Transcriptional Regulation of Adipocyte Hormone-Sensitive Lipase by Glucose

Constructs containing human adipocyte HSL 5′ flanking regions from −2,400 to −31 relative to the transcription start point and 38 bp of the 5′ untranslated region in pGL3basic vector encoding a modified firefly luciferase (Promega, Madison, WI) have been described previously (17). Mutant HSL promoter constructs were generated by synthetic nucleotide assembly in pGL3basic using the KpnI site of the vector and the XbaI site of the HSL promoter. These constructs were made using standard procedures (18) and sequenced using a Perkin Elmer Dye terminator sequencing kit and an ABI 373 sequencer. Assembly of the double-stranded oligonucleotides led to substitution of the characterized binding sites by a mutagenized box with an XbaI site. pCR3-USF1 and -USF2a expression vectors, pCMV-Sp1, and pSVsport-ADD1 and pSVsport-ADD1403, were gifts from M. Raymondjean and M. Vasseur-Cognet (INSERM U 129, Paris), A.P. Butler (M.D. Anderson Cancer Center, University of Texas, Houston, TX), and B. Spiegelman (Dana Farber Cancer Institute, Boston, MA), respectively.

In agreement with French laws on biomedical research, subcutaneous abdominal adipose tissue was obtained from female subjects undergoing plastic surgery. Human adipocytes in primary culture were differentiated as described by Hauner et al. (19), with modifications. Briefly, all visible fibrous material and blood vessels were discarded from adipose tissue samples. The remaining adipose tissue was cut into small pieces and digested in PBS containing 0.15% collagenase and 2% BSA under gentle shaking at 120 cycles/min at 37°C. After centrifugation, the stromal cell fraction was resuspended and incubated in an erythrocyte-lysing buffer (155 mmol/l NH₄Cl, 5.7 mmol/l K₂HPO₄, and 0.1 mmol/l EDTA). Cells were filtered through a nylon mesh (150 μm). After additional centrifugation, washing, and filtration (70 μm), stromal cells were suspended in Dulbecco’s modified Eagle’s medium (DMEM)/Ham’s F-12 medium supplemented with 10% fetal calf serum (FCS) and put into culture at 30,000 cells/cm². After 16- to 20-h incubation, cells were refed with a chemically defined serum-free medium consisting of DMEM/Ham’s F-12 supplemented with 15 mmol/l NaHCO₃, 15 mmol/l HEPES, 33 μmol/l biotin, 17 μmol/l panthotenate, 10 μg/ml human transferrin, and 50 μg/ml gentamicin. Insulin (66 nmol/l), 1 nmol/l triiodothyronine, 100 nmol/l cortisol, and 1 μg/ml troglitazone were added for the first 3 days. Cells were cultured for 10 days at 37°C in a humidified atmosphere of 5% CO₂. 3T3-F442A preadipocytes were cultured in DMEM containing donor calf serum (10%), penicillin (200 units/ml), and streptomycin (50 mg/l) at 37°C in a humidified atmosphere of 7% CO₂. For adipocyte differentiation, DMEM was supplemented with insulin (50 nmol/l) once cells reached confluence, and donor calf serum was replaced with 10% FCS. Medium was renewed every 2 days for 10 days. Differentiated human and murine adipocytes were treated in insulin-free, serum-free DMEM containing 25 mmol/l glucose or 1 mmol/l glucose supplemented with 0.1 mmol/l lactate and 1 mmol/l pyruvate to maintain energy balance of the cells. Use of 10 mmol/l instead of 0.1 mmol/l lactate did not modify the glucose response.

Luciferase constructs (20 μg) were stably transfected into exponentially growing 3T3-F442A cells by cotransfection with 2 μg of G418-resistant pMC1neo vector using CaPO₄-mediated transfection (18). After 2 weeks of G418 (0.6 mg/ml) selection, an average of 100 colonies were recovered and pooled. This pool of clones showed adipocyte differentiation at a frequency similar to that of the parental 3T3-F442A cell line. Primary rat white adipocytes were isolated by collagenase digestion of periepididymal fat pads from young (150–200 g) male Wistar rats. Luciferase construct (20 μg) and 5 μg of pCMVlacZ (gift of P. Furth, University of Maryland, Baltimore, MD) were introduced into rat primary adipocytes by electroporation using a Biorad apparatus and cultured as previously described (20). β-Galactosidase expression was used to normalize transfection efficiency and was quantified using an o-nitrophenyl-β-d-galactopyranoside colorimetric assay (18). HepG2 and COS7 cells were cultured in DMEM containing 10% FCS, 200 units/ml penicillin, and 50 mg/l streptomycin. The cells were transfected with Fugene-6 reagent (Roche) according to the manufacturer’s protocol using a total of 2 μg plasmid per 3.5-cm-diameter dish. In addition, 50 ng of the Renillia luciferase expression vector (pRL-SV40 from Promega) was cotransfected to check for transfection efficiency. Before harvesting, cell monolayers were rinsed with calcium- and magnesium-free PBS. Cells were then scraped in passive lysis buffer (Promega). Firefly and Renilia luciferase assays were performed as recommended by the manufacturer’s protocol (Promega) in a Labsystem luminometer.

