Mitochondrial proteases maintain cellular protein homeostasis and tissue integrity Authors: Kexin Shi, Hao Liu, Hong Xu, Weina Shang, Liquan Wang, Chao Tong Published: 2026 ABSTRACT BackgroundMitochondrial proteases are essential for mitochondrial protein import and constitute the core of the organelle’s intrinsic protein quality control system. However, their physiological functions across tissues, as well as their influence on cytosolic proteostasis, remain incompletely understood.ResultsWe generated loss- and gain-of-function alleles for 15 conserved mitochondrial proteases in Drosophila melanogaster to systematically dissect their in vivo functions. Disruption of specific proteases caused male sterility or organismal lethality, whereas tissue-specific knockouts in the eye, muscle, or fat body led to mitochondrial protein aggregates, structural defects, and age-dependent degeneration. Loss of UQCR-C1 or Afg3l2 robustly increased mitophagy, while overexpression of several proteases severely impaired muscle integrity. Loss of UQCR-C1, Mppa, or CG11771 promoted HTT72Q aggregation, and reducing UQCR-C1 or Afg3l2 markedly elevated cytosolic HTT72Q levels. Conversely, overexpressing Mppa—but with reduced efficacy in its disease-associated variants—suppressed HTT96Q aggregation and neuronal toxicity. Mppa forms a complex with UQCR-C1 to regulate mitochondrial pre-protein processing and import, indicating that enhancing mitochondrial protein import is sufficient to alleviate cytosolic proteotoxic stress caused by HTT polyglutamine (polyQ) proteins.ConclusionsThis work establishes a comprehensive in vivo resource for mitochondrial protease functions and their roles in shaping cytosolic proteostasis.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s13578-026-01612-0. Background Protein homeostasis is fundamental for cellular function and health [11]. As the central organelle for energy production and metabolism, mitochondria maintain their own proteostasis through a specialized network of mitochondrial proteases [8, 22]. These proteases, localized to distinct mitochondrial compartments, play critical roles in protein import, processing, maturation, regulation, and degradation [31]. Proper protease activity is essential for mitochondrial integrity, and its disruption often results in severe human diseases [3, 8, 31]. Recent studies revealed that certain cytosolic and disease-related proteins can be imported into mitochondria and degraded under stress conditions [17, 32, 38, 39], suggesting that cytosolic and mitochondrial protein homeostasis are closely interconnected. The activity of mitochondrial proteases therefore influences not only mitochondrial function but also the overall cellular proteome [17, 32]. Mitochondrial proteases are highly diverse and can be classified by catalytic mechanism or function [11, 31]. Based on their proteolytic chemistry, they include metalloproteases, serine proteases, and cysteine proteases [31]. Metalloproteases use a metal ion, typically zinc, for catalysis, whereas serine and cysteine proteases employ a catalytic serine or cysteine residue. In addition, mitochondria harbor pseudo-proteases that lack one or more catalytic residues; although inactive, some may perform non-proteolytic structural or regulatory roles [31]. Functionally, mitochondrial proteases can be grouped into two major classes: those involved in mitochondrial protein import and maturation, and those responsible for protein quality control [31]. Mitochondria possess a double-membrane structure consisting of the outer mitochondrial membrane (OMM) and inner mitochondrial membrane (IMM), with the intermembrane space (IMS) between them. Although mitochondria encode a small number of oxidative phosphorylation (OXPHOS) subunits, most mitochondrial proteins are synthesized in the cytosol and imported via a sophisticated transport system. These precursor proteins usually contain an N-terminal mitochondrial targeting sequence (MTS) that must be removed after import to allow proper folding and activation [2, 24]. The mitochondrial processing peptidase (MPP), located in the matrix, is essential for MTS removal [15]. It is a heterodimer composed of MPPα and MPPβ subunits, encoded by PMPCA and PMPCB in humans. MPPα is a pseudo-enzyme that forms a substrate-binding cavity with the catalytic MPPβ containing a Zn²⁺-binding site [34]. The MPP complex recognizes and cleaves the MTS, after which proteins may further processed and fold within the matrix or sorted to the IMM or IMS. After import, some matrix or IMM-embedded proteins undergo an additional maturation step catalyzed by the mitochondrial intermediate peptidase (MIP), which removes an N-terminal octapeptide [37], or by X-Pro aminopeptidase 3 (XPNPEP3), which cleaves a single amino acid [36]. Certain mitochondrial DNA (mtDNA)-encoded proteins also undergo processing by methionine aminopeptidase 1D (METAP1D), which removes the initiator methionine [33]. Proteins destined for the IMS are further processed by the mitochondrial inner membrane peptidase (IMP), which removes hydrophobic sorting signals after MTS cleavage [1]. In humans, IMP comprises IMM-localized subunits IMMP1L and IMMP2L, both containing catalytic Ser and Lys residues. IMP substrates include both nuclear- and mitochondrial-encoded proteins [25]. Mutations in PMPCA, PMPCB, or IMMP2L have been linked to cerebellar ataxia, Leigh-like syndrome, severe neurodegeneration, autism, and other neurological disorders [15, 31]. In addition to processing imported proteins, many mitochondrial proteases form a quality control network that degrades misfolded, oxidized, or damaged proteins [22, 31]. In the IMM, non-assembled or damaged respiratory chain subunits are degraded by ATP-dependent AAA (ATPases Associated with diverse cellular Activities) proteases [7, 35]. The intermembrane space AAA protease (i-AAA), represented by YME1L [26], has its zinc metalloprotease domain facing the IMS, while the matrix AAA protease (m-AAA), composed of AFG3L2 alone or with SPG7, acts toward the matrix [5]. Oma1, a metalloprotease that shares overlapping activity with the m-AAA protease, also degrades misfolded proteins within the inner membrane [21]. Within the matrix, the LONP1 serine protease degrades misfolded or oxidized proteins [6], and CLPP, together with CLPX, forms the CLPXP complex to remove damaged substrates [23]. The resulting peptides are further processed into amino acids by peptidases such as PITRM1. In the IMS, the trimeric serine protease HTRA2 and the metalloprotease ATP23 also participate in protein quality surveillance. Mitochondrial oligopeptidase M (MEP; also known as neurolysin, NLN) may also function as oligopeptidase in IMS. Mutations in AFG3L2, SPG7, CLPP, or HTRA2 are associated with various inherited neurological disorders, underscoring the essential role of mitochondrial proteases in maintaining organellar and cellular homeostasis [31]. Although many studies have investigated mitochondrial proteases, their functions remain incompletely defined due to the diversity of their substrates and the complexity of their regulatory mechanisms [31]. In addition to ensuring proper mitochondrial protein import and quality control, recent studies have revealed that mitochondrial proteases also degrade cytosolic proteins that are shuttled into mitochondria under stress conditions. In yeast, when cells experience heat shock and accumulate large amounts of misfolded proteins, these misfolded proteins bind to Tom40 and Tom70 and are subsequently imported into mitochondria for degradation [32]. In mammalian cells, when proteasome activity in the cytosol is inhibited, cytosolic proteins can be transported into mitochondria for degradation with the assistance of the mitochondrial outer membrane protein FUNDC1 [17]. These findings suggest that under stress conditions, mitochondria may utilize their protein import machinery together with proteases and folding systems to alleviate cytosolic proteostasis crises. Our previous study showed that blocking the mitochondrial protein import machinery led to cytosolic accumulation of HTT polyQ protein aggregates [19]. In recent years, studies have found that disease-causing proteins such as amyloid β protein (Aβ) in Alzheimer’s disease [10], α-synuclein in Parkinson’s disease [4], and TAR DNA-binding protein 43 (TDP-43) in amyotrophic lateral sclerosis can all enter mitochondria [20]. Moreover, TDP-43 has been reported to be imported into mitochondria, and inhibition of LonP1 was found to exacerbate its toxicity [38]. These findings raise the possibility that enhancing mitochondrial protease activity may help mitigate cytosolic proteostasis disturbances caused by misfolded disease proteins. Therefore, understanding the physiological consequences of increased mitochondrial protease activity, which has been rarely studied, is also critical. To systematically investigate the in vivo functions of mitochondrial proteases, we generated both loss- and gain-of-function alleles for 15 evolutionarily conserved mitochondrial proteases in Drosophila melanogaster (Table 1; Fig. 1B) and examined their roles across different tissues. We further assessed how alterations in these proteases