (1) most often biopsies are not available; (2) model mice for a pathology often poorly mimic it; (3) transgenic mice have been engineered so that the causative gene is deleted rather than reproducing human mutations; (4) genome organization and RNA processing sometimes differ between mice and human; (5) in vivo animal models are complex to analyze since they do not allow the study of each partner cell in the synapse. The possibility of generating muscle and motoneurons from pluripotent stem cells and make NMJ in vitro is now emerging and is creating huge opportunities to study models of human pathological synapses and develop therapeutics based upon high throughput screening. The proof of concept demonstrating that hESC can be used for disease modeling has now been provided by the group of C Martinat (iSTEM). The pathology studied is Myotonic Dystrophy Type 1 (DM1), a multisystemic disease affecting the NMJ that is caused by expanded CTG repeats in the 3′UTR of the DMPK (Myotonin-Protein Kinase) gene. These abnormal repeats trap in the nucleus RNA binding proteins involved in mRNA processing leading to abnormal splicing pattern and translation for a subset of RNA. There is no curative treatment for DM1. In a recent study, Martinat and collaborators have generated mesenchymal progenitor cells and neurons from human embryonic stem cells (hESC) of MDT1 patients. They have analyzed the splicing pattern of the insulin receptor (INSR) pre-mRNA, one of the target of DMPK protein. In these hESC derived cells, like in primary human myoblasts from DM1 patients, INSR RNA splicing is mysregulated with a defect in exon 11 inclusion. Because patients with diabetes type 2 associated to DM1 are treated with Metformin, this drug was tested for its efficacy on improved inclusion of exon 11. Metformin treatment of hESC derived cells as well as primary myoblasts from patients resulted in a significant increase in exon 11 inclusion (Laustriat et al., 2015). This successful story resulted into a drug clinical trial called MYOMET which is now in phase II to evaluate the potential of metformin in DM1 patients. In parallel with this study, the group of C Martinat has been able to generate different subtypes of motoneurons and obtain fully differentiated spinal motoneurons from hESC (Maury et al., 2015). With the recent option of obtaining muscle cells from hESC, several groups have now in hand a toolbox for studying neuromuscular disorders and motoneuron diseases such as SMA, SLA and Congenital Myasthenic Syndromes using co-cultures of human motoneurons and muscle cells. However, human iPSC derived schwann cells remain to be generated to gather the essential partners of the NMJ and complete this model synapse.

Until now, the NMJ was described as a tripartite synapse involving glial cells (terminal Schwann cell), a motoneuron and a muscle fiber. We now have to add another partner that has been unmasked by the group of R. Rudolf (Heidelberg, Germany). Cathecholamines released by the sympathetic nervous system have been shown to increase the activity of the Na+/K+ pump, and the release of Ca from the sarcoplasm, and also to promote the degradation of glycogen and control NMJ activity. However, there was no evidence that any of these effects or others on the NMJ were direct. Electron microscopy performed by the group of R. Rudolf shows that sympathetic neurons ramify around blood vessels and motoneurons, and they are present in the vicinity of the NMJ. β-2 adrenergic receptors are enriched at the NMJ where they partially co-localized with α-bungarotoxin labeled AChRs. Local ablation of sympathetic innervation leads to abnormal NMJ phenotypes and functioning, but agonists of β2 adrenergic receptors rescue these effects. One of the major pathways induced by the activation of β2 adrenergic receptors is an increase in cAMP. Levels of NMJ cAMP are quantified using a biosensor RAPSN-EPAC and FRET. RAPSN-EPAC is a construct in which rapsyn (RAPSN) is followed by the cAMP–binding domains of the Exchange protein directly activated by cAMP (EPAC). Using this technology and upon stimulation of the sympathetic chain, in vivo imaging shows an increase of cAMP at the NMJ. The increase in cAMP stimulates in turn the expression of AChR, Acetylcholinesterase and Utrophin. Interestingly, the upregulation is paralleled by an increase in the expression of peroxysome proliferator-activated receptor γ–coactivator 1α (PPARGC1A), a factor that controls NMJ postsynaptic gene expression (Handschin, C. et al., 2007). These results, which have been published in PNAS (Khan et al., 2016), are of special interest in the context of Congenital Myasthenic Syndromes (CMS) treatments. CMS are inherited neuromuscular disorders caused by mutations in genes coding synaptic proteins and caracterized by muscle weakness. Currently the most efficient treatment for a subset of CMS resulting from mutations in COLQ, MuSK and DOK7 is salbutamol, an agonist of β-2 adrenergic receptors. It should also be emphasized that salbutamol is increasingly prescribed in other CMSs with therapeutic success. So far, the mechanisms of action of this compound are unknown. With their results, R. Rudolf's group is now paving the road toward further studies.

