Modulation of Resistin Expression by Retinoic Acid and Vitamin A Status

(±) Arterenol bitartrate salt (norepinephrine), all-trans RA (ATRA); isopro-pyl-(E,E)-(R,S)-11-methoxy-3,7,11-trimethyldodeca-2,4-dienoate (metho-prene); and p-[(E)-2-(5,6,7,8-tetrahydro-5,5,8,8-tetramethyl-2-naphthalenyl)-1-propenyl]benzoic acid (TTNPB) were from Sigma (St. Louis, MO). 9-cis RA, Ro 40-6055, and Ro 25-7386 were gifts from Hoffmann-La Roche (Basel, Switzerland). Newborn calf serum and FBS were from Linus (Madrid, Spain), and other cell culture reagents were from Sigma.

3T3-L1 murine fibroblasts (American Type Culture Collection, LGC Deselaers SL, Barcelona, Spain) were maintained in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% newborn calf serum under 8% CO₂/92% air at 37°C. Differentiation was induced 2 days after confluence by incubating the cells for 48 h in DMEM containing 5 μmol/l insulin, 0.5 mmol/l 3-isobutyl-1-methyl-xanthine, 0.25 μmol/l dexamethasone, and 10% FBS, followed by 48 h in DMEM containing 10% FBS and 5 μmol/l insulin (no insulin thereafter). Adipogenesis was monitored by morphologic examination of the cells for accumulation of lipid droplets. Cells were exposed to ATRA and other compounds of interest on days 6–7 after induction of differentiation, when >80% expressed the adipocyte phenotype, and RNA was routinely extracted 24 h later. In some experiments, the culture medium was collected to determine resistin protein levels. Cytotoxicity of vehicle (DMSO) and all agonists (including ATRA) of retinoid receptors at the working concentrations used was measured by the lactate dehydrogenase method (27), using a commercial kit (Roche Diagnostics, Mannheim, Germany).

BAT precursor cells from 4-week-old NMRI mice were isolated and cultured as described previously (28). Cells were exposed to RA or norepinephrine on day 7, when they had reached confluence and started to differentiate, and RNA was extracted 24 h later.

Day 6 after hormonal onset of adipogenesis, 3T3-L1 adipocytes were treated with 0.1 μmol/l ATRA, 10 μmol/l ATRA, or vehicle (DMSO) for 24 h. Cells were then washed with PBS, submitted to cross-linking with 1% formaldehyde in PBS at 37°C for 10 min, rinsed and scraped in ice-cold PBS, centrifuged for 4 min at 700g, and resuspended in lysis buffer (1% SDS, 5 mmol/l EDTA, 50 mmol/l Tris-HCl [pH 8.1]). After a 20-min incubation on ice, samples were sonicated at 15-s pulses three times on ice. The lysates were centrifuged at 14,000g for 10 min; a small aliquot of the supernatant served as the input control, and the rest was diluted in buffer (1% Triton X-100, 2 mmol/l EDTA, 150 mmol/l NaCl, 20 mmol/l Tris-HCl [pH 8.1]) containing protease inhibitors. Samples were precleared with 2 μg of sheared salmon sperm DNA and 45 μl of protein A-Sepharose beads for 2 h.

Immunoprecipitation with antibodies against the following proteins was performed overnight: C/EBP-α, p300, cAMP response element-binding protein (CREB) binding protein (CBP), retinoblastoma protein (RB), and normal rabbit IgG (all from Santa Cruz Biotechnology, Santa Cruz, CA). Samples were then incubated with 45 μl of protein A-Sepharose beads for 1 h. Precipitates were washed as described previously (23) and extracted by incubating with elution buffer (1% SDS, 0.1 mol/l NaHCO₃) at 65°C for 6 h. DNA fragments were purified using a commercial kit (Qiagen, Hilden, Germany). Two to 5 μl of purified sample was used in 31 cycles of PCR. Primers surrounding the C/EBP-α site in the resistin promoter had sequences 5′-GTCTTGGCTCCTAGCCTTGC-3′ and 5′-GTTGACTTCTGGCCCATCC-3′ (23). PCR products were run on agarose gels, stained with ethidium bromide, recorded using an image recording system (Gelprinter; TDI, Madrid, Spain), and quantified using the Kodak 1D Image Analysis Software 3.5 for Windows (Eastman Kodak, Rochester, NY).

