ARTICLES

AdPLA ablation increases and prevents obesity induced by high-fat feeding or deficiency

Kathy Jaworski1,4, Maryam Ahmadian1,4, Robin E Duncan1,4, Eszter Sarkadi-Nagy1, Krista A Varady1, Marc K Hellerstein1, Hui-Young Lee2, Varman T Samuel2, Gerald I Shulman2, Kee-Hong Kim3, Sarah de Val1, Chulho Kang3 & Hei Sook Sul1

A main function of white is to release fatty acids from stored triacylglycerol for other tissues to use as an energy source. Whereas endocrine regulation of lipolysis has been extensively studied, autocrine and paracrine regulation is not well understood. Here we describe the role of the newly identified major adipocyte A2, AdPLA (encoded by Pla2g16, also called HREV107), in the regulation of lipolysis and adiposity. AdPLA-null mice have a markedly higher rate of lipolysis owing to increased cyclic AMP levels arising from the marked reduction in the amount of adipose E2 that binds the Gai-coupled receptor, EP3. AdPLA-null mice have markedly reduced adipose tissue mass and content but normal . They also have higher energy expenditure with increased fatty acid oxidation within adipocytes. AdPLA-deficient ob/ob mice remain hyperphagic but lean, with increased energy expenditure, yet have ectopic triglyceride storage and resistance. AdPLA is a major regulator of adipocyte lipolysis and is crucial for the development of obesity.

8 Triacylglycerol in adipose tissue is the major energy storage form hydrolysis of the sn-2 ester bond of phospholipids . The PLA2 in mammals. An imbalance between energy intake and expenditure function in remodeling of membrane phospholipids by can result in excess triacylglycerol accumulation in this tissue, acylation and deacylation cycles8. Additionally, because the sn-2 resulting in obesity1.Inmorbidobesityincreasedadipocyte position of phospholipids is typically enriched in number (hyperplasia) may occur through adipocyte differentiation and other unsaturated fatty acids, PLA2 enzymes catalyze the initial of precursor cells present in adipose tissue2. However, obesity is rate-limiting step in the production of eicosanoids9. Eicosanoids, largely attributed to adipocyte hypertrophy that occurs when including , are potent local mediators of signal trans- triacylglycerol synthesis exceeds breakdown (lipolysis), resulting duction and are known to modulate many physiological systems.

© All rights reserved. 2009 Inc. Nature America, in elevated triacylglycerol storage1,3. Indeed, unlike the triacylgly- These signaling molecules exert autocrine action, paracrine action or 4,5 cerol synthesis that occurs at high levels in other tissues, includ- both through binding to specific G-coupled stimulatory (Gas)or ing in the liver for very low density lipoprotein production, inhibitory (Gai) receptors that can, in turn, modulate a host of effects, lipolysis for the liberation of fatty acids that can then be used including lipolysis through regulation of cAMP levels10. Although as an energy source by other tissues is unique to adipocytes6. their physiological importance is unclear, there are some reports that Furthermore, lipolysis in adipocytes is tightly regulated by hor- suggest prostaglandins may modulate adipocyte differentiation mones that are secreted according to nutritional status. During in vitro11–14. In mature adipocytes, depending on the concentrations fasting,lipolysisisstimulatedbycatecholaminesthatincreasecyclic used, some prostaglandins have been reported to stimulate, inhibit or AMP (cAMP) concentrations, whereas, in the fed state, lipolysis is exert no effect on lipolysis15,16. Regardless, because AdPLA is highly inhibited by insulin1,3. Although regulation of lipolysis by these expressed only in adipose tissue, we hypothesized that it could have a endocrine factors has been extensively studied, the local regulation key role in adipose-specific processes such as lipolysis through of lipolysis in adipose tissue by autocrine and paracrine factors is modulation of arachidonic acid metabolism and its provision for not well understood. prostaglandin biosynthesis. We recently identified an adipocyte (PLA2)by Here we show that AdPLA is the major PLA2 in adipose microarray analysis that we named AdPLA7 (also called PLA2G16, tissue and that it regulates lipolysis in an autocrine and paracrine HRASLS3, HREV107, HREV107-3, MGC118754 or H-REV107-1). manner through PGE2. We report that ablation of AdPLA prevents AdPLA belongs to a newly discovered group of intracellular calcium- obesity from high fat feeding or leptin deficiency by regulating 7 dependent PLA2s . The enzymes in the PLA2 superfamily catalyze lipolysis through the PGE2-EP3-cAMP pathway.

1Department of Nutritional Science and Toxicology, 220 Morgan Hall, University of California, Berkeley, California 94720, USA. 2Department of Internal Medicine, 333 Cedar Street, Yale University School of Medicine, New Haven, Connecticut 06510, USA. 3Department of Molecular and Cell Biology, 447 Life Sciences Addition #2751, University of California, Berkeley, California 94720, USA. 4These authors contributed equally to this work. Correspondence should be addressed to H.S.S. ([email protected]). Received 29 September 2008; accepted 19 November 2008; published online 11 January 2009; doi:10.1038/nm.1904

NATURE MEDICINE ADVANCE ONLINE PUBLICATION 1 ARTICLES

RESULTS specifically expressed in adipocytes and is upregulated by feeding and AdPLA is highly expressed primarily in adipose tissue insulin, and in obesity. As we have shown previously7, the 1.3-kilobase AdPLA messenger RNA and the 18-kDa AdPLA protein are expressed in mice in large AdPLA-null mice are resistant to diet-induced obesity amounts only in (WAT) and To elucidate the physiological role of AdPLA, we used targeting (BAT) depots (Fig. 1a). AdPLA is found exclusively in adipocytes and to generate AdPLA-null mice (Supplementary Fig. 1 online) that we not in the stromal vascular fraction of adipose tissue that contains compared with wild-type littermates on a mixed genetic background preadipocytes (Fig. 1a). In humans, similar to mice, ADPLA mRNA (C57BL/6J and 129 SVJ) as well as on a C57BL/6J background (genetic was undetectable by RT-PCR in skeletal muscle and barely detectable backgrounds of mice used in specific experiments are indicated in the in liver but was abundantly expressed in WAT (Fig. 1a). We measured Methods). Although total body weights did not differ at weaning, at 7 expression of all currently known intracellular PLA2 enzymes .We 11 weeks of age, AdPLA-null mice fed a high-fat diet (HFD) began to detected AdPLA at approximately 1,000–100,000-fold higher levels gain weight at a slower rate than wild-type littermates (Fig. 1d). This (Fig. 1b), indicating that AdPLA is the major PLA2 in adipose tissue. disparity in body weight was exacerbated as mice aged, such that by We investigated AdPLA regulation in different nutritional and hor- 64 weeks of age, AdPLA-null mice fed a HFD weighed only 39.1 ± monal states. The AdPLA mRNA level in epididymal WAT was low in 0.2 g (n ¼ 3) versus wild-type littermates who weighed 73.7 ± 0.3 g fasted mice but markedly rose (by eightfold) after feeding the mice a (n ¼ 3), P o 0.001. The decreased weight gain was also observed in high-carbohydrate, fat-free diet (P o 0.01) (Fig. 1c). AdPLA mRNA AdPLA-null mice fed a standard chow diet (SD), albeit to a lesser expression was also low in adipose tissue of streptozotocin-diabetic extent (Fig. 1d). Food intakes in AdPLA-null and wild-type mice were mice, but it was similarly increased, by approximately tenfold, after equivalent, despite the differences in body weights (0.09 grams per day insulin administration (Fig. 1c). We observed substantially higher per gram body weight, n ¼ 6). AdPLA mRNA expression in WATof genetically obese ob/ob mice that The lower body weight of AdPLA-null mice was largely accounted are hyperinsulinemic and have markedly higher adipose tissue tri- for by a reduction in WAT weight (Fig. 2a). Body composition acylglycerol storage (Fig. 1c) and higher AdPLA protein expression in analysis indicated that AdPLA-null mice had decreased triacylgly- obese ob/ob and db/db mice than in lean wild-type mice in both the cerol content (Supplementary Table 1 online). At 18 weeks of fasted and refed states (Fig. 1c). AdPLA is therefore highly and age, AdPLA-null mice fed a HFD had substantially smaller WAT

