Autophagy inhibition rescues structural and functional defects caused by the loss of mitochondrial chaperone Hsc70-5/mortalin in Drosophila

We investigate in larval and adult Drosophila models whether loss of the mitochondrial chaperone Hsc70-5/mortalin is sufficient to cause pathological alterations commonly observed in Parkinson disease. At affected larval neuromuscular junctions, no effects on terminal size, bouton size or number, synapse size, or number were observed, suggesting that we study an early stage of pathogenesis. At this stage, we noted a loss of synaptic vesicle proteins and active zone components, delayed synapse maturation, reduced evoked and spontaneous excitatory junctional potentials, increased synaptic fatigue, and cytoskeleton rearrangements. The adult model displays ATP depletion, altered body posture, and susceptibility to heat-induced paralysis. Adult phenotypes could be suppressed by knockdown of DJ-1b, LRRK, p50, p150, Atg1, Atg101, Atg5, Atg7, and Atg12. The knockdown of components of the autophagy machinery or overexpression of human mortalin broadly rescued larval and adult phenotypes, while disease-associated HSPA9 variants did not. Overexpression of Pink1 or promotion of autophagy exacerbated defects.


Introduction
Parkinson disease (PD), the most prevalent movement disorder and the second most prevalent neurodegenerative disease, is characterized by resting tremor, stiffness, and slowness of movement. Analysis of genetic and environmental factors contributing to PD suggests that impairments in mitochondrial function, lysosomal degradation pathways, and synaptic transmission are of central importance to pathogenesis and progression of PD [1][2][3][4]. Due to their complex morphology and high-energy demands, neurons in the adult brain are particularly susceptible to dysregulation of mitochondrial quality control systems. These quality control systems operate at different levels protecting cells and tissues from dysfunctional mitochondria. While a significant proportion of PD cases are sporadic, mutations in at least 11 genes have been implicated in monogenic typical or atypical forms of parkinsonism [3], providing crucial insights into the cellular and molecular pathways involved in PD. In all cases, the detailed molecular mechanisms leading to disease development have not been fully elucidated. However, a number of PDassociated genes have been implicated in mitochondrial function, altered mitochondrial dynamics, or the accumulation of dysfunctional mitochondria, which are characteristic features of PD [6][7][8][9][10][11]. For example, the PD-associated genes PTEN-induced kinase 1 (Pink1) and parkin are important regulators of mitochondrial quality control and mitophagy, a form of macroautophagy [6][7][8]. Mitochondrial dysfunctions have also been implicated in various other neurodegenerative diseases such as Alzheimer disease where mitochondria are key targets of both Aβ42 and tau toxicity [12].
We have previously addressed the importance of mitochondrial quality control and selective vulnerability of dopaminergic neurons by generating a model for lossof-function of Drosophila Hsc70-5/mortalin (referred to as Hsc70-5 or mortalin hereafter) [13]. Enhanced mitophagy was identified as an early pathological feature caused by knockdown in a presymptomatic model of loss of Hsc70-5 function [13] prior to the emergence of locomotion defects. Consistent with the upregulation of mitophagy upon loss of Hsc70-5 function in a Drosophila, human fibroblasts obtained from a carrier of the PD-associated A476T Mortalin variant exhibit increased colocalization of mitochondria with autophagosomes.
In the current study, we investigate in detail the consequences of neuronal loss of Hsc70-5/mortalin function in vivo and examine epistatic interactions in this genetic background.

