Abstract

   Mitochondrial DNA (mtDNA) is exposed to multiple insults produced by
   normal cellular function. Upon mtDNA replication stress, the
   mitochondrial genome transfers to endosomes for degradation. Using
   proximity biotinylation, we found that mtDNA stress leads to the
   rewiring of the mitochondrial proximity proteome, increasing
   mitochondria’s association with lysosomal and vesicle-related proteins.
   Among these, the retromer complex, particularly VPS35, plays a pivotal
   role by extracting mitochondrial components. The retromer promotes the
   formation of mitochondrial-derived vesicles shuttled to lysosomes. The
   mtDNA, however, directly shuttles to a recycling organelle in a
   BAX-dependent manner. Moreover, using a Drosophila model carrying a
   long deletion on the mtDNA (ΔmtDNA), we found that ΔmtDNA activates a
   specific transcriptome profile to counteract mitochondrial damage.
   Here, Vps35 expression restores mtDNA homoplasmy and alleviates
   associated defects. Hence, we demonstrate the existence of a previously
   unknown quality control mechanism for the mitochondrial matrix and the
   essential role of lysosomes in mtDNA turnover to relieve mtDNA damage.
     __________________________________________________________________

   Lysosomal uptake of mtDNA controls mitochondrial quality.

INTRODUCTION

   Mitochondria are multifaceted organelles in charge of many essential
   cellular functions. Unique among cellular organelles, mitochondria
   contain their own genome, also known as mitochondrial DNA (mtDNA).
   Present in multiple copies per cell, the mitochondrial genome is
   organized within structures known as mitochondrial nucleoids. mtDNA is
   relatively small but encodes 13 critical proteins essential for the
   mitochondrial respiratory chain and a subset of RNAs and tRNAs required
   for mitochondrial protein synthesis. All these components are
   indispensable for maintaining mitochondrial function and cellular
   homeostasis ([44]1).

   One of the main subproducts generated during the coupling of the
   different mitochondrial pathways to produce adenosine triphosphate
   (ATP) are reactive oxygen species (ROS) ([45]2). ROS, including
   superoxide anions, hydrogen peroxide, and hydroxyl radicals, are
   generated when electrons leak from the electron transport chain and
   react with oxygen molecules. ROS damages not only proteins and lipids
   but also the mtDNA, modifying the biochemical properties of the
   molecule, altering base-pairing properties, and affecting the
   maintenance of the mitochondrial genome, serving as a source for
   mutations ([46]3, [47]4). These mutations impair respiratory chain
   complexes, increasing oxidative stress and triggering a self-amplifying
   cycle that results in mitochondrial dysfunction. Thus, the progressive
   accumulation of mtDNA damage influences many human diseases,
   accelerates aging, and, in marked cases, even leads to cellular death
   ([48]5).

   From energy production and ion exchange to metabolite synthesis, all
   mitochondrial functions depend on maintaining the integrity of the
   network through mitochondrial turnover. Since the discovery of
   selective degradation of mitochondria in mammals in the early 21st
   century, numerous parallel pathways have been identified ([49]6). The
   PINK1-Parkin autophagy pathway targets large mitochondrial segments for
   degradation in response to acute damage or in situations requiring
   mitochondrial adaptation, such as during cell differentiation or
   changes in nutrient availability ([50]7). Other complementary pathways
   work in parallel to selectively remove only dysfunctional mitochondrial
   pieces, thereby preserving the healthy portions of the network and
   maintaining cellular homeostasis ([51]6).

   The presence of mtDNA in autophagy-related structures, independent of
   other mitochondrial parts, suggests a specific selective quality
   control pathway for the mitochondrial genome ([52]8). Several
   mechanisms have been proposed to explain how mtDNA is degraded in
   response to stress, including mitochondrial-derived vesicles (MDVs)
   ([53]9) and piecemeal mitophagy ([54]10, [55]11). Nevertheless, the
   exact mechanism seems to depend on the initial stress. Thus, upon
   fumarate accumulation, the mitochondrial genome appears encapsulated in
   TOM20−/PDH+ MDVs ([56]9). Upon antimycin/oligomycin stress, the inner
   membrane and, eventually, the mtDNA are eliminated through
   lysosomal-dependent piecemeal mitophagy through VDAC1 pores ([57]11).
   Furthermore, cells treated with low levels of ethidium bromide separate
   mitochondrial nucleoids into inner membrane subdomains, which, upon
   fission, are eliminated in autophagosome-dependent piecemeal mitophagy
   ([58]10). In this context, the mitochondrial inner membrane PHB2 serves
   as a mitophagy receptor after the degradation of the outer
   mitochondrial membrane ([59]10, [60]12). Independent of the
   specificities of each path, all of them point toward the separation of
   a mitochondrial piece and its degradation in recycling organelles.

   Interfering with mtDNA replication leads to the accumulation of
   oxidative damage in the mitochondrial genome ([61]13). The biochemical
   changes initiated by ROS modify the architecture of the mitochondrial
   membranes and allow the translocation of the mtDNA to the endosomal
   compartment, where it will be degraded within recycling organelles
   ([62]13). We and others recently found that the extraction of the mtDNA
   responds to the coordination between mitochondrial membrane and vesicle
   trafficking proteins. The endosomal protein RAB5C physically interacts
   with Mitofusin 1 and Mitofusin 2 and assists in endosomal approach and
   mtDNA transfer ([63]14). Similarly, proper disposal of mtDNA requires
   the mitochondrial membrane protein SAMM50 ([64]13). Disturbance of this
   selective quality control pathway leads to the activation of the innate
   immune response with potential implications in aging and several human
   diseases ([65]14–[66]16).

   In the current study, we sought to elucidate the mechanisms underlying
   the elimination of the mitochondrial genome upon mtDNA replication
   stress. By spatial proteomics of the mitochondria-endosome contact
   sites, we identified a group of lysosomal proteins involved in mtDNA
   turnover. Mitochondrial matrix components, including the mtDNA, exit
   the mitochondrial network assisted by the retromer, although they do
   not follow the same path. The accumulation of the mtDNA outside the
   mitochondrial compartment is BAX dependent and enhanced by lysosomal
   inhibition with chloroquine. Using correlative light and electron
   microscopy (CLEM), we demonstrate that the piecemeal removal of mtDNA
   from mitochondria occurs with direct transfer to a recycling organelle.
   The retromer assists this process by stimulating the lysosomal function
   and mitochondrial fragmentation. Furthermore, using a Drosophila model
   with a long deletion in the mtDNA (ΔmtDNA), we validated the role of
   the retromer in promoting mtDNA quality control in vivo. Our results
   reveal a mechanism for the degradation of the mitochondrial matrix,
   including the mtDNA, and highlight the lysosomes as the central
   organelles maintaining mitochondrial function.

RESULTS

mtDNA damage changes the proteome of mitochondria-endosome contact sites

   Disrupting the mtDNA replication machinery leads to the transfer of
   mitochondrial nucleoids to the endosomal compartment ([67]13). To
   specifically impair mtDNA replication, we expressed a dominant negative
   mutation of the mitochondrial helicase Twinkle (TWNK^K319E) in HeLa
   cells ([68]Fig. 1A). The equivalent mutation in mice (TWNK^K320E) has
   been validated as a reliable tool to induce mtDNA alterations,
   oxidative stress, and replication defects ([69]17–[70]19). Expression
   of TWNK^K319E reduced mtDNA copy number 48 hours posttransfection
   ([71]Fig. 1B).

Fig. 1. mtDNA damage follows a lysosomal-dependent degradation pathway.

   [72]Fig. 1.
   [73]Open in a new tab

   (A) Confocal images of HeLa cells in the steady state and expressing
   TWNK^K319E-mCherry (mtDNA stress) labeled with α-TOM20. (B) mtDNA copy
   number analysis in HeLa cells transiently transfected with
   TWNK^K319E-mCherry. (C) Schematic representation of the SplitTurboID
   assay. The early endosomal marker RAB5C and the mitochondrial outer
   membrane protein SAMM50 were used to determine the proximity proteome
   of mitochondria and endosomes. Illustration from NIAID NIH BIOART
   Source. (D and E) Immunostaining of HeLa cells expressing (D)
   RAB5C-SplitNt-V5 and (E) SAMM50-SplitCt-HA labeled with α-RAB5 and α-V5
   and α-TOM20 and α-HA, respectively. (F and G) Pathway enrichment
   analysis with Metascape showing GO terms for proteins differentially
   enriched (F) in the steady state and (G) after expression of
   TWNK^K319E-mCherry. (H and I) Volcano plot showing proteins enriched
   after biotinylation and purification of cells expressing (H)
   SplitTurboID plasmids and (I) TWNK^K319E-mCherry. Differentially
   expressed proteins compared with cells transduced only with
   SAMM50-SplitCt-HA (significant: q value < −0.05 and absolute log[2] FC
   > 1) are highlighted in blue (steady state) or red (mtDNA stress)
   (n = 3 samples per group). P values were calculated using one-way ANOVA
   with Tukey correction for multiple comparisons. Scale bars, 10 μm. Data
   are presented as means ± SEM.

   To detect regulators involved in this pathway, we used the promiscuous
   biotinylation enzyme TurboID ([74]20). For specific detection of the
   mitochondria-endosome proteome, we used Split-TurboID ([75]Fig. 1C).
   Thus, we fused the N-terminal part of TurboID to RAB5C (TurboID amino
   acids 1 to 72; RAB5C-SplitTurboNt-V5) and the C-terminal part to SAMM50
   (TurboID amino acids 73 to 246; SAMM50-SplitTurboCt-HA), two proteins
   related to the endosomal disposal of the mtDNA ([76]13, [77]14). The
   expected subcellular localization of both transgene-encoded fusion
   proteins at endosomes and mitochondria, respectively, was validated in
   HeLa cells with confocal microscopy ([78]Fig. 1, D and E).

   To isolate a high number of biotinylated proteins, we generated human
   embryonic kidney (HEK) 293 stable cell lines expressing Split-TurboID
   plasmids by lentiviral transduction. Reconstitution of biotinylation
   activity was confirmed by Western blot and immunofluorescence (fig. S1,
   A and B). We noticed that cells expressing only SAMM50-SplitTurboCt-HA
   displayed residual enzymatic activity (fig. S1A). Therefore, to
   minimize potential background interference, we used these cells as a
   negative control and cells expressing both SAMM50-SplitTurboCt-HA and
   RAB5C-SplitTurboNt-V5 as the experimental approach.

