Lysosomal cystine mobilization shapes the response of TORC1 and tissue growth to fasting

Abstract

Adaptation to nutrient scarcity involves an orchestrated response of metabolic and signaling pathways to maintain homeostasis. We find that in the fat body of fasting Drosophila, lysosomal export of cystine coordinates remobilization of internal nutrient stores with reactivation of the growth regulator target of rapamycin complex 1 (TORC1). Mechanistically, cystine was reduced to cysteine and metabolized to acetyl-coenzyme A (acetyl-CoA) by promoting CoA metabolism. In turn, acetyl-CoA retained carbons from alternative amino acids in the form of tricarboxylic acid cycle intermediates and restricted the availability of building blocks required for growth. This process limited TORC1 reactivation to maintain autophagy and allowed animals to cope with starvation periods. We propose that cysteine metabolism mediates a communication between lysosomes and mitochondria, highlighting how changes in diet divert the fate of an amino acid into a growth suppressive program.

One Sentence Summary:

Lysosomal cysteine recycling is a metabolic break that maintains autophagy during fasting through coenzyme A synthesis.

Introduction

Organisms cope with variations in diet by adjusting their metabolism. Specific organs integrate the availability of nutrients, and respond to maintain systemic homeostasis. In fasting animals, the liver remobilizes nutrients through gluconeogenesis and β-oxidation of fatty acids to support peripheral tissue function (1, 2). Variation in nutrient availability induces parallel changes in activity of signaling pathways, which resets intracellular metabolic turnover. The (Target of rapamycin complex 1 (TORC1) signaling pathway integrates sensing of amino acids and other nutrients with signals from hormones and growth factors to promote growth and anabolism (3). Nutrient scarcity inhibits TORC1 to limit growth and promote catabolic programs, including autophagy, which recycles internal nutrient stores to promote survival (4). Autophagy sequesters cytosolic material into autophagosomes that fuse with lysosomes for cargo degradation and recycling. The lysosomal surface is also the site where nutrient and growth factor-sensing pathways converge to activate TORC1. Degradation within autolysosomes generates new amino acids that in turn fuel metabolic pathways, including the tricarboxylic acid (TCA) cycle and gluconeogenesis, and reactivate TORC1, altogether maintaining minimal anabolism and growth (5–8). However, how organisms regulate the limited pools of remobilized nutrients and balance homeostatic metabolism with anabolic TORC1 activity over the course of starvation is poorly understood.

Results

TORC1 signaling is reactivated in vivo during prolonged fasting

To study the regulation of metabolism and TORC1 signaling in vivo, we used the Drosophila larval fat body, an organ analogous to the liver and adipose tissue in mammals. The fat body responds to variations in nutrient availability through TORC1 signaling, which in turn regulates systemic larval growth rate by the secretion of growth factors from distant organs (9, 10). When larvae were fasted, i.e. completely deprived of their food source but kept otherwise hydrated, TORC1 signaling in the fat body was acutely decreased at the onset of fasting. However, prolonged fasting led to partial and progressive reactivation of TORC1 over time, as measured by phosphorylation of the specific TORC1 substrate, S6K (Fig. 1A). This process was dependent on autophagy induction (fig. S1, A and B), consistent with autophagy facilitating amino acid recycling and TORC1 reactivation in mammalian cells (6, 8). To understand how changes in nutrients and metabolites intersect with TORC1 signaling during this fasting response, we used two complementary, unbiased approaches: (i) a targeted mass spectrometry-based screen for polar metabolites altered during fasting and (ii) a larval growth screen to test the effects of individual amino acids in animals fed a low-protein diet.

Figure 1: TORC1 reactivation upon prolonged fasting correlates with increase of TCA cycle intermediates.

(A) Prolonged fasting leads to TORC1 reactivation. Phosphorylation levels of the direct TORC1 target S6K in dissected fat bodies from fasting larvae (i.e. placed on a tissue soaked in PBS). (B) Heat map metabolite levels (LC-MS/MS) from whole mid-third-instar larvae (left) or dissected fat bodies (right), fed (0h) and fasted. Lower panels are individual plots from the same dataset. (C) Schematic of anaplerosis through PC. (D) Knockdown of pcb/PC suppresses TORC1 reactivation upon prolonged fasting. For (B) and (D), data are shown as mean +/− SD. ns, P≥ 0.05; *, P≤ 0.05; **, P≤ 0.01; ***, P≤ 0.005; ****P ≤ 0.0001 (see the material and methods for details).

Interplay between TORC1 signaling and the TCA cycle during fasting

In the first screen, metabolic profiling of whole animals revealed depletion of most metabolites over the course of fasting (Fig. 1B). By contrast, multiple TCA cycle intermediates accumulated over time, in particular citrate and isocitrate. This appeared not to result from defective TCA cycle activity upon fasting because U-¹³C₆-glucose oxidation in the TCA cycle was functional and the ratios between NADH and NAD⁺ were comparable between fed and fasting conditions (fig. S2, A and B). Similar metabolomic profiles were observed in fat bodies from fed and fasted animals (Fig. 1B), indicating that TORC1 reactivation in the fat body correlates with accumulation of TCA cycle intermediates, in particular citrate and isocitrate. Next, we tested whether TORC1 activity affects the accumulation of TCA cycle intermediates during fasting. For this, we measured the concentrations of TCA cycle intermediates in fed and fasted larvae of GATOR1 (nrpl2^−/−) and GATOR2 (mio^−/−) mutants, which exhibit constitutive activation or suppression of TORC1 signaling, respectively (11, 12) (fig. S3A). Activation of TORC1 in GATOR1 mutants blunted the elevation of the concentration of TCA cycle intermediates normally observed during fasting. Conversely, inhibition of TORC1 in GATOR2 mutants raised the concentration of TCA cycle intermediates in fed animals, suggesting that the accumulation of TCA cycle intermediates during fasting requires TORC1 inhibition. We also tested whether changes in the TCA cycle during fasting affects TORC1 reactivation by targeting pyruvate carboxylase (PC). PC serves an anaplerotic function by replenishing oxaloacetate (OAA) and hence other TCA cycle intermediates (Fig. 1C), a process that regulates gluconeogenesis and regeneration of amino acids (1). Depletion of PC (pcb/CG1516) in the larval fat body increased the concentration of the PC substrate alanine in the fat body (fig. S3B) (1). Consistent with impaired TCA cycle activity, depletion of PC elevated the level of OAA and the ratio between NAD(P) and NAD(P)H, while decreasing the levels of other TCA cycle intermediates as well as cataplerotic products such as asparagine, intermediates of the urea cycle, carbamoyl aspartate, and inosine 5′-monophosphate (IMP) (fig. S3B). Depletion of PC also suppressed TORC1 reactivation during fasting (Fig. 1D), supporting the functional link between the TCA cycle and TORC1 signaling. These data suggest that at the onset of fasting, TORC1 inhibition appears necessary to raise the concentration of TCA cycle intermediates, which in turn affect TORC1 reactivation, possibly through the synthesis of building blocks downstream of PC, i.e. cataplerosis.

