The Mouse obese Gene GENOMIC ORGANIZATION, PROMOTER ACTIVITY, AND ACTIVATION BY CCAAT/ENHANCER-BINDING PROTEIN (cid:97) *

The obese gene product, leptin, regulates adiposity. Mice homozygous for a nonfunctional obese gene be-come massively obese and develop diabetes mellitus due to overeating and increased metabolic efficiency. The cDNA sequence of obese was recently reported (Zhang, Y., Proenca, R., Maffei, M., Barone, M., Leopold, L., and Friedman, J. L. (1994) Nature 372, 425–432; Correction: (1995 Nature 374, 479). We have determined the genomic organization of the 5 (cid:42) end of the mouse obese gene. The coding sequence is in exons 2 and 3. A single TATA- containing promoter was found upstream of exon 1. A minority (probably (cid:59) 5%) of the obese mRNA contained an extra, untranslated exon between exons 1 and 2. Transcription of the obese gene was detected only in adipose cells. A 762-base pair obese gene promoter driv- ing a luciferase gene yielded abundant activity in transiently transfected rat adipose cells in primary culture. The obese promoter was inactive in erythroid K562 cells. cells. The (cid:50) and CCAAT/enhancer-binding protein motifs. transcription in cell 23-fold acti- data the obese is a natural target of C/EBP 2 a 111-bp Product specificity confirmed digestion with restriction enzymes Luciferase reporter constructs on a modified pGL2-basic plasmid (Promega) which the ATG at the start of luciferase translation was changed to an Nco I site and the Bam HI enhancer insertion site was converted to a Spe I site (via cleavage with Bam HI, filling with Klenow, and insertion of a 5 (cid:57) -GGACTAGTCC link- er). p( (cid:49) 2)ob-luc (p1345) was made by insertion of the sequence 5 (cid:57) -aAGCTTGGATCCCTGCTCCAGCAGCTGCAAGGTGCAAGAAGAAG-AAGATCCCAGGGAGGACcatgg-3 (cid:57) between the Hin dIII and Nco I sites. This construct contains the complete 5 (cid:57) -untranslated region of the obese gene (from the Bam HI site), with 2 base changes that create an Nco I site at the initiator ATG. p( (cid:50) 762)ob-luc (p1494) was made by insertion of the 764-bp Bam HI fragment into p1345. p( (cid:50) 762rev)ob-luc (p1500) contains the Bam HI fragment in the reverse orientation. p( (cid:50) 456)ob-luc

Obesity is common in Western society, and the underlying molecular mechanisms are not well understood. Body weight is almost certainly regulated by a feedback control mechanism (see Ref. 1). Classic experiments with parabiotic rats showed that overeating by one animal caused its mate to starve (2). When animals subjected to forced over-or under-feeding were returned to ad libitum feeding, they adjusted their food intake appropriately to reach the same weight as control animals (3). If some adipose tissue is removed from a growing, chow-fed rat, the remaining adipose tissue enlarges, so that the rat attains the same total amount of body fat as an unoperated control (4). A recent study in humans demonstrated that upon weight gain or loss, the body's energy expenditure increases or decreases, respectively, suggesting an attempt to return to the original state (5). Taken together, these data suggest that individuals have a set point for body weight/adiposity and that a feedback control mechanism maintains the target weight.
In humans, obesity has been shown to have a large genetic component (6). However, little is known about the specific genes that are responsible. Insight into the regulation of obesity has come from the study of the ob/ob mouse (7). These mice grow massively obese and develop diabetes mellitus due to overeating and increased metabolic efficiency. Parabiotic animal experiments suggest that ob/ob animals are unable to make a satiety factor, but can respond to such a factor from a parabiotic mate. Similar experiments suggest that db/db mice make the factor missing in ob/ob mice, but cannot respond to it. Recently, Friedman's laboratory used positional cloning to isolate the obese gene encoding the leptin protein (8). The obese RNA is expressed selectively in adipose tissue. Mature, secreted leptin is 146 amino acids long and is not similar in sequence to any known protein. Treatment of ob/ob mice with leptin caused reversal of the obese phenotype (9 -11). In addition, leptin treatment caused slight weight loss in wild type mice (9 -11).