a Mouse Human b c

Diabetes 40,000 Fasting FeedingDiabetes + Insulin WT ob/ob WATBrainHeartSI KidneyLiverLungSM BATEpi Ing SVFAd.F SM Liver WAT 30,000 AdPLA ADPLA 20,000 AdPLA AdPLA 1.3 kb ACTB 200 28S 28S mRNA / 28S 2 18S 18S 18S 100 PLA Gapdh mRNA WT db/db ob/ob ND ND ND 0 α β γ δ ε ξ β γ Ren Ing Ren Ing Ren Ing 2 2 2 2 2 2 2 2 HeartLungSpleenLiverPancreasKidneySM Epi Ren Ing BAT AdPLA Fasted AdPLAcPLAcPLAcPLAcPLAcPLAcPLAiPLA iPLA AdPLA 18 kDa AdPLA Refed © All rights reserved. 2009 Inc. Nature America, d 18 weeks 32 weeks WT KO WT KO Figure 1 AdPLA tissue distribution, regulation of WT-HFD expression and body weights of AdPLA-null mice. KO-HFD (a) Top left, northern blot analysis of 10 mg of total WT-SD RNA from various mouse tissues. SI, small intestine; KO-SD SM, skeletal muscle; Epi, epididymal fat; Ing, 50 inguinal fat; SVF, stromal vascular fraction; Ad.F, *** adipocyte fraction. Top right, RT-PCR analysis of RNA 18 weeks 18 weeks *** (2.5 mg) from human SM, liver or WAT for expression WT KO WT KO ** 40 of AdPLA or b-actin. Bottom, western blot analysis Kidney for AdPLA protein in various mouse tissues. 80 mgof WAT Liver * protein was subjected to SDS-PAGE and probed with 30 antibodies to AdPLA. Ren, renal fat. (b) Quantitative Dorsal RT-PCR of RNA from wild-type (WT) renal WAT. Ventral Body weight (g) Values for PLA2 enzymes were normalized to Gapdh ISc WAT 20 Liver mRNA and then expressed relative to cPLA2a mRNA (n ¼ 5). ND, not detected. (c) Top left, northern blot of AdPLA mRNA in epididymal WAT from mice fasted Epi WAT Ing WAT 10 for 48 h, fasted and refed for 12 h or made diabetic 312151869 WT KO WT KO Age (weeks) by streptozotocin injection, with or without insulin replacement (n ¼ 3). Samples were normalized to 18S and 28S bands. Top right, AdPLA mRNA expression in inguinal WAT from WT and ob/ob mice analyzed by northern blotting (n ¼ 3). Bottom, western blotting for AdPLA in WAT depots from WT, db/db and ob/ob mice (n ¼ 3). (d) Top left, representative photographs of male WT and AdPLA-null (KO) mice at 18 and 32 weeks of age. Scale bar, 15 mm. Right, body weights of male WT and KO on either a SD (n ¼ 11) or a HFD (n ¼ 24-33). Bottom left, representative photographs of fat pads and organs of 18-week-old male KO and WT littermates. Isc, interscapular. Scale bar (left), 8 mm; scale bar (right), 10 mm. Results are means ± s.e.m.; **P o 0.01, ***P o 0.001 versus WT.

2 ADVANCE ONLINE PUBLICATION NATURE MEDICINE ARTICLES

depots compared to wild-type littermates with a combined WAT not significantly affect adipocyte differentiation (Fig. 2a,b). mRNA depot weight that was 74% lower (Fig. 2a). We observed the same expression of adipogenic transcription factors17,18 including CCAAT/ pattern of lower body and fat pad weight in AdPLA-null mice fed enhancer–binding protein-a (Cebpa) and peroxisome proliferator- a SD and in AdPLA-null mice on a C57BL/6J background (data activated receptor-g (Pparg) as well as the preadipocyte marker not shown). preadipocyte factor-1 (Pref1, also known as Dlk1) were similar in By standard pathology analysis, AdPLA-null mice showed no wild-type and AdPLA-deficient WAT, as were late markers of adipocyte evidence of any gross, microscopic or functional abnormalities, aside differentiation, including fatty acid synthase (Fasn), diglyceride acyl- from reduced adiposity (Supplementary Fig. 2 online and data not -1 (Dgat1), and adipocyte fatty acid–binding protein (Fabp4, shown). However, as described in later sections, upon HFD feeding we also known as aP2 and aFabp) (Fig. 2b). The DNA content in the fat detected enlarged livers in AdPLA-null mice with increased liver depots of AdPLA-null and wild-type mice also did not differ signifi- triacylglycerol content (Supplementary Table 1), although various cantly (140.4 ± 22.6 mg versus 123.7 ± 14.6 mg DNA per fat pad, liver enzyme levels were largely normal except for alanine amino- respectively, n ¼ 6, P ¼ 0.55).Weusedmouseembryonicfibroblasts transferase (Supplementary Fig. 2). Moreover, blood cell profile and (MEFs) from wild-type and AdPLA-null embryos as well as 3T3-L1 immunological parameters in serum and adipose tissue were not cells transfected with AdPLA or control vector. After undergoing changed in these mice compared to wild-type mice (Supplementary differentiation to adipocytes in medium containing a high concentra- Fig. 2), and the weights of other organs were similar between the two tion of insulin, there were no differences between wild-type and groups (data not shown). AdPLA-null MEFs or between AdPLA-overexpressing and control 3T3-L1 cells with regards to rounded adipocyte morphology or oil AdPLA-null mice have normal adipogenesis red O staining (Fig. 2b). mRNA expression of adipocyte markers such A decrease in adipose tissue mass can result from a reduction in as Cebpa, Pparg, Fabp4 and Pref1 were also similar between wild-type adipocyte size, a reduction in adipocyte number due to impaired and AdPLA-null MEFs, confirming the same degree of differentiation differentiation, or both. Evidence indicates that AdPLA deficiency did (Fig. 2b).

a 8 2.5 b WT KO WT KO WT KO WT-SD AdPLA 7 2.0 WT-HFD Fasn Actb KO-SD Dgat1 MEF 6 1.5 KO-HFD Fabp4 Actb Control AdPLA 5 1.0 Control AdPLA AdPLA

4 Fat pad weight (g) 0.5 *** 1.5 3T3-L1 ** WT Actb *** *** KO 3 * ** 0 Epi Ren Ing 1.5 2 1.0 WT Fat pad weight (% of BW) KO *** 1 ** *** *** ** ** 1.0

mRNA level 0.5 0 Epi Ren Ing mRNA level © All rights reserved. 2009 Inc. Nature America, 8 WT-HFD 0 0.5 Cebpa Pparg Pref1 7 KO-HFD 1.5 6 0 Cebpa Pparg Pref1 Fabp4 1.0 c WT KO

5 g per DNA)

µ 30 0.5 4 ** 20 ***** 0.0 3 TAG (mg WT KO 10 WT

2 Fat pad weight (% of BW) 0 *** 30 1 *** ***

Frequency (%) 20 0 KO Epi Ren Ing 10

Figure 2 AdPLA ablation causes a reduction in fat pad weight, triacylglycerol (TAG) content 0 3 15 27 39 51 63 75 87 99

and adipocyte size but does not affect adipocyte differentiation. (a) Left, fat pad weights as 111 123 135 147 2 a percentage of body weight (BW) and in absolute amounts (inset) from male KO and WT Cell size (× 100 µm ) littermates. Mice were fed a SD or a HFD until 18 weeks of age (n ¼ 6–16). Bottom, fat pad weights of male KO and WT littermates on a HFD at 32 weeks of age (n ¼ 8). Inset, TAG content in epididymal WAT. (b) Top left, RT-PCR for gene products involved in lipid metabolism, using RNA from epididymal WAT of male WT and KO (n ¼ 5). Bottom left, RT-qPCR for adipocyte differentiation markers, using RNA from epididymal WAT of 18 week-old male mice (n ¼ 5). Actb, b-actin. Top right, MEFs from WT and KO embryos differentiated and harvested at day 12 and stained with oil red O. 3T3-L1 cells transfected with LacZ control vector or AdPLA expression vector were also differentiated and stained for neutral lipid. AdPLA mRNA levels were determined in cells by RT-PCR using Actb mRNA as a control. Bottom right, quantitative RT-PCR of adipogenic markers using RNA from adipocytes differentiated from WT and KO MEFs (n ¼ 4). Scale bar, 6 mm. (c) Left, H&E-stained paraffin-embedded sections of epididymal WAT from 18-week-old male KO and WT mice fed a HFD. Scale bar (top), 80 mm; scale bar (bottom), 20 mm. Right, distribution of adipocyte size. Results are means ± s.e.m.; *P o 0.05, **P o 0.01, ***P o 0.001 versus WT.

