Indomethacin Treatment Prevents High Fat Diet-induced Obesity and Insulin Resistance but Not Glucose Intolerance in C57BL/6J Mice*

Background: Obesity-associated insulin resistance is linked to inflammation. Results: Indomethacin, an anti-inflammatory cyclooxygenase inhibitor, prevented diet-induced obesity, but mice became glucose-intolerant with sustained hepatic glucose output and impaired glucose-stimulated insulin secretion. Conclusion: Inhibition of cyclooxygenase activity alters the metabolic consequences of an obesogenic high fat diet. Significance: Intake of anti-inflammatory cyclooxygenase inhibitors may impair glucose tolerance. Chronic low grade inflammation is closely linked to obesity-associated insulin resistance. To examine how administration of the anti-inflammatory compound indomethacin, a general cyclooxygenase inhibitor, affected obesity development and insulin sensitivity, we fed obesity-prone male C57BL/6J mice a high fat/high sucrose (HF/HS) diet or a regular diet supplemented or not with indomethacin (±INDO) for 7 weeks. Development of obesity, insulin resistance, and glucose intolerance was monitored, and the effect of indomethacin on glucose-stimulated insulin secretion (GSIS) was measured in vivo and in vitro using MIN6 β-cells. We found that supplementation with indomethacin prevented HF/HS-induced obesity and diet-induced changes in systemic insulin sensitivity. Thus, HF/HS+INDO-fed mice remained insulin-sensitive. However, mice fed HF/HS+INDO exhibited pronounced glucose intolerance. Hepatic glucose output was significantly increased. Indomethacin had no effect on adipose tissue mass, glucose tolerance, or GSIS when included in a regular diet. Indomethacin administration to obese mice did not reduce adipose tissue mass, and the compensatory increase in GSIS observed in obese mice was not affected by treatment with indomethacin. We demonstrate that indomethacin did not inhibit GSIS per se, but activation of GPR40 in the presence of indomethacin inhibited glucose-dependent insulin secretion in MIN6 cells. We conclude that constitutive high hepatic glucose output combined with impaired GSIS in response to activation of GPR40-dependent signaling in the HF/HS+INDO-fed mice contributed to the impaired glucose clearance during a glucose challenge and that the resulting lower levels of plasma insulin prevented the obesogenic action of the HF/HS diet.

between PGs and insulin secretion is complex, and apparently conflicting results have been reported. Increased levels of PGs have been associated with impaired ␤-cell function (3). PGE 2 was reported to reduce glucose-stimulated insulin secretion (GSIS), and accordingly, administration of COX inhibitors increased GSIS and improved glucose disposal (4). By contrast, administration of the general COX inhibitor, indomethacin, a commonly used nonsteroidal anti-inflammatory drug, has been shown to decrease insulin secretion in T2DM patients (5) and to lower glucose-stimulated acute insulin response (6).
PGs have both pro-and antiobesogenic properties (7). We and others have shown that both the diet-and cold-induced appearance of brown-like adipocytes, termed "brite" or "beige" adipocytes, in white adipose tissues requires COX expression and activity (8,9). Thus, in the apparently obesity-resistant Sv129 mouse strain, treatment with indomethacin attenuated diet-induced expression of Ucp1 (uncoupling protein 1) in inguinal white adipose tissue (iWAT), thereby promoting the development of diet-induced obesity (8).
The present study was designed to examine how indomethacin affected diet-induced obesity and the associated metabolic disorders in obesity-prone male C57BL/6J mice. In sharp contrast to the effect in Sv129 mice (8), indomethacin treatment prevented the obesogenic effects of a high fat/high sucrose (HF/ HS) diet in C57BL/6J mice. Our results reveal an interesting and complex phenotype of HF/HS-fed C57BL/6J mice treated with indomethacin (HF/HSϩINDO). Compared with C57BL/6J mice fed a HF/HS diet, mice fed a HF/HSϩINDO diet were lean and remained insulin-sensitive, yet they became glucose-intolerant, a feature probably associated with impaired regulation of hepatic gluconeogenesis and impaired compensatory up-regulation of pancreatic insulin secretion.

