1. Neuroscience
Download icon

Functional limb muscle innervation prior to cholinergic transmitter specification during early metamorphosis in Xenopus

  1. Francois M Lambert  Is a corresponding author
  2. Laura Cardoit
  3. Elric Courty
  4. Marion Bougerol
  5. Muriel Thoby-Brisson
  6. John Simmers
  7. Hervé Tostivint
  8. Didier Le Ray  Is a corresponding author
  1. Université de Bordeaux, France
  2. Muséum National d’Histoire Naturelle, France
Research Article
  • Cited 1
  • Views 614
  • Annotations
Cite this article as: eLife 2018;7:e30693 doi: 10.7554/eLife.30693

Abstract

In vertebrates, functional motoneurons are defined as differentiated neurons that are connected to a central premotor network and activate peripheral muscle using acetylcholine. Generally, motoneurons and muscles develop simultaneously during embryogenesis. However, during Xenopus metamorphosis, developing limb motoneurons must reach their target muscles through the already established larval cholinergic axial neuromuscular system. Here, we demonstrate that at metamorphosis onset, spinal neurons retrogradely labeled from the emerging hindlimbs initially express neither choline acetyltransferase nor vesicular acetylcholine transporter. Nevertheless, they are positive for the motoneuronal transcription factor Islet1/2 and exhibit intrinsic and axial locomotor-driven electrophysiological activity. Moreover, the early appendicular motoneurons activate developing limb muscles via nicotinic antagonist-resistant, glutamate antagonist-sensitive, neuromuscular synapses. Coincidently, the hindlimb muscles transiently express glutamate, but not nicotinic receptors. Subsequently, both pre- and postsynaptic neuromuscular partners switch definitively to typical cholinergic transmitter signaling. Thus, our results demonstrate a novel context-dependent re-specification of neurotransmitter phenotype during neuromuscular system development.

https://doi.org/10.7554/eLife.30693.001

Introduction

The features that define a specific neuronal phenotype are generally conserved between species, are specified early during development, but can also undergo adaptive plasticity related to activity after maturation (Demarque and Spitzer, 2012; Borodinsky et al., 2014). Amongst the large variety of neuronal phenotypes, the developmental properties of the motoneuronal class are particularly well established and are shared by all vertebrates. In mammals, motoneuron (MN) specification begins in the ventral neural tube with the induction of progenitors by sonic hedgehog proteins (Roelink et al., 1995), the graded concentration of which triggers the subsequent expression of specific post-mitotic transcription factors (Goulding, 1998; Jessell, 2000). The homeodomain-containing protein Islet1 is the first molecular marker of MN differentiation (Ericson et al., 1992) and induces the later expression of homeobox Hb9, which consolidates the motoneuronal phenotype and participates in MN migration and central motor column formation (Arber et al., 1999). Soon after their specification, MNs express the two typical proteins associated with cholinergic neurotransmission, choline-acetyltransferase (ChAT) and the vesicular acetylcholine transporter (VAChT), enabling them thereafter to activate their muscle targets (Phelps et al., 1991; Chen and Chiu, 1992). The muscles, which develop concomitantly and contribute to motoneuron axon path-finding, also play a fundamental role in MN phenotyping and survival (Yin and Oppenheim, 1992; Kablar and Belliveau, 2005). Finally, MNs are considered to be fully functional once they have become affiliated to a corresponding central motor network and provide impulse-elicited excitation to muscle fibers using acetylcholine (ACh) as the neurotransmitter (Davis-Dusenbery et al., 2014).

In vertebrates, axial MNs innervating trunk muscles are distributed rostro-caudally along the spinal cord in the medial motor column (MMC), whereas fore- and hind-limb MNs are located in the brachial and lumbar lateral motor columns (LMC), respectively. In the amphibian Xenopus laevis, the axial and appendicular MNs controlling tail and limb muscles respectively are generated during two separate developmental periods. The former develop during pre-hatchling embryonic stages and control larval undulatory tail-based swimming by projecting to and exciting axial myotomes via nicotinic ACh receptor activation (van Mier et al., 1985; Sharpe and Goldstone, 2000). Thus, axial neuromuscular ontogeny takes place under conditions equivalent to those in mammals where the MNs and target muscles of a primary motor system develop simultaneously. In contrast, MNs of the neuromuscular system responsible for later limb-based locomotion differentiate and the limb buds start developing during early metamorphosis (Marsh-Armstrong et al., 2004), when the axial MNs and tail muscles are already entirely developed and operational.

Given that the primary axial and secondary appendicular MNs of Xenopus initially share the same anatomical and neurochemical environment, and because such features influence neuromuscular development (Yang and Kunes, 2004; Menelaou et al., 2015), it is conceivable that the emergence of the secondary limb MNs and their target muscles is susceptible to influences exerted by the already fully established and functional axial neuromuscular system. We thus hypothesized that the axial myotomes and their innervation constitute a potential interfering environment for the newly developing appendicular neuromuscular system, and that consequently, the latter may have to follow particular developmental rules adapted to this unusual context. In the present study, therefore, we investigated the developmental strategy employed by the early developing limb MNs and associated neuromuscular junctions, with a specific interest in exploring their neurochemical phenotype and functional capability, both in terms of spinal locomotor circuit interactions and limb bud muscle control. Our results show that during a brief pre-metamorphic period, the emerging limb MNs transiently express a non-cholinergic transmitter phenotype that involves glutamate, while exhibiting all other characteristics of typical vertebrate MNs.

Results

Spatial organization of the appendicular spinal motor column during early metamorphosis

Appendicular MNs already project into the hindlimb buds as soon as the latter begin to emerge at stage 48 (van Mier et al., 1985). At stage 50, the nascent limb consists of a tissue protrusion that is visible on the ventral side of the rostral tail myotomes, next to the larva's abdomen (Nieuwkoop and Faber, 1956). The limb bud then continues to grow and differentiate throughout the pre-metamorphic period, during which time the animal triples in size (Figure 1A1). At stage 56/57, the adult-like hindlimb is formed with differentiated thigh and leg segments along with the appearance of five webbed toes (Figure 1A2). The limb extensor and flexor muscle groups are also differentiated at this time.

Developmental stages of X. laevis and anatomical organization of the spinal motor columns.

(A) Relative size differences between stage 50 and 57 larvae (A1) and associated morphological changes of the hindlimb bud during metamorphosis onset (A2), from Nieuwkoop and Faber (1956). (B, C) Segmental (B1, C) and rostrocaudal (B2) organization of the spinal motor column region containing limb MNs at stages 50–52 (B1, B2) and 55 (C). Inset in B2 shows the location of labeled MNs in the mid-region of the spinal cord. (D) Schematic representation of the segmental organization of the appendicular and axial motor columns in the larval spinal cord. Ap./Ax. MNs, appendicular/axial motoneurons; Ap./Ax. d., Ap./Ax. dendrites; Vr, ventral root; c.c, central canal; D, dorsal; L, lateral. Scale bars: A1 and A2 = 5 mm; B1 and C = 20 µm; B2 = 100 µm.

https://doi.org/10.7554/eLife.30693.002

To investigate the central spatial organization of the motoneurons innervating the hindlimb muscles during this developmental period, we injected two different retrograde tracers into a hindlimb bud and ipsilateral axial myotomes of stage 50 to 57 animals to label the somata and dendrites of appendicular and axial MNs, respectively. Each population is located centrally in separate motor columns. Specifically, the axial motor column is distributed ventro-medially along the entire length of the spinal cord, whereas the hindlimb motor column is located more medio-laterally and restricted to spinal segments 7 to 9, identified by counting caudally the number of ventral roots from the obex (Figure 1B–D), as reported previously (Hulshof et al., 1987). In early stages 50–52 (Figure 1B1, B2), retrograde labeling from the limb bud revealed a high density of MNs with relatively small cell bodies (<10 µm) and a reduced dendritic arbor, located more dorso-laterally than the axial MNs. From stage 55 onward (Figure 1C) the appendicular motor column is positioned more laterally due to the enlargement of these spinal segments with a lower neuronal density. The appendicular MN somata then become bigger (20–50 µm) and acquire a characteristic elongated cell body shape and extended dendritic arbor, extending from a ventro-medial to dorso-lateral region of the hemicord (Figure 1C).

Delayed cholinergic transmitter phenotype expression in the developing appendicular motor column

The time course of ACh neurochemical ontogeny in appendicular MNs was first investigated using fluorescence immunochemistry for ChAT and VAChT expression in spinal cord cross-sections at different developmental stages (Figure 2A,B). At stage 53 (Figure 2A), appendicular MNs were not labeled at all with the ChAT and VAChT antibodies, whereas in the same spinal slices, axial MNs were strongly positive for the two proteins. In contrast, at later stages (e.g., stage 57 in Figure 2B) both ChAT and VAChT were clearly expressed in appendicular MNs (see lower left insets) as well as in axial MNs. Overall, ChAT and VAChT were not immuno-detected in appendicular MNs until stage 55, in contrast to axial MNs which were immunoreactive throughout the early pre-metamorphic stages.

Developmental emergence of the molecular ACh phenotype in appendicular MNs.

(A, B) Examples of fluorescence immunolabeling against ChAT (magenta) and VAChT (red) in appendicular MNs (Ap. MN) labeled with retrograde Alexa Fluor dextran 647 (AD 647, cyan) in stages 53 (A) and 57 (B) tadpoles. Insets (b) in B show x60 magnification of stage 57 appendicular MNs. (C) Variation of fluorescence (∆F/F) of axial (left plot) and appendicular (right plot) MNs for ChAT and VAChT immuno-signals at stages 49–51 (n = 4), stages 52–54 (n = 7) and stage 55–57 (n = 8). Grey dots represent the averaged ∆F/F values for ChAT (squares) and VAChT (circles) in each preparation, and the black dots are ∆F/F grand means ± SEM for all preparations in a given developmental group. ns non significant, ***p<0.001, Kruskall-Wallis test. (D) Examples of in situ hybridization (ISH) labeling for ChAT mRNA in the appendicular spinal column at stages 51 and 55. (E) Example of fluorescence immunolabeling against ChAT in stage 62 appendicular MNs that were previously labeled with AD 647 injected into the hindlimb at stage 51 (see schematic at left). All scale bars = 50 µm; D, dorsal; L, lateral.

https://doi.org/10.7554/eLife.30693.003

Semi-quantitative analysis of ChAT/VAChT fluorescence intensity in axial and appendicular MNs (see Materials and methods) was next performed on confocal image stacks in which appendicular MNs were identified by retrograde tracer labeling. The fluorescence variation (∆F/F) measured at stages 49–54 showed that the immuno-signal detected in the appendicular motor column was not significantly higher than the background signal level (ChAT: 13.5 ± 5.5 at stage 49–51; 1.2 ± 1.0 at stage 52–54; VAChT: 1.3 ± 0.8 at stage 49–51; 5.2 ± 2.0 at stage 52–54; Figure 2C, right). Immuno-labeling in the LMC became more robust with further larval development, with the appendicular/axial fluorescence variation increasing to ~20% (at stage 55–57; ChAT: 120.4 ± 31.8; VAChT: 72.4 ± 3.9; Figure 2C, right), which was consistent with the clear detection of both ChAT and VAChT in appendicular MNs at the later pre-metamorphic stages. In contrast, the fluorescence variation expressed in axial MNs (Figure 2C, left) for both ChAT and VAChT immuno-signal was significantly higher than the background signal level throughout the entire developmental period examined (ChAT: 644.7 ± 205.4 at stage 49–51; 849.1 ± 135.1 at stage 52–54; 1358.1 ± 270.9 at stage 55–57; VAChT: 273.9 ± 57.5 at stage 49–51; 267.5 ± 36.8 at stage 52–54; 872.5 ± 134.5 at stage 55–57; Figure 2C left).

In situ hybridization (ISH; n = 16) was performed for ChAT mRNA detection in a larval group ranging from stages 49 to 57. No signal was detected in the appendicular motor column area in stages 49–52 (Figure 2D left; n = 7). A ChAT ISH signal was then weakly detectable from stage 53–54 (n = 2), to become strongly evident from stage 55 onwards (Figure 2D right; n = 7). In contrast, a strong ChAT ISH signal was observed in axial MNs throughout the entire developmental period examined. Because ISH detects mRNAs that precede protein synthesis, our immunochemistry and ISH results are consistent and together confirm that appendicular MNs do not exhibit the cholinergic molecular phenotype prior to stage 55, although they innervate the hindlimb bud muscles from stage 48.

One possibility was that the non-cholinergic and cholinergic appendicular MNs observed at different developmental stages were in fact distinct populations. To address this possibility we first applied a fluorescent retrograde dye to the hindlimb bud at stage 51/52, prior to the appearance of the ACh phenotype in limb MNs (Figure 2E). Thereafter, the tracer-treated larvae were raised until they reached metamorphosis climax, by which time the appendicular system is presumably fully developed. ChAT immunolabeling then performed on spinal cross sections taken from these stage 62 animals strongly labeled the appendicular motor population (Figure 2E, middle panel). Significantly, some of these ChAT-positive MNs were also co-labeled with the retrograde fluorescent dye previously applied at earlier stage 51/52 (Figure 2E, right panel). Thus, these double-labeled MNs that innervated the hindlimb at stage 62 were already present at stage 52, before they expressed the cholinergic molecular phenotype. This finding therefore demonstrated that early non-ACh appendicular MNs constitute at least a sub-population of the cholinergic MNs that constitute the future mature appendicular motor innervation after metamorphosis.