   The proximity proteome of the mitochondria and endosomes in the steady
   state showed enrichment of proteins involved in intra-Golgi vesicle
   transport and mitochondrial protein import ([79]Fig. 1, F and H, and
   data S1). On the contrary, mtDNA replication stress elicited a complete
   rewiring of the mitochondria-endosome proteome ([80]Fig. 1, G and I,
   and data S2). Pathway analysis revealed the presence of mitochondrial
   matrix proteins (galactokinase, GALK1; pyruvate dehydrogenase
   component, PDHX; fumarate hydratase, FH) and nucleoid and mitochondrial
   inner membrane proteins (DNAJC11; prohibitin 2, PHB2; OXA1L), pointing
   to a rupture of the mitochondrial outer membrane and consequent
   biotinylation of mitochondrial matrix components. In addition, we also
   detected proteins related to coated vesicles, such as the retromer
   subunit VPS29 and RAB10, together with the lysosomal proteins MYO1B,
   MYO6, ATP6V0D1, and cathepsin D (CTSD). Thus, mtDNA stress changes the
   proximity proteome of mitochondria toward lysosomal recruitment to
   facilitate the degradation of mitochondrial components.

The retromer facilitates mitochondrial fragmentation upon mtDNA replication
stress

   The retromer is a trimeric protein complex formed by three components:
   VPS35, VPS29, and VPS26, mainly the isoform VPS26A. The main function
   of this complex is to control vesicle trafficking among endosomes and
   lysosomes and promote their maturation and degradative activity
   ([81]21). However, upon certain conditions, the retromer and primarily
   its core component, VPS35, participates in mitochondrial quality
   control. In an oxidative environment, the retromer generates
   mitochondrial vesicles directed to the peroxisomes ([82]22). Upon mtDNA
   replication stress, the mitochondrial genome transfers to retromer foci
   ([83]13). How this translocation occurs is not clear.

   To investigate how the retromer licenses mitochondrial quality control
   pathways during mtDNA stress, we first analyzed its impact on
   mitochondrial dynamics. As described, at a steady state, expression of
   the retromer component VPS29-green fluorescent protein (GFP) formed
   cytosolic foci colocalizing within the other retromer components, VPS35
   and VPS26A, but not mitochondria ([84]Fig. 2A and fig. S2A).
   Immunostaining with other vesicle proteins showed partial
   colocalization with both early (RAB5) and late endosomes (LAMP1) (fig.
   S2A). TWNK^K319E did not modify the subcellular pattern of the retromer
   within the vesicular network (fig. S2B). Nevertheless, we noticed
   higher colocalization of VPS29 with mitochondria ([85]Fig. 2, B and C).
   Expression of either VPS29-GFP, VPS35-GFP, or TWNK^K319E-mCherry alone
   had no major effect on the mitochondrial network morphology (fig. S2,
   C, D, and F). Similarly, cells coexpressing both VPS29–GFP and
   TWNK^K319E-mCherry showed no differences (fig. S2, E and F). However,
   cells coexpressing VPS35-GFP and TWNK^K319E-mCherry showed robust
   mitochondrial fragmentation, depicted by an increased number of
   mitochondria and reduced average size (fig. S2, E and F).

Fig. 2. The retromer participates in the response to mtDNA replication stress
by adapting the mitochondrial morphology.

   [86]Fig. 2.
   [87]Open in a new tab

   (A and B) Immunostaining of HeLa cells expressing VPS29-GFP and (B)
   TWNK^K319E-mCherry labeled with α-VPS35 and α-TOM20. (C) Manders’
   correlation coefficient between TOM20 and VPS29. n = 3, 10 images per
   replicate. (D to I) Western blot analysis and quantification of
   retromer components upon [(D) and (E)] VPS26A, [(F) and (G)] VPS29, or
   [(H) and (I)] VPS35 KO. α-GAPDH or α-actin was used as a loading
   control. GAPDH, glyceraldehyde-3-phosphate dehydrogenase. (J to L)
   α-TOM20 immunostaining in (J) VPS35, (K) VPS29, and (L) VPS26 KO, in
   the steady state and expressing TWNK^K319E-mCherry. (M to O)
   Quantification of mitochondrial morphology parameters. n = 3, >20 cells
   per replicate. (P) α-TOM20 immunostaining in VPS29 and VPS26A KO
   expressing TWNK^K319E-mCherry and VPS35-GFP. P values were calculated
   using Student’s t test [(C), (E), (G), and (I)] and one-way ANOVA [(M)
   to (O)]. Scale bars, 10 μm. Data are presented as means ± SEM.

   Further, we generated CRISPR-Cas9 HeLa knockout (KO) cells for each
   retromer subunit. The KO of VPS26A did not affect other retromer
   subunits ([88]Fig. 2, D and E). On the contrary, VPS29 KO decreased the
   steady-state level of VPS35 and VPS26A ([89]Fig. 2, F and G). VPS35 KO
   led to an up-regulation of VPS29 and decreased VPS26A ([90]Fig. 2, H
   and I). Then, we analyzed the effect of the KO on the mitochondrial
   morphology ([91]Fig. 2, J to L). In the steady state, all three KO cell
   lines presented an increased number of mitochondria and mitochondrial
   area ([92]Fig. 2, M to O). Opposite to what we observed in wild-type
   (WT) cells, the expression of TWNK^K319E led to mitochondrial
   elongation in VPS26 and VPS29 KO, whereas no effect was observed in
   VPS35 KO cells ([93]Fig. 2, J to O). We wondered whether the lack of
   mitochondrial fission induced by mtDNA stress and observed upon the KO
   of any of the three retromer subunits was due to the reduced levels of
   VPS35 in all KO models ([94]Fig. 2, D to I). Consistently, the
   reexpression of VPS35 in VPS29 and VPS26 KO restored mitochondrial
   fragmentation ([95]Fig. 2, M, O, and P).

   These data collectively revealed that the retromer is necessary for the
   mitochondrial response to mtDNA replication stress, and its core
   component VPS35, independently of its accessory subunits, retains the
   ability to fragment mitochondria.

The retromer facilitates the release of mitochondrial matrix components
during mitochondrial stress

   One proposed mechanism for retromer-dependent mitochondrial quality
   control is the formation of MDVs ([96]22). MDVs are intermediate
   structures formed upon different types of stresses and containing
   selective cargo. Up to now, many types of vesicles derived from the
   mitochondria have been described. MDVs contain either one of both
   mitochondrial membranes and are enriched for specific cargo ([97]23).
   The involvement of the mitochondrial fission factor Dynamin-Related
   Protein 1 (DRP1) seems to depend on the type of vesicle ([98]24,
   [99]25). MDVs are, however, independent of autophagosome-related
   proteins ([100]26).

   Thus, we first examined the presence of vesicle-like structures induced
   by the retromer and mtDNA replication stress. As an internal control,
   we included the Parkinson’s disease (PD)–associated mutation
   VPS35^D620N, known to restrict cargo selection and mitochondrial
   clearance ([101]27, [102]28). Because the mitochondrial genome resides
   in the mitochondrial matrix, we tested for the presence of vesicles
   carrying pyruvate dehydrogenase (PDH) and lacking the outer membrane
   protein TOM20 ([103]29). In the steady state, VPS35 did not colocalize
   with the mitochondrial markers PDH or TOM20 (fig. S3A). Expression of
   TWNK^K319E-mCherry induced the formation of sparse TOM20−/PDH+ foci,
   which accumulated upon lysosomal inhibition with chloroquine ([104]Fig.
   3, A and B). Very few vesicles were observed upon chloroquine treatment
   alone or expression of VPS35-GFP alone (Fig. S3B and C), suggesting
   that they rarely formed in a steady state. The coexpression of
   TWNK^K319E and VPS35 increased the accumulation of TOM20−/PDH+ foci
   ([105]Fig. 3C), which colocalize with the late endosomal/lysosomal
   marker LAMP1 (fig. S3, D and E).

Fig. 3. mtDNA replication stress and oxidation initiate the release of
mitochondrial components.

   [106]Fig. 3.
   [107]Open in a new tab

   (A) HeLa cells expressing VPS35-GFP and TWNK^K319E-mCherry and labeled
   with the mitochondrial outer membrane marker α-TOM20 and mitochondrial
   matrix α-PDH. (B to D) Quantification of α-TOM20− and α-PDH+ foci [(B)
   and (C)] in control cells expressing the indicated plasmids and (D) in
   retromer deficient cells. n = 3, >15 cells per replicate. (E)
   Immunostaining of VPS29 KO cells expressing TWNK^K319E-mCherry and
   labeled with α-TOM20− and α-PDH+ foci. (F) Fluorescence profile of
   50-pixel line. The selected area is highlighted above the graph. (G and
   H) Immunostaining of α-TOM20 and α-PDH in cells treated with (G) 200 μM
   H[2]O[2] 4 hours or (H) grown in galactose medium overnight. (I)
   Quantification of α-TOM20− and α-PDH+ foci in H[2]O[2] and
   galactose-treated cells. Where indicated, cells were treated with 10 μM
   chloroquine 4 hours before fixation to block lysosomal function (n = 3,
   >15 cells per replicate). TOM20−/PDH+ foci are encircled in magenta
   circles. P values were calculated using one-way ANOVA with Tukey
   correction for multiple comparisons. Scale bars, 10 μm. Data are
   presented as means ± SEM.

   Cells expressing VPS35^D620N-GFP or VPS35 KO cells showed reduced
   TOM20−/PDH+ vesicles ([108]Fig. 3, A, C, and D). On the other hand, the
   expression of VPS29 did not induce the formation of any type of
   mitochondrial vesicles (fig. S3F), suggesting again that VPS35 retains
   the mitochondrial fission activity. Moreover, TWNK^K319E did not induce
   any TOM20−/PDH+ vesicles without VPS29 ([109]Fig. 3D), although VPS35
   was still recruited to the mitochondria ([110]Fig. 3, E and F).

