Modeling exercise using optogenetically contractible Drosophila larvae

The pathophysiological effects of a number of metabolic and age-related disorders can be prevented to some extent by exercise and increased physical activity. However, the molecular mechanisms that contribute to the beneficial effects of muscle activity remain poorly explored. Availability of a fast, inexpensive, and genetically tractable model system for muscle activity and exercise will allow the rapid identification and characterization of molecular mechanisms that mediate the beneficial effects of exercise. Here, we report the development and characterization of an optogenetically-inducible muscle contraction (OMC) model in Drosophila larvae that we used to study acute exercise-like physiological responses. To characterize muscle-specific transcriptional responses to acute exercise, we performed bulk mRNA-sequencing, revealing striking similarities between acute exercise-induced genes in flies and those previously identified in humans. Our larval muscle contraction model opens a path for rapid identification and characterization of exercise-induced factors.


Introduction
Exercise and increased physical activity can both prevent and ameliorate the pathophysiological effects of a number of metabolic and age-related disorders. In particular, physical inactivity contributes significantly to age related loss in muscle mass and strength, is strongly associated with the onset of type 2 diabetes and cardiovascular disease, and is a reliable predictor of all causes of mortality [1][2][3][4][5][6][7][8]. Not surprisingly, increased physical activity and exercise remain some of the most prescribed interventions for fighting metabolic and age-associated diseases [3][4][5][9][10][11][12]. Increased physical activity can prevent to some extent age-associated loss in muscle strength, reduce mortality rates, improve cardiac function, and improve insulin sensitivity [3-5, 12, 13]. Additionally, exercise is beneficial to various pathologies including inflammatory disorders, certain cancers, and osteoporosis [10,[14][15][16]. Despite the well documented beneficial effects of increased physical activity, our understanding of the molecular mechanisms that mediate these beneficial effects remains limited.
Several "omics" approaches have been reported that exercise modules can induce different myokines [18], suggesting that muscles respond differently depending on the type of exercise. Such observations warrant a deeper investigation into how different myokines exert their beneficial effects on various organs. To date, a number of vertebrate species including humans have been recruited to characterize circulatory myokines in response to different exercise modules. However, in these studies it is difficult to take a systematic look at every myokine in an in vivo setting. The availability of in vivo genetic model systems would be extremely valuable to characterize the response of muscles to varying degrees of exercise. Drosophila is an established model system to understand the biology of exercise (see reviews [25][26][27]). For instance, exercise training in young flies can improve age-associated decline in mobility [28]. Of note, the fly ortholog of the vertebrate exercise response gene dPGC-1α / spargel (srl) has been implicated in the physiological effects of endurance exercise [29], suggesting an evolutionarily conserved mechanism of exercise biology. Interestingly, it has also been shown that activating octopaminergic neurons or feeding flies with octopamine can fully substitute for exercise in sedentary adult flies [30]. Therefore, the use of a genetically tractable model organism such as Drosophila may allow fast and highly reproducible functional analyses of various cell types, tissues, and genes encoding evolutionarily conserved myokines in vivo. However, most of the exercise studies have been modeled around 'endurance'-based training in adult Drosophila. The physiological effects of strength-based or acute exercise in Drosophila adults or larvae have not been well addressed. Importantly, besides addressing the physiological aspects of exercise, a systematic evaluation of the transcriptomic architecture of muscles upon exercise in Drosophila has not been reported.
Here, we used Drosophila larvae to establish an optogenetic muscle contraction model (OMC). Because larval locomotion is brought about by a series of muscle contractions which are regulated by motor neurons that innervate the body wall muscles [31][32][33][34], we employed an optogenetic tool to activate motor neurons. Optogenetics is a powerful non-invasive tool to systematically control neuronal activity, thereby controlling their effects on larval locomotion, behavior, and various other physiological aspects in Drosophila [35][36][37]. Typically, optogenetics utilizes various light-activated channelrhodopsins (ChR) including CsChrimson, ChR2, ChR2-XXL, ReaChR, GtACRs, and many others that utilize different wavelengths of light to activate or inhibit neurons [38][39][40][41][42][43]. Owing to its extreme light-sensitivity and deep tissue targeting, we used the red shifted ChR called CsChrimson [36,42,44] to induce muscle contractions in larvae by activating motor neurons using the motor neuron specific OK6-Gal4 driver [45].
Compared to optogenetics in adult Drosophila, our larval OMC model has many advantages including: 1) larvae are translucent under white light, such that light can pass through the larval cuticle, allowing easy access to motor neurons present in the central nervous system; 2) the larval cuticle is soft, making it possible to induce contraction of all of the body wall muscles at once without causing injury; and 3) larval hemolymph and tissues are easily accessible for biochemical, cell biological, and omics-based approaches. Exploiting such benefits, we demonstrate the application of OMCs to Drosophila larvae and identified several evolutionarily conserved myokine-encoding genes by RNA-sequencing and the signaling pathways that are active in muscles after acute exercise regimens. Further, we identified striking similarities between exercise-induced genes in flies and in humans. We catalog several acute exercise-induced putative myokine-encoding genes, which may aid in the systematic characterization of their roles in beneficial effects of exercise and in addressing the nature of interorgan communication networks during exercise.

