Trehalose metabolism confers developmental robustness and stability in Drosophila by regulating glucose homeostasis

Organisms have evolved molecular mechanisms to ensure consistent and invariant phenotypes in the face of environmental fluctuations. Developmental homeostasis is determined by two factors: robustness, which buffers against environmental variations; and developmental stability, which buffers against intrinsic random variations. However, our understanding of these noise-buffering mechanisms remains incomplete. Here, we showed that appropriate glycemic control confers developmental homeostasis in the fruit fly Drosophila. We found that circulating glucose levels are buffered by trehalose metabolism, which acts as a glucose sink in circulation. Furthermore, mutations in trehalose synthesis enzyme (Tps1) increased the among-individual and within-individual variations in wing size. Whereas wild-type flies were largely resistant to changes in dietary carbohydrate and protein levels, Tps1 mutants experienced significant disruptions in developmental homeostasis in response to dietary stress. These results demonstrate that glucose homeostasis against dietary stress is crucial for developmental homeostasis.

O rganisms are acutely and chronically exposed to various environmental conditions, such as temperature, humidity, radiation, and nutrition. Despite such perturbations, many organisms can develop relatively consistent phenotypes, due to their intrinsic robustness against environmental changes. The concept of developmental homeostasis refers to the ability of an organism to maintain phenotypic consistency and reproducible outcomes in response to genetic and environmental variations 1 .
Phenotypic variability during developmental processes is determined by two critical components: robustness (including synonymous terms, such as canalization) and developmental stability. Developmental robustness, or microenvironmental canalization, buffers phenotypic variations against genetic and/or environmental conditions 2 . Developmental stability buffers against local and small, random perturbations, termed developmental noise 3,4 . Developmental robustness is commonly evaluated as the phenotypic variations among individuals that share genetic and environmental conditions. In contrast, the degree of developmental stability is evaluated by using fluctuating asymmetry (FA), which is the measure of random deviations from perfect symmetry. Lower inter-individual variation (IIV) for a given trait reflects robustness, whereas reduced FA reflects the capacity of an organism to buffer against stochastic noise generated during development. Although the heritability of FA is typically very low, suggesting that FA has no or very little additive genetic basis, increased developmental stability has been associated with increased fitness 5, 6 .
The genetic basis of developmental homeostasis has been intensively studied using the wings of the fruit fly Drosophila melanogaster as a model system 7,8 . Both robustness and developmental stability have been considered to be polygenic traits whose variation is primarily determined by many genes with small impacts, instead of by few genes with substantial impacts 9,10 . In Drosophila, phenotypic variations can be assessed among genotypically identical individuals grown under controlled nutrient conditions. However, even under controlled conditions, phenotypic variations in bilaterally symmetrical structures can occur, both among and within individuals. The causes of withinindividual variations (e.g., FA) are thought to be associated with the stochastic nature of developmental noise because the left and right structures are generated under the same external and internal environmental conditions, using the same genotypes.
Robust body sizes and symmetry require highly connected genetic networks, which stabilize the precise regulation of cellular growth and proliferation during developent 7,8 . The growth of developing organs is regulated by nutrient-sensitive systemic signals and local signals, such as morphogen gradients 11,12 . Thus, appropriate control of metabolic homeostasis is crucial for postembryonic development in response to dietary fluctuations. Metabolic homeostasis is maintained by highly regulated and intricate feedback systems that ensure the delicate balance among food intake, energy expenditure, and resource storage 13,14 . The evolutionarily conserved insulin-like peptide is the primary nutrient-sensitive anabolic hormone, which promotes the absorption of postprandial blood glucose into several organs and regulates tissue growth 15,16 . In Drosophila, aerobic glycolysis is indispensable for body growth during larval development 17,18 , suggesting that glycemic control is a central component for the maintenance of developmental integrity by providing a constant supply of energy to each organ.
In Drosophila, the dominant hemolymph sugar is the disaccharide trehalose, which acts as a glucose sink in circulation 19,20 . Trehalose is synthesized by trehalose-6-phosphate synthase 1 (Tps1) in the fat body, a Drosophila organ analogous to the mammalian liver. Larvae deficient in trehalose metabolism show developmental delays and growth defects 21 . Moreover, larvae with high-sugar-or high-fat-diet-induced obesity have been reported to manifest metabolic syndromes, such as increased circulating glucose levels, higher fat content, and insulin resistance, and exhibit severe growth retardation 22,23 . Thus, in addition to nutrient-sensitive systemic signaling networks, the homeostatic control of blood glucose is vital for average body growth during development. Despite the importance of glycemic control for metabolic homeostasis, the mechanisms that govern the interplay between glucose homeostasis and developmental homeostasis remain unknown.
