Wld^S prevents axon degeneration through increased mitochondrial flux and enhanced mitochondrial Ca²⁺ buffering

Summary

Wld^S (slow Wallerian degeneration) is a remarkable protein that can suppress Wallerian degeneration of axons and synapses [1] but how it exerts this effect remains unclear [2]. Here, using Drosophila and mouse models, we identify mitochondria as a key site of action for Wld^S neuroprotective function. Targeting the NAD⁺ biosynthetic enzyme Nmnat to mitochondria was sufficient to fully phenocopy Wld^S, and Wld^S was specifically localized to mitochondria in synaptic preparations from mouse brain. Axotomy of live wild type axons induced a dramatic spike in axoplasmic Ca²⁺ and termination of mitochondrial movement—Wld^S potently suppressed both of these events. Surprisingly, Wld^S also promoted increased basal mitochondrial motility in axons before injury, and genetically suppressing mitochondrial motility in vivo dramatically reduced the protective effect of Wld^S. Intriguingly, purified mitochondria from Wld^S mice exhibited enhanced Ca²⁺ buffering capacity. We propose that the enhanced Ca²⁺ buffering capacity of Wld^S+ mitochondria leads to increased mitochondrial motility, suppression of axotomy-induced Ca²⁺ elevation in axons, and thereby suppression of Wallerian degeneration.

Results and Discussion

Mitochondria as a key site of Wld^s neuroprotective function

Remarkably, the distal fragments of severed axons survive for weeks after axotomy in the Wld^s mouse [3–6]. The Wld^s mutation resulted from the fusion of two neighboring genes and lead to the production of a novel hybrid protein (Wld^s) composed of the 70 NH-terminal amino acids of the E4 ubiqutin ligase Ube4b, a novel 18 amino acid linker domain, and the NAD⁺ biosynthetic enzyme Nmnat1 [7]. We previously found that expression of mouse Nmnat3 in Drosophila olfactory receptor neuron (ORN) axons provided protection equivalent to Wld^S 5 days after axotomy [8]. Recent work has shown that Nmnat3 expression in mouse neurons also robustly protects axons [9]. We co-expressed mouse UAS-Nmnat3::Myc and the mitochondrial marker UAS-mito::GFP in Drosophila ORNs. We found Nmnat3::Myc localized in a punctate pattern in ORN axons that precisely overlapped with mito::GFP (Figure 1A), indicating that Nmnat3 localized predominantly, if not exclusively, to mitochondria. We next assayed ORN axon preservation at 10, 20, and 50 days after axotomy. We found that Nmnat3 protected axons at levels indistinguishable from Wld^S at all time points tested (Figure 1B). Thus the N70 and W18 domains of Wld^S are dispensable for axon protection if Nmnat activity is targeted to mitochondria. By contrast, expression of Nmnat2 in ORN axons failed to suppress Wallerian degeneration, despite the fact that Nmnat2::Myc was localized throughout the axonal compartment (Supplementary Figure 1A).

Figure 1. Mitochondria as a focal point for Wld^S-mediated axon protection.

(A) Mouse Nmnat3::Myc localizes to mitochondria in Drosophila axons. 22a-Gal4 was used to drive UAS-Nmnat3::Myc and UAS-mito::GFP. Insets show boxed region.

(B) Mitochondrial Nmnat3 fully mimics Wld^S in axon protective function. 22a-Gal4 was used to drive UAS-Nmnat3 or UAS-Wld^S in a background where axons were labeled with membrane-tethered GFP (UAS-mCD8::GFP). n≥20 antennal lobes for each. *** indicates p<0.001. Error bars represent ±SEM.

(C) Wld^S-suppresses axotomy-induced termination of mitochondrial motility. Mitochondrial movement was assessed in live open-filet preparations of third instar Drosophila larvae immediately after axotomy for 5 minutes. Axotomy was induced by severing axons with a Micropoint laser ablation system and confirmed by a breakage of mCD8::mCherry-labeled axons. n≥10 live samples for each genotype and time point. ***, p<0.001.

