Integration of proteomic and genetic approaches to assess developmental muscle atrophy

The inadequate maintenance of muscle tissue mass, or muscle atrophy, has severe consequences for maintaining normal life activities and is a contributing factor to increased morbidity rates (Bonaldo and Sandri, 2013; Ebert et al., 2019; Piccirillo et al., 2014). In humans, progressive muscle wasting as a result of advanced disease states (e.g. cancer, diabetes, etc.) or aging decreases quality of life, limits economic opportunities and increases healthcare costs. While many studies have examined the triggers and resulting physiological changes associated with muscle wasting, details concerning the molecular basis of these diseases are only emerging. A better understanding of the mechanisms that contribute to muscle wasting must first be established before successful therapeutic strategies can be developed.

Numerous mouse models exist that mimic muscle disuse or disease states predicted to recapitulate atrophic conditions. Examples include mechanical ventilation to induce diaphragm atrophy, limb immobilization, hind-limb suspension, denervation and exposure to high doses of glucocorticoids (Gao et al., 2018; Malavaki et al., 2015; Romanick et al., 2013). These steroid hormones are released in times of stress as a result of renal failure, diabetes or sepsis and result in muscle atrophy by inhibiting amino acid import and protein synthesis (Braun and Marks, 2015; Schakman et al., 2013). Because glucocorticoids reduce inflammation in the treatment of autoimmune diseases, arthritis and cancer, muscle wasting is an adverse side effect of prolonged treatment. Unfortunately, complex interactions between multiple tissues and the infiltration of immune cells that promote inflammation and drive disease progression prevent a complete understanding of the muscle autonomous contribution to muscle wasting.

The normal histolysis of muscles that occurs during insect pupal development provides a unique opportunity to study muscle intrinsic properties that contribute to a reduction in muscle mass while maintaining contractile strength (Piccirillo et al., 2014; Schwartz, 2008, 2019). Studying this stereotyped developmental atrophy has several inherent advantages. First, the time course is shorter in insect models, whereby atrophy is accomplished in hours or a few days compared with weeks for mammalian induced atrophy models. Second, developmental atrophy is not lethal, which allows for observation of the complete atrophy process in addition to dissecting factors that contribute to atrophy initiation versus progression. Third, developmental atrophy is induced by a class of insect steroid hormones called ecdysteroids (Schwartz and Truman, 1983; Zirin et al., 2013); the rapid loss of muscle mass caused by the structurally related glucocorticoids and ecdysteroids suggests similar mechanisms of action that may identify mechanisms common to multiple types of induced atrophies. Finally, the set of genetic and biochemical tools available in insect models is extensive, allowing for spatial and temporal control to manipulate gene function and an arsenal of RNA interference (RNAi) lines for tissue-specific knockdown.

Herein, we utilized the capabilities of two insect models, the tobacco hornworm moth Manduca sexta and the fruit fly Drosophila melanogaster, to uncover and characterize novel proteins essential for developmental muscle atrophy. Each organism is beneficial for different reasons. The large size of M. sexta makes it an ideal biochemical model to uncover novel proteins that are upregulated or downregulated during muscle atrophy, while the functional requirement for these genes is confirmed in the in vivo platform afforded through D. melanogaster genetics. Sets of muscles in both organisms undergo atrophy during pupal development. In M. sexta, the large intersegmental muscles (ISMs) lose ∼40% of their muscle mass in the 3 days prior to adult eclosion (Bayline et al., 2005; Schwartz and Truman, 1983). Isolation of ISMs from a single organism provides enough material for biochemical analysis. In D. melanogaster, genetic tools exist to knockdown or overexpress genes in the dorsal internal oblique muscles (DIOMs) that undergo atrophy (Kuleesha et al., 2016a,b; Wasser et al., 2007; Zirin et al., 2013).

Multiple studies in M. sexta have defined factors that contribute to the ISM atrophy program. The insect molting hormone 20-hydroxyecdysone (20E) triggers atrophy, and prolonged administration of 20E is sufficient to block ISM degeneration and subsequent programmed cell death (Bayline et al., 2005; Schwartz and Truman, 1983). This reliance on hormone signaling strongly suggests transcriptional programs largely mediate the atrophy process. Indeed, analysis of individual mRNAs or genome-wide RNA sequencing has revealed a decline of muscle structural genes and upregulation of genes that promote proteasomal turnover (Haas et al., 1995; Schwartz et al., 1990), both of which are hallmarks of mammalian muscle atrophy (Bilodeau et al., 2016; Khalil, 2018; Kitajima et al., 2020). In Drosophila, inhibition of Tor signaling via RNAi knockdown of Rheb or Tor enhanced DIOM atrophy, while a reduction of the negative regulators Tsc1 and Tsc2 resulted in larger muscles that failed to undergo atrophy (Kuleesha et al., 2016a). As Tor signaling promotes muscle growth by inhibiting autophagy, it is not too surprising that RNAi silencing of the autophagy-related genes Atg5, Atg9, Atg12, Atg17 or Atg18 also prevents muscle loss.

