Maintenance of Acetylcholine Receptor Number by Neuregulins at the Neuromuscular Junction in Vivo

See allHide authors and affiliations

Science  25 Apr 1997:
Vol. 276, Issue 5312, pp. 599-603
DOI: 10.1126/science.276.5312.599


ARIA (for acetylcholine receptor–inducing activity), a protein purified on the basis of its ability to stimulate acetylcholine receptor (AChR) synthesis in cultured myotubes, is a member of the neuregulin family and is present at motor endplates. This suggests an important role for neuregulins in mediating the nerve-dependent accumulation of AChRs in the postsynaptic membrane. Nerve-muscle synapses have now been analyzed in neuregulin-deficient animals. Mice that are heterozygous for the deletion of neuregulin isoforms containing an immunoglobulin-like domain are myasthenic. Postsynaptic AChR density is significantly reduced, as judged by the decrease in the mean amplitude of spontaneous miniature endplate potentials and bungarotoxin binding. On the other hand, the mean amplitude of evoked endplate potentials was not decreased, due to an increase in the number of quanta released per impulse, a compensation that has been observed in other myasthenic states. Thus, the density of AChRs in the postsynaptic membrane depends on immunoglobulin-containing neuregulin isoforms throughout the life of the animal.

The fidelity of neuromuscular transmission depends on the extraordinarily high density of AChRs in the postsynaptic muscle membrane. Developmental studies have pointed to the important trophic influence of the motor nerve in the regulation of endplate AChR density (1), an effect strong enough to override the suppression of AChR synthesis by muscle activity. Part of the nerve’s local influence on AChR density is to promote the immobilization of AChRs, an effect mediated by the glycoprotein agrin (2). Another important local influence of the motor nerve is to increase the synthesis and insertion of AChRs into the postsynaptic membrane (3). In fact, endplate nuclei in developing and mature muscle are known to transcribe AChR subunit genes at a high rate as compared with that in nonsynaptic nuclei (4, 5). The effect of synthesis on local receptor density remains evident in mice that lack the principal cytoplasmic AChR anchoring protein rapsyn (6).

The most likely candidate for mediation of the motor neuron’s influence on endplate AChR synthesis is ARIA (for acetylcholine receptor–inducing activity), a protein purified from brain extracts on the basis of its ability to stimulate the synthesis of AChRs in cultured myotubes (7, 8). ARIA is a member of the neuregulin family of ligands for the 185-kD transmembrane receptor tyrosine kinases ErbB2, ErbB3, and ErbB4 (also called Her2, Her3, and Her4), which are closely related to the epidermal growth factor (EGF) receptor (9, 10). Neuregulins are potent activators of muscle AChR synthesis, with a median effective dose (ED50) of 25 to 50 pM (11); neuregulin mRNA can be detected in embryonic motor neurons when motor axons first invade peripheral muscle masses (12, 13) and is also abundant in adult motor neurons (13); neuregulin receptors are present in skeletal muscle cells and may be concentrated at the neuromuscular junction (14-16); neuregulin protein is concentrated in motor nerve terminals (15-18) and accumulates in the extracellular matrix of the synaptic cleft (12, 15, 18); and in mammalian muscle, neuregulin increases mRNA encoding ε (17,19), an AChR subunit that replaces the γ subunit during development. Neuregulins may therefore mediate the nerve-dependent maturation of junctional AChRs (20) as well as enhance overall AChR gene expression at developing and mature nerve-muscle synapses.

More direct evidence of the role of neuregulins at neuromuscular junctions would require the selective inhibition or elimination of neuregulin activity, or both. When mice are genetically altered so that exons encoding the EGF-like domain (21) or the immunoglobulin (Ig)-like domain (22) of neuregulin or the neuregulin receptors ErbB2 (23) and ErbB4 (24) have been deleted, homozygous animals die on or about embryonic day 10 (E10) with defects of the heart, cranial ganglia, and hindbrain. This is well before the formation of neuromuscular synapses begins on about E15 (25). Heterozygous animals appear normal and are fertile. This, of course, does not exclude a subtle defect at the neuromuscular junctions.

