Synaptic Segregation at the Developing Neuromuscular Junction

See allHide authors and affiliations

Science  20 Nov 1998:
Vol. 282, Issue 5393, pp. 1508-1511
DOI: 10.1126/science.282.5393.1508


Throughout the developing nervous system, competition between axons causes the permanent removal of some synaptic connections. In mouse neuromuscular junctions at birth, terminal branches of different axons are intermingled. However, during the several weeks after birth, these branches progressively segregated into nonoverlapping compartments before the complete withdrawal of all but one axon. Segregation was caused by selective branch atrophy, detachment, and withdrawal; the axon branches that were nearest to the competitor's branches were removed before the more distant branches were removed. This progression suggests that the signals that mediate the competitive removal of synapses must decrease in potency over short distances.

Competitive synaptic rearrangements change the number and strength of axonal connections with target cells in many parts of the nervous system (1). At each neuromuscular junction, for example, motor axons compete until all but one are removed (2). By iontophoretic application of lipophilic dye (3, 4), we observed synaptic competition between individual axon terminals at the developing neuromuscular junction of the mouse.

We began by looking at neuromuscular junctions during the second postnatal week [postnatal days 9 and 10 (P9-10)], when most muscle fibers (88%) were singly innervated and the rest (12%) were innervated by two axons (5). At the singly innervated junctions (n = 242; 20 muscles), the entire postsynaptic receptor territory was overlain by one axon, as observed by staining with lipophilic dyes or with nerve-specific antibodies (6). However, at the remaining 12% of the junctions, a labeled axon occupied only part of the receptor territory (Fig. 1, A through D). Subsequent staining with antibodies to neurofilaments or with tetanus toxin (6) showed that, in each of these cases (n = 33), the remaining part of the neuromuscular junction was innervated by another axon (Fig. 1A). In these dually innervated junctions, the labeled terminals of each individual axon occupied a variably sized but contiguous region extending from one edge of the junction toward the center (Fig. 1, A through D). Thus, in the second postnatal week, each axon's synaptic territory is confined to a separate region of the junction.

Figure 1

Confinement of one axon's terminals to part of a multiply innervated neuromuscular junction at later stages (P9-10) of development. (A) A single lipophilic dye–labeled axon (green) occupies a contiguous region of the postsynaptic site, which was labeled by staining the AChRs with fluorescently tagged α-bungarotoxin (red). (A′) Neurofilament staining of the same junction shown in (A) reveals that this junction is innervated by two axons. (B through D) Individual axons occupied variably sized but contiguous areas of multiply innervated junctions at P9-10. (E and F) In contrast, at birth, individual axons occupied multiple sites distributed throughout the junctional area. Scale bar, 20 μm [(A) through (D)] and 10 μm [(E) and (F)].

By contrast, at earlier stages (P1-3), a single axon typically occupies multiple small regions that are distributed throughout the neuromuscular junction (Fig. 1, E and F). As expected, the sites that are occupied by different axons are intermingled (n = 26) (Fig. 2A) (7). The difference in the distribution of axon terminals between P1-3 and P9-10 could mean that the areas that are occupied by different axons segregate from each other over time. Alternatively, a small number of junctions may already be segregated at birth, and these junctions may retain multiple innervation longer than intermingled ones. To explore these possibilities, we used different lipophilic dyes to separately label each axon that innervated the same neuromuscular junction, and we measured the separation between the terminal areas that were occupied by the two competing axons (8) (Fig. 2, A through C). We found no junctions at birth in which axons had completely nonoverlapping territories. Thereafter, both the absolute separation (the distance between two centroids) and the relative separation [the distance normalized to the total length of the junction (Fig. 2E)] of the terminal areas gradually increased. In most cases, the separation was only partial at P3-4 (Fig. 2, B and E) and was complete by P7-10 (Fig. 2, C and E). By P16-17, when very few junctions (1.4%) were still multiply innervated, axon territories in all of the multiply innervated junctions (8 out of 8) showed a high degree of separation with an uninnervated gap in between (9) (Fig. 2, D and E). Therefore, neuromuscular junctions become progressively compartmentalized as development proceeds.

