|Home | About | Journals | Submit | Contact Us | Français|
Axon regeneration is essential to restore the nervous system after axon injury. However, the neuronal cell biology that underlies axon regeneration is incompletely understood. Here we use in vivo single-neuron analysis to investigate the relationship between nerve injury, mitochondrial localization, and axon regeneration. Mitochondria translocate into injured axons, so that average mitochondria density increases after injury. Moreover, single-neuron analysis reveals that axons that fail to increase mitochondria have poor regeneration. Experimental alterations to axonal mitochondrial distribution or mitochondrial respiratory chain function result in corresponding changes to regeneration outcomes. Axonal mitochondria are specifically required for growth cone migration, identifying a key energy challenge for injured neurons. Finally, mitochondrial localization to the axon after injury is regulated in part by dual-leucine zipper kinase-1 (DLK-1), a conserved regulator of axon regeneration. These data identify regulation of axonal mitochondria as a new cell biological mechanism that helps determine the regenerative response of injured neurons.
Axon regeneration is a conserved mechanism in which neurons respond to axon injury by attempting to regrow and reconnect to their former postsynaptic targets. The success of axon regeneration is highly variable, even within neighboring neurons injured at the same time. It is also affected by multiple intrinsic and extrinsic factors that coordinately affect the motility, direction, morphology, and extension of the regenerating axons (Case and Tessier-Lavigne, 2005; Kerschensteiner et al., 2005). Within the injured neuron, axon regeneration involves multiple intracellular processes including gene transcription, cytoskeletal rearrangements, and growth cone formation (Bradke et al., 2012; Chisholm, 2013; Goldberg, 2003). Yet our understanding of the cell biology of axon regeneration is still incomplete. In particular, although many axon regeneration processes require energy, the functional relationship between mitochondria, nerve injury, and axon regeneration is not known.
Mitochondria are complex organelles with key functions in energy metabolism, calcium buffering, and signaling. Mitochondria change size and shape through regulated cycles of fission and fusion. In neurons, which have a highly extended morphology, mitochondrial position is regulated to satisfy local demands for energy or Ca2+ buffering, to respond to stressful conditions, and to remove damaged mitochondria (Cai et al., 2012; Chang and Reynolds, 2006; Hollenbeck and Saxton, 2005; Miller and Sheetz, 2004; Verburg and Hollenbeck, 2008; Wang et al., 2011). Mitochondrial subcellular localization and trafficking in neurons is mediated by microtubule-based motor proteins and accessory proteins such as the mitochondrial protein Miro and the adaptor Milton (Hirokawa et al., 2010; Martin et al., 1999; Sheng, 2014; Wang et al., 2011). Functionally, the spatial and temporal regulation of mitochondria positioning in neurons is critical for neuronal development (Morris and Hollenbeck, 1993; Steketee et al., 2012), cortical axon branching (Courchet et al., 2013; Spillane et al., 2013), and presynaptic transmission (Sun et al., 2013). In vivo, mitochondria respond to axon injury by increased trafficking (Mar et al., 2014; Misgeld et al., 2007). In vitro, artificially increasing mitochondrial mobility increases axon growth in cultured cortical neurons (Zhou et al., 2016). Yet the contribution of mitochondrial traffic to regeneration in vivo is not clear, and the mechanisms that regulate the response of mitochondria to injury are incompletely understood.
In this study, we define the role of mitochondria during axon regeneration using the nematode Caenorhabditis elegans (C. elegans). We show that nerve injury results in translocation of mitochondria to the injured axon. Mitochondria density is a critical determinant of regeneration success, in particular for the growth phase of regeneration. Injured axons are in a state of energy stress, and proper mitochondrial respiratory chain function is critical for axon regeneration. Finally, we show that the conserved DLK-1 regeneration pathway promotes mitochondria localization to axons.
To analyze in vivo mitochondria behavior in individual axons after axon injury, we labeled mitochondria with mCherry (Mito::mCherry) in the GABA motor neurons of C. elegans. These neurons are a well-established model for axon regeneration (Byrne et al., 2011; Hammarlund et al., 2009; Yanik et al., 2004), and co-expression of cytosolic GFP allowed for analysis of overall neuronal morphology and for single-neuron laser axotomy (Figures 1A and 1B). In intact neurons, mitochondria were abundant in the cell body, and were also enriched in puncta in the ventral (VNC) and dorsal (DNC) nerve cords likely corresponding to synaptic regions (Figure 1B). These data were consistent with previous results obtained from both fluorescent labeling and from electron microscopy reconstructions (Hall and Altun, 2008; Rawson et al., 2014). By contrast, axon commissures, which are asynaptic, contained few mitochondria. We defined mitochondria density in individual axons as the number of mitochondria puncta per 100 μm of axon, as assessed along the entire commissural axon from the ventral to the dorsal nerve cord. The average mitochondria density was 10.26 ± 2.23 per 100 μm axon length (mean ± 95% confidence intervals, n=31) (Figures 1B and 1C). Mitochondria labeled with a different fluorophore and targeting sequence (mitochondria matrix targeted-GFP) had a similar mitochondrial distribution, indicating that the observed mitochondria density and localization was independent of labeling effects (data not shown).