influence HTT polyQ aggregate formation and neurotoxicity. Notably, overexpression of wild-type Mppa, the Drosophila ortholog of PMPCA, protected neurons expressing HTT polyQ protein, whereas its disease-associated variants were less effective. Table 1Human orthologs, associated diseases, and expression patterns of Drosophila mitochondrial proteasesGene nameMammalian ortholog(s) encoding mito protease*Related DiseasesExpression patterns CG5188 METAP1D (13), METAP1 (4), METAP2 (2)Not Applied (NA)Broad Mipep/CG7791 MIPEP (14)Left ventrichlar non-compaction cardiomyopathy (LVNC) and developmental delay, seizures, hypotonia, intellectual disabilityBroad UQCR-C1/CG3731 PMPCB (14), PMPCA (2)Multiple mitochondrial dysfunctions syndrome 6Broad CG3107 PITRM1 (14)Autosomal recessive spinocerebellar ataxia-30Broad CG5131 ATP23 (14)NABroad Afg3l2/CG6512 AFG3L2 (13), YME1L (3), SPG7 (2)Dominant hereditary Spinocerebellar Ataxia Type 28 (SCA28), Spastic ataxia–neuropathy syndromeBroad, extremely high in testis Spg7/CG2658 SPG7 (10), AFG3L2 (4), YME1L (3), LONP1 (1), LONP2 (1)Spastic paraplegiaBroad YME1L/CG3499 YME1L1 (13), AFG3L2 (4), SPG7 (2), CLPX (1),Optic atrophy 11Broad, extremely high in testis Immp1/CG9240 IMMP1L (14), IMMP2L (4) NABroad Immp2/CG11110 IMMP2L (14), IMMP1L (4)Gilles de la Tourette syndrome, Autism, Alzheimer’s diseaseBroad, extremely high in testis ClpP/CG5045 CLPP (14)Perrault syndrome 3Broad, extremely high in testis HtrA2/ CG8464 HTRA2 (12), HTRA1 (10), HTRA3 (10), HTRA4 (8)Parkinson’s disease type 13, Essential tremorBroad, extremely high in testis Lon/CG8798 LONP1 (14), LONP2 (4), Spg7 (1)Multiple congenital anomalies (CODAS) syndromeBroad CG11771 NLN/MEP (7), MIPEP (3)LVNC and developmental delay, seizures, hypotonia, intellectual disabilityBroad Mppa/CG8728 PMPCA (14), PMPCB (2)Cerebellar ataxiaBroad, extremely high in testis* Only genes encoding mitochondrial proteases are listed. The numbers in parentheses indicate how many of the 14 prediction sources identified the corresponding orthologs. The ortholog with the highest score is shown in bold Table 1Human orthologs, associated diseases, and expression patterns of Drosophila mitochondrial proteasesGene nameMammalian ortholog(s) encoding mito protease*Related DiseasesExpression patterns CG5188 METAP1D (13), METAP1 (4), METAP2 (2)Not Applied (NA)Broad Mipep/CG7791 MIPEP (14)Left ventrichlar non-compaction cardiomyopathy (LVNC) and developmental delay, seizures, hypotonia, intellectual disabilityBroad UQCR-C1/CG3731 PMPCB (14), PMPCA (2)Multiple mitochondrial dysfunctions syndrome 6Broad CG3107 PITRM1 (14)Autosomal recessive spinocerebellar ataxia-30Broad CG5131 ATP23 (14)NABroad Afg3l2/CG6512 AFG3L2 (13), YME1L (3), SPG7 (2)Dominant hereditary Spinocerebellar Ataxia Type 28 (SCA28), Spastic ataxia–neuropathy syndromeBroad, extremely high in testis Spg7/CG2658 SPG7 (10), AFG3L2 (4), YME1L (3), LONP1 (1), LONP2 (1)Spastic paraplegiaBroad YME1L/CG3499 YME1L1 (13), AFG3L2 (4), SPG7 (2), CLPX (1),Optic atrophy 11Broad, extremely high in testis Immp1/CG9240 IMMP1L (14), IMMP2L (4) NABroad Immp2/CG11110 IMMP2L (14), IMMP1L (4)Gilles de la Tourette syndrome, Autism, Alzheimer’s diseaseBroad, extremely high in testis ClpP/CG5045 CLPP (14)Perrault syndrome 3Broad, extremely high in testis HtrA2/ CG8464 HTRA2 (12), HTRA1 (10), HTRA3 (10), HTRA4 (8)Parkinson’s disease type 13, Essential tremorBroad, extremely high in testis Lon/CG8798 LONP1 (14), LONP2 (4), Spg7 (1)Multiple congenital anomalies (CODAS) syndromeBroad CG11771 NLN/MEP (7), MIPEP (3)LVNC and developmental delay, seizures, hypotonia, intellectual disabilityBroad Mppa/CG8728 PMPCA (14), PMPCB (2)Cerebellar ataxiaBroad, extremely high in testis* Only genes encoding mitochondrial proteases are listed. The numbers in parentheses indicate how many of the 14 prediction sources identified the corresponding orthologs. The ortholog with the highest score is shown in bold Human orthologs, associated diseases, and expression patterns of Drosophila mitochondrial proteases * Only genes encoding mitochondrial proteases are listed. The numbers in parentheses indicate how many of the 14 prediction sources identified the corresponding orthologs. The ortholog with the highest score is shown in bold Fig. 1Mitochondrial proteases in fruit flies are important for survival or male fertility. A A scheme illustrating the mutant alleles generated by CRISPR/Cas9-mediated KO system, including the sgRNA target region (underlined nucleotides) and corresponding indels. B A list of 15 evolutionarily conserved mitochondrial protease genes analyzed in this study, including their gene names, enzyme classes, predicted functions, KO lethality, and fertility phenotypes. C Loss of CG3107, Immp2 or CG11771 led to male sterility. Fertility rates were calculated by mating individual male or female mutants with w¹¹¹⁸ females or males, respectively. n = 25 individual mutant. D The overall size of CG3107, Immp2, and CG11771 KOs’ testes remained largely similar, the seminal vesicles exhibited a notable reduction in size, which implies reduced or absent sperm storage and suggesting defects in sperm production. The seminal vesicles were indicated with red arrows. E Elongated spermatids were identified in CG3107, Immp2, and CG11771 mutants, whereas motile mature sperm were absent Mitochondrial proteases in fruit flies are important for survival or male fertility. A A scheme illustrating the mutant alleles generated by CRISPR/Cas9-mediated KO system, including the sgRNA target region (underlined nucleotides) and corresponding indels. B A list of 15 evolutionarily conserved mitochondrial protease genes analyzed in this study, including their gene names, enzyme classes, predicted functions, KO lethality, and fertility phenotypes. C Loss of CG3107, Immp2 or CG11771 led to male sterility. Fertility rates were calculated by mating individual male or female mutants with w¹¹¹⁸ females or males, respectively. n = 25 individual mutant. D The overall size of CG3107, Immp2, and CG11771 KOs’ testes remained largely similar, the seminal vesicles exhibited a notable reduction in size, which implies reduced or absent sperm storage and suggesting defects in sperm production. The seminal vesicles were indicated with red arrows. E Elongated spermatids were identified in CG3107, Immp2, and CG11771 mutants, whereas motile mature sperm were absent Methods Fly strains All Drosophila melanogaster strains were cultured on standard cornmeal food at a constant temperature of 25 °C and a constant relative humidity of 60%, with a strict 12 h light: 12 h dark photoperiod. Additional yeast supplement was provided for promoting egg-laying. Fly strains utilized in this study were sourced from Tsinghua Fly Center (Tsinghua University, Beijing, China), Bloomington Drosophila Stock Center (Indiana University, Bloomington, IN, USA), and FlyORF (University of Zurich, Zurich, Switzerland) and transgenic strains especially generated for this research. All transgenic strains were generated through plasmid constructs microinjection, performed by the Core Technology Facility of the Center for Excellence in Molecular Cell Science, Chinese Academy of Sciences (CAS, Shanghai, China). Detailed information on the genotypes and stock numbers of all strains is listed in Supplemental Table S1, S2. S2 cell culture and transfection Drosophila melanogaster S2 cells (ThermoFisher, R69007) were cultured in Schneider’s medium (Sigma-Aldrich, S0146) with 10% fetal bovine serum (Life Technologies, 10091148) and 1% penicillin‒streptomycin (Biological Industries, 03-031-1B) under constant conditions of 28 °C temperature and 56% relative humidity. The cells were purchased in 2020 and cryopreserved at low passage numbers. Cultures within 20 passages were used for experiments and were routinely tested to ensure they remained contamination-free. pUAS constructs were transiently transfected into S2 cells along with ub-GAL4 plasmids using Lipofectamine 2000 transfection reagent (Invitrogen, 11668-019). S2 cells were harvested 36–72 h post-transfection for subsequent analyses. Plasmid constructs Plasmid used to generate transgenic flies were constructed in the pUAST-attB vector (from Dr. Hugo Bellen’s laboratory, Baylor College of Medicine) and pCFD5_w vector (from Dr. Simon Bullock’s laboratory, MRC- Laboratory of Molecular Biology). Plasmid construction was performed by inserting the cDNA of each gene into the pUAST-attB vector using EcoR I (TaKaRa, 1040 A) and Xho I (TaKaRa, 1094 A) restriction enzymes. A 3 × HA or V5 tag was fused in frame to the C-terminus of each coding sequence. The pUAST-attB-Mppa A407T-V5, pUAST-attB-Mppa V283M-V5, pUAST-attB-UQCR-C1 I403T-3HA, pUAST-attB-UQCR-C1 E87Q-3HA, pUAST-attB-UQCR-C1 E174/D178A-3HA plasmids were created using site-directed mutagenesis. Protein expression from the newly constructed pUAST-attB plasmids was detected in S2 cells. For single-guide RNA (sgRNA) design targeting each gene, candidate sequences were selected within upstream coding sequences using the Clustered regularly interspaced short palindromic repeats (CRISPR) online design tool (https://www.flyrnai.org/crispr3/web), with priority given to those exhibiting low off-target