Essential cis-acting elements in rat uncoupling protein gene are in an enhancer containing a complex retinoic acid response domain.

Transgenic mice were generated with a transgene containing the 211-base pair (bp) enhancer and 0.4 kilobase pairs of 5′-flanking DNA of the uncoupling protein (ucp) gene. Expression of this transgene was restricted to brown adipose tissue and was inducible by cold exposure or treatment of transgenic mice by norepinephrine, retinoic acid (RA), or CL-316,243 β3-adrenoreceptor agonist. A search for retinoic acid response elements in the ucp gene enhancer was undertaken using mutagenesis and transfection of cultured cells with chloramphenicol acetyltransferase constructs. Deletion or mutations of several putative retinoic acid response elements were ineffective. Mutations of a TGAATCA region dramatically decreased the transcriptional activity in the presence of RA. In vitro this region was able to bind a complex containing proteins recognized by antibodies against Jun or Fos. Mutations of an adjacent region related to an inverted repeat of type 2 also markedly decreased RA effect. This region was able to bind in vitro retinoid X receptor α and retinoic acid receptor β. The two regions form an activating region between bp −2421 and −2402 (referred to as the ucp gene-activating region), which has an enhancer activity but cannot confer RA response to a promoter. This response was obtained with a larger DNA fragment (bp −2489 to −2398) constituting a complex RA response domain.

Transgenic mice were generated with a transgene containing the 211-base pair (bp) enhancer and 0.4 kilobase pairs of 5-flanking DNA of the uncoupling protein (ucp) gene. Expression of this transgene was restricted to brown adipose tissue and was inducible by cold exposure or treatment of transgenic mice by norepinephrine, retinoic acid (RA), or CL-316,243 ␤3-adrenoreceptor agonist. A search for retinoic acid response elements in the ucp gene enhancer was undertaken using mutagenesis and transfection of cultured cells with chloramphenicol acetyltransferase constructs. Deletion or mutations of several putative retinoic acid response elements were ineffective. Mutations of a TGAATCA region dramatically decreased the transcriptional activity in the presence of RA. In vitro this region was able to bind a complex containing proteins recognized by antibodies against Jun or Fos. Mutations of an adjacent region related to an inverted repeat of type 2 also markedly decreased RA effect. This region was able to bind in vitro retinoid X receptor ␣ and retinoic acid receptor ␤. The two regions form an activating region between bp ؊2421 and ؊2402 (referred to as the ucp gene-activating region), which has an enhancer activity but cannot confer RA response to a promoter. This response was obtained with a larger DNA fragment (bp ؊2489 to ؊2398) constituting a complex RA response domain.
Although the importance of this enhancer in the brown adipose tissue (BAT) specificity is debated, it is clear that it plays a significant role in transcriptional regulation of the ucp gene (16,17). An analysis of the rat ucp enhancer based on DNase footprint analysis and electrophoretic mobility shift assays, identified two footprints and revealed that retinoid X receptor (RXR) and thyroid hormone receptors (TRs) were putative transactivators of the ucp gene (14). More precisely, RXR binding was proposed to be located between bp Ϫ2348 and Ϫ2334 (14), a sequence related to a direct type 3 repeat (DR3 repeat), which represents a potential binding site of nuclear receptors (19,20). In the course of the present work, Alvarez et al. (15) reported that a large deletion of the rat enhancer abolished the RA activation of transcription of a chloramphenicol acetyltransferase (CAT) construct, and mutagenesis experiments allowed Silva and colleagues (18) to prove that the DR3 element participates in the activation of ucp gene transcription by thyroid hormones.
The present work, based on studies of transgenic mice and in vitro analysis of the effect of mutations in the rat ucp enhancer, was undertaken in order to investigate the importance of the enhancer and to test the significance of in vitro interaction analysis of RA effect (14). A minigene bearing the 211-bp enhancer was shown to be able to drive a specific and regulated expression of a reporter gene in mouse BAT. Deletion and site-directed mutagenesis of CAT-reporter gene constructs tested in cultured cells demonstrated that the DR3 repeat mentioned above does not mediate the transcriptional activation of ucp by RA. The search for other cis-acting elements mediating RA activation led to the discovery that mutations crippling the response to RA cluster in a region located at the 3Ј-boundary of the FP1 footprint. This 20-bp region, referred to as the ucp gene-activating region (UAR) is made of a short domain resembling the AP-1 binding site, attached to an inverted repeat of type 2 (IR2). UAR has an enhancer activity, but its response to RA requires other element(s) located at the 5Ј-end of the enhancer.
Cell Culture, Western Analysis, RNA Analysis, and Cellular Transfections-Primary cultures of rat brown adipocytes were carried out as described previously (10). Western analysis of ucp in mitochondria from cultured brown adipocytes was performed as described (9). An immortalized cell line, termed 1B8, was cloned from the same tumor used to derive the HIB 1B cell line (21,22) and was used in the present work. In the absence of cAMP, adrenergic agonist, or RA, this cell line does not express ucp mRNA (as estimated from Northern analysis of total RNA), but it expresses the BAT-specific mitochondrial ucp in the presence of one of these agents (14). 1B8 cells were grown and differentiated as described for HIB 1B cells (22). Northern analysis of ucp mRNA in primary cultures and in 1B8 cells was carried out as described previously (9,14). CHO K1 cells were grown in Ham's F-12 medium supplemented with 10 mM glutamine and 10% fetal calf serum. Transfection of CHO and 1B8 cells was performed in 10-cm diameter plates using the calcium phosphate precipitation method as described previously (16). The medium was changed 20 h prior to transfection. In all transfection experiments, 2 g of plasmid expressing ␤-galactosidase were included. 2 g of pSV2CAT or 10 -20 g of ucp-CAT DNA was used. When indicated, the transfection mixture contained 1 g of pSG5-RXR␣ or pSG5-retinoic acid receptor (RAR) ␣ expression vectors. The total amount of DNA was brought to the same value by the addition of salmon DNA. Prior to transfection, 1B8 cells of two plates were scraped, mixed with medium containing the DNA precipitate, and then reinoculated into three plates. After 4 h of incubation, cells were shocked with 5 ml of medium containing 15% glycerol for 2 min, washed twice with 10 ml of PBS, and fed with 10 ml of fresh medium containing 5% charcoaltreated fetal calf serum with or without RA or norepinephrine. 20 h later, the cells were rinsed and harvested in phosphate-buffered saline.