Impairment of larval locomotion upon loss of Hsc70-5/mortalin function
We have previously shown that strong pan-neuronal expression (elav-Gal4, 29°C ) of the RNAi-construct mort KK results in lethality at the late L3 larval stage [14].
In the current study, we sought to investigate cellular and functional perturbations that occur early in pathogenesis. We thus reduced silencing potency by raising larvae at 25°C, which delayed lethality to the pupal stage (elav>mort KK and elav>mort GD ). elav>mort KK larvae were sluggish compared to size-matched elav>mort GD and control larvae. Their crawling velocity was reduced and the righting reflex delayed ( Figure   1A). elav>mort GD larvae were indistinguishable from controls in their righting ability.
Hence we referred to them as presymptomatic and elav>mort KK larvae as symptomatic (Table S1). Hsc70-5 is an important regulator of mitochondrial function [13,[15][16][17]. Thus, we investigated putative mitochondrial impairments in presymptomatic and symptomatic larvae. Pan-neuronal Hsc70-5 knockdown resulted in severe reductions in mitochondrial mass at neuromuscular junctions (NMJs) of symptomatic (elav>mort KK , 25°C) but not presymptomatic (elav>mort GD , 25°C) larvae compared to control ( Figure 1B). Symptomatic larvae showed a reduction in mitochondrial number and size, and consequently, a decrease in mitochondria area fraction at the NMJ ( Figure S1). Muscle length, NMJ area, and bouton numbers were unaltered ( Figure S2) in both presymptomatic and symptomatic larvae. This suggests that elav>mort KK larvae represent an early stage in disease progression characterized by the absence of gross neurodevelopmental or neurodegeneration phenotypes.

Detection of early synaptic changes due to loss of Hsc70-5/mortalin function
To further investigate structural changes in response to loss of Hsc70-5 function, we stained for synaptic vesicles (SVs) and active zone (Az) markers. To this aim we performed confocal imaging on fixed larval NMJs. Analysis and quantification revealed that SV proteins cysteine-string protein (CSP, Figure 1C) and the vesicular glutamate transporter (VGlut, Figure 1D) were reduced at NMJs of symptomatic larvae compared to control larvae. A similar reduction was observed for the central organizing component of Azs, Bruchpilot (Brp, Figure 1E). We observed no significant differences in presymptomatic larvae for SV proteins or Az components compared to control ( Figure 1C-E).
We proceeded to investigate synapse maturation by analyzing subsynaptic reticulum (SSR) and postsynaptic density (PSD) maturation. The SSR, a complex system of membrane tubules and lamellae, is formed by the invagination of postsynaptic membranes following presynaptic innervation. Thus, an increased percentage of terminal synaptic boutons devoid of SSR can be used as a marker for delayed synapse terminal maturation [18,19]. To examine SSR folding and morphogenesis, we labeled NMJs with antibodies against Dlg and the neuronal marker HRP. Dlg reliably surrounds terminal boutons in all genotypes examined, suggesting that Hsc70-5 knockdown does not affect SSR morphogenesis ( Figure 1F).
Furthermore, quantification of the number, intensity, and size of glutamate receptor cluster per synapse ( Figure 1G) revealed no postsynaptic phenotypes.
We next assessed defects in presynaptic maturation by quantifying the percentage of PSDs unopposed by Azs. To this aim, we analyzed the fraction of PSDs apposed by Az component Brp that reliably localizes to mature synapses. Az apposed to newly forming PSDs usually recruit detectable Brp puncta within 2-4 h of their formation and, more than 95% of all glutamatergic synapses are apposed by Brp [20]. A higher percentage of Brp-negative synapses is indicative of neurodegenerative or neurodevelopmental defects [21]. elav>mort KK larvae displayed defects in the apposition of PSDs and Azs. Twice as many unaposed, putatively immature synapses were detected. This defect might be caused by impairments in presynaptic maturation or stabilization ( Figure 1G).
Defects in MT cytoskeletal organization have been associated with PD models [22,23] in particular, and with loss of synaptic mitochondria in general [24,25]. The Drosophila homolog of the leucine-rich repeat kinase 2(LRRK2), linked to familial and sporadic PD, controls MT stability at Drosophila NMJs by suppressing Futsch function [23]. Additionally, a decreased number of MT loops, which is indicative of stable MT, has been observed [24]. In this study fewer MT loops were detected in terminal boutons at the NMJs of elav>mort KK larvae (Figure 2A, Figure S3). In addition to defective MT morphology, we observed a reduction in the percentage of boutons connected to the stable MT network ( Figure S3). This might impair long-range intracellular transport. Next, we employed the synaptic footprint assay and examined neuronal membranes by labeling HRP. Synaptic footprints are biomarkers for latestage synapse retraction [26], while HRP inhomogeneity has been associated with early-stage synapse disassembly [25]. We visualized the presynaptic compartment by labeling for the membrane marker HRP and the SV marker VGlut. Neither synaptic footprints nor HRP inhomogeneity was detected in response to Hsc70-5 knockdown ( Figure 2B) suggesting that elav>mort KK larvae represent an early stage of pathology with no evidence of synapse dismantlement.

