Enhancement of the Aquaporin Adipose Gene Expression by a Peroxisome Proliferator-activated Receptor γ*

The current study demonstrates that aquaporin adipose (AQPap), an adipose-specific glycerol channel (Kishida, K., Kuriyama, H., Funahashi, T., Shimomura, I., Kihara, S., Ouchi, N., Nishida, M., Nishizawa, H., Matsuda, M., Takahashi, M., Hotta, K., Nakamura, T., Yamashita, S., Tochino, Y., and Matsuzawa, Y. (2000) J. Biol. Chem. 275, 20896–20902), is a target gene of peroxisome proliferator-activated receptor (PPAR) γ. The AQPap mRNA amounts increased following the induction of PPARγ in the differentiation of 3T3-L1 adipocytes. The AQPap mRNA in the adipose tissue increased when mice were treated with pioglitazone (PGZ), a synthetic PPARγ ligand, and decreased in PPARγ+/− heterozygous knockout mice. In 3T3-L1 adipocytes, PGZ augmented the AQPap mRNA expression and its promoter activity. Serial deletion of the promoter revealed the putative peroxisome proliferator-activated receptor response element (PPRE) at −93/−77. In 3T3-L1 preadipocytes, the expression of PPARγ by transfection and PGZ activated the luciferase activity of the promoter containing the PPRE, whereas the PPRE-deleted mutant was not affected. The gel mobility shift assay showed the direct binding of PPARγ-retinoid X receptor α complex to the PPRE. ΔPPARγ, which we generated as the dominant negative PPARγ lacking the activation function-2 domain, suppressed the promoter activity in 3T3-L1 cells, dose-dependently. We conclude that AQPap is a novel adipose-specific target gene of PPARγ through the binding of PPARγ-retinoid X receptor complex to the PPRE region in its promoter.

expressed in the human adipose tissue, plays an important role in the secretion of glycerol from the adipose tissue in the normal and the insulin-resistant conditions (3,4). In the normal mice, the AQPap mRNA was increased by fasting, and decreased by refeeding (3). These changes of AQPap mRNA were efficient to supply the glycerol into the liver for gluconeogenesis in the fasting condition. Recently, we identified negative insulin response element (IRE) in the promoter of human and mouse AQPap (5). The AQPap mRNA amounts were regulated through this IRE by fasting/refeeding. On the other hand, in the insulin resistant mice, the negative IRE of the AQPap gene was not sensitive to the high concentration of plasma insulin at fed stage. This insensitivity resulted in the high expression of AQPap mRNA, the increased concentration of plasma glycerol, and enhanced hepatic glucose production in the insulin resistant mice (3,5). Therefore, the regulation of AQPap mRNA levels in the adipose tissue is one of the determinant factors for glucose homeostasis in the whole body.
During the course of differentiation of adipocytes, AQPap mRNA was not detectable in the undifferentiated 3T3-L1 preadipocytes, and the mRNA was markedly increased during the differentiation of 3T3-L1 adipocytes (3). The expressions of adipose-specific genes are crucial for the phenotypic determination of the adipocytes (6). The transcriptional factors, which regulate the expression of these adipose-specific genes, include peroxisome proliferator-activated receptor ␥ (PPAR␥ 1 and 2) (7,8), the CCAAT/enhancer-binding protein (6), and sterolregulatory element binding proteins/adipocyte determination and differentiation factor 1 (6,9). Above all, PPAR␥ was shown to be essential for the differentiation of adipocytes (6). PPAR␥ knockout (Ϫ/Ϫ) mice had no detectable adipose tissue (10), and the fibroblasts obtained from PPAR␥ (Ϫ/Ϫ) embryos failed to differentiate into adipocytes by various stimuli (11,12). PPAR␥ forms stable heterodimers with retinoid X receptors (RXR), and this complex binds to a specific DNA sequence, designated peroxisome proliferator-activated receptor response element (PPRE), in the promoter of the target genes (13). PPRE has been identified in the promoters of several genes whose protein products are indispensable for the adipocyte characteristics, such as fatty acid transport protein 1 (FATP1) (14), lipoprotein lipase (15), acyl-CoA synthase (16), and adipocyte lipid-binding protein (17).