EXPERIMENTAL PROCEDURES
Mouse Care and Maintenance-Eight-week-old male C57BL/ 6JBomTac mice (Taconic, Ejby, Denmark) were acclimated for 1 week under thermoneutral conditions (28 -30°C) with a 12-h light and dark cycle. Mice were housed individually. Two sets of mice were assigned to three groups (n ϭ 9/group) and fed a low fat, regular diet (RD) (ssniff EF R/M Control, Germany); a HF/HS diet (ssniff S8672-E056 EF, Germany); or a HF/HS diet supplemented with indomethacin (16 mg/kg) (Sigma-Aldrich) for 7 weeks. A third set of mice was fed a HF/HS diet for 10 weeks before they were fed the HF/HS diet supplemented with indomethacin. A fourth set of mice was given saline or indomethacin (2.5 mg/kg body mass), dissolved in saline, by gavage after 5 h fasting and 1 h prior to glucose administration. In all experiments, food intake was recorded three times a week, and body weight was recorded once per week. All animal experiments were approved by the National State Board of Biological Experiments with Living Animals (Norway and Denmark).
Glucose, Insulin, and Pyruvate Tolerance Test-For the glucose tolerance test (GTT) and pyruvate tolerance test, the animals were injected intraperitoneally after 6 h of fasting with 2 g of glucose/kg of body weight or 2 g of sodium pyruvate (Sigma-Aldrich)/kg of body weight (10). For the insulin tolerance test (ITT), 0.75 units of human insulin (Actrapid)/kg of body weight was injected intraperitoneally in the fed state. For all tests, blood was collected from the tail vein of conscious animals, and blood glucose was measured using a glucometer (Ascensia Contour, Bayer) at baseline and at the indicated times.
Glucose-stimulated Insulin Secretion-Mice fasted for 3 h were injected intraperitoneally with 3 g of glucose/kg of body weight. Blood (30 l) was collected from the tail vein at baseline and 2, 5, and 10 min after glucose injection.
Indirect Calorimetry-After 6 weeks on their respective diets and a 24-h acclimatization period, O 2 and CO 2 gas exchange measurements were obtained for a 24-h period from each mouse, using the open circuit chambers of the Labmaster system (TSE Systems GmbH) (11).
Termination and Tissue Harvest-Mice were anesthetized with isoflurane (Isoba-vet, Schering-Plough) and euthanized by cardiac puncture. Blood samples were collected in tubes containing EDTA anticoagulant. Organs were immediately dissected, weighed, flash-frozen in liquid nitrogen, and stored at Ϫ80°C.
Histology Examinations-Paraffin-embedded sections of epididymal white adipose tissue (eWAT) and iWAT were stained with hematoxylin and eosin and analyzed as described previously (12). Sections of pancreatic tissue were stained for insulin, and pancreatic islet and section sizes were analyzed (13). Immunohistological detection of UCP1-positive multilocular cells was performed by an avidin-biotin peroxidase method. After deparaffination and rehydration, 5-m sections were processed through citrate buffer (pH 6), 2 ϫ 15 min (95°C). Sections were transferred to 0.3% hydrogen peroxide in methanol (30 min) before incubation with 10% goat serum (30 min) and incubation with primary antibody 1:4000 in PBS (goat serum 1%) overnight (4°C). Secondary antibody (1:200) was applied for 1 h, and tissue sections were incubated with the ABC complex (Vectastain ABC kit, Vector Laboratories), as described by the manufacturer. The 3,3Ј-diaminobenzidine (Vector Laboratories, Burlingame, CA) substrate kit was applied before counterstaining with hematoxylin and mounting. F4/80 immunohistological detection of macrophages in adipose tissue was performed by using the F4/80 sc-52664 (BM8) antibody and ImmunoCrus rat ABC staining system sc-2019 as described by the manufacturer (Santa Crus Biotechnology, Inc., Dallas, TX).
