Bendless is essential for PINK1-Park mediated Mitofusin degradation under mitochondrial stress caused by loss of LRPPRC

Cells under mitochondrial stress often co-opt mechanisms to maintain energy homeostasis, mitochondrial quality control and cell survival. A mechanistic understanding of such responses is crucial for further insight into mitochondrial biology and diseases. Through an unbiased genetic screen in Drosophila, we identify that mutations in lrpprc2, a homolog of the human LRPPRC gene that is linked to the French-Canadian Leigh syndrome, result in PINK1-Park activation. While the PINK1-Park pathway is well known to induce mitophagy, we show that PINK1-Park regulates mitochondrial dynamics by inducing the degradation of the mitochondrial fusion protein Mitofusin/Marf in lrpprc2 mutants. In our genetic screen, we also discover that Bendless, a K63-linked E2 conjugase, is a regulator of Marf, as loss of bendless results in increased Marf levels. We show that Bendless is required for PINK1 stability, and subsequently for PINK1-Park mediated Marf degradation under physiological conditions, and in response to mitochondrial stress as seen in lrpprc2. Additionally, we show that loss of bendless in lrpprc2 mutant eyes results in photoreceptor degeneration, indicating a neuroprotective role for Bendless-PINK1-Park mediated Marf degradation. Based on our observations, we propose that certain forms of mitochondrial stress activate Bendless-PINK1-Park to limit mitochondrial fusion, which is a cell-protective response.


Introduction
Mitochondria are dynamic organelles and their size varies in response to various cellular cues such as developmental [1] and stress signals [2]. This change in mitochondrial size is crucial for cellular adaptations. For example, cellular stress due to amino acid deprivation [3] or oxidative stress leads to mitochondrial hyperfusion [4]. This form of stress-induced mitochondrial hyperfusion (SIMH) is beneficial as it improves ATP production [5] and protects mitochondria from autophagy [3]. Several mitochondrial disease linked mutations have been shown to alter mitochondrial morphology such as mutations in COX10 [6] and TFAM [7] show accumulation of enlarged mitochondria, possibly due to SIMH, however the mechanism of such responses remain elusive. On the other hand, mitochondrial fragmentation occurs as a beneficial process under different cellular signals. For example, increased mitochondrial fission allows for clearance of damaged mitochondria in embryonic stem cells providing them increased resistance to apoptotic signals [8]. Mitochondrial fission enables segregation of damaged mitochondria to enable their removal through mitophagy [9][10][11][12][13], as well as reduced ROS production [14] and promoting cell survival. Hence, while a normal balance of fission-fusion exists physiologically, a change in metabolic needs or other forms of stress can tilt the balance towards either one process and this generally is required to trigger an adaptive cellular response.
Changes in mitochondrial shape and size, i.e., mitochondrial dynamics, requires regulation of GTPases essential for mitochondrial dynamics. While the Dynamin 1-like (DNM1/Drp1) protein mediates fission, Mitofusins (Mfn1 and Mfn2 in mammals, Marf in Drosophila) and Optic Atrophy 1 (OPA1) mediate the fusion of mitochondrial outer and inner membranes, respectively. Several post-translational modifications, such as phosphorylation, acetylation and ubiquitination are crucial for the activity of these proteins, and thereby play an important role in determining mitochondrial size [15,16]. Indeed misregulation of mitochondrial dynamics proteins-Mitofusin, Opa1 or Drp1 are all associated with metabolic and neurodegenerative diseases [17].
The E3 ubiquitin ligase Parkin (Park in Drosophila, PARK2 in humans) and the kinase PINK1, which are linked to autosomal recessive early-onset Parkinsonism, are known to regulate mitochondrial quality control [18]. Studies in human cancer cell lines have shown that dissipation of the mitochondrial membrane potential (MMP), increased oxidative stress or mitochondrial unfolded protein response (UPR mt ) results in the stabilization of PINK1 on the outer mitochondrial membrane (OMM). PINK1 stabilization on the OMM leads to Park recruitment, polyubiquitination of OMM proteins and mitophagy [10,[19][20][21]. Several in vivo studies have also shown a conserved role for PINK1-Park in mitophagy [22][23][24][25][26][27][28][29]. While PINK1-Park mediated mitophagy has been extensively studied in cells, how the PINK1-Park pathway is activated under physiological conditions in vivo remains elusive [30]. Additionally, in vivo studies suggest a pro-fission role of PINK1-Park [31][32][33][34][35][36], perhaps through the turnover of Mitofusin levels as loss of Pink1 or Park shows increased Marf levels [10,37]. As most of these studies utilize PINK1 and PARK2 mutants to study defects in mitochondrial dynamics, the mechanism by which they are regulated in vivo under various physiological conditions remains unresolved. Additionally, it is unclear as to how the PINK1-Park pathway may activate mitophagy, alter mitochondrial dynamics or selectively target certain OMM proteins in response to various cellular cues.
To study the regulation of Mitofusin/Marf in vivo, we undertook an unbiased genetic mosaic screen in Drosophila. Through this screen, we discovered that mutations in lrpprc2 (referred as ppr in figures) and bendless (ben) causes downregulation and upregulation of Marf levels respectively. Lrpprc2 is a homolog of human LRPPRC that is required for mitochondrial mRNA stability and translation and thus mutations in LRPPRC affect mitochondrial oxidative phosphorylation [38,39]. Loss of LRPPRC results in a neurometabolic disorder-French-Canadian Leigh Syndrome [40,41]. Studies in worms, mouse and human cells have also shown that loss of LRPPRC is associated with large mitochondrial size [42,43]. We found that loss of lrpprc2 results in proteasome-mediated Marf degradation in a PINK1-Park dependent mechanism. Further, we also discovered that mutations in bendless (ben), which encodes a K63-linked E2 ubiquitin conjugase, is essential for Marf degradation in lrpprc2 mutants. We additionally demonstrate an essential role for Ben in regulating PINK1 stability, which in turn is required for maintaining steady state Marf levels in healthy cells. Finally, we show that in lrpprc2 mutants, Ben suppresses excessive mitochondrial fusion and prevents neuronal death under mitochondrial stress.