FATP1 is one of the fatty acid-transport proteins in the adipocytes (18). Similar to AQPap, the amount of FATP1 mRNA in 3T3-L1 cells was shown to be up-regulated as a consequence of adipose conversion (19). Transcription of FATP1 is activated by PPAR␥, through the PPRE in its promoter, in the differentiation process of the adipocytes (14). The free fatty acid released from the adipocytes through the function of FATP1 is another product of lipolysis of the triglycerides and is known to be deeply involved in glucose and lipid metabolism.
In the current study, we demonstrate AQPap as an adiposespecific novel PPAR␥ target gene, by identifying PPRE in its promoter, showing the transcriptional activation of the gene by PPAR␥ agonist, and revealing the direct binding of PPAR␥⅐RXR␣ complex to the AQPap PPRE. The coordinated increases of AQPap and FATP1 gene expressions via the PPREs during adipose-differentiation enable the adipocytes to supply glycerol and free fatty acid for the whole body metabolism.

EXPERIMENTAL PROCEDURES
Reagents-Pioglitazone (PGZ) was obtained from Takeda Chemical (Osaka, Japan). Wy14643 was purchased from Calbiochem. 9-cis-Retinoic acid (RA) was obtained from Sigma-Aldrich Japan K. K. (Tokyo, Japan. Cells and Animals-A mouse 3T3-L1 cell line was obtained from Health Science Research Resources Bank (Osaka, Japan).
Eight-week-old male C57BL/KsJ mice (Clea Japan, Inc., Osaka, Japan) were fed powder chow (CRF-1, Oriental Yeast Inc., Osaka, Japan). The animals were kept at 22°C with a 12-h dark-light cycle (8 a.m. to 8 p.m.). The animals were acclimated to the new environment for a week before the experiment. The chow consisted of 53.5% (w/w) carbohydrate, 5.9% (w/w) fat, 23.1% (w/w) protein, and 3.3% (w/w) dietary fibers. At the age of 13 weeks, the mice were divided into two groups. Each group (n ϭ 4, each), was fed either powder chow or chow containing 0.01% (w/w) PGZ. Two weeks later, the mice were sacrificed after 12 h of fasting.
RNA Analysis-The total cellular RNA was isolated using the Trizol™ reagent kit (Invitrogen, Tokyo, Japan). The Northern blot analysis was performed as previously described (3,5). Briefly, 10-g aliquots of total RNA from the white adipose tissue or 3T3-L1 adipocytes were electrophoresed on 1% agarose/formaldehyde gel and transferred to a nucleic acid transfer membrane (Hybond-Nϩ, Amersham Biosciences, Inc., Buckinghamshire, United Kingdom). After fixation by ultraviolet cross-linking, the filter was prehybridized with QuikHyb hybridization solution (Stratagene) at 65°C for 0.5 h. Mouse AQPap cDNA (5) and mouse PPAR␥ cDNA containing the entire coding region were obtained by reverse transcription-polymerase chain reaction using the mouse adipose tissue RNA as a template, and were then used as probes for Northern blot analysis. The probes were labeled using the Multiprime DNA labeling system (Amersham Biosciences, Inc.) with [ 32 P]dCTP. Hybridization was carried out with the same solution at 65°C for 1 h after adding 0.2 mg/ml denatured salmon sperm DNA. Washing was performed in 2ϫ sodium saline citrate (SSC) and 0.1% sodium dodecyl sulfate (SDS) at 65°C for 10 min, then in 0.1ϫ SSC plus 0.1% SDS at 65°C for 10 min, and the filter was exposed to Kodak X-Omat x-ray film for 24 h at Ϫ80°C with an intensifying screen. Quantification of the mRNA levels was performed using a Fast Scan scanning imager (Molecular Dynamics, Buckinghamshire, United Kingdom).
