A Thyroid Hormone Response Unit Formed between the Promoter and First Intron of the Carnitine Palmitoyltransferase-I (cid:1) Gene Mediates the Liver-specific Induction by Thyroid Hormone*

Carnitine palmitoyltransferase-I (CPT-I) catalyzes the rate-controlling step of fatty acid oxidation. CPT-I con-verts long-chain fatty acyl-CoAs to acylcarnitines for translocation across the mitochondrial membrane. The mRNA levels and enzyme activity of the liver isoform, CPT-I (cid:1) , are greatly increased in the liver of hyperthyroid animals. Thyroid hormone (T3) stimulates CPT-I (cid:1) transcription far more robustly in the liver than in non-hepatic tissues. We have shown that the thyroid hormone receptor (TR) binds to a thyroid hormone response element (TRE) located in the CPT-I (cid:1) promoter. In addition, elements in the first intron participate in the T3 induction of CPT-I (cid:1) gene expression, but the CPT-I (cid:1) intron alone cannot confer a T3 response. We found that deletion of sequences in the first intron between (cid:2) 653 and (cid:2) 744 decreased the T3 induction of CPT-I (cid:1) . Upstream stimulatory factor (USF) and CCAAT enhancer binding proteins (C/EBPs) bind to elements within this region, and these factors are required for the T3 response. The binding of TR and C/EBP to the CPT-I (cid:1) gene in vivo was shown by the chromatin immunoprecipitation assay. We determined that TR can physically interact with USF-1, USF-2, and C/EBP (cid:1) . Transgenic mice were created that carry CPT-I (cid:1) -luciferase transgenes with or without the first intron of the CPT-I (cid:1) gene. In these mouse lines, the first intron is required for T3 induction as well as high levels of hepatic expression. Our data indicate that the T3 stimulates CPT-I (cid:1) gene expression in the liver through a T3 response unit of the TRE in the and C/EBP and USF, bound in the first intron.

ness (11,12). Interestingly, the first intron of the CPT-I␣ gene is also necessary for full T3 induction (11). Removal of the first intron reduces the T3 induction by 80%. Induction of CPT-I␣luciferase by T3 is more robust in HepG2 hepatoma cells compared with L6 myoblasts and cardiac myocytes. The first intron is required for the induction in transfected HepG2 hepatoma cells and not in L6 myoblasts and cardiac myocytes, suggesting that the intron contributes to the enhanced T3 induction in the liver (11). The goal of the present study was to examine more extensively the role of the first intron in liver-specific T3 induction of CPT-I␣. We identified regions within the first intron that are necessary to achieve the full T3 induction. Elements within these regions bind proteins that participate in the T3 response, including CCAAT enhancer-binding proteins (C/ EBP) and upstream stimulatory factor (USF-1 and USF-2). TR physically interacts with USF-1, USF-2, and C/EBP␣. Our data suggest that C/EBP is responsible for the liver-specific component of the induction of CPT-I␣ by thyroid hormone. Our findings show that the TR in the promoter and C/EBP and USF bound in the first intron comprise a T3 response unit that mediates the liver-selective T3 induction of CPT-I␣.

EXPERIMENTAL PROCEDURES
Electrophoretic Mobility Shift Assays-CPT-I␣ probes for electrophoretic mobility shift assays were created by labeling double-stranded oligonucleotides using Klenow enzyme and [␣-32 P]dCTP. The oligonucleotides contained sequences from the first intron and XbaI or MluI overhangs. Double-stranded unlabeled wild-type and mutant oligonucleotides were used as competitors (See Table I for oligomer sequences). Rat liver nuclear extract was prepared as described (13). The protein-DNA binding mixtures contained labeled probe (30,000 cpm) and proteins isolated from rat liver nuclei in 80 mM KCl, 25 mM Tris-HCl (pH 7.4), 0.1 mM EDTA, 1 mM dithiothreitol, and 10% glycerol (11). Poly(deoxyinosine-deoxycytidine) (double-stranded homopolymer) was added to each binding reaction as a nonspecific competitor. Antibodies for TR␣/␤, USF-1, USF-2, Sp1, C/EBP␣, C/EBP␤, Oct-1, COUP-TF, and CREB (Santa Cruz) were added to binding reactions for supershift assays. Binding reactions were incubated at room temperature for 20 min and resolved on 5% non-denaturing acrylamide gels (80:1, acrylamide/bisacrylamide) in Tris-glycine running buffer (22 mM Tris and 190 mM glycine). Electrophoresis was carried out at 180 volts for 80 min at 4°C (11). Sequence analysis of the intron for potential transcription factor binding sites was performed using the TESS transcription factor search.
