Identification of Glycerol-3-phosphate Acyltransferase as an Adipocyte Determination and Differentiation Factor 1- and Sterol Regulatory Element-binding Protein-responsive Gene

We demonstrate that the mRNA levels of glycerol-3- phosphate acyltransferase (GPAT), a mitochondrial enzyme catalyzing the initial step in glycerolipid synthe- sis, are induced during the differentiation of 3T3-L1 preadipocytes to adipocytes and following ectopic ex- pression of rat adipocyte determination and differentiation factor 1 (ADD1), a protein with high homology to the human sterol regulatory element-binding protein-1 (SREBP-1). The increase in GPAT mRNA levels that oc- curs during differentiation is partially prevented by ectopic expression of a dominant negative form of ADD1. Nucleotide sequences corresponding to the proximal promoter of the murine mitochondrial GPAT gene (Jer- kins, A. A., Liu, W. R., Lee, S., and Sul, H. S. (1995) J. Biol. Chem. 270, 1416–1421) bound SREBP-1a and NF-Y in electromobility shift assays. In addition, GPAT promot-er-luciferase reporter genes were stimulated by co-ex- pression of SREBP-1a. This increase was attenuated when either a dominant negative form of NF-Y was co- transfected into the cells or when the GPAT promoter contained mutations in the putative binding sites for SREBP-1a or NF-Y. These studies demonstrate that the regulated expression of the mitochondrial GPAT gene requires both NF-Y and -galactosidase and a luciferase reporter gene under the control of the indicated promoter in the absence or presence of 20 ng of either pSREBP-1a (pCMV-CSA10, encoding amino acids 1–490 of SREBP-1a) or pSREBP-2 (pCMV-CS2, encoding amino acids 1–485 of SREBP-2). The cells were incubated for 20 h in medium containing 10% fetal bovine serum and lysed, and enzyme assays were performed as described under “Experimental Procedures.” The luciferase activities in relative light units (RLU) are given after normalization for variations in transfection efficiency. The corrected luciferase values obtained from duplicate dishes varied by less than 10%. The numbers in parentheses indicate the -fold increase in luciferase activity relative to cells that were not transfected with plasmids encoding either SREBP-1a or SREBP-2. Similar results were obtained in duplicate experiments. The reporter genes were under the control of promoters derived from FPP synthase (pFPPS-0.247), LDL receptor (pLDL SRE) or HMG-CoA synthase (pSYN SRE) or GPAT (pGPAT-A) genes.

Sterol regulatory element-binding proteins (SREBPs) 1 and adipocyte determination and differentiation factor 1 (ADD1) are members of the helix-loop-helix-leucine zipper family of transcription factors (1)(2)(3). SREBP-1 was purified from HeLa cells based upon its ability to bind specifically to a 10-bp sequence, termed the sterol regulatory element-1 (SRE-1) (4). This element was shown to be involved in sterol-regulated transcription of the LDL receptor and HMG-CoA synthase genes (2,4). Subsequently, two human SREBP genes were identified, SREBP-1 and SREBP-2, that encode at least three proteins (SREBP-1a, SREBP-1c, and SREBP-2) of approximately 125 kDa (1,2). ADD1 was cloned independently as the result of a screen to identify rat adipocyte proteins that bound to a DNA sequence termed an E-box (3). The E-box motif (CANNTG) is present in the promoters of a number of genes including fatty acid synthase and S14 (3,5,6). ADD1 mRNA levels were shown to increase significantly during the differentiation of preadipocytes to adipocytes (3,7), consistent with a role for this transcription factor in cell differentiation. The transcription of the endogenous fatty acid synthase gene or reporter genes under the control of the fatty acid synthase promoter was induced by ADD1 and SREBP-1, consistent with a functional role for these proteins (3,5,6,8). Comparison of the nucleotide/amino acid sequence derived from rat ADD1 (3) and human SREBP-1 (1) indicates that they encode the same protein.