NATURE MEDICINE ADVANCE ONLINE PUBLICATION 3 ARTICLES

ab c d Glycerol Fatty acids

Basal StimulatedBasal Stimulated 1.4WT 0.25 2.5 2.5 2.5 2.5 KO WT WT *** 1.2 ) * 0.15 0.20 *** –1 2 KO 2 2 KO 2

1.0 g * –1 0.15 1.5 1.5 1.5 1.5 0.8 *** *** 0.10 *** *** *** 0.6 0.10 1 1 *** 1 1 mol per 100 mg tissue) 0.4 Esterification 0.05

µ ** 0.05 * mol per 100 mg tissue) *** C-palmitate per mg tissue) 0.5 0.5 0.5 *** 0.5 µ 0.2 Lipolysis (g d *** 14 *** 0.0 0.00 0.00 0 0 0 0 FFA ( FFA

Fractional TAG-glycerol synthesis TAG-glycerol Fractional Gon lng Gon lng WT KO 01234 01234 01234 01234 (nmol Glycerol ( Time (h) Time (h) Time (h) Time (h) Figure 3 AdPLA ablation increases lipolysis in vivo, ex vivo and in vitro.(a) Fractional efWT basal in vivo synthesis of triacylglycerol (TAG)-glycerol in gonadal (Gon) and inguinal (Ing) 1.4 KO basal ** WT iso WAT of 24-week-old female mice on a HFD (n ¼ 5or6).(b) In vivo lipolysis in 1.50 1.2 gonadal and inguinal WAT from 24-week-old female mice on a HFD measured as KO iso 1.25 1.0 the change in adipose tissue TAG mass per day per gram of total adipose tissue mass *** ** (g d–1 g–1;n ¼ 3–6). (c) 14C-palmitate esterification into TAG in WAT explants. 1.00 0.8 (d) Basal and stimulated (+ 100 nM isoproterenol) lipolysis, as measured by glycerol 0.75 0.6 **

(left) and fatty acids (right) released from explants of epididymal WAT from overnight mol per mg protein)

0.50 µ 0.4 fasted 16-week-old male WT and KO mice on a HFD (n ¼ 5). (e) Molar ratio of FFA to 0.25 0.2

glycerol release from WAT explants. (f) Best-fit lines of basal and stimulated lipolysis, ( FFA as measured by fatty acids released from WT and KO MEFs at day 12 after glycerol) (FFA: Molar ratio 0.00 0 differentiation into adipocytes. MEFs were incubated with or without isoproterenol (iso) WT KO WT KO 01234 Basal Stimulated Time (h) at 200 nM (n ¼ 6). Results are means ± s.e.m.; *P o 0.05, **P o 0.01, ***P o 0.001.

Histological analysis showed that epididymal WAT from AdPLA- AdPLA-null mice (Fig. 3e). Because of the heterogeneity of cell null mice contained a significantly greater frequency (P o 0.05) of populations within WAT explants, we also examined the effect of small adipocytes and a lower frequency of midsized and large AdPLA ablation on lipolysis in adipocytes differentiated from MEFs adipocytes (P o 0.01; Fig. 2c). We also found substantially decreased that were isolated from AdPLA-null embryos. Probably owing to the triacylglycerol content in adipose tissue of AdPLA-null mice (Fig. 2a). very high concentration of insulin in the medium, the basal rate of AdPLA ablation therefore causes adipocyte hypotrophy owing to lower lipolysis was very low in these cells, and we could not detect differences triacylglycerol accumulation and smaller adipocyte size rather than in basal lipolysis between adipocytes differentiated from wild-type and impaired adipocyte differentiation. AdPLA-null MEFs (Fig. 3f). This finding is in agreement with the comparable lipid content in MEF-differentiated adipocytes (Fig. 2b). AdPLA ablation increases lipolysis in adipose tissue However, isoproterenol-stimulated lipolysis, measured as free fatty acid To investigate whether the markedly decreased adiposity observed release, was significantly higher in adipocytes differentiated from in AdPLA-null mice was the result of increased lipolysis, we measured AdPLA-null MEFs (Fig. 3f). These results clearly indicate that lack in vivo triacylglycerol metabolism over a two-week period with a of AdPLA results in increased lipolysis in adipose tissue.

© All rights reserved. 2009 Inc. Nature America, recently developed heavy-water labeling technique19.Thefractional contribution of triacylglycerol-glycerol synthesis to adipose tissue AdPLA-null mice have reduced adipose PGE2 levels triacylglycerol was significantly higher in AdPLA-null mice, reflecting Phospholipase A activity in WATwas considerably reduced, by 82%, in a greater than twofold higher rate of replacement of preexisting AdPLA-null mice (from 3.71 pmol mg–1 min–1 in wild-type mice to –1 –1 triacylglycerol molecules with newly labeled triacylglycerol-glycerol, only 0.66 pmol mg min ), whereas expression of other PLA2 indicating increased triacylglycerol turnover (Fig. 3a). In agreement, enzymes were unchanged (Fig. 4a). This finding establishes AdPLA the net in vivo lipolytic rate, calculated from the absolute rate of new as the major PLA in adipose tissue. Given the inverse relationship triacylglycerol synthesis and the change in adipose tissue mass, was between PLA activity and lipolysis in AdPLA-null mice, we hypothe- also four- to sixfold higher per gram of adipose tissue in AdPLA-null sized that AdPLA may function in adipose tissue to regulate lipolysis mice compared to wild-type mice (Fig. 3b). We measured [U-14C]pal- locally by generating arachidonic acid for the production of prosta- mitate incorporation into triacylglycerol in adipose tissue explants and glandins. We first measured the abundance of prostaglandins known to found similar incorporation in wild-type or AdPLA-null mice, indi- be found in adipose tissue in wild-type mice. PGE2 was present in cating comparable WAT fatty acid esterification in these mice significantly larger amounts than any other prostaglandins, which were (Fig. 3c). Therefore, we conclude that triacylglycerol hydrolysis, but detected in amounts well below the effective concentrations required not synthesis, is altered in WAT of AdPLA-null mice. for binding to their cognate receptors20 (Fig. 4b). We next determined To further investigate the effect of AdPLA ablation on lipolysis, we the expression of the receptors for these prostaglandins in adipose compared basal and stimulated glycerol and fatty acid release from tissue and found that EP3 was more highly expressed than other explants of WAT from AdPLA-null and wild-type mice. Under basal receptors examined, including IP, DP and EP1 (all undetectable), as conditions, the rates of glycerol and fatty acid release were significantly well as EP2 and EP4, which were present in much smaller amounts higher in adipose tissue from AdPLA-null mice (Fig. 3d). Furthermore, (Fig. 4b). When we compared the expression of various prostaglandins lipolysis was also higher in AdPLA-deficient adipose tissue when in adipose tissue from AdPLA-null and wild-type mice (Fig. 4b), we 12,14 treated with isoproterenol, which stimulates lipolysis by increasing found that expression of PGF2a and 15-deoxy-D -PGJ2,whichhave cAMP levels (Fig. 3d). Notably, the molar ratio of free fatty been reported to affect adipocyte differentiation in vitro13,21,22,was acid (FFA) to glycerol released from WAT explants was lower in unchanged in AdPLA-deficient adipose tissue (Fig. 4b). However,

4 ADVANCE ONLINE PUBLICATION NATURE MEDICINE ARTICLES

WT WT ab5 1.2 cde 1.4 WT iso 2.0 KO 10 *** 30 KO WT KO WT KO 1 P-HSL KO iso ** ) 9 * –1 1.5 0.8 1.2 KO iso + PG 25 HSL 4 8 0.6

min 1.0 Desnutrin 0.4 1.0 –1 7 20 Gapdh ** 0.5 mRNA level 0.2

3 mRNA level 6 0 0.8 0.0 5 15 β γ DP IP α EP1EP2EP3EP4 2 2 2 350 *** 0.6 2 4 300 WT ** cPLA iPLA iPLA WT 10 KO mol per mg protein) 3 250 µ 0.4 KO 200 1 2 ** 150 *** 5 ( FFA 0.2 cAMP (pmol per mg protein) 100 PG content (ng per mg protein) 1 WT) (% of PLA activity (pmol mg Protein level 50 0 0 ** 0 0 0 WT KO α WT KO Desnutrin HSL P-HSL 0 1 2 34

PGE2 PGJ2 Time (h) 15-deoxy PGF2 NS PGl2 fgh** 35 *** *** 25 * 8 Figure 4 AdPLA deficiency increases lipolysis by decreasing PGE2 30 *** 7 abundance and increasing cAMP levels. (a) Total phospholipase A 20 (PLA) activity in epididymal WAT from 16-week-old male WT and 25 ) 6 cells per h) –1 cells per h) 6 6 15 5 KO mice fed a HFD (n ¼ 3). Inset, PLA2 expression in WAT. 20 4 (b) Prostaglandin (PG) content in epididymal WAT of 18-week-old 15 10 3 ** male WT and KO mice on a HFD (n ¼ 5). Inset, quantitative 10 (mg dl FFA 2 mol per 10 5 RT-PCR of prostaglandin receptors normalized to Actb,inWATof µ 5 1 WT mice (n 5). (c) cAMP abundance in epididymal WAT from ¼ 0 0 FFA ( FFA 0

2 2 cAMP (pmol per 10 male WT and KO mice on a HFD (n ¼ 5). (d)Immunoblotof 2 L826266 – – + + WT KO WT KO WT KO phosphorylated HSL (P-HSL), HSL, desnutrin and Gapdh (control) PGE2 – ++–

(top) with relative quantification (bottom). (e) Best-fit lines of KO + PGE KO + PGE KO + PGE stimulated lipolysis, as measured by fatty acid release from WT and ADA Isoproterenol KO MEFs on day 12 after differentiation into adipocytes. MEFs were incubated with 200 nM isoproterenol and 100 nM PGE2 (PG) as indicated (n ¼ 6). (f) Lipolysis in isolated adipocytes from KO or WT mice incubated with –1 1Uml adenosine deaminase (ADA) or isoproterenol (200 nM) and treated with or without 10 nM PGE2.(g) cAMP levels in isolated adipocytes from WT or AdPLA-null mice treated with or without 10 nM PGE2. NS, not significantly different. (h) Lipolysis in isolated adipocytes treated with the EP3 antagonist L826266 (10 mM) with or without 10 nM PGE2. Results are means ± s.e.m.; *P o 0.05, **P o 0.01, ***P o 0.001.