Transient Transfection of Luciferase Vectors-CPT-I␣-luciferase constructs were transiently transfected into HepG2 cells by the calcium phosphate method (9). Transfections included 3 g of CPT-I␣-luciferase vectors along with RSV-TR␤ and TK-renilla vectors. Cells were transfected in Dulbecco's modified Eagle's medium (DMEM) containing 5% calf serum/5% fetal calf serum and incubated overnight at 37°C. Following two washes with phosphate-buffered saline, the medium was replaced by DMEM containing no serum. Cells were treated with 100 nM T3 for 24 h. After T3 treatment, cells were washed twice with phosphate-buffered saline and lysed in passive lysis buffer (Promega). Cell lysates were frozen and thawed to facilitate cell lysis. Luciferase assays were conducted on extracts from cells in serum-free media and cells treated with T3. Both luciferase and renilla activity was measured. Protein content in each lysate was determined by Bio-Rad protein assay (Bio-Rad). Luciferase activity was corrected for both protein content and renilla activity to account for cell density and transfection efficiency, respectively. Data are expressed as -fold induction of luciferase in cells exposed to T3 as compared with cells that received no hormone (Fig. 3) or relative induction of luciferase in mutant construct as compared with the induction of wild-type Ϫ4495/ϩ1240 CPT-I␣-luciferase (Figs. 1, 4, and 6).
GST-pull-down assays GST and GST⅐C/EBP␣ were prepared as described by Yin et al. (14). 35 S-labeled chicken TR␣ and human RXR␣ were expressed using TNT reticulocyte lysates (Promega). GST-pull down assays were conducted as described (14). Bound proteins were eluted and resolved by SDS-PAGE and visualized by storage phosphor autoradiography. GST⅐TR␣ fusion protein was expressed in Escherichia coli BL21(DE3) (Novagen) and purified using glutathione-Sepharose (Amersham Biosciences) according to the manufacturer's protocol (14). 35 S-labeled USF-1 and USF-2 were expressed using PROTEINscript II reticulocyte lysates (Ambion). Bound proteins were eluted with 2.5 mM glutathione and resolved by SDS-PAGE.
Production and Characterization of CPT-I␣-Luc Transgenic Mouse Lines-Restriction fragments containing Ϫ6938/ϩ1240 or Ϫ6839/ϩ19 CPT-I␣-Luc genes were isolated and injected into the pronuclei of one-cell C57BL/6J ϫ SJL/JF2 hybrid (B6/SJL) mouse zygotes to produce transgenic mice (17,18). The founders were bred with B6/SJL mates to obtain a second generation of transgenics. Mice containing the luciferase gene constructs were identified by Southern analysis of genomic DNA obtained from the tail (17). Two transgenic mouse lines of each CPT-I␣ gene construct were selected for use in the T3 experiments based on initial characterization for transgene expression (data not shown). Mice were injected with 0.33 mg/Kg body weight of triiodothyronine (T3) 24 and 48 h prior to sacrifice (6). The mice were sacrificed, and pieces of the liver and heart were homogenized in luciferase assay buffer (Promega). Luciferase assays and protein determinations were conducted as described above. Fig. 1 illustrates the Ϫ4495/ϩ1240 CPT-I␣-luciferase vector containing 4495 base pairs of the promoter, exon 1, intron 1, and a segment of exon 2. Deletion of nucleotides between ϩ199/ϩ707 in the first intron reduced the T3 induction of CPT-I␣ by 50% ( Fig.  1). To define smaller regions in the intron that contribute to the T3 stimulation, additional internal deletions were made in the context of the Ϫ4495/ϩ1240 CPT-I␣-Luc vector. Removal of the ϩ515/ϩ707 region decreased the T3 response ϳ50%, as did deletion of the 80-base pair ϩ628/ϩ707 region (Fig. 1). Deletion of the ϩ628/ϩ707 region did not alter the basal expression of the CPT-I␣-Luc vector (data not shown). These data indicate that sequences within the ϩ628/ϩ707 region are required for a full response to T3.

