Current progress in the creation, characterization, and application of human stem cell-derived in vitro neuromuscular junction models

Abstract

Human pluripotent stem cells (PSCs) such as embryonic stem cells (ESCs) or induced pluripotent stem cells (iPSCs) are of great value for studying developmental processes, disease modeling, and drug testing. One area in which the use of human PSCs has become of great interest in recent years is for in vitro models of the neuromuscular junction (NMJ). The NMJ is a synapse at which a motor neuron releases acetylcholine to bind to skeletal muscle and stimulate contraction. Degeneration of the NMJ and subsequent loss of muscle function is a common feature of many neuromuscular diseases such as myasthenia gravis, spinal muscular atrophy, and amyotrophic lateral sclerosis. In order to develop new therapies for patients with neuromuscular diseases, it is essential to understand mechanisms taking place at the NMJ. However, we have limited ability to study the NMJ in living human patients, and animal models are limited by physiological relevance. Therefore, an in vitro model of the NMJ consisting of human cells is of great value. The use of stem cells for in vitro NMJ models is still in progress and requires further optimization in order to yield reliable, reproducible results. The objective of this review is 1) to outline the current progress towards fully PSC-derived in vitro co-culture models of the human NMJ and 2) to discuss future directions and challenges that must be overcome in order to create reproducible fully PSC-derived models that can be used for developmental studies, disease modeling, and drug testing.

Introduction

The neuromuscular junction (NMJ) is a cellular synapse between a motor neuron and a skeletal muscle fiber that translates electrical and chemical cues into physical activity (1). The combination of a motor neuron and the muscle fibers that the motor neuron innervates are defined as a motor unit (1). Acetylcholine is released from the motor neuron into the synaptic cleft to attach to acetylcholine receptors (AChRs) on the postsynaptic muscle fibers. AChRs are ligand-gated cation channels, so the binding of acetylcholine results in an influx of Na⁺ ions and efflux of K⁺ ions causing depolarization of the myofibers. This opens up the voltage-gated calcium channels in the sarcoplasmic reticulum, releasing Ca²⁺ ions and facilitating muscle contraction (2).

Degeneration of the NMJ is a specific pathological feature of many neuromuscular diseases including amyotrophic lateral sclerosis (ALS), spinal muscular atrophy (SMA), myasthenia gravis (MG), and muscular dystrophy (3). There is variation among neuromuscular diseases as far as the primary site of pathology. For example, some diseases such as ALS and SMA have prominent motor neuron pathology, while others such as MG directly affect the NMJ. MG patients have autoantibodies that detect acetylcholine receptors or other critical post-synaptic receptors on skeletal muscle (4). On the other hand, inclusion body myositis, congenital myopathy, and muscular dystrophy primarily influence skeletal muscle (3). Whether the primary pathology resides in motor neurons, skeletal muscle, or other cell types, many neuromuscular diseases have common ground in the loss of function of the NMJ leading to motor deficits. In fact, NMJ pathology is often an early feature of neuromuscular diseases, which has led to a dying back theory in which NMJ degeneration causes axonopathy and ultimately motor neuron cell death (5,6). Therefore, targeting the NMJ therapeutically is of great interest.

The cellular and molecular mechanisms of the human NMJ have not been thoroughly studied in vivo due to the limited accessibility of tissues. Animal models are useful for studying certain aspects of NMJ development and disease processes, however, there are limitations to the translational relevance of these models (7). For example, human NMJs are smaller and more fragmented with coin-shaped endplates compared to the large, pretzel-shaped mouse NMJ. Human motor neurons also release less acetylcholine per action potential, known as the quantal content. In order to compensate for the decreased quantal content, the human NMJ has increased folding of the postsynaptic muscle membrane which helps to amplify the signal (6). While many animal model NMJs change shape and remodel with age, the structure of the human NMJ is mostly conserved. Analysis of mouse and human NMJ proteomes shows significant differences on a molecular level as well (8). Consequently, in vitro models specific to the human NMJ are extremely useful tools to understand the mechanisms of NMJ formation in healthy development as well as degeneration in disease.