Adipose primary cell culture total RNA was extracted using the Qiagen RNeasy kit. Total RNA was stored at −80°C until analysis. Quantitation of HSL and cyclophilin mRNA levels were performed by reverse transcription-competitive PCR as described previously (21). The reverse transcription step was performed on 100 ng of total RNA with a specific antisense primer and Omniscript reverse transcriptase (Qiagen). cDNA samples were then amplified by PCR using sense and antisense primers in the presence of known amounts of a specific DNA competitor. PCR products were separated by capillary electrophoresis and quantified using the ABI PRISM 310 Genetic Analyzer system with the Genescan program (PE Applied Biosystem). HSL mRNA levels were normalized with cyclophilin mRNA levels.

3T3-F442A preadipocyte and adipocyte nuclear extracts were prepared according to standard procedures (18). Probe preparation and electromobility shift assay (EMSA) conditions were described previously (22). Briefly, 150,000 cpm/lane (0.5 ng of end-labeled DNA) and 4 μg of nuclear extracts were incubated for 15 min on ice. Upstream stimulatory factor (USF) and Sp1 competitor oligonucleotides were from Promega. The HNF1 oligonucleotide used as nonspecific competitor has been described (22). Supershift assays using antibodies against USF1 and USF2 (gift of M. Raymondjean) and against Sp1 (sc-59X; Santa Cruz) and Sp3 (sc-644X; Santa Cruz) were performed by incubating the nuclear extracts with the antibodies for 15 min on ice before addition of the probe. After electrophoresis, the gel was fixed in 10% acetic acid/10% ethanol and dried. Gels were exposed and analyzed using a Molecular Dynamics SI445 Phosphorimager.

Total RNA was isolated from perigonadal adipose tissue of fasted USF1- and USF2-deficient mice (23,24). Wild-type littermates were used as controls. Northern blot analysis was performed using mouse HSL (12) and ribosomal 18S cDNA probes as described previously (25). Specific signals were quantitated using a Phosphorimager (Molecular Dynamics).

The nonparametric Mann-Whitney U test for unpaired values was used to compare HSL mRNA levels. ANOVA with least-square difference post hoc test was used for comparisons of luciferase activities (SPSS). P < 0.05 was the threshold of significance.

To establish that human HSL was regulated by glucose, a primary culture system was used to obtain human differentiated adipocytes. Cells were treated for 48 h in a serum-free medium containing 25 or 1 mmol/l glucose. HSL mRNA levels were quantified using reverse transcription-competitive PCR. The low glucose concentration led to a decrease in the ratio of HSL to cyclophilin mRNA levels from 0.049 ± 0.003 to 0.024 ± 0.004 (P < 0.05, n = 3).

Because there is no established human preadipocyte cell line available, we used the mouse 3T3-F442A preadipocyte cell line to analyze human HSL promoter activity in response to glucose. Luciferase constructs containing 2,400 to 31 bp of the 5′ flanking region and 38 bp of the 5′ untranslated region (17) were stably transfected into preadipocytes. After 10 days of adipose differentiation, each pool of transformants was maintained for 24 h in DMEM containing 1 or 25 mmol/l glucose. In 25 mmol/l glucose medium, luciferase activity was high for the −2,400/+38, −280/+38, and −137/+38 constructs. The −86/+38, −57/+38, and −31/+38 constructs showed very weak or null promoter activity. For the −2,400/+38, −280/+38, and −137/+38 constructs, promoter activity was higher in 25 mmol/l than in 1 mmol/l glucose medium (Fig. 1A). These results indicated that the −137-bp DNA segment contained the regulatory elements mediating the glucose response and that the −137/−86 region was crucial for promoter activity in the chromatin context. A dose-response of glucose effect is shown in Fig. 1B. The maximal stimulatory effect was observed at 15 mmol/l of glucose. The addition of FCS and insulin in the culture medium containing 25 mmol/l glucose did not modify HSL promoter activity.