Here, we expand upon these previous studies to examine changes in protein levels at the initiation (D15) or conclusion (D18) of developmental atrophy in M. sexta. We found that proteins required for protein catabolism and energy production are globally induced. The tight regulation of thin filament breakdown was also evident, whereby actin disassembly proteins are increased with a corresponding reduction in actin filament promoting proteins. Surprisingly, a number of thick filament proteins remain during atrophy, suggesting an ordered breakdown of myofibril components.

Manduca sexta (Linnaeus 1763) specimens were a kind gift of Dr Mike Kanost (Department of Biochemistry and Molecular Biophysics, Kansas State University). Pupae were staged as previously described (Schwartz and Truman, 1983). Briefly, 0–1 day old male pupae were transferred to a 26°C incubator and raised on a 12 h light:12 h dark cycle. Samples were processed between 08:00 h and 10:00 h on day 15 at the beginning of muscle atrophy or on day 18 near the end of muscle atrophy, but before adult eclosion.

During dissection of staged day 15 or 18 pupae, the ISMs were removed, washed with ice-cold saline (4 mmol l⁻¹ NaCl, 40 mmol l⁻¹ KCl, 18 mmol l⁻¹ MgCl, 3 mmol l⁻¹ CaCl₂, 1.5 mmol l⁻¹ Pipes pH 6.5, 41.5 g sucrose in 500 ml ultrapure water), transferred to separate Eppendorf tubes, flash frozen in liquid nitrogen, and stored at −80°C. After thawing, samples were lysed in ∼2 ml of 50 mmol l⁻¹ Tris+1% SDS in a glass homogenizer on ice and were then centrifuged at 20,000 g for 10 min at 4°C to pellet debris. Total protein was determined using the bicinchoninic acid (BCA) assay (Pierce). Approximately 35 µg of each sample was mixed with SDS sample reducing buffer [0.5 mol l⁻¹ Tris-HCl pH 6.8, 10% glycerol, 10% (w/v) SDS, 0.5% (w/v) Bromophenol Blue] and run on a 10% SDS-PAGE gel. Proteins were visualized with Coomassie Blue (3 g l⁻¹ Coomassie Brilliant Blue R250, 45% methanol, 10% acetic acid) staining followed by destaining (45% methanol, 10% acetic acid). The gel was sent to the Recombinant DNA/Protein Core Facility at Oklahoma State University for mass spectrometry (MS) analysis. Independent gel lanes were fractionated and trypsinized as described previously (Voruganti et al., 2018). Peptides were analyzed on 75 µm×40 cm nano-columns fabricated in-house, and packed with 3 μm beads of Magic AQ C18 resin (Michrom). Peptides were injected using a vented trap column configuration, and separated on 6–35% linear acetonitrile gradients developed over a 100 min chromatography run at 250 nl min⁻¹. Columns terminated in a stainless-steel emitter for peptide ionization within a Nanospray Flex ion source (Thermo). Peptide ions were analyzed in a quadrupole-Orbitrap ‘Fusion’ mass spectrometer (Thermo) using a 3 s ‘top speed’ data-dependent MS/MS scan method. In this method, peptide precursors were measured within the Orbitrap sector at a nominal resolution of 120,000. Ions were selected for MS/MS using the quadrupole, followed by collision-induced dissociation fragmentation and analysis of the fragments using the ion trap detector. Peptides from each individual adsorption sample were subjected to three individual LC-MS/MS analyses (technical replicates). Biological replicates were performed as indicated in Fig. 1C. Peptides were identified and quantified using MaxQuant v.1.5.3.8 (Cox and Mann, 2008) to search a database of 27,403 M. sexta sequences downloaded from https://data.nal.usda.gov/dataset/Manduca-sexta-official-gene-set-v20. Database searches utilized the default MaxQuant parameters.

Manduca sexta samples were prepared as described above; 100 µg of each sample was added to 2D sample buffer (8 mol l⁻¹ urea, 2% Chaps, 0.05 mol l⁻¹ DTT, 1× Biolyte 3-10 and 0.001% Bromophenol Blue), transferred to separate lanes of a BioRad Protean IEF focusing tray containing ReadyStrip IPG strips, pH 5–8 or pH 3–6, 7 cm (BioRad), and covered with mineral oil. The strips were passively rehydrated (20°C for 12 h) and then focused (for pH 3–6: 4000 V, Rapid, 11,000 V h; for pH 5–8: 4000 V, Rapid, 9000 V h) with a Protean IEF cell (BioRad). After focusing, the strips were equilibrated in DTT-containing buffer [6 mol l⁻¹ urea, 0.375 mol l⁻¹ Tris-HCl pH 8.8, 2% SDS, 20% glycerol, and 2% (w/v) DTT] for 15 min followed by incubation in iodoacetamide-containing buffer [6 mol l⁻¹ urea, 0.375 mol l⁻¹ Tris-HCL pH 8.8, 2% SDS, 20% glycerol and 2.5% (w/v) iodoacetamide] for 15 min before rinsing briefly with standard SDS-PAGE running buffer. Molecular weight separation was achieved by running on 10% SDS-PAGE gels and bands were visualized by silver staining. Spots of interest were extracted from the gel with a plastic pipette tip and sealed in a 1.5 ml tube. Samples were sent to Oklahoma State University for MS analysis as described above.