Next, we determined whether mitochondrial localization in the commissural axons is altered by injury (Figure 1A). We severed axons at the midline between the ventral and dorsal nerve cord with a pulsed laser and imaged the injured axons at 24 hours post injury (Byrne et al., 2011; Hammarlund et al., 2009; Yanik et al., 2004). Compared to intact axons, severed axons had about two-fold increase in mitochondria density (Figures 1D, 1F, and Table S1). Increased mitochondria density in injured axons was also observed 6 hours after injury (Figures 1E, 1G and Table S1). Within the commissures of injured axons, the increase in mitochondria density within the proximal half of the commissural axon (closer to the ventral nerve cord and cell body) was similar to the increase in the distal half of the commissural axon (closer to the axon stump and eventual growth cone) (Figure 1H). The balanced increase in density makes simple spillover from the cell body unlikely, and suggests that axon injury actively triggers mitochondrial localization to axons, rather than stabilization at a particular subcellular location.
To determine whether the increase in mitochondria density depends on the mode of injury, we measured mitochondria density in unc-70/β-spectrin mutants, which have spontaneous axon breakage due to mechanical stress from body bending (Hammarlund et al., 2000; Hammarlund et al., 2007). Mitochondria density was higher in the damaged axons of unc-70/β-spectrin mutants. Further, the increased mitochondria density in unc-70 mutants was suppressed by unc-22/twitchin RNAi, which prevents body bending and axon breakage (Figures 1I-1K and Table S1). Together, these data indicate that neurons respond to axon injury by increasing mitochondria density.
Increased mitochondria density could be a result of changes in mitochondrial translocation. Alternatively, increased density could be a result of increased mitochondria fission or decreased degradation of mitochondria mediated by mitophagy. To distinguish between these possibilities, we analyzed whether axon injury increases mitochondrial translocation into commissural axons from the ventral nerve cord, which contains the neuronal cell bodies and proximal axons. We established in vivo system that allows selective labeling of VNC mitochondria using Mito::Dendra2, and that still allows visualizing axons for single-neuronal laser axotomy. Dendra2 is a photoconvertible fluorescent protein that irreversibly changes its fluorescent state from green to red in response to short wavelength light (405 nm). To test the specificity of our approach, we generated transgenic animals expressing Mito::Dendra2 in GABA neurons (Figure S1A). We selectively photoconverted the Mito-Dendra2 in the ventral nerve cord (VNC). Post-photoconversion, we found that red signal was found exclusively in the VNC (Figures S1A, S1C, S1D, and Table S1). Thus, our photoconversion approach allows us to specifically tag VNC mitochondria. Because gentle photoconversion was necessary to preserve neuronal health, only a portion of VNC mitochondria was marked with red. Next, in order to visualize axons for laser axotomy, we generated transgenic animals expressing both Mito::Dendra2 and cytosolic GFP in GABA neurons. As the green cytosolic GFP masks the green signal of unphotoconverted Mito::Dendra2, only photoconverted (red) Mito::Dendra2 can be observed. To test whether laser axotomy itself (performed at a wavelength of 440 nm) results in unwanted photoconversion, we cut axons and assessed red fluorescence immediately after injury. We found that axotomy resulted in little or no red signal, indicating that it did not significantly photoconvert Mito-Dendra2 (Figures S1B-S1D, and Table S1). Thus, this system allows the tracking of mitochondria deriving from the VNC in the context of laser axotomy.
To determine whether axon injury results in increased translocation into commissural axons from the VNC, we converted VNC mitochondria and imaged 24 hours later. At 24 hours, uninjured axons had significant numbers of red mitochondria in commissures compared to uninjured axons immediately after conversion, indicating that some VNC mitochondria translocate into the axon commissure in intact axons. By contrast, when axons were severed just after photoconversion, at 24 hours the mitochondria density in injured axons was significantly higher than in uninjured axons at 24 hours (Figures 2A-2C and Table S1). These data indicate that axon injury results in increased mitochondria translocation from the VNC to the axon commissure.
