Co-regulation of tissue-specific alternative human carnitine palmitoyltransferase Ibeta gene promoters by fatty acid enzyme substrate.

Carnitine palmitoyltransferase I (CPT-I) catalyzes the rate-determining step in mitochondrial fatty acid beta-oxidation. CPT-I has two structural genes (alpha and beta) that are differentially expressed among tissues. Our CPT-Ibeta isolates from a human cardiac cDNA library contained two different extreme 5'-sequences derived from short alternative first untranslated exons that utilize a common splice acceptor site in exon 2. Primer extension identified single dominant start sites for each transcript, and ribonuclease protection assays showed the presence of one 5'-exon in liver, muscle, and heart mRNAs, indicating that the cognate promoter U (upstream/ubiquitous) is active in each of these tissues. By contrast, mRNAs containing the alternative 5'-exon were present only in muscle and heart, indicating a muscle-specific promoter M (muscle). CPT-Ibeta mRNA levels increased markedly in tissues of fasted rats, when circulating free fatty acid concentrations are elevated. Using CPT-Ibeta promoter/reporter transient transfection of murine C2C12 myotubes and HepG2 hepatocytes, fatty acids were found to increase promoter activity in a peroxisome proliferator-activated receptor alpha (PPARalpha)-dependent fashion. A promoter fatty acid response element (FARE) was mapped, mutation of which ablated fatty acid-mediated production of both transcripts. PPARalpha/retinoid X receptor alpha formed specific complexes with oligonucleotides containing the FARE, and anti-PPARalpha antibody shifted nuclear protein-DNA complexes, confirming the role of this factor in regulating the expression of this critical metabolic enzyme gene. The constitutive repressor chicken ovalbumin upstream promoter transcription factor competitively binds at the FARE and modulates fatty acid induction of the promoters.

The carnitine shuttle is utilized in mammalian cells for entry of long-chain fatty acids into the mitochondrial matrix, where they undergo ␤-oxidation (1,2). Carnitine palmitoyltransferase I (EC 2.3.1.21; CPT-I) 1 spans the outer mitochondrial membrane and catalyzes the transfer of fatty acyl groups from coenzyme A to carnitine. Acylcarnitines thus formed traverse the inner membrane via a specific translocase, whereupon fatty acyl-CoAs are regenerated by CPT-II within the matrix. This shuttle constitutes the rate-determining process in fatty acid oxidation in all tissues and is highly regulated by virtue of inhibition of CPT-I by malonyl-CoA (1)(2)(3). This intermediate is the product of acetyl-CoA carboxylase, the first committed step in fatty acid synthesis, such that reciprocal regulation of synthesis and degradation is effected in liver. Malonyl-CoA is also the major physiological inhibitor of CPT-I in non-hepatic and non-lipogenic tissues, including heart (4 -7).
CPT-I has two structural genes (␣ and ␤) that are differentially expressed among tissues that utilize fatty acids as fuel. ␣ gene mRNAs are expressed with highest abundance in liver, pancreatic beta cells, and heart, whereas ␤ gene products predominate in skeletal muscle, adipose tissue, heart, and testis (1,8). The major CPT-I enzymes encoded by these two genes have different kinetic properties, such that the differing relative expression levels among tissues are reflected in different tissue enzyme kinetic properties (1,2). We have recently described the coexpression of two novel CPT-I␤ mRNA splicing forms in human (8) and rat (9) tissues. These forms constitute only 5-25% of total tissue CPT-I mRNA, but encode enzymes that have altered malonyl-CoA regulatory domains. Thus, changes in fatty acid oxidation rates attributable to CPT-I could result from changes in the following: 1) substrate concentrations, 2) inhibitor (malonyl-CoA) concentration, 3) total ␣ and/or ␤ enzyme mass, 4) relative proportions of ␣ and ␤ enzymes, and 5) relative levels of malonyl-CoA-sensitive and -insensitive enzymes, and therefore through regulation at allosteric, transcriptional, or pre-mRNA splicing levels.
There is a rapid increase in fatty acid oxidation in all tissues in the perinatal period, coincident with initiation of suckling (1,10,11). There is also regulation of this pathway in the adult animal, including profound up-regulation during catabolic conditions, when elevated circulating free fatty acid levels result from peripheral lipolysis (1,2). Regulation at a transcriptional level is suggested by parallel increases in fatty acid oxidative enzyme mRNAs in these circumstances (12)(13)(14). A direct role for fatty acids in this process is supported by increases in mitochondrial fatty acid ␤-oxidation enzyme mRNAs in cells cultured in medium supplemented with fatty acids (15)(16)(17). Each of these effects has been demonstrated (15) or speculated (16,17) to be transduced by the peroxisome proliferator-activated receptors (PPARs), nuclear receptor transcription factors that are regulated by fatty acid through derivative metabolites (15, 18 -21).