Statistical analysis was carried out in the freely available Perseus 1.2.0.16 software platform (https://maxquant.net/perseus/) (Tyanova et al., 2016). Label-free quantification (LFQ) intensities were imported and the raw data were filtered to remove potential contaminants or incorrect protein identifications. Expression values were converted to log₂ for data normalization before statistical analysis (Cox et al., 2014). For each biological replicate, proteins that passed the statistical threshold after being subjected to a two-sample t-test (permutation-based false discovery rate, FDR P<0.05) were kept for further analysis. Imputation was performed to replace missing values from a normal distribution. Fold-change (FC) was calculated for proteins that showed statistically significant t-test differences (P<0.05) and threshold values x>0.5 (positive FC) or x<0.5 (negative FC). Only M. sexta proteins that were considered statistically significant in 2 out of 3 biological replicates are present in Table S1. Principal component analysis (PCA) was performed in Perseus by selecting Analysis→Clustering/PCA→Principal component analysis (Benjamini–Hochberg FDR P<0.05).

Following MS peptide identification, the protein ID corresponding to each M. sexta hit was used to search the published protein sequence data from the M. sexta official gene set 2. FASTA sequences were then subjected to BLAST analysis against the D. melanogaster flybase (https://flybase.org/) database using default search parameters. The resulting D. melanogaster protein names and flybase IDs are listed in Table S1.

All fly stocks were reared and maintained on standard cornmeal medium at 25°C. RNAi crosses were performed at 25°C. The following Drosophila stocks were obtained from the Bloomington (BL) Drosophila Stock Center (BDSC) or the Vienna Drosophila Resource Center (VDRC): mef2-Gal4 (BL27390); MHC-GFP (BL38463); UAS-GFP RNAi (BL9331); UAS-DN-EcrB1 (UAS-EcR.B1-ΔC655.F645A; BL6869); UAS-atrogin1 RNAi (BL 31373); UAS-tn RNAi (v19290); UAS-awd RNAi (BL42532); UAS-skap RNAi (BL55168); UAS-tmod RNAi (BL31534); UAS-Gdi RNAi (BL 27309); UAS-CG14207 RNAi (BL64571); and UAS-cib RNAi (BL36630). A mef2-Gal4, MHC-GFP stock was generated using standard recombination procedures and crossed out to UAS-cib RNAi flies to visualize all pupal muscles.

To monitor DIOM atrophy, pupae were collected at 0, 12 and 24 h after puparium formation (apf). Whole-mount DIOMs were visualized and photographed using a Leica M165FC fluorescence stereo microscope. Flat muscle preparations were prepared by dissection in phosphate-buffered saline (PBS), followed by fixation with 4% formaldehyde in PBS for 30 min at room temperature, washed with PBS plus 0.3% Triton X-100 and stained with Alexa Fluor™ 488 Phalloidin (Thermo Fisher Scientific) for 2 h at room temperature. Images of muscle preparations were acquired using a Zeiss LSM 700 confocal microscope, processed using Zeiss Zen software, and assembled into figures in Photoshop Elements 12. DIOMs from the second abdominal segment (A2) were used for comparing the muscle area between the control and experimental groups. At least 6–10 flies per genotype were used to monitor DIOM development. Quantifications were performed using the Analyze Particles function in ImageJ, and recorded in an Excel spreadsheet. The raw data were imported into GraphPad Prism 6.0 software for the generation of either bar graphs (Fig. 6A) or mean and error plots (Figs 4D, 5C and 6D). Student's t-tests followed by Mann–Whitney tests were used to compare genotypes at each time point individually, without assuming a consistent s.d. P-values and sample size are noted in each figure legend.

qPCR was performed to verify RNAi knockdown in D. melanogaster. RNA was isolated from single whole wandering L3 larvae using the RNeasy miniprep kit (Qiagen) in triplicate. cDNA was generated from 125 ng RNA using the SuperScript III First-Strand Synthesis System Kit (Invitrogen). Each cDNA solution was diluted 1:10 and mixed with SYBR Select Master Mix for CFX (Applied Biosystems) and the appropriate primers were added. rp49 was used as the reference gene with the following primer sequences: forward 5′-GCCCAAGGGTATCGACAACA-3′ and reverse 5′-GCGCTTGTTCGATCCGTAAC-3′; atrogin1 primer sequences: forward 5′-CGTTCCAAGGTGCTCGAGT-3′ and reverse 5′-AACCCGGCTATCTCTCTGGT-3′. Reactions were run on the CFX96 Touch Real-Time PCR Detection System with CFX Manager Software (BioRad). The average of the triplicates was used to calculate the 2−ΔΔCt values (normalized fold expression) (Livak and Schmittgen, 2001).

