Abstract
In neurons, the normal distribution and selective removal of mitochondria are considered essential for maintaining the functions of the large asymmetric cell and its diverse compartments. Parkin, a E3 ubiquitin ligase associated with familial Parkinson's disease, has been implicated in mitochondrial dynamics and removal in cells including neurons. However, it is not clear how Parkin functions in mitochondrial turnover in vivo, or whether Parkin-dependent events of the mitochondrial life cycle occur in all neuronal compartments. Here, using the live Drosophila nervous system, we investigated the involvement of Parkin in mitochondrial dynamics, distribution, morphology, and removal. Contrary to our expectations, we found that Parkin-deficient animals do not accumulate senescent mitochondria in their motor axons or neuromuscular junctions; instead, they contain far fewer axonal mitochondria, and these displayed normal motility behavior, morphology, and metabolic state. However, the loss of Parkin did produce abnormal tubular and reticular mitochondria restricted to the motor cell bodies. In addition, in contrast to drug-treated, immortalized cells in vitro, mature motor neurons rarely displayed Parkin-dependent mitophagy. These data indicate that the cell body is the focus of Parkin-dependent mitochondrial quality control in neurons, and argue that a selection process allows only healthy mitochondria to pass from cell bodies to axons, perhaps to limit the impact of mitochondrial dysfunction.
SIGNIFICANCE STATEMENT Parkin has been proposed to police mitochondrial fidelity by binding to dysfunctional mitochondria via PTEN (phosphatase and tensin homolog)-induced putative kinase 1 (PINK1) and targeting them for autophagic degradation. However, it is unknown whether and how the PINK1/Parkin pathway regulates the mitochondrial life cycle in neurons in vivo. Using Drosophila motor neurons, we show that parkin disruption generates an abnormal mitochondrial network in cell bodies in vivo and reduces the number of axonal mitochondria without producing any defects in their axonal transport, morphology, or metabolic state. Furthermore, while cultured neurons display Parkin-dependent axonal mitophagy, we find this is vanishingly rare in vivo under normal physiological conditions. Thus, both the spatial distribution and mechanism of mitochondrial quality control in vivo differ substantially from those observed in vitro.
Introduction
Neurons are large, morphologically asymmetric and functionally compartmentalized cells. For this reason, they rely critically on robust, organized axonal transport to maintain the normal distribution of organelles among the somatodendritic, axonal, and synaptic compartments (Hirokawa et al., 2010; Saxton and Hollenbeck, 2012). Mitochondria play an important role in neuronal function and survival, as they supply ATP, buffer cytosolic calcium, and generate reactive oxygen species, and thus their transport and redistribution within neurons are particularly important (Saxton and Hollenbeck, 2012). They display a unique and complex life cycle that is characterized by bidirectional movement (Stowers et al., 2002; Guo et al., 2005; Glater et al., 2006; Pilling et al., 2006; Reis et al., 2009), morphological changes (Okamoto and Shaw, 2005; Chan, 2006; Song et al., 2009), and complex biogenesis and degradation (Davis and Clayton, 1996; Amiri and Hollenbeck, 2008; Maday et al., 2012; Ashrafi et al., 2014). In neurons, these components of the mitochondrial life cycle are closely interrelated: for example, the dysregulation of mitochondrial axonal transport can affect organelle morphology (Pathak et al., 2010), while altered fission–fusion balance can impair axonal mitochondrial motility (Verstreken et al., 2005; Baloh et al., 2007; Misko et al., 2010; Yu et al., 2016). Although it has been thought that the neuronal lysosomal compartment resides mainly near the nucleus, the presence, transport, and development of late endosomes (Deinhardt et al., 2006) and lysosomes (Moughamian and Holzbaur, 2012) have been demonstrated in axons. Additionally, in axons, local biogenesis (Amiri and Hollenbeck, 2008), fission–fusion (Cagalinec et al., 2013), and turnover (Maday et al., 2012; Ashrafi et al., 2014) of mitochondria are reported to occur. Thus, the regulation of the mitochondrial life cycle and the maintenance of a robust functional population, collectively termed mitochondrial “quality control” (QC; Chen and Chan, 2009; Rugarli and Langer, 2012), are likely to be more complex in neurons in vivo than in cells of more modest dimensions due to the potential distribution of functions among distant compartments.
It has been proposed that mitochondrial dynamics and QC are functionally regulated by two Parkinson's disease-related genes, PTEN (phosphatase and tensin homolog)-induced putative kinase 1 (PINK1) and the E3 ubiquitin ligase Parkin. Recent data from Drosophila suggest that PINK1/Parkin catalyze mitochondrial arrest by tagging for degradation of the mitochondria–kinesin linker protein Miro (Wang et al., 2011; Liu et al., 2012), and also prevent the fusion of senescent mitochondria with healthy ones by downregulating the mitochondrial fusion protein Mitofusin 2 (Deng et al., 2008; Poole et al., 2008, 2010). Furthermore, studies in non-neuronal cells indicate that PINK1 and Parkin target depolarized, dysfunctional mitochondria for autophagic engulfment and degradation (Narendra et al., 2008, 2010; Matsuda et al., 2010; Ashrafi et al., 2014; Lazarou et al., 2015). Although the question of whether this pathway governs mitochondrial QC in all neuronal compartments remains unsolved, these findings suggest that PINK1 and Parkin may cooperate in the maintenance of neuronal mitochondrial distribution and function by regulating mitochondrial dynamics and turnover.
Here, we used the intact Drosophila larval nervous system for in vivo study, complemented by cultured larval primary neurons as an in vitro system, to analyze the life cycle of neuronal mitochondria. We assessed the following three processes: mitochondrial axonal transport, fission–fusion, and degradation. We found that the loss of Parkin diminishes axonal mitochondrial flux without disturbing most features of movement. In addition, although Parkin-deficient motor neurons displayed greatly reduced numbers of axonal mitochondria, their organelle morphology and membrane potential were normal. However, mitochondrial morphology was disrupted in cell bodies, suggesting that this compartment houses Parkin-dependent QC in vivo. We also observed a striking difference in the extent of axonal mitophagy between in vitro motor neurons, where it was abundant, and the same neurons in vivo, where it was rare. Together, these observations indicate that Parkin-dependent mitochondrial QC differs significantly in location and abundance between in vivo and in vitro systems, and further suggest that Parkin regulates mitochondrial axonal access in vivo.
Materials and Methods
Drosophila strains and culture conditions.
All flies were reared on standard cornmeal agar medium and maintained at 25°C with a 12 h light/dark cycle. The GAL4-UAS system was used to visualize mitochondria, and autophagosomes specifically, in motor neurons. Wild-type flies expressing a single copy of mitochondrially targeted GFP (mito-GFP) or coexpressing mito-GFP with RFP-Atg8 driven by D42 driver were used as the controls (w; +/+; D42-Gal4>UAS-mitoGFP/+) and (w; UAS-RFPatg8/UAS-RFPatg8; D42-Gal4>UAS-mitoGFP/+; Pilling et al., 2006). Parkin-null mutants, (w; +/+; park25, D42-Gal4/park25, UAS-mitoGFP) and (w; UAS-RFPatg8/UAS-RFPatg8; park25, D42-Gal4/park25, UAS-mitoGFP), were used in this study (Greene et al., 2003). Genetically overexpressed Parkin in the null mutant background, (w; UAS-Parkin/+; park25, D42-Gal4/park25, UAS-mitoGFP), was used for rescue experiments. For starvation conditions, third instar larvae were collected and maintained in a distilled water-moisturized dish without food for 6 h.