Lysate preparations and the ␤-galactosidase assay were carried out as described previously (16); CAT activity was determined as described earlier (23) and normalized to the internal control ␤-galactosidase activity. Actually, when using pSV-␤-galactosidase expression vector, cotransfection of RXR or RAR expression vectors severely repressed the expression of ␤-galactosidase in CHO cells, as already noticed by Kliewer et al. (24); therefore, we used a pRSV-␤-galactosidase expression vector that was not inhibited by RXR or RAR expression vectors. Each transfection was performed a minimum of 3 times, and at least two different preparations of each DNA construct were used.
Generation of Transgenic Mice Harboring the ucp-CAT Transgene Construct-Construction of the Ϫ400-UCP-AA ϩ CAT plasmid containing the 211-bp AatII-ApaI (enhancer) fragment attached to the first 400 bp located upstream of the transcriptional start site has been described elsewhere (16). The fragment containing the 211-bp enhancer, 400 bp of the promoter, and the CAT sequence was separated from vector sequences by digestion with BglII and XhoI followed by preparative electrophoresis through agarose gel and electroelution. All microinjection and oviduct transfer procedures were carried out as described (25). Fertilized eggs were recovered from mating between B6D2F1 hybrids, microinjected with approximately 500 copies of DNA, and transferred to oviducts of C57 BL6 x CBA-F1 pseudopregnant females. The presence of the transgene in founder animals was checked by Southern blot analysis of tail DNA (26).
Plasmids, Deletions, and Mutagenesis-The parent 4551-CAT plasmid contains the 4551 bp of the 5Ј-flanking region (GenBank TM number X12925) plus the first 110 bp of the transcription unit of the rat ucp gene inserted in the EcoRI and KpnI sites of plasmid pSP73; then the CAT gene was added using KpnI and PvuII sites (16). The D0-CAT construct was made free of the Ϫ4451/ϩ110 sequence by digestion of the parent 4551-CAT plasmid with KpnI and BglII. The pUCP-AA ϩ CAT plasmid contains the minimal promoter of the rat ucp gene (bp Ϫ157 to ϩ110) fused to the 211-bp AatII-ApaI enhancer (bp Ϫ2494 to Ϫ2283) of the rat ucp gene (16); the same enhancer had also been cloned into the unique HindIII restriction site of the pTK CAT plasmid in which CAT expression is under the control of the Herpes simplex virus thymidine kinase (TK) promoter to generate pTK-AA ϩ (16). Deletion of a fragment containing a large part of the enhancer (bp Ϫ2469 to Ϫ2283) from the 4551-CAT construct was made by digestion with BclI and ApaI and religation; this plasmid was named 4551 B/A del. The CAT constructs were subjected to deletion or site-directed mutagenesis using proce-FIG. 1. Tissue-specific expression and regulation of the ؊400-AA ؉ -CAT transgene. A, schematic drawing of the construct that was used to generate transgenic mice; the Ϫ400-AA ϩ -CAT transgene was made of the 211-bp enhancer of the rat ucp gene (bp Ϫ2494 to Ϫ2283) attached to the proximal region of the ucp promoter (bp Ϫ400 to ϩ111) in front of the CAT gene. B, CAT activity in brown adipose tissue (Bat), heart (H), liver (L), white adipose tissue (W), or brain (Br) of Ϫ400-AA ϩ -CAT transgenic mice (line 1) kept at 25°C or exposed to 5°C for 16 h. C, CAT activity in brown adipose tissue (Bat), liver (L), or white adipose tissue (W) of another line of Ϫ400-AA ϩ -CAT transgenic mice kept at 25°C, exposed to 5°C for 18 h, or treated with norepinephrine (NE) or CL-316,243 (samples were collected 18 h after drug injection). D, CAT activity in brown adipose tissue (Bat), liver (L), or white adipose tissue (W) of Ϫ400-AA ϩ -CAT transgenic mice injected intraperitoneally with all-trans-retinoic acid (1.6 g, RA). Mice belonging to line 1 were used; samples were collected 16 h after drug injection.
dures based on the use of appropriate selection and mutagenic primers, essentially as described by Deng and Nickoloff (27) with the following modifications: single-strand binding protein was added after denaturation of the plasmid and annealing of primers in order to stabilize the single strand plasmid molecules, and T7 DNA polymerase was used instead of T4 DNA polymerase. Details on the different constructs are given in here. All constructs were checked by DNA sequencing using the Prism TM cyclic sequencing kits and an ABI 373 DNA sequencer.