Here, we report that trehalose metabolism acts as a metabolic buffer that sustains robust development in Drosophila melanogaster. We found that Tps1 mutants showed feeding-associated hyperglycemia and fasting hypoglycemia and consequently exhibited vulnerability to metabolic perturbations. We further revealed that Tps1 mutants showed increased IIV and FA for adult wing size. Notably, the defects on FA in Tps1 mutants were sharply exacerbated by a low-glucose (LG) diet, whereas a highglucose (HG) diet attenuated the mutant phenotype. However, both HG and low-protein diets worsened the IIV among Tps1 mutants, indicating that environmental changes differentially affected developmental robustness and stability. In conclusion, the present study provides direct evidence that glucose homeostasis impacts developmental homeostasis. Trehalose metabolism likely evolved to maximize developmental homeostasis by buffering glucose fluctuations in response to environmental variations.

Results
Trehalose metabolism functions in glucose homeostasis. To understand the significance of glucose homeostasis on developmental homeostasis, we utilized reverse genetics, which facilitates the direct manipulation of circulating sugar levels (Fig. 1a). To this end, we made use of a hypomorphic Tps1 allele, named Tps1 MI03087 , because null mutations in Tps1 result in complete lethality during the pupal stage 21 . A Minos transposon was inserted into the first intron of Tps1, just before the translational start codon (Fig. 1b), which reduces the mRNA expression level to~20% of the wild-type level in mid-third-instar larvae 21 . We hereafter refer to the Tps1 MI03087 line backcrossed with a control strain w − as Tps1 MIC . The metabolic analysis revealed that the trehalose levels in homozygous Tps1 MIC mutants were reduced to 20% of those in control at the post-feeding wandering stage (Fig. 1c). Glucose, glycogen, and triglyceride (TAG) levels did not change at the level of whole larvae.
Glucose levels fluctuate more dynamically than trehalose levels in the circulating hemolymph in response to dietary challenges 24,25 . To investigate the physiological role of trehalose metabolism in glucose homeostasis during the feeding stage, we determined the absolute concentrations of circulating sugars in mid-third-instar larvae, using liquid chromatography coupled with tandem mass spectrometry (LC-MS/MS). We found that the glucose levels in Tps1 mutants were higher than those in control larvae when fed with a normal diet (ND), containing 10% glucose (Fig. 1d). The increased glucose levels in Tps1 mutants could be reversed by chronic feeding with a diet devoid of glucose (LG diet). In contrast, HG diet containing 20% glucose further increased glucose levels in both control and Tps1 mutant larvae, indicating that circulating glucose levels reflect dietary glucose levels. Tps1 mutants also showed increased levels of the polyol pathway metabolites, sorbitol and fructose (Fig. 1a), whose levels largely corresponded with glucose levels. The diet-dependent changes in trehalose levels were less pronounced than those observed for glucose and fructose.
To further examine the changes induced by acute challenges, we transferred mid-third-instar larvae grown on ND to a wateronly diet for 4 h. The acute deprivation of diet reduced circulating glucose levels, but not trehalose levels, in control larvae (Fig. 1d). Interestingly, Tps1 MIC mutants showed lower glucose levels than control larvae, indicating that Tps1 mutants experienced postprandial hyperglycemia and fasting hypoglycemia. Moreover, acute high-glucose challenges (from LG to 20% glucose diet for 4 h) increased glucose and fructose levels more drastically in Tps1 mutants than in control larvae. Because trehalose is the dominant hemolymph sugar, with concentrations more than 100-fold those for glucose (Fig. 1d) 25,26 , we suggest that trehalose metabolism buffers the fluctuations of circulating glucose levels in response to acute or chronic dietary challenges, playing a crucial role in glucose homeostasis.