Loss of lrpprc2 results in reduced Marf levels
To identify novel regulators of mitochondrial dynamics, we performed a blind screen using a collection of Drosophila EMS induced X-chromosome lethal mutations [44,45]. This collection was initially generated to identify mutants with neurodegenerative phenotypes and has previously uncovered mutations in Marf [46] and several other genes required for mitochondrial function [39,47]. We screened these mutants for misregulation of Marf protein using an HAtagged Marf genomic construct (Marf::HA). We used the FLP-FRT mediated mitotic recombination strategy to create mutant clones (non-GFP cells) in a heterozygous background (GFP expressing cells) in the developing wing disc epithelium [48] (S1A-S1A' Fig). This allowed us to compare Marf levels in mutant and wildtype cells within the same tissue.
From this screen, we found that mutant clones of two lrpprc2 alleles (lrpprc2 A and lrpprc2 E ) show reduced Marf:HA levels compared to surrounding wildtype cells (Figs 1A-1A", 1E and S1B-S1B"). To confirm this observation, we used an independent Marf genomic rescue line, Marf::mCherry, and found reduced Marf::mCherry staining in lrpprc2 A mutant clones (S1C-S1C" Fig). To test if the reduction in Marf::HA or Marf::mCherry is caused by reduced mitochondrial content, we checked the levels of an OMM protein Tom20 using an endogenous tagged line (Tom20::mCherry). We did not observe a downregulation of Tom20::mCherry staining in lrpprc2 A mutant clones (Fig 1B-1B" and 1E). Taken together, these data suggest that downregulation of Marf in lrpprc2 mutants is not due to reduced mitochondrial content. Additionally, we also checked the levels of other proteins involved in mitochondrial dynamics -Opa1 and Drp1-using genomic tags. While we found the levels of Opa1::HA to be slightly increased in lrpprc2 A mutant clones (Fig 1C-1C" and 1E), Drp1::HA levels remained unaltered ( Fig 1D-1 D" and 1E). As mutations in lrpprc2/LRPPRC result in mitochondrial defects due to reduced stability of mtRNA [38,39,49], Marf reduction in lrpprc2 mutants could be an adaptation to segregate defective mitochondria by suppressing their fusion.

PLOS GENETICS
Since reduced Marf is expected to suppress mitochondrial fusion, we decided to check mitochondrial morphology in lrpprc2 A mutants. The cells in the wing discs are very compact making it difficult to analyze mitochondrial morphology. Hence, we checked mutant clones in the peripodial membrane, which is a squamous epithelium overlying wing discs. We used anti-Complex V staining to mark mitochondria. Interestingly, we found that mitochondrial size is increased in lrpprc2 A mutant clones (S1D-S1D" and S1E Fig). This is consistent with earlier findings that showed enlarged mitochondrial size in LRPPRC knockdown in mouse liver [42], C.elegans and mammalian cell culture [43]. As many studies have shown that mitochondrial stress can result in increased mitochondrial size [3,5,43,50], we suspect a similar mechanism results in increased mitochondrial size in lrpprc2 mutant cells, while an independent mitochondrial quality control mechanism may suppress their fusion by reducing Marf levels.

UPS dependent Marf degradation in lrpprc2 mutants
Reduced Marf levels in lrpprc2 mutant clones could be because of increased protein turnover via selective autophagy or ubiquitin-proteasomal system (UPS). We tested the possibility of autophagic degradation of Marf. We fed the larvae chloroquine, an inhibitor of autophagosome-lysosome fusion [51], and found that Marf::HA levels remain reduced in lrpprc2 A clones ( Fig 2B-2B"). To check whether chloroquine treatment itself alters Marf levels, we check endogenous Marf::mCherry levels in wing discs of chloroquine fed larvae and untreated larvae. We observe no effect of chloroquine treatment on Marf::mCherry levels (S2A-S2A" Fig). We found that the levels of p62, a protein degraded primarily via autophagy, was also not altered in lrpprc2 A mutant clones in larval wing discs (S2C-S2C" Fig). Thus we conclude that autophagy is neither enhanced nor likely the cause of Marf reduction in lrpprc2 mutant clones.
To investigate the role of UPS in Marf downregulation in lrpprc2 mutants, we fed the larvae with the proteasomal inhibitor MG132 [52,53]. MG132 treatment by itself leads to a subtle increase (Mean intensity normalized to DMSO treated cells-1.203 ± 0.06015) in endogenous Marf::mCherry levels which could be owing to reduced basal turnover of Marf by UPS (S2B-S2B" Fig). In lrpprc2 A mutant clones of DMSO fed larvae, Marf:HA levels were low as compared to the neighboring wildtype cells. However, MG132 fed larvae show no change in Marf:: HA levels between wildtype and lrpprc2 A mutant clones (Fig 2A-2A" and 2C-2C"). We further expressed a dominant negative form of Prosβ6 to inhibit UPS activity [54] and tested its effect on Marf::HA levels in lrpprc2 A mutant clones. Similar to MG132 treatment, we found that Marf::HA levels were restored in lrpprc2 A clones upon Prosβ6 1 overexpression (Fig 2D-2D"). These results suggest UPS-mediated degradation of Marf results in Marf reduction in lrpprc2 A clones.