Molecular and functional identification of early non-cholinergic appendicular neurons

The molecular identity of the non-ACh neurons innervating the newborn limbs at early pre-metamorphic stages was verified by immunochemistry against the transcription factor Islet1/2, a protein marker of motor neurons (Ericson et al., 1992). Double immunostaining in larvae younger than stage 55 revealed that the ChAT-negative neurons labeled with retrograde dye applied in the hindlimb bud were also strongly Islet1/2-positive, similar to the mature axial MNs present in the same slices (Figure 3A). Comparable results were obtained with ISH for Islet1 (Figure 3—figure supplement 1).

Figure 3 with 1 supplement see all
Motoneuronal identification of non-cholinergic limb projecting neurons.

(A) Example of fluorescence immunolabeling against ChAT (magenta) and Islet1/2 (orange) in appendicular and axial MNs (Ap. MN, Ax. MN) in a stage 52 tadpole. Appendicular MNs were previously labeled with retrograde Alexa Fluor dextran 647 (AD 647, cyan). Scale bar = 50 µm; D, dorsal; M, medial. (B) Protocol for intracellular recordings from appendicular MNs. Stage 52 MNs were identified by prior rhodamine dextran amine (RDA) retrograde labeling from the hindlimb bud and recorded with cell-attached patch-clamp in whole CNS preparations in order to preserve spinal locomotor circuitry. Alexa Fluor 488 added into the recording pipette allowed verifying the motoneuronal identity of the recorded cell. Recordings were made with either a low (black traces in C-F) or high [Cl] (blue traces in C-F) intra-pipette solution. (C) Increasing steps of injected depolarizing current elicited increasing spike discharge whereas release from hyperpolarizing current injection evoked rebound spiking. (D) Spontaneous MN (Intra) firing in the absence of axial ventral root (VRax) activity. Extended time scale is shown at right. (E-F) MNs received strong rhythmic synaptic input during spontaneous episodes of fictive axial locomotion, which triggered locomotor-related spiking in high [Cl]-recorded MNs (F). Extended time scale is shown at right. VRax traces (in yellow) are integrated transforms (Int.) of raw extracellular VR recordings.

https://doi.org/10.7554/eLife.30693.004

The electrophysiological properties of these non-ACh appendicular MNs were then tested at stage 52 by performing patch-clamp intracellular recordings of neurons identified by RDA retrograde labeling from the hindlimb bud (Figure 3B). These recorded MNs had average resting membrane potential values of −40.2 ± 8.2 mV (n = 12) and −54.8 ± 2.3 mV (n = 5) with low and high [Cl] patch solutions, respectively, and were able to produce action potentials either in response to increasing steps of depolarizing current injection (Figure 3C, left), on rebound after release from experimental hyperpolarization (Figure 3C, right) or even spontaneously (Figure 3D). Significantly, moreover, during spontaneous episodes of locomotor-like activity monitored from a more caudal spinal ventral root, correlated synaptic fluctuations were observed in all identified appendicular MNs, thereby indicating a functional synaptic coupling with other components of the spinal locomotor network (Figure 3E). In addition, the use of an elevated [Cl] solution in the intracellular electrode (see Materials and methods) revealed a capability for locomotor-related action potential firing in all recorded MNs (Figure 3F), and was consistent with the idea that chloride-mediated signaling was still not fully mature in limb MNs and remained excitatory at early developmental stages (Hanson and Landmesser, 2003; Akerman and Cline, 2006).

Altogether, these findings demonstrate that the retrogradely-labeled, non-ACh spinal neurons that prematurely innervate hindlimb buds express the motoneuronal marker Islet1/2 and present basic biophysical characteristics of functional MNs, being able to produce impulses as a function of membrane potential that in turn can be influenced via synaptic influences from central premotor circuitry.

Locomotor-related activation of hindlimb muscles by non-cholinergic motoneurons

Whether the early appendicular MNs are actually capable of driving muscle activation in the hindlimb buds before the emergence of the cholinergic phenotype was next investigated by making EMG recordings from both limb bud muscles and axial myotomes in semi-intact larval preparations (Figure 4A left panel). During episodes of spontaneous axial fictive swimming in a stage 53 preparation (n = 7; Figure 4A, right panel), locomotor commands monitored from a caudal spinal ventral root elicited rhythmic EMG activity in both the segmental myotomes and limb bud muscles. The ventral root bursts occurred in phase with EMG bursts in the recorded ipsilateral hindlimb muscle and in alternation with bursts in the contralateral myotome (Figure 4B, control). Under subsequent bath application of d-tubocurarine to block any nicotinic receptors and thereby prevent ACh-dependent synaptic transmission (Sillar and Roberts, 1992; 1993), the expression of swimming-related ventral root burst activity persisted, but associated tail myotome EMG activity was completely abolished after 15 min of antagonist perfusion (Figure 4B,d-tubocurarine). In contrast, even after >30 min d-tubocurarine perfusion the hindlimb bud muscles continued to exhibit rhythmic EMG discharges that, although reduced in amplitude (see below), remained strictly coordinated with the ongoing axial pattern (Figure 4B,d-tubocurarine; see black arrow in lower middle plot). The ability of both the tail myotomes and limb bud muscles to express locomotor-related activity recovered fully after 2 hr washout with normal saline (Figure 4B, wash). Semi-intact preparations older than stage 54 similarly exhibited EMG burst activity in their hindlimb buds and tail myotomes during fictive axial locomotion in control conditions (n = 5; e.g., stage 56 in Figure 4C, left). In these cases, however, a 15 min bath application of d-tubocurarine decreased the frequency of the centrally-generated axial swim pattern, and completely abolished, albeit reversibly, associated EMG activity in both the hindlimb bud muscles and tail myotomes (Figure 4C,d-tubocurarine; wash).

Figure 4 with 1 supplement see all
Switch from non-cholinergic to cholinergic limb muscle activation during axial swimming at different developmental stages.

(A) Rhythmic burst discharge (right panel) recorded from an axial ventral root (VRax) together with a hindlimb bud muscle (HLemg) and axial myotome (Axemg) during a spontaneous fictive swim episode in a semi-isolated stage 53 preparation (left panel). (B-C) Examples of recordings before (control), during and after (wash) bath application of d-tubocurarine to a stage 53 (B) and a stage 56 (C) preparation. Lower plots in B and C show instantaneous l-HLemg (blue) and r-Axemg (green) discharge rates (spikes/s) averaged (±SEM) over 10–20 l-VRax locomotor cycles (black). Black arrow in the middle graph of B indicates persistent limb bud EMG activity occurring in phase with the ipsilateral ventral root during d-tubocurarine application, and which was no longer present at the later developmental stage (c.f., middle plot in C).

https://doi.org/10.7554/eLife.30693.006

The substantial reduction in limb EMG activity by d-tubocurarine at developmental stages earlier than 55 (see Figure 4B) could have been due to a peripheral influence on transmission at the neuromuscular junction and/or result from an upstream action on the central activation of the limb motoneurons themselves. To distinguish between these two possibilities, we used immunolabeling and calcium imaging on completely isolated CNS preparations to assess whether functional cholinergic synapses are present on limb MNs within the spinal cord and are directly affected by nicotinic antagonist perfusion. Co-localized sites of VAChT and synapsin labeling were found on the somata of identified limb MNs (Figure 4—figure supplement 1A), which also expressed calcium fluorescence changes associated with bouts of fictive swimming (Figure 4—figure supplement 1B). Significantly, moreover, the amplitudes of these calcium signals were drastically reduced in the added presence of d-tubocurarine (Figure 4—figure supplement 1C,D). These findings thus indicated that the reduction in EMG activity observed in semi-isolated preparations (as seen in Figure 4B) might not have resulted from a peripheral action of the antagonist at the level of the neuromuscular junction (NMJ) itself, but rather, was largely a consequence of the antagonist's central nervous effects on actual limb MN activation. In this case, the unidentified premotor source of rhythmic drive to limb motoneurons can be presumed to use cholinergic signaling (Figure 4—figure supplement 1), although as suggested by the unmasking of depolarizing synaptic events by high electrode chloride concentrations (Figure 3F) and earlier findings that d-tubocurarine can block GABA-A receptors (Bixby and Spitzer, 1984), it is possible that GABA neurotransmission is also involved. The differential effects of d-tubocurarine at early and later larval stages also suggested that a developmental shift in the mechanism by which the appendicular MNs activate their target muscles during fictive locomotion occurs around intervening stages 54–55. Whereas neuromuscular transmission appears to be initially independent of ACh signaling in the pre-metamorphic tadpole, it evidently changes to a completely ACh-dependent process in older tadpoles. This in turn suggests that developmental modifications in parallel with the switch in MN transmitter signaling must also occur in the receptor phenotype of the hindlimb muscles themselves.

Developmental switch in hindlimb neuromuscular transmission

In the Xenopus embryo, immature axial MNs have been shown to co-express ACh and glutamate, although glutamate is reported to have no postsynaptic bioelectrical effects (Fu et al., 1998). Since our present data show that the appendicular MNs are able to activate hindlimb muscles before using ACh as a neurotransmitter, we asked whether glutamate may be involved in the early NMJ transmission process.

First, we explored in whole-mount hindlimb buds at various developmental stages (from 51 to 57) the putative expression and relative distribution of nicotinic and glutamate receptors with fluorescent α-bungarotoxin and the anti-NR2b antibodies, respectively. At stage 52, there was a total lack of labeling for α-bungarotoxin in hindlimb buds (Figure 5A, upper panels), indicating an absence of muscle ACh receptors at this stage. In contrast, at the same developmental stage, diffuse NR2b fluorescence, indicative of glutamate receptor labeling, was observed in the vicinity of MN axons visualized with neurofilament immuno-detection. A similar expression of NR1, another glutamate receptor subunit, was found which also paralleled synaptophysin labeling (Figure 5B). In addition, at stage 52, both the axons (Figure 5C) and cell bodies of limb MNs (Figure 5—figure supplement 1A) strongly expressed the vesicular glutamate transporter VGluT1, which again predominantly colocalized with synaptophysin and in the vicinity of postsynaptic NR1 labeling (Figure 5—figure supplement 1B1). Thus, in early pre-metamorphic development, vesicular glutamate transporters are present in hindlimb MNs, and concomitantly, glutamate receptors are expressed on hindlimb muscle fibers, close to synaptophysin-marked presynaptic terminals, consistent with the existence of functional glutamatergic NMJs.

Figure 5 with 1 supplement see all
Switch from glutamate to acetylcholine receptors in hindlimb muscles.

(A) Examples of hindlimb innervation patterns and distribution of ACh nicotinic receptors in a whole-mount hindlimb bud, revealed by fluorescence immunolabeling of neurofilament associated protein (Neurofil., red), glutamate receptor (NR2b, blue) and α-bungarotoxin labeling (α-bungaroTx, yellow) respectively, at stages 52 (upper panels) and 57 (lower panels). Inset drawings at bottom left of each panel show bud morphology at the two representative larval stages. Scale bars = 200 µm for stage 52, 100 µm for stage 57. (B) Examples of fluorescence immunolabeling against neurofilament associated protein (red), synaptophysin (Synaptoph., green) and NMDA glutamate receptor subunit 1 (NR1, blue) in whole-mount limb bud at stage 52. White arrowheads indicate sites of apposition of all three markers. Scale bar = 20 µm. (C) Examples of fluorescence immunolabeling against neurofilament associated protein (red), the glutamate vesicular transporter 1 (VGluT1, yellow) and the glutamate receptor subunit NR2b (blue) in whole-mount limb at stage 52. Scale bar = 50 µm.

https://doi.org/10.7554/eLife.30693.008

The first α-bungarotoxin positive signal was detected at stage 55 with characteristic dispersed dots (data not shown), as previously described (Marsh-Armstrong et al., 2004). By stage 57, typical clusters of nicotinic receptors were observed close to appendicular nerve terminal branches, and in further direct contrast to younger stages, NR2b (Figure 5A, lower panels) and NR1 receptor subunits (Figure 5—figure supplement 1B2) in hindlimb bud muscle as well as VGluT1 in innervating motor axons (Figure 5—figure supplement 1B2) were now absent. These findings therefore indicate that glutamate receptor subunits are no longer expressed in hindlimb muscle (and at least one vesicular transporter in motor axons) from stage 55 onwards, but rather, are superseded by nicotinic receptors in likely correspondence to the switch in NMJ transmission to a cholinergic phenotype.

