Aberrant Hepatic Expression of PPARγ2 Stimulates Hepatic Lipogenesis in a Mouse Model of Obesity, Insulin Resistance, Dyslipidemia, and Hepatic Steatosis*

Insulin-resistant apoB/BATless mice have hypertriglyceridemia because of increased assembly and secretion of very low density apolipoprotein B (apoB) and triglycerides compared with mice expressing only apoB (Siri, P., Candela, N., Ko, C., Zhang, Y., Eusufzai, S., Ginsberg, H. N., and Huang, L. S. (2001) J. Biol. Chem. 276, 46064-46072). Despite increased very low density lipoprotein secretion, apoB/BATless mice have fatty livers. We found that hepatic mRNA levels of key lipogenic enzymes, acetyl-CoA carboxylase, fatty-acid synthase, and stearoyl-CoA desaturase-1 were increased in apoB/BATless mice compared with levels in apoB mice, suggesting increased lipogenesis in apoB/BATless mice. This was confirmed by determining incorporation of tritiated water into fatty acids. Neither the hepatic mRNA of the lipogenic transcription factor, SREBP-1c (sterol-response element-binding protein 1c), nor the nuclear levels of the mature form of SREBP-1 protein were elevated in apoB/BATless mice. By contrast, hepatic levels of peroxisomal proliferator-activated receptor 2 (PPARγ2) mRNA and protein were specifically increased in apoB/BATless mice, as were hepatic mRNA levels of two targets of PPARγ, CD36 and aP2. Treatment of apoB/BATless mice for 4 weeks with intraperitoneal injections of a PPARγ antisense oligonucleotide resulted in dramatic reductions of both PPARγ1 and PPARγ2 mRNA, PPARγ2 protein, and mRNA levels of fatty-acid synthase and acetyl-CoA carboxylase. These changes were associated with decreased hepatic de novo lipogenesis and hepatic triglyceride concentrations. We conclude that hepatic steatosis in apoB/BATless mice is associated with elevated rates of hepatic lipogenesis that are linked directly to increased hepatic expression of PPARγ2. The mechanism whereby hepatic Pparγ2 gene expression is increased and how PPARγ2 stimulates lipogenesis is under investigation.

We reported previously that apoB/BATless mice, a model generated by crossing mice expressing human apolipoprotein B (apoB) 3 (1) with mice lacking brown adipose tissue (BATless) (2), have hypertriglyceridemia and hypercholesterolemia because of a 2-3-fold increase in the secretion of very low density lipoprotein (VLDL) apoB and triglycerides (TG) relative to mice expressing apoB only (3). Similar levels of apoB mRNA in the livers of apoB and apoB/BATless mice indicated that the differences in apoB secretion resulted from differences in posttranscriptional regulation of VLDL assembly and secretion (4). There were no differences in hepatic levels of microsomal triglyceride transfer protein mRNA (3), a critical factor in the early co-and post-translational regulation of apoB-containing lipoprotein secretion (5). Low density lipoprotein receptor mRNA levels in liver were higher in apoB/BATless compared with apoB mice (3), indicating that increased apoB secretion in apoB/BATless mice did not result from reduced interactions of the low density lipoprotein receptor with nascent apoB lipoproteins (6).
Associated with a rising prevalence of obesity and insulin resistance, hepatic steatosis is increasing in the United States (7). In addition to having hyperlipidemia, apoB/BATless mice are obese, insulin-resistant, and have fatty livers, providing a model for studying the molecular basis for the association between insulin resistance, VLDL secretion, and hepatic steatosis. Increased hepatic TG could result from increased TG synthesis driven by increased delivery of albumin-bound fatty acids (FA) derived from lipolysis in peripheral adipose tissue, increased hepatic uptake of triglyceride fatty acid in remnant lipoproteins, or increased de novo lipogenesis. Decreased FA oxidation in the liver could lead to accumulation of TG, even when TG synthesis and VLDL secretion are normal. In the present studies, we examined in detail the basis for the hepatic steatosis in the apoB/BATless mouse. Of note, we found increased de novo lipogenesis that was independent of both SREBP-1c (sterol-response element-binding protein 1c) (8) and carbohy-drate-responsive element-binding protein (ChREBP) (9) regulation. By contrast, we confirmed prior studies demonstrating increased hepatic peroxisomal proliferator-activating receptor ␥ (PPAR␥) gene expression and activity in mouse models of insulin resistance (10 -12), and we also demonstrated a significant, independent role for hepatic PPAR␥ in the regulation of de novo hepatic lipogenesis.

MATERIALS AND METHODS
Animals-Congenic human apoB transgenic mice on a C57BL/6J background were generated as described (1). BATless mice (2) were purchased from The Jackson Laboratory (Bar Harbor, ME). These mice were created in the FVB/N strain by using the promoter from a BAT-specific protein, uncoupling protein-1, to drive the expression of the A-chain of diphtheria toxin only in brown adipose tissue. The breeding of heterozygous apoB mice with heterozygous BATless mice resulted in the generation of F1 mice with four possible genotypes as follows: those that expressed neither transgene (wild type controls), those that expressed the transgene for either apoB or diphtheria toxin A-chain (BATless), and those that had both transgenes (apoB/BATless). The phenotype of BATless mice is similar to that of apoB/BATless mice except that plasma TG levels are not elevated, although TG secretion rates are similar in the two groups (3). Furthermore, gene profiles and liver TG content are similar in male and female apoB/BATless mice despite the fact that the females are not as insulin-resistant.