Bioluminescence Resonance Energy Transfer-based ␤-Arrestin-2 Interaction Assay-HEK293T cells were transiently transfected using polyethyleneimine with constructs encoding mouse GPR40 tagged at its carboxyl terminal with enhanced yellow fluorescence protein and ␤-arrestin-2 fused with Renilla luciferase. The ability of TUG469 and indomethacin to affect interaction between the mouse GPR40 and ␤-arrestin-2 constructs was assessed using a bioluminescence resonance energy transfer-based method (16). The resulting concentration-response data were fit to 3-parameter sigmoid curves using GraphPad Prism version 5.0 (GraphPad Software, Inc., San Diego, CA).
Real-time Quantitative RT-PCR-RNA was extracted from tissue, cDNA was synthesized, and gene expression was measured as described (19).
Tissue Lipid Extraction and Lipid Class Analysis-Total lipid was extracted from liver and muscle samples with chloroform/ methanol (2:1, v/v) and quantified on a Camaq high performance thin layer chromatography system and separated on high performance thin layer chromatography silica gel (20). After 6 weeks of treatment with the respective diets, total lipid was extracted from feces collected for 48 h (12).
Plasma Oxylipins-Plasma samples were extracted using Oasis HLB solid phase extraction cartridges (Waters Corp., Milford, MA) as described previously (21). Analytes were eluted with methanol and ethyl acetate into 6 l of glycerol and subjected to vacuum evaporation, and the resulting glycerol plug was stored at Ϫ80°C. Prior to analysis, samples were reconstituted in methanol containing a 100 nM concentration of each internal standard 1-cyclohexyl-ureido-3-dodecanoic acid and 1-phenylurea-3-hexanoic acid (Sigma-Aldrich), followed by filtration at 0.1 m and analysis by ultrahigh performance liquid chromatography-MS/MS. The analytes were separated using a reverse phase solvent gradient on a 2.1 ϫ 150-mm, 1.7-m Acquity BEH column, ionized by negative mode electrospray ionization, and detected by multireaction monitoring mode on an ABI 4000QTRAP (ABSciex, Foster City, CA) triple quadrupole mass spectrometer (22).
Statistics-All results are shown as mean Ϯ S.E. unless otherwise indicated. Statistical analyses of physiological and gene expression data were performed with GraphPad Prism version 5.0 (GraphPad Software, Inc.), and Dixon's Q-test was used to screen for outliers. One-way analysis of variance was used to compare differences between the experimental groups, followed by Fisher's least significant difference test unless other-wise indicated. A significance level of p Յ 0.05 was used for all tests. Statistical significances are denoted with asterisks as follows: *, p Յ 0.05; **, p Յ 0.01; ***, p Յ 0.001.

Indomethacin Prevents Diet-induced Obesity in C57BL/6J
Mice-Obesity-prone C57BL/6J mice were fed a HF/HS diet with or without indomethacin. Mice fed a HF/HS diet had increased weight gain, increased white adipose tissue (WAT), and increased liver masses compared with mice fed RD, whereas these increases were prevented in mice fed HF/ HSϩINDO (Fig. 1, A-D). No differences in weights of interscapular brown adipose tissue (iBAT) (Fig. 1E) or of the skeletal muscles, tibialis anterior, or soleus were observed (data not shown).
Indomethacin Reduces Feed Efficiency-To confirm that the reduced obesity in HF/HSϩINDO-fed mice was not simply due to reduced energy intake, feed intake was monitored. Energy intake was comparable between HF/HS-and HF/HSϩINDOfed mice (Fig. 1F); thus, indomethacin supplementation reduced feed efficiency (Fig. 1G). Because indomethacin has been reported to decrease bile acid secretion, with a possible effect on fat absorption (23,24), the apparent fat digestibility was calculated. Indomethacin treatment did not decrease fat absorption compared with mice fed the HF/HS diet, but both HF/HS-fed groups had increased fat absorption compared with RD-fed mice (Fig. 1H).
