Wallerian Degeneration Is Executed by an NMN-SARM1-Dependent Late Ca 2+ Inﬂux but Only Modestly Inﬂuenced by Mitochondria

Axon injury leads to rapid depletion of NAD-biosyn-thetic enzyme NMNAT2 and high levels of its sub-strate, NMN. We proposed a key role for NMN in Wallerian degeneration but downstream events and their relationship to other mediators remain unclear. Here, we show, in vitro and in vivo, that axotomy leads to a late increase in intra-axonal Ca 2+ , abolished by pharmacological or genetic reduction of NMN levels. NMN requires the pro-degenerative protein SARM1 to stimulate Ca 2+ inﬂux and axon degeneration. While inhibition of NMN synthesis and SARM1 deletion block Ca 2+ rise and preserve axonal integrity, they fail to prevent early mitochondrial dynamic changes. Furthermore, depolarizing mitochondria does not alter the rate of Wallerian degeneration. These data reveal that NMN and SARM1 act in a common pathway culminating in intra-axonal Ca 2+ increase and fragmentation and dissociate mitochondrial dysfunctions from this pathway, elucidating which steps may be most effective as targets for therapy.

Correspondence laura.conforti@nottingham.ac.uk In Brief Loreto et al. show that NAD-precursor NMN, previously reported to induce Wallerian degeneration, stimulates a Ca 2+ rise in injured axons, which requires the protein SARM1 and shortly precedes fragmentation. Changes in mitochondrial dynamics are not associated with degeneration. These results advance our understanding of axon degeneration mechanisms relevant to neurodegeneration.

INTRODUCTION
Axon degeneration is an early event contributing to symptoms in a range of age-related neurodegenerative disorders, including Parkinson's disease, Huntington's disease, Alzheimer's disease, motor neuron disease, and multiple sclerosis. Axonal damage also underpins peripheral neuropathies and is a limiting side effect of cancer chemotherapeutic agents (Conforti et al., 2014). Maintaining axon structure and function is therefore a promising therapeutic goal for these disorders.
Two genetic modifications can delay by 10-fold axon degeneration after an acute injury, a process known as Wallerian degeneration (Waller, 1850), highlighting the existence of a regulated and active axon death pathway. First, expression of the neuroprotective factor Wallerian Degeneration Slow (WLD s ), a chimeric protein which fuses the nuclear isoform of the nicotinamide adenine dinucleotide (NAD)-biosynthetic enzyme nicotin-amide mononucleotide adenylyltransferase 1 (NMNAT1) with an N-terminal sequence that relocates this enzyme from the nucleus to the cytoplasm and axons (Avery et al., 2009;Conforti et al., 2000Conforti et al., , 2009Mack et al., 2001). WLD s is effective in mice, rats, Drosophila, zebrafish and in primary human neuronal culture (Adalbert et al., 2005;Kitay et al., 2013;Lunn et al., 1989;MacDonald et al., 2006;Martin et al., 2010), indicating strong evolutionary conservation of the axon degeneration pathway with which it interferes. WLD s also ameliorates axon pathology and symptoms in some disease models (Conforti et al., 2014), highlighting shared mechanisms and translational potential. Remarkably, loss of the Toll-like receptor adaptor protein Sterile alpha and TIR motif containing 1 (SARM1) has the same robust protective efficacy of WLD s . Similar to WLD s , the protective effect of SARM1 deletion is evolutionarily conserved in mice and Drosophila (Osterloh et al., 2012).
It has been proposed that WLD s delays axon degeneration by substituting for the endogenous but short-lived axonal NMNAT isoform NMNAT2, whose enzyme activity is required for axon growth and maintenance. NMNAT2 is synthesized in the cell bodies, actively transported down the axons, and rapidly degraded after injury or block of axonal transport. WLD s , which retains the same enzymatic activity than NMNAT2 but is much more stable, prolongs axon integrity Gilley and Coleman, 2010;Milde et al., 2013).
Our previous data support a model in which WLD s /NMNAT activity maintains axon integrity by preventing the accumulation of its substrate, the pro-degenerative NAD precursor nicotinamide mononucleotide (NMN). NMN increases after injury likely as a consequence of NMNAT2 depletion and its pharmacological or genetic inhibition strongly delays Wallerian degeneration (Di Stefano et al., 2014). The robustness of axon protection phenotype conferred by SARM1 deletion resembles that obtained by scavenging of NMN. Crucially, SARM1 depletion blocks NMNAT2 loss-induced axon degeneration but does not reduce NMN, indicating that SARM1 influences the same pathway of axon death downstream of NMN, or a convergent pathway, while other weaker modulators may join onto this pathway (Conforti et al., 2014;Gilley et al., 2015;Yang et al., 2015).
NMN increases early after injury (Di Stefano et al., 2014), leaving ample scope for downstream events, which could include an effect on intra-axonal Ca 2+ . A key role for Ca 2+ in the late steps of degeneration, likely acting by activating (legend on next page) Ca 2+ -dependent proteases, has long been recognized (George et al., 1995;Wang et al., 2012). Ca 2+ could reach the axons from the extracellular environment or originate from intra-axonal organelles (Barrientos et al., 2011;Villegas et al., 2014). Recent studies suggest instead that Ca 2+ rise may occur early after injury , possibly acting upstream of SARM1 (Summers et al., 2014).
Mitochondria have been reported as central executors in Wallerian degeneration through mitochondrial permeability transition pore (mPTP) opening, which could result from a crosstalk with the endoplasmic reticulum (ER) (Barrientos et al., 2011;Villegas et al., 2014). Augmented mitochondrial motility in WLD s could increase Ca 2+ buffering capability and contribute to protection . However, Wallerian degeneration is only modestly delayed in axons devoid of mitochondria in some models and mitochondria are dispensable for WLD s protection (Kitay et al., 2013); therefore, the role of mitochondria in Wallerian degeneration and the WLD s mechanism as well as the relationships among NMN increase, SARM1, Ca 2+ , and mitochondrial dynamics remain unresolved.
Here we report that intra-axonal Ca 2+ elevation after axotomy lies downstream of NMN and immediately precedes axon fragmentation. We found that SARM1 is required for NMN to initiate axon degeneration and that Ca 2+ rise is a common downstream step. Both NMN rise inhibition and SARM1 deletion fail to prevent changes in mitochondrial dynamics occurring after axotomy, despite delaying degeneration. We also found that the rate of Wallerian degeneration is not affected by mitochondrial depolarization. By clarifying the temporal relationship between NMN rise, SARM1, Ca 2+ , and mitochondrial dynamics, our results identify critical steps of Wallerian degeneration where pharmacological intervention could be most successful.