Cysteine suppresses growth in a TORC1-dependent manner

In the second screen, we tested the effects of supplementation with single amino acids on growth of larvae fed a low-protein diet (see the supplementary materials) (Fig. 2A). Fasting and a low-protein diet may trigger distinct metabolic states with intrinsic differences, but both protocols trigger a starvation response. Supplementation of food with cysteine suppressed growth in larvae to a distinctly greater extent than any other amino acid, and this phenotype was accentuated in a low-protein diet (see the supplementary text S1 and fig. S4, A to D). The inhibition of growth by cysteine was interdependent with TORC1 signaling because cysteine treatment partially inhibited TORC1 activity in fat bodies, and constitutive activation of TORC1 in GATOR1-null (nprl2^−/−) mutants (12) partially restored growth under cysteine supplementation (fig. S5,A to E). Conversely, cysteine supplementation did not further suppress growth in mutants with constitutive suppression of TORC1 activity caused by either loss of GATOR2 (mio^−/−) or overexpression of TSC complex subunits (lpp>UAS-TSC1, UAS-TSC2) in the fat body (fig. S5, E and F). Altogether, these data suggest that cysteine has a specific inhibitory effect on growth and TORC1 signaling in larvae fed a low-protein diet, and that this growth suppression requires suppression of TORC1 signaling.

Figure 2: Lysosome-derived cysteine promotes elevation of TCA cycle intermediates and antagonizes TORC1 reactivation.

(A) Amino acid screen reveals cysteine as a growth suppressor. Time to pupariation for larvae fed a low-protein diet supplemented with the indicated amino acids all along development (see supplementary materials and methods for concentrations; the cysteine concentration is 5 mM). Scale bar, 1 mm. (B) Cysteine levels in whole control larvae (lpp>attp40). N = 5. (C) Schematic of lysosomal cystine (CySS) efflux through cystinosin/ CTNS. (D) Loss of Drosophila cystinosin in the fat body (lpp>dCTNS RNAi, dCTNS-i) leads to higher TORC1 reactivation upon prolonged fasting. P-S6K levels in dissected fat bodies from larvae fasted for the indicated time. lpp>w RNAi, control background. (E) dCTNS maintains autophagy during fasting. dCTNS^−/− clones (non-GFP, outlined) in 80h AEL larvae expressing mCherry-Atg8a fasted for 8h. Scale bar, 10 μm. (F) dCTNS overexpression suppresses TORC1 activity. P-S6K levels in dissected fat bodies from larvae fasted for the indicated time. lpp>GFP RNAi (GFP-i), control background. (G) dCTNS overexpression induces autophagy. dCTNS overexpression clones (GFP marked, outlined) in 80h AEL larvae. Scale bar 10 μm. (H) dCTNS regulates TCA cycle intermediates levels upon fasting. Relative metabolites levels measured by LC-MS/MS in 80h AEL larvae fed or fasted for 8h. Controls are dCTNS^+/− (upper graphs) or GFP-i (lower graphs). For (B), (D), (F) and (H), data are shown as mean +/− SD. ns, P≥ 0.05; **, P≤ 0.01; ***, P≤ 0.005; ****P≤ 0.0001 (see the materials and methods for details).

Lysosomal cystine transporter dCTNS regulates cysteine level during fasting

The results of our two screens led us to explore the relationship among cysteine metabolism, the TCA cycle, and TORC1 during fasting. Cysteine concentrations increased during fasting (Fig. 2B and fig. S6A) (13), and cysteine treatment increased the concentration of TCA cycle intermediates, particularly for animals fed a low-protein diet (fig. S6B). To further understand this effect of cysteine, we searched for its intracellular source during fasting. Abolishing autophagy in the fat body decreased cysteine levels upon fasting (fig. S7A), suggesting that autolysosomal function regulates cysteine balance. Thus, we focused on the role of the lysosomal cystine transporter cystinosin in recycling cysteine during fasting (Fig. 2C). Cystinosin, which is encoded by CTNS in mammals, is mutated in the lysosomal storage disorder cystinosis and has been implicated in the regulation of TORC1 signaling and autophagy (14–16). Endogenous tagging of the Drosophila ortholog CG17119 (hereafter referred to as dCTNS) confirmed its specific lysosomal localization in cells of the fat body (fig. S7, B and C). dCTNS^−/− larvae showed accumulation of cystine (fig. S7D), consistent with a role for cystinosin in lysosomal cystine transport. In fed conditions, control and dCTNS^−/− larvae showed similar abundance of cysteine, likely reflecting dietary intake as the main source of cysteine. By contrast, in fasting animals, cysteine concentration dropped in dCTNS^−/− larvae (fig. S7E). Conversely, dCTNS overexpression increased cysteine levels, which caused a developmental delay similar to cysteine treatment (fig. S7, F and G). In sum, our results indicate that dCTNS recycles cysteine from the lysosome during fasting.

Cytosolic cysteine synthesis compensates for defective lysosomal cystine export

To evaluate the contribution of cytosolic cysteine synthesis during fasting, we also analyzed the trans-sulfuration pathway, a process that requires cystathionine β-synthase (Cbs) (fig. S8A) (17, 18). Depletion of cbs in the fat body reduced cysteine levels upon fasting, but this was not the case in fed animals. Cbs and dCTNS showed additive roles in maintaining cysteine concentration in fasted animals (fig. S8B), suggesting that the phenotypic effects of dCTNS deficiency are partially masked by the trans-sulfuration pathway. Because the trans-sulfuration pathway involves the breakdown of methionine, we found that dCTNS overexpression increased the concentration of the intermediates of the trans-sulfuration pathway, particularly methionine (fig. S8C). By contrast, methionine was depleted in fasted dCTNS mutant animals (fig. S8C), suggesting that cysteine from lysosomal origin limits the activity of the trans-sulfuration pathway. Methionine is an essential amino acid, and dCTNS appears to limit its depletion upon fasting, a process possibly reminiscent of the methionine-sparing effect of dietary cystine previously observed in humans (19, 20).

