Agrin/Lrp4 signal constrains MuSK-dependent neuromuscular synapse development in appendicular muscle

Movement depends on the coordinated development of neuromuscular junctions (NMJs) between motor axons and skeletal muscle. Prior to axon arrival, the central region of muscles is ‘prepatterned’ with clustered acetylcholine receptors (AChR), which requires the receptor tyrosine kinase MuSK. Subsequently, as motor axons contact the muscle and differentiate, they release the glycoprotein Agrin, which binds to the low-density lipoprotein receptor-related protein 4 (Lrp4) on the muscle membrane to stimulate phosphorylation of MuSK. Upon activation, MuSK initiates a downstream signaling cascade to cluster AChRs in apposition to axons, thereby forming neural synapses. Mutants of lrp4, agrn and musk all fail to form NMJs in the mouse and zebrafish trunk, demonstrating their crucial and conserved role in this process (Kim and Burden, 2008; Kim et al., 2008; Zhang et al., 2008; Jing et al., 2010; Remédio et al., 2016; Gribble et al., 2018). Prior work on the Agrin/Lrp4/MuSK pathway in NMJ development has focused predominantly on trunk muscles. In contrast, significantly less is known about the mechanisms crucial for NMJ development in appendicular muscles, which move appendages, such as limbs.

Here, we used the larval zebrafish pectoral fin, analogous to tetrapod forelimbs (Mercader, 2007), to study neuromuscular synapse development in a paired appendage. At 120 h post fertilization [hpf; 5 days post fertilization (dpf)], pectoral fins comprise two antagonistic muscles, the abductor and adductor, separated by an endoskeletal disk (Fig. 1A,B). Each muscle consists of ∼50 fast-twitch (fast) muscle fibers that extend longitudinally from the proximal fin base where it attaches to the trunk, out towards the distal tip of the fin. At the fin base, the abductor and adductor muscles are each two-to-three fibers thick, and thin out to be a single fiber layer throughout most of the fin (Thorsen et al., 2004). The abductor and adductor muscles are innervated by four distinct motor nerves, which we refer to here as nerves 1-4, with cell bodies in anterior spinal cord segments 3-6 (Myers, 1985; Thorsen and Hale, 2007). Motor axons enter the fin at a dorsal (nerves 1-3) or ventral (nerve 4) plexus to sort between the abductor or adductor muscles (Thorsen and Hale, 2007). Axons then progressively defasciculate as they grow towards the distal tip of the fin, such that each muscle fiber is polyinnervated and motor axons create a patchwork pattern across the fin muscles. This innervation pattern remains unchanged until juvenile stages at 3 weeks post-fertilization (5.4-5.8 mm) when the muscles divide, nerves arborize and bone forms (Grandel and Schulte-Merker, 1998; Thorsen and Hale, 2007). The genetic tractability, availability of transgenic tools to label specific cell types, optical transparency suitable for live imaging, and behavioral readout of fin movement make the larval zebrafish pectoral fin an ideal vertebrate system to interrogate mechanisms of neuromuscular synapse development within appendicular muscles.

The axial trunk and pectoral fin neuromuscular systems differ in several key ways. First, axon innervation of trunk muscles begins between 16 and 24 hpf (Eisen et al., 1986), whereas, in appendicular/fin muscles, it starts approximately 24 h later. Second, the trunk comprises several medial layers of fast muscle fibers and a single lateral slow-twitch (slow) muscle fiber layer, which are arranged in repeating segments. In contrast, the fin muscles are only one fiber thick, comprise solely fast fibers (Patterson et al., 2008; Thorsen and Hale, 2007) and are ∼2.5 times longer than trunk muscle fibers. Additionally, after exiting the spinal cord, axons in the trunk grow perpendicular to, and along the center of, muscle fibers. In contrast, motor axons that innervate the pectoral fin grow through the body wall, sort at a plexus and then branch to create elaborate innervation patterns across fin muscle fibers. Axonal innervation of the fin is topographic, with axons from anterior spinal segments innervating the dorsal fin and axons from posterior segments innervating the ventral fin (Thorsen and Hale, 2007). Finally, in the trunk, aneural AChRs are present on slow muscle fibers, which are absent in the pectoral fin (Flanagan-Steet et al., 2005). Whether AChRs are prepatterned and how axon outgrowth occurs in the pectoral fin have not yet been described.

Although many of the signals that mediate NMJ development have been well characterized in trunk muscles, whether the same cellular and molecular mechanisms underlie NMJ development within the complicated muscle arrangement of paired appendages has remained unclear. Here, we reveal that, although some aspects of neuromuscular synapse development, such as prepatterning with aneural AChR clusters and the requirement for MuSK to form AChR clusters, are shared between the trunk and the pectoral fin, there are key muscle-specific differences. We show that both axonally released Agrin and Lrp4 on muscle cells are required to form the distributed patterning of AChR clusters in the fin. Whereas agrn and lrp4 mutants fail to form synapses in the trunk, in the pectoral fin they instead form giant AChR clusters and axonal innervation abnormalities. A developmental time course in agrn mutant fins reveals that these clusters likely arise from prepatterned AChR clusters that grow over time. These giant clusters sequester navigating growth cones, thereby disrupting the innervation patterning. Partial depletion of musk moderately suppresses the formation of these clusters in agrn and lrp4 mutants. Based on our results, we propose a model for NMJ formation in the appendicular fin muscle in which Agrin/Lrp4 signaling transitions MuSK from a prepatterning state to an axon-dependent, focal AChR-clustering state. Without Agrin or Lrp4, MuSK-dependent signaling remains active within prepatterned islands and AChR clusters grow. Importantly, this work exposes key differences in neuromuscular synapse development between axial muscles of the trunk and appendicular muscles of the pectoral fin.