prediction scores to minimize potential unintended genomic edits. All pCFD5_w sgRNA plasmids were generated by cloning sgRNA sequences into the vector using BbsI (Thermo Fisher Scientific, ER1011) restriction enzyme digestion and T4 DNA ligase (TaKaRa, 2011 A)-mediated ligation. Detailed sgRNA sequences used in this study is listed in Supplemental Table S3. Antibodies A mouse monoclonal antibody against HA (BioLegend,16B12) was used for immunofluorescence at 1:400 dilution. A rabbit monoclonal antibody against HA (Cell Signaling, 3724) was utilized with different dilutions: a 1:1 000 dilution was applied for western blotting, while a 1:400 dilution was applied for immunofluorescence. A mouse monoclonal antibody against V5 (Invitrogen, R960–25) was utilized with a 1:1 000 dilution for western blot, and a 1:400 dilution for immunofluorescence. A rabbit monoclonal antibody against V5 (Invitrogen, MA5-32053) was applied for immunofluorescence at a 1:400 dilution. A rabbit polyclonal antibody against GFP (MBL, 598) was used for western blotting at a 1:1 000 dilution. A mouse monoclonal antibody against α-Tubulin (Beyotime, AT819) was utilized for western blotting at 1:1 000 dilution. Generation of whole-body or tissue-specific knockout flies Whole-body knockout (KO) flies were generated following a germline-specific editing strategy described previously [14]. Briefly, germline-specific nanos-Cas9-flies were crossed with sgRNA transgenic stocks. F1 male progeny, in which both the nanos-Cas9 and specific sgRNA cassette inherited in their sperms, were crossed with balancers strains to transmit edited germ cells. Individual F2 males, potentially harboring mutations or frameshift alleles at the target gene, were backcrossed with balancer females to establish isogenic mutant stocks. After progeny were produced, genomic DNA was extracted from individual F2 males to identify potential mutants. Tissue-specific KO flies were generated by crossing tissue-specific GAL4 drivers with UAS-Cas9 and the corresponding sgRNA lines. Transmission electron microscopy Fat bodies from 96-h-old larvae or the indirect flight muscles (IFMs) from adult flies were dissected and fixed in 2.5% glutaraldehyde (Electron Microscopy Sciences, 16020) at 4 °C. After being fixed for 1 week, fat bodies were washed three times with 0.1 M phosphate buffered saline (PBS) (137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, 2 mM KH2PO4, pH 7.4) for 15 min each, and then post-fixed with 1% (v: v) osmium tetroxide (OsO₄; Electron Microscopy Sciences, 19152) for 1.5 h at room temperature. Next, tissues were dehydrated with 50% (v: v) ethanol for 15 min and 70% ethanol containing uranyl acetate (Electron Microscopy Sciences, Hatfield, PA, USA, 22400) for 1.5 h. Thereafter, tissues were dehydrated using a graded ethanol series (80%, 90%, 95%, 100%; 15 min each) and 100% acetone (20 min). Fat bodies were infiltrated with graded Eponate 12 resin (Embed 812 [14900], DDSA [13710], NMA [19000], and DMP-30 [13600]; all Electron Microscopy Sciences): acetone mixtures (50%, 75%, v: v) for 1 h and 3 h, respectively. Finally, tissues were embedded in freshly prepared pure Eponate 12 resin for 12 h at room temperature, then polymerized at 65 °C for 72 h. After polymerization, 80-nm semithin sections (for IFMs) or 50-nm ultrathin sections (for fat bodies, IFMs, and eyes) were cut using a Leica EM UC7 ultramicrotome (Leica). Semithin sections for IFMs were stained with toluidine blue solution on a hot plate for 1 min, and then rinsed with distilled water to remove excess stain. IFMs structures were examined under a light microscope to identify areas suitable for subsequent ultrathin sectioning. Ultrathin sections were stained with 4% uranyl acetate (Electron Microscopy Sciences, 22400) and 2.5% lead citrate (Electron Microscopy Sciences, 17800) and observed under Hitachi Model HT7700 transmission electron microscope (Hitachi, ). The protocol for retinal TEM analysis in fly eyes was similar to that used for fat bodies, with minor modifications as described previously [18]. Immunofluorescence and confocal microscopy Fat bodies dissected from 96-h-old larvae were fixed in 4% paraformaldehyde (Sigma-Aldrich, 158127) for 30 min at room temperature, and washed three times with 1× PBS containing 0.1% Triton X-100 (Sangon Biotech, T0694; 0.1% PBST). Tissues were incubated with primary antibody overnight at 4 °C, followed by three 10-min washes using 0.1% PBST. Subsequently, samples were incubated with secondary antibodies in the dark for 1 h at room temperature or overnight at 4 °C. After washing three times with 0.1% PBST, samples were mounted in 80% glycerol (Sangon Biotech, A100854) containing 5 ng/µL DAPI (Invitrogen, D-1306) and observed using confocal laser microscopy (LSM710, LSM880 and LSM900; Carl Zeiss). For IFMs dissected from adult fly thoraces, the protocol was identical except that 0.4% PBST (1× PBS with 0.4% Triton X-100) was used as the washing buffer. Co-immunoprecipitation S2 cells were collected 65 h post-transfection and washed twice with ice-cold 1× PBS. S2 cells were then lysed in S2 lysis buffer (150 mM NaCl, 50 mM Tris-HCl, pH 8.0, 10 mM NaF, 1 mM Na3VO4, 1% NP-40 [Sangon Biotech, A600385], 10% glycerol, pH 8.0) supplemented with 100× protease inhibitor cocktail (Beyotime, P1005) at 4 °C for 30 min. Cell lysates were centrifuged at 1 4000 × g for 20 min at 4 °C and 10% of the supernatant was separated as input. The remaining supernatant was incubated with V5 or HA antibody for 2 h at 4 °C on a rotator. Protein A Sepharose 4 Fast Flow beads (Cytiva, 17127901) were then added to each sample for incubation for another 2 h at 4 °C with gentle rotation. Beads were collected by centrifugation and washed 3 times with ice-cold S2 lysis buffer. Bound proteins were eluted by boiling in 2× sodium dodecyl sulfate (SDS) loading buffer (100 mM Tris-HCl [pH 6.8], 4% SDS, 20% glycerol, 0.2% bromophenol blue, 200 mM dithiothreitol (DTT)) at 98 °C for 10 min. Protein samples were either used immediately for Western blot analysis or stored at -80 °C for further experiments. Western blotting Fat bodies dissected from 96-h-old larvae in ice-cold 1× PBS were placed into pre-chilled 1.5 mL tubes. Following the addition of 2 × SDS loading buffer, fat bodies were gently minced into small pieces on ice. The samples were boiled at 95 °C for 10 min to denature proteins. Proteins were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) at 90 V for 2 h and transferred to a polyvinylidene fluoride (PVDF) membrane (Millipore, IPVH00010). After being blocked with 5% non-fat milk (Sangon Biotech, A600669) in Tris-buffered saline with Tween-20 (TBST, 20 mM Tris-HCl, 150 mM NaCl, 0.1% Tween 20 [Sangon Biotech, A100777]) for 1 h at room temperature, membranes with protein blots were incubated with primary antibodies in 5% bovine serum albumin (BSA; Sangon Biotech, A600332) / TBST overnight at 4 °C on a rotator. The membranes were washed three times with TBST, then incubated with HRP-labeled secondary antibodies (Jackson Immuno Research Laboratories Inc. 111-035-003 [Rabbit] and 115-035-003 [Mouse]) at a 1:5 000 dilution in 5% non-fat milk / TBST for 1 h at room temperature. Following washes, blots were exposed using enhanced chemiluminescence (ECL) reagents (Cyanagen, XLS3-0020). Fertility test Reproductive capacity of mutant lines was assessed following protocols as described before [12]. Briefly, 5-day-old males or virgin females (n = 25 from each mutant line) were individually mated with three age-matched w¹¹¹⁸ virgin females or males. The w¹¹¹⁸ strain served as a wild-type control. Parental flies were maintained in standard cornmeal food vial for 7 days, allowing sufficient time for mating and egg laying. After 7 days, all parental flies were removed from the vials, and the vials were incubated for an additional 10 days to allow offspring eclosion. Reproductive capacity was quantified by the proportion of mutant individuals that produced viable offspring. For the mutant lines exhibiting male-specific infertility, testes were dissected and examined by a phase-contrast microscopy (ECLIPSE Ti2; Nikon, Tokyo, Japan) to observe the stages of spermatogenesis. Statistical analysis ImageJ software was utilized to quantify fluorescence intensity, the number and size of fluorescence proteins, and mitochondrial length from confocal images. Statistical analyses and graphing were performed using GraphPad Prism 8 software, based on at least three independent biological replicates. Differences between two experimental conditions were assessed by two-tailed unpaired Student’s t test, while multiple group comparisons were analyzed by one-way Analysis of variance (ANOVA) with Bonferroni post-hoc test. In the figures, all the results, except Fig. 2C were presented as mean values ± standard error of the mean (SEM). Statistical significance is indicated as follows: ns indicates no significance; * indicates p < 0.05; ** indicates p < 0.01; *** indicates p < 0.001. Fig. 2The loss of many mitochondrial proteases led to neurodegeneration. A. TEM analysis of retinal sections from indicated mitochondrial protease KO flies at day 2 and day 30. For YME1L KO flies, the oldest individuals analyzed