Pluripotent stem cells (PSCs) such as embryonic stem cells (ESCs) or induced pluripotent stem cells (iPSCs) have the capability to self-renew and differentiate into many cell types (9). Human ESCs and iPSCs allow for the study of human development and disease processes in an in vitro setting. In particular, iPSCs are generated from adult somatic cells that have been reprogrammed to a pluripotent state by expression of specific transcription factors (10,11). Since these cells are derived from patient cells, they retain their disease-causing mutations which allows for in vitro modeling of how that mutation affects a specific cell or tissue type. They also present exciting possibilities for in vitro drug testing in a patient-specific manner. The use of CRISPR/Cas9 and other genome editing techniques allow for the creation of iPSC lines isogenic to the patient lines for confirming genotype-phenotype correlations (12).

Co-culture systems of motor neurons and skeletal myocytes can be refined to develop NMJ models in vitro that are simplified and easier to manipulate than in vivo models (13). The use of stem cells for in vitro NMJ models is still in progress and current culture systems are undergoing improvements in order to optimize the reliability and reproducibility of the results. In this review, we will outline the current progress towards in vitro human cell models of the NMJ with an emphasis on PSC models. We will also discuss future directions and challenges to creating reproducible models that can be used for drug testing and disease modeling.

Existing in vitro NMJ models: motor neuron and skeletal myocyte co-cultures

There are many factors to consider when designing an in vitro NMJ co-culture model, including cell source, culture format, how the NMJ will be characterized, and what application it will be used for (Fig. 1). To start, there have been many different combinations of cell origins used for co-cultures of motor neurons and skeletal myocytes (14,15). Early models relied on primary cells and tissue explants from animals. Eventually, cross-species co-culture systems were developed that combined human and rodent cells. Heterologous co-cultures using cells derived from different species have been beneficial for determining the contributions of each cell type to signaling components of the synapse (16). For example, one study used species-specific antibodies to stain a co-culture of human muscle cells and embryonic rat spinal cord explant and found that both skeletal muscle and motor neurons produce acetylcholinesterase at the NMJ (17). However, to study human physiology and disease, the use of human cells on both sides is necessary. Primary human cells and adult stem cells have been used, such as satellite cells to make muscle, or neural stem cells to make motor neurons (18). However, primary cells isolated from human patients often quickly lose their proliferative potential and can be difficult to genetically manipulate (19). In contrast, human pluripotent stem cells have unlimited proliferation and differentiation capacity and are more compliant to genetic modification. Several differentiation protocols for motor neurons and skeletal myocytes from PSCs have been developed in recent years, allowing for their increased use (20,21). Table 1 lists current in vitro NMJ co-culture studies in which one or more cell type is derived from human PSCs.

Figure 1. Patient-derived stem cells for in vitro co-culture neuromuscular junction (NMJ) models.

A general overview and summary of the factors to be considered when designing an in vitro co-culture to study the human NMJ using PSCs. It should be noted that these are not a definitive list; additional considerations or techniques may be used.

Table 1. Summary of studies using human pluripotent stem cells for in vitro models of the NMJ.

Only studies with one or both cell types derived from human embryonic or induced pluripotent stem cells were considered, not studies using adult stem cell-derived models. The characterization/analysis column describes what assays they used to confirm NMJ structure and functionality.