The ISMs that form in the M. sexta embryo continue to grow during the larval instar stages and undergo a 3 day period of atrophy just prior to adult eclosion, after which the ISMs undergo programmed cell death (Schwartz and Truman, 1983). To examine the relative changes in protein abundance during this pupal phase of tissue degeneration, the ISMs from individual M. sexta pupae were isolated at the initiation (day 15) and conclusion (day 18) of muscle atrophy (Fig. 1A). Analysis of day 15 and day 18 samples revealed minor differences in protein abundance visualized by SDS-PAGE electrophoresis followed by Coomassie staining (Fig. 1B, asterisks). To obtain a full profile of protein changes during the atrophy process, we used MS for peptide identification. For each biological replicate at day 15 or day 18, samples were trypsinized, ionized by electrospray and the raw MS/MS spectra were searched against the M. sexta database (Kanost et al., 2016). Analysis of LFQ values was performed using the MaxQuant software suite to obtain a list of proteins at a FDR<0.5. A total of 244 proteins emerged from these datasets with the criteria of being present in at least two biological replicates (Table S1). PCA of this core proteome set revealed biological variability between replicates, but distinct separation between day 15 and day 18 peptide groups (Fig. 1C). These findings demonstrate the high reproducibility between the developmental time points and the MS output data.

Heat map analysis was performed to evaluate expression trends during the course of ISM atrophy. Even though variability was observed between sample groups at each time point, proteins that were upregulated or downregulated separated into distinct groups (Fig. 2A). Approximately 60% of the proteins identified were increased from day 15 to day 18 samples, while ∼33% of proteins showed a decrease during this period of atrophy (Fig. 2B); 4.1% of proteins were consistently detected in day 15 or day 18 samples, but showed variability in upregulation or downregulation and were designated ‘no change’.

To understand protein groups that are required for developmental muscle atrophy, we performed gene ontology (GO) annotation enrichment analysis. As our long-term goal was to functionally evaluate genes required for atrophy in the genetically amenable Drosophila model, we first used the basic local alignment search tool (BLAST) to identify M. sexta protein orthologs in this species (Table S1). Approximately 1.2% (3/244) of the M. sexta hits did not reveal a Drosophila counterpart using this approach. The remaining orthologous Drosophila proteins were then assessed for annotations based upon predicted cellular functions to uncover major descriptor terms (Fig. 2C). As expected, the core proteome was enriched for metabolic processes that are essential in any cellular system and highlighted by an increase in proteins that contribute to energy production, including the transport and metabolism of nucleotides, lipids, carbohydrates and amino acids. Additional categories of upregulated proteins included post-translational modifications, chaperones and cytoskeletal components, which will be discussed in more detail below. To validate our global MS approach, we also performed 2D gel analysis followed by MS identification of individual protein spots. Indeed, peptide determination revealed that 5 out of 6 individual protein spots that decreased in abundance from day 15 to day 18 were present in our original MS data (Fig. S1, Table S1).

To identify specific classes of proteins that contribute to developmental muscle atrophy, we further assessed GO terms using the gene annotation and analysis resource Metascape (https://metascape.org/gp/index.html#/main/step1). Eight groups of proteins, some of which overlap, decreased in abundance from day 15 to day 18 (Fig. 3A). One class of proteins that were reduced in abundance comprised actin filaments and stress fibers, the latter describing contractile actin bundles usually found in non-muscle cells. Proteins that are destined for the plasma membrane, the extracellular space or cell–cell junctions were also decreased, suggesting that intracellular modifications are prioritized over interactions with other cells or the extracellular environment.

Notably, a number of proteins that comprise contractile sarcomeres were differentially downregulated or upregulated. A subset of proteins at the Z-disc decreased in abundance (e.g. α-actinin, Zasp52, CG14207), while others showed elevated protein levels [e.g. muscle-specific protein 300 kDa (Msp300) and muscle LIM protein at 60A (Mlp60A)]. One clear trend was the increase in A-band and/or myosin-containing proteins. Proteins of this class included myosin heavy chain (Mhc), myosin light chain 1 (Mlc1), myosin light chain 2 (Mlc2) and myofilin, among others (Fig. 3B). This result was somewhat surprising as the mRNA transcripts for some of these genes have previously been shown to be repressed during atrophy (Tsuji et al., 2020). Other protein groups that were broadly upregulated included proteins that comprise the endomembrane system and other cellular membranes (Fig. 3A). Consistent with the observed increase in proteins that influence energy production, proteins present in mitochondrial complexes, especially in the mitochondrial inner membrane, were correspondingly upregulated during atrophy.

