Hyperinnervation of Neuromuscular Junctions Caused by GDNF Overexpression in Muscle

See allHide authors and affiliations

Science  13 Mar 1998:
Vol. 279, Issue 5357, pp. 1725-1729
DOI: 10.1126/science.279.5357.1725


Overexpression of glial cell line–derived neurotrophic factor (GDNF) by muscle greatly increased the number of motor axons innervating neuromuscular junctions in neonatal mice. The extent of hyperinnervation correlated with the amount of GDNF expressed in four transgenic lines. Overexpression of GDNF by glia and overexpression of neurotrophin-3 and neurotrophin-4 in muscle did not cause hyperinnervation. Thus, increased amounts of GDNF in postsynaptic target cells can regulate the number of innervating axons.

Experimental application of growth factors can alter the density and distribution of axon branches (1); hence, growth factor release may be one means by which target cells regulate the number of synaptic connections they receive (2, 3). We sought to test this idea in a system where the complement of axon branches innervating a target cell could be visualized and functionally assessed. We generated mice in which muscle fibers synthesized excess amounts of a specific neurotrophic factor, GDNF, and studied innervation at the neuromuscular junction (NMJ). GDNF was chosen because it is perhaps the most potent survival factor for motor neurons, both in vitro and in vivo (4-6). In addition, GDNF is synthesized by muscle and Schwann cells (4, 7,8) and is internalized and specifically transported retrogradely by motor neurons through a receptor-mediated process (6).

To examine the effect of increased target-derived GDNF on NMJ development, we generated several lines of Myo-GDNF mice (9) that overexpress GDNF under a muscle-specific (myogenin) promoter (10, 11) (Fig.1A). The myogenin promoter was chosen because it drives transgene expression in muscle beginning in embryogenesis, about the time axons first approach muscle fibers, and continues expression into postnatal life (10).

Figure 1

(A) Schematic ofMyo-GDNF transgene construct. The 4.4-kb construct contains 1.6 kb of myogenin promoter, 0.7 kb of mouse GDNF (mGDNF) cDNA, and 2.1 kb of human growth hormone (hGH) gene including polyadenylation signal (pA). N, Not I; B, Bam HI; X, Xho I. (B) Southern blot analysis of Myo-GDNF transgenic lines. The 0.7-kb band corresponds to transgene GDNF cDNA. The 6- and 9-kb bands correspond to the endogenous GDNF gene. Progeny from lines 7301 and 8658 (asterisks) were used for the remaining studies. (C to E) Low-power dark-field photomicrographs of in situ hybridization at the level of the forelimb of P1 pups from Myo-GDNF transgenic lines Myo-GDNF (low) and Myo-GDNF (high), and wild-type littermates with a probe specific for GDNF mRNA. The white ring in the center (B) is artifactual hybridization to bone. The wild type and the two different transgenic lines show differing extents of GDNF mRNA expression in developing muscles (M). Scale bars, 200 μm. (F to H) High-power bright-field views of silver grains [from the same sections as (C) to (E)]. Grains are specific for muscle cells. Scale bars, 5 μm.

Ten founder lines integrated the transgene Myo-GDNF (Fig.1B) (9). On the basis of the amount of mRNA detected on postnatal day 1 (P1) (12), four of the 10 lines were chosen for further study: the two with the highest expression of GDNF mRNA (lines 8658 and 7301) (Fig. 1, D, E, G, and H) and two with GDNF mRNA expression that was indistinguishable from that of the wild type (as control). These four lines were compared to wild-type littermates (Fig.1, C and F). Because the overexpression of GDNF mRNA and protein is greater in line 8658 than in line 7301, these two high-expressing lines are denoted Myo-GDNF (high) and Myo-GDNF (low), respectively. In the two high-expressing lines but not in control muscles, quantities of GDNF protein at P3 as measured by enzyme-linked immunosorbent assay (ELISA) (13) were elevated and paralleled the increases in mRNA expression (Table1).

Table 1

Comparison of GDNF mRNA and protein expressed by neonate and adult (4 to 5 months) wild-type, Myo-GDNF(low), and Myo-GDNF (high) mice.

View this table:

Neuromuscular innervation was studied at multiple time points between birth and adulthood in all four transgenic lines and in wild-type animals. Normally in mammals, two or more motor axons innervate each NMJ at birth, but this number decreases to one over the first several postnatal weeks because of synapse elimination and associated axonal branch withdrawal (14). However, innervation of NMJs in theMyo-GDNF (high) and Myo-GDNF (low) mice was abnormal in several ways during the first several postnatal weeks. First, there were many more converging axons than normal (15) (Fig. 2). At some junctions, the number of inputs was several times the highest number of innervating axons observed in age-matched wild-type animals (Fig.3). Despite the increased innervation, there was still only one NMJ on each muscle fiber, as in control muscles (Fig. 2). Second, in addition to the increase in the number of converging axons, the period of multiple innervation persisted longer than normal, doubling from 2 weeks to 1 month in Myo-GDNF(high) mice (Fig. 3A). Interestingly, when NMJs in overexpressing muscles had lost on average all but two axons, the transition to single innervation was approximately as fast as in control animals when they were similarly innervated (Fig. 3A).

