I n Vitro Interactions between Nuclear Proteins and Uncoupling Protein Gene Promoter Reveal Several Putative Transactivating Factors Including Etsl, Retinoid X Receptor, Thyroid Hormone Receptor, and a CACCC Box-binding Protein*

Previous studies of rat ucp (uncoupling protein) gene organization carried out in this laboratory identified regulatory sequences located in the 5”flanking region. In this work, DNase I footprint analysis of the enhancer revealed two domains at base pairs (bp) -2444 to -2423 and bp -2352 to -2319. The former domain can bind in vitro, in a cooperative manner, factors related to nuclear factor 1 and Etsl; the latter domain contains a type 3 directly repeated sequence that was shown to be able to bind the retinoid X and triiodothyronine receptors. Moreover, a positive effect of retinoic acid on ucp mRNA levels in immortalized brown adipocytes was observed. DNase I footprint analysis identified two hypersensitive regions, A and B, at bp -509 to -472 and bp -403 to -350, respectively; region A contains a repeated CACCC box, and region B can bind protein related to Etsl. The A box differentially binds liver and brown adipose tissue nu- clear proteins and could be involved in uncoupling protein induction. Further analysis showed three foot- printed boxes, C-E, at bp -182 to -159, -147 to -120, and -111 to -85, able to bind in vitro proteins related to nu- clear

Following the isolation of mouse (31, rat (41, and human (5) ucp genes, the first studies of the ucp gene promoter in transgenic mice demonstrated that both tissue-specific and p-adrenergic elements are present in 3.8 or 4.5 kilobases of DNA upstream of the start site of transcription of the mouse (6) or rat (7) ucp gene, respectively. Using transgenic mice bearing ucp minigenes, Boyer and Kozak (6) proposed that a sequence between kilobases -1.2 and -3 of the 5'-start site flanking region was involved in the control of mouse UCP expression. Another study of the rat ucp promoter based on transfection of cultured brown adipocytes by CAT-DNA constructs established the importance of a strong 211-bp enhancer element (regulatory region 1 (Rl)) located between bp -2494 and -2283 (7). In the same study, a second region (regulatory region 2 (R2)) located between bp -400 and -157 was suspected to contain negative cis-acting elements; a minimal promoter was identified between bp -157 and -57 (7). Very recently, Kozak et al. (8) identified regions of the mouse ucp gene homologous to the R1 and R2 elements of the rat gene as well as a TTCC motif in R1 and several CAMP response elements (CREs) involved in UCP inducibility. In this study, we have examined in detail the DNA elements present in the promoter of the rat ucp gene and tried to identify proteins binding in vitro to these elements.

Cell Culture Conditions and
Zkansfections-Primary cultures of brown adipocytes were carried out as described previously (9). 1B8 cells correspond to a cell line derived from a brown fat tumor of a transgenic mouse previously used for the isolation of HIB 1B cells (10, 11); this cell line, when stimulated by catecholamines or CAMP, expresses brown fat-specific mitochondrial UCP. 1B8 and HIB 1B cells were grown and differentiated as described (11). Northern analysis of ucp mRNAin 1BS cells was carried out as described previously for primary cultures (9). CHO K1 cells were grown in Ham's F-12 medium supplemented with 10 m M glutamine and 10% fetal calf serum.
Nuclear Extracts and Protein Purification-Nuclear extracts from Siberian hamster brown adipose tissue, rat liver, and rat kidney were prepared as described (12) in the presence of 0.5 m M phenylmethylsulfonyl fluoride, 2.5 m M benzamidine, 10 pg/ml aprotinin, 1 pg/ml leupeptin, 1 pg/ml antipain, and 1 pg/ml pepstatin. Nuclear extracts from cells were prepared according to Shapiro et al. (13) using the dialysis buffer described by Gorski et al. (12). For protein purification, nuclear extracts were resuspended in 20 m M Hepes (pH 7.6), 100 m M KCl, 0.2 m M EDTA, 5% glycerol, and 0.5 m M phenylmethylsulfonyl fluoride. Expression in Escherichia coli and purification of GST-EZ protein were achieved according to Tsukiyama and Niwa (14).