Identification of Regions within the ϩ199/ϩ707 Region of the First Intron Required for the Full T3 Response-
Binding of Transcription Factors to the ϩ628/ϩ707 Region of the Intron-Our next studies focused on the ϩ628/ϩ707 region of the first intron. Electrophoretic mobility shift assays were conducted using double-stranded oligonucleotides that corresponded to nucleotides ϩ628/ϩ655, ϩ653/ϩ682, and ϩ674/ ϩ707 in the CPT-I␣ gene. Only the ϩ653/ϩ683 and ϩ674/ϩ707 regions bound proteins isolated from rat liver nuclei (Fig. 2, A  and B). A consensus E-box motif (CANNTG) was found at nucleotides ϩ659/ϩ664. The E-box motif binds a family of proteins that contain helix-loop-helix and leucine zipper dimerization domains, including c-Myc, sterol regulatory element binding protein (SREBP), and upstream stimulatory factor (19). Supershift assays were conducted using antibodies that recognize Sp1, C/EBP␤, Oct-1, USF-1, and USF-2. Protein binding to the ϩ653/ϩ682 region was completely disrupted by USF-1 and USF-2 antibodies, whereas the other antibodies did not alter the binding of nuclear proteins ( Fig. 2A). Western blot analysis confirmed that USF-1 and USF-2 are present in RLNE as well as in HepG2 cells, which was the cell type used in transient transfection experiments (data not shown). Competition analysis revealed that a 100-fold excess of unlabeled wild type ϩ653/ϩ682 oligomer completely competed for nuclear protein binding to the labeled probe, whereas an oligomer that contained a mutation in the E-box motif (Table I, Mut. #1) was unable to compete for protein binding ( Fig. 2A). However, competition with unlabeled oligomer that contained a mutation 3Ј to the E-box (Mut. #2) reduced protein binding as effectively as the wild type oligomer. Our data show that USF-1 and USF-2 bind within intron 1 of CPT-I␣ at the E-box motif located at ϩ659/ϩ664.
Gel shift mobility assays were conducted with an oligomer representing the ϩ674/ϩ707 region. Several complexes were formed indicating that either multiple proteins or a family of proteins bind at this site ( Fig. 2B). We analyzed the binding of nuclear factors to this site by the addition of antibodies to the gel shift assays. Antibodies to C/EBP␣ and C/EBP␤ disrupted the binding of nuclear factors (Fig. 2B). COUP-TF and TR antibodies did not alter protein binding. The ϩ677/ϩ689 region contains an AGGTCA-like motif that might interact with nuclear receptors. However, antibodies to hepatocyte nuclear factor-4, RXR, and peroximal proliferator-activated receptor ␣ did not alter the binding of nuclear proteins (data not shown). Our results indicated that C/EBP proteins could bind to this site. Competition analyses were conducted using unlabeled oligomers that corresponded to the wild type ϩ674/ϩ707 sequence as well as oligomers that contained mutations across the ϩ674/ϩ707 region. A 100-fold excess of wild type oligomer or an oligomer that contained a mutation in nucleotides ϩ695/ ϩ700 (Mut. #7) competed effectively for protein binding to the labeled oligomer. However, oligomers that contained mutations in the ϩ677/ϩ682 (Mut. #3 and #5) and ϩ684/ϩ689 (Mut. #4 and #6) sites did not compete for protein binding. Therefore, we conclude that nucleotides within the ϩ677/ϩ689 element are necessary for protein binding within the ϩ674/ϩ707 region.
Contribution of the ϩ628/ϩ707 USF and C/EBP Binding Regions to the T3 Response-To investigate the importance of elements within the ϩ628/ϩ707 region, we removed these sites from the Ϫ4495/ϩ1240 CPT-I␣-luciferase vector by making ⌬ϩ680/ϩ707 and ⌬ϩ653/ϩ707 deletions within the intron.
FIG. 1. Localization of regions within the first intron that enhance T3 responsiveness. HepG2 cells were transiently transfected with 3 g of Ϫ4495/ϩ1240 CPT-I␣-Luc constructs, 1 g of RSV-TR␤, and 1 g of TK-renilla. The deleted sequences in the first intron are indicated by the symbol (⌬). For hormone treatment, cells were incubated either in serum-free DMEM or DMEM containing 100 nM T3 for 24 h. All transfections were performed in duplicate and repeated four to six times. Luciferase and renilla assays were performed in the same tube. Luciferase activity was corrected for both protein content and renilla activity. Results are expressed as the relative induction by T3 Ϯ S.E. by comparing the T3 induction of vectors with deletions to the wild type vector, which was assigned a value of one. *, p Ͻ 0.05 versus full-length Ϫ4495/ϩ1240 CPT-I␣-luciferase.