To determine more precisely the sequences involved in the glucose responsiveness, a computer-assisted analysis of the −137-bp region revealed several potential nuclear factor binding sites (Fig. 2). A consensus CACGTG E-box motif was found between −106 and −101, and two GC-boxes were located, between −30 and −43 and between −82 and −71. E-boxes are known to bind basic helix-loop-helix leucine zipper transcription factors such as USFs. The boxes are found in several glucose response elements of hepatic genes (26). GC-boxes are bound by the Sp family of zinc finger nuclear factors. Two GC-rich sequences in promoter II of acetyl-coA carboxylase have been shown to mediate glucose responsiveness (27).

EMSA was performed with the set of oligonucleotides described in Fig. 2. EMSA of the HSL E-box performed with adipocyte nuclear extracts showed two bands (Fig. 3). Competition with unlabeled oligonuceotide showed that the DNA-protein interaction was specific for the two complexes. While the nature of the band of higher mobility (X in Fig. 3) is still elusive, the lower-mobility band clearly corresponded to binding of USF1 and USF2 dimers. Indeed, anti-USF1 and anti-USF2 antibodies induced a supershift demonstrating the presence of USF1/USF2 heterodimers as previously shown in rat liver (28). The residual binding observed in the presence of anti-USF1 and anti-USF2 antibodies corresponded to USF1 and USF2 homodimers, respectively. A similar binding pattern was resolved using preadipocyte nuclear extracts (data not shown). Some glucose-responsive genes harbor two E-boxes separated by 5 bp (29). Because a half E-box CACTGA is located 5 bp 3′ from the HSL E-box (Fig. 2), we speculated that the sequence might be weakly bound by USF factors. However, no specific binding was observed using preadipocyte and adipocyte nuclear extracts or cellular extracts prepared from COS7 cells transfected with USF1 and USF2 expression vectors (data not shown).

EMSA of the HSL 5′ GC-box (Fig. 4A) showed two bands that displayed a binding pattern characteristic of members of the Sp family of nuclear factors. The binding to the probe was specific, since competition was observed with the HSL 5′ GC-box and the Sp1 control oligonucleotides but not with the unrelated HNF1 oligonucleotide.Antibody against Sp1 induced a weak supershift of the low-mobility complex. Antibody against Sp3 was able to totally supershift binding proteins in the high-mobility complex. Combination of the two antibodies led to the disappearance of the two bands. HSL 5′ GC-box oligonucleotide did not completely abolish Sp1 binding to the probe, but competition with the Sp1 control oligonucleotide led to a complete disappearance of the bands. This indicated that Sp1 and Sp3 had a lower affinity for the HSL 5′ GC-box than for the Sp1 control competitor. This observation is in accordance with previously published results showing a better affinity of Sp1 for GGGCGGG than for GGGTGGG motifs (30). The use of the probe corresponding to the HSL 3′ GC-box with the ideal consensus sequence GGGCGGG led to the same pattern (Fig. 4B). Cross-competitions confirmed that HSL 3′ GC-box was a better competitor of Sp1 and Sp3 binding than HSL 5′ GC-box.

To evaluate the functional importance of each of the boxes on HSL promoter activity in adipocytes, we performed single and multiple site-directed mutageneses of the sites. Lack of binding of the cognate factors to the mutated sites was controlled by EMSA (data not shown). Mutant constructs were transiently transfected into primary rat adipocytes maintained in serum-free DMEM containing 25 mmol/l glucose (Fig. 5). Mutation of the HSL E-box induced a 50% decrease in luciferase activity. Mutation of the HSL 5′ GC-box had a strong negative effect, whereas mutation of the HSL 3′ GC-box had no effect. Combined mutation of the two GC-boxes led to lower luciferase activity than single mutations. Finally, mutations of the three boxes totally abolished luciferase activity.