The reduced feed efficiency and higher percentage weight loss after 12-h starvation (Fig. 1I) in HF/HSϩINDO-fed mice compared with HF/HS-fed mice could indicate increased metabolic activity; therefore, indirect calorimetry measurements were performed. O 2 consumption, CO 2 production, and the respiratory exchange ratio were similar in the HF/HS-and HF/HSϩINDO-fed groups. No significant difference was observed between HF/HS and HF/HSϩINDO, but O 2 consumption, CO 2 production, and the respiratory exchange ratio were significantly lower compared with the RD-fed mice (Fig. 2, A-C). Unexpectedly, the expression of Ucp1 in iWAT was higher in mice fed HF/HSϩINDO than in mice fed HF/HS but comparable in their iBAT and eWAT (Fig. 2D). Expression of other brown adipocyte markers was not significantly different ( Fig. 2, E and F). To evaluate if the increased mRNA expression of Ucp1 in iWAT was accompanied with UCP1-positive multilocular cells, immunohistochemical (IHC) analyses in both eWAT and iWAT were performed. The IHC staining revealed no induction of UCP1-positive cells in eWAT, irrespective of diet group; however, in agreement with the mRNA levels of Ucp1, a minute induction was observed in iWAT of mice fed HF/HSϩINDO (Fig. 2G). Despite a significant induction of Ucp1 within iWAT, the relative expression is very low compared with the expression seen in iWAT of obesity-resistant mouse strains. Thus, we conclude that the low level of UCP1 in iWAT in mice fed the HF/HSϩINDO diet is insufficient to explain energy balance and the observed differences in body weight and fat mass.
Indomethacin Increases Plasma Glycerol and NEFA in the Fasted State-The reduced feed efficiency in HF/HSϩINDO compared with HF/HS-fed mice was not associated with increased plasma levels of aspartate transaminase or alanine aminotransferase (Fig. 2H), indicating that the low dose of indomethacin in the diet did not cause liver injury (Fig. 2H). PGs, PGE 2 in particular, inhibit lipolysis (25). Hence, we asked if treatment with indomethacin would affect lipolysis. In the fed state, glycerol and NEFA levels were comparable in all groups (Fig. 2, I and J). However, in the fasted state, both glycerol and NEFA levels were higher in HF/HSϩINDO-than in HF/HS-fed mice (Fig. 2, I and J), indicating increased mobilization of fat from adipose tissue in indomethacin-treated mice. This effect might reflect a more unabated ability of the increased levels of cAMP to increase lipolysis during fasting in the absence of inhibitory PGE 2 . (26), histological examinations and gene expression analyses of macrophage infiltration markers were performed. HF/HS diet-induced hypertrophy of eWAT and iWAT was prevented by indomethacin treatment (Fig. 3A). Expression of markers for macrophage infiltration and inflammation in eWAT was not increased significantly in response to HF/HS feeding for 7 weeks, except for Ccl2 (chemokine ligand 2) (Fig. 3, B-E). On the other hand, expression of Cd68 and of Ccl2 in iWAT was significantly reduced by a HF/HS diet with indomethacin supplementation. Expression of Emr1 (p ϭ 0.08) also tended to be reduced in HF/HSϩINDOfed mice (Fig. 3, F-I).

Indomethacin Prevents Diet-induced Hypertrophy and Attenuates Expression of Inflammatory Markers in iWAT-As hypertrophy is associated with increased infiltration of macrophages and low grade inflammation
Macrophage infiltration in AT depots was evaluated by IHC staining with F4/80 antibody. The evaluation of F4/80-positive macrophages showed infiltration in the AT of mice given low fat, HF/HS, and HF/HSϩINDO and is in agreement with the mRNA expression. Despite no significant differences in the number of F4/80-positive macrophages per 100 counted adipocytes (for eWAT, RD ϭ 43.9 Ϯ 6.2, HF/HS ϭ 58.7 Ϯ 5.0, and HF/HSϩINDO ϭ 50.1 Ϯ 6.8; for iWAT, RD ϭ 31.1 Ϯ 6.5, HF/HS ϭ 34.4 Ϯ 4.8, and HF/HSϩINDO ϭ 27.7 Ϯ 2.4), IHC revealed that the formation of crownlike structures was only observed in the HF/HS-diet group (Fig. 3J).