These vectors were transiently transfected into HepG2 cells along with RSV-TR␤. Deletion of the ϩ680/ϩ707 and the ϩ653/ ϩ707 regions caused a 35% reduction in the T3 induction (Fig.  3), which was identical to the ⌬ϩ628/ϩ707 vector in this set of experiments (data not shown). These data demonstrated that the elements contributing to the T3 response are located in the ϩ680/ϩ707 region.
Role of the ϩ707/ϩ1066 Region in T3 Induction of CPT-I␣-In addition to the ϩ653/ϩ707 region, sequences between ϩ707 and ϩ1066 also participated in the T3 induction (Fig. 4).
To assess the contribution of the 3Ј-end of the intron to the T3 induction of CPT-I␣, serial deletions were created from the second exon in the Ϫ4495/ϩ1240 CPT-I␣-luciferase vector. These vectors were cotransfected with RSV-TR␤ into HepG2 cells and tested for T3 responsiveness (Fig. 4). The full T3 effect was maintained with deletion of nucleotides ϩ803 to ϩ1240. However, the T3 response decreased upon deletion of the additional nucleotides between ϩ803 and ϩ707. Deletion of the ϩ707/ϩ1240 region modestly decreased basal expression of the gene (data not shown). This stimulation was reduced further upon deletion of the intron to ϩ199. An internal deletion of the ϩ707/ϩ1066 region also diminished the T3 response by 40%. These findings show that sequences between ϩ707 and ϩ803 of intron 1 are involved in the enhancement of T3 induction.
We ligated the ϩ707/ϩ1066 sequences in front of the SV40 promoter driving the luciferase reporter gene. The reporter vector contained a Gal4 binding site. The Gal4-SV40-luciferase vectors were transfected with an expression vector for Gal4-TR␤ in which the DNA binding domain of the TR␤ was replaced with the DNA binding domain of Gal4 (20). Inclusion of the ϩ707/ϩ1066 region enhanced the T3 response 2.7-fold compared with Gal4-SV40-luciferase (Fig. 5A). Addition of the ϩ707/ϩ810 sequences allowed an additional 6.8-fold induction of the Gal4-SV40-luciferase vector. These results further demonstrate that factors binding in the ϩ707/ϩ810 region enhance the T3 induction of CPT-I␣.
Binding of USF within the ϩ707/ϩ810 Region of the Intron-Our next experiments characterized the binding of nuclear proteins within the ϩ707/ϩ803 region of the first intron. We designed three oligomers that spanned this region of the gene. Using these oligomers in gel shift mobility assays, we found that only the oligomer corresponding to the ϩ700/ϩ744 region bound proteins from rat liver nuclear extract (Fig. 5B). This region contains an E-box element. Antibodies to USF-1 and USF-2 were able to supershift the binding to this site, indicating that USF proteins can bind to this element. Competition analysis using a 100-fold excess of unlabeled wild-type and mutant oligomer that contained an altered USF binding site at FIG. 2. C/EBP and USF bind in the ؉653/؉707 region. A and B, doublestranded oligonucleotides were constructed that encompassed the ϩ653/ ϩ683 and ϩ674/ϩ707 regions of the intron. In the upper panels, the 32 P-radiolabeled double-stranded oligomers were incubated with rat liver nuclear extract (RLNE) and the antibodies (Ab) indicated. Electrophoretic mobility shift assays were conducted as described under "Experimental Procedures." In the lower panels, competition assays were conducted using double-stranded unlabeled wild type (WT) and mutant (#) oligomers. A 100-fold excess of the competitor oligomers was added to the competition assays. The oligonucleotide sequences are given in Table I. ϩ724/ϩ729 confirmed that this E-box motif is necessary for the binding of nuclear proteins (data not shown). Previously, we had identified three sites in the ϩ800/ϩ1066 region that bound nuclear proteins by DNase footprint analysis (11). The ϩ824/ ϩ842 element was analyzed in gel shift mobility assays. Several factors were able to bind to this site (Fig. 5B). Addition of antibodies to C/EBP␣ and -␤ to the binding reaction disrupted the binding of proteins to this element.
