Glucosylceramide synthase in the fat body controls energy metabolism in Drosophila.

Glucosylceramide synthase (GlcT-1) catalyzes the synthesis of glucosylceramide (GlcCer), the core structure of major glycosphingolipids (GSLs). Obesity is a metabolic disorder caused by an imbalance between energy uptake and expenditure, resulting in excess stored body fat. Recent studies have shown that GSL levels are increased in obese rodents and that pharmacologically reducing GSL levels by inhibiting GlcCer synthesis improves adipocyte function. However, the molecular mechanism underlying these processes is still not clearly understood. Using Drosophila as a model animal, we report that GlcT-1 expression in the fat body, which is equivalent to mammalian adipose tissue, regulates energy metabolism. Overexpression of GlcT-1 increases stored nutrition (triacylglycerol and carbohydrate) levels. Conversely, reduced expression of GlcT-1 in the fat body causes a reduction of fat storage. This regulation occurs, at least in part, through the activation of p38-ATF2 signaling. Furthermore, we found that GlcCer is the sole GSL of the fat body, indicating that regulation of GlcCer synthesis by GlcT-1 in the fat body is responsible for regulating energy homeostasis. Both GlcT-1 and p38-ATF2 signaling are evolutionarily conserved, leading us to propose an evolutionary perspective in which GlcT-1 appears to be one of the key factors that control fat metabolism.


Nutrient measurements
Nutrient measurements were performed as described previously ( 16 ). Data are averages ± SEM from at least three independent experiments. P values presented are from Student's t -tests. To measure total body TAG, we extracted lipids from the sample (fi ve fl ies or larvae) with 0.5 ml of chloroform-methanol (2:1, v/v). Aliquots of the extract (2 l) were analyzed by TLC (Silica gel 60 plate; Merck, Darmstadt, Germany) and a solvent system consisting of chloroform-methanol-acetic acid (98:1:1, v/v/v). TAG was visualized with charring reagent (Ceric ammonium nitrate/H 2 SO 4 solution), and TAG levels were estimated with an LAS 3000 (Fujifi lm; Tokyo, Japan) set to digitized mode. TAG purifi ed from Drosophila larvae was used as a TAG standard.
To measure the levels of carbohydrates (glucose, trehalose, and glycogen), 10 fl ies were homogenized in 100 l of 0.1% Tween-20 on ice, heated at 70°C for 5 min to inactivate endogenous enzymes, and centrifuged. Five microliters of homogenate were used in each of the assays. Glucose (HK) assay kit (Sigma) and trehalase (Sigma) were used to determine glucose and trehalose levels. The Starch UV method (Roche) was used to determine glycogen levels.

Quantitative RT-PCR
Total RNAs were extracted from the fat bodies of L3 larvae and analyzed with SYBR Green (Power SYBR Green PCR Master Mix; Applied Biosystems) and an Applied Biosystems 7900HT fast real-time PCR System, as described previously ( 16 ). See supplementary experimental procedures online for primers used.

TLC analysis
Total GSLs were extracted from the fat bodies of L3 larvae and analyzed by TLC. Total lipid extracts corresponding to two larvae were applied on a TLC plate (Merck), and the plate was developed with a solvent system of chloroform-methanol-water (65:35:4, v/v/v). GSLs were visualized with orcinol/H 2 SO 4 reagent.

dGlcT-1 controls stored nutrient levels
To characterize the function of dGlcT-1 in the fat body, we fi rst manipulated dGlcT-1 expression in the fat body using a fat body-specifi c driver, FB-GAL4 (see supplementary Fig. I). This driver enabled us to generate dGlcT-1 transgenic fl ies that overexpress dGlcT-1 in the fat body GSLs in the pathogenesis of obesity ( 9 ). Although all of these observations point to the important role of GlcT-1 in fat metabolism in adipocytes, it is diffi cult to understand the effects of adipocyte GlcT-1, since GlcT-1 inhibitor was administered orally.
GlcT-1 is involved also in the production of most complex GSLs such as gangliosides. For example, ganglioside GM3 plays a role in the pathogenesis of type 2 diabetes. Obesity increases GM3 levels, resulting in the exclusion of insulin receptor from lipid rafts or caveolae but the retention of caveolin and fl otillin ( 12 ). Membrane structural changes lead to the inhibition of insulin metabolic signaling ( 12 ). Thus, it is diffi cult to know whether the effect of GlcT-1 in fat metabolism is caused by GlcCer or by complex GSLs.
Drosophila shares most of the same basic metabolic functions found in vertebrates ( 13 ). For example, components of the insulin/insulin-like growth factor (IGF) pathway, which plays a central role in growth and metabolism, are conserved ( 14 ). Drosophila stores fat in a specialized tissue called the fat body. The fat body resembles the adipose tissue and liver of mammals, and metabolizes and stores nutrients primarily as triacylglycerol (TAG) and glycogen. Using Drosophila as an in vivo model system, we manipulated dGlcT-1 expression in the fat body and examined the resulting effects on fat metabolism. We found that the level of dGlcT-1 expression in the fat body regulates fat and sugar metabolism. Moreover, GlcCer was shown to be the dominant GSL in the fat body. GlcT-1 is highly conserved throughout evolution. Thus, our results suggest that GlcCer itself functions in fat metabolism and that manipulating GlcT-1 expression in mammalian adipose tissue may constitute a therapeutic way to counteract obesity.