Plasma Biochemistries-Blood samples were obtained from the retro-orbital plexus. Total plasma TG and cholesterol concentrations were measured by colorimetric methods using commercial kits (Wako Chemicals, Richmond, VA). FA levels were measured colorimetrically using a commercial kit (number 994-75409) from Wako Chemicals (Richmond, VA). Glucose levels were measured using an enzymatic kit (number 315-100; Sigma).
Determination of FA Turnover-To determine the rate of FA flux, [ 3 H]palmitic acid (PerkinElmer Life Sciences) was injected intravenously and its disappearance from plasma measured over time. The turnover of FA in plasma is known to be rapid; in fact, by 2 min after the injection of the tracer, plasma counts had decreased to 3-11% of the total injected volume. We bled the mice at 0.5, 1, 2, 5, and 10 min after [ 3 H]palmitic acid injection to determine radioactivity remaining in plasma. The fractional clearance rate for palmitate was determined using a two-pool model to account for recycling of FA from the extravascular space.
Determination of Liver TG Levels-Livers from the mice were collected for the measurement of hepatic TG content. Total liver lipids were extracted by a modification of the Folch method (13), and TG concentration was determined by a commercial colorimetric method (Wako Chemicals, Richmond, VA). [ 14 C]Triolein was added to each sample before lipid extraction to allow correction for recovery during the lipid extraction; final TG concentrations were adjusted accordingly. Total liver protein was extracted using T-PER TM tissue protein extraction reagent (number 78510; Pierce). A protease inhibitor mixture (number 1873580;, Roche Diagnostics) was added into the liver protein extraction reagent. Liver TG levels were expressed as milligrams of TG/g of liver protein.
Measurement of Hepatic de Novo Lipogenesis-The rate of hepatic de novo lipogenesis was determined by measuring the amount of newly synthesized FA present in the liver 1 h after intraperitoneal injection of 2 mCi of 3 H 2 O (14). 3 H-Labeled fatty acids were isolated by saponification of liver samples in KOH. After extraction of nonsaponifiable lipids, and acidification with H 2 SO 4 , the 3 H-labeled fatty acids were extracted and separated by thin layer chromatography. The plate was stained with iodine; the FA "spot" was scraped off the plate, and the isolated FA was added to scintillation fluid (Ecoscint H number LS-275; National Diagnostics, Atlanta, GA) and counted in a liquid scintillation counter (LS6500 Beckman Coulter, Fullerton. CA). The specific activity of body water was determined and used to calculate de novo lipogenesis as micromoles of 3 H 2 O incorporated into FA/h/g liver protein.
Northern Blot Hybridization with cDNA Probes-Total cellular RNA (TRNA) was isolated from livers using TRIzol reagent (Invitrogen) following the protocol provided by the company, and Northern blot hybridization was carried out with 20 g of RNA as described (1). The cDNA probes were prepared by reverse transcription-PCR from liver TRNA of male C57BL/6J mice. PCR primers used for each probe are shown in Table 1. A mouse ␤-actin riboprobe (Ambion Co., Austin, TX) was used in each experiment to normalize for loading of RNA samples. For quantification, autoradiograms were scanned with a densitometer.
RNA Probe Preparation and RNase Protection Assays-TRNA was isolated from livers as described above, and ribonuclease protection assays (RPA) were performed as described previously (1). The RNA probes were generated by amplification of target genes from TRNA of male C57BL/6J mouse by reverse transcription-PCR (15). PCR primers used are shown in Table 1. Each PCR product was cloned into a PCRII vector using a TA cloning kit obtained from Invitrogen. DNA sequences of these clones were verified by DNA sequencing using an ABI 377 automatic DNA sequencer (PerkinElmer Life Sciences).
Quantitative Real Time PCR-Liver TRNA samples were used for cDNA synthesis with oligo(dT) primers using a commercial kit (number 11904-018) from Invitrogen. The resulting cDNA samples were then quantified for each test gene using target gene-specific primers. Quantitative real time PCR (QPCR) was done using either TaqMan Universal PCR Master Mix (PerkinElmer Life Sciences) or a SYBR Green PCR Master Mix (Applied Biosystems, Foster City, CA) according to the protocols provided by the manufacturers. Detection of specific products was performed in triplicate using the Mx4000 Multiplex Quantitative PCR system (Stratagene, La Jolla, CA). Using the standard curve method, the relative quantitation of specific PCR products for each primer set was generated. For normal-ization, ␤-actin was amplified from each sample. The primers used for real time PCR are shown in Table 1. Mouse cyclophilin and GAPDH probes were synthesized by using antisense internal control templates (catalog numbers 7675 and 7431; Ambion).
Immunoblots-Equal amounts of liver protein extracts were subjected to 8% SDS-PAGE, transferred to nitrocellulose membranes, and immunoblotted using standard methods. A rabbit polyclonal antibody against mouse SREBP-1, which recognizes both mature and precursor forms, was a generous gift from Dr. Jay D. Horton (University of Texas Southwestern Medical School, Dallas). To measure the level of hepatic nuclear mature form of SREBP-1, hepatic nuclear protein was prepared immediately after exsanguinations as described previously for hamster liver (16). Hepatic FAS content was estimated using anti-FAS monoclonal antibody that reacts against the mouse protein (Pharmingen). Hepatic stearoyl-CoA desaturase (SCD) was estimated with a rabbit anti-mouse SCD polyclonal antibody that recognizes both mouse and human SCD (a generous gift from Dr. Alan Tall, Columbia University). Hepatic PPAR␥ protein level was measured with a mouse monoclonal antibody against the carboxyl terminus of human PPAR␥, which is identical to the corresponding mouse sequence (Santa Cruz Biotechnology, Santa Cruz, CA). For normalization, ␤-actin was measured with a monoclonal anti-␤-actin antibody (Sigma). Lamin B antibody was a gift from Dr. Howard Worman.