Animals were obtained from CRIFFA (Barcelona, Spain) and were treated in accordance with our institutional guidelines.

Twelve-week-old NMRI male mice that were fed regular diet (Panlab, Barcelona, Spain) received one daily subcutaneous injection of ATRA at a dose of 10, 50, or 100 mg/kg body wt during the 4 days before they were killed (seven animals/group). Controls received the vehicle (olive oil).

Four-week-old C57BL/6J (B6) male mice were fed either a normal-fat or a high-fat diet (10 and 45% of total energy as fat, mainly lard, respectively) with either the standard vitamin A content or a 40-fold excess (8 mg and 320 mg retinyl palmitate/kg diet, respectively) for 18 weeks, as previously described (17) (four animals/group). In an independent experiment, 4-week-old NMRI mice were fed the normal-fat diet supplemented or not with vitamin A for 11 weeks (six animals/group). Diets were from Research Diets (New Brunswick, NJ). The vitamin A-supplemented animals did not show any external signs of vitamin A toxicity or significant liver enlargement.

The animals were kept at 22°C under 12-h light/dark cycles (lights on at 0800) and were killed with CO₂ followed by decapitation at the start of the light cycle. Blood was collected from the neck, and serum was prepared and frozen at −20°C. Interscapular BAT (iBAT), inguinal WAT (iWAT), and epididymal WAT (eWAT) were excised in their entirety, weighed, frozen in liquid nitrogen, and stored at −70°C.

Tests were conducted in four 12-week-old NMRI male mice and in the same animals 1 week later, after they had received four subcutaneous injections of ATRA (at a dose of 100 mg/kg body wt, in 100 μl of olive oil; one daily injection the 4 days preceding the test). Glucose (3.375 g/kg animal) was administered intraperitoneally after a 6-h fast (from 2400 to 0600), and glucose levels were measured in tail blood collected at the indicated times.

Total RNA was extracted with TriPure reagent (Roche Diagnostics). Ten to 20 μg was fractionated by agarose gel electrophoresis, blotted onto a nylon membrane by capillary blotting, fixed with ultraviolet light, and hybridized as described previously (17). An antisense oligonucleotide (5′-TCCCACGAGCCACAGGCAGAGCCACAGGAGCAGC-3′) based on the mouse resistin cDNA sequence labeled at both ends with a single digoxigenin ligand was used as a hybridization probe (29). Hybridization signals were detected by chemiluminescence using CDP-Star (Roche Diagnostics) as the substrate, with membranes exposed to Hyperfilm ECL (Amersham Biosciences UK, Buckinghamshire, U.K.) for up to 60 min. Bands in films were analyzed by scanner photodensitometry and quantified as described above. Blots were stripped and reprobed for other mRNAs of interest and 18S rRNA to check equal loading of gels and transfer efficiency. Antisense oligonucleotides used to detect leptin mRNA, UCP1 mRNA, and 18S rRNA were as described previously (17).

A total of 50 μl of culture medium or 10 μl of serum was solubilized and boiled in Laemmli sample buffer containing 20% 2-mercaptoethanol, fractionated in 15% SDS-PAGE gels, and electrophoretically transferred onto nitrocellulose membranes. Blots were stained with Ponceau S solution to visualize the amount of total protein in each lane. After blocking in PBS-Tween containing 5% fat-free dried milk, membranes were incubated with a polyclonal antibody against mouse resistin (1:1,000; Linco Research, St. Charles, MO) for 1 h, washed, and incubated with a horseradish peroxidase-conjugated secondary antibody (1:10000; Research Diagnostics, Flanders, NJ). Blots were developed using an enhanced chemiluminescence Western blotting analysis system (Amersham Biosciences UK). Bands in films were analyzed and quantified as described above.

Circulating levels of triacylglycerols, glucose, and insulin were determined using commercial kits (Triglyceride [INT]20 from Sigma Diagnostics; d-glucose ultraviolet method from Roche Biopharm, Darmstadt, Germany; and rat-insulin enzyme-linked immunosorbent assay from DRG Instruments, Marburg, Germany, respectively).

