Structure and Function of the Mammalian Neuromuscular Junction

Abstract

The mammalian neuromuscular junction (NMJ) comprises a presynaptic terminal, a postsynaptic receptor region on the muscle fiber (endplate), and the perisynaptic (terminal) Schwann cell. As with any synapse, the purpose of the NMJ is to transmit signals from the nervous system to muscle fibers. This neural control of muscle fibers is organized as motor units, which display distinct structural and functional phenotypes including differences in pre- and postsynaptic elements of NMJs. Motor units vary considerably in the frequency of their activation (both motor neuron discharge rate and duration/duty cycle), force generation, and susceptibility to fatigue. For earlier and more frequently recruited motor units, the structure and function of the activated NMJs must have high fidelity to ensure consistent activation and continued contractile response to sustain vital motor behaviors (e.g., breathing and postural balance). Similarly, for higher force less frequent behaviors (e.g., coughing and jumping), the structure and function of recruited NMJs must ensure short-term reliable activation but not activation sustained for a prolonged period in which fatigue may occur. The NMJ is highly plastic, changing structurally and functionally throughout the life span from embryonic development to old age. The NMJ also changes under pathological conditions including acute and chronic disease. Such neuroplasticity often varies across motor unit types.

Introduction

The structure and function of the neuromuscular junction (NMJ) provide insight into neural communication. The term “synapse” was introduced by Sir Charles Sherrington (1897) to describe the point of contact between a neuron and its target cell. For $\alpha$ motor neurons and their target cells (skeletal muscle fibers), the synapse is the NMJ. This article focuses primarily on the mammalian NMJ, but we include important original observations on the NMJ in frog muscle (e.g., the research of Bernard Katz from the 1950s to 1970s), as well as information that is only available based on research in other nonmammalian species (e.g., seminal works by Oppenheim and Landmesser on the development of the NMJ in chick hindlimb muscle).

Using a microscope, Robert Hooke was the first to observe a living cell in 1665, which he described in his book Micrographia (98). In 1839, the concept that living organisms (both animals and plants) are composed of cells (the Cell Theory) was independently proposed by the German botanist Matthias Schleiden (410) and the German physiologist Theodor Schwann (411). Due to limitations in microscopy, it was initially unclear if the cell theory could be applied to the nervous system. However, with the introduction of the silver impregnation technique by Camillo Golgi (Golgi stain) in 1873 (167), it became possible to visualize neurons as single cells. However, Golgi continued to promote the Reticular Theory that the nervous system is a continuous interconnected network, or a reticulum, originally proposed by Gerlach in 1871 (162, 441). Based on the Cell Theory, an opposing view had emerged in the 1880s that the nervous system comprised discrete individual cells. In support, the neuroanatomist Santiago Ramón y Cajal used the Golgi staining technique to provide exquisite drawings of neural cells (154). Based on the work of Golgi and Cajal, the German anatomist Heinrich von Waldeyer-Hartz introduced the term “neuron” in 1891 to describe neural cells. The Neuron Theory (alternatively Neuron Doctrine) had important implications with respect to communication within the nervous system. In the Neuron Doctrine, a neuron communicates via synapses with other cells, for example, at the NMJ. Curiously, both Cajal and Golgi won the 1906 Nobel Prize for their work, despite their opposing views on the elemental structure of the nervous system.

During this period, the English neurologist William Gowers coined the term motor neuron (171) to describe the neurons involved in direct motor control of skeletal muscle fibers. The endpoints of motor neuron axons innervating muscle fibers are termed presynaptic terminals. The presynaptic terminal is specialized for neurotransmitter release, and the postsynaptic receptor region responds to the released neurotransmitters via selective receptors. The specialized postsynaptic region on skeletal muscle fibers is called the endplate, a term introduced by the German anatomist Wilhelm Krause as “motorische endplatte” (231). Cholinergic receptors (AChRs) at the endplate of skeletal muscle fibers are ionotropic receptors, which directly gate a cation channel in response to the binding of acetylcholine (ACh). Thus, in mammals, ACh induces depolarization of the postsynaptic membrane when it binds to the AChR.

In 1925, Edward Liddell and Sir Charles Sherrington defined a motor unit as a single motor neuron and all the muscle fibers it innervates (259). Functional differences exist across motor unit types, and in this article, we highlight these differences as an organizing principle for NMJ physiology. Different motor unit types vary in their overall activity and their frequency of activation. There must be fidelity of neuromuscular transmission at the NMJ to ensure appropriate activation and mechanical responses by motor unit muscle fibers. For example, within the diaphragm muscle (DIAm), some motor units involved in breathing are very active with a duty cycle of 30% to 50%, depending on the species. In contrast, other motor units are recruited only infrequently for higher force expulsive behaviors such as sneezing, coughing, parturition, defecation, and micturition (138, 426).

Motor Units

Motor units are the final common pathway for the neural control of muscle force and contraction and therefore movement. Once excited, an action potential transmitted via neurotransmitters across the NMJ to produce an action potential in all the muscle fibers within the unit that are then activated in an all-or-none fashion. The mechanical properties of a motor unit depend on the collective mechanical properties of the muscle fibers. Accordingly, the motor unit force will depend on the average force generated per fiber and the number of muscle fibers within the motor unit (innervation ratio). The central nervous system controls force generation by motor unit recruitment and modulation of the frequency of motor neuron action potentials.

Motor unit recruitment

Motor unit recruitment (i.e., initiation of motor neuron axonal action potentials in response to sufficient excitatory input) is dependent on the intrinsic electrophysiological properties of motor neurons. Accordingly, smaller motor neurons comprising more fatigue-resistant motor units are recruited before larger motor neurons that comprise stronger but more fatigable motor units. In his membrane theory, Julius Bernstein (1902) applied thermodynamic principles underlying the Nernst equation to describe the resting membrane potential in nerve axons. In his theory, depolarization during an action potential reflected an abrupt decrease in membrane resistance (increase in membrane conductance). With identification of the structure of the bipolar lipid cell membrane, it was recognized that the dielectric properties of the cell membrane established the ability of the cell membrane to store charge along the intra- and extracellular surfaces of the lipid bilayer, thereby forming a capacitor (417). Accordingly, cell membrane capacitance depends on surface area of the cell. The membrane surface area also relates to input resistance $\left(R_{\text{in}}\right)$ such that smaller motor neurons have higher $R_{\text{in}}$. The relation between membrane potential $\left(V_{\mathrm{m}}\right),R_{\mathrm{i}\mathrm{n}}$ and input current $I_{\text{in}}$ is described by Ohm’s Law as $V_{\mathrm{m}}=\frac{I_{\text{in}}}{R_{\text{in}}}$. However, it must be noted that Ohm’s Law represents a static state, whereas the dynamic nature of motor neuron depolarization and recruitment requires a change in $V_{\mathrm{m}}$ over time $\left(\frac{\mathrm{d}{V}_{\mathrm{m}}}{\mathrm{d}t}\right)$, which represents a change in the charge stored on the membrane $\left(C_{\mathrm{m}}\right)$ as driven by synaptic current $\left[I_{\text{syn}}\right.$; Eq. (1)]. Smaller motor neurons have less surface area and thus a lower ${\mathrm{C}}_{\mathrm{m}}$. For a given amount of ${\mathrm{I}}_{\mathrm{s}\mathrm{y}\mathrm{n}}$, there is a greater $\frac{\mathrm{d}{V}_{\mathrm{m}}}{\mathrm{d}t}$ in smaller compared with larger motor neurons making them more excitable (126, 507, 508). Whether explained through intrinsic differences in $R_{\text{in}}$ or $C_{\mathrm{m}}$, smaller motor neurons have a greater change in $V_{\mathrm{m}}$ in response to a given $I_{\text{in}}$ (Figure 1). Gasser and Grundfest (1939) found that the size of a motor neuron is approximately proportional to the diameter of its motor axon and thus, action potential conduction velocity (155). Building on their work, Elwood Henneman introduced the size principle (1957), one of the leading concepts of motor unit physiology (193, 499). He observed that smaller motor neurons with slower conduction velocities were consistently recruited before larger motor neurons with faster conduction velocities. The size principle has been largely upheld under a variety of conditions with reversal of motor unit recruitment order occurring $\le 5\mathrm{\%}$ of the time. The occurrence of recruitment order reversal is more common during more dynamic tasks likely reflecting changes in the balance of excitatory and inhibitory synaptic input (67).

Figure 1.

For a given level of glutamatergic input current (Isyn), the change in membrane potential $\left(\mathrm{d}{V}_{\mathrm{m}}/\mathrm{d}t\right)$ will be greater for smaller motor neurons due to their lower membrane capacitance (size; $C_m$) and higher input resistance $\left(R_{\text{in}}\right)$ compared to larger motor neurons. Thus, smaller motor neurons will reach the threshold for action potential generation sooner (i.e., earlier recruitment) than larger motor neurons. Smaller motor neurons also have smaller axons with slower action potential propagation velocities compared to larger motor neurons. These relationships constitute the Henneman’s size principle for motor unit recruitment.

$$\frac{\text{d}{V}_{\text{m}}}{\text{d}t}=\frac{I_{\text{syn}}}{C_{\text{m}}}$$ (1)

Distribution of motor neuron size

Generally, motor neurons innervating a skeletal muscle range in size (e.g., surface area), which underlies the selective recruitment of motor units during different motor behaviors. For example, there is an approximately five-fold range in surface areas of phrenic motor neurons in rodents (55, 137, 279, 351, 355). The distribution of phrenic motor neuron somal and total surface areas is generally Gaussian (Figure 2). Assuming a uniform innervation ratio of approximately 100 DIAm fibers innervated per phrenic motor neuron (144), the phrenic motor neuron pool can be functionally segregated into tertiles corresponding to motor unit types with the upper tertile of somal volumes likely innervating more fatigable motor units (see below). Importantly, the morphology of the motor axon and presynaptic terminal at NMJs varies proportionately with motor neuron surface area.

Figure 2.

(A) Imaging of the phrenic motor neuron pool from rats acquired by confocal microscopy after labeling with tetramethylrhodaminedextran (nerve-dip) or cholera toxin B (CTB) injected into the intrapleural space. Reused, with permission, from Mantilla CB, et al., 2009/ELSEVIER (279). (B) The distribution of phrenic motor neuron somal surface areas divided into tertiles. (C) Smaller motor neurons comprise slow (S; type I muscle fibers) and fast, fatigue-resistant (FR; type IIa muscle fibers) motor units, whereas larger motor neurons comprise fast fatigue intermediate (FInt; type IIx muscle fibers) and fast fatigable (FF; type IIb muscle fibers). (D) In the rat diaphragm muscle (DIAm), a model predicting motor unit recruitment during different motor behaviors was developed based on (i) a recruitment order dependent on motor neuron size (S first FF last), (ii) the forces generated by each motor unit type, and (iii) the relative proportion of each motor unit type. The model predictions were compared to the transdiaphragmatic pressure (Pdi) or force generated by the DIAm (273). Based on this model, the lower Pdi generated during quiet breathing (eupnea) and hypercapnia/hypoxia stimulated breathing required the recruitment of only fatigue-resistant $\mathrm{S}$ and $\mathrm{F}\mathrm{R}$ motor units. In contrast, with more forceful (higher Pdi) DIAm efforts (e.g., coughing/sneezing and voiding) require recruitment of the entire phrenic motor neuron pool, including more fatigable FInt and FF motor units.

Motor unit/muscle fiber type classification

Muscle fiber types are classified based on the expression of different myosin heavy chain (MyHC) isoforms (407, 408) as determined by immunoreactivity to specific MyHC isoform antibodies. This classification of muscle fiber types also conforms with the mechanical and fatigue properties of fibers. In adult mammalian muscles, four $\mathrm{M}\mathrm{y}\mathrm{H}\mathrm{C}$ isoforms can be identified via immunohistochemistry: ${\mathrm{M}\mathrm{y}\mathrm{H}\mathrm{C}}_{\text{slow}}$, expressed in type I (slow, fatigue-resistant) fibers; ${\mathrm{M}\mathrm{y}\mathrm{H}\mathrm{C}}_{2\mathrm{A}}$ expressed in type IIa (fast, fatigue-resistant) fibers; ${\mathrm{M}\mathrm{y}\mathrm{H}\mathrm{C}}_{2\mathrm{X}}$ and/or ${\mathrm{M}\mathrm{y}\mathrm{H}\mathrm{C}}_{2\mathrm{B}}$ expressed in type IIx/IIb (fast, fatigable) fibers (Figure 3). While ${\mathrm{M}\mathrm{y}\mathrm{H}\mathrm{C}}_{\text{Slow}}$ and ${\mathrm{M}\mathrm{y}\mathrm{H}\mathrm{C}}_{2\mathrm{A}}$ isoforms are uniquely expressed in type I and IIa muscle fibers of healthy mammals, respectively, ${\mathrm{M}\mathrm{y}\mathrm{H}\mathrm{C}}_{2\mathrm{X}}$ and ${\mathrm{M}\mathrm{y}\mathrm{H}\mathrm{C}}_{2\mathrm{B}}$ isoforms are typically co-expressed in muscle fibers in varying proportions related to their fatigability (431). For this reason, these fibers are more appropriately classified as type IIx/IIb.

Figure 3.

(A) Fiber-type classification in the rat diaphragm muscle (DIAm) is based on immunoreactivity to primary antibodies for myosin heavy chain (MyHC) isoforms (pseudo-colored in this example). Reused, with permission, from Mantilla CB, et al., 2010/ELSEVIER (273). (B) Composite table based on results from a number of studies displaying mean±SD of muscle fiber cross-sectional area, the proportion of fiber types within the DIAm, and specific force (273).

Motor units comprise many muscle fibers, with innervation ratios (number of muscle fibers per motor neuron) ranging from as few as 5 to $\ge 1000$ (495). Glycogen depletion is one technique to identify the muscle fibers innervated by a single motor neuron. In this technique, a motor axon is repeatedly stimulated to deplete glycogen stores and thus identify innervated muscle fibers (104, 107, 145, 213, 235, 427). Using glycogen depletion, Burke et al. showed that the classification of motor unit types in the cat medial gastrocnemius muscle based on mechanical and fatigue properties corresponded with the histochemical classification of muscle fibers comprising the motor unit (65). Motor units are classified as either fast or slow based on the time to peak twitch force. This also corresponds with their velocity of shortening and force generation. Motor units are further classified as fatigue resistant or fatigable based on their susceptibility to force loss during repeated stimulation. Accordingly, slow (type S) motor units comprise type I fibers, fast, fatigue-resistant (type FR) motor units comprise type IIa fibers, fast, fatigue-intermediate (FInt) motor units comprise fibers co-expressing ${\mathrm{M}\mathrm{y}\mathrm{H}\mathrm{C}}_{2\mathrm{X}}>{\mathrm{M}\mathrm{y}\mathrm{H}\mathrm{C}}_{2\mathrm{B}}$ and fast, fatigable (FF) motor units comprise fibers co-expressing ${\mathrm{M}\mathrm{y}\mathrm{H}\mathrm{C}}_{2\mathrm{X}}<{\mathrm{M}\mathrm{y}\mathrm{H}\mathrm{C}}_{2\mathrm{B}}$ (Figure 2). Importantly, the range of motor behaviors accomplished by a given muscle depends on the motor unit composition and the appropriate recruitment of motor units. For example, in the DIAm, breathing requires the recruitment of fatigue-resistant type $S$ and FR motor units, while higher force expulsive behaviors require the recruitment of more fatigable type FInt and FF motor units (138, 426). Thus, the properties of the NMJ must match these marked differences in activation history.