Fat body analysis by fl uorescence microscopy
Larval fat bodies were dissected from third (L3) instar larvae and examined. Adult fl ies or dissected larvae were fi xed in 4% paraformaldehyde for 25 min at room temperature and washed four times in PBS. To visualize lipid droplets, we stained the fi xed fat bodies with BODIPY 493/503 (Molecular Probes; Carlsbad, CA) or Nile Red (Sigma; St. Louis, MO). BODIPY 493/503 was dissolved in ethanol at 1 mg/ml, and Nile Red was dissolved in acetone at 10 mg/ml. Then BODIPY 493/503 or Nile Red was diluted (BODIPY 493/503, 1:100,000; Nile Red, 1:2,500) in PBS and applied to tissue samples. After incubation for 1 h, tissues were washed three times with PBS. Mitotic clones of dGlcT-1 knockdown cells were generated by heat shock using the following genotypes: hs-fl p UAS-GFP; tub-Gal4/UAS-dGlcT-1 IR; FRT82B/FRT82B Gal80 ( 15 ).
ing that dGlcT-1 affects lipid droplet formation in a cell-autonomous manner.
Fat body TAG levels were ‫ف‬ 18% higher in dGlcT-1overexpressing fl ies and ‫ف‬ 19% lower in dGlcT-1 knockdown fl ies than in wild-type fl ies ( Fig. 1C  Drosophila fl ies store nutrients not only in the form of lipids (TAG), but also in the form of carbohydrates (glucose, trehalose, and glycogen). Because carbohydrates are tightly regulated, we also measured stored carbohydrate levels in whole-fl y homogenates. Flies overexpressing dGlcT-1 stored signifi cantly more carbohydrates than did control fl ies ( Fig. 1D-1F ). However, dGlcT-1 knockdown fl ies displayed markedly decreased levels of trehalose com-and those that downregulate dGlcT-1 expression in the fat body (see supplementary Fig. II). Drosophila stores energy predominantly as triglycerides (TAGs) in the fat body. Thus, to evaluate intracellular TAG content, we stained the fat body with lipophilic dyes that detect lipid droplets. Lipid droplet formation was enhanced in fat bodies overexpressing dGlcT-1 (FB>dGlcT-1) compared with controls ( Fig. 1A ). By contrast, lipid droplet formation was reduced in fat bodies in which dGlcT-1 was knocked down (FB>dGlcT-1 IR) compared with controls ( Fig. 1A ).
To determine whether the regulatory function of GlcT-1 in lipid metabolism is cell autonomous, we generated mitotic clones of dGlcT-1 knockdown cells and examined lipid droplet formation. We observed that the dGlcT-1 knockdown clones contained fewer and smaller lipid droplets than the neighboring control cells ( Fig. 1B ), confi rm-

dGlcT-1 interacts with the p38-ATF2 signaling pathway
The Drosophila p38-ATF2 signaling pathway mediates fat metabolism through the activation of PEPCK transcription ( Fig. 3B ) ( 17 ). Increased FAS and decreased PEPCK mRNA expression phenotypes observed in dGlcT-1 knockdown fl ies were similar to that in dATF2 knockdown fl ies. Thus, GlcT-1 might regulate energy metabolism by regulating the p38-ATF2 signaling pathway. To examine this possibility, we fi rst assessed the level of dp38 phosphorylation in the fat body. Phosphorylation of dp38 in the fat body increased when dGlcT-1 was overexpressed but decreased when GlcT-1 expression was reduced ( Fig. 3A and supplementary  Fig. VA). To confi rm that dGlcT-1 does indeed regulate the p38-ATF2 signaling pathway, we also assessed the genetic interaction between dGlcT-1 and dATF2 in the wing, a site where the p38-ATF2 signaling pathway affects wing pattern formation. We found that dGlcT-1 did indeed regulate this signaling pathway such that dGlcT-1 overexpression or underexpression enhanced or suppressed, respectively, aberrant wing phenotypes (see supplementary Fig. VB). These results support the notion that dGlcT-1 activates the p38-ATF2 signaling pathway in Drosophila .
Given that dGlcT-1 activates the p38-ATF2 signaling pathway in the fat body, it logically follows that dGlcT-1 expression may also affect ATF2 function in fat metabolism in the fat body. To test this possibility, we manipulated dGlcT-1 expression in dATF2-overexpressing fat bodies and assessed PEPCK mRNA levels by collecting mRNAs from fat bodies and performing qRT-PCR. As expected, we observed increased dPEPCK mRNA levels when dGlcT-1 and dATF2 pared with control fl ies, even though glucose and gly cogen levels remained comparable. These observations suggest that dGlcT-1 expression in the fat body affects stored energy levels.