DNase I Protection and Electrophoretic Mobility Shift
Assays-For analysis of the R1 domain, DNA fragments containing the 211-bp R1 either in direct orientation or in reverse orientation and fused to the first 400 bp of the rat ucp promoter (7) were digested with BgZII, 3'-labeled by filling in with the Klenow fragment of DNA polymerase I in the presence of [a-32PldGTP, and cut with SpeI (at bp -173 of the promoter). The labeled fragment was isolated after polyacrylamide gel electrophoresis. For analysis of R2, a -241-CAT plasmid (7) was digested with BglII and SacI (position +53) to label one strand or was digested with BglII and BstBI (position +92) to label the other strand.
A bp -6111-235 fragment was obtained by digestion of a -611-CAT construct (7) with BglII and B a d , subcloned in pUC18 in the two orientations using repaired SacI and SphI sites, cut with EcoRI, and labeled with [a-"PldATP. Binding reactions were carried out for 20 min on ice in a final volume of 50 pl containing 0.5 ng of labeled probe (15,000-20,000 Cerenkov cpm) a t a final buffer concentration of 20 mM Hepes (pH 7.6), 100 mM KCl, 0.5 mM EDTA, 1 mM dithiothreitol, 5 mM MgCI,, 2 pg of poly(dI.dC), 2% polyvinyl alcohol, and 10% glycerol. After incubation, DNase I(lp1,50-500 ng/pl) was added, and incubation was continued for 1 min at 20 "C. A stock solution of DNase (2 mg/ml in 0.15 M NaCl and 50% glycerol, stored a t -80 "C) was freshly diluted in 25 mM Ms-HCl (pH 7.5), 2.5 mM MgCI,, 100 pg/ml bovine serum albumin, and 10% glycerol (15). The reaction was stopped by the addition of 150 pl of 120 mM NaCl, 0.6% sodium dodecyl sulfate, 12 m M EDTA, 150 pg/ml tRNA, and 90 pg/ml proteinase K, incubated for 1 h a t 50 "C; extracted with phenol/chloroform (1:lk and precipitated with ethanol. Dried DNA pellets were dissolved in denaturing buffer and boiled before loading onto a 6% denaturing polyacrylamide gel. For the gel mobility shift assay, different probes were used (see figures). Protein-DNA binding was carried out in a volume of 20 pl containing 0.5 ng of labeled oligonucleotide (20,000-40,000 Cerenkov cpm) in 10 x band-shift buffer (250 mM Hepes (pH 7.5), 50 mM MgCI,, and 340 mM KCl) with 1-4 pg of poly(dI.dC) and 0-0.6 pg of sonicated salmon sperm per reaction depending on the oligonucleotide. Electrophoresis was performed a t 150 V a t room temperature. Gels were dried and autoradiographed a t room temperature. Use of anti-ELP antibody (gift of Dr. 0. Niwa) in the gel retardation assay was as described (16). For other antibody supershift assays, anti-mRXR-a,p,y 4RX-1D12 monoclonal antibody (gift of Dr. P. Chambon) was incubated with nuclear extracts prior to the addition of the DNA probe; antiserum against v-ErbA able to immunoprecipitate c-ErbA-a and c-ErbA-p (gift of Dr. J. Ghysdael) was added to the samples after incubation with the probe, and incubation was continued for 20 min a t room temperature.