were 15 days old due to their shortened lifespan. B. TEM analysis of retinal sections at day 2 and day 30 from eye-specific KOs generated by sgRNA and GMR-GAL4–driven Cas9 expression. C and C’. Quantification of the number of rhabdomeres in the retina with indicated genotypes. All KOs’ retina appeared largely normal at day 2. At day 30, loss of CG11771, CG5131, Spg7, ClpP, Lon, or Mppa led to reduced photoreceptor numbers. n = 10 ommatidia from 3 flies of each genotype. D and D’. The statistics of mitochondrial length in the retina. Loss of Immp1, CG11771, or Mppa led to elongated mitochondria at day 2. Loss of Afg3l2 or ClpP resulted mitochondria elongation in aged flies, while YME1L loss caused shorter mitochondria in 15-day-old flies. n = 50 mitochondria from 3 flies of each genotype. E and E’. Quantification of the ratio of mitochondria with less cristae (red arrows in A and B). Mitochondria in the retina of young CG5188 or CG11771 mutants, as well as aged flies loss of ClpP, Lon, or CG11771, exhibited less cristae. n = 5 TEM images from 3 flies of each genotype. F and F’. Quantification of the ratio of mitochondria with aggregates (green arrows in A and B). Loss of CG5131, Spg7, ClpP, Lon, or Mppa led to mitochondrial aggregate formation in both young and old flies, whereas mitochondrial aggregates were detected exclusively in older eyes of CG11771 mutants. n = 5 TEM images from 3 flies of each genotype. All data were presented as mean ± SEM. ns: not significant; *p < 0.05; **p < 0.01; ***p < 0.001 The loss of many mitochondrial proteases led to neurodegeneration. A. TEM analysis of retinal sections from indicated mitochondrial protease KO flies at day 2 and day 30. For YME1L KO flies, the oldest individuals analyzed were 15 days old due to their shortened lifespan. B. TEM analysis of retinal sections at day 2 and day 30 from eye-specific KOs generated by sgRNA and GMR-GAL4–driven Cas9 expression. C and C’. Quantification of the number of rhabdomeres in the retina with indicated genotypes. All KOs’ retina appeared largely normal at day 2. At day 30, loss of CG11771, CG5131, Spg7, ClpP, Lon, or Mppa led to reduced photoreceptor numbers. n = 10 ommatidia from 3 flies of each genotype. D and D’. The statistics of mitochondrial length in the retina. Loss of Immp1, CG11771, or Mppa led to elongated mitochondria at day 2. Loss of Afg3l2 or ClpP resulted mitochondria elongation in aged flies, while YME1L loss caused shorter mitochondria in 15-day-old flies. n = 50 mitochondria from 3 flies of each genotype. E and E’. Quantification of the ratio of mitochondria with less cristae (red arrows in A and B). Mitochondria in the retina of young CG5188 or CG11771 mutants, as well as aged flies loss of ClpP, Lon, or CG11771, exhibited less cristae. n = 5 TEM images from 3 flies of each genotype. F and F’. Quantification of the ratio of mitochondria with aggregates (green arrows in A and B). Loss of CG5131, Spg7, ClpP, Lon, or Mppa led to mitochondrial aggregate formation in both young and old flies, whereas mitochondrial aggregates were detected exclusively in older eyes of CG11771 mutants. n = 5 TEM images from 3 flies of each genotype. All data were presented as mean ± SEM. ns: not significant; *p < 0.05; **p < 0.01; ***p < 0.001 Key resources The key resources and their identifier number could be found in Table S4. Results Mitochondrial proteases in fruit flies are important for survival or male fertility To systematically study the functions of mitochondrial proteases in flies, we selected 15 genes whose human orthologs encode mitochondrial proteases (Table 1; Fig. 1B). Among them, eight (CG5188, Mipep/CG7791, UQCR-C1/CG3731, CG3107, CG5131, Afg3l2/CG6512, Spg7/CG2658, YME1L/CG3499) encode metalloproteases, and five (Immp1/CG9240, Immp2/CG11110, ClpP/CG5045, HtrA2/CG8464, Lon/CG8798) encode serine proteases. Mppa/CG8728 encodes a pseudo-metalloprotease lacking catalytic activity. In mammals, its ortholog PMPCA encodes a regulatory subunit that functions together with PMPCB, the catalytic component of the complex, to process newly imported mitochondrial proteins. CG11771 encodes a protein characterized as an M3 non-peptidase homolog, sharing amino acid sequence similarity with M3 metallopeptidases but lacking one or more essential catalytic residues. It is unclear whether CG11771 functions as a protease. Since it shows some homology to the mitochondrial proteases MEP and MIPEP, we also included this gene in our analysis. To study the functions of these genes, we generated KO flies using the CRISPR/ CRISPR-associated protein 9 (Cas9) system (Fig. 1A). Flies lacking CG5188, CG3107, YME1L, Immp1, Immp2, HtrA2, or CG11771 were viable, indicating that these genes are not essential for fly survival. However, loss of CG3107, YME1L, Immp2, HtrA2, or CG11771 led to male sterility, suggesting that these genes may play important roles in spermatogenesis (Fig. 1B and C). Knockouts of the remaining genes (Mipep, UQCR-C1, CG5131, Afg3l2, Spg7, ClpP, Lon, and Mppa) resulted in lethality (Fig. 1B). For all genes except Spg7, we obtained multiple KO alleles with lesions introducing premature stop codons. For Spg7, all mutant alleles carried deletions of three or multiples of three nucleotides. The largest deletion removed 21 nucleotides and caused lethality, though it remains unclear whether this allele represents a complete loss of function. For the male-sterile KO flies, we dissected adult testes to examine overall morphology and germ cell development. As male sterility in YME1L [30] and HtrA2 [41] mutants has been reported previously, we did not pursue further analysis of these lines. In CG3107, Immp2, and CG11771 KOs, the testes appeared largely normal in size, but the seminal vesicles were noticeably smaller, indicating reduced or absent sperm storage and suggesting defects in sperm production (Fig. 1D). In Immp2 KO flies, the seminal vesicles were smaller than in controls, while the adjacent testis region appeared enlarged. When the testes were gently squashed and examined under a microscope, elongated spermatids were observed in CG3107, Immp2, and CG11771 mutants, but no motile mature sperm were detected as in controls (Fig. 1E). These findings suggest that these genes are required for sperm maturation in flies. The loss of many mitochondrial proteases lead to neurodegeneration Since the nervous system is highly sensitive to mitochondrial defects, we examined whether loss of mitochondrial proteases would cause neurodegenerative phenotypes in fly eyes. For mitochondrial protease KO flies that survived to adulthood, we analyzed the retina at day 2 and day 30 by transmission electron microscopy (TEM) (Fig. 2A). YME1L KO flies survived only up to 15 days; thus, the oldest samples examined for this mutant were 15 days old. In wild-type flies, each ommatidium contains eight photoreceptor cells surrounded by pigment (glial) cells, and seven tightly packed rhabdomeres are visible in both young and old flies. All viable mutants displayed seven rhabdomeres and appeared largely normal at day 2 (Fig. 2A and C). At day 30, most flies showed no signs of degeneration except for CG11771 mutants, which frequently retained only six photoreceptors (Fig. 2A and C). For KO lines that were lethal before adulthood, we performed tissue-specific KOs by expressing sgRNA ubiquitously and driving UAS-Cas9 expression in the developing eye with GMR-GAL4 (Fig. 2B). Loss of CG5131, Spg7, ClpP, Lon, or Mppa resulted in reduced photoreceptor numbers in old but not young flies, indicating age-dependent retinal degeneration (Fig. 2B and C’). In contrast, loss of Afg3l2 or Mipep did not cause detectable eye degeneration (Fig. 2B and C’). We attempted to generate whole-eye clones of the UQCR-C1 KO alleles using the Flippase (FLP)/ Flippase recognition target (FRT) system, but the resulting flies had very small eyes and showed no obvious UQCR-C1 KO clones, suggesting a cell-lethal phenotype. In addition to photoreceptor numbers, we examined mitochondrial morphology within the photoreceptors (Fig. 2A, B, D and D’). Loss of Immp1, CG11771, or Mppa led to elongated mitochondria in young flies, but the average mitochondrial length appeared normal in older flies of these genotypes (Fig. 2A, B, D and D’). Loss of Afg3l2 or ClpP resulted in elongated mitochondria in aged flies, whereas YME1L loss caused shorter mitochondria in 15-day-old flies (Fig. 2A, B, D and D’). Occasionally, mitochondria with empty spaces were observed in young CG5188 or CG11771 mutants, but most mitochondria appeared normal in older CG5188 KO flies (red arrows in Fig. 2A and E). In contrast, some mitochondria lost cristae and became swollen in aged flies lacking ClpP, Lon, or CG11771 (red arrows in Fig. 2A and B, and Fig. 2E, and 2E’). Interestingly, loss of CG5131, Spg7, ClpP, Lon, or Mppa led to the formation of aggregates within mitochondria in both young and old flies (Fig. 2B and F’). These aggregates varied in size and abundance—some were small and sparse, while others were large and numerous (green arrows in Fig. B). In most cases, mitochondrial aggregates accumulated with age. In CG11771 KO flies, mitochondrial aggregates were detected only in older eyes (Fig. 2A and