ROS regulation of axonal mitochondrial transport is mediated by Ca2+ and JNK in Drosophila

Mitochondria perform critical functions including aerobic ATP production and calcium (Ca2+) homeostasis, but are also a major source of reactive oxygen species (ROS) production. To maintain cellular function and survival in neurons, mitochondria are transported along axons, and accumulate in regions with high demand for their functions. Oxidative stress and abnormal mitochondrial axonal transport are associated with neurodegenerative disorders. However, we know little about the connection between these two. Using the Drosophila third instar larval nervous system as the in vivo model, we found that ROS inhibited mitochondrial axonal transport more specifically, primarily due to reduced flux and velocity, but did not affect transport of other organelles. To understand the mechanisms underlying these effects, we examined Ca2+ levels and the JNK (c-Jun N-terminal Kinase) pathway, which have been shown to regulate mitochondrial transport and general fast axonal transport, respectively. We found that elevated ROS increased Ca2+ levels, and that experimental reduction of Ca2+ to physiological levels rescued ROS-induced defects in mitochondrial transport in primary neuron cell cultures. In addition, in vivo activation of the JNK pathway reduced mitochondrial flux and velocities, while JNK knockdown partially rescued ROS-induced defects in the anterograde direction. We conclude that ROS have the capacity to regulate mitochondrial traffic, and that Ca2+ and JNK signaling play roles in mediating these effects. In addition to transport defects, ROS produces imbalances in mitochondrial fission-fusion and metabolic state, indicating that mitochondrial transport, fission-fusion steady state, and metabolic state are closely interrelated in the response to ROS.


Introduction
Mitochondria perform functions that are critical for neuronal survival, such as ATP production and Ca 2+ homeostasis. However, the mitochondrial electron transport chain is also a major source of reactive oxygen species (ROS) production [1,2]. ROS are generated from incomplete reduction of oxygen, and include the superoxide anion (O 2 -), hydrogen peroxide (H 2 O 2 ) and the hydroxyl radical (HOÁ). Under physiological conditions, ROS serve as important signaling molecules [3,4]. However, excess ROS induce oxidative stress that harms cells PLOS

Results
Mitochondrial transport in axons is reduced in response to ROS Impairment of the axonal transport of mitochondria has been shown to result in neurodegenerative diseases [11,14]. However, we know little about how ROS production might circle back to regulate mitochondrial transport, an issue that may be important in the pathophysiology of neurodegeneration. Although diminished overall mitochondrial transport under oxidative stress conditions has been shown in cultured neurons [25], it remains unknown whether oxidative stress affects mitochondrial transport in in the more complex in vivo environment, in which homeostatic control of ROS is probably more robust and physiological pO 2 is lower.
To test whether ROS affect mitochondrial transport in axons in vivo, we quantified all parameters of mitochondrial transport in 3rd instar Drosphila larvae expressing mito-GFP in motor neurons and treated with 20 mM paraquat for 24 hrs [27]. Using larval preparation as we have previously described (Shidara et al, 2010;Devireddy et al, 2014), we observed mitochondrial traffic in axons still connected to their cell bodies and synapses in the central nervous system. In axons, moving mitochondria are categorized into anterograde and retrograde populations by their dominant directions, which are easily discernible despite pauses and brief reversals of direction ( Fig 1A). We first quantified flux, which is a gross indicator of movement representing how many mitochondria pass a fixed point per unit time. We found reduced organelle flux in both the anterograde and retrograde populations after paraquat treatment ( Fig 1B). To parse this effect further and determine the possible components producing the reduced flux, we examined more specific parameters of mitochondrial motility: velocity, duty cycle, run length, percentage of moving mitochondria, and density [28]. Under oxidative stress conditions, we found reduced instantaneous velocity in both directions (Fig 1C). In addition, we found a modestly-reduced duty cycle in the retrograde direction (Fig 1D), indicating that retrograde mitochondria spent less time moving in the dominant direction. Retrograde run length was also modestly reduced, perhaps as a consequence of the reduced velocity and duty cycle ( Fig 1E). Moreover, an increased percentage of stationary mitochondria ( Fig 1F) also contributed to the reduced flux. Mitochondrial density is not necessarily directly linked to motility, but reduced density can of course contribute to reduced flux [28]. However, we found no change in the mitochondrial density of segmental nerve axons under oxidative stress conditions (Fig 1G), indicating that mitochondrial motility itself is reduced under these conditions in vivo.