Recently Yahata et al. (9) reported that in mouse neurons Wld^S protein is located in mitochondria, cytosol, perosixome/lysosome, and ER/Golgi-enriched cell fractions. We revisited Wld^s localization in isolated mouse striatum from control and Wld^s mice by separating the tissues into 3 fractions: non-synaptic striatal tissue, striatal synaptosomes without mitochondria, and synaptic mitochondria. We found Wld^S was detectable in the non-synaptic fraction, as would be expected from its predominantly nuclear localization. In addition, we detected Wld^S in synaptic mitochondria, but not in mitochondria-free synaptic preparations (Supplementary Figure 1B). These data are consistent with a primarily mitochondrial localization of Wld^S in axons and synapses in vivo in mouse brain.

Wld^s suppresses termination of mitochondrial motility after injury and axotomy-induced increases in axonal Ca²⁺

We assayed mitochondrial dynamics in live Drosophila axons using the Tdc2-Gal4 driver, which is expressed in only 3 axons per segment of larval peripheral nerve, by driving UAS-mCD8::mCherry (to label axonal membranes) and UAS-mito::GFP (to label mitochondria; Figure 3A). In uninjured axons we found no differences in the total number of mitochondria, mitochondrial morphology, or mitochondrial size when we compared control and Wld^S-expressing axons (Supplementary Figure 1A,B). In control animals we found ~35% of mitochondria were motile before injury, but all motility terminated after laser axotomy (Supplementary Movie 1). In striking contrast, we found that laser axotomy of Wld^S-expressing axons had no effect on mitochondrial movement (Supplementary Movie 2; Figure 1C).

Figure 3. Wld^s increases mitochondrial flux, which is essential for neuroprotective function.

(A) Mitochondrial flux was assayed in tdc2-Gal4-expressing neurons in live preparations of third instar larvae. Representative kymographs of mitochondrial movement are shown for control and Wld^S-expressing axons. Anterograde is to the right, retrograde is to the left.

(B) Mitochondrial flux was assayed in axons expressing each of the following molecules: Nmnat1, DN16-Wld^S, Wld^S-dead, Wld^S, N16-Nmnat1, and Nmant3. n≥10 live samples for each genotype. *, p<0.05; ***, p<0.001.

(C) Quantification of the movement of individual mitochondrial by binning into mobile, docked, or pausing/releasing during a 5 minute window. n=5 movies for each genotype.

(D) ORN axotomy assays in Wld^S backgrounds and miro mutants. A single antennal lobe where ORN axons were severed is shown.

(E) Quantification of data from (D). Age-matched uninjured controls at the same time points are shown at right. n≥15 samples for each genotype and time point. **, p<0.01; ***, p<0.001.

Axon injury in mammals leads to extracellular Ca²⁺ entry, which is necessary and sufficient for Wallerian degeneration [10]. Mitochondrial motility is known to be potently modulated by Ca²⁺ [11, 12]. We therefore sought to determine whether Drosophila axons showed axotomy-induced changes in axonal Ca²⁺, and whether axonal Ca²⁺ signaling was modulated by Wld^S. We drove the expression of the genetically-encoded Ca²⁺ indicator GCaMP3 in axons and measured changes in GCaMP3 signals in distal axon segments after laser-induced axotomy. In control animals we found a rapid increase in Ca²⁺ levels within seconds after axotomy, with Ca²⁺ levels peaking within 1 minute after cut, and then returning towards baseline levels over the next hour (Figure 2A,B,C; Supplementary Movie 3). However, >1 hour after axotomy axonal Ca²⁺ levels remained significantly elevated above baseline (~20% increase). Strikingly, while baseline Ca²⁺ levels in Wld^S+ axons were indistinguishable from those in controls (Supplementary Figure 3C), in Wld^s+ axons injury-induced Ca²⁺ bursts were almost completely eliminated: Ca²⁺ levels only rose to ~15% control levels and returned to baseline within 1.5 minutes (Figure 2C; Supplementary Movie 4).

Figure 2. In vivo laser axotomy induces a dramatic rise in axonal Ca²⁺ that is suppressed by Wld^S.

(A) tdc2-Gal4 labels 3 axons in each peripheral nerve, only one segment is illustrated.