PGE2 expression was reduced to 12% of wild-type expression in into ob/ob mice to generate double-knockout mice. On a SD, double- AdPLA-deficient adipose tissue (Fig. 4b). PGI2 amounts were also knockout mice gained substantially less weight than did ob/ob significantly reduced (Fig. 4b). However, PGI2 concentrations were mice (Fig. 5a,b). Noteworthy differences in body weights were well below those required for receptor activation, and IP was not apparent by as early as 6 weeks of age and became even more detected in adipose tissue (Fig. 4b). Because PGE2 and EP3 were found pronounced with age (Fig. 5b). To our surprise, the differences were in adipose tissue at the highest concentrations, and because PGE2 levels not attributable to a reduction in food consumption, as food intakes were the most markedly affected of the prostaglandins in AdPLA-null were, in fact, somewhat increased in double-knockout mice compared

© All rights reserved. 2009 Inc. Nature America, adipose tissue, we postulated that decreased activation of EP3 by PGE2, to ob/ob mice (Fig. 5b). However, double-knockout mice showed may have been a major contributor to the observed effects. reduced adiposity, with a marked reduction in the weight of WAT If AdPLA regulates the production and release of PGE2, which can depots compared to ob/ob mice (Fig. 5c). Other organ weights, except bind Gai-coupled EP3, then cAMP levels should be increased in for the increased liver weight, were similar between the two groups adipose tissue of AdPLA-null mice. Indeed, we detected an approx- of mice (data not shown). Carcass analysis showed that body and imate twofold increase in cAMP concentrations relative to wild-type carcass weights of AdPLA-null and double-knockout mice at 40 weeks mice (Fig. 4c) and found significantly higher phosphorylation of of age were less than those of either wild-type or ob/ob mice, whereas hormone-sensitive (HSL) in the absence of a change in HSL or the percentage of water and lean tissue mass was increased (Fig. 5d). desnutrin (also called ATGL) protein expression (Fig. 4d). Our results Percentage lipid was reduced by 81% in AdPLA-null mice and 69% support the idea that AdPLA expression induced by feeding and in double-knockout mice compared to wild-type and ob/ob insulin inhibits lipolysis by increasing PGE2 levels, which, in turn, mice, respectively, reflecting the substantially leaner phenotype in activates EP3 and thereby decreases cAMP levels. In this regard, AdPLA deficiency. addition of exogenous PGE2 decreased lipolysis to wild-type levels We observed significantly higher lipolysis in double-knockout in differentiated adipocytes derived from AdPLA-null MEFs (Fig. 4e) mice under both basal and stimulated conditions compared to ob/ob and in WAT explants (data not shown), as well as in isolated mice (Fig. 5e) that was accompanied by an 86% reduction in PGE2 adipocytes from AdPLA-null mice (Fig. 4f). The addition of PGE2 amounts (Fig. 5f) and a fourfold increase in cAMP levels (Fig. 5g). also restored cAMP levels in adipocytes isolated from AdPLA-null In ob/ob mice, PGE2 concentrations were 2.6-fold higher com- mice (Fig. 4g). Addition of L826266, an EP3 antagonist, prevented the pared to wild-type mice, possibly reflecting their higher AdPLA levels antilipolytic effect of PGE2 in adipocytes (Fig. 4h), providing evidence (Fig. 5f). Notably, PGE2 expression in double-knockout mice did not that EP3 mediates the antilipolytic effect of PGE2. differ from that in AdPLA-null mice, indicating that AdPLA dom- inantly regulates PGE2 levels in adipose tissue (Fig. 5f). Taken AdPLA ablation prevents obesity in ob/ob mice together, these results indicate that AdPLA deficiency markedly To test whether AdPLA deficiency could prevent genetic obesity such influences PGE2 and cAMP levels to modulate lipolysis and adiposity, as that caused by leptin deficiency, we introduced AdPLA deficiency even in genetic obesity caused by leptin deficiency.

NATURE MEDICINE ADVANCE ONLINE PUBLICATION 5 ARTICLES

AdPLA ablation causes insulin resistance AdPLA-null and ob/ob a ob/ob DKO b 80 Changes in adiposity are often associated WT 75 with alterations in glucose and insulin home- ob/ob 70 AdPLA-null ostasis. AdPLA-null mice fed a SD showed an AdPLA-heterozygous 65 and ob/ob attenuated response to insulin during an 60 insulin tolerance test (ITT) but showed glu- 55 cose tolerance similar to wild-type mice 50 9 8

(Supplementary Fig. 3 online), probably 45 )

–1 7 owing to elevated insulin secretion (Supple- 40

Body weight (g) Body weight 6 mentary Table 2 online). AdPLA-null mice 35 5 fed a HFD had significantly impaired glucose 30 4 Liver 3 25 clearance during a glucose tolerance test 2 Food intake (g d intake Food (GTT) and an impaired response to insulin Epi 20 1 WAT during ITT compared to wild-type litter- 15 0 5 6789101112 13 14 15 16 17 ob/ob DKO mates (Fig. 6a). In addition, we found that Age (weeks) AdPLA ablation further impaired glycemic 3.0 120 c *** d *** WT control in ob/ob mice on both a SD (data not KO 2.5 *** 100 *** shown) and a HFD, making double-knock- ob/ob DKO out mice even more glucose and insulin 2.0 *** 80 ****** *** intolerant than the already impaired ob/ob 1.5 60 *** mice (Fig. 6b). In agreement with our find- *** *** ings from the GTT and ITT, AdPLA-null 1.0 ** 40 Fat pad weight (g) pad weight Fat *** ** *** mice fed a SD showed normal fasting glucose 0.5 *** 20 but elevated serum insulin, whereas HFD-fed 0 0 mice had elevated fasting serum glucose and Epi Ren Ing Body Carcass Water Lipid Lean weight weight (%) (%) tissue insulin compared to wild-type mice (Supple- e 1.6 ob/ob basal (g) (g) (%) mentary Table 2). Serum glucose and insulin DKO basal 1.4 ob/ob iso * levels were also elevated in double-knockout DKO iso fg 1.2 * compared with ob/ob mice on either diet 3.0 30 ** 1.0 (Supplementary Table 2). * 2.5 25 0.8 * To discern the impact of AdPLA ablation 2.0 20 0.6 * on peripheral and hepatic insulin action, we 1.5 * 15 0.4 performed a hyperinsulinemic-euglycemic change) (fold 2 1.0 10 0.2 clamp. The glucose infusion rate necessary (mmol per 100 mg tissue) FFA PGE 0.5 5 to maintain euglycemia in AdPLA-null mice 0 024 0 cAMP (pmol per mg protein) 0 was 77% lower than in wild-type mice, Time (h) WT KO ob/ob DKO WT KOob/ob DKO indicating severely blunted insulin-stimu-