Gel Mobility Shift Assay-Gel mobility shift assays were performed by using consensus double-stranded oligonucleotides with the following sense strand sequences: CYP7A1 oligonucleotides that contain the mouse LXR-response element (LXRE) (5Ј-AGCTTGCTCTGGTCACCCAAGTTCAAGT-TAC-3Ј) (17) and the mouse sterol regulatory element (SRE) in the Fas promoter (5Ј-CGCGCGCGGGCATCACCCCAC-CGACGG-3Ј) (18). The probes were labeled with [␥-32 P]ATP and T4 polykinase. Gel mobility shift assays were performed using gel shift core system (Promega, Madison, WI). Hepatic nuclear proteins were prepared with a nuclear extraction kit (Active Motif, Carlsbad, CA). Mouse SREBP2 protein, used as a positive control, was a generous gift from Dr. Timothy Osborne (University of California, Irvine). The mouse LXR␣ and retinoid X receptor-␣ proteins, used as positive controls, were gen- 5Ј-GGT GGG CCA GAA TGG CAT CT-3Ј PPAR␥ Antisense Oligonucleotide Experiment-An antisense nucleotide (ASO) complementary to murine PPAR␥ (Gen-Bank TM accession number U09138.1), ISIS 141941, 5Ј-AGTG-GTCTTCCATCACGGAG-3Ј, was injected intraperitoneally twice a week into apoB/BATless mice at the dosage of 100 mg/kg/week for 4 weeks. To demonstrate the specificity of inhibition, an in-house universal control, ISIS 141923:,5Ј-CCTTC-CCTGAAGGTTCCTCC-3Ј, was injected simultaneously into a parallel group of apoB/BATless mice. The body weight and plasma lipid profile were measured before and after ASO treatment. At the end of the experiment, de novo lipogenesis, liver TG content, and mRNA levels of PPAR␥1, PPAR␥2, CD36, aP2, FAS, ACC, SREBP-1c, ChREBP, and LPK were determined as described above.
Kupffer Cell Depletion by Liposome-encapsulated Clodronate-Liposome-encapsulated clodronate, which is known to deplete Kupffer cells, was administered intravenously on 2 consecutive days (300 l), and livers were harvested 48 h later (19). PBS encapsulated in multilamellar liposomes was used as the control. Sixteen mg of cholesterol (C3045; Sigma) and 172 mg of phosphatidylcholine (P3556; Sigma) were dissolved in chloroform in a round bottom flask. The chloroform was evaporated at 37°C in a rotary evaporator under vacuum until a thin lipid film formed. Dicholoromethylene diphosphonate, 2 g (clodronate, D4434; Sigma), dissolved in 10 ml of PBS or 10 ml of PBS (to prepare empty liposomes) were added to the lipid film and shaken at 250 rpm for 30 min. The solution was sonicated for 3 min at room temperature in a water bath sonicator (50 watts). The liposomes were centrifuged at 100,000 ϫ g for 1 h and resuspended in 8 ml of PBS. Kupffer cell depletion was assessed by liver F4/80 staining and by using QPCR to quantify F4/80 and Nramp1 mRNA. The primers were as follows: mouse F4/80, forward 5Ј-CTT-TGGCTATGGGCTTCCAGTC-3Ј and reverse 5Ј-GCAAG-GAGGACAGAGTTTATCGTG-3Ј (20); mouse Nramp 1, forward 5Ј-ATCCTGCCCACTGTGTTGGT-3Ј and reverse 5Ј-GCGAAGGGCAGCAGTAGACT-3Ј (21). Results were normalized by mRNA levels of ␤-actin.
Statistics-All data are presented as means Ϯ S.D. Differences in the mean values between two groups were assessed by two-tailed Student's t test. Differences in mean values among more than two groups were determined by analysis of variance. p Ͻ 0.05 was considered to be statistically significant.

RESULTS
Male and female mice were used for all experiments except the direct measurement of lipogenesis with 3 H 2 O, where only male mice were studied. Although female apoB/BATless mice are less insulin-resistant than male apoB/Batless mice and have lower plasma TG levels despite similar TG secretion rates (3,22), they were otherwise identical in terms of hepatic gene expression and hepatic lipid metabolism. We have therefore presented only male data. Studies presented used mice between the ages of 12 and 30 weeks; the lipid and glucose profiles of apoB and apoB/BATless mice weaned onto a Western-type diet are stable over that period of time.
In our prior study (3), male apoB/BATless mice were more obese than apoB mice and were hyperinsulinemic, hypertriglyceridemic, and hypercholesterolemic; those results were recapitulated in the present studies. Thus, by about 20 weeks of age, after weaning onto a Western-type diet (see "Materials and Methods"), male apoB/BATless mice were ϳ15 g heavier than male apoB mice (57.0 Ϯ 8.7 versus 41.4 Ϯ 4.0 g). Also consistent with our previously published results (3), fasting blood analyses of 15 male apoB/BATless and 13 male apoB mice revealed hyperinsulinemia (6.7 Ϯ 2.9 versus 1.6 Ϯ 0.3 ng/ml), hypertriglyceridemia (206.6 Ϯ 60.9 versus 127.0 Ϯ 23.5 mg/dl), and hypercholesterolemia (389.6 Ϯ 97.4 versus 332.2 Ϯ 68.4 mg/dl) in the apoB/BATless mice. As observed previously, plasma glucose concentrations were not different in male apoB/BATless compared with male apoB mice (data not shown). A key finding in our first paper was that the hypertriglyceridemia was because of significantly elevated rates of hepatic secretion of VLDL triglyceride, apoB100, and apoB48 into plasma in male apoB/BATless mice compared with apoB mice; those results were confirmed in the present studies (data not shown). Of relevance to the present studies, apoB/BATless mice had greater hepatic TG compared with age-matched apoB mice (Fig. 1A), despite greater rates of secretion of both TG and apoB at the age of 22-26 weeks.