Data are expressed as mean ± SE. Statistical significance was assessed by Student’s t test or by one- or two-way ANOVA followed by least-significant difference and Student’s t test post hoc comparisons, respectively. The minimum significance level was set at P < 0.05.

Resistin mRNA levels were dose-dependently downregulated by ATRA and 9-cis RA both in primary brown adipocytes differentiated in culture and in 3T3-L1 white adipocytes (Fig. 1A and B). In both cell systems, ATRA was more effective than 9-cis RA. As little as 0.01 μmol/l ATRA reduced resistin mRNA levels by 80% in primary brown adipocytes and by 60% in 3T3-L1 white adipocytes. In the brown adipocytes, this dose was insufficient to trigger the induction of the UCP1 gene, which was triggered, as expected (12,13), by higher doses of the retinoids (Fig. 1A). The inhibitory effect of RA isomers on resistin expression in brown adipocytes was only partially reproduced by 1 μmol/l norepinephrine, which, similar to retinoids, induced the expression of the UCP1 gene, as expected (Northern blot in Fig. 1A). The inhibitory effect of retinoids on resistin mRNA levels exhibited a relatively lengthy time course, with a half-maximal reduction observed after ∼10 h of exposure of 3T3-L1 adipocytes to 1 μmol/l ATRA (Fig. 1C), and was followed by reduced resistin accumulation in the culture medium (Fig. 1D). Expression of resistin transcripts in differentiated primary brown adipocytes agrees with the recently reported differentiation-dependent expression of resistin in the brown adipocyte cell line T37i (30).

We assayed whether nonisomerizable synthetic compounds that selectively bind and activate RARs or RXRs could reproduce the inhibitory effect of RA on resistin gene expression in 3T3-L1 adipocytes (Table 1). Both TTNPB, which activates all RAR subtypes, and Ro 40-6055, a selective RAR-α agonist, reduced resistin mRNA abundance in a dose-dependent manner. The pan-RAR agonist was more effective than the selective RAR-α agonist, suggesting that different RAR subtypes mediate the inhibitory effect of RA on the resistin gene. Although with lower potency than ATRA and TTNPB, methoprene, a pan-RXR agonist, and Ro 25-7386, a selective RXR-α agonist, also reduced resistin mRNA abundance in a dose-dependent manner; the two compounds were similarly effective, reproducing the inhibitory effect of 1 μmol/l ATRA at 100 μmol/l, which for methoprene is the concentration known to activate RXR in mammalian cells (31). Only TTNPB and Ro 25-7386 at the highest concentration tested (100 μmol/l) had a significant cytotoxic effect over vehicle (DMSO), leading to a 2.6–2.8 increment of lactate dehydrogenase activity in the medium (P < 0.001 for both compounds, Student’s t test).

We reasoned that the inhibitory effect of RA on resistin expression could result, at least in part, from impaired binding of C/EBP-α and/or its coactivators to the resistin gene promoter. To test this, we performed chromatin immunoprecipitation assays measuring C/EBP-α, p300, CBP, and RB association with a proximal region of the resistin promoter containing the C/EBP-α binding site (23) in mature 3T3-L1 adipocytes treated with ATRA (0.1 and 10 μmol/l) or vehicle for 24 h. p300 coactivates C/EBP-α (32), and both p300 and the closely related coactivator CBP were reported to be recruited to the mouse resistin gene promoter in differentiated 3T3-L1 mature adipocytes through their interaction with DNA-bound C/EBP-α (23). Hypophosphorylated RB also coactivates C/EBP-α, this activity being of importance for adipocyte terminal differentiation (33–35). At 10 μmol/l, ATRA treatment resulted in reduced association of C/EBP-α, p300, CBP, and RB with the resistin gene promoter by 45 ± 9%, 33 ± 10%, 27 ± 6%, and 46 ± 12%, respectively (n = 4; Fig. 2). The reduced association of C/EBP-α and its coactivators to the resistin gene promoter was already evident or even more marked in the cells that were treated with 0.1 μmol/l ATRA (Fig. 2).

Resistin mRNA levels in eWAT, iWAT, and iBAT (Fig. 3) and serum resistin levels (Fig. 4) were severely reduced in ATRA-treated NMRI mice. In accordance with previous reports (15,36), RA administration also triggered a reduction of adipose tissue leptin expression that was evident in the three depots (Fig. 3) and a reduction of body weight and of the mass of all fat pads analyzed (Table 2).