Functional consequences of loss of Hsc70-5/mortalin function
Next, we investigated whether the aforementioned morphological changes affect synaptic function. To examine whether symptomatic elav>mort KK larvae have impaired synaptic transmission; we recorded evoked excitatory junctional potentials (eEJPs) by using current-clamp recording at NMJ from muscle 6 in segment A5. The amplitude of eEJPs in elav>mort KK was reduced compared to control ( Figure 3A). We also recorded spontaneous events ( Figure 3B and C) and the response of nerve terminals during and following high-frequency stimulations ( Figure 3D-F). We noted a decrease in the amplitude of miniature excitatory junctional potential (mEJP) and an increase in their frequency in elav>mort KK larvae compared to control. The ratio of eEJP to mEJP or quantal content ( Figure 3C) was reduced in mutants compared to control. 10Hz stimulation at mutant terminals caused no changes in EJP amplitude ( Figure 3D and E) during a 10Hz stimulation paradigm. However, a drastic timedependent increase in failure in mutant compared to control ( Figure 3D and F) was observed.
Next, we co-overexpressed Hsc70-5 or HSPA9 with mort KK and performed mitochondrial morphological analysis at the larval NMJ. Overexpression of Hsc70-5 or HSPA9 rescued number, area fraction of mitochondria at the NMJ. Moreover, normal mitochondrial size and morphology were restored. In contrast, HSPA9 R126W , HSPA9 A476T, and HSPA9 P509S failed to rescue these phenotypes ( Figure 4C, and C').
These results suggest that Hsc70-5 or HSPA9 rescued pupal lethality and larval locomotion defects by restoring mortalin function.
We utilized the Gal4/Gal80 system (elav>mort KK ,tub-GAL80 ts ) to achieve lateonset conditional knockdown. Raising larvae at 18°C before transferring them to 25°C for Gal4 expression 5 days after egg laying (AEL) prevent pupal lethality. We thus could analyze behavioral defects in flies. 4-days post-eclosion, elav>mort KK ,tub-GAL80 ts flies exhibited severe climbing defects and abnormal wing posture ( Figure   4D and E). We referred to these flies as symptomatic (Table S1). Importantly, late-onset knockdown of Hsc70-5 caused a reduction of ATP levels in fly heads ( Figure   4F). Overexpression of either Hsc70-5 or HSPA9 in the elav>mort KK ,tub-GAL80 ts background improved climbing ability, wing posture, and restored ATP levels in symptomatic model ( Figure 4D-F). Overexpression of HSPA9 R126W , HSPA9 A476T and HSPA9 P509S variants did not rescue any of the aforementioned defects ( Figure 4D-F). ATP availability is crucial for the regulation of intracellular calcium at the presynapse, in particular upon exposure to temperature-dependent cellular stress [29][30][31]. Furthermore, ATP is required for reversal of membrane potential following depolarization and propagation of action potentials along axons [32]. Flies with mutations in mitochondrial protein Cytochrome c Oxidase have been reported to suffer from temperature-induced paralysis [33]. We assayed flies at 39.5°C. elav>mort KK ,tub-GAL80 ts flies paralyzed faster than control flies ( Figure 4G). This defect was rescued by overexpression of either Hsc70-5 or HSPA9 but not HSPA9 R126W , HSPA9 A476T or HSPA9 P509S variants ( Figure 4G).