Nuclear Extracts and Electrophoretic Mobility Shift Assays-Purification of nuclear proteins and DNA binding reactions were made as reported (14). Antibodies against RXRs and RARs were a kind gift from Dr. P. Chambon and were used as described (14). c-Jun/AP-1 (polyclonal antibody, Santa Cruz Biotechnology, catalog number sc-44), c-Fos (polyclonal antibody, Santa Cruz Biotechnology, catalog number sc-253), and ATF-1 (monoclonal antibody, Santa Cruz Biotechnology, catalog number sc-270) antibodies were used as recommended by the supplier (2 l/20-l reaction volume, added subsequently to 32 P-labeled oligonucleotide probe and incubated for 30 min at room temperature).
Statistical Analysis-Results are expressed as the mean Ϯ S.E. Significance was assessed using the nonparametrical Mann and Whitney's test.

RESULTS
Specific and Inducible Expression of Ϫ400-AA ϩ -CAT DNA in Brown Adipose Tissue of Transgenic Mice-Eight positive founder mice bearing the Ϫ400-AA ϩ -CAT DNA (Fig. 1A), as identified by Southern analysis, were outbred to generate heterozygous lines. Four lines of transgenic mice out of six analyzed expressed CAT activity. In the four lines, a low CAT activity was detected in interscapular brown adipose tissue but was undetectable in heart, liver, or brain (Fig. 1B). A higher level of CAT activity was observed in brown adipose tissue when mice were either exposed to a cold environment for 16 h (Fig. 1B) or kept at room temperature but injected with norepinephrine, the ␤3-adrenoreceptor agonist CL-316,243 (28) or all-trans-RA ( Fig. 1, C and D). No such CAT induction was observed in liver, brain, or muscle of these mice. In some mice, treatment by norepinephrine or CL-316,243 also induced CAT activity in white adipose tissue (data not shown). We attribute this induction to the presence of a small number of brown adipocytes as well as dormant brown adipocytes in white fat depots (29). The addition of all-trans-RA to in vitro incubated brown fat fragments from transgenic mouse increased CAT activity (data not shown). The data obtained from Ϫ400-UCP-AA ϩ -CAT transgenic mice demonstrated that a DNA fragment made of the 211-bp enhancer fused to 400 bp of the proximal promoter contains sequences that can confer both specific transcription in brown fat and induction or activation by the cold, adrenergic agents or RA. Although the promoter alone linked to the CAT gene was not tested in transgenic mice (see "Discussion"), these data strengthened the interest in analyzing mutants of the ucp gene enhancer using cell transfection experiments.
Retinoic Acid Activates ucp Gene Transcription in 1B8 Cells-1B8 cells were used to analyze mutants of the ucp gene enhancer. The 1B8 immortalized cells do not express the ucp gene unless they are activated by norepinephrine or RA (14). Fig. 2A shows the time course of ucp mRNA induction in 1B8 cells treated with 1 M all-trans-RA for 6 -30 h. RA also induced ucp in 1B8 cell mitochondria ( Fig. 2A). As soon as 2 h after the RA addition, ucp mRNA induction was observed in 1B8 cells. This early response, most logically explained by transcriptional activation, was prevented by the addition of actinomycin D (Fig. 2B). Moreover, in agreement with Alvarez et al. (15), we also observed RA stimulation in the presence of cycloheximide (at 60 M); this effect must be independent of protein synthesis (data not shown). High concentrations of all-trans-RA or 9-cis- RA were necessary to raise ucp mRNA (Fig. 2C). The quantitative effects of RA or isoproterenol on ucp mRNA in 1B8 cells were similar (Fig. 2D). RA also increased ucp mRNA in primary cultures of rat brown adipocytes; in this system RA was as potent as CL-316,243 (Fig. 2D, inset). A dose effect of alltrans-RA similar to that observed on the endogenous gene was observed when the 4551-CAT plasmid was transfected in 1B8 cells; RA induced higher CAT expression than norepinephrine (Fig. 2E).
The Region between bp Ϫ2469 and Ϫ2283 in the ucp Gene Enhancer Mediates the Positive Effects of Retinoic Acid-The CAT vector containing no ucp promoter or only the minimal promoter (from bp Ϫ157 to ϩ110) did not respond to the RA addition (Fig. 3); similar data were obtained with the Ϫ400-   (14). Binding sites of Ets1, nuclear factor 1 (NF1), RXR, or TR (DR3 element) were derived from electrophoretic mobility shift experiments (14). A sequence related to an AP-1 binding site is underlined. tat-ind-HIV-LTR identifies a sequence identical to that implicated in Tat-induced activation of the HIV-LTR (32). The UAR is made of a domain related to AP-1 binding site and a downstream inverted repeat of type 2. TREs were identified between bp Ϫ2391 and Ϫ2376 and between bp Ϫ2348 and Ϫ2334 (18). Horizontal arrows identify repeated elements. The complete sequence (GenBank TM number X12925) of the gene, including the 5Ј-flanking region containing the enhancer, was published by Bouillaud et al. (30). UCP-CAT construct (data not shown). The 211-bp enhancer, attached either to the minimal promoter of the ucp gene (pUCP-AA ϩ plasmid) or to the promoter of the TK gene (pTK-AA ϩ plasmid), conferred inducibility by RA to the reporter gene when transfected into 1B8 cells, even in the absence of cotransfection with a plasmid expressing high levels of RAR or RXR (Fig. 3); similar data were obtained with the Ϫ400-UCP-AA ϩ -CAT construct (data not shown). RA strongly stimulated expression of the basic 4551-CAT plasmid in 1B8 cells; this effect was not enhanced when RARs or RXRs were overexpressed, suggesting that 1B8 cells contain endogenous RA receptors. Deletion of bp Ϫ2469 to Ϫ2283 (4551-CAT B/A del plasmid), which encompasses the DR3 element (bp Ϫ2348 to Ϫ2334) previously located by electrophoretic mobility shift assay analysis (14), abolished the RA effect, even when RXRs or RARs were overexpressed (Fig. 3).