Dissected larval preparation for observing neuronal mitochondria in vivo.
To image mitochondria in live motor neurons, late third instar larvae were collected, dissected, and prepared as previously described (Devireddy et al., 2014). In brief, larvae were pinned ventral side down on a Sylgard plate and dissected with a handheld scissors (catalog #15000-00, Fine Science Tools) from posterior to anterior along the dorsal body wall in HL6 buffer containing 0.6 mm CaCl2 and 4 mm[scap] l-glutamate. Fat bodies and intestine were then removed to reveal the intact ventral ganglion (VG), segmental nerves (SNs), and neuromuscular junctions (NMJs; Pilling et al., 2006; Louie et al., 2008; Shidara and Hollenbeck, 2010). The dissected larva was then placed dorsal side up on a glass slide and covered by a glass coverslip using dental wax as a spacer to form a chamber. This was filled with HL6 buffer and sealed with Valap for imaging. Mitochondria in both cell bodies and axons of larval motor neurons were observed using laser-scanning confocal microscopy (LSCM; C1 on an Eclipse 90i microscope, Nikon) within 20 min of dissection to ensure normal in vivo conditions of the nervous system (Shidara and Hollenbeck, 2010; Devireddy et al., 2014).
Time-lapse microscopy for organelle axonal transport analysis.
To analyze axonal transport of organelle in motor neurons, time-lapse images of mito-GFP and RFP-Atg8 were acquired from the longest SN in the following three different regions: proximal (segment A2); middle (segment A4); and distal (segment A7) to the VG. Time-lapse confocal images were obtained using 5% laser power with a 488 nm band for mito-GFP and 10% laser power with a 561 nm band for RFP-Atg8 through the smallest pinhole (30 μm). The region imaged was 50 μm in length, and frames for individual channels (513/30 for mito-GFP and 590/50 for RFP-Atg8) were obtained at 1 s time intervals for 2 min. The following parameters of transport were quantified essentially as previously described by the manual tracking function of ImageJ software (Devireddy et al., 2014). Briefly, mitochondrial and autophagosomal flux were measured for both anterograde and retrograde movement by observing the number of moving organelles passing a defined point per unit time over 2 min. Since some diffuse RFP-Atg8 signals were seen throughout the cytosol, only vacuoles above a threshold intensity (1000 on a 12 bit scale) were designated as autophagic vacuoles (AVs). Mitochondrial velocity was measured from continuous mitochondrial movements of >3 s duration in one direction. One pixel represents 0.1 μm in our confocal images, and only net velocities of >0.1 μm/s or < −0.1 μm/s for at least three consecutive frames were selected as bona fide anterograde and retrograde transport, respectively. During 2 min time-lapse movies, only mitochondria that moved for >60 s were considered for velocity analysis. Mitochondrial moving and stationary percentage was measured from the total population of mitochondria within a field. Since the population of mitochondria within the assigned axonal regions included a changing set of moving mitochondria, the second frame of each movie was selected for designating mitochondrial total population. Mitochondria in the population were categorized and marked as anterograde, retrograde, or stationary by ImageJ/Cell Counter to quantify the percentage of moving and stationary mitochondria in segmental axons. The mitochondrial duty cycle was measured for moving mitochondria by observing their movement or pause time over the observation interval. The percentage of the time that mitochondria moved in each particular direction or paused was quantified, and again only mitochondria that moved for >60 s were considered for duty cycles. Mitochondrial run length was defined as the distance per individual run between stops. Mitochondrial net runs in the dominant direction only for both anterograde and retrograde movement were used for the run length measurement. Mitochondrial flux with photo-bleaching was measured from 50 μm regional length of middle (segment A4) SNs with 30 μm photo bleached for 30 s with full intensity of 488 nm light from the confocal laser. Time course images were acquired at a rate of 1 frame every 2 s for 5 min. Every 30 s, the number of steady-state moving mitochondria was quantified from the bleached regions. An exponential series was applied to estimate the number of steady-state moving mitochondria in bleached regions. Thirty and 29 steady-state moving mitochondria were estimated in control and park25, Parkin animals respectively, while 17 steady-state moving mitochondria were estimated in park25 mutants from the 30 μm photo-bleached regions, as follows: control from (y = 0.033e63.612x), park25 from (y = 0.059e58.643x), and park25, Parkin from (y = 0.035e63.741x). These estimated numbers were nearly consistent with the numbers calculated from the percentage of moving mitochondria, as follows: ∼27 steady-state moving mitochondria were calculated in controls, while ∼16 steady-state moving mitochondria were calculated in park25 mutants from the bleached regions (data not shown).
Determination of organelle density and colocalization.
MetaMorph version 7.6.5 software was used to measure the mitochondrial density in SNs based on the intensity of the mito-GFP signal (Devireddy et al., 2014). The second frame of each movie was selected, and the number of pixels with GFP signal above threshold was used to determine the density of mitochondria in the region of observation. To clarify the shape of the mitochondrial boundary, the selected frames of confocal images are converted to binary images with 400 minimum-intensity thresholds on a 12 bit scale of 0–4095. Mitochondrial density in NMJs was measured by NIS-Elements AR 3.2 software to analyze the area. In each NMJ, the polygonal ROI tool was used to draw the region around the synaptic boutons, and the area that was covered by mitochondria was calculated with the total area. A 515/30 (mito-GFP) channel was used for measuring mitochondrial density, whereas a 590/50 (anti-HRP) channel was used for measuring the area of synaptic boutons. The images were converted by representative thresholds (400 minimum intensity for mitochondria, and 200 minimum intensity for anti-HRP staining) to clarify the shape of the boundaries. Colocalization of AVs with mitochondria was quantified by NIS-Elements AR 3.2 software. The intensity profiles from two channels, 515/30 (mito-GFP) and 590/50 (RFP-Atg8), were used to analyze the colocalization in each region.
Immunostaining.
For immunostaining of NMJs, late third instar larvae were partially dissected and fixed with 4% paraformaldehyde for 20 min at room temperature (Shidara and Hollenbeck, 2010). Fixed samples were washed with PBT (PBS containing 0.1% Triton X-100), and blocked with PBTB (PBT containing 0.2% BSA), then stained with mouse anti-HRP (1:1000) that was conjugated with Alexa Fluor 594 goat anti-mouse (a gift from J.C. Clemens, Purdue University, West Lafayette, IN). Immunostained samples were washed with PBTB before mounting and were observed by LSCM (C1 on an Eclipse 90i microscope, Nikon). Images of immunostained NMJs were taken between muscles 6 and 7 from A4 SNs.