We next examined the top five groups of proteins that were either downregulated (Fig. 3C) or upregulated (Fig. 3D) during atrophy using Protein Analysis THrough Evolutionary Relationships (PANTHER). The largest class of reduced proteins were chaperone complexes (GO:0101031), composed largely of seven of the eight members of the Drosophila chaperonin-containing T-complex (CCT1-4, CCT6-8; GO:0005832). As CCT proteins facilitate the folding of actin and tubulin proteins, their absence further supports the observations that both of these cytoskeletal monomer subunits are decreased (Table S1). The second class of downregulated proteins were the larval serum complex proteins (GO:0005616). This decrease in larval serum complexes may provide an additional source of amino acids during atrophy to prevent wasteful duplication of metabolic resources (Roberts et al., 1991). Degradation of muscle tissue predicts that muscle-specific complexes will be lost. Indeed, a reduction in troponin/thin filament proteins (GO:0005861, GO:0005865), Z-disc components (GO:0030018) and actin filaments (GO:0097517, GO:0032432, GO:0005884) support this prediction. Troponins and tropomyosins are thin filament-associated proteins that bind to sarcomeric actin filaments (Fig. 3B). Different classes of proteins that are increased during developmental atrophy can be simplified into three main groups. Upregulation of proteasome (GO:0000502) and peptidase complexes (GO:1905368) ensures turnover of proteins to reduce muscle mass. The 20- to 40-fold increase in ATP synthase components (GO:0006091) and mitochondrial respiratory chain proteins (GO:0098803) support increased energy demands for protein turnover, in addition to providing ribosomal subunits (GO:0005840) for protein translation, suggesting that atrophy requires a substantial shift in energy metabolism to promote muscle catabolism.

Holometabolous insects make two sets of muscles during their life cycle, one in embryogenesis for larval movement and the other during pupation for adult life (Bothe and Baylies, 2016). During metamorphosis, remodeling of the larval musculature ensures muscles are functional for post-emergent functions, including walking, mating and flight. While the M. sexta ISMs are ideal for biochemical studies, this organism lacks the ability for routine targeted genetic manipulations, such as RNAi to assess protein function (Garbutt et al., 2013). To functionally test the validity of our MS dataset and to directly compare conserved roles for proteins in developmental muscle atrophy across species, we chose to examine muscle remodeling during Drosophila pupation.

Two groups of muscles that undergo morphogenesis during the Drosophila pupal transition are the dorsal external oblique muscles (DEOMs) and the dorsal internal oblique muscles (DIOMs) (Kuleesha et al., 2016a,b; Wasser et al., 2007; Zirin et al., 2013). Both of these muscle groups are present in abdominal segments A1 to A7 and can be visualized with a myosin heavy chain (MHC)-GFP transgene (Fig. 4A). The DEOMs are subject to histolysis and degenerate 12–24 h apf depending on their anatomical location (Fig. 4A,B). In contrast, the DIOMs undergo two distinct phases: a period of atrophy followed by hypertrophy in the second half of pupation before adult eclosion. While the entire period of atrophy continues until ∼50 h apf, the reduction in muscle area is readily apparent by 24 h apf and was thus used as an endpoint to assess the presence or absence of normal atrophy (Fig. 4A,B). In vivo analysis of GFP RNAi control muscles labeled with phalloidin showed a ∼40% and 70% decrease in muscle area by 12 h and 24 h, respectively (Fig. 4C,D). As ecdysone regulates remodeling of the M. sexta dorsal external oblique body wall muscles (Hegstrom et al., 1998), we first tested whether muscle expression of a dominant-negative ecdysone receptor (mef2>DN-EcR) prevents DIOM degeneration. Consistent with Zirin et al. (2013), we confirmed the persistence of the DIOMs at both 12 and 24 h apf (Fig. 4C,D). These data validate the Drosophila DIOMs as a genetic model to verify hits obtained in our M. sexta MS dataset.

Elevated protein turnover is a conserved hallmark of muscle atrophy (Bilodeau et al., 2016; Khalil, 2018; Kitajima et al., 2020). As shown in Fig. 3D, protein subunits that comprise the proteasome were the most over-represented as atrophy progressed. We mapped the proteins in the ‘Proteasome regulatory’ and ‘Peptidase’ complexes using the STRING (https://string-db.org/) and Cytoscape (https://cytoscape.org/) software programs (Fig. 5A; Table S1). The resulting protein interaction network revealed coordinated upregulation of proteasomal proteins (light blue boxes) that physically interact with deubiquitinating enzymes (DUBs; yellow ovals) and peptidases (light purple ovals). DUBs typically remove ubiquitin (Ub) independent of, or in conjunction with, proteasomal proteolysis (Dikic, 2017). Ubiquitin-specific protease 14 (Usp14) reversibly associates with the 19S component of the proteasome (Wertz and Murray, 2019), reflected in our interactome data as a protein that links other DUBs, including ubiquitin carboxy-terminal hydrolase (Uch), ubiquitin-specific protease 5 (Usp5) and ubiquitin-specific protease 7 (Usp7), to the proteasome. We found that 29 out of 33 proteasome subunits were enriched after the onset of atrophy; their relative locations within the proteasome are shown in Fig. 5B. Each α or β subunit of the 20S core particle (CP) possesses catalytic activity for substrate degradation, while subunits that comprise the regulatory particle (RP), including the regulatory particle triple-A ATPases (Rpt1–6) and regulatory particle non-ATPases (Rpn1–15), assist in the recognition and unfolding of client proteins (Tanaka, 2009). These data confirm the overall sensitivity of the MS approach, as over 87% of the expected proteasome components were identified as having altered protein levels during atrophy.