dCTNS affects TORC1 reactivation and the TCA cycle during fasting

Next, we analyzed the effect of cysteine recycling by dCTNS on TORC1 reactivation and the TCA cycle. Although dCTNS depletion in the larval fat body did not affect TORC1 inhibition at the onset of fasting, it slightly increased TORC1 reactivation upon prolonged fasting, as indicated by increased S6K phosphorylation (Fig. 2D) and cytosolic accumulation of Mitf/TFEB (fig. S9A). Analysis of dCTNS^−/− fat body clones showed that increased TORC1 signaling was cell autonomous and sufficient to compromise maintenance of autophagy during fasting (Fig. 2E and fig. S9B). Accordingly, treatment with the TORC1 inhibitor rapamycin restored autophagy in dCTNS-deficient cells (fig. S9C). Conversely, dCTNS overexpression caused downregulation of TORC1 in fed and fasting animals, and induced ectopic autophagy in fed animals (Fig. 2, F and G). Metabolic profiling of dCTNS^−/− animals showed a depletion of TCA cycle intermediates specifically during fasting, whereas they accumulated after overexpression of dCTNS in fed and fasted animals (Fig. 2H).

dCTNS is required for animal fitness during starvation

We also examined the role of dCTNS on starvation resistance. In normally fed animals, dCTNS deficiency delayed larval development but appeared to have no effect on the life span of adult flies (Fig. 3, A and B). However, dCTNS-deficient animals had an increased developmental delay when raised in a low-protein diet, and adults died more quickly from starvation (Fig. 3, B and C). Depletion of dCTNS specifically in the fat body did not affect development in fed animals but caused a developmental delay on a low-protein diet, consistent with the importance of dCTNS function in the fat body during fasting (Fig. 3B). The starvation sensitivity of dCTNS-deficient animals was decreased by low concentrations of rapamycin (that did not elevate cysteine concentration), indicating a possible role for altered TORC1 signaling and autophagy in mediating dCTNS growth phenotypes (Fig. 3D, fig. S9D; see supplementary text S2). The role of dCTNS during starvation was dependent on its cystine transport function as treatment with either cysteamine [which facilitates cystine export out of the lysosome independently of cystinosin (21)] or low concentrations of cysteine (which did not affect development of control animals) decreased the starvation sensitivity of dCTNS^−/− animals (Fig. 3, E to G). Altogether, we demonstrate the physiological importance of cysteine recycling from the lysosome through dCTNS during fasting, which supports the accumulation of TCA cycle intermediates and antagonizes TORC1 reactivation to maintain autophagy.

Figure 3: dCTNS controls resistance to starvation through cysteine efflux and TORC1.

(A) dCTNS does not affect life span in fed condition. Life span of control (w¹¹¹⁸) and dCTNS^−/− animals fed a standard diet. N = 2. (B) dCTNS in the fat body controls starvation resistance during development. Shown is the fold change time to pupariation for larvae of indicated genotype grown on control (fed) or low-protein diet. Controls are dCTNS+/− (left panel) or white RNAi (control-i, right panel). (C) dCTNS controls starvation resistance of adult animals. Survival of control (w¹¹¹⁸) and dCTNS^−/− animals fed a chemically defined starved diet composed only of physiologically relevant ions, including bio-metals (see the materials and methods). (D to F) Low dose of rapamycin and cysteine treatments rescue starvation sensitivity of dCTNS^−/− animals. Show is the developmental time of larvae raised on a low-protein diet supplemented with the indicated concentration of rapamycin (D) or 0.1 mM cysteine (F) and survival of adult flies on chemically defined starved diet with or without 1 mM cysteine (E). Controls are dCTNS^+/− [(D) and (F)] and w¹¹¹⁸ (E). (G) Cysteamine treatment restores starvation resistance of dCTNS^−/− animals. Show is the life span of control (w¹¹¹⁸) and dCTNS^−/− animals fed a chemically defined starved diet supplemented with 0.5 mM cysteamine or vehicle. For (B) to (G), data are shown as mean +/− SEM. ns, P≥ 0.05; *, P≤ 0.05; **, P≤ 0.01; ***, P≤ 0.005; ****, P≤ 0.0001 (see the materials and methods for statistics details).

dCTNS connects with the TCA cycle through acetyl-CoA metabolism

To search for a mechanistic link between cysteine metabolism and the TCA cycle, we performed heavy isotope-labeled cysteine tracing experiments by feeding larvae either U-¹³C-cysteine or ¹³C₃¹⁵N₁-cysteine. Consistent with cysteine metabolism in mammals, this revealed three major metabolic fates for cysteine: glutathione (GSH), taurine, and coenzyme A (CoA) (Fig. 4A and B). CoA derived from labeled cysteine was subsequently used to form acetyl-CoA in the fat body of larvae fed a control and low-protein diet (Fig. 4, A and B, and fig. S10, A and B). We found that the starvation sensitivity of dCTNS^−/− animals was partially suppressed by dietary treatment with pantethine, a disulfide intermediate of the CoA biosynthetic pathway downstream of cysteine incorporation (fig. S11, A and C), but not with pantothenic acid, the vitamin precursor upstream of cysteine incorporation during CoA synthesis (fig. S11B). This suggests that deficiencies in CoA synthesis caused by limiting cysteine availability contributes to the dCTNS phenotypes. To analyze how this affects the TCA cycle, we further focused on the direct flow of carbons from cysteine into acetyl-CoA. Upon feeding, the pyruvate dehydrogenase complex (PDHc) harvests the acetyl moiety from pyruvate and transfers it to CoA to generate acetyl-CoA (22) (fig. S12A). Upon fasting, PDHc activity decreases, draining pyruvate towards PC to support biosynthesis, whereas β-oxidation of fatty acids may provide the acetyl moiety for acetyl-CoA synthesis (23) (Fig. 5A and fig. S12A). In fed larvae overexpressing dCTNS in the fat body, depletion of the PDHc activator pyruvate dehydrogenase phosphatase (pdp) (24) normalized growth, TORC1 activity and the concentration of TCA cycle intermediates (fig. S12, B to D). By contrast, when dCTNS-overexpressing larvae were fed a low-protein diet or fasted, PDHc inhibition failed to restore development or fully restore TORC1 activity respectively, presumably reflecting the use of substrates other than pyruvate for acetyl-CoA synthesis under these conditions, such as acetate and fatty acids (fig. S12, B and C). In turn, depletion of CPT1/whd, the acylcarnitine transferase required for β-oxidation of fatty acids, restored growth of dCTNS-overexpressing larvae to a larger extent on a fed diet than on a low-protein diet, suggesting that increased cysteine metabolism is sufficient to promote β-oxidation upon feeding (fig. S12F). Accordingly, dCTNS overexpression increased the abundance of acetyl-CoA and decreased the abundance of acyl-carnitines and fatty acids in both fed and fasted animals. In contrast, in dCTNS^−/− animals, acetyl-CoA levels were decreased whereas acyl-carnitines and fatty acids showed higher levels, specifically during fasting, which could be normalized by cysteamine treatment (Fig. 5, B and C, and fig. S13, A and B). In addition, dCTNS overexpression was sufficient to deplete triglyceride stores in fed animals, whereas they slightly accumulated after depletion of dCTNS in the fat body upon fasting (fig. S13C). Altogether, our data suggest that during fasting, lysosomal cysteine mobilization is potentially rate limiting for de novo CoA synthesis, which in turn may promote acetyl-CoA production through β-oxidation of fatty acids remobilized from triglyceride stores. The precise contribution of acetyl moieties by fatty acids and other substrates for the synthesis of acetyl-CoA during fasting remains to be determined.