Axonal mitochondria in adult fly wing.
To observe axonal mitochondria in the wing nerve, 5-d-old adult flies were collected and paralyzed on a CO2 anesthetizing pad. The fly wings were visualized under a dissection stereomicroscope and cut at the end of wing root to preserve a whole nerve tract along the humeral crossvein (HCV) with the first lateral vein (LV1) and costal vein of LV0 (Fang et al., 2012). Severed wings were pretreated with detergent (20%, Triton X-100 in dH2O) to avoid trapping air bubbles, briefly rinsed with 1× PBS (Fang et al., 2013), placed dorsal side up on a glass slide, and covered by a glass coverslip using dental wax as a spacer. The chamber was filled by PBS and sealed with Valap for imaging. Images were taken using LSCM within 10 min to ensure normal in vivo conditions of the wing neuronal system. Axons of HCV from the wing arch were monitored to analyze axonal mitochondrial density. In all experiments, only the right wing from the female adult was used.
Morphological analysis of mitochondria.
Image quantification was performed using NIS-Elements AR 3.2 software. Nonoverlapping axonal mitochondria from each SN (segment A2, segment A4, and segment A7) were selected for both length and area measurements, while in primary cultured neurons only length measurements were performed. Since 1 pixel represents 0.1 μm in the NIS-Elements AR 3.2 software images, the minimum standard distance unit was 0.1 μm. To determine the boundary of mitochondria for the length and area measurements, images were thresholded at a standard intensity of 200 to identify pixels belonging to mitochondria. Mitochondrial morphology in cell bodies was analyzed using both individual sections and three-dimensional image stacks. The latter were built from 20 μm (X, height) by 20 μm (Y, width) with 5 μm (Z, depth) of 100 z-stacks centered on the middle of the cell body. Morphological analyses were performed blinded to the experimental condition.
Measurement for mitochondrial membrane potential.
Tetramethylrhodamine methyl ester (TMRM; Life Technologies), a lipophilic cationic fluorescent dye, was used for the determination of mitochondrial transmembrane potential both in vivo (Devireddy et al., 2014) and in vitro (Verburg and Hollenbeck, 2008). Thin neurons and NMJs adjacent to the middle SNs in larval A4 segment and projected axons of primary cells were selected for measuring neuronal mitochondrial membrane potential (ΔΨm) in in vivo and in vitro conditions, respectively. TMRM 200 nm in prepared medium (HL6 buffer or Schneider's medium) was added to the dissected larvae or cultured cells for 20 min, then replaced with 50 nm TMRM for imaging. Only nonoverlapping axonal mitochondria were considered for quantification. Images were obtained sequentially by 488 nm excitation with 5% laser power for mito-GFP and by 561 nm excitation with 5% laser power for TMRM. The ratio of mitochondrial fluorescence intensities (Fm) to cytoplasmic fluorescence intensities (Fc) was used to determine ΔΨm (Fm/Fc; Verburg and Hollenbeck, 2008; Shidara and Hollenbeck, 2010; Devireddy et al., 2014).
Primary neuronal cell culture.
Brain lobes and ventral ganglia were taken from third instar larvae and transferred into Schneider's medium. Tissue was incubated in 0.7 mg/ml collagenase for 1 h at room temperature and then dissociated to individual cells by siliconized pipette triturating. After dissociation, ∼250 μl of cell solution was transferred to each glass coverslip in a cell culture dish and incubated for 1 h for cell adhesion. Glass coverslips were coated with 20 μg/ml concanavalin A for cell adhesion. Five to seven sets of CNS were used for two 35 mm cell culture dishes. After 1 h of incubation, 2 ml of Schneider's medium was added, and culture dishes were incubated for ∼96 h to allow neurons to extend their processes. Antimycin A (Ant A; 5 μm) was treated for 20 min to depolarize mitochondria. For oxidative stress conditions, 100 μm H2O2 was added to the culture dishes 1 h before taking images.
Statistical analysis.
To analyze data from park25 mutants compared with controls, p values were calculated using an unpaired homoscedastic t test. For multiple group comparisons, statistical significance was calculated by a one-way ANOVA with Bonferroni correction post hoc test. Significance between two populations was determined by two-sample Kolmogorov–Smirnov (K–S) test for analyzing the critical distribution. Correlations between two parameters were assessed by Pearson's correlation coefficient R2 values. In all cases, at least three independent experiments were performed. All statistical significance was verified at *p < 0.05, **p < 0.01, and ***p < 0.001 using GraphPad Prism.
Results
Parkin loss affects axonal mitochondrial flux without disturbing most features of movement
The PINK1/Parkin pathway has been proposed both to support mitochondrial QC and turnover (Narendra et al., 2008, 2010; Matsuda et al., 2010), and to regulate mitochondrial motility (Wang et al., 2011; Liu et al., 2012; Saxton and Hollenbeck, 2012). These two functions are likely to be related, as the return of senescent mitochondria to the cell body would require targeted net retrograde traffic. Thus, one expectation is that Parkin deficiency will inhibit the turnover and/or retrograde traffic of mitochondria, causing an accumulation of senescent organelles in the axon. To test this hypothesis in vivo, we quantified the axonal transport of mitochondria in motor neurons of the intact Drosophila larval nervous system (Pilling et al., 2006; Russo et al., 2009). Using mito-GFP driven by D42, transport was quantified in both directions in the proximal (segment A2), middle (segment A4), and distal (segment A7) regions of the SNs, and was compared between control (wild-type with D42-Gal4>UAS-mitoGFP) and park25 (parkin-null) animals (Fig. 1A,B; Devireddy et al., 2014). Consistent with previous analysis, mitochondrial flux in control motor axons was greater in the anterograde than in the retrograde direction, and declined with distance from the cell body in SNs (Shidara and Hollenbeck, 2010). In park25 SNs, flux was attenuated in both directions (Fig. 1C). However, anterograde flux was more severely impaired than retrograde in park25 SNs: anterograde flux dropped to 48% (A2), 41% (A4), and 53% (A7) of control, while retrograde flux dropped to 70% (A2), 53% (A4), and 61% (A7). These deficits were rescued by Parkin overexpression in the null background (Fig. 1B,D).
Parkin perturbation alters mitochondrial axonal traffic by reducing flux without affecting motility properties. A, Confocal image of dissected late third instar larva expressing mito-GFP in motor neurons. Mitochondrial movements are monitored at three different regions in the axons of SNs: proximal (A2), middle (A4), and distal (A7). Thin branch nerves, adjacent NMJs, and cell bodies in the VG are also analyzed. Scale bar, 200 μm. B, Representative kymographs of mitochondrial axonal transport from the A4 SNs. The red arrows indicate anterograde movements, while blue arrows indicate retrograde. Scale bar, 10 μm. C, Mitochondrial flux in different parkin genotypes in A2, A4, and A7 SNs. park25 mutants display attenuated mitochondrial flux in both directions throughout SNs. (n/n), Number of axonal regions/number of animals. D, Mitochondrial flux in parkin deletion and rescue in A4 SNs. The attenuated mitochondrial flux in park25 mutants is restored by UAS-Parkin expression. (n/n), Number of axonal regions/number of animals. E, Mitochondrial net velocity in different parkin genotypes. park25 mutants show retain normal mitochondrial velocities in both directions throughout the SNs. (n/n), Number of mitochondria/number of animals. F, The percentages of moving and stationary mitochondria are unaffected in park25 SNs. n, number of animals. Error bars indicate the mean ± SEM. Significance is determined by Student's t test (C, E, and F) or by one-way ANOVA with Bonferroni correction (D). **p < 0.01 and ***p < 0.001.
