Synaptic silencing of fast muscle is compensated by rewired innervation of slow muscle

See allHide authors and affiliations

Science Advances  08 Apr 2020:
Vol. 6, no. 15, eaax8382
DOI: 10.1126/sciadv.aax8382


For decades, numerous studies have proposed that fast muscles contribute to quick movement, while slow muscles underlie locomotion requiring endurance. By generating mutant zebrafish whose fast muscles are synaptically silenced, we examined the contribution of fast muscles in both larval and adult zebrafish. In the larval stage, mutants lacked the characteristic startle response to tactile stimuli: bending of the trunk (C-bend) followed by robust forward propulsion. Unexpectedly, adult mutants with silenced fast muscles showed robust C-bends and forward propulsion upon stimulation. Retrograde labeling revealed that motor neurons genetically programmed to form synapses on fast muscles are instead rerouted and innervate slow muscles, which led to partial conversion of slow and intermediate muscles to fast muscles. Thus, extended silencing of fast muscle synapses changed motor neuron innervation and caused muscle cell type conversion, revealing an unexpected mechanism of locomotory adaptation.


It is generally accepted that the skeletal muscle in vertebrates consists of two types of muscle fibers: slow and fast (1). An intermediate type, with mixed characteristics of slow and fast fibers, is also reported in multiple species (2, 3). The functional roles of fast and slow muscle in locomotion have been conjectured based on classical studies (4). With the advent of molecular techniques, mutants that fail to develop fast or slow muscles were analyzed (57). These mutants enabled analysis of locomotor activities under free-moving conditions without recording electrodes. These studies (57) corroborated that fast muscles contribute to quick movement, while slow muscles underlie certain types of locomotion requiring endurance. However, a combination of anatomical and metabolic abnormalities accompanying genetic mutations complicated the interpretation of results. Moreover, these mutants died prematurely, and the long-term effects of silencing fast or slow muscle fibers remained unexplored.

In the present study, we addressed these problems using genome-editing techniques for genes expressed in synapses. Zebrafish as a model animal have various advantages in studying synapses in slow and fast muscles (8, 9). Slow and fast muscle cells in zebrafish are spatially segregated and can easily be distinguished by their anatomical and histological characteristics (10, 11). Acetylcholine receptors (AChRs) in the neuromuscular junction (NMJ) are nicotinic, as opposed to muscarinic AChR receptors that work though G proteins (12). Nicotinic AChRs are pentamers composed of 2α1s, β1, δ, and ε (or γ) subunits, which are all specific to skeletal muscles (13). In zebrafish as well as in mammals, γ subunits are expressed in larvae and change to ε as the animal matures (14, 15). Recent studies showed that AChRs in slow muscles of zebrafish lack ε/γ subunits and are composed of only α, β, and δ subunits (13). To further explore the significance of these newly identified AChR compositions, we generated knockout (KO) zebrafish lines that lacked γ and ε subunits. By analyzing locomotion and synaptic traits in these mutants, we investigated the functional contribution of fast and slow muscles in both larval and adult zebrafish.


γ/ε AChR subunit KO zebrafish lines

We generated a γ subunit gene KO zebrafish (γKO) using CRISPR-Cas9 (Fig. 1A) and an ε subunit gene KO zebrafish (εKO) using transcription activator–like effector nucleases (TALEN) (Fig. 1A). The γKO zebrafish did not show obvious phenotypes during development and matured in a fashion indistinguishable from wild-type (WT) siblings (fig. S1). In contrast, εKO fish generally failed to form swim bladders, and most of them died prematurely within 2 weeks after fertilization. However, a fraction of εKO fish (approximately 25%) survived to adulthood. A double KO (DKO) line was generated by crossing γKO and εKΟ lines. DKO larvae also failed to form swim bladders (Fig. 1B) and died within 2 weeks after fertilization.

Fig. 1 Generation of γ/ε DKO zebrafish.

(A) Schematic diagram of targeted genes. Arrowheads indicate targeted regions of genome editing. Each box and line indicates an exon and an intron, respectively. Alignment of genomic DNA sequences of WT and KO lines showed a 7–base pair (bp) insertion in the AChR γ subunit gene chrng and a 1-bp insertion in the AChR ε subunit gene chrne. (B) Photograph showing WT and γ/ε DKO larva at 6 dpf. Notice the lack of swim bladder (arrowheads) in DKO. Scale bar, 1 mm. (C) Trunk regions of a WT larva (6 dpf) and a DKO larva (6 dpf) were stained with α-BTX conjugated with Alexa Fluor 488 (green). In WT, AChRs were distributed in myoseptal regions (arrows) and in punctae in middle regions (arrowhead). DKO had α-BTX signals only in myoseptal regions. Scale bars, 100 μm.

We histologically analyzed the expression of AChRs in the trunk region of 6 days post-fertilization (dpf) larvae by using α-bungarotoxin (α-BTX) conjugated with Alexa Fluor 488, a toxin that specifically binds to the assembled AChR (Fig. 1C). AChR clusters in DKOs were observed only in boundary regions between body segments (Fig. 1C), where slow muscles form NMJs (16). We initially expected that AChRs in fast muscles of DKO larvae would convert to the slow muscle–type AChRs, comprising only α, β, and δ subunits. This conversion of subunit composition would not cause a change in AChR distribution visualized by α-BTX, because both types of AChRs bind to α-BTX. However, α-BTX signals were absent in fast muscles, which suggested that fast muscles could not express AChRs composed of α, β, and δ subunits.

