Axon Regeneration Requires a Conserved MAP Kinase Pathway

See allHide authors and affiliations

Science  06 Feb 2009:
Vol. 323, Issue 5915, pp. 802-806
DOI: 10.1126/science.1165527


Regeneration of injured neurons can restore function, but most neurons regenerate poorly or not at all. The failure to regenerate in some cases is due to a lack of activation of cell-intrinsic regeneration pathways. These pathways might be targeted for the development of therapies that can restore neuron function after injury or disease. Here, we show that the DLK-1 mitogen-activated protein (MAP) kinase pathway is essential for regeneration in Caenorhabditis elegans motor neurons. Loss of this pathway eliminates regeneration, whereas activating it improves regeneration. Further, these proteins also regulate the later step of growth cone migration. We conclude that after axon injury, activation of this MAP kinase cascade is required to switch the mature neuron from an aplastic state to a state capable of growth.

Severed neurons can regenerate. After axons are cut, neurons can extend a new growth cone from the axon stump and can attempt to regrow a normal process. Most invertebrate neurons are able to regenerate, as are neurons in the mammalian peripheral nervous system. By contrast, neurons in the mammalian central nervous system have limited regenerative capability (1). Regeneration is thought to be initiated by signals arising from the injury, including calcium spikes and the retrograde transport and nuclear import of regeneration factors (2). These mechanisms lead to increased cyclic adenosine monophosphate (cAMP) levels, local and somatic protein synthesis, and changes in gene transcription that, in turn, promote remodeling of the cytoskeleton and plasma membrane at the site of injury. The ability of specific neurons to regenerate is determined in part by the balance between proregeneration signals and cellular pathways that inhibit regeneration. For example, regeneration in the mammalian CNS is inhibited by extrinsic signals from myelin and chondroitin sulfate proteoglycans; these signals activate pathways in the damaged neuron that prevent regrowth (3). But CNS regeneration can be achieved even in the presence of inhibitory signals. A conditioning lesion to a peripheral process results in increased regeneration of the CNS branch of dorsal root ganglion neurons, presumably by triggering injury signals that result in an overall increase in regenerative potential (4). Thus, intrinsic regeneration signals can influence regenerative success, and these signaling processes represent potential targets for therapies to enhance regeneration.

We identified the mitogen-activated protein (MAP) kinase kinase kinase (MAPKKK) dlk-1 as essential for regeneration in the course of a large screen for genes required for regeneration. This screen was conducted in a β-spectrin mutant (unc-70) background (5). Neurons in β-spectrin mutant nematodes break because of mechanical strain induced by locomotion. The γ-aminobutyric acid (GABA)–releasing motor neurons respond to breaks by regenerating toward their targets in the dorsal cord (6). Axon guidance during regeneration is imperfect, resulting in axons in mature animals with branching and other abnormalities (Fig. 1). We found that RNA interference of dlk-1 eliminates regeneration in unc-70 mutant animals. Neither unc-70 (6) nor dlk-1 (7) is essential for axon outgrowth during development of the GABA neurons (Fig. 1) (8). The unc-70 dlk-1 synthetic phenotype for axon morphology suggests that dlk-1 may function specifically in regeneration.

Fig. 1.

dlk-1 is required for axon regeneration in β-spectrin mutant nematodes. (A) Cartoon showing development of axon morphology in control β-spectrin mutant (left) and in β-spectrin mutant lacking hypothetical regeneration gene, e.g., dlk-1 (right). (B and C) GABA neurons in representative L4 stage β-spectrin mutants (unc-70), expressing green fluorescent protein, under control conditions or after dlk-1 RNA interference. Scale bars, 20 μm. (D) High-magnification view of boxed region in (C). Arrows indicate inert axon stumps. Scale bar, 10 μm. (E and F) GABA neurons in representative L4 stage wild-type and dlk-1 mutant animals. Scale bars, 20 μm.

To demonstrate that dlk-1 functions in regeneration independently of unc-70, we used laser axotomy to trigger regeneration. The GABA motor neurons can regenerate after laser axotomy (9). We used a pulsed 440-nm laser to cut axons (10). In larval stage 4 (L4) wild-type animals, 70% of severed axons initiated growth cones within 24 hours after axotomy (Fig. 2 and table S1). But when axons were cut in L4 stage dlk-1(ju476) null mutants, growth cones were never observed. These severed neurons appeared healthy after surgery; both the stump of the remaining axon and its cell body showed no decrease in green fluorescent protein (GFP) expression or other signs of injury. Nevertheless, these neurons failed to regenerate. To test whether regeneration was merely delayed, we monitored some severed axons for 5 days; regeneration still was not observed. Thus, dlk-1 is essential for axon regeneration after spontaneous breaks and after laser surgery but is dispensable for axon outgrowth during development.