Overexpression of SOD attenuates ROS-induced defects of mitochondrial transport
To further ensure that this decreased mobility was specifically caused by oxidative stress, we took advantage of the observation that SOD overexpression has been shown to relieve oxidative stress in Drosophila [29,30]. We found that the reduced mitochondrial flux and velocity in both directions produced by paraquat treatment were rescued by SOD1 or SOD2 overexpression (Fig 1B and 1C). In the retrograde direction, both reduced duty cycle and run length were also rescued (Fig 1D and 1E). In addition, the defects in the percentage of moving mitochondria produced in response to ROS were fully rescued by SOD1 or SOD2 overexpression ( Fig  1F). These results indicate that a reduction of the excess ROS levels can diminish or eliminate the defects in mitochondrial transport. On the other hand, the axonal mitochondrial density was comparable to controls in larva with SOD1 or SOD2 overexpression alone (Fig 1G), suggesting that mitochondrial biogenesis and/or the anterograde-retrograde traffic steady state are not affected by the overexpression. Altogether, these results indicate that excess ROS do cause specific defects in the axonal transport of mitochondria in vivo.
Transport of neuropeptide-bearing large dense core vesicles (DCVs) is barely affected by the presence of ROS To test whether the defects of transport in response to ROS are specific to mitochondria, we expressed ANF-GFP, targeted to the lumen of neuropeptide-bearing large DCVs, in motor neurons [31], and examined their transport under oxidative stress conditions. Detailed parameters of DCV transport, including velocity, duty cycle, run length, and density were analyzed (Fig 2). All parameters of DCV transport in response to ROS were comparable to controls ( Fig  2B-2D), indicating that transport of DCVs is barely affected by oxidative stress. In addition, Anterograde transport is toward the right; retrograde transport is toward the left. (B) Paraquat treatment produces a decrease in mitochondrial flux, which is rescued by overexpression of SOD1 or SOD2 in both directions. (C) Velocity is reduced with paraquat treatment, and this is rescued by SOD1 or SOD2 overexpression in both directions. (D) Paraquat treatment shows a small reduction in retrograde duty cycle with an increase of pause time and a decrease of moving time; overexpression of SOD1 shows a small increase of pause time. (E) Retrograde run length is modestly reduced with paraquat treatment. (F) Paraquat treatment shows an increase of the percentage of stationary mitochondria, which is rescued by SOD overexpression. The percentage of anterograde moving the axonal DCV density was similar to controls (Fig 2E), implying an unchanged steady state of DCV biogenesis, transport, and degradation. Altogether, these data indicate that oxidative stress impedes mitochondrial transport more specifically, rather than the general axonal transport of organelles.

ROS reduce mitochondrial transport by the elevation of Ca 2+ levels
Elevated intracellular Ca 2+ levels decrease mitochondrial transport by binding the EF domain on Miro [15][16][17]. In addition, since there are significant interactions between Ca 2+ and redox signaling [32][33][34], we asked whether the defects in mitochondrial transport caused by oxidative stress might be mediated by the elevation of intracellular Ca 2+ levels. To test this, we used primary neuronal cells as our system since experimental manipulation of Ca 2+ levels in vivo is difficult.
mitochondria increases with SOD2 overexpression. (G) Mitochondrial density is comparable to control with paraquat treatment or SOD overexpression. The number of larvae analyzed is shown on the bars. Error bars indicate mean ± SEM. Significance is determined by one-way ANOVA with Bonferroni's post-test. *p<0.05, **p < 0.01, and ***p < 0.001.
https://doi.org/10.1371/journal.pone.0178105.g001 We first ensured that mitochondrial transport was also reduced in response to ROS in our in vitro system in Drosophila. To assess this, primary neuron cell cultures obtained from brains and ventral ganglia of third instar larvae expressing mitochondrially-targeted GFP in motor neurons were treated with 100 μM H 2 O 2 for 1 hr. In control cells,~40% of mitochondria were mobile, whereas the percentage of moving mitochondria in H 2 O 2 -treated cells decreased significantly (Fig 3). Interestingly, mitochondrial ROS did not show significant increase after 1 hr using ROS indicators roGFP expressed in mitochondria. Instead, mitochondrial ROS levels did increase after 2 hrs, but the motility defects did not become worse (S1 Fig). These data show that mitochondrial ROS are not required to induce motility defects, suggesting that cytosolic ROS are sufficient to induce motility defects. In addition, similar to the results in vivo (Fig 1), the defects in the percentage of moving mitochondria produced in response to ROS were fully rescued by SOD1 or SOD2 overexpression (Fig 3), indicating that oxidative stress impair mitochondrial transport both in vitro and in vivo.