RESULTS
DNase I Protection Analysis of Rl: Delineation of Boxes FPl and FP2-We delineated earlier a 211-bp enhancer region (Rl, bp -2494 to -2283) using transfection of cultured differentiated brown adipocytes (7); however, this enhancer was shown to be able to activate the minimal promoter of the ucp gene or of a heterologous promoter both in brown adipocytes and in cell lines that do not express UCP. To dissect R1 further, nuclear extracts prepared from either Siberian hamster brown adipose tissue (BAT) or rat liver were used for DNase footprint analysis with a DNA fragment including the 211 nucleotides of R1. Two footprints were observed on both strands of DNA. These footprints, termed FP1 and FP2, span 21 nucleotides from bp -2444 to -2423 (FP1) and 33 nucleotides from bp -2352 to -2319 (FP2) in the coding strand ( Fig. 1). A similar picture of FP1 and FP2 was obtained using proteins extracted from rat kidney nuclei (data not shown). Box FP1 includes potential binding sites for NF-1 (17) and Etsl (18) factors. When footprint analysis was performed in the presence of oligonucleotides corresponding to the binding sites for these factors, competition was observed with the NF-1 oligonucleotide; this competition was abolished when using a FP1 nucleotide containing a mutation of 2 bases essential to NF-1 binding, FP1 mut NF-1 (Fig. lB 1. In the presence of the Etsl oligonucleotide, no competition was seen (data not shown).
Gel Shift Analysis of Protein-DNA Interactions a t Box FPl: Participation of NF-1 and Etsl Factors-The potential NF-1and Etsl-binding sites present in the FP1 region were studied using band-shift experiments (Fig. 2). Retarded complexes were observed using a labeled FP1 probe and liver, CHO cell, 1B8 cell (immortalized brown adipocytes), or BAT protein extracts. In competition experiments, the FP2, AP1, C/EBP, and (A) and FPl ( E ). DNase I footprinting of the UCP 5"flanking region was carried out with the -2494/-2283 AatII-ApaI enhancer fused to the first 400 bp upstream of the transcription site (7) labeled in the coding strand. The probe was incubated with nuclear extracts from rat liver, Siberian hamster BAT, or rat kidney and treated with DNase I under the conditions described under "Materials and Methods." Numbers above the lanes correspond to micrograms of nuclear proteins. Competition experiments were carried out in the presence of a 200-fold mass excess of nonradioactive FP2, FP1, NF-1, or FP1 mut NF-1 oligonucleotide; the sequences of the FP1, NF-1, and FP1 mut NF-1 oligonucleotides are given in the legend of Fig.   2; the sequence of the FP2 oligonucleotide is in Fig. 3. G/A, Maxam-Gilbert sequence reactions. Vertical burs delineate footprints termed FP1 and FP2, which were also observed when using the opposite strand as a probe.
HNF-1 oligonucleotides were ineffective, whereas the FP1, AFP4 (which contains an NF-1-binding site and a C/EBP-binding site), and NF-1 oligonucleotides inhibited the formation of complexes. These data are in agreement with the footprint experiment shown in Fig. 1 and strongly suggest that a protein related to NF-1 binds in vitro to the FP1 box. Although an Etsl probe was unable to compete with the FP1 probe in footprint experiments, competition was noticed in the band-shift analysis with BAT or liver proteins (low molecular weight complex in Fig. 2, C and D); mutations of 2 essential bases in the Etslbinding site (Etsl mut) suppressed the competition (Fig. 20). Interestingly, a better competition effect of the NF-1 or Etsl oligonucleotide was obtained when both oligonucleotides were present, suggesting a cooperative effect of the two factors. To demonstrate binding of Etsl to FP1, purified Etsl protein was added to the labeled Etsl probe; under these conditions, the FP1 oligonucleotide behaved as a competitor, and mutations of the Etsl-binding site of FP1 (FP1 mut Etsl) prevented this competition (Fig. 2E).