Downregulation of autophagy was protective
ATP and proper mitochondrial function are important maintain synaptic transmission at increased temperatures [29]. Mutants with compromised mitochondrial function paralyze faster compared to controls. We thus performed a candidate-based screen to identify modifiers of Hsc70-5 loss-of-function ( Figure 5A).
Downregulation of synaptic proteins involved in neurotransmission and endocytic machinery enhanced heat-stress induced paralysis observed upon loss of Hsc70-5.
Knockdown of Parkinson disease associated DJ-1b and LRRK, as well as components of retrograde transport machinery, were suppressors of Hsc70-5 knockdown induced paralysis.
We identified five members of the autophagy machinery Atg1, Atg5, Atg7, Atg12, and Atg101 as suppressors of Hsc70-5 loss-of-function induced paralysis ( Figure 5A). To verify the RNAi efficiency of modifiers isolated in the screen, we tested the Atg1, Atg5, Atg7, Atg12, and Atg101 RNAi strains for their ability to inhibit starvation-induced autophagy ( Figure 5B). Overexpression of RNAi against Atg1, Atg5, Atg7, Atg12, and Atg101 in the larval fat body using a mosaic system [34,35] for genetic analysis in cells expressing GFP completely suppressed starvation-induced autophagy ( Figure 5B). After validation of these constructs, we tested if knockdown of autophagy-related genes could ameliorate phenotype caused by loss of Hsc70-5. We used UAS-RNAi strains against Lame duck and white with mort KK (mort KK +white-RNAi and mort KK +Lame duck-RNAi) as controls to compare larvae bearing mort KK in combination with a second UAS site expressing RNAi against autophagic components. Pan-neuronal expression of RNAi strains targeting components of the autophagic machinery, Atg1, Atg5, Atg7, Atg12, and Atg101 in the elav>mort KK background restored the righting reflex to control levels ( Figure 5C). Next, we extended the analysis to flies. As previously demonstrated 4-day-old elav>mort KK ,tub-GAL80 ts symptomatic flies displayed severe defects in climbing and wing posture ( Figure 5D and E) [14]. Knockdown of autophagic components reversed impairments in locomotion, wing posture and ATP levels ( Figure 5

D-F) induced by
Hsc70-5 knockdown in young flies. Knockdown of autophagy genes alone did not reveal any significant differences compared to controls ( Figure 5C-F). These results suggest that inhibiting autophagy is sufficient to rescue Hsc70-5 knockdown associated impairments in larval and adult symptomatic models.

Autophagy induction in symptomatic flies exacerbated Hsc70-5/mortlain knockdown induced defects
Loss of Hsc70-5 function has been associated with increased mitophagy in Drosophila and human fibroblasts [14]. Consistent with a protective role for autophagy, Parkin overexpression is sufficient to rescue alterations in mitochondrial morphology and increased apoptosis [36] caused by mortalin silencing in HeLa cells and the dopaminergic SH-SY5Y cell line. To address whether promoting autophagy is detrimental or protective in vivo, we at first modulated autophagic flux genetically.