Because this pathway has been proposed to specifically detach kinesin from the mitochondrial surface (Wang et al., 2011), we next examined whether reduced mitochondrial flux in park25 SNs derived from the alteration of particular elements of mitochondrial motility. We analyzed the mitochondrial velocities, percentage moving, duty cycle, and run length (Devireddy et al., 2014). In park25 SNs, mitochondrial velocities in both directions were similar to those of controls throughout the axons (Fig. 1E), and the percentages of moving and stationary mitochondria in motor axons were unaffected by Parkin deletion (Fig. 1F). This eliminated the major movement parameters as sources of the flux deficit, but we further measured the duty cycle and run length of moving mitochondria, and found that park25 mutants showed a modestly increased percentage of time pausing during mitochondrial movement in both directions, caused by the reduced proportion of movements in the dominant direction (Table 1). However, the run lengths of moving mitochondria were generally maintained with only a modest decrease in retrograde in the A4 region of SNs (dropped to 81%; Table 1). These data indicate that although the loss of Parkin significantly reduces mitochondrial flux in the axon, mitochondrial motility behaviors and the percentage of moving mitochondria are nearly unchanged.
Duty cycle, run length, and area of axonal mitochondria from Drosophila larval motor neurons
Reduced mitochondrial flux in park25 motor axons results mainly from reduced organelle density
If mitochondrial motility was essentially normal in park25 mutants, what could explain the large decrease in flux? One possibility is that Parkin deficiency reduces the number of mitochondria in the axon. This would run contrary to expectations, since the proposed role of the PINK1/Parkin pathway in mitochondrial turnover predicts that Parkin deficiency would cause the accumulation of senescent mitochondria (Narendra et al., 2008, 2010; Matsuda et al., 2010). Nonetheless, we assessed this possibility by quantifying total mitochondrial density (both moving and stationary) in SN axons (Devireddy et al., 2014) and found that there was indeed a pronounced loss of axonal mitochondria throughout park25 motor axons, as follows: the mitochondrial densities in axons were reduced to 52% (A2), 61% (A4), and 56% (A7) of control levels in park25 SNs (Fig. 2A,B). This deficit was restored in A4 of park25 SNs by Parkin overexpression (Fig. 2A,C). To examine this deficit in park25 adult animals, we examined mitochondria in the wing, where a subset of sensory neurons expresses genes driven by D42, and these are situated along the wing margin (L0 vein and L1 vein) with L3 vein (Fang et al., 2012). We visualized D42>mito-GFP mitochondria in these sensory axons by observing the HCV of LV1 (Fig. 2D). As in the larval nervous system, HCV axons of park25 mutant wings showed reduced mitochondrial density; unlike the larval SNs, the deficit was not significantly restored by Parkin overexpression (Fig. 2E).
Reduced mitochondrial density is found in park25 motor axons. A, Representative binarized images of mitochondria from the axon of A4 SNs. Scale bar, 10 μm. B, Axonal mitochondrial density measurements from different parkin genotypes in the A2, A4, and A7 regions show that density is decreased by half in park25 mutants throughout the axons. C, Axonal mitochondrial density from different parkin genotypes in A4 SNs shows that the decreased density in park25 axons is restored by UAS-Parkin expression. D, Axonal mitochondria from 5-d-old adult fly wing. Axons from the HCV (red box) of the LV1 are monitored. Scale bars, 100 μm. E, Axonal mitochondrial density in the HCV from different genotypes. Binarized images of mitochondria from the outlined area (e′, e″, and e‴) are quantified, and park25 mutants display reduced mitochondrial density in HCV. The deficit is not restored by UAS-Parkin expression in adult HCV. Scale bars, 50 μm. n, number of animals. Error bars indicate the mean ± SEM. Significance is determined by Student's t test (B) or by one-way ANOVA with Bonferroni correction (C, E). **p < 0.01 and ***p < 0.001.
Because the majority of axonal mitochondria in SNs are persistently stationary (Fig. 1F), we assessed separately the density of the moving mitochondria. We photo bleached the mito-GFP in a 30 μm region of SN, quantified the entry of mitochondria into it over time in both directions (Fig. 3A), and extrapolated the densities to steady state using both linear and hyperbolic plots with good agreement (Fig. 3B,C). The park25 axons showed a density of moving mitochondria <60% that of controls, and this was partially rescued by Parkin overexpression. It was notable that, despite the lower densities of moving organelles, equal times were required to fill a bleached region to the steady state; both control and park25 animals required ∼210 s to reach a plateau (Fig. 3B). This confirms the previous observations (Fig. 1) that the reduced mitochondrial flux seen in park25 SN axons results from reduced organelle density without significant changes in motility behavior.
Reduced number of motile mitochondria is the cause of altered mitochondrial flux in park25 motor axons. A, Images from 5 min time lapse of a photo-bleached region of axon in A4 SNs. The steady-state numbers of moving mitochondria are quantified by monitoring their entry into the bleached regions (30 μm). Red arrows indicate the anterograde direction, blue indicates retrograde. Scale bar, 10 μm. B, Number of moving mitochondria in the bleached region reaches steady state at ∼5 min. Mitochondria are counted every 30 s. C, Double-reciprocal plots of the data in B. The steady-state mitochondrial number is estimated from the extrapolated y intercepts. A total of 30 and 29 steady-state moving mitochondria are estimated in control and park25, Parkin animals, while 17 steady-state moving mitochondria are estimated in park25 mutants from the 30 μm photo-bleached regions. (n/n), Number of axonal regions/number of animals. Error bars indicate the mean ± SEM.
