L-PGDS-produced PGD2 in premature, but not in mature, adipocytes increases obesity and insulin resistance

Lipocalin-type prostaglandin (PG) D synthase (L-PGDS) is responsible for the production of PGD2 in adipocytes and is selectively induced by a high-fat diet (HFD) in adipose tissue. In this study, we investigated the effects of HFD on obesity and insulin resistance in two distinct types of adipose-specific L-PGDS gene knockout (KO) mice: fatty acid binding protein 4 (fabp4, aP2)-Cre/L-PGDS flox/flox and adiponectin (AdipoQ)-Cre/L-PGDS flox/flox mice. The L-PGDS gene was deleted in adipocytes in the premature stage of the former strain and after maturation of the latter strain. The L-PGDS expression and PGD2 production levels decreased in white adipose tissue (WAT) under HFD conditions only in the aP2-Cre/L-PGDS flox/flox mice, but were unchanged in the AdipoQ-Cre/L-PGDS flox/flox mice. When fed an HFD, aP2-Cre/L-PGDS flox/flox mice significantly reduced body weight gain, adipocyte size, and serum cholesterol and triglyceride levels. In WAT of the HFD-fed aP2-Cre/L-PGDS flox/flox mice, the expression levels of the adipogenic, lipogenic, and M1 macrophage marker genes were decreased, whereas those of the lipolytic and M2 macrophage marker genes were enhanced or unchanged. Insulin sensitivity was improved in the HFD-fed aP2-Cre/L-PGDS flox/flox mice. These results indicate that PGD2 produced by L-PGDS in premature adipocytes is involved in the regulation of body weight gain and insulin resistance under nutrient-dense conditions.

There are two distinct types of PGD synthase (PGDS). One is lipocalin-type PGDS (L-PGDS) and the other is hematopoietic PGDS (H-PGDS). The L-PGDS gene is highly expressed in the brain, heart, and male genital organs 24 . Whereas H-PGDS is responsible for the synthesis of PGD 2 in inflammatory and immune cells, i.e., macrophages, mast cells, and Th2 cells 25,26 . PGD 2 exerts its physiological effects through two G protein-coupled receptors, the PGD 2 receptors (DP), DP1 receptors and chemoattractant receptor-homologous molecule expressed on Th2 cells (CRTH2), DP2 receptors 27 .
L-PGDS-produced PGD 2 enhances lipid accumulation in 3T3-L1 cells 11,14 through suppression of lipolysis via the DP2 receptors 28 . In vivo studies have been carried out using the gene-manipulated mice of the L-PGDS gene. PGD 2 -overproducing mice fed a high-fat diet (HFD) became obese 29 . L-PGDS gene knockout (KO) mice showed glucose intolerance and insulin resistance, and increased fat mass in the aorta under HFD conditions 30 . Adipose cells of the L-PGDS KO mice were larger than those of mice fed low-fat diet (LFD) 30 . L-PGDS KO mice showed no change in body weight, but improved glucose tolerance under HFD conditions 31 . In contrast, no glucose or insulin intolerance was observed in L-PGDS KO mice, but body weight gain and atherosclerotic lesions in the aorta were increased 32 . Thus, the roles of L-PGDS in obesity and obesity-related phenotypes in the L-PGDS gene-manipulated mice remain controversial. PGD 2 is involved in the regulation of various physiological events and L-PGDS is widely expressed in the body 33 . The disruption of the L-PGDS gene throughout the whole body may cause the unexpected effects and/or the unexplained phenotypes.
To address these concerns, we investigated the adipose-specific functions of L-PGDS and PGD 2 by the use of adipose-specific L-PGDS KO mice under the control of fatty acid binding protein 4 (Fabp4, aP2)-Cre transgene (aP2-Cre/L-PGDS flox/flox ) or adiponectin (AdipoQ)-Cre transgene (AdipoQ-Cre/L-PGDS flox/flox ) through the Cre-loxP system. The aP2-Cre/L-PGDS flox/flox mice exhibited decreased body weight gain with the reduction of fat mass, and improved insulin sensitivity under HFD conditions. Therefore, L-PGDS may be a target for the development of anti-obesity medicine and the treatment of obesity-mediated insulin resistance.