These data indicated that bp Ϫ2469 to Ϫ2283 mediate RA effect. In transfected CHO cells, the same data were obtained, but stimulation by RA was weaker, and the CAT activity was one-tenth of the activity recorded in 1B8 cells. Moreover, RA stimulation of CHO cells was higher in the presence of exogenous RXRs (Fig. 3). In all subsequent experiments with CHO cells, an RXR expression vector was added.
None of the Putative Retinoic Acid Response Elements (RAREs) Present in the Rat ucp Enhancer Mediates RA Response: Effect of Large Deletions on Response to RA- Fig. 4 gives the sequence (30) and organization of the AatII/ApaI domain (bp Ϫ2494 to Ϫ2283) forming the whole rat ucp enhancer (16). In gel shift analysis of rat ucp enhancer, we previously identified a 15-bp region AGGGCAGCAAGGTCA (bp Ϫ2348 to Ϫ2334 in Fig. 4), referred to as the RXR/TR DR3 element, that was able to bind RXR or TR in vitro in such a way that the retarded complex could be supershifted by anti-RXR or anti-TR antibodies (14). Moreover, mutagenesis of the two G residues in the downstream half-site completely inhibited RXR binding (14). Therefore, the contribution of DR3 to RA stimulation was tested in 1B8 cells by transiently transfecting CAT constructs (Fig. 5). In fact, deletion of 15 bp forming the DR3 motif did not impair RA responsiveness (4551 DR3 del in Fig.  5), indicating that the DR3 element by itself was not responsible for the RA effect. The sequence of the enhancer shows that the DR3 element is bound on the 5Ј-side with the sequence AGGCTC (from bp Ϫ2356 to Ϫ2351, Fig. 4) that may constitute a DR2 element in association with the upstream half-site of the DR3 element (see Fig. 4). Such a DR2 was putatively able to mediate RA effect. Actually, this hypothesis was ruled out by the analysis of the FP2 deletion (bp Ϫ2357 to Ϫ2319 were deleted) in the parent 4551-CAT plasmid that showed no inhibition of RA stimulation (Fig. 5).
Reexamination of the enhancer sequence revealed two other putative RAREs located between bp Ϫ2398 and Ϫ2376 (see Fig.  4). Nucleotides Ϫ2398 to Ϫ2386 may form a DR1 element on the opposite strand (CGGCCTCACCCCT), and nucleotides Ϫ2391 to Ϫ2376 may constitute an inverted repeat of type 4 (ACCCCTACTGAGGCAA). Interestingly, Rabelo et al. (18) have shown that this latter region, termed upstream thyroid hormone response element (TRE) by the authors, contributes to responsiveness of the ucp gene to triiodothyronine. RA responsiveness of the region between bp Ϫ2398 and Ϫ2376 was investigated. The three C residues at positions Ϫ2389 to Ϫ2387, known to be protected from methylation in the presence of purified triiodothyronine R␤ (18), were mutated into TTA in the parent CAT construct (Fig. 5). This triple mutation did not alter the responsiveness of the 4551-CAT construct (4551-CAT mut TRE) to RA in 1B8 cells (Fig. 5). Moreover, the same triple mutation was introduced in a 4551-CAT construct also mutated at the level of the two G residues of the downstream half-site of the RXR/TR DR3 element to give the 4551 TRE/DR3 The DR3 motif (AGGGCGCAAGGTCA from bp Ϫ2348 to Ϫ2334) was deleted to create the 4551 DR3 del plasmid. A longer deletion (4551 FP2 del) was also made in the basic construct in order to eliminate the FP2 region (from bp Ϫ2357 to Ϫ2319), which encompasses the DR3 element. The 3 C residues at positions Ϫ2389, Ϫ2388, and Ϫ2387 belonging to a TRE (18) were mutated into T, T, and A, respectively, to create the 4551 TRE mut. This plasmid was mutated in the DR3 region (the 2 G residues of the half-site changed to 2 A residues) to generate the 4551 TRE/DR3 mut. Deletions of bp Ϫ2420 to Ϫ2319 and bp Ϫ2457 to Ϫ2421 were made to obtain 4551 3Ј del and 4551 5Ј del constructs, respectively. The plasmids were transiently expressed in 1B8 cells before the CAT assay. Data represent mean Ϯ S.E. (four experiments). **, significant stimulation by all-trans-RA (p Ͻ 0.01); *, significant stimulation by all-trans-RA (p Ͻ 0.05). mut plasmid. The responsiveness to RA of this latter plasmid in 1B8 cells remained high (Fig. 5). In conclusion, the TRE described by Silva and colleagues (18) does not contribute to RA effect on the ucp gene. The absence of effect of the 3-C mutation at bp Ϫ2389 on RA responsiveness also invalidates a potential role of the type 1 direct repeat located on the opposite strand between bp Ϫ2398 and Ϫ2386 (Fig. 4).
We decided to create two promoter mutants, each having a large deletion from either bp Ϫ2420 to Ϫ2319 (4551 3Ј del plasmid) or bp Ϫ2457 to Ϫ2421 (4551 5Ј del plasmid). With these 2 plasmids, RA responsiveness was strongly decreased and remained only slightly positive (Fig. 5).