Figure 4: Cysteine fuels de novo CoA/acetyl-CoA metabolism.

(A) Mean fractional enrichment +/− SD of U-¹³C-cysteine (N = 10), ¹³C₃_¹⁵N₁-cysteine (N = 5) or unlabeled samples (N = 10) in indicated metabolites measured by LC-MS/MS in whole larvae fast overnight with 5 mM tracer. m+n refers to the number of ¹³C atoms (+n) added to the expected mass spectra of each measured isotopomer (m); m+0 means unlabeled. (B) Schematic of cysteine metabolism and labelling patterns from U-¹³C-cysteine and ¹³C₃-¹⁵N₁-cysteine tracers. Red arrows indicate main cysteine flux.

Figure 5: Cysteine metabolism to acetyl-CoA affects the concentration of fatty acids and increases carbon flux through PC and the TCA cycle.

(A) Schematic of acetyl-CoA synthesis during fasting. (B and C) Metabolite levels in whole 3^rd instar larvae showing dCTNS^−/− and dCTNS overexpression in the fat body. (D) Schematic of alanine carbon flux into the TCA cycle upon fasting. E) Alanine flux ratio. Shown in the fold change lpp>dCTNS /control (lpp>attp40) for the indicated TCA cycle intermediates isotopomers measure by LC-MS/MS in dissected fat bodies from 3^rd instar larvae fed a low-protein diet with 25 mM U-¹³C-alanine for 6 hours.

Lysosomal-derived cysteine may control TORC1 indirectly through the flow of amino acids in and out of the TCA cycle

Increased CoA synthesis from cysteine might enlarge the TCA cycle carbon pool in at least two ways: first, by providing the CoA required to accept increasing amounts of carbon from fatty acids to form acetyl-CoA, and second, by promoting anaplerosis of alternative carbon sources through the allosteric activation of PC by acetyl-CoA (25). To test whether increased production of acetyl-CoA supported by lysosomal cysteine efflux could increase anaplerosis, we analyzed alanine anaplerosis in the TCA cycle in animals overexpressing dCTNS in the fat body. We supplemented a low-protein diet with a [U-¹³C]alanine tracer and followed the anaplerotic flux of alanine in the TCA cycle in dissected fat bodies (Fig. 5D). We used a tracer amount that had a negligible contribution to the total alanine pool and also did not affect cysteine metabolism to acetyl-CoA (fig. S14, A and B). dCTNS overexpression increased by more than two-fold alanine anaplerosis as well as oxidative flux in the TCA cycle through citrate synthase (that consumes OAA and acetyl-CoA to generate citrate; Fig. 5E). We therefore propose that during fasting, cysteine recycling and metabolism to acetyl-CoA in the fat body supports anaplerosis through PC and flux through citrate synthase, thereby contributing to the accumulation of TCA cycle intermediates, in particular citrate and isocitrate.

During fasting, given the fixed and pre-established level of carbons available, increased abundance of TCA cycle intermediates may indicate the retention of anaplerotic inputs in the TCA cycle at the expense of their extraction for biosynthesis (i.e. cataplerosis). Because cataplerosis promotes amino acid synthesis (1), which in turn affects TORC1 signaling (6), we analyzed the effect of lysosomal cysteine recycling on individual amino acid pools. dCTNS overexpression in the fat body led to depletion of aspartate and downstream nucleotide precursors (IMP and uridine 5′-monophosphate), as well as, to a lesser extent, asparagine and glutamate (Fig. 6A and fig. S15, A and B). The abundance of aspartate and IMP were increased in fasted dCTNS^−/− animals, and cysteamine treatment partially normalized their concentration (fig. S15C). Aspartate is a cataplerotic product of OAA (1), a process that involves glutamate oxaloacetate transaminase 2 (Got2). Cysteine metabolism may transiently trap anaplerotic carbons into the TCA cycle in fasted animals, away from their immediate extraction for biosynthesis. We analyzed whether cysteine metabolism could regulate growth and antagonize TORC1 reactivation through limiting the availability of glutamate and aspartate, because these amino acids are critical regulators of cell growth and have previously been implicated in TORC1 signaling (6, 26). For this analysis, we replenished aspartate and glutamate concentrations in dCTNS-overexpressing animals through dietary supplementation of anaplerotic amino acids. Treatments with combinations of alanine, aspartate, asparagine, glutamate, and proline [which replenishes glutamate in flies (27)] restored normal levels of aspartate and glutamate without compromising dCTNS-induced elevation of TCA intermediates (fig. S15, D and E). These treatments rescued the developmental delay induced by dCTNS overexpression in the fat body (Fig. 6B). Similarly, co-supplementation of cysteine with excess of single amino acids including alanine, aspartate, asparagine, glutamate, and proline each rescued cysteine-induced growth suppression upon fasting (Fig. 6C). In addition, treatments with a combination of amino acids restored TORC1 activity upon fasting after dCTNS overexpression in the fat body (Fig. 6D). Consistent with the importance of aspartate synthesis, clonal loss of Got2 induced autophagy (6), inhibited TORC1 activity as indicated by decreased levels of phosphorylated 4E-BP and increased levels of the reporter Unk-GFP (28), and inhibited cell growth as shown by decreased nucleus size (fig. S15F). Thus, cysteine metabolism appears to regulate anaplerotic carbon flow in the TCA cycle and the level of cataplerotic products such as aspartate. We propose that lysosomal-derived cysteine converts the TCA cycle into a reservoir of carbons in the mitochondria while limiting their extraction for biosynthesis. This process may spare nutrients to allow animals to survive starvation while resetting TORC1 activity to a threshold that maintains minimal growth without compromising autophagy.