RESULTS
Pharmacological and Genetic Data Indicate that NMN Induces a Rise in Intra-axonal Ca 2+ Rapidly Followed by Axon Fragmentation To address events downstream of NAD-precursor NMN (Figure S1A), which we proposed to induce axon degeneration after injury (Di Stefano et al., 2014), we first tested whether NMN accumulation evokes changes in intra-axonal Ca 2+ concentration. This is important since WLD s , which blocks NMN accumulation, has been suggested to protect degenerating axons by inhibiting Ca 2+ increase or enhancing its buffering Avery et al., 2012). To analyze the temporal correlation between Ca 2+ rise and axonal fragmentation, we microinjected superior cervical ganglia (SCG) neurons with plasmids encoding the Ca 2+ indicator GCaMP5 and the fluorescent marker DsRed2. One day after microinjection, the axons were separated from the cell bodies and changes in GCaMP5 fluorescence signal in distal stumps were measured and normalized to their corresponding uncut controls (Figure 1; Figures S1B and S1C). Consistent with previous studies , we found a marked increase in Ca 2+ levels in distal injured axons starting 2-3 hr after axotomy (Figures 1A and 1B; Movie S1). This Ca 2+ rise just preceded the appearance of morphological damage and axon integrity was lost 10-25 min after Ca 2+ started to increase (Figures 1A and 1C). To test whether this Ca 2+ rise was induced by NMN accumulation, we treated axons with FK866, a potent and specific inhibitor of NMN-synthesizing enzyme NAMPT ( Figure S1A). FK866 completely abolished the Ca 2+ rise and, consistent with our previous report (Di Stefano et al., 2014), preserved axonal integrity for the entire duration of the experiment ( Figures 1A-1C; Movie S1). Co-administration of NMN fully reverted the effects of FK866, restoring Ca 2+ rise and axonal fragmentation ( Figures 1A-1C; Movie S1).
We previously showed that scavenging NMN by expressing the bacterial enzyme NMN deamidase ( Figure S1A; Galeazzi et al., 2011) remarkably delays axon degeneration after injury (Di Stefano et al., 2014). We then asked whether this genetic intervention to reduce NMN could also block intra-axonal Ca 2+ rise and found that expressing NMN deamidase completely abolished this increase and preserved axonal integrity for the entire duration of the experiment ( Figures 1D-1F; Movie S1).
Thus, NMN increase after injury evokes a late intra-axonal Ca 2+ rise, which can be blocked by NAMPT inhibitor FK866 or by genetically scavenging NMN.
NMN Requires SARM1 to Induce Degeneration and Intra-axonal Ca 2+ Increase Because injured Sarm1 À/À nerves have high NMN levels but do not degenerate (Gilley et al., 2015), it is likely that SARM1 is required for NMN to induce axon death. We tested this possibility  A) Plasmids encoding for the Ca 2+ indicator GCaMP5 and for the axonal marker DsRed2 were microinjected into wild-type SCG nuclei; axons were cut 1 day after microinjection and left untreated or treated as shown with FK866 and NMN, added immediately after cut. Time-lapse fluorescent images were acquired every 5 min up to 260 min. Scale bar represents 50 mm. (B) Quantification of Ca 2+ peak intensities (see Experimental Procedures). All measurements were normalized to uncut controls. FK866, which blocks NMN and confers protection after axotomy, completely abolished injury-induced Ca 2+ rise; NMN addition re-established Ca 2+ increase and reverted FK866-mediated protection. Mean ± SEM; n = 6, two fields in three independent experiments; one-way ANOVA followed by Bonferroni post hoc test; ***p < 0.001. (C) Representative traces showing Ca 2+ responses in axon fragments distal to the injury site over time in correlation with the percentage of healthy axons. Note that Ca 2+ rise shortly preceded axonal fragmentation in both the untreated and FK866+NMN-treated samples. No Ca 2+ increase was detected in the FK866treated sample. (D) Fluorescent images of wild-type SCG axons co-injected with plasmids encoding GCaMP5, DsRed2, and either NMN deamidase or Empty vector, at the indicated time points after transection. Scale bar represents 50 mm. (E) Quantification of Ca 2+ peak intensities. All measurements were normalized to uncut controls. No Ca 2+ increase was detected in NMN deamidase expressing axons. Mean ± SEM; n = 6, two fields in three independent experiments; one-way ANOVA followed by Bonferroni post hoc test; ***p < 0.001. (F) Representative traces showing no Ca 2+ increase in NMN deamidase distal axon fragments in correlation with the percentage of healthy axons. See also Figures S1 and S2. by adding NMN to transected Sarm1 À/À axons and found that they remained intact for at least 4 days, even in the presence of high concentrations of exogenous NMN (Figures 2A and  2B), which we even tested up to 5 mM (data not shown). In contrast, FK866-protected wild-type neurites degenerate within 3 hr when exposed to concentrations of NMN 200-fold lower (Di Stefano et al., 2014). Thus, NMN requires SARM1 to induce axon degeneration.
We then asked whether SARM1 is also required for the increase in intra-axonal Ca 2+ after cut. This is important in view of a recent report that SARM1 acts downstream of Ca 2+ (Summers et al., 2014). However, in our experimental conditions we did not find any increase after axotomy in Ca 2+ signals in the distal stumps of Sarm1 À/À axons (Figures 2C and 2D;Movie S2), which remained similar to the uninjured neurites (Figures S1D and S1E). In parallel, axonal integrity was preserved for the entire duration of the experiment ( Figures 2C and 2E). NMN administration failed to induce Ca 2+ rise in Sarm1 À/À injured axons (Figures 2C-2E; Movie S2).
Thus, SARM1 is required for NMN to induce Wallerian degeneration and Ca 2+ increase represents a common downstream step. (B) Degeneration index was calculated from three fields per sample in three independent experiments. Mean ± SEM; n = 9; one-way ANOVA followed by Bonferroni post hoc test. (C) Fluorescent images of Sarm1 À/À SCG transected axons expressing GCaMP5 and DsRed2, untreated or treated with 1 mM NMN. Scale bar represents 50 mm. (D) Quantification of Ca 2+ peak intensities. All measurements were normalized to uncut controls. Mean ± SEM; n = 6, two fields in three independent experiments; one-way ANOVA followed by Bonferroni post hoc test, p = 0.3698. (E) Representative traces showing no increase in Ca 2+ responses in distal Sarm1 À/À axons cut and left untreated or cut and treated with 1 mM NMN as shown, in correlation with the percentage of healthy axons. See also Figures S1 and S2.
A Rapid Ca 2+ Increase near the Injury Site Is Not Abolished in Sarm1 À/À , NMN Deamidase, or FK866-Treated Axons An increase in intra-axonal Ca 2+ in proximity to the cut rapidly follows the transection injury, and it is most likely due to an influx from the cut site . This immediate and localized Ca 2+ rise leads to rapid degeneration of a short axonal segment adjacent to the site of cut, an event known as acute axon degeneration (Kerschensteiner et al., 2005). We recorded an early rise in Ca 2+ levels in Sarm1 À/À axons near the cut site, followed by degeneration of short axonal segments. This was also observed in wild-type controls untreated or treated with FK866 or with FK866+NMN and in NMN deamidase-expressing axons ( Figure S2) and it is consistent with direct influx of Ca 2+ from the injury site. This suggests that Sarm1 À/À and NMN deamidase axons, as well as axons treated with FK866, retain the capability to degenerate in response to Ca 2+ , but that this early, proximal Ca 2+ rise is distinct from the delayed increase in the distal stump that occurs just before frank fragmentation ( Figure 1).
Inhibiting NMN Synthesis with FK866 Delays Injury-Induced Intra-axonal Ca 2+ Increase In Vivo To assess the relationship between NMN and intra-axonal Ca 2+ in vivo, we used zebrafish larvae 48-54 hours postfertilization (hpf), transiently expressing DsRed along with the calcium sensor GCaMP-HS in trigeminal and Rohon-Beard somatosensory neurons. These were treated with FK866 for 6 hours prior to two-photon laser axotomy ( Figures 3A and 3B). As we previously reported, FK866 strongly delays axon fragmentation in this model organism (Figures 3Bm-3Bo and 3C;(Di Stefano et al., 2014)). Consistent with our in vitro observations, we detected a marked Ca 2+ increase within axons distal to the injury site (Figures 3Be-3Bh and 3D). The first signs of axon damage appeared as soon as 4 min after Ca 2+ rise ( Thus, NMN synthesis inhibitor FK866 blocks the increase in Ca 2+ that immediately precedes axonal fragmentation in vivo in a vertebrate model organism. The Extracellular Environment Is the Main Source of NMN-Induced Intra-axonal Ca 2+ Rise Because both extracellular influx of Ca 2+ and its release from intracellular stores have been implicated in Wallerian degeneration (Wang et al., 2012), we asked what is the primary source of the Ca 2+ rise induced by NMN. SCG explant axons were cut and immediately treated with FK866. In order to exclude intracellular conversion of NMN to NAD mediated by any residual axonal NMNAT2, we added NMN to FK866-protected neurites 12 hr after cutting ( Figure 4A), a time when little or no NMNAT2 remains . In these conditions, NMN induces complete degeneration of FK866-protected axons within 3 hr from its addition (Figures 4Be-4Bl and 4C;(Di Stefano et al., 2014)). We used EGTA to chelate extracellular Ca 2+ and saw a remarkable delay of axon degeneration after NMN addition (Figures 4Bm-4Bp and 4C). Similar results were obtained using L-type Ca 2+ channels blockers verapamil (Figures 4Bq-4Bt and 4C) and nifedipine  and S3B). In addition, also KB-R7943, an inhibitor of the reverse Na + /Ca 2+ exchanger, significantly delayed NMNinduced degeneration, although its effect was weaker (Figures 4Bu-4Bx and 4C). EGTA was protective also in dorsal root ganglion (DRG) cultures ( Figure S4).
A release of Ca 2+ from axonal ER, leading to mPTP opening and release of intra-mitochondrial Ca 2+ has been observed in the execution phase of Wallerian degeneration (Villegas et al., 2014). Therefore, we tested whether this sequence of events is induced by NMN. We observed that ryanodine receptor antagonist ryanodine significantly delayed NMN-induced degeneration, although its effect was weaker than that obtained with EGTA and the other Ca 2+ -channel blockers used  and S3B; Figures 4B and 4C). We also found that cyclosporine A and ruthenium red, which block mPTP opening (Barrientos et al., 2011), more modestly influence the rate of NMN-induced axon degeneration in our model ( Figures S3C (i-p) and S3D). These results indicate that the main source of intra-axonal Ca 2+ stimulated by NMN is the extracellular environment, with a weaker contribution from intracellular organelles.
To investigate the temporal relationship between Ca 2+ influx and its internal release and to detect any Ca 2+ increase in areas surrounding mitochondria, Ca 2+ buffering organelles which have been proposed to release Ca 2+ during Wallerian degeneration Barrientos et al., 2011;Villegas et al., 2014), we microinjected SCG neurons with GCaMP5 and the mitochondrial marker TagRFP-mito. We then cut the axons and immediately added FK866 and NMN. As expected, NMN administration induced a diffuse intra-axonal Ca 2+ increase in FK866-treated axons, rapidly followed by the formation of Ca 2+ -rich axonal swellings which started to appear as early as 10 min after Ca 2+ rise (Figures 4Da, 4Db, 4Dd, and 4E; Figure S5A; see also Figure 1). Eighty percent of these swellings contained mitochondria (Figures 4Db and  4Dd)). The relative Ca 2+ fluorescence intensity inside axonal swellings was significantly higher compared to the signals measured inside adjacent axonal segments (Figures S5A and S5B), suggesting an accumulation of Ca 2+ at the sites of morphological damage. EGTA addition drastically reduced NMN-induced Ca 2+ rise (Figures 4De and 4E) and blocked axon degeneration (Figures 4Bm-4Bp and 4C). However, we still observed a weak increase in Ca 2+ levels confined in some axonal areas where mitochondria were present (Figures 4De and 4E). This localized Ca 2+ rise was considerably weaker than that observed following extracellular Ca 2+ influx (Figure 4E); small swellings also appeared within the same confined axonal regions; nevertheless, frank degeneration of the distal axon stump was delayed for many hours (Figures  4Dg and 4Dh).
Together, these results indicate that NMN stimulates a late influx of Ca 2+ from the extracellular environment. A more modest Ca 2+ release from intracellular stores also occurs, but this is not sufficient to trigger fragmentation.