Fly stocks and husbandry
The OK6-Gal4 bc line was generated by backcrossing the original OK6-Gal4 (BDSC:64199) line to a w 1118 control fly line for 7 generations. The UAS-CsChrim bc line was generated by backcrossing the original UAS-CsChrim (BDSC:55136) to the same w 1118 fly line for 7 generations. All stocks were maintained at 25 °C and experiments were performed at room temperature unless specified otherwise. The endurance training paradigm was performed at 20 °C to prolong the larval stage, allowing for a longer regimen of muscle contractions. Egg laying cages were set up by crossing ~ 150 virgin females and ~ 120 males in 100 ml volume cages. Crosses were maintained on grape juice plates supplemented with yeast paste and were kept at 25 °C under 60% humidity. Egg collections were started from day 4 onwards post mating. Collections were done for a 3-hour time period every day with a small dab of yeast paste on the grape juice plates. The dab of yeast paste was removed after egg collection and the plates were incubated upside down at 25 °C until larvae started hatching. First instar larvae were collected 24 hours after the end of egg collection and either 60 animals were transferred to standard food vials (for acute exercise training) or 30 animals were transferred to all-trans retinal (ATR) containing food (for endurance training).

Optogenetic muscle contraction (OMC) model using Drosophila larvae
Our setup consists of a series of high-power light emitting diodes (LEDs) that are controlled by an Arduino microprocessor controller via a number of Buckpuck power regulators and switches. Details on the components used to make the OMC setup are available in Table 1 and the circuit diagram for the setup is shown in Fig. 1A (also see Fig. S1A for the entire setup). Adapter rings that allow attachment of standard narrow Drosophila food vials without interrupting air flow were made by mounting laser-cut acrylic pieces around the LEDs. Scaled diagrams for the adapter ring setup are depicted in Fig. S1B-B′. Optogenetic muscle contractions were performed on animals that were either kept in 30 mm cell culture plates in a thin layer of 2% sucrose or in food vials containing 2-3 mm of food per vial. The adapter ring setup allows attachment of up to 12 vials in a single experiment.
Our standard workflow for an OMC experiment is represented in Fig. 1B. Briefly, Ok6-Gal4 bc /+; UAS-CsChrim bc /+ (OK6-Gal4 bc > CsChrim bc ) embryos were collected for a window of 3 hours on grape juice plates. Twenty-four hours after egg lay (AEL), 60 first instar larvae were transferred to standard food vials that were wrapped in aluminum foil to prevent exposure to light and reared at 25 °C for 3 days. The evening before induction of OMCs, mid third instar larvae were floated out of the food using 20% glycerol and 30 size matched larvae were transferred to new food vials containing 100 μM ATR, a substrate for Channelrhodopsin. Next, the vials were wrapped in foil and the larvae reared at 25 °C overnight. The following morning, size matched experimental and control larvae were again selected and placed in 30 mm cell culture plates for inducing OMCs.