Study	Motor neuron source	Skeletal muscle source	Characterization/ Analysis	Disease modeling
Marteyn et al. 2011	hESC-derived	Human Mu2bR3 cells	Immunocytochemistry	Myotonic dystrophy type I (hESC-derived motor neurons)
Corti et al. 2012	hiPSC-derived	Human myoblasts	Immunocytochemistry	SMA (hiPSC-derived motor neurons from SMA patients and genetically corrected controls)
Demestre et al. 2015	hiPSC-derived	hiPSC-derived	Immunocytochemistry	N/A
Maury et al. 2015	hiPSCs and hESCs-derived	Human myoblasts	Immunocytochemistry	N/A
Shimojo et al. 2015	hESC-derived	Human myoblasts	Immunocytochemistry	N/A
Yoshida et al. 2015	hiPSC-derived	Mouse C2C12 cells	Immunocytochemistry	SMA (hiPSC-derived motor neurons from SMA patients and controls)
Steinbeck et al. 2016	hESC-derived, transduced with Channelrhodopsin2	Human primary myoblasts from adult and fetal donors	Immunocytochemistry. Optogenetic stimulation, blocked with vercuronium. Calcium imaging and microelectrode recordings	Myasthenia gravis (Added MG patient IgG and complement)
Maffioletti et al. 2018	hiPSC-derived	hiPSC-derived	Immunocytochemistry	3D muscle constructs from hiPSCs of patients with Duchenne, limb-girdle type 2D, and LMNA-related muscular dystrophies and healthy donors (but control lines only for co-cultures)
Osaki et al. 2018	hiPSC- and ESC-derived, transfected with channelrhodopsin-2	hiPSC-derived skeletal myoblasts in 3D collagen/Matrigel mixture	Immunocytochemistry. Glutamic acid, electrical, and light stimulation. Contractions blocked with BTX, visualized with calcium imaging, and quantified by pillar deflection.	Two models of ALS: 1) Excess glutamic acid; 2) motor neurons derived from sporadic ALS patient iPSCs
Santhanam et al. 2018	Human spinal cord stem cell or iPSC-derived	Human skeletal muscle myoblasts	Immunocytochemistry, electrical stimulation and treatment with NMJ toxins	N/A
Bakooshli et al. 2019	hESC-derived	Primary myogenic progenitors from patient biopsies	Immunocytochemistry. Glutamate-induced contractions viewed by calcium imaging (cells were transduced with GCaMP6 calcium reporter) recorded with electrophysiology and inhibited by BOTOX and d-tubocurarine	Myasthenia gravis (treated co-culture with IgG from MG patients) and congenital myasthenic syndromes (Waglerin-1 treatment to block the AChR epsilon subunit)
Lin et al. 2019	hiPSC and hESC-derived, with channelrhodopsin	hiPSC-derived	Immunocytochemistry and electron microscopy. Contractions stimulated by light, viewed by calcium imaging (Fluo-8), blocked with dantrolene or curare	SMA (used shRNA to knock down SMN)
Picchiarelli et al. 2019	hiPSC-derived	hiPSC-derived	Immunocytochemistry	Amyotrophic lateral sclerosis (mutations in FUS)
Vila et al. 2019	Human primary muscle cells reprogrammed to iPSCs and transfected with channelrhodopsin-2	Human skeletal muscle stem cells	Immunocytochemistry. Electrical and light stimulation of contractions, blocked with BTX treatment	Myasthenia gravis (added serum from myasthenia gravis patients)
Guo et al. 2020	hiPSC-derived	hiPSC-derived	Immunocytochemistry and electrophysiology.	N/A
Martins et al. 2020	hESC and hiPSC-derived axial stem cells	hESC and hiPSC-derived axial stem cells	Immunocytochemistry and electron microscopy. Measured spontaneous and glutamate-stimulated contractions with MEA and visualized with calcium imaging. Blocked with curare.	Myasthenia gravis (treated with MG patient serum)
Mazaleyrat et al. 2020	hiPSC-derived codifferentiation	hiPSC-derived codifferentiation	Immunocytochemistry, electron microscopy, RNA sequencing, calcium imaging, contraction stimulation with glutamate and inhibition with BTX and tetrodotoxin. Also drug tested muscle relaxants.	Duchenne muscular dystrophy, facio scapulo humeral dystrophy, myotonic dystrophy, limb girdle muscular dystrophy type 2A

Abbreviations: hESC, human embryonic stem cell; hiPSC, human induced pluripotent stem cell; SMA, spinal muscular atrophy; N/A, not applicable; MG, myasthenia gravis; LMNA, lamin A/C; BTX, bungarotoxin; ALS, amyotrophic lateral sclerosis; NMJ, neuromuscular junction; IgG, immunoglobulin; AchR, acetylcholine receptor; shRNA, short hairpin ribonucleic acid; SMN, survival motor neuron; MEA, multielectrode array.