Figure 6: Cysteine metabolism regulates TORC1 and growth through cataplerotic amino acids levels.

(A) Relative levels of aspartate in 85h AEL larvae dCTNS^−/− (whole larvae and fat body) and after dCTNS overexpression in the fat body. (B) Amino acid supplementation suppresses the developmental delay induced by dCTNS overexpression in the larval fat body. Shown is the fold change time to pupariation for larvae fed a low-protein diet with or without supplementation with the indicated amino acids (Ala, Pro and Glu: 5 mM; Asp: 10 mM). Red asterisks show the significance between dCTNS-overexpressing animals treated with vehicle vs. amino acids. (C) Photographs of aged matched animals fed a low-protein diet with or without 5 mM cysteine with or without the indicated metabolites (25 mM each). Scale bar, 1 mm. (D) dCTNS-induced TORC1 inhibition is reversed by supplementation with the indicated amino acids. P-S6K levels in fat body from fed and fasted (6h) larvae of indicated genotypes. The 70 to 72h AEL larvae were transferred to a low-protein diet with or without supplementation with the indicated amino acids (Ala and Pro: 20 mM; Asp and Glu: 10 mM). Control is GFP-i. For (A), (B), and (D), data are shown as mean +/− SD. ^ns, P≥ 0.05; *, P≤ 0.05; **, P≤ 0.01; ***, P≤ 0.005; ****P≤ 0.0001 (see the materials and methods for details).

Discussion

Maintaining cellular homeostasis upon nutrient shortage is an important challenge for all animals. Decreased activity of TORC1 is necessary to limit translation, reduce growth rates, and promote autophagy. Conversely, minimal TORC1 activity is required to promote lysosomal biogenesis, thus maintaining autophagic degradation necessary for survival (8). Using Drosophila as an in vivo model, we found that TORC1 reactivation upon fasting integrates the biosynthesis of amino acids from anaplerotic inputs into the control of growth. The regulation of aspartate abundance appears to be critical during this process, possibly because it serves as a cataplerotic precursor for various macromolecules, including other amino acids and nucleotides, which in turn impinge on TORC1 activity (29). Cysteine recycling through the lysosome may fuel acetyl-CoA synthesis and prevent reactivation of TORC1 above a threshold that would compromise autophagy and survival during fasting. Reactivation of TORC1 during fasting was not passively controlled by the extent of amino acid remobilized from the lysosome. Instead, cysteine metabolism supported an increased incorporation of the carbons from these remobilized amino acids into the TCA cycle. We therefore propose that the remobilized amino acids may be transiently stored in the form of TCA cycle intermediates compartmentalized in the mitochondria, thereby restricting their accessibility. The regulation of TORC1 activity over a fasting period appears to be a combination of activating and suppressing cues that conciliate autophagy with anabolism. This process is self-regulated by autophagy, because autophagic protein degradation controls cystine availability through the lysosomal cystinosin transporter. Thus, in contrast to fed conditions, in which amino acid transporters at the plasma membrane maintain high cytosolic concentration of leucine and arginine that can directly be sensed by members of the TORC1 machinery (3), TORC1 reactivation in prolonged fasting is regulated indirectly by lysosome-mitochondrial cross-talk. Because cystinosin has also been shown to physically interact with several components of lysosomal TORC1 in mammalian cells (14), additional layers of regulation are conceivable during this process.

Multiple functions of cysteine impinge on cellular metabolism, including transfer RNA thiolation, the generation of hydrogen sulfide, the regulation of hypoxia-inducible factor (HIF), and its antioxidant function through glutathione synthesis (30–32). Supplementation with cysteine or modified molecules such as N-acetyl-cysteine (NAC) can be used to efficiently buffer oxidative stress, and perhaps alleviate symptoms of diseases that promote oxidative stress or glutathione deficiency, including cystinosis (33–36). Cysteine or NAC treatment extends the life span in flies, worms, and mice, and mice fed NAC show a sudden drop in body weight similar to that caused by dietary restriction (18, 37). Our results indicate that cysteine may not only act through its antioxidant function but also by restricting the availability of particular amino acids and limiting TOR activity, processes known to extend life span. Moreover, we show that CoA is a main fate of cysteine that affects oxidative metabolism in the mitochondria, which is the main source of reactive oxygen species (ROS). Thus, the antioxidant function of cysteine also might be coupled to its effects on the mitochondria to buffer ROS production.

In summary, we demonstrate that cysteine metabolism acts in a feedback loop involving de novo CoA synthesis, the TCA cycle, and amino acid metabolism to limit TORC1 reactivation upon prolonged fasting. This pathway may be particularly important for developing organisms that must maintain autophagy and balance growth and survival during periods of food shortage.