Plasma Oxylipin Profiles in Mice Fed HF/HS and HF/ HSϩINDO-To ascertain the efficacy of indomethacin treatment and obtain an overview of changes in plasma eicosanoids, plasma non-esterified oxylipin profiles were assessed. Cyclooxygenase-dependent metabolite levels in plasma were significantly reduced in mice given HF/HSϩINDO compared with mice given a HF/HS diet. The levels of prostaglandins 6-keto PGF 1␣ and PGD 2 as well as thromboxane B2 were significantly reduced by indomethacin supplementation, suggesting impacts on both COX1-and COX2-dependent metabolism ( Table 1).

Indomethacin Supplementation Attenuates HF/HS-induced Diacylglycerol and Triacylglycerol Accumulation in Liver and
Skeletal Muscle-To investigate if the reduced adipose tissue mass in HF/HSϩINDO-fed mice was associated with increased ectopic fat accumulation, different lipid classes were quantified in liver and muscle tissue. Indomethacin attenuated HF/HSinduced accumulation of triacylglycerol (TAG) and diacylglycerol (DAG) in both liver and tibialis anterior muscle (Fig. 4,  A-D). The levels of phospholipids were not affected in liver or muscle by any diet (data not shown).  reduced insulin sensitivity in HF/HS-fed mice, whereas mice fed HF/HSϩINDO exhibited HOMA-IR and QUICKI comparable with those of RD-fed mice (Fig. 4, E and F). Importantly, the ITT demonstrated significantly reduced insulin sensitivity in HF/HS-fed mice, whereas mice fed HF/HSϩINDO exhibited insulin sensitivity comparable with that of mice fed RD (Fig.  4G). Indomethacin Does Not Prevent Hyperglycemia and Impaired Glucose Tolerance Associated with an HF/HS Diet-Despite being lean and insulin-sensitive, HF/HSϩINDO-fed mice exhibited elevated levels of plasma glucose in the fed and fasted state compared with RD-fed mice (Fig. 5A). Furthermore, HF/HSϩINDO-fed mice were as glucose-intolerant as those fed HF/HS, indicating that HF/HSϩINDO-fed mice were unable to cope with the glucose challenge imposed by the GTT (Fig. 5B). Despite a reduction in glucose tolerance in mice given HF/HSϩINDO, no compensatory enhancement of glucosestimulated insulin secretion (GSIS) was observed 15 min after glucose injection compared with that observed in mice fed a HF/HS diet (Fig. 5C). To evaluate whether increased hepatic glucose output might contribute to the high plasma glucose values in HF/HSϩINDO mice, a pyruvate tolerance test was performed. Mice fed the HF/HS diet had increased blood glucose levels after a pyruvate injection compared with RD-fed mice, and this was not attenuated by indomethacin (Fig. 5D). Expression of G6pc (glucose-6-phosphatase) and Pck1 (phosphoenolpyruvate carboxykinase 1) was significantly increased in both HF/HS-and HF/HSϩINDO-treated mice, compared with mice fed a low fat RD (Fig. 5, E and F).
Indomethacin Prevents HF/HS-induced Glucose-stimulated Insulin Secretion-To further investigate the reduced glucose tolerance in HF/HSϩINDO-fed mice, glucose tolerance and GSIS were examined in a second set of mice after 1, 2, and 3 weeks of feeding (Fig. 5, G-L). Interestingly, the HF/HSϩINDO-fed mice exhibited clear signs of glucose intolerance already after 1 week of feeding with no alterations in GSIS as measured 15 min after glucose injection and no significant difference in body weight (Fig. 5, G and H). After 3 weeks of feeding, both the HF/HS-fed and the HF/HSϩINDO-fed mice were markedly glucose-intolerant, but only the HF/HS-fed mice exhibited a compensatory increase in GSIS, reflecting the insulin-resistant state of these mice (Fig. 5, K and L). This indicates that indomethacin exerted an early effect on glucose intolerance that temporally preceded changes in GSIS and increased glucose intolerance in the HF/HS fed mice. Moreover, these results suggest that indomethacin did not directly inhibit GSIS.