As expected, due to the fact that gentle photoconversion results in incomplete mitochondria labeling, the density of commissural mitochondria we observe with the photoconversion approach (both in uninjured and injured axons; Figures 2B and 2C) is lower than the density observed with stable fluorophores under the same conditions (Figure 1F). However, this result also leaves open the possibility that a fraction of the mitochondria increase is due to other mechanisms besides translocation. We tested the idea that reduced mitophagy or increased mitochondrial division could contribute to increased mitochondria density in commissural axons after injury. To determine the effect of mitophagy, we examined axonal mitochondria density in a mitophagy defective mutant. pdr-1 is the C. elegans Parkin homologue and is required for mitophagy (Palikaras et al., 2015). We found that axonal mitochondria density in intact axons of pdr-1 mutant animals was comparable to that of control animals (Figure S2A-S2C and Table S1), suggesting that blocking mitophagy is not sufficient to increase mitochondria density in axons. Furthermore, axon injury increased axonal mitochondria density in pdr-1 mutant animals, indicating that mitophagy is not required to increase axonal mitochondria density after injury (Figure S2A-S2C and Table S1).
Next, to determine whether mitochondria division is increased by axon injury, we measured the size of mitochondria puncta. Mitochondria fission reduces the size of mitochondria (Detmer and Chan, 2007). Thus, if fission contributes to increased mitochondria density after axon injury we would expect mitochondria size to be reduced. However, injury had no effect on mitochondria puncta size (Figures S2D-S2F), suggesting that fission is not increased in response to injury and does not significantly contribute to increased mitochondria density in response to injury. We conclude that relocalization of mitochondria from the VNC to the commissure is the primary cause of increased mitochondria density after injury.
AMP-activated protein kinase (AMPK) senses cellular energy status to maintain energy homeostasis and acts as a major regulator of mitochondria (Hardie, 2007; Kahn et al., 2005). The C. elegans genome encodes two orthologs of the AMPK α catalytic subunits, AAK-1 and AAK-2. However, AAK-2 appears to be functionally more important, as aak-2 mutations affect longevity, stress response, and reproduction, while mutation of aak-1 does not cause obvious defects (Apfeld et al., 2004; Curtis et al., 2006; Lee et al., 2008). Consistent with previous findings, we found that mutation of aak-2 caused reduced axon regeneration, whereas mutation of aak-1 had no effect, (Figures S3A and S3B) (Hubert et al., 2014; Nix et al., 2014). However, aak-2 mutants exhibited indistinguishable mitochondria density in axons with or without injury compared to controls (Figures S3C, S3D, and Table S1). Similar results were observed for aak-1 mutants. These data indicate that the axon regeneration phenotype of aak-2 mutants is not due to changes in mitochondria density or translocation.
The increase in mitochondria density in axons after injury suggests that increased mitochondrial localization to the injured axon may support axon regeneration. Like mitochondrial localization (Figure 1), axon regeneration after injury is a highly variable process at the level of individual neurons. Under standard experimental conditions, approximately 70% of injured GABA axons manifest a successful growth cone, while 30% fail (Hammarlund et al., 2009; Yanik et al., 2004). We examined mitochondria density in individual severed axons 12 hours after injury, and also assessed axon regeneration of the same individual neurons (Figure 3A). While on average injured axons had increased mitochondria density compared to intact axons, in individual axons there was substantial overlap in mitochondria density between intact and injured axons. However, when only injured axons were considered and partitioned into those that had regenerated and those that had not, we observed a striking difference in mitochondria density comparing individual regenerating and non-regenerating axons. In regenerating axons, the distribution of mitochondria densities was skewed away from the lower end. Only 1 out of 21 individual regenerating axons had a mitochondria density less than the median density in intact control axons (11 puncta per 100 μm), whereas 14 of 45 non-regenerating axons had a mitochondria density less than 11 per 100 μm (Figures 3A and 3B). Thus, axons with low mitochondria density are unlikely to regenerate (6.6 percent, or 1 of 15 axons with density lower than 11 per 100 μm), compared to axons with high mitochondria density (39.2 percent, or 20 of 51 axons with density higher than 11 per 100 μm) (Figure 3C). Thus, increased mitochondria density is necessary for robust regeneration. Despite these differences, in both regenerating and non-regenerating axons, average mitochondria density increased compared to uninjured axons, and a substantial number of individual axons in both categories achieved high mitochondria density (Figure 3A). These data indicate that increased mitochondria density is necessary, but not sufficient for successful regeneration: even axons with high individual mitochondria density may fail to regenerate, but axons with low mitochondria density are very unlikely to do so.
To confirm the idea that low mitochondria density results in poor regeneration, we tested whether inhibition of mitochondrial transport in axons affects axon regeneration after injury. Miro is a mitochondrial outer membrane GTPase that is specifically required for mitochondrial transport (Fransson et al., 2003; Guo et al., 2005). The C. elegans genome encodes two Miro homologues, miro-1/K08F11.5 and miro-2/C47C12.4. 97% of the MIRO-2 amino acid sequence is conserved in MIRO-1, suggesting these genes may have a redundant function in mitochondrial transport (Figure 4A). To inhibit both miro-1 and miro-2 function exclusively in GABA neurons, we performed GABA-specific RNAi (Firnhaber and Hammarlund, 2013), against both miro-1 and miro-2 (Figure 4A). miro RNAi animals showed reduced mitochondria density in GABA commissures (Figure 4B, 4C and Table S1). Thus, C. elegans Miro has a conserved function in mitochondrial axon transport. Next, we assessed axon regeneration in miro RNAi animals, and found that reduction of Miro resulted in reduced regeneration across the dorsal-ventral body midline and reduced regrowth to the dorsal nerve cord (Figures 4D and 4E). The overall length of regenerated axons was shorter in miro RNAi animals (Figures 4D-4F). Therefore, mitochondrial transport is critical for axon regeneration. Detailed analysis of regeneration showed that although miro RNAi results in defective axon extension, miro RNAi and control animals have similar growth cone formation (Figure 4G). Thus, axonal mitochondria are dispensable for injury signaling and growth cone initiation, but are required for robust regenerative growth.