At 120 hpf, zebrafish pectoral fin motor axons have an elaborate pattern across the abductor and adductor muscles. To compare the organization between nerve and muscle within the fin, we used transgenic mnx1:GFP or Xla.tubb:dsRed to visualize most, if not all, motor axons innervating the pectoral fin, and α-actin:GFP to label pectoral fin muscle fibers (Fig. 1A-C). Although the fin comprises both abductor and adductor muscles with independent innervation fields (Fig. 1D,H), for simplicity we include only abductor innervation unless otherwise noted (Fig. 1E-G). Labeling of nicotinic AChRs with α-bungarotoxin (α-Btx) revealed hundreds of small, evenly spaced, en passant neuromuscular synapses juxtaposed to motor axons (Fig. 1F,G). The largest postsynaptic AChR clusters were associated with the major nerve branch closest to the proximal fin base, whereas AChR clusters became smaller as the finer nerve branches defasciculated towards the distal fin tip.

Previous work has described the development of the fin (Yano et al., 2012; Siomava et al., 2018), as well as the innervation and musculature of the fin after 120 hpf (Thorsen and Hale, 2007), a time point at which the larval innervation pattern is relatively stable. To our knowledge, the development of the zebrafish pectoral fin musculature and its complex innervation pattern prior to 120 hpf have not been described. To observe this process dynamically, we used time-lapse imaging of transgenic embryos to observe muscles [Tg(α-actin:GFP)] and axons [Tg(Xla.tubb:dsRed)]. By approximately prim-25 (36 hpf), motor axons from nerves 1-3 coalesced at the future dorsal plexus, and nascent muscle fibers in the pectoral fin bud, located laterally to the axons, had just started expressing α-actin:GFP (Fig. 2A,B; Movie 1). Muscle fibers continued to reorganize through the long-pec stage as the fin moved further medially, closer to the plane of the dorsal plexus; motor axons began to grow into the abductor and then adductor muscles beginning around the long-pec stage at 46 hpf. Concurrently, axons in nerve 4 made a sharp turn dorsally to innervate the fin via the ventral plexus. Thick axon bundles first grew perpendicular to muscle fibers near the proximal fin base and, subsequently, axons turned posteriorly to grow mostly parallel to muscle fibers and towards the fin tip. As muscle fibers elongated, branching motor axons followed close behind, forming a diffuse innervation network. By ∼68 hpf, a simplified innervation pattern was recognizable (Fig. 2B,G) that became more complex through 120 hpf. Thus, pectoral fin neuromuscular development is a highly dynamic yet tightly coordinated process.

Pectoral fins start moving rhythmically around 72 hpf (Uemura et al., 2020), prompting us to examine when neuromuscular synapses form in the pectoral fin. To observe neuromuscular synapse development in pectoral fins, we fixed transgenic mnx1:GFP larvae (to label presynaptic motor axons) at various time points and labeled AChRs with α-Btx. In vertebrates, ‘prepatterned’ AChR clusters form on muscle fibers prior to motoneuron innervation (Lin et al., 2001; Yang et al., 2001), but whether or when this occurs in appendicular muscle of pectoral fins is unclear. At the high-pec stage (∼42 hpf), when the pectoral fin bud is located lateral to the nascent dorsal plexus and axons have not yet entered the fin, AChR clusters were undetectable (Fig. 2C). By 46 hpf, as axons had just started to grow past the plexus onto the fin musculature, we observed small AChR clusters near the base of the fin that were not yet associated with labeled axons (Fig. 2D). At 51 hpf, axons growing from the dorsal and ventral plexi extended towards each other along the fin base and innervated nearby AChR clusters, whereas remaining aneural AChR clusters at the not-yet-innervated medial region of the fin increased in size (Fig. 2E). By 60 hpf, all AChR clusters at the proximal fin base were associated with axons (Fig. 2F).

The entry of axons into the fin initiated a second phase of neuromuscular synapse formation. As axons grew beyond the fin base and branch to extend along muscle fibers, new AChR clusters emerged. These new AChR clusters were always associated with axons (Fig. 2F,G). The number of AChR clusters increased as the innervation pattern became more complex and, by 120 hpf, axons were associated with hundreds of regularly spaced AChR clusters that remained relatively small (<5 µm²) (Fig. 2G). In contrast, clusters at the fin base that preceded axon outgrowth continued to increase in size, such that, by 120 hpf, the biggest AChR clusters in the fin (>20 µm²) were in this proximal region. Thus, neuromuscular synapse development in appendicular muscle of the pectoral fin mirrors that in axial muscle of the trunk, with a prepatterned phase and an axon-associated phase of synapse formation.

Given the similarities in AChR prepatterning and axon-associated AChR clustering, we investigated whether the well-established genetic pathway that regulates neuromuscular synapse development in axial muscle is also required for this process in pectoral fin muscle. In vertebrate axial muscles, AChR prepatterning and the formation of neuromuscular synapses requires the receptor tyrosine kinase MuSK (DeChiara et al., 1996; Lefebvre et al., 2007). By staining sibling and musk mutant pectoral fins with α-Btx to label AChR clusters, we first determined whether musk is also required for AChR prepatterning within larval zebrafish appendicular muscle. At the long-pec stage (46 hpf), when wild-type (WT) sibling animals had developed robust prepatterned aneural AChR clusters, staining of musk-mutant pectoral fins failed to reveal any prepatterned AChR clusters (Fig. S1). To account for a delay in pectoral fin development or aneural cluster formation, we also examined musk mutants at later time points. At 51 hpf and 60 hpf, the extent of axon growth in the fin was comparable between siblings and musk mutants. At these time points, WT axons had innervated prepatterned AChR clusters and new clusters had formed, whereas in musk mutant fins α-Btx staining remained diffuse and muscles lacked discernible AChR clusters. Thus, as in trunk axial muscle, musk is required for AChR prepatterning in appendicular fin muscle.