Although brain-purified ARIA contained Ig-domain amino acid sequences, Ig-containing isoforms may represent a small fraction (about 10 to 20%) of the total that are present in motor neurons (13,26). Other forms, in which a cysteine-rich region replaces the Ig-like domain (Fig. 1A), predominate. We studied mice deficient in Ig-containing neuregulins because these isoforms may be particularly important at the neuromuscular junction. The Ig-like domain binds heparin (8, 27), an affinity that may be responsible for the observed association of neuregulin with the extracellular matrix of the synaptic cleft (12, 15, 18). Thus, by accumulating at the endplate, Ig-containing neuregulins may exert a major effect on endplate AChR synthesis. In fact, molecules associated with the synaptic basal lamina have been shown to promote the synthesis of AChRs in denervated adult muscle (28), and they appear to have a selective effect on the ε-containing (adult) type of AChR (5, 29).

Figure 1

Heterozygote mice are deficient in Ig-containing neuregulin isoforms. (A) The two major splice variants of neuregulin that are NH2-terminal to the EGF-like domain. One isoform contains an Ig-like domain, and in the other the Ig-like domain is replaced by a region that contains eight cysteine residues. (B) Nuclease protection of spinal cord RNA from seven heterozygote and five wild-type (WT) animals with a probe corresponding to the Ig-like domain (30). A GAPDH probe was used to measure equivalent loading. Msp, an Msp I digest of pBR322 DNA; MW, a 100–base pair molecular weight ladder (the numbers at left and right indicate length in nucleotides); neg., protection of probes when hybridized to an equivalent weight of yeast RNA. (C) Nuclease protection with a probe derived from the EGF-like domain. A simultaneous assay for GAPDH was used to establish equivalence of RNA loading (not shown).

All neuregulin-deficient mice used in this study were derived from founder animals described in the original publication on the disruption of the Ig-like domain of the neuregulin gene by homologous recombination (22). Heterozygous (+/–) mice as well as wild-type (+/+) littermates were genotyped by polymerase chain reaction (PCR) analysis of genomic DNA (22). The abundance of Ig-containing neuregulin transcripts in the spinal cord was quantified by S1-nuclease protection assay (30). Protected fragments that represent Ig-containing neuregulin mRNA were reduced in each of seven heterozygotic animals as compared to five wild-type animals (Fig.1B). The mean reduction was 45%. The concentration of glyceraldehyde phosphate dehydrogenase (GAPDH) mRNA, which was measured simultaneously as a control, was identical in the two groups (Fig. 1B). Although RNA from the entire spinal cord was analyzed in this experiment, in situ hybridization studies of spinal cord sections have shown that neuregulin mRNA resides largely in motor neurons (8, 10, 12, 13,31). Therefore, expression of Ig-containing neuregulins is reduced in the motor neurons of the heterozygote animals. In contrast, total neuregulin expression in the spinal cord, as measured by nuclease protection with a probe from the EGF-like domain, was not significantly reduced (Fig. 1C). The expected small reduction (on the order of 5 to 10%) would not have been detected in this assay.

Neuromuscular transmission, assayed by recording compound muscle action potentials (CMAPs) during repetitive nerve stimulation (32), was more sensitive to curare in heterozygous animals than in their wild-type littermates. Incremental doses of d-tubocurarine, an AChR antagonist, were intraperitoneally administered at 10-min intervals while trains of stimuli were delivered at 1-min intervals. In the experiment shown in Fig. 2, CMAPs recorded from the heterozygous mouse began to decrease in amplitude 6 min after the 280 nmol per kilogram of body weight (nmol/kg) pulse of curare, whereas those recorded from the wild-type littermate did not decrease until 6 min after the 1120 nmol/kg dose. A decrement in CMAP amplitude reflects the progressive dropout of individual muscle fibers as the evoked endplate potential drops below the threshold of action potential generation. We observed CMAP decreases at significantly lower curare concentrations in the heterozygote mouse in four of the five littermate pairs studied. In one pair, the difference was minimal (only 4 min earlier in the heterozygote animal) and probably insignificant. In the absence of curare, there was no significant decrease in the CMAP amplitude during 3- to 30-Hz trains of 10 stimuli delivered to mice of either genotype.