To determine which sites in the ϩ653/ϩ850 region are important for the T3 induction, we disrupted each by site-directed mutagenesis. The sites were altered in the context of Ϫ4495/ ϩ1240 CPT-I␣-Luc by introducing the mutations that had been FIG. 3. C/EBP and USF binding sites participate in the T3 induction. HepG2 cells were transiently transfected with 3 g of CPT-I␣-Luc vector, 1 g of RSV-TR␤, and 1 g of TK-renilla. The Ϫ4495/ϩ1240 CPT-I␣-luciferase vectors contained deletions in the intron. Cells were treated for 24 h with 100 nM T3 and luciferase assays performed as described in Fig. 1. Experiments were conducted in duplicate and repeated four to six times. Results are expressed as-fold induction by T3 Ϯ S.E. by comparing luciferase activity of untreated cells with that of cells exposed to hormone. †, p Ͻ 0.005 versus full-length Ϫ4495/ϩ1240 CPT-I␣-luciferase.
FIG. 4. The ؉707/؉1240 region contributes to the T3 response. HepG2 cells were transiently transfected with 3 g of Ϫ4495/ϩ1240 CPT-I␣-Luc or serial deletion constructs, 1 g of RSV-TR␤, and 1 g of TK-renilla. The ovals represent elements that were previously identified as transcription factor binding sites by DNase I footprinting (11). Cells were treated with 100 nM T3 in serum-free DMEM for 24 h before cells were harvested. Luciferase assays were performed as described in the legend to Fig. 1. All transfections were conducted at least four times. Results are expressed as relative induction Ϯ S.E. by comparing the T3 induction of the vectors containing deletions with the Ϫ4495/ϩ1240 CPT-I␣ Luc, which was assigned a value of one. *, p Ͻ 0.05 versus full-length Ϫ4495/ϩ1240 CPT-I␣-luciferase.

FIG. 5. Contribution of the ؉707/ ؉810 region to the T3 induction. A,
Gal4-SV40-Luc vectors were constructed that contained the ϩ707/ϩ1066 and the ϩ707/ϩ810 regions of the intron. HepG2 cells were cotransfected with 3 g of Gal4-SV40-Luc, 100 ng of Gal4-TR␤ expression vector, and 1 g of TK-renilla. Cells were treated for 24 h with 100 nM T3. Luciferase assays were conducted as described in the legend to Fig. 1. All transfections were repeated four times in duplicate. Results are expressed as relative induction by T3 Ϯ S.E. compared with Gal4-SV40-Luc. *, p Ͻ 0.05 versus full-length. The Gal4-SV40-Luc was induced 7-fold by T3. B, electrophoretic mobility shift assays were conducted using 32 P-labeled oligomers that corresponded to the ϩ700/ϩ744 and the ϩ824/ϩ842 regions of the intron. Oligomers were incubated with proteins isolated from rat liver nuclei (RLNE). Supershift assays were conducted using antibodies (Ab) to USF-1, USF-2, C/EBP␣, C/EBP␤, COUP-TF, and CREB.
shown to disrupt binding in gel shift mobility assays. Mutation of either USF binding site and the ϩ677/ϩ689 C/EBP site decreased the T3 induction, strongly suggesting that these factors contributed to the T3 induction (Fig. 6). Disruption of the ϩ827/ ϩ842 C/EBP binding site did not alter the T3 induction, indicating that not all C/EBP binding sites contribute to the response of the CPT-I␣ gene to T3. Our data show that USF and C/EBP are accessory factors in the T3 induction of the CPT-I␣ gene.
To determine the TR motif through which the physical interaction with C/EBP␣ occurs, we conducted pull-down assays using truncated 35 S-labeled TR␣ proteins and GST⅐C/EBP␣. Removal of the first 50 amino acids had no effect on the interaction between TR and C/EBP (Fig. 7C). Further deletion of amino acids through 120 completely abolished binding to C/EBP␣. Isolated polypeptides corresponding to amino acids 1-118 and 1-157 interacted with C/EBP␣. However, removal of the residues 1-50 diminished its ability to interact with C/EBP␣. Our results indicate that the interaction between TR␣ and C/EBP␣ occurs through a region of the TR that encompasses the DNA binding domain and is independent of T3. We have also found that TR␤ can interact with C/EBP␣ (data not shown). Physical interactions between TR and the accessory factors, USF and C/EBP, may contribute to the T3 induction of the CPT-I␣ gene.