Hepatic TG stores are derived from three sources as follows: 1) uptake of plasma albumin-bound FA; 2) uptake of VLDL or chylomicron remnant triglyceride fatty acid; or 3) de novo lipogenesis. Our previous published results demonstrated that fasting plasma FA levels were not increased in apoB/BATless mice (3). Because steady state levels might not provide a true estimate of FA flux, we determined the fractional clearance rate of plasma FA with [ 3 H]palmitate. There was no difference in the fractional clearance rate of plasma FA in apoB/BATless compared with apoB mice (1.70 Ϯ 0.4 versus 1.97 Ϯ 0.8 pools per min, respectively; p ϭ 0.2). When the fractional clearance rates were multiplied by the FA pool sizes, which were larger in the more obese apoB/BATless mice, there was a nonsignificant trend toward a higher total FA flux in the apoB/BATless compared with the apoB mice (2.32 Ϯ 0.6 versus 1.63 Ϯ 0.9 mEq/min; p ϭ 0.10).
Next, we examined de novo hepatic lipogenesis in apoB/ BATless and apoB mice. Fig. 1B depicts results of studies to measure hepatic lipogenesis using 3 H 2 O. Incorporation of 3 H 2 O into FA was increased ϳ2-fold in apoB/BATless mice. Therefore, the doubling of hepatic TG mass (Fig. 1A) was associated with marked increases in de novo hepatic lipogenesis.
Consistent with the mRNA levels, protein levels of FAS in the liver were also increased 2-3-fold in apoB/BATless mice (Fig. 3A). Furthermore, hepatic SCD protein levels were significantly higher in apoB/BATless than in apoB mice (Fig. 3B). Because SREBP is regulated significantly by post-transcriptional cleavage of the full-length protein to an amino-terminal segment that is transcriptionally active, we determined levels of both the full-length precursor and the truncated mature forms of this protein in apoB/BATless mice. Levels of the precursor and mature forms of SREBP-1 protein in whole hepatic lysates were not different between apoB/BATless and apoB mice (Fig. 3C). Furthermore, the levels of the mature form of hepatic SPEBP-1 protein in nuclear extracts were not different in apoB/BATless versus apoB mice (Fig. 3D).
Because both SREBP-1c and LXR play such prominent roles in hepatic lipogenesis, and because they are ligandactivated, we carried out gel mobility shift assays to support the mRNA and protein data. Male apoB/BATless and apoB mice of similar ages were used for hepatic nuclear extract preparation. These studies revealed that there were no differences in binding activities of hepatic nuclear extracts from apoB/BATless versus apoB mice to either an SRE oligonucleotide found in the promoter of Fas (18) (Fig. 4A) or to an LXRE found in the promoter of Cyp7A1 (17) (Fig. 4B). Together with the mRNA results for both genes, and protein data for SREBP-1c, these results rule out a significant role for SREBP1-c and LXR in the increased lipogenesis observed in apoB/BATless mice.  DECEMBER 8, 2006 • VOLUME 281 • NUMBER 49

JOURNAL OF BIOLOGICAL CHEMISTRY 37607
ChREBP is a recently described transcription factor that can stimulate both glycolysis and lipogenesis (9). Neither the mRNA levels of ChREBP nor those of its specific hepatic target gene, LPK (9), were different between apoB/BATless and apoB mice (Fig. 5).
It had been reported that PPAR␥2 is increased in livers of BATless mice fed a high fat diet (28). Quantitative real time-PCR revealed that there were no significant differences in PPAR␥1 mRNA levels between apoB/BATless and apoB mice (Fig. 6A, left side). By contrast, hepatic mRNA levels of PPAR␥2 were increased by 2-3-fold in apoB/BATless (apoB/BATless,: Equal amounts of liver protein extracts were subjected to 8% SDS-PAGE, transferred to nitrocellulose membranes, and immunoblotted using standard methods. B, hepatic SCD protein was measured using a rabbit polyclonal antibody that recognizes both mouse and human SCD. Equal amounts of liver protein extracts were subjected to 8% SDS-PAGE, transferred to nitrocellulose membranes, and immunoblotted using standard methods. C, the levels of hepatic SREBP-1 precursor and mature protein forms were measured in extracted liver protein by Western blot using a rabbit polyclonal anti-mouse SREBP-1 IgG. The average volumes of densitometry scans from apoB mice were set as 100%. Open bars, apoB mice. Closed bars, apoB/BATless mice. Three representative immunoblots of both forms in each mouse type are shown below the bar graphs. D, the levels of hepatic nuclear SREBP-1 were measured in liver nuclear extracts by Western blots using the same antibody mentioned above. Loading of the gels was normalized using nuclear lamin B levels. Contamination of the nuclear preparations with the precursor, cytosolic form of SREBP is observed. The bar graphs represent the means Ϯ S.D. for the average volumes of densitometry scans of the mature form of SREBP-1 from each group, with the values for apoB mice set as 100%. Open bars, apoB mice. Closed bars, apoB/BATless mice. Representative immunoblots are shown for two mice in each group.

