Thyroid Hormone Regulates Carnitine Palmitoyltransferase Iα Gene Expression through Elements in the Promoter and First Intron*

Carnitine palmitoyltransferase I (CPT-I) catalyzes the transfer of long chain fatty acyl groups from CoA to carnitine for translocation across the mitochondrial inner membrane. CPT-Iα is a key regulatory enzyme in the oxidation of fatty acids in the liver. CPT-Iα is expressed in all tissues except skeletal muscle and adipose tissue, which express CPT-Iβ. Expression of CPT-Iα mRNA and enzyme activity are elevated in the liver in hyperthyroidism, fasting, and diabetes. CPT-Iα mRNA abundance is increased 40-fold in the liver of hyperthyroid compared with hypothyroid rats. Here, we examine the mechanisms by which thyroid hormone (T3) stimulates CPT-Iα gene expression. Four potential T3 response elements (TRE), which contain direct repeats separated by four nucleotides, are located 3000–4000 base pairs 5′ to the start site of transcription in the CPT-Iα gene. However, only one of these elements functions as a TRE. This TRE binds the T3 receptor as well as other nuclear proteins. Surprisingly, the first intron of the CPT-Iα gene is required for the T3 induction of CPT-Iα expression, but this region of the gene does not contain a TRE. In addition, we show that CPT-Iα is induced by T3 in cell lines of hepatic origin but not in nonhepatic cell lines.

A rate-controlling step in the pathway of mitochondrial fatty acid oxidation is the formation of acylcarnitines that are required for the transport of long chain fatty acids into the mitochondria (1). This reaction is catalyzed by carnitine palmitoyltransferase I (CPT-I) 1 located in the outer mitochondrial membrane. CPT-I transfers the acyl moiety from acyl-CoA to carnitine (1). Acylcarnitine is translocated across the inner mitochondrial membrane in exchange for carnitine by carnitine acylcarnitine translocase (2) and then re-esterified with CoA by the inner mitochondrial membrane CPT (1). CPT-I␣ is inhib-ited by malonyl-CoA, which is a substrate for fatty acid synthesis (3). Insulin increases the sensitivity to inhibition by malonyl-CoA (4). Two isoforms of CPT-I have been identified (5,6). The cDNA for the "liver" isoform (CPT-I␣) was cloned first from a rat liver cDNA library (5), and the "muscle" isoform (CPT-I␤) was subsequently cloned from rat brown adipose tissue and rat heart libraries (6,7). The cDNAs for the human isoforms of CPT-I␣ and CPT-I␤ have been described (8,9). The rat ␣ and ␤ isoforms share 63% identity at the amino acid level (1). The liver isoform is present in all tissues except skeletal muscle and adipose tissue (1).
Thyroid hormone (T3) has profound effects on metabolism and fatty acid oxidation in the liver (10). The oxidation of long chain fatty acids is increased in the hyperthyroid state. We found that CPT-I␣ activity, sensitivity to malonyl-CoA, and mRNA levels were altered by T3 (11). In hyperthyroid rats, there was a 5-fold increase in CPT-I␣ mRNA abundance in the liver over euthyroid rats. CPT-I␣ mRNA levels in the liver decreased 80% in hypothyroidism (11). CPT-I␣ enzyme activity was increased in hyperthyroid and decreased in hypothyroid rats (11). The sensitivity of CPT-I␣ to malonyl-CoA inhibition was greatly increased in hypothyroidism. CPT-I␣ expression in the liver is altered by diet and during development. Oxidation of long chain fatty acids is increased in response to high fat diets and fasting, as well as in disease states such as diabetes (1,12). Glucagon and long chain fatty acids stimulate whereas insulin inhibits CPT-I␣ transcription (4,13). During fetal development, CPT-I␣ is not expressed in the liver, but transcription of the gene is induced at birth (14). Therefore, CPT-I␣ gene expression is controlled by hormonal and nutritional factors.
The effects of T3 are mediated through the binding of the liganded thyroid hormone receptor (TR) to T3 response elements (TRE) in the promoters of genes (15,16). The consensus TRE contains a direct repeat of the AGGTCA motif separated by 4 nucleotides (DR4). Small variations in the AGGTCA motifs will affect the affinity of the TR for the DR4 motif. In addition, the TR can bind to single repeats and repeated motifs separated by various spacing (17,18). The TR binds to DNA primarily as a heterodimer with the retinoid X receptor (RXR) (19). On DR4 motifs, RXR binds to the 5Ј motif, whereas the TR binds to the 3Ј repeat (19). There are two isoforms of the TR: TR␣ and ␤ (15). The ␤ isoform of the TR appears to mediate many of the effects of T3 in the liver. The T3 induction of several hepatic genes including S14 and malic enzyme is blunted in TR␤ knockout mice (20).
Previously, we cloned the promoter of the CPT-I␣ gene (21). The CPT-I␣ promoter is TATA-less, and basal expression is driven by Sp1 and NF-Y (22). Here, we wished to examine the mechanisms by which T3 stimulates CPT-I␣ gene expression. Our results demonstrate that there is a TRE in the promoter of the CPT-I␣ gene. However, T3 induction is dependent on sequences within the first intron.