© All rights reserved. 2009 Inc. Nature America, lated glucose uptake and metabolism Figure 5 AdPLA deficiency prevents obesity in ob/ob leptin-deficient mice. (a) Top, representative (Fig. 6c and Supplementary Fig. 3 online). photographs of 16-week-old male ob/ob and double-knockout mice fed a SD. Scale bar, 8 mm. Bottom, Compared with wild-type mice, AdPLA-null representative photographs of their livers and epididymal WAT. Scale bar, 6 mm. (b) Left, body weights mice showed a 50% decrease in whole-body of female mice on a SD. Right, food intake in 12-week-old male mice fed a SD. (c) Comparison of weights of WAT depots from WT, KO, ob/ob and double-knockout mice. (d) Carcass analysis of 40-week- glucose uptake (Fig. 6d) and 44% and 65% old male mice fed a HFD. (e) Basal and stimulated lipolysis measured by fatty acid release from decreases (P o 0.05) in and glyco- explants of epididymal WAT in 12-week-old male ob/ob and double-knockout mice fed a HFD. gen synthesis, respectively (data not shown), (f,g)PGE2 (f) and cAMP (g) abundance in WAT of 12-week-old male WT, KO, ob/ob and double- indicating decreased glucose metabolism. In knockout mice fed a HFD. Results are means ± SEM; *P o 0.05, **P o 0.01, ***P o 0.001. addition, we found that suppression of hepa- tic glucose production by insulin during the clamp was severely Lipid staining with oil red O also showed higher intramyo- blunted in AdPLA-null mice (Fig. 6e), indicating hepatic insulin cellular triacylglycerol content in skeletal muscle, although we found resistance. We found no difference in gastrocnemius 2-deoxyglucose no difference in skeletal muscle diacylglycerol content (data uptake between wild-type and AdPLA-null mice (Fig. 6f). Notably, not shown) or insulin receptor substrate-1 (IRS-1) phosphorylation epididymal WAT 2-deoxyglucose uptake per gram was higher in (Supplementary Fig. 3). AdPLA-null mice, although the substantial decrease in adipose tissue We found that circulating amounts of leptin and adiponectin were mass resulted in a 72% reduction in total insulin-stimulated glucose decreased in AdPLA-null mice on either a SD or a HFD (Supple- uptake by this tissue (Fig. 6g), which may explain the lower net whole mentary Table 2). Relative mRNA expression of these adipokines in body glucose metabolism in these mice. WAT from HFD-fed mice was also decreased (1.0 ± 0.30 versus 0.25 ± The livers of AdPLA-null mice were pale tan in color and enlarged 0.07, P o 0.01 for leptin; 1.0 ± 0.26 versus 0.28 ± 0.05, P o 0.01 for with numerous, lipid-laden vacuoles in hepatocytes (Supplementary adiponectin, in wild-type and AdPLA-null mice, respectively). Circu- Fig. 3). Liver diacylglycerol concentrations, which have been asso- lating amounts of adiponectin were lower in ob/ob mice compared to ciated with insulin resistance23, were also significantly increased in double-knockout mice (Supplementary Table 2). Adiponectin has AdPLA-null mice compared with wild-type mice (3.05 ± 0.46 mgmg–1 been shown to increase fatty acid oxidation in skeletal muscle24.We tissue versus 1.47 ± 0.38 mgmg–1 tissue, respectively, P o 0.05). found decreased phosphorylation of AMP-activated protein kinase

6 ADVANCE ONLINE PUBLICATION NATURE MEDICINE ARTICLES

abcd500 300 KO 700 e * 700 DKO 100 100 25 ** WT ob/ob WT 250 600 600 * ) ) ** 400 ) * * KO

** –1 –1

–1 80 80 20 ) ** ) 200 ** 500 500 * ** –1 –1

300 min ** 400 400 –1 60 60 15 min 150 *** min –1 *** –1 200 * 300 300 100 40 40 10 200 200 * Glucose (mg dl Glucose (mg dl

100 (mg kg 50 (mg kg 100 100 20 ** 20 5 GINF (mg kg 0

0 Hepatic glucose production

0 0 Whole-body glucose uptake 0 30 60 90 120 0 30 60 90 120 0 306090120 0 30 60 90 120 0 0 0 Time (min) Time (min) Time (min) Time (min) WT KO WT KO Basal Clamp fg h3,500 WT ob/ob i j 25 KO dTKO * 200 25 ) 3,000 * –1 20 6 WT * ) 2,400

min *** 15 –1 2,500 KO

–1 5 20 h ) 150 10 –1 ) * –1 2,000 –1

Glucose uptake 4 15 (nmol g 5 * 1,600 cells min) min

1,500 6 –1 100 0 (ml kg WT KO 2 3 10 1,000 VO mRNA level (nmol min 2

Glucose uptake 800 (nmol g 50 500 5 Total glucose uptake Total

* C]palmitic acid oxidation 1 0 14 (c.p.m. per 10 (c.p.m.

0 0 [U- 0:311:43 3:315:19 7:078:19 WT KO WT KO 19:4320:5522:43 10:0711:1913:0714:55 0 0 Dio2 Ucp1 Ppard WT KOob/ob DKO Time of day

Figure 6 AdPLA deficiency impairs glycemic control, increases energy expenditure and promotes fatty acid oxidation in WAT. (a) GTT (left) and ITT (right) in 18-week-old male WT and KO mice fed a HFD (n ¼ 7). (b) GTT (left) and ITT (right) in 14-week-old male ob/ob and double-knockout mice fed a HFD (n ¼ 8or9).(c–g) Results from hyperinsulinemic euglycemic clamp performed in 12-week-old male WT and KO mice fed a HFD (n ¼ 4or5).(c) Average glucose infusion rate (GINF). (d) Whole-body glucose uptake. (e) Hepatic glucose production (HGP) under basal and clamp conditions. (f) Glucose uptake by skeletal muscle (gastrocnemius). (g) Total glucose uptake and glucose uptake per gram of tissue (inset) in epididymal WAT. (h) Oxygen consumption rate (VO2) determined via indirect calorimetry during the light (7 a.m.–7 p.m.) and dark (7 p.m.–7 a.m.) period in 18-week-old male KO and WT mice on a SD (n ¼ 3–6). (i) Quantitative RT-PCR for Ucp1, Dio2 and Ppard, using RNA from epididymal fat from 20-week-old male WT and KO mice fed a SD (n ¼ 3 14 14 or 4). (j) Oxidation of [U- C]palmitate to CO2 by adipocytes isolated from WT, KO, ob/ob and double-knockout mice (n ¼ 3). Results are means ± s.e.m.; *P o 0.05, **P o 0.01, ***P o 0.001.

(AMPK) at Thr172 and decreased expression of acyl-CoA oxidase in found that fatty acid oxidation was increased by 37% in isolated skeletal muscle of AdPLA-null mice, as well as lower oxidation of adipocytes from AdPLA-null mice compared to wild-type mice [U-14C]-palmitate (Supplementary Fig. 3). (Fig. 6j). Fatty acid oxidation was significantly lower in adipocytes Despite increased lipolysis in AdPLA-null mice, serum nonesterified from ob/ob mice than in those from wild-type mice but was restored fatty acid (NEFA) levels were not higher but lower on both a SD and a to wild-type levels in double-knockout mice. We observed a decrease HFD (Supplementary Table 2). Serum were also lower in in the molar ratio of FFA to glycerol released from WAT explant AdPLA-null and double-knockout mice, despite pronounced steatosis studies of AdPLA-null mice compared to wild-type mice (Fig. 3e).

© All rights reserved. 2009 Inc. Nature America, and hepatic insulin resistance (Supplementary Table 1). We found This change in the molar ratio of fatty acid to glycerol in AdPLA-null increased triacylglycerol clearance in mice upon oral feeding of lipids mice indicates increased use of FFA within adipocytes and supports but no difference in serum triacylglycerol when we first injected the our finding of increased fatty acid oxidation in adipose tissue. These pharmacologic agent WR1339, which inhibits lipoprotein removal findings help to explain not only increased energy expenditure but from the circulation (Supplementary Fig. 4 online). Lipoprotein also decreased NEFA in the circulation, despite higher lipolysis in lipase expression was twofold higher in livers from AdPLA-null mice AdPLA-null mice. (Supplementary Fig. 4). Taken together, the results show that fatty acid uptake by the liver seems to be higher in AdPLA-null mice. DISCUSSION AdPLA is highly expressed only in adipose tissue, where lipolysis AdPLA ablation raises energy expenditure and fat oxidation is a major function. We postulated that, because PLA2 enzymes Total oxygen consumption was higher in AdPLA-null and double- catalyze the initial rate-limiting step in prostaglandin synthesis9, knockout mice compared with wild-type and ob/ob mice, respectively AdPLA may regulate lipolysis locally by controlling the provision of (Fig. 6h), and these differences were not attributable to changes in arachidonic acid for the production of prostaglandins. As AdPLA is ambulatory activity (data not shown). Thus, we next examined BAT induced by feeding and insulin, as well as in the ob/ob and db/db but found no significant difference in the morphology or weight of models of obesity, we predicted that AdPLA probably has an inhibi- interscapular BAT, nor did we find any change in mRNA levels of tory role in lipolysis. Indeed, we found that AdPLA has a major role in involved in thermogenesis in BAT when mice were housed at modulating adipose tissue lipolysis by regulating PGE2 abundance. As 4 1C (data not shown). In WAT from AdPLA-null mice, however, we a result, ablation of AdPLA in mice prevented obesity induced by detected a 5.5-fold upregulation in mRNA levels of uncoupling feeding on a HFD or by leptin deficiency. protein-1 (Ucp1), as well as upregulation of other genes involved in The local regulation of lipolysis in adipose tissue by autocrine oxidative metabolism, including 3.2-fold and fivefold increases in and paracrine factors such as PGE2 has not been studied extensively expression in mRNA levels of peroxisome proliferator-activated compared to the regulation of this process by endocrine factors receptor-d (Ppard), and deiodinase-2 (Dio2), respectively (Fig. 6i). such as catecholamines and insulin25. We found that AdPLA is 14 14 We determined the production of CO2 from [U- C]palmitate and expressed in adipocytes at a much higher level than any other