In this study, we evaluated the expression of CPT-I␤ mRNAs in rat tissues in fed and fasted states in order to explore mechanisms of fatty acid oxidation regulation. We show the existence of alternative human CPT-I␤ transcripts that arise from two closely approximated promoters that have tissuespecific or -preferred activities, each of which is alternatively spliced to generate the three previously identified isoforms ␤1, ␤2, and ␤3 (8,9). A cis-acting element that confers fatty acid regulation of both transcripts is identified that binds and is activated by PPAR/retinoid X receptor (RXR) heterodimers. Together with our previous findings, these results indicate that there is enzyme substrate-responsive regulated expression of alternative CPT-I␤ transcripts and alternative pre-mRNA splicing variants in tissues.

EXPERIMENTAL PROCEDURES
Genomic Cloning-A P1 clone was obtained (Genome Systems, St. Louis, MO) by hybridization screening with previously isolated human CPT-I␤1 cDNA (8). A MluI-AvrII fragment corresponding to ϳ5 kilobases upstream of the exon 2 splice acceptor was cloned into pBS (Stratagene) and verified by dideoxy sequencing (22). The sequence of this clone was identical to a portion of BAC clone CIT987SK-384D8. 2 Primer Extension-Total RNA from human heart, skeletal muscle, or liver (CLONTECH) was used in extension reactions (22) with [␥-32 P]ATP end-labeled primer U cDNA (5Ј-GGGGGTTGGTCGGCAC-.CTTC-3Ј) or M cDNA (5Ј-AGCTCGGGTTCACTCCTGTCCG-3Ј), followed by product resolution on 6% denaturing polyacrylamide gel. Parallel sequencing reactions with M cDNA and U GEN (5Ј-gaaccagaacccacccac-.CTTC-3Ј) on the genomic DNA template were resolved simultaneously.
Ribonuclease Protection-Fragments extending from the 5Ј-end to the PvuII site (ϩ182) of human CPT-I␤1 cDNA clones containing either exon M or exon U were isolated and cloned into pBS for use in generating hM␤ and hU␤ cRNA probes. Templates for hM␤2 and hU␤2 probes were created by fusing a PvuII-HindII fragment of human CPT-I␤2 (8) into these plasmids. A SalI-EcoRI fragment of rat CPT-I␤2 (9) was cloned into pBS for rat ␤2 cRNA probe template production. Complementary RNA probes were synthesized, and ribonuclease protection assays (RPAs) were performed as described (8,9).
Cell Transfection-CDM expression vectors for PPAR␣, RXR␣, and COUP-TF and DEAE-dextran HepG2 cell transfection procedures were as described previously (15). C 2 C 12 cells maintained at subconfluence in Dulbecco's modified Eagle's medium supplemented with 20% fetal bovine serum were differentiated after transfection by serum withdrawal (24), except where otherwise noted. C 2 C 12 cells were cotransfected using LipofectAMINE according to the manufacturer's recommendations (Life Technologies, Inc.) or by calcium phosphate/DNA co-precipitation (25). Cells were harvested after 48 h, at which time Ͼ90% of the cells were differentiated myotubes. Luciferase activities were corrected for transfection efficiency by normalizing to ␤-galactosidase activity produced by cotransfected RSV-␤-galactosidase. Fatty acids, POCA, gemfibrozil, and LG69 were prepared as stock solutions in Me 2 SO as described (15,26).