PINK1 and Park dependent Marf regulation in lrpprc2 mutants
Several E3 ubiquitin ligases have been linked to Mitofusin degradation. For example, Mitofusin degradation by HUWE1 occurs under genotoxic stress or under altered fat metabolism conditions [55,56], while Mitofusin degradation by Park occurs on mitochondrial membrane depolarization [9,11] (Fig 3B-3B"). Since park is known to function  [58,59], and previously a role for PINK1 and Park in Marf downregulation has been established [10], we tested whether PINK1 is also required for Marf degradation in lrpprc2 A mutant clones. We generated lrpprc2 A Pink1 5 double mutant clones and found that these clones do not show reduction in Marf::HA levels ( Fig 3C-3C"), suggesting that mitochondrial impairment in lrpprc2 mutant cells could cause PINK1-Park activation and subsequently Marf downregulation. Our observations may relate to reports of downregulation of Mfn1 and Mfn2 upon CCCP treatment as a mechanism to suppress mitochondrial fusion prior to PINK1-Park mediated mitophagy [9,11].

UPR mt may not be sufficient to induce Marf downregulation
The role of the PINK1-Park pathway in mitochondrial quality control is well known. However, the exact mechanism of PINK1-Park activation in in vivo contexts remains unclear. We first checked PINK1 levels in lrpprc2 A mutant clones using a genomically tagged PINK1::Myc line, we found no significant difference in total PINK1::Myc levels between lrpprc2 A mutant clones and neighboring wildtype cells (S3C-S3C" Fig). In cancer cell lines, dissipation of MMP and increased oxidative stress have been shown to activate PINK1-Park on the OMM leading to mitophagy [19,60]. However, we have earlier shown that lrpprc2 mutants do not have increased oxidative stress as compared to control [39]. We checked MMP in lrpprc2 A mutant clones using TMRE, a dye that reversibly stains mitochondria in a membrane potential-dependent manner. We observed that TMRE intensity in lrpprc2 A mutant clones is similar to that of wildtype cells (S3D-S3D" Fig). These observations rule out the possibility that PINK1-Park is activated due to oxidative stress or altered MMP in lrpprc2 mutants.
Mitochondrial unfolded protein response (UPR mt ), which is a cellular response to altered mitochondrial proteostasis, has also been shown to activate PINK1-Park leading to mitophagy [61]. Therefore, we tested for UPR mt activation in lrpprc2 mutants by checking levels of Hsp60, which is reported to be increased due to elevated UPR mt [62]. We found increased levels of Hsp60A protein in lrpprc2 mutant clones suggesting elevated UPR mt (S4A-S4A" Fig). Activation of UPR mt upon the loss of LRPPRC has also been observed in C.elegans and mammalian cells and could be evolutionarily conserved [63]. Increased UPR mt could activate PINK1-Park, leading to Marf degradation. We genetically suppressed the UPR mt response pathways and checked its impact on Marf levels in lrpprc2 mutants. We generated lrpprc2 A mutant clones in the background of crc, foxo or dve knock downs, which are transcription factors mediating UPR mt [64][65][66][67]. None of these interventions affected Marf::HA downregulation in lrpprc2 A clones, suggesting that the activation of these UPR mt pathways may not be causing PINK1-Park activation (S4B-S4D" Fig). However, these interventions would not change the altered mitochondrial proteostasis in lrpprc2 mutants, which can still activate PINK1-Park.
Since, to the best of our knowledge, there is no reported method to suppress mitochondrial proteostasis defects, we asked whether inducing mitochondrial proteostasis defects is sufficient to cause Marf degradation. To induce mitochondrial proteostasis defects, we expressed a mutant form of ornithine transcarbamylase (ΔOTC) that accumulates in an unfolded state in the mitochondrial matrix and is shown to trigger UPR mt in flies [67]. We expressed ΔOTC in the posterior half of the wing disc (marked by RFP) using En>Gal4 (En>Gal4/+; UAS-ΔOTC/ MG132(C-C'). (D-D') lrpprc2 A mutant clones (non green cells, D and dashed white line, D') on overexpression of Prosβ6 1 under Actin>Gal4, wing discs immunostained for Marf::HA (red). Scale bar represents 20μm. (A", B", C" and D") Quantification for relative fluorescence intensities of Marf::HA in lrpprc2 A mutant clones on treatment with, DMSO (A",n = 24), chloroquine (B",n = 23), MG132 (C", n = 10) and on overexpression of Prosβ6 1 under Actin>Gal4 (D",n = 15). Graphs represent intensity values normalized to that of control cells. Two tailed unpaired t-test between control and lrpprc2 A mutant cells. Significance represented by n.s. non significant, p<0.001***. https://doi.org/10.1371/journal.pgen.1010493.g002

PLOS GENETICS
+) and found that ΔOTC expression shows increased Hsp60 levels indicating UPR mt (Mean increase is 1.166 ± 0.0313 times) (S4E-S4E" Fig). However, the Hsp60 level increase was to a lesser extent compared to lrpprc2 mutant clones (Mean increase is 1.376 ± 0.02939 times). We checked Marf::HA levels on wildtype OTC and ΔOTC expression using En>Gal4 and found no change in Marf::HA levels in the posterior half (marked by RFP) as compared to the anterior half of the wing discs for both (S4F-S4G" Fig). Although these observations do not rule out a role for mitochondrial proteostasis in activating PINK1-Park pathway in lrpprc2 mutants, our data suggest that UPR mt induced by expression of ΔOTC is not sufficient to cause Marf degradation.