Changes in Mitochondrial Membrane Potential and Morphology Mark Different Stages of Axonal Degeneration
Changes in mitochondrial function and structure have been previously shown to occur after injury (Sievers et al., 2003;Villegas et al., 2014); however, it is still unclear whether mitochondria are major players, modest regulators, or simply markers of degeneration. As we previously reported early changes in mitochondrial membrane potential, an index of functionally active mitochondria, in injured neurites of young DRG cultures (Sievers et al., 2003), we asked whether these changes also occur in our SCG model of NMN-induced degeneration. Surprisingly, using the mitochondrial membrane potential indicator tetramethylrhodamine methyl ester perchlorate (TMRM), we found that mitochondrial membrane potential remained stable for at least 4 hr after axotomy, a time point where axons still appear intact in SCG explant cultures. Similar results were observed after FK866 treatment and when NMN was re-added to FK866treated axons (Figures 5A and 5B). A visible decrease in TMRM signal only appeared 6 hr after injury, when clear signs of degeneration were already present. FK866 abolished the drop in TMRM signal visible at this time point, which was then re-established by NMN re-addition ( Figures S5C and S5D). We also found that SARM1 deletion delayed the drop in TMRM signal after injury, which remained stable up to 32 hr after axotomy ( Figures S5E and S5F). (C) Degeneration index was calculated from three fields per sample in three independent experiments (mean ± SEM; n = 9; one-way ANOVA followed by Bonferroni post hoc test; ****p < 0.0001; **p < 0.01). (D) Fluorescent and bright-field images of wild-type SCG axons expressing GCaMP5 and TagRFP-mito. Neurites were cut, treated with 100 nM FK866 + 1 mM NMN (a-d) or with 100 nM FK866 + 1 mM NMN + 5 mM EGTA (e-h) and imaged at the indicated time points. Arrows in (b) and (d) indicate mitochondria colocalizing with axonal swellings. Localized increases in Ca 2+ levels after EGTA treatment were confined in some of the axonal areas where mitochondria were present (red box and arrowheads in e-h). The enlargement represents the boxed area in panel (e). Scale bar represents 10 mm. (E) Quantification of Ca 2+ peak intensities (see Experimental Procedures). All measurements were normalized to uncut controls (mean ± SEM; n = 6 axons, three independent experiments; one-way ANOVA followed by Bonferroni post hoc test; *p < 0.05, **p < 0.01). See also Figures S3, S4, and S5.
Mitochondrial morphological changes have been reported to be among the earliest detectable alterations during Wallerian degeneration (Vial, 1958;Webster, 1962). We thus tested whether NMN accumulation after injury could influence mitochondrial morphology. No significant change was observed among the different treatment groups in the size distribution of mitochondria immediately after cut (Figures 5C and 5D). We found a small but significant increase in the number of shorter mitochondria 4 hr after injury only in the sample treated both with FK866 and NMN. However, no change was detected in the untreated sample, or the sample treated with FK866 alone (Figures 5C and 5D). Conversely, we observed a dramatic reduction in mitochondrial length in Sarm1 À/À axons 48 hr after injury ( Figures 5E and 5F), long before frank fragmentation occurs (Figures 2A and 2B). Interestingly, we also found a significant difference (K-S p = 0.0005) in the size distribution of mitochondria between Sarm1 À/À and wild-type axons immediately after cut (compare Figures 5D and 5F), perhaps reflecting a direct influence of SARM1 on the maintenance of normal mitochondrial shape.

