During DIOM remodeling, autophagy occurs during atrophy, which lasts from 12 hr to 2 d APF (after puparium formation). Then, organelle and myofibril reformation accompanies hypertrophy from 2 to 4 d APF (Figure 1A). We previously reported the dynamics of several organelles in DIOM remodeling (Fujita et al., 2017; Murakawa et al., 2020); however, the overall mechanism is still enigmatic. To gain molecular insights into DIOM remodeling, we performed time-course RNA-seq of isolated DIOMs from four animals at five time points (Figure 1A). Dissected animals were fixed with methanol to avoid cell death induction (Wang et al., 2021), and six DIOMs were isolated from each animal to prep total RNA (Figure 1B). The mRNAs were amplified, and the cDNA libraries were prepared using CEL-Seq2, a sensitive protocol for single-cell RNA-seq (Hashimshony et al., 2016). Then, they are subjected to next-generation sequencing. As a result, ~8000 genes were counted in the DIOMs above background (Supplementary file 1, ctrl). High similarities among four replicates at each time point show the reproducibility of our protocol (Figure 1—figure supplement 1A). We performed DESeq2 principal component analysis (PCA) of all samples to reduce the dimensionality of the data (Figure 1C, Figure 1—figure supplement 1B). In the PC1–PC2 plane, which explains 43% of expressed genes in DIOMs, the transcriptome changed dynamically from 3IL to 4 d APF (Figure 1C). Of note, there was a substantial difference between before (3IL) and after (4 d APF) remodeling in both PC1–2 and PC3–4 (Figure 1C), suggesting the functional importance of remodeling larval muscle to adult abdominal muscle during this epoch.

To identify genes with similar temporal expression profiles, we used fuzzy c-means to categorize normalized read counts into 12 clusters (Figure 1—figure supplement 1C), each with 379–937 genes (Figure 1—figure supplement 1D). The 12 clusters represent unique temporal expression dynamics, with each time point showing a distinct profile of clusters with high expression (Figure 1D, Figure 1—figure supplement 1D). We also noticed that a time-dependent shift in gene ontology (GO) terms is apparent among the clusters, reflecting the sequential events during DIOM remodeling (Figure 1F). The principal component coefficients (loadings) of individual clusters to PC1–4 are shown in Figure 1E. The direction and length of the arrows indicate the contribution of each cluster to principal components (PC1–4). Briefly, upon induction of muscle atrophy (1–2 d APF), the ubiquitin–proteasome system, developmental signaling, autophagy, and histone modification-related genes were upregulated. Then, chromatin remodeling-related genes were elevated around 2 d APF. Myofibril, mitochondrion, and ribosome biogenesis-related genes required for muscle hypertrophy were upregulated around 3 d APF (Figure 1F). The data illustrates the sequential events involved in DIOM remodeling, aligning with previously observed morphological changes (Fujita et al., 2017; Murakawa et al., 2020).

Autophagy deficiency impacts both atrophy and myofibril regeneration during hypertrophy (Fujita et al., 2017; Murakawa et al., 2020; Figure 2A). Given that autophagy is known to influence gene expression through both direct and indirect mechanisms (Morishita and Mizushima, 2019; Sánchez-Martín et al., 2019), we hypothesized that the loss of autophagy also alters gene expression in DIOMs. To investigate this possibility, we performed a comparative time-course RNA-seq of DIOM remodeling in Atg18a RNAi, FIP200 RNAi, or Stx17 RNAi DIOMs and compared the results with control data (Figure 2B). Atg18a and FIP200 are essential factors for autophagosome formation, and Stx17 is an autophagosomal SNARE protein required for autophagosome–lysosome fusion (Lőrincz and Juhász, 2020; Nakatogawa, 2020). Thus, their knockdown blocks autophagy at early or late stages, respectively (Figure 2B, bottom). Similar to the experiment in Figure 1, we performed RNA-seq at five time points from 3IL to 4 d APF (Figure 2B) and confirmed high similarities among the four replicates in all 20 conditions (Figure 2—figure supplement 1A and Supplementary file 1). We confirmed that RNAi efficiently knocked down targeted genes (Figure 2—figure supplement 1B–D). In contrast to our prediction, the knockdown of Atg18a, FIP200, or Stx17 only had a slight impact on transcriptomic dynamics in DIOM remodeling (Figure 2C), with only minor changes detected (Figure 2—figure supplement 2G).