(B) Axons were labeled with mCD8::mCherry, axonal Ca²⁺ was monitored by co-expressing GCaMP3 in the tdc2-Gal4⁺ subset of neurons. Note the breakage of the axon after laser axotomy (red, mCherry). Axonal Ca²⁺ levels one minute after axotomy (green, GCaMP3).

(C,D) Representative traces showing Ca²⁺ responses in axon fragments distal to the injury site over time. Genotypes as indicated.

(E) Quantification of peak Ca²⁺ intensities and time to ½ recovery from average peak intensity for each genotype listed. n≥5 live samples for each genotype and time point versus control. *, p<0.05; **, p<0.01; ***, p<0.001.

In previous work we generated a collection of UAS-regulated Wld^s-derived molecules that suppress Wallerian degeneration to varying degrees [8] (Supplementary Figure 1C). We assayed axotomy-induced changes in GCaMP3 fluorescence in axons expressing each of these molecules in live larval preparations. As with Wld^s, we found no evidence for changes in axonal mitochondrial number, morphology, or size in these backgrounds (Supplementary Figure 2A,B). However, we found a striking correlation between axon protective function and suppression of axotomy-induced increases in axonal Ca²⁺: Wld^s and Nmnat3 strongly suppressed Wallerian degeneration and post-injury axonal Ca²⁺ increases; Nmnat1, ΔN16::Wld^s, and Wld^s-dead partially suppressed Wallerian degeneration and post-injury axonal Ca²⁺ increases; and Nmant1^dead, which lacks NAD⁺ biosynthetic activity and provided no protection from Wallerian degeneration [8], did not affect post-injury axonal Ca²⁺ increases. These changes were evident in both the distal and proximal axon segment, and affected both peak axonal Ca²⁺ intensities and recovery times to baseline Ca²⁺ levels (Figure 2C,D,E; Supplementary Figures 3A,B).

Wld^s enhances mitochondrial movement, which is essential for maximal axonal protection after injury

Since mitochondrial motility is Ca²⁺-modulated we reasoned that changes in axonal Ca²⁺ buffering might affect axonal mitochondrial motility. We therefore assayed mitochondrial flux in axons expressing Wld^S-derived neuroprotective molecules. Surprisingly, we found a significant change in basal mitochondrial motility: in control animals ~35% of total axonal mitochondria were motile, however ~65% of mitochondria were motile in Wld^s+ axons. Moreover, we found that molecules that provide partial suppression of Wallerian degeneration (Nmnat1, ΔN16::Wld^S, and Wld^S-dead) led to a modest, but significant, increase in the number of motile mitochondria, while molecules that maximally suppress Wallerian degeneration (Wld^S, N16::Nmnat1, and Nmnat3) led to a robust increase in the number of motile mitochondria (Figure 3B). This change in mitochondrial flux in Wld^S axons appears to represent a decrease in docked mitochondria, an increase in motile mitochondria, but no significant change in pause/release rates for individual mitochondria (Figure 3C).

Is increased mitochondrial flux critical for Wld^S-mediated axon protection? The adapter protein Miro functions to tether mitochondria to cytoskeletal motor proteins and modulate mitochondrial movement in a Ca²⁺-dependent fashion [12]. Impressively, mutations in miro dominantly decrease mitochondrial motility [13]. We therefore crossed strong alleles of miro (miro^SD32 and miro^SD26) and a UAS-regulated version of Miro (UAS-myc::Miro), which when overexpressed acts as a dominant negative [13] into the Wld^S background and assayed mitochondrial flux. We found that miro mutants or expression of Myc::Miro dominantly suppressed mitochondrial movement in controls. In addition, we found that loss of Miro function also decreased mitochondrial movement in the presence of Wld^S to levels found in control animals (Supplementary Figure 2C,D). Remarkably, reduced miro function also dominantly suppressed the neuroprotective effects of Wld^S in ORN axotomy assays. In animals with reduced miro function, axon loss was evident by 5 days after axotomy, with synaptic regions showing significant degeneration, and by 30 days after axotomy the protection afforded by Wld^S is almost completely blocked (Figure 3D,E).