Next, we expressed the genetically encoded Ca 2+ indicator GCaMP6 in motor neurons using the D42-Gal4 driver and determined the response of Ca 2+ levels to ROS [35]. We found that cells with H 2 O 2 treatment for 1 hr showed about 50% higher GCaMP6 intensity compared to controls (Fig 4A and 4B), indicating that excess ROS can induce elevated intracellular Ca 2+ levels. We then examined whether elevated Ca 2+ levels decrease mitochondrial transport and/ or affect DCV transport. We treated cells with the Ca 2+ ionophore, ionomycin, which produced a dramatic increase in Ca 2+ levels ( Fig 4A and 4B), and examined both mitochondrial transport and DCV transport ( Fig 4C). Notably, dramatically increased Ca 2+ levels only affected mitochondrial transport (Fig 4C and 4D). When we examined DCV transport, we found no significant difference in velocity of DCV transport between control and ionomycintreated cells (Fig 4C and 4E). Although mitochondrial transport was reduced in ionomycintreated cells, Ca 2+ levels were much higher than physiological or ROS-treated levels. Thus, we used thapsigargin, an ER Ca 2+ ATPase inhibitor that increases only intracellular Ca 2+ levels, to mimic Ca 2+ levels in response to ROS (Fig 4A and 4B). Under these conditions, the percentage of moving mitochondria decreased to a level similar to H 2 O 2 -treated cells (Fig 4F and 4G), and velocity of DCV transport was comparable to controls (Fig 4F and 4H). Thus, elevated Ca 2+ levels reduced mitochondrial transport, as expected [15][16][17] and did so more specifically.
These data (Fig 4A-4H) support the hypothesis that ROS decrease mitochondrial transport via the elevation of Ca 2+ levels. To test this directly we treated cells with Ca 2+ -free medium containing EGTA and exposed them to H 2 O 2 , to examine whether the defects of mitochondrial transport caused by ROS would be rescued. First we confirmed that the elevated Ca 2+ levels caused by H 2 O 2 treatment were reduced in Ca 2+ -free medium with EGTA using the GCaMP6 indicator (Fig 4I and 4J). Next, we examined mitochondrial transport under these conditions. In the Ca 2+ -free medium containing EGTA, there was no increase of the percentage of moving mitochondria above normal (Fig 4K and 4L) since the Ca 2+ levels are comparable to controls (Fig 4I and 4J). Moreover, under ROS treatment with Ca 2+ held to control levels, the ROS-induced reduction of moving mitochondria was partially rescued (Fig 4K and  4L), indicating that ROS-induced defects of mitochondrial transport are mediated in large part by the elevation of Ca 2+ levels.
Since EGTA treatment, rescuing Ca 2+ to physiological levels, rescued ROS-induced defects of mitochondrial transport, we wondered whether further reduction of Ca 2+ levels could increase mitochondrial transport. To examine this, we first treated cells with BAPTA-AM, which is a Ca 2+ chelator that can cross the cell membrane. However, we verified that the BAPTA-AM only was not enough to chelate both extracellular and intracellular Ca 2+ by treating cells with BAPTA and ionomycin (S2 Fig). Therefore, we treated both BAPTA and EGTA to chelate both extracellular and intracellular Ca 2+ . Under these conditions, intracellular Ca 2+ levels were reduced compared to controls (Fig 5A and 5B). In addition, under oxidative stress conditions, Ca 2+ levels decreased in the Ca 2+ -free medium containing BAPTA-AM and EGTA (Fig 5C and 5D), indicating that combined intracellular and extracellular Ca 2+ chelation indeed reduces Ca 2+ levels lower than physiological levels. Surprisingly, mitochondrial transport was decreased by treatment with both BAPTA-AM and EGTA ( Fig 5C and  5D), indicating that Ca 2+ levels that are below physiological levels can also inhibit mitochondrial transport. All together, these data suggest that modulation of Ca 2+ levels mediates the effect of ROS on mitochondria, and emphasize that Ca 2+ homeostasis is important for mitochondrial transport.