Results
Expression profile of the L-PGDS gene. For the various tissues of LFD-and HFD-fed mice, the expression of the L-PGDS gene was the highest in the brain, followed by the heart in both LFD-and HFD-fed mice and in white adipose tissue (WAT) in HFD-fed mice (Fig. 1a). The mRNA level of the L-PGDS gene in WAT of HFDfed mice was selectively enhanced approximately 2.3-fold as compared with LFD and was almost unchanged in other tissues (Fig. 1a).
We then examined the expression of the L-PGDS gene and two adipogenic marker genes, aP2 (Fabp4) and adiponectin (AdipoQ) during adipogenesis of mouse adipocyte 3T3-L1 cells (Fig. 1b). The transcription of the aP2 gene was induced in premature adipocytes even at 2 days after the initiation of adipogenesis and was gradually enhanced during adipogenesis, whose profile closely resembled that of the L-PGDS gene. On the other hand, the expression of the AdipoQ gene was very low at 2 days in premature adipocytes and was induced at 4 days in the mature stage of adipogenesis, indicating that the AdipoQ gene was selectively expressed in mature adipocytes and that its expression came later than those of the L-PGDS and aP2 genes.
The null L-PGDS allele was detected in visceral WAT (vWAT) of the HFD-fed aP2-Cre/L-PGDS flox/flox mice, but not in the AdipoQ-Cre/L-PGDS flox/flox mice (513-bp), although the floxed L-PGDS was detected in all tissues of both L-PGDS flox/flox and aP2-Cre/L-PGDS flox/flox mice (2513-bp; Supplemental Fig. S1). The body weight gains of wild-type and L-PGDS flox/flox mice were almost the same under LFD or HFD conditions (Supplemental Fig. S2a). Moreover, the expression levels of the L-PGDS mRNA in the brain, liver, and vWAT of the L-PGDS flox/flox mice were almost the same as those of wild-type mice under LFD or HFD conditions (Supplemental Fig. S2b).
The Cre transgene was abundantly expressed in vWAT under the control of the aP2 promoter/enhancer or AdipoQ promoter, but not in the brain and liver (Fig. 2b). In addition, the mRNA for the Cre transgene was not detected in wild-type mice (Fig. 2b). The mRNA level of the L-PGDS gene was significantly reduced in vWAT of the HFD-fed aP2-Cre/L-PGDS flox/flox mice, but unchanged in their brains and livers, as compared with that of the control L-PGDS flox/flox mice (Fig. 2b). In contrast, the adipose-specific decrease in the L-PGDS mRNA level was not detected in vWAT of the AdipoQ-Cre/L-PGDS flox/flox mice, although the Cre transgene was expressed in WAT in these mice under the control of the AdipoQ promoter (Fig. 2b). Furthermore, to confirm a decrease in L-PGDS protein, we carried out Western blot analysis. The expression of L-PGDS protein in the brain and liver of the HFD-fed aP2-Cre/L-PGDS flox/flox or AdipoQ-Cre/L-PGDS flox/flox mice was almost the same as those in the L-PGDS flox/flox mice (Fig. 2c). In contrast, L-PGDS expression was clearly lowered in vWAT of HFD-fed aP2-Cre/L-PGDS flox/flox mice, but not in vWAT of the HFD-fed AdipoQ-Cre/L-PGDS flox/flox mice (Fig. 2c).