Tissue dissections Body wall muscle
Dissections were done in 9-well depression plates with the wells half filled with silicone elastomer to allow positioning of the larvae and fillet using insect pins. Dissections were started by adding 1 ml of relaxing buffer [(RB) 1x PBS, 5 mM MgCl 2 and 5 mM EGTA] to the well and placing a washed and cleaned 3rd instar larvae in the well. The animal was positioned ventral side up and the head and tail of the animals were pinned using insect pins with a gentle stretch along the body. A pair of fine micro scissors was used to cut along the ventral midline of the larvae. A 1 ml pipette was used to squirt the RB inside the animals until all the muscles were visibly relaxed. Four additional insect pins were used to open the fillet and lay the larval cuticle flat. Internal organs were carefully removed making sure not to injure the muscle attached to the cuticle. Samples were fixed in 4% paraformaldehyde in RB for 20 mins in the same well while the cuticle was still pinned to the silicone bed. Post fixation, the insect pins were removed, and the fillets were transferred to low adhesion tubes for further processing. For experimental larvae that show a strong contraction under the white light, we followed the following additional steps. First, we used a lower intensity of white light to do the dissections. Second, we waited at least 30 seconds to allow the larvae to relax after the initial strong contraction under the white light. Third, after the incision along the ventral midline and once the RB had relaxed all the muscles, we readjusted the insect pins to stretch out any slack that was caused by relaxation of the body wall muscles.

Fat body
Dissection was performed in a deep-well dish with the well half filled with silicone elastomer. A 20 μm diameter insect pin that was bent by 90 degrees was inserted in the silicone to create a hook that was used to invert the larvae. One milliliter of RB was added to the well, and stage and size matched subject animals were washed and placed in the well. The posterior end of the animal was cut with a pair of micro scissors and two sharp forceps were used to gently invert the animal on the insect pin keeping all internal organs to one side of the body. Subsequently the gut was removed while making sure other organs were not damaged. Tracheae were gently pulled out by holding the cut end of the primary tracheal tubes and pulling them along the length of the body. The carcass with the fat body still attached was then unmounted from the insect pin and transferred to RB in non-adhesion tubes. Samples were fixed in 4% paraformaldehyde in RB for 20 mins before use for subsequent processing. Final dissection of the fat body was performed on slides prior to mounting.

Oenocytes
Sample preparation for oenocytes was performed identically to muscle preparation with the exception that animals were mounted with the dorsal side facing up and the incision was made along the dorsal midline.

Antibody staining
Third instar muscle fillets were permeabilized in 0.1% Triton-X100 in PBS (PBT) for 30 mins. Samples were then blocked in PBT + 5% BSA blocking solution (BS) for 2 hours at RT. Subsequently primary antibody (antiglycogen Mouse anti-IgM: 1/100, anti-ATP5α: 1/500) diluted in BS was added to the samples. Samples were then incubated over night at 4 °C with gentle rotation.
The following day, the samples were washed in PBT 4 times (15 mins each). Samples were then incubated with an appropriate secondary antibody (with 1/500 dilution) The circuit components allow rapid switching of the high-power LEDs using a microprocessor controller such as an Arduino board. (BOTTOM) Schematic of the heat sink, to which the LEDs were attached, and the attached assembly used as a holder for the larval vials. Holders were assembled from laser-cut acrylic pieces and are designed to prevent light leak between tubes while allowing air flow to support oxygen supply. B Schematic of a basic experimental set up for inducing optogenetic contractions in third instar larvae. The images show snapshots of larvae over-expressing CsChrimson in their motor-neurons (OK6-Gal4 > CsChrim) when the LEDs were either OFF or ON, demonstrating the synchronized contraction of the larval body wall muscles (see Video 1 for the larval OMC). L1 and L3 correspond to first and third larval instar stages, respectively, and ATR is all-trans retinal in BS for 2 hours at RT. Next, the samples were washed again with PBT 4 times (15 mins each) and mounted in VectaShield with DAPI (VectaShield 1200). Samples were mounted on a slide with the body wall muscles facing a coverslip. The coverslip was bridged to avoid squishing of the sample by sandwiching two stacked pieces of scotch tape (3 M) between the coverslip and the slide on both ends of the coverslip. Samples were imaged using a Zeiss LSM780 confocal microscope. Images were acquired at RT.