   Next, we asked whether the formation of these vesicular structures
   relies on the MDVs machinery, specifically Miro1 (RHOT1) and DRP1
   ([111]6). To test this, we generated Miro1 KO HeLa cells and expressed
   TWNK^K319E-mCherry along with VPS35-GFP (fig. S3, G to I). Although we
   detected sparse TOM20−/PDH+ foci, their number remained unchanged upon
   mtDNA replication stress (fig. S3F). Similarly, DRP1 KO HeLa cells
   showed minimal independent TOM20−/PDH+ foci (fig. S3, G to I). Notably,
   DRP1 KO cells maintained an elongated mitochondrial network, even upon
   TWNK^K319E expression (fig. S3J). Likewise, Miro1 KO cells displayed a
   tubulated network, which became aggregated under mtDNA replication
   stress (fig. S3J).

   Furthermore, we asked whether the elimination of mitochondrial content
   mediated by the retromer also occurred when interfering with
   mitochondrial function by other means. Given that mtDNA replication
   stress causes oxidative damage ([112]13) and this is one of the most
   common forms for natural acquisition of mtDNA mutations ([113]30), we
   treated our cells with a sublethal concentration of H[2]O[2]. In
   parallel, we also grew the cells in galactose, a condition that
   supports mitochondrial metabolism and oxidation ([114]31). Thus, we
   observed the formation of TOM20−/PDH+ foci enriched upon the inhibition
   of the lysosomal function with chloroquine ([115]Fig. 3, G to I).
   Again, these particles were not detectable upon the expression of the
   PD-associated mutant VPS35^D620N ([116]Fig. 3, G to I).

   Recently, a mechanism involved in the elimination of the mitochondrial
   inner membrane, and occasionally mtDNA, was described to follow vesicle
   formation through VDAC1 pores ([117]11). However, upon mtDNA
   replication stress, immunostaining of the inner mitochondrial membrane
   subunit of the complex V ATPase (adenosine triphosphatase), ATP5A,
   showed no foci independent of the outer membrane protein TOM20 (fig.
   S3K).

   Together, these findings confirm that the downstream response to the
   mtDNA replication stress includes mitochondrial vesiculation controlled
   by Miro1 and DRP1, independent of vesicles derived from the inner
   membrane (VDIMs) ([118]11). mtDNA replication stress follows the
   activation of lysosomal-dependent quality control mechanisms for the
   mitochondrial matrix that is related to vesicle formation, resembling
   the MDV pathway.

mtDNA ejection during mtDNA stress occurs directly to vesicles from the
endosomal system

   We sought to investigate whether the elimination of the mitochondrial
   genome follows the same path observed for mitochondrial matrix
   components. We found VPS35-GFP in the exit sites of both PDH and the
   mtDNA ([119]Fig. 4, A and B). To explore the role of the retromer in
   the extraction of mtDNA, we generated cells stably expressing VPS35-V5.
   Quantitative polymerase chain reaction (qPCR) analysis of mtDNA genes
   in mitochondria-free fractions showed increased MT-ND1 and MT-ND5 in
   cells with mtDNA replication stress ([120]Fig. 4C and fig. S4A). VPS35
   overexpression further increased extraction, whereas no changes were
   observed in cells with the dominant negative VPS35^D620N. Consistently,
   coexpression of TWNK^K319E and VPS35 increased the number of cytosolic
   double-stranded DNA (dsDNA) foci, which were further enriched upon
   lysosomal inhibition with chloroquine ([121]Fig. 4, D and E, and fig.
   S4, B to D). In contrast, cells expressing VPS35^D620N exhibited less
   extra-mitochondrial dsDNA ([122]Fig. 4, D and E). Unexpectedly, Miro1
   KO cells also displayed nonmitochondrial dsDNA foci ([123]Fig. 4, F and
   G; Fig. S4E). Notably, MDV analysis in cells transfected with
   mtDNA-Kaede (fig. S4F) showed either TOM20−/PDH+ foci or TOM20−/mtDNA+,
   and almost none TOM20−/PDH+/dsDNA+ vesicles ([124]Fig. 4, H and I, and
   fig. S4G). These data suggest that, although mitochondrial matrix MDVs
   and mtDNA release are activated under mtDNA replication stress, mtDNA
   release occurs independently of the vesiculation of the mitochondrial
   matrix.

Fig. 4. mtDNA replication stress triggers Bax-dependent mtDNA release.

   [125]Fig. 4.
   [126]Open in a new tab

   (A) HeLa cells expressing TWNK^K319E-mCherry and VPS35-GFP labeled with
   α-TOM20 and α-dsDNA or α-PDH. (B) Fluorescence profile of 50-pixel line
   labeled in (A). (C) qPCR quantification of ND1 and ND5 mtDNA genes in
   mitochondria-free cytosolic fractions (n = 3). (D to G) Analysis and
   quantification of cytosolic dsDNA foci upon mtDNA replication stress by
   immunostaining with α-TOM20 and α-dsDNA in [(D) and (E)] VPS35-GFP
   expressing cells and [(F) and (G)] Miro1 KO cells. (n = 3, >15 cells
   per replicate). (H) HeLa cells expressing mtDNA-Kaede and
   TWNK^K319E-mCherry labeled with α-TOM20 and α-PDH. (I) Quantification
   of TOM20- Vesicles containing PDH, dsDNA or both (n = 3, >20 cells per
   replicate). (J and K) Analysis and quantification of the involvement of
   mitochondrial membrane pores in mtDNA release. Where indicated, 24
   hours after transfection, cells were further treated 4 hours with 10 μM
   chloroquine, 25 μM VBIT-12 for VDAC1 inhibition, or 100 μM V5 peptide
   for the inhibition of BAX oligomerization (n = 3, >15 cells per
   replicate). TOM20−/PDH+ foci are encircled by magenta circles.
   TOM20−/dsDNA^+ and TOM20−/mtDNA-Kaede^+ foci are encircled by magenta
   circles. DMSO, dimethyl sulfoxide. P values were calculated using
   one-way ANOVA with Tukey correction for multiple comparisons [(C), (E),
   and (I)] and Student’s t test [(G) and (K)]. Scale bars, 10 μm, except
   for (H), 5 μm. Data are presented as means ± SEM.

   Thus, we investigated whether mitochondrial pores contribute to the
   mtDNA release. Cells treated with the VDAC1-specific inhibitor VBIT-12
   ([127]32) in combination with chloroquine showed no changes in the
   number of cytosolic dsDNA foci ([128]Fig. 4, J and K). In contrast,
   treatment with the BAX inhibitor peptide V5 ([129]33) showed a
   significantly reduced number of nonmitochondrial dsDNA foci ([130]Fig.
   4, J and K).

   Our data challenge the notion of vesicle-mediated mtDNA release,
   suggesting that, upon mtDNA replication stress, the mitochondrial
   genome may be released directly into the cytoplasm. To visualize the
   cytosolic dsDNA foci in high resolution and gain information about the
   cellular context, we performed CLEM and combined it with electron
   tomography. We stained chloroquine-treated cells with SYBR Gold for DNA
   and PK Mito Deep Red for mitochondria (fig. S5A). In contrast to
   control cells, cells expressing TWNK^K319E-mCherry showed cytosolic DNA
   foci ([131]Fig. 5A and fig. S5B). These cells presented reduced
   mitochondrial size and fewer cristae per mitochondria (fig. S5, C to
   E). The correlation of confocal images with transmission electron
   microscopy and electron tomography showed that areas containing
   cytosolic dsDNA were enriched in recycling organelles at different
   stages of maturation ([132]Fig. 5, B and C; fig. S4, F and G; and movie
   S1). None of the cells analyzed with CLEM (n = 6) showed DNA inside a
   small independent free cytosolic vesicle or membrane-free in the
   cytoplasm. Next, we performed the same approach in cells expressing
   TWNK^K319E-mCherry and VPS35–cyan fluorescent protein (CFP) ([133]Fig.
   5D and fig. S4H). Correlation experiments showed the presence of
   VPS35-CFP in the same type of recycling organelle containing
   electron-dense material and in close apposition with mitochondria
   ([134]Fig. 5, E and G, and movie S2). Electron tomography
   reconstitution of one mtDNA exit site confirmed the transfer of mtDNA
   to one of these nearby vesicles also labeled with VPS35-CFP ([135]Fig.
   5, F and H, and movie S3). Consistently, immunostaining of
   TWNK^K319E-mCherry cells showed an increased number of dsDNA in contact
   with the lysosomal/late endosomal marker LAMP1 (fig. S5, I and J).

Fig. 5. mtDNA release occurs through direct transfer to a recycling
organelle.

   [136]Fig. 5.
   [137]Open in a new tab

   (A) CLEM of cells expressing TWNK^K319E-mCherry and labeled with SYBR
   Gold and PK MitoDeep Red. Boxes labeled with (a) and (b) indicate the
   areas used for CLEM. (B and C) Volumetric reconstitution of electron
   tomographies of cytosolic mtDNA areas. The organelle containing dsDNA
   is shown in green, other recycling organelles in lilac, mitochondrial
   outer membrane in gray, and mitochondrial cristae in cyan. (D)
   TWNK^K319E-mCherry cell further expressing VPS35-CFP. Boxes labeled
   with (c) and (d) indicate the areas used for CLEM. (E and F)
   Correlation of VPS35-CFP foci and (F) mtDNA exit site. (G and H)
   Volumetric reconstitution of electron tomographies in VPS35-CFP cells.
   In all cases, cells were treated with 10 μM chloroquine for 4 hours.
   For CLEM, all dyes were loaded for 1 hour before fixation. Scale bars,
   500 nm for reconstituted tomograms [(B), (C), (G), and (H)], 1 μm [(E)
   and (F)], and 10 μm [(A) and (D)].

Active lysosomes facilitate mtDNA turnover

   One of the main functions of the retromer complex is the maturation of
   early endosomes toward late endosomes/lysosomes. In HeLa cells, the
   retromer promotes the processing of lysosomal hydrolases, influencing
   lysosomal morphology and function ([138]34). Lysosomes are highly
   dynamic organelles that move fast in cells, adapting the degradative
   capacity of the cells to cellular needs ([139]35). The positioning of
   lysosomes within the cell affects their activity and interactions with
   other organelles ([140]36). Perinuclear lysosomes are typically
   involved in degradation, whereas those in the periphery function in the
   endocytic processes. Rab guanosine triphosphatases (GTPases) coordinate
   with the retromer complex to regulate lysosomal positioning and
   trafficking ([141]37).