The JNK pathway plays a role in the regulation of mitochondrial transport in axons by oxidative stress The involvement of Ca 2+ suggests that the effects of ROS exposures on mitochondrial transport probably reflect not just generalized oxidative damage, but a disturbance of normal ROSbased signaling. To investigate this, we examined the JNK pathway, since it is both activated under oxidative stress conditions [6,36] and involved in regulating axonal transport [20,37]. Previous studies have shown that activation of the JNK pathway causes phosphorylation of kinesin heavy chain (Khc), which causes the kinesin motor to release from microtubules, and thus reduces anterograde axonal transport in squid axoplasm [20]. We tested the hypothesis To examine this, we activated the JNK pathway by overexpressing Hep B2 (JNK kinase in Drosophila), or down-regulated it by RNAi knockdown of Bsk (JNK in Drosophila). We expected that activation of the JNK pathway would decrease mitochondrial transport and down-regulation of the JNK pathway would increase it. Indeed, we found that anterograde flux was reduced by JNK activation. In addition, flux was not further reduced in the presence of JNK activation by the addition of oxidative stress (Fig 6A and 6B), which implicates the JNK pathway as a downstream effector of the influence of oxidative stress on anterograde mitochondrial flux. We also found that the defects in anterograde flux produced by paraquat treatment were partially rescued by JNK knockdown (Fig 6A and 6B). These data suggested that the JNK pathway plays a role, albeit not exclusive, in mediating the inhibition of mitochondrial anterograde transport by oxidative stress. Activation of the JNK pathway produced reduced anterograde velocity (Fig 6A and 6C), and down-regulation of JNK partially rescued the reduced anterograde velocity that occurred in response to ROS (Fig 6A and 6C), suggesting altered organelle velocity is one of the factors contributing to the difference in flux. For retrograde movement, either activation or down-regulation of the JNK pathway showed reduced mitochondrial flux and velocity under oxidative stress conditions (Fig 6A-6C), suggesting that other regulatory systems might be responsible for the effects of oxidative stress on retrograde traffic. While the regulatory mechanisms are directionally different, optimal levels of activity in the JNK pathway are nonetheless essential for both. Other parameters of mitochondrial motility such as duty cycles, run length, and % of moving mitochondria were only modestly affected by either overexpression or knockdown of the JNK pathway (Fig 6D-6F   our results demonstrate that the JNK pathway, activated by oxidative stress, is involved in the regulation of mitochondrial transport in the anterograde direction, consistent with directionspecific regulation of mitochondrial transport. Since ROS-induced defects of mitochondrial transport are mediated by Ca 2+ and the JNK pathway, we further determined whether Ca 2+ and the JNK pathway have interaction in the ROS regulation of mitochondrial transport. However, neither activation nor down-regulation of the JNK pathway affects intracellular Ca 2+ levels (S3 Fig), indicating that the JNK pathway is not upstream of the Ca 2+ signaling. This suggests that the defects of mitochondrial transport caused by ROS-induced Ca 2+ levels are not mediated by the JNK pathway, and also raises the possibility that Ca 2+ and JNK signaling regulate mitochondrial transport in parallel.

Mitochondrial fission-fusion balance, inner membrane potential, and transport have interrelated response to ROS
The different elements of the mitochondrial life cycle are closely interrelated, and conditions that compromise mitochondrial motility can also alter the mitochondrial metabolic state, diminishing mitochondrial inner membrane potential [11,38]. In turn, disruption of the fission-fusion steady-state by mutation of the mitochondrial fusion protein, mitofusin 2, induces mitochondrial fragmentation as well as decreased mitochondrial transport [22,23,39,40]. To determine whether the reduced mitochondrial transport produced by ROS is accompanied by effects on metabolic state and/or fission-fusion balance, we examined both mitochondrial morphology and TMRM uptake under oxidative stress conditions in vitro. We found that after 1 hr of H 2 O 2 treatment, mitochondrial shapes became more rounded and their lengths were reduced (Fig 7A and 7B). SOD1 or SOD2 overexpression rescued these defects (Fig 7A and  7B), confirming that they are induced by oxidative stress. Mitochondrial membrane potential was assessed using the intensity ratio of TMRM staining between the mitochondria and neighboring cytosol [12,13,28,41]. We found reduced mitochondrial membrane potential in response to H 2 O 2 (Fig 7A and 7C), indicating that mitochondrial metabolic function is disrupted under oxidative stress conditions. However, the diminished mitochondrial membrane potential was not rescued by SOD1 or 2 overexpression (Fig 7A and 7C), suggesting that excess SOD is not sufficient to rescue all mitochondrial defects produced by oxidative stress. Since the main function of SOD is convert O 2 to H 2 O 2 [6], other cellular or mitochondrial ROS scavengers such as catalase or GSH