BODIPY staining of oenocytes and fat body
Dissected and fixed tissues were permeabilized in 0.1% PBT for 30 mins at RT. Samples were washed three times, 5 mins each, in PBS to remove all Triton-X100. A 1 mg/ ml stock of BODIPY in DMSO was diluted at 1/500 in PBS and added to the samples. Staining was performed for 30 mins at RT with constant rotation. The samples were then washed with PBS three more times, 5 mins each, and mounted on slides in VectaShield with DAPI (VectaShield 1200). The coverslip was bridged to avoid squishing of the sample by sandwiching two stacked pieces of scotch tape (3 M) between the coverslip and the slide on both ends of the coverslip.

RNA sequencing and analysis
Total RNA was extracted from larval cuticle preparations using a Qiagen RNAeasy mini prep kit as per the manufacturer's instructions. RNA-sequencing (RNA-seq) on mRNA-selected samples were performed by the Sulzberger Columbia Genome Center, Columbia University, using an Illumina HISEQ 2500 instrument. We obtained approximately 30 million single-end reads per biological replicate and three independent biological replicates were used for both control and exercised conditions. The quality of the sequencing reads was checked using the FASTQC package and reads were mapped onto the Drosophila genome (release 6) using the STAR aligner package. The read counts were calculated using feature-Counts. The subsequent read counts matrix was further processed in R software using the Deseq2 package for normalization and differential gene expression analysis. Principal Component Analysis (PCA) was performed by plotPCA function from Deseq2 package using top variance genes. The heatmap for Fig. 3B was generated by R package pheatmap. The p-value was adjusted (padj) for multiple testing using the Benjamini and Hochberg method and the criteria used for differentially expressed gene selection are padj< 0.05 and fold change > 1.5. We used Gprofiler2 and cluster profiler to measure gene ontology (GO) enrichment and the resulting illustration was done by emapplot function (Figs. 3D, 4B) or by cnetplot function (Fig. 4C, S3B and S4B) from clusterProfiler package. Gene interaction networks in supplemental Figs. S3C and S4A were visualized using Cytoscape [46] with the edges for protein-protein or genetic interactions derived from MIST [47]. The annotation of the core components of signaling pathways was obtained from GLAD [48] and enrichment analysis was done using a program written in house based on hypergeometric distribution. The heatmap illustration of pathway enrichment result was done using TM4 software suite (http:// mev. tm4. org/). For Figs. 5B and S5, we used the gene log2 Fold-Change and pathway information as input and plotted the bar plot using ggplot2 package and used coord_polar function to convert the standard bar plot to the circle plot. The raw RNA-seq reads have been deposited in the NCBI Gene Expression Omnibus (GEO) database with the accession number GSE196850.