NATURE MEDICINE ADVANCE ONLINE PUBLICATION 7 ARTICLES

known PLA2 enzymes, and total phospholipase A activity was greatly these prostaglandins are not detected at high enough concentrations in reduced in adipose tissue of AdPLA-null mice whereas expression of adipose tissue to effectively bind their receptors20,whichwerealso other known PLA2 enzymes remained unchanged, revealing that barely detectable in adipose tissue. Therefore, prostaglandins other AdPLA is the major PLA2 enzyme in this tissue and may provide than PGE2 probably do not have a substantial role in modulating arachidonic acid for prostaglandin synthesis. Among the prostaglan- adipocyte processes, including differentiation. In our study, PGE2 was dins that have previously been detected in adipose tissue, we found the most abundant prostaglandin found in adipose tissue. However, that PGE2 is present at levels one to two orders of magnitude higher our results clearly show that adipocyte differentiation, either in vitro or than those of other prostaglandins. Furthermore, PGE2 levels are in vivo, was not affected by AdPLA. markedly decreased in AdPLA-null and double-knockout mice, The increased cAMP concentrations in WAT of AdPLA-null providing evidence for a key role for AdPLA in regulating PGE2 and AdPLA and leptin double-knockout mice resulted in increased synthesis. We found that the Gai coupled receptor, EP3, which inhibits lipolysis and, consequently, decreased adipocyte size with lower and therefore negatively regulates cAMP levels, was triacylglycerol content, despite similar food intakes. Although predominant in adipose tissue. We also found that a specific inhibitor increased lipolysis in AdPLA-null mice resulted in ectopic triacylgly- of EP3 prevented the antilipolytic effect of PGE2. These findings cerol accumulation in the liver and skeletal muscle, this effect could suggest that PGE2 suppresses lipolysis by decreasing cAMP levels not fully account for the loss of triacylglycerol from adipose tissue. through EP3 activation during feeding when AdPLA is strongly Indeed, oxygen consumption was increased in AdPLA-null and induced by insulin. Indeed, we found that cAMP levels were elevated double-knockout mice. Although we did not detect any changes in WAT of AdPLA-null and double-knockout mice, and PGE2 treat- in BAT, to our surprise, Ucp1, Dio2 and Ppard mRNA expression ment restored cAMP levels in AdPLA-null adipocytes to wild-type in WAT of AdPLA-null mice was substantially increased, suggesting levels. In AdPLA-null WAT, HSL phosphorylation was increased, higher oxidation and thermogenesis in WAT. Ectopic expression of whereas total amounts of HSL and desnutrin were unchanged, Ucp1 in WAT has been reported to cause resistance to diet-induced suggesting that phosphorylation of HSL through cAMP-mediated obesity with increased fatty acid oxidation in adipocytes32. Consistent activation of protein kinase A is probably a key mediator of increased with this, we found significantly increased fatty acid oxidation in lipolysis in these mice. adipocytes from AdPLA-null and double-knockout mice compared In vivo, lipolysis was markedly higher per gram of adipose tissue with wild-type and ob/ob mice, respectively, indicating that, at least in in AdPLA-null mice compared to wild-type mice. Furthermore, part, increased fatty acid utilization within adipocytes contributed to both basal and isoproterenol-stimulated lipolysis were increased in the increased energy expenditure. explants of AdPLA-deficient adipose tissue and in adipocytes AdPLA-null and double-knockout mice are extremely lean but isolated from AdPLA-null mice, and stimulated lipolysis was insulin resistant. Results from hyperinsulinemic-euglycemic clamping higher in adipocytes differentiated from AdPLA-null MEFs. studies indicate that insulin resistance in AdPLA-null mice is due Exogenous PGE2 rescued lipolytic rates in AdPLA-deficient adipocytes to hepatic insulin resistance as well as reduced peripheral glucose in all three model systems, showing AdPLA regulation of lipolysis metabolism. Notably, there was no difference in insulin-stimulated in adipocytes via PGE2 production. The inhibitors skeletal muscle glucose uptake between wild-type and AdPLA-null NS-398 (ref. 26) and indomethacin27 have been shown to enhance mice, but total insulin-stimulated glucose-uptake was lower in WAT basal and stimulated lipolysis, respectively, in adipose tissue, as a result of the marked decrease in the mass of this tissue. It is consistent with our present findings that AdPLA has a major part noteworthy that, despite severe insulin resistance and increased lipo- in regulating lipolysis through production of PGE2. Furthermore, lysis in AdPLA-null mice, serum NEFA concentrations were lower in

© All rights reserved. 2009 Inc. Nature America, we show that an EP3 inhibitor prevents the antilipolytic effects these mice. The molar ratio of FFA to glycerol release from adipose of PGE2, which is consistent with our proposed model of tissue was considerably lower in AdPLA-null mice, suggesting increased lipolysis regulation by AdPLA through PGE2 signaling. Owing to utilization of fatty acids within adipose tissue. Consistent with this increased phosphorylation of HSL and unsuppressed lipolysis, finding, we observed increased fatty acid oxidation in adipose tissue. AdPLA-null mice have markedly decreased triacylglycerol content in Ectopic storage of triacylglycerol in liver and skeletal muscle suggests adipose tissue. This indicates that suppression of lipolysis by the that removal of FFA from the circulation by these tissues may local PGE2 produced by adipocytes is crucial in regulating potentially have been increased and therefore may also have contrib- adipocyte lipolysis and shows a role for the AdPLA-PGE2-EP3- uted to decreased serum NEFA abundance. However, it is most cAMP signaling pathway in development of excess adipose tissue likely that even with the higher rate of lipolysis, the markedly decreased mass, triacylglycerol storage and obesity. In support of this, PGE2 adipose tissue mass in AdPLA-null mice resulted, overall, in lower net abundance was found to be higher in adipose tissue of obese human FFA liberation. Indeed, other mouse models with increased lipolysis subjects compared to normal weight controls28. Overall, the present and decreased adipose tissue mass also reported unchanged or reduced study shows that AdPLA has a crucial regulatory role in the adipocyte serum NEFA33–36. dominant function of lipolysis through PGE2 in an autocrine manner, AdPLA expression in humans is also adipose specific (Supplemen- a paracrine manner or both. tary Fig. 5 online). Currently, little is known regarding the patholo- Fatty acids such as arachidonic acid that are released by the action of gical phenotype of individuals lacking AdPLA. A database search for PLA2 enzymes have been shown to either stimulate or inhibit adipocyte single-nucleotide polymorphisms (SNPs) in the PLA2G16 gene in differentiation29–31. Similarly, selective cyclooxygenase-2 inhibitors12,14 humans identified one SNP within the coding region resulting in a or prostaglandins themselves11,13,21,22 were reported to induce or Ser48Ala substitution. An additional 230 SNPs have also been detected 12,14 inhibit adipogenesis. PGF2a,viaitsFPreceptor,and15-deoxy-D - within noncoding regions of the PLA2G16 gene. It is not known PGJ2, which is a ligand for PPAR-g, have been reported to affect in vitro whether these result in altered protein function or abundance, and 13,21 12,14 adipocyte differentiation .PGD2 may generate 15-deoxy-D - clinical associations for these SNPs have not been reported, most likely PGJ2 and PGI2 and, through its IP receptor, may also affect adipocyte because our study is to our knowledge the first to characterize the 11,22 differentiation in vitro . However we have found that, unlike PGE2, physiological role of AdPLA. Many questions remain regarding the