Figure 2

The safety factor of neuromuscular transmission is reduced in heterozygote mice as compared with that in wild-type mice. CMAPs recorded from forelimb flexor muscles of a heterozygote mouse (A) and of a wild-type littermate (B) in response to 30-Hz stimulation of the ipsilateral brachial plexus (8, 10,12, 13, 31). Both records were obtained 4 min after the 560 nmol/kg dose of intraperitoneal curare. (C) Complete, timed, incremental dose-response curves. Incremental doses of curare (bold numbers) were administered at 10-min intervals, whereas CMAP responses were recorded at 2-min intervals. The percent decrement is based on the amplitudes of the second and sixth CMAPs during the 30-Hz trains. CMAPs do decrease in wild-type animals, but do so after a delay and an additional, higher dose of curare.

In a separate set of experiments, single doses of d-tubocurarine were administered to heterozygote and wild-type mice, and trains of CMAP amplitudes were recorded at 1-min intervals for a period of 30 min. In six of seven heterozygote mice (86%), a single intraperitoneal dose of 300 nmol/kg produced at least a 10% decrement between the second and sixth CMAP, whereas only two of seven wild-type animals (29%) exhibited the same CMAP decrease at that dose. Thus, the safety factor of synaptic transmission at heterozygote endplates was reduced, which is consistent with a deficit in the postsynaptic density of AChRs in these animals.

We tested the postsynaptic sensitivity to ACh by measuring the amplitude of spontaneous miniature endplate potentials (MEPPs) with intracellular microelectrodes in isolated diaphragm-phrenic nerve preparations (33). As illustrated in Fig. 3, A and B, the mean MEPP was reduced at heterozygote endplates as compared with wild-type controls. Table 1 summarizes the results from four wild-type and four heterozygous mice. On average, there was an approximately 30% reduction in MEPP size (v 1). This difference could not be attributed to a disparity of resting membrane potential (V m), nor could it be due to differences in placement of the recording electrode relative to the endplate, because the mean rise times of the synaptic potentials (t 10–90) were virtually identical (Table 1).

Figure 3

AChR number is reduced at heterozygote endplates as compared with wild-type endplates. (A) Examples of stimulus-evoked and spontaneous synaptic potentials recorded from endplates of wild-type (+/+) and heterozygote (+/–) mice. MEPPs were often seen following nerve-evoked EPPs. The MEPP amplitudes (v 1) can be directly compared at these two endplates, because the resting membrane potentials are equivalent. (B) Amplitude histograms of MEPPs recorded at the same endplates. (C) Autoradiography of teased muscle fibers labeled with 125I–α-bungarotoxin. (D) Estimates of AChR density obtained at individual endplates are shown as open symbols, and the means adjacent to each column are shown as solid circles. Bars indicate standard errors. The ordinate shows the number of pixels below a preselected threshold (34).

Table 1

Summary of synaptic potentials recorded from wild-type and heterozygous mice.