Bendless, a K63-linked E2 ubiquitin conjugase, is a regulator of Marf
In addition to lrpprc2, through the genetic mosaic screen we also identified ben as a regulator of Marf. We found a subtle but consistent increase in Marf::HA levels in mutant clones of two independent EMS allele of ben (ben A and ben B ) (Figs 4A-4A", S5C-S5C" and S5G), which is similar to that of park Δ21 and Pink1 5 mutant clones (S5A-S5B" Fig). We further confirmed increased Marf levels in ben mutants by western blot using whole larval extracts ( Fig 4E-4E'). In a previous study, ben knockdown by RNAi did not alter Marf levels [68] possibly due to inefficient knockdown-the efficacy of RNAi may not be comparable with the two independent loss of function alleles we used. Ben is a fly homologue of the E2 conjugase UBE2N/ UBC13 with a marked similarity from yeast to humans (S5E Fig). We ruled out the possibility that the increase in Marf::HA levels upon loss of Ben is due to increased mitochondrial content as there was no difference in Tom20::mCherry levels between ben mutant clones and control ( Fig 4B-4B"). Further, we did not find an increase in Marf mRNA levels in ben mutants suggesting that the increase in Marf protein levels is not a consequence of increased transcription ( Fig 4F). These data suggest that Ben regulates Marf levels post-transcriptionally.
Next, we asked whether Ben overexpression can induce Marf degradation. To test this, we generated a C-terminal V5-tagged Ben (UAS-ben::V5) transgenic line for tissue-specific expression of ben. We first confirmed that the fusion protein is biologically functional by complementing the lethality associated with the ben A mutant allele (S5F Fig). We then expressed ben::V5 in the posterior half of the wing disc (marked by green) using the En>Gal4 and found that ben::V5 overexpression did not affect Marf::HA levels ( Fig 4C-4C"). Additionally, we overexpressed an N-terminal HA-tagged Ben (UAS-HA::ben) using En>Gal4 and found no change in Marf::mCherry levels (S6A-S6A" Fig). These data suggest that Ben is necessary but not sufficient for Marf regulation. Since loss of ben, Pink1 or park results in mild upregulation of Marf, we hypothesize that Ben acts in the PINK1-Park pathway to regulate the steady state levels of Marf.

Bendless is essential for Marf downregulation in lrpprc2 mutants
Given that Marf undergoes proteolytic degradation in lrpprc2 mutants, we wanted to check if Ben is involved in Marf degradation not only basally but under mitochondrial stress as well, similar to PINK1 and Park. We thus created lrpprc2 and ben double mutant clones and found that lrpprc2 A ben A and lrpprc2 A ben B double mutant clones showed no reduction in Marf::HA levels, unlike lrpprc2 A mutant clones (Figs 4D-4D" and S5D-S5D"). This suggests that Ben is essential for Marf degradation in lrpprc2 mutant cells.

Ben is required for PINK1 mediated Marf degradation
To study the role of Ben in PINK1-Park mediated regulation of Marf, we tested the functional interaction between ben and Pink1. Since PINK1-Park activity is suppressed on PINK1 We then overexpressed Park in ben A mutant discs-interestingly we found loss of ben could not suppress reduction of Marf::mCherry due to Park overexpression ( Fig 5E-5E") suggesting that overexpression of Park can override loss of ben. Overall, these experiments suggest that Ben is essential for PINK1 mediated regulation of Marf and acts genetically upstream to Park.

Ben regulates PINK1 stability
To understand how Ben regulates PINK1, we first checked Ben and PINK1 protein interaction using co-immunoprecipitation. We used the UAS-Ben::V5 and genomically tagged PINK1:: Myc fly lines (w;PINK1::Myc/+;Actin>Gal4/UAS-ben::V5) and pulled down PINK1:Myc. As shown in Fig 6A, probing for Ben::V5 on pull down of PINK1::Myc shows presence of Ben::V5 indicating Ben and PINK1 directly interact. Further, we checked the effect of loss of ben on PINK1 levels. We performed western blots using whole larval extracts from control and ben A mutants containing genomic tagged PINK1::Myc. We found a significant downregulation of full length PINK1::Myc in ben A mutants, but an increase in low molecular weight PINK1::Myc bands ( Fig 6B-6C'). This suggests that Ben is required for stabilizing full length PINK1. The low molecular weight bands might be products of PINK1 degradation by mitochondrial dashed white line, D'), wing discs immunostained for Marf::HA (red). Scale bar represents 20μm. (A", B", C" and D") Quantification for relative fluorescence intensities of Marf::HA in ben A mutant clones (A", n = 15), Tom20::mCherry in ben A mutant clones (B", n = 7), Marf:: HA levels on Ben::V5 overexpression (C", n = 6) and in lrpprc2 A ben A double mutant clones (D", n = 16). Graphs represent average intensity values normalized to that of control cells. Two-tailed unpaired t-test between control and mutant cells/cells expressing UAS-ben:: V5. (E) Representative western blot for ben mutant (y w ben A FRT19A) and control (y w FRT19A) larval lysate probed for Marf::HA and Actin. (E') Quantification for intensity of Marf::HA band normalized to Actin band intensity for ben A mutant and control larvae (n = 5). Two-tailed unpaired t-test between control and ben A mutant larvae. (F) Quantification of Marf mRNA levels in third instar ben mutant (y w ben A FRT19A) larvae compared to control (y w FRT19A) (n = 4). Two tailed unpaired t-test between control and ben A mutant larvae. Error bars represent S.E.M. Significance represented by n.s.-non significant, p<0.05*, p<0.01**, p<0.0001***.