Px might be required to fully rescue these defects. These data indicate that oxidative stress disrupts the mitochondrial fission-fusion steady state and metabolic function and, furthermore, that mitochondrial fission-fusion balance, membrane potential, and mobility have closely interrelated responses to oxidative stress. Since modulation of JNK signaling and Ca 2+ levels can mediate the effects of ROS on mitochondrial transport, we further examined whether JNK or Ca 2+ -modulated mitochondrial transport are also related to mitochondrial length control or membrane potential. Although JNK signaling had mild effects on mitochondrial length and membrane potential (S4 Fig), we found that ROS-induced mitochondrial length changes were rescued by EGTA treatment but were not rescued by both BAPTA-AM and EGTA treatment (Fig 7D and 8E), which was consistent with the transport data (Figs 4K, 4L, 5C and 5D), suggesting that mitochondrial fission-of Bsk in response to paraquat shows a slightly increase of pause and a decrease of moving time. (E) Retrograde run length is modestly reduced by paraquat treatment. Either overexpression of Hep B2 of knockdown of Bsk does not show significant difference compared to controls. (F) Paraquat treatment shows an increase of the percentage of stationary mitochondria, while neither overexpression of Hep B2 nor knockdown of Bsk shows any difference. The number of larvae analyzed is shown on the bars. Error bars indicate mean ± SEM. Significance is determined by one-way ANOVA with Bonferroni's post-test. *p<0.05, **p < 0.01, and ***p < 0.001.
https://doi.org/10.1371/journal.pone.0178105.g006 Representative images of mitochondrial length and membrane potential. Mitochondrial lengths are measured using mitoGFP signals and mitochondrial membrane potential is measured using TMRM staining by the intensity ratio of mitochondrial fluorescence to cytosolic fluorescence. Scale bars indicate 10 μm. (B) Quantitative results of mitochondrial length. ROS treatment shows a decrease of mitochondrial length, which is rescued by SOD1 or SOD2 overexpression. (C) Quantitative results of mitochondrial membrane potential. Mitochondrial membrane potential is reduced under oxidative stress conditions. SOD1 or SOD2 overexpression does not rescue these defects. (D) Representative images of mitochondrial length measured fusion steady state is closely interrelated to transport. Moreover, these data also indicate that Ca 2+ homeostasis is important for both mitochondrial transport and mitochondrial morphology.

Discussion
While excess ROS produce varied damage to cells [1,2], at physiological levels they are important signaling molecules [3,4]. ROS are produced by mitochondria [2,10,42] and also impact them: since they affect mitochondrial transport and quality control processes in vitro [25,[43][44][45], it is important to find underlying mechanisms among these effects. In this study, we have probed the relationship between ROS and mitochondrial transport, and have demonstrated mechanistic connections between these two. We find that: 1) ROS preferentially impair mitochondrial transport, rather than the general axonal transport of organelles; 2) ROS increase neuronal Ca 2+ levels, and subsequently inhibit mitochondrial transport; 3) The JNK pathway plays a role in the regulation of anterograde transport in response to ROS; 4) Mitochondrial fission-fusion balance, membrane potential, and mobility have closely interrelated responses to oxidative stress.
We first found that mitochondrial axonal transport is reduced in response to ROS in vivo, mainly due to reduced velocity (Fig 1). More important, ROS have a more specific effect on mitochondria, as transport of DCVs is barely affected (Fig 2). This is consistent with a previous in vitro study [25], and indicates that mitochondria are a prime target of ROS among the axonal transport cargoes.
What could explain the specific sensitivity of mitochondrial transport to ROS? Unlike axonal vesicles, mitochondria are linked to kinesin motor proteins via a complex of Miro and Milton. Ca 2+ binds to the two EF-domains of Miro, causing a conformational change that disrupts the ability of kinesin to move mitochondria, but interestingly, inhibiting both directions of mitochondrial transport [15][16][17]. Thus, Ca 2+ regulation is a major difference in the axonal transport of mitochondria versus other organelles. The JNK pathway might also affect mitochondrial transport more specifically. Some studies indicate that the JNK pathway affects axonal transport generally via phosphorylation of kinesin or superior cervical ganglion 10 (SCG10), a microtubule binding protein in axons [20,46]. However, in APLIP1 mutants, the Drosophila homologue of a JNK-interacting protein (JIP), transport of vesicles is reduced in both directions, but mitochondrial transport is reduced only in the retrograde direction, suggesting a unique regulatory feature of mitochondrial transport [47]. In addition, the alpha subunit of the heterotrimeric G protein G12 (Gα12) has been shown to stimulate JNK activity [48]. Notably, Gα12 is targeted to mitochondria and itself affects mitochondrial motility [49]. These studies suggest that the JNK pathway exerts some specific effects on axonal transport of mitochondria.