As axonal mitochondria promote regeneration, we considered whether increasing axonal mitochondria density above normal levels might improve regeneration. Overexpression of Miro has been shown to influence motility or subcellular distribution of mitochondria in neuronal cultures and Drosophila neurons (Ahmad et al., 2014; Guo et al., 2005; Saotome et al., 2008). Further, overexpressing Miro results in increased mitochondrial motility and increased growth in cultured cortical neurons (Zhou et al., 2016). Consistent with these results, we found that overexpression of miro-1 in GABA neurons resulted in increased mitochondria density (Figures 5A-5C and Table S1). Next, we tested the effect of miro-1 overexpression on axon regeneration. After injury, miro-1 overexpression did not affect growth cone formation (Figure 5E), but resulted in longer axon lengths compared to controls (Figures 5D, 5F and 5G). Thus, manipulation of the mitochondria transport factor Miro results in changes in axonal mitochondria density, and in corresponding changes in axon regeneration.
As an alternative approach to Miro manipulation to test the importance of axonal mitochondria for axon regeneration, we examined animals with abnormal mitochondrial fission. Dynamin-related protein-1 (DRP-1) is a member of the dynamin family of large GTPases that regulate mitochondrial fission (Labrousse et al., 1999). In Drosophila neurons, loss of DRP-1 function results in hyperfused mitochondria that fail to traffic to synapses (Verstreken et al., 2005). However, mitochondrial membrane potential and cellular ATP level remain largely unaffected in multiple studies (Lathrop and Steketee, 2013; Verstreken et al., 2005; Wakabayashi et al., 2009). We found that in the GABA neurons of drp-1 mutants, mitochondria were mainly confined to the cell body. Mitochondria were largely depleted from axons and synapses. Elongated mitochondria were visible in some axons, consistent with hyperfusion (Figure S4A). The abnormal distribution and shape of mitochondria distribution in drp-1 mutants were rescued by GABA-specific expression of wild type drp-1 (Figure S4A). Despite their mitochondrial defects, drp-1 null mutants exhibit normal GABA neuron morphology. We found that after single-neuron laser axotomy, drp-1 mutants had defects in axon regeneration similar to miro RNAi. While growth cone formation in drp-1 mutants was similar to control animals (Figure S4C), drp-1 mutants had reduced axon regeneration over the dorsal-ventral body midline and to the dorsal nerve cord (Figure S4B, and S4D). The overall length of regenerated axons was also shorter in drp-1 mutants (Figures S4E and S4F). Consistent with its effect on mitochondrial localization (Figure S4A), GABA-specific expression of wild type drp-1 rescued the axon regeneration defects in drp-1 null mutants (Figures S4D and S4G). Taken together, our data on Miro and drp-1 demonstrate that manipulation of mitochondria density causes corresponding changes in long-distance axon regeneration, consistent with the idea that axon regeneration is critically dependent on increased mitochondria density.
Mitochondria are multifunctional organelles that mediate energy metabolism, calcium homeostasis, regulation of reactive oxygen species (ROS) level, and other functions (Lin and Sheng, 2015). We hypothesized that mitochondrial function in energy metabolism is specifically important for injured axon regrowth, and that axons that lack sufficient mitochondria have reduced axon regeneration due to energy deficits. To test the hypothesis that axon regeneration depends on energy production by mitochondria, we assessed axon regeneration in animals with defective mitochondrial respiratory chain function. nuo-6 (a subunit of complex I) and isp-1 (a subunit of complex III) encode subunits of mitochondrial respiratory complexes. Partial loss of function mutations in these genes results in reduced energy production, as assessed by decreased ATP levels and oxygen consumption. Of the two mutants, isp-1(qm150) has lower ATP levels than nuo-6(qm200), making this a useful system for testing the relationship between energy levels and axon regeneration (Yang and Hekimi, 2010; Yee et al., 2014). We analyzed axon regeneration in these mutant animals 24 hours after axotomy. Similar to the regeneration defects caused by alterations in mitochondria density (Figures 3--5),5), the isp-1(qm150) and nuo-6(qm200) mutations resulted in reduced regrowth over the midline and less full regeneration to the dorsal nerve cord (Figures 6A to 6B). Further, isp-1 mutants had a lower axon regeneration rate than nuo-6 mutants, consistent with the different ATP levels between isp-1 mutants and nuo-6 mutants (Figures 6A to 6B) (Yang and Hekimi, 2010; Yee et al., 2014). Neither isp-1 nor nuo-6 mutants had abnormal mitochondrial positioning in the axon (Figures S5A-S5C and Table S1). Like control animals, axon injury increases mitochondrial density in nuo-6 mutants (Figures S5D-S5F). These observations indicate that positioning of mitochondria with proper energy production in axons is required for efficient axon regeneration after injury, and support the idea that the major role of axonal mitochondria during regeneration is to supply adequate levels of ATP.