Synaptic transmission at NMJs of fast and slow muscles

To correlate the AChR expression pattern observed by the α-BTX staining with the synaptic function, we analyzed synaptic activities of fast and slow muscles in the DKO line at 6 dpf. We recorded spontaneous synaptic currents from muscle cells using the whole-cell patch clamp technique (Fig. 2, A to C). Traces show miniature endplate currents (mEPCs) from muscles of WT or DKO larvae (Fig. 2A). Slow muscles in the DKO line exhibited mEPCs. The frequency (14.5 ± 3.1 Hz in WT, 15.5 ± 3.2 Hz in DKO) and the amplitude of slow muscle mEPCs (260.0 ± 74.1 pA in WT, 491.7 ± 105.2 pA in DKO) showed no differences between WT and DKO lines (Fig. 2, B and C). However, fast muscles in DKO failed to produce mEPCs. To confirm that the lack of mEPCs is caused by the absence of functional receptors, we recorded currents in muscles generated by puff application of ACh (Fig. 2D). While fast muscles in WT larvae showed ACh-induced currents (756.4 ± 138.6 pA), those in DKO larvae failed to show any response (0 ± 0 pA). These results, in conjunction with the α-BTX staining (Fig. 1C), showed that fast muscles of DKO larvae do not express any AChRs and receive no synaptic input.

Fig. 2 Synaptic functions of the DKO line.

(A) mEPC traces from fast or slow muscles of WT and DKO larvae (6 dpf) by whole-cell patch-clamp recordings. Fast muscle cells in DKO failed to exhibit mEPCs. (B and C) Frequencies (B) and amplitudes (C) of mEPCs were plotted for each muscle (n = 8 cells). (D) Representative traces of voltage-clamped slow and fast muscles in DKO larvae in response to the application of 30 μM ACh. Calibration: 1 s, 500 pA. Amplitudes of ACh-induced currents in slow (n = 7 cells) and fast muscles (n = 7 cells) are shown. Each dot represents a muscle cell. (E) Construct used for Ca2+ imaging. Top: The GCaMP7a coding sequence was fused to the promoter region of the α-actin promoter pαact. Bottom: Schematic illustration showing the experimental procedure. The gene construct was injected into eggs of DKO at the one cell stage. Ca2+ response was analyzed at 6 dpf. Representative traces showing the increase of ΔF/F in a fast muscle (black line) and a slow muscle (red line) during spontaneous contractions. (F) Overexpression of the ε subunit fused with an EGFP (ε-EGFP) in WT (3 dpf). Top panels: ε-EGFPs were expressed under the control of a slow muscle–specific promoter, psmyhc. EGFP signals (green), expressed in the superficial slow muscles, filled the cytoplasm and did not colocalize with α-BTX (magenta) signals. Bottom panels: ε-EGFPs were expressed under the regulation of pαact. In deeper layer fast muscles, the clusters of EGFP and α-BTX colocalized (arrowheads). Scale bars, 50 μm.

We performed in vivo Ca2+ imaging in the DKO larvae at 6 dpf to further support the result of synaptic current recordings. We designed a gene construct in which a pan-muscle promoter, α-actin promoter, drives the expression of a Ca2+ indicator, GCaMP7a (17), and injected the construct into fertilized eggs (Fig. 2E). In DKOs, we recorded Ca2+ response associated with spontaneous locomotion activities, induced by the application of N-methyl-d-aspartate (50 μM) (18). The results showed that slow muscle cells exhibited Ca2+ transients, while fast muscle cells did not generate any Ca2+ response.

Considering that fast muscles do not allow composition of α, β, and δ subunits, we next examined whether slow muscles conversely allow incorporation of ε subunits in the AChR pentamer, by overexpressing the ε subunit in slow muscles. We designed a gene construct that expressed an ε subunit fused with enhanced green fluorescent protein (ε-EGFP) under the regulation of a slow muscle–specific promoter, psmyhc (19). We injected the construct into fertilized WT eggs and observed the expression of EGFP at 3 to 4 dpf. EGFP signals typically filled the cytoplasm of the slow muscle cells and never colocalized with α-BTX signals (Fig. 2F). In a control experiment, in which ε-EGFP was driven by the pan-muscle promoter (α-actin promoter), the ε-EGFP signals made clusters in fast muscles, colocalizing with α-BTX signals in deeper layers of the trunk region where fast muscles form NMJs. Together, fast muscles and slow muscles express specific types of AChR, and the alternate composition of subunits is prohibited.

Analysis of locomotion in larvae

To examine how silencing of synapses in fast muscles affect locomotion, we next analyzed swimming of WT and DKO larvae at 6 dpf. We induced escape responses by gentle tactile stimuli. Locomotion was recorded with a high-speed camera, and we measured angles between head and tail trajectories throughout each escape response (Fig. 3A and movie S1). WT fish turned their heads 120° to 140° in the initial stage of escape. The typical startle response of teleosts generally begins with a large turn of the head (termed C-bend), followed by a robust forward propulsion as described in previous studies (20).