Figure 2

NMJs in mice expressing Myo-GDNFare hyperinnervated during development. Images are low-power fluorescence photomicrographs of sternomastoid muscles from a wild-type mouse (A) and a Myo-GDNF (high) mouse (B) at P8. Individual muscle fibers are running diagonally from top right to bottom left; postsynaptic AChRs are labeled red with TRITC-αBTX, and presynaptic axon neurofilaments and nerve terminal synaptophysin are immunolabeled green. Nerve terminals overlying AChRs at the NMJ appear yellow. The majority of muscle fibers in the wild-type animal are contacted by only one axon at this age. Note the axon in the process of being eliminated, which appears as a retracting bulb (*). The majority of muscle fibers in the Myo-GDNF (high) animal are multiply innervated, often by three or more axons. Scale bar, 10 μm.

Figure 3

(A) Delay of synapse elimination. Upper panel: NMJs in mice expressingMyo-GDNF are innervated by more than one motor axon (multiply innervated) for a longer period of development than are junctions in normal mice. Axons innervating the sternomastoid muscle inMyo-GDNF overexpressors (solid circles) and wild-type (open squares) littermates were immunolabeled with antibodies to neurofilament and synaptophysin and were viewed by confocal microscopy. The results show a delay of about 2 weeks in the loss of multiple innervation for the Myo-GDNF (high) line. Lower panel: Graph of the average number of axons converging to innervate single muscle fibers as a function of postnatal age, showing a dose dependence of hyperinnervation on the amount of GDNF for wild-type,Myo-GDNF (low), and Myo-GDNF (high) animals. (B) Bar graph showing distribution of axons per NMJ at P8 in wild-type and Myo-GDNF (high) animals. (Cand D) Examples of confocal images of immunolabeled axons converging to single NMJs for Myo-GDNF animals during the first two postnatal months. Relative to age-matched control mice, muscle fibers in young mice expressing Myo-GDNF are innervated by more axons. In (C), an NMJ innervated by eight axons (arrows) at P4 is shown; this number of converging axons is greater than we have observed in normal animals at any postnatal age. Scale bar, 5 μm. In (D), examples of NMJs at different postnatal ages are shown: P8 (NMJ with five converging axons), P14 (NMJ innervated by three axons), P22 (endplate innervated by three axons), and P63 (100% of muscle fibers are singly innervated). Scale bar, 5 μm (P8–P22), 8.3 μm (P63).

The prolonged period of multiple innervation in these animals did not appear to be caused by a systemic maturational delay: Although the transgenic animals' weight tended to be slightly lower than that of control littermates during the first few postnatal weeks (90% of control at P0, 85% at P8, 70% at P14, and 83% at P23), eye opening, fur growth, weaning, and reproduction all occurred at appropriate times. We found that by adulthood, amounts of GDNF protein from the transgene dropped somewhat in both Myo-GDNF (high) andMyo-GDNF (low) mice (Table 1). Thus, we do not yet know if the eventual loss of multiple axonal convergence could be prevented altogether by maintained elevation of GDNF. The observed hyperinnervation appeared to be related to the dose of GDNF because both the number of converging axons per multiply innervated endplate and the time required for an entire muscle to become uniformly singly innervated correlated with the amount of GDNF expressed in theMyo-GDNF (high) and Myo-GDNF (low) lines (Fig.3A).

To determine whether GDNF overexpression may have induced innervation of muscle fibers by sensory or autonomic axons, which express the GDNF receptor Ret (16) but would not be expected to be functional, we recorded synaptic responses from diaphragm muscle fibers of transgenics and wild-type controls while stimulating the phrenic nerve with varying voltages (17). At P9 to P10, 80% of muscle fibers (n = 51, three animals) in theMyo-GDNF (high) line were multiply innervated. Of these, 56% (23 of 41) were contacted by two axons, and 44% (18 of 41) were contacted by three or more axons (Fig.4). In some cases, we found that as many as three inputs to a muscle fiber were sufficiently strong to drive the muscle fiber to contract (Fig. 4, inset). In contrast, only 20% of muscle fibers (n = 55, two animals) in wild-type controls were multiply innervated, and in all cases by two axons. Therefore, the extra axonal inputs form functional contacts and almost certainly arise from motor neurons.

Figure 4

Extra innervation of muscle fibers in Myo-GDNF overexpressors is functional. Examples of intracellular recording traces of synaptic potentials elicited by gradual recruitment of motor axon innervation by progressively increasing the strength of stimulus to the phrenic nerve. Superimpositions of traces from a diaphragm muscle fiber are shown for wild-type and Myo-GDNF (high) mice at P9. Graded stimulation of the phrenic nerve evoked a single endplate potential from the wild-type fiber, indicating that the muscle fiber is singly innervated. Graded stimulation of the phrenic nerve evoked three distinct endplate potentials from the Myo-GDNF muscle fiber, indicating that the fiber is innervated by three separate axons. All three of these axons were capable of driving the muscle fiber to threshold (insert; scale represents 5 mV, 5 ms).