In a final step, the electrophysiological response properties of neuromuscular transmission were tested on isolated hindlimb preparations over a similar range of developmental stages (from 52 to 57; Figure 6). Single pulse electrical stimulation applied to a limb bud motor nerve elicited EMG responses in the hindlimb bud with a mean latency of 7.8 ± 1.9 ms at stages 52–54 (n = 6) and 5.1 ± 2.1 ms at stages 55–57 (n = 6), and a mean amplitude of 0.30 ± 0.01 mV at stages 52–54 and 2.50 ± 0.99 mV at stages 55–57 in normal Ringer’s saline. Either d-tubocurarine or CNQX/AP5 cocktail was then bath-applied to test whether these neuromuscular responses were ACh- or glutamate-mediated, respectively. At stages 52–54, the appendicular nerve stimulus-evoked EMG response was not diminished by subsequent d-tubocurarine application, but rather, was increased by 35 ± 6% (Figure 6B, left; Fig; 6C, left black bar, p>0.01, U = 642.0). [Note that such a d-tubocurarine-induced enhancement of post-junctional responses has already been reported in other animal models and is unrelated to its antagonistic effects on nicotinic receptors themselves (Egan et al., 1993; Baron et al., 1996)]. Conversely, the hindlimb muscle response was decreased by 44 ± 3% following bath-application of CNQX/AP5 (Figure 6B, left; Figure 6C, left grey bar, p<0.001, U = 39.5). At stages 55–57 on the other hand, the nerve stimulus-evoked EMG response was decreased by 80 ± 2% under d-tubocurarine perfusion (p<0.001, U = 74.0) but was not significantly affected (p=0.5, U = 522.0) in the presence of CNQX/AP5 (Figure 6B, right; Figure 6C, right black and grey bars, respectively).

Functional switch in hindlimb neuromuscular transmission.

(A) EMG recordings (HLemg) from an isolated hindlimb preparation (schematic at left) in response to single pulse (10 µs) electrical stimulation (Stim) of a limb motor nerve (HL nerve) in stages 53 and 57 larvae. In each case, the black arrow indicates the beginning of the EMG response and the black trace represents the mean profile of 6 superimposed responses. Note that the trace illustration for stage 53 was taken from a 1 mM Mg2+ saline experiment. (B) Integrated motor nerve-evoked HLemg responses in control (black), and under d-tubocurarine (d-tubo.; red) or CNQX + AP5 (purple) bath application to stage 53 and 57 preparations. Thick lines represent the mean response profile (± SEM); the area under each curve (grey) was used to measure the response size. (C) Mean (± SEM) EMG response area as percentage of control response during d-tubocurarine (black) or CNQX/AP5 (grey) bath application at stages 52–54 and 55–57, respectively. **p<0.01 and ***p<0.001, Mann–Whitney U-test.

https://doi.org/10.7554/eLife.30693.010

Altogether, these immunochemical and pharmacological results showed that glutamate, but not ACh, is initially involved in hindlimb muscle activation, whereas NMJ transmission becomes ACh-mediated and increasingly efficient (increase in response amplitude, reduction in transmission delay) from stage 55 onwards. This therefore confirms the occurrence of a functional switch in Xenopus hindlimb muscle properties that is commensurate with the switch in appendicular MN neurotransmitter phenotype during the pre-climax phase of metamorphosis.

Discussion

In this study, we report that as early as pre-metamorphosis (i.e., prior to stage 55), de novo appendicular MNs express the Islet transcription factor, exhibit characteristic MN electrical properties and synaptic influences, are involved in locomotor bouts of activity, and project to limb bud muscles to evoke typical post-junctional responses. Unexpectedly, however, these MNs do not at this premature stage use standard cholinergic neurotransmission to activate their target muscles, but it is not until stage 55 that the limb MNs and their NMJs undergo a simultaneous switch from non-cholinergic to acetylcholine-dependent signaling. Thus our data show for the first time that in metamorphosing Xenopus, MNs that innervate the newly emerging hindlimbs first develop through the employment of a transient, but functional, alternative transmitter mechanism before a conventional and definitive cholinergic phenotype appears.

Classically in vertebrates, developing MNs are perpetually cholinergic and extend their axons out of the spinal cord towards target muscles to form functional NMJs (Phelps et al., 1991; Goulding, 1998). However, in Xenopus the ontogeny of appendicular MNs does not obey this well-established developmental pattern, since the present data indicate that before the acquisition of a cholinergic phenotype, these MNs are able to transmit centrally-generated motor commands to the developing limb bud muscles using a different transmitter effector(s). Several lines of evidence indicate that glutamate plays a major role in this precursor signaling at early metamorphic stages. Immunolabeling of stage 52 preparations revealed the presence of the vesicular glutamate transporter VGluT1 in identified (retrogradely-labeled) hindlimb-bud MNs, both in their cell bodies, as has been reported in mammalian CNS neurons (Nakamura et al., 2005; Yang et al., 2014), and peripheral axons (Melo et al., 2013). Significantly, we found the axonal VGluT1 to be co-localized with the synaptic protein, synaptophysin, indicative of a presence at the actual NMJ presynaptic terminals.

Also essential to synaptic signaling is the expression of appropriate receptors at the post-junctional level in correspondence with the type of pre-synaptic neurotransmitter released. Accordingly, we observed that the use of glutamate prior to ACh as a peripheral neurotransmitter is matched by the presence of glutamate receptors prior to the appearance of cholinergic receptors at the limb NMJs. Such a parallel developmental sequence is therefore consistent with a neurotransmitter phenotype switching, whereby intrinsic molecular processes lead to a change of one (or several) transmitter(s) to another within the same presynaptic neuron and a concomitant modification in associated postsynaptic receptor subtype in order to ensure synapse functionality. This loss/gain in pre-synaptic neurotransmitter/post-synaptic receptor phenotype has been reported to occur in developmental, post-lesional or activity-dependent contexts and is considered to be a major underlying feature of neuronal plasticity (Spitzer, 2017). It can either maintain or invert synaptic sign (Borodinsky et al., 2004; Borodinsky and Spitzer, 2007) and is generally associated with observable behavioral changes (Sillar et al., 1998; Demarque and Spitzer, 2010).

Our immunohistochemical evidence for an early role of glutamate at the developing limb NMJ was also supported by electrophysiological data on the effects of CNQX/AP5 on stimulus-evoked EMG responses in isolated motor nerve/limb muscle preparations. In direct contrast to older, post-stage 55 preparations where these glutamate antagonists were without any effect on evoked muscle potentials, at younger stage 52–54, EMG amplitudes in response to nerve electrical stimulation were strongly reduced. However, that this blockade remained only partial (ca. 50%, in the presence of CNQX/AP5 alone or in combination with the cholinergic receptor antagonist d-tubocurarine) might have been due to the low concentrations of glutamate ionotropic receptor antagonists used in our experiments (cf., Dale and Roberts, 1985) or that the antagonists were not fully effective on the still immature limb neuromuscular system. Alternatively, metabotropic glutamatergic receptors may be present (Pinard et al., 2003) and contribute to neurotransmission at the early developing limb NMJ or (an)other, as yet unidentified, neurotransmitter(s) may be involved.

In Xenopus, limb MNs are born during pre-metamorphosis, then their number diminishes after the establishment of NMJs (Hughes, 1961; Prestige, 1967). This in turn raises the possibility that the initial non-cholinergic motoneuron population we identify here constitutes a short-lived subclass of LMC neurons that project transiently into the limb buds at pre-metamorphic stages, but then is totally replaced by a different cholinergic population as metamorphosis proceeds. However, our finding that limb bud MNs previously retrogradely-labeled at stage 51 are still present at stage 62 when the population of appendicular MNs is fully established and uniformly cholinergic (Baldwin et al., 1988), including our early-labeled neurons, argues against this hypothesis. The possibility that the initial non-cholinergic phenotype may be restricted to a specialized motoneuronal sub-population that innervates limb muscle spindles is also unlikely since frog spindles are mainly innervated by collateral branches of skeletal MNs (Katz, 1949; Gray, 1957). Although a specific fusimotor innervation has been reported in the bull frog (Fujitsuka et al., 1987), since these MNs appear to be restricted to the semitendinosus muscle only, they would constitute a very discrete motor subset, the small proportion of which does not correspond to the large number of early non-cholinergic LMC neurons that we find in pre-metamorphic Xenopus. Thus, our data strongly indicate that the same motoneuronal population in the developing appendicular motor system of Xenopus does indeed utilize consecutively two different neurotransmitter signaling mechanisms in association with the emergence of hindlimb motility (Hughes and Prestige, 1967) and forthcoming limb-based locomotor and postural control (Combes et al., 2004; Beyeler et al., 2008).

Given that limb MNs acquire their cholinergic phenotype around stage 55 and remain cholinergic throughout adulthood (Baldwin et al., 1988), the existence of an early, brief but functional non-cholinergic phenotype is at first sight puzzling. In a broader context, a number of studies have reported the co-existence of various neurotransmitters in vertebrate MNs. For instance, axial MNs of Xenopus embryos co-express glutamate and ACh (Fu et al., 1998), while developing myotome fibers initially express a variety of receptors, including both cholinergic and glutamate subtypes, until NMJ formation when solely cholinergic receptors are preserved (Borodinsky and Spitzer, 2007). Co-released glutamate regulates the development and function of cholinergic neuromuscular synapses in larval zebrafish (Todd et al., 2004) and Xenopus (Fu et al., 1998) by potentiating ACh release through an activation of presynaptic receptors on the MN terminals themselves. Moreover, mammalian spinal MNs co-release glutamate and ACh centrally to activate Renshaw cells (Mentis et al., 2005; Nishimaru et al., 2005; Lamotte d'Incamps and Ascher, 2008), as well as at their peripheral terminals (Waerhaug and Ottersen, 1993; Rinholm et al., 2007) where post-synaptic muscle fibers possess both ACh and glutamate receptors (Mays et al., 2009). Here again, however, in no such case has glutamate been reported to produce muscle activation per se, but rather, this transmitter acts indirectly by regulating ACh's own impact at the NMJ (Vyas and Bradford, 1987; Malomouzh et al., 2003; Pinard et al., 2003) via the activation of post-synaptic NMDA receptors (Pinard and Robitaille, 2008; Petrov et al., 2013). On the other hand, glutamate is the predominant excitatory neurotransmitter in the vertebrate CNS, and supraspinal glutamatergic neurons can re-specify functional glutamatergic NMJs from otherwise purely cholinergic synapses on mammalian skeletal muscles following grafting with transected peripheral motor nerve (Brunelli et al., 2005). Given such a short-term reorganizing capability and the fact that glutamate is a major excitatory neurotransmitter at the NMJ of phylogenetically distant invertebrates (Gerschenfeld, 1973), it is possible that the transient employment of glutamatergic neuromuscular transmission in pre-metamorphosing Xenopus is representative of a latent ancestral step in the evolutionary transition of intrinsic molecular programming of the NMJ to cholinergic-dependent signaling.

A further and more appealing possibility is that the switching process is related to early appendicular MN axon path-finding and initial NMJ formation because of the unusual context of frog metamorphic development where secondary limb MNs axons must grow to their muscle targets with the primary axial neuromuscular apparatus already in place, fully functional and using ACh as the NMJ transmitter (Figure 7). Transplantation experiments in Xenopus have demonstrated the ability of MNs to innervate novel tissue targets (Elliott et al., 2013). Moreover, amongst the many guidance cues that orient axon elongation during development, both target-derived signals and axon growth cone-released ACh participate in target reaching and initial synapse formation (Yin and Oppenheim, 1992; Erskine and McCaig, 1995; Yang and Kunes, 2004). On this basis, therefore, it is conceivable that the pre-existing axial neuromuscular system in Xenopus provides a potential disturbing environment that could attract appendicular motor axons, if they were cholinergic, to make improper neuromuscular connections during pre-metamorphic development. On the contrary, a matched expression of glutamatergic neurotransmitter and receptors in appendicular MNs and limb muscles during pre-metamorphosis could ensure the correct orientation of the developing axon growth cones towards the limb buds and the establishment of synaptic contacts appropriate for the control of limb movements (Figure 7A). Supporting this hypothesis are previous findings that morphologically normal neuromuscular connections can still develop despite the experimental suppression of either pre- or postsynaptic cholinergic partners (Westerfield et al., 1990; Misgeld et al., 2002) and that glutamatergic CNS neurons can replace cholinergic MNs and reinnervate skeletal muscle fibers after a peripheral nerve lesion (Brunelli et al., 2005). It is also relevant that in addition to neural signaling, glutamate has been implicated in different neurotrophic functions, including cell growth and migration, during nervous system development (Nguyen et al., 2001). In Xenopus, once the initial appendicular NMJs are established, both pre- and postsynaptic partners could thereafter be instructed to switch to their definitive cholinergic phenotype (Figure 7B).

Schematic representation of neurotransmitter phenotype switching associated with the establishment of functional limb muscle innervation at different stages of Xenopus metamorphic development.

See text for further explanation. AChRs: nicotinic ACh receptors; GluRs: glutamate receptors; Ap. MN: appendicular MNs; Ax. MN: axial MNs; c.c: central canal.

https://doi.org/10.7554/eLife.30693.011

The signal for such neurotransmitter re-specification remains to be determined, but thyroid hormones, which control the developmental expression of ChAT (Patel et al., 1987; Gould and Butcher, 1989) and probably also VAChT since both proteins share the same gene locus (Eiden, 1998), are most likely to be involved. Consistent with this possibility is that the increase in thyroid hormone levels at metamorphosis onset, which starts at stage 54–55 in Xenopus (Shi, 2000), triggers a variety of gene-switching molecular programs required for the development of limb muscles and associated motor circuitry through the regulation of spinal cord neurogenesis and functions (Das et al., 2002; Marsh-Armstrong et al., 2004; Brown et al., 2005). Sensory feedback from new proprioceptors in the developing limbs may also participate in the respecification process, as found in the adult rat brain where the occurrence of novel sensory information can trigger neurotransmitter phenotype switching in postsynaptic central neurons (Dulcis et al., 2013).