RESULTS
Alternative Human CPT-I␤ Transcripts Exist-The presence of two divergent sequences at the 5Ј-end of our CPT-I␤ isolates from a human cardiac cDNA library (Fig. 1a) suggested the possibility of alternative gene promoters. CPT-I␤ structural gene sequence (27) 2 indicated that our cDNAs correspond to alternative 5Ј-exons (Fig. 1c). Using human heart and skeletal muscle total RNAs, we performed primer extension assays to determine transcription start sites (TSSs). Reactions with an antisense primer corresponding to sequence within exon M (M cDNA ) (Fig. 1b, right panel) gave one strong product, indicating a single start site for this transcript at Ϫ523 relative to the initiation codon adenine. Similarly, assays using U cDNA , which corresponded to sequence spanning the exon U/2 junction, gave a single product (Fig. 1b, left panel). Parallel genomic sequencing with antisense primer U GEN , which has the identical 5Ј-end and is the same length as U cDNA , indicated that the major start site for this transcript is at Ϫ702. Ribonuclease protection assays with human heart RNA confirmed this, but indicated several minor upstream U TSSs between Ϫ710 and Ϫ745. Both 5Ј-exon splice donor signals and the exon 2 acceptor sequence conform to the consensus sequence (28) (Fig. 1d). RPAs and polymerase chain reaction-based analyses failed to detect the existence of mRNAs corresponding to previously reported cDNAs that extend immediately 5Ј of exon 2 (29) and excluded use of the 5Ј-untranslated exons in combination. Neither fulllength mRNA contains an AUG codon upstream of that found in exon 2. Thus, the human CPT-I␤ gene has two closely ap-proximated promoters that give rise to mRNAs with an identical initiation codon.
Alternative CPT-I␤ Promoters Are Differentially Utilized among Tissues-Variable intensity of primer extension products in reactions using various tissue RNAs suggested differing relative abundance of the two CPT-I␤ mRNA transcripts among tissues. To further evaluate this and as a first step in analyses of regulation of the promoters, we performed RPAs with human tissue RNA using cRNA probes that spanned the transcript-specific splice junctions. Thus, the M␤ probe was derived from a cDNA containing exon M, whereas the U␤ probe contained the exon U 5Ј-sequence (Fig. 2a). In assays with human heart RNA, two protected probe fragments were obtained when using each probe (Fig. 2b, H lanes). Protected fragments of the M␤ probe corresponded to mRNA derived from the M transcript (M␤) and from one or more different transcripts with alternative 5Ј exon(s) spliced to exon 2 (␤). Similarly, the pattern of U␤ probe protection was consistent with the presence of the U transcript mRNA (U␤) as well as alternative(s) (␤). A similar pattern was seen in RPAs using skeletal muscle RNA (Fig. 2b, S lanes), although the ratio M:U was 1:1, unlike the 4:1 ratio in heart. In contrast, a Ͼ20:1 preference of U transcript-derived mRNA was present in liver (Fig. 2b, L lanes) as well as in pancreas (data not shown). The relative intensities of the two M␤ probe bands were reciprocal to those of U␤ (i.e. [M␤/␤] M␤ ϭ [␤/U␤] U␤ ), indicating that any possible additional transcripts in which other upstream exons splice into the exon 2 acceptor could contribute only a very minor fraction to the total CPT-I␤ mRNA in the tissue samples. Thus, the two alternative CPT-I␤ promoters, M (muscle) and U (upstream, ubiquitous), show tissue-specific or -preferred utilization, and these are the major promoters active in the tissues examined.
Both CPT-I␤ Transcripts Are Alternatively Spliced-We have previously described alternative splicing of human (8) and rat (9) CPT-I␤ pre-mRNAs that gives rise to ␤1, ␤2, and ␤3 isoforms. ␤3 mRNA results from exon 5 skipping and represents a minor component of human and particularly rat total CPT-I␤ mRNAs in tissues. ␤2 message is derived from use of a cryptic exon 3 splice donor in combination with exon 4 skipping and contributes 5-25% of total tissue CPT-I␤. Expression of these mRNAs also shows tissue preference (8), and all independent human ␤3 cDNA isolates contained exon M sequence. To explore whether these findings reflected differential splicing pattern preference of the alternative transcripts, we performed RPAs using cRNA probes that spanned both the M or U transcript-specific splice junction and the ␤2 isoform-specific seam (Fig. 3a). These assays indicated that both the M and U transcripts are alternatively spliced to generate ␤1 and ␤2 mRNAs (Fig. 3b). The splicing patterns that give rise to the six CPT-I␤ mRNAs are depicted in Fig. 3c.
