Scientific Understanding of Consciousness
Consciousness as an Emergent Property of Thalamocortical Activity

Axonal Remodeling by Visual Stimulation


Science 23 May 2014:  Vol. 344  no. 6186  pp. 904-909 

Rapid Hebbian axonal remodeling mediated by visual stimulation

Martin Munz,

Department of Neurology and Neurosurgery, Montreal Neurological Institute, McGill University, 3801 Rue University, Montreal, QC H3A 2B4, Canada.

Geisel School of Medicine, Dartmouth College, Hanover, NH 03755, USA.

Brain Research Centre, University of British Columbia (UBC), 2211 Westbrook Mall, Vancouver, BC V6T 2B5, Canada.


We examined how correlated firing controls axon remodeling, using in vivo time-lapse imaging and electrophysiological analysis of individual retinal ganglion cell (RGC) axons that were visually stimulated either synchronously or asynchronously relative to neighboring inputs in the Xenopus laevis optic tectum. RGCs stimulated out of synchrony rapidly lost the ability to drive tectal postsynaptic partners while their axons grew and added many new branches. In contrast, synchronously activated RGCs produced fewer new branches, but these were more stable. The effects of synchronous activation were prevented by the inhibition of neurotransmitter release and N-methyl-d-aspartate receptor (NMDAR) blockade, which is consistent with a role for synaptic NMDAR activation in the stabilization of axonal branches and suppression of further exploratory branch addition.

Neuronal activity and molecular cues cooperate to form precise neuronal circuits. Experimental blockade of action potential firing or synaptic transmission particularly involving N-methyl-d-aspartate receptors (NMDARs) degrades axonal projections in the developing nervous system. The precise pattern of neuronal firing is believed to be important for instructing connection refinement because disrupting the temporal correlation of firing between neighboring neurons, even while sparing overall activity levels, results in axons with diffuse terminal arbors. Hebbian plasticity, an appealing model for activity-dependent refinement of circuits, posits that synapses may be strengthened or stabilized when the presynaptic cell participates in making its postsynaptic partner fire. Convergent inputs firing synchronously would cooperatively excite the postsynaptic neuron to fire. Thus, Hebbian plasticity in principle should aggregate coactive inputs, effectively leading to circuit refinement. The detailed mechanisms by which such remodeling actually occurs remain poorly understood.

The developing retinotectal system of the albino Xenopus laevis tadpole is amenable both to live imaging and in vivo electrophysiological characterization. RGC axons in Xenopus normally project to the contralateral optic tectum. Occasionally, a single mistargeted ipsilaterally projecting retinal ganglion cell (RGC) axon is observed. Using post mortem intraocular 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (DiI) injections to label all RGCs in stage-46 to -48 tadpoles when the retinotectal projection is established but still refining (16), we detected no ipsilateral RGC axon in the majority of cases (61%). However, animals with one (21%), two (9%), or more (9%) ipsilaterally projecting axons occasionally were observed. Ipsilaterally projecting axons are thus unlikely to represent a specific class of RGCs but rather reflect random pathfinding errors at the optic chiasm. We exploited the fact that the lone ipsilateral and surrounding contralateral RGCs could be independently visually stimulated in order to test the role of correlated activity on synaptic maintenance and axonal refinement.

These experiments demonstrate how correlated neural activity helps orchestrate the morphological remodeling of developing axons into precisely organized maps. Sensory stimulation promotes exploratory branching and outgrowth of RGC axons within their target structure. Axons that may have extended into inappropriate territory, where their firing patterns do not match those of nearby inputs, would fail to maintain stable functional and structural contacts and continue actively elaborating in search of appropriate partners. In contrast, axons that form synaptic contacts onto partners that receive other highly coactive inputs will engage cooperative mechanisms to stabilize those contacts and the branches on which they reside. Our experiments confirm that glutamate release and activation of NMDARs are critical steps in initiating a branch-stabilizing signal, although its molecular identity remains unknown. Because a single axon firing out of synchrony with numerous synchronized inputs does not appear to benefit from the stabilization signals that these coactive axons presumably receive, the stabilizing signal must be very precisely spatially or temporally restricted, which rules out long-lived, highly diffusible molecules as plausible candidates. The idea that this signal likely originates in the postsynaptic cell is supported by earlier experiments that demonstrated that calcium/calmodulin–dependent protein kinase II activity in tectal neurons can retrogradely modulate RGC axon growth. However, our data do not exclude a potential contribution by the surrounding glial cells or the possibility that putative presynaptic NMDARs on RGCs may be involved.

Our observation that visual stimulation drives a rapid increase in branching and growth is consistent with earlier studies in the retinotectal projections of zebrafish and mouse, in which suppression of RGC firing through expression of inward-rectifying potassium channels inhibited the dense elaboration of branches. Similarly, the enlarged arbors reported in zebrafish RGCs expressing TeNT-Lc match our findings that this treatment prevents the down-regulation of branching during correlated activity. These authors argued for a model based on activity-dependent competition to explain their data; however, our results suggest that correlation detection may offer an important alternative explanation for these findings. Activity-dependent competition is useful to explain pathological conditions such as amblyopia but likely plays a minor role in normal map development. A correlated stimulation approach similar to ours was recently reported that used transgenic mice expressing channelrhodopsin-2 in RGCs for stimulation during early postnatal development. The consequences of synchronous and asynchronous stimulation in that system were highly consistent with our findings, with synchronous stimulation leading to ectopic ipsilateral eye projections presumably stabilized in contralateral eye territory.

Here, we present live observations of axonal structural plasticity directed by patterned visual stimuli in vivo that support the Hebbian prediction that coactive inputs are stabilized. Our data also demonstrate the up-regulation of exploratory growth over days in the absence of correlated firing (26) and growth suppression when inputs are coactive. The speed with which physiological visual stimuli can drive such changes—reducing the strength of synaptic currents evoked through the ipsilateral eye after just 10 min of asynchronous visual stimulation and significantly increasing the rate of new branch addition in under 20 min—was unanticipated. Identifying the specific signals that mediate correlation-dependent structural plasticity will be greatly facilitated by exploiting the experimental protocol presented here.

[end of paraphrase]


Return to — Neural Network