Frequency modulation of motor units

The force generated by a motor unit depends on the contractile properties (contractile protein composition) of muscle fibers as well as the frequency of activation by the motor neuron. Muscle fibers produce force through a process of excitation-contraction coupling (see below). Neuromuscular transmission at the NMJ triggers excitation via an increase in cytosolic ${\mathrm{C}\mathrm{a}}^{2+}$ concentration $\left({\left[{\mathrm{C}\mathrm{a}}^{2+}\right]}_{\text{cyt}}\right)$ leading to cross-bridge recruitment and muscle contraction (Figure 4).

Figure 4.

(A) An incoming motor neuron action potential is transduced to muscle fiber action potential, intracellular ${\mathrm{C}\mathrm{a}}^{2+}$ concentration $\left({\left[{\mathrm{C}\mathrm{a}}^{2+}\right]}_{\mathrm{i}}\right)$, and force response in a 1:1 ratio. At lower stimulus intensities, the ${\left[{\mathrm{C}\mathrm{a}}^{2+}\right]}_{\mathrm{i}}$ and force responses do not summate, but as the frequency of activation increases, both the ${\left[{\mathrm{C}\mathrm{a}}^{2+}\right]}_i$ and force responses summate until they fuse. (B) In permeabilized muscle fibers, the relationship between $\left[{\mathrm{C}\mathrm{a}}^{2+}\right]\left(\mathrm{p}\mathrm{C}\mathrm{a}\right)$ and muscle force exhibits a sigmoidal relation that is shifted left in type I fibers compared with type II fibers (159). (C) The relationship between cat diaphragm muscle (DIAm) motor unit force and phrenic nerve stimulation frequency also displays a sigmoidal curve, which is shifted leftward for slow motor units as compared to fast motor units. Fast fatigue intermediate (FInt) and fast fatigable (FF) motor units display the most rightward shifted force-frequency response curves (144, 425). (D) Single motor unit action potentials or compound summated action potentials in the rat diaphragm muscle (DIAm) were recorded using electromyographic (EMG) electrodes. Single motor unit action potentials were identified by their constant waveform. In this example, the discharge profiles of three single motor units in the rat DIAm were discriminated. (E) Once recruited, the discharge rate of DIAm motor units increases as inspiratory efforts proceed. As inspiratory drive increases, the difference between onset and peak motor unit discharge rate increases, reflected frequency modulation of force generation to accomplish different behaviors (415).

The force response of a muscle fiber to a single axonal action potential is the twitch force (Figure 4A). As the discharge rate increases, the force signals summate in time resulting in the fusion of twitch responses. There is a sigmoidal relation between motor unit force and the frequency of activation induced by electrical stimulation of motor axons (Figure 4C). At lower frequencies, there is high fidelity between axonal action potentials and the muscle fiber response (131, 213). The sigmoidal relation between the force response and the frequency of activation (force/frequency response) is due to differences in ${\mathrm{C}\mathrm{a}}^{2+}$ sensitivity and cross-bridge formation (excitation-contraction coupling). The binding affinity of troponin $\mathrm{C}\left(\mathrm{T}\mathrm{n}\mathrm{C}\right)$ for ${\mathrm{C}\mathrm{a}}^{2+}$ is higher in type II DIAm fibers, which shifts their force-${\mathrm{C}\mathrm{a}}^{2+}$ response to the left compared with type I fibers (Figure 4B) (158, 428). Accordingly, the ${\mathrm{C}\mathrm{a}}^{2+}$ concentration $\left({\left[{\mathrm{C}\mathrm{a}}^{2+}\right]}_{\text{cyt}}\right)$ that produces $50\mathrm{\%}$ of maximum force $\left({\mathrm{p}\mathrm{C}\mathrm{a}}_{50}\right)$ is higher in type II DIAm fibers compared with type I. The ${\mathrm{C}\mathrm{a}}^{2+}$ sensitivity of force generation is indexed by ${\mathrm{p}\mathrm{C}\mathrm{a}}_{50}$. The differences in force-frequency response of different fiber types are explained by essential differences in ${\mathrm{C}\mathrm{a}}^{2+}$ sensitivity and the force-${\mathrm{C}\mathrm{a}}^{2+}$ response (Figure 4C).

The range of motor unit discharge rates across behaviors can be measured in vivo with electromyography using intramuscular or subcutaneous electrodes (Figure 4D) (38). In the rat DIAm, motor units display frequency coding with progressively increasing discharge rates during activation (e.g., lower onset discharge rate compared with peak discharge rates for higher forces; Figures 4D and 4E). The onset discharge rate of DIAm motor units during lower force ventilatory behaviors (e.g., eupnea, hypoxia-hypercapnia, and sighs) is fairly consistent at approximately 28 Hz (112, 273, 415, 470). During inspiratory efforts, the discharge rate of DIAm motor units progressively increases to approximately $40\mathrm{H}\mathrm{z}$ during eupnea, to approximately $45\mathrm{H}\mathrm{z}$ during hypoxia-hypercapnia, and to approximately $60\mathrm{H}\mathrm{z}$ during sighs. This range of discharge rates corresponds to the steepest portion of the force-frequency response curve of slow and FR motor units (Figure 4C). During higher force behaviors of the DIAm, including airway occlusion ($\sim 40\mathrm{\%}$ maximum force), the onset discharge rate of motor units is higher at approximately $40\mathrm{H}\mathrm{z}$ with a peak discharge rate is approximately $55\mathrm{H}\mathrm{z}$ (Figure 4E). Thus, across all motor behaviors of the DIAm, the range of motor unit discharge rates corresponds with the steepest portion of the sigmoidal force-frequency response curve (143).

Motor units are recruited to accomplish different DIAm motor behaviors that vary considerably in their duty cycles. Differences in motor unit activation (discharge rate and duty cycle) may drive structural and functional changes at the NMJ. Type $\mathrm{S}$ and FR DIAm motor units consisting of type I and IIa muscle fibers are recruited to maintain breathing across the lifespan (Figure 4B) with duty cycles of approximately $40\mathrm{\%}$ but at lower discharge rates (273). More fatigable DIAm motor units consisting of type IIx/IIb muscle fibers are recruited relatively infrequently for higher force and velocity expulsive behaviors (duty cycles $<1\mathrm{\%}$), but at higher discharge rates [Figure 6; (138, 225, 226).

Figure 6.

(A) Postsynaptic responses to spontaneous or evoked synaptic vesicle release can be measured electrophysiologically. In this example, the response to the spontaneous exocytosis of ACh was observed as miniature endplate potentials (mEPPs) and more concerted release as evoked endplate potentials (EPPs) were recorded using an intracellular microelectrode in the muscle fiber. The microelectrode (glass micropipette) was inserted into a rat diaphragm muscle (DIAm) fiber near the neuromuscular junction identified by labeling the presynaptic terminal using FM4-64. (B) Quantal content (QC determined as the ratio of EPP amplitude to mEPP amplitude) was significantly greater in type IIx/IIb DIAm fibers compared with type I and IIa fibers (*P < 0.01) (389). (C) In rat DIAm, the average mEPP amplitude recorded in type IIx/IIb fibers was significantly smaller than that in type I and IIa fibers (*P < 0.05) (389). (D) The frequency of spontaneous mEPPs was comparable among type I, IIa, and IIx/IIb fibers in the rat DIAm (389).

Neuromuscular Junctions

The NMJ comprises a presynaptic terminal, the postsynaptic endplate, and the terminal Schwann cell (Figure 5). The presynaptic terminal comprises structures and mechanisms for the mobilization, release, and reuptake of synaptic vesicles as well as support for local metabolic demands. The synaptic cleft ($\sim 50\mathrm{n}\mathrm{m}$ across) is the space between the pre- and postsynaptic membranes and contains a basal lamina layer that is continuous with that of the muscle fibers and local Schwann cells. The mammalian postsynaptic endplate consists of junctional folds with AChRs clustered at the peaks of folds (124, 504) and voltage-gated ${\mathrm{N}\mathrm{a}}^{+}$ channels in the troughs (127). The terminal Schwann cell plays a fundamental supporting role in the development and maintenance of the NMJ (123). ACh released at the presynaptic terminal induces cation conductance and depolarization at AChRs, which opens voltage-gated ${\mathrm{N}\mathrm{a}}^{+}$ channels, generating an action potential that propagates along the muscle fiber membrane.

Figure 5.

(A) Pre- and postsynaptic elements of neuromuscular junctions (NMJs) on rat diaphragm muscle (DIAm) fibers can be visualized by confocal microscopy. Phrenic motor axon and presynaptic terminals were labeled by fluorescence immunostaining using neurofilamin antibody (red). Motor endplates were labeled using fluorescently tagged $\alpha$-bungarotoxin (green), which binds to cholinergic receptors. DIAm fiber types were distinguished by immunofluorescence using antibodies specific to different myosin heavy chain $\left(\mathrm{M}\mathrm{y}\mathrm{H}\mathrm{C}\right)$ isoforms. In this example, immunoreactivity for anti-${\mathrm{M}\mathrm{y}\mathrm{H}\mathrm{C}}_{2\mathrm{B}}$ was used. The $\mathrm{M}\mathrm{y}\mathrm{H}\mathrm{C}2\mathrm{B}$ isoform is co-expressed with $\mathrm{M}\mathrm{y}\mathrm{H}\mathrm{C}2\mathrm{X}$ in rat DIAm fibers (159) so they are classified as type IIx/IIb. Accordingly, the two NMJs shown in this image are on type IIx/IIb DIAm fibers. (B) Electron micrographic (EM) image of a NMJ on a type $\mathrm{N}\mathrm{l}/\mathrm{l}\mathrm{l}\mathrm{b}$ rat DIAm fiber in which it is possible to clearly distinguish the presynaptic terminal containing an active zone, synaptic vesicles, and the motor endplate containing junctional folds. Synaptic vesicles docked near the active zone hypothetically form a readily releasable pool, whereas non-fused synaptic vesicles form a reserve pool. It is assumed that the availability of synaptic vesicles to fuse with the presynaptic terminal membrane near the active zone depends on the distance from the active zones.

Presynaptic terminals

Motor axons traverse varying distances originating from somatotopically organized segments within the spinal cord in mammals (17, 18) and ending as the presynaptic terminal at specialized regions along muscle fibers, the endplate (433). The presynaptic terminal of the motor neuron axon primarily subserves signaling mechanisms for the transduction of membrane depolarization and the release of ACh through synaptic vesicles, that is, the impetus for excitation-contraction coupling. Presynaptic terminals also provide structures that enable retrograde communication from muscle fibers back to the presynaptic terminal and up toward the input region of the motor neurons. Presynaptic terminals undergo remodeling throughout the lifespan, from the embryo to old age.

Synaptic vesicles and synaptic vesicle release

From the 1950s to the 1980s, Bernard Katz and his colleagues provided seminal observations in NMJs of frog hindlimb muscles characterizing the role of synaptic vesicle release at the presynaptic terminal in neuromuscular transmission. These studies leveraged intracellular electrophysiological recordings using micropipette electrodes (261, 311), which they modified to measure depolarization of the postsynaptic membrane of NMJs in frog muscle fibers (118). Coincident with the advances provided by electrophysiological techniques, the electron microscope (EM) provided visualization of the presynaptic terminal. Using EM, De Robertis and Bennett observed that the presynaptic terminals of frog sympathetic ganglia and earthworm nerve cord neuropil contained many spherical electron translucent/transparent vesicles, which they termed synaptic vesicles (92). These investigators concluded that these synaptic vesicles were related to the particulate or granular fractions reported by others to contain ACh and catecholamines. In 1956, Palay (338) proposed that the vesicles observed by EM were the structural source of the small amplitude, subthreshold postsynaptic depolarizations that had been previously observed (103, 118, 119) (Figure 6), which Fatt and Katz termed miniature end-plate potentials (mEPPs) (119). Accordingly, the physiologically based hypothesis of quantal transmitter release had a structural correlate. Subsequently, support for the concept that synaptic vesicles are the structural unit of neurotransmitter release (quanta) came from studies using freeze-fracture EM, which visualized the fusion and release events of a synaptic vesicle at the presynaptic terminal in the frog cutaneous pectoris muscle (196).

The quantal theory of synaptic vesicle release was based on observations of the relationship between the average amplitude of spontaneous mEPPs compared to the amplitude of evoked endplate potentials (EPPs) (93, 199, 283, 454). Importantly, it was shown that the amplitude of EPPs was a multiple of the mean amplitude of mEPPs (93). Katz and colleagues recognized that $\mathrm{m}\mathrm{E}\mathrm{P}\mathrm{P}\mathrm{s}$ and EPPs were not action potentials but followed passive cable properties such that the amplitude of the mEPP and EPP decayed depending on the distance from the endplate region (determined by a length constant).

Consistent with the passive properties of synaptic events at the NMJ, Katz demonstrated that mEPP amplitude varied with fiber size in the frog digitorum longus muscle (223). In agreement, in the rat DIAm, it was shown that the mean amplitude of mEPPs is greater in type I and IIa fibers compared with type IIx/IIb fibers (113, 389) (Figure 7), reflecting the smaller fiber cross-sectional area of type I and IIa fibers, which affects their intrinsic electrophysiological properties $\left(R_{\mathrm{i}\mathrm{n}},C_{\mathrm{m}}\right)$ (389). The mean amplitude of evoked EPPs is similar across fiber types in the rat DIAm (113, 389) (Figure 7). Quantal content $\left(\mathrm{Q}\mathrm{C}\right)$ is calculated as the ratio of EPP to mean mEPP amplitudes, and thus reflects the number of synaptic vesicles (i.e., mEPPs) contributing to an EPP (93). The evoked QC of NMJs at type IIx/IIb DIAm fibers is greater than that at type I and IIa fibers (113, 389).

Figure 7.