8 ADVANCE ONLINE PUBLICATION NATURE MEDICINE ARTICLES

effect of partial or total PLA2G16 gene ablation in humans, and (Perkin-Elmer) at a rate of 0.05 mCi min–1 for 2 h to assess basal glucose further research in this area will yield advances in understanding the turnover, followed by the hyperinsulinemic-euglycemic clamp for 140 min with pathology of human obesity as well as of type 2 diabetes. a primed and continuous infusion of human insulin (300 pmol kg–1 prime (43 mU kg–1)) over 3 min followed by a 42 pmol kg–1 min–1 (6 mU kg–1 min–1) infusion (Novo Nordisk), a continuous infusion of [3-3H]glucose (0.1 mCi METHODS –1 Cell culture. We isolated, maintained and differentiated MEFs and 3T3-L1 cells min ) and a variable infusion of 20% dextrose to maintain euglycemia B –1 as previously described37. ( 100–120 mg dl ). We obtained plasma samples from the tail and measured tissue-specific glucose uptake after injection of a bolus of 10 mCi of 2-deoxy-D- 14 41 RNA analysis. We subjected total RNA isolated with Trizol reagent (Invitrogen) [1- C]glucose (Perkin Elmer) at 85 min . We analyzed our results as 42 to northern blotting (Amersham) or to quantitative RT-PCR or RT-PCR. previously described . We prevalidated primers used with the ABI PRISM 7700 sequence fast Histological analysis. We embedded tissues in paraffin and stained 6-mm-thick detection system (PE Applied Biosystems) for efficiency of amplification sections with H&E and determined adipocyte cell size with Image J software that was reported to be the same for all and essentially 100% by (US National Institutes of Health), measuring at least 300 cells from Applied Biosystems. each sample. Western blotting. We separated proteins by 12% SDS-PAGE, transferred the Phospholipase A activity. We assayed supernatants of WAT for PLA activity by proteins to nitrocellulose and probed the membrane with primary antibodies monitoring the liberation of [14C]palmitate from 1,2-di[1-14C]palmitoyl-sn- against AdPLA or desnutrin (Genemed Synthesis), IRS-1, phospho–IRS-1 glycero-3-phosphocholine as previously described43. (Ser307), phospho-AMPK-a (Thr172), AMPK-a, glyceraldehyde 3-phosphate dehydrogenase (Gapdh) or b-actin (Santa Cruz Biotechnology), HSL (Cell Indirect calorimetry and body temperature. We measured oxygen consump- Signaling), or phosphoserine (Calbiochem). tion (VO2) using the Oxymax system (Columbus Instruments).

Lipolysis. We performed lipolysis studies in adipocytes isolated as previously 2H O Labeling and gas chromatography–mass spectrometry analysis. We 38 2 described from gonadal WAT, in explants from freshly removed epididymal fat extracted lipids from gonadal fat pads of mice administered 2H Oindrinking B 2 pads ( 20 mg) and in MEFs differentiated to adipocytes. We incubated samples water for a 2-week period, trans-esterified the lipid phase by incubation with in Krebs-Ringer buffer (12 mM HEPES, 121 NaCl, 4.9 mM KCl, 1.2 mM MgSO4 3 N methanolic HCl and separated glycerol from fatty acid–methyl esters with and 0.33 mM CaCl2) with 3.5% fatty acid-free BSA and 0.1% glucose (KRB), the Folch technique40. We lyophilized and derivatized the aqueous phase with or without 200 nM isoproterenol (Sigma), PGE2 (Cayman), adenosine containing free glycerol to glycerol triacetate by incubation with acetic deaminase (Calbiochem), 0.5 mM Triacsin C (Biomol) or L826266 (Merck Frosst anhydride-pyridine (2:1). We used a model 6890 gas chromatographer with Canada) and measured glycerol (Sigma) and fatty acid (Wako) content. We a 5973 mass spectrometer (Agilent Technologies), fitted with a DB-225 determined prostaglandin abundance by competitive enzyme immunoassay fused silica column (J&W). We analyzed glycerol-triacetate under chemical (R&D) after solid-phase extraction by C18 cartridges (Cayman), and we ionization conditions by selected ion monitoring of mass-to-charge ratios determined cAMP levels by competitive immunoassay (R&D). (m/z) of 159–161 (representing M0–M2) and fatty acid–methyl esters as described elsewhere19. Mouse maintenance. All studies received approval from the University of We measured fractional triacylglycerol-glycerol synthesis as previously California at Berkeley Animal Care and Use Committee. We used mice on described19,44, with the following equation (see refs. 19,44 for details): a pure C57BL/6J background, after backcrossing for ten generations, for studies

of lipolysis, energy expenditure, hyperinsulinemic-euglycemic clamping, ftriacylglycerol ¼ EM1triacylglycerolglycerol=A11 triacylglycerolglycerol fatty acid oxidation, triacylglycerol clearance and for all studies on double- knockout mice. In all other experiments, we compared AdPLA-null and We estimated net lipolysis from ftriacylglycerol synthesis and adipose mass wild-type littermates on a mixed genetic background (C57BL/6J and 129 as follows:

© All rights reserved. 2009 Inc. Nature America, SVJ), but we also confirmed the results in a C57BL/6J background. We ½f ðadipose mass = labeling timeÞðDadipose mass=labeling timeÞ=fat pad mass provided either a SD or a HFD (45% of kcal from fat, 35% of kcal from triacylglycerol carbohydrate and 20% of kcal from protein, Research Diets) ad libitum.To We excluded values with net lipolysis equivalent to zero from analysis. generate ob/ob mice deficient in AdPLA, we bred C57BL/6J mice heterozygous for the ob mutation (Jackson Laboratory) with AdPLA-null mice on a C57BL/ Fatty acid oxidation. We determined fatty acid oxidation in isolated adipocytes 45 6J background, resulting in heterozygotes that we interbred to produce double- as previously described . knockout mice. Statistical analyses. We assessed the results by Student’s t-test to compare two groups or by one-way analysis of variance with Dunnett’s post-hoc test for Carcass and tissue analysis. We homogenized frozen, eviscerated carcasses multiple comparisons and expressed them as means ± s.e.m. We analyzed from 40-week-old mice fed a HFD in water, dried the homogenates to a adipocyte size distribution by the Wilcoxon signed rank Test. constant weight and estimated the lipid content by a previously described 39 method . We extracted tissue neutral lipids by a previously described meth- Accession codes. Pla2g16 is located on 19 in the mouse genome. 40 od , isolated triacylglycerol and diacylglycerol by thin-layer chromatography Reference sequence identifiers for mouse Pla2g16 in Gene are as follows: and quantified lipids with Infinity reagent (Thermo Trace) after sonification in gene ID, 225845; gene, NC_000085.5; protein, NP_644675; mRNA, 1% Triton X-100. NM_139269.

Glucose and insulin tolerance tests. For the GTT, we injected mice intraper- Note: Supplementary information is available on the Nature Medicine website. itoneally with D-glucose (wild-type or knockout, 2 mg per g body weight; double-knockout or ob/ob mice, 0.625 mg per g body weight) after an ACKNOWLEDGMENTS overnight fast and monitored the tail blood glucose levels. For ITT, we injected This work was supported in part by DK75682 from the US National Institutes mice with insulin (humulin, Eli Lilly) at a level of 0.5 mU per g body weight of Health to H.S.S. and DK59635 to G.I.S. R.E.D. and K.A.V. are recipients of (wild-type and knockout mice on a SD), 0.75 mU per g body weight (wild-type postdoctoral fellowships from the Natural Sciences and Engineering Research and knockout mice fed a HFD) or 1.75 mU per g body weight (ob/ob and Council of Canada. R.E.D. is a recipient of a postdoctoral fellowship from the double-knockout mice) after a 5-h fast. Canadian Institutes of Health Research. The authors would like to thank O. Barauskas for technical help; D. Frasson for fatty acid oxidation measurement; Hyperinsulenemic-euglycemic clamp. We implanted jugular venous catheters Y. Wang for performing WR1339 injections; J. Lu, J. Chen, R. Mantara and 7 d before the study. After an overnight fast, we infused [3-3H]glucose N. Nag for assistance with animal maintenance; A. Birkenfled and D. Frederick