Gel Shift Analysis of Protein-DNA Interactions a t Box FP2: Binding of Factors Belonging to Superfamily of Nuclear Hormone Receptors and Related to RXR and T3 Receptor-A FP2 probe incubated with proteins from several tissues or cells was retarded during gel migration (Fig. 3). Competition was only observed with the FP2 oligonucleotide; an unrelated oligonucleotide such as FP1 did not compete. A FP4 probe, shorter than the FP2 probe, was also retarded (Fig. 3). In fact, both the FP2 and FP4 sequences contain a 5'-AGGTCA-3' element that has the P box consensus sequence of the first zinc finger of the steroid hormone receptor superfamily, which also includes thyroid hormone receptors, vitamin D, receptors, retinoic acid receptors, and peroxisome proliferation-activated receptors (19); this element is itself included in the sequence CAAGGTCA, which is a binding site for ELP (16), and is included in a DR3  3) did not compete with FPA for the formation of the FP4-protein complex, indicating that the AGGTCA consensus sequence is essential. This was the first indication of binding of a nuclear receptor to the FP2 region of the UCP enhancer. When a double mutation (FP7 oligonucleotide) (Fig. 3) of the first element of the putative DR3 sequence was used, competition with the FP4 oligonucleotide, although weaker, was also observed. We also observed that when the FP7 oligonucleotide was used as a probe, the complex obtained with the FP2 or FP4 oligonucleotide was partially destabilized (data not shown). We concluded that the first element of the presumed DR3 sequence was also involved in protein binding. Several nuclear receptors bind to genomic regulatory elements by forming a heterodimer with RXR (20-23). To determine if this was indeed the case, a monoclonal antibody against RXR was used in gel shift experiments.
In the presence of this antibody, the complex was supershifted ( Fig. 31, whereas anti-RAR-a, -p, and -y had no effect (data not shown), indicating that the FP4 region of the ucp promoter could be a binding site for a member of the superfamily of receptors that is able to dimerize with RXR. In agreement with this observation, retinoic acid alone was able to induce ucp mRNA in 1B8 cells; in addition, retinoic acid promoted the effect of adrenergic agonists on UCP (Fig. 4). No effect of retinoic acid on a-glycerophosphate dehydrogenase, a marker of adipocyte differentiation, was observed. The same observations were made with primary cultures of brown adipocytes.' T, receptors are known to form heterodimers with RXR (20, 23). To test the hypothesis of binding of the T, receptor to the DR3 sequence, which, in fact, is not a consensus sequence for the thyroid hormone-binding site, the FP4 probe was incubated with recombinant v-ErbA protein (Fig. 5). This experiment demonstrated that FP4 can bind in vitro the thyroid hormone receptor. The addition of anti-T, receptor antibodies to BAT nuclear proteins incubated with the labeled FP4 probe induced a supershifted complex and confirmed that the FP4 region of the ucp promoter contains a binding site for the T, receptor (Fig. 5).
In other experiments, since a consensus sequence for ELP (which is a member of the superfamily of nuclear receptors) binding is also present in FP2, a FP3 oligonucleotide with a double point mutation a t positions known to be necessary for the specific recognition of ELP (16) 1 pg), or HIB 1B cells (4 pg). Competitor oligonucleotides (mass excess) were used as indicated above the lanes. The labeled FP4 probe was incubated with liver or BAT nuclear extracts in the presence or absence of a 200-fold mass excess of FP4, FP8, or FP7 oligonucleotide. The same probe was incubated with nuclear extracts (3 pg of liver protein and 1 pg of BAT protein) in the presence or absence of 1 pl of anti-mRXR-a,P,y monoclonal antibody. Nuclear extraction and band-shift assays were carried out as described under "Materials and Methods." Nucleotide sequences (which also contained ATTC/G EcoRI extremities) are given on the left-hand side; the tandem arrows indicate a DR3 sequence, and the broken arrow indicates a potential ELP-binding site. The FP1 oligonucleotide sequence is given in the legend of Fig. 2. No competition was observed when using the AP1, C/EBP, or AFP4 oligonucleotide (data not shown). otide could not compete with the FP2 probe in the presence of nuclear proteins (Fig. 6B). These data suggest that a protein related to ELP can bind in vitro the FP2 region of the ucp promoter. To investigate this hypothesis, the interaction of FP2 with ELP was studied using recombinant GST-EZ protein. This protein was purified from bacteria transformed with a recombinant plasmid expressing 154 amino acids of ELP (one-third of ELP encompassing the zinc finger domain) fused to the C terminus of glutathione S-transferase (14). Significant binding of GST-EZ to FP2 was observed (Fig. 6A); however, the impossibility of supershifting the retarded complex with antibodies against ELP ruled out the hypothesis of ELP binding to the FP2 region of the ucp gene (data not shown). Accordingly, a syn-thetic oligonucleotide containing only one AGGTCA unit could not compete with the labeled FP2 probe, and Northern analysis under stringent conditions of brown adipocyte and liver mRNAs did not reveal ELP mRNA (data not shown).