Decrease of body weight gain in the HFD
Histological analysis showed that HFD increased adipocyte size approximately 3.9-fold in the L-PGDS flox/ flox mice, but only about 1.8-fold in the aP2-Cre/L-PGDS flox/flox mice (Fig. 4d,e). These results indicate that body weight gain and fat mass increases by HFD were reduced in the aP2-Cre/L-PGDS flox/flox mice.      (Table 1). Under LFD conditions, serum levels of total cholesterol and total lipid were significantly decreased in the aP2-Cre/L-PGDS flox/flox mice (Table 1). Further investigation is needed to understand the reason why these levels were decreased in the LFD-fed aP2-Cre/L-PGDS flox/flox mice. Moreover, the levels of total cholesterol, HDL-cholesterol, LDL-cholesterol, glucose, and TG were significantly lower in the aP2-Cre/L-PGDS flox/flox mice than in the L-PGDS flox/flox mice under HFD conditions (Table 1). These results indicate that serum levels of cholesterols, glucose, and TG were lowered in the HFD-fed aP2-Cre/L-PGDS flox/flox mice.  Fig. 6 and Supplemental Fig. S4). In contrast, HFD elevated the mRNA levels of the F4/80 and CD11c genes in vWAT and sWAT, about 6.3-and 5.0-, and 5.6-and 4.7-fold, respectively, in the L-PGDS flox/flox mice. However, the mRNA levels of the F4/80 and CD11c genes were not significantly increased in vWAT and sWAT of the aP2-Cre/L-PGDS flox/flox mice by HFD feeding (Fig. 6 and Supplemental Fig. S4). The mRNA levels of the F4/80 and CD11c genes were decreased to about 44% and 62%, and 56% and 46%, respectively, in vWAT and sWAT of the aP2-Cre/L-PGDS flox/flox mice, as compared with the L-PGDS flox/flox mice ( Fig. 6 and Supplemental Fig. S4). In contrast, the mRNA levels of the M2 macrophage marker genes; e.g., CD163, CD204, and CD206 were not significantly altered or rather increased in the aP2-Cre/L-PGDS flox/flox mice as compared with the L-PGDS flox/flox mice in the HFD-fed condition ( Fig. 6 and Supplemental Fig. S4). These results suggest that adipose-specific L-PGDS is associated with the elevation of inflammation in WAT.  (Fig. 7a). However, serum insulin levels in the HFD-fed aP2-Cre/L-PGDS flox/flox mice were slightly higher than those of the L-PGDS flox/flox mice (Fig. 7a).   (Fig. 7b). In contrast, when fed an HFD, serum glucose levels in the aP2-Cre/L-PGDS flox/flox mice were lower than those of the L-PGDS flox/flox mice (Fig. 7b). These results reveal that adipose L-PGDS is associated with the impairment of insulin sensitivity in mice.

Discussion
L-PGDS is widely expressed in various mouse tissues (Fig. 1a) and likely plays many different types of physiological and pathological functions 33 . L-PGDS is a bifunctional protein: one is to act as a PGD 2 -producing enzyme that catalyzes the isomerization of PGH 2 to produce PGD 2 34 , and the other is as a carrier protein for small lipophilic molecules such as retinal and retinoic acid 35 , biliverdin 36 and bilirubin 37 , and gangliosides 38 . In adipocytes, PGD 2 and its metabolites, Δ 12 -PGJ 2 and 15-deoxy-Δ 12,14 -PGJ 2 , accelerate lipid accumulation through the DP2 receptors 28 and PPARγ 12-14 , respectively. As shown in Fig. 4a, the PGD 2 content is decreased in WAT of the aP2-Cre/L-PGDS flox/flox mice to about 50% of the L-PGDS flox/flox mice. Moreover, AT-56, an L-PGDS inhibitor, suppresses adipogenesis in mouse 3T3-L1 cells 14 . These results, taken together, indicate that L-PGDS acts as a PGD 2 -producing enzyme in adipocytes. The other half of PGD 2 in WAT is considered to be produced by L-PGDS in non-adipocytes, such as endothelial cells 39

of the blood vessels, or by H-PGDS in mast cells and other inflammatory cells 25 within WAT.