To establish whether indomethacin acutely affected glucose clearance and GSIS, RD-fed mice were orally treated with a single dose of indomethacin. This treatment acutely impaired glucose disposal despite no significant impairment in GSIS (Fig.  6, A-C). Neither glucose tolerance nor GSIS was affected by acute administration of indomethacin in obese, glucose-intolerant HF/HS-fed mice (Fig. 6, D-F). To evaluate if long term treatment with indomethacin was able to influence glucose tolerance and GSIS in the setting of an RD feeding, we fed mice RD Ϯ INDO for 7 weeks. Body weight (data not shown) and WAT mass in RDϩINDO-fed mice were comparable with levels for those fed RD (Fig. 6G). In this context, indomethacin supplementation did not lead to glucose intolerance (Fig. 6H) or alterations in GSIS (Fig. 6I).
Finally, we aimed to investigate if treatment with indomethacin could reverse elevated HF/HS-induced GSIS in obese and insulin-resistant mice. To increase fat mass and reduced insulin sensitivity, C57BL/6J mice were fed a HF/HS diet for 10 weeks; then a group was switched to an HF/HSϩINDO diet, and another was maintained on HF/HS for an additional 8 weeks. Mice fed an RD were used as a reference. No reduction in body weight, feed efficiency, and lean and fat mass was observed when obese mice were fed HF/HSϩINDO (Fig. 6, J-M). Furthermore, we observed no effect of INDO treatment regarding glucose tolerance and GSIS in already obese animals (Fig. 6, N  and O).
Indomethacin Treatment Combined with Activation of GPR40 Attenuates GSIS-To corroborate the notion that mice fed HF/HSϩINDO failed to compensate for a sustained hepatic glucose production, insulin levels were measured in both the fasted and fed states, and ␤-cell mass was quantified after 8 weeks of feeding. Despite the marked glucose intolerance, the HF/HS diet-induced hyperinsulinemia in both the fasted and fed states was not observed in HF/HSϩINDO-fed mice (Fig.  7A). Quantification of pancreatic ␤-cell mass demonstrated no reduction in ␤-cell mass in the HF/HS-fed mice (Fig. 7B). Evaluation of insulin secretion upon an intraperitoneal injection of glucose demonstrated that insulin secretion in mice fed the HF/HSϩINDO diet was comparable with that in the RD-fed mice, whereas it was increased in the HF/HS-fed mice (Fig. 7C). Fatty acid-dependent modulation of GSIS depends on both ␤-cell fatty acid metabolism and signaling via GPR40/FFA1 (30 -32). To examine whether indomethacin cell-autonomously modulated GSIS when GPR40/FFA1 was activated, we analyzed the effect of indomethacin on GSIS in the presence and absence of a synthetic GPR40 agonist, TUG469, in mouse MIN6 cells. No effect on cell survival and growth in response to treatment with up to 10 M indomethacin was observed (Fig.  7D), and indomethacin alone did not impair GSIS in MIN6 cells. However, in response to GPR40 activation in the presence of indomethacin, GSIS was inhibited at high glucose concentration (20.0 mM). Of note, no reduction in GSIS was observed at the low (2.8 mM) glucose concentration (Fig. 7E). Based on its structure, it was possible that indomethacin might act as an antagonist of GPR40. To investigate this possibility, we employed a bioluminescence resonance energy transfer-based ␤-arrestin-2 interaction assay, detecting the interaction between mouse GPR40 and ␤-arrestin-2 (32). As predicted, the addition of TUG469 potently increased GPR40-␤-arrestin-2 interaction. However, indomethacin did not antagonize TUG469 activation of GPR40; rather, indomethacin behaved like a low potency agonist of GPR40 (EC 50 ϭ 4.5 Ϯ 1.3 M), acting additive at submaximal concentrations of TUG469 (Fig.  7F). Taken together, these results indicate that indomethacin, in combination with fatty acid-dependent activation of GPR40, impairs GSIS, which at least in part may explain the lack of a  compensatory increase in GSIS to cope with the sustained hepatic glucose output in HF/HSϩINDO-fed mice.