Trehalose catabolism cell-autonomously regulates organ growth. We next investigated the consequences of impaired trehalose metabolism on organ growth. Although the underlying mechanisms remain to be clarified, sexual dimorphism for survival rate was observed among Tps1 MIC mutants: approximately 30% of males and 70% of females were eclosed under wellnourished ND conditions 21 . The overall morphology of eclosed mutant flies was indistinguishable from those of control flies in both males and females (Fig. 2a). Detailed analyses revealed that the wing areas of Tps1 MIC mutants were smaller than those of heterozygous flies in both males and females (Fig. 2b), indicating that the chronic reduction in trehalose levels decreases organ size.
Organ size variations result from changes in cell sizes and/or cell numbers. To determine whether the reductions in wing size observed in Tps1 MIC mutants were caused by reduced cell LG, low-glucose diet; HG, highglucose diet; Glc, glucose. Composition of diets is shown in Fig. 5a. *p < 0.05, **p < 0.01, ***p < 0.001; unpaired two-tailed Student's t-test (c), unpaired two-tailed Student's t-test with Bonferroni correction (d). Results are presented as the mean ± SD. The numbers indicate the number of biological replicates.
numbers, cell sizes, or both, we determined the wing hair density. The adult wing is composed of a two-layered epithelial sheet. Because each cell in the wing generates a single hair, hair density is inversely correlated with cell size. The hair density increased in Tps1 mutants compared with that in heterozygous flies in both males and females, whereas the calculated cell numbers did not change (Fig. 2b). Thus, Tps1 MIC mutant wings are small because they contain smaller cells with minimal changes in cell numbers.
We next investigated the cell-autonomous contributions of trehalose catabolism to organ growth using tissue-specific genetic manipulations. The gene locus of the trehalose hydrolysis enzyme trehalase (Treh) produces two types of proteins with distinct localizations through alternative splicing: cytoplasmic Treh (cTreh) and secreted Treh (sTreh) (Fig. 1a). cTreh and sTreh are expressed in various tissues during development and are functionally redundant 21,27 . The knockdown of Treh at the posterior compartment of wing imaginal discs using en-Gal4 did not affect the overall wing morphology of adults (Fig. 2c). However, the knockdown of Treh significantly reduced the size of the posterior region compared with the size of the anterior region (Fig. 2d). In contrast, the overexpression of cTreh, but not of sTreh, increased the size of the posterior region. The knockdown of Tps1 in wing discs had no effect on wing size because Tps1 is exclusively expressed in the fat body 21 and, thus, served as an additional negative control. These results suggest that trehalose hydrolysis affects organ growth in a cell-autonomous manner.
In addition to trehalose, branched polysaccharide glycogen functions as a form of intracellular glucose storage. Larvae deficient in glycogen metabolism manifest various metabolic alterations and growth defects 28 . However, the wing morphologies and sizes in glycogen synthase (GlyS) and glycogen phosphorylase (GlyP) knockdown adults were indistinguishable from those in control flies (Fig. 2c, d), suggesting that glycogen metabolism is dispensable for wing development and morphogenesis under normal growth conditions. Defects in trehalose metabolism reduce developmental homeostasis and fitness. We next investigated developmental homeostasis in Tps1 mutant flies. We examined two components: size variation among individuals (IIV), as a measure of developmental robustness/canalization, and asymmetric size variation within an individual (FA), as a measure of developmental stability. We found that Tps1 MIC mutants showed significantly increased IIV and FA for both sexes compared with genetically matched w − control and Tps1 MIC heterozygous flies ( Fig. 3a-c). Moreover, these changes were canceled by the presence of a single Tps1 genomic rescue construct. These results indicate that homozygous mutations in Tps1 reduce both developmental robustness and stability.
Although FA is not necessarily a general bioindicator of fitness 5,6 , FA in males has been linked to sexual selection in Drosophila 29,30 . Therefore, we next tested the reproductive fitness of Tps1 mutant males by a mating competition assay, in which we crossed a virgin female from a wild-type strain, either Canton S or Oregon R, with age-matched w − control and Tps1 mutant males and then determined the genotypes of the offspring, to identify which males mated successfully. This assay revealed that Tps1 mutant males had a lower mating success rate than the control males (Fig. 3d). In the absence of any competitor,~60% of Tps1 mutant males generated offspring, with a reasonable number of progenies (Fig. 3e). Thus, more than half of the mutant males were fertile, although a noticeable population of males was sterile. Although a causal relationship between reduced developmental homeostasis and reproductive ability remains unclear, these results reveal that Tps1 mutant males show reduction in both reproductive fitness and developmental homeostasis.