NATURE MEDICINE ADVANCE ONLINE PUBLICATION 9 ARTICLES

for assistance with clamping studies; C. Lange and J. Chithalen for assistance protein kinase and peroxisome proliferator-activated receptor. Diabetes 55, with graphics; and Merck Frosst Canada for the kind gift of L826266. 2562–2570 (2006). 25. Johansson, S.M., Yang, J.N., Lindgren, E. & Fredholm, B.B. Eliminating the anti- Published online at http://www.nature.com/naturemedicine/ lipolytic adenosine A1 receptor does not lead to compensatory changes in the Reprints and permissions information is available online at http://npg.nature.com/ antilipolytic actions of PGE2 and nicotinic acid. Acta. Physiol. Scand. 190,87–96 (2007). reprintsandpermissions/ 26. Fain, J.N., Leffler, C.W. & Bahouth, S.W. Eicosanoids as endogenous regulators of leptin release and lipolysis by mouse adipose tissue in primary culture. J. Lipid Res. 1. Duncan, R.E., Ahmadian, M., Jaworski, K., Sarkadi-Nagy, E. & Sul, H.S. Regulation of 41, 1689–1694 (2000). lipolysis in adipocytes. Annu. Rev. Nutr. 27, 79–101 (2007). 27. Girouard, H. & Savard, R. The lack of bimodality in the effects of endogenous and 2. Gregoire, F.M., Smas, C.M. & Sul, H.S. Understanding adipocyte differentiation. exogenous prostaglandins on fat cell lipolysis in rats. Prostaglandins Other Lipid Physiol. Rev. 78, 783–809 (1998). Mediat. 56, 43–52 (1998). 3. Jaworski, K., Sarkadi-Nagy, E., Duncan, R.E., Ahmadian, M. & Sul, H.S. Regulation of 28. Fain, J.N., Madan, A.K., Hiler, M.L., Cheema, P. & Bahouth, S.W. Comparison of the triglyceride metabolism. IV. Hormonal regulation of lipolysis in adipose tissue. Am. release of adipokines by adipose tissue, adipose tissue matrix, and adipocytes from J. Physiol. Gastrointest. Liver Physiol. 293, G1–G4 (2007). visceral and subcutaneous abdominal adipose tissues of obese humans. Endocrinology 4. Dircks, L. S.H. Acyltransferases of de novo glycerophospholipid biosynthesis. Prog. 145, 2273–2282 (2004). Lipid Res. 38, 461–479 (1999). 29. Gaillard, D., Negrel, R., Lagarde, M. & Ailhaud, G. Requirement and role of arachidonic 5. Yet, S.F., Lee, S., Hahm, Y.T. & Sul, H.S. Expression and identification of p90 as the acid in the differentiation of pre-adipose cells. Biochem. J. 257, 389–397 murine mitochondrial glycerol-3-phosphate acyltransferase. J. Biochem. 32, (1989). 9486–9491 (1993). 30. Massiera, F. et al. Arachidonic acid and prostacyclin signaling promote adipose 6. Vance, D.E. & Vance, J.E. Biochemistry of lipids, lipoproteins and membranes. tissue development: a human health concern? J. Lipid Res. 44, 271–279 277–303 (Elsevier, Oxford, 2008). (2003). 7. Duncan, R.E., Sarkadi-Nagy, E., Jaworski, K., Ahmadian, M. & Sul, H.S. Identification 31. Petersen, R.K. et al. Arachidonic acid–dependent inhibition of adipocyte differentia- and functional characterization of adipose-specific phospholipase A2 (AdPLA). J. Biol. tion requires PKA activity and is associated with sustained expression of cyclooxy- Chem. 283, 25428–25436 (2008). genases. J. Lipid Res. 44, 2320–2330 (2003). 8. Schaloske, R.H. & Dennis, E.A. The phospholipase A2 superfamily and its group 32. Kopecky, J., Hodny, Z., Rossmeisl, M., Syrovy, I. & Kozak, L.P. Reduction of dietary numbering system. Biochim Biophys Acta (2006). obesity in aP2-Ucp transgenic mice: physiology and adipose tissue distribution. Am. J. 9. Yuan, C., Rieke, C.J., Rimon, G., Wingerd, B.A. & Smith, W.L. Partnering between Physiol. 270, E768–E775 (1996). monomers of cyclooxygenase-2 homodimers. Proc. Natl. Acad. Sci. USA 103, 33. Hertzel, A.V. et al. Lipid metabolism and adipokine levels in fatty acid–binding protein 6142–6147 (2006). null and transgenic mice. Am. J. Physiol. Endocrinol. Metab. 290, E814–E823 10. Richelsen, B. Release and effects of prostaglandins in adipose tissue. Prostaglandins (2006). Leukot. Essent. Fatty Acids 47, 171–182 (1992). 34. Lucas, S., Tavernier, G., Tiraby, C., Mairal, A. & Langin, D. Expression of human 11. Aubert, J. et al. Prostacyclin IP receptor up-regulates the early expression of C/EBPb hormone-sensitive lipase in white adipose tissue of transgenic mice increases and C/EBPd in preadipose cells. Mol. Cell. Endocrinol. 160, 149–156 (2000). lipase activity but does not enhance in vitro lipolysis. J. Lipid Res. 44, 154–163 12. Fajas, L., Miard, S., Briggs, M.R. & Auwerx, J. Selective cyclo-oxygenase-2 inhibitors (2003). impair adipocyte differentiation through inhibition of the clonal expansion phase. 35. Martinez-Botas, J. et al. Absence of perilipin results in leanness and reverses obesity in J. Lipid Res. 44, 1652–1659 (2003). Lepr(db/db) mice. Nat. Genet. 26, 474–479 (2000). 13. Forman, B.M. et al. 15-Deoxy-D12, 14-prostaglandin J2 is a ligand for the adipocyte 36. Tansey, J.T. et al. Perilipin ablation results in a lean mouse with aberrant adipocyte determination factor PPAR g. Cell 83, 803–812 (1995). lipolysis, enhanced leptin production, and resistance to diet-induced obesity. Proc. 14. Yan, H., Kermouni, A., Abdel-Hafez, M. & Lau, D.C. Role of COX-1 and Natl. Acad. Sci. USA 98, 6494–6499 (2001). COX-2 in modulating adipogenesis in 3T3–L1 cells. J. Lipid Res. 44, 424–429 (2003). 37. Kim, K.H., Lee, K., Moon, Y.S. & Sul, H.S. A cysteine-rich adipose tissue-specific 15. Cohen-Luria, R. & Rimon, G. can bimodally inhibit and stimulate the secretory factor inhibits adipocyte differentiation. J. Biol. Chem. 276, 11252–11256 epididymal adipocyte adenylyl cyclase activity. Cell. Signal. 4, 331–335 (1992). (2001). 16. Kather, H. & Simon, B. Biphasic effects of prostaglandin E2 on the human fat cell 38. Viswanadha, S. & Londos, C. Optimized conditions for measuring lipolysis in murine adenylate cyclase. J. Clin. Invest. 64, 609–612 (1979). primary adipocytes. J. Lipid Res. 47, 1859–1864 (2006). 17. Smas, C.M. S.H. Pref-1, a protein containing EGF-like repeats, inhibits adipocyte 39. Bligh, E.G. & Dyer, W.J. A rapid method of total lipid extraction and purification. Can. J. differentiation. Cell 73, 725–734 (1993). Biochem. Physiol. 37, 911–917 (1959). 18. Latasa, M.J. G.M., Moon YS, Kang C, Sul HS. Occupancy and funtion of the –150 SRE 40. Folch, J., Lees, M. & Sloane Stanley, G.H. A simple method for the isolation and and –65 E-box in nutritional regulation of the fatty acid synthase gene in living purification of total lipides from animal tissues. J. Biol. Chem. 226, 497–509 animals. Mol. Cell. Biol. 23, 5896–5907 (2003). (1957). 2 19. Turner, S.M. et al. Measurement of TG synthesis and turnover in vivo by H2O 41. Youn, J.H. & Buchanan, T.A. Fasting does not impair insulin-stimulated glucose uptake © All rights reserved. 2009 Inc. Nature America, incorporation into the glycerol moiety and application of MIDA. Am. J. Physiol. but alters intracellular glucose metabolism in conscious rats. Diabetes 42, 757–763 Endocrinol. Metab. 285, E790–E803 (2003). (1993). 20. Bell-Parikh, L.C. et al. Biosynthesis of 15-deoxy-(12,14–PGJ2 and the ligation of 42. Samuel, V.T. et al. Targeting foxo1 in mice using antisense oligonucleotides PPARg. J. Clin. Invest. 112, 945–955 (2003). improves hepatic and peripheral insulin action. Diabetes 55, 2042–2050 21. Reginato, M.J., Krakow, S.L., Bailey, S.T. & Lazar, M.A. Prostaglandins promote and (2006). block adipogenesis through opposing effects on peroxisome proliferator-activated 43. Lucas, K.K. & Dennis, E.A. Distinguishing phospholipase A2 types in biological receptor g. J. Biol. Chem. 273, 1855–1858 (1998). samples by employing group-specific assays in the presence of inhibitors. Prostaglan- 22. Vassaux, G., Gaillard, D., Ailhaud, G. & Negrel, R. Prostacyclin is a specific effector of dins Other Lipid Mediat. 77, 235–248 (2005). adipose cell differentiation. Its dual role as a cAMP- and Ca2+-elevating agent. J. Biol. 44. Chen, J.L. et al. Physiologic and pharmacologic factors influencing glyceroneogenic Chem. 267, 11092–11097 (1992). contribution to triacylglyceride glycerol measured by mass isotopomer distribution 23. Savage, D.B., Petersen, K.F. & Shulman, G.I. Disordered lipid metabolism and the analysis. J. Biol. Chem. 280, 25396–25402 (2005). pathogenesis of insulin resistance. Physiol. Rev. 87, 507–520 (2007). 45. Bansode, R.R., Huang, W., Roy, S.K., Mehta, M. & Mehta, K.D. Protein kinase Cb 24. Yoon, M.-J. et al. Adiponectin increases fatty acid oxidation in skeletal muscle cells deficiency increases fatty acid oxidation and reduces fat storage. J. Biol. Chem. 283, by sequential activation of AMP-activated protein kinase, p38 mitogen-activated 231–236 (2008).

10 ADVANCE ONLINE PUBLICATION NATURE MEDICINE