DISCUSSION
In this study, we report that HF/HS diet-induced obesity, GSIS, and insulin resistance, but not glucose intolerance, in C57BL/6J mice were prevented by the general COX inhibitor, indomethacin. In addition, fatty acid-dependent up-regulation of GSIS was perturbed in HF/HSϩINDO-fed mice. Together, our findings point to a complex network controlling glucose tolerance and insulin secretion regulated by an intricate relationship between HF/HS feeding and indomethacin supplementation, where the lack of a compensatory up-regulation of GSIS in combination with sustained elevated hepatic gluconeogenesis resulted in a state of glucose intolerance in the otherwise lean and insulin-sensitive INDO-supplemented mice.
Indomethacin prevented increased adipose tissue mass and hypertrophy induced by a HF/HS diet in C57BL/6J mice. This is in striking contrast to the obesity-promoting action of indo- methacin together with a very high fat diet in Sv129 mice, normally considered obesity-resistant (8).
A high concentration of indomethacin has previously been reported to enhance PPAR␥-dependent transactivation and thereby increase adipocyte differentiation (34,35). However, the dose of indomethacin used in the present study was low (16 mg/kg of diet) compared with other studies and considered insufficient to raise plasma levels sufficiently high to acti- vate PPAR␥, and accordingly, expression of PPAR␥ and PPAR␥-responsive genes in adipose tissues was not significantly altered, comparing indomethacin-supplemented and unsupplemented mice (data not shown). The effects of PGs in lipolysis are complex. It has been shown that lipolysis is reciprocally affected by PGE2 and PGI2 and possibly also modulated by other PGs (36). Exogenous PGE2 has been shown to reduce lipolysis, whereas PGI2 have been shown to antagonize the anti-lipolytic effect of PGE2. Inhibition of prostaglandin H synthesis by indomethacin reduces production of both PGE2 and PGI2, and in agreement with (36), the levels of plasma glycerol and free fatty acid in the fed state indicate that indomethacin did not seem to affect lipolysis (Fig.  2I). On the other hand, our results demonstrate that mice treated with indomethacin had a significant increase in plasma free fatty acid after 16 h of fasting. This may at least in part reflect an increased ability of fasting-induced increases in cAMP levels to promote lipolysis in the absence of inhibitory prostaglandins, and additionally, indomethacin might change the balance between lipolysis-promoting and -antagonizing PGs.
The observed protection against diet-induced obesity was associated with significantly reduced feed efficiency and weight gain, but surprisingly, this was not accompanied by detectable changes in O 2 consumption or CO 2 production. However, we cannot exclude the possibility that O 2 consumption and CO 2 production might be affected at a different time points during the experiment.
Insulin resistance is associated not only with obesity but also with low grade inflammation in the adipose tissue (1, 2), ectopic fat accumulation, and increased DAG accumulation in both liver and skeletal muscle (37). COX-2 is necessary for the acute inflammatory response (38), and COX-2 deficiency attenuates age-dependent inflammation and infiltration of macrophages in adipose tissue (39). In the present study, inclusion of indomethacin reduced the circulating levels of 6-keto-PGF 1␣ , PGD 2 , and TXB2 and reduced expression of markers of macrophage infiltration and inflammation in adipose tissues. Further- more, indomethacin supplementation abolished HF/HS-induced accumulation of TAG and DAG in liver and muscle. Reflecting their lean phenotype and suppressed expression of inflammatory markers in adipose tissue, the HF/HSϩINDOfed mice remained insulin-sensitive. However, they developed glucose intolerance within 1 week of feeding. Similar to the obese HF/HS-fed mice, mice fed HF/ HSϩINDO had elevated sustained expression of genes involved in hepatic gluconeogenesis, suggesting a sustained high hepatic glucose output, correlated with high fasting and fed plasma glucose. This suggests a certain degree of insulin resistance in the liver, which, however, was undetectable in the HF/HSϩINDOfed mice using whole body ITT. A hyperinsulinemic-euglycemic clamp experiment would be needed to draw a firm conclusion on the precise contribution of hepatic glucose output to the observed hyperglycemia. Nevertheless, our pyruvate tolerance test data convincingly demonstrated that hepatic glucose output was increased in both obese HF/HS-fed and lean HF/HSϩINDO-fed mice, and the early development of glucose intolerance in the lean HF/HSϩINDO-fed mice indicated that these mice, despite being insulin-sensitive, were unable to compensate for the increased glucose output during a GTT.