F). The presence of aggregates, rather than changes in mitochondrial morphology, correlated strongly with retinal degeneration, suggesting that aggregate accumulation within mitochondria may be particularly disruptive to neuronal integrity. Loss of mitochondrial proteases causes mitochondrial morphology defects in flight muscles and fat bodies Indirect flight muscles, like neurons, are highly energy-demanding tissues. In wild-type flies, mitochondria with densely packed cristae align neatly along muscle fibers to sustain energy production. To assess the roles of mitochondrial proteases in muscle, we examined mitochondrial ultrastructure in adult indirect flight muscles deficient for these proteases. For viable KOs, muscle morphology was analyzed directly (Fig. 3A). For lethal mutants, muscle-specific KOs were generated by ubiquitously expressing sgRNAs together with Mef2-GAL4–driven UAS-Cas9 (Fig. 3B). Notably, muscle-specific KO of UQCR-C1 or Mppa also caused lethality, indicating essential functions during muscle development (Fig. S1). Fig. 3Loss of mitochondrial proteases causes mitochondrial morphology defects in flight muscles and fat bodies. A TEM analysis of indirect flight muscles at day 30 from viable mitochondrial protease KO flies. B TEM analysis of indirect flight muscles from 30-day-old adults with muscle-specific KO generated by sgRNA and Mef2-GAL4–driven Cas9 expression. C The proportions of mitochondria with normal morphology (gray), less cristae (blue), or swirl-like structure (orange) in the indirect flight muscles of flies with indicated genotypes. n = 10 images from 3 flies per genotype were analyzed. Data were presented as mean values. ns: not significant; ***p < 0.001. D Quantification of the ratio of mitochondria with aggregates (green arrows in A and B). Protein aggregates were detected within mitochondria loss of CG3107 or Lon. n = 5 TEM images from 3 flies per genotype. E The confocal and TEM images of larval fat bodies with fat body-specific protease loss generated by sgRNA and Cg-GAL4–driven Cas9 expression. MitoGFP serves as a mitochondrial marker. F Quantification of mitochondrial length from confocal images of larval fat bodies with the indicated genotypes. Loss of UQCR-C1, CG5131, Afg3l2, Spg7, YME1L, Immp1, or Lon resulted in shorter mitochondria, whereas loss of HtrA2 caused mitochondrial elongation. 10 larvae were dissected and n = 5 images were analyzed for each genotype. G Quantification of the ratio of mitochondria with less cristae in TEM images (red arrows in E). Mitochondria deficient in Mipep, UQCR-C1, Afg3l2, or Spg7 showed reduced cristae. n = 5 TEM images from 3 larval fat bodies of each genotype. H Quantification of the ratio of mitochondria with aggregates in TEM images (green arrows in E). n = 5 TEM images from 3 larval fat bodies of each genotype. All data, except 3 C, were presented as mean ± SEM. ns: not significant; *p < 0.05; **p < 0.01; ***p < 0.001 Loss of mitochondrial proteases causes mitochondrial morphology defects in flight muscles and fat bodies. A TEM analysis of indirect flight muscles at day 30 from viable mitochondrial protease KO flies. B TEM analysis of indirect flight muscles from 30-day-old adults with muscle-specific KO generated by sgRNA and Mef2-GAL4–driven Cas9 expression. C The proportions of mitochondria with normal morphology (gray), less cristae (blue), or swirl-like structure (orange) in the indirect flight muscles of flies with indicated genotypes. n = 10 images from 3 flies per genotype were analyzed. Data were presented as mean values. ns: not significant; ***p < 0.001. D Quantification of the ratio of mitochondria with aggregates (green arrows in A and B). Protein aggregates were detected within mitochondria loss of CG3107 or Lon. n = 5 TEM images from 3 flies per genotype. E The confocal and TEM images of larval fat bodies with fat body-specific protease loss generated by sgRNA and Cg-GAL4–driven Cas9 expression. MitoGFP serves as a mitochondrial marker. F Quantification of mitochondrial length from confocal images of larval fat bodies with the indicated genotypes. Loss of UQCR-C1, CG5131, Afg3l2, Spg7, YME1L, Immp1, or Lon resulted in shorter mitochondria, whereas loss of HtrA2 caused mitochondrial elongation. 10 larvae were dissected and n = 5 images were analyzed for each genotype. G Quantification of the ratio of mitochondria with less cristae in TEM images (red arrows in E). Mitochondria deficient in Mipep, UQCR-C1, Afg3l2, or Spg7 showed reduced cristae. n = 5 TEM images from 3 larval fat bodies of each genotype. H Quantification of the ratio of mitochondria with aggregates in TEM images (green arrows in E). n = 5 TEM images from 3 larval fat bodies of each genotype. All data, except 3 C, were presented as mean ± SEM. ns: not significant; *p < 0.05; **p < 0.01; ***p < 0.001 Loss of most mitochondrial proteases, except CG5131 and HtrA2, led to mitochondrial abnormalities in aged (30-day-old) muscles (Fig. 3A and C). Based on morphology, mitochondria were categorized as normal, less cristae (sparse cristae or empty spaces), or swirl (cristae arranged in swirl-like patterns). Deficiency of YME1L (15-day-old) or Lon produced both “less cristae” and “swirl” types (Fig. 3A and C), whereas loss of CG5188 or CG11771 resulted mainly in swirl-like mitochondria (Fig. 3A and C). In contrast, deficiency of Mipep, CG3107, Afg3l2, Spg7, Immp1, Immp2, or ClpP primarily caused mitochondria with reduced cristae (Fig. 3A and C). Protein aggregates were also observed within mitochondria lacking CG3107 or Lon (Fig. 3A and B green arrows, 3D). The larval fat body, functionally analogous to the mammalian liver and adipose tissue, contains large cells enriched with mitochondria and other organelles, providing an excellent model for studying mitochondrial morphology. Mitochondria were labeled with MitoGFP, and protease loss was induced by expressing sgRNAs together with Cg-GAL4–driven UAS-Cas9. Confocal and transmission electron microscopy revealed that most fat body–specific KOs showed normal mitochondrial morphology (Fig. 3E). However, loss of UQCR-C1, CG5131, Afg3l2, Spg7, YME1L, Immp1, or Lon caused reduction of mitochondrial length (Fig. 3E and F). Mitochondria in HtrA2-deficient fat bodies appeared elongated and interconnected (Fig. 3E and F). TEM revealed that mitochondria lacking Mipep, UQCR-C1, Afg3l2, Spg7, or ClpP exhibited reduced cristae (red arrows in Fig. 3E, and Fig. 3G). Furthermore, loss of Mipep, UQCR-C1, CG3107, Afg3l2, Immp1, or Lon led to the accumulation of protein aggregates (green arrows in Fig. 3E and H), indicating that these proteases are crucial for maintaining mitochondrial protein homeostasis in larval fat bodies. Loss of certain proteases increases mitophagy Mitochondrial proteases and mitophagy both contribute to mitochondrial quality control. To determine whether loss of mitochondrial proteases influences mitophagy, we examined this process in the larval fat body, a tissue with very low basal mitophagy and thus suitable for detecting upregulation. Mitophagy was monitored using mito-QC, a reporter consisting of tandemly fused mCherry, GFP, and a fragment of the mitochondrial outer membrane protein Fis1 (aa 101–152) [16]. Mito-QC-labeled mitochondria display both green and red fluorescence, but because GFP is quenched while mCherry remains stable in the acidic lysosomal environment, mitochondria undergoing mitophagy appear as red-only puncta. The number of red-only foci therefore reflects the level of mitophagy. We expressed sgRNAs targeting individual mitochondrial proteases ubiquitously and drove UAS-Cas9 expression in the fat body using Cg-GAL4 (Fig. 4A). Reduction of most proteases did not alter mitophagy, but reduction of UQCR-C1 or Afg3l2 caused a marked increase in red-only puncta (arrows in Fig. 4A and B). Red puncta were also detected upon loss of CG5188 or YME1L, though the changes were not statistically significant (Fig. 4B). These findings indicate that reduced activity of certain mitochondrial proteases, particularly UQCR-C1 and Afg3l2, can enhance mitophagy in the fat body. Fig. 4Loss of certain proteases increases mitophagy. A Representative images of mito-QC patterns in the early third instar larval fat bodies from tissue-specific KOs generated by sgRNA and Cg-GAL4–driven Cas9 expression. Red-only signals indicated lysosomal engulfed mitochondria (white arrows). B Quantification of the number of red-only foci per cell in larval fat bodies with indicated genotypes. Loss of UQCR-C1 or Afg3l2 resulted in a marked increase in red-only puncta. 10 larvae were dissected and n = 5 images were analyzed for each genotype. All data were presented as mean ± SEM. ns: not significant; ***p < 0.001 Loss of certain proteases increases mitophagy. A Representative images of mito-QC patterns in the early third instar larval fat bodies from tissue-specific KOs generated by sgRNA and Cg-GAL4–driven Cas9 expression. Red-only signals indicated lysosomal engulfed mitochondria (white arrows). B Quantification of the number of red-only foci per cell in larval fat bodies with indicated genotypes. Loss of UQCR-C1 or Afg3l2 resulted in a marked increase in red-only puncta. 