Under HFD conditions, WAT was the third most enriched organ for L-PGDS mRNA expression followed by the brain and heart (Fig. 1a), and was the largest organ in the body. Thus, WAT is the most active organ in the total amount of L-PGDS gene expression under HFD conditions. The roles of L-PGDS in obesity have been identified by several in vivo studies [29][30][31][32] . PGD 2 -overproducing mice become obese under HFD conditions 29 . L-PGDS gene KO mice showed glucose intolerance and insulin resistance, and increased fat mass in the aorta under HFD conditions 30 . L-PGDS-ablated mice showed an improvement in glucose tolerance under HFD conditions 31 . In contrast, glucose intolerance or insulin resistance was not observed in the L-PGDS KO mice, but body weight gain and atherosclerotic lesions were increased in the aorta 32 . The roles of L-PGDS and/or PGD 2 in obesity are controversial, because L-PGDS and PGD 2 carry various functions in the body. Therefore, ablation of the L-PGDS gene or overproduction of PGD 2 in the whole body may not be suitable for the evaluation of their roles in peripheral adipose tissue. Therefore, we employed the adipose-specific L-PGDS KO mice through the Cre-loxP system    to find the functions of adipose L-PGDS and PGD 2 , and finally demonstrated that L-PGDS-produced PGD 2 in premature adipocytes regulates body weight gain and insulin resistance under HFD conditions. In this study, we used two distinct types of Cre-expressing mice under the control of adipocyte-specific promoters to generate the adipose-specific KO mice, aP2-Cre/L-PGDS flox/flox and AdipoQ-Cre/L-PGDS flox/ flox mice. Between these two adipocyte-specific conditional KO mice, the aP2-Cre/L-PGDS flox/flox mice showed HFD-induced depletion of L-PGDS in adipocytes, whereas the AdipoQ-Cre/L-PGDS flox/flox mice did not show such a phenotype. When we examined the time course of L-PGDS expression in 3T3-L1 cells during their development from fibroblasts to adipocytes, the L-PGDS expression was similar to that of aP2 and earlier than that of AdipoQ (Fig. 1b). These results are supported by previous reports that showed the expression of the AdipoQ mRNA occurred slightly later than that of the aP2 mRNA in 3T3-F442A and 3T3-L1 cells 40 . Thus, L-PGDS and aP2 are expressed even in preadipocytes but AdipoQ was only expressed in mature adipocytes. Therefore, the aP2-Cre/L-PGDS flox/flox mice may be useful to delete L-PGDS under HFD conditions. In fact, we succeeded to disrupt the L-PGDS gene in adipose tissue under HFD conditions only by the use of the aP2 promoter-driven Cre, but not, of the AdipoQ promoter-driven one (Fig. 2b,c).
The HFD-fed aP2-Cre/L-PGDS flox/flox mice showed decreased body weight gain with the reduction of fat mass (Fig. 3a,b). The WAT of the aP2-Cre/L-PGDS flox/flox mice were smaller in size than the control L-PGDS flox/flox mice (Fig. 4d,e), suggesting that adipose L-PGDS and PGD 2 are associated with the enhancement of obesity, together with the enlargement of adipose cells. However, the loss of the L-PGDS gene in WAT and decreased body weight gain were not observed in the HFD-fed AdipoQ-Cre/L-PGDS flox/flox mice (Fig. 3a,b), although the Cre transgene was expressed in WAT in these mice (Fig. 2b). Further expression profile analysis demonstrated that induction of the expression of the AdipoQ gene came later than those of the L-PGDS and aP2 genes in mouse adipocytic 3T3-L1 cells (Fig. 1b). These results suggest that the delayed induction of AdipoQ gene expression did not disrupt the L-PGDS gene in WAT of the HFD-fed AdipoQ-Cre/L-PGDS flox/flox mice and that L-PGDS in WAT had already been produced in premature adipocytes. Mature adipocytes with active gene expression of AdipoQ did not induce L-PGDS by the HFD feeding and were not mainly involved in L-PGDS-mediated increases in body weight and The aP2 gene is also expressed in macrophages 3,41 , liver, and brain. However, the expression level of the aP2 gene in macrophages is about 10 −4 -fold lower than that in adipocytes 3 . In the HFD-fed aP2-Cre/L-PGDS flox/flox mice, the transcription level of the Cre transgene was negligible in brain and liver. (Fig. 2b). The expression of L-PGDS was not affected by HFD in those organs of the aP2-Cre/L-PGDS flox/flox mice (Fig. 2b,c). The expression level of the L-PGDS gene was very low in the stromal vascular fraction (SVF) of obese adipose tissue and peritoneal macrophages prepared from LFD-and HFD-fed wild type mice (data not shown). All of these results, taken together, indicate that the HFD-induced upregulation of L-PGDS occurs not in macrophages, but predominantly in adipocytes. In this study, we have not yet identified the cells that express L-PGDS in obese adipose tissue. As SVF consists of a heterogeneous population that includes endothelial cells, erythrocytes, fibroblasts, and lymphocytes as well as pre-adipocytes, and adipocyte progenitor cells, we will undertake further analyses using pure SVF to identify the L-PGDS-expressing cells in obese adipose tissue. Moreover, we will investigate the function of L-PGDS in adipose macrophages by using macrophage-specific L-PGDS gene KO mice, lysozyme M (LysM)-Cre/L-PGDS flox/flox mice.