In Vivo Binding of TR and C/EBP to the CPT-I␣ Gene-Previously, we showed that TR binds to the CPT-I␣ TRE in vitro (11). To investigate if such binding occurs in vivo we conducted chromatin immunoprecipitation (ChIP) assays. Rat hepatocytes were treated with 1% formaldehyde to cross-link DNA and proteins. Immunoprecipitations were performed using an antibody that recognized both the TR␣ and -␤ isoforms as well as antibodies to C/EBP␣ and C/EBP␤. IgG was used as a control in these experiments. PCR reactions were conducted using primer sets that corresponded to nucleotides Ϫ6473/ Ϫ6450 and Ϫ6076/Ϫ6099, Ϫ3079/Ϫ3056 and Ϫ2802/Ϫ2825 within the CPT-I␣ promoter as well as ϩ200/ϩ220 and ϩ750/ ϩ730 within intron 1. The promoter primers Ϫ3079/Ϫ3056 and Ϫ2802/-2825 encompassed the CPT-I␣ TRE, which is located at nucleotides Ϫ2938/Ϫ2923. Antibodies to the TR, C/EBP␣, and C/EBP␤ immunoprecipitated sequences in the promoter and intron 1 (Fig. 8). IgG failed to pull down promoter or intron sequence. The Ϫ6473/Ϫ6450 and Ϫ6076/Ϫ6099 primers, which were our upstream controls, produced no PCR product in our experiments. Our results show that TR and C/EBP interact with sequences within intron 1 of the CPT-I␣ gene and the promoter at the TRE region.
Regulation of CPT-I␣-Luciferase Genes in Transgenic Mice-To test whether the CPT-I␣ intron was important for the T3 response in vivo, we created transgenic mice that expressed either the Ϫ6839/ϩ1240 or Ϫ6839/ϩ19 CPT-I␣-Luc transgenes. We initially characterized five independent transgenic lines with the 6839/ϩ1240 CPT-I␣-Luc transgenes and two lines with the Ϫ6839/ϩ19 CPT-I␣-Luc transgenes for liver-  Table I  GST proteins were immobilized on glutathione-conjugated-Sepharose beads. Binding reactions were conducted in the presence (ϩ) or absence (Ϫ) of 100 nM thyroid hormone (T3). Eluted proteins and 10% of the radiolabeled USF-1 or USF-2 used in each binding reaction (10% Input) were resolved by SDS-PAGE and visualized by storage phosphor autoradiography. B, pull-down assays were also conducted with 35 Slabeled TR␣ or RXR␣ and bacterially expressed GST and GST⅐C/EBP␣ fusion protein in the presence (ϩ) or absence (Ϫ) of 1 M T3 or 9-cisretinoic acid. C, additional pull-down assays were conducted using truncated forms of 35 S-labeled TR that were incubated with GST and GST⅐C/EBP␣ proteins immobilized to glutathione-conjugated-agarose. On the left panel, 10% of the 35 S-TR in the binding reaction is shown. On the right panel, the 35 S-TR pulled down by GST or GST⅐C/EBP␣ is shown. Binding reactions were conducted in the presence or absence of 1 M T3. Bound proteins were resolved by SDS-PAGE and visualized by storage phosphor autoradiography. specific expression of luciferase (data not shown). From the initial seven transgenic lines, we tested two independent transgenic mouse lines expressing each luciferase reporter gene for the present studies. Several important observations were made regarding the regulation of the CPT-I␣ gene using these mice. First, the expression of the Ϫ6839/ϩ1240 CPT-I␣-Luc gene is at least 100-fold higher than the expression of Ϫ6839/ϩ19 CPT-I␣-Luc transgenes that do not contain the intron. The basal expression of lines one and two were 2.1 Ϯ 0.3 and 1.1 Ϯ 0.7 units/mg protein, whereas lines three and four expressed 0.02 Ϯ 0.02 and 0.01 Ϯ 0.1 units/mg protein, respectively (Fig.  9). These data suggest that the intron is important for the basal expression of the gene in the liver. Also, the expression of the Ϫ6839/ϩ1240 CPT-I␣-Luc gene is much greater in the liver than in the heart (data not shown). Second, both of the lines expressing CPT-I␣-Luc genes containing the intron responded to T3. Line one was induced 3-fold, whereas line two was increased 9-fold ( Fig. 9). However, the expression of the Ϫ6839/ ϩ19 gene was so low in lines 3 and 4 that the extent of the T3 induction was difficult to evaluate. These data are consistent with the concept that sequences within the intron are vital for hepatic expression of the CPT-I␣ gene in vivo.