In addition to acting as a reserve for glucose homeostasis, circulating trehalose plays potential roles in various biological processes, such as the maintenance of water homeostasis and the regulation of autophagy 27,31 . Moreover, Tps1 can prevent apoptosis in yeast by acting as a "moonlighting" protein 32  determine whether the reduced developmental homeostasis observed in Tps1 mutants was caused by decreased trehalose/ Tps1 levels or by defects in sugar homeostasis, we analyzed hypomorphic Treh mutants because null mutations in Treh resulted in complete lethality during the pupal stage 27 , precluding any analyses of these mutations in adults. We found that homozygous mutations in cTreh (Treh MIC ) showed increased FA in males, similar to the phenotypes observed in Tps1 MIC mutants (Fig. 3b, c). Moreover, Treh MIC mutants showed increased IIV in females (Fig. 3a). In contrast, homozygous mutations in sTreh (Treh s1 ) resulted in no increases in either IIV or FA. Thus, the wing phenotype observed in Tps1 mutants was partially recapitulated by Treh mutant flies, suggesting that appropriate glucose homeostasis during the growth phase plays a role in developmental robustness and stability.
Phenotypic specificity and severity on developmental homeostasis by perturbations of trehalose metabolism. To further examine the specificity of genetic mutations involved in trehalose metabolism on developmental homeostasis, we analyzed two wild-type strains and compared them with a control strain, w − , which we used as a genetically matched control. We found that Canton S and Oregon R showed almost equivalent levels of IIV and FA to those observed for the control w − in both sexes (Fig. 4a, b), indicating that IIV and FA are relatively stable in at least these three genotypes. We next investigated the contribution of glycogen metabolism to developmental homeostasis using GlyS and GlyP null mutants 28 . GlyS mutants slightly increased IIV in females but did not result in increased FA (Fig. 4a, b). In contrast, GlyP mutants displayed no changes in either IIV or FA compared with control flies. These results   indicate that the contribution of glycogen metabolism to developmental homeostasis is limited.
We noticed that the wing area in GlyP mutants was smaller than that in control flies for both males and females (Fig. 4c). Because GlyP mutants are smaller than controls, as assessed by both pupal volume and adult weight 28 , the size reductions observed in GlyP mutant wings likely occur in a non-cellautonomous manner. Tps1 MIC mutant flies also had smaller wings (Fig. 2b), and the degrees of reduction were comparable to those observed in GlyP mutants. These results imply that the reduction in organ size is unrelated to the reduction in developmental robustness and stability.
We further examined the phenotypic severity of the size deviations observed in the Tps1 mutant flies. To this end, we reevaluated dilp8 mutants, which demonstrated reduced developmental robustness and stability 33 . Dilp8 is a damage-inducible peptide secreted from imaginal discs and is a strong candidate for the long-sought hormone that stabilizes growth across the body and ensures robust symmetry and proportionality 8 . Consistent with previous reports, we observed increased size deviations in dilp8 MIC mutants, although sex differences were also detected in our experimental conditions; dilp8 MIC mutants significantly increased IIV in females but not in males, whereas these mutants increased FA in males but not in females (Fig. 4a, b). The reductions in developmental robustness and stability observed in Tps1 mutants are approximately comparable to those observed in dilp8 mutants, further supporting the conclusion that trehalose metabolism is crucial for developmental homeostasis.