Little is known about the regulation of the obese gene. The ob C57Bl /ob C57Bl mice have a R105term nonsense mutation and a 20-fold increased obese RNA level (8). This suggests that these mice have an intact mechanism to sense adiposity and to transcribe the obese gene, but do not make functional leptin. Thus, obese is probably subject to regulation at the level of transcription and/or RNA stability, and comprehension of the regulation of obese will increase our knowledge of the adiposity sensor. As a step toward understanding the regulation of the obese gene, we have cloned and sequenced the wild-type obese promoter and characterized its activity in transient expression assays in primary cultures of rat adipose cells.
Cloning-P1 clones were obtained from Genome Systems (St. Louis, MO; clone addresses 42 and 233) using PCR 1 primers x248 and x249, which yield a 141-bp product. Clone 42 was shotgun subcloned using BamHI, EcoRI, or HindIII into pBluescriptII SKϪ (Stratagene) and subclones were selected by hybridization with oligonucleotide probes.
A mouse epididymal fat pad cDNA library (Clontech, ML3005b) was * This work was supported in part by a research award grant from the American Diabetes Association (to M. J. Q.) and by a grant from the Lucille P. Markey Charitable Trust (to M. R.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM  screened to obtain the 5Ј end of the obese mRNA. PCR was first carried out (0.5-1 l of library in 50 l, using 1.5 mM MgCl 2 , 0.4 M primers, 94°C, 4 min followed by 25 cycles of 94°C, 2 min, 61°C, 1 min, 72°C, 1 min) with primers for the obese coding region (x251) and the left (x225) or right (x226) arms. Nested PCR on 0.5 l used primers x249 (obese coding) and x200 (left ) or x201 (right ), 30 cycles, and 55°C annealing, otherwise as above. The PCR products were cloned (PCRscript, Stratagene), and the clones containing the longest inserts were sequenced using an Applied Biosystems, Inc. model 373 DNA sequencer with the fluorescent primer or dideoxy kits (Applied Biosystems). Sequence searches used BLASTN on the NCBI server (12) and FASTA (13).
Reporter Constructs-Luciferase reporter constructs were based on a modified pGL2-basic plasmid (Promega) in which the ATG at the start of luciferase translation was changed to an NcoI site and the BamHI enhancer insertion site was converted to a SpeI site (via cleavage with BamHI, filling with Klenow, and insertion of a 5Ј-GGACTAGTCC linker). p(ϩ2)ob-luc (p1345) was made by insertion of the sequence 5Ј-aAGCTTGGATCCCTGCTCCAGCAGCTGCAAGGTGCAAGAAGAAG-AAGATCCCAGGGAGGACcatgg-3Ј between the HindIII and NcoI sites. This construct contains the complete 5Ј-untranslated region of the obese gene (from the BamHI site), with 2 base changes that create an NcoI site at the initiator ATG. p(Ϫ762)ob-luc (p1494) was made by insertion of the 764-bp BamHI fragment into p1345. p(Ϫ762rev)ob-luc (p1500) contains the BamHI fragment in the reverse orientation. p(Ϫ456)ob-luc (p1505) and p(Ϫ161)ob-luc (p1508) were made by digestion of p(Ϫ762)ob-luc with KpnI and XhoI, respectively, followed by religation.
pob-luc-RSV (p1518) was made by insertion into SpeI-cut p(Ϫ762)obluc of the 479-bp HgiAI/EcoRI fragment of the RSV long terminal repeat, which contains enhancer but no promoter activity (the fragment used contained multilinker-derived restriction sites at either end and was inserted as a XbaI fragment). A clone with the EcoRI end of the enhancer away from the luciferase gene was chosen.