The SNARE protein complex mediates the docking and fusion of synaptic vesicles to the presynaptic terminal membrane. In response to a nerve action potential and depolarization of the presynaptic terminal, this process is initiated by the influx of ${\mathrm{C}\mathrm{a}}^{2+}$ (represented by the grey circles) via voltage-gated ${\mathrm{C}\mathrm{a}}^{2+}$ channels. The elevated intracellular ${\mathrm{C}\mathrm{a}}^{2+}$ concentration $\left({\left[{\mathrm{C}\mathrm{a}}^{2+}\right]}_i\right)$ leads to increased ${\mathrm{C}\mathrm{a}}^{2+}$ binding to synaptotagmin, which then triggers the signaling cascade responsible for synaptic vesicle fusion to the presynaptic terminal membrane near active zones. The SNARE signaling cascade involves synaptobrevin, mammalian uncoordinated-18 (Munc18), complexin, and syntaxin working together to fuse the synaptic vesicle membrane to the presynaptic terminal membrane. ${\mathrm{C}\mathrm{a}}^{2+}$ channels are also tethered close to the synaptic vesicle by Munc13, rab3-interacting molecules (RIM), and RIM binding protein (RIM-BP).

While Fatt and Katz were the first to describe the distribution of mEPP amplitude and frequency at frog NMJs, Gage and Hubbard (1965) extended and confirmed similar findings in several different species including the mammalian NMJ. The frequency of mEPPs is related to the probability of spontaneous release and thus likely reflects the probability of synaptic vesicle fusion to the presynaptic terminal membrane and subsequent release (53, 287, 389). At rest, $\mathrm{m}\mathrm{E}\mathrm{P}\mathrm{P}$ frequency is similar across different fiber types in the rat DIAm (113, 389). A small yet thorough body of research has examined whether mEPP frequency truly represents a spontaneous, random process (stochastic Poisson distribution). Cohen et al. tested the Poisson distribution of mEPP frequency from 11 different datasets in the sartorius muscle from a frog using 5 different goodness of fit tests (81). Additionally, they tested the independence of the interval between mEPPs using both autocorrelation analysis and a comparison of the unsmoothed power spectrums given a Poisson expectation. They concluded that the frequency of mEPPs was not accurately described by a true Poisson distribution. Segal et al. explored the distribution of $\mathrm{m}\mathrm{E}\mathrm{P}\mathrm{P}$ frequency through Campbell’s theorem, which calculates the expected value and variance of a random point process but does not inherently require it to be a Poisson distribution, allowing for skewed distribution (414). This approach accurately predicted the distribution of $\mathrm{m}\mathrm{E}\mathrm{P}\mathrm{P}$ frequencies except at high frequencies. Conclusions about the nature of $\mathrm{m}\mathrm{E}\mathrm{P}\mathrm{P}$ frequency are partly obscured by observations of a nonquantal release of ACh (leakage) (222) and limited by damage caused by micropipettes, effects of any added stimulating agents, redistribution of ions (i.e., ${\mathrm{K}}^{+},{\mathrm{C}\mathrm{l}}^{-}$) in isolated preparations, accumulation of $\mathrm{A}\mathrm{C}\mathrm{h}$ in the synaptic cleft, and high rates of leakage (222, 414). Furthermore, Poisson distributions follow assumptions of pure independence such that any two events cannot occur at the same time, and the occurrence of one event will not affect the probability that a second event will occur (nondeterministic). Given the 3-D nature of NMJs and the singular recording micropipette within the muscle fiber, these assumptions seem too limited to accurately describe the release dynamics at NMJs.

Synaptic vesicles are relatively spherical with a lipid bilayer membrane interposed with protein structures to transport neurotransmitters and ions across the synaptic vesicle membrane (94, 180), trafficking proteins, and docking proteins. The neurotransmitter protein complex exchanges neurotransmitter for either protons ( 2 protons: 1 neurotransmitter) or through a process that acidifies the synaptic vesicle (44). Each synaptic vesicle at presynaptic terminals of motor neurons has been estimated to contain approximately 5,000 to $\mathrm{10,000}\mathrm{A}\mathrm{C}\mathrm{h}$ molecules in the cutaneous pectoris muscle of frogs and the external oblique muscle of snakes (239). Trafficking proteins mediate the intracellular movement of the synaptic vesicles including axonal transport, synaptic vesicle pool formation, and docking in the active zone. Whereas the docking proteins allow fusion with the presynaptic terminal membrane and release of $\mathrm{A}\mathrm{C}\mathrm{h}$ via exocytosis.

Some synaptic vesicles in the presynaptic terminal are docked to active zones, identified by electron-dense regions on the presynaptic terminal membrane. The synaptic vesicles docked to active zones are thus hypothesized to be readily available to release ACh upon stimulation of the presynaptic terminal (termed the “readily releasable pool”). This readily releasable pool of synaptic vesicles is distinguished structurally by their location in proximity to active zones, which are structurally aligned with the AChR dense region of the junctional folds at the endplate (Figure 6). The size of the readily releasable pool of synaptic vesicles is also defined functionally by the initial amplitude of the evoked EPP (QC—see below). Notably, while at higher stimulation frequencies, quantal estimates suggest total vesicular exocytosis at active zones, at low stimulation frequencies, exocytosis of the readily releasable pool is smaller. This suggests that not all active zones participate in exocytosis during stimulations in the physiological range (394). The presynaptic terminal also appears to contain other pools of synaptic vesicles that are not readily releasable (375). These synaptic vesicles are located throughout the presynaptic terminal and can dock with the presynaptic terminal membrane only when a site becomes available (94, 389). It is thought that synaptic vesicles that are closer to active zones, but not docked, constitute an “immediately available” pool (374), but the functional definition of this pool is not as well defined. Structurally, the number of synaptic vesicles in the immediately available pool can be arbitrarily determined based on distance (e.g., 200 nm) from the active zone. Although not clearly defined, the synaptic vesicles in the immediately available pool can be functionally determined by changes in QC during repetitive stimulation (see below). Other synaptic vesicles are bound within the presynaptic terminal and must be freed before they are available for docking and release, thus comprising a reserve pool. The number of docked synaptic vesicles can be counted to estimate the size of the readily releasable vesicle pool based on location, and QC can be determined electrophysiologically (see below). In the rat DIAm, the number of docked synaptic vesicles per active zone is not different across motor unit types. However, presynaptic terminals at type IIx/IIb DIAm fibers have a greater surface area with more active zones and thus, a larger readily releasable pool of synaptic vesicles (389). The number of synaptic vesicles in the structurally defined immediately available pool can only be estimated as the mean number of vesicles contained within an arbitrarily defined distance (e.g., 200 nm) from presynaptic terminal active zones (389, 409). Those synaptic vesicles outside of this arbitrary range constitute a reserve pool. With larger presynaptic terminals, the size of both the immediately available and reserve pools of synaptic vesicles is larger at type IIx/IIb DIAm fibers.

Mechanisms of synaptic vesicle release

Synaptic vesicle release involves a complex process requiring a cascade of intracellular mechanisms to mobilize, dock, fuse, and release ACh collectively termed the SNARE (soluble $N$-ethylmaleimide-sensitive factor-associated protein receptor) complex. The fusion of synaptic vesicles and release of ACh occurs at active zones, which consist of bands of electron-dense “active zone material”—aggregates of structural macromolecules (ribbons, T-bars, beams, etc.) bound to the membrane-with the precise ordering of the structures that bind synaptic vesicles (307, 450, 484). This complex docks and primes synaptic vesicles for exocytosis. Synaptotagmin allows for ${\mathrm{C}\mathrm{a}}^{2+}$-evoked synaptic vesicle fusion and exocytosis at the active zones. Close association of postsynaptic receptors to active zones is ensured via transsynaptic cell-adhesion molecules. Active zones comprise a set of core constituent proteins consisting of rab3-interacting molecule (RIM), RIM-binding protein (RIM-BP), mammalian uncoordinated-13 (Munc13), $\alpha$-liprin, and a protein rich in the amino acids glutamine, leucine, lysine, and serine (ELKS) (218). RIM, RIM-BP, and Munc13 are discussed below in relation to their role in neurotransmitter release via the SNARE complex. The proteins $\alpha$-liprin and ELKS are cytoskeletal and scaffolding proteins involved in active zone assembly (262, 447). In addition, ELKS are implicated in the regulation of ${\mathrm{C}\mathrm{a}}^{2+}$ influx within the presynaptic terminal (262). Active zones are situated immediately across the synaptic cleft from the region of the postsynaptic membrane that contains the highest density of AChRs. Based on studies in rat and mouse limb muscles and DIAm, there are several hundred (ranging from $\sim 200$ to 900) active zones per presynaptic terminal (74). Proteins comprising the SNARE complex are localized directly across the synaptic cleft from the region of the endplate containing AChRs (i.e., within the active zones) of frog NMJs.

In an elegant series of studies in NMJs at frog sartorius muscle, Katz and Miledi clearly demonstrated that synaptic vesicle release at the presynaptic terminal requires ${\mathrm{C}\mathrm{a}}^{2+}$ influx and a rise in ${\left[{\mathrm{C}\mathrm{a}}^{2+}\right]}_{\text{cyt}}$ (219–221). The rise in ${\left[{\mathrm{C}\mathrm{a}}^{2+}\right]}_{\mathrm{c}\mathrm{y}\mathrm{t}}$ at the presynaptic terminal triggers synaptic vesicle fusion and exocytosis at the active zones, thereby releasing $\mathrm{A}\mathrm{C}\mathrm{h}$ into the synaptic cleft. The presence of ${\mathrm{C}\mathrm{a}}^{2+}$ triggers synaptic vesicle docking in <100 μs (396), which some take as evidence that the SNARE complex is at least partially formed and possibly primed prior to synaptic vesicle docking (449). Voltage-gated ${\mathrm{C}\mathrm{a}}^{2+}$ channels at the presynaptic terminal consist of an $\alpha 1$ subunit (pore-forming) and two auxiliary subunits (β and $\alpha 2\delta$). The $\alpha 1$ subunit consists of four homologous domains, each with six transmembrane segments. Segments 1 through 4 are sensitive to changes in $V_{\mathrm{m}}$ and segments 5 and 6 comprise the channel pore (496). Voltage-gated ${\mathrm{C}\mathrm{a}}^{2+}$ channels are tethered to the presynaptic terminal membrane and the SNARE complex by the proteins RIM, RIM-BP, and Munc13, which also play roles in docking the synaptic vesicles (Figure 7) (449). Voltage-gated ${\mathrm{C}\mathrm{a}}^{2+}$ channels are tethered less than $100\mathrm{n}\mathrm{m}$ away from the cellular membrane-bound SNARE within the active zones of frog NMJs (382, 383).

The SNARE complex provides the structure necessary to overcome the electrostatic and hydration repulsive forces between the membranes of the synaptic vesicle and the presynaptic terminal. Synaptic vesicle fusion and release of ACh require the coordination and interaction of multiple SNARE proteins (300, 419). Mutual negative charges on the exposed proteins will prevent fusion by electrostatic repulsion, and binds to ${\mathrm{C}\mathrm{a}}^{2+}$ sensitive proteins it provides ares of reversed polarity (47, 339). The density of SNARE proteins determines fusion kinetics via the extent of liposome fusion (210).

There are two lines of evidence for the existence of the SNARE complex in the presynaptic terminal of mammals: (i) studies using botulinum toxin, which destabilizes or prevents the formation of SNARE, provide a useful tool to study the SNARE complex in mammalian NMJs (487), and (ii) histochemical techniques using fluorescently labeled proteins to directly study the presence of and changes in key SNARE complex proteins. The proteins associated with the docking and release of synaptic vesicles directly form the SNARE complex and provide additional tethering between the SNARE complex and the membranes (Figure 8). The core SNARE proteins are synaptobrevin, synaptosomal-associated protein, 25 kDa (SNAP25), and syntaxin. Synaptobrevin is a highly prevalent transmembrane synaptic vesicle protein (~70/vesicle), whereas SNAP25 and syntaxin are associated with the cell membrane and provide an ionic coupled link with neighboring SNARE complexes (292). Synaptotagmin, complexin, Munc18, and Munc13 are regulatory proteins related to the SNARE complexes (Figure 8). Synaptotagmin is evolutionarily conserved and is the primary ${\mathrm{C}\mathrm{a}}^{2+}$ detector. In response to a rise in ${\left[{\mathrm{C}\mathrm{a}}^{2+}\right]}_{\text{cyt}}$, synaptotagmin, as well as complexin, bind to syntaxin (34, 60). Complexin mediates SNARE oligomerization and organizes SNARE complexes into zig-zag patterns (240, 466). If complexin is absent, there is a suppression of fast, synchronous exocytosis and conversely an increase in spontaneous exocytosis (367). Finally, Munc18 interacts with syntaxin, which triggers a conformational change that allows it to bind with synaptobrevin (453). The absence of Munc18 entirely prevents synaptic vesicle fusion and leads to neuron degeneration (189). The formed SNARE complex undergoes a progressive zippering-like effect destabilizing the hydrophilic surfaces of the lipid bilayer and allowing the opening of a fusion pore (184). Finally, synaptophysin is another highly prevalent protein in the synaptic vesicle membrane but its functional role is poorly understood. Some evidence suggests that synaptophysin acts as a negative regulator of neurotransmitter release by modifying the probability of release (362). Differences in synaptophysin may explain fiber type differences in the probability of synaptic vesicle release (see below).

Figure 8.

(A) The average decline in quantal content (QC) over a 10-min period of continuous 20-Hz stimulation at type I or IIa vs type IIx/IIb fibers in the rat diaphragm muscle (DIAm). At both fiber types, there was a rapid early decline in QC (1st 100 s) followed by a slower (late) decline (beyond 300 s). Inset: Shows the immediate decline in QC occurring within the first $0.5\mathrm{s}$. (B) The relative change in QC (normalized to the initial QC) is similar across type I or IIa and type IIx/IIb DIAm fibers in both the early and late phases of decline. However, there was a fiber type difference in the immediate decline in $(<2.5\mathrm{s}$; inset), with a greater relative change at type I or IIa fibers compared to that at type $\mathrm{I}\mathrm{l}\times \mathrm{I}\mathrm{l}\mathrm{l}\mathrm{b}$ fibers $(P<0.05)$. (C) The rate of synaptic vesicle replenishment from the reserve pool during repetitive phrenic nerve stimulation (20 or $50\mathrm{H}\mathrm{z};0.5\mathrm{m}\mathrm{s}$ pulse duration) was estimated based on the difference between the predicted depletion of the readily- releasable pool of synaptic vesicles and the actual QC measurements at DIAm fibers. A higher rate of synaptic vesicle replenishment was observed at type IIx/IIb fibers compared to type I and IIa fibers $\left({}^{\mathrm{*}}P<0.05\right)$. Figure used with permission from (389).