Another important finding in this study was that the adipose-specific disruption of the L-PGDS gene showed an anti-inflammatory effect. In obese adipose tissue, at least two different macrophages, M1 and M2, are found 42 . M1 macrophages make up the majority of adipose macrophages that exist in WAT of obese 42 . However, it is still unclear whether obesity induces the recruitment of monocytes that become M1 macrophages, or if HFD changes the phenotype of the tissue contianing M2 macrophages. In WAT of the HFD-fed aP2-Cre/L-PGDS flox/ flox mice, the expression levels of the M1 macrophage marker genes were all decreased ( Fig. 6 and Supplementary  Fig. S4), whereas the transcription levels of the M2 macrophage marker genes were either enhanced or not altered under HFD conditions ( Fig. 6 and Supplementary Fig. S4). The loss of adipose L-PGDS during obesity prevents HFD-induced inflammation. Obesity and insulin resistance are closely associated with inflammation in adipose tissue [43][44][45] . Accelerated de novo adipogenesis and lipogenesis with repressed lipolysis are closely associated with insulin sensitivity 3 . M1 macrophages in adipose tissue produce pro-inflammatory cytokines such as TNFα, which induces insulin resistance and suppresses the expression of PPARγ 3 . PGD 2 may be involved in enhancing inflammation in WAT of the HFD-fed mice. The HFD-fed aP2-Cre/L-PGDS flox/flox mice showed improved insulin sensitivity ( Fig. 7b) with lowered expression of TNFα in WAT ( Fig. 6 and Supplementary Fig. S4). In a previous study, when COX activity was inhibited by indomethacin in the HFD-fed mice, insulin resistance was prevented by the decreased plasma PGD 2 level and reduced expression of the macrophage marker genes in adipose tissue 46 . Thus, the absence of adipose L-PGDS and PGD 2 may prevent the phenotypic pro-inflammatory state that is induced under HFD. Macrophages express H-PGDS and infiltrate into the enlarged adipose tissue 26 . Thus, PGD 2 may be produced by H-PGDS in macrophages that have infiltrated the enlarged adipose tissue. The roles of H-PGDS-produced PGD 2 in macrophages that have infiltrated obese adipose tissue should be further elucidated.
As summarized in Fig. 8, adipose L-PGDS enhances body weight gain with the elevation of fat mass under HFD conditions. Adipose-specific disruption of the L-PGDS gene in the aP2-Cre/L-PGDS flox/flox mice under HFD shows an improvement in insulin sensitivity. The molecular mechanism for the decrease of adiposity in the HFD-fed aP2-Cre/L-PGDS flox/flox mice is still unclear. An in vitro study demonstrated that PGD 2 suppressed the lipolysis through the DP2 receptors in adipocytes 28 . Thus, adipocyte PGD 2 might be related to the regulation of lipolysis in vivo. Further in vivo studies are needed to elucidate the whole molecular mechanism of PGD 2 -regulated adiposity. In this study, we conclude that adipocyte-specific inhibition of L-PGDS or the DP2 receptors is potentially useful for the treatment of obesity and obesity-mediated insulin resistance. . Littermates lacking the aP2-Cre or AdipoQ-Cre transgene (L-PGDS flox/flox ) were used as the control. In the present study, we used only male mice to exclude the effects of female hormonal imbalance.
Mice were maintained with a 12-h light/12-h dark photoperiod in a humidity-and temperature-controlled room (55% at 24 °C). Water and food were available ad libitum. The animals were fed either LFD (FR-2, 4.8% fat; Funabashi Farm, Chiba, Japan) or HFD (35% fat; Research Diets, New Brunswick, NJ, USA).
The animal study was approved by the Animal committee of Osaka University of Pharmaceutical Sciences. Animals were handled in accordance with the principles and guidelines established by the respective committee. Every effort was made to minimize the number of animals used in these studies and their suffering.