EXPERIMENTAL PROCEDURES
DNase I Footprint Analysis-The probe used for DNase I footprint analysis of the CPT-I␣ TRE was generated by PCR amplification of sequence between nucleotides Ϫ3143 and Ϫ2684 in the CPT-I␣ promoter. The sequence of the CPT-I␣ promoter, first exon and first intron has been entered into GenBank TM under accession number AF020776 (21). The PCR reaction contained the forward primer 5Ј-TCTCTCGAG-GCAAACGCTGCC-3Ј and the reverse primer 5Ј-TTCGGATCCGCCTC-CTTTCCTGTTGGCAGCACC-3Ј. The Ϫ4495/ϩ19 CPT-I␣-Luc vector was digested with KpnI and EcoRI. The fragment from nucleotides Ϫ4495 to Ϫ1653 was isolated from a 1% (w/v) agarose gel and used as the template. The forward primer was labeled with [␥-32 P]ATP and T4 polynucleotide kinase before addition to the PCR reaction. PCR conditions included 1 cycle at 95°C for 45 s and 35 cycles of 95°C for 45 s, 52°C for 45 s, and 72°C for 1 min. The labeled 460-base pair PCR product was purified from a 3% (w/v) NuSieve 3:1 agarose gel (FMC BioProducts, Rockland, ME).
Recombinant TR␤ and RXR␣ proteins were incubated with the radiolabeled CPT-I␣ probe, and the probe was then digested with DNase I as described by Park et al. (23). The reactions were resolved on a denaturing (5% w/v) acrylamide gel in 1ϫ TBE (89 mM Tris-HCl, pH 8.0, 89 mM borate, 2 mM EDTA) running buffer.
A DNase I footprint probe within intron 1 was prepared by digesting the Sma-Mlu-Gal4-SV40 vector with BglII. The resulting overhang was treated with alkaline phosphatase and labeled with [␥-32 P]ATP and T4 polynucleotide kinase. The linearized labeled vector was then digested with SacI to create a top strand labeled probe containing nucleotides ϩ708 through ϩ1066. The probe was purified from a (1% w/v) agarose gel.
Preparation of Recombinant Proteins in Escherichia coli-Synthesis of histidine-tagged rTR␤ (His-rTR␤) was performed by digestion of the rat TR␤ cDNA (24) with BamHI and PstI and ligation of this fragment into pQE30 (Qiagen, Valencia, CA). In this vector, the first 63 amino acids were removed from the amino terminus of rTR␤, but the DNA and ligand binding domains were intact. Synthesis of histidine tagged hRXR␣ (His-hRXR␣) has been described previously (24). The histidine tagged receptors were expressed in the M15 pREP4 E. coli strain. Purification of His-rTR␤ and His-hRXR␣ has been described previously (24).
The labeled probe (30,000 cpm) was combined with 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 (24). The nonspecific competitor added to each binding reaction was 1 g of a 1:1 ratio of poly(dI⅐dC):poly(dA⅐dT). In the supershift assays, antibodies were added to the binding reaction prior to the addition of nuclear protein extract. TR␤ antibody was purchased from Upstate Biotechnology, Inc. (Lake Placid, NY). Preimmune serum was added in a control binding reaction. Binding reactions were incubated at room temperature for 20 min and resolved on a 5% nondenaturing acrylamide gel (80:1, acrylamide/bisacrylamide) in 0.5ϫ Tris-glycine running buffer (22 mM Tris and 190 mM glycine) at 4°C (22). Proteins from rat liver nuclei were isolated by the method of Gorski et al. (25).
Construction of Luciferase Vectors-Regions of the CPT-I␣ promoter and intron 1 were ligated into the pGL3-basic luciferase vector (Promega, Madison, WI). Numbering of the CPT-I␣ gene begins at the start site of transcription as ϩ1, and the promoter extends upstream to nucleotide Ϫ6839. Exon 1 contains nucleotides ϩ1 through ϩ27. Numbering continues through the first intron, and exon 2 begins at nucleotide ϩ1201. Construction of Ϫ1653/ϩ19 CPT-I␣-Luc, Ϫ4495/ϩ19 CPT-I␣-Luc, and Ϫ6839/ϩ19 CPT-I␣-Luc was described previously (22). Nucleotides from Ϫ1653 in the promoter to ϩ1240 in exon 2 were removed from the original P1 clone (21) by EcoRI-SalI digestion and ligated into EcoRI/SalI of pBS II K/S ϩ/Ϫ (Stratagene, La Jolla, CA). This Ϫ1653/ϩ1240 CPT-I␣-pBS vector was digested with XhoI and partially digested with SacI to remove the Ϫ1653/ϩ1240 CPT-I␣ fragment that was ligated into pGL3-basic at XhoI/SacI to create Ϫ1653/ϩ1240 CPT-I␣-Luc. The Ϫ4495/ϩ1240 CPT-I␣-Luc vector was constructed by ligating nucleotides Ϫ1653/ϩ1240 into EcoRI/BglII of Ϫ4495/ϩ19 CPT-I␣-Luc. The additional 5Ј region of the promoter (nucleotides Ϫ6839/ Ϫ1653) was initially isolated from the P1 cone and inserted into the pBS II K/S ϩ/Ϫ multiple cloning site at SalI/EcoRI. The Ϫ6839/ϩ1240 CPT-I␣-Luc vector was made by ligating the Ϫ6839/Ϫ1653 fragment from pBS and ligating into KpnI/EcoRI of Ϫ1653/ϩ1240 CPT-I␣-Luc.