Materials and Methods

Fly stocks and maintenance

All flies were reared at 25°C and 60% humidity with a 12-hours on/off light cycle on standard laboratory food. N.P’s standard laboratory food: 12.7 g/L deactivated yeast, 7.3 g/L soy flour, 53.5 g/L cornmeal, 0.4% agar, 4.2 g/L malt, 5.6% corn syrup, 0.3% propionic acid, 1% tegosept/ethanol. M.S’s standard laboratory food: 18 g/L deactivated yeast, 10 g/L soy flour, 80 g/L cornmeal, 1% agar, 40 g/L malt, 5% corn syrup, 0.3% propionic acid, and 0.2% 4-hydroxybenzoic acid methyl ester (nipagin)/ethanol. Density was standardized for at least one generation before the experiments. For experiments, larvae were reared on freshly made food. Lpp-gal4 was a gift from P. Léopold. UAS-tsc1, UAS-tsc2 was a gift from C. Mirth (38). yw,hs-Flp; mCherry–Atg8a; Act>CD2>GAL4, UAS–nlsGFP/TM6B was a gift from E. Baehrecke. hsFlp; act>CD2>Gal4, UAS nlsGFP is a stock from N.P.’s laboratory (39). yw, hsFlp, Tub-Gal4>UAS-nlsGFP/FM6;;neoFRT82B, TubGal80/TM6,Tb,Hu was a gift from A. Bardin. nprl2¹ and mio² were a gift from M. Lilly. hsFlp; R4-Gal4, UAS-mCherry-Atg8a; FRT82B UAS-GFP/TM6b was a gift from G. Juhász. The following stocks were obtained from the Bloomington Drosophila Stock Center (BDSC) at Indiana University: UAS-w^RNAi (HMS00045), UAS-w^RNAi (HMS00017), UAS-GFP^RNAi (#9330) attp40 (#36304), attp2 (#36303), UAS-mCherry-nls (#38425), UAS-Atg1^RNAi (HMS02750), UAS-Atg18a^RNAi (JF02898), UAS-TSC2^RNAi (HM04083), w¹¹¹⁸, UAS-dCTNS^RNAi (HMS00213), UAS-Got2^RNAi (HMJ21924), UAS-CPT1/whd^RNAi (HMS00033), UAS-Cbs^RNAi (GL01309), UAS-pdp^RNAi (HMS018888), UAS-pcb^RNAi (HMC04104). When comparing the effects of RNA interference (RNAi) knockdown or protein overexpression through induction of UAS-dsRNA or UAS-cDNA expression, UAS-w^RNAi (HMS00045), UAS-w^RNAi (HMS00017), attp2 (#36303) and attp40 (#36304) were used as controls for the TRIP collection (http://www.flyrnai.org/TRiP-HOME.html), and UAS-GFP RNAi for dCTNS overexpression. dCTNS knockout flies were generated with CRISPR/Cas9 technology according to (40). Two single guide RNAs (sgRNAs) using oligos (one after the ATG start codon in exon 3: GGTGATGTCATGGGAATCGA, and the other before the translation of the first transmembrane domain in exon 4: GGGCAGTACTCGAAATCAGT) were produced by polymerase chain reaction (PCR) and in vitro transcribed into RNA using the MEGAscript^™ T7 Transcription Kit (Thermo Fisher Scientific). RNA was injected into act-Cas9 flies (from F. Port and S. Bullock). F₀ flies were crossed to w;; TM3,Sb/TM6,Tb balancer flies and F₁ progenies were screened through PCR using oligos flanking the targeted genome region. Any indel difference >3 bp was visualized in 4% agarose gel in heterozygote F₁ progeny. To generate dCTNS-mKate2 fusion allele, an mKate2 open reading frame was inserted at the C-terminus of dCTNS using a CRISPR/Cas9 endogenous tagging strategy wih vectors kindly provided by Y. Bellaiche (Curie Institut, Paris). In brief, two 1-kb-long homology arms (HR1 and HR2) of the dCTNS gene flanking the sgRNA-guided Cas9 cutting site were cloned into a vector flanking the ATG/STOP-less mKate2 allele (HR1-linker-mKate2-loxP-mini-white-loxP-linker-HR2). In addition, two vectors for the expression of sgRNA (sgRNA-1: CCACCGTGACCGATGTTCAAAAT, sgRNA-2: CCGAGCGAAGTGACGACTGAGAA) targeting the C-terminal coding region of dCTNS were generated. All three vectors were injected into vas-Cas9 flies (BDSC#55821) embryos by Bestgene. Progenies were screened for the red eyes (selection marker mini-white) and crossed to Cre-expressing flies to remove the mini-white by loxP/Cre excision. For overexpression of dCTNS, dCTNS cDNA was cloned into the Gateway destination vector pUASg-HA.attB (GeneBank: KC896837) according to (41). The plasmid was injected by Bestgene into embryos (BDSC#24482) for ϕ31-mediated recombination at a attP insertion site on the second chromosome.

Fly food and starvation protocols

Heavy isotope tracers were from Cambridge Isotope Laboratories. All other amino acids and compounds used were from Sigma-Aldrich. In N. P’s laboratory, compounds in solution were added to the following food mixture: 60 g/L sucrose, deactivated yeast as a source of total protein (2 g/L for low-protein diet, 4 g/L for mild low-protein diet, 20g/L for fed fly food); 80 g/L cornmeal, 0.35% agar, 0.3% propionic acid, and 1% Tegosept (100g/L in ethanol). Glucose tracing was done without sucrose, and alanine tracing without yeast. In M.S’s laboratory, fasting food had to be adapted to 6 g/L of deactivated yeast to match control fast (2 g/L) developmental rates observed in the N.P.’s laboratory. For control food in the M.S. laboratory, the standard lab food was used (see protocol above). In fig. S4 specifically, 100% amino acid was 17g/L deactivated yeast. For fasting experiments, larvae were placed on paper wipes soaked in phosphate-buffered saline (PBS) in a petri dish.

Generation of clones

For autophagy experiments, clones were generated by crossing yw, hs-flp; mCherry–Atg8a; Act>CD2>GAL4, UAS–nlsGFP/TM6B with the indicted UAS lines. Progeny of the relevant genotype was reared at 25°C, and spontaneous clones were generated in the fat body because of the leakiness of the heat-shock flipase (hs-flp). For dCTNS^−/− clones, autophagy was analyzed by crossing w;;neoFRT82B, dCTNS^−/− to hs-flp; R4-Gal4, UAS-mCherry-Atg8a; FRT82B UAS-GFP/TM6b. For P-4EBP1 experiments, clones were either generated by crossing hs-flp; act>CD2>Gal4, UAS nlsGFP with the indicted UAS lines or, for dCTNS^−/− clones, by crossing w;;neoFRT82B, dCTNS^−/− to yw, hs-flp, tub-Gal4>UAS-nlsGFP/FM6;;neoFRT82B, tubGal80/TM6,Tb,Hu. F1 embryos collected overnight were heat shocked for 2h at 37°C the following morning.

Amino acid screen

Larvae were fed a diet with reduced yeast extract (50% of normal diet), individual amino acids were systematically added to the food, and the time to pupariation was monitored as a proxy for growth rate. The following mixture was diluted 1:1 with amino acids solutions in water: 10 g/L agar; 120 g/L sucrose; 17 g/L deactivated yeast extract; 80 g/L cornmeal; 6 mL/L propionic acid; 20 mL/L Tegosept. The amount of amino acids added to the food was determined based on those used in tissue culture growth supplements (see table S1)(42–44).

Food intake

Larvae were synchronized in L1 and reared on the indicated food types until the mid-2^nd instar stage. They were then transferred on the same food type supplemented with 0.5% weight/volume erioglaucine disodium salt (Sigma-Aldrich) for 2 hours. Samples were homogenized in 200 μl of PBS, and absorbance of the dye in the supernatant was measured at 625 nm. Results were normalized to protein content.