Quantification of larval locomotion revealed that pan-neuronal overexpression of Atg1
at 25°C is sufficient to cause a sluggish righting reflex. Concomitant overexpression of elav>mort KK did not further exacerbate this phenotype ( Figure 6A). Using the Gal4/Gal80 system, we investigated the effect of Atg1 overexpression on Hsc70-5 knockdown related phenotypes in flies by performing the longevity assay ( Figure 6B).
Knockdown of Hsc70-5 caused a significant reduction in median and maximum life expectancy compared to control flies. Atg1 overexpression in the control background enhanced median but not maximum survival. Notably, concomitant overexpression of Atg1 with mort KK caused a reduction in both median and maximum lifespan compared to mort KK alone ( Figure 6B). Atg1 overexpression in the elav>mort KK ,tub-GAL80 ts background exacerbated the loss of Hsc70-5 induced climbing impairment ( Figure 6C); however, a slight improvement was observed in wing phenotypes ( Figure   6D). Atg1 overexpression alone did not impair adult locomotion of 4-day-old flies examined using the climbing ability nor induce abnormal wing phenotypes ( Figure 6C and D). These results suggest that overexpression of Atg1 did not modify Hsc70-5 knockdown induced defects at the larval stage but reduced lifespan and climbing ability in flies.
Rapamycin has been used to induce autophagy in Drosophila [37]. Feeding rapamycin is sufficient to enhance autophagy as demonstrated by increased accumulation of autophagy markers Atg8a and relative levels of Atg8a-II, in addition to reduced levels of Ref (2)P, homolog of mammalian p62 ( Figure 6E and F). Hsc70-5 knockdown caused the same effects in fly heads ( Figure 6E and F). We investigated whether supplementing rapamycin could further enhance autophagy caused by PINK1 has been implicated in familial PD. It regulates the degradation of old and dysfunctional mitochondria in health and upon exposure to the mitochondrial uncoupler CCCP [6][7][8]38,39]. Hence, we examined if Pink1 overexpression could modulate the loss of Hsc70-5 phenotypes. Pink1 overexpression was sufficient to induce a sluggish righting reflect at the larval stage ( Figure 6J). In elav>mort KK larvae it rescued locomotion defects ( Figure 6J). Ectopic Pink1 expression in the elav>mort KK ,tub-GAL80 ts genetic background reduced lifespan ( Figure 6K) and exacerbated the climbing impairment and wing phenotypes in symptomatic adult flies ( Figure 6L and M).

observed in symptomatic larvae
Atg1 is required for the initiation of autophagosome formation [40]. We thus characterized the cellular changes caused by modulation of Atg1 expression in the elav>mort KK background. Atg1 knockdown reversed the loss of mitochondria observed at the NMJs of elav>mort KK larvae ( Figure 7A) and rescued the decrease in mitochondria area fraction, mitochondrial number, size, and altered morphology at the NMJ in elav>mort KK larvae ( Figure 7B).
NMJs of early symptomatic elav>mort KK larvae were characterized by reduced presence of SV marker proteins and Az components, impaired synapse maturation, and alterations in the MT cytoskeleton. We next assessed MT abundance as a marker for synapse stability. Atg1 knockdown restored usual MT abundance in elav>mort KK larvae ( Figure 7C). Furthermore, Atg1 knockdown in elav>mort KK larvae alleviated impairments in VGlut ( Figure 8A and B) and Brp abundance ( Figure 8C and D), and synapse maturation impairment ( Figure 8E and F). Overexpression of Atg1 failed to rescue defects in MT abundance ( Figure 7C), SV protein abundance ( Figure 8A and B), and synaptic maturation defects caused by loss of Hsc70-5 at the NMJ ( Figure 8E and F). These findings indicate that restoring mitochondrial mass by Atg1 knockdown could rescue loss of Hsc70-5 induced synaptic defects.

Autophagy suppression was protective against oxidative stress
Next, we tested whether knockdown of Hsc70-5 enhances vulnerability to oxidative stress. We treated young flies with hydrogen peroxide (H2O2). H2O2 induces generalized oxidative stress which is less specific than the effects of the mitochondrial complex inhibitors like rotenone or paraquat [41]. DJ-1 and LRRK are crucial for protection against oxidative stress [41,42]. H2O2 has also been shown to reduce the lifespan of flies that display a reduced function of the PD-associated proteins DJ-1 and LRRK. elav>mort KK ,tub-GAL80 ts flies were more vulnerable to H2O2 treatment than control flies ( Figure 9A). While Atg1 knockdown caused a minor reduction of lifespan upon exposure to oxidative stress compared to control, it proved to be effective in restoring the diminished stress resistance in the elav>mort KK ,tub-GAL80 ts background ( Figure 9A). We thus conclude that suppression of autophagy in the elav>mort KK ,tub-GAL80 ts background might be beneficial for survival under oxidative stress.