FIGURE 4. Lack of evidence for increased transcriptional activity of LXR or SREBP-1c.
Electrophoretic mobility shift assays were performed using synthetically labeled oligonucleotides for either the region of the Fas promoter containing an SRE (A) or the region of the Cyp7A1 promoter containing an LXRE (B), and nuclear extracts from livers of apoB/BATless and apoB mice. For comparison of nuclear binding activities between apoB/ BATless and apoB mice, the same amount of nuclear protein (ϳ10 g), based on the content of nuclear lamin B, was used. In all the even lanes of A, and lanes 1 and 3 of B, the shifted bands were efficiently competed away by the presence of unlabeled mouse probe, indicating that the nuclear proteins were binding specifically to the labeled consensus nucleotides. Mouse recombinant SREBP2, with a FLAG epitope attached, was used as a positive control in A (unnumbered lane at far right). Mouse LXR and retinoid X receptor recombinant proteins were used as a positive control in B. A is representative of five experiments; B is representative of three experiments. (Fig. 6A, right side). Northern blots showed the same patterns of hepatic Ppar␥1 and Ppar␥2 gene expression (data not shown). Immunoblotting showed that hepatic protein levels of PPAR␥2 were also increased in apoB/BATless compared with apoB mice (apoB/BATless, 167.4 Ϯ 41.8%, n ϭ 8, versus apoB, 100 Ϯ 21.7%, n ϭ 7; p Ͻ 0.01) (Fig. 6B). In most experiments, PPAR␥1 was not clearly separated from a nonspecific band (Fig. 6B), and densitometry was not performed. However, there was no apparent difference in PPAR␥1 protein levels in apoB/BATless versus apoB. To determine whether increased PPAR␥2 mRNA and protein in the liver were physiologically active, we measured hepatic mRNA levels of two downstream targets, aP2 and CD36. Hepatic mRNA levels of aP2 were increased 2-fold in apoB/BATless, whereas mRNA levels of CD36 were increased 3-fold in apoB/BATless versus apoB mice (Fig. 6C). Adipsin mRNA, another PPAR␥ target, was also increased in the apoB/ BATless mice (data not shown).
To determine the cellular site of increased expression of hepatic PPAR␥ and its targets genes, liposome-encapsulated clodronate was administered intravenously over 2 days to deplete hepatic Kupffer cells (19). After clodronate treatment, F4/80 immunostaining demonstrated that Kupffer cells had been ablated (Fig. 7A). QPCR also showed that hepatic mRNA levels of two macrophage-specific markers F4/80 (20) and Nramp1 (21) were dramatically decreased (Fig. 7B, left and right sides, respectively). In contrast, there were no changes in hepatic mRNA levels of either PPAR␥1 or PPAR␥2 or their targets, aP2 and CD36 (Fig. 7C), after clodronate treatment. Importantly, ablation of hepatic Kupffer cells did not affect hepatic mRNA levels of FAS and ACC (Fig. 7C).
To directly demonstrate the role of PPAR␥ overexpression in the increased de novo lipogenesis present in livers of apoB/ BATless mice, either PPAR␥ ASO or control oligonucleotide was injected intraperitoneally, twice a week for 4 weeks, to inhibit hepatic PPAR␥ activity in apoB/BATless mice. All the mice (7-8/group) tolerated the treatment well. Body weights at the start of the experiment were 46.7 Ϯ 7.7 and 48.6 Ϯ 6.7 g in the PPAR␥ and control ASO groups, respec-tively. After 4 weeks of treatment, mice receiving PPAR␥ ASO had a weight of 47.8 Ϯ 6.7 g, whereas those receiving control ASO had a weight of 54.2 Ϯ 6.3 g. There was a nonsignificant trend (p ϭ 0.08) for a greater weight gain in the control ASO group.
Hepatic mRNA levels of PPAR␥1 and PPAR ␥2 were dramatically decreased (80 -90%) in PPAR␥ ASO-treated mice (Fig. 8A). Consistent with those changes, hepatic levels of PPAR␥ protein and mRNA levels of CD36 and aP2 were significantly decreased in PPAR␥ ASO-treated mice (data not shown). Importantly, hepatic mRNA levels of FAS and ACC were also dramatically reduced in PPAR␥ ASO-treated mice (Fig. 8B). However, hepatic mRNA levels of SREBP-1c (Fig. 8C), CHREBP (Fig. 8D, left), and LPK (Fig. 8D, right) were not affected by ASO treatment. To determine whether the reductions in ACC and FAS mRNA were physiologically important, we measured lipogenesis in male apoB/BATless mice after inhibition of PPAR␥ with ASO. Incorporation of 3 H 2 O into FA was significantly decreased in PPAR␥ ASO versus control oligonucleotide-treated apoB/BATless mice (Fig. 8E), and this was associated with a significant 50% reduction in liver TG content in PPAR␥ ASO-treated mice (PPAR␥ ASO, 202.4 Ϯ 99.1, versus control ASO, 332.7 Ϯ 91.4 mg/g liver protein; p ϭ 0.04) (Fig. 8F). Of interest, plasma TG levels were not affected by 4 weeks of ASO treatment (212 mg/dl at the start of the PPAR ␥2 ASO treatment and 205 mg/dl at the end of treatment).
A similar experiment was performed with PPAR␥ ASO and control ASO in apoB mice (n ϭ 2 per group) on Western diets. mRNA levels for PPAR␥1 and PPAR␥2 were reduced more than 90%, whereas FAS and ACC mRNA levels fell about 90%. Consistent with the findings in apoB/BATless mice, liver TG was 89.8 Ϯ 12.5 mg/g liver protein in PPAR␥ ASO-treated mice compared with 296.3 Ϯ 77.8 mg/g liver protein in control ASOtreated mice.