Ben regulates mitochondrial dynamics under mitochondrial stress
Given the role of Ben in Marf regulation under steady state conditions as well as in lrpprc2 mutants, we sought to investigate and compare mitochondrial morphology between control, ben, lrpprc2 and lrpprc2 ben double mutants. First, we compared mitochondrial morphology in mutant clones in peripodial cells of wing discs using mitotracker red staining and live imaging (Fig 7A-7D). We found that the mitochondrial morphology in ben mutant cells is comparable to wildtype cells-they both show a filamentous network of mitochondria. A previous study, however, has shown increased mitochondrial size due to ben knockdown in the fat body -the difference in the phenotype could be due to tissue specific differences in mitochondrial physiology [71]. In lrpprc2 A mutant cells, we observe filamentous mitochondria along with large aggregated mitochondria and ring-shaped mitochondria (Fig 7A-7F). Further, we found that the large aggregated mitochondria and ring-shaped mitochondrial phenotype are worsened in lrpprc2 A ben A double mutant cells (Fig 7C-7F). Compared to fixed samples (S1D-S1D" Fig), live imaging showed more tubular and networked mitochondria (Fig 7A-7D). This difference in mitochondrial morphologies could be owing to the difference in sample preparations as also documented earlier [72]. However, in both scenarios lrpprc2 A mutants consistently show presence of large mitochondria as compared to wildtype cells. We further investigated mitochondrial morphology in larval muscles using Complex V antibody staining and we found wildtype and ben A mutants show a comparable filamentous network of mitochondria (Figs 7G-7H and S8A-S8B); lrpprc2 mutants show distinctive large globular mitochondria along with filamentous and ring shaped mitochondria (Figs 7I and S8C); lrpprc2 ben double mutants rarely show filamentous mitochondria, instead, we observed a significant increase in the size and frequency of large globular and ring-shaped mitochondria as compared to lrpprc2 (Figs 7J and S8D). We also observed that in lesser frequency mitochondria in lrpprc2 A ben A double mutants form clusters, especially around the nucleus which is not observed in either lrpprc2 A or ben A mutants (S8D Fig). To report these mixed phenotypes we have documented several images for each genotype in Figs 7 and S8.
To further resolve the mitochondrial morphology and quantify various features of individual mitochondria in larval muscles we used mitochondrial photoactivatable GFP (Mito-PA-GFP). This allows visualization of individual mitochondria and its network within a cell by activating GFP fluorescence, using 405 nm laser, in a region of interest [73,74]. We found a comparable filamentous network of mitochondria in control and ben A mutants as they show similar branch numbers and aspect ratios (Fig 8A-8B" and 8E-8G). In lrpprc2 A mutants we observed globular shaped mitochondria characterized by larger area and lower aspect ratio.
https://doi.org/10.1371/journal.pgen.1010493.g005 lrpprc2 A mutants also show marked reduction in mitochondrial network as characterized by reduced branch number (Fig 8C-8C" and 8E-8G). In lrpprc2 A ben A double mutants we observed the presence of globular shaped mitochondria characterized by larger area and lower aspect ratio (Fig 8D-8D" and 8E-8G). As compared to lrpprc2 A , in which globular mitochondria remain isolated, in lrpprc2 A ben A we observed globular mitochondria are interconnected (Fig 8D-8D"). This is also reflected in the increase in branch numbers in the case of lrpprc2 A ben A double mutants when compared to lrpprc2 A mutants (Fig 8G). Overall, our results suggest that mitochondrial dysfunction in lrpprc2 may induce the formation of globular mitochondria and Ben mediated regulation of Marf suppresses their fusion (Fig 9E).  (Fig 9A-9D). This suggests that Marf regulation by Ben is a neuroprotective mechanism. Additionally, we also investigated adult wing phenotype in lrpprc2 ben double mutants. We used Ubx-FLP to generate large clones in the developing wing. Most of the wings in lrpprc2 A and ben A single mutants were normal, with occasional minor defects in bristle pattern in both mutants (S9A- S9C Fig). However, in lrpprc2 A ben A double mutants we see the following: most flies eclose with improperly folded wings, additionally these mutant wings show wing patterning defects such as presence of ectopic veins and dark patches on the wing blade (S9D Fig). Together these data suggest a protective role of Bendless upon mitochondrial stress and this becomes prominent in the case of neurons, possibly due to their high energy requirements.