In conclusion, the development of the limb neuromuscular apparatus and limb-based locomotion during Xenopus metamorphosis occurs in successive stages that involve a close functional relationship with the preexisting axial motor system until full autonomy is achieved (Combes et al., 2004) and, as shown here, is associated with changing underlying molecular patterning. Amongst the latter, neurotransmitter phenotype switching at the NMJ may enable limb motor axons to reach their appropriate muscle targets, which constitutes a novel and fundamental role for this process during motor innervation development (Spitzer, 2017) and adds to our understanding of NMJ development in general.

Materials and methods

Animals

Experiments were conducted on the South African clawed toad X. laevis obtained from the Xenopus Biology Resources Centre in France (University of Rennes 1; http://xenopus.univ-rennes1.fr/). Animals were maintained at 20–22°C in filtered water aquaria with a 12:12 hr light/dark cycle. Developmental stages were sorted according to external body criteria (Nieuwkoop and Faber, 1956), and experiments were performed on larvae from stage 49 to 57. All procedures were carried out in accordance with, and approved by, the local ethics committee (protocols #68–019 to HT and #2016011518042273 APAFIS #3612 to DLR).

Motoneuron retrograde tracing

Request a detailed protocol

Procedures used for neuronal retrograde tracing were as described previously (Bougerol et al., 2015). Briefly, animals were anesthetized in a 0.05% MS-222 water solution and transferred into a Sylgard-lined Petri dish. In order to backfill MNs from their muscle targets, the skin covering the muscles of interest was dried before a tiny incision was made and crystals of fluorescent dextran amine dyes were applied intramuscularly (Forehand and Farel, 1982; van Mier et al., 1985; Roberts et al., 1999). In most cases, only hindlimb MNs were labeled with either 3 kD rhodamine (RDA) or 10 kD Alexa Fluor 647 (Thermo Fisher, Illkirch, France), except in the experiments illustrated in Figure 1B,C where axial MNs were also labeled using 10 kD Alexa Fluor 488. Excess dye was washed out with cold Ringer solution (75 mM NaCl, 25 mM NaHCO3, 2 mM CaCl2, 2 mM KCl, 0.5 mM MgCl2, and 11 mM glucose, pH 7.4). After recovering from anesthesia, larvae were kept in a water tank for 24–48 hr to allow tracer migration into MN cell bodies and dendrites. In a series of experiments (n = 4), the hindlimb buds of stages 51–52 larvae were injected and the animals were kept for several days in a separate aquarium until reaching metamorphic climax (Figure 2E), in order to verify that early stage MNs were preserved through later development. Generally, such a labeling approach stains a large proportion of neurons that project processes within the bud muscles and thus potentially labels both MNs and sensory neurons (Forehand and Farel, 1982; van Mier et al., 1985; Roberts et al., 1999). However, since MNs have centrally located cell bodies, our retrograde tracings allowed the confident identification of MN somata only within the spinal cord.

Immunofluorescence labeling

Request a detailed protocol

After MN retrograde labeling, spinal cords were dissected out and fixed in 4% paraformaldehyde (PFA) for 12 hr at 4°C. Preparations were incubated in a 20% [in phosphate-buffered saline (PBS) 0.1%] sucrose solution for 24 hr at 4°C, then embedded in a tissue-tek solution (VWR-Chemicals, Fontenay-sous-Bois, France) and frozen at −45°C in isopentane. 40 µm cross-sections were cut using a cryostat (CM 3050, Leica, Nanterre, France). Fluorescence immunohistochemistry was carried out on these spinal cross-sections using the same protocol as described previously (Bougerol et al., 2015). Briefly, after several rinsing steps and the blocking of non-specific sites (using a solution with PBS, Triton X-100 0.3%, bovine serum albumin 1%; Sigma, St. Quentin Fallavier, France) samples were incubated with the primary antibody for 48 hr at room temperature. After rinsing, cross-sections were incubated for 90 min at room temperature with a fluorescently labeled secondary antibody, and washed again before mounting in a homemade medium containing 74.9% glycerol, 25% Coon’s solution (0.1M NaCl and 0.01M diethyl-barbiturate sodium in PBS), and 0.1% paraphenylenediamide. The primary antibodies used were goat anti-ChAT (1:100; Millipore), rabbit anti-VAChT (1:1000; Santa-Cruz) and mouse anti-Islet1/2 (1:250; Developmental Studies Hybridoma Banks (DHSB), University of Iowa, Iowa city, US). Secondary antibodies were donkey anti-goat and anti-rabbit or anti-goat and anti-mouse IgGs coupled to Alexa Fluor 488 and 568 (1:500; Life Technologies).

Fluorescent immunohistochemistry was carried out on entire or 20 µm sliced hindlimb buds from developmental stages 51 to 57 after overnight fixation in PFA 4%. The same labeling protocol was used as described above for spinal cross-sections. Alexa Fluor 488-conjugated α-bungarotoxin (10 μg/ml; Life Technologies) was used to label neuromuscular junctions. The primary antibody, mouse anti-neurofilament associated protein (3A10; 1:100; DHSB), was used to label nerve branches innervating the limb bud. The primary antibodies, rabbit anti-synaptophysin (1:500; abcam, Paris, France), goat anti-NMDA subunit one receptor (NR1; 1:100; abcam) or rabbit anti-NR2b (1:200; abcam), and guinea pig or mouse anti-VGluT1 (1:100; respectively from Millipore, France and Synaptic Systems, Germany), were used to label synapses, glutamate receptors, and vesicular glutamate transporters, respectively. Note that comparable results were obtained with both anti-VGluT1 antibodies. The bud preparations were then incubated with secondary antibodies donkey anti-mouse, anti-rabbit and anti-goat IgGs coupled to Alexa Fluor 488, 568 and 647 (1:500, Thermo Fisher). For microscope imaging whole-mount limb buds were mounted on cavity slides in a homemade medium (see above).

Image acquisition and fluorescence quantification

Request a detailed protocol

Whole-mount preparations and cross-sections labeled with fluorescent material were imaged using an Olympus FV1000 confocal microscope equipped with 488, 543 and 633 nm laser lines. Images were processed using Fiji and Photoshop softwares. Multi-image confocal stacks with 1 μm z-step intervals were generated using a 20x/0.75 oil objective and with 0.3 µm z-step intervals using a 60x/1.4 oil objective. Figure images were obtained by orthogonal projection from multi-image stacks with artificial fluorescent colors using the freeware Fiji.

ChAT, VAChT and Islet1/2 fluorescence quantifications were performed on original images from 0.3 μm z-step stacks. Fluorescence intensity was measured automatically from 3 ROI in single planes with the same size, defining the slice background, and the axial and appendicular MNs fluorescence signals, respectively. The variation of fluorescence (∆F/F = (F-F0)/F0) was calculated for both axial and appendicular MNs ROI relative to background, the latter being acquired in a ventral spinal region devoid of axial and limb MN cell bodies, yet where cholinergic terminals were present. This calculation was performed on five consecutive confocal planes per slice where the appendicular motor column was identified by retrograde labeling (usually 2 to 5 40 µm slices) and where background noise was maximal. Preparations with concomitant ChAT and VAChT fluorescent labeling were combined for statistical quantification (Figure 2C).

In situ hybridization

Request a detailed protocol

Because of its tetraploid status, X. laevis possesses two genes for ChAT (Genbank accession numbers: XM_018225547 and XM_018239425 for ChAT1 and ChAT2, respectively) that exhibit about 87% nucleotide identity. This prompted us to generate two ChAT probes. To synthesize these probes, two PCR fragments of 529 (ChAT1) and 512 (ChAT2) bp were amplified from adult X. laevis brain and spinal cord RACE-ready cDNA (Bougerol et al., 2015), then subcloned into pGEM-T easy (Promega, Charbonnières, France). The pairs of primers used were: 5′-TTTGCTGCCAACCTTATCTCTG-3′ and 5′-ATGAAGTAACTGCAAAGCCCTG-3′ for ChAT1, and 5′-TCTGGAGTGCTGGATTACAA-3′ and 5′-ATGAAGTAACTGCAAAGCCCTG-3′, for ChAT2. Sense and antisense digoxigenin-labeled riboprobes were synthesized from the linearized plasmid with the RNA polymerases T7 or Sp6, using the RNA Labeling Kit (Roche Diagnostics, Mannheim, Germany).

The trunk region of tadpoles from stages 49 to 55 was dissected, fixed with 4% PFA in 0.1 M PBS overnight at 4°C and rinsed in 0.1 M PBS. Fixed samples were cryoprotected in 15% then 30% sucrose/PBS and embedded in Tissue-Tek (Sakura, Netherlands). Frontal sections (20 µM) of the trunks were cut at −20°C using a cryostat, collected on Superfrost Plus slides (O. Kindler, Freiburg, Germany), dried at room temperature for 24 hr and stored at −80°C until use.

The in situ hybridization protocol used in the present study was adapted from earlier studies (Buresi et al., 2012; Bougerol et al., 2015) and consisted of the following steps. Briefly, sections were rinsed 2 × 5 min with PBS at room temperature, 15 min in five times concentrated sodium chloride and sodium citrate solution (5X SSC), then were incubated for 2 hr in prehybridization buffer (50% formamide, 5X SSC, 50 μg/ml heparin, 5 mg/ml yeast RNA, 0.1% Tween) at 65°C. When prehybridization was complete, the prehybridization solution was removed and replaced with the same buffer containing a mix of the two heat-denaturated digoxigenin-labeled ChAT1 and ChAT2 riboprobes. Hybridization was carried out overnight at 65°C. Sections were rinsed 3 × 30 min in 2X SSC at 65°C then 1 hr in 0.1 X SSC at 65°C. Two final washes were performed for 5 min in MABT (maleic acid 100 mM, pH 7.2, NaCl 150 mM, Tween 0.1%) at room temperature. Sections were transferred to a blocking solution [5% blocking reagent (Roche), 5% normal goat serum in MABT] and incubated at room temperature for 1 hr, before addition of the alkaline phosphatase coupled anti-digoxigenin antibody (1/4000) for overnight storage at 4°C. Sections were again washed 3 × 10 min in MABT and 2 × 5 min in PBS at room temperature, then incubated 10 min in staining buffer (100 mM Tris-HCl, pH 9.5, 50 mM MgCl2, 100 mM NaCl, and 0.1% Tween) and transferred to BM Purple (Roche) for colorimetric detection. Finally, they were washed twice in PBS to stop the reaction, and then mounted on gelatin-coated slides in Mowiol. Sections were imaged using a Leica DM5500 B microscope connected to LAS V4.1 software. The specificity of the hybridization procedure was verified by incubating sections with the sense riboprobes with which only background signals typical of this type of chemical reaction could be observed.

Patch-clamp recording of appendicular motoneurons

Request a detailed protocol

Retrograde labeling of appendicular MNs was performed using 3kD RDA dextran on stages 51, 52 larvae. The day after, patch-clamp electrophysiological recordings of labeled MNs were made on isolated brainstem-spinal cord in vitro preparations (n = 7). After anesthesia in 0.05% MS-222, the brainstem and spinal cord, including spinal segmental ventral roots, were dissected out in cold oxygenated (95% O2, 5% CO2) Ringer solution. The preparation was then placed in a recording chamber and continuously superfused with oxygenated Ringer solution (~2.5 mL, ~17°C, rate of ~2 mL/min). Spontaneous fictive locomotor episodes were recorded from a caudal ventral root (between segments 12 and 15; Figure 3B) using a borosilicate glass suction electrode (tip diameter, 100 nm; Clark GC 150F; Harvard Apparatus) filled with Ringer solution. The recorded signal was amplified (A-M system), rectified and integrated (time constant 100 ms; Neurolog System). RDA-positive appendicular MNs were identified with a standard epifluorescent illumination system (Cy3 filter) within the whole-mount spinal cord (dorsal-side opened) and subsequently visualized using a differential interference contrast microscope with an infrared video camera to facilitate the patch electrode trajectory (Figure 3B). Using an Axoclamp 2A amplifier (Molecular Devices, Berkshire, UK), whole-cell patch-clamp recordings were made with a borosilicate glass electrode (pipette resistance, 5–6 MΩ; Clark GC 150TF; Harvard Apparatus) filled with a solution containing (in mM) 100 K-gluconate, 10 EGTA, 2 MgCl2, 3 Na2ATP, 0.5 NaGTP, 10 HEPES, pH 7.3. In additional experiments (n = 2), to impose an elevated intracellular chloride concentration corresponding to that of immature neurons (Ben-Ari, 2002), including those in Xenopus (Akerman and Cline, 2006), this low [Cl] recording solution was replaced by a high [Cl] version that contained (in mM) 70 K-gluconate, 30 KCl, 10 EGTA, 2 MgCl2, 3 Na2ATP, 0.5 NaGTP, 10 HEPES, pH 7.3. Alexa Fluor 488 (Life Technologies), which was added in the patch pipette to fill the recorded neurons, allowed subsequent verification that recorded cells were indeed RDA-positive (Figure 3B). All electrophysiological signals were computer-stored using a digitizer interface (Digidata 1440; Pclamp10 software; Molecular Devices) and analyzed offline with Clampfit software (Molecular Devices).