Next, we examined the role of MuSK in the formation of neural AChR clusters characteristic of neuromuscular synapses. In the zebrafish trunk, motor axons branch and form synapses distributed along myofibers throughout the myotome. Consistent with previous work (Lefebvre et al., 2007; Jing et al., 2010), at 120 hpf, compared with sibling controls, trunk axial muscle fibers in musk mutants exhibited diffuse α-Btx staining with fewer and smaller AChR clusters (Fig. 3A,B,D). Similar to trunk axial muscle fibers, appendicular muscle fibers in musk mutants displayed mostly diffuse α-Btx staining and lacked WT-like AChR clusters (Fig. 3C). Whereas musk mutants exhibited some AChR clusters that form in apposition to axons, they were fewer and smaller in size compared with sibling controls. These circular clusters resembled the dystroglycan-dependent clusters that form on zebrafish axial muscle fibers in the absence of musk (Lefebvre et al., 2007). Moreover, compared with WT controls, motor axons in musk mutant pectoral fins were less fasciculated, similar to that reported previously for trunk muscle fibers in zebrafish and mice lacking MuSK (Kim and Burden, 2008; Jing et al., 2009).

Finally, we asked whether MuSK acts through its well-established downstream effector Rapsyn during appendicular neuromuscular development. Like musk, rapsyn is required for AChR clustering in mouse and zebrafish axial muscle (Gillespie et al., 1996; Ono et al., 2002). Similar to its role in axial muscle, rapsyn is required for AChR clustering in the pectoral fin, as rapsyn mutants displayed diffuse α-Btx signal throughout pectoral fin muscles (Fig. S2). In contrast to musk mutants, in which pectoral fin axons were defasciculated and overgrown (Figs 3C and 7B), motoneuron innervation in the pectoral fin of rapsyn mutants was indistinguishable from that of WT (Fig. S2), similar to observations of rapsyn-mutant trunk innervation (Zhang et al., 2004; Gribble et al., 2018). Thus, musk and rapsyn are required for NMJ development in the pectoral fin, suggesting that postsynaptic mechanisms crucial for synapse formation are shared between trunk and appendicular muscle.

NMJ development requires bidirectional coordination between axons and muscles. To determine the effect of motor axon-derived signals on the postsynaptic innervation pattern, we laser-ablated the motoneurons innervating the dorsal portion of the pectoral fin muscle. Specifically, we ablated the cell bodies of motoneurons in spinal segments 3-5 at 42 hpf, prior to the growth of motor axons into the pectoral fin bud, and then re-ablated any newly formed motoneurons 24 h later. Importantly, we left intact all motoneurons in spinal segment 6, which innervate the ventral muscle region of the fin, to serve as an internal control when comparing the innervation patterns of innervated and nerve-deprived muscle fibers. At 120 hpf, AChR patterning in the innervated ventral region of the fin was indistinguishable compared with non-ablated controls, with small, evenly spaced AChR clusters apposed to axons (Fig. 4A-C). Surprisingly, in the dorsal region of motoneuron-ablated fins that had never been innervated, fewer but much larger AChR clusters formed (Fig. 4C). These clusters were globular and evenly dispersed throughout the non-innervated musculature and differed vastly in size and distribution from those observed in the absence of MuSK or Rapsyn (Fig. 3; Fig. S2). Formation of these abnormal clusters requires MuSK, as motoneuron-ablated musk-mutant pectoral fins exhibited diffuse α-Btx staining similar to innervated musk-mutant fins (Fig. S3). Thus, whereas lack of the postsynaptic signal transduction machinery blocks the formation of AChR clusters almost completely, the absence of axonal-derived signals results in unpatterned yet exuberantly sized AChR clusters, which we refer to as giant AChR clusters. We conclude that axonal-derived signals are crucial for AChR patterning and for limiting AChR cluster size.

The opposing consequences of blocking MuSK-dependent postsynaptic signaling versus eliminating all presynaptic signaling prompted us to examine the role of signaling components that activate MuSK. Axon-derived Agrin coordinates MuSK-dependent AChR clustering between nerve terminals and muscle fibers (Kim et al., 2008; Zhang et al., 2008), and zebrafish mutants lacking the motoneuron-derived Agrin isoforms lack synaptic AChR clusters along axial trunk muscle fibers (Fig. 5A) (Gribble et al., 2018).

Therefore, we hypothesized that Agrin has a similar crucial role in inducing neural AChRs in pectoral fin muscle. Throughout the fin of siblings, hundreds of <5 µm² AChR clusters were evenly distributed along muscle fibers. In contrast, AChR clusters in agrn mutants were reduced in number but increased in size (>20 µm²) (Fig. 5B,E), indistinguishable from the giant AChR clusters in nerve-deprived fins (Fig. 4C). To quantify the number of neural AChR clusters and their size distribution across genotypes, we focused on abductor muscle fibers, although quantification of adductor muscles across all genotypes revealed similar results (Fig. S4). Sibling abductor muscle fibers exhibited 294.4±21.5 α-Btx-positive AChR clusters per fin with a median cluster size of 3.9±0.6 µm² (Fig. 5C-E). In contrast, agrn-mutant abductor muscle fibers had significantly fewer clusters per fin (51.6±5.9, unpaired t-test P<0.0001) but exhibited a vastly increased median cluster size of 18.4±4.7 (unpaired t-test P<0.0001). Thus, in striking contrast to agrn-mutant trunk muscle fibers, which exhibited smaller AChR clusters [median cluster size, siblings: 3.5±0.7 µm² (n=17), agrn^−/−: 1.7±0.7 µm² (n=9), unpaired t-test P<0.0001], agrn-mutant appendicular muscle fibers exhibited greatly enlarged AChR clusters.