Plasmids-The mouse AQPap promoter-luciferase reporter plasmids were constructed by excising the promoter fragment from the genomic clone of AQPap (5) and inserting it into the MluI and XhoI site of the control pGL3 basic luciferase expression vector (Promega). A mutation of nucleic acids and deletion mutant of pGL3-AQPap luciferase plasmid were constructed using the QuickChange site-directed mutagenesis kit (Stratagene). PPRE-deleted construct was designed to lack the PPRE consensus region (Ϫ93/Ϫ77) from the wild-type construct (Ϫ141/ϩ63). The plasmids for transfection were purified using the Qiagen plasmid kit. PCR-generated fragments of full-length PPAR␥2, RXR␣, PPAR␣, and activation function-2 (AF-2)-deleted mutant ⌬PPAR␥ were subcloned into the XhoI site of the pcDNA3.1 expression vector (Invitrogen, Groningen, Netherlands). The pcDNA 3.1-PPAR␥ expression vector was a generous gift from Dr. David Mangelsdorf (University of Texas Southwestern Medical Center, Houston, TX). The pCMX-PPAR␣ expression vector was a generous gift from Dr. Ronald M. Evans (20). The ⌬PPAR␥ mutant construct lacks 11 amino acids (PLLQEIYKDLY) in the AF-2 domain, at its carboxyl terminus. The integrities of all plasmids were verified by DNA sequencing.
Cell Culture, Transfection, and Luciferase Assays-3T3-L1 preadipocytes were grown to confluence and were induced to differentiate into adipocytes according to the modified method of Rubin et al. (5,21). Briefly, 3T3-L1 cells were grown on a 9-cm culture dish or 12-well culture dish in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum (FCS). The cells were grown to confluence and were differentiated by incubation in DMEM with 10% FCS containing 0.5 mM 1-metyl-3-isobutylxanthine, 1 M dexamethasone, and 5 g/ml insulin for 48 h. The differentiated cells were maintained in DMEM with 10% FCS until the transfection experiments. Typically, for each 12-well culture plate, 1 g of firefly (Photinus pyralis) luciferase plasmids constructed from pGL3-basic luciferase expression vector and 10 ng of a sea pansy (Renilla reniformis) luciferase, pRL-SV40 plasmid (Promega) were complexed with LipofectAMINE™ 2000 (Invitrogen, Tokyo, Japan) following the manufacturer's protocol and used for transfection. The medium was removed 24 h later after transfection, and the cells were maintained in DMEM containing 10% FCS for 24 h before harvest for reporter assays. Luciferase activities were measured with the Dual Luciferase reporter assay system (Promega) according to the manufacturer's protocol.
Gel Electromobility Shift Assays-To study the binding of the nuclear hormone receptors to the putative PPRE, a double-stranded oligonucleotide, PPREwt, spanning nucleotides Ϫ98 to Ϫ64 of the mouse AQ-Pap gene upstream sequence were 32 P-radiolabeled with polynucleotide kinase (Promega). A 15-l reaction solution containing end-labeled PPRE oligonucleotide probe (2 ϫ 10 5 cpm) and 1 l of in vitro translation reaction was incubated for 20 min at 25°C and 15 min at 4°C in a buffer containing 20 mM Hepes (pH 8.0), 60 mM KCl, 1 mM dithiothreitol, 10% glycerol, and 1 g of poly(dI-dC). The DNA-protein complexes were resolved from the free probe by electrophoresis on a 4% polyacrylamide gel in 0.5ϫ TBE buffer (1ϫ TBE ϭ 9 mM Tris, 90 mM boric acid, 20 mM EDTA). The gels were dried and autoradiographed at Ϫ80°C. Double-stranded oligonucleotides composed of the following sequences were used for binding and competition analysis. AQPap PPREwt, 5Ј-TTCTGTTGTGCTTCTCCAGGGGAGAGGTCAGTAGG-3Ј; AQPap PPREmut, 5Ј-TTCTGTTGTGCTTCTCCAGGGGtGtcGTCAGTAGG-3Ј. PPRE sequence is underlined. The mutated bases are shown in lowercase letters.