The ⌬Sma Ϫ4495/ϩ1240 CPT-I␣-Luc vector was created by digesting Ϫ4495/ϩ1240 CPT-I␣-Luc with SmaI and religating the vector. Similarly, the Mlu-Mlu region of the intron was removed (⌬Mlu Ϫ4495/ ϩ1240) by Mlu digestion and religation. The Sma-Sma region of the first intron (nucleotides ϩ199/ϩ707) was ligated into the XbaI site of the SV40-Luc vector and the Mlu-Mlu region (ϩ130/ϩ1066) was ligated into SV40-Luc at the Mlu site. A Gal4 DNA-binding site corresponding to the sequence 5Ј-TCGGAGTACTGTCCTCCGT-3Ј was ligated into the SV40 enhancerless pGL3-promoter vector (Gal4-SV40). The Sma-Sma region of the CPT-I␣ first intron, the Mlu-Mlu region, and the Sma-Mlu region were ligated into the Gal4-SV40 vector at the SacI and BglII sites. Deletions within the Sma-Mlu region were created by PCR with the following reverse primer containing a BglII restriction site: 5Ј-AATAGATCTTCGAAAAAGC TTCTAGACTC-3Ј. The forward primers contained a Mlu I site as follows: Mlu(ϩ983) 5Ј-CAGACGCGTGTC-CAAGGACAGATTTTAG-3Ј and Mlu(ϩ900) 5Ј-GTGACGCGTGCTA-GAT ACAATCAGGCTC-3Ј. The PCR products of the intron regions ϩ983/ϩ1066 and ϩ901/ϩ1066 were ligated into the MluI and BglII sites of the Gal4-SV40-Luc vector.
Serial deletions from the 5Ј end of Ϫ4495/ϩ64 CPT-I␣-Luc were created through a PCR based approach. A reverse primer was generated beginning at the CPT-I␣ promoter EcoRI restriction site, 5Ј-GCT-GAATTCATTAGAAGGAGGACT-3Ј. Resulting PCR fragments were digested with KpnI/EcoRI and ligated into Ϫ1653/ϩ19 CPT-I␣-Luc, generating a series of 5Ј promoter deletions (as indicated by the nucleotide number with each forward primer listed above). The first intron was introduced into these serial deletion constructs by removing nucleotides Ϫ1653/ϩ19 (with EcoRI and BglII digestion) and replacing them with nucleotides Ϫ1653/ϩ1240. All serial deletion vectors were confirmed by sequence analysis.
Site-directed mutagenesis was conducted using the DpnI Quick Change method (Stratagene, La Jolla, CA). The Ϫ4495/Ϫ1653 region of the promoter was ligated into pBluescript using the KpnI and EcoRI sites. Oligonucleotides were generated to sequentially disrupt four potential TREs in the CPT-I␣ promoter. Each potential TRE was mutated at two nucleotides in each half-site of the DR4 as follows: Ϫ4251 Mut 5Ј-CCTGGGATCTAGTGCAATCCTCTGATCAGGGGGCACTTGCAC-3Ј, Ϫ3925 Mut 5Ј-CAATCCCAGTACAGGGAGGTAGTGGCAATATTT-CAGGAGTTCAAGGTCAC-3Ј, Ϫ3225 Mut 5Ј-CTAAAGTGAAGAAACT-GAGGTTTACACAATATTAGGAGGGGCTCACACTG-3Ј, and Ϫ2923 Mut 5Ј-GTGACAAACGCCGCTGTTTTCATGGAATATGGTGACGCTG-GCTG-3Ј. Mutations in the CPT-I␣ promoter were confirmed by sequence analysis. The mutated Ϫ4495/Ϫ1653 CPT-I␣ fragment was ligated into the KpnI and EcoRI sites of the Ϫ1653/ϩ19 CPT-I␣-Luc vector. The first intron was introduced into these vectors as described above for the serial deletion constructs.
Cell Culture, Transfections, and Luciferase Assays-Transient transfections were conducted in HepG2 cells using calcium phosphate precipitation as described previously (22). The calcium phosphate precipitate contained 3 g of CPT-I␣-Luc and either 1 g of RSV-TR␤ or 100 ng of a mammalian expression vector for the Gal4-TR␤. HepG2 cells were maintained in Dulbecco's modified Eagle's medium containing 5% fetal bovine serum and 5% calf serum. Cells were trypsinized and replated at 60 -70% confluence. After 4 h for cell reattachment, the calcium phosphate DNA precipitate was added to each plate and incubated overnight. The cells were washed with phosphate-buffered saline and incubated in 10% charcoal stripped fetal calf serum and 100 nM T3 for 18 h. Rat L6 myoblasts were transfected under identical conditions as were HepG2 cells.
HepG2 and L6 cells were harvested and luciferase activity was determined as described previously (21). Luciferase activity was measured in a TD 20/20 luminometer (Turner Designs, Sunnyvale, CA). Protein concentration of the cell lysate was measured with a colormetric assay using Bio-Rad protein dye reagent. Luciferase data are expressed as luciferase activity corrected for protein concentration in the cell lysate. Each transfection was conducted in duplicate and repeated three to six times.
Primary culture of Percoll-purified neonatal cardiac ventricular myocytes was conducted as described by Bahouth (26) Ventricles from 1-day-old rats were minced in buffer containing 0.075% Viocase. Cardiac myocytes were separated from nonmuscle cells by centrifugation through Percoll gradients and were cultured on collagen-coated plates in 68% Dulbecco's modified Eagle's medium, 17% medium-199, 10% horse serum, 5% fetal bovine serum. Beating cardiomyocytes were maintained in culture for 5-7 days before transfection. Transient transfections of cardiac myocytes were carried out using the LipofectAMINE Plus system from Life Technologies, Inc. using the manufacturer's recommendations. Myocytes were harvested, and luciferase activity was determined for firefly luciferase and Renilla luciferase (transfection control) using the Dual luciferase assay system (Promega).