There has also been variation in how the cell types are cultured together (13). Historically, the most commonly used method is a simple co-culture in which myocytes and motor neurons are plated directly on top of or adjacent to each other and the motor neurons extend axons to form functional NMJs with the myocytes. This method has been widely utilized for its simplicity and ease of use, but it has some drawbacks. First, being 2-dimensional (2D) in nature it lacks the 3-dimensional (3D) cues that would be present in vivo. This includes signaling from other cell types as well as interactions with the extracellular matrix (22). In addition, the unorganized mixture of cells in the direct co-culture model can make it difficult to discern individual cell types and observe axon outgrowth. However, this method does not require advanced manufacturing of a culture substrate, is sufficient to induce the formation of functional neuromuscular junctions, and allows for relatively simple analysis.

3D co-culture models have the potential to create a more physiologically relevant NMJ in vitro. These models often include extracellular matrix components which contribute to the development and maintenance of NMJ structure and function in vivo. For example, proteins in the basal lamina help guide axons to the proper site during development and post-injury. The basal lamina also facilitates signaling between the axon and the muscle fiber by allowing diffusion of acetylcholine and by acting as an anchor for acetylcholinesterase which degrades excess acetylcholine (23). While there has been great progress in 3D in vitro skeletal muscle models (24), their innervation is a more recent development (25–36). 3D NMJ models allow for a more physiologic model of muscle contractile function. 3D skeletal muscle models often include anchor points on either end of the construct, similar to how muscle is anchored by tendons and bone in vivo. This provides tension and prevents the cells from detaching upon contraction. A popular method is to create a 3D muscle construct by combining cells within a hydrogel that is connected to two flexible posts. The post deflection during contraction of the muscle construct is used to measure contraction force (25,28,29,31,32,34–37). Alternatively, strips of velcro can be used to anchor either end of the construct (33,38). Some potential drawbacks to 3D in vitro NMJ models include difficulties in reproducibility and scaling up for high throughput analysis. Additionally, some analysis methods such as imaging are more difficult with 3D models. Therefore 2D models may still be necessary in some cases such as for single fiber analyses.

A microfluidic co-culture system can be a compromise between 2D and 3D culture options by culturing cells in a more physiologically relevant environment than 2D cultures but on a smaller, more precisely defined scale than some 3D models. Microfluidic culture systems are devices in which only small amounts (on the μl, nl, or pl scale) of liquid are needed in order to culture cells in a defined space (39). In current microfluidic NMJ models, the myotubes are plated in one compartment, the motor neurons in another, and there are small channels between the compartments through which the motor neuron axons can extend through to form NMJs (25,34,40–46). By having the cell types in individual compartments, you can establish separate unique microenvironments that are more relevant to what the cells would encounter in vivo. This also allows for the study of localized treatments. For example, a drug or growth factor can be added either to the neuronal compartment or the muscle/NMJ compartment to determine where it has a maximum effect (42). One study found that exposing the muscle/NMJ compartment to oxidative stress resulted in more cytotoxicity than when the motor neuron soma compartment was treated. The same study also found that glial cell line-derived neurotrophic factor (GDNF) treatment of the NMJ compartment but not the soma compartment facilitated axon growth and the formation of active NMJs (42). Microfluidic devices also allow for the creation of a chemotactic gradient or a fluid gradient. This can be achieved by adding more media to one side than the other to cause fluid to flow only in one direction towards the lower side (43). Overall, microfluidic devices allow for more precise manipulations of the culture microenvironment and can save on resources as they require fewer cells and less media. Microfluidic devices hold a lot of potential but are still relatively new and require a high level of optimization (39).