Synaptic vesicle recycling

The presynaptic terminal membrane is in a constant state of turnover, which serves to release and reform synaptic vesicles, as well as provide a mechanism for retrograde transport of signaling molecules taken up through endocytosis (442). Importantly for synaptic transmission, synaptic vesicle recycling is required to replenish the readily releasable pool of synaptic vesicles to sustain quantal release (370, 377, 448) after approximately the first s of repetitive stimulation at type IIx/IIb fibers and after approximately 0.5 s at type I and IIa fibers in the rat DIAm (389) (Figure 8). The recycling of synaptic vesicles can be visualized using styryl dyes such as FM4-64 (272) (Figure 9). After synaptic vesicle fusion with the presynaptic terminal membrane, ACh is released via exocytosis thereby exposing the synaptic vesicle membrane to styryl dyes in the extracellular fluid. Styryl dyes such as FM4-64, introduced by William J. Betz in 1992 and named after Fei Mao (40), bind to the entire presynaptic terminal membrane, including the fused synaptic vesicle membranes. When the synaptic vesicle is recycled via endocytosis, the FM4-64 labeled membrane is internalized in the presynaptic terminal (37, 40, 366). After washing out the extracellular FM4-64, the recycled and internalized synaptic vesicles can be visualized using confocal fluorescence microscopy (Figure 10). The recycling of synaptic vesicles is faster in presynaptic terminals of type I fibers (366). By imaging FM4-64 uptake, the extent of synaptic vesicle recycling following stimulation at 10 Hz was reported to be greater in type I and IIa DIAm fibers compared with type IIx/IIb fibers. With repetitive stimulation at 50 Hz, the extent of FM4-64 “destaining” was greater in type I and IIa DIAm fibers compared to type IIx/IIb fibers, again indicating an increased rate of synaptic vesicle recycling (272). Two weeks of phrenic motor neuron and DIAm inactivity imposed by cervical spinal cord hemisection at ${\mathrm{C}}_2$ results in an increase in the rate of synaptic vesicle recycling across all fiber types. In contrast, blocking action potential propagation along the phrenic nerve for two weeks using tetrodotoxin (TTX) markedly disrupted synaptic vesicle recycling. These results indicate that the synaptic vesicle pools at presynaptic terminals display considerable plasticity that can affect the efficacy of neuromuscular transmission.

Figure 9.

(A) Confocal imaging of presynaptic terminals at three different rat diaphragm muscle (DIAm) fibers labeled by the uptake of the styryl dye FM4-64. (B) After loading the terminals with FM4-64, the dye was washed from the bath and the terminals were destained by repetitive phrenic nerve stimulation at 10 Hz (0.5 ms supramaximal pulses with a 67% duty cycle for a 20-min period) compared with presynaptic terminals at NMJs on type IIx/IIb fibers. Values are mean±standard error for single-exponential fitted curves.

Figure 10.

FM4-64 is applied extracellularly and is taken up into the presynaptic terminal through synaptic vesicle endocytosis. Confocal imaging during this initial process provides information on synaptic vesicle recycling. After loading the presynaptic terminal with FM4-64, and rinsing the remaining extracellular FM4-64, synaptic vesicle binding and release can be assessed by the decline in FM4-64 fluorescence.

The fastest proposed mechanism for synaptic vesicle endocytosis is a process termed “kiss-and-run” whereby the synaptic vesicle fuses, the pore opens, and is immediately closed again (376). While a fairly straightforward hypothesis, the ‘kiss-and-run’ mechanism of synaptic vesicle fusion and ACh release is the most difficult to verify due to the speed of the ‘kiss-and-run’ actions and limitations in approaches to directly study it (442).

Synaptic vesicle uptake occurs in a frequency-dependent manner (168). The nontoxic subunit of cholera toxin B (CTB), which binds to gangliosides on the motor neuron membrane can be used to label internalized membrane after endocytosis. When CTB is conjugated with a fluorescent marker, the internalized CTB can be imaged using confocal microscopy, which has been used to identify the uptake of lipid rafts. In unstimulated motor neurons or under conditions of low-frequency stimulation, endocytosis (lipid raft uptake) is not different across rat DIAm fiber types. However, at higher frequencies of stimulation, endocytosis increases (168). In general, motor units comprising type IIx/IIb fibers discharge at higher frequencies compared with motor units comprising type I and IIa (416, 426). Therefore, higher rates of lipid raft uptake through bulk endocytosis occur in larger motor neurons that innervate type IIx/IIb fibers.

Neurotrophic and myotrophic anterograde/retrograde communication

In addition to synaptic vesicles, there are other vesicles that are important in communication between muscle fibers and motor neurons via endocytosis and axonal transport. These other vesicles transport molecules that do not easily diffuse in the hydrophilic extracellular matrix (268) and can carry proteins, lipids, DNA, mRNA, and micro RNA that are involved in transmembrane transport, intracellular signaling pathways, cell dynamics, and metabolism (387). Clathrin is a large protein that lines the inner surface of the cell membrane and assists in the budding process to create endocytosed vesicles (340). Clathrin-mediated endocytosis is part of the process of recycling synaptic vesicles in which de novo vesicles are filled with ACh via acetylcholine transferase (AChT) and returned to the reserve pool synaptic vesicles (252). Clathrin-mediated endocytosis also mediates the internalizing of signaling molecules, that are bound to the presynaptic terminal membrane. Clathrin-mediated endocytosis is not directly affected by discharge rate (19, 195). Bulk endocytosis occurs through a process of lipid raft uptake, which accounts for synaptic vesicle recycling at higher frequencies (168). Bulk endocytosis is not molecule specific and requires postendocytic sorting mechanisms (442).

Retrograde signaling pathways from the presynaptic terminal back toward the motor neuron soma mediate myotrophic influences at the motor neuron as well as provide communication regarding the efficacy of neuromuscular transmission. Retrograde communication is mediated via the axonal transport of endocytosed vesicles. Direct evidence for the importance of retrograde signaling pathways at the NMJ predominantly come from research using drosophila; however, studies in mice that have manipulated protein expression in the muscle and observed changes in the motor neuron provide evidence for retrograde communication across the NMJ. For example, signaling cascades involving the secretion of Wnt glycoproteins are important mediators of NMJ development and maturation. Recently, researchers used mice with a muscle-specific $\beta$-catenin (a downstream protein in the Wnt canonical pathway) knockout and found a decrease in presynaptic ${\mathrm{C}\mathrm{a}}^{2+}$ sensitivity at the motor neuron presynaptic terminal (258).

Retrograde signaling by brain-derived neurotrophic factor (BDNF) via its high affinity tropomyosin receptor kinase B (TrkB) receptor may also play an important role in maintaining the integrity of the presynaptic terminal. The motor neuron produces BDNF, which is released in synaptic vesicles after which it binds to its high affinity TrkB receptor. Subsequently, the BDNF/TrkB complex is internalized by endocytosis and then retrogradely transported to the motor neuron soma where it mediates signaling via phosphorylation of the transcription factor cAMP response element-binding protein (CREB). Endocytosis of BDNF/TrkB has a distinct advantage in the presence of cargo-binding adaptors that enable the specificity of molecule uptake (230, 476, 510).

Energetic demands of the presynaptic terminal

Repeated activation of the presynaptic terminals at NMJs has an energetic cost that is met by local ATP production by mitochondria. The volume density of mitochondria for a given surface area of a neuron predicts the ATP requirements for maintaining the membrane potential (492). The presynaptic terminal consumes energy (ATP hydrolysis) via the ${\mathrm{N}\mathrm{a}}^{+},{\mathrm{K}}^{+}$-ATPase pump to counter ion flow from action potentials and leakage currents to restore and maintain the resting membrane potential. ATP requirements for the ${\mathrm{N}\mathrm{a}}^{+},{\mathrm{K}}^{+}$-ATPase pump have been estimated based on recordings of ${\mathrm{N}\mathrm{a}}^{+}$ conductance $\left({\mathrm{g}}_{\mathrm{N}\mathrm{a}}\right)$ during an action potential. The rate of ATP consumption was estimated as a function of the current produced by the ${\mathrm{N}\mathrm{a}}^{+},{\mathrm{K}}^{+}$-ATPase pump, and Faraday’s constant (F; the charge per 1 mole of electrons). The ATP consumption rate was assumed to be directly proportional to the probability that an action potential would occur (187, 250). In the rat DIAm, phrenic motor neurons innervating type I and IIa fibers are highly active being recruited to sustain breathing with a duty cycle of $40\mathrm{\%}$ to 50% (138, 426). Thus, it is likely that presynaptic terminals of NMJs at type I and IIa DIAm fibers exhibit higher mitochondrial volume densities compared to type IIx/IIb fibers to support the higher energetic demands at these NMJs.

The main isoform of the ${\mathrm{N}\mathrm{a}}^{+},{\mathrm{K}}^{+}$-ATPase pump in neuronal projections, and presumably the presynaptic compartment of the NMJ, is the $\alpha 3$ subunit (51, 506), which appears to be optimized for high-intensity neuronal firing (16, 46, 85). Indeed, in motor neurons innervating type FF motor units (type IIx/IIb muscle fibers), which exhibit higher discharge rates upon recruitment, the expression of the $\alpha 3$ subunit has been reported to be greater than that found in motor neurons innervating type S (type I muscle fibers) motor units (390). However, the utility of such selectivity is somewhat opaque, as it was also suggested that $\alpha 3$ subunit expression was selective to gamma motor neurons (105). Like many molecular-based classifications of motor neurons, the fidelity of labeling in both motor neurons and their presynaptic terminals at the NMJ is difficult to disambiguate and the reproducibility in mixed motor units has not been established (280). Regardless, at type-identified presynaptic terminals, it remains to be seen whether ${\mathrm{N}\mathrm{a}}^{+},{\mathrm{K}}^{+}$-ATPase pump isoforms are specific to fiber type. Despite these caveats, in diseases such as Alzheimer’s (324), Parkinson’s (421), and importantly amyotrophic lateral sclerosis (ALS)—where marked presynaptic terminal/NMJ dysfunction occurs, ${\mathrm{N}\mathrm{a}}^{+},{\mathrm{K}}^{+}$-ATPase pump function is inhibited (106, 390), which may reflect mitochondrial dysfunction.

ATP is also hydrolyzed to provide the energy required to disassemble the SNARE complex and recycle synaptic vesicles (187, 482). Synaptic vesicles are filled with neurotransmitter via proton exchange, and an ATPase pump maintains the vesicular proton concentration (406). The SNARE complex must be disassembled and recycled to maintain the synaptic vesicle pool. A single molecule of ATP is hydrolyzed to power a “spring-loaded” protein unfolding for a single SNARE complex (395). As with ATP hydrolysis for the ${\mathrm{N}\mathrm{a}}^{+},{\mathrm{K}}^{+}$-ATPase pump to maintain the presynaptic terminal membrane potential, the ATP requirements for maintaining the synaptic vesicle pool are also dependent on motor neuron activity and thus the release of synaptic vesicles. In addition, there are motor unit/muscle fiber type differences in QC that would affect the energetic requirements of the SNARE complex.

Mitochondria within the presynaptic terminal serve two functions. First, mitochondrial oxidative phosphorylation provides ATP for local energy demands at the presynaptic terminal (30, 31, 479). The ATP demand depends on motor neuron discharge rate, as well as the presynaptic terminal surface area and the size of the synaptic vesicle pool. Second, as would be expected, with increases in stimulation frequency, ${\left[{\mathrm{C}\mathrm{a}}^{2+}\right]}_{\text{cyt}}$ also rises (88). If mitochondria are depolarized prior to stimulation, the rise in ${\left[{\mathrm{C}\mathrm{a}}^{2+}\right]}_{\mathrm{c}\mathrm{y}\mathrm{t}}$ is much greater. Therefore, mitochondrial ATP production serves to buffer ${\left[{\mathrm{C}\mathrm{a}}^{2+}\right]}_{\mathrm{c}\mathrm{y}\mathrm{t}}$ by direct influx into mitochondria via the mitochondrial ${\mathrm{C}\mathrm{a}}^{2+}$ uniporter (MCU) as well as ${\mathrm{C}\mathrm{a}}^{2+}$ sequestration via the sarcoplasmic-endoplasmic ${\mathrm{C}\mathrm{a}}^{2+}$ (SERCA) pump. A rise in ${\left[{\mathrm{C}\mathrm{a}}^{2+}\right]}_{\text{cyt}}$ is coupled to an increase in mitochondrial $\left[{\mathrm{C}\mathrm{a}}^{2+}\right]\left({\left[{\mathrm{C}\mathrm{a}}^{2+}\right]}_{\text{mito}}\right)$, which stimulates the tricarboxylic acid (TCA) cycle and leads to an increase in ATP production (99, 160)—excitation-energy coupling. Interestingly, mitochondrial volume density within the presynaptic terminal increases with continuous electrical stimulation, at least in Drosophila (468). Whether this adaptive response is also present in mammals has yet to be determined.

Mitochondrial sequestration of ${\mathrm{C}\mathrm{a}}^{2+}$ can either reduce or increase the probability of synaptic vesicle release by affecting ${\left[{\mathrm{C}\mathrm{a}}^{2+}\right]}_{\text{cyt}}$ (89). However, cell death can occur if ${\left[{\mathrm{C}\mathrm{a}}^{2+}\right]}_{\text{cyt}}$ is not tightly regulated across both the cytosolic and mitochondrial membranes. For example, if ${\left[{\mathrm{C}\mathrm{a}}^{2+}\right]}_{\text{mito}}$ is not regulated and thus increases, a general increase in mitochondrial permeability occurs, including substances to which mitochondria are normally impervious. This increase in mitochondrial permeability can lead to an increase in internal mitochondrial pressure resulting in mitochondrial swelling and eventual rupture. Mitochondrial rupture results in the release of cytochrome $\mathrm{C}$ triggering apoptotic signals within the cell and in some cases neuronal degeneration (318).

Amyotrophic lateral sclerosis reflects motor-unit specific neuronal degeneration

Degeneration and loss of motor neurons are pathognomonic for ALS, with muscle weakness and death by respiratory complications usually occurring within approximately 2 to 3 years of diagnosis (129, 473). Although only a minor fraction of ALS patients have a known genetic cause (the basis for rodent mutant models), there are no observed differences between pathology and prognosis between genetic and sporadic ALS (156). The common pathophysiological disturbances of motor neurons reflect synaptic dysfunction and dysmorphisms that precede cell death and symptoms (9, 72, 313, 384, 444). Across motor units, larger motor neurons comprising type FInt and FF motor units (type IIx/IIb fibers) are more vulnerable, with smaller motor neurons comprising type S and FR motor units (type I and IIa fibers) surviving relatively unscathed (100, 128, 135, 136, 190, 191, 227, 418).

Morphological derangements of the presynaptic terminal are often an early observation in rodent models of ALS employing a variety of genetic mutations and include frank denervation and alterations of distal axons and may exhibit sex differences (5, 286). These alterations include a reduction in the size of the synaptic vesicle pool, reduced colocalization of synapse-stabilizing proteins (such as dystrophin and rapsyn), and increases in distal axon arbors (i.e., collateral sprouting) (70, 77, 285, 313). Similar disturbances, including denervation, are also present in human ALS patients (62).