DISCUSSION
In this study, we have examined the mechanisms by which T3 induces the CPT-I␣ gene. We show that the first intron is required for the basal expression and T3 induction of CPT-I␣ in the liver. We determined that the CPT-I␣ gene contains a T3 response unit consisting of a TRE at nucleotides Ϫ2938/Ϫ2923 in the promoter, USF binding sites at ϩ659/ϩ664 and ϩ724/ ϩ729, and a C/EBP binding site at 677/ϩ689. Each of these binding sites is independently required to obtain a full T3 response. The importance of accessory factors in the hormonal regulation of gene expression is becoming increasingly apparent (21). Accessory factors may contribute to the control of gene expression by modulating the actions of liganded receptors, recruiting coactivators to the promoter, and in the case of the CPT-I␣ gene enhancing the T3 induction of CPT-I␣ in the liver.
The first component of the T3 response unit is the CPT-I␣-TRE. We have found that this TRE contains a DR4 motif which binds purified TR-RXR heterodimers (11). In the current studies, we have used the ChIP assay to show that TR␤ is associated with the CPT-I␣ gene in vivo. Utilizing the ChIP assay, Fondell and coworkers have demonstrated that the TR is associated with the TREs of genes in the presence and absence of T3 (22). The addition of ligand leads to association of coactivators with the TR (22). The CPT-I␣-TRE is contained within a DNase I hypersensitive site in the CPT-I␣ promoter, indicating that the TRE is in a transcriptionally active region (23). Louet and et al. (23) identified a CREB binding site immediately adjacent to the TRE. In addition, there is a DR1 element located within 100 base pairs of the TRE that binds both peroxisomal proliferator-activated receptor ␣ and HNF-4 (23,24). Also, they identified a C/EBP binding site within the hypersensitive region (24). In the liver, the TR␤ is more highly expressed than TR␣ (25). To determine whether the TR␤ was more effective than TR␣ in mediating a T3 induction, we cotransfected CPT-I␣-Luc with expression vectors for both TR␣ and TR␤. Both TR isoforms induced CPT-I␣-Luc to a similar extent, indicating that the greater induction of CPT-I␣ by T3 in the liver than in the heart is not due to the preponderance of the TR␤ isoform (data not shown).
The second component of the CPT-I␣ T3 response unit is C/EBP. Two isoforms of C/EBP (␣ and ␤) are expressed in a variety of tissues including liver, lung, adipose, and intestine (26,27). C/EBP␣ is not expressed in the heart, and C/EBP␤ is expressed only at low levels (26). Both C/EBP isoforms contribute to the regulation of gene expression by a variety of hormones including T3, glucocorticoids, cAMP, and insulin (16,21,28,29). Previously, we have found that both C/EBP␣ and -␤ participate in the T3 induction of the PEPCK gene (16). It has been shown that C/EBP␣ is important for the T3 induction of malic enzyme (30). These results raise the possibility that C/EBP proteins are accessory factors for multiple hepatic genes that are stimulated by T3. T3 induces C/EBP␣ gene expression, and there is a TRE in the promoter of the C/EBP␣ gene (31). C/EBP is essential for the tissue-selective induction of CPT-I␣ in the liver. However, we believe that T3 stimulates CPT-I␣ in Immunoprecipitations were conducted using antibodies that recognize C/EBP␣, C/EBP␤, and TR. Immunoprecipitations with IgG were used as controls. Immunoprecipitated proteins were incubated with protein A-Sepharose. Precipitated ChIP products were used in PCR reactions. Input lanes contain PCR products from reactions using total cross-linked DNA and protein as a template along with specific primer sets. PCR products were resolved on ethidium bromide-stained 3% NuSeive-agarose gels. Sequences of PCR primers are listed under "Experimental Procedures." Experiments were performed three times with identical results. association with C/EBP proteins that are already bound to the intron of the CPT-I␣ gene.
C/EBP proteins have a crucial role in the regulation of gluconeogenesis. C/EBP␣ null mice die at birth from hypoglycemia and other complications (32). The C/EBP␤ null mice have a complicated phenotype in which 50% die shortly after birth and the remaining mice survive to adulthood (33). Our data indicates that C/EBP proteins regulate some aspects of CPT-I␣ gene expression and suggest that as in gluconeogenesis C/EBPs may contribute to the regulation of hepatic ketogenesis because CPT-I␣ is a rate-controlling step in this process.