   One of the proteins that we identified by proximity proteomics was
   RAB10 ([142]Fig. 1I). Upon different stimuli, RAB10 influences the
   distribution of lysosomes between the perinuclear region and the cell
   periphery, thereby affecting cellular processes like autophagy and
   lysosomal degradation ([143]38). Upon mitochondria depolarization,
   RAB10 binds to the autophagy receptor optineurin and promotes its
   accumulation, thereby facilitating mitophagy ([144]39).

   To determine how RAB10-lysosomes participate in the downstream response
   to mtDNA stress, we used the expression of three different alleles,
   RAB10 WT, the gain-of-function allele RAB10^Q68L, and the dominant
   negative RAB10^T23N. The activity of RAB10 is influenced by its binding
   status to GTP (guanosine triphosphate) ([145]40). Although the WT
   allele can shift from an active to an inactive form, the
   gain-of-function always remains active, in contrast to the dominant
   negative mutation, which always remains inactive. Therefore, the
   changes induced by RAB10 are expected to be enhanced for the RAB10^Q68L
   allele.

   Consistently, mtDNA replication stress increased the association of
   RAB10^Q68L-GFP with the retromer, pointing to the activation of a
   lysosome-dependent path to degradation ([146]Fig. 6, A and B).
   Quantitative analysis of mitochondrial morphology showed that, upon
   mtDNA replication stress, cells expressing WT RAB10 or the
   gain-of-function allele RAB10^Q68L present an increased number of
   mitochondria and reduced average size, more prominent for the
   gain-of-function allele. However, no changes were detectable in cells
   expressing the dominant negative variant ([147]Fig. 6, C to E).

Fig. 6. The small GTPase RAB10 promotes mitochondrial fragmentation and mtDNA
degradation in lysosomes.

   [148]Fig. 6.
   [149]Open in a new tab

   (A) Immunostaining of HeLa cells expressing the constitutive active
   protein RAB10^Q68L-GFP labeled with α-VPS35. (B) Manders’ correlation
   coefficient between RAB10 and VPS35. (C and D) Confocal images of cells
   expressing WT RAB10-GFP, constitutive active RAB10^Q68L-GFP, dominant
   negative RAB10^T23N-GFP, in the steady state, and (D) expressing
   TWNK^K319E-mCherry, labeled with α-TOM20. (E) Quantification of the
   mitochondrial morphology in RAB10 expressing cells (n = 3, >20 cells
   per replicate). (F and G) Cells expressing RAB10^Q68L-GFP and (G)
   TWNK^K319E-mCherry labeled with α-LAMP1 and α-dsDNA. Arrows depict
   RAB10-LAMP1-dsDNA foci. (H) Manders’ correlation coefficient between
   RAB10-GFP and LAMP1 and LAMP1 and dsDNA (n = 3, 10 images per
   replicate). (I) RAB10-GFP coimmunoprecipitation in the steady state and
   cells expressing TWNK^K319E-mCherry with the lysosomal protein LAMP1. P
   values were calculated using one-way ANOVA with Tukey correction for
   multiple comparisons. Scale bars, 10 μm. Data are presented as
   means ± SEM.

   Then, we investigated the subcellular location of RAB10 upon mtDNA
   stress. In the steady state, RAB10 partially localized with late
   endosomes, but no dsDNA could be detected in these particles ([150]Fig.
   6F). Upon mtDNA replication stress, RAB10 accumulated in LAMP1-positive
   late endosomes together with extra-mitochondrial dsDNA ([151]Fig. 6G).
   In agreement, colocalization between RAB10 and LAMP1, as well as
   between LAMP1 and dsDNA, was enhanced ([152]Fig. 6H). Pull-down
   experiments with RAB10-GFP confirmed a strong association with LAMP1
   upon mtDNA stress ([153]Fig. 6I). In summary, these data confirm that
   the downstream response associated with mtDNA replication stress is
   regulated by RAB10-active lysosomes.

Generation of long mtDNA deletion in Drosophila larval epidermis

   Our findings suggest that the retromer enhances specific quality
   control mechanisms for the mtDNA. We hypothesized that overexpression
   of VPS35 might also be beneficial in vivo, by reducing the
   mitochondrial burden associated with mtDNA mutations. To investigate
   our hypothesis, we used a well-established Gal4/upstream activated
   sequence (UAS) Drosophila model based on nucleases ([154]41).
   Drosophila mtDNA harbors two target sites for the restriction enzyme
   Afl III. Coexpression of mitochondria-targeted Afl III (mitoAfl III)
   and the DNA ligase from phage T4 (mitoT4lig) ([155]Fig. 7A; UAS-mitoAfl
   III, UAS-mitoLigase) results in the generation of a 2564–base pair (bp)
   deletion in the mtDNA, including the CytB gene (mtDNA^Δ10.789-13.372).

Fig. 7. VPS35 overexpression recovers mtDNA homoplasmy in Drosophila.

   [156]Fig. 7.
   [157]Open in a new tab

   (A) Schematic representation of the approach to generate mtDNA deletion
   covering 2564 bp in larval epidermis. The restriction enzyme Afl III
   and the T4-DNA ligase directed to mitochondria were both expressed
   under the UAS promotor. A-58 Gal4 was used to drive the expression of
   the transgenes in the larval epidermis. WT, wild-type mtDNA.
   Illustration from NIAID NIH BIOART Source. (B) Bright-field images of
   transgenic L3 larvae used in this study. (C) CytB/ND5 ratio showing
   mtDNA heteroplasmy obtained by qPCR from total DNA extracts of L3
   larvae (Control, n = 8; Epi-ΔmtDNA, n = 9). (D) mtDNA copy number
   quantification using ND5 gene and TUBB as a mitochondrial and nuclear
   gene, respectively (n = 9). (E and F) Western blot analysis (E) and
   quantification (F) of total protein extracts of L3 larvae for SDHB and
   ATPA. Ponceau S (Ps) was used as a loading control (Control, n = 3;
   Epi-ΔmtDNA, n = 4). (G and H) mtDNA copy number (G) and mtDNA
   heteroplasmy quantification (H) upon genetic manipulations (Control,
   n = 9; Epi-ΔmtDNA, n = 9; Epi-ΔmtDNA-Vps35^OE, n = 9). (I) Electron
   microscopy images of larval epidermis. Mitochondria were pseudocolored
   in yellow and lysosomes in blue (dark content). Control refers to the
   A58Gal4/+ genotype (B) and specific UAS control in other
   quantifications. P values were calculated using the Student’s t test
   [(C), (D), and (F)] and one-way ANOVA with Tukey correction for
   multiple comparisons [(G) and (H)]. Scale bars, 1 mm (B) and 500 nm
   (I). Data are presented as means ± SEM.

   Simultaneous induction of both UAS constructs (UAS-mitoAfl III,
   UAS-mitoLigase) by most tested Gal4 drivers (neurons, epithelial,
   muscles, or fat bodies) leads to embryonic or early larval lethality
   ([158]41). Consistently, upon induction mediated by the larval
   epidermal driver A58 at 25°C (A58-Gal4 > UAS-mitoAflIII,
   UAS-mitoLigase; from now on Epi-ΔmtDNA), only a few larvae could
   survive to the pupal stage. To overcome the lethality, we maintained
   the crosses at 18°C. At this temperature, the Gal4 activity is at its
   lowest levels, leading to larval survival till the pupal stage
   ([159]42). At this temperature, we noticed the presence of restricted
   chromogenic foci but larvae could reach pupation, although only ~1% of
   pupae emerged ([160]Fig. 7B). Hence, this temperature and L3 larvae
   were selected for further experiments.

   Expression of mitoAfl III was confirmed to peak in the early L3 instar
   stage and decay in the mid-stage (L3-E versus L3-M; fig. S6A).
   Conventional PCR using oligos flanking mtDNA^Δ10.789-13.372 confirmed
   the presence of the mtDNA deletion in Epi-ΔmtDNA (fig. S6B).
   Accordingly, the CytB/ND5 ratio was reduced ([161]Fig. 7C) and the
   mtDNA copy number decreased, to a bigger extent when CytB, included in
   the mtDNA^Δ10.789-13.372 deletion, was used for mtDNA copy number
   quantification ([162]Fig. 7D and fig. S6C). Western blot analysis from
   total protein extracts showed equal levels of the mitochondrial complex
   II subunit succinate dehydrogenase B (SDHB) and depletion of complex V
   subunit ATPA ([163]Fig. 7, E and F). Given that mitochondrial complex
   II is encoded by nuclear genes and complex V by both nuclear and
   mitochondrial genomes, this result confirmed that mtDNA depletion was
   not related to reduced mitochondrial mass but due to a lower number of
   mtDNA copies. Thus, the coexpression of mitoAfl III and mitoT4lig in
   the epidermis of Drosophila larvae efficiently deletes mtDNA and
   increases mtDNA heteroplasmy.

VPS35 promotes mtDNA clearance in vivo

   To test whether Vps35 elevation supports mtDNA clearance in vivo, we
   overexpressed Vps35 in the larval epidermis by crossing the UAS-Vps35
   line with the A58-Gal4 driver (Epi-Vps35^OE). Epi-Vps35^OE larvae did
   not exhibit macroscopic abnormalities with no observable morphological
   or developmental defects ([164]Fig. 7B). qPCR quantification of total
   RNA extracts revealed a ninefold increase in Vps35 mRNA (Fig. S6D). In
   addition, Epi-Vps35^OE larvae had a moderate but significant increase
   in mtDNA copy number (fig. S6E). Next, we generated
   Epi-ΔmtDNA-Vps35^OE. In this background, Vps35 mRNA levels were
   increased fivefold on average compared to Epi-ΔmtDNA (fig. S6F). We did
   not observe major macroscopic and morphological abnormalities in larvae
   and the chromogenic foci initially observed in Epi-ΔmtDNA larvae
   disappeared ([165]Fig. 7B). Quantification of mtDNA copy number and
   CytB/ND5 ratio showed normalized mtDNA copies in Epi-ΔmtDNA-Vps35^OE
   compared to control larvae ([166]Fig. 7, G and H). Moreover, Western
   blot data confirmed higher levels of the complex V protein ATPA and no
   changes in the nuclear-encoded complex II protein SDHB (fig. S6, G and
   H).