Axial and hindlimb EMG recordings in semi-intact preparations

Request a detailed protocol

Semi-intact preparations from stages 52 to 57 larvae were used to simultaneously record EMG activity from axial myotomes and hindlimb bud muscles (Figure 4; n = 12). Brainstem-spinal cord preparations were dissected out in the same way as for patch-clamp recording, but tail myotomes (7-10) and hindlimb buds were left attached to the spinal cord. Semi-intact preparations were fixed in a Sylgard-lined recording chamber, continuously superfused with oxygenated Ringer solution (1.3–2.1 mL/min) and maintained at 18 ± 0.1°C with a Peltier cooling system. In some experiments, d-tubocurarine (30 µM; Sigma) was exogenously applied to block nicotinic receptor-type cholinergic synapses. Fictive locomotion sequences were generated either spontaneously or triggered by electrical stimulation of the caudal region of the brainstem (Grass stimulator S88), and the spinal swimming pattern was monitored from a caudal ventral root (between segments 12 and 15) with a suction electrode as described above. EMG activities in rostral myotomes (7–10) and hindlimb buds were recorded simultaneously using pairs of 50 µm insulated wire electrodes connected to a differential AC amplifier (A-M System). Both nerve and EMG activities were digitized at 10 kHz (CED 1401, Cambridge Electronic Design, UK), and displayed and stored on computer for offline analysis with Spike2 software (CED).

Discharge rates in individual nerve and EMG recordings were measured by setting an amplitude threshold to count all impulses in such multi-unit recordings. Firing rates (in spikes/s) were averaged over 10–20 locomotor cycles. Cycle period was taken as the interval between the onsets of two consecutive ventral root bursts. These consecutive burst onsets were used as a trigger for averaging the discharge rates of each EMG channel over cycle duration.

EMG recordings and drug applications in isolated nerve-limb bud preparations

Request a detailed protocol

Isolated hindlimb-bud preparations with the sciatic and crural nerves still attached (Figure 6; n = 17) were used to record the EMG activity evoked by electrical stimulation of either appendicular nerve branch at stages 52 to 57. Under MS-222 anesthesia, appendicular nerves were disconnected from the spinal cord and separated from tail myotomes, taking care not to detach them from the rest of the bud. The bud and attached nerves were fixed with small pins in a Sylgard-lined recording chamber and superfused with oxygenated Ringer solution. A small incision was made at the distal extremity of the bud to allowing insertion of the EMG electrodes. Either limb nerve branch was stimulated with a glass suction electrode connected to a Grass stimulator S88 through a photoelectric stimulus isolation unit (PSIU 6; Grass Instruments). Note that stimulating either branch provided similar results, and no distinction was made in this report. Single pulses (70–300 µA; 10 µs) were delivered every 100 s, and polarity inversion tests were performed in order to distinguish the stimulation artifact from the muscle response. EMG signals were amplified, integrated (time constant 10 ms) with Spike2 software and stored as described above. EMG responses were measured as the area under the integrated recording traces.

30 µM d-tubocurarine or a cocktail of 20 µM CNQX +10 µM AP5 were added to the perfusion solution to block nicotinic or glutamatergic receptors, respectively (all drugs from Sigma). Because of the typical difficulty in washing-out these drugs in such experiments, a second antagonist was generally added while the first one was still present in the bath. Control experiments where only one drug was applied showed effects similar to those of combined drug application, and thus data from either approach were pooled in this study. In some experiments on early stages, the normal Ringer solution was replaced by a solution containing half (1 mM) of the normal concentration of MgCl2 in order to unmask any NMDA receptor-mediated component of the EMG’s glutamatergic response. Despite causing a noticeable increase in the control EMG amplitude, such a low Mg2+ solution had no influence on the effects resulting from subsequent antagonist application.

Statistics

After signal processing in Spike2, electrophysiological data were analyzed using Prism5 (GraphPad, USA) and OriginPro8 (OriginLab Corporation, USA). Data are shown as means and standard errors of the mean (± SEM), unless stated otherwise. For ChAT and VAChT immunofluorescence signals (Figure 2C), differences between preparation means were tested using the Kruskall-Wallis ANOVA multi-comparison test (Student-Newman-Keuls method; statistical significance is indicated in Figure 2C). For integrated EMG signals, differences of means were tested using the unpaired two-tailed Mann–Whitney U-test (statistical significance and U values are indicated in the corresponding Result section).

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
    Coexistence of glutamate and acetylcholine in the developing motoneurons
    1. WM Fu
    2. HC Liou
    3. YH Chen
    4. SM Wang
    (1998)
    The Chinese journal of physiology 41:127–132.
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
    The spindle and extrafusal innervation of a frog muscle
    1. EG Gray
    (1957)
    Proceedings of the Royal Society B: Biological Sciences 146:416–430.
    https://doi.org/10.1098/rspb.1957.0021
  35. 35
  36. 36
  37. 37
    Cell degeneration in the larval ventral horn of Xenopus laevis (Daudin)
    1. A Hughes
    (1961)
    Journal of Embryology and Experimental Morphology 9:269–284.
  38. 38
    The distribution of motoneurons supplying hind limb muscles in the clawed toad, xenopus laevis
    1. JB Hulshof
    2. R de Boer-van Huizen
    3. HJ ten Donkelaar
    (1987)
    Acta Morphologica Neerlando-Scandinavica 25:1–16.
  39. 39
  40. 40
  41. 41
    The efferent regulation of the muscle spindle in the frog
    1. B Katz
    (1949)
    The Journal of Experimental Biology 26:201–217.
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
    Roles of neurotransmitter in synapse formation: development of neuromuscular junctions lacking choline acetyltransferase
    1. T Misgeld
    2. RW Burgess
    3. RM Lewis
    4. JM Cunningham
    5. JW Lichtman
    6. JR Sanes
    (2002)
    Neuron 36:635–648.
  50. 50
  51. 51
  52. 52
    Normal Table of Xenopus Laevis (DAUDIN). a Systematical and Chronological Survey of the Development From the Fertilized Egg Till the End
    1. PD Nieuwkoop
    2. J Faber
    (editors) (1956)
    Amsterdam: North Holland Publishing Co.
  53. 53
  54. 54
  55. 55
  56. 56
  57. 57
  58. 58
  59. 59
    The control of cell number in the lumbar ventral horns during the development of xenopus laevis tadpoles
    1. MC Prestige
    (1967)
    Journal of Embryology and Experimental Morphology 18:359–387.
  60. 60
  61. 61
  62. 62
  63. 63
  64. 64
  65. 65
    Amphibian Metamorphosis: From Morphology to Molecular Biology
    1. YB Shi
    (2000)
    New York: Wiley-Liss Inc.
  66. 66
  67. 67
  68. 68
  69. 69
  70. 70
  71. 71
  72. 72
  73. 73
  74. 74
  75. 75
  76. 76
  77. 77

Decision letter

  1. Carol A Mason
    Reviewing Editor; Columbia University, United States

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your article "Functional limb muscle innervation prior to cholinergic transmitter specification during early metamorphosis in Xenopus" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Eve Marder as the Senior Editor.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

The reviewers found your study on changes in limb motor neuron transmitter phenotype and the composition of receptors at the frog neuromuscular junction of interest and well-executed. The reviewers welcomed your evidence that during metamorphosis when the limbs are forming and becoming innervated, motoneurons initially release glutamate, and then acetylcholine at their synapses on limb musculature. They agree that the manuscript is clearly written and for the most part well-illustrated, especially the use of retrograde labeling to identify and distinguish appendicular and axial motoneurons.

However, the three reviewers concurred on amending four aspects, all of which would greatly strengthen your study and begin to shed light on mechanism. They expected that these amendments could be accomplished within eLife's expected timeline for revisions:

1) Demonstration of the presence of any of the known vesicular glutamate transporters in motoneurons at the expected ages.

2) Better explanation in the text of the large effect of d-tubocurarine. There is a discrepancy in the reported effects of applying t-tubocurarine at Stage 53, leading to a large decrease in amplitude of EMG activity, and the localization of receptors. Reviewer 2 asks whether you could test the effects of tubocurarine or glutamate antagonists at an earlier stage (e.g., 52).

3) Inclusion in the Results section of the electrophysiological recordings showing the effects of glutamatergic antagonists. This experiment was alluded to in the Materials and methods section in which a cocktail of CNQX and AP5 was used to block glutamatergic receptors and are critical to the arguments of the paper.

4) Better and more complete statistical analyses: Reviewer 3 notes a lack of statistical analysis and an indication of the number of experiments used to generate the quantitative data, a lack of statistical comparison between the fluorescence levels at the different ages, and confusion over standard error of the mean and standard deviation used to illustrate the variability.

Reviewer #1:

1) Subsection “Locomotor-related activation of hindlimb muscles by non-cholinergic motoneurons”, "is initially largely independent of ACh signaling in…" The amplitude is reduced by more than 50% by d-tubocurarine. How about "is initially only partially suppressed in…."? Could the residual signals be stimulus artifacts? In contrast to Control and Wash, the amplitudes are constant in amplitude. The Materials and methods section describes experiments in which a cocktail of CNQX and AP5 was used to block glutamatergic receptors. These experiments should be illustrated here. They are critical to the story.

Related to this point, subsection “Developmental switch in hindlimb neuromuscular transmission”, "indicating the absence of anatomically detectable muscle ACh receptors" would be more appropriate. Figure 4B suggests that they are detectable by physiological / pharmacological methods.

Again, related to this point, in the Discussion section, I find "indicative of a complete neurotransmitter phenotype switching…" too great a claim, given that d-tubocurarine substantially reduces the amplitude of the l-HLemg in Figure 4B.

2) Discussion section, the authors argue against a muscle spindle innervation role for the limb motoneurons they have studied, on the basis of the sizes of the muscle spindle- and muscle fiber-innervating populations of neurons. However, in the absence of information about the sizes of these populations this is not convincing.

3) Quantification is inadequate. The numbers of animals in which observations were made should be provided in each figure legend. Contrary to the assertion in the Transparent Reporting Form, the number of replicates is largely omitted in the Results section and Figure Legends.

Reviewer #2:

1) Figure 4: It is argued that the switch from glutamatergic to cholinergic transmission occurs around stages 54-55. However, the results presented in Figure 4 clearly show that cholinergic transmission is at place already at stage 53 and possibly even at earlier stages. There is a large decrease in the amplitude of the EMG activity by d-tubocurarine.

Could these experiments be done at stage 52? Alternatively, is it experimentally possible to apply glutamatergic antagonists locally on the limb bud?

The results of Figure 4 show clearly that the EMG activity at stage 53 is largely mediated by t-tubocurarine-sensitive receptors. This is at odds with the main conclusion of this study.

2) Figure 5C,D: The EMG activity recorded at stage 53 is very weak and this may indicate that the neuromuscular junction is not fully developed.

The authors should provide electrophysiological traces showing the effects of glutamatergic antagonists. Also, the weak EMG activity recorded at stage 53 might be due to the absence of activation of NMDA receptors. I suggest to repeat these experiments using a Mg-free extracellular solution.

3) Overall, it is not clear if limb motor neurons switch the transmitter phenotype or if they use both glutamate and acetylcholine as co-transmitters with a dominance of the one or the other at different developmental stages. In this regard, the authors could test if limb motor neurons co-express transporters for glutamate and acetylcholine in different proportions at different stages.

Reviewer #3:

The central thesis of this paper is that motoneurons transiently release glutamate instead of acetylcholine at the neuromuscular junction of developing hindlimb muscles at the onset of metamorphosis. The main evidence in support of this conclusion comes from the immunocytochemistry showing that appendicular motoneurons do not express either ChAT or VAChT immunoreactivity and exhibit very little ChAT mRNA at stage 53. Moreover, at stage 52, the developing limb muscle has very low levels of α-bungarotoxin but does express NMDA receptor subunits some of which are co-localized with neuronal synaptophysin (although many are not).

Unfortunately, the physiological experiments do not completely support the anatomy. For example, at stage 53 bath application of d-tubocurarine greatly reduces the amplitude of the rhythmic EMG signals from the limb bud during locomotor episodes despite the apparent absence of acetylcholine receptors in the muscle and cholinergic markers in limb motoneurons. The authors do not explain why the EMG is reduced at this age. Similarly, in the nerve/limb preparation illustrated in Figure 5, bath application of CNQX/AP5 only reduces the stimulus-evoked EMG by 50%. Why is the block incomplete? Again, this is not discussed by the authors. The authors showed only the presence of NMDA receptor subunits in the muscle but did not show the effect of NMDA antagonists alone on the evoked EMG to validate their presence.