Fig. 2.

dlk-1 is required in severed axons for growth cone initiation. (A) Regenerating axons 18 to 20 hours after laser surgery in a wild-type animal. Both severed axons have generated a growth cone (arrows). Scale bar, 10 μm. (B) Axons in dlk-1 mutants fail to generate growth cones 18 to 24 hours after surgery. Scale bar, 10 μm. (C) DLK-1 acts cell intrinsically to mediate regeneration. (D) RPM-1 controls DLK-1 activity in axon regeneration. (E) Regeneration requires dlk-1 at all ages and overexpression of DLK-1 rescues age-associated decline. DD and VD are motor neurons of different lineages (8). (F) DLK-1 acts at the time of injury to mediate regeneration. Time of heat shock relative to surgery is indicated in hours. No heat shock is indicated by “no.” “L2” indicates surgery at the L2 stage. Axotomy at L4 stage unless indicated otherwise. (C) to (F) Percentage of axons that initiated regeneration and 95% confidence interval (CI). *P < 0.05, **P < 0.01, ***P < 0.001.

Mosaic experiments demonstrate that the DLK-1 protein acts in the damaged cell rather than in the surrounding tissue. To determine whether DLK-1 acts cell-autonomously to promote regeneration, we expressed dlk-1 under the GABA-specific promoter Punc-47 in the dlk-1 null background. Neurons were severed by laser surgery, and regeneration was assayed after 18 to 24 hours. We found that expressing DLK-1 in the GABA neurons restored regeneration to dlk-1 null mutants (Fig. 2C). Further, mutations that affect DLK-1 levels also affect regeneration (Fig. 2D). In Caenorhabditis elegans, DLK-1 levels are negatively regulated by RPM-1 (7). Overexpression of RPM-1 reduced regeneration after surgery to levels similar to those of dlk-1 loss-of-function mutants. Conversely, initiation of regeneration was enhanced in rpm-1 mutant animals. Initiation of regeneration was also enhanced in animals lacking FSN-1, an F-box protein that functions with RPM-1 to promote DLK-1 degradation (11). However, loss of GLO-1 (Rab) or GLO-4 (Rab guanine nucleotide exchange factor), both of which mediate ubiquitin-independent functions of RPM-1 (11), did not have strong effects on regeneration. These results confirm that changes in DLK-1 protein abundance can determine regenerative ability.

Although regeneration is age-dependent, dlk-1 is required at all stages, and overexpression of DLK-1 can rescue some age-dependent decline (Fig. 2E). We analyzed regeneration at various developmental stages. We found that regeneration declines significantly with age, and only a few axons in old adults regenerated (10). Despite these differences, dlk-1 is required for all regeneration, even in very young animals. Because growth cones in the very young animals had not yet reached their targets, the dependence on dlk-1 is not correlated with target contact or synaptogenesis. Further, dlk-1 is required for regeneration of both presynaptic and postsynaptic processes (8).

To mediate regeneration, DLK-1 is required at the time of injury (Fig. 2F). We expressed the DLK-1 protein at different times using the heat shock promoter Phsp-16.2. DLK-1 expression at the time of injury was sufficient for regeneration. Applying heat shock hours before or hours after surgery resulted in less regeneration, and when heat shock was applied either 11 hours before or 48 hours after surgery, little or no regeneration was observed. This effect was independent of age, because surgery in L2 stage larvae failed to elicit regeneration when heat shock was applied 48 hours later. Thus, DLK-1 must function within a short temporal window near the time of injury to mediate regeneration, rather than establishing a permissive state for regeneration during development. These data suggest that DLK-1 signaling must coincide with other proregeneration signals, such as calpain activation (12) or cAMP elevation (1315), for regeneration to occur.