Sequencing-DNA was sequenced with the Dye terminator cycle sequencing system (PerkinElmer Life Sciences) and analyzed at the Molecular Resource Center of the University of Tennessee. Additional sequencing was conducted at the Center for Biotechnology, St. Jude Children's Research Hospital (Memphis, TN).

RESULTS
Our initial experiments were designed to define broad regions of the CPT-I␣ gene that were involved in the stimulation of expression by T3. To conduct these experiments, the CPT-I␣ promoter was ligated in front of the luciferase reporter gene (CPT-I␣-Luc). Previously, we had cloned the promoter of the CPT-I␣ gene and characterized the proximal promoter of the gene (21). The Ϫ6839/ϩ1240 CPT-I␣-Luc vector contains the full-length promoter, the first exon, the first intron and a portion of the second exon (Fig. 1A). Exon 1 is only 27 nucleotides long. The Ϫ6839/ϩ19 CPT-I␣-Luc vector contains the CPT-I␣ promoter and a portion of exon 1. The Ϫ6839/ϩ1240 CPT-I␣-Luc vector was transiently transfected into HepG2 cells in the presence or absence of T3. Addition of T3 stimulated transcription of Ϫ6839/ϩ1240 CPT-I␣-Luc 5.5 Ϯ 0.1-fold (Fig.  1B). The Ϫ4495/ϩ1240 CPT-I␣-Luc vector maintained T3 responsiveness with a 5.8 Ϯ 0.6-fold induction by T3. Deletion of the promoter between Ϫ4495 and Ϫ1653 resulted in a loss of T3 responsiveness. Also, deletion of the first intron as in the Ϫ4495/ϩ19 greatly reduced T3 stimulation from a 5.5 Ϯ 0.1-to a 2.0 Ϯ 0.1-fold induction. These observations indicate that two widely separated elements are required for the T3 induction of the CPT-I␣ gene. As a control, one or two copies of an idealized TRE, which consists of DR4, were ligated in front of the enhancerless SV40 promoter driving the luciferase reporter gene. The vector containing two copies of the DR4 was stimulated 5.7 Ϯ 0.7-fold by T3.
To further analyze the T3 induction of CPT-I␣ gene expression, we identified the TRE in the CPT-I␣ gene. Initially, serial deletions were created in the promoter of the CPT-I␣ gene The nucleotides 3Ј to the start site of transcription are labeled with positive numbers so that exon 1 contains nucleotides ϩ1 to ϩ27 and exon 2 starts at ϩ1201. Intron 1 and the translation start site (ATG) in exon 3 are labeled. The promoter and first exon (Ϫ6839/ϩ19) or the promoter, first two exons and intron (Ϫ6839/ϩ1240) were ligated in front of the Luc reporter gene as shown. B, HepG2 cells were transiently cotransfected with 3 g of CPT-I␣-Luc vectors, 1 g of a mammalian expression vector RSV-TR␤, and 1 g of SV40-␤gal as a transfection control. Some HepG2 cells were cotransfected with SV40-luciferase vectors, 1 g of RSV-TR␤, and 1 g of SV40-␤gal. The SV40 promoter is the enhancerless SV40 promoter driving the luciferase reporter gene. Either one or two copies of a direct repeat separated by four nucleotides (DR4) were ligated in front of the SV40 promoter. After an 18-h transfection, cells were incubated 24 h in the presence of 100 nM T3. Cells were harvested, and luciferase assays were conducted. The fold T3 induction was determined, and luciferase activity was corrected for protein concentration in the dish. Each transfection was conducted in duplicate and repeated four to six times. The data are presented as fold induction by T3 Ϯ S.E. between Ϫ4495 and Ϫ1653 to define a narrow region required for T3 responsiveness. The serial deletions removed approximately 50 -400-base pair fragments from the 5Ј end of the promoter in the Ϫ4495/ϩ1240 CPT-I␣-Luc vector (Fig. 2). The average 6-fold induction by T3 was lost with removal of the nucleotides between Ϫ2971 and Ϫ2912, suggesting that this region contained a TRE.
Examination of the sequences between Ϫ4495 and Ϫ1653 for potential TREs revealed the presence of four DR4-like motifs that are shown in Table I. Each of these elements appeared to be good candidates for TR␤-RXR␣ binding and raised the possibility that multiple TREs were involved in T3 action. However, deletion of the three most 5Ј motifs had no effect on T3 responsiveness. The fourth DR4 fell between nucleotides Ϫ2971 and Ϫ2912 in the T3-responsive region, suggesting that this element was most likely the functional TRE.
To specifically determine which of these elements was involved in the T3 induction of CPT-I␣ transcription, each of the DR4 motifs was altered by site-directed mutagenesis in the context of the Ϫ4495/ϩ1240 CPT-I␣-Luc vector (Fig. 3). The specific mutations that were introduced are shown in Table I. In agreement with the data from our serial deletions, disruption of the three most 5Ј elements did not diminish the 5.8 Ϯ 0.6-fold T3 induction seen with the wild type vector. However, mutation of the most 3Ј DR4 (Ϫ2938 Mut) resulted in a complete loss of T3 responsiveness (Fig. 3). These data demonstrate that T3 stimulates CPT-I␣ gene expression through a single TRE and not by synergism between several weak TREs. It is not clear why this DR4 motif serves as the TRE as opposed to the other DR4-like sequences.