Mitochondria from Wld^s-expressing axons show enhanced Ca²⁺ buffering capacity and resistance to formation of the permeability transition pore

Mitochondria are major sinks for cellular Ca²⁺ in both axons and synapses [14]. A powerful mechanism by which Wld^S could exert all of these effects would be by altering mitochondrial Ca²⁺ buffering capacity. We therefore assessed mitochondrial Ca²⁺ cycling/capacity in cortical mitochondria isolated from young (~p25) wild type (NJ) and Wld^S mice. Mitochondrial isolations [15, 16] from both control and Wld^S animals yielded healthy, well-coupled mitochondria (Figure 4A,B). No apparent difference was observed in the Ca²⁺ uptake rates in mitochondria isolated from Wld^S versus control mice (Figure 4A,B). In contrast, the threshold for mitochondrial permeability, indicated by the loss of membrane potential (Figure 4A) and mitochondrial release of Ca²⁺ (Figure 4B), was significantly greater in mitochondria from Wld^S mice (Figure 4C). Thus, following increases in cytoplasmic Ca²⁺, Wld^S mitochondria isolated from mouse brain buffer higher loads of Ca²⁺ before releasing it back into the cytoplasmic compartment via the mitochondria permeability transition pore (PTP).

Figure 4. Wld^S brain mitochondria display higher Ca²⁺ load capacity than age matched wild type (NJ) controls.

(A,B) TMRE (membrane potential indicator) and CaG5N (extra-mitochondrial Ca²⁺ indicator) fluorescence were monitored over time simultaneously for each sample of non-synaptic mitochondria. As illustrated in TMRE traces for the first three minutes, the addition of pyruvate and malate (PM) an oxidative substrate, causes a marked downward deflection at 1 minute due to increased mitochondrial membrane potential (Δψm). Following ADP (A) addition, the loss of Δψm is indicated by upward deflection at 2 mins as Δψm is utilized to phosphorylate ADP to ATP via proton flow thru the ATP synthase. The ATP synthase inhibitor, oligomycin (O) addition at 3 min results in maximum Δψm as proton flow is inhibited. The Ca²⁺ infusion began at 5 min (infusion rate 160 nmol of Ca²⁺/mg protein/min) and was monitored by CaG5N fluorescence, and is illustrated by the initial upward deflection followed by constant signal due to mitochondrial Ca²⁺ uptake into the matrix. The subsequent rise in CaG5N fluorescence accompanied by a loss of membrane potential signifies mitochondrial permeability transition and subsequent release of mitochondrial Ca²⁺.

(C) Quantification of mitochondrial Ca²⁺ buffering capacity (nmols/mg protein) indicates that Wld^S non-synaptic mitochondria sequestered significantly higher amounts of Ca²⁺ compared to control group (n=6/group, * p<0.05, unpaired t-test).

Conclusions

The mechanistic action of Wld^s has remained controversial, but recent work has established a non-nuclear role for Wld^s [2] after injury [17]. In this study we show that Wld^S is localized to mitochondria in vivo. It is important to note that protein localization studies with Wld^s must be interpreted cautiously—the primarily nuclear localization of Wld^s suggested a nuclear role for Wld^s and initially misled the field [2]. However, we also find that Wld^s increases mitochondrial Ca²⁺ buffering capacity, and results in maintained mitochondrial motility after axotomy. Taken together, these data argue strongly that the mitochondrial compartment is a key site of action for Wld^S in vivo.

We have shown that axonal injury in live Drosophila preparations leads to a dramatic and transient rise in axonal Ca²⁺. Increased axonal Ca²⁺ has been observed in mammals after acute nerve crush [18] and entry of extracellular Ca²⁺ is necessary and sufficient for Wallerian degeneration [10]. Impressively, Wld^s expression resulted in a striking suppression of this axotomy-induced rise in axonal Ca²⁺. The most plausible explanation for this enhanced buffering is that increased ATP/energy production observed in Wld^s+ mitochondria [9]—presumably via increased mitochondrial NAD⁺ production, though we cannot formally exclude essential roles for other substrates of Nmnat—is linked to increased mitochondrial membrane potential (Δψm), and thereby increased Ca²⁺ entry through the Δψm–regulated mitochondrial Ca²⁺ uniporter [19]. This model is supported by our observation that Wld^s-expressing mitochondria isolated from mouse brain exhibit an enhanced ability to maintain their membrane potential and avoid PTP formation in the face of increasing extramitochondrial Ca²⁺. In the future it will be important to confirm that such changes are also observed in Drosophila axonal mitochondrial physiology in vivo in Wld^s-expressing neurons.