Fig. 3 Locomotion of WT and KO lines.

(A) Escape behaviors in WT and DKO lines at 6 dpf in response to tactile stimuli. Images of representative larva on the left show superimposed frames of the complete escape response (the duration of movement is indicated in the top right corner). Scale bars, 2 mm. Kinematics for representative traces of 10 larvae are shown for the initial 50 ms of the response. Middle panels represent averaged traces. In the right panels, each trace represents a different larva. Body angles are shown in degrees, with 0 indicating a straight body, and positive and negative values indicating body bends in opposite directions. Scale bars, 10 ms. (B to D) Maximum turn angles, time to reach the maximum angle, and post-startle swimming speed were calculated for each group of fish (6 dpf). In DKO, the turn angle and the swimming speed were notably reduced, and it took longer to reach maximum angles (n = 10 fish). (E and F) Analyses of spontaneous locomotion. Images of representative larva (left) for WT or DKO showed superimposed frames of spontaneous swim bouts (the duration of movement indicated in the bottom right corner). Swimming speed was calculated for WT (n = 5 fish) and DKO (n = 5 fish), which showed no significant difference. Scale bars, 2 mm.

The initial turns of the DKO larvae were in sharp contrast to WT. Averaged maximum head turn angles in DKOs were markedly smaller compared to WT larvae (116.0 ± 5.8° in WT, 20.2 ± 4.0° in DKO; P < 0.001) (Fig. 3B), and time to reach the maximum angle was increased (8.7 ± 0.2 ms in WT, 15.8 ± 0.8 ms in DKO; P < 0.001) (Fig. 3C). In addition to the absence of C-bends, the post-startle swimming speed of the DKO line was also notably slower (84.9 ± 8.1 mm/s in WT, 12.8 ± 1.3 mm/s in DKO; P < 0.001) (Fig. 3D).

In addition to the escape response, we also analyzed spontaneous locomotion, which corresponds to the “slow swim” described by Budick and O’Malley (21) or “scoot” reported by Burgess and Granato (22) (Fig. 3, E and F). Significant difference in swimming speed was not observed between WT and DKO (16.1 ± 1.60 mm/s in WT, 13.2 ± 0.9 mm/s in DKO; P = 0.20) (Fig. 3F). Thus, the contribution of fast muscles in spontaneous swimming is relatively small. These results strongly suggest that fast muscles in larval zebrafish play a key role in executing quick escape responses including the C-bend and fast forward propulsion behaviors, which corroborate earlier studies (23).

Locomotion in adults

DKO fish die prematurely and do not develop into adults. However, εKOs that reached the adult stage are expected to lack both γ and ε subunits, because γ subunit expression terminates early in development.

To dismiss the possibility of compensatory up-regulation of the γ subunit in adult εKOs, we analyzed the expression of γ subunit mRNA with digital droplet polymerase chain reaction (ddPCR). γ Subunit mRNA was not detected in adult εKOs, which were 3 to 5 months old (Fig. 4A). Interestingly, γ subunit mRNA was strongly up-regulated in larval εKOs (Fig. 4B), which may account for functional escape response behavior at 6 dpf (fig. S1). Thus, our findings suggest that compensation by the γ subunit expression occurs only in larval εKOs and not in adults.

Fig. 4 AChR expression restricted to slow muscles of 3- to 5-month-old adult εKO zebrafish.

(A) Quantification of γ or ε subunit mRNA in adult muscles. γ Subunit was not detected in WT. γ or ε subunit mRNA was not detected in εKO (n = 6 fish in WT, n = 5 fish in εKO). Sample numbers are shown in parentheses. (B) mRNA expression of γ subunit in 1-dpf larvae. γ Subunit was highly up-regulated in the εKO (n = 5 fish) compared to WT (n = 5 fish). Sample numbers are shown in parentheses. (C) Schematic illustration of a transverse section of the trunk region. The area shown in micropictograms is indicated with a box. The distribution of AChRs in adults, WT or εKO, was visualized by α-BTX conjugated with Alexa Fluor 488 (green). Broken lines indicate the boundary of fast muscle area (arrowheads). Fast muscles in the εKO fish lack α-BTX signals. (D) Sections of adult fast muscles of WT and εKO, stained with the fast muscle–specific F310 antibody. Fast muscles in εKO fish did not display atrophy. In the right panel, diameters of fast muscles in WT and εKO were calculated (87 fibers, n = 3 fish). There was no significant difference. Scale bars, 100 μm.

The expression of AChR in adult εKO fish, visualized by α-BTX, was consistent with the lack of γ compensation (Fig. 4C). Transverse sections of the trunk region were labeled with α-BTX. Slow, intermediate, and fast muscles are spatially segregated (11). Slow muscles are located closest to the surface. WT fish displayed universally distributed, positive α-BTX signals. In sharp contrast, α-BTX signals in the εKO fish were detected only in shallow, lateral regions, and fast muscles of the adult εKO lacked AChR expression.

In spite of the absence of α-BTX–positive signals, fast muscle fibers in εKO fish unexpectedly lacked signs of prominent atrophy (24). A fast muscle–specific F310 antibody used via immunohistochemistry allowed the visualization and diameter measurements of fast muscle fibers. Statistical analysis revealed no difference between εKO and WT fiber size (58.7 ± 0.5 μm in WT, 58.3 ± 0.7 μm in εKO; P = 0.945) (Fig. 4D).