Next, we determined whether TR␤ could bind to the TRE in the CPT-I␣ promoter. Because the TR binds with higher affinity as a heterodimer with the RXR, TR␤ and RXR␣ were prepared by overexpression in an E. coli expression system. These proteins contained a histidine tag on the amino terminus and were purified by nickel affinity chromatography (24). Both DNase I footprinting and gel shift mobility assays were used to examine binding to the TRE. A DNase I footprinting probe was generated from nucleotides Ϫ3143 to Ϫ2684. The probe was incubated with recombinant TR␤ and RXR␣ as described under "Experimental Procedures." Nucleotides Ϫ2938 to Ϫ2923, which contain the most 3Ј DR4 motif in the CPT-I␣ promoter, bound the TR␤-RXR␣ heterodimer and were protected from DNase I digestion (Fig. 4).
To further analyze the binding of the TR␤ to the CPT-I␣ TRE, electrophoretic mobility shift assays were conducted with a labeled double-stranded oligonucleotide containing the TRE. As shown in Fig. 5A, the probe was incubated with TR␤ and RXR␣. Double-stranded oligonucleotides containing various mutations in the TRE were used in competition studies. The sequences of these oligomers are shown in Table II. The shifted DNA-protein complex was competed with the unlabeled CPT-I␣ TRE oligomer. An oligomer with a mutation in both half-sites of the DR4 (Mut1) was unable to compete for TR␤-RXR␣ binding. This mutation was identical to the mutation that was introduced into Ϫ4495/ϩ1240 CPT-I␣-Luc and eliminated T3 responsiveness. Similarly, an oligomer containing only a 3Ј half-site disruption (Mut2) was unable to compete. A mutation in the 5Ј-flanking sequence of the CPT-I␣ TRE (Mut3) and an oligomer consisting of an idealized TRE (Cons) were both able to compete for protein binding. Additional assays showed no binding of TR␤ monomers or homodimers to the CPT-I␣ TRE probe (data not shown). These data indicate that TR␤, as a heterodimer with RXR␣, can bind to the TRE in the CPT-I␣ promoter.
To determine whether the TR in a nuclear extract could bind the CPT-I␣ TRE, the binding of proteins isolated from rat liver nuclei (RLNE) to this element was examined using gel shift mobility assays. The CPT-I␣ TRE probe was incubated with RLNE, and three DNA-protein complexes were observed (Fig.  5B). Competition analysis was conducted with the same oligomers that were used in Fig. 5A. All three complexes were competed away by a 100-fold excess of wild type probe. The most slowly migrating (top) complex was not competed by the consensus DR4 or the Mut3 oligomer, indicating that this complex does not contain the TR. The competition pattern of the middle complex was similar to that seen with the TRE-TR␤-RXR␣ complex in Fig. 5A, suggesting that this complex might contain the TR. Binding of the most rapidly migrating (bottom) complex was eliminated by unlabeled oligomers containing mutations in the DR4. To determine whether the middle complex contains the TR, supershift analyses with the TR␤ antibody were conducted (Fig. 5C). Addition of the TR␤ antibody to the binding reaction disrupted the middle DNA-RLNE complex. Antibody alone produced no shifted band. These data indicate that TR␤-RXR␣ binds the TRE in the CPT-I␣ promoter. In addition, other proteins present in rat liver nuclei can bind this element.
Our data in Fig. 1 indicated that sequences in the first intron were required for the full T3 induction of CPT-I␣ gene expression. These data raised the possibility either that there was a second TRE in the first intron or that an accessory factor site in the first intron was required for the T3 stimulation. Our next experiments examined the role of the first intron in the T3 induction of CPT-I␣ gene expression. Internal deletions in the first intron were introduced in the context of the Ϫ4495/ϩ1240 CPT-I␣-Luc vector. The Ϫ4495/ϩ1240 CPT-I␣-Luc vector containing the Sma to Sma (⌬Sma) and Mlu to Mlu (⌬Mlu) deletions within the first intron were tested for T3 responsiveness (Fig. 6A). The ⌬Sma deletion that removes nucleotides (ϩ199/ ϩ707) reduced the T3 induction to 2.5-fold, and the ⌬Mlu vector, which eliminates nucleotides (ϩ130/ϩ1066), was stimulated only 1.4-fold by T3. These data suggest that at least two elements in the intron contribute to the T3 response.
To determine whether the intron contained a TRE, the Sma to Sma (ϩ199/ϩ707) or Mlu to Mlu (ϩ130/ϩ1066) regions were

FIG. 2. Localization of a T3-responsive region in the promoter of the CPT-I␣ gene. Serial deletions of the CPT-I␣ promoter between Ϫ4495 and
Ϫ1653 were ligated in front of the luciferase reporter gene. These vectors contain the first intron. Transfections into HepG2 cells and additions of T3 were conducted as described in Fig. 1. Each transfection was conducted in duplicate and repeated four to six times. The data are presented as fold induction by T3 Ϯ S.E. ligated in front of the SV40 promoter (Fig. 6B). Neither of these segments of DNA conferred a T3 response upon the SV40 promoter. These data suggest that the intron contains elements that bind accessory factor(s) required for the T3 induction of CPT-I␣. In addition, there are no DR4-like elements within the first intron. Finally, we tested the ability of the CPT-TRE to confer T3 responsiveness on a neutral promoter (Fig. 6C). Either one or two copies of the CPT-TRE was ligated in front of the SV40 promoter driving the luciferase reporter gene. T3 stimulated the vector containing two copies of the CPT-TRE to a similar extent as two copies of the idealized DR4 motif. These results indicate that the isolated CPT-TRE can confer a T3 induction on a neutral promoter. However, when the CPT-TRE is placed in the context of the CPT-I␣ gene, elements within the first intron are essential for the T3 effect.