Could the ROS effects on mitochondrial transport be mediated via changes in Ca 2+ levels? We show that excess ROS do increase Ca 2+ levels (Fig 4A and 4B) and that the ROS-impaired mitochondrial mobility is rescued by Ca 2+ chelation (Fig 4I-4L), consistent with ROS acting on mitochondrial transport via Ca 2+ . It is notable that H 2 O 2 -induced Ca 2+ levels are rescued to physiological levels by the extracellular Ca 2+ chelator EGTA (Fig 4I and 4J), suggesting that this Ca 2+ increase results from Ca 2+ influx from the extracellular environment. Several Ca 2+ using the mitoGFP signal. Scale bars indicate 10 μm. (E) H 2 O 2 and/or EGTA/BAPTA reduce mitochondrial length, which is consistent with the results of mitochondrial transport (Figs 4K, 4L, 5C and 5D). The number of cells analyzed is shown on the bars. Error bars indicate mean ± SEM. Significance is determined by one-way ANOVA with Bonferroni's post-test. *p<0.05, **p < 0.01, and ***p < 0.001. https://doi.org/10.1371/journal.pone.0178105.g007 channels on the plasma membrane are regulated by redox status [34]. For example, H 2 O 2 induces Ca 2+ influx through L-type or T-type voltage-dependent Ca 2+ channels (VDCCs) [50,51]. ROS are also known to stimulate Ca 2+ influx through channels involved in receptorinduced Ca 2+ signals such as transient receptor potential (TRP) channels and store-operated Ca 2+ channels (SOC), mediated by Orai channel proteins [52,53]. Thus, we propose that elevated Ca 2+ levels are induced by ROS through activation of Ca 2+ channels on the plasma membrane. In addition, while most studies focus on the effects of elevated Ca 2+ levels on mitochondrial transport [15-17, 54, 55], experimentally decreased intracellular Ca 2+ levels reduce mitochondrial transport (Fig 5), suggesting that optimal Ca 2+ levels are important for mitochondrial transport. This might explain a previous study showing that while the impaired mitochondrial transport caused by MPP + can be rescued by the thio-antioxidant N-acetylcystein (NAC), it is not rescued by a high concentration of EGTA (2.5 mM), probably capable of reducing Ca 2+ levels below physiological levels [56].
Although Ca 2+ regulates mitochondrial transport, ROS-impaired mitochondrial transport is not fully rescued by reducing Ca 2+ (Fig 4A-4D), suggesting that other factors might be involved in ROS regulation of mitochondrial transport. Indeed, several signaling pathways have been implicated in the regulation of axonal transport. Axonal mitochondria are recruited in response to Nerve Growth Factor (NGF), and this recruitment may be regulated via phosphoinositide 3-kinase (PI3K) or MAPK signaling [18]. Glycogen synthase kinase 3 (GSK3) can phosphorylate kinesin light chain (Klc) to inhibit axonal transport, while activation of AKT/ GSK3β stimulates axonal mitochondrial transport [19,21]. Finally, activation of the JNK pathway phosphorylates Khc, reducing the interaction between kinesins and microtubules and inhibiting anterograde axonal transport [20]. Here we show that activation of the JNK pathway reduces anterograde mitochondrial transport, and these defects are rescued by specific knockdowns (Fig 6A-6C). This anterograde regulation might result from the phosphorylation of kinesin by JNK [20]. In the retrograde direction, knockdown of the JNK pathway decreases mitochondrial transport (Fig 6A-6C). Since JIPs are required for stress-induced JNK activation [57][58][59], and mutation of a Drosophila JIP homologue reduces mitochondrial transport only in the retrograde direction [47], it is likely that disruption of the JNK pathway also produces retrograde-specific defects of mitochondrial transport.