Figure 5. Changes in Mitochondrial Membrane Potential and Morphology Mark Different Stages of Axonal Degeneration
(A) Fluorescent and bright-field (BF) images of wild-type SCG axons, labeled with the mitochondrial membrane potential indicator TMRM (50 nM), at the indicated time points after cut. Where indicated, 100 nM FK866 and 1 mM NMN were added immediately after cut. Scale bar represents 5 mm. (B) Quantification of relative TMRM intensity (see Experimental Procedures); all measurements were normalized to uncut controls to which an arbitrary value of 1 was assigned (mean ± SD; n = 3, one-way ANOVA followed by Bonferroni post hoc test). (C) Fluorescent images of wild-type SCG axons, labeled with the mitochondrial marker mitotracker red, at the indicated time points after injury. Where indicated, 100 nM FK866 and 1 mM NMN were added immediately after cut. Scale bar represents 4 mm. (D) Quantification of mitochondrial length in wild-type SCG axons after cut. The results are expressed as the percentage of mitochondria that fall within the indicated range of length (total number of mitochondria analyzed per condition: cut untreated, 0 hr n = 359, 4 hr n = 350; Cut + FK866, 0 hr n = 410, 4 hr n = 445; cut + FK866 + NMN, 0 hr n = 362, 4 hr n = 395; Kolmogorov-Smirnov test). (E) Fluorescent images of Sarm1 À/À SCG axons, labeled with the mitochondrial marker mitotracker red, at the indicated time points after injury. Scale bar represents 4 mm. (F) Quantification of mitochondrial length in Sarm1 À/À SCG axons after cut. The results are expressed as detailed above (total number of mitochondria analyzed per condition: 0 hr n = 418, 48 hr n = 223; Kolmogorov-Smirnov test). See also Figure S5.
Thus, in this experimental system, a decrease in mitochondrial membrane potential marks the late stages of degeneration. Furthermore, while mitochondrial size is subject to modifications following an injury, these changes do not appear to correlate with the extent of degeneration.