Discussion
To identify novel regulators of mitochondrial fusion in an in vivo system, we screened fly mutants for altered Marf levels and identified mutations in lrpprc2 causing reduction in Marf levels ( Fig 1A). We found that in lrpprc2 mutants, Marf is degraded by the UPS (Fig 2C-2D) in a PINK1-Park dependent mechanism (Fig 3A-3C). In the screen, we also identified mutations in the E2 conjugase ben, causing subtle Marf upregulation (Fig 4A). We found that Ben is essential for PINK1 stability (Fig 6B), regulates Marf levels ( Fig 4D) and mitochondrial morphology (Figs 7D, 7J and 8D) in lrpprc2 mutants. We also found that a combined loss-of-function mutation of lrpprc2 and ben in the eyes results in accelerated retinal degeneration ( Fig  9D) and developmental abnormalities in wings (S9D Fig). Indicating that under mitochondrial stress induced by loss of lrpprc2, Ben mediated regulation of mitochondrial dynamics is a protective response (Fig 9E). represent S.E.M. A one-way ANOVA-Tukey's multiple comparison test was used to calculate the significance between the samples in graph (E) and (F). Significance represented by n.s-non significant, p<0.05*, p<0.01**, p<0.0001*** (G-J) Confocal sections of third instar larval muscles immunostained for Complex V (gray) in control(G), ben A (H), lrpprc2 A (I) and lrpprc2 A ben A (J) larvae. Representative individual mitochondrial morphology is marked by different colors: filamentous (red), large globular (yellow) and ring (blue). Scale bar represents 5μm. https://doi.org/10.1371/journal.pgen.1010493.g007

PLOS GENETICS
Increased mitochondrial size and globular mitochondrial phenotype, as observed in lrpprc2 mutants (Figs S1E, 7C and 8E), have been observed in certain metabolic diseases [75,76]. Increased mitochondrial size has also been observed upon loss of lrpprc2 homologs in C.elegans, mouse and human cell lines [42,43] as well as in other mutants where the ETC is compromised [62,77,78]. The mechanism of such responses, and how these unusual shaped mitochondria contribute to cell physiology and disease progression, is not clear. As SIMH has been observed in a bid to increase oxidative phosphorylation under various cellular and mitochondrial stresses [5,79,80], we hypothesize that reduced ETC activity and mitochondrial stress in lrpprc2 [38,39,43] can induce mitochondrial enlargement (Figs 6D and S1E), similar to SIMH, through an unknown mechanism. Since SIMH increases ATP synthesis and inhibits mitophagy [3,5,81,82], increased mitochondrial size could be a compensatory adaptation in lrpprc2 mutants in response to a bioenergetic deficit or mitochondrial stress. Further, globular shaped mitochondria could also be a favorable adaptation as it is recently shown that, in comparison to elongated mitochondria, globular mitochondria contain densely packed cristae membranes with high curvature. As ATP synthase is known to localize at intense curvature in cristae, globular mitochondria might possess better energetic capabilities [83,84]. Further ultrastructural analysis along with mitochondrial activity assays may shed light on this possibility.
Despite the increased mitochondrial size (S1E Fig), we observed Marf downregulation in lrpprc2 mutants (Fig 1A). We hypothesize that, while an adaptive mechanism may induce SIMH (cellular response), MQC may induce Marf degradation to suppress the fusion of dysfunctional mitochondria (mitochondrial response) (Fig 8D). Indeed we observed that Marf reduction in lrpprc2 mutants is correlated with the presence of large globular mitochondria that remain isolated (Fig 8E-8G). A similar scenario of isolation of dysfunctional mitochondria prior to mitophagy has been proposed earlier [9][10][11]. Alternatively, increased mitochondrial size in lrpprc2 mutant cells may induce the PINK1-Park pathway to limit mitochondrial fusion by Marf degradation. A similar hypothesis was also proposed by Yamada et al. wherein loss of Drp1 results in Parkin dependent Mitofusin downregulation [85].
In steady state conditions, PINK1 is imported into the mitochondria and cleaved by mitochondrial peptidases, it then retro translocates to the cytoplasm and is degraded by UPS to limit PINK1-Park activity [70,89,90]. The initial remarkable discovery by Narendra et al. that CCCP which dissipates MMP, induces PINK1-Park-dependent mitophagy in cancer cells provided an unparalleled assay to investigate the mechanism further [19,91]. Further studies also show that increased oxidative stress or UPR mt stabilizes full-length PINK1, which then recruits Park leading to ubiquitination of OMM proteins and mitophagy [10,19,61,92,93]. Given no change in MMP (S3D Fig) and oxidative stress in lrpprc2 mutants [38,39,49], we suspected that impaired mitochondrial proteostasis activates PINK1-Park to downregulate Marf. However, activation of UPR mt by ΔOTC expression did not result in Marf degradation suggesting that activation of UPR mt alone may not be sufficient to activate PINK1-Park mediated Marf degradation in vivo (S4G Fig). Identification of the nature of the mitochondrial stressors leading to PINK1-Park activation in lrpprc2 will require further investigation.
Several regulators of PINK1 stability and activity have been identified. For example, CHIPmediated K48-ubiquitination promotes PINK1 turnover [94], while BAG2, a chaperon, prevents ubiquitination and promotes PINK1 stability [95,96]. We found that Marf degradation in lrpprc2 mutants or by Pink1 overexpression is completely suppressed in the absence of the K63-linked E2 conjugase Ben (Figs 4D, 5C and S5D). However, park overexpression could cause Marf downregulation even in the absence of ben (Fig 5E). This suggests that Ben regulates PINK1 mediated Marf degradation and that Park may not be directly regulated by Ben. Previous studies have observed that the mammalian homolog of Ben, UBE2N, is dispensable for mitophagy but facilitates the clustering of mitochondria during CCCP-induced mitophagy [68,97,98]. We also found that the loss of ben does not alter developmental mitophagy during larval midgut remodeling (S7B Fig), which has been shown to be dependent on PINK1-Park [22,23].
K63 ubiquitination of PINK1 by the Traf6-SARM1 complex is shown to stabilize PINK1 on depolarized mitochondria in mammalian cells [99]. As Ben protein interacts with PINK1 ( Fig  6A) and loss of ben results in reduced PINK1 levels (Fig 6B), Ben is likely to increase the stability of PINK1 by K63 ubiquitination. Indeed, human PINK1, in cell culture systems, is known to be ubiquitinated at K137 by both K48 and K63 linkages [100]. While K48 chains are linked with PINK1 degradation; the significance of the K63 linkage is not obvious. K63 ubiquitination is suggested to protect proteins from proteasomal degradation [101]. Overall, Ben-mediated K63 ubiquitination could be responsible for PINK1 stability and remains to be tested. We hypothesize that absence of Ben could lead to increased import of PINK1 into the mitochondria hence reducing its full-length levels, even on overexpression of Pink1. A similar observation was made by Sekine et. al. with reference to Tom7 [102]. Wherein loss of Tom7, a component of the TOMM complex, resulted in mitochondrial import of PINK1 and cleavage by OMA1 [102]. Additionally, whether Ben regulates PINK1 activity needs further study.
Ben-PINK1-Park regulation of Marf appears to be a homeostatic function which is further activated in response to aberrant mitochondrial function. Compared to lrpprc2, lrpprc2 ben mutants show increased number of ring shaped, globular and large aggregated mitochondria. We also observed that large globular mitochondria are interconnected in lrpprc2 ben double mutants whereas globular mitochondria remain isolated in lrpprc2 mutants, possibly due to the reduction of Marf. This Ben-PINK1-Parkin mediated Marf degradation in lrpprc2 appears to be a cell protective mechanism as lrpprc2 ben double mutants show accelerated retinal degeneration and worsened adult wing phenotype as compared to lrpprc2 mutant (Figs 9 and S9). Given that mutations in LRPPRC result in Leigh syndrome, it is likely that Ben/Ubc13-PINK1-Park may regulate Mfn1 and Mfn2 in Leigh syndrome as well in other mitochondrial diseases. Indeed, altered mitochondrial dynamics has been reported in many mitochondrial diseases [76,[103][104][105]. Thus, further studies on the mechanisms of Ben/Ubc13-PINK1-Park activation will be crucial for understanding mitochondrial quality control in mitochondrial disease.