Because of the importance of the HSL E-box in promoter activity and its in vitro binding to USF, we wished to determine the transactivation properties of USF1 and USF2 on the HSL promoter. Cotransfection of the −137/+38 construct with USF1 and USF2 expression vectors in HepG2 cells resulted in a marked increase in luciferase activity (Fig. 6). Similar results were obtained in COS7 cells (data not shown). USF1 and USF2 overexpression showed reduced effect on the −86/+38 construct, which does not contain the E-box. We also studied the effect of ADD1/SREBP1c, a basic helix-loop-helix leucine zipper transcription factor that has been proposed to play a role in the activation of hepatic genes by insulin and glucose (31,32). ADD1-403 is a mutant that lacks the membrane-attachment domain and, therefore, enters the nucleus directly and shows enhanced transcriptional activity (33). ADD1/SREBP1c (data not shown) and ADD1−403 expression had no effect on the −137/+38 and −86/+38 construct activity. This is consistent with the lack of binding of ADD1/SREBP1c observed on the double E-box motif of the human uncoupling protein-2 promoter (34).

To further assess the role of USF1 and USF2 in adipocyte HSL gene expression, HSL mRNA levels were determined by Northern blot analysis of adipose tissue from USF1- and USF2-deficient mice (23,24). The amount of HSL mRNA was reduced to 49% in USF1-deficient mice and to 14% in USF2-deficient mice compared with wild-type mice (Fig. 7).

To define the DNA elements involved in the glucose response, the mutant constructs were stably transfected into 3T3-F442A preadipocytes. Transfectant pools were subjected to glucose deprivation for 24 h after 10 days of adipocyte differentiation (Fig. 8). Mutations of the GC-boxes had no effect on glucose-induced luciferase activity. Mutation of the E-box resulted in a complete loss of the glucose response. In accordance with this observation, simultaneous mutations of the GC-boxes and the E-box led to an absence of glucose response. Mutation of the half E-box CACTGA (Fig. 2) did not alter the glucose response (data not shown). Therefore, the glucose response element could be attributed to the E-box in the human adipocyte HSL promoter.

To determine the metabolic pathway(s) involved in the glucose response, we tested several glucose analogs and metabolites (Fig. 9). Compounds were added for 24 h after a 24-h period in 1 mmol/l glucose medium. Supplementation of the medium with 25 mmol/l glucose restored luciferase activity. Mannitol, which is not transported into cells, did not mimic glucose effect. Hence, the effect of glucose deprivation was not due to difference in osmotic pressure. 3-O-methylglucose, which is transported inside the adipocyte but not phosphorylated, had no effect. A lack of effect was also observed for 2-deoxyglucose, which is transported and phosphorylated into 2-deoxyglucose-6-phosphate but not further metabolized. Therefore, glucose-6-phosphate is not the metabolite mediating the glucose response. Fructose and mannose, which enter glycolysis at the glucose-6-phosphate level, partially restored luciferase activity. Dihydroxyacetone, which is converted into dihydroxyacetone phosphate, had no effect. Similarly, xylitol, an intermediate of the nonoxidative branch of the pentose phosphate pathway, and glucosamine, which enters the hexosamine pathway, could not substitute for glucose. Increased glucosamine and xylitol concentrations up to 20 mmol/l did not modify the promoter activity (data not shown).

Glucose-mediated regulation of gene expression has been reported for a growing set of genes involved in metabolic and energetic pathways in the cell (35). We have recently shown that HSL, the gene encoding the enzyme catalyzing the breakdown of triacylglycerol into fatty acids in adipose tissue, is positively regulated by glucose in a mouse adipocyte cell line (13). Here, we show that HSL mRNA is also responsive to glucose in primary culture of human subcutaneous adipocytes. Because the molecular mechanisms underlying the glucose response are largely unknown in the adipocytes, we sought to determine the cis-acting elements in the human adipocyte HSL promoter mediating upregulation by glucose.

Determination of reporter gene activity in differentiated 3T3-F442A adipocytes showed that the proximal 137 bp of the HSL promoter contained a glucose-responsive region. Interestingly, this glucose-responsive region contains two functional GC-boxes binding the Sp family of nuclear factors and an E-box binding USF. Sp1 has been implicated in the glucose response of the acetyl CoA carboxylase and leptin genes in adipocytes (27,36). It was proposed that glucose-induced Sp1 dephosphorylation enhanced Sp1 binding and transcriptional activity of the acetyl CoA carboxylase promoter II (37). Glucose was also shown to promote glucosamine-mediated O-linked glycosylation of Sp1, which protects the transcription factor from proteolytic degradation (38). The glucosamine pathway is involved in the regulation of the leptin gene (39). However, our data do not support a role for Sp1 or Sp3 in the glucose response of the HSL promoter. Indeed, individual and double mutations of the GC-boxes did not alter the glucose response of several pools of stably transfected 3T3-F442A adipocyte clones. Moreover, in our experimental conditions, no difference in Sp1 and Sp3 DNA binding activity was observed with nuclear extracts prepared from either glucose-deprived or glucose-fed adipocytes (data not shown). However, we showed that the GC-boxes are necessary for full HSL promoter activity in adipocytes.