View this table:

More direct evidence for a reduction in overall endplate AChR density was obtained in α-bungarotoxin binding experiments (34). Freshly dissected diaphragms were exposed to saturating concentrations of 125I-bungarotoxin. In one set of experiments, the diaphragms were cut into endplate-rich strips, and the radioactivity was determined in a gamma counter. There were approximately 30% fewer specific counts per minute (total counts per minute minus counts per minute bound in the presence of excess unlabeled α-bungarotoxin) in the endplate-rich hemidiaphragm strips of heterozygote animals as compared with those of wild-type animals [8756 ± 765 cpm, mean ± SEM, n = 6; versus 12,462 ± 949 cpm,n = 6 (P = 0.012)]. In another set of experiments, the diaphragms were teased into fascicles of fewer than five muscle fibers and processed for autoradiography. Clusters of grains were less dense in diaphragms from heterozygote animals than in those from wild-type animals (Fig. 3C). Morphometric analysis of 49 endplates from three wild-type animals and 69 endplates from three heterozygote animals showed some overlap between the two samples. However, the ratio of the mean grain densities was 51%, and this reduction was highly significant (P < 0.001). Each measurement and the means of each group are shown in Fig. 3D. Our results show that AChRs are reduced in number at the neuromuscular junctions of mice that are deficient in Ig-containing neuregulins as compared with those of wild-type mice.

Nerve-evoked endplate potentials (EPPs) were studied under conditions of elevated (12 mM) extracellular Mg2+ (33), which reduces the probability (p) of ACh release. Surprisingly, the mean evoked EPP response was not reduced at heterozygote endplates (Table 1). Because quantal size is reduced, this implies that more quanta are released per impulse from heterozygote than from wild-type motor nerve terminals. In fact, the mean quantal content calculated from the ratio of mean EPP/mean MEPP was twice as large in heterozygote as compared with wild-type endplates (m 1 in Table 1). The difference in mean quantal content is also reflected in the paucity of failures at heterozygote endplates as compared with wild-type endplates (Fig.4A). In fact, many heterozygote endplates showed no failures at all during 300 pulses, precluding the calculation of mean quantal content (m) by the method of failures (m = ln N/n 0) based on Poisson statistics. We used another estimate of m that does not rely on the use of failure number. If Poisson statistics apply, then the mean quantal content (m 2) is equal to 1/CV 2, where CV is the coefficient of variation. Estimates of m 2 showed an approximately twofold increase at heterozygote endplates as compared with wild-type endplates (Table 1). Figure 4B illustrates the disparity between wild-type and heterozygote endplates. The close correlation between the values for m 1 andm 2 suggests that Poisson statistics did apply under our conditions. On the basis of independent estimates ofm, we conclude that presynaptic nerve terminals appear to compensate for reduced postsynaptic ACh sensitivity by increasing the number of quanta released per pulse.

Figure 4

The number of quanta released per impulse is increased at heterozygote endplates as compared with wild-type controls. (A) Amplitude histograms of nerve-evoked EPPs recorded in high extracellular Mg2+. The shaded bar indicates the number of failures (no response after nerve stimulation). (B) Mean quantum content calculated as 1/(CV) 2 (m 2) as a function of mean quantum content calculated as meanV epp /meanv 1(m 1). The quantal content calculated by either method is higher at heterozygote endplates (open triangles) than at wild-type endplates (solid ovals). The slope of this relation is slightly greater than 1, probably because the smallest MEPPs disappear into the noise, leading to an underestimate ofm 1; and because the smallest evoked responses are also lost, leading to an overestimate ofm 2. (C) Overall endplate size, revealed by cholinesterase histochemistry, does not differ between wild-type and heterozygote animals.

The increase in mean quantal content might be due to the compensatory sprouting of motor nerve terminals. However, we found no evidence of an increase in endplate size, as estimated by cholinesterase staining of teased muscle fibers (35) (Fig. 4C).

Despite the increase in mean quantal content and the relatively normal EPP amplitudes at heterozygote endplates, the population response (CMAP) declined during repetitive stimulation at lower doses of curare in the heterozygote animals. This is not unexpected, given the direct relation between the rate of decline of transmitter release at neuromuscular junctions and the initial mean quantal content (36). When stimulated at a rate of 25 Hz, neuromuscular junctions in the heterozygote mice fatigued more rapidly than in wild-type mice (37) (Fig. 5). In these experiments, extracellular Mg++ was reduced from 12 mM to physiologic levels (1 mM), so that neurotransmitter release was not suppressed. The size of EPPs was reduced with concentrations of d-tubocurarine just sufficient to block action potential generation.