Measurement of RNA level. Extraction of RNA and synthesis of first-strand cDNAs were performed as described previously 49 . Measurement of the mRNA levels by qPCR was conducted using the LightCycler System (Roche Diagnostics, Mannheim, Germany) and ABI 7500 Real-Time PCR System (Thermo Fischer Scientific, Waltham, MA, USA) with THUNDERBIRD SYBR qPCR Mix (Toyobo, Osaka, Japan) and Power SYBR Green PCR Master Mix (Thermo Fischer Scientific), and primers (Supplemental Table S1). Transcription level of the desired gene was normalized to that of TATA-binding protein (TBP) as the internal control.

Western blot analysis.
Proteins from tissues were prepared as follows. Tissues were disrupted in RIPA buffer {50 mM Tris-Cl (pH 8.0), 150 mM NaCl, 0.5%(w/v) sodium deoxycholate, 0.1%(v/v) SDS, 1% (v/v) NP-40} containing 1%(v/v) Triton X-100 and a protease inhibitor cocktail (Nacalai Tesque) by a Bead beater-type homogenizer (TAITEC, Saitama, Japan). After centrifugation to remove debris, protein concentrations of the supernatants (crude extracts) were determined by using a Pierce BCA Protein Assay Kit (Thermo Fisher Scientific). Proteins were separated by SDS-PAGE, followed by blotting onto the PVDF membranes (Immobilon; Merck, Kenilworth, NJ, USA). Further analysis by Western blotting was carried out as described previously 50 . Mouse L-PGDS polyclonal antibody and β-actin monoclonal antibody (Sigma), and anti-mouse or anti-rabbit IgG antibody conjugated with horseradish peroxidase (Santa Cruz Biotech., Dallas, TX, USA) were used in this study. enzyme immunoassay (eIA). The PGs in WAT were extracted as described previously 29 . Production of PGD 2 was measured by using a PGD 2 MOX EIA Kit (Cayman Chemical, Ann Arbor, MI, USA) according to the manufacturer's instructions.
Computed tomography (Ct). Mice were anesthetized with Nembutal (50 mg/kg of body weight, i.p.; Abbott Laboratories, North Chicago, IL, USA). CT analysis was carried out by a micro-CT scanner (LaTheta LCT-100; Hitachi, Tokyo, Japan). The analysis of CT data was carried out by the use of LaTheta software (Hitachi). The vWAT and sWAT, and muscle weights were measured from images at the level of the umbilicus. Subcutaneous WAT was defined as the extraperitoneal fat between skin and muscle. The intraperitoneal part with the same density as the subcutaneous fat layer was defined as vWAT. Proportions of vWAT and sWAT were determined by automatic planimetry as described previously 29 . Histological analysis. Tissues were fixed in 4%(v/v) paraformaldehyde and embedded in Tissue-Tek O.C.T.
Compound (Sakura Finetek, Torrance, CA, USA). Frozen sections (10 μm-thickness) were stained with hematoxylin and eosin. The sections were observed using an ECLIPSE E600 microscope (Nikon, Tokyo, Japan). At least three discontinuous sections were used for evaluation. serum biochemical parameter. Mice were fasted for 16 h prior to the collection of blood samples. Blood was collected from the abdominal aorta. Serum TG levels were determined by using Triglyceride Test Wako (Wako Pure Chemical, Osaka Japan), and insulin levels were measured by using ELISA kits (SHIBAYAGI, Gunma, Japan), according to the manufacturer's instructions. Serum TG, NEFA, total cholesterol, low-density lipoprotein (LDL), and high-density lipoprotein (HDL) levels were determined by using L-Type TG M test, NEFA-C, Cholesterol M, L-Type LDL-C, and L-Type HDL-C Kits (Wako Pure Chemical) according to the manufacturer's instructions.
Insulin tolerance test. Mice were fasted for 16 h prior to intraperitoneal (i.p.) injection of insulin (0.75 IU/ kg of body weight; HUMULIN ® ; Eli Lilly, Indianapolis, IN, USA). Blood was collected from the tail vein and glucose levels were immediately measured by the use of a MEDISAFE MINI Blood Glucose Monitoring System (Terumo, Tokyo, Japan). Blood glucose levels were measured at 0, 15, 30, 60, 90, and 120 min after injection of insulin.