The strikingly divergent appendicular NMJ phenotypes observed in musk mutants characterized by an almost-complete loss of AChR clusters versus the giant AChR clusters present in agrn mutants prompted us to examine the role of the Agrin receptor Lrp4. Upon binding Agrin, Lrp4 induces MuSK phosphorylation to initiate synaptic differentiation (Kim et al., 2008; Zong et al., 2012). To determine whether lrp4 mutants display the phenotype exhibited by its ligand Agrin or by its co-receptor MuSK, we examined the role of Lrp4 in appendicular neuromuscular development. Similar to musk and agrn, lrp4 is required for NMJ development in axial muscles of the zebrafish trunk, because mutants form fewer and smaller synapses (Remédio et al., 2016). Identical to the agrin-mutant fin phenotype and in contrast to the lrp4-mutant phenotype in axial trunk muscles, at 120 hpf, lrp4-mutant pectoral fin muscles displayed large AChR clusters (Fig. 6A,B,D-F). Moreover, in fins of lrp4 mutants lacking the intracellular domain (Saint-Amant et al., 2008), we also observed giant AChR clusters (Fig. S5), providing compelling evidence that Lrp4 acts through a ligand-dependent mechanism to regulate appendicular NMJ development. Importantly, the overall number and morphology of Tg(α-actin:GFP)-positive pectoral fin muscle fibers in lrp4 mutants were indistinguishable from that observed in siblings, indicating that appendicular muscle development is unaffected in lrp4 mutants (Fig. S5). Finally, expressing Lrp4 using a muscle-specific promoter Tg(α-actin:lrp4-GFP) (Gribble et al., 2018) in otherwise lrp4-mutant animals fully restored appendicular synapse development (Fig. 6C-F), indicating that Lrp4 functions in muscle. Thus, loss of Agrin or Lrp4, although associated with a significant reduction in neural AChR cluster numbers in both axial and appendicular muscle, leads to an increase in AChR cluster size on appendicular muscle, distinct from their mutant phenotypes in axial muscle.

The giant AChR clusters we observed on agrn- and lrp4-mutant appendicular muscles resembled those previously described in Xenopus and chick muscle cells grown in the absence of axons as AChR ‘hot spots’, which disperse upon innervation (Bekoff and Betz, 1976; Moody-Corbett and Cohen, 1982; Peng, 1986). Therefore, we investigated whether the giant AChR clusters we observed in agrn- or lrp4-mutant pectoral fins are caused by the lack of axonal innervation. Analysis of axonal innervation patterns at 120 hpf revealed that, in both agrn and lrp4 mutants, the overall pectoral fin innervation pattern was comparable to that of WT, indicating that axon guidance mechanisms are largely intact. However, unlike in siblings, in both agrn- and lrp4-mutant fins, we observed large mnx1:GFP-positive innervation ‘swellings’ (Fig. 5B; Fig. 6B). These swellings colocalized with the enlarged postsynaptic AChR clusters, supporting the notion that, rather than being aneural AChR hot spots, these giant AChR clusters are indeed innervated and do not disperse upon axon contact. Moreover, expressing Lrp4 using the muscle-specific promoter Tg(α-actin:lrp4-GFP) (Gribble et al., 2018) in otherwise lrp4-mutant animals fully suppressed formation of presynaptic axonal swellings (Fig. 6C-F), indicating that muscle-derived Lrp4 signaling plays a role in establishing both presynaptic and postsynaptic neuromuscular synapse patterning.

To further investigate the nature of these innervation swellings, we used the Znp-1 antibody against the presynaptic marker Synaptotagmin 2. In control pectoral fins, the Znp-1 signal concentrated at α-Btx-positive postsynaptic areas, demarcating presumptive presynaptic sites. Similarly, in agrn-mutant pectoral fins, the Znp-1 signal colocalized with the giant AChR clusters (Fig. S6). Furthermore, localization of the postsynaptic protein Dystrophin (Dmd) to synaptic regions was maintained in lrp4-mutant pectoral fins (Fig. S7). Thus, whereas in agrn and lrp4 mutants, AChR cluster size and distribution were altered, both presynaptic and postsynaptic proteins were recruited to these giant clusters. Finally, compared with WT siblings, movement was largely unaffected in lrp4-mutant pectoral fins (Fig. S8), suggesting that giant AChR clusters represent bona fide functional synapses.

We next examined the prominent presynaptic ‘swellings’ in lrp4 and agrn mutants that form in opposition to enlarged AChR clusters at the single axon level. To do so, we sparsely labeled axons using mnx1:mKate in the context of the entire population of motor axons (mnx1:GFP). In siblings, we found that individual axons branched and fasciculated with other axons to form complex patterns. Axons terminated abruptly, with endings approximately the same diameter as the rest of the labeled axon (Fig. 6G,H). In lrp4 mutants, most individually labeled axons exhibited similarly complex trajectories, were similar in diameter to sibling controls and occasionally formed simplified endings. However, we also observed individual mnx1:Kate-positive axons that formed bulbous and swollen structures. These globular endings of mnx1:mKate axons were part of larger mnx1:GFP swellings, indicating that multiple axons contribute to these swellings. These swellings colocalized with α-Btx (Fig. 6B), consistent with the idea that, despite their abnormal morphology, they represent axonal terminals. In further support of this conclusion, electron microscopy revealed that, in lrp4 mutants, multiple vesicle-positive axonal processes contributed to these enlarged terminals, reminiscent of presynaptic regions (Fig. S9). These results strongly suggest that, during appendicular neuromuscular development, Agrin/Lrp4-dependent signaling not only promotes the formation of postsynaptic AChR clusters, but also limits their size. In addition, Agrin/Lrp4 signaling also influences presynaptic patterning. Thus, our data reveal that the roles of Agrin and Lrp4 in zebrafish appendicular fin are distinctly different from their well-characterized functions in trunk axial muscle.