Modeling acute exercise using OMC in Drosophila larvae
There are two modalities of physical exercise, endurance training and acute training, which differ in the intensity and duration of the regimens [24]. Endurance training involves low intensity but highly repetitive bouts of training regimens, whereas acute or strength-based exercise typically involves a short duration of high intensity bouts of exercise [49]. Most studies in Drosophila focused mainly on endurance-based exercise regimens using adult flies [26] and work addressing acute exercise is relatively scant. Hence, we reasoned that our OMC model could be utilized to mimic acute exercise in larvae, which are ideal for passing light through the translucent cuticle for optogenetics and subsequent induction of body wall muscle contractions. Because acute exercise training is based on the intensity and frequency of exercise bouts, followed by depletion of muscle glycogen [50,51], we intended to induce acute exercise by inducing rapid muscle contractions and use muscle glycogen as a readout to determine the acute exercise program in larvae. For measuring glycogen levels, we used a previously validated immunofluorescence-based method to visualize muscle glycogen [52,53]. For re-validating the anti-glycogen antibody in our experimental setup, we first tested the specificity of the antibody after treating the body wall muscles with amyloglycosidase, an enzyme known to deplete glycogen [52]. As expected, anti-glycogen antibody failed to detect any glycogen in the muscles that were treated with amyloglucosidase, suggesting that the antibody-based method to detect glycogen is reliable (Fig. S2A-B). Next, we induced an acute exercise program with a series of repetitive muscle contractions over a period of 2 hours. Briefly, size matched mid-third instar experimental (OK6-Gal4 bc > CsChrim bc ) and control (w 1118 ) larvae were kept on ATR containing food overnight. On the day of the experiment, the animals were floated out of food using 20% glycerol and transferred to 30 mm petri dishes that contained 2 ml of a 2% sucrose solution in PBS. It is important at this stage to wipe the sides of the plates with a cotton swab dipped in acetone to create a hydrophobic surface that prevented formation of a concave meniscus. The above two conditions prevented the animals from wandering off the plate while making sure that they did not drown in the liquid. Representative images and the video of animals that are either relaxed (LED off ) or are contracted (LED on) in response to 660 nm light stimulation are shown in Fig. 1B and Video 1, respectively. The acute exercise paradigm consisted of 8 rapid contractions, consisting of 1 sec each of light stimulation interspersed by 2 seconds of no light stimulus, followed by a 30 second break (rest). This cycle was repeated 3 times before giving the animals a 5 min of rest period. The entire exercise cycle was repeated in a loop for 2 hours (see schematic in Fig. 2A). At the end of the acute exercise paradigm the larvae remained inactive and devoid of any contractile action for a short period of time. The animals were, however, alive as evidenced by a beating heart tube. When transferred to fresh food and dark conditions, the exercised animals recovered and developed to pupae and then adults. Interestingly, adults that emerge from these pupae failed to expand their wings, a phenotype that is very similar to what is seen in flies that are mutant for the genes bursicon (burs) and partner of bursicon (pburs) [54,55]. Burs and pburs encode cysteine knot proteins that dimerize to form the neurohormone Bursicon that regulates cuticular pigmentation and wing expansion in adult flies. These genes are produced and released from the CCAP neurons in the fly brain and exert their effects on a distinct set of neuronal network that regulate the complex behavior of wing expansion [56]. The OK6-Gal4 driver used in our experiments is expressed in a subset of the CCAP neurons and the similarity in phenotypes indicate that the activation of these neurons possibly impairs the production of Bursicon or leads to premature release of Bursicon, leaving the adult flies devoid of Bursicon and hence incapable of inflating their wings [57]. This observation highlights one of the potential drawbacks of our OMC model, i.e., that non-specificity of the neuronal driver might cause systemic physiological effects by impacting neuro-endocrine signals.
To determine the physiological impact of our acute exercise paradigm, we monitored muscle glycogen levels in control and exercise animals using the anti-glycogen antibody-based method of detecting glycogen reserves. Relative to control animals, we observed a strong and significant decrease in muscle glycogen levels in exercised animals ( Fig. 2B-C). Interestingly, when we changed the exercise regimen to endurance-like training over a period of 72 hours (Fig. S2C), we observed that the muscles tend to accumulate glycogen levels, in contrast to those of acute exercised animals (Fig. S2D, E). This differential depletion or accumulation of glycogen levels is quite similar to that of muscle glycogen levels in humans depending on the exercise regimen [58,59], suggesting that our larval OMC model is relevant in the context exercise physiology in humans.
The acute exercise regimen also led to a significant reduction in the size of lipid droplets in the larval adipose tissue, indicating lipid mobilization from the lipid stores (Fig. 2D). Interestingly, whole animal triacylglycerol (TAG) levels did not change in response to exercise, suggesting that the change in lipid droplet size reflects a switch to a lipid mobilization mode whereby large lipid droplets are broken into smaller ones, and not necessarily an overall lipid depletion. Another hallmark of lipid mobilization in Drosophila larvae is the accumulation of lipid droplets in the hepatocyte-like oenocytes [60]. Induction of lipid mobilization from the larval adipose tissue, either by starvation or adipose tissue specific over-expression of the brummer lipase, leads to immediate accumulation of lipid droplets in the oenocytes [60]. While the exact mechanism behind this shuttling of lipids is not well characterized, a number of recent studies indicate the pivotal role of the oenocytes in lipid partitioning and systemic regulation of lipid metabolism [61,62]. Interestingly, larvae subjected to the acute exercise regimen also showed a significant increase in deposition of lipid droplets in the oenocytes, further demonstrating that lipid mobilization is induced in response to acute exercise (Fig. 2E-F). It is known that acute exercise may also lead to a metabolic flexibility where nutrient stores are repartitioned from storage sites such as the liver and adipose tissue to help fuel skeletal muscle activity [63]. These physiological changes are known to trigger changes in gene expression that are associated with acute exercise [7]. For instance, acute exercise is known to dramatically increase the expression of interleukin-6 (IL6) in the muscle, one of the very first exercise-induced myokines identified in vertebrates [64] and a known activator of JAK/STAT pathway [65]. There are three cytokines in Drosophila: unpaired (upd) 1, upd2 and upd3, which are known to activate the JAK/STAT pathway in Drosophila [66,67]. Furthermore, upd3 has been recently described as a cytokine with homology to interleukin-6 [68]. Hence, we tested if the expression levels of upd genes are regulated following acute exercise. We dissected the body wall muscle/cuticle complexes of control and exercised larvae and measured the levels of upd1, upd2 and upd3 using qRT-PCR and observed a dramatic increase in the expression levels of upd2 in response to acute exercise (Fig. 2G). Expression of upd3 also increased significantly but was not as pronounced as upd2, and the levels of upd1 remained unchanged (Fig. 2G).
Taken together, these observations support the validity of the OMC model to mimic acute exercise-like phenotypes in Drosophila larvae.