Although many of the principles expounded upon in this article apply to all mammals in general, there are some well-established differences in human NMJ structure and function compared to those of other mammals. Overall, human NMJs are amongst the smallest that exist in mammals, in both absolute terms and relative to size (in terms of both body mass and muscle fiber diameter), with a “nummular” (coin-shaped) appearance compared to rodents, cats, and dogs with morphologies more akin to “pretzel” shapes (49, 216, 434). One caveat to this is that the fiber-type dependence of NMJ size in many of these species, including humans remains uncharacterized or poorly documented. Despite the smaller cross-sectional areas of human NMJs, their surface area increases around 8-fold, due to the deepened folds of human NMJs compared to smaller mammalian counterparts (1, 435). This greater surface area effectively acts as an amplifier of acetylcholine postsynaptic responses, perhaps to compensate for the lower QC for a given stimulus frequency of humans compared to smaller mammals (86, 166, 435, 493, 494).

Despite motor unit-specific analyses being relatively uncommon, there are suggestions that morphological changes are selective to vulnerable FInt and FF motor units (type IIx/IIb muscle fibers) (148). However, the morphology of both presynaptic terminal and endplate is unlikely to be a useful surrogate for more sensitive assessments as denervation of endplates is seen in a remarkable number of normal, healthy human NMJs (43, 505) and can be misleading regarding the fidelity of signal transduction (130, 139, 488). Due to these confounding factors, simple estimates of synaptic invasion have gone out of vogue as, without the inclusion of a functional assessment, they appear to have no bearing in relation to the fidelity of neuromuscular transmission.

Functional neuromuscular transmission deficits, including QC and mEPP frequency abnormalities, and decreased facilitation of neuromuscular transmission are readily apparent in rodent models of ALS (9, 72, 310, 384) and may explain some of the EMG phenomena observed in clinical cohorts (91, 192, 217). The deficits in QC and mEPP frequency are more apparent in vulnerable FInt and FF motor units (type IIx/IIb muscle fibers) (9, 417). There is much debate regarding the timing of motor axon degeneration compared with motor neuron pathophysiology (483). Impaired axonal transport (480) and diminished trophic signaling (489, 490) undoubtedly contribute to a steady escalation of motor unit pathology (451, 490) concomitant or immediately following motor neuron loss and denervation (313, 444). In addition, marked loss of Schwann cell function is characteristic of ALS. Healthy Schwann cells, which respond to motor neuron discharge patterns, increased ACh release among other physiological changes to presynaptic terminal structure and function (9, 15, 285, 465, 471), contribute to neuromuscular impairment in ALS in a motor unit-specific manner (471).

Overall, in ALS, motor axon presynaptic terminals undergo remarkable remodeling in the attempt to ameliorate the effects of motor neuron loss and denervation on neuromuscular transmission fidelity. Indeed, these processes may be engaged in response to aberrant axonal signaling prior to denervation (285, 286). It should also be noted that studies show remarkably few muscle defects in ALS, aside from denervation-induced atrophy (75), providing an ideal substrate for recovery provided motor unit innervation can be effectively restored and motor neuron loss blunted (7, 474). Looking forward, interventions focused on preserving presynaptic terminal contact with the endplate in ALS may also benefit retrograde neurotrophic support derived from muscle-associated factors. Indeed, efforts to promote AChR clustering in ALS models using muscle-specific kinase (MuSK) promoting antibodies lead to the preservation of NMJ innervation, NMJ function, the preservation of motor neurons (prior to endstage), and gross muscle strength (69, 343). Efforts aimed at enhancing docking protein 7 (DOK7), which activates MuSK (203, 325) achieves similar ends, with improvements in NMJ innervation, ameliorates muscle atrophy, and increases life span, but not motor neuron survival (298). Much effort remains to characterize how these attempts to improve NMJ outcomes with ALS relate to the known vulnerabilities of FInt and FF-type motor units.

Postsynaptic endplates

The endplate consists of junctional folds in vertebrates, the peaks of which are paired tightly to the active zones of the presynaptic terminal. The local depolarization at the endplate is electrotonically conducted to the voltage-gated ${\mathrm{N}\mathrm{a}}^{+}$ channels in the troughs of the junctional folds. The shape of the junctional folds at endplates function to facilitate the endplate potential (284) by guiding the depolarizing current toward voltage-gated ${\mathrm{N}\mathrm{a}}^{+}$ channels that are densely populated in troughs of the junctional folds (68, 127). The folded shape is maintained by the basal lamina. This is supported by evidence that dissolution of the basal lamina alters the shape and depth of the folds. The junctional folds in the membrane change the membrane and axial resistance, both dependent on geometry, and therefore the length constant (an electrical constant that describes the exponential decay in passive depolarization over a distance). The length constant is greater for the folded area of the membrane compared with the unfolded membrane (392). This spatial arrangement plays an integral role in producing a muscle fiber action potential (115) to facilitate the excitation-contraction coupling process that results in force production (176, 401).

Slow (type I fibers) and fast fatigue-resistant (type IIa fibers) motor units in the rat DIAm have smaller and simpler endplate structures (e.g., $50\mathrm{\%}$ of the surface area compared with more fatigable fast (type IIx/IIb fibers) motor units), which both reflects and supports their typically lower discharge rates and smaller muscle fibers (131, 168, 353). Endplates at type IIx/IIb fibers are larger, more complex, and more prone to the effects of remodeling. However, this does not appear to be directly generalizable across different muscles. Comparisons of endplate area across muscles in mice showed that NMJs in the soleus (predominantly slow) are smaller than NMJs in the extensor digitorum longus (predominantly fast) but bigger than NMJs in the extensor digitorum communis (25).

Nicotinic acetylcholine receptors

In 1905, John Langley was the first to introduce the concept that receptors are responsible for the cellular responses to applied substances (Receptor Theory), and he specifically described the responses of muscle fibers to the application of nicotine and curare (248). Nearly 30 years later, Henry Dale and Otto Loewi won the Nobel Prize for their shared work on the chemical transmission at neuronal synapses. Notably, Dale spent approximately 20 years working with ACh and in 1929 confirmed the presence of ACh in mammals. In 1934, he introduced the terms cholinergic and adrenergic nerve fibers to describe the substances released by these neurons. An interesting division in the concept of neurological signal transmission occurred with the emerging research from Dale and researchers such as John Eccles, colloquially known as “soup vs spark.” While Dale and colleagues proposed that neural signals were transmitted between cells via chemical substances, Eccles and others firmly held that synapses relied on electrical impulses transmitted to the target cell. In 1951, Eccles “came to accept it unreservedly by what Sir Henry regards as the scientific equivalent of a religious conversion (458).” Curiously, it was not until 1955 that David Nachmansohn theorized that there was a protein acting as the receptor for ACh, and the AChR was finally biochemically characterized in 1970 by Changeux and colleagues (73).

Nicotinic AChRs are pentameric, ionotropic receptors ($2\alpha ,\gamma ,\beta ,\delta$ subunits) activating nonselective cation channels, which produce relatively large synchronous depolarizing currents across the peaks of the junctional folds at the endplate (257). The ACh-binding site at the nicotinic AChR lies at the junction of the $\alpha$ subunit and either the $\gamma$ (mature endplate) or $\epsilon$ (developing endplate) subunit (170). The density of AChRs across mammalian species (bat, mouse, rat) is surprisingly consistent with approximately $8700\mathrm{A}\mathrm{C}\mathrm{h}\mathrm{R}\mathrm{s}/\mu {\mathrm{m}}^2$ (3) and was first described by Salpeter & Eldefrawi in 1973 (399). The EEP resulting from AChR activation is typically greater than the voltage threshold required to generate an action potential via activation of voltage-dependent ${\mathrm{N}\mathrm{a}}^{+}$ channels (i.e., safety factor). The safety factor is affected by the number of AChRs available and the amount of ACh released (i.e., QC). A safety factor is required to ensure the fidelity of generating propagating muscle fiber action potentials in response to neural signals (493). The AChR-mediated depolarizing current is greater in type IIx/IIb fibers compared with type I and type IIa fibers due to differences in QC and the number of postsynaptic AChRs (391).

While mEPPs and EPPs are the results of the synchronous activation of many AChRs, the depolarization events represent the summation of current responses from each activated AChR channel. It is estimated that a single AChR has a 70% to 80% probability of opening in response to neurotransmitter release from the presynaptic terminal (245). The probability that an AChR will open partially depends on the concentration of ACh available to bind to the receptor (14, 224). High concentrations or repeated activation by ACh lead to desensitization of the AChR, which is measured as increased silent periods, where despite sufficient presence of ACh, the receptor channel does not open (97, 224). Based on data with currents elicited by ejecting ACh from a glass pipette, AChR channels open at a fast rate (246) (Figure 11), which is consistent with the steep rise observed during mEPPs. The initial depolarization resulting from AChR currents is sustained, directed, and amplified by the junctional folds toward the voltage-gated ${\mathrm{N}\mathrm{a}}^{+}$ channels to trigger a muscle fiber action potential.

Figure 11.

(A) Current traces from acetyl choline receptors (AChRs) at motor endplates of frog NMJs: (a) currents measured immediately above the AChR clusters and currents measured a short distance from the junctional folds at weaker (b) and stronger (c) stimulus pulses. Figures reproduced with permissions from (269). (B) The current-voltage (I-V) relationship for junctional (open) and extrajunctional (filled) AChRs at motor endplates of frog NMJs recorded using the patch-clamp technique. The slope of I-V relationship represents the conductance of the AChR channel. The reversal potential was $0\mathrm{m}\mathrm{V}$ indicating the AChR at frog NMJs is a non-selective cation channel.

Myasthenia gravis reflects postsynaptic pathophysiology

Myasthenia gravis is caused by the production of autoantibodies to either the AChR itself or to endplate AChR-clustering proteins MuSK, Agrin, LDL receptor-related protein 4 (LRP4), MuSK, Dok7, and Rapsyn (Figure 12), resulting in a breakdown of neuromuscular transmission and symptoms of muscle weakness and increased fatigue (164, 197, 198). In some forms, this auto-immune attack of AChRs or AChR-associated proteins is congenital (110). By contrast to aging and ALS, myasthenia gravis has effects spanning all muscle fiber types (79, 80, 282). Structural NMJ changes in myasthenia gravis include reduced number of AChRs (117, 163, 302, 303, 359), loss of presynaptic innervation of endplates (83, 359), and the derangement of postjunctional folds (108). Functional NMJ changes in myasthenia gravis include reduced safety factor (393), and a reduction of QC (EPP amplitude) (111, 255, 316, 371) despite increased QC (86, 299, 347, 348).

Figure 12.

(A) Averaged images from electron microscopy (EM) were used to reconstruct the nicotinic acetylcholine receptor (AChR) at 4.6 A resolution. The top right panel is the intracellular perspective, and the bottom right is extracellular (297). (B) AChRs are more densely clustered, typically at the peak of the junctional folds (red denotes greater density) compared with the troughs. Although this image is of a cultured rat myotube, which does not consistently exhibit junctional folds. Reproduced with copyright permissions from (452). (C) Agrin is released from the presynaptic cleft and binds to muscle-specific kinase (MUSK) receptors on the muscle fiber. The binding of agrin to MuSK triggers an intracellular cascade that clusters the AChRs. Reproduced with copyright permissions from (308).

In relation to morbidity and mortality in myasthenia gravis, the prime mover is the health of the DIAm, with the major contributors to mortality being respiratory-tract infections such as pneumonia and influenza being major contributors to death (8, 294, 336, 346). Indeed, for patients with MuSKassociated myasthenia gravis, a clear respiratory muscle involvement occurs (242, 312, 337). A variety of treatments are effective for myasthenia gravis, with combined approaches aimed at increasing neuromuscular transmission (e.g., cholinesterase inhibitors) and immunomodulators, to reduce auto-immune activity (346, 422).

Efforts to understand the molecular underpinning of myasthenia gravis have involved a variety of animal model paradigms, including immuno-sensitization to the AChR and MuSK, which replicate the weakness, impaired neuromuscular transmission, and endplate fragmentation apparent in human clinical cases (59, 82, 83, 108, 109, 301, 346, 372). The muscle specificity of symptoms has been associated with developmental timelines of endplate formation, with those muscles that swiftly form synapses after initial in utero motor neuron to skeletal muscle contact being more sensitive to experimentally induced myasthenia gravis (358, 500). In the rodent DIAm, neuromuscular transmission failure is readily apparent (475), along with diminished mEPPs (66), a reduction in AChRs (293), and denervation (244). Some findings at human and rat endplates have suggested a compensatory increase in QC (347).

A key component of future efforts will be to ensure that as individuals are successfully treated for myasthenia gravis age, their adjunctive therapy for confounding factors should be considered. For example, vigilance regarding potential indications of respiratory neuromotor deficits occurring with sarcopenia (125), cardiovascular illness (363, 486), or age-associated infection risks (i.e., COVID-19) (305, 397, 440) are essential.

Terminal Schwann cells

Nonmyelinating terminal Schwann cells cap the NMJ and are a significant component of the NMJ. Louis-Antoine Ranvier first described terminal Schwann cells in 1878, but it has only been in the last approximately 30 years that their importance to the NMJ has been thoroughly explored. Terminal Schwann cells influence growth and stabilization of the NMJ, respond to motor neuron activity, and support NMJ remodeling. There is a moderate-to-strong linear relation between the number of terminal Schwann cells per NMJ (range of $\sim 1-5$) and the size of the endplate (263). The number of terminal Schwann cells present at the NMJ changes with the size of the NMJ through developmental stages. Terminal Schwann cells proliferate during development (12), axonal degeneration due to injury or disease (78, 400, 501), and axonal regeneration (11, 266, 503).

Terminal Schwann cells influence growth and stabilization of the NMJ

Terminal Schwann cells guide outgrowing motor axons, provide support to maintain established NMJs, and are key to the transition from the premature polysynaptic innervation to mature monosynaptic innervation of muscle fibers (Figure 13). Both nerves and Schwann cell precursors develop from neural crest cells. Schwann cell precursors then form immature Schwann cells at E15-E17 in rats and E13-E15 in mice (208, 209, 296). The transition from Schwann cell precursors to immature Schwann cells is synchronized with vascularization of the nerves and the developing perineurium. Motor neurons will still innervate muscles in frog models where Schwann cells have been ablated indicating that terminal Schwann cell development is not necessary for the outgrowth of motor neurons (364). Neurite outgrowth closely follows Schwann cell precursor outgrowth in tadpoles, leading some researchers to propose that Schwann cells guide nerve outgrowth (194). There is supporting evidence in mice and rats showing that after the application of neuregulin1 (NRG1), which causes Schwann cell sprouting, nerve processes extend along the Schwann cell sprouts (188, 469).

Figure 13.

(A) Perinatal developmental timeline of the development of phrenic motor neurons and their innervation of muscle fibers in the rat diaphragm muscle (DIAm). (B) Photomicrographs of motor axons and presynaptic terminals labeled using an anti-body specific for neurofilamin (green) and Schwann cells labeled using an anti-body for S-100. The merged image shows the close association of motor axons to Schwann cells. (C) During late embryonic and early postnatal development, DIAm fibers are innervated by more than one phrenic motor neuron (i.e., polyneuronal innervation). Subsequently, polyneuronal innervation disappears via synapse elimination leaving each muscle fiber innervated by only one motor axon.