Next, to examine whether loss of autophagy affects protein synthesis in DIOMs, we used the quantification of DIOM volume changes as a surrogate measurement and compared controls with FIP200 RNAi from 3IL to 4 d APF (Figure 2D, E). As expected, muscle atrophy observed until 2 d APF was affected by FIP200 RNAi (Figure 2A, D). Notably, DIOMs in the FIP200 RNAi condition grew from 2 to 3 d APF at levels comparable to the control (Figure 2E), indicating that protein synthesis occurs independently of autophagy. Consistent with this observation, transcription of both ribosomal RNAs (rRNAs) and proteins, established indicators of the target of rapamycin complex 1 (TORC1) activity (Saxton and Sabatini, 2017), did not drop but instead was slightly elevated in conditions lacking autophagy (Figure 2—figure supplement 2). The synthesis of rRNA was upregulated in ATG RNAi between 2 and 3 d APF (Figure 2—figure supplement 2A). It was confirmed that the number of nuclei remained unchanged by the loss of autophagy (Figure 2—figure supplement 2B, C). The size of the nucleolus, which positively correlates with TORC1 activity (Grewal et al., 2007), was increased by FIP200 and Stx17 RNAi (Figure 2—figure supplement 2D–F). Moreover, most ribosomal protein large subunit (RpLs) and ribosomal protein small subunit (RpSs) mRNA levels were higher in ATG RNAi conditions (Figure 2—figure supplement 2G). These results suggest that TORC1 activity and protein synthesis capacity are comparable between control and autophagy-deficient DIOMs. Altogether, we conclude that gene expression occurs independently of autophagy in DIOMs.

The above results suggest that the autophagic degradation of cytosolic components in itself is critical for DIOM remodeling. We have previously reported that FIP200 RNAi or Stx17 RNAi induce the accumulation of mitochondria or mitophagosomes, respectively (Figure 3A, B; Fujita et al., 2017). In addition, mitophagosomes were often observed in wild-type DIOMs at 1 d APF, when autophagy is strongly induced (Figure 3C; Fujita et al., 2017). These results suggest that mitochondria are major autophagy cargo, and impaired mitophagy contributes to the severe loss of autophagy phenotype.

We tried to identify relevant mitophagy receptors directly via our RNA-seq approach. The time-course RNA-seq data (Figures 1 and 2) indicated that, among the known mitophagy regulators, only BNIP3 was robustly expressed in 1 d APF DIOMs. In contrast, Zonda, CG12511, Pink1, Park, Key, Ref(2)P, and IKKε—the Drosophila orthologs of FKBP8, FUNDC1, PINK1, Parkin, Optineurin, p62, and TBK1, respectively—showed little or undetectable expression at this stage (Figure 3D). BNIP3 is the sole fly ortholog of mammalian NIX and BNIP3. BNIP3 mRNA levels were relatively high at 3IL and 4 d APF when mitophagy was minimally induced. We noticed that FBXL4 is expressed in 3IL and 4 d APF but not in 1 d APF DIOM (Figure 3E). BNIP3 is known to undergo ubiquitination and degradation through a mechanism mediated by FBXL4, a ubiquitin E3 ligase (Niemi and Friedman, 2024). Previous data suggest that transcriptional and post-translational mechanisms regulate BNIP3 protein levels. Consistent with expression levels, BNIP3 RNAi led to mitochondrial accumulation in 4 d APF DIOMs, whereas RNAi targeting other known mitophagy regulators did not (Figure 3—figure supplement 1). We generated BNIP3 knockout (KO) flies lacking all exons using two guide RNAs (gRNAs) to characterize BNIP3 further (Figure 3F). BNIP3 KO was compatible with viability and did not markedly impair mobility (Figure 3—figure supplement 2A–C). Similar to RNAi, mitochondria accumulated significantly in BNIP3 KO DIOMs at 4 d APF (Figure 3G).

At the ultrastructural level, BNIP3 KO induced the accumulation of mitochondria and loss of myofibrils (Figure 4A–C), consistent with our confocal data (Figure 3G). To test the role of BNIP3 in mitophagosome formation more directly, we performed transmission electron microscopy (TEM) in the absence of Stx17 to block the autophagosome–lysosome fusion. As expected, mitophagosomes (pink-colored) were frequently observed in the Stx17 RNAi condition (Figure 4D, D’, F). In stark contrast, mitophagosomes were rare, and unengulfed mitochondria (green-colored) accumulated in the combined condition of Stx17 RNAi and BNIP3 KO (Figure 4E, E’, F). We also confirmed there was no significant difference in the number of autophagosomes (blue-colored) between the two conditions (Figure 4G), indicating that BNIP3 is dispensable for autophagosome formation. These data show that BNIP3 is critical for mitophagosome formation in DIOMs.