FK866 and SARM1 Deletion Do Not Prevent Early
Changes in Mitochondrial Motility after Cut WLD s was suggested to confer axon protection after injury by maintaining mitochondrial axonal transport, although this is a modest and short-lived effect in some studies O'Donnell et al., 2013). Therefore, we next investigated whether a reduction in mitochondrial motility could underline NMN-induced axon degeneration.
We tested whether blocking NMN synthesis with FK866 had an impact on mitochondrial dynamics after injury. To this purpose, we labeled mitochondria with the mitochondrial marker mitotracker red ( Figure 6A). Consistent with previous reports O'Donnell et al., 2013), we found that axotomy caused a marked reduction in the percentage of motile mitochondria 4 hr after cut in SCG explant cultures ( Figure 6B; Movie S5). We also observed a decrease in the number of motile mitochondria in axons treated with FK866, similar to that in untreated cut axons ( Figure 6B; Movie S5), despite FK866-treated axons maintaining their integrity during this time ( Figure 6C), and for at least another 30 hr (Di Stefano et al., 2014). A similar reduction in mitochondrial motility was also observed in axons treated with both FK866 and NMN (Figure 6B; Movie S5). Interestingly, FK866-treated axons showed an almost complete arrest of mitochondrial transport 24 hr after axotomy, a time point when they still have overtly normal morphology (Movie S6).
We also analyzed mitochondrial transport in Sarm1 À/À SCG neurites, where protection against Wallerian degeneration is even stronger than that achieved with FK866. Remarkably, a substantial reduction in mitochondrial motility was also detected as early as 4 hr after injury, when Sarm1 À/À axons are intact (Figures 6D and 6E; Movie S7); mitochondrial transport completely stopped 48 hr after injury (Movie S8), long before axons fragment (Figures 2A and 2B).
Thus, axon protection by FK866 and SARM1 deletion is not mediated by an effect on mitochondrial dynamics.