   Last, we investigated changes in the mitochondria ultrastructure
   triggered by these genetic modulations. Electron microscopy preparation
   of the larvae epidermis showed cristae aberrations in the epidermis of
   Epi-ΔmtDNA animals ([167]Fig. 7I). In addition, we noticed the
   accumulation of lysosomes and multivesicular bodies intercalated with
   mitochondria ([168]Fig. 6I, blue pseudocolored). Epi-Vps35^OE in
   Epi-ΔmtDNA background showed a complete recovery of mitochondrial
   ultrastructure similar to control larvae.

Elevated Vps35 reverses transcriptome profile in the epidermis of ΔmtDNA
larvae

   For an unbiased and genome-wide overview of different pathways
   activated in ΔmtDNA epidermis, we conducted a transcriptome analysis in
   isolated epidermis from L3 ΔmtDNA larvae (fig. S7, A and B). As
   expected, due to the presence of the deletion, mtDNA^Δ10.789-13.372,
   mitochondrial mRNAs for CytB, ND1, and lrRNA, genes included in the
   deletion, were strongly reduced (fig. S7C). Then, we performed
   differential expression analysis using four independent bioinformatic
   methods (EdgeR-QFL, EdgeR-LRT, limma-voom, and DESeq2) and identified
   541 genes with significant up-regulation, whereas 311 genes were
   down-regulated ([169]Fig. 8A and data S3). Pathway analysis with DAVID
   (Database for Annotation, Visualization, and Integrated Discovery) and
   Gene Set Enrichment Analysis (GSEA) with Gene Ontology (GO) pathways
   uncovered up-regulation of proteolysis, innate immune response, and
   redox control routes ([170]Fig. 8, B and C; fig. S7, D and E; and data
   S3). To further validate these data, we selected some of the
   up-regulated targets and performed qPCR analysis in independent RNA
   isolates. We found consistent enrichment of the glutathione
   S-transferase genes GstD2, GstD5, GstD9, GstE5, GstE7, and Gst9, along
   with the mitochondrial proteases Spg7 and Afg3l2 ([171]Fig. 8D).
   Epi-Vps35^OE in the background of Epi-ΔmtDNA larvae (fig. S6I) reversed
   all mRNA levels for the selected targets to normal ([172]Fig. 8D).

Fig. 8. VPS35 expression restores mitochondrial defects associated with
ΔmtDNA.

   [173]Fig. 8.
   [174]Open in a new tab

   (A) Venn diagram for genes differentially expressed (FDR of 5%) between
   control and Epi-ΔmtDNA shared through four differential expression
   methods (Limma-voom, EdgeR-QLF, EdgeR_LRT, and DESeq2) (n = 4 per
   genotype). (B) DAVID pathway enrichment analysis for genes
   differentially expressed. (C) Heatmap for log FC showing gene
   identities for the top four categories of differentially enriched
   pathways. (D) mRNA quantification of selected genes in Epi-ΔmtDNA and
   Epi-ΔmtDNA-Vps35^OE larvae (n = 4 per genotype). Control refers to the
   A58-Gal4/+ genotype. P values were calculated using one-way ANOVA with
   Tukey correction for multiple comparisons. Data are presented as
   means ± SEM.

VPS35 elevation supports lysosomal function without affecting macroautophagy

   To understand the molecular mechanisms by which VPS35 restores the
   mitochondrial function, we turned once again to human cells. Given that
   the retromer participates in lysosomal homeostasis, we hypothesized
   that higher levels of VPS35 would improve the cellular degradation
   pathways. LysoTracker staining in cells expressing VPS35-V5 and
   TWNK^K319E-mCherry showed a higher number of acidic particles than in
   control cells (fig. S8, A and B). In line with this, in VPS35 KO cells,
   the number of acidic foci was down-regulated (fig. S8, A and B).

   Next, we wondered whether these effects on lysosomal pH related to
   VPS35 overexpression were also linked to changes in macroautophagy.
   Western blot analysis of autophagy markers showed increased levels of
   LC3B-II and p62 at the steady state in cells overexpressing VPS35-V5
   (fig. S8, C and D). Autophagy flux experiments in chloroquine-treated
   cells revealed no differences for LC3B-II accumulation, but
   VPS35-overexpressing cells showed a subtle increase in the adaptor
   protein p62 and VPS35 itself. At the same time, the lysosomal marker
   LAMP1 was highly enriched in VPS35 cells (fig. S8, E and F).

   Confocal microscopy showed an increased association of mitochondria
   with late endosomes LAMP1 in VPS35 overexpressing cells (fig. S8, G and
   H). We wondered whether this increased association translated into an
   increase in the macroautophagy of the complete mitochondria. Thus, we
   transduced our cells with mitoKeima, a fluorescence reporter with a
   pH-dependent fluorescence excitation profile. In the mitochondria,
   mitoKeima has an excitation peak at 448 nm, whereas in the lysosomes,
   due to the acidic pH, the excitation peak shifts to 561 nm. In both
   cases, the fluorescence emission remains at 610 nm. The ratiometric
   analysis of mitoKeima fluorescence emission, when excited sequentially
   at 586 nm/440 nm, showed a decrease in bulk mitophagy for cells
   overexpressing VPS35 (fig. S8I).

   All these data together confirm that VPS35 elevation does not affect
   the general autophagy pathways, including bulk mitophagy. However, it
   stimulates the lysosomal function upon stress, accelerating the
   piecemeal removal of mitochondrial components and facilitating recovery
   upon mtDNA damage.

DISCUSSION

   Mitochondrial quality control serves as a salvage pathway to eliminate
   ill-functioning mitochondria. Like other mitochondrial components, the
   mitochondrial genome is also prone to accumulating damage. However,
   persistent mtDNA replication stress does not trigger bulk mitochondrial
   removal but rather leads to the degradation of mitochondrial nucleoids
   in the endosomal compartment, serving as a selective degradation route
   for the mitochondrial genome ([175]13).

   The existence of alternative or cargo-specific mitophagy pathways
   highlights the complexity and adaptability of cellular systems.
   Dysfunctional mitochondrial pieces can be precisely eliminated by
   preserving the rest of the mitochondrial network. Different studies
   showed that these pathways share components with the vesicle
   trafficking system. For instance, in an oxidative environment, the
   retromer induces the formation of MDVs containing oxidized cargo and
   directed to peroxisomes ([176]43). Nevertheless, although the retromer
   and the mitochondrial genome appeared in cellular degradation
   compartments upon mtDNA stress ([177]13), it was not clear whether this
   pathway followed MDVs.

   To identify the specific mechanism involved in the mitochondrial genome
   turnover upon mtDNA replications stress, we used directed proximity
   proteomics between mitochondria and endosomes. Consistent with our
   previous findings, we detected an enrichment of vesicle and
   lysosomal-related proteins, including RAB10, ATP6V0D1, MYO1B, BIRC6,
   and the retromer component VPS29. All these proteins participate in
   lysosomal maturation and acidification ([178]34, [179]38,
   [180]44–[181]46), suggesting a prominent role of lysosomes in the
   elimination of the mitochondrial genome.

   Our data support a model where the events following mtDNA stress
   proceed through parallel pathways, strongly regulated by the vesicular
   system. mtDNA replication stress primarily leads to mitochondrial
   fragmentation, a process that may aid in the separation of
   dysfunctional mitochondrial components. Thus, the fission of the
   mitochondrial network is inhibited in cells expressing the dominant
   negative allele of RAB10^T23N and in retromer-deficient cells.
   Nevertheless, the role of these two candidates in adapting
   mitochondrial dynamics upon stress is not new. On the one hand, RAB10,
   traditionally associated with vesicle transport and endoplasmic
   reticulum morphology ([182]47), also controls mitochondrial turnover
   upon depolarization ([183]39). On the other hand, the retromer induces
   mitochondrial fragmentation in a DRP1/DLP1-dependent manner. Thus, the
   dominant negative mutation VPS35^D620N causes a VPS26-dependent
   mitochondrial fragmentation due to an enhanced turnover of
   mitochondrial DRP1 ([184]48). Consistently, VPS35 KO in HeLa cells
   suppressed the mitochondrial fragmentation caused by mtDNA replication
   stress. The fragmentation of the network was recovered upon
   reexpression of VPS35, independently of the accessory subunits VPS29 or
   VPS26.

   Furthermore, mtDNA replication stress triggers the formation of
   canonical MDVs containing the mitochondrial matrix component PDH. The
   formation of these vesicles also depends on the retromer and requires
   the mitochondrial fission protein DRP1 and transport protein Miro1,
   both critical for the formation of MDVs ([185]29, [186]49). Notably,
   these matrix-derived vesicles are distinct from the recently described
   VDIMs, which form independently of DRP1 ([187]11). Unexpectedly, the
   delivery of the mtDNA to the endosomes does not follow an MDV-related
   path. In contrast, mtDNA turnover depends on the retromer and, at
   least, on the Bcl-2–associated X protein BAX. Thus, the recruitment of
   retromer-vesicles to the mitochondrial surface might activate the
   formation of BAX-BAK1 macropores, enabling mitochondrial inner membrane
   herniation and extrusion of big mtDNA fragments ([188]50, [189]51),
   followed by vesicle capture. Then, the retromer and other
   vesicle-related proteins such as RAB10 may promote lysosomal maturation
   and the degradation of damaged mtDNA ([190]Fig. 9).

Fig. 9. Proposed model for retromer function upon mtDNA stress.

   [191]Fig. 9.
   [192]Open in a new tab

   The retromer enhances mitochondrial fragmentation and mtDNA turnover.
   mtDNA ejection occurs in a BAX-dependent manner, targeting
   RAB10-VPS35-positive lysosomes, and independent of MDVs. Stimulation of
   these pathways restores mitochondrial function and mitigates defects
   associated with mtDNA damage in vivo.