Terminal Schwann cells are important for motor neuron survival and health after outgrowth and contact with the muscle fibers. If Schwann cell precursors are absent, such as through a mutant mouse model, most motor neurons die by E18 (373). In mouse models where immature Schwann cells are ablated, motor nerves display the aberrant formation of nerve bundles on outgrowing axons (fasciculation), synaptic growth is stunted, and subsequently, a substantial portion of the nerve terminals retract (71). NRG1 release by motor neurons and ErbB receptor activity on terminal Schwann cells has been implicated as key to this motor neuron and terminal Schwann cell mutually dependent survival during development.

Finally, terminal Schwann cells mediate the transition from polysynaptic innervation to monosynaptic innervation of muscle fibers by motor neurons. Curiously, around P2 in mice, terminal Schwann cells begin to attack the NMJs along muscle fibers. The terminal Schwann cells do not determine which NMJ remains but begin to compete for direct contact with the AChR dense surfaces on muscle fibers to phagocytose displaced motor neuron terminals (438). The motor neuron that is successfully capable of maintaining the greatest contact with the endplate during this competition phase appears to remain as the single surviving presynaptic terminal. The most successful neuron is likely to be the most active neuron. Terminal Schwann cells randomly phagocytose presynaptic terminals during the transition from polysynaptic innervation to monosynaptic innervation at NMJs and the “axosomes”—fragments of the terminal containing intact organelles—are consumed by the terminal Schwann cells and integrated into the cytoplasm (42). The process is at least partially modulated by neuron-derived type III NRG1 during its peak in postnatal development (254) as manipulating the type II NRG1 signal also affects the rate of synaptic pruning. The high level of NRG1 triggers the terminal Schwann cells to intrude into the synaptic cleft.

Terminal Schwann cells and neuromuscular transmission in adults

Terminal Schwann cells sense NMJ activity through muscarinic AChRs as well as receptors for ATP, adenosine (A1 receptor), and substance P (52, 161, 380, 386). In 1992, Jahromi and colleagues as well as Reist and Smith found that terminal Schwann cells respond with an increase in ${\left[{\mathrm{C}\mathrm{a}}^{2+}\right]}_{\text{cyt}}$ after neurotransmitter release from the presynaptic terminal in the cutaneous pectoris muscle of frogs (204, 368). This increase in ${\left[{\mathrm{C}\mathrm{a}}^{2+}\right]}_{\mathrm{c}\mathrm{y}\mathrm{t}}$ in the terminal Schwann cell is dependent on the type of stimulation. During continuous high-frequency stimulation, the increased ${\left[{\mathrm{C}\mathrm{a}}^{2+}\right]}_{\text{cyt}}$ evoked in Schwann cells (204, 368, 386) leads to synaptic potentiation (381). During repeated trains of high-frequency stimulation of the motor nerve, oscillatory ${\left[{\mathrm{C}\mathrm{a}}^{2+}\right]}_{\mathrm{c}\mathrm{y}\mathrm{t}}$ responses occur, and synaptic depression is observed in the mouse soleus muscle (465). When terminal Schwann cells are ablated at frog NMJs, their electrophysiological and force-producing capabilities are significantly altered. After 1 week, EPP amplitude and mean QC are both decreased. The depression after high-frequency continuous stimulation and paired-pulse facilitation is reduced. Finally, twitch tension via nerve stimulation is significantly weaker (364).

Neuromuscular Transmission

The raison d’être of neuromuscular transmission is to faithfully propagate neural signals from the motor neuron to skeletal muscle fibers. To do this, ACh released at the presynaptic terminal induces cation conductance and depolarization at AChRs, which opens adjacent voltage-gated ${\mathrm{N}\mathrm{a}}^{+}$ channels generating an action potential that propagates along the muscle fiber membrane. The cascade initiated by muscle fiber action potentials that leads to muscle fiber contraction is known as excitation-contraction coupling. The effectiveness of neuromuscular transmission governs the initiation of excitation-contraction coupling and thus motor system mechanical output.

Excitation-contraction coupling

Muscle fiber depolarization due to an action potential is passively transmitted down transverse tubules (t-tubules), where it activates voltage-sensitive dihydropyridine (DHP) receptors. The DHP receptors are fused with ryanodine receptor (RyR) channels in the sarcoplasmic reticulum (SR) that open to release ${\mathrm{C}\mathrm{a}}^{2+}$ into the cytosolic space. The increase ${\left[{\mathrm{C}\mathrm{a}}^{2+}\right]}_{\text{cyt}}$ increases binding to troponin C on the actin filament that interacts with troponin I to expose binding sites for myosin heads. The binding of myosin to actin (cross-bridge formation) underlies force production and contraction (176, 401).

Efficacy of neuromuscular transmission

Functionally, the purpose of the NMJ is effective neuromuscular transmission. The measure of the efficacy of neuromuscular transmission is the safety factor or safety margin, defined as the ratio or difference, respectively, between the amplitude of the evoked EPP and the threshold for generation of a muscle fiber action potential (493). In the rat DIAm, the safety factor at type IIx/IIb fibers during a single evoked response is higher compared with type I and IIa fibers. The increased safety factor in type IIx/IIb fibers is in line with the greater QC and the higher density of voltage-gated ${\mathrm{N}\mathrm{a}}^{+}$ channels at the endplate in IIx/IIb muscle fibers (295, 389). However, with repetitive stimulation, EPP amplitude (QC) decreases to a greater extent in type IIx/IIb fibers, compromising their safety factor, and potentially leading to neuromuscular transmission failure.

The efficacy of neuromuscular transmission declines as the duration of motor neuron activity increases. The amplitude of EPPs (QC) decreases over time if the nerve is continuously stimulated, and this is likely due to a decrease in both probabilities of release and number of synaptic vesicles available to release (389) (Figure 9). In the rat DIAm, the absolute level of decline is not different but when normalized to initial QC the decline in QC is greater in the first few s of stimulation in type I and IIa fibers compared to IIx/IIb fibers. Based on estimates of the number of synaptic vesicles in the readily releasable pool at presynaptic terminals, it was determined that the initial decline in QC at type IIx/IIb fibers could be solely attributed to the depletion of the readily releasable pool (389). However, for presynaptic terminals at type I and IIa DIAm fibers, the decline in QC was much faster than would be predicted by a depletion of the readily releasable pool, which was interpreted as a decline in probability of synaptic vesicle release (Figure 14).

Figure 14.

In the rat diaphragm muscle (DIAm), quantal content (QC) decreases with repeated phrenic nerve stimulation (at 20 or $50\mathrm{H}\mathrm{z};0.5\mathrm{m}\mathrm{s}$ pulse duration) in all fiber types (A. type I and IIa; B. type IIx/IIb) (389). A model of the replenishment of the readily releasable pool (RRP) of synaptic vesicles was developed to explain the initial decline of QC and the subsequent leveling of QC. For type I/IIa fibers, the initial decline in QC is steeper than can be explained by the depletion of the RRP, so at least part of this depletion was attributed to a decrease in the probability of synaptic vesicle release. For type IIx/IIb fibers the initial decline in QC was parallel to the depletion of the RRP, which suggests that the depletion is causal. All fiber types exhibit an ability to continuously release synaptic vesicles despite a continual decline in the RRP, therefore the sustained release is due to the repletion of synaptic vesicles through synaptic vesicle recycling. The probability of release and repletion of the RRP is difficult to directly measure so these remain suggested mechanisms, hence the question mark in the figure. The probability of release remains relatively unchanged at low compared to high frequencies for type I and IIa fibers. In contrast, higher frequencies substantially decrease the probability of release at type IIx/IIb DIAm fibers.

In the frog, Del Castillo and Katz consistently observed neuromuscular transmission failure at the presynaptic terminal, reflected by the absence of a postsynaptic response despite sufficient nerve stimulation. These basic observations were accentuated by exposure to a low ${\mathrm{C}\mathrm{a}}^{2+}$/high ${\mathrm{M}\mathrm{g}}^{2+}$ medium (93). Neuromuscular transmission failure can also result from a failure of the nerve action potential to propagate down all branches of the motor neuron (branch-point failure), a failure to release sufficient synaptic vesicles from the presynaptic terminal, or a failure of the motor endplate to respond to ACh release (e.g., reduced AChR sensitivity; Figure 15). Branch-point failure is frequency dependent and contributes more to neuromuscular transmission failure in the developing neuromuscular system, likely due to polyneuronal innervation, but plays less of a role in mature motor axons (142). With 2-min stimulation of the phrenic nerve at 10 Hz, the incidence of axonal propagation failure is approximately 4% in rats during the first postnatal week but is not observed at all in adult rats (Figure 16). By contrast, at 75 Hz the incidence of phrenic axonal propagation failure rate after 2-min is approximately 66% during the first postnatal week and approximately 26% in adult rats (Figure 16). There are two proposed mechanisms behind branch-point failure, relating to either: (i) transmembrane ${\mathrm{K}}^{+}$ concentration or (ii) axonal geometry (2, 179, 429). As the membrane branches, membrane capacitance decreases while membrane resistance increases in the smaller branches. More importantly, axonal resistance increases reducing the length constant and decreasing the probability of action potential propagation (445, 446). Fatigable motor units typically have higher innervation ratios with each motor neuron having more axonal branches and are therefore more likely to fail. This was tested in the rat DIAm using the technique of glycogen depletion in which repetitive stimulation depletes muscle fiber glycogen. If axonal propagation fails to activate muscle fibers, this is reflected by reduced glycogen depletion. It was observed that the incidence of axonal propagation failure (lack of glycogen depletion) was more evident in type IIx/IIb DIAm fibers and increased with stimulation rate (213). At both high and low frequencies, type IIx/IIb fibers were found to be approximately 30% more susceptible to neuromuscular transmission failure than type I and IIa fibers despite the greater safety factor and QC exhibited by the type IIx/IIb fibers (213).

Figure 15.

Neuromuscular transmission failure (NMTF) can occur due to failures at both the presynaptic and postsynaptic components of the neuromuscular junction (NMJ). At the presynaptic component, NMTF can be subdivided into branch point failure and a failure to release synaptic vesicles, which can be caused by either a decrease in the probability of release or depletion of the synaptic vesicle pools.

Figure 16.

(A) One approach to test neuromuscular transmission failure (NMTF) is to repeatedly stimulate motor axons and compared the evoked forces to those evoked by intermittent direct muscle stimulation. NMTF is quantified as the difference between the amplitude of the forces evoked by direct muscle vs nerve stimulation. In the rat diaphragm muscle (DIAm), the extent of NMTF is dependent on the frequency of phrenic nerve stimulation such that at lower frequencies (B. 10 Hz) there is less divergence in the forces evoked by direct muscle versus nerve stimulation (i.e., less NMTF) than at higher frequencies (C. 75 Hz) (213).

At higher frequencies (40, 75, and 80 Hz), neuromuscular transmission failure could be due to both a failure to release synaptic vesicles or AChR desensitization. Synaptic vesicle release can fail due to a decrease in either the available pool of readily releasable synaptic vesicles or the probability of synaptic vesicle release. With repetitive stimulation, the probability of synaptic vesicle release decreases ((389); Figure 15). The fidelity in the propagation of action potentials along the axon may also relate to the level of myelin surrounding the axon, with deficits in myelination increasing the likelihood of axonal propagation failure (22, 149, 150, 412). In the continued presence of ACh, AChRs desensitize resulting in a reduced depolarizing current (224). There is no evidence that AChR differs across fiber types but the presence of ACh in the synaptic cleft increases with repeated stimulation at higher frequencies. Therefore, AChR desensitization is unlikely to explain the differences in neuromuscular transmission failure across fiber types.

A common paradigm to test neuromuscular transmission failure is to compare the force response evoked by nerve stimulation to that produced by direct muscle stimulation (bypassing the nerve) (Figure 16) (4, 28, 238). In previous studies using this technique, we reported a number of conditions that affect neuromuscular transmission failure at the DIAm in adult rats and mice, including aging (131, 175, 178), cardiopulmonary bypass (114), testosterone (45), corticosteroids (430), and hypothyroidism (350).

As mentioned above, we found that the neonatal rat DIAm is far more susceptible to neuromuscular transmission failure during repetitive activation than the adult (33, 122, 142, 357, 365). Neuromuscular transmission failure in the embryonic and early postnatal DIAm may be due in part, to the more extensive of phrenic axon branching before the elimination of polyneuronal innervation. Failure of action potential propagation at axonal branch points has been well established (Table 1) (142, 234, 236). Additionally, myelination of the motor axons develops over the first several weeks postnatally in mice, which strongly influences both intrinsic electrophysiological properties and action potential propagation (432). It is also possible that there are developmental differences in QC due to fewer synaptic vesicles in the readily releasable pool and differences in SNARE proteins that mediate fusion and exocytosis of synaptic vesicles. In the rat DIAm during the early postnatal period, the frequency and amplitude of spontaneous mEPPs are reduced and the amplitude of evoked EPPs is also reduced as compared to the adult DIAm of mice and rats (142, 304). In addition, neonates have smaller mEPP and EPP amplitudes, either due to fewer synaptic vesicles or fewer receptors (122). Both mEPP amplitude and safety factor are reduced in mouse genetic models with reduced neuregulin expression (402). Heterozygous mice with disruptions in neuregulin show a 50% reduction in AChR but also potentially compensate for fewer receptors through an increase in QC.

Table 1.

Incidence of Axonal Propagation Failure and Extent of Decrement in EPP Amplitude During Repetitive Stimulation at Different Ages (142)

	Stimulation rate (pps)
	10	20	40	75	10	20	40	75
Age	Propagation failure rate (%)				Decrement in EPP amplitude (%)
1 week	4.4±1.7a	14.5±3.5a	40.5±3.8a	65.6±2.2	52.2±10.8	70.0±8.2	87.2±7.7a	91.3±4.4
3 week	0±0	2.0±1.0	16.7±4.2b	37.5±4.8b	63.7±9.7	61.2±5.2	78.5±3.1	77.4±2.8
Adults	0±0	2.5±0.5	4.0±2.4	26.0±5.1	61.3±7.8	62.8±6.2	73.2±2.7	75.1±1.3

Values are mean±SD. Propagation failure rate calculated as the proportion of absent evoked EPPs in the first 10 pulses of each train.

^aSignificantly different from older animals (P < 0.01).

^bSignificantly different from adults (P < 0.05).

Importantly, increased neuromuscular transmission failure occurs with motor unit expansion, whether associated with distal NMJ factors (axon terminals, synaptic cleft, or synaptic vesicles), more proximal disturbances such as branch-point failures, or both. If motor unit expansion occurs during perinatal development, NMJs per se, remain morphologically sound (130), despite marked motor neuron loss (54, 55). In this scenario, increased neuromuscular transmission failure is likely due to increased innervation ratios in the absence of denervation/reinnervation and pre- or postsynaptic NMJ defects (130, 139). Conversely, in scenarios such as age-associated sarcopenia, where motor neuron loss occurs, increased neuromuscular transmission failure (131) is likely associated with motor unit expansion (increased innervation ratio) thereby increasing branch point failures due to spared motor neurons re-innervating denervated fibers (137) and the frank degeneration of pre- or postsynaptic NMJ compartments (131, 354).