Statistical Analysis-The results were expressed as mean Ϯ S.E. The significance of the difference between the groups was evaluated by Student's t test or analysis of variance (ANOVA) with Fisher's protected least significant difference test. Fig. 1A shows the expressions of AQPap and PPAR␥ mRNAs during the differentiation of 3T3-L1 adipocytes. In the undifferentiated state, 3T3-L1 cells had no detectable transcripts of either AQPap or PPAR␥. AQPap mRNA expression increased from day 4 after induction of differentiation. PPAR␥ mRNA induction was seen within 2 days after differentiation of 3T3-L1 cells. Transient transfection assays were performed to know which promoter region of the mouse AQPap gene is important for induced expression of its mRNA during the differentiation of 3T3-L1 adipocytes (Fig. 1B). A 0.5-kb DNA fragment of the mouse AQPap gene promoter including the transcription start site (GenBank accession no. AB056092; Ref. 5), and the serial 5Ј truncations were placed before a luciferase reporter gene. These reporter constructs were transfected into 3T3-L1 preadipocyte (preconfluent state) and fully differentiated 3T3-L1 adipocytes (day 7 after differentiation-induction). Comparative analysis of luciferase activities between the preadipocytes and adipocytes revealed that the majority of differentiation-induced activation of AQPap gene promoter required sequences between Ϫ141 and Ϫ14 (Fig.  1B). Inspection of this region of the mouse AQPap gene promoter revealed a putative PPRE of the direct repeat 1 type at Ϫ77 to Ϫ93, which is similar to consensus PPRE ( Fig. 1, C and D) (11)(12)(13)(14)(15).

Identification of a Putative PPRE in the Promoter of Mouse AQPap Gene-
Effects of Thiazolidinedione and PPAR␥ Deficiency on AQ-Pap mRNA Expression in Mice-Enhancements of AQPap mRNA that occurred subsequent to PPAR␥ induction during the differentiation of 3T3-L1 cells and the existence of putative PPRE site in its promoter strongly suggested that AQPap is a primary target gene of PPAR␥. To elucidate this hypothesis in vivo, we treated mice with 0.01% (w/w) pioglitazone (PGZ), synthetic PPAR␥ ligand (22) for 2 weeks and measured the mRNA levels of AQPap and PPAR␥ in the adipose tissue (Fig.  2, A and B). PGZ treatment caused 2.2-fold increase of AQPap mRNA expression in the adipose tissue. As the other side of the spectrum, we examined the AQPap mRNA in PPAR␥-deficient mice. The levels of AQPap mRNA expression were significantly decreased in the adipose tissue of PPAR␥ ϩ/Ϫ heterozygous knockout mice (Fig. 2, C and D) (12).
Thiazolidinedione-mediated Induction of AQPap mRNA and Promoter Activity in 3T3-L1 Adipocytes-We investigated the effect of thiazolidinediones on the regulation of AQPap gene transcription in 3T3-L1 adipocytes. Incubation with thiazolidinediones augmented the amount of AQPap mRNA and the promoter activity (Ϫ497/ϩ63) in a dose-dependent fashion (Fig.  3, A and B). To determine the function of the AQPap gene promoter as well as its putative PPRE, transient reporter assays using various AQPap gene promoter constructs were performed in 3T3-L1 adipocytes (Fig. 3C). Serial deletion showed that the basal levels and the PGZ-mediated induction of luciferase activities were highly maintained when the constructs contained the region of Ϫ141/Ϫ14. The construct Ϫ14/ϩ63, which lacked the region of Ϫ141/Ϫ14, including the putative PPRE, reduced the basal activity and abolished the response to PGZ (Fig. 3C).
To determine the further significance of the putative PPRE in the promoter of AQPap gene, we generated a construct lacking PPRE region (Ϫ93/Ϫ77) specifically from the wild-type construct (Ϫ141/ϩ63) (Fig. 4A), and compared the response to various stimuli between the wild-type (Ϫ141/ϩ63) and pAQPap PPRE-deleted constructs (Fig. 4A). Fig. 4B showed the effect of PGZ, 9-cis-RA, and Wy14643, which are the specific ligands for PPAR␥, RXR␣, and PPAR␣, respectively, on the luciferase activities of three constructs in 3T3-L1 preadipocytes and adipocytes. These compounds did not induce the reporter activities in the preadipocytes (Fig. 4B, left). PGZ enhanced the luciferase activity of the wild-type construct in 3T3-L1 adipocytes (Fig. 4B, right), similar to the data in Fig. 3B, and further increase was observed when 9-cis-RA was added (Fig. 4B,  right). Neither 9-cis-RA nor Wy14643 alone induced the luciferase activity in the wild-type construct (Fig. 4B, right). The pAQPap PPRE-deleted construct had a lower basal luciferase activity, and the responses to the treatment of PGZ were totally abolished (Fig. 4B, right). Fig. 4C clarified the effect of each reagent on the wild-type AQPap gene promoter construct (Ϫ141/ϩ63) in 3T3-L1 preadipocytes. The ectopic expression of PPAR␥ increased the basal luciferase activities of the wild-type construct (lane 9 versus lane 1). PGZ treatment induced the luciferase activities when the preadipocytes were transfected with PPAR␥ expression construct (lanes 11, 12, 15, and 16). Co-expression of RXR with PPAR␥ further increased the PGZinduced luciferase activities (lanes 15 and 16). Addition of 9-cis-RA augmented this increase (lane 12 compared with lane  11, and lane 16 compared with lane 15). These results indicate that the wild-type AQPap promoter activities were enhanced by PGZ in the preadipocyte when they expressed PPAR␥.
