The research team, led by Anyi Zhang, Luca Masin, and Lieve Moons of KU Leuven together with collaborators in Paris, used a transgenic zebrafish line that sparsely labels individual retinal ganglion cells with a fluorescent membrane marker, making it possible to trace the full dendritic tree of single neurons in the adult retina. By combining this imaging approach with optic nerve crush injury, quantitative morphometrics, transcriptomic analysis, and pharmacological manipulation, they mapped how dendritic architecture changes as axons regenerate, and identified the molecular switch that governs the process.

The first surprise came from the anatomy. After optic nerve crush, the dendrites of retinal ganglion cells did not remain intact while their axons regrew. Instead, by three days post-injury the cells had lost roughly thirty percent of their total dendritic length and branch points, and by six days, when axons were just beginning to re-enter the brain, the number of bifurcations had fallen to around sixty percent of uninjured levels. Crucially, the overall footprint of the dendritic field was largely preserved, indicating that the remodeling involved the selective elimination of terminal branches rather than wholesale collapse of the arbor. As the axons re-established connections with the optic tectum, the brain’s visual processing center, dendritic complexity rebounded, approaching uninjured architecture by twenty-one days after injury.

This transient simplification was not random degeneration. Gene set enrichment analysis of a bulk RNA sequencing dataset from isolated adult zebrafish retinal ganglion cells showed that transcripts associated with synapses, dendrites, and synaptic transmission were coordinately downregulated between one and six days post-injury, precisely the window in which axons are extending through the optic nerve. The same window saw a sweeping suppression of the insulin-PI3K-Akt-mTOR cascade, the signaling pathway long regarded as a master regulator of cellular growth and a major target of regenerative medicine. Receptor and adapter components such as insra, irs1, and irs2a declined, as did regulatory PI3K subunits and key nodes of the mTOR machinery including rptor, rictor, rheb, and tsc genes.

Because RNA levels are an imperfect proxy for pathway activity, the researchers also examined protein phosphorylation in retinal tissue. They used antibodies against phosphorylated ribosomal protein S6, a well-established readout of mTORC1 activity, the growth-controlling complex downstream of Akt. The results revealed a striking biphasic pattern. In the first two days after injury, the fraction of S6-phosphorylated cells in the ganglion cell layer surged, reaching roughly fifty percent for the Ser235/236 site and up to eighty-five percent for the Ser240/244 site, suggesting an early burst of mTORC1 activity that likely helps initiate the regenerative program. Then, from day three onward, the signal collapsed to near-undetectable levels, coinciding exactly with the onset of dendritic pruning and sustained axonal elongation.

The team also tested whether Gsk3β, a kinase previously implicated in dendrite pruning, was driving the remodeling. Although total Gsk3β protein increased after injury, its inhibitory Ser9 phosphorylation rose even more, indicating a net reduction in kinase activity. The pruning program, in other words, was associated primarily with the repression of mTORC1 signaling rather than with activation of the destructive kinase, sharpening the causal picture.

The decisive experiment came from pharmacology. When the researchers injected recombinant human insulin into the eye on days two through four after injury, they kept the mTOR pathway active during the window when it would normally be suppressed. Insulin treatment did exactly what the model predicted: it partially preserved dendritic complexity, with treated retinal ganglion cells retaining greater dendritic length, area, and branching at six days post-injury. But the preserved dendrites came at a cost. Anterograde tracing with biocytin showed that insulin-treated fish reinnervated their optic tectum roughly fifty percent less than vehicle-treated controls, a clear delay in axonal regrowth. Keeping the dendrites intact, it seemed, actively hindered the rebuilding of the axon.

Follow-up experiments confirmed that this effect runs through mTORC1. Rapamycin, a selective mTORC1 inhibitor, reversed the dendritic preservation caused by insulin, restoring dendritic metrics to vehicle levels. Conversely, MHY1485, a small-molecule mTOR activator, mimicked insulin: it increased dendritic complexity, including more than a fifty percent rise in branch points, and likewise delayed tectal reinnervation by about half. Together, these results establish that mTORC1-dependent maintenance of dendritic branching is a limiting factor for sustained axonal regrowth in the regenerating zebrafish visual system.

Why would a neuron sacrifice its input structures to rebuild its output? The authors propose a resource-allocation model. In mature mammalian neurons, mitochondria become locked near synapses and recycling endosomes are confined to the somatodendritic compartment, whereas regenerating zebrafish retinal ganglion cells transiently relocate mitochondria from dendrites into elongating axons. Insulin receptors sit at the postsynaptic density and mTOR is enriched in dendrites, where it drives local protein synthesis. By pruning dendrites and repressing dendritic mTOR signaling, the neuron may free up protein synthesis capacity, membrane material, and metabolic energy, redirecting these limited resources toward the growing axon. The sequence also echoes development: in both fish and mice, retinal ganglion cells extend their axons first and only elaborate dendrites after the axon has reached its brain target, suggesting that the adult regenerative program recapitulates an ancient developmental order to avoid competition between the two compartments.

The findings carry an uncomfortable implication for mammalian regenerative medicine. Strategies that hyperactivate mTOR, such as PTEN deletion, reliably induce robust axonal regrowth in injured mouse optic nerves, yet extensive functional reinnervation of central brain targets remains largely unachieved. The zebrafish data raise the possibility that sustained mTOR activity may itself hinder the later phases of repair, and that precisely timed, biphasic control of the pathway, an early burst to initiate regeneration followed by deliberate suppression to permit dendritic remodeling and resource reallocation, could be the missing ingredient. Translating that temporal logic to the mammalian central nervous system will be challenging, and the authors caution that their study relied on pharmacological tools rather than genetic perturbation, did not examine contributions from microglia or Müller glia, and leaves open how dendritic pruning affects the electrical activity of regenerating neurons. Still, the message from the fish is clear and provocative: sometimes the best way to rebuild a nerve is to let part of the cell go first.

Subject of Research: Dendritic pruning and mTORC1 signaling regulation during optic nerve regeneration in adult zebrafish retinal ganglion cells

Article Title: Repressed mTORC1 signaling and transient dendritic pruning support axonal regeneration in adult zebrafish

Article References: Zhang, A., Masin, L., Bergmans, S., Putti, E., Albadri, S., Poulain, F. E., Del Bene, F., & Moons, L. (2026). Repressed mTORC1 signaling and transient dendritic pruning support axonal regeneration in adult zebrafish. iScience, 29(11), Article 117826. https://doi.org/10.1016/j.isci.2026.117826

Image Credits: AI Generated

DOI: 10.1016/j.isci.2026.117826

Keywords: zebrafish, optic nerve regeneration, retinal ganglion cells, dendritic pruning, mTORC1, insulin signaling, axonal regrowth, optic tectum reinnervation, neuroregeneration, PI3K-Akt-mTOR, rapamycin, CNS repair