We observed escape responses induced by objects dropping on water and subsequently analyzed C-bend angles and the swimming speed during escape (Fig. 5A) (25). We compared the maximum C-bend angles between the focal genetic lines. Similar to WT larvae (Fig. 3), WT adults start the escape response with the initial extreme head turn. Unexpectedly, we found that εKO adult fish also display robust C-bends (Fig. 5, A and B). Although smaller in amplitude (103.0 ± 7.5° in WT, 53.4 ± 2.5° in εKO), their time course did not exhibit any delay compared to WT. This is in sharp contrast to the complete loss of C-bend behavior observed in larval DKOs (Fig. 3). The duration of first turn also showed no significant difference between WTs and εKOs (38.9 ± 3.8 ms in WT, 46.6 ± 4.9 ms in εKO).

Fig. 5 Locomotion of the εKO fish in the adult stage.

(A) Escape behaviors in WT and εKO adults (3 to 4 months old). The startle response was induced by dropping objects on water. Images of representative fish to the left show superimposed frames of the complete escape response (the duration of movement is indicated in the bottom right corner). Kinematics for representative traces from 10 or 9 fish are shown for the initial 50 ms of response. Middle panels represent averaged traces. In right panels, each trace represents a different fish. Body angles are shown in degrees, with 0 indicating a straight body. Positive and negative values indicate body bends in opposite directions. (B) First turn angles were calculated for each group of fish (n = 10 fish in WT, n = 9 fish in εKO). Turn angles were reduced in the εKO fish. Sample numbers are shown in parentheses. (C) Post-startle swimming speed and total distance traveled were calculated for the first 120 ms. There was no significant difference between WT (n = 10 fish) and εKO (n = 9 fish) adults.

Furthermore, the forward propulsion during escape of the εKO adult zebrafish was almost intact. When the distance traveled was plotted against the time after stimulation, the curves for WT and εKO nearly overlapped (Fig. 5C). The swimming speed (31.7 ± 1.3 cm/s in WT, 25.5 ± 3.0 cm/s in εKO; P = 0.08) and total distance traveled (4.0 ± 0.2 cm in WT, 3.2 ± 0.4 cm in εKO; P = 0.08) were similar between WT and εKO adults.

Suspecting that compensation of locomotion occurred at the level of neural projection, we examined the projections of motor neurons by retrograde labeling using a fluorescent tracer, dextran conjugated with Alexa Fluor 488 (Fig. 6, A to C). We injected the tracer into muscles of WT and εKΟ fish following a method described in a previous report (26). Spinal motor neurons in adult zebrafish are classified on the basis of morphological features. Dorsomedial motor neurons with larger cell somas, which are called primary motor neurons (pMNs), specifically innervate fast muscles. Ventrolateral motor neurons with smaller somas, called secondary motor neurons (sMNs), are grouped in distinct populations depending on the innervation target: fast, intermediate, and slow muscles (2729). We analyzed the location of motor neuron somas in the spinal cord (Fig. 6B) by measuring the distance from the center of spinal cord to cell somas. In WT adults, fast muscles were innervated mainly by dorsomedial motor neurons (located close to the center), and slow muscles were innervated by ventrolateral motor neurons (Fig. 6, A and B).

Fig. 6 Retrograde labeling of motor neurons show changed innervation.

(A) Schematic illustration of a transverse section of the trunk region showing the sites of dye injections. Right panels showing cell bodies of labeled motor neurons (arrowheads) in spinal cords. Broken lines indicating outlines of spinal cords. Scale bars, 50 μm. (B) A graph showing the distance from the center of the spinal cord to cell bodies of motor neurons. In WT, motor neurons located close to the center innervate fast muscles, and ventrolateral motor neurons innervate slow muscles. In εKO, slow muscles were innervated by motor neurons located close to the center. Numbers of labeled cells are shown in parentheses. (C) Graph showing the size of cell somas of motor neurons. In WT, large motor neurons innervate fast muscles, and smaller neurons innervate slow muscles. In εKO, slow muscles were innervated by large motor neurons. (D) Schematic illustration of a transverse section of the trunk region showing the locations of the DiI crystal insertion. The right panel displays cell body of labeled pMN (arrowhead) in the spinal cord. The broken line indicates the outline of the spinal cord. Scale bar, 50 μm. (E) Presynaptic structures were visualized by SV2A antibody. Broken lines indicate the boundary of slow muscle area (left side). Note the reduced signal in the fast muscles of the εKO fish. Scale bars, 100 μm. (F and G) Fast muscle–specific myosins labeled by F310 antibody in WT (F) and εKO (G). In (G), the boxed area is enlarged in the right panel. Broken lines indicate the boundary of slow muscle area (left side). Arrowheads indicate muscle cells with F310 signals in the slow muscle region. While a small number of slow muscle cells in WT sometimes showed immunoreactivity, the cell number was markedly increased in εKO. Scale bars, 100 μm. (H and I) Glycolytic muscle fibers were visualized by αGPD staining in WT (H) and εKO (I). Black broken lines indicate the boundary between slow and intermediate muscles, and the red broken line indicates the boundary between intermediate and fast muscles. Fast, intermediate, and slow muscle areas are labeled with F, I, and S, respectively. Note that the intermediate muscle region in εKO is hard to distinguish from the fast muscle region, blurring the boundary (I). Arrowheads in the right panel indicate muscle cells with αGPD signals in the slow muscle region. Scale bars, 100 μm. (J) Schematic illustration showing the rerouted innervation of pMNs. In εKO adults, synaptic silencing of fast muscles led to the innervation of fast muscle–specific pMNs on slow muscle. This reinnervation caused conversion of slow to fast muscles. The projections of sMNs that innervate fast muscles may not change.