Transient Expression-Transient expression in primary rat adipose cells was performed by electroporation as described (15). Transfections usually contained 6 g of DNA: 5 g of a luciferase plasmid and 1 g of RSV-cat (16) per cuvette. Each transfection was performed in triplicate. After incubation (ϳ20 h; sometimes ϳ44 h also), cells were transferred to a 1.5-ml tube and allowed to float at 1 ϫ g. The infranatant was discarded, 150 l of Lysis buffer (Promega) was added, incubated at room temperature for 10 min with periodic vortexing. Chloramphenicol acetyltransferase (17) and luciferase (Promega Luciferase Assay System) activities were assayed using 50 and 16 l, respectively. Transient expression in K562 cells was performed as described elsewhere. 2 Briefly, 20 g of plasmid DNA (5 g of luciferase plasmid, 3 g of RSV-␤-galactosidase, and 12 g of pUC18 carrier) were electroporated (in 400 l of Dulbecco's phosphate-buffered saline, 450 V, 500 microfarads, resulting ϭ 7.7 ms) into 5 ϫ 10 6 cells, and cultured for 46 -48 h before assay. ␤-Galactosidase assays were performed as described (14).

RESULTS AND DISCUSSION
Genomic Organization of obese-As a first step toward understanding the transcriptional regulation of the obese gene, we defined its genomic organization. Using the reported coding sequence, a genomic P1 clone was obtained, mapped, subcloned, and partially sequenced (Fig. 1). The 5Ј end of the mRNA was isolated using nested PCR on a mouse epididymal fat pad cDNA library (see "Materials and Methods"). Comparison of the cDNA and genomic sequences revealed the genomic organization shown in Fig. 1. The 5Ј ends of five independent cDNA clones matched the published obese cDNA sequence and 2 O. Gavrilova and M. Reitman, manuscript in preparation.

FIG. 2.
A, primer extension to map obese transcription start sites. Primer extension was performed on 10 g of RNA using avian myeloblastosis virus reverse transcriptase and primer x249, and then subjected to denaturing gel electrophoresis. RNA was from peritoneal fat, brain, testes, liver, kidney, or adrenal of FVB/N mice, from peritoneal fat of ob/ob mice, or yeast tRNA, as indicated. The size, in nucleotides, of MspI-digested pBR322 marker is at the left. B, RT-PCR showing tissue specificity and alternate exon use. RT-PCR was performed using RNAs from FVB/N mice and the products were electrophoresed under denaturing conditions (see "Materials and Methods"). The left panel shows the results of 30 amplification cycles using the exon 1/alternate exon primers, the center panel contains the products of 24 amplification cycles using the exon 1/exon 2 primers, and the right panel shows the results from 30 cycles using the alternate exon/exon 2 primer set. Below each set of lanes is a diagram showing the primers and expected amplification product sizes from mRNAs without or with the alternate exon. Exon 1 is solid, exon 2 is hatched, and the alternate exon is open. Controls include no RNA or tRNA in the reverse transcription reaction and RNA from adipose tissue without reverse transcriptase (Fat, no RT). Fat-1 and Fat-2 are independently isolated RNA samples. PCR products of the predicted sizes were detected only in RNA isolated from fat. Product specificity was confirmed by restriction enzyme digestion as detailed under "Materials and Methods." contained 5, 9, 12, 20, and 21 bp of the 3Ј end of exon 1. The first exon was located ϳ7.5 kb upstream of the 175-bp exon 2. Intron 2 was ϳ1.7 kb long, in agreement with the recently deposited 1731-bp sequence (GenBank accession number U22421). Exon 3 was at least 2.5 kb in size; its 3Ј end was not mapped.
Evidence for Alternate Splicing-One clone of the 5Ј end of the mRNA had the last five bases of the first exon, followed by 93 bp of unrelated sequence and the second exon. These data suggested the existence of an alternate exon. This was confirmed by identification of the 93-bp sequence, flanked by splice acceptor and donor sites, in genomic DNA ϳ4 kb downstream of the first exon.
To further characterize the obese mRNAs, primer extension using an exon 2 primer was performed ( Fig. 2A). In RNA from ob/ob adipose tissue, a single strong band was observed at ϳ201 nucleotides, corresponding to a first exon size of ϳ26 bp. A weak band (ϳ20-fold less signal, but clearly visible on the original) was observed at ϳ290 nucleotides, the size expected for mRNAs containing the alternate exon. With RNA from control FVB/N mice (18), the 201-nucleotide band was also present in adipose tissue (at a much lower level, ϳ20-fold weaker than in ob/ob mice). It is likely that a faint band at ϳ147 nucleotides results from incomplete extension by the reverse transcriptase, although we cannot formally rule out another minor RNA specie.
