Figure 1.
Time course of removal of laser-target input and takeover by another at developing neuromuscular junction.
(A) In vivo image of region in a P7 mouse sternomastoid muscle. Boxed area shows a neuromuscular junction innervated by two axons (green, YFP-filled axons; red, α-bungarotoxin-tagged acetylcholine receptors). The axon on the right in the boxed region was irradiated with a mode-locked infrared laser at the site of the circle. (B) The same region of muscle reimaged 27 h later showing that the irradiated axon has completely disappeared. (C) Time-lapse imaging reveals that the remaining axon grows to occupy the sites that were vacated by the damaged axon (arrow points to site of irradiation). Within 0.5 h of irradiation the right axon's terminal branches swell and clearly reveal that the undamaged axon, that remained brightly fluorescent, terminated in a bulb. The swelling of the damaged axon deformed the shape of the intact axon's terminal bulb (panels at 0.8, 1, and 1.5 h). At 1.8 h, the bulb recovered its original shape, presumably because the damaged input had lost its turgor presumably due to membrane leakage. Over the next hour, the fluorescence in the damaged terminals became fainter, its axon fragmented (see 1.8 h), and it largely became invisible. Reimaging the junction 27 h after damage showed that the remaining axon branched to occupy many of the sites previously occupied by the laser irradiated axon. Scale bar, 20 µm.
Figure 2.
Laser targeted axons die-back to a proximal branch point.
(A) To understand the consequences of laser axotomy we irradiated YFP expressing axons at multiply innervated neuromuscular junctions such as the green tinted axon here in a P7 sternomastoid muscle of a living mouse (circle shows site of irradiation). Note that the proximal terminal branch point is visible (arrow). Another axon (grey) with smaller caliber also innervated this junction (and a third recently retracted axon with a bulb ending can also be seen). (B) Six hours after laser irradiation the distal end of the damaged axon swelled, its intensity became fainter, and fragmentation was visible back to the point from which the terminal axon branched off of the parent axon. The relatively small area occupied by the remaining axon is visible at the right edge of the neuromuscular junction site and appears largely unchanged at this time (red, α-bungarotoxin-tagged acetylcholine receptors). Scale bar, 10 µm.
Figure 3.
Small axonal inputs take over developing neuromuscular junctions after laser-targeted removal of larger inputs.
(A, B) We identified doubly innervated neuromuscular junctions in sternomastoid muscles of P7-P8 mice in vivo, which were innervated by both a small caliber and a large caliber axonal input. Previous studies have shown that in most cases the small caliber input ultimately is eliminated from such junctions. In the cases shown here, laser irradiation of the larger caliber inputs (circles and arrows mark the sites of laser-induced damage) caused (A) their swelling or (B) rapid disappearance within 1 h of laser irradiation. After the damage, the synaptic area occupied by the un-irradiated smaller input could be clearly seen: (a) ∼20% of the total postsynaptic territory (lower left portion of junction) and (b) ∼5% of the area (lower middle portion of junction). The animals were allowed to recover from surgery and then re-anesthetized the next day and the neuromuscular junctions relocated. In each case, the small remaining input grew to occupy the entire neuromuscular junction at some time prior to 24 h. In addition in both cases the caliber of the axon entering the junction enlarged. The axons were imaged in singly and doubly transgenic mice that had constitutive cytoplasmic expression of (A) YFP (green) and (B) both CFP (pseudo-colored green) and YFP (red), respectively. In the latter there was a greater proportion of CFP in the smaller axon, making it appear greener (in the first panel) and thus possible to distinguish its territory from the sites occupied by the other axon even before laser axotomy. Scale bar, 20 µm.
Table 1.
Synaptic takeover at developing neuromuscular junctions following laser-induced damage of an axonal input.
Figure 4.
Previously retracted axons reinnervate their former neuromuscular junctions after laser-targeted removal of the innervating axon.
(A) Shown is a singly innervated neuromuscular junction with a nearby recently retracted axon (tipped by a bulb) in the sternomastoid muscle of P7 mouse in vivo. The pulsed laser irradiated the axon innervating the neuromuscular junction at the circle marked by a large arrow. After 1 h, the irradiated axon was mostly invisible, leaving the bulb-tipped retracting input unchanged. By the next day (1 d), the retracting input had reversed direction and nearly completely reinnervated the junction. After 2 d (2 d) the caliber of the reinnervating axon and its terminal branches increased. In this experiment it was unambiguous that the regenerated axon was the former withdrawing axon because the two axons could be distinguished by their relative concentrations of YFP and CFP. The bulb-tipped axon could also be re-identified after it reinnervated the junction because of the location of its proximal branch point (indicated by the small arrow in each image). Scale bar, 20 µm. (B) Graph showing the incidence of return of retracting axons as a function of its distance from the junction at the time of laser axotomy of the innervating axon (also see Table 1B). The percentages were computed relative to the total number of retracting axons studied. Retracting axons 10 µm or closer appeared more likely to return than ones farther away.