Mitochondrial Depolarization Does Not Alter the Rate of Wallerian Degeneration
Next, we asked whether direct damage to mitochondria could affect the rate of Wallerian degeneration. This is possible because damaged mitochondria could release mediators of cell and axon death that could cause the axon demise after injury (Court and Coleman, 2012). Therefore, we treated axons with the protonophore Carbonyl cyanide m-chlorophenyl hydrazone (CCCP), which causes an uncoupling of the proton gradient and disruption of mitochondrial membrane potential (Ly et al., 2003).
When exposed to 50 mM CCCP, mitochondria in axons immediately lost membrane potential, as shown by a remarkable, instantaneous decrease in TMRM fluorescent signal ( Figure 7A). Despite this, frank axonal fragmentation only occurred 32-48 hr after CCCP addition (Figures 7A and 7B). This is a much slower time course than that of Wallerian degeneration, which in our experimental system is completed within 8 hr. This remarkable time difference enabled us to test whether mitochondrial depolarization affects the rate of Wallerian degeneration. We added CCCP to SCG explants immediately prior to axotomy ( Figure 7C) and then followed the time course of axon degeneration. Despite the much earlier, almost instantaneous loss of TMRM signal, cut axons treated with CCCP underwent Wallerian degeneration at the same rate as cut, vehicle-treated axons (Figures 7Da and 7E). Furthermore, CCCP treatment did not affect axon protection conferred by FK866 or reversion by NMN administration after axotomy (Figures 7Db, 7Dc, 7F, and   7G). Similarly, the rate of axon degeneration in transected Sarm1 À/À axons treated with CCCP did not differ from that of the untreated axons, despite the rapid depolarization of Sarm1 À/À mitochondria upon CCCP addiction, as expected ( Figures S6A and S6B).
Together, these results dissociate the loss of mitochondrial function from the onset of Wallerian degeneration. (A) Uninjured wild-type SCG explants were treated with 50 mM CCCP or vehicle (DMSO); 50 nM TMRM was added to confirm full depolarization of axonal mitochondria following CCCP addition. Representative bright-field (BF) and representative fluorescent images of wild-type SCG axons were acquired at the indicated time points after CCCP addition. Scale bars represent 50 and 10 mm. (B) Degeneration index was calculated from three fields per sample in six independent experiments (mean ± SEM; n = 18; one-way ANOVA followed by Bonferroni post hoc test; ****p < 0.0001). (C) Schematic representation of the experimental design (not drawn to scale). Wild-type SCG axons were incubated with 50 nM TMRM to monitor mitochondrial membrane potential and treated with CCCP (added 10 min before axotomy) or DMSO. (D) Representative bright-field and representative fluorescent images of transected wild-type SCG axons treated with CCCP or vehicle (DMSO) (a-c) and with 100 nM FK866 (b) or with 100 nM FK866 + 1 mM NMN (c). The dramatic loss of TMRM signal following CCCP addition confirmed full depolarization of axonal mitochondria. Scale bars represent 50 and 10 mm. (E-G) Degeneration index was calculated from three fields per sample in three independent experiments (mean ± SEM; n = 9; one-way ANOVA followed by Bonferroni post hoc test). See also Figure S6.