   Because the overexpression of the retromer core component VPS35
   increases mtDNA extraction in vitro, we hypothesized that this would
   also be beneficial to eliminate mitochondrial burden in vivo.
   Noteworthy, overexpression of VPS35 has been shown to restore cellular
   homeostasis associated with lysosomal dysfunction ([193]52, [194]53).
   Thus, the presence of ΔmtDNA changes the mitochondrial redox state,
   which is translated into activation of detoxifying pathways. These
   changes activate a specific transcriptome profile to counteract
   cellular damage but also to modify mitochondria structure. Thus,
   besides the up-regulation of genes for glutathione metabolism and
   innate immune response, we detected up-regulation of many transcripts
   encoding for membrane carriers, such as members from aquaporin (Drip),
   ABCB transporters (Mdr49 and Mrp4), ion channels (aralar1, MME1, sCaMC,
   and TrpA1), and mitochondrial proteases (Lon, Afg3L2, and Spg7). The
   last three ones are all involved in the degradation of damaged
   mitochondrial proteins to maintain mitochondrial integrity ([195]54).
   In this context, overexpression of Vps35 in Drosophila restores the
   mitochondrial defects associated with damaged mtDNA while also
   restoring the cellular transcriptome, thereby suggesting a recovery of
   cellular health ([196]Fig. 9).

   Together, our data underscore the intricate relationship between
   mitochondria and lysosomes and how these two organelles coordinate to
   maintain cellular health. Accordingly, mitochondrial impairment leads
   to an increase in lysosomal biogenesis, lysosomal acidification, and
   mitochondrial turnover ([197]55). Similarly, we found that, in cells,
   expression of VPS35 does not influence macroautophagy, but directly
   impacts on lysosomal acidification upon mtDNA stress burden.

   In conclusion, we have provided a mechanistic view of a complex
   mechanism. mtDNA replication stress follows a sophisticated downstream
   response that synergizes to prevent the accumulation of damage in the
   mtDNA and ensure mitochondrial integrity. First, we identified a
   prominent role of the vesicular system in controlling mitochondrial
   dynamics. Second, MDVs directed to the lysosomal compartment eliminate
   the mitochondrial matrix. Last, leaked mtDNA requires the formation of
   mitochondrial BAX membrane pores. However, the expelled DNA does not
   exist free in the cytoplasm and is captured in a recycling organelle.
   We demonstrate in vivo that the retromer can drive the elimination of
   mutated mtDNA, removing the source of cellular burden and thereby
   modulating the downstream response. Hence, our data confirm that
   modulation of VPS35 levels is an effective strategy to counteract
   mitochondrial damage associated with ΔmtDNA.

MATERIALS AND METHODS

Study design

   The primary goal of this study was to understand how the mitochondrial
   genome ends up in recycling organelles upon mtDNA replication stress.
   We first designed an approach to decipher the proximity proteome of
   mitochondria and endosomes. To interfere with mtDNA replication, we use
   the expression of a missense mutation of the mtDNA helicase, knowing
   that, in replicative cells, it leads to lysosomal-dependent mtDNA
   depletion ([198]13). Our results brought up a subset of candidate
   proteins involved in the regulation of this process, highlighting the
   lysosomes. Thus, we selected the retromer and evaluated its role in
   mitochondrial dynamics and the extraction mechanisms for mitochondrial
   matrix components. Last, we assayed in vivo the ability of the retromer
   to increase mitochondrial turnover, for which we used a Drosophila
   model with specific alteration in the mtDNA.

Plasmid generation

   RAB10 plasmids were obtained from Addgene (WT, Addgene #49472; Q68L
   Addgene #49544; and T23N Addgene #130885). RAB10^T23N was subcloned in
   pEGFP-C1 using Bgl II and Eco RI. VPS35-EGFP and VPS35-CFP were
   generated by subcloning VPS35 from pLenti/V5/DEST-VPS35 plasmid
   (Addgene #21691) in pEGFP-N1 and pECFP-N1 plasmid using Xho I and Bam
   HI sites. Twinkle Homo sapiens open reading frame (ORF) was subcloned
   from Twinkle-APEX2-V5 (Addgene #129705) in pN1-mCherry using Bgl II and
   Eco RI sites. VPS35 p.D620N and Twinkle p.K319E were generated by
   site-directed mutagenesis. VPS29-YFP (yellow fluorescent protein) was
   obtained from Addgene (#174520) and subcloned in pEGFP-C1 using Eco RI
   and Bam HI cloning sites. MitoKeima was amplified from
   mKeima-Red-Mito-7 (Addgene #56018) and subcloned in pLenti-CMV-Puro in
   EcoR V sites. All generated plasmids were verified by Sanger
   sequencing. To generate stable cell lines, VPS35 or VPS35 p.D620N ORFs
   were subcloned in pLenti/V5/DEST using EcoR V sites. mtDNA-Kaede
   plasmid was a gift from T. Shutt (University of Calgary, Canada).

Generation of HeLa cells KO cell lines

   All KO cell lines were generated by CRISPR-Cas9 and cloning of specific
   guide RNAs (gRNAs) in pSpCas9(BB)-2A-Puro (PX459) V2.0 (Addgene #62988)
   in the Bbs I site. All gRNAs were designed using
   [199]http://crispor.tefor.net/crispor.py. VPS35 KO was generated by
   targeting exon 13 (ATATGAATTCATGTCC); VPS26A was knocked out with a
   gRNA specific for exon 4 (GCACCGATGTAAGATTCATA) and VPS29, by targeting
   the isoform 1 (GGCAAACTGTTGCACCGGTG), isoform 2 (GGACATCAAGTTATTCCATG),
   and isoform 3 simultaneously (GACTATCTCAAGACTCTGGC). Miro1 (RHOT1) KO
   was generated by simultaneously targeting exon 10
   (GGGACTGTGCTTCGACGATT) and exon 12 (GTGGCCCCCAAGGTATGTAA). In all
   cases, transfected cells were selected using puromycin (2.5 μg/ml;
   Sigma-Aldrich, #P8833) for 72 hours after transfection, followed by
   isolation of monoclonal cell lines.

Cell culture and chemical treatments

   Cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM) with
   glucose (4.5 g/liter) + GlutaMAX (Gibco, #10566016) supplemented with
   10% fetal bovine serum (FBS), 1x Pen/Strep (Gibco, #15070063), and
   Plasmocin (2.5 μg/ml; InvivoGen, #ant-mpp). Transfection of HeLa cells
   was achieved using FuGENE HD (Promega, #E2311) following the
   manufacturer’s instructions and changing the growing medium 6 hours
   after transfection. The lysosomal activity was inhibited with 10 μM
   chloroquine (Sigma-Aldrich, #C6628) for 4 hours where indicated.
   Oxidative damage was induced by treating cells with 200 μM H[2]O[2] in
   a complete medium for 4 hours. For stimulating mitochondrial
   metabolism, cells were grown overnight in nonglucose DMEM supplemented
   with 10 mM galactose (Sigma-Aldrich, #1287700), 10% FBS, 1x Pen/Strep,
   1x GlutaMAX (Gibco, #35050061), and 1x pyruvate (Gibco, #11360070). For
   inhibiting mitochondrial membrane pore formation, 24 hours after
   transfection, cells were treated with 25 μM VBIT-12 (MedChemExpress,
   #HY-135885) or 100 μM Bax Inhibitor peptide V5 (MedChemExpress,
   #HY-P0081) alongside with chloroquine for 4 hours prior to fixation.

   For the generation of stable cell lines, HeLa cells were transduced
   with pLenti/V5/DEST-VPS35, pLenti/V5/DEST-VPS35 p.D620N, and
   pLenti-CMV-Puro-mitoKeima using psPAX2 and pMD2.G as helper plasmids
   and HEK293 cells for viral packaging. The supernatant containing
   lentivirus was filtered through a 0.45-μm filter and supplemented with
   polybrene (10 μg/ml; Sigma-Aldrich, #H9268). Transgenic expression of
   VPS35 was confirmed by Western blot. Transduced clones were selected
   and maintained in a medium containing blasticidin S (20 μg/ml;
   InvivoGen, #ant-bl-05) for VPS35 overexpression cell lines and
   puromycin (2.5 μg/ml) for mitokeima. HeLa DRP1 KO cells were kindly
   provided by M. Escobar (University of Cologne, Germany).

Drosophila stocks

   The Gal4/UAS system was used for tissue-specific expression ([200]56).
   All Drosophila experiments were carried out using A58-Gal4 driver lines
   to induce UAS-constructs in the epidermis of Drosophila larvae
   ([201]57). A58-Gal4 is expressed in the epidermis from the first larval
   instar stage onward ([202]42, [203]58, [204]59). UAS-constructs used in
   this work are as follows: UAS-mitoT4lig, UAS-mitoAflIII (BL 84979),
   UAS-mitoAflIII, UAS-mitoT4lig/CyO (kindly provided by B. A. Hay)
   ([205]41), and UAS-Vps35.HA ([206]60) (BL67152). All stocks were in a
   white-eyed genetic background and maintained at 25°C under a
   12-hour/12-hour light/dark cycle on standard fly food. All crosses were
   set up and maintained at 18°C under a 12-hour/12-hour light/dark cycle
   on standard fly food. The complete genotypes of all strains generated
   and used in this study are listed in table S1. All molecular
   experiments using Drosophila were conducted using complete larvae,
   excluding transcriptome analysis, where larval epidermis was isolated.

mtDNA copy number and qPCR

   Total DNA was isolated using the DNeasy Blood & Tissue Kit (Qiagen,
   #69504) according to the manufacturer’s instructions. Twenty nanograms
   of total DNA was used for the analysis of threshold amplification
   differences between mtDNA and nuclear DNA [delta C(t) method]. For
   human (HS) cell–based analysis, mtDNA copy number was obtained with
   oligos amplifying the mitochondrial gene ND1 and the nuclear gene APP.