10 larvae were dissected and n = 5 images were analyzed for each genotype. All data were presented as mean ± SEM. ns: not significant; ***p < 0.001 Overexpression of mitochondrial proteases disrupts mitochondrial and tissue homeostasis Mitochondria can serve as a protein degradation center under stress conditions [32]. For instance, the disease-related protein TDP-43 can be imported into mitochondria, where the protease LonP1 degrades it to maintain mitochondrial homeostasis [38]. These findings raise the possibility that elevating mitochondrial protease expression might benefit cytosolic proteostasis and mitochondrial function. To test this, we overexpressed various mitochondrial proteases in Drosophila tissues and examined their expression and effects on mitochondrial morphology and tissue integrity. Using Cg-GAL4 to drive expression in the larval fat body, we found that protease expression levels varied despite being under the same driver (Fig. 5A). CG5131, Immp1, and YME1L were barely detectable, while Mipep, ClpP, and CG11771 showed strong expression, likely reflecting differences in protein stability. Overexpression of HtrA2 caused larval lethality (Fig. S2A), indicating strong cytotoxicity. Most proteases co-localized with MitoGFP and were evenly distributed in mitochondria, but Mppa formed long fiber-like structures within certain mitochondrial regions, and Immp2 preferentially accumulated in mitochondria near the nucleus. Fig. 5Overexpression of mitochondrial proteases disrupts mitochondrial and tissue homeostasis. A HA or V5 tagged mitochondrial proteases (red) were co-expressed with MitoGFP (green) in the fat bodies using Cg-GAL4 driver. Early third instar larval fat bodies were dissected and stained with anti-HA or V5 antibodies. DAPI (blue) labels nuclei. B V5 or HA tagged mitochondrial proteases (red) were co-expressed with MitoGFP (green) in indirect flight muscles using Mef2-GAL4 driver. The indirect flight muscles in 2-day-old flies were dissected and stained with anti-HA or V5 antibodies. C TEM analysis of adult indirect flight muscles overexpressing Mipep, Afg3l2, Spg7, Immp2, or Lon. Severe mitochondrial defects and muscle damage were observed in young (2 days) flies, and these phenotypes progressively worsened with age (30 days) Overexpression of mitochondrial proteases disrupts mitochondrial and tissue homeostasis. A HA or V5 tagged mitochondrial proteases (red) were co-expressed with MitoGFP (green) in the fat bodies using Cg-GAL4 driver. Early third instar larval fat bodies were dissected and stained with anti-HA or V5 antibodies. DAPI (blue) labels nuclei. B V5 or HA tagged mitochondrial proteases (red) were co-expressed with MitoGFP (green) in indirect flight muscles using Mef2-GAL4 driver. The indirect flight muscles in 2-day-old flies were dissected and stained with anti-HA or V5 antibodies. C TEM analysis of adult indirect flight muscles overexpressing Mipep, Afg3l2, Spg7, Immp2, or Lon. Severe mitochondrial defects and muscle damage were observed in young (2 days) flies, and these phenotypes progressively worsened with age (30 days) When proteases were overexpressed in muscle, Mipep, Afg3l2, Spg7, Immp2, and Lon caused severe muscle degeneration, resulting in abnormal wing postures (erect or droopy) (Fig. S2B) and extensive loss of indirect flight muscles, which worsened with age (Fig. S2C). The remaining muscles exhibited severely damaged mitochondria (Fig. 5C). Muscle-specific overexpression of HtrA2 also caused larval lethality (Fig. S2D). Similar to fat body results, YME1L and Immp1 were expressed at very low levels, whereas CG5131 localized within mitochondria but showed uneven distribution. In addition, Lon expression was also barely detectable in the residual muscle fibers, suggesting that cells expressing high levels of this protease were eliminated. Unlike their uniform distribution in fat body mitochondria, CG5188, Mipep, Afg3l2, and CG11771 displayed uneven intramitochondrial localization. Mppa overexpression in muscle again induced thin, fiber-like mitochondrial structures enriched with Mppa (Fig. 5B). Together, these findings demonstrate that excessive expression of many mitochondrial proteases compromises mitochondrial integrity and tissue homeostasis, with tissue-specific differences in sensitivity and protease localization patterns. Certain mitochondrial proteases facilitate Huntington polyQ protein degradation Our previous work showed that blocking mitochondrial protein import promotes aggregation of the Huntington disease (HTT) protein [19]. Given that disease-related proteins such as TDP-43 can be imported into mitochondria and degraded by the protease LonP1 [38], we examined whether mitochondrial proteases modulate HTT aggregation. HTT.72Q.Emerald was expressed in the fat body, and individual mitochondrial proteases were knocked out (Fig. 6A and B). In control tissues expressing Cas9 alone, HTT.72Q.Emerald was diffusely distributed without aggregates. In contrast, loss of UQCR-C1, CG11771, or Mppa induced HTT72Q aggregation (Fig. 6A–6C’). Loss of UQCR-C1 also led to a dramatic increase in HTT72Q levels (Fig. 6B and D’). Depletion of Afg3l2, Spg7, ClpP, or Lon did not increase aggregate formation but elevated overall HTT72Q levels, with Afg3l2 KO causing the most pronounced increase (Fig. 6A, B, D and D’). These data suggest that UQCR-C1, CG11771, Mppa, Afg3l2, Spg7, ClpP, and Lon contribute to the clearance of HTT72Q, either directly or indirectly. Fig. 6Certain mitochondrial proteases facilitate Huntington polyQ protein degradation. A and B. HTT.72Q.Emerald proteins (green) were expressed in larval fat bodies with tissue-specific KOs generated by sgRNA and Cg-GAL4–driven Cas9 expression. Two pUAS-Cas9 lines (inserted on the 2nd or 3rd chromosome) were used in the experiments because the sgRNAs were located on different chromosomes. The fluorescent intensity and number of HTT72Q puncta varied slightly between the two pUAS-Cas9–expressing groups. Each pUAS-Cas9 line co-expressed with HTT.72Q.Emerald was used as the control for tissue-specific KO lines generated with the same pUAS-Cas9 background (A: pUAS-Cas9 2nd, B: pUAS-Cas9 3rd). C and C’: Quantification of the number of HTT72Q puncta in the tissues with the indicated genotypes. 10 larvae were dissected and n = 5 images were analyzed for each genotype. D and D’: Quantification of relative fluorescence intensity of HTT72Q in the fat bodies with indicated genotypes. 10 larvae were dissected and n = 5 images were analyzed for each genotype. E. HTT.96Q.Cerulean protein (green) was co-expressed in larval fat bodies with the indicated genes. LacZ served as a control. F. Quantification of the number of HTT96Q puncta in the fat bodies with the indicated genotypes. 25 larvae were dissected and n = 5 images were analyzed for each genotype. G. Quantification of the size of HTT96Q puncta in the fat bodies with the indicated genotypes. 25 larvae were dissected and n = 5 images were analyzed for each genotype. H. GMR-GAL4 drove the co-expression of HTT96Q and the indicated proteins in the eye. TEM analysis of adult fly retinas was performed at day 2, day 15, and day 30. I. Quantification of rhabdomere number per ommatidium in retinas of the indicated genotypes. MigamuRFP overexpression served as a control. n = 10 ommatidia from 3 flies of each genotype. All data were presented as mean ± SEM. ns: not significant; *p < 0.05; **p < 0.01; ***p < 0.001 Certain mitochondrial proteases facilitate Huntington polyQ protein degradation. A and B. HTT.72Q.Emerald proteins (green) were expressed in larval fat bodies with tissue-specific KOs generated by sgRNA and Cg-GAL4–driven Cas9 expression. Two pUAS-Cas9 lines (inserted on the 2nd or 3rd chromosome) were used in the experiments because the sgRNAs were located on different chromosomes. The fluorescent intensity and number of HTT72Q puncta varied slightly between the two pUAS-Cas9–expressing groups. Each pUAS-Cas9 line co-expressed with HTT.72Q.Emerald was used as the control for tissue-specific KO lines generated with the same pUAS-Cas9 background (A: pUAS-Cas9 2nd, B: pUAS-Cas9 3rd). C and C’: Quantification of the number of HTT72Q puncta in the tissues with the indicated genotypes. 10 larvae were dissected and n = 5 images were analyzed for each genotype. D and D’: Quantification of relative fluorescence intensity of HTT72Q in the fat bodies with indicated genotypes. 10 larvae were dissected and n = 5 images were analyzed for each genotype. E. HTT.96Q.Cerulean protein (green) was co-expressed in larval fat bodies with the indicated genes. LacZ served as a control. F. Quantification of the number of HTT96Q puncta in the fat bodies with the indicated genotypes. 25 larvae were dissected and n = 5 images were analyzed for each genotype. G. Quantification of the size of HTT96Q puncta in the fat bodies with the indicated genotypes. 