We tested whether elements within the intron could enhance T3 responsiveness outside the context of the CPT-I␣ gene. A Gal4 site was ligated in front of the SV40-luciferase (Gal4-SV40-Luc) vectors containing portions of the first intron. These vectors were cotransfected with a mammalian expression vector for TR␤ in which the TR␤ DNA-binding domain was replaced by the Gal4 DNA-binding domain (Gal4-TR␤) (27). The Gal4-TR␤ vector was used to eliminate the possibility that the TR␤ was binding to a TRE in the intron.
T3 stimulated the Gal4-SV40-Luc vector 7.8 Ϯ 0.8-fold (Fig.  7A). The T3 induction of the intron vectors were compared with Gal4-SV40-Luc, which was normalized to 1.0. Addition of the Sma to Sma (ϩ199/ϩ707) region to Gal4-SV40-Luc did not further stimulate the T3 responsiveness. Gal4-SV40-Luc vectors containing the Mlu to Mlu (ϩ130/ϩ1066) and the Sma to Mlu (ϩ707/ϩ1066) regions were stimulated an additional 3.1 Ϯ 1.0 and 3.4 Ϯ 0.3-fold by T3, respectively (Fig. 7A). The Mlu to Mlu vector without the Gal4 DNA-binding site was not stimulated by T3 when cotransfected with the Gal4-TR␤ expression vector (data not shown). These data indicate that an accessory factor site between the ϩ707 and ϩ1066 region of the first intron enhances the T3 induction.
To further characterize the involvement of this 360-base pair intron region, we first conducted DNase I footprint analysis. A probe containing nucleotides ϩ707 through ϩ1066 (Sma to Mlu) was labeled on the sense strand and incubated with proteins isolated from rat liver nuclei. Four protein-binding regions were protected from DNase I digestion (Fig. 8).
Next, we further defined regions within the Sma to Mlu portion of the intron that were involved in the T3 induction. The positions of the DNase I protected sites within the Sma to Mlu region are indicated by shaded ovals in Fig. 7B. Regions containing each protein-binding site were deleted in the Gal4-SV40 Luc vector as described previously for Fig. 7A. The 3.4fold enhancement of T3 responsiveness was completely lost upon deletion of nucleotides ϩ707 to ϩ901 (Fig. 7B). However, the ϩ707 to ϩ901 region alone was not able to enhance the T3  Table I were introduced into the promoter of the Ϫ4495/ϩ1240 CPT-I␣-Luc vector. The positions of four potential TREs corresponding to DR4 motifs are indicated with a shaded box. The most 5Ј nucleotide of the mutated DR4 motif is delineated by the nucleotide number followed by Mut. The ؋ indicates a disrupted DR4 motif. The luciferase vectors were transiently transfected into HepG2 cells and exposed to T3 as described in the legend to response. The results of these transfections suggest that several protein-binding domains in the ϩ707 to ϩ1066 region will participate in the T3 response.
The final experiments examined the ability of T3 to stimulate CPT-I␣ gene expression in various cell lines. CPT-I␣-luciferase vectors were transfected into either HepG2, L6 rat myoblasts, or primary cardiac myocytes, and cells were treated with T3 (Fig. 9). Expression of CPT-I␣-Luc vectors was stimulated 1.5-2-fold by T3 in L6 cells and cardiac myocytes regardless of the presence or absence of the first intron. The T3 responsiveness of the Ϫ4495/ϩ1240 CPT-I␣-Luc vector in L6 myoblasts and cardiac myocytes was similar to the T3 induction of Ϫ4495/ϩ19 CPT-I␣-Luc in HepG2 cells. These data suggest that binding of a specific accessory factor in the first intron allows CPT-I␣ gene expression in hepatic cells to be more responsive to T3 more than in muscle cells. As a control, the DR4 ϫ 2 SV40-Luc was transfected into each cell type. Expression of this vector was well stimulated in all cell types, indicating that T3 was capable of inducing transcription in these cells. These data agree with the observation that T3 produces a minimal increase in the CPT-I␣ mRNA abundance in cardiac myocytes. 2

DISCUSSION
The regulation of gene expression by T3 has profound effects on hepatic metabolism. We show here that T3 regulates the transcription of a gene encoding a key enzyme in long chain fatty acid ␤-oxidation. The promoter of the CPT-I␣ gene contains a TRE located 3000 base pairs upstream of the transcriptional start site. DNA-protein binding studies reveal that TR␤ binds this element as a heterodimer with RXR␣. To achieve the maximum stimulation by T3, the first intron of the gene is required. Thus, our results indicate that TR␤ binds a TRE in the CPT-I␣ 5Ј-flanking region and works in concert with accessory factor(s) bound to sequences in the first intron.