Increased FA availability (from any source) for incorporation into VLDL TG could also be a consequence of decreased FA oxidation in the liver. To rule out this possibility, PPAR␣, CPT-1, and AOX mRNA levels were measured. None of these were increased in apoB/BATless mice, suggesting that FA oxidation was not different between the two groups of mice (Fig. 9).

DISCUSSION
Hepatic steatosis without excessive alcohol intake, called nonalcoholic fatty liver disease, is a common finding in association with obesity and insulin resistance (7,29,30). The basis for accumulation of TG in the liver is complex and can include reduced assembly and secretion of VLDL because of defects in the synthesis of apoB (31) or in microsomal triglyceride transfer protein activity (32). However, VLDL assembly and secretion are typically increased above normal in animals (3,4,(33)(34)(35) and people (36 -39) with insulin resistance, diabetes mellitus, and hepatic steatosis, suggesting that other abnormalities are playing key roles.
Increased TG synthesis could lead to hepatic steatosis even if VLDL assembly and secretion were increased. Hepatic TG synthesis and VLDL TG secretion are regulated by the delivery of plasma albumin-bound FA (40 -42). In apoB/BATless mice, despite the presence of insulin resistance, we did not find increased fasting plasma FA levels, even FIGURE 6. Hepatic levels of PPAR␥2 mRNA are increased in apoB/BATless mice. A, hepatic mRNA levels of PPAR␥1 and PPAR␥2 were measured by QPCR with mouse ␤-actin as internal control. The bar graphs represent the means Ϯ S.D. derived from each group of mice with the average value from apoB mice set as 100%. Open bars, apoB mice. Closed bars, apoB/BATless mice. B, hepatic PPAR␥ protein levels were measured by Western blot with a mouse anti-PPAR␥ monoclonal antibody (Santa Cruz Biotechnology). ␤-Actin was used as a control for the amount loaded on the gel. The bar graph represents the means Ϯ S.D. of PPAR␥2 protein levels derived from each group of mice with the average value from apoB mice set as 100%. Representative immunoblots are shown in each mouse type. C, hepatic mRNA levels of CD36 and aP2, two target genes of PPAR␥, were measured by QPCR using mouse ␤-actin as an internal control. The bar graph represents the means Ϯ S.D. derived from each group of mice with the average value from apoB mice set as 100%. * is p Ͻ 0.05. Open bars, apoB mice. Closed bars, apoB/BATless mice.

FIGURE 7. The increase in PPAR␥2 expression in livers of apoB/BATless mice is in hepatocytes, not Kupffer cells.
A, clodronate or PBS was administered as described under "Materials and Methods." After clodronate treatment for 2 days, mice were sacrificed, and liver was used to determine levels of F4/80 immunostaining (brown color). Liver from a PBS-treated mouse is shown in the left panel; a clodronate-treated liver is in the right panel. B, after clodronate treatment, hepatic mRNA levels of two macrophage-specific markers F4/80 and Nramp1 were measured by QPCR with mouse ␤-actin as internal control. The bar graphs represent the means Ϯ S.D. derived from each group of mice with the average value from PBS-treated mice as 100%. * is p Ͻ 0.05. Open bars, PBS-treated. Closed bars, clodronate-treated. C, after ablation of hepatic Kupffer cells in apoB/BATless, mRNA levels of PPAR␥1 and PPAR␥2, and their targets aP2 and CD36, and mRNA levels of FAS and ACC were measured by QPCR with mouse ␤-actin as internal control. The bar graphs represent the means Ϯ S.D. derived from each group of mice with the average value from PBS-treated mice set as 100%. Open bars, PBS-treated. Closed bars, Clodronate-treated. after extended fasting (3). In this study, plasma FA turnover, determined using [ 3 H]palmitic acid, tended to be higher in apoB/BATless mice compared with apoB mice, but this was not statistically significant. Taken together, these results argue against a major role for increased delivery of plasma albumin-bound FA to the liver in the hepatic steatosis observed in the apoB/BATless mouse.
Based on the above findings, we were left with three other possible causes for the accumulation of hepatic TG in the apoB/BATless mouse as follows: decreased oxidation of hepatic FA, increased hepatic uptake of TG-enriched VLDL and/or chylomicron remnants, and increased lipogenesis. We examined oxidation of FA in the liver by determining the expression of PPAR␣, a key transcription factor regulating several genes important for ␤-oxidation of fatty acids by the liver. We found that mRNA levels for PPAR␣, and two of its target genes, Cpt-1 and Aox, were not different in apoB/ BATless mice compared with apoB mice. These results suggest that there was no significant difference in hepatic fatty acid ␤-oxidation between apoB/BATless and apoB mice.
Although we have not directly determined the role of hepatic remnant lipoprotein uptake in apoB/BATless mice, our previously published study indicated that post-heparin plasma lipoprotein and hepatic lipase in the apoB/BATless mice were the same as in apoB mice (3). Those data suggested that lipolytic clearance of VLDL TG was efficient in apoB/BATless mice. Our finding that VLDL secretion was increased 2-3-fold indicated that the hypertriglyceridemia in the apoB/BATless mice was not because of reduced clearance of VLDL TG. Overall, it is unlikely that there was a greater return of TG-enriched remnants to the liver in apoB/ BATless mice.