PPAR␥⅐RXR␣ Complexes Bind to the AQPap PPRE-PPAR␥'s partner for transcriptional activation is RXR␣ (23). To determine whether PPAR␥ binds to the AQPap PPRE as complexes with RXR␣, gel mobility shift assays were performed with double-stranded oligonucleotides containing the AQPap PPRE (Fig. 5, A and C). Transient reporter assays using AQPap PPREwt or PPREmut, which is 3-base substitution mutant constructs, were performed in 3T3-L1 preadipocytes (Fig. 5, A  and B). These 3-base substitutions have been described previously to diminish the binding of PPAR␥ to PPRE in the FATP1 gene promoter (14). Transfection of PPAR␥⅐RXR␣ led to 3.5-fold stimulation of luciferase activity of the wild-type PPRE re- The putative PPRE was denoted by a closed box; wild-type, pAQPap wild-type construct (Ϫ141/ϩ63); PPRE DEL, pAQPap PPRE-deleted construct. B, luciferase activities after the treatment with PGZ, 9-cis-RA, and Wy14643. 3T3-L1 preadipocytes (left) and adipocytes (right) were transfected for 24 h with the reporter constructs of pGL3 basic, pAQPap wild-type or pAQPap PPRE-deleted construct. The cells were then treated with 10 M PGZ, 10 M 9-cis-RA, combination of PGZ and 9-cis-RA, or 10 M Wy14643 for 24 h before harvest. Luciferase activities were normalized to pRL-SV40 activity, and the value of pGL3-basic vector was arbitrarily set as 1.0. The data are shown as mean Ϯ S.E. (n; 3 wells). C, synergistic activation of AQPap gene promoter by PPAR␥ and RXR␣. In 3T3-L1 preadipocytes, 1 g of pAQPap wild-type (Ϫ141/ ϩ63) reporter constructs was transfected with or without 500 ng of PPAR␥ and/or RXR␣ expression vectors, and treated with or without 10 M PGZ and/or 10 M 9-cis-retinoic acid. After transfection, the cells were incubated for 48 h. The total amount of DNA added was adjusted to 2.01 g using empty pcDNA3.1. The normalized luciferase activities are shown as mean Ϯ S.E. (n ϭ 3) and are expressed as -fold induction relative to the pAQPap wild-type (Ϫ141/ϩ63) promoter activity in the pcDNA3.1 expression vector (empty) (lane 1). The data are shown as mean Ϯ S.E. (n; 3 wells). Similar results were obtained in the other two independent experiments performed in triplicate.
porter construct (Ϫ141/ϩ63) (Fig. 5B, lane 3). Addition of PGZ and 9-cis-RA resulted in further 3-fold increase of luciferase activity (lane 4). The 3-base substitution mutation in the AQ-Pap PPRE abolished the response to PPAR␥⅐RXR␣ expression, and addition of PGZ/9-cis-RA (Fig. 5B, lanes 7 and 8). The 32 P-radiolabeled double-stranded PPREwt oligonucleotides were incubated with in vitro translated PPAR␥ protein (Fig.  5C). Neither PPAR␥ nor RXR␣ alone bound to AQPap PPRE (lanes 2 and 3). When PPAR␥ and RXR␣ were produced together, the mobility of 32 P-radiolabeled PPRE oligonucleotides were shifted to higher range, which indicated the binding of PPAR␥⅐RXR␣ complex to the AQPap PPRE (lane 4). The addition of excessive unlabeled PPREwt oligonucleotides reduced the signal of the 32 P-radiolabeled PPREwt oligonucleotides' binding to PPAR␥⅐RXR␣ complex in a dose-dependent manner (lanes 5-7). On the other hand, addition of PPREmut oligonucleotides, which were not supposed to bind to the PPAR␥⅐RXR␣ complex, did not reduce the specific signal (lanes 8 -10). These results demonstrate the specific binding of PPAR␥⅐RXR␣ complex to the AQPap PPRE.