Drosophila BNIP3 harbors LIR, MER, and TMD domains, similar to mammalian NIX and BNIP3 (Figure 5A). Although the function of the MER domain has been unknown since it was identified (Zhang et al., 2012), recent studies have shown that the MER of mammalian NIX and BNIP3 directly interacts with WIPIs, the mammalian orthologs of Atg18 (Adriaenssens et al., 2024; Bunker et al., 2023). AlphaFold 3 (Abramson et al., 2024) predicted that this interaction is evolutionarily conserved in Drosophila in a manner analogous to their mammalian counterparts (Figure 5B; Adriaenssens et al., 2024; Bunker et al., 2023). In this prediction, residues 31–57 of Drosophila BNIP3 (slate blue and yellow) interact with the surface formed by blades 2 (green) and 3 (pink) of the β-propeller structure of Drosophila Atg18a. Notably, residues 42–53 (yellow), which correspond to the MER designated for NIX (Bunker et al., 2023; Zhang et al., 2012), fold into an α-helix, docking into the groove formed by blades 2 and 3 and forming substantial hydrophobic interactions as well as defined polar contacts. We confirmed that HA-tagged Drosophila Atg18a co-immunoprecipitated with GFP-tagged full-length Drosophila BNIP3, and that this interaction was attenuated by the deletion of the MER (residues 42–53) (Figure 5C).

Next, to verify the importance of the LIR and MER motifs in BNIP3, we expressed a series of BNIP3 mutants in BNIP3 KO flies (Figure 5A). ΔLIR (W16A/L19A) lacks critical consensus residues in LIR due to double mutations (Johansen and Lamark, 2020). The MER mutant (MER^mut, L49A) substituted residue L49 with an A, which is predicted to be deeply buried at the interface with Atg18a (Figure 5C). The corresponding L75A mutant in human NIX has been shown to lose its affinity for WIPI2/Atg18 (Bunker et al., 2023). The ΔMER mutant lacks an entire short α-helix (G42 to Q53) of the MER motif (Figure 5C, shown yellow-orange). The ΔLIR + ΔMER is a combined construct of ΔLIR and ΔMER. All BNIP3 constructs retain an intact TMD at the C-terminus, ensuring their localization to the outer mitochondrial membrane (Figure 5—figure supplement 1A).

As shown in Figure 5D, the expression of the full-length (Full) construct almost completely suppressed the accumulation of mitochondria in BNIP3 KO at 4 d APF, confirming the rescue of BNIP3 function using our construct-based approach. To our surprise, the BNIP3 ΔLIR construct rescued the BNIP3 KO phenotype comparable to Full, indicating LIR on its own may be redundant for mitophagy in DIOMs. In contrast, mitochondria accumulated in MER^mut and ΔMER reconstituted DIOMs. Furthermore, the combination of ΔLIR and ΔMER resulted in mitochondria accumulation nearly identical to that observed in BNIP3 KO flies (Figure 5D, D’). Consistent with these findings, Atg101-dependent autophagic degradation of GFP-BNIP3 was strongly suppressed by the ΔLIR + ΔMER mutations (Figure 5E). Atg101 is an essential subunit of the Atg1/ULK1 kinase complex (Guo et al., 2019; Hegedűs et al., 2014; Noda and Mizushima, 2016). From these results, we conclude that both MER and LIR motifs contribute to BNIP3-mediated mitophagy, with some redundancy in their contributions to selectivity.

The data above suggest that a BNIP3-mediated mechanism degrades larval muscle-derived mitochondria; however, in previous figures, we only observed the terminal phenotype of BNIP3 KO at 4 d APF. To observe the changes in the number of mitochondria and cell shape during DIOM remodeling, we first performed a time-course experiment in control and BNIP3 KO conditions (Figure 6A, Figure 6—figure supplement 1A, B). There was no significant difference in the number of mitochondria at 3IL; however, mitochondria started accumulating at 1 d APF (Figure 6A, A’), suggesting that BNIP3-mediated mitophagy degrades larval muscle mitochondria. Further, we tested the Mito-QC probe (Lee et al., 2018), a tandem GFP-mCherry fusion protein targeted to the OMM, to compare mitophagy flux at 1d APF with and without BNIP3. When the probe is delivered to lysosomes via autophagy, the GFP is quenched due to the acidic environment of the lysosome. Robust mitophagy activity was detected in control DIOMs at 1 d APF. In sharp contrast, BNIP3 KO blocked mitophagy flux at a comparable level with FIP200 RNAi (Figure 6B, B’).