Transcriptional responses in the larval muscle in response to acute exercise
Exercise is known to induce robust transcriptional responses within skeletal muscles [7,69] and proper assessment of such transcriptional changes may depend on the timing, duration, and gender after the exercise bouts [70][71][72]. To identify gene expression changes that respond immediately to an exercise bout, we chose to assess the transcriptional responses that happen in the body wall muscles in response to our larval acute exercise regimen. Note that in order to maintain consistency across all experiments performed in endurance and acute exercise regimens, we did not separate male and female larvae. While sexing larvae is easily doable for third instar larvae that were used for acute exercise, sexing first instar larvae used for endurance training is difficult. Thus, we performed RNA-seq on polyA-selected mRNA derived from dissected larval body wall muscles from mixed-gender control and exercised animals at the end of the 2 hours exercise regimen (Fig. 3A). Principal component analysis (PCA) of the transcriptome data shows the separation of the control samples from the samples of exercised animals (Fig. S3A). The analysis of the differentially expressing genes revealed a total of 2004 genes that changed significantly between the control and experimental samples with adjusted P value < 0.05. Of these, 1190 genes showed a change of 1.5 fold or more, with 672 genes showing an upregulation of 1.5 fold or more and 518 genes showing a downregulation of 1.5 fold or more (Supplementary Table 1, Fig. 3B). A heatmap representing these 1190 significantly changing genes show some variability in their expression within the control and experimental replicates (Fig. 3B), which may have led to the identification of a lower number of significantly up or downregulated genes than expected. Nevertheless, we did see a significant increase in the expression of both upd2 and upd3 as expected from the qRT-PCR results (Fig. 2G). To obtain a systems-wide understanding of the biological processes affected by our acute exercise models, we performed a Gene Ontology (GO) enrichment analysis of the genes that showed a statistically significant fold-change of 1.5 or more in expression (Fig. 3C). Additionally, we looked at the relationship between the top enriched GO categories and found that these categories primarily mapped to processes such as regulation of growth and cellular component size, response to stimuli or signaling pathways, and response to starvation (Fig. 3D). Additionally, we identified some significantly enriched GO categories that have physiological relevance to exercise. These categories include processes such as response to wound healing, response to oxidative stress, response to glutathione synthesis, and cellular response to unfolded protein (Fig. S3B). Interestingly, we also identified 'open tracheal system development' as one of the categories enriched in our dataset (Fig. S3B). The tracheal system in flies is analogous to the respiratory tract or lungs that includes trachea in vertebrates [73,74]. It has been shown how various intensities of swimming exercise contribute to lung development and tracheal reactivity in humans and rodents, respectively [75,76]. Thus, genes identified in this category may provide insights into the mechanisms of how trachea and the respiratory tract in general are affected by various exercise modalities in humans.
As noted above, our exercise regime induced changes in expression of genes that are involved in signaling pathways, growth, and cellular compartment size (Fig. 3C-D,  S3C). These genes most likely represent early responses that lead to physiological changes such as increase in muscle mass, mitochondrial density, and vascularization. Relatedly, the genes that comprise the GO terms related to growth, cell signaling, and tracheal system development might play roles in manifesting these exercise-induced changes. Interestingly, multiple GO terms related to 'morphogenesis' were also over-represented in our dataset (Fig. 3C-D). Some of the upregulated genes in this category include vg and Mrtf (Table S2), the human orthologs of which (VGLL2 and MRTF, respectively) directly relate to exercise in humans [7,77]. In holometabolous insects, critical weight is a developmental represents the adjusted p-values of the enrichment analysis relative to the other displayed terms (brighter red is more significant) and the size of the terms represent the number of genes that consist the term. D An enrichment GO plot showing the relationship between the top 25 most significantly enriched GO terms based on adjusted p-value. Similar terms are grouped together. The color represents the adjusted p-values of the enrichment analysis relative to the other displayed terms (brighter red is more significant) and the size of the terms represent the number of genes that consist the term checkpoint associated with the third instar larvae and determines whether the animal has enough energy reserves to survive metamorphosis [78,79]. We used mid third instar larvae for the acute exercise experiments. These animals had already cleared the critical weight checkpoint. When starved or stressed, animals that have crossed the checkpoint hasten entry into metamorphosis [79,80]. Since our animals were exercised in petri dishes that only contained a small volume of 2% sucrose, the animals may have been starved and thus initiated entry into metamorphosis. Although both our control and experimental animals were kept under identical conditions, the additional stress of being contracted could have triggered a stronger starvation response in these animals. Further reducing the duration of the acute exercise paradigm and/or using pre-critical weight larvae might mitigate these changes in metamorphosis genes.