isp-1 and nuo-6 mutants have mild increases in mitochondrial superoxide levels, but have decreased oxidative damage to proteins, increased expression of the superoxide dismutases (Yang and Hekimi, 2010; Yang et al., 2007), and increased resistance to paraquat treatment (Yang and Hekimi, 2010). These results suggest that oxidative damage or the cellular response to ROS is not the cause of reduced regeneration in these animals (Hwang and Lee, 2011; Lee et al., 2010; Yee et al., 2014). We further tested whether a mild increase in ROS levels could affect axon regeneration (Figure S6A). Treatment with 0.5 mM paraquat has been shown to be sufficient to increase the expression of superoxide dismutases and mitochondrial chaperones, and to cause a life span extension that recapitulates the life span extension of nuo-6 and isp-1 mutants (Lee et al., 2010; Runkel et al., 2013), suggesting that this treatment results in appreciable levels of ROS. Consistent with this, we found that treatment with 0.5 mM paraquat resulted in increased expression of an oxidative stress reporter, in which GFP is expressed under the promoter of gst-4 (Glutathione S-transferase 4) (Figure S6B) (Tawe et al., 1998). However, paraquat treatment did not alter axon regeneration (Figure S6C). Consistent with our finding that ROS do not inhibit axon regeneration in C. elegans, it has been shown that in zebrafish, hydrogen peroxide promotes axon regeneration of peripheral sensory neurons (Hwang and Lee, 2011; Lee et al., 2010; Rieger and Sagasti, 2011; Yee et al., 2014).
Mitochondria also play critical roles in cytoplasmic Ca2+ buffering and Ca2+ homeostasis. Under normal conditions, the mitochondrial calcium uniporter (MCU) complex regulates Ca2+ entry into the mitochondrial matrix (Baughman et al., 2011; De Stefani et al., 2011). mcu-1 is the sole MCU homologue in C. elegans, and loss of mcu-1 strongly reduces mitochondrial Ca2+ uptake induced by epidermal wounding (Xu and Chisholm, 2014). We found that mcu-1 mutants had normal levels of axon regeneration, indicating that impaired mitochondrial Ca2+ buffering after axotomy does not lead to axon regeneration defects (Figures S7A-S7D).
Together, these results support the model that energy production is the primary function of axonal mitochondria during axon regeneration. If this model is true, we hypothesized that regenerating axons might be in a state of energy crisis, consistent with our findings that alterations in mitochondria distribution and energy production have major effects on regeneration. We assessed energy status in axons using a GFP reporter of phosphofructokinase localization (PFK-1.1::eGFP). Phosphofructokinase is a key glycolytic enzyme, and PFK-1.1::eGFP forms clusters in neurons under diverse conditions that cause energy stress (Jang et al., 2016). Consistent with previous results, we found that PFK-1.1::eGFP clusters were rarely observed in intact GABA neurons under normal conditions. By contrast, after axon injury we found high levels of PFK-1.1::eGFP clusters, indicating that injured axons are under energy stress (Figures 6C-6E). However, disrupting glycolysis by mutating pfk-1.1 did not affect either long-distance axon regeneration or growth cone formation after injury (Figures 6F and 6G). Thus, injured axons are in a state of energy crisis, and mitochondria— rather than glycolysis—are required to supply sufficient energy for regeneration.