DISCUSSION
Our proposed model of Wallerian degeneration predicts that axonal integrity is lost as a consequence of the accumulation of the pro-degenerative molecule NMN (Di Stefano et al., 2014), whose levels in axons are normally limited by the NADbiosynthetic enzyme NMNAT2 . Here, using pharmacological and genetic experimental approaches in in vitro mammalian primary neurons and an in vivo vertebrate model organism, we show that NMN initiates a Ca 2+ -mediated execution program of axon degeneration after injury, which requires the presence of the recently identified pro-axon death protein SARM1. Furthermore, inhibition of NMN synthesis and SARM1 deletion both fail to prevent early alterations in mitochondrial motility after cut and disrupting mitochondrial membrane potential does not affect the rate of injuryinduced axon degeneration; thus, a change in mitochondrial homeostasis appears to be a marker of Wallerian degeneration rather than a causative event.
Although our NMN-induced axon degeneration model is supported by pharmacological and genetic studies in vitro and in vivo (see Figures 1 and 3 and Di Stefano et al., 2014), some recent studies find instead a beneficial effect of NMN, explained by its intra-axonal conversion to NAD (Wang et al., 2015). The difference in our experimental settings (mouse intact explants instead of rat dissociated neurons) precludes a direct comparison. However, in our hands, NMN shows a rapid pro-degenerative effect when added up to 12 hr (SCG explants) or 24 hr (DRG explants) post-axotomy (see Figure 4, Figure S4, and Di Stefano et al., 2014); at these time points, NMN conversion to NAD is prevented by the complete degradation of NMNAT2, whose activity could remain in the experiments performed by (Wang et al., 2015). Careful comparison studies will need to address these discrepancies to reconcile these apparent contradictions.
Intra-axonal Ca 2+ increase after axotomy follows a biphasic pattern, showing a first rise soon after and near the site of injury followed by second, delayed one along the whole axon length; however its temporal relationship with axonal fragmentation is still debated Avery et al., 2012;Gerdts et al., 2011;Villegas et al., 2014;Yang et al., 2013). Our finding that NMN induces a late intra-axonal Ca 2+ increase after injury, which is rapidly followed by axonal degeneration, are consistent with a model in which the late Ca 2+ rise is the central and causative event marking the execution phase of degeneration on which several other mediators may converge.
Consistent with a recent report that SARM1 is required for degeneration downstream of NMNAT2 loss (Gilley et al., 2015), our data suggest that Ca 2+ lies downstream of both NMN and SARM1 in a main, unique axon death pathway ( Figure S6C; Conforti et al., 2014). A recent study proposed that SARM1 acts downstream of intra-axonal Ca 2+ increase (Summers et al., 2014). The early Ca 2+ rise has been suggested as a causative step in Wallerian degeneration . Our data, however, indicate that the early Ca 2+ increase in the proximity of the cut site and its associated acute degeneration occur indistinguishably in wild-type rapidly degenerating axons as well in Sarm1 À/À , NMN deamidase and FK866-treated axons, all condi-tions where Wallerian degeneration is dramatically delayed. Similarly, WLD s does not prevent the early rise of Ca 2+ near the cut site . Furthermore, Ca 2+ blockers delay degeneration induced by NMN in axons injured as long as 12 hr earlier but maintained intact by FK866. Crucially, WLD s -sensitive Wallerian degeneration can be induced without injury in vitro or in vivo by NMNAT2 selective depletion Hicks et al., 2012) and does not require an influx of Ca 2+ from the cut site to initiate it. All these observations dissociate initial Ca 2+ influx from degeneration.
Two ways of preventing Ca 2+ influx from the extracellular environment confer strong protection from NMN-induced axon degeneration: (1) sequestering extracellular Ca 2+ with EGTA, and (2) blocking Ca 2+ influx via Ca 2+ channels. While strongly implicating extracellular Ca 2+ in the late execution stages of Wallerian degeneration, these observations suggest that the Ca 2+ sensitive step could be at a point where morphologically protecting axons may not restore their function. Therefore, upstream mediators of the axon death pathway may prove therapeutically more useful targets.
Extracellular Ca 2+ influx may be due to a dysregulation of Ca 2+ channels following a depolarization of the axonal membrane, which itself could be a consequence of energy depletion (Mishra et al., 2013;Wang et al., 2005). This could be triggered by activation of a MAPK signaling pathway by SARM1 (Yang et al., 2015) or even by a downstream effect on NAD levels (Gerdts et al., 2015). Consistent with previous reports (Barrientos et al., 2011;Villegas et al., 2014), our data also indicate a small contribution of intracellular organelles to the total intraaxonal Ca 2+ rise after injury, noticeable following EGTA treatment ( Figures 4D and 4E). Despite being insufficient to trigger fragmentation on its own in the rapid timescale of Wallerian degeneration, this could nevertheless contribute to the degeneration process. Our observation that mitochondria localize at the sites of Ca 2+ accumulation suggests they could be one of the sources of this weaker Ca 2+ increase. Consistently, we found a protective effect when interfering with mPTP opening ( Figures S3C and S3D), even if more modest to what previously reported ex vivo (Barrientos et al., 2011), where Schwann cells and, more specifically, an effect of the pharmacological treatments on Schwann cell mitochondria, could have a stronger influence on axonal survival.
Whether mitochondria play a central role in Wallerian degeneration, and whether reported changes in mitochondrial shape, dynamics, and membrane potential Sievers et al., 2003;Villegas et al., 2014) are a cause or consequence of degeneration or are general markers of compromised axons remain unclear. In our model of NMN-induced axon degeneration, we find that mitochondrial membrane potential drops after cut, but only when morphological damages to axons are already present, and may even be a consequence of Ca 2+ influx. Crucially, disrupting mitochondrial membrane potential before injury does not alter the rate of degeneration or the effect of FK866 and SARM1 deletion. Consistent with our data, Wallerian degeneration occurs also in axons devoid of mitochondria and WLD s protection is independent of axonal mitochondria (Kitay et al., 2013;Rawson et al., 2014). The depletion of other axonal proteins after injury may explain the effects on mitochondrial homeostasis. For instance, SCG10, which is involved in the regulation of mitochondrial dynamics, is also rapidly lost after injury and its overexpression confers a modest protective phenotype (Shin et al., 2012), raising the possibility that mitochondrial changes may follow the depletion of such a regulator. In addition, mitochondrial arrest could also be a response to local Ca 2+ changes (Wang and Schwarz, 2009). Mitochondria depend on the integrity of a nuclear-mitochondrial communication pathway regulated by NAD metabolism (Fang et al., 2014), and could be affected when this control is lost as consequence of the injury. Intriguingly, axonal mitochondria appear shorter and more static in the absence of SARM1. While supporting that Wallerian degeneration is independent of mitochondrial shape and motility, these observations raise a series of questions on the role of SARM1 in mitochondria. Indeed, SARM1 contains an N-terminal mitochondrial localization signal and is, at least in part, associated with mitochondria (Osterloh et al., 2012). SARM1 mitochondrial localization may be dispensable for the execution of Wallerian degeneration; however, it is required for viral-induced cell death (Mukherjee et al., 2013). The relevance of mitochondrial homeostasis in SARM1 biological function therefore needs to be further explored.
In our hands treatment of axons with CCCP to induce mitochondrial depolarization does not affect the rate of Wallerian degeneration, but it does initiate a slower axonal degeneration process in intact axons. A previous study showed that CCCP toxicity in uninjured axons was insensitive to WLD s (Ikegami and Koike, 2003), which may appear in contrast to our result. However, we used CCCP at a lower concentration and, most importantly, in the context of axotomy where CCCP action could differ. Intriguingly, other reports show that mitochondrial dysfunctions activate a pathway of degeneration sensitive to regulators of Wallerian degeneration (Press and Milbrandt, 2008;Summers et al., 2014). This mechanism could be different from the axon degeneration induced by depletion of axonal mitochondria (Fang et al., 2012;Rawson et al., 2014), which in contrast is insensitive to NMNAT/WLD s pathway (Fang et al., 2012). Furthermore, in C. elegans mitochondrial and transport defects leading to axon degeneration are observed as a consequence of a-tubulin acetyltransferase gene mec-17 loss (Neumann and Hilliard, 2014). Explaining the mechanistic links between mitochondrial dysfunction-induced axon degeneration and Wallerian degeneration will be important in view of the causative role of mitochondrial dysfunction in a variety of neurodegenerative disorders characterized by axonal degeneration (Conforti et al., 2014).
In summary, our data identify crucial steps of an axon death pathway initiated by NMN accumulation. We show that NMN requires SARM1 to activate a destruction process where extracellular Ca 2+ influx is the main executor ( Figure S6C). We reveal an additional, minor contribution of mitochondrial Ca 2+ release but our data do not support a central role for mitochondria in NMN-induced axon degeneration. Because Ca 2+ appears to be temporally distant from the point of NMN accumulation, leaving time for signaling events, it is now important to clarify what are these signals and how NMN interacts with SARM1. These studies should help identifying effective targets to delay axon degeneration in injury and disease.