Circulating levels of insulin, glucose, and triacylglycerols in ad libitum feeding conditions trended lower in the ATRA-treated NMRI mice (Table 2), suggesting an improvement of insulin sensitivity. To address this further, we performed glucose tolerance tests in a separate group of NMRI mice before and after ATRA administration. The treatment triggered a 13.3% reduction of body weight. Circulating glucose levels at 0, 30, 90, and 180 min after the glucose load were, respectively, 4.0 ± 0.5, 28.9 ± 1.9, 16.6 ± 2.8, and 9.4 ± 0.6 mmol/l before the ATRA treatment and 4.6 ± 0.2, 21.0 ± 3.7, 9.6 ± 1.9, and 4.4 ± 0.8 mmol/l afterward. The difference was of statistical significance at t = 180 min (P = 0.002, Student’s t test). The area under the glucose curve was 39.5 ± 3 cm² before ATRA treatment and 25.6 ± 3.6 cm² after it, indicating a statistically significant (P = 0.025, Student’s t test) enhancement of glucose disposal upon ATRA treatment.

We analyzed resistin expression in mice that were chronically fed diets with a 40-fold excess of vitamin A relative to the usual content (Table 3). A similar dietary manipulation was previously shown to result in a high vitamin A status in rats, as revealed by the increment of liver retinol concentration (37). Our assumption was that supplementation could result in increased availability of RA to tissues and could reproduce, at least in part, the effects of acute RA treatment. In fact, in both rats and mice, vitamin A supplementation was shown to reproduce the effects of acute RA doses downregulating leptin expression and upregulating UCP1 expression (17,37).

Resistin mRNA levels in eWAT and circulating resistin levels were reduced in both B6 and NMRI mice that were chronically fed a normal-fat high-vitamin A diet (for 18 and 11 weeks, respectively) compared with controls that were fed the same diet with the normal vitamin A content. Vitamin A supplementation did not significantly affect body weight or adiposity of either B6 (17) or NMRI mice, although in the latter, the two parameters trended lower after supplementation (by 9.7 and 30%, respectively). In the obesity-prone B6 mice, high-fat diet feeding for 18 weeks caused 26% overweight, as reported previously (17), and a significant reduction of both eWAT resistin mRNA levels and serum resistin levels. Supplementation of the high-fat diet with vitamin A did not affect the final obesity reached (17) and did not lead to further significant decreases of eWAT resistin mRNA levels or serum resistin levels.

The present study shows that RA inhibits resistin expression in adipocyte cell model systems and in vivo, after its administration to intact mice, both at mRNA and protein levels. Both RAR and RXR agonist reproduced the inhibitory effect of RA on resistin expression in 3T3-L1 adipocytes, strongly suggesting that both RAR- and RXR-dependent signaling pathways mediate the RA effect. Liganded RXR may be acting as a heterodimer with PPAR-γ, which, although results are conflicting (38–40), has been previously implicated in the inhibition of the resistin gene (1,23,24,41–43). Indeed, the finding that both PPAR-γ agonists and RXR agonists—including the endogenous ligand, as first demonstrated here—inhibit resistin gene transcription reinforces the idea of the involvement of PPAR-γ:RXR, because this entity is known to respond to ligands of both partners (21). In any case, our results are the first to suggest that not only liganded RXR but also liganded RARs can repress resistin gene transcription.

Liganded retinoid receptors may act on the resistin gene through different, nonmutually exclusive mechanisms. The fact that the inhibitory effect of both ATRA (Fig. 1C) and the TZD PPAR-γ ligand rosiglitazone (23) on resistin expression in 3T3-L1 adipocytes follows a relatively lengthy time course suggests that liganded PPAR-γ:RXR may be inducing a protein or proteins that repress resistin gene transcription. Our chromatin immunoprecipitation assay results (Fig. 2) indicate that impairment of C/EBP-α activity on the resistin promoter may also contribute to the ATRA effect. We favor the idea that this impairment is mediated by RARs because liganded RARs are known to interfere with the transcriptional activity of C/EBPs (26); in addition, rosiglitazone treatment did not affect the occupancy of the resistin promoter by C/EBP-α or p300/CBP (23), making it unlikely that the interference detected here is mediated by liganded PPAR-γ:RXR. Liganded retinoid receptors may also repress resistin gene expression through direct binding to negative RA and/or PPAR response elements in its promoter or in an enhancer. No conventional RA response elements were reported in the promoter region of either the mouse or the human resistin gene analyzed to date (23,25), and no functional PPAR-γ binding site was found within 6.2 kb upstream of the transcriptional start site of the mouse resistin gene (23), but it remains possible that negative response elements are present at other locations.