Generation of Dominant Negative PPAR␥ Construct and Its Effect on the AQPap PPRE-To further confirm the PPAR␥-dependent enhancement of AQPap gene expression, we generated the dominant negative PPAR␥ expression constructs (Fig. 6A). In the other nuclear receptor-type transcriptional factors, the mutant construct lacking carboxyl AF-2 domain possessed the dominant negative effect on the transcription of the target genes (24,25). We generated a mutant PPAR␥ expression construct, designated ⌬PPAR␥, which lacked the last 11 amino acids in AF-2 domain (Fig. 6A). As shown in Fig. 6B 12  and lanes 19 -24). In 3T3-L1 adipocytes expressing abundant PPAR␥ endogenously, the basal AQPap gene promoter activities were inhibited gradually by transfection of the increasing amount of ⌬PPAR␥ construct (Fig. 6D, lanes 1-7). PGZ-induced AQPap gene promoter activity was significantly reduced when these cells were transfected with the increasing amount of ⌬PPAR␥ expression construct (Fig. 6D, lanes 8 -14). These data also confirmed the specific activation of AQPap gene transcription by PPAR␥.
The Effect of PPAR␣-mediated Pathway on the AQPap PPRE-To elucidate the mechanism by which the promoter activity of AQPap gene was specifically enhanced by PGZ, not by Wy14643 (Fig. 4B), we examined the effect of PPAR␣-mediated pathway on the AQPap gene promoter. Northern blotting showed that PPAR␣ mRNA was abundantly expressed in liver, but that PPAR␣ mRNA expression was not detectable in mouse white adipose tissue, and 3T3-L1 preadipocytes and adipocytes (Fig. 7A, left). The expression level of AQPap mRNA was not altered in 3T3-L1 adipocytes by the treatment of Wy14643 (Fig.  7A, right). Fig. 7B examined the effect of PPAR␣ expression and Wy14643 on the AQPap promoter activity in 3T3-L1 adipocytes. Without the transfection of PPAR␣ expression plasmid, there was no activation of AQPap gene promoter containing PPRE in adipocytes (Fig. 7B, lane 2 versus lane 1), as was shown in Fig. 4B. When PPAR␣ was exogenously expressed in 3T3-L1 adipocytes, basal luciferase activity of AQPap gene promoter was increased (Fig. 7B, lane 3 versus lane 1). Wy14643 further enhanced the AQPap gene promoter activity (Fig. 7B, lane 4 versus lane 3). These enhanced promoter activities were totally abolished when the PPRE was selectively deleted from the AQPap gene promoter (Fig. 7B, lanes 5-8). Gel mobility shift assay demonstrated the PPAR␣⅐RXR␣ complex had an ability to bind to radiolabeled AQPap PPRE oligonucleotides (Fig. 7C, lane 7 versus lane 6), and this binding was competed specifically by excessive amount of nonradiolabeled AQPap PPRE oligonucleotides (Fig. 7C, lane 8 versus lane 7), FIG. 5. Specific binding of PPAR␥⅐RXR␣ complex to the PPRE in the AQPap promoter. A, oligonucleotide sequences around the PPRE derived from the wild-type mouse AQPap gene promoter region (PPREwt) and the mutant mouse AQPap promoter region (PPREmut). PPRE sequences are underlined, and the mutated bases are in lowercase (closed circles are put above the characters.). B, the AQPap-PPREwt (wild-type) or -PPREmut reporter constructs were transfected in 3T3-L1 preadipocytes with or without the transfection of the expression vectors of PPAR␥ and RXR␣, and treated with or without 10 M PGZ and 10 M 9-cis-retinoic acid. After transfection, the cells were incubated for 48 h and harvested. The normalized luciferase activities are shown as mean Ϯ S.E. (n ϭ 3) and are expressed as -fold induction relative to the AQPap-PPREwt (Ϫ141/ϩ63) promoter activity with the pcDNA3.1 expression vector (empty) (lane 1). C, direct and specific binding of PPAR␥⅐RXR␣ complex to the AQPap PPRE. Electrophoretic mobility shift assays were performed as described under "Experimental Procedures." The 32 P-radiolabeled PPREwt oligonucleotide was incubated with in vitro synthesized PPAR␥ and/or RXR␣ proteins. The competitive gel mobility shift assay was performed using 32 P-radiolabeled PPREwt as input probe and unlabeled oligonucleotides (PPREwt or PPREmut) as competitors at 2-, 10-, and 50-fold molar excess.