Developmental timing

Three-day-old crosses were used for 3- to 4 hour periods of egg collection on standard laboratory food. Newly hatched L1 larvae were collected 24 hours later for synchronized growth using the indicated diets at a density of 30 animals per vial. The time to develop was monitored by counting the number of animals that underwent pupariation every 2 hours in fed conditions or once or twice a day in starved conditions. The time at which half the animals had undergone pupariation was recorded.

Adult survival experiments

To generate age-synchronized adult flies, larvae were raised on laboratory food at low density, transferred to fresh food upon emerging as adults, and mated for 48h. Animals were anesthetized with low levels of CO₂, and males were sorted at a density of 10 per vial. Each condition contained 8 to 10 vials. Each experiment was repeated at least 3 times, and the average values of each experiment were used for statistical analysis. Flies were transferred to fresh food vials three times per week, at which point deaths were scored. After 10 days, deaths were scored every day. Chemically defined (holidic) fasting medium was prepared following the protocol described in (43), which contains all physiologically relevant ions (including bio-metals) but lacks energy sources such as sugar, proteins, amino acids, lipids, and lipid-related metabolites, as well as nucleic acids and vitamins.

Growth curves and pupal weight

Synchronized, newly-hatched L1 larvae were immediately weighed or placed on the indicated food at a density of 30 to 50 animals per vial. Pools of 20 to 80 animals were weighed every 24 hours using an analytical scale (Mettler Toledo), and the weight was reported +/− SEM. For pupal weight, 2-day-old pupae from vials at a density of 30 animals were weighed in batches of 5 to 10 pupae. The weights of different batches of larvae from the same vials were averaged and counted as N=1.

Metabolite profiling

For whole-body metabolic profiling, 25 to 38 mid-second-instar or 8 to 15 mid-third-instar larvae per sample were collected, snap-frozen in liquid nitrogen and stored at −80°C in extraction buffer (4 to 6 biological replicates/experiment). For fat body metabolic profiling, fat bodies from 35 to 40 larvae 96 hours after egg laying (AEL) were dissected in 20 μl PBS, diluted in 300 μl cold extraction buffer and snap frozen. Tissues were homogenized in extraction buffer using 1 mm zirconium beads (Next Advance, ZROB10) in a Bullet Blender tissue homogenizer (model BBX24, Next Advance). Metabolites were extracted using 80 % (v/v) aqueous methanol (two sequential extractions with 300 to 600 μl) and metabolites were pelleted by vacuum centrifugation. Pellets were resuspended in 20 μl of high-performance liquid chromatography (HPLC) grade water, and metabolomics data were acquired using targeted liquid chromatography tandem mass spectrometry (LC-MS/MS). A 5500 QTRAP hybrid triple quadrupole mass spectrometer (AB/SCIEX) coupled to a Prominence UFLC HPLC system (Shimadzu) was used for steady-state analyses of the samples. Selected reaction monitoring (SRM) of 287 polar metabolites using positive/negative switching with hydrophilic interaction LC (HILIC) was performed. Peak areas from the total ion current for each metabolite SRM Q1/Q3 transition were integrated using MultiQuant version 2.1 software (AB/SCIEX). The resulting raw data from the MultiQuant software were normalized by sample weight for whole animals. Fat body samples were normalized by the mean protein content measured from duplicate dissection for each condition. Data were analyzed using Prism informatic software. Alternatively, collected larvae were rinsed with water, 70% ethanol and PBS to remove food and bacteria, snap-frozen in liquid nitrogen and stored at −80°C until extraction in 50% methanol, 30% acetonitrile, and 20% water. The volume of extraction solution added was adjusted to larvae mass (40 mg/ml), samples were vortexed for 5 min at 4°C, and then centrifuged at 16,000 g for 15 minutes at 4°C. Supernatants were collected and analyzed by LC-MS using a QExactive Plus Orbitrap mass spectrometer equipped with an Ion Max source and a HESI II probe and coupled to a Dionex UltiMate 3000 UPLC system (Thermo Fisher Scientific, USA). An SeQuant ZIC-pHilic column (Millipore) was used for liquid chromatography separation (45). The aqueous mobile-phase solvent was 20 mM ammonium carbonate plus 0.1% ammonium hydroxide solution, and the organic mobile phase was acetonitrile. The metabolites were separated over a linear gradient from 80% organic to 80% aqueous for 15 min and detected across a mass range of 75–1,000 m/z at a resolution of 35,000 (at 200 m/z) with electrospray ionization and polarity switching mode. Lock masses were used to ensure mass accuracy <5 ppm. The peak areas of different metabolites were determined using TraceFinder software (Thermo Fisher Scientific) using the exact mass of the singly charged ion and known retention time on the HPLC column. In total, each metabolic profiling experiment was performed at least two times with 3 to 7 biological replicates per genotype.

TCA cycle isotopomer method from U-¹³C-cysteine, ¹³C₃-¹⁵N₁-cysteine, U-¹³C-glucose and U-¹³C-alanine

Fed and starved mid-third-instar animals were supplemented with the indicated concentrations of U-¹³C-cysteine, ¹³C₃-¹⁵N₁-cysteine, U-¹³C-glucose, U-¹³C-alanine or vehicle in the food for the indicated times. For tracing experiments on the low-protein diet (starved), animals were pre-starved 1 hour on PBS before being transferred to the relevant tracer and food. Samples were collected (at least 5 biological replicates for the labelled conditions, and at least 4 biological replicates for the unlabeled condition), and intracellular metabolites were extracted using 80% (v/v) aqueous methanol. Q1/Q3 SRM transitions for incorporation of ¹³C-labelled metabolites were established for polar metabolite isotopomers, and data were acquired by LC-MS/MS. Peak areas were generated using MultiQuant version 2.1 software. Peak areas from unlabeled conditions were used for background determination in each experiment. In figures in Fig. 5E and figs. S2A, S10A and B, and S14, data for both labelled and unlabeled conditions were corrected for natural isotope abundance before normalization using the R package IsoCorrectoR GUI 1.9.0. In figure 4A, data for both labelled and unlabeled conditions were not corrected for natural isotope abundance. In Fig. 5E, ¹³C₃-alanine incorporation into the indicated isotopomer of TCA cycle intermediates was measured in control and lpp>dCTNS-overexpressing animals. Values were corrected for natural isotope abundance before normalization, and data present the fold change ¹³C-labelling in TCA cycle intermediates in dCTNS overexpressing versus control animals.

Cysteine measurement (other than LC-MS/MS)

25 to 40 mid-second-instar animals were homogenized in cold PBS with 0.1% Triton X-100 (PBST) and centrifuged at 4°C. Cysteine measurement was performed in triplicate from the supernatant using the MicroMolar Cysteine Assay Kit (ProFoldin, CYS200) according to the manufacturer’s instructions. Data were normalized to protein content.