Analysis of genes that are upregulated by the acute exercise regimen
To determine whether specific biological processes were activated or suppressed in response to exercise, we further separately analyzed genes that were either up-or down-regulated in response to the acute exercise paradigm. Interestingly, GO term enrichment analysis of the 518 downregulated genes yielded very few significantly enriched GO term categories. For example, enrichment analysis of downregulated genes using the widely used GO profiler algorithm identified only 3 broad categories: ion binding; membrane-related genes, and metabolic genes. Similarly, analysis of these genes using another GO term enrichment algorithm, cluster-Profiler [81], yielded a single enriched category: drug catabolic process. In contrast, analysis of the up-regulated genes revealed a number of significantly enriched terms. The top 25 highly enriched GO terms show a significant over-representation of genes involved in wound healing, immune response, glutathione metabolic process, and protein folding (Fig. 4A). Analysis of the relationships among the various enriched GO terms together with their gene interaction networks show that they cluster primarily around processes such as protein folding, amino acid and modified amino acid metabolic processes, and immune response processes (Fig. 4B, S4A). The top 10 enriched GO categories, including the genes that comprise these groups and the degree of upregulation of each of these genes, is shown in Fig. 4C. A similar depiction of genes that comprise an additional six categories that were hand selected for their potential involvement in exercise biology, is shown in Fig. S4B. These categories include innate immune response, negative regulation of immune response, muscle attachment, tube morphogenesis, response to hypoxia, and negative regulation of proteolysis. Some of these processes, such as enhanced wound healing, are known to benefit from regular exercise [82], and there is increasing evidence suggesting that repeated physical exercise improves immunosurveillance and immunocompetence [83]. Furthermore, one category ("response to hypoxia") was enriched for several heat shock protein (Hsp) coding genes, which are known to peak after an initial exposure to heat generated as a consequence of exercise in higher vertebrates (see review [84]). Of note, the human ortholog of fly Hsp70Bc (HSPA1A/Hsp72), is known to be upregulated in response to acute exercise in many organs including skeletal muscle, suggesting that the induction of Hsp genes in response to acute exercise is highly conserved [85]. Moreover, analysis of the genes that are upregulated revealed several secreted protein coding genes that may represent putative 'myokines' (Table 2). Altogether, the gene sets that are activated in response to acute exercise in our OMC model are relevant in the context of human exercise biology.