Tissue CPT-I␤ Gene Expression Is Up-regulated with Fasting-We examined the influence of fasting on steady-state CPT-I␤ mRNA levels in rat myocardium, skeletal muscle, and liver. Under these catabolic conditions, circulating free fatty acid concentrations are elevated due to peripheral lipolysis, and fatty acid oxidative activity is brisk in these tissues (2,30,31). As shown in Fig. 4, in 2-and 5-day fasted rats, there was a marked increase in steady-state CPT-I␤ mRNA levels in each tissue compared with those in rats fed ad libitum as determined by RPA. Increases in each CPT-I␤ isoform mRNA occurred with the fast, although the relative proportions of these messages appeared to change, particularly in liver (data not shown). Thus, there is a coordinate increase in both malonyl-CoA-sensitive (␤1) and -insensitive (␤2) 3 CPT-I expression under catabolic conditions, and this is likely to contribute to the observed accelerated fatty acid ␤-oxidation rates.
Fatty Acids Regulate CPT-I␤ Gene Expression through PPAR␣ Activation-To begin to evaluate the mechanism of the observed effect on CPT-I␤ mRNA in vivo, we performed transient transfection experiments with C 2 C 12 myotubes, HepG2 hepatocytes, and other cell lines. For initial studies, a fragment of the human CPT-I␤ gene extending from Ϫ1 to Ϫ1062 relative to the initiation codon adenine was fused to a luciferase reporter gene to generate U/M/Int 1 -LUC (Fig. 5a). This reporter includes both the M and U TSSs and retains the first intron (Int 1 ) and the exon 2 splice acceptor. Because circulating counter-regulatory hormone and free fatty acid concentrations are elevated with fasting compared with the fed state, we first evaluated the influence of these factors. Glucagon, adrenergic agonists, and adenylate cyclase activators did not affect U/M/ Int 1 -LUC activity. However, small but reproducible effects were seen when transfected cell culture medium was supplemented with fatty acids such as oleate. Fatty acids can regulate transcription through the PPAR nuclear receptor transcription factors. Since the PPAR␣ isoform is the dominant isoform in these tissues (15,21), we evaluated the role of PPAR␣ in mediating fatty acid-induced CPT-I␤ promoter activity using cotransfection of C 2 C 12 myotubes. As shown in Fig. 5b, cotransfection of PPAR␣ produced a 17-fold activation of the promoter when incubated in medium stripped of low molecular weight hydrophobic moieties, including fatty acid acids and retinoids (15,26). This apparent constitutive PPAR activity is attributable to activation of this transcription factor by endogenous fatty acids, as is typically seen (15,19,26). Fatty acid-mediated CPT-I␤ promoter activation up to 50-fold above basal activity was seen in PPAR␣-cotransfected cells incubated in medium supplemented with long-chain (oleate, C18:1) and mediumchain (decanoate, C10:0) fatty acids as well as gemfibrozil, POCA (a CPT-I inhibitor) (1,15), and the PPAR␣-specific agonist (8S)-hydroxyeicosatetraenoic acid (19).
Fatty Acids Co-regulate the U and M Transcripts-To evaluate differential or co-regulation of the U and M transcripts by fatty acids as well as other factors, we constructed additional reporters (Fig. 5a). In U-LUC, the U promoter is fused to the correctly spliced untranslated portions of exons U and 2 and to a luciferase reporter gene. M-LUC contains the structural gene sequence 5Ј of the M TSS fused to the untranslated portions of exons M and 2 and the luciferase gene. PPAR␣ activated both U-LUC (12.4-fold) and M-LUC (9.6-fold) in transfected C 2 C 12 cells (Fig. 5c), and these activations were not significantly different from that seen with U/M/Int 1 -LUC. Independent confirmation of co-regulation of the promoters was provided using assays of RNA obtained from cells transfected with U/M/Int 1 -LUC with or without PPAR␣ cotransfection. Here, coordinate increases in the intensity of protected probe bands corresponding to exons M and U were visualized (data not shown). These studies also confirmed correct exon U/2 and M/2 splicing of the plasmid reporter-derived mRNA. Our results indicate that the 3 G.-S. Yu and T. Gulick, manuscript in preparation.
FIG. 3. Alternatively spliced mature mRNA isoforms arise from both CPT-I␤ transcripts. a, the locations of probes used in RPA analyses are superimposed on schematics of partial CPT-I mRNA sequences (8). Expanses of various protected probe fragments are indicated. b, RPA was carried out with human skeletal muscle total RNA using the cRNA probes. Protected probe fragments corresponding to each transcript isoform are indicated. Each band was also visualized in assays with cardiac RNA. c, a schematic of the CPT-I␤ gene shows patterns of alternative splicing used to generate each mRNA. ␤2 and ␤3 are as described previously (8).

FIG. 4. CPT-I␤ mRNA increases in tissues with fasting.