DLK-1 is required for growth cone formation rather than the earlier step of filopodial extension. We used time-lapse microscopy to monitor morphological changes in axons after surgery. We found that in wild-type animals, newly severed axons repeatedly extend short, transient filopodia from the axon stump (Fig. 3 and movie S1). The first filopodium appears with an average delay of more than 3 hours. In animals that successfully initiate regeneration, a single filopodium eventually persists and is transformed into a growth cone. Growth cone formation in wild-type animals occurs with an average delay of 7 hours after surgery. In dlk-1 mutants, transient filopodia appear at approximately the same time and the same rate as in the wild type. However, growth cones were never observed in these mutants. These data demonstrate that DLK-1 is required to transform exploratory filopodia into growth cones.

Fig. 3.

dlk-1 controls growth cone initiation and morphology during axon regeneration. (A) Transient filopodium in a wild-type animal. Images were taken at 165 (left), 170 (center), and 180 (right) minutes after surgery. Scale bar, 5 μm. (B) Transient filopodium in a dlk-1 mutant animal. Images were taken at 475 (left), 480 (center), and 490 (right) minutes after surgery. Scale bar, 5 μm. (C) Representative axons in a wild-type animal 120 min after axotomy. Proximal and distal ends have retracted away from site of surgery, but proximal ends (arrows) show no evidence of regeneration. Scale bar, 10 μm. (D) A representative axon in an animal overexpressing DLK-1 120 min after axotomy. The proximal end (arrow) has already regenerated past the retracted distal end. Scale bar, 10 μm. (E) Representative growth cones in a wild-type animal. Although these axons successfully initiated regeneration, the growth cones (arrows) have a dystrophic morphology. Scale bar, 10 μm. (F) Representative growth cones in an animal overexpressing DLK-1 under the unc-47 promoter. These growth cones (arrows) have a compact morphology similar to growth cones observed during development. Scale bar, 10 μm. (G) Distribution of all times of filopodia initiation in wild type and dlk-1. Each dot represents a filopodium. (H) Time of first filopodium initiation in wild type and dlk-1. Means ± SEM. (I) Rate of filopodia initiation in wild type and dlk-1. Means ± SEM. (J) Time to initiate regeneration after surgery in wild type and dlk-1 overexpressing (OE) animals. Initiation is defined as the appearance of the filopodia that becomes a growth cone. Each dot represents a single axon. (K) Percentage of wild-type and dlk-1 overexpressing (OE) regenerating axons that reached the dorsal cord after 18 to 24 hours. Error bars indicate 95% CI.

Increased expression of DLK-1 in wild-type animals accelerates the formation of growth cones and improves migration success. We overexpressed DLK-1 in GABA neurons and found that time of growth cone initiation by axon stumps was advanced relative to the wild type (Fig. 3 and movie S1). Also, more axons initiated growth cones (Fig. 4). In addition to these effects on growth cone initiation, DLK-1 overexpression improved growth cone performance. In wild-type animals, regenerating growth cones often have a branched, dystrophic morphology. Dystrophic growth cones migrate poorly, and most never reach the dorsal nerve cord in 24 hours (Fig. 3). These dystrophic growth cones resemble dystrophic growth cones observed in failed regeneration in the mammalian CNS (16). By contrast, regenerating growth cones in neurons that overexpress DLK-1 have a compact shape, similar to growth cones observed during initial axon development (17). These compact growth cones were much more likely to reach the dorsal nerve cord. Growth cone migration during regeneration in C. elegans shares some genetic requirements with developmental axon guidance, including components of the netrin and slit signaling pathways (18) and the ephrin pathway (10). It is possible that DLK-1 overexpression improves regeneration by affecting the response of the growth cone to such signals. Thus, DLK-1 acts at two steps of regeneration: It is required for growth cone formation, and it also controls growth cone morphology and behavior.

Fig. 4.

MAP kinase signaling is required for axon regeneration. (A) Regeneration is eliminated by mutations in the DLK-1/MKK-4/PMK-3 MAP kinase module. (B) Other MAP kinase elements contribute to regeneration, but are not essential. (C) Activated DLK-1 has targets in addition to MKK-4 and PMK-3. (D) Model for function of MAP kinase signaling during axon regeneration. (A) to (C) Error bars indicate 95% CI.

