Synapses Form in Skeletal Muscles Lacking Neuregulin Receptors

See allHide authors and affiliations

Science  24 Jun 2005:
Vol. 308, Issue 5730, pp. 1920-1923
DOI: 10.1126/science.1108258


The formation of the neuromuscular junction (NMJ) is directed by reciprocal interactions between motor neurons and muscle fibers. Neuregulin (NRG) and Agrin from motor nerve terminals are both implicated. Here, we demonstrate that NMJs can form in the absence of the NRG receptors ErbB2 and ErbB4 in mouse muscle. Postsynaptic differentiation is, however, induced by Agrin. We therefore conclude that NRG signaling to muscle is not required for NMJ formation. The effects of NRG signaling to muscle may be mediated indirectly through Schwann cells.

Motor neurons, terminal Schwann cells, and muscle fibers express two neuregulin genes, nrg-1 and nrg-2 (13). NRGs activate receptor tyrosine kinases of the ErbB family (4) that are expressed in the same cells. Motor neuron–derived NRG1β, originally isolated from brain as acetylcholine receptor (AChR)–inducing activity (ARIA) in cultured myotubes (1), is widely accepted to induce AChR gene transcription in subsynaptic myonuclei by activating ErbB receptors; the receptors stimulate the same regulatory elements in AChR gene promoters as the motor nerve (57). Agrin, another nerve-derived signal for NMJ formation (8), activates its receptor MuSK in muscle (9) and thus recruits AChRs, muscle-derived NRGs, ErbBs, and other synaptic components to the subsynaptic muscle membrane. Whereas the importance of Agrin in NMJ formation has been demonstrated by genetic means (10), the analysis of NRG/ErbB signaling has been less obvious. Mice lacking nrg1, erbB2, or erbB4 die during development because of heart malformation, which prevents analysis of their function in synapse development. Transgene rescue experiments or muscle-specific gene ablation have been used to circumvent embryonic lethality. However, because Schwann cells depend on NRG signals from motor neurons for their survival and differentiation (11) and signals from Schwann cells regulate motor neuron function (12), the effect of NRG on subsynaptic muscle differentiation could not be resolved in the genetically rescued mice. Moreover, whereas ErbB2 is required for muscle spindle development (13), neither ErbB2 nor ErbB4 receptors alone are essential for NMJ formation (14). However, they may have redundant functions on synapse-specific gene expression (3).

To abolish all NRG signaling to muscle, we inactivated both the ErbB2 and ErbB4 receptors selectively in muscle without affecting their expression in Schwann cells. To this end, we crossed mice double-homozygous for loxP-flanked alleles of the erbB2 (13) and erbB4 (15) genes with a mouse line that begins to express Cre recombinase under the control of the human skeletal actin promoter in muscle precursors at embryonic day 9 (E9) (16), i.e., before neuromuscular synapse formation at E13. Mutant mice (HSA-Cre+/–;erbB2fl/–;erbB4fl/fl) were born in the expected Mendelian ratio. They were viable, but their body weight was reduced by about 30% compared with littermates (fig. S1), which suggested a role of ErbBs in muscle growth.

Several lines of evidence indicate that NRG signaling to mutant muscle was effectively inactivated. First, cultured myotubes did not respond to NRG1β (Fig. 1A), and in fetal myotubes at E13, loxP-flanked erbB alleles were recombined (fig. S2). Second, only 41% of all nuclei in mouse gastrocnemius muscle are myonuclei (fig. S3B); because Cre expression is selective for myonuclei (fig. S3A), the maximal fraction of recombined erbB alleles will be 41%. This is in good agreement with the 40% recombination of the mutant erbB2 and erbB4 alleles as determined from Southern blots of DNA from mutant muscle (fig. S3B), which indicates complete recombination in all myonuclei. However, even if recombination occurred in only 80% of myonuclei, combinatorial considerations predict that the probability for a nonrecombined nucleus to be located at the synapse, which would allow ErbB-mediated AChR gene expression, would be too low to explain normal maintenance of synapses in all fibers (see supporting online text). Third, consistent with the complete recombination of loxP-flanked erbB alleles in myonuclei observed above, >98% of myonuclei were Cre positive (fig. S3C). Fourth, erbB2 mRNA was reduced to <10% normal amounts, and erbB4 mRNA to below the detection limit (fig. S3D), whereas erbB1 and erbB3 mRNAs were not affected (17). The remaining erbB2 mRNA likely originates from Schwann cells (11), because ErbB2 protein from nerves rather than muscle fibers accounted for the protein detected in mutant muscle, and ErbB2 could not be detected in nerve-free muscle segments (fig. S3E; ErbB4 protein could not be resolved). Combined, these data strongly suggest that undetected failure of recombination, if present at all, cannot explain ErbB-dependent maintenance of neuromuscular synapses in all fibers.

Fig. 1.