The hyperinnervation of muscle found in Myo-GDNF mice could have arisen because of a greater number of motor neurons innervating the muscle as a whole (as might occur if GDNF spared motor neurons from naturally occurring cell death) or because of a greater number of axonal branches in the innervating nerve. However, retrograde labeling of the motor neurons innervating the sternomastoid muscle with Fluoro-Gold (18) at P11 and counts of myelinated axons in the nerve to the sternomastoid at P9 to P10 (19) showed that the number of motor neurons or axon branches was only slightly higher in Myo-GDNF mice than in wild-type animals (Table2). Thus, it seems more likely that the hyperinnervation is a consequence of more axonal branching within the muscle (that is, larger motor units).

Table 2

Comparison of the numbers of motor neurons, axons, muscle fibers, and motor units in wild-type and Myo-GDNF(high) mice (n.s., not significant).

View this table:

Motor unit sizes were determined from twitch tension measurements at P10 (20). Maximal muscle tension was elicited inMyo-GDNF (high) animals by activation of only 66% of the number of motor axons needed to elicit maximal muscle tension in wild-type controls, indicating a greater degree of overlap in muscle fiber innervation by different axons (Table 2). Because excess GDNF did not alter the number of muscle fibers (Table 2), we conclude that motor units in P10 Myo-GDNF transgenic mice are at least 1.5 times the size of those in wild-type mice, and that these motor axons coinnervate the same NMJs more often than in wild-type mice.

During the period of greatest hyperinnervation (birth to 3 weeks postnatal), Myo-GDNF mice exhibit a tremor (Fig.5). At neonatal ages, the shaking is sufficiently obvious that transgenic animals can be distinguished from their littermates without error. The tremor severity wanes as multiple innervation diminishes. The magnitude of the tremor was greater in amplitude and persisted over a longer developmental period in theMyo-GDNF (high) line than in the Myo-GDNF (low) line. Normal rodent neonates have a tremor that is most obvious during the first few postnatal days and gradually subsides over the next week (21). This tremor may be analogous to “jitteriness” in human neonates (22). Tremor disappearance corresponded to the loss of multiple innervation in each transgenic line, as it did in wild-type animals (23). The tremor therefore may be a reflection of the large number of muscle fibers innervated by each axon during early development, such that each motor axon impulse resulted in an obvious muscle twitch.

Figure 5

Myo-GDNF animals display an activity-dependent high-frequency tremor during the period of hyperinnervation. The photograph (exposure, 1 s) shows a P4Myo-GDNF mouse standing next to its wild-type littermate.

Muscle-specific overexpression of neurotrophin-3 (NT-3) with the myogenin promoter (Myo–NT-3) did not cause extra innervation of muscle fibers at the NMJ (24), even though motor neurons expressed the NT-3 receptor trkC and responded to NT-3 with increased survival (25-27) and even though these animals had an increased number of proprioceptive sensory axons and target muscle spindles (11). Muscle-specific overexpression of NT-4 (Myo–NT-4) also did not cause extra innervation of muscle fibers (24), even though motor neurons express the trkB receptor (26-28) and NT-4 causes sprouting in adult rodents (29).

The source of excess GDNF affects the amount of hyperinnervation. Overexpression of GDNF by glial cells under the glial fibrillary acidic protein (GFAP) promoter (30) had no effect on the number of axons converging on muscle fibers or the time course of synapse elimination, even though motor neurons in these animals are less susceptible to axotomy-induced cell death as a consequence of GDNF (31). However, we do not yet know whether glial cells at the NMJ (terminal Schwann cells) express the GFAP-GDNFtransgene.

The hyperinnervation seen with muscle-specific overexpression of GDNF is more extreme than that described after parenteral administration of neurotrophins and other neurotrophic factors. These previous manipulations had modest effects on the time course of synapse elimination and had no effect on the number of innervating axons each target cell received (32). The effect with muscle-derived GDNF is also more pronounced than that seen in the mutant mouseparalysé, in which a deficit in neuromuscular activity is thought to delay synapse elimination but apparently does not cause extra axonal convergence (33).

It is not known how GDNF causes hyperinnervation. Muscle-derived GDNF may act as a synaptotrophin to prolong the maintenance of synaptic connections that were established during early development (3). Alternatively, GDNF may have caused hyperinnervation by inducing motor axons to establish extra terminal branches that are capable of forming synapses. In the latter case, the period of multiple innervation might be prolonged because of the additional time necessary to eliminate the abnormally large number of axons at each NMJ. In either case, our results demonstrate that enhanced trophic factor expression by postsynaptic cells can increase the amount of innervation they receive. Moreover, this synaptic plasticity is mediated by a growth factor outside the prototypical neurotrophin family.

  • * These authors contributed equally to this report.

  • To whom correspondence should be addressed. E-mail: jeff{at}


View Abstract

Navigate This Article