Axonal Ca²⁺ spikes could result solely from entry of extracellular Ca²⁺ into the axon after injury. This would be consistent with the observation that blocking Ca²⁺ channels inhibits Wallerian degeneration [10, 18]. Mitochondria are a well-established sink for Ca²⁺ in axons [14] and here we show that Wld^s+ mitochondria exhibit enhanced Ca²⁺ buffering capacity and resistance to Ca²⁺-induced formation of the permeability transition pore (PTP). Indeed PTP formation appears to be a key final execution step in Wallerian degeneration [20–22]. We therefore favor a model whereby extracellular Ca²⁺ enters the axon after axotomy and normally acts as a switch to activate Wallerian degeneration. In Wld^s axons this Ca²⁺ is instead rapidly buffered by mitochondria, thereby blocking induction of axonal destruction. Consistent with this model, uncoupling mitochondria, which suppresses mitochondrial Ca²⁺ uptake [23, 24], completely abrogates the protective effect of Wld^S in vitro [25].

Wld^S-expressing neurons exhibit a roughly 2-fold increase in the number of motile versus stationary mitochondria compared to wild type controls, which could result from changes in mitochondrial Ca²⁺ buffering. Notably, genetic suppression of enhanced mitochondrial flux using mutations in miro also resulted in a remarkable suppression of Wld^s-mediated axonal protection in vivo. However, since this suppression was only partial, additional factors beyond increases in mitochondrial motility must also contribute to Wld^s-mediated axonal protection. For example, axonal energy supplies are likely closely intertwined with mitochondrial transport and bioenergetics. Wld^s+ mitochondria are known to exhibit an enhanced ability to generate ATP [9]. This change in bioenergetics, coupled with increased mitochondrial motility in Wld^s+ axons might enhance distribution of ATP or other mitochondrially-derived metabolites. At the same time, enhanced mitochondrial motility could also speed the removal of metabolic byproducts normally processed by mitochondria. Similarly, increased mitochondrial motility in axons could further enhance mitochondrial Ca²⁺ buffering in Wld^s+ axons since motile mitochondria would be predicted to traverse more — axonal space, and perhaps be exposed to more Ca²⁺ than stationary mitochondria. Together these could have the combined effect of increasing energy delivery, removing harmful byproducts, and increased buffering of Ca²⁺, a signal that can potently activate axonal degeneration.

A role for mitochondria in the Wld^s neuroprotective mechanism is intriguing since defects in mitochondria respiration and dynamics are emerging as critical underlying factors in a number of neurological disorders [26]. For example, in mouse models of ALS (SOD1 transgenics) anterograde [27] and retrograde [28] mitochondrial transport is reduced; altered mitochondrial trafficking has been observed in models of Alzheimers disease [29]; and mutant, but not wild type Huntington protein blocks mitochondrial movement in cortical neurons [30]. However, in the majority of models, whether these mitochondrial alterations are a cause or consequence of disease remains an open question [26]. Our study shows, reciprocally, that enhanced mitochondrial flux is associated with, and is required for maximal axon protection by Wld^s.

Highlights.

• Axotomy induces dramatic Ca²⁺ bursts and termination of mitochondrial motility in axons

• Wld^s suppresses axotomy-induced Ca²⁺ bursts and axons maintain mitochondrial motility after axotomy

• Wld^s is localized to mitochondria and enhances mitochondrial Ca²⁺ buffering, resistance to mPTP, and mitochondrial motility in vivo

• Enhanced mitochondrial motility is essential for Wld^s neuroprotective function