Parkin mutation affects neither gross morphology nor ΔΨm of axonal mitochondria
The Parkin pathway has also been proposed to keep senescent, depolarized mitochondria from fusing with normal ones by promoting the degradation of Mitofusin (Deng et al., 2008; Poole et al., 2010; Ziviani et al., 2010). Thus, another prediction is that park25 mutants will display longer mitochondria with diminished inner ΔΨm. To test this hypothesis, we first quantified the length of both stationary and moving axonal mitochondria, and found that neither was affected by parkin perturbation (Fig. 4A). Stationary mitochondria measured 1.54 ± 0.01 μm in controls versus 1.55 ± 0.01 μm in park25 mutants, while moving mitochondria were 1.38 ± 0.03 μm in controls versus 1.28 ± 0.03 μm in park25 mutants. To determine whether a length difference in park25 mutants might be manifested not in the population average, but as a small outlying population of longer mitochondria, we examined the length distributions by a two-sample Kolmogorov–Smirnov test and found no difference between controls and park25 mutants for either moving or stationary mitochondria (data not shown). We further measured the projected area of axonal mitochondria, to confirm whether park25 mutants show any shape differences in axonal mitochondria, and found that here too park25 measurements were comparable to those for controls throughout the axons (Table 1; 0.82 ± 0.01 μm2 with 0.56 μm mean height in controls vs 0.79 ± 0.01 μm2 with 0.56 μm mean height in park25 mutants). Thus, Parkin seems to be dispensable for maintaining normal mitochondrial morphology in motor axons. We did not observe any correlation between morphology and motility in SN axons, so we further regressed mitochondrial velocities against their lengths to divine any relationship between the two. Although lengths were distributed from 0.35 to 4.65 μm, the net velocities of mitochondria were consistent in both directions, and there was no correlation between length and velocity or direction of movement (data not shown). Although this information seems counterintuitive on its face, it is consistent with the findings of previous studies that have shown no effect of mitochondrial length on axonal motility except at extreme, nonphysiological lengths (Amiri and Hollenbeck, 2008).
Parkin-deficient animals do not accumulate senescent mitochondria in motor axons and NMJs. A, Mean length of axonal mitochondria in different parkin genotypes. Stationary and moving mitochondria are measured in axons of A2, A4, and A7 SNs. Normal mitochondrial length is preserved in park25 mutants throughout the SNs. B, Representative TMRM stained images of axonal mitochondria with mito-GFP. Mitochondria are stained with TMRM (red) to indicate inner ΔΨm, while mito-GFP (green) reveals all axonal mitochondria. Note that mito-GFP here confirms the lower mitochondrial density in park25 axons compared with controls. Scale bars, 10 μm. C, Mean Fm/Fc intensity ratio of TMRM from neurons in different parkin genotypes. Normal mitochondrial inner membrane potential is retained in park25 mutant axons. (n/n), Number of mitochondria/number of animals. D, Representative TMRM staining images of mitochondria with mito-GFP from NMJs. NMJs are taken between muscles 7 (M7) and 6 (M6) from A4 SNs, and mitochondria in synaptic boutons are selected for ΔΨm quantification. Scale bars, 10 μm. E, Mean Fm/Fc intensity ratio of TMRM from NMJs in different genetic backgrounds. Normal ΔΨm is retained in synaptic boutons of park25 NMJs. (n/n), Number of mitochondria/number of animals. F, Images of mito-GFP (green) and anti-HRP staining (red) show mitochondria in motor synaptic boutons. Scale bars, 10 μm. G, Mitochondrial density in motor NMJs. park25 mutants display reduced mitochondrial density in synaptic boutons. (n/n), Number of boutons/number of animals. Error bars indicate the mean ± SEM. Significance is determined by Student's t test. ***p < 0.001.
Next, we assessed the ΔΨm of mitochondria from intact larval motor neurons to verify whether the sparse axonal mitochondria in park25 mutants nonetheless display the diminished membrane potential that is characteristic of senescence. TMRM was used for quantitative determination of transmembrane potential using the Fm/Fc ratio in mitochondria in thin peripheral motor nerves and NMJs adjacent to the larval A4 segment in vivo (Verburg and Hollenbeck, 2008; Shidara and Hollenbeck, 2010; Devireddy et al., 2014, 2015). Surprisingly, the ΔΨm values in park25 SNs (Fig. 4B,C) and NMJs (Fig. 4D,E) was indistinguishable from those in controls, and the reduced mitochondrial density observed in SN axons was also apparent in these thin nerves (Fig. 4B) and synaptic boutons (Fig. 4F,G) of park25 mutants. Our results thus indicate that Parkin deletion fails to produce the predicted accumulation of longer and senescent mitochondria in motor axons in vivo; park25 mutant animals display normal status for their axonal mitochondria.
Parkin deficit produces abnormal mitochondrial morphology in motor cell bodies
Parkin deficiency not only failed to produce the changes in axons predicted by its proposed role in mitochondrial QC—that is, an accumulation of senescent, longer mitochondria—but actually generated a dramatically reduced density of axonal mitochondria. This raised the possibility that the cell body is the locus of Parkin-dependent turnover and/or Parkin-dependent mitochondrial biogenesis, perhaps including the regulation of axonal entry. Thus, we analyzed the morphology of mitochondria in motor cell bodies of the larval nervous system; these give rise to the SN motor axons observed above. Interestingly, abnormal tubular, interconnected mitochondria were found in park25 cell bodies (Fig. 5A). To quantify these complex differences in organelle morphology, we categorized the cell body mitochondria into the following four different phenotypes—fragmented (1), wild-type (2), tubular (3), and hyperfused (4)—and scored them while blinded to the genotype. Compared with controls, the park25 cell bodies showed a much higher frequency of tubular (56% vs 9.6%) and hyperfused (24% vs 0%) mitochondria (Fig. 5B). Overexpression of Parkin in the park25 mutant background largely restored normal mitochondrial morphology (Fig. 5A,B). Together with Figure 3, these data indicate that Parkin is essential for both normal mitochondrial morphology in the cell body and the entry of mitochondria into the axon, and suggest the following likely relation of the two: that normal somatic organelle morphology, reflecting a normal fission-fusion balance, is important for producing and/or launching axonal mitochondria.
Parkin perturbation produces tubular, interconnected mitochondria in cell bodies. A, Representative plane and three-dimensional reconstructed images of the mitochondria in motor cell bodies. Reconstructed images are generated from sections of 20 μm width (x), 20 μm height (y), and 5 μm depth (z) using 100 frames to construct the z-stacks. Scale bars, 5 μm. B, Mitochondrial morphology in cell bodies of different parkin genotypes. Mitochondrial morphology is categorized as follows: 1, fragmented; 2, wild-type; 3, tubular; or 4, hyperfused. Tubular and hyperfused morphological defects of mitochondria are observed much more frequently in park25 cell bodies. Morphological defects of mitochondria in park25 cell bodies are partially restored by UAS-Parkin expression. (n/n), Number of cell bodies/number of animals. Data are generated in three independent analyses blinded to the genotype in the images and assessed by a two-sample Kolmogorov–Smirnov test (morphological changes of mitochondria caused by Parkin-null are significantly different from control; D = 0.7306 > critical D = 0.2209 with a corresponding p < 0.001).