Whereas a good deal of information is now available about the mechanisms of synapse development, the molecular basis and the sequence of events that underlie nerve terminal regeneration at the NMJ remain to be understood. Useful tools for studying these mechanisms include the presynaptic neurotoxin α-latrotoxin from black widow spider venom and anti-polysialoganglioside antibodies plus complement complex. Both agents induce a localized and reversible motor axon terminal degeneration, and their action mimics the cascade of events that leads to nerve terminal degeneration in injured patients and in other neurodegenerative conditions (Duregotti et al., 2015; Rodella et al., 2016). Anti-polysialoganglioside autoimmune antibodies are found in patients affected by a group of immune-mediated peripheral-nerve disorders collectively named Guillain-Barré syndrome (GBS). In most cases patients experienced preceding airway or gastrointestinal infections; this observation led to the hypothesis that a molecular mimicry between human gangliosides and bacterial LPS is the mechanism that triggers the autoimmunity (Yuki and Hartung, 2012). The syndrome is commonly characterized by a progressive acute flaccid areflexic (or hyporeflexic) paralysis, usually starting from distal limbs and ascending to proximal ones. GBS subtypes can have highly different prognoses, from spontaneous complete recovery to a poorer outcome. The Rigoni group has shown that in response to α-latrotoxin or to anti-GQ1b plus complement (the anti-polysialoganglioside antibody most frequently associated with a GBS subtype named Miller Fisher syndrome), primary neurons release ATP as an early sign of the process of degeneration. In turn, ATP contributes to the activation within Schwann cells in co-cultures with degenerating neurons of a number of intracellular pathways of crucial importance for regeneration to occur, such as Ca²⁺, cAMP and MAP kinases (Negro et al., 2016; Rodella et al., 2017). It is likely that the mechanisms described in this context could be extended to other forms of peripheral neuropathies. In addition, understanding the molecular basis of regeneration will highlight common and different pathways between this process and developmental mechanisms.

Once the motoneuron has contacted the muscle postsynaptic domain, agrin is released and stimulates AChR cluster formation through the activation of MuSK. But prior to that step, what determines synapse position? How does the axon terminal find the prepatterned domain in the center of the muscle? Although we are still far from fully answering this question, we now have a few hints about this issue. We know that axon attraction to the prepatterned zone and AChR aneural clustering are MuSK-dependent (Flanagan-Steet et al., 2005; Panzer et al., 2006). Since the nerve is not yet present, what is the mechanism of MuSK activation during this prepatterning step? One possibility is that MuSK is highly expressed in a discrete domain and is autophosphorylated. The other hypothesis, which is not exclusive from the first one, is that another ligand of MuSK is activating MuSK phosphorylation. In the context of this second hypothesis, it is interesting to note that MuSK contains a cysteine-rich motif (CRD) in its extracellular domain. This motif is homologous to the one found in the Wnt canonical receptors Frizzled. However, there are 19 Wnts in vertebrates, and the challenge was to identify those which would have a role on MuSK induced AChR clustering. An extensive analysis of Wnts on AChR clustering was performed by the L Mei group and showed that a number of Wnts bind MuSK CRD and induce AChR clustering on C2C12 muscle cells in vitro (Zhang et al., 2012; Barik et al., 2014). The Granato group provided evidence in zebrafish that Wnt11 interacts with the CRD and phosphorylates the receptor, stimulates the formation of aneural AChR clusters, and guides motor axon to the muscle central region (Jing et al., 2009). C Legay's group showed that in mammals Wnt4 plays a similar although more discrete role compared to Wnt11 in zebrafish (Strochlic et al., 2012). However, the phenotype observed in Wnt11 knock-down zebrafish seems more severe than the Wnt4 null phenotype in mammals. The differential effect of Wnt4 and Wnt11 was confirmed in zebrafish where Wnt4a morpholino had little effect on AChR prepatterning in zebrafish, but the injection of Wnt4a morpholino in Wnt11 knockdown embryos completely abolishes the presence of prepatterned AChR clusters (Gordon et al., 2012). To test the consequence of a complete absence of Wnt binding on MuSK, the MuSK CRD was removed in knock-out mice (MuSKΔCRD). Controversial results were obtained by the C Legay and S Burden groups. The first group observed profound defects of the NMJ presynaptically and postsynaptically that were still present in adult muscle, whereas the second group did not notice any clear changes in NMJ phenotype although non-innervated ectopic muscle islands were detected in the diaphragm (Messeant et al., 2015; Remedio et al., 2016). One of the reasons for this discrepancy could lie in the relatively high level of MuSKΔCRD expression at the membrane compared to WT MuSK. MuSKΔCRD mRNA are not upregulated and this membrane overexpression could be due to the absence of Wnt binding since Wnts induce endocytosis in their receptors. In this context, an explanation could be that the overexpression of the mutated protein but not the mutation itself would lead to the observed defects. Genetic evidence does not support this hypothesis, indicating that overexpression of MuSKΔCRD is not responsible for the phenotype (L Strochlic, unpublished results). Besides, the different genetic backgrounds of the mice used by Remedio et al. (2016) and Messéant et al. (2015) could account for the differences in phenotypes. Indeed, a detailed analysis of the genotype-phenotype relationships recently showed that opposite phenotypes can be found depending on the genetic background (Sittig et al., 2016). The role of specific Wnts in the different steps of NMJ formation and their structural as well as functional relationships with the other ligands of the same complex MuSK/LRP4 remain to be clarified.

Abbreviations

NMJ

neuromuscular junction

AChR

acetylcholine receptor

MuSK

muscle-specific kinase

LRP4

low-density lipoprotein receptor-related protein 4

DoK7

docking protein 7

DM1

myotonic dystrophy type 1

DMPK

myotonic protein kinase

CMS

Congenital Myasthenic Syndromes

hESC

human embryonic stem cells

COLQ

Collagen like tail subunit of asymmetric acetylcholinesterase