Figure 5.
Synaptic withdrawal precedes takeover during naturally occurring synapse elimination.
Shown are four views of a multiply innervated neuromuscular junction from the sternomastoid muscle of a living mouse viewed over 2 d starting at P10. AChRs were lightly labeled with fluorescently tagged α-bungarotoxin (pseudo-colored blue). Both axonal inputs contain CFP (pseudo-colored green) and YFP (red) but in different proportions, allowing one input (red+green = yellow) to be distinguished from the other (orange). At the first view (0 h) the orange-colored axon occupied the lower part of the junction and a small branch at the top. Insets below show axons (left) and fluorescently tagged AChRs (right, in gray) of the bottom region of the junction. To more clearly see the extent of the territory occupied by the yellow axon, the fluorescent protein in orange-colored input was bleached at the nerve entry zone for several minutes without damaging the axon (see Figure S2) using visible continuous wave laser light. A day later (22 h) the orange-colored input is no longer at the junction, but a remnant of the retracting axon is visible (arrow). Several sites that had previously been occupied by the orange axon are now vacant (dashed ellipse). However, by 45 h, the remaining input grew into the vacated territory. The delay between the withdrawal of the orange axon and the takeover by the yellow one is similar to time course observed following removal of an input by laser irradiation (approximately 1 d). Scale bar, 10 µm.
Figure 6.
Small synaptic vacancies within innervated neuromuscular junctions induce axon growth.
Singly innervated neuromuscular junctions were imaged in vivo in the sternomastoid muscle of adult transgenic mice expressing green fluorescent protein (GFP) in all motor neurons. After pulsed laser-targeted irradiation (circles and arrows indicate sites of laser-induced damage) the nerve terminals were re-imaged at 1 h to confirm damage to the targeted branches. The size of the denervated synaptic area relative to the total area of the junction was approximately (A) 30%, (B) 10%, and (C) less than 5%. AChRs of each neuromuscular junction are shown in the insets of the rightmost panels. The same nerve terminals were imaged again 2–10 d later. In (A) two branches of the same axon innervate neighboring regions of the neuromuscular junction (asterisk). Reinnervation of the bottom portion was complete within 2 d of laser irradiation. In (B) an axon branch that innervated the upper part of the junction was irradiated. Reinnervation of the unoccupied receptor sites was complete at some point between 3 d and 10 d after laser axotomy. In (C) the terminal boutons overlying two small synaptic sites were removed via laser irradiation and then became reinnervated. Given the large safety factor for muscle activation it is unlikely that the small regions that were denervated had any effect on muscle fiber activity; nonetheless, the nerve grew to reoccupy these sites. Scale bars, 20 µm.
Figure 7.
Model of synaptic competition based on piecewise withdrawal from synaptic sites.
Inspired by models of population dynamics of species competing to occupy the same niche, we produced a simple graphical model of neuromuscular synaptic competition that gives rise to many of the features that have previously been observed (see Materials and Methods for model details). (A) We represent a junction by a set of synaptic sites that, in this simulation, are randomly distributed among six innervating axons each with a different color (see left panel). Based on the findings in this article showing that synaptic vacancies can induce nearby axonal branches of either the same or a different axon to grow, we simulated synaptic competition as an iterative process of axon withdrawal from a randomly selected site followed by takeover of the vacated site (see also Video S1). As shown in the first two panels the process starts with the pink axon losing a synaptic contact. The vacated site is subsequently reoccupied by one of the immediately neighboring axons (in this case the dark blue one; see right panel). (B) This stepwise process is repeated approximately 1,000 times until all the sites become innervated by the same axon and the junction is in its mature singly innervated state. Over the course of this simulation, the synaptic contacts of each axon become progressively clustered, and eventually, when the junction has only two remaining inputs, the axons become completely segregated as has been seen in normal development [29]. (C) Graph showing that in this model there is a rapid loss of inputs in the first postnatal week (when many axons converge on each junction) compared to the second postnatal week (when junctions are innervated by at most two axons). This result is consistent with recent experimental observations [45]. (D) The model also shows that an axon's territory does not necessarily change in a monotonic way and that the final outcome cannot be predicted with certainty by the relative amounts of territories that each axon occupies. Notice, for example, that the axon with the least amount of territory at time 0 (the light blue axon) is the ultimate winner. These results are consistent with the “flip flop” described and observed in vivo [4].