Both the location and the size of motor neuron somas suggested that slow muscles in εKO adults were innervated by large motor neurons, which innervate only fast muscles in WT adults (Fig. 6C). Ventrolateral neurons did not seem to innervate slow muscles in εKOs, as they were absent in retrograde labeling (Fig. 6, B and C). When we injected the tracer into fast muscles of εKO adults, pMNs were not labeled (fig. S2). Motor neurons labeled in these preparations were presumably fast sMNs (26).

To rule out the possibility that pMN axons are inadvertently damaged by dye injections into slow muscles of εKO adults, we used another method of retrograde labeling using a lipophilic tracer DiI (or DiIC18), which has a minimal possibility of causing pressure injection damage (30). After gently placing crystals of DiI onto slow muscles of εKO adults, we found that pMNs were labeled in spinal cords of εKO adults (Fig. 6D). We also analyzed the presynaptic input in muscles of WT and εKO adults using SV2A antibody to visualize presynaptic proteins (Fig. 6E). The results showed that positive signals within fast muscles were reduced in εKO compared to WT adults. Thus, fewer motor neurons innervated fast muscles in εKO fish.

The muscle cell type is determined by the motor neuron input (31). Suspecting the signals from pMNs may convert the properties of slow muscles into those of fast muscles in adult εKO fish, we examined the characteristics of slow muscle fibers. To do so, we analyzed the F310 antibody immunohistochemistry in adult εKO fish, which labels fast muscle–specific myosin (Fig. 6, F and G) (19). We also examined the α-glycerophosphate dehydrogenase (α-GPD) activity, which is a well-established method to visualize glycolytic muscles, i.e., fast muscles (Fig. 6, H and I) (32). Some tissue located in slow muscle regions stained positive for F310 (n = 3 fish; Fig. 6G) and α-GPD signals (n = 3 fish; Fig. 6I), suggesting that some slow muscles expressed the fast muscle–type isoform of myosin light chain and obtained glycolytic ability. Intermediate muscle fibers in εKO also showed higher glycolytic ability compared to WT (Fig. 6, H and I). Thus, a subpopulation of slow and intermediate muscles was converted to fast muscles, presumably due to the innervation of fast muscle motor neurons (31).

In summary, the absence of AChRs in developing εKOs is presumed to drive motor neuron axon innervation of fast muscles to instead reroute to slow muscles. These rewired pMNs presumably predominate over original axons in slow muscles, as a result of synaptic competition, and convert some slow and intermediate muscles to fast muscles (Fig. 6J).


This is the first report to analyze an adult animal whose synaptic contact on fast muscles is selectively silenced. Although overexpression of slow muscle–specific transcriptional regulators produced mice with an increased proportion of slow muscle cells, the conversion was not complete (33, 34). Synaptic silencing in this study showed that fast muscles are necessary for making C-bends in the larval stage. However, adult εKO fish displayed unexpected adaptation to synaptic silencing: rewiring of motor axons and conversion of slow and intermediate muscles to fast muscles.

An additional important finding was that fast muscle could not express AChRs of the slow muscle type, comprising α, β, and δ subunits; conversely, slow muscle AChRs could not incorporate the ε subunit. Although the distinct compositions between fast and slow muscle AChR subunits were recently described (13), the mechanism remains unresolved. Ahmed and Ali reported that slow muscles expressed mRNA of γ and ε subunits (35). The absence of alternate subunit composition reported here also strongly supports the idea that the subunit composition is regulated at the posttranslational level. The conversion of slow muscle properties observed in εKO adults (Fig. 6, F to I) poses an interesting phenomenon. If the change to fast muscle is complete, AChRs will no longer express without γ and ε subunits, which contradicts the result of α-BTX staining in εKOs (Fig. 4). Therefore, the conversion seems partial; the myosin light chain type and metabolism are changed, while the AChR subunit composition is spared.

Motor neurons that innervate slow and fast muscles are distinct, in terms of both the synaptic drive from interneurons and the passive electrical properties (36). In larval zebrafish, pMNs innervate fast muscles, while a subgroup of sMNs called intermyotomal sMNs innervates slow muscles. These intermyotomal sMNs are silenced during fast speed swimming behavior (37, 38). In addition, the dorsoventrally projecting subgroup of sMNs innervates both fast and slow muscles (36).

In the adult spinal cord, fast pMNs purely innervate fast muscles. Whereas the fast, intermediate, and slow classes of sMNs innervate fast, intermediate, and slow muscles, respectively (26). Adult slow sMNs, in contrast to larval slow muscle–specific sMNs, are active in both slow and fast swimming (26).