The remaining pathway, de novo lipogenesis, was clearly increased in the apoB/BATless mice compared with apoB mice, as evidenced by a 2-fold increase in the incorporation of 3 H 2 O into fatty acids and increased expression of three key genes regulating FA and TG synthesis: Fas, Acc, and Scd1. Furthermore, protein levels of FAS and SCD were increased in livers of apoB/BATless mice. Lipogenesis is a modest but significant source of TG for VLDL secretion in rodents (33,43) and in humans with obesity and insulin resistance (44,45). Studies from the laboratory of Goldstein and Brown have characterized in detail the role of the basic/ helix-loop-helix/leucine zipper transcription factor, SREBP-1c, in the regulation of hepatic lipogenesis (8). More recently, their laboratory has extended this work to clarify the link between insulin and SREBP-1c regulation of lipogenesis (46,47). Importantly, hepatic Srebp-1c gene expression is increased in several rodent models of insulin resistance and hyperinsulinemia, even when insulin signaling through the insulin receptor substrate pathway is reduced (48). In the present study, however, apoB/BATless mice did not have increased levels of hepatic SREBP-1c mRNA, despite the presence of insulin resistance and hyperinsulinemia. Furthermore, levels of LXR␣ mRNA, a nuclear receptor that can stimulate Srebp-1c gene expression in the presence of increased insulin (47,49) and can also directly stimulate transcription of Fas (50), were not increased in the livers of apoB/BATless versus apoB mice. SREBP-1c transcriptional activity is also regulated at the post-translational level by enzymatic cleavage of the precursor transmembrane form, allowing the amino-terminal peptide to translocate into the nucleus (8). Our demonstration that both the precursor and mature forms of SREBP-1 protein were not different in apoB/BATless versus apoB mice is further evidence that the SREBP-1c pathway did not play a role in the increased lipogenesis present in apoB/BATless mice. Finally, gel mobility shift assays using hepatic nuclei isolated from apoB/BATless and apoB mice confirmed the absence of increased transcriptional activity for either an SRE from the Fas promoter or an LXRE from the Cyp7A1 promoter in apoB/BATless mice. Altogether, these different experimental approaches rule out a role for SREBP-1c in the stimulated lipogenesis in apoB/BATless mice.
The lack of evidence linking increased SREBP-1c transcriptional activity to increased lipogenesis in the apoB/ BATless mouse was surprising and led us to look for another stimulus for lipogenesis. One such a candidate was the newly identified transcription factor ChREBP, which binds to specific carbohydrate-responsive elements of several lipogenic genes and is regulated by glucose and cyclic AMP (9). In transfected hepatocytes, ChREBP stimulates the transcription of reporter constructs containing carbohydrate-responsive elements of the Acc and Fas genes (51). Additionally, mice lacking ChREBP have reduced lipogenesis without changes in SREBP activity (52). In apoB/BATless mice, fasting plasma glucose levels were not elevated compared with apoB mice (3), although the former mice are glucose-intolerant and insulin-resistant. In any event, hepatic mRNA levels of ChREBP, and its specific hepatic target gene LPK (53), were not different in apoB/BATless versus apoB mice. Thus, the ChREBP pathway for lipogenesis, like the SREBP-1c pathway, does not seem to account for increased lipogenesis in apoB/BATless mice versus apoB mice.
In the absence of evidence for either SREBP-1c or ChREBP playing important roles in the elevated lipogenesis in apoB/ BATless mice, we turned our attention to PPAR␥. This nuclear receptor, which is a key regulator of adipogenesis, is present in two isoforms, PPAR␥1 and PPAR␥2 (54). PPAR␥1 is expressed in many tissues, whereas significant PPAR␥2 expression is thought to be limited almost exclusively to adipose tissue. In adipose tissue, Ppar␥2 gene expression is regulated by insulin and nutrients. Importantly, increased expression of either or both isoforms has been observed in livers of obese, insulin-resistant rodents (55,56). In fact, specific induction of hepatic Ppar␥2 gene expression was demonstrated in the BATless mouse on a high fat diet (28). We have confirmed that finding in the present studies, and we have also demonstrated induction of several Ppar␥ gene targets in livers of these mice. Studies with clodronate confirmed that the increase in Ppar␥2 gene expression and activity was in hepatocytes and not Kupffer cells. Most importantly, using PPAR␥ antisense, we have directly linked reductions in hepatic Ppar␥ gene expression and protein to reductions in hepatic FAS and ACC mRNA levels, hepatic TG content, and hepatic lipogenesis. The effects of PPAR␥

Insulin-resistant Mouse with PPAR␥2-stimulated Lipogenesis
antisense on hepatic FAS and ACC mRNA levels occurred in the absence of any changes in the expression of SREBP-1c, CHREBP, or LPK. PPAR␥ ASO treatment had similar effects on lipogenic gene expression and hepatic TG content in apoB mice.