SREBPs/ADD1 are unique transcription factors containing two transmembrane domains that anchor the proteins to the endoplasmic reticulum and the nuclear envelope (9). Low cellular sterol levels stimulate a complex set of proteolytic cleavage events of the precursor form of SREBPs to release the mature 68-kDa amino-terminal domain from the endoplasmic reticulum (10,11). This soluble, transcriptionally active form of SREBP is targeted to the nucleus where it binds to cis-elements in the promoters of SREBP-responsive genes (2,4). Three distinct cis-elements have been identified: SRE-1 in the promoters of genes encoding the LDL receptor and HMG-CoA synthase (12), SRE-3 in the promoter of the gene encoding FPP synthase (13), and an E-box motif in the promoters of the fatty acid synthase and S14 genes (3,5,8). Motifs corresponding to SRE-1 and SRE-3 have been identified in the promoter of the squalene synthase gene (14). However, the functional roles of these elements in this latter gene have not been reported.
Increased transcription of the LDL receptor, HMG-CoA synthase, and FPP synthase genes requires a second DNA-binding protein in addition to SREBP. Studies with FPP synthase and HMG-CoA synthase promoter-reporter genes identified NF-Y as the second crucial transcription factor (15). The binding of NF-Y to an inverted CCAAT box motif in the proximal promoter of the FPP synthase gene stimulated the binding of SREBP to an adjacent SRE-3 and allowed active transcription of an FPP synthase promoter-reporter gene (16). A different ubiquitous transcription factor, Sp1, is necessary for high expression of the LDL receptor gene (17). The binding of SREBP to SRE-1 in the promoter of the latter gene stimulated the binding of Sp1 to an adjacent low affinity binding site for Sp1 (17,18).
Based on these studies, we hypothesized that other SREBP/ ADD1-responsive genes would require NF-Y or Sp1 in addition to SREBP/ADD1 for high transcription. Consequently, we searched a number of data bases to identify genes involved in lipid biosynthesis that contained potential NF-Y and SREBP/ ADD1-binding sequences in their proximal promoters. This analysis identified a number of genes, including the one encoding glycerol-3-phosphate acyltransferase (GPAT).
GPAT catalyzes the acylation of the sn-1 hydroxyl of glycerol-3-phosphate, the initial step of glycerolipid synthesis. The product, 1-acyl-sn-glycerol-3-phosphate is a committed intermediate in the biosynthesis of both triglycerides and phospholipids (19). There are two major mammalian forms of GPAT that are localized to either the endoplasmic reticulum or mitochondria (19). In animal studies, the activities of both GPAT isoenzymes were regulated by changes in the nutritional and hormonal status of the animal (20). However, only the protein (21), cDNA (22), and gene (23) encoding the mitochondrial GPAT have been isolated to date. Jerkins et al. (23) identified 1447 bp of the murine GPAT proximal promoter and demonstrated that the expression of GPAT promoter-reporter genes in cultured cells was regulated by glucose and insulin, presumably via a carbohydrate response element. In addition, reporter genes containing a minimum of 86 bp of the GPAT promoter were expressed at 2-fold higher levels in 3T3-L1 adipocytes as compared with preadipocytes (23). The reason for the increased expression of the reporter gene in adipocytes was not determined.
In the current study, we report the identification of distinct cis-elements in the GPAT proximal promoter that interact with either SREBP/ADD1 or NF-Y. We demonstrate that the expression of GPAT promoter-reporter genes in transiently transfected cells is increased by co-expression of mature SREBP-1a or SREBP-2. We also demonstrate that high levels of expression of GPAT promoter-reporter genes, in response to co-expressed SREBP-1a, are attenuated when the binding sites for either SREBP or NF-Y in the promoter are mutated or when cells are co-transfected with a plasmid encoding a dominantnegative form of NF-Y. Finally, we show that the endogenous GPAT mRNA levels in 3T3-L1 cells increase during differentiation of preadipocytes to adipocytes and in response to ectopic expression of ADD1.