   For cytosolic mtDNA detection, HeLa cells stably expressing VPS35 or
   VPS35^D620N were transfected with TWNK^K319E-mCherry. Forty-eight hours
   posttransfection, cells were collected and equally divided for internal
   control and fractionation experiments. The total DNA for internal
   control samples was obtained by solubilizing cell pellets in 10 mM
   NaOH, followed by boiling at 98°C for 30 min and neutralization of the
   pH with Tris (pH 8) to a final concentration of 150 mM. For
   mitochondria-free fractions, cell pellets were incubated in digitonin
   (25 μg/ml; Roth, #4005.2), 150 mM NaCl, and 50 mM Hepes (pH 7.4) for 10
   min on ice. Unbroken cells and mitochondria were cleared by sequential
   differential centrifugation first at 1000g and resultant supernatant at
   14,000g. DNA was further isolated using the DNeasy Blood & Tissue Kit
   and eluted in equal volumes. Three microliters of pure DNA was used to
   amplify the nuclear gene APP and mitochondrial genes ND1 and ND5 in
   both fractions. Cytoplasmic content of mtDNA was obtained after
   normalization of each Ct value from mitochondria-free fractions to
   normalized value of total sample Ct (Ct Mitochondrial^MITOC.Free) − (Ct
   Mitochondrial^TOTAL − Ct Nuclear^TOTAL) = DDCt. Fold change (FC) values
   were obtained with 2^−DDCt.

   For mtDNA heteroplasmy quantification and mtDNA copy number in
   Drosophila, 20 ng of total DNA isolated from five larvae was used to
   amplify ND5 and CytB, mitochondrial genes outside and inside the
   deleted region, respectively, and 3RTub (TUBB) as a nuclear gene. Both
   Ct values from mitochondrial genes were normalized with the Ct for TUBB
   and the ratio between CytB/ND5 was calculated. Quantitative real-time
   PCR for gene expression was performed using cDNA retrotranscribed from
   1 μg of total RNA using PowerUP SYBR green (Thermo Fisher Scientific,
   #A25780).

   All qPCR experiments were performed in Quant Studio 1 (Thermo Fisher
   Scientific) and PowerUP SYBR Green (Thermo Fisher Scientific, #A25780).
   Conventional PCR was performed using either cDNA or total DNA,
   specified in the corresponding figure. Primer sequences are specified
   in table S2.

Western blot and immunofluorescence

   For protein isolation, cells were pelleted at the indicated time points
   and resuspended in radioimmunoprecipitation assay (RIPA) buffer [150 mM
   NaCl, 1% Triton X-100, 0.1% SDS, 50 mM tris-HCl (pH 8), 1 mM EDTA, and
   0.5% Na-deoxycholate] containing protease inhibitor (Sigma-Aldrich,
   #11697498001). Protein concentration was quantified using Bradford
   reagent (Bio-Rad, #5000001), and equal amounts of protein were loaded
   onto SDS–polyacrylamide gel electrophoresis (PAGE), transferred to
   polyvinylidene difluoride membrane (Bio-Rad), blocked in TBST–5%
   fat-free milk, and blotted with the indicated antibodies. For Western
   blot involving Drosophila, five L3 larvae were homogenized in lysis
   buffer [150 mM NaCl, 1% Triton X-100, 50 mM tris-HCl (pH 8), 0.01%
   NP-40, and 1 mM EDTA] containing a protease inhibitor, using a Teflon
   pestle. Equal amounts of proteins were directly loaded onto SDS-PAGE
   and proceeded as described above.

   Antibodies used for Western blot were as follows: polyclonal α-SDHB
   (10620-1-AP; 1:1000), α-TOM20 (11802-1-AP; 1:1000), α-optineurin
   (10837-1-AP; 1:1000) α-LC3B (14600-1-AP; 1:1000), α-GFP (50430-2-AP;
   1:1000), α-VPS26A (12804-1-AP; 1:1000), α-Miro1 (#84055-5-RR; 1:1000),
   and monoclonal α-GAPDH (60004-1-Ig; 1:4000), α-GFP (66002-1-Ig;
   1:1000), and α-p62 (66184-1-Ig; 1:1000) from Proteintech; polyclonal
   α-RAB5 (ab109534; 1:1000), α-VPS29 (ab236796; 1:1000), and monoclonal
   α-ATP5A (ab14748; 1:1000) from Abcam; polyclonal α-LAMP1 (#9091;
   1:1000), α-V5 (#13202; 1:1000), and monoclonal α-V5 (#80076; 1:1000)
   from Cell Signaling; monoclonal α-VPS35 (sc-374372; 1:1000) and α-RAB5
   (sc-46692; 1:500) from Santa Cruz; monoclonal α-DLP1(DRP1) from BD
   Bioscience (#611112; 1:500). All membranes were incubated with
   appropriate secondary antibodies combined with horseradish peroxidase
   (HRP) (Jackson ImmunoResearch; goat anti-mouse #115-035-003; goat
   anti-rabbit #111-035-003), developed using Western Lighting Plus ECL
   (PerkinElmer, #NEL104001EA), and visualized using LAS500 CCD camera (GE
   Healthcare).

   For immunofluorescence, cells were seeded on glass coverslips, fixed
   for 15 min in 4% paraformaldehyde/phosphate-buffered saline (PBS), and
   stored at 4°C until use. Permeabilization was achieved using 0.2%
   Triton/PBS followed by 1-hour blocking (5% fat-free milk, 10% FBS, 0.1%
   Triton, and 0.5% BSA in PBS). Primary antibodies were incubated
   overnight at 4°C. Antibodies used for immunofluorescence were as
   follows: polyclonal α-TOM20 (11802-1-AP; 1:1000) and α-VPS26A
   (12804-1-AP; 1:1000) from Proteintech; monoclonal α-dsDNA (ab27156;
   1:1000), α-PDHX (ab110333; 1:500), α-ATP5A (ab14748; 1:500), and
   polyclonal α-RAB5 (1:500) from Abcam; polyclonal α-LAMP1 (#9091;
   1:500), α-V5 (#13202; 1:500), and monoclonal α-V5 (#80076; 1:500) from
   Cell Signaling; monoclonal α-VPS35 (sc-374372; 1:500) from Santa Cruz.
   After washing the coverslips with PBS, samples were incubated with
   appropriate secondary antibodies (1:1000, 1 hour at room temperature)
   from Thermo Fisher Scientific: goat α-rabbit combined with Alexa-405
   (#A-31556), goat α-rabbit Alexa-488 (#A-11008), goat α-mouse Alexa-488
   (#A-11001), goat α-rabbit Alexa-546 (#A-11035), goat α-mouse Alexa-546
   (#A-11030), goat α-rabbit Alexa-647 (#A-21245), and goat α-mouse
   Alexa-647 (#A-21235) Coverslips were mounted using Fluoromount
   containing 4′,6-diamidino-2-phenylindole (DAPI) (Thermo Fisher
   Scientific, #00-4959-52).

Confocal microscopy and flow cytometry

   For confocal microscopy, images were acquired using a Leica SP8
   microscope with a 63x/1.40 oil PL Apo objective and a PerkinElmer
   Spinning-disk confocal microscope UltraVIEW VoX with 60x objective. All
   images were deconvoluted using ImageJ (NIH) before analysis to increase
   resolution. Brightness was adjusted equally in the entire image. All
   image analyses were performed using ImageJ (NIH). Quantification of
   mitochondrial morphology was achieved using as particle analysis tool
   with nonexclusion criteria. Mitochondrial vesicle quantification was
   performed using an internal IMAGE J plugin by counting TOM20−/PDH+ or
   TOM20−/dsDNA^+ particles. Manders’ correlation coefficient was
   calculated using the JACOP plugin ([207]61). The fluorescence profile
   was obtained by drawing a line covering 50 pixels and using the RGB
   Profiler plugin.

   For LysoTracker experiments, cells were seeded onto μ-Dish 35 mm, high
   Glass Bottom (Ibidi, #81158) and transfected with TWNK^K319E-mCherry
   plasmid. Twenty-four hours after transfection, cells were loaded with
   50 nM LysoTracker Green (Thermo Fisher Scientific, #L7526) for 30 min,
   following a medium change and immediately imaged using the spinning
   disk at 37°C. Only cells transfected with TWNK^K319E-mCherry were
   considered for the analysis.

   Analysis of bulk mitophagy was performed in cells stably expressing the
   mitophagy reporter Mitokeima. Flow cytometry analysis was performed on
   a CytoFLEX-S (V4-B2-YG4-R3), Beckman-Coulter (Acquisition Software
   CytExpert V2.4.0.28). After setting the cell gate (FSC-A/SSC-A), FSC-H
   and FSH-W were used for doublet discrimination. The signals from the
   mt-Keima construct were measured in the Violet 610 (excitation laser
   405 nm; 610/20 bandpass filter) and the ECD (excitation laser 561;
   610/20 bandpass filter) detector channels.

Electron microscopy

   The epidermis from Drosophila L3 larvae was isolated and immersion
   fixed in 2% formaldehyde, 2% glutaraldehyde, and 3 mM CaCl[2] in 0.1 M
   sodium cacodylate buffer for 48 hours. Samples were washed with 0.1 M
   sodium cacodylate buffer and postfixed with 2% OsO[4] in 0.1 M
   cacodylate buffer, washed 3x5 min with ddH[2]O, followed by dehydration
   in an ascending ethanol series for 15 min each (50, 70, 90, and 3x100%)
   at 4°C. Samples were incubated for 15 min with 50% ethanol/propylene
   oxide, followed by two times 15 min pure propylene oxide. Then, samples
   were infiltrated with a mixture of 50% Epon/propylene oxide and 75%
   Epon/propylene oxide for 2 hours each and pure Epon overnight at 4°C.
   The next day, samples were incubated with fresh Epon for 2 hours and
   placed into flat embedding molds and cured at 60°C.

   Ultrathin sections of 70 nm were cut using an ultramicrotome (Leica
   Microsystems, UC6) and a diamond knife (Diatome, Biel, Switzerland) and
   stained with 1.5% uranyl acetate and 3% Reynolds lead citrate solution.
   Images were acquired using a JEM-2100 Plus Transmission Electron
   Microscope (JEOL) operating at 80 kV equipped with a OneView 4K camera
   (Gatan).