To show the degradation of larval muscle-derived mitochondria more directly, we labeled larval muscle mitochondria specifically by using the temperature-sensitive mutant of GAL80 (GAL80^ts), a repressor of GAL4 (McGuire et al., 2004). At the restrictive temperature (29°C), the GAL80^ts mutant is inactive, allowing GAL4 to drive the expression of the UAS-fused construct, UAS-Mito-GFP. Conversely, at the permissive temperature (18°C), the GAL80^ts is active, suppressing GAL4 activity. Using this system, Mito-GFP was expressed in the larval muscle and subsequently suppressed throughout the entire pupal stage (Figure 6C). Anti-ATP5A antibodies were used to visualize total mitochondria. The amount of Mito-GFP-positive mitochondria was comparable in control and BNIP3 KO at 3IL muscle. However, at 4 d APF, the Mito-GFP signal was nearly absent in the control, while a strong Mito-GFP signal was observed in the BNIP3 KO (Figure 6D, D’), indicating that BNIP3-mediated mitophagy degrades larval muscle mitochondria during muscle remodeling.

The following antibodies were used: Mouse monoclonal anti-ATP5A (1:300, ab14748, Abcam, Cambridge, UK), Rat monoclonal anti-HA (1:1000; 11867423001, Roche, Basel, Switzerland), Rabbit polyclonal anti-GFP (1:2000; 598, MBL, Nagoya, Japan), Rabbit polyclonal anti-Fibrillarin (1:300; ab5821, Abcam, Cambridge, UK), anti-rabbit IgG Alexa Fluor 594 conjugate (1:400; A11012, Thermo Fisher Scientific, Waltham, USA), anti-mouse IgG Alexa Fluor 488 conjugate (1:400; A11001, Thermo Fisher Scientific, Waltham, USA), anti-mouse IgG Alexa Fluor 594 conjugate (1:400; A11005, Thermo Fisher Scientific, Waltham, USA), HRP-conjugated AffiniPure Goat Anti-Rabbit IgG (1:10,000; 111-035-144, Jackson ImmunoResearch, West Grobe, USA), HRP-conjugated AffiniPure Donkey Anti-Rat IgG (1:5000; 712-005-153, Jackson ImmunoResearch, West Grobe, USA), Bisbenzimide H33342 Fluorochrome Trihydrochloride DMSO Solution (1:10,000; 04915-81, Nacalai Tesque, Kyoto, Japan), Alexa Fluor 633 Phalloidin (1:200; A22284, Thermo Fisher Scientific, Waltham, USA), and GFP-Trap Agarose (gta, Chromotek, Planegg-Martinsried, Germany).

Flies were reared under standard conditions at 25°C unless otherwise stated. The w¹¹¹⁸ strain was used as a control for knockout strains. For muscle-targeted gene expression, Mef2-GAL4 was used. UAS-LacZ was used as a control for UAS transgenes. All genetic combinations were generated by standard crosses. Detailed genotypes are described in Supplementary file 3. Genotypes of flies used in this study include the following: (1) y¹ w*; P{w^+mC = GAL4-Mef2.R}3 (Bloomington Drosophila Stock Center, BL 27390; DMef2-GAL4), (2) w¹¹¹⁸; P{w^+mC = UAS-lacZ.B}Bg4-1-2 (BL 1776; UAS-LacZ), (3) w^*; P{w^+mC = UAS-Rheb.Pa}2 (BL 9688), (4) w¹¹¹⁸; P{w^+mC = UAS-mito-HA-GFP.AP}2/CyO (BL 8442; Mito-GFP), (5) w¹¹¹⁸; P{y^+t7.7 w^+mC = UAS-mito-QC}attP16 (BL 91640, Mito-QC), (6) w¹¹¹⁸; P{w^+mC = UAS-Dcr-2.D}10 (BL 24651, UAS-Dcr2), (7) y^* w^*; P{w^+mC = UAS-tdTomato.mito}2 (Kyoto DGGR 117016, tdTomato-Mito), (8) w; UAS-IR-Atg18a^KK100064 (VDRC 105366; Atg18a RNAi), (9) w; UAS-IR-Stx17^KK100034 (VDRC 108825; Stx17 RNAi), (10) w; UAS-IR-BNIP3^KK111246 (VDRC 107493; BNIP3 RNAi), (11) w; UAS-IR-Pink1^KK101205 (VDRC 109614; Pink1 RNAi), (12) w; UAS-IR-Park^KK107919 (VDRC 104363; Park RNAi), (13) w; UAS-IR-CG12511^KK109618 (VDRC 101785; CG12511 RNAi), (14) w; UAS-IR-Zonda^KK107775 (VDRC 110620; Zonda RNAi), (15) y¹ sc^* v¹ sev²¹; P{y^+t7.7 v^+t1.8=TRiP.HMS01611}attP2/TM3, Sb¹ (TRiP, BL 36918; FIP200 RNAi), (16) y¹ v¹; P{y^+t7.7 v^+t1.8=TRiP.JF01937}attP2 (TRiP, BL 25896; Stx17 RNAi), (17) y¹ sc^* v¹ sev²¹; P{y^+t7.7 v^+t1.8=TRiP.GL00156}attP2 (TRiP, BL 35578; mTor RNAi), (18) w; UAS-tub-GAL80^ts (from E. Kuranaga), (19) w; UAS-mCD8:GFP, (20) w; DMef2-GAL4, UAS-Dcr2 (Fujita et al., 2017), (21) w; UAS-Dcr2; DMef2-GAL4, sqh-YFP:Mito^BL7194/TM6C Sb Tb (Murakawa et al., 2020). New genotypes generated during this study include the following: (22) w*;; BNIP3 KO CRISPR{3xP3-RFP}/TM6B, Tb¹, (23) w¹¹¹⁸, Atg101 KO CRISPR{3xP3-RFP}/FM7a, (24) w¹¹¹⁸; UASt-GFP-BNIP3_full^attP40, (25) w¹¹¹⁸; UASt-GFP-BNIP3_ΔLIR (W16A/L19A)^attP40, (26) w¹¹¹⁸; UASt-GFP-BNIP3_MER mutant (L49A)^attP40, (27) w¹¹¹⁸; UASt-GFP-BNIP3_ΔMER (Δ42–53) ^attP40, and (28) w¹¹¹⁸; UASt-GFP-BNIP3_ ΔLIR+ΔMER ^attP40.