The CPT-I␣ TRE loosely conforms to the consensus DR4 motif (AGGTCAXXXXAGGTCA), which is found in several genes including the myosin heavy chain ␣ and malic enzyme genes (28,29). The bottom strand DNA sequence of the CPT-I␣ TRE is AGGTTCcatgAGGACA, differing from the consensus DR4 by three nucleotides. Three additional DR4-like motifs are located upstream of the TRE and contain sequences suggesting that TR␤ would bind. Interestingly, these elements did not contribute to the T3 responsiveness. Although it is unclear as to why these elements do not function as TREs, subtle sequence alterations may modulate their affinity for TR. Another possibility is that other factors may bind these elements and preclude the binding of TR. Katz and Koenig (17) reported that the optimal DNA-binding half-site for the TR is the octomeric halfsite TAAGGTCA. The 3Ј half-site of the CPT-I␣ TRE conforms to the optimal half-site by 6 of 8 nucleotides. Spacing of halfsites and variations of the nucleotides within the sequence of this octomer can alter TR binding as well as change ligand responsiveness of the receptor and its ability to transactivate (17,18,30).
Variability of the TRE sequence plays a role in transcriptional regulation through several mechanisms. Alterations in the flanking sequence of the AGGTCA half-site can increase the requirement for RXR␣ heterodimerization with TR for T3mediated gene expression to occur (31). Using gel shift assays, we did not observe monomeric or homodimeric binding of TR␤ to the CPT-I␣ TRE (data not shown). Coactivator involvement in TR-mediated responses may also be dictated by the sequence of the TRE. For instance, SRC-1 greatly stimulates TR-mediated gene expression from optimal TREs but plays a lesser role at suboptimal sites where TR/RXR␣ heterodimers bind (31). The architecture of the TRE can affect the release of TRassociated corepressors (32). In addition, divergence from the consensus DR4 may allow other proteins to bind these sites. Our studies reveal that proteins from rat liver nuclei other than the TR␤ can bind the CPT-I␣ TRE. Supershift assays conducted with the CPT-I␣ TRE, nuclear proteins from rat liver and an antibody to COUP-TF indicated that COUP-TF did not bind this element (data not shown).
Although the regulation of gene expression by T3 requires the binding of TR to the gene, other transcription factors, called accessory factors, modulate the effect of T3. The participation of accessory factors in hormone responsiveness occurs in many genes and with numerous hormones. We have shown that the T3 stimulation of the PEPCK gene requires the TR and the CCAAT enhancer-binding protein (C/EBP) (24). Cotransfection of the Ϫ4495/ϩ1240 CPT-I␣-Luc vector with a dominant negative expression vector for C/EBP did not affect the T3 induction, although basal expression was reduced (data not shown).
These results indicate C/EBP is not involved in the T3 regula-2 G. Cook, unpublished observations.  Table II. Further details concerning the DNA-protein binding conditions and gel assays are provided under "Experimental Procedures." B, the gel shifts were repeated using proteins isolated from RLNE. C, a gel mobility supershift analysis was conducted with the CPT-I␣ TRE probe. The probe was incubated with proteins isolated from rat liver nuclei, and TR␤ antibody was added to the binding reaction as indicated above the lane. tion of CPT-I␣ gene expression. In the heart, myocyte-specific enhancer factor 2 is involved in the T3 induction of the ␣-cardiac myosin heavy chain gene (33). In the liver, NF-Y is required for the stimulation of S14 gene transcription by T3 (34). The S14 TREs are located between Ϫ2700 and Ϫ2500, while the NF-Y binding site is near the start site of transcription (34,35). The induction of human placental lactogen B (hCS-B) and rGH genes by T3 is dependent on the pituitary factor, Pit-1 (36,37).
The first intron is required for full T3 induction of the CPT-I␣ gene. The expression of several genes has been documented to involve a cooperation between the 5Ј-flanking sequence and the first intron. The hepatocyte-specific expression of the rat glutamine synthetase gene involves enhancer elements in the first intron in addition to 5Ј upstream regulatory regions (38). HNF-3-mediated enhancer activity of intron 1 cooperates with the 5Ј-flanking sequence to regulate the expression of glutamine synthetase during adipocyte differentiation of 3T3-L1 cells (39). In the chicken, the first intron of the liver-specific very low density apolipoprotein II (apoVLDLII) gene contributes to a 4-fold increase in estrogen responsiveness (40). T3-responsive elements have been described that are 3Ј to the transcriptional start site in several genes. Transcription of the ␤ 1 -adrenergic receptor gene is stimulated by T3 in cardiac ventricular myocytes (41). We have demonstrated that although the TRE of the ␤ 1 -adrenergic receptor gene is located 3Ј to the transcriptional start site, the induction by T3 is dependent on promoter elements located more that 2,000 base pairs upstream (41). The NRGN gene in the human brain is regulated by T3 through a TRE located in the first intron of the gene (42). The rat growth hormone gene contains a complex TRE in  Fig. 5 are listed in the left column in parentheses. The top strand oligonucleotide of the CPT-I␣ promoter TRE is provided in the right column. The DR4 motif is underlined, and mutated nucleotides are indicated in bold type. The wild type (WT) competitor corresponds to the TRE present in the CPT-I␣ promoter. Mutation 1 (Mut1) contains mutated nucleotides in both the 5Ј and 3Ј half-sites. Mutation 2 (Mut2) contains a mutation in the 3Ј half-site. Nucleotides in the 5Ј flanking sequence of the DR4 are mutated in mutation 3 (Mut3). The sequence of a consensus DR4 (Cons) is shown. All oligonucleotides were generated with XbaI restriction enzyme overhangs.