How does axon injury increase mitochondria density in injured axons? We hypothesized that increases in mitochondria density may be mediated by the DLK injury signaling pathway. DLK-1 is a dual leucine zipper-bearing MAP kinase kinase kinase (MAPKKK) that is a key determinant of axon regeneration in C. elegans, mice and flies (Hammarlund et al., 2009; Nakata et al., 2005; Shin et al., 2012; Xiong et al., 2010; Yan et al., 2009) (Figure 8A). In C. elegans GABA neurons, loss of dlk-1 blocks axon regeneration, while DLK-1 overexpression enhances axon regeneration (Hammarlund et al., 2009; Yan et al., 2009). We tested whether dlk-1 regulates mitochondria translocation. In intact axons, dlk-1 mutants had mitochondria density indistinguishable from controls. By contrast, at 24 hours after axotomy, injured axons in dlk-1 mutants had significantly lower mitochondria density than injured axons in controls (Figures 7A-7C and Table S1). While mitochondria density did increase in response to injury in dlk-1 mutants, the average increase was smaller: control axons showed about 1.9 fold increased mitochondrial density after injury, and dlk-1 mutant axons showed about 1.5 fold increase. Further, in dlk-1 mutants 27.4 percent of individual injured axons had a mitochondria density below 10, compared to 6.9 percent in controls (Figure 7C). Axons with such a low mitochondria density are highly unlikely to regenerate (Figure 3C). Thus, DLK signaling is required for axons to translocate sufficient mitochondria into axons for regeneration competence, and multiple injury signaling mechanisms, including DLK-1 signaling, regulate mitochondria density in injured axons.
To confirm the finding that DLK-1 signaling increases mitochondria density in response to injury (Figures 7A-7C), we tested whether DLK-1 activity is sufficient to increase mitochondria density even in the absence of axon injury. Overexpression of the active form of DLK-1 results in constitutive activation and enhanced regeneration (Hammarlund et al., 2009; Yan and Jin, 2012). We found that overexpression of the active form of DLK-1 (DLK-1 OE) increased mitochondria density in uninjured axons (Figures 7D-7F, and Table S1). By contrast, DLK-1 overexpression did not affect the axonal distribution or appearance of other organelles such as lysosomes, Golgi outposts, and synaptic vesicle precursors (Figures S8A-S8C). Thus, DLK-1 activity appears to specifically affect mitochondrial positioning in axons, rather than global axonal transport. A major mediator of DLK-1 signaling is the bZip domain-containing protein CEBP-1 (Yan et al., 2009). We found that loss of cebp-1 suppressed the increase in mitochondria density in animals overexpressing DLK-1 (Figures 7D-7F,and Table S1). We conclude that activation of the canonical DLK-1 pathway is sufficient to increase mitochondria density in axons.
Both DLK overexpression and Miro overexpression result in increased mitochondria density and improved axon regeneration (Figures 5 and and7).7). DLK overexpression functions via cebp-1 (Figure 7) (Yan et al., 2009), but downstream mechanisms are largely unknown. We tested the idea that DLK overexpression affects mitochondria and regeneration by increasing Miro-based mitochondria transport. We examined miro-1; miro-2 double mutant animals, either with or without dlk-1 overexpression. Consistent with miro(RNAi) animals (Figure 4), miro-1; miro-2 double mutants exhibited reduced mitochondrial density compared to wild-type animals (Figures 8A-8C). Further, dlk-1 overexpression in miro-1; miro-2 double mutants increased mitochondrial density in intact axons (Figures 8A-8C). Thus, Miro itself and Miro-dependent mechanisms of mitochondria transport are not required for the effect of DLK overexpression on axonal mitochondria density in intact neurons. Next, we severed axons and assessed regeneration. Consistent with previous studies, dlk-1 overexpression in GABA neurons increased overall axon length, and a larger fraction of axons extended to near the dorsal nerve cord (between 75% to 100% of relative axon length) (Hammarlund et al., 2009). miro-1; miro-2 double mutants, like miro RNAi (Figure 4), had reduced regeneration. However, dlk-1 overexpression in miro-1; miro-2 double mutants resulted in increased regeneration, similar to dlk-1 overexpressing animals alone (Figures 8D-8F). Together, these results indicate that DLK-1 overexpression and Miro regulate both axon regeneration and mitochondria localization via two independent mechanisms.
We have investigated the relationship between axon injury, axonal mitochondria, and regeneration. We find that axon injury results in increased mitochondria density in injured axons. High mitochondria density in individual axons is required for regeneration, and experimentally altering mitochondria density results in corresponding effects on regeneration. Injured axons are in energy crisis, and of all mitochondria functions only respiratory chain function is critical for axon regeneration, indicating that increased axonal mitochondria provide energy required for regeneration. Finally, the DLK-1 injury signaling pathway specifically promotes increased mitochondria density, potentially by mechanisms independent of the Miro-based mitochondria trafficking complex. Together, these data establish functional links between nerve injury, mitochondria localization, and axon regeneration.
We find that C. elegans axon injury triggers a robust increase in mitochondria density in injured axons in vivo. Mitochondria density increases quickly after injury, and is maintained to a late time point, when axon regeneration is complete. The distribution of mitochondria is uniform throughout the entire injured axon, including the growth cone area. Two different modes of injury—laser axotomy and mechanical breakage—trigger increases in mitochondria density, indicating that mitochondria density is responding to injury itself, rather than to side effects of the injury method. Further, we found that injury results in increased mitochondria translocation from neuronal cell bodies and proximal axons to the more distal commissures, indicating that overall mitochondria flux during this period is anterogradely directed. Our data are consistent with studies in mice that demonstrated increased axonal mitochondria traffic after peripheral nerve injury (Mar et al., 2014; Misgeld et al., 2007), suggesting that a general response to axon injury across species is to increase translocation of mitochondria into injured axons through regulation of mitochondrial transport and flux. An open question is whether mitochondria translocation into axons is accompanied by mitochondrial biogenesis. One regulator of mitochondrial biogenesis is AMPK, and we found that mutation of either C. elegans AMPK homolog did not affect mitochondria density or translocation, but further experiments are necessary to determine whether axon injury triggers a mitochondrial biogenesis response.