Both resistin and leptin mRNA levels were reduced in adipose tissues of ATRA-treated NMRI mice (Fig. 3). These results add to the striking similarity between the regulation of these two adipokines, both of which seem to be negatively controlled by PPAR-γ (23,24,44) and positively controlled by C/EBP-α (23,24,45) at the transcriptional level. Acute RA treatment causes in rodents a reduction of body weight and adiposity, not explained solely by the observed changes in food intake, that has been reproduced here (Table 2) and was previously shown to correlate with a depressed adipogenic/lipogenic potential in adipose tissues (16) and an increased expression of UCPs in BAT and muscle (13,15,17). Teleologically, reduction of both a signal that inhibits food intake and enhances energy expenditure (leptin) and a signal that may function as an inhibitor of adipocyte differentiation and as a resistance factor to the anabolic action of insulin (resistin) may be part of a security mechanism to avoid fat depletion in the context of an overall effect of high RA doses promoting fat mobilization.

An improvement of glucose tolerance paralleled the reduction of adipose tissue resistin mRNA and circulating resistin levels in ATRA-treated NMRI mice, which is compatible with resistin’s acting as an insulin-resistance factor in normal mice. However, it is likely that the improvement of glucose tolerance is, at least in part, a secondary response to the reduction of body weight elicited by ATRA treatment. In addition, other effects of ATRA that may contribute to changes in glucose tolerance—including effects on the expression and circulating levels of other adipokines potentially related to insulin sensitivity, such as tumor necrosis factor-α or adiponectin—cannot be discarded. Further studies are needed to evaluate precisely the relationship among vitamin A status, insulin sensitivity, and adipokine expression.

Not only acute ATRA treatment but also chronic dietary vitamin A supplementation reduced resistin expression, as evidenced in two different strains of mice (Table 3). Remarkably, although resistin expression was reduced in high-fat diet-fed B6 mice as compared with their lean controls, the inhibitory effects of high-fat diet and vitamin A supplementation on the resistin gene were not additive (Table 3), suggesting that common mechanisms are involved. Reduced resistin mRNA levels in fat depots of obese animals is well established (reviewed in 2–4). Concerning circulating resistin, our finding of reduced levels in obese B6 mice is in line with results of Rajala et al. (46) and Maebuchi et al. (47) and in contrast with the increased serum levels reported by Steppan et al. (1) in rodents with genetic and diet-induced obesity and diabetes. Changes of circulating resistin in human obesity are also controversial (48,49). The explanation for these differences is unclear: for diet-induced obesity models, it may be related to differences in the composition of the high-fat diets used. Reduced circulating resistin in obese animals apparently conflicts with resistin’s being a critical link between obesity and insulin resistance. However, as noted previously (4), the correlation between circulating levels of a bioactive peptide and its effects may not be linear, because receptor and postreceptor mechanisms (which for resistin are largely unknown) may be critical.

Gene-nutrient interactions are likely to contribute to the current epidemic of obesity and type 2 diabetes in industrialized societies. The inhibition of resistin expression by vitamin A characterized in this work may be a specific gene-nutrient interaction of interest in this context, in view of the potential relationship of resistin with insulin resistance, which is currently the subject of an intense debate (2–4).

This work was supported by the Spanish Government (Dirección General de Investigación, BFI2000-0988-C06-01; Ministerio de Sanidad y Consumo, FIS 01/1379 and G03/028), and the European Union (COST Action 918; DLARFID project, QLRT-2001-00183). F.F. was the recipient of a doctoral fellowship from the Spanish Government.

We thank Dr. Francisca Serra for defining the resistin probe used in the Northern blot analysis.