Competitors
Oligonucleotide The ⌬Sma deletion removes nucleotides ϩ199 to ϩ707, and the ⌬Mlu deletion removes nucleotides ϩ130 to ϩ1066. These vectors were transfected along with 1 g of RSV-TR␤ into HepG2 cells and assessed for T3 responsiveness as described in Fig. 1. B, regions from the first intron of the CPT-I␣ gene were ligated in front of the enhancerless SV40 promoter. These vectors were transfected along with RSV-TR␤ into HepG2 cells as described above. C, one or two copies of the CPT-TRE were ligated in front of the SV40-Luc vector. These vectors were transfected with 1 g of RSV-TR␤ as described in Fig. 1.   FIG. 7. Contribution of specific intron regions in the T3 induction of CPT-I␣. A, a Gal4 DNA-binding site was introduced in front of the enhancerless SV40-Luc. Either the Sma to Sma (ϩ199/ϩ707), Mlu to Mlu (ϩ130/ϩ1066), or the Sma to Mlu (ϩ707/ϩ1066) regions of the first intron were included in these vectors. These constructs were cotransfected into HepG2 cells with 100 ng of a mammalian expression vector for Gal4-TR␤ into HepG2 cells and exposed to T3 as described for Fig. 1. The Gal4-SV40-Luc vector was induced 7.8-fold by T3. The T3 induction of the intron vectors were compared with Gal4-SV40-Luc, which was normalized to 1.0. Each transfection was conducted in duplicate and repeated four to six times. The data are presented as relative induction by T3 Ϯ S.E. B, the positions of the protein-binding sites within the Sma to Mlu region of the first intron are indicated by shaded ovals. Deletions within this region were introduced into the Gal4-SV40-Luc vector and analyzed for T3 induction as described for A. Each transfection was conducted in duplicate and repeated four to six times. The data are presented as relative induction by T3 Ϯ S.E. normalized to the Gal4-SV40-Luc vector. the 5Ј-flanking sequence, and evidence exists for an additional TRE in the third intron, which causes a further increase in the T3 stimulation of gene expression (43,44). However, with a TRE located so far 5Ј and with an accessory factor site in the first intron, the architecture of the CPT-I␣ gene is unique.
Looping of the DNA may bring the TR␤ and accessory factor(s) bound to the first intron closer to each other and to the transcriptional start site. In addition to a direct interaction between the TR␤ and accessory factor(s), various coactivators and basal transcription factors may be involved in the T3 responsiveness of the gene. In our transient transfections with the CPT-I␣-Luc vector, overexpression of two coactivators, SRC-1 and CREB-binding protein (CBP/p300), did not increase the T3 induction of CPT-I␣-Luc constructs (data not shown). However, overexpression of SRC-1 did stimulate basal transcription. The possibility of basal transcription factor involvement in the T3 induction of CPT-I␣ was also explored. Previ-ously, we constructed mutations in the Sp1 and NF-Y sites of the CPT-I␣ proximal promoter (22). The disruption of protein binding to these elements greatly decreased basal gene expression. The mutation of the Sp1 and NF-Y sites in the Ϫ4495/ ϩ1240 CPT-I␣-Luc construct did not decrease the T3 induction of the gene (data not shown).
The first intron of the CPT-I␣ gene does not contain a TRE, and homology searches have not identified any consensus sequences for TR binding. Experiments with vectors containing internal deletions of the intron have localized two broad regions that are important in T3 responsiveness. At least one of these regions (nucleotides ϩ707 to ϩ1066) can stimulate T3mediated transcription outside the context of the CPT-I␣ promoter. Nucleotides ϩ707 to ϩ1066 were characterized by DNase I footprint analysis, and four protected regions were identified (Fig. 8). Whether any of these elements are involved in T3 responsiveness of the CPT-I␣ gene is not known.
CPT-I␣-Luc constructs containing the first intron stimulate T3-mediated gene expression in hepatoma cell lines but not in other cell lines tested. Because the accessory factor appears to enhance T3 responsiveness only in hepatoma cells, it is possible that the accessory factor may be predominantly expressed in the liver. Future studies will not only identify the intron factor involved in T3 induction of CPT-I␣ but will also characterize its role in cell type-specific regulation. These experiments will provide further insight into the mechanisms underlying how metabolic disease states such as hyperthyroidism affect fatty acid oxidation. FIG. 9. Effects of T3 on CPT-I␣ gene expression in various cell lines. HepG2 cells, L6 myoblasts, or primary cardiac myocytes were transfected with CPT-I␣-Luc or SV40-Luc vectors. The HepG2 and L6 cells were treated with 100 nM T3 for 24 h, whereas the cardiac myocytes were exposed to 20 nM T3 for 18 h. The cells were harvested, and luciferase activity was assessed as described for Fig. 1. Each transfection was conducted in duplicate and repeated three to six times. The data are presented as fold induction by T3 Ϯ S.E.
FIG. 8. Binding of nuclear proteins to the ؉707 to ؉1066 region of the CPT-I␣ first intron. A DNase I footprint probe was constructed from nucleotides ϩ707 to ϩ1066 (Sma to Mlu) within the first intron. The sense strand of the Sma to Mlu region was labeled with [␥-32 P]ATP and incubated with proteins isolated from RLNE before DNase I treatment. The G-ladder is shown in the first lane. Protected sequences are indicated by boxes to the right of the footprint and are labeled with the corresponding nucleotides.