Our results reveal that, unlike in trunk muscle of mice and zebrafish, musk-mutant appendicular muscle displayed an almost-complete loss of AChR clusters, whereas agrn- and lrp4-mutant appendicular muscle displayed a strikingly divergent phenotype characterized by enlarged AChR clusters. This led us to first ask whether, in the context of appendicular NMJ development, Agrin and Lrp4 act through MuSK. Indeed, we found that musk;lrp4 double mutants recapitulated the musk-mutant phenotype because they failed to cluster AChRs and display axon overgrowth, confirming that Lrp4 acts through MuSK in NMJ development in appendicular muscle (Fig. 7A,B).

Given that MuSK is necessary for AChR clustering in the pectoral fin in both WT and lrp4 mutants, and MuSK expression is sufficient to induce AChR clusters (Kim and Burden, 2008), we hypothesized that MuSK drives the formation of giant AChR clusters in the absence of agrn or lrp4. This would suggest, unexpectedly, that, in appendicular muscle, Agrin/Lrp4 restricts MuSK-dependent AChR clustering. If this were the case, we would predict that reducing MuSK expression in lrp4 mutants would suppress the giant AChR cluster phenotype. To this end, we examined lrp4 mutants that lacked one copy of musk (lrp4^−/−;musk^−/+). Indeed, these lrp4-mutant animals exhibited a less-severe phenotype compared with lrp4 homozygous mutants. Whereas fins in lrp4^−/−;musk^+/− larvae still contained some giant AChR clusters, portions of the fins in these animals contained smaller, evenly-dispersed neural clusters resembling the sibling patterning. Additionally, the portions of the fin with smaller AChR clusters also lacked presynaptic innervation swellings. Compared with lrp4^−/− mutants, lrp4^−/−;musk^+/− mutants displayed an increase in the number of α-Btx-positive AChR clusters per fin (Fig. 7E), a rescue of the median cluster size (Fig. 7F) and a rescue of the overall distribution of cluster sizes in the fin (Fig. 7G).

In lrp4^−/−;musk^+/− pectoral fins, smaller clusters were mostly present in the distal fin, where AChR clusters formed later in development (Fig. S10). In contrast, ‘non-rescued’ giant clusters often formed closer to the proximal fin base where, during development, the first prepatterned AChR clusters emerge (Fig. 7D). This further supports the idea that Agrin/Lrp4 signaling restrains MuSK-dependent prepatterned AChR cluster growth within appendicular muscle.

To further explore the idea that, unlike in axial muscle, Agrin restrains MuSK-dependent cluster growth in appendicular NMJ development, we examined the progression of appendicular NMJ development in siblings and agrn mutants. We predicted that, prior to the arrival of motor axons, the Agrin-independent formation of aneural prepatterned AChR clusters would be indistinguishable between siblings and agrn mutants. Indeed, we found that, at 46 hpf (long-pec stage), agrn-mutant fins exhibited prepatterned AChR clusters, indistinguishable from controls (Fig. 8A). Moreover, similar to sibling controls, navigating axons in agrn mutants tended to occupy the prepatterned region near the proximal fin prior to extending towards the distal fin (Fig. 8B). Subsequently, in siblings, the arrival of motor axons and the release of Agrin induced the formation of small neural clusters, akin to the process previously described in axial muscle (Panzer et al., 2006).

If nerve-derived Agrin indeed restrains prepatterned cluster growth, we predicted that, in agrn mutants, these initially aneural clusters would grow over time, despite the arrival of motor axons. We also predicted that lack of Agrin-mediated local MuSK activation would result in the failure to form new small neural clusters that normally emerge along growing axons. Indeed, at 60 hpf, there was a prominent difference in cluster size, number and distribution between genotypes. In siblings, the AChR cluster field mirrored the innervation pattern; even the furthest-reaching axons were associated with small AChR clusters, suggesting that, as axons grow, new clusters are rapidly formed. In contrast, in agrn mutants, AChR clusters had increased in size and remained globular (Fig. 8C). Unlike in siblings, presynaptic swellings apposed to AChR clusters become apparent by 72 hpf in agrn mutants, with stretches of axon deprived of any discernible AChR clusters (Fig. 8D). This supports the idea that navigating growth cones are inappropriately attracted to, and sequestered by, these prepatterned ‘islands’. Between 72 and 120 hpf, both sibling and mutant axons continued to grow to occupy the entire muscle territory. In siblings, the majority of the AChR clusters remained small (<5 µm²) and evenly dispersed throughout the appendicular muscle. In contrast, both presynaptic axonal swellings and neural AChR clusters continued to grow throughout the entire appendicular muscle in agrn mutants (Fig. 8D-F). Thus, lack of agrn leads to a progressive size increase in prepatterned AChR clusters. Together, our results support a model for appendicular NMJ development in which Agrin/Lrp4 signaling, in addition to its sole role in axial muscle to activate MuSK and promote NMJ development, also restrains MuSK-dependent prepatterning to size and pattern neuromuscular synapses appropriately.