Characterization of in vitro neuromuscular junctions

Regardless of the type of culture platform, in vitro NMJ models must first be properly characterized before being used to study a developmental process or disease phenotype. Ideal characterization should include both structure and functionality (Fig. 2). Microscopy is often used to show the proximity of axons to the skeletal myocytes. In early studies this was done using phase contrast imaging, but it is now common to use immunostaining for specific components of the synapse. Markers to identify skeletal muscle include myosin heavy chain or sarcomeric ɑ-actinin. Motor neuronal markers commonly used are choline acetyltransferase (ChAT) (31,47–50), HB9 (34,42,47,48,50–52), class III β-tubullin (Tuj1) (18,25,29,31,49,51,53–56), Isl1 (51), MAP2 (41), and Neurofilament M and/or H (SMI-32) (25,32,33,41,47–49). Finally, AChRs are identified using fluorescent-tagged α-bungarotoxin (BTX). An overlap of AChRs with the neuronal marker is commonly shown as proof of NMJ formation. Some papers go even further in depth characterizing various aspects of pre- and post-synaptic terminals by staining for muscle contractile components (dihydropyridine receptor, ryanodine receptor, titin), pre-synaptic components (Syne-1, Synapsin I, synaptotagmin, Bassoon), and proteins important for NMJ stability and maintenance [Muscle specific tyrosine kinase (MuSK), Rapsyn] (7,32,54). The maturation level of AChRs can be tested by monitoring expression of the embryonic gamma subunit and the adult epsilon subunit of the AChRs (7,33,50). In addition to immunofluorescence, imaging techniques such as electron microscopy and stimulated emission depletion (STED) microscopy are also used to study the precise structure of the NMJ (57).

Figure 2. Characterization of in vitro NMJs on a structural and functional level.

The left side of the diagram lists key structural components of the NMJ that are often used to confirm proper differentiation of the source cells into motor neurons and skeletal myocytes, as well as the formation of an NMJ. The illustration on the right depicts aspects of determining the functionality of an in vitro NMJ. At a fully functioning NMJ, acetylcholine is released from the motor neuron and crosses the synaptic cleft to bind to acetylcholine receptors on the skeletal muscle membrane. This leads to a sodium ion influx into the muscle, calcium release from the sarcoplasmic reticulum, and contraction of the sarcomeres. Contractions may be stimulated through methods such as patch-clamp electrophysiology or modifying motor neurons to express channelrhodopsin (ChR2) and therefore create an action potential from light stimulation. There are also drugs and growth factors that can be added to either stimulate (e.g. glutamate or acetylcholine) or block (e.g. Botulinum toxin A, α-bungarotoxin, tubocurarine, or rocuronium) contractions.

While an NMJ may appear to be fully formed through microscopy, functional tests are needed to confirm that a legitimate synapse has formed. In order to prove functionality of the NMJ, there should be confirmation of muscle contraction induced by motor neurons which can be demonstrated in a variety of ways. Glutamate can be added to the culture to stimulate motor neurons and observe resulting muscle contraction. Many studies combine this with the use of a substance that either blocks acetylcholine from being released (such as botulinum toxin A (58)) or from binding to the AChRs on muscle (such as α-bungarotoxin, rocuronium, or tubocurarine (7,16)). If the contractions cease, this further confirms that the earlier glutamate-stimulated contractions were due to motor neuron input and not just spontaneous muscle contractions. Tetrodotoxin can also be used to inhibit spontaneous muscle contractions by blocking voltage-gated sodium channels (16). Alternatively, traditional electrophysiological methods can be used to stimulate motor neurons and record skeletal muscle action potentials. However, this technique requires a lot of skill and training (29). A newer electrophysiological technique is the multielectrode array (MEA). The MEA can stimulate and record action potentials from a population of cells plated onto a surface embedded with electrodes (59). There are limitations to this method, in that when using a mixed population of cells such as motor neurons and myocytes, it would be difficult to distinguish what cell type the action potential is being recorded from unless the boundaries between cell types are clearly defined. Another method used to stimulate contractions is to genetically modify motor neurons to express channelrhodopsin-2, a light-sensitive ion channel that will trigger an action potential in response to light stimulation, causing the skeletal myocytes to contract (25,33–35,37,50).

There are also several different methods used to record and analyze muscle contractions. Visualization of intracellular calcium flux is commonly used by adding calcium indicators such as Fluo-3 AM (42,43), Fluo-4 AM (32,55), Fluo-8 AM (25,49), or fura-2 (50) directly to the cells or genetically modifying the cells with a calcium indicator protein GCamp6 (33). Software analysis of brightfield videos can be used to quantify contractions by analyzing pixel displacement (28,37). Several studies have used the deflection of flexible posts or cantilevers to quantify muscle contraction (25,28,29,31,32,34–37,60).