Cystine measurements

Larvae were washed three times (in water, quickly in 70% ethanol, and finally in PBS), dried on tissue paper, and 10 larvae per sample were shock frozen in liquid nitrogen and stored at −80°C until lysis. Larvae were lysed in 80 μl of 5.2 mM N-ethylmaleimide, centrifuged for 10 min at 4°C, and 75 μl of supernatant was deproteinized by the addition of 25 μl of 12% sulfosalicylic acid. Protein-free supernatants were kept frozen at −80°C until analysis and centrifuged at 1,200g before use. Cystine quantification was performed using the AccQ-Tag Ultra kit (Waters) on a UPLC-xevoTQD system (Waters) according to the manufacturer’s recommendations. Ten microliters of samples was mixed with 10 μl of a 30 μM internal standard solution (stable isotope of cystine), 70 μl of borate buffer, and 20 μl of derivative solution and incubated at 55°C for at least 10 minutes. Derivatized samples were diluted with 150 μl of ultrapurified water, and 5 μl of the final mixture was injected in the triple quadrupole mass spectrometer in positive ion mode. Transitions used for derivatized cystine quantification and the internal standard were 291.2>171.1 and 294.2>171.1, respectively. Cystine values were normalized by protein content using the Lowry’s method on protein pellets.

Immunostaining

Fat body tissues from 68 to 85 hours AEL larvae were dissected in PBS with 2% formaldehyde at room temperature, fixed 20 to 30 min in 4% formaldehyde, washed twice for 10 min each in PBST 0.3%, blocked for 30 min (PBST, 5% bovine serum albumin (BSA), 2% fetal bovine serum, 0.02% NaN₃), incubated with primary antibodies in the blocking buffer overnight, and washed four times for 15 min each. Secondary antibodies diluted 1:200 or 1:500 in PBST were added for 1 hour and tissues were washed four times before mounting in Vectashield/ 4′,6-diamidino-2-phenylindole (DAPI). Rabbit anti P-4EBP1 from Cell Signaling Technologies (CST 236B4, #2855) was diluted 1:500, rabbit anti-tRFP was from Evrogen (#AB233) and was used against mKate2 to stain dCTNS-mKate2. Purified polyclonal antibody against Mitf was generated in guinea pig by the company Eurogentec using the epitope: CRRFNINDRIKELGTL. Samples were imaged using Zeiss LSM 780 and Leica TCS SP8 SMD confocal systems with a 40x water or oil immersion objective, and images were processed with Fiji software.

Western blots

Tissues from 10 to 30 animals were dissected in CST lysis buffer (Cat#9803) containing 2x protease inhibitor (Roche, #04693159001) and 3x phosphatase inhibitor (Roche, #04906845001), and homogenized using 1mm zirconium beads (Next Advance, #ZROB10) in a Bullet Blender tissue homogenizer (model BBX24, Next Advance). Protein content was measured to normalize samples, 2x Laemmli sample buffer (Bio-rad) was added and samples boiled 6 min @ 95°C. Lysates were resolved by electrophoresis (Mini-PROTEAN TGX Precast Gels, BioRad, PAGEr EX Gels, Lonza, or home-made SDS gels), proteins transferred onto polyvinylidene difluoride (PVDF) membranes (Immobilon P, Millipore), blocked in Tris-buffered saline with or without 0.1% Tween-20 buffer containing 3 to 6% BSA or 5% milk, and probed with P-S6K antibody (1:1000, CST 9209). After P-S6K was revealed, membranes were stripped for 5 to 30 min (Restore PLUS Buffer, Thermo Scientific #46430), washed, blocked in PBS Tween-20 buffer containing 5% dry milk, and probed with S6K antibody [1:10000, a gift from A. Teleman (46)]. For normalization blots were probed with GADPH antibody (1:5000, GeneTex #GTX100118). Data show representative results from at least 2 or 3 biological replicates (see quantification plots). Horseradish peroxidase (HRP)-conjugated secondary IgG antibodies (1:10000) were used together with the SuperSignal West Dura Extended Duration Substrate (Thermo Scientific #34076) to detect the protein bands.

Statistics

Experiments are presented with whisker plots or show the mean +/− SD or SEM. P values and significance were as follows: ns, P≥ 0.05; *, P≤ 0.05; **, P≤ 0.01; ***, P≤ 0.005; ****, P≤ 0.0001. For the life span experiments shown in Fig. 3A, N = 2; for Fig. 3C, E, and G, and fig. S11, N≥ 3. Significance was determined by a two-tailed t-test (Mann-Whitney). N = 1 means average of 8 to 10 vials per genotype and condition in 1 experiment. For the larval development (Pupariation assay) results shown in Fig. 3B and fig. S7G, significance was determined by a two-tailed t test (Mann-Whitney); for Figs. 3D and F; Fig. 6B; fig. S12C and E, one-way ANOVA followed by a Bonferroni’s multiple-comparisons test. For cysteine measurements (profoldin kit) shown in Fig. 2B, significance was determined by t test; for fig. S6A, one way ANOVA followed by Dunnett’s multiple-comparisons test was used; for fig. S7A, E and F two-tailed t test (Mann-Whitney) was used; and for fig. S8B and S9D one-way ANOVA followed by Bonferroni’s multiple-comparisons test was used. For Western blots shown in Fig 1A, the spline curve represents the trend over multiple experiments; for those shown in Fig. 1D and fig. S1A, significance ws determined by two way ANOVA followed by Sidak’s multiple-comparisons test; for Fig. 2, D and F; Fig. 6D; and fig. S12B one-way ANOVA followed by a Bonferroni’s multiple-comparisons test was used. For the metabolomics shown in Fig. 1B (left plots), fig. S12D, Fig. 13B, and fig. S15, C and D, significance was determined by one-way ANOVA followed by a Bonferroni’s multiple comparisons test; in fig. S3A, two-way ANOVA followed by Sidak’s multiple comparisons test was used; for Fig. 1B (right plots); Fig. 2H; Fig. 5, B, C and E; Fig. 6A; fig. S8, C and D; fig. S13, A and C; fig. S15, A, B and E and fig. S3B, two-tailed t test (Mann-Whitney) was used. For pupal weights shown in fig. S4C and food intakes shown in fig. S4D, significance was determined by unpaired two-tailed t test.

Data and materials availability:

All data are available in the main text or the supplementary materials.