Energy requirements of neuromuscular transmission

The consumption of ATP at the muscle fiber relates to countering active and resting membrane currents $\left({\mathrm{N}\mathrm{a}}^{+},{\mathrm{K}}^{+}\right.$-ATPase pump), cross-bridge cycling mechanics, and re-sequestering of ${\mathrm{C}\mathrm{a}}^{2+}$ following activation. Sadly, despite the abundance of information about ${\mathrm{N}\mathrm{a}}^{+},{\mathrm{K}}^{+}$-ATPase pump, and SERCA function in skeletal muscle, including the DIAm, there is a paucity of knowledge regarding the endplate of the NMJ. It is surprising that although many of the conditions that afflict the ${\mathrm{N}\mathrm{a}}^{+},{\mathrm{K}}^{+}$-ATPase pump, and sarcoplasmic reticulum ${\mathrm{C}\mathrm{a}}^{2+}$ transport ATPase (SERCA pump) function have an NMJ phenotype, the ${\mathrm{N}\mathrm{a}}^{+},{\mathrm{K}}^{+}$-ATPase pump, and SERCA at the NMJ have not been the focus of the investigation to date. Similar to their role in other cells, including the presynaptic component of the NMJ (see above), the maintenance of the postsynaptic membrane potential depends on ATP hydrolysis via the ${\mathrm{N}\mathrm{a}}^{+},{\mathrm{K}}^{+}$-ATPase pump. In skeletal muscle, a suite of ${\mathrm{N}\mathrm{a}}^{+}/{\mathrm{K}}^{+}$-ATPase pump isoforms exist, with fatigue-resistant type I and type IIa fibers having increased expression of $\alpha 1$ and $\beta 1$ isoforms and reduced expression of $\beta 2$ isoform compared with type IIx/IIb fibers (146, 202, 462). In a variety of conditions, including aging (511), muscular dystrophy (233), hypoxia, and chronic obstructive pulmonary disease (172, 291), the isoform-specific expression of the ${\mathrm{N}\mathrm{a}}^{+},{\mathrm{K}}^{+}$-ATPase pump is altered. Notably, the ${\mathrm{N}\mathrm{a}}^{+},{\mathrm{K}}^{+}$-ATPase pump does not appear to be concentrated at the NMJ endplate (10), although the fiber type dependence of this remains unexplored.

ATP hydrolysis is also necessary to re-sequester ${\mathrm{C}\mathrm{a}}^{2+}$ into the SR via the SERCA pump, and as SERCA facilitates relaxation its function is associated with muscle fatigue. In skeletal muscle, the predominant isoform is SERCA2A in type I fibers and SERCA1 in type II fibers (265, 306, 497). In DIAm, increased SERCA2A and/or decreased SERCA1 is correlated with improvements in fatigue resistance, while decreased SERCA2A and/or increased SERCA1 isoform expression results in decreased endurance (13, 212). Impairment of SERCA has been shown to occur in circumstances ranging from aging (50, 361, 461), hyperthyroidism (183), immobilization (398, 455), hypoxia and chronic obstructive pulmonary disease (335, 420), ALS (360), and muscular dystrophy (169, 289).

Following a single depolarizing action potential, or a single contraction, the magnitude of ATP consumption is likely to be greater in an individual type IIx/IIb fibers compared with type I and IIa fibers. However, allowing for the major differences in activation history across fiber types in the DIAm, ATP requirements for maintaining membrane potential and sequestering ${\left[{\mathrm{C}\mathrm{a}}^{2+}\right]}_{\text{cyt}}$ are much greater in type 1 and IIa fibers than in type IIx/IIb fibers. Thus, the cumulative rate of ATP consumption across a given time period (as opposed to a unitary single activation) would be much greater in I and IIa DIAm fibers.

Plasticity of NMJs Across the Age Spectrum

Across the aging spectrum, there are major changes in the structure and function of NMJs that affect neuromuscular transmission. During embryonic development, the arrival of motor axons coincides with the formation of muscle fibers from myoblasts and myotubes (myogenesis). During this embryonic and subsequent early postnatal period in mammals, muscle fibers can be influenced by more than one motor neuron (polyneuronal innervation) (Figure 13). In addition, during the perinatal period, there is a marked culling (loss) of motor neurons and elimination of synapses until the adult motor unit is established (Figure 13). On the other end of the aging spectrum, motor neurons are also lost (similar to ALS) and muscle fibers are denervated as a result. Reinnervation of denervated muscle fibers by surviving motor neurons expands motor unit innervation ratio. The mechanisms underlying NMJ plasticity during perinatal development and aging are not fully understood but have been intensely studied over the years.

Myogenesis and innervation coincide during development

In the embryonic rodent DIAm, the phrenic nerve arrives around embryonic day 11 (E11 in mice) and 13 (E13 in rats; Figure 13) (6), at a time when myogenesis is incomplete and the DIAm consists of myoblasts and myotubes. Myogenesis consists of two major phases: a determination phase, when myoblasts emerge from mesodermal progenitor cells, and a terminal differentiation phase, when myoblasts fuse to form myotubes followed by the formation of muscle fibers that express contractile proteins structurally organized in sarcomeres (173). The timing of myogenesis is influenced by the expression of muscle regulatory factors (MRFs), including myogenic differentiation factor 1 (MyoD) (90), myogenin (56), myogenic factor-5 (Myf-5) (57), and myogenic regulatory factor-4 (MRF4) (369). These MRFs are also involved in regulating MyHC isoform expression and the formation of adult muscle fiber types (200). The process of myogenesis and formation of adult fiber types is not complete until the emergence of ${\mathrm{M}\mathrm{y}\mathrm{H}\mathrm{C}}_{2\mathrm{X}}$ and ${\mathrm{M}\mathrm{y}\mathrm{H}\mathrm{C}}_{2\mathrm{b}}$ isoforms in type IIx/IIb fibers. In the rodent DIAm, type IIx/IIb fibers do not emerge until after the second postnatal week, and even then, type IIx/IIb fibers display differential growth into adulthood. Thus, the period between E11-E13 and postnatal day 14 (P14) represents a critical time for the organization of motor units that are essential for the diversity of muscle functions.

In myogenesis, myoblast fusion forms multinucleated myotubes and is essential for the subsequent formation of muscle fibers with expression of contractile proteins organized into sarcomeres. It is important to note that nerve outgrowth and arrival at the muscle are coincident with the critical period of contractile protein organization during myogenesis (275). However, the terminal differentiation of myoblasts and the formation of myotubes and muscle fibers are not dependent on innervation. Using rat DIAm injected with $\beta$-botulinum toxin at E14, Harris concluded that primary myotubes can develop absent from any support by the nervous system; however, secondary myotube development depends on innervation (185). The transition from primary to secondary myotubes is a staggered process and innervation of the burgeoning secondary myotubes can occur through the transfer of a nerve terminal from a more densely populated primary myotube (102). Developing myofibrils must contract during the development or generation of secondary myotube ceases (185). Myotube and muscle fiber formation are primarily influenced by mechanical properties that affect protein expression in the myoblast cell membrane (e.g., a1-integrin), the cytoskeleton (e.g., actin and desmin), the basal lamina (e.g., muscle cell adhesion molecule—M-CAM), and the extracellular matrix (e.g., fibronectins, laminins, M-cadherin, and neural cell adhesion molecule—N-CAM) (228, 232). Thus, the terminal differentiation of myoblasts and their subsequent fusion to form multinucleated myotubes and muscle fibers involves complex processes.

During late embryonic and early postnatal development, the expression of MyHC isoforms in muscle fibers undergoes substantial transition with the early expression of ${\mathrm{M}\mathrm{y}\mathrm{H}\mathrm{C}}_{\text{Slow}}$ and ${\mathrm{M}\mathrm{y}\mathrm{H}\mathrm{C}}_{2\mathrm{A}}$ isoforms together with embryonic $\left({\mathrm{M}\mathrm{y}\mathrm{H}\mathrm{C}}_{\mathrm{E}\mathrm{m}\mathrm{b}}\right)$ and neonatal $\left({\mathrm{M}\mathrm{y}\mathrm{H}\mathrm{C}}_{\mathrm{N}\mathrm{e}\mathrm{o}}\right)\mathrm{M}\mathrm{y}\mathrm{H}\mathrm{C}$ isoforms (275). During this time, the size of motor neurons innervating muscle fibers is relatively homogeneous (132). After the second postnatal week in rodents, expression of ${\mathrm{M}\mathrm{y}\mathrm{H}\mathrm{C}}_{2\mathrm{X}}$ and ${\mathrm{M}\mathrm{y}\mathrm{H}\mathrm{C}}_{2\mathrm{B}}$ isoforms emerges, even in the absence of innervation (424). Singular expression of MyHC isoforms does not occur in rat DIAm fibers until the second postnatal week, and the final adult proportions of fiber types do not appear until the fourth postnatal week (157). However, after the second postnatal week in the rat DIAm, there is an approximately 2- to 3-fold growth of type IIx/IIb fibers compared with type I and IIa fibers (214, 349). With the postnatal emergence in type IIx/IIb DIAm fibers, there is a corresponding increase in specific force (force per cross-sectional area) and velocity of shortening (157). If the DIAm is denervated before the second postnatal week, the differential growth of type IIx/IIb fibers is blunted, suggesting a trophic influence of innervation. The period of postnatal growth of type IIx/IIb fibers in the rat DIAm also corresponds with an increase in size of the phrenic motor neurons (351, 356).

Polyneuronal innervation and synapse elimination during perinatal development

During embryonic and early postnatal development, each muscle fiber can be innervated by more than a single motor neuron—polyneuronal innervation (206, 320, 491) (Figure 13). Early observations regarding polyneuronal innervation were based on intracellular electrophysiological recordings of EPPs, where EPPs were complex and consisted of 2 to 4 “units” distinguishable by time and amplitude (365). Upon arrival of the phrenic nerve in the rat DIAm at E13 in rats, motor axons branch extensively but only start to form synapses after about 24 h (36, 173, 186). The reason for the short delay in branching and initial synapse formation is unclear, but it may be that cervical motor neurons are not electrically excitable and capable of generating and transmitting action potentials until E14 (256). In the DIAm, electrical stimulation of the phrenic nerve evokes an AChR-dependent postsynaptic response by E14 (182). The pattern of intramuscular axonal branching establishes a somatotopic organization in which phrenic motor neurons from higher cervical segments innervate the sternal region and more ventral aspects of the costal and crural regions (145, 249).

Emerging motor axons are accompanied by Schwann cells, which may be involved in the guidance and outgrowth of axons (123). Both Schwann cells and nerve terminals display coated vesicles suggesting that there may be the vesicular release of trophic factors (see below). Among the regulatory processes that may be involved are those associated with properties of the extracellular matrix, the release of trophic factors, and/or chemotactic substances. Insight into axonal outgrowth and branching derives from in vitro studies in rats and zebrafish, which suggest that the perineural sheath functions as a scaffold for targeting motor axons toward muscle fibers (237, 270, 463). Unfortunately, exploring these processes under in vivo conditions is difficult.

Trophic influences derived from both motor neurons (neurotrophic) and maturing muscle fibers (myotrophic) appear to be important in matching innervation with muscle fiber development (140, 141, 327, 334, 477, 498). During development, a complex interplay between programmed motor neuron cell death and the development of muscle fibers and their nascent innervations occurs in the third week of gestation (26, 309, 317, 457). Spontaneous motor unit activity has been proposed to regulate motor neuron survival and innervations in two ways: first, by the activity-dependent production of trophic factors (326), and second, by the regulation of neuromuscular synapse number, in line with Hebbian concepts (247). Experimentally, when in utero motor neuron pools have reduced activity (27, 87), excessive axonal branching ensues and an increased profligacy of nascent presynaptic terminals is observed (27, 87, 140, 141, 328, 329, 332, 333, 345). Conversely, in cases with increased motor neuron activity (27, 133, 134, 331), axonal branching and the number of NMJs are markedly restricted (27, 140, 141, 331).

In cases where neurotrophins may bind to multiple receptors, for example, BDNF binding to its high affinity TrKB receptor as well as the lower affinity p75NTR receptor, the overall signaling effect of BDNF depends on the concentration of BDNF and the relative expression of each receptor (319) (Figure 17). Notably, for BDNF, receptor-dependent responses involve the potentiation of canonical trophic effects via TrKB or the triggering of apoptosis by p75NTR (23, 32, 319, 341). There is a striking decline in the abundance of p75NTR receptors in motor neurons during embryonic development corresponding to the period of programmed cell death of motor neurons (101, 151–153). At birth and thereafter the expression of the p75NTR receptor in motor neurons is at a very low level. The developmental importance of the interplay between BDNF and p75NTR receptors is underscored by p75NTR knock-out mice, which exhibit delays in presynaptic terminal maturation and neurotransmission efficacy (207, 342). Interestingly, p75NTR receptors are expressed again in adult motor neurons as a result of injury (axotomy) and diseases afflicting motor neuron axons, such as ALS (215, 413). It remains to be determined if the adult expression of p75NTR contributes to the phenotype of developmental disorders of motor units (e.g., early-onset spasticity), other scenarios where motor neuron denervation occurs (e.g., DIAm sarcopenia), or if there is a motor unit type-dependence to p75NTR expression during pathological conditions.

Figure 17.

Neurotrophin 4 (NT-4) and brain-derived neurotrophic factor (BDNF) are predominantly produced and released by motor neurons, although there is some evidence suggesting a myogenic pathway for these neurotrophins. Both NT-4 and BDNF bind to a high affinity tropomyosin kinase B (TrkB) receptor located both pre- and postsynaptically. Some evidence suggests that TrkB and BDNF in the bound state are endocytosed and transported retrogradely via the axon to provide signaling to the soma. Republished, with permission, from Mantilla CB and Sieck GC, 2003 (274).