HF/HS-fed mice exhibited the expected correlation between obesity and hyperinsulinemia. By contrast, plasma insulin levels in the lean HF/HSϩINDO-fed mice were comparable with those in RD mice. These differences impinge on the important question as to whether hyperinsulinemia is a compensatory action to counteract obesity-elicited inflammation and peripheral insulin resistance or whether increased plasma insulin precedes obesity development, inflammation, and insulin resistance. Whereas the canonical view favors the first possibility (1, 40), we and others have recently argued that a high level of insulin is a prerequisite for HF diet-induced obesity (41)(42)(43). In keeping with this view, adipose-specific insulin receptor knock-out mice are protected against diet-induced obesity (44). Thus, the leanness of the HF/HSϩINDO-fed mice may reflect a similar dependence on high insulin levels to elicit the obesogenic action of the HF/HS diet. High plasma insulin levels also exert positive feedback on ␤-cell mass expansion in response to HF/HS feeding, and similarly, the effect of glucose on ␤-cells may well depend on the increased insulin secretion. Accordingly, ␤-cell mass expansion and insulin hypersecretion may eventually result in systemic insulin resistance (33,42).
The paradoxical situation of glucose intolerance in insulinsensitive mice developed only in the context of HF/HS feeding. This observation invited speculations that indomethacin in combination with an HF/HS diet perturbed the normal regulation of insulin synthesis and/or secretion from ␤-cells. Our finding that, compared with HF/HS fed mice, mice fed HF/HSϩINDO had decreased fasting and fed insulin levels and a reduced insulin secretion during a GTT suggested that indomethacin impaired HF/HS-induced GSIS.
Initially, the regulatory effect of fatty acids on GSIS was assumed to be related to fatty acid metabolism in ␤-cells, but now it is known that a large part of the effect depends on fatty acid activation of GPR40 (31). Using MIN6 cells as a model, we demonstrated that the combined action of a selective GPR40 agonist, TUG469, and indomethacin inhibited GSIS. The augmented GSIS in HF-fed C57BL/6 mice in response to intraperitoneal glucose injection is reported to require GPR40 (32). In line with the finding that indomethacin does not influence GSIS in RD-fed mice, GSIS is also unaffected in RD-fed GPR40 Ϫ/Ϫ C57BL/6 mice (32). Thus, although indomethacin did not act as a GRP40 antagonist, the finding that the combined action of indomethacin and TUG469 inhibited GSIS in MIN6 cells suggests a mechanism by which indomethacin attenuates the compensatory increased insulin secretion in HF/HS-fed mice.
In conclusion, our results demonstrate that indomethacin, a commonly used COX inhibitor, prevented HF/HS-induced obesity and insulin resistance but not glucose intolerance and increased hepatic glucose production associated with a HF/HS diet. Indomethacin did not per se inhibit insulin secretion, but the indomethacin in combination with activation inhibited GSIS. In a situation of sustained hepatic glucose output, the inhibition of fatty acid enhancement of GSIS created a situation where pancreatic insulin secretion became insufficient to fully handle the glucose challenge of a GTT. Remaining important questions concern elucidating the precise molecular mechanisms by which indomethacin in combination with GPR40 activation inhibits GSIS and whether the effects of indomethacin and other nonsteroidal anti-inflammatory drugs can be translated into a human setting. Considering the worldwide common use of nonsteroidal anti-inflammatory drugs, these questions warrant further investigation.