To characterize the influence of nonautonomous tissues on axon degeneration and regeneration, we developed a reproducible method for severing the zebrafish pLL nerve in vivo and monitoring its behavior with time-lapse imaging. Imaging allowed us to visualize disconnected pLL axon fragments undergoing Wallerian degeneration and to monitor the entire process of axon regeneration, from injury to the reestablishment of pLL functionality. This system thus offers an unprecedented opportunity for analyzing the regulation of axon degeneration and regeneration in vivo and dissecting the roles of cellular and molecular regulators of these processes.
WD in the zebrafish pLL was extremely fast compared to axon degeneration in many previously characterized models. For example, the lag phase in the dorsal root ganglion nerve of the mouse lasts one to one and a half days after axotomy, a phenomenon that in the pLL of zebrafish larvae takes approximately three hours, similar to what was recently described for spinal motor neurons. The elimination of axonal fragments took an average of around five hours in the pLL and twenty-four hours in spinal motor neurons. Both of these phases are slightly but reproducibly faster in zebrafish trigeminal axons. This observation indicates that WD in peripheral axons differ even in the same animal, suggesting either intrinsic or extrinsic differences between them. The fact that we were able to assess the success of WD just five to six hours after axotomy raises the possibility that medium-to-high throughput strategies for drug assays or genetic screens may be a feasible method for dissecting WD in this system. Imaging fragmentation along pLL axons (which can be longer than 1000 μm) demonstrated that WD proceeds from the distal end of an axon fragment to the proximal end, even though in a given section of a detached axon, fragmentation can appear to be synchronous. Surprisingly, this retrograde progression more closely resembles the outcome of a crush injury to the mouse sciatic nerve than the anterograde progression seen when the sciatic nerve is severed. The signals mediating propagation along a severed axon fragment in one direction or the other are still unknown, but our results provide an additional model for investigating retrograde progression.
Zebrafish larvae can regenerate many tissues and structures, including the pLL nerve. By performing time-lapse imaging of regenerating axons, we observed new growth cones arising from the proximal axon stump immediately after degeneration of the distal fragment. In both pLL and trigeminal axons, regeneration starts only after debris from the original axon is cleared. It is still not clear whether regeneration is dependent on fragment removal per se since, in trigeminal neurons, regeneration occurs on schedule when axon degeneration is delayed by WldS expression. Nonetheless, these neurons seem to have evolved so that degeneration and regeneration occur in orderly succession.
Regeneration of the pLL nerve followed its original pathway, similar to what has been described for regeneration of zebrafish spinal motor neurons. Analyzing regeneration of single axons labeled in transient transgenic larvae, we verified that reconstitution of a functional nerve corresponded to bona fide axon regeneration, rather than axon growth arising from new neurons. Regeneration of pLL axons qualitatively resembled axon regeneration described in invertebrates and in some other vertebrate axons[60, 61]. Regenerating pLL axons in wild-type animals grew at a speed of approximately 0.5 μm/minute, reaching all of the pLL neuromasts by 24 hours postaxotomy. These axons thus regenerated much faster than regenerating axons in the mouse central and peripheral nervous systems, where axons grow at 0.2 μm/minute and 0.07 μm/minute, respectively.
Axon fragments are known to be phagocytosed by at least two cell types: Schwann cells and immune cells. Ample evidence indicates that in the mouse PNS macrophages and glial cells jointly regulate the clearance of axons and myelin[13–16]. We found that in the absence of innate immune cells, pLL axon fragments persisted for significantly longer than in controls (Figure2). Elimination of each cell type separately suggests that leukocytes are the main participants in the removal of axonal debris. Our findings are complementary to those of Rosenberg et al., whose work suggested that debris clearance after motor axon transection is mediated by macrophages, and that Schwann cells are dispensable for this process. We also suggest that at least one other cell type must have the capacity to remove axonal debris because in the absence of both leukocytes and Schwann cells, axon fragments disappeared on schedule. Epidermal skin cells could fulfill this role as they associate closely with the pioneering pLL axons during their development and, in the absence of Schwann cells, pLL axons remain within the epidermis.
Interestingly, we found that Schwann cells are essential for proper navigation of the regenerating pLL nerve. In the absence of these cells, axons grew erratically, even though they extended at normal speeds. In the mammalian inner ear, several studies indicate that hair cells secrete factors that induce axon growth of acoustic ganglion neurons after injury (reviewed in[22, 26]). Moreover, after Wallerian degeneration, the endoneurium that had surrounded the distal axon remains, leaving an endoneurial tube around which Schwann cells proliferate, forming bands of Büngner[61, 62] that serve as a structural guide by which new axons navigate. In addition to this structural role, after nerve injury Schwann cells express neurotrophic factors, which could contribute to axon guidance. The erratic growth observed in regenerating pLL axons in the absence of Schwann cells, may thus be explained by the absence of a structural guide for regeneration or the lack of appropriate guidance molecules.
A recent report identified a fundamental role for glial cell line-derived growth factor (GDNF) in axonal guidance during development and regeneration of the pLL nerve. In contrast with our results, that study found that Schwann cell absence did not affect regeneration of the pLL nerve, but that regeneration was dependent on GDNF production by interneuromastic cells. However, in that report, only a subset of Schwann cells were ablated (specifically, Schwann cells between neuromasts I and II). It is possible that some Schwann cells escaped ablation (for example, by dedifferentiating and losing GFP transgene expression) or that intact Schwann cells located beyond the second neuromast continued to exert an effect on the regenerating nerve.
Our study also addressed whether neuromast hair cells, the targets of innervation by the pLL nerve, are necessary for proper regeneration or pathfinding of axons in the pLL. Chemical elimination of neuromast cells (using copper toxicity) resulted in erratic growth of regenerating pLL axons, but axons reestablished the correct pathway upon regeneration of the neuromasts. This result suggests that the target cell environment influences the guidance of regenerating axons. However, when hair cells were eliminated genetically or physically with a laser, pLL regeneration trajectory and velocity were both normal, suggesting that hair cells themselves are not required for this process. A possible explanation for these contradictory results is that although copper treatment selectively damages neuromasts, without affecting surrounding glial cells, axons, or other tissues in the pLL, it can cause damage to multiple cell types in the neuromast, such as supporting cells and mantle cells, in addition to hair cells. These other cell types in the neuromast were not affected by laser or genetic ablation. Thus, it is possible that the aberrant behavior of regenerating pLL neurites observed after copper treatment is due to neuromast accessory cell elimination. We cannot, however, exclude the possibility that copper may have an unexpected effect on the regenerating axons themselves or on the interstitial space through which they navigate. It will be important to use alternative methods of ablating neuromasts to further test their role in pLL axon regeneration.
While we did not find that innervation was required for the survival of most hair cells, it is possible that viability of these mechanosensory cells would be impaired after long-term denervation by afferent fibers, as is clearly the case in the mammalian inner ear. We observed occasional hair cell death after axotomy, but most dying cells identified by TUNEL staining were located among accessory cells in denervated neuromasts. While it was known that pLL neuromast hair cells can develop and differentiate in the absence of a pLL nerve[47, 48], this is the first study to examine their viability when denervation occurs postembryonically, after hair cells have developed and integrated normally into the lateral line. Nonetheless, reinnervation occurred within 24 hours after injury, due to the robust regeneration of the proximal axon, indicating that the hair cells are deprived of afferents for only a short period. It would be interesting to permanently block axonal regrowth and assess the survival of hair cells in the denervated state.