We used a more sensitive RT-PCR assay for further characterization of the splicing products. RT-PCR with an exon 1/exon 2 primer set yielded the 195-bp product expected for direct splicing of exons 1 and 2 (Fig. 2B, center panel). Under the conditions used, the 288-bp product expected from RNAs containing the alternate exon was not observed. However, PCR reactions specific for splicing of exon 1 to the alternate exon and of the alternate exon to exon 2 detected the expected 111and 270-bp products, respectively (Fig. 2B, left and right  panels). Both types of obese mRNA were detected only in adipose tissue, and not in the brain, testes, liver, kidney, or adrenal samples. While the PCR assay was not strictly quantitative, the alternate exon products appeared less abundant, requiring six extra cycles to be amplified to a level comparable to the product without this exon. Thus, the alternative exon is also limited to adipose tissue and is present in a minor fraction of the obese mRNAs.
The alternate exon is the second example of alternate splicing in the obese gene, the first being the variable inclusion of the first codon in exon 3 (8). Since the alternate exon sequence (Fig. 3) does not contain an ATG, inclusion of this exon does not change the predicted protein product which is initiated downstream in exon 2.
Promoter Sequence-The DNA sequence of the promoter is shown in Fig. 3. With the exception of a simple sequence repeat upstream of the promoter, none of the new DNA sequence showed statistically significant similarity to those already in GenBank. As predicted by the single strong start site, the promoter contained a TATA motif at Ϫ29 to Ϫ34. A match to the Sp1 consensus sequence (GGGCGG; Ref. 19) occurs at Ϫ95 to Ϫ100. A number of transcription factors of the Sp1-like zinc-finger family may bind to this motif. Between Ϫ49 and Ϫ58 is a short palindrome that is predicted to bind C/EBP (20).
Promoter Function in Adipose Cells and Erythroid Cells-Promoter function was tested using transient expression in primary rat adipose cells that express obese mRNA (data not shown; Ref. 21). The p(Ϫ762)ob-luc construct, containing 762 bp of promoter sequence produced reporter activity ϳ100-fold greater than the assay background (Table I). In contrast, the p(Ϫ762rev)ob-luc plasmid, with the promoter region in the inverted orientation, showed only background activity, the same as the promoterless p(ϩ2)ob-luc and pGL2-basic constructs. Inclusion of the RSV enhancer in the p(Ϫ762)ob-luc reporter plasmid increased transcription 15-fold, demonstrating that the obese promoter can interact with this enhancer.
The obese gene mRNA has been detected only in adipose tissues. We examined the tissue specificity of the obese promoter using transient expression in the erythroid K562 cell line (22). To allow comparison of the adipose cell and K562 data, the results were expressed relative to pCIS-luciferase, a highly expressed plasmid that is active in both types of cells (Table I). In K562 cells, the promoterless reporter plasmids expressed luciferase at ϳ0.01% of the level of pCIS-luciferase. Addition of the obese promoter did not increase luciferase above this background. The level of expression of p(Ϫ762)ob-luc, after normalization to pCIS-luciferase, was 162-fold greater in adipose cells than in K562 cells. Thus, there is sufficient information contained in the 762-bp promoter to allow expression in adipose cells, but not in K562 cells.
To map its functional regions, various lengths of the obese promoter were tested for their ability to drive the expression of a luciferase reporter gene (Fig. 4A). Deletion of regions upstream of Ϫ161 did not reduce the promoter activity. We conclude that the region up to Ϫ161 functions as a promoter in adipose cells and that addition of another 600 bp of upstream DNA did not increase the promoter activity.
Transactivation by C/EBP␣-The putative C/EBP␣ binding site in the proximal promoter is intriguing since C/EBP␣ promotes adipocyte differentiation (23,24) and transactivates the promoters of many adipose-specific genes (25)(26)(27). Coexpression of a CMV-driven C/EBP␣ expression plasmid caused a 23-fold increase in obese reporter expression, with no change in expression of the cotransfected RSV-cat plasmid (Fig. 4B).