Mutations in a TGAATCA Region Inhibit RA Responsiveness- Fig. 6A shows that the two large deletions analyzed in Fig. 5 split the bp Ϫ2421 to Ϫ2415 TGAATCA motif, a sequence resembling the consensus AP-1 binding site (TGA(C/G)TCA; Ref. 31). This motif is located just at the 3Ј-end of the footprinted FP1 region and is bound to a sequence previously identified as tat-ind-HIV-LTR (32), which is itself inserted in a domain resembling an inverted repeat of type 2 referred to as IR2 (see Figs. 4 and 6A). To further analyze the possible role of this putative AP-1 binding site, we made two types of mutation, preventing Jun and Fos binding (33). The first mutation (4551 AP-1 mut 1) was introduced either in the 4551 CAT DNA or in the 4551 FP2 del plasmid. In the presence of RA, the CAT activity generated from the 4551 AP-1 mut 1 plasmid or the 4551 AP-1 mut 1/FP2 plasmid was strongly decreased in 1B8 cells (Fig. 6B). The second mutation (4551 AP-1 mut 2) also decreased RA responsiveness (Fig. 6B). In CHO cells, a similar inhibition of RA effect on mutants was observed. These experiments demonstrated that integrity of a TGAATCA site, located between bp Ϫ2421 and Ϫ2415, was required for enhancer activity in the presence of RA.
Mutation of Nucleotides Downstream of TGAATCA Also Inhibits RA Responsiveness: Delineation of UAR, an Activating Region-We were surprised by the finding that mutations of the element related to AP-1 markedly decreased RA responsiveness, since a role for AP-1 factors in transcriptional activation by RA had never been described. A hypothesis explaining the function of TGAATCA in RA activation of ucp transcription is that the RA effect is mediated by an unidentified RARE, which may be inhibited when the AP-1-type site is mutated. This element could be the sequence immediately downstream of the TGAATCA domain: a DNA stretch that is similar to a sequence important in Tat-induced activation of HIV-LTR (31) and that is contained in IR2 (Fig. 6A). Another repeated region related to a nuclear receptor binding site is also present in FP1, upstream of the TGAATCA domain. In other respects, the comparison of mouse (17) and rat (30) enhancer sequences shows that the AP-1-type TGAATCA sequence of the rat gene is replaced by a GAAATCA sequence in the mouse gene. This mouse genomic fragment is not an AP-1 motif and did not compete in a gel shift experiment (see use of mutant oligonucleotide shown in Fig. 7A, right part). However, the sequences that are on both sides of the (T/G)(G/A)AATCA region are highly conserved in the two species. This suggests that these regions contain important cis-acting elements.
In order to delineate cis-acting elements close to the TGAATCA domain, several mutations were made upstream of this domain at position Ϫ2427, Ϫ2429, Ϫ2433, Ϫ2435, Ϫ2441, Ϫ2443, Ϫ2451, or Ϫ2457 in CAT constructs. RA responsiveness of these CAT constructs was not impaired (data not shown). Then we mutated bp Ϫ2407 to Ϫ2404 in the IR2 element to create the 4551 IR2 mut CAT DNA (Fig. 6B). The IR2 mutation was also introduced in the 4551 CAT DNA already mutated at the level of the putative AP-1 binding site to give the 4551 AP-1/IR2 mut. The RA responsiveness of 4551 CAT, 4551 AP-1 mut 1, 4551 IR2 mut, and 4551 AP-1/IR2 mut was compared in transfection experiments. Fig. 6B shows that mutation of the IR2 domain decreased RA responsiveness by 80%, indicating that this mutation was almost as effective as mutation of the TGAATCA domain; moreover, mutation of both TGAATCA and IR2 domains decreased RA responsiveness by 85-90%. TGAATCA and IR2 domains form a ucp gene-activating region referred to as UAR (see Fig. 6A).

DNA Mobility Shift Binding Activity of TGAATCA in the ucp Enhancer Reveals Binding of Factors Related to Jun and Fos-
The ability of bp Ϫ2422 to Ϫ2416 of the ucp gene enhancer to bind nuclear factors was analyzed in vitro using an electrophoretic mobility shift assay (Fig. 7). A probe referred to as probe 24 and corresponding to bp Ϫ2424 to Ϫ2407 was synthesized (see Fig. 4). Probe 24 was labeled with 32 P and incubated with nuclear proteins from various tissues. Whereas a very   FIG. 6. Mutations of a putative AP-1 binding site or of IR2 in the ucp gene enhancer strongly impair RA responsiveness. A, sequence of boundary between the two large deletions tested (see Fig. 5, 4551 5Ј del and 4551 3Ј del). These deletions split a domain related to the AP-1 binding site (sequence shown in boldface type). Downstream of this site is a sequence referred to as IR2, related to an inverted repeat of type 2. This region encompasses a domain that in HIV is involved in Tat-induced activation of the LTR (32). The UAR is made of an AP-1type domain and IR2. The arrow above FP1 indicates the 3Ј-limit of the DNase 1 footprinted region FP1 (14). B, transient transfection experiments of 1B8 cells with CAT constructs made of entire (4551-CAT) or mutated 5Ј-flanking region of the ucp gene treated or not with 10 Ϫ6 M all-trans-RA. Two types of mutations of the site related to the AP-1 binding site were introduced in the parent construct (the putative AP-1 binding site TGAATCA was mutated either to GAATTCA or TGAACAA to give 4551 AP-1 mut 1 or mut 2, respectively). The element referred to as IR2, which is located immediately downstream of the AP-1 binding site, was mutated in the 4551 CAT parent construct to give the 4551 IR2 mut plasmid. The IR2 mutation was also introduced in this latter construct to give the 