A typical E-box motif, CACGTG, was located between −106 and −101 in the HSL promoter. USF1/USF2 heterodimers from adipocyte nuclear extracts bound to the E-box. Moreover, overexpression of the transcription factors increased the −137/+38 construct-mediated luciferase activity, and mutation of the E-box lowered promoter activity in rat adipocytes. The importance of USF was also observed in vivo. Indeed, HSL gene expression was decreased in adipose tissue of USF1- and USF2-deficient mice compared with wild-type mice. Together, these data demonstrate that USF1 and USF2 could participate in the transactivation of the HSL gene in adipose tissue. Mutation of the HSL E-box abolished the glucose response in stably transfected 3T3-F442A adipocytes. E-boxes have been identified as parts of glucose response elements for several hepatic genes (26). In USF1 and USF2 knockout mice (23–25), as well as in USF2-deficient hepatocytes in primary culture (S.V., unpublished results), endogenous USF was shown to be important for carbohydrate induction of hepatic genes. However, as previously reported (26), these transcription factors cannot by themselves explain the transcriptional regulation by glucose. USF binding to the HSL E-box did not differ between nuclear extracts from adipose cells grown in the presence or absence of glucose (data not shown). Our data show that the HSL E-box is an essential part of the glucose response element. Whereas USF may be indirectly involved in the transcriptional activation by glucose, the data indicate that other factors are part of the glucose sensor complex. A good candidate is the recently identified carbohydrate response element binding protein, which is a member of the basic helix-loop-helix leucine zipper family of transcription factors expressed in liver (40). It is not known whether the factor is expressed in adipocytes and plays the same function as in liver.

Several metabolic pathways have been proposed for glucose-mediated transcriptional activation (35). We therefore tested the effect of glucose analogs and metabolites on the activity of the −137/+38 construct. In adipose tissue, hexokinase-dependent metabolism of glucose into glucose-6-phosphate is required for the glucose-mediated induction of fatty acid synthase and acetyl CoA carboxylase gene expression (41). The lack of effect of 3-O-methyl glucose and 2-deoxyglucose on HSL promoter activity suggests that a metabolite downstream to glucose-6-phosphate in the glycolytic pathway was involved. However, dihydroxyacetone, which has been shown to increase hepatic glucagon receptor mRNA expression (42), did not restore HSL promoter activity. It has been proposed that glucose acts in liver though the nonoxidative branch of the pentose phosphate pathway to induce the l-pyruvate kinase gene expression (43). Xylitol, which is converted into xylulose 5-phosphate, was unsuccessful in restoring HSL promoter activity. The hexosamine pathway may also mediate the effect of glucose on gene expression, as proposed for leptin expression in skeletal muscle and adipose tissue (39). The lack of glucosamine effect excluded this pathway and its end product UDP-N-acetylglucosamine in HSL transcriptional activation. Together, the data suggest that a metabolite generated during glycolysis between glucose-6-phosphate and triose phosphate could regulate the transcription of the human HSL gene. The fructose and mannose stimulatory effects are in agreement with this scenario, since they enter glycolysis at the glucose-6-phosphate level. Hence, the metabolic steps involved in the regulation of HSL promoter activity define a novel mechanism for the transcriptional effect of glucose.

In conclusion, we have shown that an E-box and two GC-boxes are critical cis-acting elements of the human adipocyte HSL promoter. USF, through binding to the E-box, may participate in the promoter transactivation. The E-box also mediates the glucose-mediated induction of HSL gene expression. Metabolism of glucose at a step between glucose-6-phosphate and triose phosphates is required. Further work is needed to identify the metabolites and the transcription factor complex involved in the glucose response.

This work was supported by grants from Fondation pour la Recherche Medicale (to P.R.) and Produits Roche France (to D.L.). F.S. was supported by Groupe Danone.

We are indebted to J.P. Salier (INSERM U 519, Rouen, France) and P.J. Romanienko (National Institutes of Health, Bethesda, MD) for critical reading of the manuscript. We also thank Corinne Lefort and Lydia Pinard for excellent technical assistance.