Analysis of Region Overlapping 5'-Extremity of R2: Delineation of Hypersensitive Regions; Liver and BAT Nuclear Factors Differentially Bind to bp -5091-472 Region, Which Contains CACCC Elements-Using a probe corresponding to the bp -6W
-235 domain of the ucp promoter, two hypersensitive regions (region A, bp -509 to -472; and region B, bp -403 to -350) were detected during footprint analysis (Fig. 7). When the opposite strand of the probe was used, a bp -4671-430 footprint was obtained with BAT nuclear extracts and was inhibited in the presence of excess CIEBP oligonucleotide (data not shown); this bp -4671-430 footprinted region had been previously shown to contain a CIEBP-binding site (24). Both the hypersensitive regions A and B were particularly obvious when using nuclear extracts prepared from BAT. Band-shift analysis confirmed that nuclear factors can interact with the bp -5091-472 and -4031-350 boxes (Fig. 8). At the level of region A, several complexes were retarded, and a different picture was obtained when using liver factors or BAT factors. The pattern obtained with nonstimulated 1B8 cells, which do not transcribe the ucp gene, was almost similar to the pattern obtained with liver extracts; interestingly, a pattern similar to that obtained with BAT factors was observed when using nuclear extracts from 1B8 cells treated with norepinephrine ( Fig. 8 A ) , which therefore transcribe the ucp gene (Fig. 4). Competition was observed with an excess of homologous oligonucleotide. This 41-bp oligonucleotide contains several CACCC elements, and a 24-bp oligonucleotide containing almost only CACCC elements was also able to compete with the bp -5091-472 probe (Fig. 8 A ) . When shift DNA binding assay. Band-shift assays and supershift experiments were carried out as described under "Materials and Methods." The labeled FP4 oligonucleotide (sequence is given in Fig. 3) was incubated with liver nuclear extracts, BAT nuclear extracts, or recombinant v-ErbA protein expressed in baculovirus (1 pl of baculovirus extract corresponding to 30 ng of v-ErbA protein was used) in the presence (+) or absence (-) of antibodies against v-ErbA. using the labeled region B, competition was observed with the Etsl or FP1 oligonucleotide, the latter which contains an Etslbinding site (see above); accordingly, no competition was obtained when using oligonucleotides with mutations of the Etslbinding site (Fig. 8B). : Delineation of bp -1801-159, -1471-120, and  -1111-85 Boxes Binding NF-1, CREB, and  Respectively-Nuclear extracts prepared from BAT, liver, or kidney were used for DNase I footprint analysis with a DNA fragment that included nucleotides -241 to +53 of the ucp gene (Fig. 9). Two main regions a t bp -182 to -158 and bp -111 to -85 were protected from DNase I by nuclear proteins; a more discrete box was also observed between bp -147 and -120. No competition was obtained with the A F ' 1 or HNF-1 oligonucleotide, whereas the AFP4 oligonucleotide containing HNF-1-, NF-1-, and C/EBP-binding sites competed with the probe at the level of the window at bp -182 to -158. Obvious competition was observed a t this level with an NF-1 oligonucleotide, demonstrating that the bp -180/-158 box is a binding site of a factor related to NF-1. Accordingly, the FP1 oligonucleotide, which contains an NF-1-binding site, competed. In fact, the nucleotide sequence of this box (AGGAN,GCCAA) is very close to that described for the NF-1-binding site (T/CGGA/CN,GCCAA) (17). Another region identified in DNase I footprint experiments (bp -111 to -85) contains a potential binding site for S p l (ACGCCC) (17). The binding of NF-1 to the bp -1821-158 box was confirmed using band-shift analysis (Fig. 10). This type of analysis also revealed an Spl-binding site in the bp -111/-85 box. The third box identified between these two regions and located between bp -147 and -120 contains a TGACG sequence closely related to that of CRE. Use of an oligonucleotide corresponding to the CREB-binding site in band-shift experiments confirmed the presence of a CRE between bp -147 and -120 of the ucp gene minimal promoter (Fig. 10).