The mouse diaphragm and the larval zebrafish trunk are powerful in vivo models to study neuromuscular synapse development. Genetic studies using these models have converged on an evolutionarily conserved pathway in which axonally released Agrin binds its receptor Lrp4 to locally activate the receptor tyrosine kinase MuSK and cluster AChRs. Here, we used the larval zebrafish pectoral fin as a genetically tractable system in which to study neuromuscular development in appendicular muscle. We report that Agrin, Lrp4 and MuSK are required to form neural AChR clusters in appendicular muscle, similar to their roles in axial muscle. In addition, we provide compelling in vivo evidence that Agrin and Lrp4 play an additional, previously unappreciated role to suppress the growth of aneural AChR clusters selectively in appendicular muscle. In addition, we show that Agrin/Lrp4 signaling influences presynaptic axonal patterning, because multiple axons inappropriately innervated these giant AChR clusters in lrp4 and agrn mutants. Thus, our work reveals similarities and differences in the regulation of neuromuscular synapse development between two major divisions of the muscular system.

Compared with the neuromuscular system in axial muscle, little is known about either the steps by which the intricate innervation pattern of appendicular muscle arises or the genetic pathways that control this process. Our work reveals key differences in neuromuscular synapse development between axial muscle in the zebrafish trunk and appendicular muscle in the pectoral fin. First, innervation of the trunk myotome occurs much earlier than that of the pectoral fin; trunk motor axon outgrowth begins at 16 hpf (Panzer et al., 2006), whereas we first observed axon sorting at the fin plexus around 46 hpf. Second, both trunk and pectoral fin musculature were prepatterned with MuSK-dependent aneural AChR clusters. Surprisingly, prepatterned AChR clusters only form on adaxial, slow muscle fibers in the trunk (Flanagan-Steet et al., 2005; Panzer et al., 2006), and have not been reported in fast muscle fibers of the trunk, whereas, in the pectoral fin, which lacks slow muscle fibers, prepatterned AChR clusters formed exclusively on fast muscle fibers. This demonstrates that both slow and fast muscle fibers have the capacity to develop AChR prepatterning, and a crucial next step will be to identify mechanisms that selectively promote AChR prepatterning in trunk slow muscle and/or suppress it in adjacent fast muscle fibers. Third, in both zebrafish trunk and mouse diaphragm muscle, prepatterned AChRs are restricted to the central region of muscle fibers. We found that AChR prepatterning in pectoral fins was also restricted, but to the proximal region of individual muscle fibers. Without the ability to track AChR clusters in live animals, we cannot exclude the possibility that prepatterned AChR clusters initially arise in the ‘center’ of nascent pectoral fin muscle fibers and, as these fibers elongate, clusters become off-center and shift towards the proximal region of muscle fibers. Independent of this possibility, our analysis of appendicular NMJ development reveals that AChR prepatterning is not strictly restricted to the center but instead can localize to other muscle fiber areas.

Finally, early motor axon outgrowth in axial muscle is confined to the center muscle fibers, in which growth cones contact prepatterned AChR clusters that form in this region (Panzer et al., 2006). Similarly, we found that, during appendicular NMJ development time points, extending axons selectively grew towards AChR prepatterned muscle regions (Fig. 8). Yet, beyond the prepatterned region, axons in the pectoral fin formed intricate patterns across muscle fibers. This growth pattern is more similar to later axon outgrowth in the trunk in which motor axons branch to innervate fast muscle fibers deep in the myotome (Beattie, 2000). These differences in prepatterning and axon outgrowth might be a consequence of the anatomy of the pectoral fin, in which axons converge at a dorsal or ventral plexus prior to topographically innervating longitudinal muscle fibers. Indeed, pectoral fin neuroanatomy is similar to that of the muscles of tetrapod forelimbs, in which axons sort at the brachial plexus to innervate distinct muscles. Although we identified a conserved requirement for MuSK to establish prepatterning across muscles, these key anatomical and developmental differences between appendicular and axial muscles indicate that the signals that determine the location of AChR prepatterning and direct axon pathfinding are differentially regulated in appendicular muscle, leading to open questions regarding additional molecular mechanisms and pathway components that orchestrate NMJ development selectively in appendicular muscle.

Navigating growth cones respond to local extrinsic cues to determine where to form a synapse. This ‘stop signal’ requires MuSK signaling, because musk mutants in mouse and fish have an overgrown, defasciculated axon pattern (DeChiara et al., 1996; Zhang et al., 2004; Kim and Burden, 2008). In the zebrafish, the role of MuSK in axon patterning is independent of AChR clustering, because rapsyn mutants that lack clustered AChRs have normal axon patterning (Zhang et al., 2004; Gribble et al., 2018) (Fig. S2). Thus, MuSK independently clusters AChRs postsynaptically and regulates axon patterning, although the mechanism is poorly understood. Unexpectedly, we found that, in the pectoral fin, the presynaptic consequence of agrn or lrp4 loss is for multiple axons to form mature synapses in apposition to giant postsynaptic AChR clusters, suggesting an overactive ‘stop signal’ in these regions. These giant clusters might reflect local ‘islands’ of enhanced MuSK activity, because depletion of musk suppressed the formation of presynaptic swellings and giant AChR clusters. Taken together, these results suggest that Lrp4 signaling, induced by the arrival of axonally released Agrin, inhibits the MuSK-dependent ‘stop signal’ in the appendicular muscle of the pectoral fin. Such a mechanism could be a way to signal to new waves of navigating growth cones that this synaptic region is occupied.