Figure 5

Evoked EPPs decay more rapidly at heterozygote endplates (open triangles) than at wild-type endplates (solid circles). These experiments were recorded in normal Mg2+ (1 mM) with synaptic transmission blocked by curare. EPP amplitudes were normalized to the highest amplitude obtained during the tetanus, which was delivered at 25 Hz. Each point on the wild-type curve represents the mean of 29 endplates in nine mice, and each point on the heterozygote curve represents the mean of 35 endplates in nine mice. Asterisks mark significant differences (P < 0.05).

We have shown that neuregulin-deficient mice have a reduced safety factor of neuromuscular transmission. Intracellular microelectrode recordings of MEPP amplitudes, as well as measurements of125I–α-bungarotoxin binding, indicate that the myasthenia is due to the loss of AChRs from the postsynaptic muscle membrane. Our data provide the most direct evidence to date that neuregulins play a crucial role in maintaining AChR number at motor endplates in vivo.

The heterozygote knockout animals studied here appear to express diminished levels only of neuregulin isoforms with an Ig-like domain. These isoforms represent a small minority of the total, and the remainder (that is, forms that substitute the Ig-like domain with the Cys-rich motif) are presumably biologically active. Our results suggest that Ig-containing isoforms of neuregulin are present in limiting amounts at neuromuscular junctions and are crucial in maintaining postsynaptic AChR density at neuromuscular junctions. This may be the result of the immobilization of this isoform to extracellular matrix components of the synaptic cleft. Thus, although the EGF-like domain appears to be sufficient for activating neuregulin receptors, the free ligand may be less effective in influencing synapse formation in vivo, perhaps because it cannot be concentrated near its site of action.

We observed an increase in the release of ACh at the neuromuscular junctions of heterozygous animals, an apparent compensatory response to the loss of postsynaptic ACh sensitivity, which is mediated by an unknown retrograde signal. A similar phenomenon has been observed in an experimental model of myasthenia gravis (38) and in biopsies of human muscle from patients with myasthenia gravis (39). The augmented transmitter release, although adequately compensating for the loss of AChRs during low rates of motor nerve stimulation, increases the likelihood of synaptic failure during prolonged, high rates of stimulation. The compensation was not caused by an enlargement of the endplate. Thus, the increased quantal content of nerve-evoked ACh release in the heterozygote animals probably reflects an increase in the amount of ACh released per unit length of nerve terminal. It remains to be determined whether this is the result of an increase in the number of release sites or of an increase in the probability of release at each site.

Our results imply that the continued activation of AChR genes in subsynaptic nuclei by neuregulin is critical for the maintenance of postsynaptic AChR density at adult endplates, despite the decrease in AChR turnover during development and despite the ability of agrin to immobilize AChRs in the synapse. In fact, neuregulin-like immunoreactivity remains high at adult neuromuscular junctions (15, 18). The fact that the decrease in neuregulin gene expression is roughly proportional to the loss of AChR number at endplates implies that postsynaptic ACh responses are closely linked to the rate of AChR synthesis and insertion into the postsynaptic membrane (2). We do not yet know when the deficit in neuromuscular transmission first becomes evident in the neuregulin-deficient mice during development. Indeed, neuregulin genes are expressed in motor neurons early enough to affect initial neuromuscular synapse formation (8, 12, 13). Studies of neuregulin-deficient mice during development will shed light on when neuregulin action first becomes essential at neuromuscular junctions.

  • * These authors contributed equally to this work.

  • Present address: Program in Neuroscience, Florida State University, Tallahassee, FL 32306, USA.


View Abstract

Stay Connected to Science

Editor's Blog

Navigate This Article