Autophagy suppression improved healthspan but did not extend lifespan
To investigate the long-term impact of Atg1 knockdown, we examined the lifespan of flies under normal conditions ( Figure 9B). Atg1 knockdown led to a reduction of the lifespan in elav>mort KK ,tub-GAL80 ts flies. As a minor decrease in lifespan had been observed, we thought to assessed the effect of Atg1 knockdown at various pathological stages. Atg1 knockdown was beneficial in 4-day old symptomatic flies and improved locomotion. In 10-day old flies, concomitant knockdown resulted impairment in climbing using a less challenging ( Figure 9C). We referred to these 10-day old flies as late-symptomatic flies (Table S1). We conclude that reduced autophagy flux is beneficial for improving locomotion ability in Hsc70-5 reduced background in young flies, but impairs climbing ability and lifespan in elav>mort KK ,tub-GAL80 ts flies in the long run.

Investigating the impact of rare Hsc70-5/mortalin mutations in vivo
Three mutations have been identified in mortalin in a congenital disease termed epiphyseal, vertebral, ear, nose, plus associated finding) syndrome or EVEN-PLUS [28]. Biallelic mutations were identified in three individuals from two families, including the previously reported R126W variant [17]. Functional studies investigating the impact of variant HSPA9 P509S revealed a lower ATP catalysis rate and reduced chaperoning activity compared to HSPA9 WT [17]. Next, morphological studies in human fibroblasts from a heterozygous carrier of variant HSPA9 A476T showed mitochondrial defects compared to HSPA9 WT homozygous sibling [17]. Another study investigated the effects of human variant R126W, and P509S in yeast by generating analogous substitutions in the Saccharomyces cerevisiae ortholog of Hsc70-5, the SSC1 gene [16]. The substrate-binding domain mutation SSC1 A453T was shown to cause mitochondrial dysfunction, enhanced ROS levels, and reduced ability to prevent aggregate formation of unfolded substrates [16].
In this study, we used a Drosophila model of loss of Hsc70-5/mortalin function to test the functional relevance of rare mortalin mutations HSPA9 R126W , HSPA9 A476T, and HSPA9 P509S in vivo. Overexpression of human and Drosophila mortalin (HSPA9 and , in a genetic background with reduced Hsc70-5, unlike mutant variants HSPA9 R126W , HSPA9 A476T and HSPA9 P509S , were able to rescue the loss of synaptic mitochondria, locomotion defects, abnormal wing phenotype, and reduced ATP levels. This study provides evidence that the investigated amino acid replacements in the ATPase and substrate-binding domain impair Mortalin function. In vitro overexpression studies have suggested no difference in overexpression levels or localization of the mutant variants [17] and future studies should address this in in vivo setting.

Hsc70-5/mortalin knockdown induced synaptic defects
Neuronal mitochondria have been implicated in diverse functions including spine formation [43], synaptic plasticity [44], and axonal branching [45]. Loss of Miro dramatically excludes mitochondria from distal compartments and causes a gradual time-dependent reduction in EJP amplitude following 10Hz stimulation, although no changes were observed in EJP or mEJP amplitude under baseline conditions [24].
Similar findings were reported in Drp1 mutants where dramatic synaptic mitochondria reduction was observed [46].
Loss of neuronal mitochondria as a consequence of Hsc70-5 silencing causes a reduction in EJP and mEJP amplitudes and quantal content ( Figure 3A-C). There was an increase in the frequency of mEJP events ( Figure 3C), possibly due to the elevation of intracellular calcium levels following loss of Hsc70-5. More frequent synaptic failure was observed upon stimulation at 10Hz ( Figure 3F). Failed responses that are caused by the depletion of synaptic vesicles are preceded by a gradual decline in EJP amplitude [46]. However, in our case, the amplitude of the remaining EJPs was not affected. Thus, failures might instead be caused by impaired propagation of action potentials as described for Atpα mutant [32].
Cellular analysis of larval NMJ in Hsc70-5 knockdown animals did not reveal gross impairments in the morphology of distal compartments ( Figure S2) or inhomogeneity of presynaptic membrane or synaptic footprints ( Figure 3B) that are biomarkers for early and late stages of synapse disassembly, respectively [25,26].
Importantly, defects in the availability of synaptic components, such as CSP, VGlut, and Brp (Figure 1), were observed at a stage in which no major [47] neurodegeneration was noted. Hence, the protective effects of blocking mitophagy described in this study were likely related to mitochondrial function rather than susceptibility to cell death. At this stage of disease progression, RNAi against Atg1 was sufficient to reverse defects in development, regulation of synaptic vesicles, synaptic terminal stability, and function (Figures 7 and 8). Furthermore, Atg1-RNAi restored mitochondria abundance. In addition, it also reversed the only putatively degenerative change we found in Hsc70-5 knockdown larvae: MT cytoskeletal alterations ( Figure 7B).