E3 ubiquitin ligase proteins append K48-linked Ub chains to proteins destined for proteasomal turnover. Three E3 proteins have been extensively studied in mammalian muscle atrophy: muscle RING-finger 1 (MuRF1), Atrogin-1/MAFbx and tripartite motif-containing protein 32 (TRIM32). Protein levels of MuRF1 and Atrogin-1 are rapidly induced in all types of atrophy (Bodine and Baehr, 2014; Gomes et al., 2001), while TRIM32 functions during atrophy upon fasting (Cohen et al., 2012). As none of these proteins, nor any other E3 ligases, were upregulated in our experimental conditions during M. sexta atrophy, we wondered whether this organism possesses homologous proteins. Indeed, BLAST searches of the corresponding mammalian protein sequences revealed modest sequence similarities to proteins present in both M. sexta and D. melanogaster. TRIM9, the protein with the most identity to MuRF1 (M. sexta – 20.67%; D. melanogaster – 21.65%), is highly expressed in the Drosophila nervous system (Morikawa et al., 2011; Song et al., 2011). In contrast, publications by us and others have shown that both Drosophila Atrogin-1/CG11658 and TRIM32 [encoded by thin (tn)] are enriched in muscle tissue (Bawa et al., 2020; Domsch et al., 2013; LaBeau-DiMenna et al., 2012; Vishal et al., 2018). Muscle knockdown of either atrogin1 or tn mRNA transcripts both showed a partial block of DIOM degeneration (Fig. 5C; Fig. S2). It is not clear why these proteins were not detected in the M. sexta MS dataset as we identified a clear orthologous protein for both Atrogin-1/F-box only protein 32 (Fig. 5D) and TRIM32 (Fig. 5E) in both Drosophila and M. sexta. Moreover, M. sexta possesses two isoforms of each protein, each of which is slightly more closely related to their mammalian counterparts than to the Drosophila proteins. These results together convincingly support a role for proteasome components, DUBs and E3 ligase proteins in increased protein turnover during muscle atrophy.

Next, we wanted to assess whether the Drosophila DIOM system could be a good experimental model to validate proteins that were either upregulated or downregulated during M. sexta atrophy. Six genes for which we could obtain RNAi lines were tested for inhibition of muscle atrophy. Knockdown of three genes, abnormal wing discs (awd), succinyl-coenzyme A synthetase β subunit (skap) and tropomodulin (tmod), all underwent normal DIOM atrophy similar to controls (mef2-Gal4/+ or mef2>GFP RNAi) by 24 h apf (Fig. 6A). In contrast, inducing RNAi against three other targets, GDP dissociation inhibitor (Gdi), the chaperone CG14207 or ciboulot (cib), blocked DIOM atrophy to different extents (Fig. 6A).

Proteins that comprise the thin filament and/or interact with actin are predominantly downregulated from day 15 to day 18 (Fig. 3A,C). Exceptions include proteins that negatively regulate actin filament formation (Gelsolin, Cofilin, and Capulet/CAP), which would be expected to increase if destruction of actin filaments is essential for sarcomere breakdown. Mapping the actin interactome using STRING and Cytoscape revealed three distinct clusters of proteins (Fig. 6B). The multiprotein complex CCT consists of 8 subunits, 7 of which are represented in the MS data. Downregulation of Act57B, Act88F, αTub84D and βTub56D in addition to the CCT complex is consistent with a major role for this molecular chaperone in assisting with the folding of newly synthesized actin and tubulin subunits, although it is likely that other substrates exist.

The largest group that forms the core of the actin interactome network are proteins that are part of, or associated with, actin thin filaments. Act57B and Act88F are muscle actin proteins predominant in Drosophila larval and pupal muscles (Fyrberg et al., 1980; Tobin et al., 1980). Actin-binding proteins, including Tropomyosin 1, Tropomyosin 2, Troponin C at 25D, Troponin C at 73F, and tropomodulin, all decrease in abundance from day 15 to day 18. As blocking Cib function qualitatively showed the strongest phenotype (Fig. 6A) and was reduced in our label-free (Table S1) and 2D gel (Fig. S1) MS experiments, we chose to examine this further. Cib has sequence similarity similar to mammalian β-thymosin. While β-thymosin binds to G-actin to sequester monomer addition, Cib functionally behaves like Profilin to promote growth at barbed but not pointed ends of actin filaments during Drosophila brain metamorphosis (Boquet et al., 2000). To functionally test the requirement for Cib in DIOM atrophy, we again employed RNAi knockdown during pupal development. Compared with control GFP RNAi muscles, knockdown of cib prevented muscle atrophy, not just in the abdominal segment shown (A2; Fig. 6C, middle panel) but in the DIOMs of all segments (Fig. 6D, bottom panel; cf. Fig. 4A). These data together confirm for the first time a functional role for Cib in muscle remodeling during metamorphosis.