   For CLEM, cells were seeded onto carbon-coated plates containing a
   numerated grid. Cells were transfected with TWNK^K319E-mCherry and
   VPS35-CFP plasmids as described previously. The lysosomal function was
   blocked for 4 hours with 10 μM chloroquine. One hour before fixation,
   cells were loaded with 1x SYBRGold (Thermo Fisher Scientific, #S11494)
   and 250 nM PK mito Deep Red (Spirochrome, #SC055). Cells were fixed for
   30 min at room temperature and 30 min at 4°C in 2% glutaraldehyde, 2.5%
   sucrose, and 100 mM CaCl[2] in 0.1 M Hepes (pH 7.4) and washed with 0.1
   M Hepes buffer. Transfected cells showing cytoplasmic DNA were scanned
   using the SP8 confocal microscope (Leica) with 63x/1.40 oil objective,
   and bright-field images were used to localize the coordinates of cells
   of interest. Following light imaging, cells were incubated with 1%
   Osmiumtetroxid for 30 min at 4°C, washed in 0.1 M cacodylate buffer,
   and dehydrated using ascending ethanol series (50, 70, 90, and 100%) at
   4°C. Cells were infiltrated with a mixture of 50% Epon/ethanol for 1
   hour, 66% Epon/ethanol for 2 hours, and pure Epon overnight at 4°C.
   TAAB capsules filled with Epon were placed upside down onto the glass
   bottom and cured for 48 hours at 60°C. Glass bottom was removed by
   alternatingly putting the dish into boiling water and liquid nitrogen.
   Block face was trimmed to the previously noted coordinates using a 90°
   trim tool (Diatome, Biel, Switzerland), and ultrathin sections of 70 nm
   were cut using an ultramicrotome (Leica Microsystems, UC6) and a
   diamond knife (Diatome, Biel, Switzerland) and stained with 1.5% uranyl
   acetate for 15 min at 37°C and lead citrate solution for 4 min.

   For electron tomography, ultrathin sections of 300 nm were cut and
   incubated with 10 nm of protein A gold (CMC, Utrecht) diluted 1:60 in
   ddH[2]O. Sections were stained with 2% uranyl acetate for 20 min and
   Reynolds lead citrate solution for 3 min. Images for the tilt series
   were acquired using Serial EM with a pixel size of 1.108 nm from −60°
   to 60° with a 1° increment on the same JEM-2100 Plus Transmission
   Electron Microscope (JEOL) operating at 200 kV. Reconstruction was done
   using Imod, MiB, and Imaris.

Proximity biotinylation

   Proximity biotinylation was performed in transduced cells expressing
   Split TurboID constructs RAB5C-SplitNt-V5 and SAMM50-SplitCt-HA. Cells
   expressing only SAMM50-SplitCt-HA were used as a negative control.

   For protein isolation, HEK293 cells were incubated for 4 hours with 250
   μM biotin. For inducing mtDNA damage, cells were previously transfected
   with TWNK^K319E-mCherry 24 hours before the biotin incubation. Pellets
   were scraped from plates right after the biotin incubation and
   solubilized in RIPA buffer. Three hundred micrograms of total protein
   extracts were purified using Streptavidin Sepharose High-Performance
   beads (GE Healthcare Bio-Sciences AB). For Western blot analysis, beads
   were washed three times with RIPA buffer, resuspended in 2x Laemmli
   buffer, boiled at 95°C for 5 min, and loaded onto SDS-PAGE.
   Biotinylated proteins were visualized with α-Streptavidin–HRP antibody
   (Jackson ImmunoResearch, 1:2000, #016-030-084). For mass spectrometry,
   beads were washed three times with ABC buffer [50 mM ammonium
   bicarbonate (pH 7.8)] and incubated for 10 min with 50 μl of urea
   buffer (8 M urea), 1 hour with 5 mM dithiothreitol, and 30 min with 40
   mM indole-3-acetic acid at room temperature and in the dark. Protein
   digestion was achieved by incubating with Lys-C in a ratio of 1:100,
   for 2 to 3 hours. Samples were diluted with ABC buffer to a final urea
   concentration of 2 M and incubated with trypsin overnight in a ratio of
   1:100. Digestion was stopped with 1% formic acid and peptides loaded
   into preequilibrated stage tips and stored at 4°C until used for mass
   spectrometry analysis.

Liquid chromatography–mass spectrometry analysis

   Affinity-enriched proteomics samples were analyzed in positive mode
   using data-dependent acquisition either by an Easy-nLC 1000–Q Exactive
   Plus or an Easy-nLC 1200–Orbitrap Eclipse tribrid system (Thermo Fisher
   Scientific). Online chromatography was directly coupled to the mass
   spectrometry systems using a nanoelectrospray ionization source.
   Peptides were separated by reversed-phase chromatography with a binary
   buffer system of buffer A (0.1% formic acid in water) and buffer B
   (0.1% formic acid in 80% acetonitrile) using a 60-min chromatographic
   gradient. Separation was performed on a 50-cm-long in-house packed
   analytical column filled with 1.9 μM C18-AQ Reprosil Pur beads (Dr.
   Maisch). Using the 60-min chromatographic gradient, peptide separation
   based on their hydrophobicity was performed by linearly increasing the
   amount of buffer B from initial 13 to 48% over 35 min followed by an
   increase in B to 95% for 10 min. The column was washed for 5 min, and
   initial column conditions were achieved by equilibrating the column for
   10 min at 7% B. Full mass spectrometry spectra [300 to 1750 mass/charge
   ratio (m/z)] were acquired with a resolution of 70,000, a maximum
   injection time of 20 ms, and an automatic gain control (AGC) target of
   3 × 10^6. The top 10 most abundant peptide ions were isolated (1.8 m/z
   isolation windows) for subsequent HCD fragmentation (NCE = 28), and
   tandem mass spectrometry (MS/MS) recorded at a resolution of 35,000, a
   maximum injection time of 120 ms, and an AGC target of 5 × 10^5.
   Peptide ions selected for fragmentation were dynamically excluded for
   20 s.

Data processing and analysis for proteomics

   All recorded RAW files were processed with the MaxQuant software suite
   75 (1.5.3.8 for IP data, 1.6.14 for pSILAC). For peptide identification
   and scoring, MS/MS spectra were matched against the mouse UniProt
   database (downloaded 15 August 2019) using the Andromeda search
   algorithm 76. For the affinity-enriched samples, multiplicity was set
   to one and trypsin/P was selected as digestive enzyme.
   Carbamidomethylation was set as a fixed modification, and methionine
   oxidation or N-terminal acetylation was selected as a variable
   modification. Peptides were identified with a minimum amino acid length
   of seven and a false discovery rate (FDR) cutoff of 1% on the peptide
   level. Proteins were identified with FDR < 1% using unique and razor
   peptides for quantification. Label-free quantification (LFQ) was
   performed using the standard settings of the maxLFQ algorithm. Match
   between runs was activated.

   Statistical analysis and visualization were done with the Perseus
   (1.6.5) and InstantClue software suites 77 and 78. LFQ intensities of
   immunoprecipitation samples were log[2] transformed and filtered for
   proteins identified in at least two replicates of one condition.
   Missing values were imputed with the Perseus plugin ImputeLCMD using a
   deterministic minimal value approach (MinDet, q value = 0.001) to
   simulate the lower detection limits of the mass spectrometer. To
   evaluate the principal components responsible for the variances between
   samples, we performed a principal components analysis. Furthermore, we
   performed a two-sided Student’s t test to identify significantly
   regulated proteins (S0 = 0.1, permutation-based FDR = 0.05, 500
   randomizations). If not otherwise indicated, significantly enriched
   proteins were determined by a combination of FDR-corrected P values and
   log[2] protein FCs (q value < 0.05 and absolute log[2] FC > 1). Pathway
   analysis was performed using Metascape ([208]62).

Transcriptomic analysis

   For RNA next-generation sequencing, L3 larval epidermis from five
   larvae was first isolated and pooled. Total RNA was extracted as
   explained previously, and 2 μg of RNA was sequenced using Poly
   A + selection. A total of 25 million to30 million reads per sample were
   obtained using paired-end 100-bp read length (Illumina NovaSeq 6000). A
   quality control analysis of the RNA-seq raw data was performed with
   FastQC v0.12.0 (Babraham Bioinformatics) and MultiQC v1.18 software
   ([209]63). 3′ poly(A) tail and the 3′ adapter from Illumina were
   trimmed with the TrimGalore 0.06.10 tool (Babraham Bioinformatics) to
   reach a PhredScore greater than 35 and adapter content lower than 1%.
   Trimmed reads were then mapped onto the Drosophila melanogaster genome
   (Drosophila BDGPG.32, ENSEMBL) by using the RNA-seq aligner STAR
   2.7.10b ([210]64), followed by a quality assessment of the alignment
   with Qualimap v2.2.1 ([211]65). To estimate gene expression levels,
   mapped reads for all transcript variants of a gene (gene counts data)
   were counted with RSEM v1.3.3 software.

   Quality control, trimming, mapping, and counting steps were performed
   in an Ubuntu 22.04.2 LTS environment. Further analysis was run through
   the RStudio 2023.03.0 +386 (R version 4.2.2) software. Thus, gene
   counts were annotated with the AnnotationHub Bioconductor package, and
   four differential expression analyses were performed both by applying
   the negative binomial distribution with EdgeR-LRT (likelihood-ratio
   test), EdgeR-QL (quasi-likelihood) ([212]66), and DESeq2 ([213]67)
   bioconductor packages or by applying empirical Bayes model using
   Limma-Voom analysis ([214]68, [215]69). Genes with a common
   differential expression among the four methods (cpm < 0.5; P < 0.05)
   were selected for the GSEA with GO pathways by using the
   clusterProfiler Bioconductor package ([216]70) and DAVID Knowledgebase
   (NIH).

Statistical analysis

   Statistical analysis was conducted using GraphPad Prism version 10
   (GraphPad Software Inc., San Diego, CA, USA). The significance of
   differences among the two or more samples was calculated using unpaired
   and parametric Student’s t test or one-way analysis of variance (ANOVA)
   test, according to the experimental design and data distribution. The
   precise number of samples analyzed in each experiment and the
   respective statistical test used for analysis are reported in figure
   captions. Statistical significance is determined by P values, which are
   shown in each figure.

Acknowledgments