Combinations of the methods above can be used to develop more advanced ways to analyze functional data. One group using an optogenetic co-culture model developed a system which can control light stimulation and record and analyze the resulting contractions concurrently (37). This system uses pulses of controlled length and frequency to stimulate contractions, and includes an image processing algorithm which can differentiate between spontaneous or triggered muscle contractions. Systems such as these are helpful to increase reproducibility and decrease user bias (37).

Applications of human iPSC-based in vitro NMJ models

Developmental studies

An in vitro NMJ model using human stem cell-derived motor neurons and skeletal myocytes can allow us to better understand the complex signaling that occurs during development such as what guides the axons to a specific innervation site, what drives competition between nearby axons, how the cells transition from polyinnervation to a single axon per myofiber, and how these connections are matured (1). Understanding these processes could help to develop methods to encourage regeneration in cases of injury or disease. Studies that simultaneously differentiate both skeletal myocytes and motor neurons from the same PSC source could be more developmentally relevant, as signaling between skeletal muscle and motor neurons is important for NMJ development in vivo (49,54,61). Neuromuscular organoids in which human PSCs are simultaneously directed down multiple lineages could be especially useful to characterize human developmental processes. One such study differentiated human PSC lines into neuromesodermal progenitors and then grew them in suspension to form self-organizing trunk neuromuscular organoids. The organoids contained functional NMJs and several cell populations which were characterized as neural, mesodermal, endothelial, epithelial, and sclerotome (54).

Disease modeling

Disease modeling is another major application for patient-oriented iPSC models of the NMJ. There have been many studies using cells or tissue explants from animal models or human patient primary cells for disease modeling (14), but so far there have only been a handful of co-culture studies using motor neurons or skeletal myocytes derived from human stem cells (see “Disease modeling” column of Table 1). A commonly used proof of concept that an in vitro NMJ model can be used for disease modeling has been to add MG patient IgG with human serum containing active complement to the culture to model MG pathology. Studies which have used this technique have found decreased contractile activity following treatment with MG patient IgG (33,37,50,54). Some also showed a recovery of contractile function after a return to regular media, representative of patient plasmapheresis (37,50).

Characterization of NMJs formed from iPSCs containing specific disease mutations or backgrounds is of great interest to the field but has been less commonly achieved. A co-culture using healthy human myocytes with motor neurons differentiated from myotonic dystrophy type I (DM1) ESCs found increased neurite outgrowth and decreased NMJ formation compared to healthy control lines (51). Two separate studies cultured SMA patient iPSC-derived motor neurons with either human (48) or mouse (47) myocytes. Both found fewer and smaller AChR clusters in the SMA co-cultures compared to controls. Genetic modification of the patient cell lines to promote SMN exon 7 retention and therefore increase the amount of SMN protein was able to reverse the AChR clustering phenotype (48,62). Another group used a different approach to study SMA, by using shRNA to knock down SMN in healthy iPSCs (49). The cultures with SMN knockdown formed fewer and smaller NMJs than the isogenic control. The SMN knockdown co-cultures also had decreased contraction synchronicity and speed, as well as changes in mitochondrial morphology (49). One group used a protocol for simultaneous differentiation of myocytes and motor neurons from patient iPSC lines with backgrounds of Duchenne muscular dystrophy (DMD), myotonic dystrophy (DM1), limb girdle muscular dystrophy type 2A (LGMD2A), and facio scapulo humeral dystrophy (FSHD2) (61). They found structural abnormalities including reduced fiber size in DMD cells, increased fiber size in FSHD2 cells, and sarcolemmal disorganization in all, including vacuoles in DMD and electron dense inclusions in LGMD2A. Functionally, they found changes in intracellular calcium signaling, with an increase in calcium handling amplitude in the DMD cultures and an increase in the others (61).