Plasmid vectors were constructed using standard molecular biology techniques. The DNA polymerase KOD One (TOYOBO, Osaka, Japan) and PrimeSTAR GXL Premix (Takara Bio, Shiga, Japan) were used for PCR amplification. The DNA fragments were assembled using the Gibson Assembly. Construction of the expression vectors of GFP-tagged BNIP3_full under UAS control (pUASt-attB-GFP-BNIP3) was as follows: Drosophila BNIP3 CDS was amplified from template RE48077 (DGRC Stock 9148; https://dgrc.bio.indiana.edu//stock/9148; RRID:DGRC_9148) using the following primer sets: 5′-ATGTCTACGACACCAAAATCGAG-3′ and 5′-TCAGTCAATGACCACACGG-3′. This fragment of BNIP3 CDS was ligated into the linearized pUAST-attB-GFP vector, which is amplified with the following primer sets: 5′-CGTGTGGTCATTGACTGAGCGGCCGCGGCTCGAGGGTACC-3′ and 5′-TTGGTGTCGTAGACATGGTGAAGGGGGCGGCCGCGGAG-3′. To generate plasmids with mutations or deletions in the BNIP3 coding sequence (pUAST-attB-BNIP3_ΔLIR, pUAST-attB-BNIP3_MER^{mut (L49A)}, and pUAST-attB-BNIP3_ΔMER), the plasmid pUAST-attB-GFP-BNIP3_full was used as a template. For the construction of pUAST-attB-BNIP3_ΔLIR, these primer sets: 5′-CTGCGATCGAAGCGAGCACAACAGCTGCGATG-3′ and 5′-CTCGCTTCGATCGCAGATTCGCCCAGCAAATC-3′ were used to introduce the W16A/L19A mutations. For the construction of pUAST-attB-BNIP3 MER^{mut (L49A)}, these primer sets: 5′-AGACTTGCGCGCGAGGCCCAGCGCGAG-3′ and 5′-CTCGCGCGCAAGTCTCAGGTACTCCTC-3′ were used to introduce the L49A mutation. For the construction of pUAST-attB-BNIP3_ΔMER, these primer sets: 5′-CAACAATCGCGAGTCGAACCAGTCG-3′ and 5′-GACTCGCGATTGTTGAATGGCAACGG-3′ were used to delete G42 to Q53. To generate a construct that harbors both ΔLIR and ΔMER (pUAST-attB-BNIP3_ΔLIR+ΔMER), the plasmid pUAST-attB-GFP-BNIP3_ΔMER was used as a template. The same primer sets as pUAST-attB-BNIP3_ ΔLIR construction were used for amplification. Each resultant linear DNA fragment was ligated into a circular plasmid. Generation of the mammalian expression constructs—pMRX-IRES-puro-GFP-BNIP3_full, pMRX-IRES-puro-GFP-BNIP3_ΔMER, and pMRX-IRES-puro-3xHA-mCh-Atg18a—was performed by amplifying the corresponding inserts from pUAST-attB-GFP-BNIP3_full, pUAST-attB-GFP-BNIP3_ΔMER, or LD38705 (DGRC Stock 2722; https://dgrc.bio.indiana.edu//stock/2722; RRID:DGRC_2722), respectively. The amplified fragments were subsequently ligated into the linearized pMRX-IRES-puro vector. All the resultant vectors were validated by DNA sequencing.