RNAs harvested from skeletal muscle, heart, and liver of adult male rats fed ad libitum (0) or fasted for 2 or 5 days were analyzed by RPA using a probe from the rat CPT-I␤ cDNA (8). Loading was normalized by preliminary comparisons on Northern blots probed with an 18 S rRNA probe and using a glyceraldehyde-3-phosphate dehydrogenase (GAPDH) cRNA probe. two promoters are co-regulated by fatty acids through PPAR activation, consistent with the observed elevations in CPT-I␤ mRNA levels in heart, skeletal muscle, and liver. Furthermore, the cis-acting FARE responsible for this activity is necessarily located 5Ј of the U TSS.
The CPT-I␤ FARE Is a Prototypical DR1 Nuclear Receptor Response Element-The CPT-I␤ gene FARE was mapped using a U-LUC 5Ј-deletion series. PPAR␣ activation was maintained in constructs deleted for promoter sequences to Ϫ787 or a mere 85 base pairs from the U TSS (Fig. 6a). Inspection of gene sequence in this region revealed the existence of an antisense consensus nonsteroid-type nuclear receptor binding site, AG-GTCA (32). Flanking sequence was compatible with an imperfect direct hexamer repeat with single base pair separation (DR1; AGGGAAaAGGTCA) that might be permissive for receptor binding and activity (Fig. 6b). An analogous reporter (U-LUC FARE ) containing dinucleotide substitution within the consensus hexamer (AGGGAAaAccTCA) had unchanged basal reporter activity, but failed to respond to PPAR␣ or fatty acids (Fig. 6c). Identical FARE substitution mutations in the other reporters (U/M/Int 1 -LUC FARE and M-LUC FARE ) also abrogated PPAR responsiveness. Thus, the CPT-I␤ M and U transcripts are co-regulated from a single necessary and sufficient element centered at Ϫ769 relative to the AUG initiation codon, 67 base pairs upstream of the major U TSS and 246 base pairs upstream of the M TSS. In muscle cells, the basal activity of M-LUC was virtually identical to that of U/M/Int 1 -LUC, whereas that of U-LUC was substantially lower. Full maintenance of activity in M-LUC as compared with U/M/Int 1 -LUC indicated that there is no influence of intron 1 sequences on promoter activity in myocytes.
COUP Represses CPT-I␤ Promoter Activity from the FARE-Basal activities of the CPT-I␤ promoter reporters containing FARE mutations were only slightly lower than those of the wild-type promoters in fully differentiated myotubes (Fig. 6c). This was not the case in myoblasts, where the FARE mutation resulted in increased basal activity. Furthermore, PPAR␣ was less effective in activating the promoters in these undifferentiated cells (data not shown). These findings suggest the existence of a functional FARE-binding repressor that is preferen-  6. PPAR␣ co-regulates the U and M transcripts from a U promoter FARE. a, the ratios of CDMmPPAR␣-to CDM-cotransfected cell luciferase activities in C 2 C 12 cells transfected with a U-LUC 5Јdeletion series are shown. In these experiments, the medium was devoid of fatty acid supplements. Data are from a single representative experiment that was repeated three times. b, the human CPT-I␤ proximal U promoter contains an imperfect direct repeat of a nuclear receptor RGKTCA (reverse strand ϭ TGAMCY)-binding site (26,32). Numbers refer to locations with respect to the initiation codon (ATG ϭ ϩ1). The dinucleotide substitution incorporated into the FARE reporters is also depicted. c, luciferase activities in C 2 C 12 myocytes transfected with reporters containing the wild-type or a mutant FARE are shown. Reporter structures indicated schematically on the left correspond to those shown in Fig. 5a. Transfections were performed and analyzed as described for Fig. 5b. RLU, relative light units. tially expressed in the undifferentiated myoblasts. One likely candidate was COUP-TF (33), which is known to bind and repress transcription from DR1-type (33,34) and other (15) elements. Expression of this factor is robust in poorly differentiated myoblasts and declines with differentiation (34). Exogenous overexpression of COUP␣ in differentiated C 2 C 12 cells almost entirely attenuated PPAR␣ activity on U/M/Int 1 -LUC (Fig. 7) as well as the FARE in a heterologous promoter context (data not shown). With higher levels of COUP␣ expression, the basal activity of the promoter was diminished. Thus, COUP-TF transrepresses and modulates the fatty acid responsiveness of the CPT-I␤ promoter.