The third component of the CPT-I␣ T3 response unit is USF-1 and USF-2. These factors are expressed ubiquitously (35). F. B. Hillgartner and associates (34) reported that COUP-TF and E-box-binding proteins enhance the T3 responsiveness of the malic enzyme gene in avian hepatocytes. However, the authors did not find that USF proteins were binding to the E-box in the malic enzyme gene. Previous groups have identified USF as an important component of glucose response complexes in the L-type pyruvate kinase (PK) gene and the glucagon receptor gene, although others have reported that carbohydrate response element-binding protein rather than USF-1 is important in the glucose stimulation of gene expression (35,36,37). USF-1 and USF-2 have also been shown to participate in the regulation of the fatty acid synthase and acetyl-CoA carboxylase-␣ genes by insulin (38). To our knowledge, our studies are the first indicating that USF proteins are involved in T3 responsiveness.
Although our analysis of the CPT-I␣ gene indicates that C/EBP and USF factors are involved in T3 induction, other factors can serve as accessory factors in the stimulation of gene expression by T3. The sterol regulatory element-binding protein-1 (SREBP-1C) interacts with the TR to enhance acetyl-CoA carboxylase (ACC-I) transcription in hepatocytes (14). Furthermore, the homeodomain proteins, PBX and MEIS1, are accessory factors that enhance T3 induction of malic enzyme gene expression in hepatocytes (39). In the liver, NF-Y binds near the start site of transcription and is required for the stimulation of S14 gene transcription (40). The S14 TREs are located far upstream between nucleotides Ϫ2700 and Ϫ2500 (41). In the heart, myocyte-specific enhancer factor 2 (MEF2) contributes to the T3 induction of the ␣-cardiac myosin heavy chain (␣-MHC) gene (42). The induction of human placental lactogen B (hCS-B) and rGH genes by T3 is dependent on the pituitary factor, Pit-1 (43,44). Accessory factors function in a gene-and tissue-specific manner to modulate hormone responses.
A unique aspect of the CPT-I␣ gene is that the accessory factors are located in the first intron, whereas the hormone response element is located in the promoter. One question raised by our studies is how the accessory factors located 3,600 base pairs from the TR bound to the TRE cooperate in the T3 induction. In these studies, we have demonstrated that the TR␣ can physically interact with USF-1 and USF-2 as well as with C/EBP␣ through the DNA binding domain of TR. We also show by ChIP assay that TR interacts with sequences within intron 1, which suggests that the DNA loops so that the TR can interact with USF and C/EBP in the intron. Previously, we have ligated a Gal4 site in front of several C/EBP binding sites. The inclusion of C/EBP binding sites enhanced the induction by T3 and a Gal4-TR␤ protein that does not have the TR DNA binding domain (16). However, these observations do not rule out direct physical interactions between the TR and C/EBP. We do not know if multiple USF sites can potentiate T3-mediated transcriptional induction out of the context of the CPT-I␣ gene.
It is possible that coactivator proteins are involved in the interactions between the accessory factors and the TR␤. Recent studies have shown that there is a sequential order of coactivator recruitment to the liganded TR (21). First SRC-1 and CBP/p300 are recruited to the liganded receptor followed by the recruitment of the TRAP/mediator complex. The accessory factors may help to recruit or stabilize the coactivators that are associated with the CPT-I␣ gene. C/EBP␤ interacts with p300 through its amino terminus and the E1A binding region of p300 (45). We have found that overexpression of CBP, the functional homologue of p300, modestly increases the basal expression of the CPT-I␣ gene (data not shown). We have recently reported that CBP enhances the T3 induction of the PEPCK gene (16). In addition, we have observed weak physical interactions between C/EBP␣ and CBP, although these interactions might be stabilized in the context of the CPT-I␣ gene (16). Furthermore, hepatocyte nuclear factor-4 can interact with SRC-1, so other proteins adjacent to the CPT-I␣-TRE as well as those in the intron may assist in the recruitment of coactivators (46).
Our studies have established the necessity of intron 1 in the transcriptional regulation of the CPT-I␣ gene. The present work has defined a unique regulatory arrangement for thyroid hormone involving cooperation between transcription factors in the intron and promoter. The interaction of TR with sequences in the intron and its ability to physically interact with C/EBP and USF present a novel model for cooperation in gene induction between nuclear receptors and accessory factors. Coactivators may also participate by providing a physical link between the TR in the promoter and accessory factors in intron 1. Future studies will investigate the role of coactivators in this regulation.