Combinations of healthy and diseased cells can be used to determine the specific influence of each cell type towards NMJ instability in the disease state. For example, co-cultures of mouse primary motor neurons and skeletal muscle cells with or without the SOD1^G93A mutation were used to investigate the relative contributions of skeletal muscle and motor neurons to NMJ pathology in ALS. This study found that in vitro NMJ formation was hindered when the mutation was present in both cell types. If the mutation was only present only in skeletal muscle, functional NMJs still formed but the contraction frequency was decreased (63). A similar experimental method using all human iPSC-derived cell types examined the effects of mutations in FUS, a cause of familial ALS (64). Myotubes differentiated from FUS-ALS patient iPSCs had impaired endplate maturation as well as decreased expression of the AChR alpha 1 subunit compared to healthy and isogenic control lines. Co-culture of healthy myotubes with FUS-ALS motor neurons also caused a decrease in mature endplates. The phenotype was worsened when both cell types were derived from FUS-ALS iPSCs. Together this suggests that both myotubes and motor neurons contribute to NMJ pathology in ALS (64). Studies such as these could be used to reveal the relative contributions of certain cell types to NMJ degeneration in a variety of neuromuscular diseases.

Drug screening

In vitro NMJ models created from patient-derived stem cells allow an exciting opportunity to preclinically test drug efficacy or possible side effects in a patient-specific manner. While many groups have used iPSC-derived motor neuron cultures for drug testing [reviewed in Chang 2020 (65)], this has been achieved less commonly using in vitro iPSC-derived NMJ models as these are not often reproducible in a high throughput manner. However, one study using healthy human stem cell-derived motor neurons and human myoblasts showed the feasibility of drug testing with in vitro NMJ models by generating dose-response curves to drug treatments. The co-cultures were treated with varying dosages of bungarotoxin, Botox, and curare. Contraction videos were recorded and used to determine the frequency and amplitude of contractions, which were then used to generate a dose-response curve for each drug (44). Another study of iPSC-derived NMJ co-cultures were treated with muscle relaxant drugs Salbutamol, Carisoprodol, and Mexiletine. Calcium imaging of the treated co-cultures showed decreased contractions followed by a recovery (61). In addition, a microfluidic-based NMJ model with motor neurons derived from a sporadic ALS patient iPSC line was used to test drug candidates (25). Compared to the healthy control, the co-culture with ALS motor neurons had fewer muscle contractions, increased motor neuron degradation, and increased apoptosis in the skeletal myocytes. Treatment with rapamycin and bosutinib decreased apoptosis and improved muscle contraction force (25). To increase the physiological relevance of the model, the drugs were applied through an endothelial cell barrier, to mimic in vivo drug delivery obstacles such as the blood brain barrier. Studies in which NMJ models can reproducibly show a disease phenotype that can be rescued with drug treatment could become a useful pre-clinical method for drug screening. Having an all-human cell assay to test for drug efficacy or side effects would be beneficial for several reasons. First, it would be ethically advantageous by reducing the need for animal studies. Second, differences between laboratory animals and humans may be a contributing factor to the failure of many drugs in human trials that had succeeded in preclinical animal trials. Therefore, using an in vitro culture system to test drugs would hopefully reduce the frequency at which drugs fail in clinical trials.

In vivo implantation

Finally, 3D constructs of skeletal muscle innervation originally created in vitro can be used for transplantation purposes, such as grafts for healing volumetric muscle loss. Attempts to treat a rat model of tibialis anterior muscle defect injury with transplantation of a 3D skeletal muscle graft had previously been unsuccessful in restoring muscle function due to delayed innervation of the graft in vivo (53). However, grafts made with a mixture of muscle progenitor cells and neural stem cells led to an increase in cell survival and muscle differentiation within the graft. Eight weeks after transplantation, the combination graft showed a full restoration of muscle volume and muscle force, and integration with the host nerves (53).

Challenges and future directions for human stem cell-derived in vitro NMJ models

While there are exciting possibilities for the application of iPSC-derived NMJ models, in some cases additional optimization of the co-cultures or their analysis methods are warranted. As discussed above, characterizing in vitro NMJs can be complicated and variable between labs. Both structure and function should be well characterized and defined to increase reproducibility. There is also a need for increased throughput in order to utilize in vitro NMJ models for drug screening.