PPAR␣/RXR Heterodimers and COUP-TF Bind the CPT-I␤ FARE-Binding of tissue nuclear PPAR␣ to the CPT-I␤ FARE was confirmed in electrophoretic mobility shift assays. Using a probe that included U promoter sequence from Ϫ780 to Ϫ740, substantial binding of liver and cardiac nuclear proteins was observed (Fig. 8a). The existence of a PPAR␣⅐DNA complex was established by the appearance of a band representing a "supershifted" complex when binding reactions supplemented with anti-PPAR␣ antibody were resolved. Purified recombinant PPAR␣ also bound to the CPT-I␤ probe, but only in the presence of equimolar concentrations of recombinant RXR␣ (Fig.  8b), as is typical of PPAR binding to other gene response elements (15,21,26,32,35), and each PPAR isoform (␣, ␥, and ␦) bound the FARE with similar affinity (data not shown). The precise PPAR/RXR-binding site was delineated in studies using competitors containing partial probe sequences (S2) as well as a second probe corresponding to sequence from Ϫ757 to Ϫ780 (atcggTGACCTtTTCCCTacatt), which also produced PPAR␣⅐ RXR␣⅐DNA complexes (data not shown). COUP-TF also bound to the FARE probes as a monomer ((COUP) 1 ) or homodimer ((COUP) 2 ) (Fig. 8c). Complexes representing PPAR/RXR and (COUP) 2 were both observed in co-incubation assays unless the probe concentration was limiting, in which case there was mutual competitive binding (data not shown).
RXR Differentially Regulates the CPT-I␤, ACO, and MCAD FAREs-Cooperative binding of PPARs with RXRs on the CPT-I␤ FARE led us to evaluate the response of this promoter and element to retinoid activators of RXR (36) as well as combinatorial receptor and cognate ligand activities. In these studies, we compared the function of U-LUC with that of the pleiotropic NRRE1 of human MCAD (15) as well as the prototypical DR1 PPAR response element of the ACO gene promoter (35) in HepG2 cell cotransfections. CPT-I␤ U-LUC failed to respond to the RXR-specific retinoid ligand LG69, even with RXR␣ cotransfection (Fig. 9a), whereas this ligand stimulated the MCAD (Fig. 9b) and ACO (data not shown) elements in an RXR-dependent manner. There was some constitutive PPAR␣ activity on all elements, but this was most robust for the CPT-I␤ FARE. Each element showed additive activities of PPAR-and RXR-specific activators/ligands in PPAR-cotransfected cells. However, overexpression of RXR unexpectedly severely blunted CPT-I␤ promoter activity while potentiating PPAR stimulation of the ACO and MCAD elements, indicating that CPT-I␤ FARE element structure dictates novel regulatory properties. DISCUSSION We have demonstrated that the human CPT-I␤ gene has two transcripts that generate mRNAs with differing 5Ј-untranslated regions. These transcripts show tissue-enriched or -specific expression, and each is alternatively spliced within the coding region. One transcript is present in heart and muscle (M), whereas the other is ubiquitously (U) expressed in tissues with high fatty acid oxidation rates. Fatty acids regulate production of both transcripts through activation of PPAR␣, which binds as a heterodimer with RXR on a FARE located in the proximal U promoter region. The physiological significance of this regulation is reflected in a robust increase in CPT-I␤ mRNA levels in rat tissues under catabolic conditions, when elevated circulating free fatty acids are presented as fuel and are available to activate gene expression.
Both the CPT-I␤ U and M promoters are TATA-less. This is somewhat surprising, as this feature is more typical of "housekeeping" genes and those whose products are required early in development (37). The M promoter does have two proximal GC boxes that may facilitate transcription factor IID binding and RNA polymerase II activity (37,38). The location of the major U transcript TSS at Ϫ702 differs from that previously reported by Yamazaki et al. (39), who suggested that it was located at Ϫ745 based on 5Ј-rapid amplification of cDNA ends. Although we saw no band consistent with this start site using primer extension, RPA did confirm this and other secondary start sites between Ϫ702 and Ϫ745. Alternative U transcript start sites are of interest because additional functional regulatory elements overlap or are contained within the Ϫ702 to Ϫ745 region. 4 Furthermore, the FARE is located a mere 18 -30 bases upstream of the Ϫ745 TSS, where it would be expected to interfere with the binding of polymerase II components. The existence of numerous proximal U promoter regulatory elements in the absence of a TATA box suggests that regulated activities of cognate factors may influence the precise locus of polymerase II binding as well as its activity.