similarly to the specific binding of PPAR␥⅐RXR␣ complex to AQPap PPRE (lanes 3-5). These data indicate that small amounts of PPAR␣ in adipocytes caused no enhancements of AQPap gene promoter activity through its PPRE by Wy14643, although PPAR␣ had an ability to bind to and activate the AQPap PPRE.

DISCUSSION
The current study showed that aquaporin adipose (AQPap) is the adipose-specific novel PPAR␥ target gene. The amounts of AQPap mRNA and its promoter activity were almost negligible in 3T3-L1 preadipocytes, which did not express PPAR␥. Once the cells were differentiated into the adipocytes, PPAR␥ mRNA was induced, and subsequently AQPap gene expression was enhanced. The AQPap mRNA levels in the adipose tissue were increased in mice treated with PGZ, a synthetic PPAR␥ agonist, and decreased in PPAR␥ ϩ/Ϫ heterozygous knockout mice. In 3T3-L1 adipocytes, PGZ augmented the mRNA amount and promoter activity of AQPap gene, although Wy14643, a PPAR␣ agonist, had a trivial effect on the AQPap gene promoter activity. Serial deletion analysis of 5Ј-flanking promoter region of AQPap gene revealed the putative PPRE site at Ϫ93/Ϫ77. When 3T3-L1 preadipocytes were transfected with PPAR␥ expression vector, the luciferase activity driven by the promoter containing PPRE was markedly enhanced with PGZ treatment, although it did not occur in 3T3-L1 preadipocytes not transfected with PPAR␥ construct. An additional effect of the RXR␣ activator, 9-cis-retinoic acid, on PPAR␥-induced transcription of AQP gene was shown in 3T3-L1 preadipocytes and adipocytes. Moreover, we demonstrated the direct binding of PPAR␥⅐RXR␣ complex to the PPRE sequence in the AQPap  Fig. 4A), with or without 100 ng of PPAR␣ expression vector for 24 h, and then the medium was supplemented with or without 10 M Wy14643 for 24 h. The total amount of DNA added was adjusted to 1.11 g using empty pcDNA3.1. The value of pAQPap wild-type luciferase activity in lane 1 was arbitrarily set as 1.0. The normalized luciferase activities are shown as mean Ϯ S.E. (n ϭ 3) in the results, which are represented by a column and bar graph. C, direct and specific binding of PPAR␣⅐RXR␣ complex to the AQPap PPRE. Electrophoretic mobility shift assays were performed as described under "Experimental Procedures." The 32 Pradiolabeled AQPap PPREwt was incubated with in vitro synthesized PPAR␥, PPAR␣, and/or RXR␣ proteins. The competitive gel mobility shift assay was performed using 32 P-radiolabeled PPREwt as input probe and unlabeled oligonucleotides (PPREwt) as competitors at 10fold molar excess. gene promoter, which indicated the usage of classical PPAR⅐RXR system for the PPAR␥-mediated activation. Recently, we have also identified a PPRE at Ϫ62/Ϫ46 in the promoter region of the human AQPap gene, which was 100% homologous to that in the mouse AQPap promoter (data not shown). These results indicate that the inducible expression of PPAR␥ is critical for enhancement of AQPap mRNA in the mouse and human adipose tissue via its PPRE in the promoter.