DLK-1 functions in a MAP kinase signaling cascade that also includes the MAP kinase kinase (MAPKK) MKK-4, and the p38 MAP kinase PMK-3 (7). We tested whether this entire MAP kinase signaling module functions in regeneration by examining null mutants in mkk-4 and pmk-3. Like dlk-1, neither of these mutants has appreciable defects in axon outgrowth during development. But after axotomy, both mutant strains fail to initiate regeneration (Fig. 4A). These data suggest that MKK-4 and PMK-3 are the downstream targets of DLK-1 for regeneration. Inhibition of p38 also reduces regeneration of cultured vertebrate neurons (19), which suggests that the function of p38 MAP kinases in regeneration is conserved. Do other MAP kinase cascades also contribute to regeneration? We tested a sampling of C. elegans MAP kinase components and found that mutations in these genes did not eliminate regeneration (Fig. 4B and table S1). Initiation of regeneration was not affected by loss of the MAPKKK nsy-1 or its target MAPKK sek-1. Loss of the MAPKK jkk-1 also did not affect regeneration. By contrast, loss of the MAPKKK mlk-1 reduced initiation of regeneration (although some regeneration still occurred), as did loss of its downstream target mek-1. MLK-1 and MEK-1 are thought to activate a second C. elegans p38 MAP kinase, PMK-1 (20), which suggests that multiple p38 family members contribute to regeneration. (Because null mutations in pmk-1 are lethal, we were unable to test its function directly.) Loss of the MAP kinase jnk-1 increased initiation of regeneration. Thus, whereas the DLK-1/MKK-4/PMK-3 MAP kinase cascade is required to initiate regeneration, other MAP kinase pathways also regulate this process. Consistent with these data, mutations in mkk-4 or pmk-3 did not eliminate the stimulation of regeneration by DLK-1 overexpression, which suggests that cross-talk between MAP kinase modules may contribute to regeneration (Fig. 4C). However, the modest phenotype of other MAP kinase mutants and the inability of DLK-1 overexpression to bypass the requirement for mkk-4 and pmk-3 suggest that the DLK-1/MKK-4/PMK-3 module is the major MAP kinase pathway for axon regeneration.

What stimulates DLK-1 function when an axon breaks? In the simplest model, the axon break interrupts trafficking of DLK-1. Local DLK-1 accumulation in the injured neuron then leads to homodimerization and activation, followed by activation of the downstream targets MKK-4 and PMK-3 (Fig. 4D). Alternatively, specific regulatory mechanisms activate DLK-1 after injury, such as scaffolding proteins like Jip1 (21); phosphatases such as PP1, PP2a, and calcineurin (22); or regulators of the proteasome (23). The strict requirement for the DLK-1 pathway in regeneration suggests that mature neurons have intrinsic barriers to growth that are not present during development. Once axons have reached their target and have begun synaptogenesis, termination of growth signals by mechanisms like RPM-1 may down-regulate growth to allow synapse maturation and to stabilize neuronal architecture. Indeed, mutations in RPM-1 or its homologs cause overgrowth of axons in worms (24), aberrant sprouting in Drosophila (25), and aberrant growth cone initiation on axon shafts in mouse (26). We found that dlk-1 is required for regeneration even in neurons that are actively growing at the time of injury. Further, overexpressing DLK-1 partially prevented the loss of regeneration in old animals (Fig. 2E). Thus, barriers to growth are quickly erected in axons, and post-injury signaling via DLK-1, MKK-4, and PMK-3 is required to drive the neuron back to its prelapsarian, embryonic state.

How does the MAP kinase PMK-3 stimulate regeneration? The DLK-1 pathway is first required for growth cone formation about 7 hours after a break occurs—a process likely to be mediated by the polymerization of microtubules. Activated p38 MAP kinase regulates microtubule dynamics (26), and microtubule remodeling is required for growth cone initiation during regeneration (27). Further, defects in microtubule dynamics contribute to the axon outgrowth phenotype of Phr1 mutant mice (26). Activated p38 may also control other targets that facilitate axon regeneration. p38 regulates local protein synthesis (28), which is required for regeneration (19). p38 is also likely to have functions in the nucleus, because it contributes to injury-induced changes in gene transcription (29). Activated p38 may reach the nucleus by retrograde transport. Retrograde transport in general is critical for regeneration (30), and transport of activated MAP kinases from axons to the cell body following axotomy has been observed in Aplysia sensory neurons (31) and in rodent sciatic nerves (32, 33). Thus, regeneration may require activated PMK-3/p38 at the site of the break to regulate microtubule stability and protein expression and also may require PMK-3 to traffic to the nucleus to regulate gene transcription (Fig. 4D). The DLK-1 signaling pathway thus provides a critical link between axon injury and the process of regeneration.

Supporting Online Material

Materials and Methods

Table S1


Movie S1

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article