Although the U and M TSSs are closely approximated, the existence of mutually exclusive first exons, as well as preferential utilization among tissues, indicates that these are bona fide alternative promoters. Vigorous M promoter activity in muscle and heart, but not in liver and nonmuscle tissues, has two mechanistic explanations. First, muscle-specific factors may transactivate cis-acting elements that selectively control the M promoter. Alternatively, factors in nonmuscle cells may transrepress to eliminate M promoter activity. Candidate elements include GC boxes in the proximal M promoter, which may regulate gene expression through Sp1/Sp3 competition, for example (37,40,41). However, regardless of mechanism, the distinction of M and U promoters does not dictate that element(s) that specifically regulate M expression are necessarily located only in the region between the two TSSs. Pertinent to 4  this is the existence of a conserved E box and a consensus myocyte enhancer factor-2 site in the proximal U promoter region of human and rat genes. 4 The activity of the FARE indicates that that the converse also applies: a cis-acting element within the U promoter is capable of controlling both promoters. Differential regulation of CPT-I␤ M and U promoters by various transcription factors, functioning alone or through cooperative interactions, may ramify in altered CPT-I enzyme kinetics through preferential transcript splicing patterns (see below).
We have recently described alternative mRNA splicing within the coding region of CPT-I␤ (8,9). ␤2 and ␤3 mRNAs are produced by cryptic splice donor utilization and/or exon skipping and encode enzymes that are deleted for a transmembrane domain or an adjacent region that is necessary for physiological allosteric regulation of CPT-I by malonyl-CoA (1,42). 3 Because of precedent for differential splicing patterns of alternative transcripts (28), we investigated this possibility for this gene. Although we demonstrate that both the M and U transcripts are alternatively spliced, we cannot exclude the possibility that coding region alternative splicing is co-regulated with alternative promoter utilization or activity. However, even if this is true, the relative abundance of mRNAs containing M versus U exons and of ␤1 versus ␤2 versus ␤3 mRNAs in heart, muscle, and liver (8) indicates that alternative splicing is also independently regulated.
During the preparation of this manuscript, Mascaro et al. (43) reported regulation of the CPT-I␤ promoter by PPAR␣ and PPAR␥ isoforms and cognate activators in cotransfected CV1 cells. These investigators mapped the identical FARE, but found less vigorous PPAR effects (3-4-fold) than we report here, possibly due to the different transfected cell type. Regulation of the alternative promoters was not addressed, and PPAR isoform-specific activators were evaluated rather than the ordinary dietary and endogenous fatty acids of our study. Although much recent attention has focused on high affinity isoform-specific PPAR ligands (19,20,44), activities of oleate and palmitate at concentrations that are routinely achieved in the circulation suggest that the physiologically relevant activators of PPAR␣ are these common fatty acids, regardless of the higher affinity of rare metabolites, as we have previously suggested (15). This point is underscored by the demonstrated efficacy of decanoic acid in activating PPAR␣ on this FARE. Since medium-chain fatty acids make up a major component of milk, this implies a role for this factor in the observed upregulation of CPT-I and other genes encoding fatty acid oxida- FIG. 8. PPAR/RXR heterodimers and COUP homodimers compete for FARE binding. a, nuclear extracts (Nuc Ext) of rat heart (H) or liver (L) were incubated with the FARE probe and resolved by polyacrylamide gel electrophoresis. Binding reactions included anti-PPAR␣ antibody (Ab) as indicated or 50-fold molar amounts of nonspecific (NS) or specific (S1 and S2) unlabeled competitor (Comp) probes. S1 is identical to the labeled probe, and S2 contains probe sequence that flanks but does not include the FARE. Bands corresponding to PPAR␣-containing complexes are indicated. b, bacterially overproduced purified hexahistidine-tagged PPAR␣ and RXR␣ were incubated with a CPT-I␤ FARE oligonucleotide probe with or without anti-PPAR␣ antibody prior to electrophoresis. c, nuclear extracts from HeLa cells infected with recombinant vaccinia expressing COUP␣ (25) were combined with purified PPAR␣ and RXR␣ in FARE oligonucleotide binding assays as indicated. COUP binds as a dimer ((COUP) 1 ) and a homodimer ((COUP) 2 ) on this element.

FIG. 9. RXR exerts differential effects on the CPT-I␤ FARE compared with other PPAR response elements.