The large size of insect ISMs provides a distinctive advantage in garnering enough material for biochemical analyses and subsequent protein identification using MS. Indeed, a previous study uncovered 15 proteins using 2D gel electrophoresis that showed differential expression of skeletal muscle proteins in the silkworm Bombyx mori from the last day of larval development compared with the end of pupal metamorphosis (Zhang et al., 2007). Here, we expand upon these previous findings to directly assess overall protein composition during the period of muscle atrophy in the ISMs. Our conservative estimation predicts that <3% of proteins show increased or decreased protein levels to ensure tight regulation of developmental atrophy. More importantly, these findings highlight conserved aspects of atrophy across lepidopteran and dipteran insects.

Most of the published M. sexta ISM studies have examined mRNA levels as a readout during atrophy (Haas et al., 1995; Hu et al., 1999; Schwartz et al., 1990; Sun et al., 1996), including a recent comprehensive study examining differential gene expression during skeletal muscle atrophy (Tsuji et al., 2020). Here, we examined global changes in the skeletal muscle proteome, thus providing an opportunity to compare relative transcript and corresponding protein levels. Clear similarities were obvious in the two datasets, including a decrease in components that positively regulate actin filament assembly and actin-mediated cell contraction. Similar upregulated RNA and protein levels were observed with proteasome components and some elements of metabolic activity, including catalytic and transport activity. More surprising were notable differences between transcript and protein levels for myosin and other thick filament proteins. For example, Mhc and Mlc1 mRNA levels have been shown to decrease during developmental muscle atrophy (Tsuji et al., 2020), while we observed an increase in both Mhc and Mlc1 protein levels, along with other thick filament-associated proteins (Myofilin, Flightin).

Multiple explanations may account for the preferential loss of thin filament proteins and concomitant enrichment of thick filament proteins during atrophy. Foremost, the relative half-lives for major contractile proteins may differ. If both genes are transcriptionally repressed, but Mhc has a longer half-life than actin, it may appear that Mhc protein abundance is higher. Alternatively, increased protein stability may reflect a biological necessity to undergo ordered destruction of the sarcomere. Myosin binding protein-C (MyBP-C), and myosin light chains 1 and 2 (MyLC1 and MyLC2) are important for thick filament stability and their selective loss during atrophy precedes degradation of mammalian myosin heavy chain (MyHC) (Cohen et al., 2009). Thus, the elevated levels of Mlc1 and other thick filament proteins may serve to delay Mhc protein degradation. Tight regulation of sarcomere breakdown is consistent with opposing changes in the levels of proteins that either comprise or regulate thin filament homeostasis. Monomeric actin subunits (Act57B, Act88F), actin folding chaperones (CCTs) or thin filament-associated proteins (tmod, Tm1, Tm2) are reduced in protein abundance. In contrast, elevated levels of the actin filament depolymerizing factors Cap/CAP and Tsr/Cofilin ensure disassembly of the thin filament, independent of thick filament dynamics.

Supported by these observations, we postulate a model whereby ordered, sequential steps lead to myofibril breakdown. In this model, disassembly of the thin filament precedes the destruction of thick filament proteins. It is likely that accessory proteins, such as ubiquitin ligases, proteases and/or chaperones may facilitate the removal of proteins from intact myofibrils for subsequent degradation (Aweida and Cohen, 2021). Our data show upregulation of a number of putative proteases (dipeptidyl aminopeptidase III, tripeptidyl-peptidase II, puromycin sensitive aminopeptidase) and valosin-containing protein (VCP)/p97, an ATPase complex that is required during mammalian muscle atrophy for the accelerated degradation of muscle substrates (Piccirillo and Goldberg, 2012). It would be valuable to extend the genetic analysis in the D. melanogaster DIOM model to further examine the temporal and spatial contribution of thick filament, thin filament and additional chaperone proteins [CTT1-4,6-8, l(2)efl/CryAB, Hsp60A, etc.] to further define our atrophy model.

Mammalian skeletal muscle atrophy is triggered by numerous, diverse pathological conditions, including renal failure, diabetes, sepsis and cancer cachexia (Cohen et al., 2015; Ebert et al., 2019; Gao et al., 2018; Piccirillo et al., 2014). Normally required to maintain the balance between anabolic and catabolic pathways, elevated levels of pro-inflammatory cytokines, such as tumor necrosis factor-α (TNF-α), interleukin (IL)-1, IL-6 and myostatin, leads to the destructive breakdown of skeletal muscle proteins. Multiple studies support a role for the mammalian steroid hormone glucocorticoid as a general mediator of muscle wasting in multiple disease states to promote muscle protein breakdown (Braun and Marks, 2015; Menconi et al., 2007; Schakman et al., 2008). Structurally similar to glucocorticoids, the arthropod steroid hormone ecdysone developmentally regulates muscle atrophy (Hegstrom et al., 1998; Schwartz and Truman, 1982; Zirin et al., 2013). An additional shared feature between developmental atrophy and glucocorticoid-induced mammalian muscle atrophy is that loss of muscle mass does not affect normal physiological properties such as tetanic contractions, excitation–contraction coupling, or the amount of force generated per cross-sectional area (Laszewski and Ruff, 1985; Ruff et al., 1982; Schwartz and Ruff, 2002).