The neural network underlying the escape response has been extensively studied (39). In response to tactile or acoustic stimuli, Mauthner neurons in the hindbrain activate and send signals down the spinal cord. In larvae, motor neurons subsequently fire and cause trunk muscle contraction, leading to the C-bend behavior, whose maximum angle occurs within <12 ms after the initiation of muscle contraction (39). Adult sMNs that innervate slow muscles are disengaged by the Mauthner neuron excitation and remain suppressed for over 0.5 s (40). In light of these studies, the peak bending of the trunk in the εKO adults, which occurred in less than 20 ms, was unexpected (Fig. 5). We therefore hypothesized that slow muscles in εKOs were innervated by fast pMNs, which was corroborated by retrograde labeling and the conversion of muscle type from slow to fast (Fig. 6). Notably, the peak amplitude of C-bend behavior was smaller in adult εKOs than WTs (Fig. 5). This is presumably due to the reduction of muscle cell population responding to the motor neuron firing (Fig. 6J).

εKO mice show muscle atrophy and die prematurely due to muscle weakness (24), in contrast to zebrafish εKOs, which survived to adulthood and displayed relatively normal swimming behavior. The difference between mice and zebrafish phenotypes may stem from several factors. First, it remains to be determined whether slow muscles in mammals express AChRs without γ or ε. εKO mice may therefore be unable to express any AChRs in muscles after the decrease of the γ subunit expression, in which case rewiring of motor neuron axons will be futile. Second, fish muscle did not show signs of atrophy, while εKO mice suffered from muscle weakness and atrophy. The lack of atrophy may result from the relatively young age of adult fish (3 to 5 months) used in the study. Many developmental factors are involved in the formation or growth of muscle fibers (41). It is possible that these developmental factors remain functional in fast muscles of εKO fish and prevent atrophy. Older εKO fish may present atrophy similar to mice and has yet to be explored and characterized. While the precise cause remains unknown, zebrafish εKOs also have a low survival rate (~25%), and it is possible that factors leading to premature death of εKO mice may have a role in zebrafish fatality. Thus, the difference between εKO zebrafish and mice phenotypes may actually be subtle.

The molecular and physiological mechanisms leading to the rerouted innervation of pMNs, which is not genetically programmed and caused by synaptic silencing of fast muscles, are highly exciting. This phenomenon can be divided into three processes. First, pMN axons abandon nonresponsive fast muscles. Second, pMN axons explore, locate, and make synaptic contacts with active slow muscles. Third, pMNs compete with previously established sMNs and predominate innervation.

The third process presumably results from competition of axons innervating a single target: a well-established concept called synaptic elimination (42, 43). In mice diaphragm muscles, axons with more activity are favored over less active axons, resulting in a single innervation per muscle fiber.

The first and second processes are unique, lacking directly comparable experimental paradigms to the best of our knowledge. In developing εKOs, only slow muscles function after the termination of γ subunit expression. It is feasible that the higher activity of slow muscles relative to silenced fast muscles attracted motor neuron axons as the animal matured. The information of muscle activity can be transmitted over a distance. For example, injured peripheral nerves in mammals displayed activity-dependent enhancement of axon regeneration and functional recovery (44, 45). The molecular identity of this signal, which results in the attraction of pMN axons toward active muscle cells and the establishment of synaptic contacts against the genetic programming, will await further studies.


Fish lines and maintenance

Zebrafish were maintained in the self-circulating systems at the National Institute on Alcohol Abuse and Alcoholism (NIAAA)/National Institutes of Health (NIH) and the Osaka Medical College. All methods were carried out in accordance with relevant guidelines and regulations. All experimental protocols were approved by NIAAA/NIH and Osaka Medical College.

Genome editing

A frameshift mutant of the AChR γ subunit gene (chrng) was generated by the CRISPR-Cas9 system. For generation of guide RNA (gRNA), oligo DNA for gRNA (Eurofins Genomics, Tokyo, Japan) was annealed and ligated with gRNA expression vector (DR274; Addgene), followed by digestion with Bsa I (New England Biolabs, Ipswich, MA) (46). After cloning and digestion by Dra I (New England Biolabs), gRNA was transcribed by T7 polymerase (Roche, Molecular Biochemicals GmbH, Mannheim, Germany). The Cas9 mRNA was transcribed using Pme I–digested Cas9 expression vector (MLM 3613; Addgene) by an mMESSAGE mMACHINE T7 ULTRA kit (Thermo Fisher Scientific, Waltham, MA). CRISPR-Cas9 solution containing gRNA (12.5 ng/μl) and Cas9 mRNA (300 ng/μl) was injected into the cytoplasm of fertilized one cell–stage eggs. Genomic DNA of F1 fish was extracted from the caudal fin. Amplicon that includes the target region of each gene was generated by PCR. Amplicons of the candidate fish were sequenced by a commercial company (Eurofins Genomics). F1 fish that had mutations were crossed. Homozygous KO F2 fish were selected by genome sequence (fig. S1 shows phenotypes of single KO lines). Frameshift mutants of the AChR ε subunit gene (chrne) were established similarly, except that TALEN was used instead of CRISPR-Cas9. Complementary DNA (cDNA) encoding TALEN was designed by Life Technologies Corporation (Carlsbad, CA). mRNAs were synthesized in vitro and injected at one cell stage into fertilized embryos. After checking somatic mutations, germline transmission was confirmed and established lines were used to generate homozygous fish.