In both experiments, it would have been useful to combine the glutamatergic and cholinergic antagonists to see if this produced a complete blockade of the muscle electrical activity.

Finally, in Figure 5D, the results for control and d-tubocurarine are shown, but it would also be of relevance to show a recording for the glutamate antagonists. Here, the number of experiments is missing, and no p-values are reported.

The intracellular recordings of the locomotor activity illustrated in Figure 3E are puzzling. The recordings reveal rhythmic hyperpolarizing potentials and the cells do not fire. Was this true of all of the motoneurons that were recorded? If so, how are the muscles activated if the motoneurons do not spike? For the 11 recorded motoneurons, the authors should indicate from how many preparations they were obtained, and if the motoneurons were only identified by their labeling or whether they tested for an antidromic response. If it is the former, did the authors ever use a second dye applied intracellularly to confirm that the labelled motoneuron was indeed the one that was recorded from.

At the earliest stages of limb formation, it was stated that motoneurons were small cells. However, the resistance of the cell shown in Figure 3C was between 3 to 5 MOhm suggesting either a very leaky cell or a large one. What was the input impedance of the other motoneurons?

As already alluded to, a major concern is the lack of statistical analysis and an indication of the number of experiments used to generate the quantitative data. For example, in subsection “Delayed cholinergic transmitter phenotype expression in the developing appendicular motor Column” there is no statistical comparison between the fluorescence levels at the different ages nor is the number of experiments shown. Similar problems apply to the quantification in other parts of the paper. The authors often refer to the mean and the SE. Do they mean the standard error of the mean (S.E.M). Why was the standard error of the mean rather than the standard deviation used to illustrate the variability?

The quantification of the fluorescence in Figure 2C was confusing. In the text, it states that ChAT and VAChT were not immuno-detected in appendicular MNs until stage 55. The figures in Figure 2A confirm this. However, in the plot of 2C (52-54), which shows the AP/AX fluorescence ratio, the values range from ~0-14% for ChAT. It is difficult to see how these values are generated if AP is zero. The authors should indicate how the ratio was calculated.

Also, despite an increase in the ratio at later stages, it is still about 20% of the axial fluorescence (when appendicular MNs are supposed to be bigger). This implies that the majority of appendicular motoneurons are still not cholinergic. It would be helpful to show that once the metamorphosis is complete, this ratio is 100% thereby validating the measure.

I found the in-situ hybridization results shown in Figure 2D, difficult to interpret. It is difficult to know the extent of the ChAT ISH labelling in these Figures For example, the extent of the Axial motor column in D 53 looks much more limited than the ChAT labelling in A 53. Furthermore, it is not clear that appendicular MNs at stage 53 are negative. The background is very high. Sections with dextran labelling would have made clear where the motoneurons are located. This figure does not agree with the statement that ISH is more sensitive than immunodetection. How many times was this experiment performed?

[Editors' note: further revisions were requested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Functional limb muscle innervation prior to cholinergic transmitter specification during early metamorphosis in Xenopus" for further consideration at eLife. Your revised article has been favorably evaluated by Eve Marder (Senior Editor), a Reviewing editor, and three reviewers.

The manuscript has been improved and reviewers 1 and 2 were satisfied with your amendments, but there are some remaining issues raised by reviewer 3 that remain to be addressed, indicated in the appended reviews. The other reviewers and I in consultation now agree on the need for these revisions:

1) The verity of the VGluT1 localization, in synaptic terminals rather than in cell bodies; the predicted restriction of vGluT1 to motor neuron terminals and colocalization with synaptic proteins, and mRNA localization in motor neuron cell bodies; and a comparison to IHC of vGluT2. These aspects were deemed essential to your next revision.

The next two sets of comments could be dealt with in the text of the Discussion section:

2) incomplete blockade of evoked muscle action by glutamate antagonists, that could be explained by the low concentration of glutamate antagonists used;

3) physiology experiments showing that the reduction of the calcium transients generated during fictive swimming by tubocurarine may be due to blockade of GABA-A or glycinergic receptors rather than cholinergic receptors.

Reviewer #1:

The authors have addressed my concerns satisfactorily. Demonstration of the presence of vGluT1 and the effective use of glutamate receptor antagonists were key. The data in Figure 6B,C are nicely characterized as a "functional switch in hindlimb neuromuscular transmission".

Reviewer #2:

The authors revised this manuscript according to the criticisms raised by the reviewers. The responses provided reasonable explanations and interpretations.

Reviewer #3:

In response to the reviewers' concerns the authors have attempted to demonstrate that motoneurons express VGluT1. Unfortunately, I did not find their presentation of the data convincing. For example, VGluT1 expression at stage 52 (Figure 5—figure supplement 1) is widespread throughout the ventral cord and appears, surprisingly, to label cell bodies. This is in contrast to numerous other immunocytochemical studies of VGluT1 labelling in the spinal cord that show the labelling is restricted to synaptic terminals. (Betley et al., 2009; Schultz et al., 2017). The authors should illustrate an image of the whole transverse section of the cord to show how widespread the labelling is. Was this antibody tested or has it been described in other papers to show that it is indeed targeting VGluT1? What experiments did the authors perform to examine the specificity of the antibody? The VGluT1 labelling shown in Figure 5B appears coextensive with the muscle nerve but there is no evidence of the more punctate labelling that characterizes the neuromuscular junction and is shown in Figure 5A (α-bungarotoxin). Furthermore, the authors should show the VGluT1 labelling at stage 57 when glutamatergic transmission is absent. Finally, it was not clear if the authors had tried to label the cells with VGluT2 whose mRNA has been reported in some studies to be expressed in the neonatal mammalian motoneurons.

A more convincing approach to demonstrate the presence of a vesicular protein in motoneurons would be to show that it co-localizes with a synaptic protein – such as synaptophysin at the neuromuscular junction or to demonstrate the presence of the mRNA in motoneuronal cell bodies.

https://doi.org/10.7554/eLife.30693.016

Author response

Essential revisions:

1) Demonstration of the presence of any of the known vesicular glutamate transporters in motoneurons at the expected ages.

New experiments on the immuno-detection of VGluT1 have demonstrated the presence of the glutamate vesicular transporter in early stage tadpoles, both at the level of motor axons in the developing hindlimbs (illustrated in new Figure 5C) and the limb motoneuron cell bodies themselves (illustrated in new Figure 5—Figure supplement 1).

2) Better explanation in the text of the large effect of d-tubocurarine. There is a discrepancy in the reported effects of applying t-tubocurarine at Stage 53, leading to a large decrease in amplitude of EMG activity, and the localization of receptors. Reviewer 2 asks whether you could test the effects of tubocurarine or glutamate antagonists at an earlier stage (e.g., 52).

Evidently there was a general misinterpretation of the results illustrated in Figure 4 and original Figure 5, and we thank the three referees for having brought this lack of clarity to our attention. In old Figure 5 (new Figure 6), we used isolated hindlimb preparations to stimulate directly the limb motor nerve, and therefore the antagonists’ effects on the EMG responses could be concluded to be exclusively peripheral, at the neuromuscular junction. In Figure 4, in contrast, the experiments were performed on semi-intact preparations where spontaneously-active spinal CPG circuitry was still present and able to drive motoneurons and EMG activity in both the axial and limb muscles. In this case, any observed effects of d-tubocurarine could be attributable to an action at either central or peripheral levels, or both. In comparing the results of the experiments illustrated in Figure 4 and old Figure 5, it seemed reasonable to us to interpret the EMG amplitude changes seen under d-tubocurarine at stage 53 as being centrally-mediated in Figure 4 and peripherally-mediated in old Figure 5. Manifestly, however, this was not obvious to the reader. We now also provide new data (in Figure 4—figure supplement 1) showing that cholinergic synapses are present on limb motoneuron cell bodies, that they play an important role in the excitation of these motoneurons during centrally-generated swimming and are directly affected by nicotinic antagonist perfusion. This is demonstrated by new calcium imaging experiments that show a strong reduction in the calcium signal in limb motoneurons during fictive swimming in the presence of d-tubocurarine. These supplementary results, which are briefly described in subsection “Locomotor-related activation of hindlimb muscles by non-cholinergic motoneurons”, thus strongly support the conclusion that d-tubocurarine was indeed having a major central effect on limb motoneuron activation in the experiments of Figure 4B (as corroborated in old Figure 5/new Figure 6 by the responses that could only occur at the level of the neuromuscular junction).

3) Inclusion in the Results section of the electrophysiological recordings showing the effects of glutamatergic antagonists. This experiment was alluded to in the Materials and methods section in which a cocktail of CNQX and AP5 was used to block glutamatergic receptors and are critical to the arguments of the paper.

These data are now illustrated in new Figure 6. We initially believed that illustrating the cholinergic drug effects was important for emphasizing more the non-cholinergic nature of the transmitter phenotype present at early developmental stages. However, we agree that illustrating the parallel effects of glutamate antagonists does strengthen the phenotype switching message of the article.

4) Better and more complete statistical analyses: Reviewer 3 notes a lack of statistical analysis and an indication of the number of experiments used to generate the quantitative data, a lack of statistical comparison between the fluorescence levels at the different ages, and confusion over standard error of the mean and standard deviation used to illustrate the variability.

The lack of numbers of experiments in the original manuscript was due to an unfortunate editing error for which we apologize. This information is now reestablished, where appropriate, in the text and/or figure legends. The lack of statistical comparison pointed out by reviewer 3 relates specifically to the changes in ChAT and VAChT fluorescence levels found in limb motoneurons during development (Figure 2C). Although we originally performed such a comparison, we didn't report it because of uncertainty about its validity due to prepto-prep differences, for example, in the extent of retrograde label uptake, fluorescence intensities, and tissue penetration by both primary and secondary antibodies. However, as requested by the reviewer, this comparative analysis is now provided in Figure 2.

We have followed conventional procedure in providing means ± SEM in the text. In original Figure 2C, individual values (circled dots) were presented for single experiments in each age group, allowing their dispersion to be visualized, with the error bars representing the standard error of the overall mean (black dots) of these values per group. Note that this analysis has been redone according to reviewer 3’s remarks and that we have changed the data display to improve clarity.

Reviewer #1:

1) Subsection “Locomotor-related activation of hindlimb muscles by non-cholinergic motoneurons”, "is initially largely independent of ACh signaling in…" The amplitude is reduced by more than 50% by d-tubocurarine. How about "is initially only partially suppressed in…."? Could the residual signals be stimulus artifacts? In contrast to Control and Wash, the amplitudes are constant in amplitude. The Materials and methods section describes experiments in which a cocktail of CNQX and AP5 was used to block glutamatergic receptors. These experiments should be illustrated here. They are critical to the story.

This part of the Results section has been rewritten to incorporate new immunohistochemical and electrophysiological data (Figure 4—figure supplement 1) which indicate that the reduction in limb muscle EMG under d-tubocurarine in pre-stage 55 preparations was most likely due to a central nervous effect on motoneuron activation rather than a peripheral effect on the neuromuscular junction. Note also that the activity recorded in Figure 4 is spontaneous fictive swimming without involving any electrical stimulation (and thus the possibility of stimulus artifacts).

Hindlimb EMG responses to motor nerve electrical stimulation in the presence of glutamatergic antagonists are now illustrated in new Figure 6 for both early and late developmental stages, as requested by the three reviewers and the corresponding text have been modified (subsection “Developmental switch in hindlimb neuromuscular transmission").

Related to this point, subsection “Developmental switch in hindlimb neuromuscular transmission”, "indicating the absence of anatomically detectable muscle ACh receptors" would be more appropriate. Figure 4B suggests that they are detectable by physiological / pharmacological methods.

Again, related to this point, in the Discussion section, I find "indicative of a complete neurotransmitter phenotype switching…" too great a claim, given that d-tubocurarine substantially reduces the amplitude of the l-HLemg in Figure 4B.

In order to be more circumspect, we have rewritten as “indicating an absence of muscle ACh receptors…” subsection “Developmental switch in hindlimb neuromuscular transmission”.

For the same reason, “indicative” has been replaced by “consistent with” in the Discussion section.

We agree that Figure 4B could suggest ACh to be involved in the neuromuscular transmission at stage 53. However, as stated in our response to the reviewer’s preceding comment, another explanation is that the observed effect resulted from the d-tubocurarine acting centrally on limb motoneurons during fictive axial swimming (see new Figure 4—figure supplement 1 and the new paragraph in subsection “Locomotor-related activation of hindlimb muscles by non-cholinergic motoneurons). Specifically, these new data show the presence of cholinergic synapses in the vicinity of limb MN cell bodies and a suppressive effect of the nicotinic antagonist on the calcium dynamics of limb MNs during fictive swimming. Taken together with our other results, therefore, these new data add strong support to the conclusion of a complete switch in the neurotransmitter phenotype of limb motoneurons during early metamorphosis.

2) Discussion section, the authors argue against a muscle spindle innervation role for the limb motoneurons they have studied, on the basis of the sizes of the muscle spindle- and muscle fiber-innervating populations of neurons. However, in the absence of information about the sizes of these populations this is not convincing.