To generate Drosophila mutants with a knockout of BNIP3 (CG5059) or Atg101 (CG7053), the CRISPR/Cas9 system was used, following a modified method based on Kondo and Ueda, 2013. gRNA sets for BNIP3 KO and Atg101 KO were designed as follows: for BNIP3 KO, upstream gRNA sequences, 5′-CGTCAACACCAAAGATAACT[TGG]-3′ and downstream gRNA sequences 5′-CTTTCAGTCAATGACCACAC[GGG]-3′; for Atg101 KO, upstream gRNA sequences 5′-GTCCACCTGACGACCCTCCA[TGG]-3′ and downstream gRNA sequences 5′-CTCGCAATGTGACGGGCTGT[CGG]-3′. Each gRNA was individually cloned into a plasmid under the control of the U6 promoter. A donor DNA cassette containing a 3x P3-RFP selection marker, loxP sites, and two homology arms was cloned into the pUC57-Kan plasmid for DNA repair. The gRNAs targeting BNIP3 or Atg101, and the hs-Cas9 coding DNA plasmids, along with the donor plasmid, were microinjected into embryos of the w¹¹¹⁸ strain. F1 files were screened by a selection marker of 3x P3-RFP and validated by genomic PCR and sequencing (WellGenetics Inc). To generate a series of UASt-GFP-BNIP3 transgenic flies, the vectors described above were injected into embryos for phiC31-mediated insertion (WellGenetics Inc).

For collecting DIOMs or their precursors, larvae or pupae were dissected and fixed with methanol. Six DIOMs were isolated from each animal using forceps and a pipette under a stereomicroscope (SZX16, EVIDENT, Tokyo, Japan). The collected muscle cells were lysed with BLA buffer (ReliaPrep Tissue RNA Miniprep System, Promega) and stored at –80°C. Following sample preparation, the frozen samples were thawed, and total RNA was extracted and eluted in RNase-free water according to the manufacturer’s protocol (ReliaPrep Tissue RNA Miniprep System, Promega). Subsequently, cDNA libraries were prepared using CEL-Seq2, a single-cell RNA-seq protocol (Hashimshony et al., 2016). Briefly, reverse transcription was performed using poly(T) primers with a T7 promoter to generate single-stranded DNA from mRNA, which was then converted into double-stranded DNA through second-strand synthesis. Amplified RNA was produced via in vitro transcription using the MEGAscript T7 kit (Thermo Fisher Scientific, Waltham, MA) and subsequently reverse transcribed to construct the cDNA library. Library sequencing was performed using the NovaSeq6000. Cell barcode and UMI in Read1 were extracted by using UMI-tools with the following command ‘umi_tools extract --bc-pattern=CCCCCCCCCCCCCCCCNNNNNNNNNN --stdin read1 --stdout read1_out --read2-in read2 --read2-out=read2_out --whitelist=whitelist’. To extract barcodes, umi_tools whitelist was used. Barcodes in the whitelist were validated by comparing them to barcodes used in the reverse transcription. Reads were mapped to the BDGP6 reference using HISAT2. Read counts for each gene were obtained by featureCounts. The expression levels of genes in each sample were determined from the UMI counts after a normalization process using the DESeq2 package in R. The R programming language was used for each type of analysis, utilizing the pheatmap package for cluster analysis, the prcomp package for PCA, and the Mfuzz package for fuzzy c-means clustering analysis (Kumar and Futschik, 2007).

Total RNA was extracted as described in the ‘RNA-seq analysis of DIOMs’ section. The quality and composition of the extracted RNA were assessed using the Agilent 2100 Bioanalyzer system with the RNA 6000 Pico Kit (Agilent, Santa Clara, USA). rRNA content in the DIOMs was quantified by measuring the peak areas corresponding to 18S and 28S rRNA within the total RNA profile.

Muscle preparations were performed as previously described (Ribeiro et al., 2011). 3IL or pupae were pinned on a Sylgard-covered Petri dish in dissection buffer (5 mM HEPES, 128 mM NaCl, 2 mM KCl, 4 mM MgCl₂, 36 mM sucrose, pH 7.2). The animals were pinned flat and fixed (4% PFA, 50 mM EGTA, PBS) at room temp for 20 min. Then, the samples were unpinned and blocked (0.3% bovine serum albumin, 0.6% Triton X-100, PBS) at room temp for 5 min, incubated with primary antibody overnight at room temperature, washed with PBS, and incubated with Alexa Fluor 488 or 594-conjugated secondary antibody (Thermo Fisher Scientific, Waltham, MA) for 2 hr at room temperature. The immunostained samples were washed and mounted in FluorSave reagent (Merck Millipore, Darmstadt, Germany).