EXPERIMENTAL PROCEDURES
All studies conformed to the institution's ethical requirements in accordance with the 1986 Animals (Scientific Procedures) Act.
Explant and Dissociated Cell Cultures C57BL/6 or CD1 (referred to as wild-type, Charles River, UK) and Sarm1 À/À (kindly provided by Dr. Michael Coleman, the Babraham Institute) mouse SCG explants, were dissected from P0-2 mouse pups and DRG explants were dissected from E16 mouse embryos. Explants were cultured for 6-7 days and then axotomized as previously described (Di Stefano et al., 2014) and as detailed in the Supplemental Experimental Procedures.
Wild-type and Sarm1 À/À SCG dissociated neurons were obtained as described (Di Stefano et al., 2014) and as detailed in Supplemental Experimental Procedures. Neurites were allowed to extend for 3 days in culture before microinjection and were axotomised the following day using a scalpel. We refer to time 0 as the first time point acquired immediately after cut. In our experimental system (Di Stefano et al., 2014), neurites originating from wild-type SCG explants degenerate within 8 hr after cut. For the purpose of comparison, no morphological changes occur in wild-types within the first 4 hr after cut; clear morphological damages and frank fragmentation occur at 6-8 hr after cut. Neurites originating from dissociated cells degenerate within $3 hr after cut; this difference could be due to the different time in culture (Buckmaster et al., 1995).

Microinjection
Dissociated cells were microinjected as described in (Di Stefano et al., 2014) and in the Supplemental Experimental Procedures.

Ca 2+ Imaging and Quantification
Wild-type and Sarm1 À/À dissociated SCG neurons were cut and then imaged in a controlled environment (37 C and 5% CO 2 ). Treatments were applied 30 min before cut or immediately after cut and time-lapse fluorescent images and bright-field images were acquired analyzed and quantified as detailed in the Supplemental Experimental Procedures.

Ca 2+ Imaging and Quantification in Zebrafish Larvae
Zebrafish larvae were grown, transfected, axotomised, and imaged as described (Di Stefano et al., 2014) and as detailed in the Supplemental Experimental Procedures.

SCG and DRG Explant Treatment and Acquisition of Bright-Field Images
Wild-type SCG neurites were cut and drugs were administered as described in (Di Stefano et al., 2014) and in Supplementary Methods. Bright field images were acquired and analyzed as detailed in the Supplemental Experimental Procedures.

Mitochondrial Motility Imaging and Quantification
Wild-type and Sarm1 À/À SCG explants were pre-incubated with 100nM Mitotracker red CMXRos (Invitrogen) for 15 min at 37 C before axotomy. Drug treatment and image acquisition is detailed in the Supplemental Experimental Procedures.
Quantification of mitochondrial motility was performed using the ImageJ plugin Difference tracker (Andrews et al., 2010). Differences in contrast between each movie were minimized by applying automated contrast normalization to all images using ImageJ.

Mitochondrial Membrane Potential and Length Quantification
For mitochondrial membrane potential, wild-type and Sarm1 À/À SCG neurons were incubated with the cell-permeant dye TMRM (VWR International) for 30 min at 37 C before axotomy and imaging. When used, 100 nM FK866 and 1mM NMN were administered immediately after axotomy. For mitochondrial length, wild-type and Sarm1 À/À SCG explants were incubated with 100 nM Mitotracker red CMXRos (Invitrogen) for 15 min at 37 C before cutting. Drug treatment, image acquisition and image analysis are detailed in the Supplemental Experimental Procedures.

CCCP Treatment
Uncut and cut wild-type or cut Sarm1 À/À SCG neurons were treated with 50 mM CCCP or vehicle (DMSO) just prior to imaging. To check that CCCP treatment was triggering full depolarization of axonal mitochondria neurons were incubated for 30 min at 37 C with TMRM.

Statistical Analysis
Data are expressed as mean ± SEM or SD. Statistical analysis was performed using ANOVA, Kolmogorov-Smirnov test, or Student's t test with p values less than 0.05 being considered significant for any set of data.