4551 AP-1/IR2 mut plasmid. Each bar corresponds to the mean Ϯ S.E. of four independent experiments. **, significant stimulation by RA (p Ͻ 0.01); *, significant stimulation by RA (p Ͻ 0.05). The numbers above the bar, representing the ϩRA data, denote the -fold induction of CAT activity (with RA/without RA). Similar data were obtained when transfection experiments were made in CHO cells (data not shown). faint binding was observed when using liver factors, a sharp retardation was observed with 1B8 cell proteins; a similar retardation was also obtained with nuclear proteins of hamster brown adipose tissue. Oligonucleotides corresponding to footprinted FP1 and FP2 did not compete with the probe 24. Competition was observed in the presence of an excess of 24, AP-1, or cAMP response element-binding protein (CREB) oligonucleotides (Fig. 7A, left). AP-1 oligonucleotide containing the motif TGAGTCA was a better competitor than CREB oligonucleotide containing the sequence TGACGTCA (Fig. 7A, middle). Moreover, a higher amount of probe was retained by proteins prepared from norepinephrine-stimulated 1B8 cells, whereas treatment of cells by RA did not increase retardation (Fig. 7A,  right). Such an increase in binding with factors from norepinephrine-stimulated 1B8 cells was not obtained when using other probes such as Sp1 and CREB probes (data not shown). In order to approach the identity of protein(s) retained by the 24 DNA, this fragment was tested in the presence of nuclear proteins extracted from NIH 3T3 cells stimulated by fetal calf serum or transformed by RAS; both treatments induce AP-1 binding factors (34). Retardation was increased when the 24 probe was incubated with factors from activated NIH cells. A 50-fold excess of 24 oligonucleotide competed effectively (Fig.  7B, left). The band obtained with the 24 probe in the presence of extracts from 1B8 cells was also observed when using the AP-1 oligonucleotide as a probe; in that case, the complex was sensitive to the addition of antibodies against AP-1 factors (Fig.  7B, middle). Moreover, the complex obtained between the 24 probe and 1B8 extracts was inhibited in the presence of antibodies broadly reactive with c-Jun, Jun B, and Jun D proteins FIG. 7. Gel shift analysis of the putative AP-1 binding site in the ucp gene enhancer. A, in the left part, the labeled probe 24, encompassing the putative AP-1 binding site of the rat enhancer, was incubated with nuclear factors from rat liver (10 g), control 1B8 cells (1B8c; 5 g), or BAT (2 and 5 g). In the right part, the probe 24 was incubated with nuclear factors from 1B8 cells treated with 10 Ϫ5 M norepinephrine for 4 h (1B8 ϩ NE; 5 g), or 1B8 cells treated with 10 Ϫ6 M all-trans-RA for 2 or 4 h (1B8 ϩ RA; 5 g). mu oligonucleotide competitor corresponds to the mouse ucp gene fragment located at a position equivalent to the AP-1 binding site of rat gene when enhancers of both species are aligned. In the left and right parts, competitor oligonucleotides were used at a 50-fold mass excess. In the middle part, binding of 1B8 nuclear factors to labeled probe 24 was competed with increasing concentrations of either competitor oligonucleotide 24 (20-and 50-fold mass excess) or AP-1 or CREB (20-, 50-, and 100-fold mass excess). B, probe 24 or AP-1 was incubated with nuclear factors prepared from NIH 3T3 cells untreated (NIH), treated by serum (NIHs), or transformed by Ras (NIHr) NIH 3T3 cells. The same probes were also incubated with nuclear proteins prepared from control (1B8c) or norepinephrine-treated 1B8 cells (1B8 ϩ NE). Competition with cold 24 or AP-1 probe is shown. Antibodies used were broadly reactive against Jun family, Fos family, or CREB family (Atf) or specific for RXR␣ (Rx) or RAR␤ (Ra). On the right, the arrow indicates a supershifted complex. Nuclear extraction and band shift assays as well as details of use of antibodies were as described under "Experimental Procedures." The sequences of synthetic oligonucleotides were as follows: 24, AATTCATGAATCAGGCTCTCTG; AP-1, AGCTTGATGAGT-CAGCCG; FP1, AATTCCTTTCCACGCTTCCTGCCAGAGCATGAG; FP2, TCTGAGGGCAGCAAGGTCAGCCCTTTCTTTGGA; CREB, AGAGAT-TGCCTGACGTCAGAGAGCTAG; mu, AATTCAGAAATCAGACTCTCTG.
FIG. 8. Gel shift analysis of IR2 in ucp gene enhancer. In the right panel, the labeled probe IR2 was incubated with nuclear factors from 1B8 cells (1B8; 5 g); competition with cold IR2 or 24 probe (50-fold excess) is shown. In the left panel, the IR2 probe was incubated with nuclear factors from 1B8 cells (5 g) or rat liver (L; 10 g). Antibodies used were specifically reactive against RXR␣, RAR␣, RAR␤, or RAR␥ or broadly reactive against TRs, CREB family (Atf), or Jun family (Jun). The use of antibodies was as described under "Experimental Procedures." The sequences of synthetic oligonucleotides were as follows: IR2, AATTCAGGCTCTCTGGGGATACCG; 24, AATCATGAA-TCAGGCTCTCTGG. and was supershifted in the presence of antibodies broadly reactive with c-Fos, Fos-B, Fra-1, and Fra-2; in the same experiment, neither ATF-1 antibodies reactive with ATF-1, CREB-1, and CREM-1 proteins nor antibodies against RAR␤ altered the shift (Fig. 7B, right); however, in the same experiment, antibodies against RXR␣ were slight inhibitors (Fig. 7B, right; see "Discussion"). The binding of 1B8 factors (sensitive to anti-ATF-1 antibodies) to a CREB probe was not stimulated by factors from norepinephrine-treated 1B8 cells, since it was observed with the 24 or AP-1 probe (data not shown).