How might Ca 2+ and the JNK pathway interact in the ROS regulation of mitochondrial transport? Some studies suggest that elevated Ca 2+ levels activate the JNK pathway, and subsequently induce apoptosis or neurodegeneration [60][61][62]. However, it is likely that the JNK pathway is not simply downstream of Ca 2+ signaling in ROS regulation of mitochondrial transport. ROS could increase Ca 2+ influx by activating Ca 2+ channels [50][51][52][53], and activate the JNK pathway via apoptosis signal-regulating kinase (ASK1) [6,36,63], providing parallel, largely independent pathways of ROS-activated Ca 2+ influx and JNK signaling. Activation or down-regulation of JNK signaling does not affect Ca 2+ levels (S3 Fig), supporting the paralleleffect hypothesis. In addition, Ca 2+ inhibits mitochondrial transport in both directions [15][16][17], but activation of JNK signaling inhibits mitochondrial transport anterogradely (Fig 6). Since Ca 2+ influx and the JNK pathway could be induced by ROS via parallel pathways and produce different effects on mitochondrial transport, it is likely that they regulate mitochondrial transport in parallel.
In addition to elucidating mechanistic connections between ROS and mitochondrial transport, our studies provide evidence that mitochondrial quality control is related to transport. Excess ROS and abnormal Ca 2+ homeostasis not only impair mitochondrial transport (Figs 1, 3, 4 and 5) but also reduce mitochondrial length (Fig 7). In addition, when mitochondrial transport is rescued (Figs 1, 3, 4K and 4L), mitochondrial lengths are also rescued (Fig 7). Previous studies have shown that mutations of Mitofusin reduce mitochondrial length, impair mitochondrial transport and increase Ca 2+ levels in axons [22,23,39], consistent with our findings, and together reinforcing that mitochondrial fission-fusion are highly interrelated. Moreover, both mitochondrial membrane potential and transport are reduced in response to ROS (Figs 3 and 7C), suggesting the association between mitochondrial metabolic activity and transport. This idea is supported by studies showing that some neurodegenerative disease models, such as reduced frataxin expression or PINK1 mutation [12,13], show reduction of both mitochondrial transport and membrane potential.
Due to the complicated interrelation in mitochondrial transport, morphology, and quality control, ROS-induced defects of mitochondrial transport would not be the only factor leading to neurodegeneration. ROS-induced deficits in mitochondrial motility are partially recovered after ROS insult, but the axonal degeneration is irreversible [25], supporting the hypothesis that instead of one factor, axonal degeneration could be caused by the combination of defects of mitochondrial transport, the imbalance of fission/fusion and metabolic state.
In conclusion, we provide evidence that ROS specifically impairs the axonal transport of mitochondria. This ROS-impaired mitochondrial transport is caused by an imbalance of Ca 2+ homeostasis and activation of the JNK pathway. We also show the close interrelationship of mitochondrial transport, fission-fusion balance, and metabolic state in response to ROS. Thus, this study provides probable mechanistic links between ROS and mitochondrial transport, and suggests a new role for ROS in the induction of neurodegeneration.

Preparation of primary neuronal cell culture and dissected larvae for live imaging
For in vitro cell culture, 4 brains and ventral ganglia were dissected from third-instar larvae in Schneider's medium and were incubated in Ca 2+ /Mg 2+ free saline (137mM NaCl, 2.7mM KCl, 0.36mM NaH 2 PO 4 ÁH 2 O, 11.9mM NaHCO 3 , 5.6 mM glucose) with 0.7 mg/ml collagenase for 1 hr at room temperature. Samples were centrifuged at 300 g for 3 min and washed once with Schneider's medium. We triturated the tissue with siliconized pipette tips to disperse it into individual cells, which were then plated on concanavalin A-coated coverslips and incubated in Schneider's medium containing 10% FBS in a humidified chamber at 25˚C for 72 hrs. Oxidative stress was induced in primary cell cultures with 100 μM H 2 O 2 treatment for 1 hr, followed by staining and imaging [25]. Mitochondria were imaged using the 60x oil-immersion objective lens of an upright laser scanning confocal microscope (LSCM, Nikon Eclipse 90i) with a 488 nm laser. To induce oxidative stress in vivo, second to third instar larvae were starved for 3 hrs and then maintained on normal fly food containing 20 mM paraquat for 24 hrs [27]. To image mitochondria or DCVs in vivo, third instar larvae were prepared using previously described methods [13,28,64]. Briefly, third instar larvae were pinned and opened through the cuticle on the dorsal side in HL6 buffer with 4 mM L-glutamate. Fat bodies and muscles were removed exposing the intact nervous system, including ventral ganglia and segmental nerves. The dissected larva was moved to a glass slide and covered with a coverslip using double-sided tape as a spacer. Mitochondria or DCVs in segmental nerves were observed using the 60x oil-immersion objective lens of an upright time-lapse laser confocal microscope (LSCM, Nikon Eclipse 90i) with a 488 nm laser.