A specific fusimotor innervation has not previously been described in frogs, other than by Fujitsuka et al., (1987) who reported the existence of a small population of fusimotor MNs innervating exclusively muscle spindles, and in only one muscle (the semitendinosus). Although not quantified, our data on Islet immunoreactivity (and now from in situ hybridization; Figure 3—figure supplement 1) have revealed the existence of a relatively large population of putative appendicular MNs, at least the greater proportion of which was identified (by back-filling) as projecting into the limb bud. Thus, it seems reasonable to conclude that this extensive population of non-cholinergic limb MNs does not correspond to a possible discrete fusimotor population. The text has been rewritten (Discussion section) to try and clarify this point.

3) Quantification is inadequate. The numbers of animals in which observations were made should be provided in each figure legend. Contrary to the assertion in the Transparent Reporting Form, the number of replicates is largely omitted in the Results section and Figure Legends.

We apologize for this omission, which has been corrected. All numbers of animals and cells are now provided where appropriate in the text and/or figure legends.

Reviewer #2:

1) Figure 4: It is argued that the switch from glutamatergic to cholinergic transmission occurs around stages 54-55. However, the results presented in Figure 4 clearly show that cholinergic transmission is at place already at stage 53 and possibly even at earlier stages. There is a large decrease in the amplitude of the EMG activity by d-tubocurarine.

Could these experiments be done at stage 52? Alternatively, is it experimentally possible to apply glutamatergic antagonists locally on the limb bud?

The results of Figure 4 show clearly that the EMG activity at stage 53 is largely mediated by t-tubocurarine-sensitive receptors. This is at odds with the main conclusion of this study.

This important point was also raised by reviewer 1. Our response is the same and is as follows: We agree that Figure 4B could suggest that ACh is involved in transmission at the NMJ of the de novo limb muscles at stage 53. However, another (and we think more likely) explanation is that the decreased EMG response to d-tubocurarine is due to the drug acting centrally in diminishing the premotor activation of limb motoneurons during fictive axial swimming in these semi-isolated (CNSaxial/appendicular muscle) preparations. In support of this idea, we provide new immuno-anatomical labeling evidence for the presence of VAChT co-localized with synapsin around limb motoneuron cell bodies, indicating the presence of central cholinergic synapses on these neurons. That these synapses are involved in driving the limb MNs was confirmed with calcium imaging experiments on isolated CNS preparations demonstrating that their activation is substantially decreased during fictive swimming in the presence of d-tubocurarine. These additional results, which are illustrated in Figure 4—figure supplement 1 and briefly described in subsection “Locomotor-related activation of hindlimb muscles by non-cholinergic motoneurons” are therefore consistent with our study’s main conclusion that a switch in neurotransmitter phenotype of limb motoneurons occurs during early metamorphosis.

Finally, applying glutamatergic antagonists to the limb buds is certainly a potentially instructive approach and one that we had already attempted. Unfortunately, due the extremely small size of the limb buds in early stage animals (50-53), injecting drugs without destroying bud neuromuscular tissue and while maintaining a muscle recording electrode in place was found to be technically impossible.

2) Figure 5C,D: The EMG activity recorded at stage 53 is very weak and this may indicate that the neuromuscular junction is not fully developed.

The authors should provide electrophysiological traces showing the effects of glutamatergic antagonists. Also, the weak EMG activity recorded at stage 53 might be due to the absence of activation of NMDA receptors. I suggest to repeat these experiments using a Mg-free extracellular solution.

The effects of glutamate antagonists on EMG responses to nerve stimulation are now shown in new Figure 6. A series of experiments was also done in saline containing half the normal MgCl2 concentration. As implied by the reviewer, the EMG response at stage 53 was increased (an illustration of the response of a stage 53 preparation in new Figure 6 (A, middle trace) is now taken from these experiments), but the effect of the glutamatergic antagonists (about a half-reduction) remained similar to that under normal saline.

3) Overall, it is not clear if limb motor neurons switch the transmitter phenotype or if they use both glutamate and acetylcholine as co-transmitters with a dominance of the one or the other at different developmental stages. In this regard, the authors could test if limb motor neurons co-express transporters for glutamate and acetylcholine in different proportions at different stages.

The expression of the vesicular transporter for glutamate VGluT1 in motor axons at early stages is now illustrated for a stage 53 tadpole in new Figure 5C. In addition, we provide as Figure 5—figure supplement 1 an illustration of VGluT1 immuno-detection in limb motoneuron cell bodies at stage 52, when ChAT and VAChT are not expressed. However, the expression of this transporter does not seem to persist at later stages, thus paralleling the lack of expression of NR1 and NR2b receptor subunits in the limb muscles.

Reviewer #3:

The central thesis of this paper is that motoneurons transiently release glutamate instead of acetylcholine at the neuromuscular junction of developing hindlimb muscles at the onset of metamorphosis. The main evidence in support of this conclusion comes from the immunocytochemistry showing that appendicular motoneurons do not express either ChAT or VAChT immunoreactivity and exhibit very little ChAT mRNA at stage 53. Moreover, at stage 52, the developing limb muscle has very low levels of α-bungarotoxin but does express NMDA receptor subunits some of which are co-localized with neuronal synaptophysin (although many are not).

Unfortunately, the physiological experiments do not completely support the anatomy. For example, at stage 53 bath application of d-tubocurarine greatly reduces the amplitude of the rhythmic EMG signals from the limb bud during locomotor episodes despite the apparent absence of acetylcholine receptors in the muscle and cholinergic markers in limb motoneurons. The authors do not explain why the EMG is reduced at this age. Similarly, in the nerve/limb preparation illustrated in Figure 5, bath application of CNQX/AP5 only reduces the stimulus-evoked EMG by 50%. Why is the block incomplete? Again, this is not discussed by the authors. The authors showed only the presence of NMDA receptor subunits in the muscle but did not show the effect of NMDA antagonists alone on the evoked EMG to validate their presence.

As explained in our responses to similar comments made by reviewers 1 and 2, the experiments in Figure 4 were performed on semi-isolated CNS/muscle preparations where EMG activity in both the axial and limb muscles was driven spontaneously by the spinal locomotor network. We now provide new anatomical and calcium imaging evidence (Figure 4—figure supplement 1; subsection “Locomotor-related activation of hindlimb muscles by non-cholinergic mot”) that cholinergic synapses are present on limb motoneuron cell bodies and are substantially involved in the excitation of these cells during centrally-generated swim episodes. Thus, it is likely that the diminution of the limb EMG strength evident in Figure 4B resulted from a direct central effect of dtubocurarine on limb motoneuron excitation rather than at the peripheral neuromuscular junction.

On the other hand, the use of glutamatergic antagonists on the semi-intact preparation totally abolishes fictive swimming and so an equivalent EMG analysis could not be done with these antagonists. However, both cholinergic and glutamatergic sets of antagonists could be used on the isolated motor nerve-hindlimb preparation with nerve stimulation (old Figure 5, new Figure 6). The lack of a complete block by glutamatergic antagonists at stage 53 (Figure 6B, C) has several possible explanations. The first and simplest is that glutamate antagonists are not fully effective in the amphibian (especially in such an immature neuromuscular system) and cause a partial blockade only, as already reported (e.g., work from Robitaille’s group; Fu et al.,; Jamieson…). Another explanation is that glutamate is not the only neurotransmitter used at the limb neuromuscular junctions in these early stages. In relation to this possibility, we did an additional series of experiments using other neurotransmission blockers, but no clear results emerged (for example, picrotoxin was used to block chloride channels associated with GABA and Glycine receptors, on the basis that GABA has been found in immature Xenopus axial motoneurons in culture; e.g., work from N. Spitzer’s group). In addition, combining antagonists (notably, d-tubocurarine + glutamate antagonists) never led to a full blockade of stimuli-evoked EMG responses. We chose not to report these additional experiments in the revised manuscript to avoid over-complicating the main message and especially since it seems more generally that glutamate antagonists are unable to completely block glutamate-mediated transmission in amphibians.

In both experiments, it would have been useful to combine the glutamatergic and cholinergic antagonists to see if this produced a complete blockade of the muscle electrical activity.

Finally, in Figure 5D, the results for control and d-tubocurarine are shown, but it would also be of relevance to show a recording for the glutamate antagonists. Here, the number of experiments is missing, and no p-values are reported.

As explained in our previous response, additional experiments in which we applied the glutamatergic and cholinergic antagonists in combination still failed to produce a complete blockade of stimulus evoked EMG responses. In any case, because usually neither drug could be fully washed out in our previous experiments, the first drug was still present if a second was applied sequentially. We also did control experiments where only one drug was applied, and the effects were similar to those obtained with two antagonists applied in combination. This latter observation is not illustrated but is stated in the Materials and methods section. The effects of glutamatergic antagonists are now illustrated for both early and late stages in new Figure 6. The numbers of experiments and p values are also now given in the corresponding text (subsection “Developmental switch in hindlimb neuromuscular transmission”).

The intracellular recordings of the locomotor activity illustrated in Figure 3E are puzzling. The recordings reveal rhythmic hyperpolarizing potentials and the cells do not fire. Was this true of all of the motoneurons that were recorded? If so, how are the muscles activated if the motoneurons do not spike? For the 11 recorded motoneurons, the authors should indicate from how many preparations they were obtained, and if the motoneurons were only identified by their labeling or whether they tested for an antidromic response. If it is the former, did the authors ever use a second dye applied intracellularly to confirm that the labelled motoneuron was indeed the one that was recorded from.

At the earliest stages of limb formation, it was stated that motoneurons were small cells. However, the resistance of the cell shown in Figure 3C was between 3 to 5 MOhm suggesting either a very leaky cell or a large one. What was the input impedance of the other motoneurons?

The absence of firing in our recorded motoneurons was due to the internal solution that we generally used in our patch-clamp electrodes. We have performed a new set of experiments in which the patch pipette solution was changed for a high chloride solution, as usually used for recording immature motoneurons (see Akerman and Cline, now cited in the Materials and methods section). In this latter condition, neurons were able to fire action potentials spontaneously or in response either to current injection or spontaneous CPG activity. These experiments are illustrated in Figure 3C,D and F, and both the Results section and Materials and methods section have been updated accordingly.

Recorded motoneurons were indeed identified according to their labeling from hindlimb muscles, although fluorescent dye was also sometimes added in the patch pipette for later verification. This is now stated in the text and illustrated in Figure 3B. The number of preparations in each case is indicated in the Materials and methods section.

Early stage appendicular motoneurons have somata of 15-20 µm and were classified as 'small' neurons compared to the large axial motoneuron cell bodies (that are about three-fold larger than those of limb motoneurons). We are fairly confident that the intracellular recordings reported in this study were not ‘leaky’ (in fact, leaky cells were excluded from analysis), but since we have no comparison with other studies in this developmental range, we cannot draw any conclusions about the significance of the resistance values observed.

As already alluded to, a major concern is the lack of statistical analysis and an indication of the number of experiments used to generate the quantitative data. For example, in subsection “Delayed cholinergic transmitter phenotype expression in the developing appendicular motor Column” there is no statistical comparison between the fluorescence levels at the different ages nor is the number of experiments shown. Similar problems apply to the quantification in other parts of the paper. The authors often refer to the mean and the SE. Do they mean the standard error of the mean (S.E.M). Why was the standard error of the mean rather than the standard deviation used to illustrate the variability?

A statistical comparison was not reported for the fluorescence levels because we doubted the validity of this type of analysis. However, we have changed the way the fluorescence data are displayed, and statistics are also given. Note that this analysis has been redone according to the following comment.

The quantification of the fluorescence in Figure 2C was confusing. In the text, it states that ChAT and VAChT were not immuno-detected in appendicular MNs until stage 55. The figures in Figure 2A confirm this. However, in the plot of 2C (52-54), which shows the AP/AX fluorescence ratio, the values range from ~0-14% for ChAT. It is difficult to see how these values are generated if AP is zero. The authors should indicate how the ratio was calculated.

Cholinergic terminals are widely present in the spinal cord, including around limb motoneuron somata (now illustrated in Figure 4—figure supplement 1), which can sometimes affect the fluorescence ratio. We have redone the analysis in choosing as background a region without motoneurons but enriched in cholinergic terminals. This is now stated in the Materials and methods section and we are grateful to the reviewer for pointing out this analytical bias. The results of this analysis are presented differently from the original Figure 3C: we now show ∆F/F for axial and appendicular motoneurons separately, which also allowed reporting the statistics.

Also, despite an increase in the ratio at later stages, it is still about 20% of the axial fluorescence (when appendicular MNs are supposed to be bigger). This implies that the majority of appendicular motoneurons are still not cholinergic. It would be helpful to show that once the metamorphosis is complete, this ratio is 100% thereby validating the measure.

Our evidence indicates that all limb motoneurons become cholinergic from stage 55 onwards. However, axial motoneurons always displayed much more fluorescence than limb motoneurons, and this is even more evident in the new data representation in Figure 2C. Even in juvenile adults, i.e. after metamorphosis completion, the ratio remains largely in favor of axial motoneurons. Presumably this indicates a higher concentration of cholinergic proteins in axial motoneurons, but we do not know why this is so.