Two evolutionarily conserved biological processes that control the degradation of cellular material during muscle atrophy are the ubiquitin–proteasome system (UPS) and the autophagy–lysosome system (ALS) (Bonaldo and Sandri, 2013; Dikic, 2017; Khalil, 2018; Kitajima et al., 2020). Many genes that comprise the UPS or ALS are classified as so-called atrogenes and are upregulated in various catabolic conditions (Taillandier and Polge, 2019). Blocking the UPS in numerous mammalian studies reduces muscle atrophy associated with pathological conditions (Kandarian and Jackman, 2006; Tawa et al., 1997). The atrogenes MurF1 and Atrogin-1/MAFbx were the first two muscle-specific ubiquitin E3 ligases to be discovered that are upregulated in nearly all catabolic conditions tested thus far (Bodine and Baehr, 2014; Gomes et al., 2001). Additional E3 ligases, including TRIM 32, Cbl-b, TRAF6 and UBR4, also play essential roles in the prevention of muscle wasting (Cohen et al., 2012; Nakao et al., 2009; Paul et al., 2010). As in mammals, developmental atrophy in insects is associated with upregulation of all the components of the UPS system (Reynolds et al., 1995; Haas et al., 1995, 2007; Hastings et al., 1999; Jones et al., 1995; Löw et al., 1997; Schwartz et al., 1990). Here, we confirm that the Atrogin-1 and Trim32 E3 ubiquitin ligases play a similar role in regulating developmental atrophy. As one of the major ubiquitin ligases, MurF1, is missing in the insect system (Piccirillo et al., 2014), there may be other novel proteins that replace MurF1 function during developmental atrophy.

Both TRIM32 and TRAF6 have been recently linked to autophagy through regulation of Unc-51-like autophagy activating kinase (ULK1)/Autophagy related 1 (ATG1) (Di Rienzo et al., 2019; Paul et al., 2012), part of a complex that initiates autophagy (Dikic, 2017). Additional FOXO-mediated atrogenes that regulate autophagy during muscle atrophy include Cathepsin L, LC3 and GabarapL1 (Lecker et al., 2004; Zhao et al., 2007). Similarly, developmental atrophy in M. sexta is marked by a severe increase in the number of both lysosomes and autophagic vesicles (Beaulaton and Lockshin, 1977; Schwartz, 2008). This functional requirement has been substantiated in the Drosophila DIOMs, whereby blocking the function of five different autophagy-related genes, Atg5, Atg9, Atg12, Atg17 and Atg18, reduced developmental muscle atrophy (Kuleesha et al., 2016a). Although mammalian and insect studies have highlighted the importance of the autophagy/lysosomal pathway in muscle atrophy, there are still a lot of unanswered questions to further understand the connection between the UPS and the ALS. The DIOMs provide an excellent model for further genetic analysis and/or to perform suppressor/enhancer screens to identify new components that contribute to muscle atrophy. It would also be interesting to test whether actin regulatory proteins, such as Cib, play a role in mammalian atrophy.

The global MS approach we undertook in this study did not assay for post-translational modifications. Thus, it is possible that altered protein functions (e.g. phosphorylation, ubiquitination, etc.), without changes in overall protein levels, were not detected. Another limitation was analysis of only two time points: day 15 and day 18. Time points subjected to high-resolution RNAseq by the Schwartz group were collected each day of M. sexta pupal development, from day 13 to day 18, and thus provide a more comprehensive view of gene regulation during initiation, progression and the conclusion of atrophy (Tsuji et al., 2020). In the wild, M. sexta pupate in the ground and their circadian clock is entrained by temperature. It is possible that the 12 h:12 h photoperiod and constant temperature used for developmental coordination in the laboratory setting contributed to individual variability observed among biological replicates. An additional caveat is over-interpreting the present MS data to make conclusions about the relative classes of protein that are upregulated or downregulated. For example, if the contractile proteins are preferentially degraded during atrophy and the same mass of ISM tissue is present in each experiment, there could be an automatic enrichment of cytoplasmic and/or nuclear proteins as well as components of the endomembrane system.

In conclusion, work presented here further highlights some parallels between developmental atrophy in insects and disease-induced mammalian atrophy. Thus, a combination of studies in Manduca and Drosophila will provide valuable insights for understanding the mechanism regulating muscle atrophy in mammals.

A special thank you to Mike Kanost for providing M. sexta specimens and to Lisa Brummet for assistance with M. sexta pupal staging. Thank you to Dr Steve Hartson and Janet Rogers at the Recombinant DNA/Protein Resource Facility at Oklahoma State University for technical advice and assistance with the MS experiments. We are grateful to the BDSC (https://bdsc.indiana.edu/) and VDRC (www.vdrc.at) for providing fly stocks. The BDSC is supported by a grant from the Office of the Director of the National Institutes of Health under Award Number P40OD018537.