dGlcT-1 alters the expression of key energy metabolism genes
Because dGlcT-1 expression in the fat body affected stored energy levels, we assessed the expression of key genes involved in energy storage ( Fig. 2A ). Acetyl-CoA carboxylase (ACC) is one of the rate-limiting enzymes in FA synthesis, a source of TAG. We quantifi ed the changes in ACC mRNA expression levels in the fat bodies of L3 larvae. We used larval fat bodies for the quantitative RT-PCR (qRT-PCR) assay because adult fat body cells are attached to the cuticle, and it is diffi cult to isolate the fat body without contamination from other tissues. We found that ACC mRNA expression was dramatically decreased in dGlcT-1 knockdown fat bodies ( Fig. 2B ). Western blot analysis using anti-ACC antibody revealed that ACC protein expression was also decreased in dGlcT-1 knockdown fat bodies ( Fig. 2C ).
We also examined other rate-limiting enzymes of FA synthesis and found that decreased phosphoenolpyruvate carboxykinase (PEPCK) mRNA expression increased fatty acid synthase (FAS) expression in dGlcT-1 knockdown fat bodies ( Fig. 2B ). Thus, decreased ACC and PEPCK expression may, in part, underlie the diminished energy storage phenotype of dGlcT-1 knockdown fl ies. Fig. 2. ACC, FAS, and PEPCK mRNA expression is altered in dGlcT-1 knockdown fat bodies. A: Schematic diagram of the pathway and critical enzymes for TAG and glycerol biosynthesis. B: qRT-PCR was used to measure ACC, FAS, and PEPCK mRNAs involved in TAG and glycerol biosynthesis. FB-Gal4 was used to express UAS-dGlcT-1 IR specifi cally in third instar larval fat body; n = 3 for each genotype. C: Western blot analysis using anti-ACC indicated that changes in ACC mRNA expression are in good agreement with the levels of ACC enzyme protein. Proteins were collected from third instar larval fat body. The error bars represent SD.

GlcCer is a prominent GSL in the fat body
Although we demonstrated that dGlcT-1 expression levels affect stored nutrient levels, we needed to clarify whether this was due to the amount of GlcCer or due to more-glycosylated GSLs such as mactosylceramide (MacCer) (see supplementary Fig. VIA). Thus, we examined the GSL composition in the fat body by TLC and LC/MS of larvae fat bodies. We detected only GlcCer ( Fig. 4A ). In the wild-type fat body, however, GSLs such as MacCer or more-glycosylated GSLs were barely detectable. Consistent with this fi nding is that qRT-PCR detected very low levels of endogenous egghead ( egh ) mRNA expression in these fat bodies (see supplementary Fig. VIB). Egh protein catalyzes the transfer of mannose to GlcCer from GDP-mannose to produce MacCer. Furthermore, overexpression or underexpression of egh in the fat body did not affect stored TAG levels (see supplementary Fig.  VIC). Therefore, these results strongly indicated that GlcCer were coexpressed but decreased dPEPCK mRNA levels when dGlcT-1 expression was knocked down ( Fig. 3C ).
Overexpression of dATF2 increases TAG levels, whereas dATF2 knockdown decreases TAG levels ( 17 ). Thus, we also examined whether dGlcT-1 expression affects dATF2mediated effects on stored TAG levels by staining fat bodies with lipophilic dyes. Lipophilic dye signals were stronger when dGlcT-1 was overexpressed with dATF2 in the fat body than when only dATF2 was overexpressed in the fat body ( Fig. 3D ). Furthermore, whole-body TAG levels were ‫ف‬ 45% higher in fl ies that overexpressed dGlcT-1 in the fat body than in fl ies that overexpressed only dATF2 in the fat body ( Fig. 3E ). On the other hand, lipophilic dye signals were lower when dGlcT-1 expression was reduced ( Fig. 3D ), even though whole-body TAG levels were not affected ( Fig. 3E ). These results demonstrate that dGlcT-1 expression in the fat body regulates p38-ATF2 signaling activity. Fig. 3. Genetic interaction between dGlcT-1 and the p38-ATF2 signaling pathway. A: Schematic of the p38-ATF2 signaling pathway. B: Lsp2-Gal4 was used to overexpress or suppress dGlcT-1 specifi cally in the fat body of larvae, and the level of dp38 phosphorylation was analyzed by Western blotting with phospho-p38 antibody. C-E: Effect of dGlcT-1 overexpression or knockdown on PEPCK mRNA (C) and on stored lipid levels (D, E) in dATF2-overexpressing fat bodies of larvae. As described above, Lsp2-Gal4 was used to express dATF2 specifi cally in the fat body. In C, PEPCK mRNA levels were measured by qRT-PCR; n = 3 for each genotype. In E, whole-body TAG levels of third (L3) instar larvae were measured; n = 5 for each genotype. The error bars represent SD.