Swimming analysis

High-speed image capturing of larval and adult zebrafish was performed with a Photron camera (Photron, Tokyo, Japan) at 1000 frames/s. Captured images were saved as JPEG and processed with ImageJ. For larvae (6 dpf), the trunk region was touched gently with a pipette to induce escape behavior (47). For adults, we dropped a 0.5-ml tube weighted with 1.5 g onto a point close to the animal’s head from 30 cm above the water (25). The head angle (C-bend) for each frame of the response was measured. In addition, swimming speed and total distance traveled during the escape were calculated. The frame at 0 ms was chosen before the movement was first detected. Measured angles were plotted against time.


Briefly, adult zebrafish (3 to 5 months old) were deeply anesthetized with tricaine (Tokyo Chemical Industry, Tokyo, Japan) before decapitation. Trunks were dissected into small segments and fixed with 4% paraformaldehyde/phosphate-buffered saline (PBS) at 4°C for 24 hours. The specimens were washed with PBS and immersed in 30% sucrose/PBS at 4°C overnight. The fixed specimens were embedded in optimal cutting temperature (OCT) compound (Sakura Finetek Japan, Tokyo, Japan) and frozen. Specimens were frontally cryosectioned at 20 μm using a cryostat (CM 3050S; Leica Microsystems, Wetzlar, Germany) and mounted onto MAS-GP type A–coated glass slides (Matsunami, Osaka, Japan). The sections were incubated with α-BTX and Alexa Fluor 488 conjugate (Thermo Fisher Scientific) [1:400; with PBS containing 0.3% Tween 20 (PBST)] for 2 hours and rinsed twice with PBS. The sections were coverslipped with DAPI (4′,6-diamidino-2-phenylindole) Fluoromount-G (SouthernBiotech, Birmingham, AL) and observed under a TCS SP8 confocal microscope (Leica Microsystems).

For immunohistochemistry, adult zebrafish (3 to 5 months old) were fixed and mounted similarly. Specimens were frontally cryosectioned at 10 μm using a cryostat (CM 3050S; Leica Microsystems, Wetzlar, Germany) and mounted onto MAS-GP type A–coated glass slides (Matsunami, Osaka, Japan). The sections were incubated with F310 [1:50; Developmental Studies Hybridoma Bank (DSHB), Iowa City, IA] or synaptic vesicle glycoprotein 2A (SV2A) antibody (1:50; DSHB) overnight. After wash with PBST, the sections were incubated with secondary antibody, goat anti-mouse immunoglobulin G (H+L) Alexa Fluor 555 (Thermo Fisher Scientific) or 488 (1:500), for 2 hours. After a wash with PBST, sections were coverslipped with DAPI Fluoromount-G (SouthernBiotech) and observed under a confocal microscope (Leica TCS SP8, Leica Microsystems).

For αGPD staining, cryosectioned samples were incubated for 45 min at room temperature in 0.05 M tris buffer at pH 7.4 with 0.02% menadione, 0.05% nitroblue tetrazolium, and 0.01 mol disodium α-glycerophosphate. The sections were subsequently rinsed with PBS, coverslipped with DAPI Fluoromount-G (SouthernBiotech), and observed under a BZ-X700 microscope (KEYENCE, Osaka, Japan).

Histology for larval zebrafish was performed as previously described with some modifications (48). Larvae were anesthetized first in 10% Hank’s solution with 0.02% tricaine. The fish were kept in 100% Hank’s solution containing α-BTX and Alexa Fluor 488 conjugate (1:400) for 1 hour. Then, the fish were washed for 2 hours in toxin-free Hank’s solution to remove nonspecific binding of the toxin. Subsequently, larvae were embedded in 1% agarose and mounted on glass-bottom petri dish for observation under a TCS SP8 confocal microscope (Leica Microsystems).

Electrophysiological recordings

Recording of mEPC in the NMJ was performed as previously described with some modifications (47). Skinned larvae were pinned down to the recording chamber coated with Sylgard and immobilized by bathing in 10 μM nifedipine (Merck, Kenilworth, NJ). Patch-clamp recordings were made by the whole-cell ruptured technique of muscle cells. The pipette solution used for the voltage-clamp recording was 120 mM KCl, 5 mM BAPTA, and 5 mM Hepes (pH 7.2). The extracellular solution contained 112 mM NaCl, 2 mM KCl, 2 mM CaCl2, 1 mM MgCl2, 3 mM glucose, and 5 mM Hepes (pH 7.4). Membrane currents were recorded with an EPC10 amplifier and PatchMaster. For mEPC recordings, muscle cells were voltage-clamped at −90 mV, and 1 μM Tetrodotoxin (TTX) was added to the recording solution. The currents were sampled at 50 kHz and filtered at 3 to 5 kHz before analysis. Capacitive transients were compensated manually.

For puff application, a glass electrode [opening, ∼30 μm; filled with bath solution containing 30 μM ACh and dextran–Alexa Fluor 488 (Thermo Fisher Scientific, Waltham, MA)] was placed near the voltage-clamped muscle cell and positive pressure was applied using Picospritzer II (Parker Hannifin; 30 ms, 1 psi). ACh-induced currents were recorded by the whole-cell ruptured technique.