Consistent with previous work, we found that loss of Agrin or Lrp4 led to a significant reduction in neural AChR clusters on appendicular muscle fibers, demonstrating a crucial and conserved role for both genes in promoting the formation of synaptic AChR clusters (Fig. 5C and 6D). Examining agrn/lrp4 mutants and nerve-deprived pectoral fins revealed a previously unappreciated role for Agrin/Lrp4 signaling. Whereas appendicular muscle fibers lacking Agrin or Lrp4 displayed a reduced number of neural AChR clusters, the clusters that formed were significantly larger. A developmental time course is consistent with the notion that these giant clusters are derived from MuSK-dependent prepatterned AChR clusters, which, in the absence of Agrin/Lrp4 signaling, expand over time. Combined, our findings strongly support the idea that Agrin and Lrp4 play a dual role to both promote the formation of axon-induced AChR clusters and, unexpectedly, constrain the growth and development of MuSK-dependent aneural clusters into synapse-associated AChR clusters.

One attractive mechanism for how MuSK might fulfil this dual role is through Wnt signaling. MuSK can bind Wnts through its extracellular cysteine-rich domain (CRD) (Jing et al., 2009; Strochlic et al., 2012; Zhang et al., 2012). Indeed, in axial muscle of the zebrafish trunk, Wnt4a and Wnt11r binding through the MuSK CRD are required for AChR prepatterning and axon guidance (Jing et al., 2009; Gordon et al., 2012). Although a functional role for Wnt:MuSK signaling to establish prepatterning in mouse is more controversial (Messéant et al., 2015; Remédio et al., 2016), Wnt proteins in mammals regulate both AChR clustering (Henriquez et al., 2008; Wang et al., 2008) and axon guidance (reviewed by Zou, 2004). Interestingly, given that the MuSK CRD domain can adopt two distinct conformations with differential abilities to bind Wnts (Stiegler et al., 2006), Guarino et al. speculate that Agrin/Lrp4 binding to MuSK can promote a conformational change that renders MuSK unable to bind Wnts, thereby shifting downstream MuSK signaling (Guarino et al., 2020). Therefore, it is tempting to speculate that, prior to axon innervation of appendicular muscle in the pectoral fin, MuSK:Wnt signaling promotes prepatterning of both AChRs and local extracellular matrix (ECM) cues to restrict navigating growth cones to future synaptic sites. Once an axon arrives, Agrin binding to Lrp4 induces a conformational change in MuSK, preventing Wnt binding. Given that there are 23 Wnts in the zebrafish genome (Lu et al., 2011) with dynamic and differential expression throughout larval development, relevant Wnt ligands are expected to be expressed in the pectoral fin but not in the trunk. Future studies will be necessary to identify the precise upstream and downstream signaling events that enable Agrin/Lrp4 signaling to simultaneously potentiate and attenuate MuSK signaling in appendicular muscle at different regions on the same muscle fiber.

One obvious question is how Agrin/Lrp4/MuSK signaling differs between axial and appendicular muscle. At the molecular level, it appears likely that either axial muscle in the trunk is exposed to an unknown signal (redundant with Agrin/Lrp4) that attenuates MuSK-dependent prepatterning or that the appendicular muscle in the fin is exposed to a signal that potentiates MuSK-dependent prepatterning. Independently of the underlying mechanism, our results uncover a novel role for Agrin/Lrp4 signaling in modulating MuSK-dependent AChR clustering in appendicular muscle. At a broader level, the noncanonical nature of our findings reveal diversity in the molecular pathways that mediate vertebrate neuromuscular synapse development and highlight zebrafish as a powerful model to study this process beyond axial muscle.

Protocols and procedures involving zebrafish (Danio rerio) were in compliance with the University of Pennsylvania Institutional Animal Care and Use Committee regulations. All transgenic lines were maintained in the Tübigen or Tupfel long fin genetic background and were raised as previously described (Mullins et al., 1994). The following transgenic lines were used: Tg(mnx1:GFP)^ml2 (Flanagan-Steet et al., 2005), Tg(α-actin:Lrp4-GFP)^p159Tg (Gribble et al., 2018), Tg(Xla.Tubb:DsRed)^zf148 (Peri and Nüsslein-Volhard, 2008), Gt(dmd-citrine)^ct90a [a kind gift from Dr Sharon Amacher (The Ohio State University, Columbus, OH, USA)] (Ruf-Zamojski et al., 2015) and Tg(α-actin:GFP) (Higashijima et al., 1997). The following mutant strains were used: agrn^p168 (Gribble et al., 2018), lrp4^p184 (Remédio et al., 2016), lrp4^mi36 (Saint-Amant et al., 2008), musk^tbb72 (Granato et al., 1996; Zhang et al., 2004) and rapsyn (two^th26) (Ono et al., 2002). Mutants homozygous for these genes can be phenotyped at ∼36 hpf because they all display motor defects when prodded with a probe. The lrp4^p184, two^th26 and agrn^p168 alleles were genotyped using Kompetitive Allele Specific PCR (KASP, LGC Biosearch Technologies). Animals were staged as previously published (Grandel and Schulte-Merker, 1998; Kimmel et al., 1995). Within all experiments, sibling animals from the same crosses were compared following genotypic and/or phenotypic analysis. Given that our experiments in larval zebrafish occurred prior to sex determination, sex was not a relevant biological variable in this study (Kossack and Draper, 2019).