DISCUSSION
In this paper, we describe how GlcT-1 functions in energy storage. First, our data showed that GlcT-1 in the fat body regulates energy storage. Overexpression of dGlcT-1 in the fat body increased the levels of stored lipid (TAG) and carbohydrates (glucose, trehalose, and glycogen). On the other hand, dGlcT-1 knockdown fl ies displayed markedly decreased levels of TAG and trehalose.
Drosophila stores lipid (TAG) as a source of energy, mainly in the fat body, and reserves this energy until it is required for behavioral activities such as fl ight. In addition to TAG, trehalose released from the fat body is a major source of energy in the hemolymph and thorax muscles itself is responsible for metabolic regulation in the fat body (see supplementary Fig. VII).
Recent drug and in vitro studies have suggested that suppression of GlcT-1 function may present a novel strategy for treating diseases such as obesity (9)(10)(11). In Drosophila and mammals, GlcT-1 function and p38-ATF2 signaling are conserved. Thus, we believe that our in vivo data will be useful for future studies examining the detailed molecular mechanisms underlying the GlcT-1-mediated regulation of energy metabolism. and is consumed during fl ight. Thus, our results suggest that GlcT-1 in the fat body may be one of the key factors involved in energy homeostasis.
A recent study has shown that p38-ATF2 signaling leads to the accumulation of stored energy in Drosophila ( 17 ). In the present study, we showed that dGlcT-1 activates p38-ATF2 signaling. Supporting our results is the fi nding that GlcT-1 activates p38 in mammalian cultured cells ( 21 ). Furthermore, dp38 mutant fl ies are more sensitive to starvation stress than are wild-type fl ies, which is indicative of dp38 involvement in energy metabolism ( 22 ). However, how GlcT-1 regulates dp38 is as yet unclear. We lack empirical evidence for a candidate molecule that interacts with newly synthesized GlcCer. Thus, further analysis is required to elucidate the upstream signaling mechanism that regulates the p38-ATF2 pathway.
Drosophila GlcT-1 generates GlcCer, a precursor glycolipid of most GSLs, which are important components of lipid rafts. Lipid rafts have been suggested to be involved in a variety of dynamic membrane processes such as signal transduction, membrane traffi cking, and stress sensing ( 23 ). Thus, GlcT-1 may affect not only p38-ATF2 signaling but also other signaling pathways that regulate energy metabolism. This may account for the differences in the phenotypes between GlcT-1-and ATF2-transgenic fl ies. Indeed, regulation of body trehalose levels varies across different transgenic fl ies ( 14 ). Knockdown of dATF2 did not change trehalose levels, but knockdown of dGlcT-1 did. Furthermore, we observed increased FAS mRNA levels in dGlcT-1 knockdown fl ies, even though TAG levels decreased. FAS gene transcription is under tight nutritional and hormonal control. Insulin, sterol regulatory element binding protein, and upstream stimulatory factor have been reported to be regulatory elements of FAS transcription ( 24 ). Further studies are necessary to elucidate the specifi c contribution of fat body dGlcT-1 in energy homeostasis.
We showed that GlcCer is a prominent GSL in the fat body. Indeed, MacCer or more-glycosylated GSLs are barely detectable in the fat body. Altering the expression of egh, which catalyzes the reaction that produces MacCer from GlcCer, did not affect stored TAG levels. Thus, it seems likely that GlcCer expression in the fat body is suffi cient to regulate stored energy levels.
We observed that dGlcT-1 overexpression caused the formation of an additional ceramide species, dihydroceramide (GlcCer having d14:0/20:0 and d14:0/22:0). Why was such an unusual GlcCer formed? A plausible explanation for this phenomenon is that excess dGlcT-1 protein localizes to the endoplasmic reticulum (ER) when dGlcT-1 is overexpressed. Under normal conditions, dihydroceramide is an intermediate lipid that is quickly converted to ceramide by desaturase, an enzyme found in ER membranes. Possibly, dGlcT-1 overexpressed in the ER utilizes dihydroceramide as substrate. We have shown previously that dGlcT-1 is expressed not only in Golgi membranes but also in the ER in photoreceptor cells ( 3 ).
Finally, it has been reported that the effects of schlank, a Drosophila ceramide synthase, on TAG metabolism are inde-