These data demonstrate that C/EBP␣ can activate the promoter of the obese gene. Point mutations causing loss of C/EBP transactivation are needed for definitive identification of the C/EBP-responsive cis-element(s).
A number of adipose and hepatic genes are known to be regulated in response to hormones and metabolic state (28 -30). For example, insulin and/or high glucose stimulate transcription of the genes for glyceraldehyde-3-phosphate dehydrogenase (31), fatty acid synthase (32), S 14 (33), stearoyl-CoA desaturase I (34), and pyruvate kinase and decreases the transcription of the phosphoenolpyruvate carboxykinase (35). Recent evidence suggests that obese people have slightly elevated obese mRNA levels (36). The transcription factor binding motifs in the obese promoter, such as the putative C/EBP site, may play a role in mediating this regulation.

TABLE I Reporter activity in rat adipose and K562 cells
The indicated reporter plasmids were transiently expressed in primary rat adipose cells and K562 cells as described under "Materials and Methods." Results are luciferase activity corrected for transfection efficiency (using RSV-CAT for the adipose cells and RSV-␤-galactosidase for K562 cells). The expression of all plasmids was normalized to pCIS-luciferase (31) at 1,000,000 units. Results Ͻ200 for adipose cell and Ͻ20 for K562 assays are not clearly greater than the luminometer background. Thus, the luciferase activity of pGL2-basic, p(ϩ2)ob-luc, and p(Ϫ762rev)obluc cannot be distinguished from background. Each independent experiment was performed in triplicate.  4. Transient expression in rat adipose cells. A, the indicated reporter plasmids were transiently expressed in primary rat adipose cells as described under "Materials and Methods." Reporter constructs contained obese (f) and luciferase (Ⅺ) sequences as indicated and are shown to scale except for the length of the luciferase. Results are luciferase activity corrected for transfection efficiency and normalized to p(Ϫ762)ob-luc ϭ 100 and are the mean Ϯ standard error (number of independent experiments). Results Ͻ1 are not clearly greater than the luminometer background. B, effect of a C/EBP␣ expression vector on obese expression. Transfections contained 2 g of p(Ϫ762)ob-luc, 1 g of RSV-cat, the indicated amount of FlagCMV80-C/EBP␣ expression vector, and pUC18 filler to give a total of 8 g of plasmid DNA. The raw RSV-cat data (q) and the p(Ϫ762)ob-luc data corrected for chloramphenicol acetyltransferase expression (E) are expressed relative to no C/EBP␣ vector. Data are the mean of two independent experiments, except for the 4 g point which is the mean Ϯ standard error from five independent experiments.
Given the postulated role of the obese gene as an adiposity sensor, transcription of the obese gene may be sensitive to lipid status. For example, polyunsaturated fatty acids decrease transcription of hepatic pyruvate kinase (37) and fatty acid synthase (38). With varying precision and certainty, candidate cis regulatory elements and their cognate trans factors have been identified as mediators of these events. Two different direct mechanisms are known to mediate the transcriptional response to intracellular hydrophobic ligands. The sterol regulatory element-binding proteins remain membrane-bound under sterol-replete conditions but are proteolytically cleaved to release an active transcription factor under low sterol conditions (39). Ligand binding to steroid hormone superfamily receptors causes the receptors to bind DNA and activate transcription. In particular, peroxisome proliferator-activated receptor ␥ is known to increase transcription in response to linoleic acid, clofibric acid, and thiazolidinediones (40,41) and has been proposed to regulate obese (40). The mouse obese promoter does not contain sequences that are identical to the reported sterol regulatory element or peroxisome proliferator-activated receptor regulatory element. However, obese may be controlled indirectly by these factors, or directly via distant elements and/or elements that do not conform to the classical sequence motifs. Elucidation of the regulatory mechanisms controlling expression of the obese gene and of the promoter elements that confer adipose specific expression will be important for understanding the regulation of body fat in the normal state and the pathogenesis of obesity.