25 larvae were dissected and n = 5 images were analyzed for each genotype. H. GMR-GAL4 drove the co-expression of HTT96Q and the indicated proteins in the eye. TEM analysis of adult fly retinas was performed at day 2, day 15, and day 30. I. Quantification of rhabdomere number per ommatidium in retinas of the indicated genotypes. MigamuRFP overexpression served as a control. n = 10 ommatidia from 3 flies of each genotype. All data were presented as mean ± SEM. ns: not significant; *p < 0.05; **p < 0.01; ***p < 0.001 Because HTT72Q overexpression produces minimal aggregate formation in wild-type tissues under our experimental conditions, it is not suitable for assessing suppression of polyQ aggregation. In contrast, HTT96Q robustly forms aggregates when overexpressed in wild-type fat bodies, making it a more appropriate model for testing whether protease overexpression can reduce polyQ aggregation. Therefore, to examine the effects of protease overexpression on HTT polyQ aggregation, UQCR-C1, CG11771, Mppa, or Afg3l2 were co-expressed with HTT.96Q.Cerulean in the fat body (Fig. 6E). Whereas control tissues co-expressing LacZ showed robust HTT96Q aggregates, overexpression of CG11771, Mppa, or Afg3l2 markedly reduced aggregate formation, while UQCR-C1 had no effect (Fig. 6E and G). Because HTT96Q expression in the eye causes neurodegeneration, we further tested whether mitochondrial protease overexpression could mitigate toxicity. HTT.96Q.Cerulean was co-expressed with UQCR-C1, CG11771, Mppa, or Afg3l2, with mutant Miga (MigamuRFP, a protein whose overexpression has no effect on eye development or maintenance) as a control (Fig. 6H). In control eyes, HTT96Q caused early loss of rhabdomeres that worsened with age. Co-expression of CG11771 or Mppa preserved retinal structure at day 2, and Mppa retained partial protection at day 15, though the effect was lost by day 30 (Fig. 6H and I). These results indicate that CG11771 and Mppa provide transient protection against HTT96Q-induced neurotoxicity, with Mppa exhibiting a more sustained protective effect. Mppa forms a complex with UQCR-C1 to regulate mitochondrial protein homeostasis, and disease-related mutants are less active The mammalian ortholog of Mppa, PMPCA, encodes a pseudo-mitoprotease that forms a complex with PMPCB, the catalytic subunit responsible for processing mitochondrial targeting peptides. In Drosophila, UQCR-C1 is the ortholog of PMPCB. Interestingly, loss of either Mppa or UQCR-C1 enhanced HTT72Q aggregate formation (Fig. 6A–6C’), whereas overexpression of Mppa, but not UQCR-C1, reduced HTT96Q aggregation and toxicity (Fig. 6E and I). To test whether these two proteins function together, we co-expressed Mppa and UQCR-C1 in S2 cells and performed immunoprecipitation (IP). Each protein co-precipitated with the other, indicating complex formation (Fig. 7A). To assess their role in mitochondrial pre-protein processing, we expressed matrix-localized MitoGFP in larval fat bodies and analyzed its electrophoretic mobility by western blot. Control tissues displayed a single band corresponding to mature MitoGFP, whereas tissues with reduced Mppa or UQCR-C1 showed both precursor and mature forms (Fig. 7B), demonstrating impaired signal peptide cleavage. These data indicate that both proteins are required for mitochondrial pre-protein processing. Fig. 7Mppa forms a complex with UQCR-C1 to regulate mitochondrial protein homeostasis. A Immunoprecipitation experiments in S2 cells. B Larval fat bodies with the indicated genotypes were collected for western blot. The double GFP bands represent the precursor (upper band) and mature (lower band) forms of MitoGFP. α-Tubulin was a loading control. C Cg-GAL4 driven co-expression of indicated proteins with MitoGFP in fat bodies. C1–C4. The V5 tagged Mppa wild type or mutants (red) were co-expressed with MitoGFP (green). LacZ was overexpressed as a control (C1). C7–C10. The HA tagged UQCR-C1 wild type or mutants (red) were co-expressed with MitoGFP (green). C11. Co-expression of HA-tagged UQCR-C1 (gray) with V5-tagged Mppa (red in C11) or co-expression of mCherry-mito (red) with Mppa (gray in C5) (C5). C12. UQCR-C1 sgRNA was co-expressed with UAS-Cas9, MitoGFP (green), and V5-tagged Mppa (red in C12) or mCherry-mito (red in C6) (C6, as a control for C12) in larval fat bodies. D Confocal and TEM images of adult indirect flight muscles with the indicated gene overexpression. Fiber-like structures were indicated by white arrows. E Quantification of the ratio of mitochondria with aggregates (green arrows) in muscle TEM images. n = 5 TEM images from 3 animals of each genotype. F Cg-GAL4 driven the co-expression of HTT.96Q.Cerulean (green) with indicated proteins. G Quantification of the number of HTT96Q puncta in the fat bodies with the indicated genotypes. 25 larvae were dissected and n = 5 images were analyzed for each genotype. H Quantification of the size of HTT96Q puncta in the fat bodies with the indicated genotypes. 25 larvae were dissected and n = 5 images were analyzed for each genotype. I TEM analysis of adult fly retinas. J Quantification of rhabdomere number per ommatidium in retinas. n = 10 ommatidia from 3 flies of each genotype. Data were presented as mean ± SEM. ns: not significant; *p < 0.05; **p < 0.01; ***p < 0.001 Mppa forms a complex with UQCR-C1 to regulate mitochondrial protein homeostasis. A Immunoprecipitation experiments in S2 cells. B Larval fat bodies with the indicated genotypes were collected for western blot. The double GFP bands represent the precursor (upper band) and mature (lower band) forms of MitoGFP. α-Tubulin was a loading control. C Cg-GAL4 driven co-expression of indicated proteins with MitoGFP in fat bodies. C1–C4. The V5 tagged Mppa wild type or mutants (red) were co-expressed with MitoGFP (green). LacZ was overexpressed as a control (C1). C7–C10. The HA tagged UQCR-C1 wild type or mutants (red) were co-expressed with MitoGFP (green). C11. Co-expression of HA-tagged UQCR-C1 (gray) with V5-tagged Mppa (red in C11) or co-expression of mCherry-mito (red) with Mppa (gray in C5) (C5). C12. UQCR-C1 sgRNA was co-expressed with UAS-Cas9, MitoGFP (green), and V5-tagged Mppa (red in C12) or mCherry-mito (red in C6) (C6, as a control for C12) in larval fat bodies. D Confocal and TEM images of adult indirect flight muscles with the indicated gene overexpression. Fiber-like structures were indicated by white arrows. E Quantification of the ratio of mitochondria with aggregates (green arrows) in muscle TEM images. n = 5 TEM images from 3 animals of each genotype. F Cg-GAL4 driven the co-expression of HTT.96Q.Cerulean (green) with indicated proteins. G Quantification of the number of HTT96Q puncta in the fat bodies with the indicated genotypes. 25 larvae were dissected and n = 5 images were analyzed for each genotype. H Quantification of the size of HTT96Q puncta in the fat bodies with the indicated genotypes. 25 larvae were dissected and n = 5 images were analyzed for each genotype. I TEM analysis of adult fly retinas. J Quantification of rhabdomere number per ommatidium in retinas. n = 10 ommatidia from 3 flies of each genotype. Data were presented as mean ± SEM. ns: not significant; *p < 0.05; **p < 0.01; ***p < 0.001 When overexpressed in the fat body, Mppa formed fiber-like structures that transformed parts of mitochondria into elongated, filamentous forms (Fig. 7C2, S3A). IP analysis using V5- and HA-tagged Mppa in S2 cells revealed self-association (Fig. S3B), suggesting that Mppa fibers may arise from self-interaction. Overexpression of UQCR-C1 alone did not alter mitochondrial morphology (Fig. 7C7), but when co-expressed with Mppa, the Mppa fibers disappeared, and both proteins co-localized evenly throughout mitochondria (Fig. 7C11). In contrast, co-expression of Mppa with mCherry-MitoOMM did not affect its fiber formation (Fig. 7C5), indicating that UQCR-C1 specifically modulates Mppa distribution. Mppa fiber formation persisted in UQCR-C1 KO cells (Fig. 7C12), showing that it does not require endogenous UQCR-C1. Mutations in human PMPCA and PMPCB cause cerebellar ataxia and multiple mitochondrial dysfunctions syndrome 6 (MMDS6). Because catalytic and disease-associated residues are conserved, we generated Drosophila lines expressing mutant forms of Mppa and UQCR-C1 that mimic human disease variants or catalytic inactivation (Fig. S3C). All mutants localized to mitochondria, indicating that the point mutations did not affect targeting (Fig. 7C3, 7C4, 7C8, 7C9, 7C10). The disease-mimicking Mppa A407T formed fibers similar to wild-type Mppa (Fig. 7C3), whereas Mppa V283M failed to form fibers or alter mitochondrial morphology (Fig. 7C4), suggesting disrupted conformation or activity. In muscle, overexpression of wild-type Mppa or Mppa A407T induced fiber-like mitochondrial structures (white arrows in Fig. 7D2, 7D3). Mppa A407T overexpression also led to aggregate accumulation in mitochondria (green arrows in Fig. 7D3’ and Fig. 7E). In contrast, Mppa V283M expression caused disorganized muscle fibers (Fig. 7D4, D4’), loss of mitochondrial cristae, and protein aggregate formation in mitochondria (green arrows in Fig. 7D4’ and Fig. 7E). Overexpression of wild-type UQCR-C1 or disease-mimicking UQCR-C1 I403T did not alter mitochondrial morphology (Fig. 7D5, 7D5’, 7D6, 7D6’), but catalytic-site (UQCR-C1 E87Q) or pre-peptide recognition-site (UQCR-C1 E174/D178A) mutants caused muscle fiber disorganization (Fig. 7D7, 7D8). UQCR-C1 E87Q overexpression led to abnormal wing posture and progressive indirect flight muscle loss with age (Fig. S3D). Mitochondrial