Are these axonal mitochondria important for regeneration? In mice, peripheral axon injury causes a global increase in the axonal transport of other factors in addition to mitochondria, including cytoskeleton components and metabolic enzymes (Mar et al., 2014; Misgeld et al., 2007). We used multiple approaches to focus specifically on axonal mitochondria. By measuring mitochondria and regeneration in single axons, we found that axons with low mitochondria density have poor regeneration. By experimentally manipulating mitochondrial localization using two independent methods that are relatively specific to mitochondria (Miro and DRP1), we found that increasing mitochondria density is sufficient to increase regeneration, while reducing mitochondria density reduced regeneration. Furthermore, we found that experimentally inhibiting mitochondrial respiratory chain function also reduces regeneration. These results are consistent with a recent study demonstrating that the growth of cultured cortical neurons is enhanced by increasing mitochondrial motility and by ATP supplementation (Zhou et al., 2016). Together, these results indicate that while multiple cellular components may be trafficked after nerve injury, axonal mitochondria have a specific and important function in axon regeneration.
Axon regeneration is a complex process, requiring injured neurons to generate injury signals, initiate a growth cone, and drive growth cone migration over extended distances toward the post-synaptic target (El Bejjani and Hammarlund, 2012). We found that experimentally altering axonal mitochondria density does not affect growth cone formation. Thus, injury signals and elaboration of growth cones after injury do not depend on axonal mitochondria. Rather, our data place the requirement for axonal mitochondria in the growth phase of the regeneration program, when injured axons are attempting to drive growth cone migration.
Animals with an impaired mitochondrial respiratory chain have reduced oxygen consumption and reduced ATP levels, indicating reduced energy production. We found that these mutants have reduced axon regeneration. We also found that injured axons activate a reporter of energy stress, consistent with the idea that ATP generation by axonal mitochondria is critical to support regeneration. These data suggest that supplying energy is the major function of mitochondria during axon regeneration. Recent work in cultured cortical neurons supports this idea, and further shows that axon injury can depolarize nearby mitochondria (Zhou et al., 2016). In regenerating axons, local ATP supply may depend on mitochondrial translocation within axons due to local mitochondrial depolarization (Zhou et al., 2016), the extended structure of axons, the limited diffusion ability of ATP (Hubley et al., 1995; Rostovtseva and Bezrukov, 1998; Sun et al., 2013), and ATP’s short biological half-life (less than 1 sec) (Mortensen et al., 2011). Axon regeneration requires a number of membrane trafficking steps (Bradke et al., 2012; Tuck and Cavalli, 2010), and these steps may be particularly sensitive to ATP levels (Rangaraju et al., 2014). Thus, although other aspects of axon regeneration may also require ATP, the growth phase depends strongly on ATP that must be supplied by axonal mitochondria.
Axon injury generates injury signals that are received and transduced by signaling pathways, including the conserved DLK-1 regeneration pathway (Hammarlund et al., 2009; Shin et al., 2012; Xiong et al., 2010). However, the cellular effects of these signaling pathways, and how they mediate axon regeneration, are incompletely understood. Here we show that the DLK-1 pathway affects mitochondria density in injured axons. In the absence of injury, DLK-1 activation is sufficient to increase mitochondria density, but does not change the distribution of other intracellular organelles. The ability of DLK-1 signaling to increase mitochondria density is largely dependent on the cebp-1 transcription factor, suggesting that transcriptional regulation of target genes contributes to this process.
In injured neurons that lack DLK-1 signaling, the post-injury increase in mitochondria density is reduced, but not eliminated. Thus, DLK-1 signaling is part of the cellular pathway linking injury to mitochondria density, but other, unknown injury pathways function in parallel to perform this role. Beyond DLK-1 and AMPK, axon regeneration has been linked to diverse signaling factors, including NGF (Lindsay, 1988) , Nogo (GrandPre et al., 2000), semaphorins (De Winter et al., 2002; Montolio et al., 2009), Roundabout/Slit (Chen et al., 2011), hippo (Nix et al., 2014), and PTEN (Byrne et al., 2014; Park et al., 2008; Sun et al., 2011). Many of these regulators also play roles in the regulation of mitochondria dynamics, function, or transport (Chada and Hollenbeck, 2004; Hollenbeck and Saxton, 2005; Nagaraj et al., 2012; Schwamborn et al., 2004; Sutendra et al., 2011; Thomas and Cookson, 2009). Further studies are necessary to determine the role of other pathways in the regulation of mitochondria density.