NRG signaling is blocked in mutant myotubes; fetal neuromuscular junctions in HSA-Cre;erbB2fl/–;erbB4fl/fl mice develop normally. (A) Wild-type, but not ErbB-deficient myotubes express AChRϵ mRNA in response to 5 nM NRG1β (means ± SEM, from three mutant and four wild-type cultures). P values in two-tailed t tests are <0.01 for wild-type and >0.2 for mutant cultures. (B) Wild-type and mutant diaphragm muscles stained for AChR (red), neurofilament, and synaptophysin (green) at E14.5. Arrowheads denote nerve-free AChR clusters. Scale bar, 50 μm.

We next compared synapses in wild-type and mutant animals. In wild-type diaphragm, nerve-free AChR clusters developing as part of an intrinsic developmental program can be seen from E13 up to E16.5; they are then consolidated into mature synaptic contacts opposed by nerve terminals (18, 19). Innervated as well as nerve-free AChR clusters were also observed in mutant muscles of the same age (Fig. 1B at E14.5). Another hall-mark of developing endplates is a synapse-specific switch in AChR subunit composition from α2βγδ to α2βϵδ, beginning in leg muscle during the first postnatal week (20). A similar switch occurred in ErbB-deficient muscle (fig. S4). Taken together these experiments suggest that the development of the AChR clusters, the timing of innervation, and the developmental maturation of the endplates were not affected by the lack of NRG/ErbB signaling to muscle.

Whereas initial stages of the formation of postsynaptic structures might be independent of transcriptional regulation by the nerve, the maintenance of postsynaptic function in electrically active muscles is dependent on AChR gene expression in subsynaptic myonuclei (21). We therefore determined whether NRG signaling is required for synaptic maintenance in adult muscle fibers. Synapses in ErbB-deficient muscles appeared morphologically normal (Fig. 2A). Synaptic AChR levels, as estimated from their fluorescence after staining with Texas red–α-bungarotoxin, and the amplitudes of miniature endplate currents (MEPCs) were reduced by about 10%, and the MEPC rise and decay times were slightly prolonged (Fig. 2A; Table 1). Nevertheless, the short MEPC decay time constant in the mutants indicates that synaptic AChRs contained ϵ subunits, as observed in wild-type synapses. No obvious defects were observed in the appearance of nerve terminals (22), the number and morphology of terminal Schwann cells, and the innervation pattern (Fig. 2, C and D; Table 1), which suggested that reciprocal signals from muscle regulating presynaptic differentiation were not significantly affected.

Fig. 2.

Neuromuscular junctions in adult muscle of HSA-Cre;erbB2fl/–;erbB4fl/fl mice are marginally affected. (A, top) Endplate AChRs and sample MEPCs. Scale bar, 5 μm. The rapid MEPC decay at mutant endplates indicates synaptic AChRϵ channels. (B) Relative levels of AChRϵ, AChRδ, and MuSK mRNAs in synaptic (syn) and extrasynaptic (esyn) regions of wild-type (wt) and mutant (ko) diaphragm muscles (means ± SEM, n = 6 to 8; the relatively high extrasynaptic expression of MuSK mRNA combined with limited accuracy of synaptic dissection obscures true differences in synaptic versus extrasynaptic mRNA levels). (Bottom) In situ hybridization indicates focal synaptic accumulation of AChRϵ mRNA. Scale bar, 10 μm. (C) Terminal Schwann cells stained with antibody to S-100 (red, arrowheads) are similar in morphology and number in mutant and wild-type muscles (Table 1). Nuclei are blue. Scale bar, 10 μm. (D) Acetylcholinesterase staining of endplate bands in diaphragm. Scale bar, 3 mm.

Table 1.

Physiological and morphological parameters of mutant versus wild-type synapses. The rapid MEPC decay time constant in mutant endplates indicates the presence of adult AChRϵ channels. Endplate band width in left hemidiaphragm. Means ± SEM. N, number of synapses examined.

Parameter Wild type N Mice (n) HSA-Cre;erbB2fl/-; erbB4fl/flN Mice (n) P
Relative AChR density (%) 100 ± 2 25 5 92 ± 2 27 5 <0.05
MEPC amplitude (nA) 3.07 ± 0.11 20 3 2.70 ± 0.14 20 3 <0.05
MEPC decay time constant (ms) 1.00 ± 0.05 20 3 1.35 ± 0.07 20 3 <0.02
MEPC rise time (μs) 262 ± 5 20 3 321 ± 9 20 3 <0.01
Endplate band width (μm) 366 ± 38 4 321 ± 33 4 >0.4
Schwann cells/endplate 4.35 ± 0.28 30 2 4.68 ± 0.30 30 2 >0.5

In agreement with the minor reduction in the levels of synaptic AChR numbers and MEPC amplitudes, in mutant muscle fibers we observed a minor reduction in the levels of synaptic AChRϵ, AChRδ, and MuSK mRNA levels (Fig. 2B). Endplate bands of diaphragm analyzed separately from extrasynaptic segments revealed a reduction by 20 to 30% in AChR mRNAs in mutant muscles, but the marked synapse-specific expression of AChR mRNAs was maintained. Thus, the nerve induces close-to-normal AChR and musk expression in fibers lacking ErbB receptors.