Dietary conditions strongly influence developmental homeostasis in Tps1 mutants. We next examined the impacts of dietary conditions during the larval period on developmental homeostasis (Fig. 5a). Reductions in dietary yeast, a significant source of protein in the diet, decrease the growth rate and result in smaller body sizes 34 . When the amount of yeast was reduced by five-fold (1/5Y), the lethality among Tps1 MIC mutants increased for both sexes, but this effect was not observed for a three-fold reduction in yeast (1/3Y) (Fig. 5b). Compared with flies fed with the ND, control flies fed with reduced-yeast diets had smaller wings and increased IIV in a dose-dependent manner (Fig. 5c, d). Interestingly, under yeast-poor conditions, Tps1 MIC mutants displayed more drastically decreased wing sizes and significantly increased IIV compared with control flies. However, FA did not change in either control or Tps1 MIC mutants under yeast-poor conditions (Fig. 5e), indicating that reduced dietary protein levels worsened size deviations between individuals but not within the individuals. The genotype-by-environment interaction demonstrates that Tps1 mutants are sensitive to environmental perturbations.  We further tested the effects of LG and HG diets. Changing the amount of dietary glucose had limited impacts, if any, on wing size for both control and Tps1 MIC mutant flies when compared with changing the amount of dietary protein (Fig. 5c). We found that control flies did not show changes in IIV or FA when grown on either LG or HG diets, indicating that control flies are mostly resistant to changing levels of dietary glucose with regards to developmental homeostasis. Remarkably, the LG diet drastically increased FA in Tps1 MIC mutant males but not in females (Fig. 5e). In contrast, the HG diet ameliorated the increase in FA observed in Tps1 MIC mutants grown on the ND (Fig. 5e). Importantly, the changes in FA were well-correlated with changes in survival rate; the LG diet sharply increased and the HG diet rescued the lethality in Tps1 mutants (Fig. 5b). However, Tps1 MIC mutants showed significantly increased IIV when grown on the HG diet in both sexes (Fig. 5d), indicating that LG  LG  the HG diet enhances IIV and dampens FA in Tps1 MIC mutants. Collectively, these results strongly support the idea that defects in glucose homeostasis reduce developmental homeostasis in Tps1 mutants.
Asymmetric organ size in Tps1 mutants is caused by changes in both cell size and cell number. To further investigate the cellular basis of the asymmetry observed in mutant flies, we selected two Tps1 MIC mutant individuals that showed drastic deviations between left and right wings when fed the LG diet (Fig. 6a, b). We determined the wing hair densities of five different regions, on both the dorsal and abdominal sides (10 areas in total) of each wing (Fig. 6c). The total cell numbers in each wing were then estimated based on the cell size values for each area. We found that the small-sized wings had smaller cell sizes and fewer cell numbers than the larger wings (Fig. 6d, e). In contrast, randomly selected Tps1 MIC mutants that showed relatively symmetric wing sizes displayed comparable values for both cell sizes and cell numbers between left and right wings (Fig. 6f, g). These results indicate that changes in both cell size and cell number produce the asymmetric size deviations in Tps1 MIC mutants.

Discussion
Understanding the impacts of genotype-by-environment interactions is fundamentally important for unraveling the mechanisms that drive developmental homeostasis. Several genes have been reported to contribute to the developmental robustness and stability in the genetically tractable model organism Drosophila, including microRNAs, heat-shock protein chaperones, the growth regulator dMyc, the transcription regulator CycG, and the interorgan signal dilp8 7,8,35 . Although blood glucose is an essential fuel source that supplies many energy-demanding events, little is known regarding the impacts of metabolic homeostasis on developmental homeostasis. The genetic manipulation of trehalose   metabolism allowed us to directly evaluate the impacts of altered glucose buffering capacities during development. We propose that trehalose metabolism is a noise-buffering mechanism that confers developmental robustness and stability in response to environmental perturbations and developmental noise. The adjustment between cellular growth and proliferation contributes to developmental homeostasis by achieving consistent organ sizes 36,37 . A proteome-wide association study has revealed that adult wing size positively correlates with the protein levels of many glycolytic enzymes in wing discs 38 , suggesting that larger wing discs require increased glucose metabolism. Consistently, we found that the level of trehalose hydrolysis impacts organ size, although the observed effects are relatively mild. The tissue-specific knockdown of glycolytic genes in eyes and wing imaginal discs have been reported to have little, if any, impacts on normal morphogenesis and organ size 39,40 , suggesting that other metabolic pathways compensate for reduced glycolysis during normal development. However, reductions in glycolytic activity can attenuate JNKinduced cell death and TGFβ/Hipk-driven tissue overgrowth in imaginal discs 39,41,42 . Thus, glucose metabolism appears to play protective roles against developmental perturbations.
We found that wild-type flies are widely capable of adapting to various dietary conditions with regards to developmental homeostasis, except for the increase in IIV under poor-yeast diet conditions. Reductions in robustness may be caused by competition between individuals over a limited nutrition source. Despite the increase in IIV, FA remained unchanged in control flies following the nutritional variations examined in this study. These results are consistent with previous reports that variations in temperature strongly affect the IIV, but not the FA, of wing sizes in Drosophila 43 . Interestingly, larvae fed with a high-carbohydrate diet display more symmetric sex combs, accompanied by reduced survival rates and slower growth 44 . Moreover, the existence of both independent and shared biological mechanisms for robustness and stability have been demonstrated 9,10,43 . Our results strongly support the view that reductions in developmental robustness do not always coincide with reductions in developmental stability.