I found the in-situ hybridization results shown in Figure 2D, difficult to interpret. It is difficult to know the extent of the ChAT ISH labelling in these Figures For example, the extent of the Axial motor column in D 53 looks much more limited than the ChAT labelling in A 53. Furthermore, it is not clear that appendicular MNs at stage 53 are negative. The background is very high. Sections with dextran labelling would have made clear where the motoneurons are located. This figure does not agree with the statement that ISH is more sensitive than immunodetection. How many times was this experiment performed?

Islet1 in situ hybridization has been performed on the same preparations as ChAT hybridization. We now also illustrate the results for Islet1 (in Figure 3—figure supplement 1) in order to compare the location and extent of the appendicular motor column at key developmental stages. Whereas limb motoneurons were always Islet1 mRNA-positive, ChAT mRNAs were never detected at early stages (stage 51 now illustrated in Figure 2D). New images are displayed in which the axial motor column is clearly visible in all hybridizations, and the number of experiments has been added in the text (subsection “Delayed cholinergic transmitter phenotype expression in the developing appendicular motor column”).

mRNAs were detected with in situ hybridization before their translation into proteins. Thus, at pivotal stages 53-54, a light ChAT mRNA labeling could be observed in some preparations whereas ChAT proteins were never immuno-detected. This again is consistent with a delayed acquisition of the cholinergic phenotype by limb motoneurons.

[Editors' note: further revisions were requested prior to acceptance, as described below.]

The manuscript has been improved and reviewers 1 and 2 were satisfied with your amendments, but there are some remaining issues raised by reviewer 3 that remain to be addressed, indicated in the appended reviews. The other reviewers and I in consultation now agree on the need for these revisions:

1) The verity of the VGluT1 localization, in synaptic terminals rather than in cell bodies; the predicted restriction of vGluT1 to motor neuron terminals and colocalization with synaptic proteins, and mRNA localization in motor neuron cell bodies; and a comparison to IHC of vGluT2. These aspects were deemed essential to your next revision.

A major remaining concern of reviewer 3 was our finding of VGluT1 localization in the immature limb motoneuron cell bodies, where he/she argued (along with two reference citations) that VGluT1 should be found exclusively in axon terminals, in the vicinity of synaptic contacts. However, we respectfully point out that there is evidence to indicate otherwise. Most previous studies on the subject (including the two reports cited by the reviewer) investigated glutamatergic projections onto non-glutamatergic neurons and accordingly, a strong concentration of VGluT1 immunoreactivity was detected in the vicinity of the latters' cell bodies, but which themselves unsurprisingly were nonreactive. In contrast, other studies on glutamatergic signaling in various brain and spinal structures (e.g., Malet and Brumovsky, 2015; Melo et al., 2013; Nakamura et al., 2005; Yang et al., 2014) have reported the expression of VGluT1 both within the somata of glutamatergic neurons, as well as along their axons. Although these studies were on rodents, and we have not found previous reports of similar findings in amphibians, the results do provide a clear precedence for our observation of cell body and axonal VGluT1 expression in immature Xenopus limb motoneurons.

Nevertheless, we performed a new set of immunolabeling experiments using a different VGluT1directed antibody (from mouse, instead of the guinea pig antibody originally used.). Here again, we obtained a very similar labeling in early limb motoneuron cell bodies (now illustrated in new Figure 5—figure supplement 1A) as with the guinea pig antibody. Additionally, VGluT1 expression was found along MN axons, as well as co-localized with synaptophysin and NR1 immuno-labeling (new Figure 5—figure supplement 1 B1) as would be expected with a functional glutamatergic NMJ. As also requested by the reviewer, we now illustrate that VGluT1 labeling is lost in stages older than 54 (Figure 5—figure supplement 1B2), when glutamatergic signaling disappears.

Unfortunately, after many attempts with various protocols (including that published by Morona et al., 2017 and even using their primer, which they kindly sent to us) we failed to produce VGluT hybridization, which if successful would have obviously reinforced our immunohistological observations. However, we believe that (1), finding very similar results with VGluT1directed antibodies originating from two different mammalian sources, (2), the general endorsement from the literature, and (3), the combination of the experimental data provided by our present study support the conclusion that before becoming cholinergic, de novo limb motoneurons at premetamorphosis initially use glutamate as a major transmitter at the NMJ.

The impetus for our VGluT1 immunolabeling experiments was based on a previous report (Gleason et al., Gene Expression Patterns, 2003) demonstrating the transporter's presence in the early developing Xenopus laevis spinal cord. Although our experiments using mammalian VGluT1 antibodies gave positive results, this of course doesn't mean that the Xenopus and guinea pig or mouse VGluT1 isoforms are identical. We also tested for VGluT2 immunolabeling (again using a mammalian-targeted antibody) in spinal cord cross sections and this failed completely. In this context, however, it is relevant that according to a phylogenetic analysis by Villar-Cervino et al., (2010), vesicular glutamate transporters in Xenopus are thought to exist as two related subtypes (a and b) of VGluT1, rather than as homologues of the distinct VGluT1 and 2 isoforms. In any case, we should emphasize that the aim of our VGluT immunolabeling experiments was to seek supporting evidence for (or against) glutamatergic synaptic signaling playing a role in the early developing limb system, rather than attempting to identify the specific transporter isoforms that might be involved.

The next two sets of comments could be dealt with in the text of the Discussion section:

2) incomplete blockade of evoked muscle action by glutamate antagonists, that could be explained by the low concentration of glutamate antagonists used;

A new paragraph has been added to the Discussion section where we now address the incomplete blockade of evoked EMG responses by glutamatergic antagonists. We have also rewritten much of the Discussion section’s first two paragraphs in order to avoid implying that glutamate is alone responsible for neurotransmission at the early limb NMJ.

3) physiology experiments showing that the reduction of the calcium transients generated during fictive swimming by tubocurarine may be due to blockade of GABA-A or glycinergic receptors rather than cholinergic receptors.

This interesting and plausible possibility is now mentioned in the Results section. Indeed, the possible contribution of a d-tubocurarine-sensitive, GABA-A receptor-mediated drive to limb motoneurons reinforces our hypothesis that in spontaneously active semi-intact preparations, d-tubocurarine causes a central blockade of motoneuron activation, rather than acting peripherally at the limb NMJ.

Reviewer #3:

In response to the reviewers' concerns the authors have attempted to demonstrate that motoneurons express VGluT1. Unfortunately, I did not find their presentation of the data convincing. For example, VGluT1 expression at stage 52 (Figure 5 supplement 1) is widespread throughout the ventral cord and appears, surprisingly, to label cell bodies. This is in contrast to numerous other immunocytochemical studies of VGluT1 labelling in the spinal cord that show the labelling is restricted to synaptic terminals. (Betley et al., 2009; Schultz et al., 2017). The authors should illustrate an image of the whole transverse section of the cord to show how widespread the labelling is. Was this antibody tested or has it been described in other papers to show that it is indeed targeting VGluT1? What experiments did the authors perform to examine the specificity of the antibody? The VGluT1 labelling shown in Figure 5B appears coextensive with the muscle nerve but there is no evidence of the more punctate labelling that characterizes the neuromuscular junction and is shown in Figure 5A (α-bungarotoxin). Furthermore, the authors should show the VGluT1 labelling at stage 57 when glutamatergic transmission is absent. Finally, it was not clear if the authors had tried to label the cells with VGluT2 whose mRNA has been reported in some studies to be expressed in the neonatal mammalian motoneurons.

A more convincing approach to demonstrate the presence of a vesicular protein in motoneurons would be to show that it co-localizes with a synaptic protein – such as synaptophysin at the neuromuscular junction or to demonstrate the presence of the mRNA in motoneuronal cell bodies.

As detailed in our response to editorial comment 1 above, our observation of a diffuse VGluT1 immuno-reactivity in the somata of early Xenopus limb motoneurons finds strong echoes in the literature. Similarly, the expression of VGluT1 along glutamatergic neuron axons was also reported in these studies, which also corresponds to our findings in immature limb motoneuron axons.

The new series of VGluT1 immuno-detection experiments using a different antibody (generated from mice) provided a test for antibody specificity. The finding of VGluT1 expression in MN somata was the same as with the original guinea pig antibody, and these data are now illustrated in new Figure 5—figure supplement 1A. In addition, combined VGluT1, synaptophysin and NR1 immuno-detection in early hindlimbs showed co-localized labeling, consistent with functional neuromuscular synapses (Figure 5—figure supplement 1B1). Finally, we provide new images illustrating the parallel loss of VGluT1 and NR1 immuno-detection in hindlimbs in stages older than 54 (Figure 5—figure supplement 1B2).

According to a phylogenetic analysis by Villar-Cervino et al., (2010) a homologue of the mammalian or fish VGluT2 gene doesn't not exist in Xenopus, and what is usually termed VGluT2 is in fact a subtype of the VGluT1 isoform due to the tetraploid nature of the animal. Contrary to what is observed in other species, the amino acid sequences differ very little between the two resulting proteins, thus corresponding to VGluT1a and VGluT1b. Although no distinction could be made in our experiments between these two VGluT1 subtypes, this would explain our failure to see any VGluT2 labeling using a mammalian antibody.

Frustratingly, our repeated attempts (and with different protocols) failed to achieve VGluT hybridization. However, as stated in our response to editorial comment 1, we believe that the combination of our existing immunolabeling and electrophysiological data provide convincing evidence that pre-metamorphic limb motoneurons initially use glutamate as a neurotransmitter before becoming cholinergic.

https://doi.org/10.7554/eLife.30693.017

Article and author information

Author details

  1. Francois M Lambert

    Institut de Neurosciences Cognitives et Intégratives d’Aquitaine, Université de Bordeaux, Bordeaux, France
    Contribution
    Conceptualization, Resources, Data curation, Software, Formal analysis, Supervision, Validation, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing
    For correspondence
    francois.lambert@u-bordeaux.fr
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-8655-2652
  2. Laura Cardoit

    Institut de Neurosciences Cognitives et Intégratives d’Aquitaine, Université de Bordeaux, Bordeaux, France
    Contribution
    Data curation, Formal analysis, Methodology
    Competing interests
    No competing interests declared
  3. Elric Courty

    Institut de Neurosciences Cognitives et Intégratives d’Aquitaine, Université de Bordeaux, Bordeaux, France
    Contribution
    Data curation, Formal analysis, Methodology
    Competing interests
    No competing interests declared
  4. Marion Bougerol

    Evolution des Régulations Endocriniennes, Muséum National d’Histoire Naturelle, Paris, France
    Contribution
    Data curation
    Competing interests
    No competing interests declared
  5. Muriel Thoby-Brisson

    Institut de Neurosciences Cognitives et Intégratives d’Aquitaine, Université de Bordeaux, Bordeaux, France
    Contribution
    Conceptualization, Data curation, Formal analysis, Methodology, Writing—original draft, Writing—review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-3214-1724
  6. John Simmers

    Institut de Neurosciences Cognitives et Intégratives d’Aquitaine, Université de Bordeaux, Bordeaux, France
    Contribution
    Writing—original draft, Writing—review and editing
    Competing interests
    No competing interests declared
  7. Hervé Tostivint

    Evolution des Régulations Endocriniennes, Muséum National d’Histoire Naturelle, Paris, France
    Contribution
    Conceptualization, Data curation, Formal analysis, Supervision, Methodology, Writing—original draft, Writing—review and editing
    Competing interests
    No competing interests declared
  8. Didier Le Ray

    Institut de Neurosciences Cognitives et Intégratives d’Aquitaine, Université de Bordeaux, Bordeaux, France
    Contribution
    Conceptualization, Resources, Data curation, Software, Formal analysis, Supervision, Validation, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing
    For correspondence
    didier.le-ray@u-bordeaux.fr
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-2089-9861

Funding

Fondation pour la Recherche Médicale (Equipe FRM DEQ20170336764)

  • Muriel Thoby-Brisson

Muséum National d'Histoire Naturelle (Actions thématiques du Museum)

  • Hervé Tostivint

Centre National de la Recherche Scientifique

  • Didier Le Ray

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We are grateful to Gilles Courtand (UMR 5287) and Cyril Willing (Centre de Microscopie de fluorescence et d’Imagerie numérique du Muséum) for technical assistance in image analysis, Feng Quan, Anne-Laure Gaillard (both UMR 7221) and Boudjema Imarazene (UMR 7208) for technical assistance with in situ hybridization, and Lionel Parra-Iglesias (UMR 5287) for animal care. This work was supported by grants from the Centre National de la Recherche Scientifique, by the Actions Thématiques du Muséum and the Fondation pour la Recherche Médicale (Equipe FRM DEQ20170336764 M. Thoby-Brisson).

Ethics

Animal experimentation: All procedures were carried out in accordance with, and approved by, the local ethics committee (protocols no. 68-019) to H. Tostivint and no. 2016011518042273 APAFIS no. 3612 to DLR)

Reviewing Editor

  1. Carol A Mason, Columbia University, United States

Publication history

  1. Received: July 25, 2017
  2. Accepted: May 6, 2018
  3. Accepted Manuscript published: May 30, 2018 (version 1)
  4. Version of Record published: June 12, 2018 (version 2)

Copyright

© 2018, Lambert et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 614
    Page views
  • 121
    Downloads
  • 1
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, Scopus, PubMed Central.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)