Drosophila culture
Flies were cultured on standard media containing sucrose, malt, yeast and corn flour at room temperature. Crosses were maintained at 25˚C. Crosses involving RNAi were maintained at 28˚C. Drosophila larvae expressing UAS-Prosβ6 1 were maintained at 25˚C till 3 rd instar stage and were then transferred to 28˚C for 24 hours before dissection, to avoid cell death observed on prolonged inhibition of proteasomal activity. To activate the FLP-FRT system, heat shock was given during first instar larval stages at 37˚C for 1hr. Genotypes used are as listed in Table 1.
For drug treatments, 3 rd instar larvae were transferred to food containing 3mg/ml chloroquine [106], 100μM MG132, or DMSO (vehicle control) for 24 hours prior to dissection. For western blot and qPCR, 3 rd instar larvae were used. We observed that development of lrpprc2 A mutant larvae is substantially delayed. Therefore, we used size matched 3 rd instar lrpprc2 A mutant larvae that are obtained after 14-15 days post hatching.

Immunofluorescence and imaging
Larvae were dissected in 1X PBS, followed by fixing in 4% paraformaldehyde (Himedia-TCL-119 -100ml) for 30 minutes at room temperature and three washes in 1X PBS with 0.2% Tri-tonX-100 (Himedia-MB031, 1X PBST). Primary antibodies were incubated overnight at 4˚C. Followed by blocking in 5% normal goat serum (Himedia-RM10701) for 1h at room temperature and then secondary antibody incubation followed by washing and dissection. Samples were mounted in Vectashield (VectorLabs-H100) and imaged under 40X or 63X oil immersion Leica Stellaris 5 or Olympus FV3000 confocal microscopes. Images were processed using Fiji. All antibody dilutions and the blocking solution were made in 1X PBST; details of antibodies and their dilutions used are listed in Table 2. Wing clones were made by crossing heterozygous mutant flies with using UbiGFP frt19A; Ubx-FLP. The flies were anesthetised and fixed in 70% ethanol and stored at 4˚C. For mounting, the wings were dissected from the flies and transferred to 100% ethanol. The wings were then mounted in DPX (Sigma 06522-100mL). The DPX media was allowed to evenly spread by applying weight on the coverslip and incubating the slides at 60˚C. The samples were imaged using transmitted light in a Leica-DMi8 inverted microscope using a 4X objective.

Western blot
3 rd instar larvae were crushed in RIPA lysis buffer [50mM Tris,150mM NaCl, 0.2% Triton X 100 and 1X protease and phosphatase inhibitor cocktail (Thermo Fisher-A32965, A32957 respectively)], followed by centrifugation at 16,000g for 10 mins at 4˚C. Clear fat-free supernatant was used for total protein estimation by BCA assay (Thermo Scientific-23227). Lysate was mixed with equal volume of 1X Laemmli buffer (0.004% bromophenol blue, 20% glycerol, 4% SDS and 0.125M Tris-HCl pH 6.8) having 5% beta-mercaptoethanol and heated at 98˚C for 5 minutes, centrifuged, and 25μg of protein was loaded in each well and resolved on 4-15% gradient Tris-Glycine gel (Bio-Rad-4561086). Semi-dry transfer was done onto 0.2μm Nitrocellulose membrane as per Trans-BlotTurbo Kit (Bio-Rad-1704270) for seven minutes. Blocking in either 5% Blotto (Santa Cruz sc-2325) or 5% BSA made in 1X TBS with 0.1% Tween-20 (1X TBSTw20) for 1 hour at room temperature followed by primary antibody incubation overnight at 4˚C. After washing thrice in 1X TBSTw20, membranes were incubated in HRP conjugated secondary antibodies (Table 2) for 2 hours at room temperature. After washing, they were developed using Clarity Western ECL Substrate (Bio-Rad-1705061) and visualized using Vilber-Lourmat chemidoc. Band intensities were quantified using Fiji and normalized with Actin.