In contrast with wild-type flies, Tps1 mutants are vulnerable to nutritional variations during development. Because Tps1 mutants display feeding-associated hyperglycemia and fasting hypoglycemia, we suggest that hypoglycemia is the likely cause of reduced developmental stability, whereas hyperglycemia is the likely cause of reduced robustness. In support of this reasoning, asymmetry in the wing appears to develop during the post-feeding developmental stage, including the pupal stage, which must be completed using internal stored resources and, thus, represents a metabolically vulnerable period. Strikingly, most of the stored trehalose during the larval stage is consumed during early metamorphosis 21 . In Drosophila, the tubular hearts and aorta, which lie along the dorsal side, function to transport body fluid directionally by pumping hemolymph forward in the body cavity 45,46 . Peristaltic muscular movements during locomotion also facilitate hemolymph movement during the larval period. Both sides of the body are assumed to be exposed to nearly identical concentrations of circulating sugars due to the open circulatory system. However, during pupation, part of the larval heart tube is lost, and the remaining cells must be remodeled into the adult heart tube 47 . The remodeling of the circulatory system and peristaltic movement quiescence during metamorphosis may cause local energy supply fluctuations between sides, leading to asymmetric organ sizes due to differences in cell size and cell number. Moreover, subtle differences in glucose availability between sides could amplify developmental noise or dampen noise-canceling mechanisms by disrupting cellular homeostasis.
We detected a statistically significant reduction in IIV among heterozygous Tps1 mutant males and a significant reduction in FA among sTreh mutant females compared with genetically matched w − controls. Because both developmental robustness and stability are polygenic and heterozygosity increases developmental homeostasis 4,7,48 , the observed changes might be caused by genetic background differences near the intended locus, due to incomplete backcross. Alternatively, these genotypes may have sex-specific beneficial effects on developmental homeostasis and organ growth for unknown reasons.
We also detected sex-specific unfavorable effects on developmental homeostasis. For example, the increased FA associated with cTreh and dilp8 mutations were only observed in males, but not females. Likewise, the drastic increase in FA observed in Tps1 mutants fed with the LG diet was only observed in males, although Tps1 mutants demonstrated increased FA in both sexes under ND conditions. A similarly male-biased increase in FA for morphological traits has been reported in Drosophila 10 , suggesting that developmental stability in males may be more sensitive to genetic and environmental perturbations than in females. The overall sexual dimorphism in body size is caused by femalespecific growth rates and weight loss under the regulation of systemic insulin-like growth factor signaling, the cell-autonomous function of the sex determination gene Sex-lethal (Sxl), and dMyc [49][50][51][52] . These female-biased signaling pathways may be involved in the suppression of developmental noise caused by genetic and environmental perturbations. Interestingly, growth impairments, associated with poor glycemic control, are more severe in males than females among children with insulindependent diabetes, although the pathophysiological mechanisms underlying these differences are not fully understood 53 . The relationship between sex-specific developmental homeostasis and glucose homeostasis would be an interesting issue for further research.
Trehalose metabolism has been suggested to play protective roles against various stresses, such as cold and desiccation 20,27 . Thus, determining whether the buffering function of trehalose metabolism on developmental homeostasis is generalizable to non-dietary environmental variations will also be interesting to explore. Further insight into the molecular mechanisms that underly the observed reductions in developmental robustness and stability will reveal how genotypes and environments interact, providing a more comprehensive understanding of developmental homeostasis.
Fly diets. The animals were reared on fly food (normal diet, ND) that contained 8 g agar, 100 g glucose, 45 g dry yeast, 40 g cornflour, 4 ml propionic acid, and 0.45 g butylparaben (in ethanol) per liter (1× recipe). Low/high-glucose and low-yeast diets were prepared, as shown in Fig. 5a. For the acute high-glucose challenge and starvation experiments, shown in Fig. 1d, mid-third instar larvae (approximately 24 h after the second ecdysis) grown on ND or LG diet were transferred to vials containing either 20% glucose and 0.8% agar in H 2 O or 0.8% agar in H 2 O. No yeast paste was added to the fly tubes in any experiments. All the experiments were conducted under non-crowded conditions at 25°C.