DNA Mobility Shift Binding Activity of IR2 Reveals Binding of Factors Related to RXR␣ and RAR␤-The ability of IR2 to bind nuclear factors was analyzed in vitro using an electrophoretic mobility shift assay (Fig. 8). A weak binding was obtained with liver nuclear factors, whereas a strong and specific complex was obtained with nuclear factors from 1B8 cells. The probe 24 corresponding to the putative AP-1 binding site of the rat ucp gene enhancer did not compete with IR2 probe. When using a series of antibodies against transcriptional factors, the complex retained by the IR2 probe was inhibited (and not supershifted) in the presence of antibodies against RXR␣ and RAR␤. Antibodies against RAR␣, RAR␥, TR receptors, or members of CREB or Jun families were ineffective (Fig. 8). These data demonstrated that proteins related to RXR␣ and RAR␤ can bind to IR2. Moreover, treatment of brown adipocytes by RA induced the binding to IR2 probe (data not shown).
Nucleotides Upstream of UAR and FP1 Are Required for RA Responsiveness: Enhancer Activity of UAR-Since several independent mutations in UAR strongly decreased the transcriptional activity of the ucp gene in presence of RA, transient transfection experiments were performed to test whether it was able to mediate a RA response in 1B8 cells. In fact, TGAATCA alone (data not shown), UAR (bp Ϫ2421 to Ϫ2402), or three tandem copies of UAR did not mediate RA effect (Fig.  9). The addition of FP1 (bp Ϫ2444 to Ϫ2423) to UAR did not confer RA responsiveness. We deleted nucleotides from the 5Ј-extremity of the enhancer (bp Ϫ2494) to the BclI site (bp Ϫ2469) and observed a significant but partial reduction of the RA effect, compared with the response of the pUCP-AA ϩ -CAT construct (Fig. 9). These data indicated that nucleotides forming the BclI site or upstream of this site participate in the RA effect. It was decided to assay the RA response of a 92-bp DNA spanning the 5Ј-extremity of the enhancer to the 3Ј-end of the UAR. This 92-bp DNA conferred a strong response to RA both in the context of ucp gene minimal promoter and TK promoter. Similar data were obtained when transfecting CHO cells (data not shown).
In Fig. 9, it is shown that the basal activity of a plasmid bearing three tandem copies of UAR was 83% of the activity obtained with the whole 211-bp enhancer. These data point to the enhancer activity of UAR, at least in the context of the ucp gene minimal promoter. DISCUSSION We had shown earlier that cis-acting elements of the rat ucp gene are present in a 211-bp enhancer at Ϫ2.4 kilobase pairs (14). By performing cell transfections with constructs containing various deletions of the mouse ucp 5Ј-flanking region, Kozak et al. have proposed that the enhancer is essential for brown fat specificity (17). In order to further analyze the role of the rat enhancer in tissue specificity and to identify the cisacting elements present inside this enhancer that are responsible for both brown fat specificity and regulation by hormonal factors such as retinoids, we have set up a strategy based both on creation of transgenic mice and transient expression of CAT constructs in cell lines. FIG. 9. UAR alone does not confer a RA response, but a larger fragment of 92 bp does confer it in the context of a homologous or a heterologous promoter-enhancer activity of UAR. Transient transfection experiments of CAT constructs were made in 1B8 cells. The upper part shows constructs made using the 157-bp ucp gene minimal promoter (pUCP). The lower part shows constructs made using the minimal TK promoter (pTK). The pUCP-AA ϩ -CAT plasmid contains the entire rat ucp gene enhancer. UAR alone or attached to FP1 was ligated upstream of the promoter-CAT reporter construct. One (UAR) or three (3 ϫ UAR) tandem copies of the UAR were ligated upstream of the pUCP-CAT construct. A 92-bp DNA fragment corresponding to bp Ϫ2489 to Ϫ2398 was ligated in the correct (92 bp) or inverse (R92bp) orientation upstream of pTK-CAT construct. The inverted 92-bp fragment, in the inverse orientation, was also ligated upstream of the pUCP-CAT construct. The data are expressed as an average -fold induction of CAT activity (with RA/without RA) Ϯ S.E. of three separate experiments. *, significant statistical difference (p Ͻ 0.01) relative to the activity of pUCP-CAT or pTK-CAT plasmid, respectively; **, significant statistical difference (p Ͻ 0.05) relative to activity of pUCP-AA ϩ -CAT plasmid. Numbers in the upper right corner correspond to CAT activity obtained with pUCP, pUCP-AA ϩ , 1 ϫ UAR-pUCP, and 3 ϫ UAR-pUCP-CAT reporter plasmids in the absence of RA, the CAT activity of the pUCP-AA ϩ construct being taken as 100%.
The specific and regulated expression of the Ϫ400-AA ϩ -CAT DNA in brown adipose tissue of several lines of transgenic mice allowed the location of essential cis-acting elements, since no transgene expression occurred in tissues other than BAT, even in animals treated with adrenergic compounds or RA. The marked effect of CL-316,243 on transgene transcription in brown fat agrees with the known ability of ␤3-adrenoreceptors to activate ucp gene transcription (36,37). The rather weak stimulation by norepinephrine and RA observed in transgenic mice compared with the CL-316,243 effect may result from different degradation rates of the inducers or from dose effect.