Parkin-dependent mitophagy is undetectable in motor neurons in vivo
Studies in embryonic neurons in vitro have shown that autophagosomes initiate distally and undergo maturation during their retrograde transport (Maday et al., 2012; Maday and Holzbaur, 2014). Since the PINK1/Parkin pathway is thought to be necessary for local mitophagy in the distal axon (Ashrafi et al., 2014), the lack of evidence for Parkin-dependent axonal mitochondrial turnover that we found here was surprising. To probe further the role of Parkin in mitophagy in vivo, we visualized simultaneously mitochondria and AVs in SN motor neurons, using D42-driven mito-GFP and RFP-Atg8 (Barth et al., 2011). Because we had observed altered mitochondrial morphology in park25 cell bodies, we first quantified the colocalization of AVs with mitochondria there to determine whether Parkin plays a critical role in targeting dysfunctional mitochondria for mitophagy. Consistent with our previous data, tubular and elongated mitochondria were found in park25 cell bodies. However, mitochondria failed almost completely to colocalize with AVs in either control or park25 cell bodies (Fig. 6A,B). Studies using in vivo imaging of autophagosomes in mouse Purkinje cells in the cerebellar cortex provide evidence that macroautophagy is induced by starvation (Alirezaei et al., 2010; Chen et al., 2015). Thus, to increase autophagy above basal conditions, we prevented larvae from feeding for 6 h before observation, triggering starvation-induced autophagy (Gomes et al., 2011; Rambold et al., 2011; Ghosh et al., 2012). Although this treatment caused mitochondrial fragmentation (Fig. 6A), which should render mitochondria even better candidates for engulfment, it did not generate detectable colocalization of AVs with mitochondria in either control or park25 cell bodies (Fig. 6A,B). The mito-GFP and RFP-Atg8 signals clearly represented distinct organelles in virtually every case. Thus, in cell bodies we could not test the hypothesis that Parkin deficiency would reduce the rate of mitophagy in vivo, because there was no detectable mitophagy even in controls, and under either normal physiological or starvation-induced autophagic conditions.
Mitochondria do not colocalize significantly with AVs in vivo. A, Representative images of cell bodies in different parkin genotypes. mito-GFP fluorescence is used to identify mitochondria (green) and RFP-Atg8 to identify AVs (red). Autophagy is induced by 6 h of starvation. White arrows indicate lines along which organelle colocalization is analyzed for basal (a′, a″, b′, and b″) and starvation (c′, c″, d′, and d″) conditions. Scale bars, 10 μm. B, Respective signal intensity profiles from the line scans indicate in A. Distinct mito-GFP and RFP-Atg8 signals reveal no clearly detectable mitophagy events in motor cell bodies from either genotype. C, Representative images of axonal mitochondria with AVs in A2, A4, and A7 SNs from control animals. Organelle colocalization is represented by oblique (A2) or rectilinear (A4 and A7) line scan analysis. Scale bars, 10 μm. D, Signal intensity profiles of line scans of images in C. E, Representative images of axonal mitochondria with AVs in A2, A4, and A7 SNs from park25 mutant animals. Organelle colocalization is determined by oblique (A2 and A7) or rectilinear (A4) line scan analysis. Scale bars, 10 μm. F, Signal intensity profiles of line scans of the images in E. Mito-GFP and RFP-Atg8 signals from both genotypes show distinct, nonoverlapping distributions of mitochondria and AV signal throughout SNs. Angle brackets denote corresponding points in images (C, E) and line scan analysis for intensity profiles (D, F). AU, Arbitrary units.
We then observed autophagosomes along motor axons to elucidate the organization of autophagy and mitophagy in that compartment. As expected from previous studies (Maday et al., 2012), axonal AVs mainly moved retrogradely and rarely changed their direction, yielding a retrograde/anterograde flux ratio of almost 4, quite distinct from the saltatory bidirectional movements of mitochondria (Fig. 7A). The majority of retrograde AVs displayed faster movement than mitochondria (data not shown). When we simultaneously analyzed axonal transport of mitochondria and AVs, we found that, consistent with our previous data, mitochondrial flux was diminished in both directions by Parkin deficiency (Fig. 7B). AV flux was also substantially reduced in park25 mutants, but specifically in the retrograde direction; park25 mutants showed anterograde AV flux equal to that of controls, while retrograde flux was reduced by 50% (Fig. 7C). However, as in cell bodies, in axons we observed a striking lack of colocalization of AVs with mitochondria in either control or park25 mutant animals (Fig. 6C–F). These results indicate that while AVs are obvious and their axonal transport is consistent with them having a Parkin-dependent origin and/or maturation in the distal axon, mitophagy is so rare as to be undetectable in motor neurons in vivo. This was true even when starvation (Fig. 6A,B) produced large numbers of small, fragmented mitochondria that are highly suitable for autophagic engulfment.
Parkin perturbation impairs retrograde transport of AVs. A, Representative kymographs of mitochondrial and autophagosomal axonal transport from the A4 SNs axons of control animals. Red arrows indicate anterograde moving mitochondria, and blue arrows indicate retrograde moving organelles. Scale bar, 10 μm. B, Mitochondrial flux in different parkin genotypes. C, Autophagosomal flux in different parkin genotypes. Reduced AV flux is observed only in the retrograde direction, while reduced mitochondrial flux is observed in both directions in park25 mutant axons. (n/n), Number of axonal regions/number of animals. Error bars indicate the mean ± SEM. Significance is determined by Student's t test. *p < 0.05 and ***p < 0.001.
Mitophagy is common in Drosophila neurons in vitro and requires Parkin
Although we did not observe mitochondrial colocalization with AVs in Drosophila neurons in vivo, we note that most of the previous studies that have supported a role for Parkin in mitochondrial turnover and mitophagy were performed in vitro, with either neuronal cells (Ashrafi et al., 2014) or non-neuronal cells (Narendra et al., 2008, 2010; Matsuda et al., 2010; Lazarou et al., 2015). Thus, we examined whether autophagosomes and mitophagy can be detected in neurons in vitro in the same system that we used for our in vivo studies. We grew Drosophila primary motor neurons in culture, using the mito-GFP larvae as our source of cells (Bai et al., 2009; Pathak et al., 2010). In contrast to our in vivo results, we observed numerous AVs that colocalized with mitochondrial signals in vitro (Fig. 8A). Not only did control neurons show this evidence for mitophagy in vitro, but the frequency of AV–mitochondrial colocalization was substantially reduced in neurons from park25 mutant larvae; frequencies were reduced to 33% of control levels in the cell bodies and 27% in the axons (Fig. 8B). This suggests that mitochondrial engulfment into autophagosomes is Parkin-dependent in vitro. We also observed more elongated axonal mitochondria in park25 mutant neurons in vitro (Fig. 8A,C), which is predicted by current models of the role of Parkin in mitochondrial turnover (Deng et al., 2008), but was not observed in vivo. However, the number of AVs per cell in neurons cultured from park25 mutants was not reduced relative to controls in either the cell bodies or the axons (data not shown), arguing against Parkin control of general autophagy in vitro.
Parkin is essential for mitochondrial engulfment by autophagosomes in vitro. A, Images of in vitro primary motor neurons expressing mito-GFP and RFP-Atg8 after 96 h in culture. For stress conditions, cells are exposed to Ant A (5 μm, 20 min) or H2O2 (100 μm, 1 h) before imaging. White angle brackets denote the colocalization of mitochondria with AVs. Scale bars, 10 μm. n, Number of cells. B, Frequency of autophagosomes that contain mitochondria under each condition. Mitochondrial engulfment into AVs is inhibited in park25 mutant neurons. C, Mean length of axonal mitochondria in different parkin genotypes in exposed conditions. park25 mutant neurons show elongated axonal mitochondria. Axonal mitochondrial elongation in park25 neurons is inhibited by either treatment (Ant A or H2O2). n, Number of mitochondria. Error bars indicate the mean ± SEM. Significance is determined by Student's t test (B) or by one-way ANOVA with Bonferroni correction (C). *p < 0.05, **p < 0.01, and ***p < 0.001.