HepG2 cells were transfected with CPT-I␤ U-LUC (a) or MCAD(NRRE1) 3 -tkLUC (b) (15), with or without PPAR␣ or RXR␣ cotransfection, and incubated in charcoal-stripped serum with or without 250 M decanoic acid (C10:0) supplementation and with or without 1 M LG69 RXR-specific ligand (36) as indicated. Luciferase activities were normalized for transfection efficiency using RSV-␤-galactosidase and to that seen in cells incubated without receptor cotransfection and without medium supplementation (ϭ1.0). Data are from representative experiments repeated at least twice. RLU, relative light units. tive enzymes during suckling (12)(13)(14).
The CPT-I␤ FARE sequence conforms to the conventional PPAR/RXR DR1-type element (26,32), and although the 5Јhalf-site (antisense AGGGAA) deviates from the consensus sequence, its flanking region is A/T-rich, consistent with the structure of other PPAR response elements (45). There are unique features of this element with respect to RXR and retinoid activities compared with other DR1 and complex elements, as exemplified by those of the ACO (35) and MCAD (15) genes, respectively. For these and other previously described elements (26,35), both PPAR and RXR are cognate ligandresponsive, and the receptors function cooperatively. In the case of the CPT-I␤ FARE, despite cooperative binding of the receptors, overexpression of RXR attenuates PPAR and fatty acid responsiveness. Furthermore, the FARE is not activated by the RXR-specific ligand LG69 (36) in the absence of PPAR coexpression. These distinct activities may involve polarity of receptors on the element (46), lack of capacity for RXR homodimer binding, or co-repressor or co-activator interactions (32, 46 -49). Alternatively, overexpression of RXR may enable recruitment of a heterodimer partner that represses either constitutively or in the absence of cognate ligand (32,46). Regardless of mechanism, our findings indicate that RXR may act in cells and in vivo as a ligand-independent repressor on the CPT-I␤ FARE, providing a mechanism to discriminate fatty acid-and retinoid-responsive expression of this gene.
COUP transcription factors are widely expressed in tissues, where they compete with various nuclear receptors to modulate constitutive and hormone-regulated gene expression (25,33,34). Our finding that COUP␣ represses transcription from the CPT-I␤ FARE is not surprising, as the DR1-type element is the preferred COUP homodimer-binding site and functional element (32,33). This and other nuclear receptors that regulate DR1 elements (32), each with differential and regulated tissue expression, are candidate influences on CPT-I␤ gene expression, and each has the potential to modulate the vitality of promoter activation in response to fatty acids. In particular, this may be reflected in COUP-mediated inhibition in developing tissues and undifferentiated cells or in pathophysiological states associated with retrogression in gene expression programs (33,34,41).
Regulated expression of CPT-I␤ by fatty acids in muscle and heart, where PPAR␣ is the dominant PPAR isoform (15,21), indicates a mechanism by which fuel availability can control fuel preference and utilization in these tissues. This also is true in liver, where our clear demonstration of regulated expression of CPT-I␤ contrasts with previous reports of exclusive CPT-I␣ expression (1). Of particular note, hepatic induction of CPT-I␤2 mRNA with fasting may explain the recognized decrease in hepatic CPT-I sensitivity to malonyl-CoA inhibition in catabolic states as well as the increase in V max (1,2). CPT-I␤ expression in adipose tissue, where PPAR␥ is the dominant isoform (44,50,51), was not explored in our study. There is precedent for both ligand-mediated transrepression (50) and transactivation (44,51) of different PPAR elements and gene promoters in this tissue, such that unique CPT-I␤ regulation may be manifest. Speculation (43) that activation of skeletal muscle CPT-I␤ expression by PPAR␥ agonists provides a mechanism for the hypoglycemic action of thiazolidinediones (44,51) requires superimposition of an additional layer of complexity: the resulting increase in fatty acid oxidation would be accompanied by a decrease in glucose disposal by a Randle mechanism (2,23). This paradox might be explained by partial agonist activity of the drug in a context of endogenous receptor ligands in vivo, by competitive activities of the PPAR isoforms, or by regulation of CPT-I and fatty acid oxidation in adipose tissue with secondary effects at sites of glucose disposal. In any case, regulation of CPT-I␤ by PPAR␣ is consistent with the expanded role that we first proposed for this factor (15) as a nutrient-and metabolic state-responsive governor of the coordinate expression of mitochondrial fatty acid oxidative enzymes that is essential for metabolic homeostasis.