The Ϫ400-AA ϩ DNA is the smallest fragment known to be able to drive expression of a reporter gene specifically in brown adipose tissue. In fact, the present data belong to a detailed and not yet finished analysis of cis-acting elements of ucp gene in transgenic mice. Since we have no data yet with transgenic mice bearing the Ϫ400 bp promoter only, the type of construct used in Ϫ400-AA ϩ -CAT transgenic mice did not allow us to discriminate between the respective roles of the Ϫ400 bp promoter and the 211-bp enhancer located at Ϫ2.4 kilobase pairs. However, since Boyer and Kozak (35) were unable to detect any expression of a mouse ucp minigene in transgenic mice in which only 1.2 kilobase pairs of the 5Ј-flanking region was used, it may be concluded that the enhancer itself is essential for specific expression in brown adipose tissue. The creation of new transgenic mice containing either the enhancer alone attached to a minimal promoter or the 4551-bp DNA deleted from bp Ϫ2494 to Ϫ2283 will definitively establish the essential role of the ucp enhancer in tissue specificity.
Retinoids are positive effectors of adipose cell differentiation, and growing adipoblasts contain RAR and RXR mRNAs (38,39). Moreover, differentiated adipocytes respond to RA with a rapid increase in S14 (40) and phosphoenolpyruvate carboxykinase gene transcription (41). A goal of the present work was a functional analysis in transfected cells of cis-acting elements present in rat ucp enhancer and involved in RA activation of ucp gene transcription.
Transfection of 1B8 or CHO cells with deleted and mutated CAT constructs demonstrated that neither the DR3 region nor the FP2 region mediates RA effect. This may be explained by the fact that the DR3 sequence differs from that of typical RARE. Moreover, it is known that RAR⅐RXR heterodimers and RXR homodimers strongly activate transcription from directly repeated RGGTCA separated by 1, 2, 5, 10, or more bp and weakly activate transcription from directly repeated RGGTCA separated by 3 bp (42)(43)(44). Our data also indicate that other putative RAREs do not mediate RA effect. Moreover, polymers of footprinted FP1 and FP2 regions (14) attached to a minimal promoter fused to the CAT gene and transfected in 1B8 cells responded very weakly to RA (data not shown).
UAR is the only region in which mutations strongly impaired RA responsiveness. This region can in vitro bind RXR␣ and RAR␤ precisely at the IR2 level and probably constitutes a nonconventional site for RA effect in the ucp gene. Mutations of the TGAATCA sequence revealed that the integrity of this domain was required for high transcriptional activity in the presence of RA. Although electrophoretic mobility shift assays and use of various antibodies suggested that the TGAATCA region may be related to an AP-1 binding site, our data do not demonstrate that this domain present in the enhancer is a functional AP-1 binding site. Such a role for the AP-1 binding site in transcriptional activation by RA has never been reported. Moreover, we were unable to see any increase in ucp mRNA or 4551-CAT activity in 1B8 cells treated with phorbol esters (data not shown). Retinoid effects are mediated by the RARs and RXRs, both of which induce transcriptional activa-tion through specific RAREs (19,20,38). However, a number of cases have been documented in which RA affects transcription through an AP-1 binding site (45)(46)(47)(48); in that situation, RA represses transcription and RAR does not bind the AP-1 site. An interesting situation is that of the interleukin-6 promoter, which can be strongly stimulated by RA, although it does not contain elements similar to previously described RAREs (49). In other respects, although the consensus sequence for AP-1 sites is TGA(G/C)TCA, a TGAATCA sequence has already been shown to bind Jun (33); a similar AP-1 binding site was also described in the steroid 11␤-monooxygenase gene (50).
The simplest explanation of how the TGAATCA site could intervene in the response to RA of the ucp gene is that RA directly activates or induces factors that bind to this site. However, this hypothesis is contradicted by our gel shift experiments, since RA treatment of 1B8 cells did not activate the binding of factors to probe 24 (Fig. 7A, right). Intriguingly, the complex bound to probe 24 was partially inhibited in the presence of antibodies against RXR␣ (Fig. 7B, right). This observation cannot be easily explained. This effect may be related to the probe 24 that encompasses the 5Ј-moiety of the IR2 probe; in other respects, Sista et al. (51) have described a functional interaction between RXR␣ and BZLF-1, which is an AP-1-like protein. A stimulation of binding to AP-1 probe occurred with factors from 1B8 cells treated by norepinephrine. This may be related to the induction of c-Fos mRNA in brown adipocytes stimulated by norepinephrine (52).
An interesting feature of UAR is that, in three tandem copies, it could induce a 27-fold increase of the activity of ucp gene minimal promoter (Fig. 9). Therefore, UAR contributes strongly to the enhancer activity of the AatII-ApaI region of the ucp gene. This agrees with the fact that mutations of UAR impair activation of ucp gene transcription by noradrenaline. 2 Although the presence and integrity of UAR was required for RA responsiveness of the ucp gene, we observed that UAR alone was unable to confer RA responsiveness and therefore does not constitute a RARE. To achieve RA responsiveness, it was necessary to use 92 bp containing UAR and forming the 5Ј-moiety of the enhancer. It can be concluded that this enhancer contains noncanonical RAREs that span up to 92 bp. Such a complex situation was described already for several genes such as the mouse laminin B1, rat growth hormone, and human oxytocin genes (described in Ref. 53). Our data demonstrate that, in addition to UAR, nucleotides forming the BclI site, upstream of this site, or located between bp Ϫ2469 and Ϫ2445 participate in RA activation of rat ucp gene transcription. Interestingly, it was very recently reported that this region in the enhancer of the mouse ucp gene can bind RXR␣ and peroxisome proliferator-activated receptor (54).