Ca2+ imaging

We amplified a zebrafish genomic DNA fragment containing the 5′-flanking region of the α-actin gene by PCR using PrimeSTAR (Takara, Shiga, Japan). The primers used had restriction sites (Sph I and Sal I), and the amplicon was inserted into the pT2RUASGCaMP7a plasmid (supplied by K. Kawakami’s laboratory), which swapped the upstream activation sequence with the promotor sequence of α-actin to drive the expression of GCaMP7a.

We injected the constructed plasmid into one cell–stage embryos from γ−/− ε+/− pairs. The genotype of the injected embryos (at 6 dpf) was determined by responses to touch, and the embryos were anesthetized with 0.02% tricaine (Tokyo Chemical Industry) and pinned to a Sylgard dish. Tricaine was washed out, and embryos were bathed in Evans solution containing a muscle myosin inhibitor, N-benzyl-p-toluene sulfonamide at 5 mM (Merck), which does not affect the Ca2+ regulation (49). N-methyl-d-aspartate (50 μM) was applied to activate swimming (18). We performed line scanning at 26 Hz (38 ms/scan) of a fluorescent muscle cell. Fluorescence intensity was divided by an averaged (background) intensity. The resting intensity before [Ca2+] rise was chosen as the zero point. After recording, genotypes of the embryos (at 6 dpf) were determined by genomic DNA sequencing.

Digital droplet PCR

Total RNA was isolated from the whole body of WT or εKO larvae (n = 5 each) or from muscles of trunk regions of adult WT (n = 6 fish) or εKO fish (n = 5 fish) using NucleoSpin RNA XS (Takara). cDNA was synthesized using the PrimeScript RT Reagent Kit (Perfect Real Time, Takara). ddPCR was carried out using QX100 droplet digital PCR system following the manufacturer’s instructions (Bio-Rad, Pleasanton, CA). Briefly, for each 20-μl reaction, 0.5 μl of cDNA was mixed with 1 μl of target primers/probe (FAM) (20×), 10 μl of Bio-Rad’s ddPCR supermix for probes (2×), and 8.5 μl of nuclease-free water. We designed primers/probes in exon-exon boundary regions for each target gene; γ gene, ε gene, and β-actin gene (Supplementary Materials). The entire reaction mixture was loaded onto a disposable plastic cartridge (Bio-Rad) together with 70 μl of droplet generation oil (Bio-Rad) and placed in the droplet generator (Bio-Rad). After processing, samples were transferred to a 96-well PCR plate. PCR amplification was carried out on a T100 thermal cycler (Bio-Rad) using a thermal profile: 95°C for 10 min, 40 cycles of 94°C for 30 s and 60°C for 60 s, 98°C for 10 min. After amplification, the plate was loaded on the droplet reader (Bio-Rad), and the droplets from each well of the plate were read automatically at a rate of 32 wells per hour. ddPCR data were analyzed with QuantaSoft analysis software (Bio-Rad), and the quantification of the target molecule was presented as the number of copies per microliter of PCR mixture after normalization with the β-actin control.

Retrograde labeling of motor neuron

Retrograde labeling of motor neurons was performed as previously described with some modifications (26). After anesthesia by 0.02% tricaine (Tokyo Chemical Industry), crystals of dextran–Alexa Fluor 488 (molecular weight, 10,000; Thermo Fisher Scientific) were dissolved in distilled water and injected into fast or slow muscles using glass pipettes. Five to 6 hours after recovery from anesthesia, the animals were again deeply anesthetized and decapitated. For experiments using lipophilic fluorescent tracer DiI [or DiIC18(3)] (Biotium, Fremont, CA), crystals of DiI were inserted gently into the slow muscle area of εKO adults using a glass pipette. After placing the fish back in water for 24 hours, fish were deeply anesthetized and decapitated. Trunks were dissected and fixed with 4% paraformaldehyde/PBS at 4°C overnight. The specimens were washed with PBS and immersed in 30% sucrose/PBS at 4°C overnight. The specimens were embedded in O.C.T. compound (Sakura Finetek Japan) and frozen. Specimens were frontally cryosectioned at 20 μm using a cryostat (CM 3050S; Leica Microsystems) and mounted onto MAS-GP type A–coated glass slides (Matsunami) and coverslipped with DAPI Fluoromount-G (SouthernBiotech) for observation under a Leica TCS SP8 confocal microscope (Leica Microsystems).


Unpaired t test (two-tailed) was performed for statistical analysis. For multiple comparisons, t tests were performed with Bonferroni correction. Averages and SEM are displayed.


Supplementary material for this article is available at

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: We thank N. Tanaka for excellent care of the fish and for help in making figures. We thank A. Nishino for critical reading of the manuscript. We thank S. Higashijima and K. Kawakami for providing plasmids. Funding: This study was supported by Grant-in-Aid for Scientific Research (KAKENHI 18K06882 to F.O.) from the Japanese Ministry of Education, Culture, Sports, Science and Technology and internal grants from NIAAA/NIH and Osaka Medical College. Author contributions: Conceptualization: F.O.; investigation: B.Z.; supervision: F.O.; resources: Y.Y.; writing the manuscript: B.Z., T.W., and F.O. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All materials used in the analysis are available to any researcher for purposes of reproducing or extending the analysis.

Stay Connected to Science Advances

Navigate This Article