One potential obstacle to disease modeling is that often motor neurons and skeletal myocytes derived from human stem cells retain an embryonic phenotype. This may not be an issue for modeling early onset diseases such as SMA, but it could be a limiting factor in the ability of these cells to prove useful for in vitro NMJ models of late onset diseases such as ALS. Methods for promoting the maturation of these cell types may be necessary. However, it seems that merely culturing these cell types together promotes their mutual maturation. For example, when comparing rat muscle fibers cultured alone to muscle fibers cultured with motor neurons, the co-cultured muscle fibers had more frequent contractions with a higher force, and also trended towards increased expression of adult myosin heavy chain isoforms (27). In addition, co-culture of human PSC-derived motor neurons and skeletal myocytes promotes the expression of the more mature epsilon subunit of the AChR (33,49). One study which co-differentiated the myocytes and motor neurons together from the same cell source found increased expression of adult myosin heavy and light chain isoforms, and were even able to maintain the culture for up to a year (61).

Other challenges to using PSC-derived cultures include low differentiation efficiency which leads to contamination of other cell types. This is undesirable for studies that aim to dissect specific interactions between a motor neuron and myotube. However, in some cases the presence of other cell types may be desirable in order to create a more physiologically-relevant model. In the human body, there are many cell types that can influence motor neuron survival or signaling, either directly or indirectly. These include astrocytes (66), microglia (67), interneurons (68), Schwann cells (69,70), kranocytes (71), and myocytes (72). There are many other additional interactions that are key to skeletal muscle function in vivo and are beginning to be represented in in vitro models, including the myotendinous junction (73) and the vascularization of muscle tissue (74). These features may also need to be considered in creating a physiologically relevant model of a motor unit. One study in particular demonstrated vascularization of their NMJ model by including isogenic differentiation of motor neurons, skeletal myocytes, vascular endothelial cells, and pericytes (32). Some NMJ models include glial cells to support NMJ formation and maintenance (41,75). While there are existing protocols to differentiate Schwann cells from human PSCs (76,77), there are not currently any protocols that can specifically differentiate terminal Schwann cells of the NMJ. However, protocols which use a simultaneous co-differentiation of both myocytes and motor neurons from the same cell source have noted the presence of Schwann cells which may be functioning as terminal schwann cells (49,61).

Many of the in vitro co-culture techniques that have been developed for modeling the NMJ could be applied towards modeling other types of innervation such as sympathetic innervation of cardiac muscle (78), or different types of skeletal muscle innervation. Depending on progress made with the specificity of motor neuron and skeletal myocyte differentiation protocols, in vitro NMJ models could examine differences between NMJs of different motor neurons (such as alpha, beta, and gamma motor neurons) and fiber types (including fast-twitch fatigable, fast-twitch fatigue-resistant, and slow-twitch fatigue-resistant) as they show differences in vulnerability to aging and disease (79). One group developed a model of the proprioceptive neuromuscular reflex arc including sensory neurons and gamma motor neurons (80). The 3D human trunk neuromuscular organoid mentioned previously can also be used to study central pattern generator-like circuits, which are the neural networks of rhythmic locomotion (54,78). An interesting future application from studying the neuromuscular reflex arc in vitro is the development of bioactuators, or “soft robots” which can be used to study locomotion at a more systemic level. For example, one model has been used to study the development of a central neural network innervating multiple muscle tissues (36).

Conclusions

The use of human pluripotent stem cells for creating in vitro NMJ models holds great potential for examining the intricacies of the human NMJ during development, aging, injury, and disease. There has been a wide variety of culture techniques used to create in vitro NMJs, each with their own benefits and drawbacks. While many NMJ models are still in the development phase, the field is moving towards exciting applications including developmental studies, disease modeling, drug testing, transplantation, and even bioactuators/robotics. In all, the use of human PSC-derived cell types for in vitro NMJ models is a relatively recent development, but holds immense potential for studying neuromuscular development and disease.