Measurement of protein, TAG, and sugar levels. Measurements of protein, trehalose, glucose, glycogen, and TAG levels in whole larvae were performed as described previously 28,54 .
In brief, larvae were rinsed several times with PBS to remove all traces of food. Frozen samples in tubes were homogenized using a pellet pestle in 100 μl of cold PBS containing 0.1% TritonX-100, immediately heat inactivated at 80°C for 10 min, and then cooled to room temperature (RT). Samples were further crushed to obtain uniform homogenates with 1× ϕ3 mm zirconia beads using an automill (Tokken Inc.) at 41.6 Hz for 2 min. The homogenate samples were used to determine the TAG and glycogen levels, and the cleared samples after centrifugation at 12,000 × rpm (13,000 × g) at RT for 10 min were used to determine the trehalose and glucose levels. Ten μl of the homogenate was mixed with 10 μl of a triglyceride reagent (Sigma-Aldrich), incubated at 37°C for more than 30 min, and then cleared by centrifugation. Five μl of the supernatant was used for the measurement of TAG by free glycerol reagent (Sigma-Aldrich). A triolein equivalent glycerol standard (Sigma-Aldrich) was used as the standard. Five μl of the sample was incubated with PBS containing amyloglucosidase (Roche) or bacterially produced recombinant His-tagged cTreh at 37°C overnight. Ten μl of the sample was incubated with PBS without enzymes in parallel for the determination of glucose levels. The reaction was carried out in a 15 μl assay mixture. The amount of glucose was determined by a glucose assay kit (Sigma-Aldrich). A serial dilution of glucose was used as standard. The trehalose and glycogen concentrations for each sample were determined by subtracting the values of free glucose in the untreated samples. The amounts of TAG, trehalose, glycogen, and glucose were normalized to the total protein level as described below.
For determination of total protein levels in larval samples, the homogenate samples were mixed with two volumes of 0.2 N NaOH, vortexed for 10 min at RT, heated at 95°C for 15 min to solubilize proteins. The cleared samples after centrifugation at 12,000 rpm (13,000 × g) for 10 min at RT were used to quantify protein using a BCA protein assay kit (Thermo).
Measurement of circulating sugar levels by LC-MS/MS. Staged larvae were collected, rinsed with PBS, and dried on tissue paper. The cuticle was carefully torn to release the hemolymph onto a parafilm membrane. One microliter hemolymph was collected using a micropipette, mixed with 300 μl cold methanol, and stored at −80°C until extraction. The samples were mixed with 300 μl H 2 O and 100 μl CHCl 3 , vortexed for 20 min at RT, and centrifuged at 15,000 rpm (20,000 × g) for 15 min at 4°C. The supernatant (550 μl) was transferred to a new 1.5-ml tube, dried in a vacuum concentrator, re-dissolved in 1 mM ammonium bicarbonate, and analyzed by LC-MS/MS.
Chromatographic separation and mass spectrometric analyses were performed essentially as described previously 27,55 . Chromatographic separation was performed on an ACQUITY BEH Amide column (100 mm × 2.1 mm, 1.7 μm particles, Waters) in combination with a VanGuard precolumn (5 mm × 2.1 mm, 1.7 μm particles) using an Acquity UPLC H-Class System (Waters). Elution was performed at 30°C under isocratic conditions (0.3 mL/min, 70% acetonitrile, and 30% 10 mM ammonium bicarbonate, pH 10.0). The mass spectrometric analysis was performed using a Xevo TQD triple quadrupole mass spectrometer (Waters) coupled with an electrospray ionization source in the negative ion mode. The multiple reaction monitoring transitions were as follows: m/z 341.2-89.0 for trehalose, m/z 179.1-89.0 for glucose and fructose, and m/z 181.1-89.0 for sorbitol. Analytical conditions were optimized using standard solutions. Sample concentrations were calculated from the standard curve obtained from a serial dilution of the standard solution.
Wing morphometric analysis. Adult flies were photographed under a Zeiss Stemi 2000-C stereomicroscope (Zeiss) equipped with a Canon PowerShot G15 digital camera (Canon). Adult flies with indicated genotypes and dietary conditions were fixed in isopropanol in a 1.5-ml tube. Left/right wings were removed in 1× PBS under a Zeiss Stemi 2000 stereomicroscope (Zeiss) and mounted dorsal side up on a glass slide using 25% glycerol/25% isopropanol. Digital images were acquired with a Zeiss Primo Star stereomicroscope equipped with AxioCam ERc (Zeiss). The wing area, as indicated in Fig. 2c, was manually traced and measured using either ImageJ or AxioVision (Zeiss) software.