To date, adipocyte lipid-binding protein (15) is the only gene known to be the direct target of PPAR␥ and is expressed exclusively in the adipose tissue. In this sense, AQPap is the second adipose-specific PPAR␥ target gene. The mRNAs of the other PPAR target genes, including FATP1 (14), lipoprotein lipase (15), acyl-CoA synthase (16), phosphoenolpyruvate carboxykinase (26), liver fatty acid-binding protein (27), acyl-CoA oxidase (28,29), and hydroxymethylglutaryl-CoA synthase (30), are expressed in the other tissues and organs. There was no sequence specificity in the PPRE of AQPap gene promoter, in comparison with those of other target genes. It remains to be elucidated how the mRNA expression of AQPap is restricted to the adipose tissue. Recently, Dr. Spiegelman and colleagues have isolated two co-activators of PPAR␥, designated PPAR␥ co-activator 1 (31) and 2 (32) (PGC-1 and PGC-2). PGC-1 mRNA is expressed in the brown adipose tissue, heart, kidney, and brain. PGC-1 plays a key role in adaptive thermogenesis, oxidative metabolism, and glucose uptake. PGC-1 can function as a potent transcriptional coactivator for PPAR␥. PGC-1 binds to DNA-binding and hinge domains of PPAR␥ (31). PGC-2, expressed in most tissues, augments the transcriptional and adipogenic activities of PPAR␥. PGC-2 binds to AF-1-transactivating domain of PPAR␥ (32). We also showed that the protein small heterodimer partner, which lacks a DNA-binding domain (33), is also expressed in the adipose cells, and interacts with PPAR␥ (41). Mutual interaction of all these coactivators for PPAR␥ may be partially accountable for the adipose-specific expression of AQPap gene through its PPRE.
Mutations within the AF-2 domain, which is located at the carboxyl-terminal ligand binding region, have been shown to abolish ligand-activated transcription in several other receptors including the glucocorticoid receptor (34), estrogen receptor (35), retinoic acid receptor (36), thyroid hormone receptor (37), retinoid X receptor (38), and liver X receptor (39). This domain is likely to be a key region of transcriptional regulation with ligand binding. In this study, we demonstrated that this AF-2 domain is also required for PPAR␥ activation, for the first time. We confirmed the direct binding of ⌬PPAR␥⅐RXR␣ complex to the PPRE sequence in the AQPap gene promoter, similar to the wild-type PPAR␥⅐RXR␣ heterodimer (23). Nevertheless, in the mature 3T3-L1 adipocytes, ⌬PPAR␥ repressed the AQPap gene promoter activity by competing with the endogenous PPAR␥ to form a complex with RXR␣ and to bind with the PPRE in AQPap gene promoter. These results further confirmed that AQPap is directly regulated by PPAR␥⅐PPRE system.
The differentiated adipocytes actively conduct both synthesis and hydrolyzation of triglyceride. The free fatty acids, the product of triglyceride hydrolyzation, can be re-esterified into triglycerides and re-stored in the adipocytes. However, all glycerols, another product of the triglyceride hydrolyzation, enter the general circulation because of no system to re-uptake into the adipocytes. The released glycerol from the adipocytes becomes a substrate for hepatic gluconeogenesis (1,2,40). Therefore, AQPap, the regulator to transport the hydrolyzed glycerol, should be one of the determinant factors for the whole body glucose homeostasis. Our recent study revealed that the transcription of AQPap gene is regulated in response to the nutri-tional conditions through the negative IRE in its promoter (5). Increased AQPap in the insulin resistant condition was associated with high concentration of glycerol in plasma (3), which might result in the augmentation of hepatic glucose production.
Adipose-specific induction of AQPap through PPAR␥⅐PPRE, which is a master regulatory system of the adipocyte phenotype, strongly suggests that the release of glycerol is important as the function of the adipocytes. Coordinated induction of AQPap and FATP1 via PPRE during adipogenesis implies the physiological significance of these pathways supplying glycerol and free fatty acids into the blood. In the differentiation process from preadipocyte to mature adipocyte, insulin might drive the expression of AQPap mRNA probably in an indirect fashion. Further, once the cells are fully differentiated into adipocytes, AQPap gene transcription is negatively regulated directly by insulin through its IRE in the promoter, in accordance with the nutritional conditions. Further analysis on the regulation of AQPap gene transcription should lead to clarification of the mechanism to determine the plasma glycerol concentrations that modify glucose homeostasis.