PLOS GENETICS
glycerophosphate, 50mM NaF, 1mM orthovanadate and 0.3% CHAPS) and incubated for 30 mins on ice. Then centrifuged at 16,000g for 10 mins at 4˚C and the supernatant was aspirated out carefully. Protein estimation was done by using the BCA method (Thermo Scientific-23227). To the remaining beads 300ul co-IP buffer was added and incubated with equilibrated 35μl Myc beads (Sigma-E6654) for 4 hrs at 4˚C. It was then centrifuged and the supernatant discarded. Beads were washed thrice with the co-IP buffer. The beads were then incubated with 50μl anti-Myc peptide and incubated for 4 hrs at 4˚C. After that the sample was centrifuged and supernatant having elute was collected carefully in fresh tube and processed for Western blot as given above.

Real-time PCR
3 rd instar larvae were used for RNA isolation using TRIzol (Ambion life tech-15596018) method. cDNA conversion for 1μg of RNA was carried out using a cDNA conversion kit (Thermo Fisher-4368814). qPCR was carried out in 96 well plates in three technical replicates for each of the three biological replicates. Marf qPCR was done using the iTaq SYBR Green supermix (Bio-Rad -1725121) using LightCycler 96 ( Fig 5F).

Mitochondrial morphology analysis
Wing discs immunostained for Complex V were imaged using Leica Stellaris 5 confocal microscope at 63X oil objective. The mitochondria were segmented on Fiji using the Trainable Weka segmentation plugin [108]. The segmented images were then used to find out the mitochondrial area using Particle Analyze Tool on Fiji. Blind test: for qualitative assessment of mitochondrial morphology in larval muscle, we renamed a set of images containing mitochondria from larval muscles with random numbers. The images from different genotypes (control, ben A , lrpprc2 A , and lrpprc2 A ben A ) were pooled and were assessed for the presence of different mitochondrial morphologies, including presence or absence of mitochondria network, large globular mitochondria, ring-shaped mitochondria and mitochondrial aggregates. Multiple images were used for the assessment, 40 images from 11 larvae for control, 24 images from 7 larvae for ben A , 27 images from 7 larvae for lrpprc2 A , and 32 images from 9 larvae for lrpprc2 A ben A .

Mitochondrial morphology analysis using mito-PA-GFP
Third instar larvae expressing C57>Gal4, UAS-mito-dsRed, and UAS-mito-PA-GFP were fileted in Schneider's Drosophila Medium. Before imaging, fresh Schneider's medium with 5mM glutamate was added to block neurally evoked muscle contractions. Imaging was done using Leica Stellaris 5 confocal using a 63X water dipping objective. To activate GFP, a 405nm laser at 100% power was used at designated ROIs for 10 iterations. After activation, 5 images were acquired at 3min intervals in RFP and GFP channels. The images were segmented using the Fiji-Trainable Weka segmentation plugin. The segmented images were used to get various mitochondrial morphology parameters using the Fiji-Mitochondria Analyzer plugin. Parameters including branch number (proxy for network), aspect ratio-ratio of the major axis to the minor axis of a mitochondria (proxy for globular v/s tubular) and Area (proxy for size) were then compared using one-way ANOVA-Bonferroni's Multiple Comparison Test. The photoactivation protocol was modified from Chowdhary et. al. 2017 [73].

Statistics analysis
At least three independent experiments were used for all quantifications, the n values for each experiment is indicated in their respective figure legends. n represents the number of clones/ regions used for the analysis. Two-tailed unpaired t-test was used to analyze data obtained from clonal analysis; One sample t-test was used to analyze the data in S5 Fig. Two-tailed unpaired t-test was used to analyze all other data sets. ANOVA-Bonferroni's Multiple Comparison Test was used to compare data sets in Figs 5C", 7I, 7J, 8E, 8F and 8G. Significance of the data was represented as * for p<0.05, ** for p<0.01, and *** for p<0.0001. All statistical analyses were carried out using GraphPad Prism software version 9.  " and B") Quantification for relative fluorescence intensities of Marf::mCherry in control (A", n = 12) and chloroquine treated larvae (A",n = 10) and DMSO (B",n = 12) and MG132 (B",n = 12) treated larvae. Graphs represent average intensity values normalized to control/DMSO. Two tailed unpaired t-test between control and chloroquine and between DMSO and MG132 treatments. (C-C') lrpprc2 A mutant clones (non green cells, C and dashed white line, C'), wing discs immunostained for endogenous p62 (red). Scale bar represents 20μm. (C") Quantification for relative fluorescence intensities of p62 in lrpprc2 A mutant clones (n = 12). Graphs represent average intensity values normalized to that of control cells. Two tailed unpaired t-test between control and lrpprc2 A mutant cells. Significance represented by n.s.-non significant, p<0.05 *. (TIF) S3 Fig. (A-A')