In neuronal (Ashrafi et al., 2014) as well as non-neuronal (Matsuda et al., 2010; Narendra et al., 2010) environments, mitochondria that undergo mitophagy mostly show an isolated, fragmented morphology. Thus, we further asked whether the size of mitochondria alone affects mitophagy, as visualized by AV–mitochondrial colocalization. To remodel mitochondria independently of Parkin perturbation, neurons were treated with 5 μm Ant A, an inhibitor of respiratory complex III that depolarizes neuronal mitochondria (Slater, 1973; Cai et al., 2012). Applying 5 μm Ant A for 20 min caused a significant decrease in mitochondrial size in park25 mutant cells (Fig. 8C). However, mitochondrial engulfment into autophagosomes was still inhibited in park25 mutant neurons; the frequency of AVs containing mitochondrial signal from park25 mutants was reduced to 28% of control levels in cell bodies and 49% in axons (Fig. 8A,B). In addition, we tested the role of Parkin in mitochondrial colocalization with AVs under conditions of experimental autophagic induction. Cells were treated with 100 μm H2O2, which induces autophagy through oxidative stress (Court and Coleman, 2012), and should promote mitophagy. Again, mitochondrial size was significantly decreased in both control and park25 mutant cells by H2O2 treatment for 1 h (Fig. 8C). However, even under conditions of oxidative stress, we saw the inhibition of mitochondrial engulfment in park25 mutant cells. In H2O2-treated cells, the frequency of AVs colocalizing with mitochondria was reduced to 18% of control levels in cell bodies and to 39% in axons (Fig. 8A,B). Our results suggest that Parkin regulates mitophagy in vitro independently of the modulation of mitochondrial size. Furthermore, since Ant A and H2O2 treatments cause the fragmentation of mitochondria (Fig. 8C) but no increase in AVs (data not shown), we also conclude that Parkin is dispensable for the remodeling of depolarized mitochondria.
Since we observed Parkin regulation of autophagosomal engulfment of mitochondria in vitro, we also asked whether axonal mitochondria in park25 mutant cells display normal membrane potential. The intensity ratio of TMRM staining showed that park25 mutant neurons indeed show diminished axonal ΔΨm (Fig. 9A,B). Consistent with previous analysis, axonal mitochondria were also elongated in park25 mutant cells (Fig. 9C). Thus, we regressed mitochondrial ΔΨm against their lengths to determine whether neurons in vitro display a key predicted consequence of Parkin-mediated QC: the accumulation of longer, less polarized mitochondria in the absence of Parkin. We found that park25 motor axons contained a small outlying population of longer mitochondria, and those mitochondria generally displayed diminished ΔΨm (Fig. 9D,E). Thus, as previously reported, we too find a significant in vitro role for Parkin in regulating axonal mitochondrial morphology and metabolic state; however, we do not observe this mechanism in the same neurons studied in vivo.
Parkin perturbation produced elongated mitochondria with diminished mitochondrial ΔΨm in vitro. A, Representative images of motor neurons stained with TMRM in vitro after 96 h in culture. Scale bars, 10 μm. B, Mean Fm/Fc intensity ratio of TMRM from cultured neurons of different parkin genotypes. C, Mean length of axonal mitochondria in vitro. D, Length distribution of axonal mitochondria in vitro. E, Relationship between mitochondrial ΔΨm and length in axons in vitro. Correlation coefficients: control (R2 = 0.0049) and park25 (R2 = 0.0004). park25 mutant neurons in vitro show decreased ΔΨm and elongated axonal mitochondria. (n/n), Number of mitochondria/number of cells. Error bars indicate the mean ± SEM. Significance is determined by Student's t test (B, C; *p < 0.05 and ***p < 0.001) or by two-sample Kolmogorov–Smirnov test (D; K–S statistics show significant difference; D = 0.2107 > critical D = 0.1274 with a corresponding p < 0.001).
Discussion
It seems likely that the cellular maintenance and long-term survival of neurons require adequate mitochondrial QC. Because neurons are highly compartmentalized cells, with structurally and functionally distinct somatodendritic and axonal compartments, it is important to understand not just whether and how dysfunctional mitochondria are turned over, but also where this occurs. In cells in vitro, the PINK1/Parkin pathway has been proposed both to target dysfunctional mitochondria for turnover and, in a related fashion, to regulate mitochondrial transport and dynamics. We have tested these hypotheses about Parkin-dependent mitochondrial QC in the Drosophila nervous system both in vivo and in vitro, and show here the following: (1) Parkin-deficient animals have dramatically reduced numbers of axonal mitochondria, but that the status and transport of these organelles remain normal; (2) the in vivo mitochondrial fission–fusion balance is regulated in a Parkin-dependent but somatically restricted manner; and (3) in vivo, mitophagy is rare in motor neurons and absent from their axons, but is both readily apparent and Parkin dependent in the same neurons in vitro.
Axonal transport of mitochondria has long been held to be intimately related to other organelle functions such as fission–fusion, metabolism, and turnover (Miller and Sheetz, 2004; Baloh et al., 2007; Pathak et al., 2010; Arduíno et al., 2012; Saxton and Hollenbeck, 2012). The PINK1/Parkin pathway has been proposed to coordinate mitochondrial motility with organelle turnover by specifically dissociating kinesin, the microtubule-based anterograde motor, from the organelle surface through degradation of its linker protein Miro (Wang et al., 2011). This model, however, does not explain how or why the modulation of the pathway disrupts mitochondrial movement in both the anterograde and retrograde directions. Since disruptions targeted to one direction of microtubule-based axonal transport often alter movement in both directions (Pilling et al., 2006; Park et al., 2009; Russo et al., 2009), it seems likely that the regulation of transport includes intricate coupling of both motors, kinesin and dynein. In any case, contrary to the proposed model of PINK1/Parkin dependent mitochondrial arrest, here we observed nearly unchanged mitochondrial motility behavior in vivo when Parkin expression was perturbed. Thus, although we found that Parkin deletion dramatically reduced mitochondrial flux in motor axons in vivo (Fig. 1C), this was derived almost entirely from a reduction in mitochondrial density to one-half that of normal axons (Figs. 2, 3), with only very modest, directionally balanced changes in motility. This is the opposite of the result expected for Parkin deletion, if this pathway were driving clearance of senescent mitochondria from the axon. It is, however, consistent with our previous demonstration that manipulation of PINK1 also fails to affect mitochondrial arrest (Devireddy et al., 2015). Since the PINK1/Parkin pathway is proposed not only to target damaged mitochondria for local turnover (Narendra et al., 2008, 2010; Matsuda et al., 2010; Ashrafi et al., 2014) but also to inhibit organelle fusion (Deng et al., 2008; Poole et al., 2010; Ziviani et al., 2010), we also expected Parkin deletion to produce elongated and metabolically compromised mitochondria in vivo. Surprisingly, our quantitative analyses demonstrated that mutant animals preserve both normal mitochondrial morphology and ΔΨm in their axons and NMJs (Fig. 4). On the other hand, in the cell bodies of Parkin-deficient motor neurons in vivo, organelle morphology was dramatically altered (Fig. 5), which suggests a compartmentalized in vivo role for Parkin in mitochondrial QC.