Figure 2

Axons at multiply innervated neuromuscular junctions gradually segregate as development proceeds. (A) Two intermingled competing nerve terminals at this P1 junction labeled with DiO (green) and DiI (red). Small white circles indicate centroids of each labeled axon (8). (B) A multiply innervated junction at P4 shows less intermixing than seen at birth. Arrows (red and green correspond to DiI- and DiO-labeled branches, respectively) show thin branches of each axon in the territory that is dominated by the other axon. (C) A multiply innervated junction at P10 in which the two inputs are separated by an uninnervated gap (at most junctions at this age, however, the inputs are separate but juxtaposed). (D) At P17, two competing axons at this rare remaining multiply innervated junction show a large separation between two inputs. In this case, both axons were stained with an antibody to neurofilament (green) and the postsynaptic AChRs were labeled with fluorescently tagged α-bungarotoxin (red). (E) The segregation index (8) increases through early postnatal life. Each circle represents an individual junction; horizontal bars are the means of the segregation index. Scale bar in (D), 10 μm [shows scale for (A) through (D)].

To learn how this segregation of terminals occurred, we examined multiply innervated junctions at P5-7 before segregation was complete in all junctions (n = 62). We found that, at multiply innervated junctions, individual axons often (30 out of 62) possessed two types of branches, thick and thin (Fig. 3, A and B). The thin axon branches were always nearest to or within the territory occupied by the competing axon, whereas the thickest branches were located on the perimeter of the junction at the greatest distance from the competing axon's territory. Some thin branches (6 out of 30) had swollen ends (arrows inFig. 3, A and B) that were reminiscent of “retraction bulbs,” which were previously observed during the final stages of synapse elimination when an axon withdraws completely from a junction (7, 10) (see also Fig. 4B). Neurofilaments were detectable in the thick branches but not in the thin branches (insets in Fig. 3, A and B). The thick branches were closely apposed to the postsynaptic membrane. In contrast, the thin branches were often several microns above the postsynaptic receptor plaque (Fig. 3C). Together, these data suggest that branches of the same axon within the same neuromuscular junction behave differently; whereas the thin branches undergo retraction, the thick branches remain intact. These observations also suggest that segregation of competing axon terminals is caused by a progressive process of branch atrophy, detachment, and withdrawal that begins with the branches nearest the territory dominated by a competing axon.

Figure 3

Atrophy and detachment of the branches nearest to the synaptic region occupied by a competitor result in segregation at developing neuromuscular junctions. (A andB) Two P7 junctions in which only one of the two nerve terminals was stained with DiO (green). The postsynaptic AChRs were labeled with fluorescently tagged α-bungarotoxin (red). Very thin terminal branches ending in swellings (arrows) were located close to the other axon's territories. Insets show neurofilament staining, which reveals the presence of a second axon in each case. The very thin branches, however, were not detectable with an antibody to neurofilament. (C) Stereo pairs of a P5 junction showing that the atrophic branch is detached from the postsynaptic site as it crosses the width of the junction, whereas the thicker branches are closely apposed to the postsynaptic sites. A three-dimensional movie of this junction can be seen Scale bar, 20 μm [(A) and (B)] and 10 μm (C).

Figure 4

Synapse elimination continues after segregation. (A) DiO staining of two axons innervating the same neuromuscular junction at P7. The postsynaptic AChRs were labeled with fluorescently tagged α-bungarotoxin (red). The two axons differ in caliber. The area occupied by the thinner axon is smaller and is adjacent to a faintly stained patch of AChRs (arrow). (B) A retracting thin axon (arrowheads) terminates in a retraction bulb, which is near a very low density AChR site (arrow). These receptors were presumably occupied by that axon before its withdrawal. Thus, in development as in synaptic competition in adult muscle (5,20), the loss of receptors continues after the nerve has vacated a synaptic site. Scale bar, 15 μm (A) and 10 μm (B).