In summary, our results indicate that axon injury increases mitochondria density in injured axons, and these axonal mitochondria supply energy required for robust axon extension during axon regeneration. Our findings suggest that insufficient mitochondrial transport or energy production is one cause of failed axon regeneration, and raise the possibility that promoting energy production might help to improve axon regeneration after injury.
Laser axotomy was performed as described previously (Byrne et al., 2011). Axons were scored for full regeneration to the dorsal nerve cord or regrowth over the dorsal-ventral midline. To determine the relative axon length, the distance from the ventral nerve cord to the injured axon end was normalized by the distance between ventral and dorsal nerve cords. Details can be found in the Supplemental Experimental Procedures.
Transgenic animals expressing Mito::GFP or mCherry were analyzed at the young adult stage (see table S1). To assess mitochondria density in GABAergic neurons after axotomy, selected axons in transgenic animals expressing Mito::GFP or mCherry were cut using a Micropoint laser from Photonic Instruments (10 pulses, 20 Hz). The axotomized animals were recovered to NGM plates and cultured at 20°C. 6, 12 or 24 hours later, images were acquired as 0.3~0.5 μm z-stacks at room temperature on an UltraVIEW Vex (PerkinElmer) spinning disc confocal microscope (Nikon Ti-E Eclipse inverted scope; Hamamatsu C9100-50 camera) with a 60× CFI Plan Apo numerical aperture (NA) 1.4 oil-immersion objective using Volocity software (Improvision). To score mitochondria density, Mito::mCherry puncta in individual GABAergic axons were counted and then the axon length was measured. The number of mitochondria was normalized by the axon length and converted to density per 100 μm axon length. To compare the volume of Mito::mCherry puncta, Volocity software (Improvision) was used to measure the volume of individual puncta. All animals within each experiment were imaged on the same day with identical conditions including camera gain, exposure settings, and fluorescence filters.
The Dendra2 plasmid was generously provided by Dr. David Sherwood. Photoconversion was performed as previously described (Ihara et al., 2011). The ventral nerve cord of live transgenic animals was photoconverted and imaged using an UltraVIEW Vex (PerkinElmer) spinning disc confocal microscope (Nikon Ti-E Eclipse inverted scope; Hamamatsu C9100-50 camera) with a 40x objective and Volocity FRAP Plugin (Improvision). Selected ventral nerve cord and cell bodies were scanned with a 405 nm laser at 1% power. The photoconverted red-Mito-Dendra2 within axons was captured immediately after photoconversion as 0.5 μm z-stacks with a 60x objective. For experiments involving injured axons, GABA neurons were severed using a pulsed laser axotomy and then photoconverted within about 5 minutes. Imaging immediately after axotomy did not show red-Mito::Dendra2 signal within the injured axon commissures. To test mitochondrial trafficking from the ventral nerve cord to the axons, animals with or without axotomy were recovered from the agar pad after photoconversion, cultured at 20°C for 24 hour s, and then reimaged with identical settings using the 60x objective.
After axotomy, animals were recovered on fresh NGM plates. Twenty four hours after axotomy, animals were mounted, and images were acquired as 0.3 μm z-stacks at room temperature within 10 minutes, to minimize hypoxic effects. PFK-1.1::eGFP clustering was analyzed as described previously (Jang et al., 2016). Briefly, segmented line scans were performed to measure fluorescence values for PFK-1.1::eGFP in individual axons using Volocity software (Improvision). Fluorescence peaks with higher values than 5,000 (A. U.) as a threshold were considered PFK-1.1::eGFP clusters.
We thank the Caenorhabditis Genetics Center (funded by the NIH Center for Research Resources) and the Mitani lab (Tokyo Women’s Medical University School of Medicine) for strains, Dr. Cori Bargmann, Dr. Daniel Colon-Ramos, Dr. Siegfried Hekimi and Dr. David R. Sherwood for plasmids or strains. We thank the Yale C. elegans community and the Yale axon regeneration group for advice on the project. We thank Sori Jang, Dr. Hieu Hoang, Dr. Sejin Lee, and members of the Hammarlund lab for advice and for sharing materials. S.M.H was supported by the James Hudson Brown - Alexander Brown Coxe Postdoctoral Fellowships in the Medical Sciences, and a Department of Genetics Research Fellowship from Yale University.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Han et al. find that axon injury and activation of DLK-1 MAP kinase increase axonal mitochondria density. Robust mitochondria density is required to produce adequate ATP for axon regeneration.
Lead contact: Marc Hammarlund
Author contributions: S.M.H. and M.H. designed the experiments, analyzed data, and wrote the manuscript. S.M.H. and H.S.B conducted the experiments.