DISCUSSION
This work was based on the previous identification (7) of three regions essential to ucp promoter activity: a 211-bp enhancer located at kilobase -2.4 (Rl), a more proximal region containing a silencing element (R2), and a minimal promoter (bp -157 to -57). Similar regions were also recently identified in the murine ucp gene (8). The main goal of this work was a detailed in vitro analysis of R1, R2, and regions flanking R2 to delineate short DNAregions of interest and to identify putative transactivating proteins on which future studies will be based.

Role for Etsl, R X R , and T3 Receptor at Level of R1 Enhancer
Element-A partial alignment of the rat R1 enhancer (4)  FIG. 9. Footprint analysis of protein-DNA interactions at ucp gene R2 proximal domain, and minimal promoter reveals three boxes: bp -182 to -158, -147 to -120, and -111 to -85. DNase I footprinting of the UCP 5'-flanking proximal region was carried out with the bp -241/+53 DNA fragment. The probe was incubated with nuclear extracts from rat liver and treated with DNase I under the conditions described under "Materials and Methods." A similar pattern of footprinting was obtained when using brown fat or kidney nuclear extracts. The same windows were observed when using the opposite strand of the probe. Numbers above the lanes correspond to micrograms of nuclear proteins. Competition experiments were carried out in the presence of a 200-fold mass excess of nonradioactive FP1, NF-1, AFP4, HNF-1, or AP1 oligonucleotide. Vertical burs delineate footprints. comp, competitor.
the mouse enhancer (8) is given in Fig. 11, which shows a high level of conservation (86% identity). In fact, the FP1 domain of the rat gene is homologous to the CRE-2hrown fat regulatory element 1 element in the mouse gene, and rat FP2 is homolo-gous to the mouse EcoRV-XbaI region also termed the Xmrown fat regulatory element 2 element (8). We have shown that the rat homolog of mouse brown fat regulatory element 1 can bind an Etsl-related protein; such binding had been previously speculated by Kozak et al. (8). Interestingly, mutagenesis of 2 nucleotides in the mouse homolog Etsl region resulted in 95% reduction in CAT expression of the entire ucp promoter transfected into the norepinephrine-stimulated brown adipocyte cell line B-7 (8). It can therefore be concluded that this Etsl-binding site plays a crucial role in ucp gene transcription. In the rat DNA region that is homologous to murine brown fat regulatory element 2 and that also contains a TTCC motif, we were unable to detect Etsl binding. The Ets oncogene family encodes a class of sequence-specific DNA-binding proteins that are involved in cell growth and differentiation (18,25).
We have attached either the multimerized FP1 or FP2 region to the minimal promoter of the rat ucp gene and to the cat gene. When each plasmid was transfected into brown adipocytes, only a weak CAT transcription could be measured, much lower than that observed with the entire 211-bp R1 enhancer (data not shown). A similar observation was made by Kozak et al. (8). Therefore, it can be concluded that the full activity of the enhancer requires both FP1 and FP2 elements.
The main feature of the 33-bp footprinted FP2 element is the presence of a canonical AGGTCA sequence common to the binding site of all members of the superfamily of nuclear hormone receptors. Although the FP2 probe interacted with a fusion protein containing one-third of ELP, experiments with anti-ELP antibodies did not confirm a possible role for ELP. FP2 contains a potential binding site for ELP; however, this site is unlikely to bind ELP since this protein is known to bind a DNA element that does not contain repeated units (16). In fact, FP2 contains a potential DR3 sequence, which is generally considered as a potential binding site for the receptor of vitamin D, (19,26) receptor; an alignment of the rat FP1 and FP2 domains (this work) with corresponding regions of the mouse ucp promoter ( 8 ) is given. A and R indicate hypersensitive zones particularly obvious during DNase I experiments in the presence of BAT nuclear factors; box A contains a repeated CACCC element previously identified in the 6-globin promoter (29), and box B could bind protein related to Etsl. Footprinted boxes C-E can bind NF-1, CREB, and Spl, respectively. Two C/EBP-binding sites were previously detected (24).