Epistatic interaction of Hsc70-5/mortalin with the autophagic machinery
Mitophagy, a type of macro-autophagy that targets specifically mitochondria, is increased upon partial loss of mortalin function in Drosophila and human fibroblasts [14]. Increased autophagic flux can be likened to a double-edged sword that protects neurons from chronic oxidative stress but might accelerate mitochondrial loss under conditions of premature mitophagy or impaired mitochondrial biogenesis [13,48]. The beneficial effects of Parkin overexpression in mouse embryonic fibroblasts treated with siRNA against mortalin were dependent on intact autophagic machinery, suggesting a potential therapeutic benefit of upregulating autophagy [17]. However, the interpretation of suppressed apoptosis in HEK293 cells and tumor-derived SH-SY5Y cells is complicated by the fact that (like most tumor-derived culture cells) these cells primarily generate ATP via aerobic glycolysis and are therefore not dependent on oxidative phosphorylation [49]. Thus, the balance between costs and benefits of effective mitochondria removal might differ in vivo.
We previously demonstrated that Hsc70-5 knockdown in Drosophila caused a severe loss of synaptic mitochondria and cellular ATP depletion [13]. The Drosophila NMJ has proved to be a particularly useful model as it is easily accessible and morphologically complex. These characteristics allowed us to model a highly active synaptic population that is vulnerable to impairment in intracellular trafficking because it is located very distantly to the neuronal soma. The in vivo data obtained in this study did not provide evidence for the therapeutic benefit of promoting autophagic flux following impaired mitochondrial function. Ectopic Atg1 expression, which was sufficient to induce autophagy in Drosophila [50] did not rescue impaired locomotion or mitochondrial mass following Hsc70-5 knockdown in the symptomatic larval model ( Figure 7A). Besides, Atg1 overexpression did not reverse alterations in synaptic development at the larval NMJ (Figure 8). In symptomatic flies, Atg1 overexpression exacerbated climbing defects and shortened lifespan caused by loss of Hsc70-5. The inability of Atg1 to exacerbate Hsc70-5 knockdown associated mitochondrial abundance or cellular defects at the larval stage might be explained by the fact that overexpression of Atg1 alone does not reduce mitochondrial abundance ( Figure 7A).
It is also possible that the loss of mitochondria upon Hsc70-5 knockdown is so extreme that further Atg1 overexpression is unable to exacerbate the already severe defect (Figures 7 and 8).
The analysis of Atg1 overexpression in larval stages was complicated by the fact that Atg1 overexpression alone caused a sluggish righting phenotype ( Figure   6A). Pink1 overexpression alone also caused a sluggish righting phenotype in larvae and reduced life expectancy in adult flies ( Figure 6). Nevertheless, both Atg1 and  Figure 1B). Similar pathological induction of autophagy has been reported for hypoxic-ischemic brain injury and two toxin-based PD models in which blockade of autophagy proved to be protective [54].

Stimulation of autophagy under conditions of oxidative stress
Oxidative and nitrosative stress is associated with various neurodegenerative  (Figures 7 and 9) presented in this study. The in vivo evidence presented here suggests that reduced rates of autophagy might be protective for neurons compromised by pathologically increased levels of mitophagy. However, we observed that detrimental side effects exceeded protective benefits in the long-term (Figure 9).