What could be restricting Parkin-dependent mitochondrial QC to cell bodies in vivo? One possibility is the dynamic interplay with other organelles, such as endoplasmic reticulum (ER). ER tubules form membrane contact sites with a number of other organelles (Rowland and Voeltz, 2012; Rowland et al., 2014), and such contacts regulate diverse neuronal functions, including motor protein-dependent organelle transport and neurite outgrowth (Raiborg et al., 2015). Recently, it has been proposed that ER tubules also play a regulatory role in mitochondrial division by contacting mitochondria and causing their constriction before the recruitment of the mitochondrial fission protein Drp1 (Friedman et al., 2011). Although ER elements are found in the axon and contribute to the transport and distribution of axonal cargos (González and Couve, 2014), ER tubules apparently maintain their contact sites with other organelles independent of microtubules (Rowland et al., 2014), and their major network is located in the cell body (Berridge, 1998; Verkhratsky and Petersen, 1998). Thus, it is plausible that the role of ER in mitochondrial morphology restricts the compartmental localization of the defects caused by Parkin perturbation in vivo. In addition, if the somatic compartment is the main site of mitochondrial turnover (Safiulina and Kaasik, 2013), then the entry of mitochondria into the axon from the cell body, and perhaps their subsequent return, are critically important in regulating the neuronal mitochondrial life cycle. In this context, we propose that mature neurons possess a QC barrier between the cell body and the axon that restricts axonal entry to those mitochondria of appropriate size, morphology, and metabolic status. This proposal is consistent with recent evidence that the preaxonal exclusion zone in the hillock provides a cytoplasmic boundary for the entry of vesicles into the axonal domain (Farías et al., 2015). Thus, the disruption of the fission–fusion balance in cell bodies by Parkin perturbation, which caused abnormal, interconnected mitochondria in the cell body, may physically disrupt the “launching process” of dysfunctional organelles into the axon.
If new mitochondria to supply the axon are generated largely in the cell body in vivo, then the importance of axonal transport looms even larger, especially anterograde movement. Indeed, it is notable that the same reduced-axonal-density phenotype that we observed has previously been reported in motor axons of Drosophila SNs with disrupted motor protein expression (Pilling et al., 2006). In that study, it was remarkable that only kinesin heavy chain mutants displayed reduced mitochondrial density in SNs, while mutations in dynein retained the normal compliment of axonal mitochondria. Furthermore, our previous report demonstrated that Parkin selectively restored the abnormal mitochondria of PINK1 mutants only in cell bodies, not in axons (Devireddy et al., 2015). In addition, it is noteworthy that a previous study (Cai et al., 2012) of cultured cortical neurons in vitro shows the restricted Parkin-mediated mitophagy only in the somatodendritic compartment. Together, these findings suggest a compartmentalized in vivo role of PINK1/Parkin in neuronal mitochondrial turnover and provide further evidence that the mitochondrial supply from the cell body is critically important for preserving organelle numbers and distribution in axons.
We did observe the morphological remodeling of mitochondria under stress conditions both in vivo and in vitro: starvation in vivo (Fig. 6A), as well as Ant A and H2O2 treatments in vitro (Fig. 8C) both produced fragmented mitochondria, even in the absence of Parkin. It is possible that mitochondrial fission is regulated independently of Parkin and/or upregulated by stress conditions. However, we observed several dramatic distinctions in the mitochondrial life cycle between motor neurons in vivo and in vitro. First, we rarely observed AVs that colocalized with mitochondria anywhere in motor neurons in vivo, whereas Parkin-dependent mitophagy was obvious in the same neurons in vitro. Second, axonal mitochondria under Parkin perturbation were elongated and displayed diminished ΔΨm in motor neurons in vitro, but had virtually normal status in vivo. Thus, while our observations from in vitro neurons supported the current models for the role of Parkin in axonal mitochondrial QC, targeting senescent mitochondria and promoting mitochondrial fission, data from the same neurons in vivo expressly did not.
What could explain these differences? One possibility is that fully connected neurons in vivo have established a physical filter for mitochondria between the cell body and axon that remains less developed in vitro. In support of this idea, we noted that the most axonal mitochondria in vivo were morphologically nearly uniform and approximately two times shorter than axonal mitochondria in cultured neurons (Figs. 4A, 9C). In addition, axonal mitochondria in vitro often extended continuously from the cell body into the axon (Fig. 8A), a feature that is rarely, if ever, seen in vivo, suggesting that the filter between the cell body and axon is absent or more permissive in cultured neurons. If mitochondrial entry into the axon is more restricted in vivo than in vitro, it could also explain the greatly downregulated axonal mitophagy that we observed in vivo. It is also important to note that mature neurons in vivo are functionally associated with surrounding cells and tissues. Recent evidence from mouse optic nerve head in vivo (Davis et al., 2014) shows the extrusion and transcellular degradation of axonal mitochondria, “transmitophagy,” a process that could provide a different avenue for mitochondrial turnover in the periphery, and could contribute to the lack of axonal mitophagy observed in vivo. It is also likely that neurons in vitro, subjected to hyperoxic conditions, are more susceptible to environmental stress. Thus, they may upregulate their mechanisms for mitochondrial clearance.
In summary, comparing Drosophila motor neurons in vivo and in vitro, we find a dramatic difference that indicates a highly compartmentalized in vivo role for Parkin in mitochondrial QC. Unlike in vitro conditions, the in vivo role of Parkin does not appear obviously related to mitophagy, and is restricted to the cell body where it seems to modulate mitochondrial fission–fusion balance and, indirectly, the mitochondrial composition of the axon. Our results re-emphasize the intricate interdependence of mitochondrial dynamics and axonal transport, and suggest that neuronal mitochondrial QC occurs in the cell body, and that a critical step is the initial access of mitochondria to the axon after organelle biogenesis.
Footnotes
This work was supported by National Institutes of Health/National Institute of Neurological Disorders and Stroke Grant 5R01-NS-027073. We thank L.J. Pallanck (University of Washington, Seattle, WA) for providing flies (D42-Gal4>UAS-mitoGFP), (park25, D42-Gal4/TM6B), (park25, UAS-mitoGFP/TM6B), and (UAS-Parkin/Cy0); E. Hafen (ETH, Zürich, Switzerland) for providing flies (UAS-RFPatg8/Cy0); J.C. Clemens (Purdue University, West Lafayette, IN) for providing anti-HRP; and Elisabeth Garland-Kuntz for assisting with blinded analysis of mitochondrial morphology.
The authors declare no competing financial interests.
- Correspondence should be addressed to Peter J. Hollenbeck, Department of Biological Sciences, Purdue University, 915 West State Street, West Lafayette, IN 47907. phollenb{at}purdue.edu