To image live pupal DIOMs, staged pupae were removed from their pupal cases and mounted between a slide glass and a cover glass following the protocol described by Zitserman and Roegiers, 2011. Imaging was performed through the dorsal cuticle. Both live and fixed samples were observed using a confocal microscope (FV3000, EVIDENT, Tokyo, Japan) equipped with either a ×60 silicone/1.30 NA UPlanSApo or a ×4/0.16 objective lens (EVIDENT, Tokyo, Japan). FLUOVIEW (EVIDENT, Tokyo, Japan) was used for image acquisition, and the exported images were processed and analyzed with ImageJ (NIH, Bethesda, USA).

Staged pupae (1 or 4 d APF) were removed from pupal cases, pinned on a Sylgard-covered Petri dish, and dissected directly in fixative (2% paraformaldehyde, 2.5% glutaraldehyde, 150 mM, 5 mM calcium chloride, sodium cacodylate, pH 7.4). They were then fixed for 2 hr at room temperature followed by overnight at 4°C. The dissected fillets were washed with 0.1 M phosphate buffer pH 7.4, post-fixed in 1% OsO₄ buffered with 0.1 M phosphate buffer for 2 hr, dehydrated in a graded series of ethanol, and embedded flat in Epon 812 (TAAB, Aldermaston, UK). Ultrathin sections (70 nm thick) were collected on copper grids covered with Formvar (Nisshin EM, Tokyo, Japan), double-stained with uranyl acetate and lead citrate, and then observed using a JEM-1400Flash transmission electron microscope (JEOL, Tokyo, Japan).

For immunoblots, five larval fillets were collected for each sample. The samples were lysed directly in 1.5× sample loading buffer [100  mM Tris-HCl (pH 6.8), 3% SDS, 15% glycerol, and 0.0015% Bromophenol Blue]. Equal amounts of proteins per sample were subjected to SDS–PAGE and transferred to Immobilon-PSQ PVDF transfer membrane (Merck Millipore, Darmstadt, Germany) with Trans-Blot Turbo Transfer System (Bio-Rad, Hercules, CA, USA). The membranes were blocked with TBS-T buffer [10 mM Tris-HCl (pH 8.0), 150 mM NaCl, 0.1% Tween 20, PBS] containing 5% skim milk for 1 hr, and were then incubated with primary antibodies in Can Get Signal Solution 1 (TOYOBO, Osaka, Japan) overnight at 4°C. Membranes were washed three times with TBST, incubated with HRP-conjugated secondary antibodies in the blocking buffer for 2 hr at room temperature, and washed five times with TBST. Immunoreactive bands were detected using Clarity Western ECL Substrate (Bio-Rad, Hercules, CA, USA) and the ChemiDoc MP Imaging System (Bio-Rad, Hercules, CA, USA). The images were processed using ImageJ.

For immunoprecipitation, HEK293 cells were transfected with plasmids using jetPRIME transfection reagent (Polyplus-transfecton, Illkirch, France) according to the manufacturer’s instructions. The following plasmids were used: pMRX-IRES-puro (-), pMRX-IRES-puro-GFP-BNIP3_full (Full), pMRX-IRES-puro-GFP-BNIP3_ΔMER (MERΔ), and pMRX-IRES-puro-3xHA-mCh-Atg18a. After 24 hr of transfection, cells were washed with PBS and lysed in lysis buffer [20 mM HEPES (pH 7.4), 1 mM EDTA, 150 mM NaCl, 0.1% Triton X-100, 10% glycerol, 1 mM PMSF, and protease inhibitor cocktail (Nacalai Tesque, Kyoto, Japan)]. Cell lysates were centrifuged at 17,000 × g for 1 min at 4°C, and the supernatants were incubated with GFP-Trap agarose beads (Chromotek, Planegg-Martinsried, Germany) for 1.5 hr at 4°C with gentle rotation. After incubation, beads were washed with Wash Buffer [20 mM HEPES (pH 7.4), 150 mM NaCl, 0.1% Triton X-100, 10% glycerol] to remove non-specific proteins. Bound proteins were eluted by boiling in 2× sample buffer for 5 min, and subsequent steps were performed as described above for immunoblots.

The amino acid sequences of Drosophila BNIP3 (Q9VPD6) and Atg18a (Q9VSF0) were retrieved from the UniProt Knowledgebase (https://www.uniprot.org/uniprotkb/) and submitted to the AlphaFold Server (https://alphafoldserver.com/) to predict the structure of their complex. The PyMOL Molecular Graphics System (Version 2.5.4; Schrödinger, LLC) was then used to analyze the resulting five predictions and visualize the top-ranked structure.