The wing sizes shown in Figs. 4c and 5c indicate the average sizes between the right and left wings for individuals. Inter-individual variation (IIV) for wing size was expressed as the variance among the average individual wing sizes within populations. Fluctuating asymmetry (FA) was estimated using FA index 6, as described previously 3 , which is based on a signed difference between sides and scaled by average individual size. Namely, FA is the variance expressed as (L − R)/ [(L + R)/2], where L and R are the left and right wings, respectively. The presence of directional asymmetry (DA) was investigated by a one-sample Student's t-test for each data sample [Ho = mean(L − R) = 0]. We detected statistical significance (p < 0.01) for several genotypes and dietary conditions, suggesting the presence of DA. Of note, the FA index 6 remains unaffected by DA 3 . The biological significance and molecular mechanisms underlying DA remain unknown. Experiments were repeated with at least two independently reared populations, and all data were combined. IIV and FA were evaluated by an f-test with a Bonferroni correction for multiple comparisons.
To evaluate the measurement error, we performed two measurements per side for each sample during our first experiment. The average variances for the measurement error were 16.6 % (control flies) and 13.4% (Tps1 MIC mutant flies) of the respective variance calculated for FA (p < 0.0001 by the f-test in all cases). We also determined whether the between-sides variations were larger than the measurement error using a two-way analysis of variance (ANOVA). The interaction term was significant with side and individual as factors (p = 0.0001 in control flies and p = 0.0015 in the Tps1 MIC mutant flies). Thus, the contribution of measurement error to overall variation was relatively small; therefore, we decided to use one measurement per side to calculate FA. All measurements per data set were performed by the same person.
The measurements of cell size and cell number in the adult wing were performed as described previously 56 with minor modifications. The dorsal and ventral sides of the five defined regions, as indicated in Fig. 6c, were acquired separately with a Zeiss Primo Star microscope equipped with AxioCam ERc (Zeiss). Cell density was manually analyzed by counting the number of wing hairs in a 0.01-mm 2 area, and cell size and approximate cell number per wing were calculated. For data shown in Fig. 2b, a single region in the third posterior cell was analyzed per wing. For wing compartment analysis, as shown in Fig. 2d, the wing area and the posterior region (the area between L4 and the wing margin) were manually measured in ImageJ. The anterior region was calculated by subtracting the posterior region from the wing area.
Mating competition assay. Male flies were collected within 12 h of eclosion and were maintained on ND for 3 days. Two males (a control and a Tps1 mutant) and a virgin female (Oregon R or Canton S) were placed into a vial together for mating and oviposition. Parents were discarded 3 days later, and the subsequent F1 males were collected for genotyping to identify the parent male. In brief, genomic DNA was extracted by homogenizing a single fly in 50 μl of buffer [10 mM Tris/HCl (pH 8.0), 1 mM EDTA, 25 mM NaCl, 2 μg/ml Proteinase K (Roche)] using a pellet pestle. The samples were incubated at 37°C for 30 min, heat-inactivated at 95°C for 10 min, and centrifuged at 15,000 rpm (20,000 × g) for 10 min at RT. The supernatant was subjected to PCR amplification to determine the presence of the Minos insertion. Experiments were repeated twice using independently reared populations, and all data were combined.
Fertility test. Male flies were collected within 12 h of eclosion and maintained on ND for 3-7 days. Single males (Tps1 MIC heterozygote or homozygote) and three virgin females (w − ) were placed into a vial for mating and were transferred to fresh vials every day until 3 days after mating. Male fertility was assessed by the presence of larvae. The average number of progenies was determined by counting adult flies from three consecutive vials. Experiments were repeated twice using independently reared populations, and all data were combined.
Statistics and reproducibility. The experiments were replicated at least twice using independently rared populations to ensure reproducibility. The experiments were not randomized, and the investigators were not blinded to fly genotypes during experiments. Sample sizes for each experiment are indicated in the figures and figure legends. Statistical tests were performed in Microsoft Excel and GraphPad Prism 7 software. The statistical tests used are described in the figure legends. Statistical significance was presented as follows: *p < 0.05, **p < 0.01, ***p < 0.001.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.