to response elements containing directly repeated motifs separated by 2, 3, 4, or 5 bp, and they concluded that there is no simple rule defining the binding specificities of RxRs and RARs. RXRs are known to bind DNA by forming heterodimers with RARs, T, receptors, vitamin D, receptors, or peroxisome proliferation-activated receptors. On one hand, we report here that all-trans-retinoic acid stimulates ucp mRNA production in brown adipocytes, and on the other hand, it is known that T, has a positive effect on ucp gene transcription (9,281. Moreover, neither clofibrate nor 1,25-hydroxycholecalciferol activated ucp gene transcription or R1 enhancer-driven CAT transcription in brown adipocytes.3 Thus, although the RXR-binding element present in FP2 has a DR3 structure that is a priori distinct from that of the T, receptor-binding box, it was reasonable to investigate a role for the thyroid hormone receptor. Experiments described here (Fig. 5) establish that the T, receptor can bind the ucp promoter i n vitro, probably through forming a heterodimer with RXR.

Silencer R2 and Flanking Domains: Region Containing Repeated CACCC Elements Interacts Differentially in Vitro with
Liver or BAT Nuclear Extracts and Could Be Involved in ucp Gene Dunscription Znduction-In addition to CiEBP-binding sites at bp -457 to -440 and bp -335 to -318 (24), this region was shown here to contain five potential regulatory elements termed A-E (Fig. 11). Boxes A and B, at bp -509 to -472 and bp -403 to -350, respectively, were detected as hypersensitive regions during DNase I footprint analysis and were shown to be able to bind proteins by using band-shift analysis. Analysis of sequence corresponding to the A and B boxes did not reveal any potential binding site, except for a CCACACCC sequence at bp -495 in region A previously described in the P-globin promoter (29). In fact, two copies of the CACCC sequence plus one inverted repeat are present in region A, and it has been shown that such copies may function as important general transcription factors and may also interact with other transcription factors (30).
In other respects, the A box is the only region in the ucp promoter that did not bind liver and BAT factors in a similar manner. As previously observed with HIB 1B cells (10, ll), 1B8 cells do not transcribe the ucp gene, but the addition of norepinephrine strongly activates this transcription (see Fig. 4).
Here it was shown that the induction of ucp gene transcription in 1B8 cells activates proteins able to interact with the A box (which does not contain CRE) in a manner similar to that observed with nuclear extracts from BAT. Therefore, a detailed analysis of the A box could constitute a strategy toward the understanding of the cell-specific transcription of the ucp gene.
In this work, we did not intend to identify and demonstrate the functional importance of CRE, as has been done recently for the mouse ucp gene (8). The CREB-binding site identified in the minimal promoter of the rat ucp gene at bp -147 to -120 (3'-GCGCGTCA-5') is probably equivalent to CRE-4 identified in the minimal promoter of the murine gene (5'-GCGCGTCA-3') and is considered to have more of a promoter function than an enhancer function (8).
In conclusion, several cis-elements and nuclear factors controlling the ucp promoter have been proposed. Two small regions, FP1 and FP2, were identified inside the R1 enhancer using DNase I footprint and band-shift assays. The FP1 element was shown to be able to bind in vitro factors related to NF-1 and Etsl (in a cooperative manner), whereas FP2 can bind RXR and thyroid hormone receptors. Two DNase I-hypersensitive sites, A and B, particularly obvious in the presence of BAT nuclear extracts, were detected on both sides of the 5'extremity of R2; liver and BAT nuclear extracts bind differentially to box A, which contains repeated CACCC elements and could play a crucial role in ucp gene transcription. Binding of an Etsl-related protein to box B of R2 was suggested. Binding sites for NF-1, CREB, and Spl factors were detected in the proximal part of R2 and in the minimal promoter. The next step will be the analysis of the functional importance of precise cis-elements and trans-factors involved in the specific expression of UCP in brown adipocytes.