Given the intermingling of axons throughout all regions of neuromuscular junctions at birth, the complete absence of overlap in the second postnatal week implies that both inputs (the eventual winner and loser) lose branches in many junctions. We observed doubly innervated junctions in which both axons possessed thin branches in the territories dominated by their competitors and thick branches at a distance from the other axon's terminals (Fig. 2B, arrows). Therefore, at some junctions (at least), both axons appear to be relatively effective competitors that are capable of removing some branches of the other axon.

After segregation, branches of one of the two inputs must have continued to withdraw because the average percentage of multiply innervated junctions changed from 42% at P7 to 4% at P12 (5) and to 0% in adult animals (in this study, 0 out of 354). At most of the segregated multiply innervated junctions (77 out of 90) at P7-12, we observed no distinctions between the axons that would suggest which one would eventually be eliminated. Some segregated junctions (13 out of 90), however, did show characteristics that may be related to the final stages of synaptic competition [the relatively low incidence of observation of these characteristics may mean that the final events of branch withdrawal are rapid (11)]. At these junctions, one axon was thinner than the other axon (Fig. 4A). In addition, in some junctions, the postsynaptic acetylcholine receptor (AChR) density associated with the thin axon was lower than that associated with the thicker axon (Fig. 4A). In some cases, a thin axon possessing a retraction bulb was located near but not apposed to a very dimly stained AChR site, which presumably was previously occupied by the withdrawing axon (Fig 4B). Thus, axons apparently compete even when separated by some distance, resulting in the disassembly of both the pre- and postsynaptic sites.

The studies reported here provide several insights regarding synaptic competition. First, even within the confines of a neuromuscular junction that is ∼10 to 50 μm long, some branches of a single axon are maintained, whereas others are lost. Thus, the elimination of different branches of an axon is regulated independently at the level of each individual branch. Second, departing axonal branches undergo a stereotyped process of atrophy, detachment, and retraction within a neuromuscular junction. This program of branch removal seems analogous to the final stage of synapse elimination, when the losing axon terminates in a retraction bulb as it withdraws from the end plate (compare Figs. 3 and 4) (7, 10). Therefore, the process of branch removal may illuminate the entire process of axon withdrawal. Finally, the withdrawal process is spatially regulated so that the branches of an axon that are nearest to the competitor's territory undergo retraction but branches of the same axon several micrometers more distant from the competitor's territory appear unaffected. It is this pattern of distance-dependent loss that leads to the segregation of competing axons. The local regulation of axon branch removal does not support the idea that signals that maintain their strength over distance (such as action potentials) are mediators of synaptic competition (12).

Axons may compete by generating activity-mediated signals that destabilize synaptic sites associated with other inputs (13). Our results indicate that such destabilizing signals would have to be spatially regulated and decrease in potency over distances of <50 μm. The nature of the spatially localized signals in this process is not known. Experiments in Xenopus nerve muscle cocultures suggest that activation at one synaptic site can decrease synaptic efficacy at nearby sites that are occupied by different inputs through a steeply decreasing intracellular Ca2+ signal (14). It is therefore possible that a diffusible signal cascade triggered by local calcium entry [for example, through activated AChRs or other Ca2+ channels (15)] could initiate (16), or itself be, the spatially decreasing signal at developing neuromuscular junctions. Although such a spatially decreasing signal could in principle explain segregation, it would not necessarily result in the elimination of an input when it had obtained a sufficient distance from the competitor. Therefore, we presume that the strength or extent of the short-range destabilizing signals increases during synaptic maturation [as synaptic efficacy is increasing (11)] or that the destabilization process is itself incremental (taking longer to remove distant synapses than nearby ones), or that both processes occur. Because the decreasing signals may become totally ineffective when the competitors are sufficiently separated (17), the normal confinement of all the incoming axons to a small AChR plaque may thus be a strategy that encourages strong competition among nearby synapses, which rapidly results in single innervation at developing neuromuscular junctions. Given the existence of competitive synaptic reorganization on neurons (18) and the evidence for the restriction of axonal innervation to parts of a dendritic arbor (19), it seems likely that analogous short-range signals operate throughout the nervous system.

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


Stay Connected to Science

Navigate This Article