Analysis of mitochondrial and DCV transport
Time lapse confocal images of mitochondrial transport were acquired in the middle (segment A4) of larval segmental nerves using LSCM at 1 s intervals for 200 s. For cell culture, the time resolution is 1 frame/s for 120 s. We used the Manual Tracking plug-in with ImageJ software to track individual organelle. The percentage of moving mitochondria was analyzed using the Cell Counter plug-in with Image J software for both in vitro and in vivo experiments. For in vivo studies, four additional parameters were analyzed: flux, velocity, duty cycle, and run lengths. Flux is defined as the number of mitochondria that cross an assigned point in oneminute interval. If the average velocity in one interval is larger than 0.1 μm/s or smaller than -0.1μm/s and the motion is sustained in at least three frames, we considered this a "run". The percentage of moving mitochondria is the fraction of mitochondria that move either anterogradely or retrogradely. Velocity is defined as the average velocity in the total run. Duty cycle is defined as the percentage of time spent moving in a particular direction. Run length is the average distance moved per run [13,28,64]. All kymographs of mitochondrial transport were prepared using Nikon NIS-Elements software [13,28]. For transport of DCVs, time-lapse images were acquired using LSCM with 1 frame/s for 120s in vitro or 200s in vivo. Velocity were analyzed from kymographs using MetaMorph software in vitro. For in vivo analysis of DCV transport, the Manual Tracking plug-in with ImageJ software was used to track individual DCVs.

Measurement of Ca 2+ levels
Neuronal cultures expressing GCaMP6 in motor neurons were washed and incubated in S2 saline (120 mM NaCl, 5 mM KCl, 8 mM MgCl 2 , 2 mM CaCl 2 , 10 mM HEPES). To chelate extracellular Ca 2+ or both extracellular and intracellular Ca 2+ , cells were incubated in Ca 2+ free saline (S2 saline without CaCl 2 ) containing 0.1 mM EGTA or both 0.1 mM EGTA and 10 μM BAPTA-AM for 30 min, respectively [60,65,66]. To increase intracellular Ca 2+ levels, cells were treated with S2 saline containing 10 μM ionomycin or 2 μM thapsigargin [65]. Mean intensity of GCaMP6 fluorescence was thresholded and measured in whole cells with Nikon NIS-Elements software. To examine the effects of Ca 2+ levels on mitochondrial or DCV transport, the percentage of moving mitochondria was determined under conditions described above.

Quantification of mitochondrial density and lengths
Mitochondrial density was determined by assessing the number of mitochondrial pixels contained in fixed lengths of segmental nerves using MetaMorph software. The second frame of the time lapse images was used to reduce background noise and sharpen images [28]. Then images were thresholded, binarized, and density was measured by obtaining the integrated intensities (total number) of white mitochondrial pixels within the nerve. Mitochondrial lengths were measured with Nikon NIS-Elements software. Only non-overlapping mitochondria in neural processes were selected for length measurements.

Measurement of mitochondrial membrane potential
Mitochondrial membrane potential was measured with tetramethylrhodamine methyl ester (TMRM) [12,13,28,41]. Cells were incubated in Schneider's medium with 25 nM TMRM for 10 minutes, and then were imaged in medium with 6.25 nM TMRM to maintain dye equilibrium. The TMRM images were thresholded and the intensity of mitochondrial fluorescence (Fm) was measured with Nikon NIS-Elements software. The fluorescence intensity in cytosol (Fc) was determined in 5 pixel x 5 pixel regions next to the mitochondria. The mitochondrial membrane potential was then expressed as the intensity ratio of mitochondrial fluorescence (Fm) and cytosolic fluorescence (Fc) which is logarithmically related to the membrane potential.

Statistics
All statistical analysis was performed using GraphPad Prism 5 software. All error bars in graphs indicate mean ± SEM. In the in vivo studies, the experimental unit is an individual larva. In the in vitro studies, the experimental unit is a cell. Data were analyzed using a oneway ANOVA with Bonferroni's post-test for multiple group comparisons, paired or unpaired Student's t-test for two group comparisons. In all cases, at least three independent experiments were performed.