The Adult lifespan assay was performed according to a previously reported protocol (Linford et al., 2013). Briefly, age-synchronized adult flies were collected, sorted by sex, and housed in groups of 30 per vial. Flies were maintained at 25°C under a 12 hr light/12 hr dark cycle and transferred to fresh food vials every 2–3 days without anesthesia. Mortality was recorded at each transfer, and statistical significance was determined using the log-rank test.

The climbing assay was conducted according to the protocol previously published (Gevedon et al., 2019). Adult flies were gently transferred without anesthesia into a vertical climbing chamber composed of two connected vials, marked at 5 cm from the bottom. After a brief recovery period, flies were tapped to the bottom of the chamber, and the number of flies that climbed above the 5 cm mark within 10 s was recorded. The climbing success rate was calculated as the percentage of flies that reached or surpassed the mark. Each group was tested in five trials, with a rest period of several minutes between trials. The median success rate across trials was used as the representative value for each group.

Late-stage pupae were collected and maintained at 25°C. The eclosion process was recorded, and the time from initial rupture of the pupal case to the complete emergence of the adult fly was measured. For each group, 7–10 individuals were analyzed, and the average elapsed time was calculated.

For cell volume measurement (Figure 2E), the DIOMs expressing GFP were imaged at 1 μm intervals. The muscle cell volume was calculated by multiplying the area of each XY slice by the interval length and summing the results across all slices. To measure nucleolus and nucleus volume (Figure 2—figure supplement 2E, F), DIOMs stained for DNA (H33342) and Fibrillarin were imaged at 0.5 μm intervals. DAPI- and anti-Fibrillarin-stained areas were extracted by binarization using ImageJ. This process was repeated for the z-stacks, and the nucleolus and nucleus volume were determined by multiplying the area of each slice by the interval length between slices. For mitochondrial area measurement in TEM images (Figure 4C), each mitochondrion and the corresponding DIOM region were manually segmented, and the total mitochondrial area and DIOM area per image were calculated for each image using ImageJ. The total mitochondrial area was then normalized to the DIOM area. For each genotype, at least nine images from multiple animals were analyzed. Quantification of the number of autophagosomes, mitophagosomes, or mitochondria in TEM images was conducted as follows (Figure 4D, E): Double-membrane-bound structures containing undigested cytoplasmic contents (autophagosomes), autophagosomes enclosing mitochondria (mitophagosomes), and unengulfed mitochondria were manually counted. At least 17 images derived from multiple animals were analyzed for each genotype. Mitochondrial area in DIOMs is quantified in Figure 5D’ and 6A’, D’. Cellular regions containing F-actin were manually segmented, and tdTomato (Figure 5D’) or GFP (Figure 6A’, D’) channel intensity within these segmented regions was measured. For Figures 5D’ and 6A’, image preprocessing steps, including filtering, background subtraction, and intensity binarization, were applied to the tdTomato (Figure 5D’) or GFP (Figure 6A’) channel using ImageJ to reduce noise. The area of each detected object was measured, and the total mitochondrial area was defined as the sum of these object areas. Mitochondrial area values were then normalized to the corresponding cellular region area. For Figure 6D’, GFP intensity quantification was performed as follows: because GFP signals in the 3IL BWM images were weaker than those at 4 d APF, the GFP channel intensity was uniformly enhanced across all images at a constant scale. All images were preprocessed using filtering and background subtraction. The mean GFP intensity within each segmented area was then measured and normalized, with the mean GFP intensity of the control for each developmental stage set to 1.

For Figure 5D’, at least 46 DIOMs from 5 animals were analyzed per genotype. For Figure 6A’, D’, at least five DIOMs per fly were analyzed, with three biological replicates per genotype.

For the Mito-QC assay, a representative DIOM was imaged for each genotype and analyzed by ImageJ (Figure 6B). Image preprocessing included filtering and background subtraction for both GFP and mCherry channels. Dot plots were then generated, and Pearson’s correlation coefficients for GFP and mCherry intensities were calculated using the Coloc2 plugin in ImageJ (Figure 6B, B’).

Each experiment was performed with at least three different cohorts of unique flies analyzed. One exception was for TEM analyses, performed on two parallel replicates with multiple animals each. All experiments were performed in parallel with controls. Raw data files corresponding to all quantified results are provided in Supplementary file 2. Error bars show the standard deviation for bar charts. When more than two genotypes were used in an experiment, Sidak’s test, Kruskal–Wallis test, or Dunnett’s T3 multiple comparisons test was used (GraphPad, Prism9 version 9.5.1). Mann–Whitney tests were used to compare two means. p < 0.05 was regarded as statistically significant (*p < 0.05, **p < 0.001, ***p < 0.0001, and ****p < 0.00001).