Effect of acute exercise on major signaling pathways
Genes involved in cellular signaling pathways are expected to play a key role in mediating both cell autonomous and non-autonomous impacts of acute exercise. Indeed, enrichment of terms such as regulation of response to stimuli and regulation of cell communication in our GO term enrichment analysis indicates that the acute exercise paradigm does impact signaling pathway activities (Fig. 3C-D, S3C). We analyzed our data to determine the effect of the acute exercise regimen on expression levels of an exhaustive list of genes that are components of the core signaling pathways in Drosophila. An enrichment analysis looking for representation of the core signaling pathway components amongst all significantly changing genes (log2FC ≤ − 0.6 and ≥ 0.6) show that FGFR signaling and Insulin signaling pathways were most affected (Fig. 5A). Similarly, analysis looking for representation of the core signaling pathway components amongst all significantly upregulated genes (log2FC ≥ 0.6) show that JNK, Imd, and TNF-α signaling pathways were strongly affected (Fig. 5A). The change in expression of 335 core signaling pathway genes in response to the acute exercise regimen were grouped into 15 core signaling pathways. Of these, 10 pathways had at least 1 gene that showed a significant change of 1.5 fold or more (Fig. S5). While many of these genes could be involved in more than one pathway, each gene (with a fold change of 1.5 or more) was represented in only one of the pathways for simplicity in Fig. 5B. Importantly, gene sets that are commonly reported as upregulated

Concluding remarks
In summary, we have established an optogenetic muscle contraction (OMC) model in Drosophila larvae that provides a fast and effective method for induction of acute exercise regimens. Although exercise models have been developed in adult flies, most of the studies mainly focused on the physiological effects of endurance training. Importantly, studies on the molecular pathways that are activated in muscles upon exercise have not been addressed. Hence, compared to existing exercise models in Drosophila, our larval OMC model is unique in 1) genetically inducing exercise in larvae and 2) characterizing muscle-specific transcriptional responses to acute exercise and the molecular pathways that are activated in the exercised muscles. Importantly, the transcriptomic analysis of exercised muscles revealed striking similarities between acute exercise-induced genes in flies and in humans. Altogether, our larval muscle contraction model opens a path for rapid identification and characterization of exercise-induced factors such as myokines.