Molecular biology
cDNA constructs for HSPA9 WT , HSPA9 R126W , HSPA9 A476T , and HSPA9 P509S were received from Rejko Krüger (University of Luxembourg). The constructs were recloned and inserted into a modified pUAST attB vector (using BamH I and Xho I).
The full-length Drosophila Hsc70-5 (GM13788) was inserted into a modified pUAST attB vector. Details of fly strains have been provided in the supplementary information (Materials and Methods, Table S2).

Bar charts
The standard error of mean (SEM) and standard deviation (SD) are shown as a box and black line.

Staining and imaging larval neuromuscular junctions
Dissection and labeling of size-matched mid-3 rd instar larvae were performed using previously described protocols [25]. For more information on antibody source and concentrations, refer to supplementary information.

Fat body assay for investigating autophagy
Assay to validate functionality of UAS-Atg1 overexpression or RNAi constructs against autophagy related proteins were performed using a mosaic genetic analysis in larval fat body cells essentially as previously described [34,35]. Refer to supplementary information for details.

ATP measurements
ATP levels in head homogenates were measured and normalized using a luciferase-based bioluminescence assay as described earlier [13]. Five heads of female flies were homogenized in 6M guanidine-HCl and frozen in liquid nitrogen.

Locomotion analysis
Climbing assays were, unless otherwise noted, conducted as previously described [14]. Climbing male flies was monitored by analyzing their performance to climb 6 cm (challenging assay, used in Figure 4D or 5D) or 3 cm (less challenging assay used in Figure 9C for 10-day old flies) within 14s. A successful attempt was scored as 1, and failure to reach the top as 0. Each fly was assessed three times to calculate the average climbing score. At least 40 flies per genotype were analyzed.
Larval locomotion was investigated by examining larvae crawling speed and righting assay. Larvae locomotion speed was quantified as previously described [25].

Longevity assay
Drosophila larvae were transferred from 18°C to 25°C to boost UAS-transgene

Fly strains
Transgenic fly stocks were obtained either from Indiana University Stock Center

Fly culture conditions
Flies were raised on standard cornmeal/agar medium. To circumvent pupal lethality caused by Hsc70-5/mortalin knockdown and to analyze behavioral defects in adult flies, we utilized the Gal4/Gal80 system (elav>mort KK ,tub-GAL80 ts ). This allowed us to achieve late-onset conditional knockdown, by raising larvae at 18°C before transferring them to 25°C at 5 days after egg laying (AEL). Flies were kept at 25°C during development for analysis of wing phenotype, climbing defects, longevity, ATP levels from heads, and temperature-induced paralysis at appropriate ages.

Morphological analysis
Analyses of mitochondria and NMJs were performed as previously described [1]. The area fraction occupied by mitochondria was used as the index of mitochondrial mass. To quantify Futsch loops, the number of loops located to the two most distal boutons at each terminal of the NMJ was scored. The total number of loops in these regions was normalized to the number of terminals.
Images were scaled by a factor of 2 before Gaussian blur filtering was applied (pixel radius=2). Gamma values were set to 0.75. For quantitative comparisons of intensities, standard imaging settings were chosen that avoided oversaturation.
ImageJ Software Version 1.43e (National Institutes of Health, Bethesda, MD, USA) was used for image processing.

Larval fat body assay for investigating autophagy
Starvation of 2 nd instar stage larvae was performed for 5-6 hours in fresh empty vials on a filter paper soaked with H2O. For imaging, 2 nd instar larvae that were well fed or starved for 4-5 hours were cut open and turned inside out like a 'sock' before removal of fat bodies. Fat body tissues were stained with Dapi for 2 minutes, briefly washed in PBS, and mounted on a glass slide using Vectashield and imaged immediately using a Leica SP5 II confocal imaging system using a 40× Plan-Apochromat 1.4 N.A. oil objective. The entire procedure from fat body isolation to imaging was completed within 30 minutes.

Morphological analysis
Analyses of mitochondria and NMJs were, unless otherwise noted, performed as previously described [1]. The area fraction occupied by mitochondria was used as the index of mitochondrial mass. To quantify Futsch loops, the number of loops located to the two most distal boutons at each terminal of the NMJ was scored. The total number of loops in these regions was normalized to the number of terminals.                   Table S1