protein aggregates were abundant in UQCR-C1 E87Q, but rare in other variant overexpression (green arrows in Fig. 7D6–7D8 and Fig. 7E). Despite their phenotypic differences, mutant and wild-type Mppa and UQCR-C1 retained mutual interaction in IP assays (Fig. S3E, S3F). Finally, we examined whether the mutant forms could suppress HTT aggregation as their wild-type counterparts did. Similar to wild-type Mppa, both Mppa A407T and Mppa V283M reduced the number, but not the size, of HTT96Q puncta in fat body cells, whereas overexpression of either wild-type or mutant UQCR-C1 had no such effect (Fig. 7F and H). In the eye, both Mppa A407T and Mppa V283M mutants provided transient protection against HTT96Q-induced photoreceptor loss at day 2, but only wild-type Mppa retained protection at day 15 (Fig. 7I and J). These results suggest that disease-associated Mppa mutants partially preserve, but do not fully maintain, wild-type activity. Discussion Although several proteases, such as Afg3l2 [27], Spg7 [28], HtrA2 [13], YME1L [30], and Lon [29], have been previously studied in flies, their tissue-specific functions remain unclear, and systematic side-by-side comparisons are lacking. We found that loss of CG3107, YME1L, Immp2, HtrA2, or CG11771 caused male sterility, whereas knockout of Mipep, UQCR-C1, CG5131, Afg3l2, Spg7, ClpP, Lon, or Mppa resulted in lethality. Previous studies showed that all examined mitochondrial proteases are broadly expressed, with several being highly enriched in the testis (Table 1, source data are from Flybase, http://www.flybase.org). Consistent with this, the testis-enriched proteases YME1L, Immp2, and HtrA2 caused male sterility when deleted. Interestingly, although CG3107 (ortholog of PITRM1) and CG11771 (ortholog of MEP, with low homology to MIPEP) are broadly expressed, their KOs also led to male sterility. PITRM1 encodes a matrix peptidase that degrades peptides into amino acids, while MEP is an intermembrane space oligopeptidase and MIPEP is a matrix enzyme required for precursor protein maturation by octapeptide removal. Although CG11771 is the ortholog of MEP and with low homology to MIPEP, it lacks key catalytic residues for protease activity. Further study is required to elucidate whether CG11771’s function is related to mitochondrial protease. CG3107, CG11771, YME1L, Immp2, and HtrA2 participate in distinct protein degradation or cleavage pathways and may therefore disrupt spermatogenesis through different mechanisms. Our data indicated that the late steps of spermatogenesis seemed very sensitive to the loss of mitochondrial proteostasis. In this study, we examined mitochondrial morphology in multiple tissues upon loss of various mitochondrial proteases. Interestingly, mitochondrial phenotypes were highly consistent for the loss of certain genes but varied among tissues for others. Loss of Lon caused accumulation of protein aggregates within mitochondria in photoreceptor cells, muscles, and fat bodies. Loss of CG3107 led to mitochondrial aggregates in muscles and fat bodies but not in photoreceptor cells. Loss of CG11771, CG5131, Spg7, ClpP, or Mppa resulted in aggregates only in photoreceptor cell mitochondria, whereas no aggregates were detected in muscles or fat bodies (muscle-specific KO of Mppa was lethal). In contrast, loss of Mipep, Afg3l2, or Immp1 induced mitochondrial aggregates exclusively in fat bodies, but not in muscles or photoreceptor cells. Regardless of their predicted functions—protein import, activation, or quality control—the loss of many of these proteases led to the formation of mitochondrial protein aggregates. There was no clear pattern indicating which specific step in protein processing was disrupted to cause aggregate accumulation. These results suggest that different tissues rely on distinct sets of mitochondrial proteases to maintain proteostasis. It is not clear why the same protease exerts different effects across tissues. We speculate that tissue-specific differences in protease expression levels, mitochondrial abundance, metabolic demands, and proteostasis capacity may contribute to the distinct phenotypes observed upon loss or gain of these proteases. In addition, different tissues may contain distinct protease substrates, leading to differential dependence on specific mitochondrial proteases. Further studies will be required to elucidate these tissue-specific functions and regulatory mechanisms of mitochondrial proteases. In the fly eye, loss of mitochondrial proteases affected mitochondrial length, cristae morphology, or aggregate formation. Among these, only aggregate accumulation correlated well with retinal degeneration. Alterations in mitochondrial length or cristae structure were sometimes observed in young but not in aged flies, suggesting adaptive remodeling during aging. We also found that loss of UQCR-C1 or Afg3l2, but not other proteases, led to elevated mitophagy in the fat body. In mammalian cells, four mitochondrial proteases—PMPCB, PARL, AFG3L2, and ClpP—have been reported to participate in PINK1 degradation [9]. Consistent with this, Rho-7, the Drosophila ortholog of PARL, has been implicated in Pink1-dependent mitophagy [40], although it was not analyzed in this study. Notably, loss of ClpP did not enhance mitophagy, whereas loss of UQCR-C1 or Afg3l2 likely increased mitophagy due to their roles in Pink1 import and degradation. In this study, we found that loss of UQCR-C1, Mppa, or CG11771 increased the formation of HTT72Q aggregates in the fat body. Loss of UQCR-C1, Afg3l2, Spg7, Lon, or CLpP elevated total HTT72Q protein levels, with UQCR-C1 and Afg3l2 showing the strongest effects. The precise trigger for HTT72Q aggregation remains unclear, as increased protein abundance alone was insufficient to induce aggregate formation. Overexpression of Afg3l2, CG11771, or Mppa, but not UQCR-C1, reduced HTT96Q aggregate formation. Mppa and UQCR-C1 encode the two subunits of MPP, which removes MTS from imported proteins. Although loss of either subunit promoted HTT72Q aggregation, only Mppa overexpression—but not UQCR-C1—suppressed HTT96Q aggregation, suggesting that endogenous UQCR-C1 levels in fat body are sufficient and that its overexpression does not further enhance MPP activity. Afg3l2 is an m-AAA protease that degrades misfolded proteins within the inner mitochondrial membrane, either alone or in complex with Spg7 [5]. Loss of both proteases increased HTT72Q levels, though the effect of Spg7 loss was milder, indicating that both may contribute—directly or indirectly—to HTT72Q degradation. The role of CG11771 in HTT polyQ aggregation remains unclear, as it appears to be a pseudo-protease lacking key catalytic residues. It may retain residual proteolytic activity or function as a regulatory subunit, similar to the role of Mppa in modulating protease complexes [31]. There is currently no evidence that HTT polyQ proteins enter mitochondria and are directly degraded by the proteases identified in this study. Therefore, the observed changes in HTT polyQ aggregation and toxicity may instead reflect indirect effects mediated by mitochondrial protein import-dependent alterations in cytosolic protein homeostasis. Enhancing mitochondrial protease activity has been proposed as a therapeutic strategy to mitigate proteotoxic stress caused by misfolded disease proteins such as HTT polyQ and TDP-43. However, we observed that overexpression of many mitochondrial proteases caused severe defects. For example, HtrA2—previously reported to promote apoptosis [13]—caused lethality when overexpressed in either muscle or fat body. Similarly, overexpression of Mipep, Afg3l2, Spg7, Immp2, or Lon in muscle led to extensive muscle tissue loss and mitochondrial damage. These findings highlight the need for precise regulation of protease activity to avoid deleterious effects. We also found that overexpressed Mppa formed fiber-like structures within mitochondria, converting them into elongated, fiber-shaped organelles. Co-expression of UQCR-C1 with Mppa inhibited fiber formation, suggesting that Mppa may assemble into fibers only when unbound to UQCR-C1. It remains unclear whether fiber formation affects Mppa or MPP function. Notably, one disease-associated Mppa mutant failed to form fibers, while another retained this ability. Further investigation is needed to determine whether Mppa fiber formation has physiological significance. Conclusions In this study, we systematically analyzed 15 mitochondrial proteases across multiple tissues of Drosophila melanogaster. Our side-by-side comparison of these mitochondrial proteases provides new insights into the mitochondrial degradation network. Through extensive loss- and gain-of-function analyses, we revealed both shared and distinct physiological roles among these proteases. Interestingly, manipulation of mitochondrial protease activity markedly affected the formation and toxicity of HTT polyQ protein aggregates, suggesting that mitochondrial proteases may serve as potential targets for modulating not only mitochondrial function but also cytosolic protein homeostasis. Supplementary Information Supplementary Material 1. Supplementary Material 1. Supplementary Material 2. Supplementary Material 2.