These novel findings add significant and important insights to results from recent studies in several laboratories that suggested an important, if not direct, role of PPAR␥ in regulating hepatic lipogenesis. Adenovirus-induced overexpression of PPAR␥1 in livers of PPAR␣ knock-out mice was sufficient to induce adipose tissue-specific gene expression and hepatic steatosis (57). However, it was not clear if overexpression of PPAR␥1 in that study would have caused increased lipogenesis without concomitant absence of PPAR␣. Stable expression of PPAR␥2 in a cultured cell line of hepatocytes resulted in similar changes in hepatic lipid metabolism and gene expression (58). Unlike our findings in apoB/BATless mice, the latter study reported significantly increased hepatic expression of SREBP-1c as well (58). Importantly, specific targeted hepatic deletion of PPAR␥ resulted in reduced hepatic lipogenesis in both a lipodystrophic model of insulin resistance and steatosis (11) and in ob/ob mice (10). The mRNA levels of the hepatic lipogenic genes, Fas, Acc, and Scd1, were reduced in both models of hepatic PPAR␥ deletion without obvious changes in Srebp-1c gene expression, which was increased in both of those mouse models. Inhibition of PPAR␥ expression by adenovirus-delivered RNA interference reduced lipogenic gene expression and hepatic TG content in mice that had reduced levels of cyclic AMP-responsive transcription factor (CREB) (12). Hepatic Srebp-1c gene expression was not altered in the latter study. In a recent report, adenoviral overexpression of PPAR␥2 was associated with steatosis in C57BL/6 mice; short hairpin RNA suppression of hepatic PPAR␥ decreased hepatic TG levels in db/db mice (59). Thus, taking all of the published data together, it appears that hepatic PPAR␥ can play a critical role in the regulation of hepatic lipid synthesis. However, in all of those studies, either SREBP-1c was elevated concomitantly or some other key hepatic regulatory gene (PPAR␣ and CREB) was altered. In our model, loss of brown adipose tissue leads to the complex phenotype of obesity, insulin resistance, and altered hepatic lipid metabolism that is regulated by aberrant expression of PPAR␥2.
Why is PPAR␥2 increased in the livers of apoB/BATless mice, and how might PPAR␥2 stimulate lipogenesis? It is possible that the modest increase in FA flux to the liver that we demonstrated, together with hyperinsulinemia, would induce specific expression of PPAR␥2 (60). Insulin signaling can regulate hepatic Ppar␥ gene expression; increased insulin action in livers with targeted deletion of PTEN (a phosphatidylinositol 3-kinase phosphatase) is associated with increased Ppar␥ (the authors did not differentiate between PPAR␥1 or PPAR␥2) gene expression (61), whereas hepatic Ppar␥ gene expression is decreased in mice lacking hepatic insulin receptors. 4 If insulin can directly elevate hepatic PPAR␥2 levels in the apoB/BATless mouse, it would indicate that insulin signaling to PPAR␥2 was normal despite the presence of significant insulin resistance through the phosphatidylinositol 3-kinase pathway in apoB/ BATless livers. 4 Compartmentalization of hepatic insulin resistance has been observed to be related to intact insulinmediated stimulation of Srebp-1c gene expression in the otherwise insulin-resistant liver of the ob/ob mouse (48). A recent study indicated that LXR is able to induce Ppar␥ gene expression in adipocytes (62); it is not known if LXR can activate PPAR␥ in the liver. However, we did not find any evidence of increased LXR transcriptional activity in the livers of apoB/ BATless mice. CREB appears to inhibit Ppar␥ gene expression by stimulating expression of the transcriptional repressor called Hairy Enhancer of Split gene (Hes-1) (12), but neither CREB mRNA nor HES-1 mRNA levels were reduced in apoB/ FIGURE 8. Antisense-mediated reductions of PPAR␥1 and PPAR␥2 mRNA levels resulted in reductions of mRNA levels of FAS and ACC, hepatic TG content, and hepatic lipogenesis. A, apoB/BATless mice were treated for 4 weeks with either specific PPAR␥ ASO or control ASO by intraperitoneal administration. Hepatic mRNA levels of PPAR␥1 and PPAR␥2 were measured by QPCR with mouse ␤-actin as internal control. The bar graphs represent the means Ϯ S.D. derived from each group of mice with the average value from apoB/Batless mice that received control ASO set as 100%. B, hepatic mRNA levels of FAS and ACC were measured by QPCR with mouse ␤-actin as internal control mice after 4-week intraperitoneal PPAR␥ ASO versus control ASO administration. The bar graphs represent the means Ϯ S.D. derived from each group of mice with the average value from apoB/Batless mice that received control ASO set as 100%. C, hepatic mRNA levels of SREBP-1c were measured in male apoB/BATless mice after 4-week intraperitoneal PPAR␥ ASO versus control ASO administration. The bar graphs represent the means Ϯ S.D. derived from each group of mice with the average value from apoB/Batless mice that received control ASO set as 100%. D, hepatic mRNA of ChREBP and L-PK were measured in male apoB/BATless after 4 weeks intraperitoneal PPAR␥ ASO versus control ASO administration. Open bars, control ASO. Closed bars, PPAR␥ ASO. The bar graphs represent the means Ϯ S.D. derived from each group of mice with the average value from apoB/Batless mice that received control ASO set as 100%. E, hepatic de novo lipogenesis was measured by determining incorporation of tritiated water into fatty acids as described under "Materials and Methods." Lipogenesis in apoB/BATless mice treated with PPAR␥ ASO was normalized against lipogenesis in mice treated with control ASO. F, hepatic triglyceride content was determined as described under "Materials and Methods." TG is depicted as milligram/g of liver protein. * is p Ͻ 0.05. BATless mice (data not shown). Thus we cannot, at this time, determine why Ppar␥2 gene expression is uniquely and aberrantly increased in the apoB/BATless mouse. Even less clear is the manner in which increased Ppar␥2 gene expression would induce lipogenesis; although adipocyte PPAR␥ expression is critical for full adipogenesis, there are no known PPAR-response elements in the lipogenic genes we have studied. Of note, however, is the recent report that a PPAR␦ agonist increased hepatic expression of the genes for FAS and ACC and that a putative PPAR-response element was present in the Acc promoter (27). Whatever the basis for our findings, further delineation is clearly required, particularly because we have demonstrated that reductions in PPAR␥ expression in apoB/ BATless mice were associated with improvement in hepatic steatosis.