Zebrafish embryos or larvae were immobilized with tricaine (A5040, Sigma-Aldrich, MS-222) and then fixed for 1 h at room temperature in sweet fix (4% paraformaldehyde with 125 mM sucrose in PBS) plus 0.1% Triton X-100 (Fisher, BP151). The embryos or larvae were then washed in phosphate buffer and incubated overnight at 4°C in either primary chicken anti-GFP antibodies (1:2000, Aves Labs, GFP-1010) or mouse anti-Znp-1 antibodies (1:200, Developmental Studies Hybridoma Bank, AB_2315626) in incubation buffer [2 mg/ml bovine serum albumin (BSA; Sigma-Aldrich, A2153), 0.5% Triton X-100 (Thermo Fisher Scientific, BP-151), 1% normal goat serum (S-1000, Vector Laboratories)]. After washing in phosphate buffer, animals were incubated overnight at 4°C in Alexa Fluor 488 donkey anti-chicken secondary antibodies (1:1000, Jackson ImmunoResearch, 703-545-155), Alexa Fluor 594 goat anti-mouse secondary antibody (1:1000, Invitrogen, A-21125) and/or α-Btx Alexa Fluor 594 conjugate (1:500, Thermo Fisher Scientific, B13423) in incubation buffer. Animals were then mounted in agarose in a glass-bottomed dish and imaged in 1.5 µm slices using a 40× or 63× water immersion lens on a Zeiss LSM880 confocal microscope using Zen Software (Fig. 2C and 8) or a 40× water immersion lens on an ix81 Olympus spinning disk confocal microscope using SlideBook Software. The representative images within each figure include only sibling animals that were fixed and immunostained in parallel.

A DNA vector encoding mnx1:mKate was injected as previously described (Gribble et al., 2018; Thermes et al., 2002) into one-cell-stage embryos. Embryos were screened at 1 dpf for sparse mKate expression in the anterior spinal cord. At 5 dpf, animals were mounted in agarose and imaged live with a 40× lens using an Olympus spinning disk confocal microscope if they had sparse mKate-expressing axons innervating the pectoral fin.

Embryos expressing both α-actin:GFP and Xla.Tubb:DsRed were anesthetized with tricaine and mounted in agarose at around 35 hpf. Animals were time-lapse imaged with a 40× lens using an ix81 Olympus spinning disk confocal microscope in a temperature chamber set to 28°C as previously described (Rosenberg et al., 2012). Stacks through the developing fin bud were captured in 1.5 µm slices at 30 min intervals. Animals were imaged continuously for up to 3 days, with some adjustments to account for drifting and the pectoral fin moving out of frame.

mnx1:GFP animals were mounted in agarose and motoneurons from spinal cord segments 3-5 were ablated using an Ablate! 532 nm attenuable pulse laser (3i Intelligent Imaging Innovations) beginning at 2 dpf, prior to axons innervating the pectoral fin bud. Neurons were considered ablated when there were no GFP+ cell bodies present in the ablated spinal cord region and axons showed signs of fragmentation. Ablation was repeated at 3 dpf to prevent any regenerated motoneurons from innervating the fin. Fins were visually inspected to confirm the absence of GFP+ motor axon signals within the denervated region of the fin. Animals were fixed and stained with α-Btx at 5 dpf.

At the time of fixation, 5 dpf lrp4 and sibling larvae were immobilized with tricaine and tails were removed for genotyping, whereas the anterior portion of the larvae was fixed with 2.5% glutaraldehyde, 2.0% paraformaldehyde in 0.1 M sodium cacodylate buffer, pH 7.4, overnight at 4°C. After washing for 10 min in 0.1 M sodium cacodylate buffer, the samples were post-fixed in 2.0% osmium tetroxide for 1 h at room temperature and then rinsed twice in distilled water prior to en bloc staining with 2% uranyl acetate. After dehydration through a graded ethanol series, the tissue was infiltrated and embedded in EMbed-812 (Electron Microscopy Sciences). Thin sections were stained with uranyl acetate and lead citrate and examined with a JEOL-JEM 1010 electron microscope fitted with a Hamamatsu digital camera and AMT Imaging Advantage image capture software.

To simplify data visualization and quantification, signal from abductor or adductor innervation was manually separated from stacks through pectoral fins. Individual channel image stacks were opened in Fiji (Schindelin et al., 2012), the background was subtracted, the channels were merged and the image was changed to RGB. Stacks were visualized using the 3D viewer plugin and rotated to a top-down view so that the separation between abductor and adductor innervation was distinct. Axon signal from the opposite innervation field and other fluorescent signal from the larval body wall were selected and filled, resulting in the corresponding region filled with black on the RGB stack. Any residual signal from the opposing innervation field or trunk was removed directly on the RGB stack. This resulted in signal specific to either the abductor or adductor muscles, as specified. Stacks were converted to maximum projections. For quantification of α-Btx puncta, a custom CellProfiler (Lamprecht et al., 2007) pipeline was created to detect and measure the area of α-Btx punta per fin. Mutants that did not form distinct α-Btx puncta (musk and rapsyn) could not be quantified using CellProfiler pipelines. Fin images were excluded from analysis if the maximum projection did not include the whole fin, the fin was damaged or abnormally small, or if measurements were clear outliers from other fins of the same genotype in the dataset. For analysis of α-Btx cluster size relative to fin position (Fig. S10), the x-axis pixel position for the center point of clusters was binned. The area of all individual clusters in each bin was averaged. All figures show only abductor innervation except for early developmental stages in Figs 2 and 7, which are a maximum projection of the entire pectoral fin bud or elsewhere, as noted.

Data were imported into GraphPad Prism for statistical analysis. Groups were compared using an unpaired t-test or one-way ANOVA with either Dunnett's or Sidak's multiple comparisons tests. For histograms of cluster sizes, the area of all AChR clusters measured in each genotype were pooled and binned into 5 µm² bins, with any cluster over 20 µm² included in the same bin. Distributions were compared using a Kruskal–Wallis test with Dunn's multiple comparisons test. In figures in which the control group is labeled as ‘siblings’, we pooled WT and heterozygous animals because there was no statistical difference between these groups.

The authors thank members of the Granato lab for helpful feedback on the manuscript. Finally, we thank the Penn Sanger sequencing core, the Penn Cell and Developmental Biology microscopy core, and the Electron Microscopy Resource Lab.