The small changes in synaptic physiology did not affect sustained muscle strength, as indicated by the ability of wild-type and mutant mice to cling upside-down onto a horizontal wire mesh without falling off (inverted screen test). Both groups (n = 8) performed equally up to 120 s when the test was discontinued. Based on the fact that myasthenic mice fall off the wire mesh between 30 and 60 s (23), we conclude that the mutant mice studied here had no myasthenic condition.

To analyze whether synapses form de novo in adult ErbB-deficient muscle, we induced the formation of ectopic endplates. The proximal stumps of the transsected fibular nerve were transplanted onto the endplate-free region of mutant soleus muscle. Sectioning the soleus nerve led to the formation of new ectopic endplates in wild-type and ErbB-deficient muscle (Fig. 3, A to C). Exogenous stimulation of the foreign motor axons elicited endplate potentials and action potentials followed by vigorous twitch contractions. Thus, functional ectopic endplates form in mutant adult muscle for which quantitative recombination has been demonstrated (fig. S3).

Fig. 3.

Formation of ectopic postsynaptic membranes by nerves and by Agrin in adult muscle of HSA-Cre;erbB2fl/–;erbB4fl/fl mice. (A) Experimental procedure. (B) Ectopic endplates (top and middle) and original endplate (bottom) from same muscle, 40 days after foreign nerve transplantation and 18 days after soleus nerve section. Scale bar, 20 μM. (C) Miniature endplate potentials, synaptic potentials, and action potential recorded from ectopic innervation region. Calibration bars: 0.5, 1, 40 mV; 2, 20 ms. (D to H) Expression plasmids for neural chicken Agrin (pcAgrin748) and green fluorescent protein tagged with nuclear localization signal (pnlsGFP) injected intracellularly into extrasynaptic regions of single fibers induce nerve-free endplate membranes in ErbB-deficient muscles. (D) Experimental procedure. (E) Original endplate and Agrin-induced ectopic AChR cluster from same muscle. Scale bar, 5 μM. (F) Mean AChR densities (± SEM, nine ectopic and six original endplates in two muscles) are lower, but overlapping at Agrin-induced sites (diamonds). (G) Agrin-induced AChR clusters colocalize with Agrin deposits. Scale bar, 20 μM. (H) Injection of expression plasmid for neural Agrin (Agrin748), but not for nonneural Agrin (Agrin700) activates an AChRϵ subunit gene promoter–luciferase reporter construct (relative luciferase activity, means ± SEM, n = three muscles each).

If NRG signaling to the muscle is not required for synapse formation, the question arises how the neural induction of synapse-specific AChR transcription is mediated. In wild-type electrically active muscles, recombinant neural Agrin can induce ectopic postsynaptic membranes, a process depending on ectopic AChR gene transcription (24). Neural Agrin induced ectopic membranes also in ErbB-deficient fibers (Fig. 3, D to G), and it activated expression of a luciferase reporter gene driven by an AChRe subunit gene promoter-fragment, consistent with AChRe subunit gene induction (Fig. 3H). Ectopic AChR densities were on average lower than those in the nerve-induced endplate of the same muscle (Fig. 3F). However, at 3 out of 16 Agrin-induced ectopic sites examined, AChR densities were equal to or larger than those at the original endplates with the lowest AChR density. Thus, overexpressed Agrin alone can account for normal subsynaptic AChR density. The lower mean density could be related to the fact that Agrin is deposited by the nerve more focally than by transfected muscle fibers, and/or that accumulation of AChRs may depend on additional factors of limited abundance in the muscle, as suggested by the wider spread of ectopic Agrin deposits and MuSK aggregates compared with the corresponding AChR aggregates (25). Alternatively, physical interactions between the nerve terminal and the muscle fiber (26) may be involved.

Our findings demonstrate that the development and maintenance of neuromuscular synapses are only marginally affected in muscles that lack the ErbB2 and ErbB4 receptors. Neural Agrin can induce postsynaptic differentiation, a process requiring the induction of genes for synaptic components such as MuSK, Rapsyn, and AChRϵ. Taken together, our results indicate that the nerve regulates synapse-specific gene transcription, at least in part by the secretion of Agrin, and that neuromuscular NRG signaling is redundant with that by Agrin (but not vice versa). Finally, NRG is not required to activate the reverse signaling mechanisms from muscle that control differentiation of the presynaptic nerve terminal.

The changes in synaptic physiology in ErbB-deficient muscles are more subtle than those observed in mice heterozygous for a certain NRG1β isoform (27). Because Schwann cells depend on NRG-mediated signaling from motor neurons for survival (12), the stronger defects in the NRG1β mutants could be explained by impaired Schwann cell function, which in turn may impair the motor neuron's ability to maintain the synaptic muscle membrane.

Given that NRG is not required for synapse-specific gene transcription at the NMJ, the proposed function of NRG in regulating glutamate receptor expression during synapse development in the brain (28) warrants further examination.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S5

References and Notes

References and Notes

View Abstract

Navigate This Article