Initially, each myotube/myofiber can be contacted by multiple motor neurons (polyneuronal innervation). Subsequently, polyneuronal innervation disappears through the process of motor nerve terminal withdrawal and pruning of axonal branches, rather than complete degeneration of motor neurons, which in the mouse and rat DIAm is completed by about the second postnatal week (35, 36, 61, 344, 403). The rate of synapse elimination is initially rapid but slows until only one remains (459). The process of synapse elimination is not fully understood, but it has been suggested that competition among motor neurons for target cell innervation depends on activity and/or the fidelity of neuromuscular transmission (Hebbian competition) (29, 38, 39, 41, 63, 64, 120, 121, 206, 344, 365, 385). Accordingly, nerve terminals of more active motor neurons might persist, while those of less active motor neurons are eliminated. Synapse elimination is affected by imposed increases or decreases in neuromuscular transmission such that reducing activity also reduces the rate of synapse elimination and patterned high-frequency stimulation accelerates synapse elimination (321, 460, 464). However, the simple theory of activation does not support the fact that in adults, motor units with the largest innervation ratio (i.e., number of muscle fibers innervated by a single motor neuron) are often those that are least active (e.g., type IIx/IIb fibers comprising fast-fatigable motor units) (48, 144, 423). It is likely that the activity patterns of motor neurons change markedly during postnatal development with the growth of motor neurons and changes in intrinsic electrophysiological properties dictating excitability. For example, myelination of motor neurons, which has profound effects on electrophysiological properties, is not complete until the second to third postnatal week in mice (432). The most dramatic period of motor neuron growth occurs after synapse elimination is complete (i.e., after the second postnatal week in rats) (349). Another possibility is that the preservation of a presynaptic terminal depends on the fidelity of neuromuscular transmission. It has been shown that the incidence of neuromuscular transmission failure is much higher in the neonatal rat DIAm as compared with the adult (142). Therefore, presynaptic terminals from those phrenic motor neurons that fail to elicit a postsynaptic response are eliminated consistent with Hebbian competition. However, what underlies the precise matching of motor neurons and their presynaptic terminals with muscle fiber types is still unknown. The pattern and fidelity of activation may affect the secretion of synaptotrophins or synaptotoxins that contribute to the maintenance or elimination of synapses (403). Another possibility is that muscle fibers use intracellular signals called synaptomedins to maintain contact with effective nerve terminals, which may or may not be dependent on activity (21). At present, none of these possibilities can be excluded, and it is likely that a combination of mechanisms is responsible for the final pattern of motor unit innervation. However, since the properties of motor neurons and muscle fibers are closely linked, it is likely that the mechanisms underlying synapse elimination are also linked to muscle fiber lineage. Ultimately, the motor unit composition may also be affected by ongoing postnatal myogenesis and the formation of new myotubes and muscle fibers, which occurs until the third postnatal week in the rat DIAm (349). Thus, the final innervation ratio of motor units is not established in the rat DIAm until about postnatal day 28.

In rats, respiratory movements have been observed at E17 (174), which indicates that functional synapses and neuromuscular transmission are present at least within a few days after initial innervation. In the embryonic DIAm, we observed that transient cytosolic ${\mathrm{C}\mathrm{a}}^{2+}\left({\left[{\mathrm{C}\mathrm{a}}^{2+}\right]}_{\text{cyt}}\right)$ and contractile responses were evoked by electrical stimulation of the phrenic nerve as early as E12.5 in mice and E14 in rats. Thus, neuromuscular synapses are functional soon after the arrival of the presynaptic terminal. Importantly, the cytosolic ${\mathrm{C}\mathrm{a}}^{2+}$ and contractile responses were blocked by d-tubocurarine and $\alpha$-bungarotoxin indicating dependence on ACh release and activation of AChRs. Yet, in the embryonic rat DIAm just after initial innervation, there is no apparent postsynaptic endplate with limited clustering of AChRs until E17. The weak evoked contractile response in DIAm myotubes/myofibers is consistent with the emergence of sarcomeric organization of contractile proteins that are necessary for a mechanical response. Thus, phrenic nerve activity can elicit a mechanical response even before well-defined endplates are present. These primitive mechanical responses may be important in the maturation of myotubes into DIAm fibers (i.e., sarcomeric organization of contractile proteins and their alignment in myofibers).

Motor neuron loss in old age—Denervation and reinnervation

Neural plasticity refers to the ability of the nervous system to adapt to new and repeated experiences. While work in synaptic plasticity began in the field of psychology with Donald Hebb in the 1940s, his observations that repeated stimulation of a presynaptic neuron could facilitate the response in the postsynaptic neuron have had profound impact within the realm of neurophysiology (478). Plasticity is important during development as it reinforces and maintains useful synapses, ensuring a parsimonious neural network. During adulthood, plasticity allows the NMJ to adapt to environmental changes and stressors. Remodeling of the NMJ in response to degeneration, whether due to injury, disease, or aging, can restore or maintain neuromuscular function.

Aging is associated with a loss of motor neurons resulting in the denervation of muscle fibers (137, 467). We have found that age-related loss of phrenic motor neurons was restricted to larger motor neurons that innervate type IIx/IIb DIAm fibers in rats (137). The loss of motor neurons mandates NMJ reinnervation of surviving muscle fibers (131, 354). With reinnervation, there is increased axonal branching and de novo regeneration of presynaptic terminals at these fibers. There is greater fragmentation of presynaptic terminals associated with reinnervation in both rodent and human studies (84, 243, 322, 354, 379, 485), which is associated with spatial uncoupling between the active zone and receptor location across the synaptic cleft (131, 253, 354, 378). There have been similar reports of an age-related increase in complexity of nerve terminal arborization at type IIx/IIb fibers (353, 354). The age-related increase in the complexity of presynaptic nerve terminals results in an increase in presynaptic terminal surface area with aging, which may not improve neuromuscular transmission (354). Unfortunately, age-related changes in the ultrastructure of presynaptic terminals have not been fully characterized. However, the increased fragmentation of the NMJ in DIAm of older mice was not associated with any change in QC but was associated with a greater endplate current amplitude after repeated stimulation at ≤5 Hz (488). Similarly, the number of active zones is decreased with aging in presynaptic terminals of motor neurons innervating the sternomastoid muscle in mice (74). It is unknown whether there are age-related changes in SNARE protein mediated vesicle release and/or synaptic vesicle recycling. Such changes may compromise QC and the fidelity of neuromuscular transmission in old age. Indeed, increased DIAm neuromuscular transmission failure in older animals has been reported (131, 175).

Synaptic plasticity is modulated through neurotrophic signaling pathways. Neurotrophins are released by both neurons (neurogenic) and muscle fibers (myogenic; Figure 13) (509). Two of the most researched neurotrophins are BDNF and NT-4. Both BDNF and NT-4 enhance neuromuscular transmission during repetitive stimulations (278). Evidence suggests that both BDNF and NT-4 act through the high-affinity TrkB receptor. Indeed, inhibition of TrkB kinase reverses the facilitation of neuromuscular transmission in the presence of BDNF and NT-4. In aging, inhibition of the TrkB receptor leads to reduced overlap of presynaptic terminals and endplate regions and increases the proportion of denervated NMJs by approximately 20% (177).

Glial-derived neurotrophic factor (GDNF) also has profound effects on motor neurons. GDNF knockout mice exhibit 20% to 30% fewer motor neurons as well as altered axonal branching patterns (181). In contrast, overexpression can lead to increased motor neuron survival during development and extensive axonal sprouting patterns in mice (315, 330). Finally, neural growth factor (NGF) can augment evoked potentials, depolarize postsynaptic resting membrane potential, and limit the effects of AChR blockers.

Evidence of myotrophic signaling is provided by studies using nerve-muscle co-cultures from cloned cells and frog NMJs. Research with nerve-muscle co-cultures observed clustering of synaptic vesicles and synaptic vesicle-associated antigens closest to muscle tissue but not when the neurites were grown near other types of tissue suggesting a muscle-specific effect (264). There is also evidence that when motor neurons reinnervate muscle fibers, they do so at preexisting junctional folds as observed in the cutaneous pectoris of frogs (405), which provides muscle-derived signaling at the endplate. Myotrophic substances such as MyoD, myogenin, and MRF-4 have been implicated in the determination of fiber type (274). However, both MyoD and myogenin are also affected by innervation, thus blurring the distinction between myotrophic and neurotrophic influences (201).

Aggregation of AChRs at postsynaptic endplates during development

The progression in expression and aggregation of AChRs is concurrent with the outgrowth of motor axons and presynaptic contact with myotubes/muscle fibers, around E16 in mice (288). In rat DIAm, mEPPs and EPPs are measured as early as E14 (95), providing evidence for functional synaptic transmission at the NMJ. In the mouse DIAm, aggregation and clustering of AChRs are apparent at approximately E13. Results from studies on chick embryos have found that AChR aggregation occurs within a few hours of contact with motor axon terminals (147). Postnatally, the density and clustering of AChRs at endplates change from a plaque-to-pretzel morphology in mammals (281). The process of aggregation of AChRs has received considerable attention and critically involves agrin secretion by the nerve terminal (Figure 16). Agrin is a nerve-derived glycoprotein that binds to MuSK receptors on the muscle fiber membrane and activates a rapsyn-mediated clustering of AChR (58, 241, 275, 314, 404). Activation through intracellular pathways allows a single agrin molecule to cluster many AChRs (290). MuSK phosphorylation appears to be critical for AChR clustering, as clustering is absent in MuSK double mutant mice and in mice treated with an siRNA designed to block MuSK transcription (229). The expression of MuSK receptors varies along the length of muscle fibers, being most abundant in their central regions where MuSK can undergo autoactivation (267), leading to AChR clustering in the absence of neural influence. In neighboring sections of the muscle fiber, where MuSK is not as richly expressed, agrin released by motor axons can also lead to AChR clustering (260). Importantly, nerve-derived agrin release by presynaptic terminals and MuSK expression in muscle fibers appear to coordinate the formation of pre- and postsynaptic components of the NMJ and account for the apparent “targeting” of motor axons to “preformed” AChR clusters (481, 502), as well as the induction of new clusters by nerve-derived influences (211). Agrin is important not only for the development of the endplate but also for the development of mechanisms underlying excitation-contraction coupling in myotubes including the expression of ryanodine receptors and voltage-gated L-type ${\mathrm{C}\mathrm{a}}^{2+}$ channels (24, 456).

Another important neurotrophic influence is that of neuregulin, which is a member of the larger epidermal growth factor family (116, 402, 472, 512). Neuregulin mediates its influence through ErbB receptors to activate tyrosine kinases in maturing muscle fibers (251). Neuregulin is both a nerve-derived and muscle-derived trophic influence for myotube-to-muscle fiber differentiation during embryonic and early postnatal development, but the interactive roles of different trophic factor signaling pathways (e.g., agrin vs. neuregulin) in the emerging NMJ remains to be determined.

Following AChR aggregation, synaptic folds from and complex endplate structures differentiate with characteristics of adult NMJs in the rat DIAm. At birth, AChRs appear in irregular patches at an endplate that more closely resembles an oval or “spoon” (281). Within a few days, the endplate begins to form gutters (junctional folds), and AChRs not located at the peaks of junctional folds are degraded. Previously, we demonstrated that in the adult rat DIAm, endplate morphology varies across different fiber types, being far more complex in type IIx/IIb fibers compared with type I and IIa fibers (271, 352). This is consistent with work in other mammalian muscles (323, 443). By contrast, postsynaptic motor endplates in the fetal and neonatal DIAm are much smaller and far less complex compared with adults. The progression of endplate growth with development tightly matches muscle fiber growth (20). There is very little information regarding the mechanisms responsible for the development of fiber type-specific differences in endplate morphology (276). Importantly, type IIx/IIb fibers in the rodent DIAm do not emerge until after the second postnatal week (277). Furthermore, type IIx/IIb DIAm fibers display faster postnatal growth compared with the other fiber types beginning at the third postnatal week into adulthood (277). With an increase in motor endplate complexity, there is an increase in surface area with a greater number of AChRs to elicit an endplate potential. It is likely that the mechanisms underlying the growth of type IIx/IIb fibers and the increasing complexity of motor endplates are linked.

Changes related to aging in postsynaptic endplates

At the other end of the lifespan, there are also significant changes at the endplate that predominantly affect type IIx/IIb fibers (354). It is possible that trophic influences that affect normal remodeling are altered with aging (131, 354). It should be pointed out that an increase in the size of endplates does not necessarily correlate with improved synaptic efficacy. A possible reflection of the instability of NMJ remodeling may be the increased expression of neural cell adhesion molecule (NCAM) that occurs with aging (439). NCAM is involved in pre- and postsynaptic communication at the NMJ (76). Curiously, despite the increased prevalence with age, the NCAM response after denervation is suppressed with old age, and therefore, researchers have suggested that an impaired ability to appropriately modulate NCAM may be responsible for failed reinnervation potential (not all muscle fibers are successfully reinnervated) (165). In addition, there is an age-related gradual decrease in the number of AChRs at the motor endplate and increased extra-junctional receptors in the sternomastoid and DIAm of aged rats (84, 436, 437). In addition, the endplate increases in size and complexity on type IIx/IIb fibers but not on type I or IIa fibers in aged rats and mice (354, 388). Endplate growth is also paired with AChR disaggregation (205). Additionally, the voltage-gated sodium channels exhibit slower activation and inactivation kinetics and there is an increase in extrajunctional ${\mathrm{N}\mathrm{a}}^{+}$ current (96). Disordering of the laminar structures of the NMJ also impairs neuromuscular transmission of the DIAm, which is exacerbated in aging (253). The effects of motor unit type on disaggregation and changes in voltage-gated channel physiology have not been explored.

Conclusion

The NMJ is the sole interface between the motor control centers (the motor neuron pools and their associated presynaptic circuits) and skeletal muscle fibers. As such, the NMJ is a potential Achilles’ heel for the uninterrupted and orderly activation of motor units. Many of the redundancies of the NMJ structure and function attempt to lessen the likelihood of signal transduction failure (see the sections on the readily releasable pool, vesicular recycling, and the safety factor). When these mechanisms fail, or if the motor neuron innervating muscle fibers dies and reinnervation and remodeling are impaired or incomplete, a suite of NMJ dysfunctions may occur. Often, these pathophysiological changes are specific to a subset of motor unit types, with selective vulnerability of FInt and FF motor units, as evidenced in aging and ALS. By contrast, myasthenia gravis is not selective, with pathology across multiple motor unit types.

Throughout this article, we have highlighted the plasticity and remodeling relation between the presynaptic and postsynaptic elements of motor units as an organizing principle of the NMJ. These relations extend beyond structural maintenance and include the coordination of neuro- and myotrophic factors, the maintenance of vesicular availability to demand for release, and the balance of ATP availability to consumption. Key advances in the field will likely be related to the motor unit type-specifics of these phenomena, the importance of which has recently been unveiled in concert with disease-associated studies (e.g., Schwann cells and the tripartite synapse in ALS). Such gains in understanding may also serve as a template for efforts focused on the central synapses within the CNS.

We conclude that with the ever-increasing array of experimental tools at our disposal, the molecular underpinnings of state-dependent NMJ functions (e.g., the paradoxical effects of BNDF and NMJ autophagy via the p75NTR receptor) will be unveiled. These discoveries will hopefully usher in a new era of NMJ therapeutics for a variety of developmental conditions and degenerative diseases.

Didactic Synopsis.

Major teaching points

• The information contained in this article will help with teaching NMJ physiology using different motor unit/muscle fiber types as an organizing principle.

• A detailed section on motor unit physiology provides insight into the unique mechanical and fatigue properties displayed by different muscle fiber types and how they are activated via neuromuscular transmission.

• The NMJ comprises the presynaptic axon terminal, the postsynaptic receptor region at the muscle fiber, and the perisynaptic (terminal) Schwann cell that provides support and is essential for maintenance of the NMJ.