EXPERIMENTAL PROCEDURES
Materials-DNA restriction and modification enzymes were obtained from Life Technologies, Inc. 32 P-Labeled nucleotide triphosphates were obtained from Amersham Corp. pRSETB (Invitrogen) containing both a partial sequence of SREBP-1a (amino acids 1-490) and T7 plus polyhistidine tags and pCMV-CSA10, which encodes amino acids 1-490 of SREBP-1a, and pCMV-CS2, which encodes amino acids 1-485 of SREBP-2, were kindly provided by Dr. T. Osborne (Department of Molecular Biology and Biochemistry, University of California, Irvine). Lipoprotein-deficient fetal calf serum was purchased from PerImmune. pNF-YA29, which encodes a dominant negative form of the A subunit of NF-Y, and antibodies to human NF-YA and NF-YB were generous gifts from Dr. R. Mantovani (University of Milan, Italy). Antibodies to CBF-HSP were a generous gift of Dr. B. Wu (Northwestern University). The T7 antibody was from Invitrogen. pFPPS-0.247 is a promoter-reporter construct containing 247 bp of the FPP synthase proximal promoter (24). The reporter genes under the control of promoters derived from FPP synthase (pFPPS-0.247), LDL receptor (pLDL SRE), or HMG-CoA synthase (pSYN SRE) genes have been described (15,17). The sources of all other reagents and plasmids have been given (13,15,24,25).
Cell Culture-HepG2 cells were cultured as described previously (24). Following transient transfection, the cells were incubated for 20 h in media supplemented with either 10% lipoprotein-deficient calf serum and 5 M compactin in the absence (inducing media) or presence (repressing media) of sterols (10 g/ml cholesterol and 1 g/ml 25-hydroxycholesterol) (13,15) or 10% fetal bovine serum, as indicated in the figure legends.
3T3-L1 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum at 5% CO 2 and 37°C. Differentiation of preadipocytes to adipocytes was achieved by allowing the cells to reach confluence before the addition of Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 0.5 mM 3-isobutyl-1-methylxanthine (Sigma), 0.25 M dexamethasone (Sigma), and 1 g/ml insulin (Sigma). After 2 days, and every 2 days thereafter, fresh medium (Dulbecco's modified Eagle's medium plus 10% fetal bovine serum) was added to the cells.
GPAT Promoter-reporter Gene Constructs-The GPAT promoter was obtained by polymerase chain reaction using mouse genomic DNA as a template. The 3Ј-primer corresponded to nucleotides Ϫ12 to ϩ17 of the published genomic sequence ( Fig. 1) (23) and an additional 10 nucleotides that contained a BglII restriction site. The 5Ј-primer contained an SstI restriction site and nucleotides that corresponded to Ϫ200 to Ϫ176 of the published GPAT genomic sequence ( Fig. 1) (23). The polymerase chain reaction product was subcloned directionally into pGL2 basic luciferase vector (Promega) to produce pGPAT-A and sequenced to confirm the wild-type GPAT sequence. The subcloned 217-bp fragment of the GPAT promoter was used as a template in subsequent polymerase chain reactions using the same 3Ј-primer and 5Ј-primers corresponding to Ϫ176 to Ϫ160, Ϫ160 to Ϫ142, or Ϫ78 to Ϫ58 of the genomic sequence. The products were subcloned into pGL2 to produce pGPAT-B, pGPAT-C, and pGPAT-D, respectively. Two additional 5Ј-primers were also used, 5Ј-TACCGAGCTCCTTGGCTGATCAAGCTCAGCC-3Ј and 5Ј-TACCGAGCTCATTGGCTGATCAAGCTAAACCTAGCAGACC-3Ј that contained mutations (underlined) in the putative NF-Y or SREBP binding sites, respectively. The promoter-reporter constructs containing these mutations were pGPAT-Y and pGPAT-S, respectively. All polymerase chain reaction products were sequenced after subcloning into pGL2 in order to confirm the presence of wild-type or mutant sequences.
Oligonucleotides-Oligonucleotides were either synthesized by Life Technologies, Inc., or were obtained by polymerase chain reaction using the GPAT promoter fragment subcloned into pGL2. Double-stranded DNA corresponding to GPAT cDNA was obtained by the polymerase chain reaction following the synthesis of cDNA from mouse liver mRNA using Superscript II (Life Technologies, Inc.) and a 3Ј-primer (CAG-CACCACAAAACTCAGAA) corresponding to the carboxyl terminus of the mouse GPAT protein. The subsequent polymerase chain reaction was carried out with the same 3Ј-primer and a 5Ј-primer (TCCA-GAAACCCAGCAGACGA) complementary to the mouse cDNA (22). The resulting oligonucleotide was cut with EcoRI and PstI, gel-purified, and subcloned into Bluescript SK (Stratagene). The construct was sequenced in order to confirm the sequence of the GPAT insert. The EcoRI/PstI restriction fragment was gel-purified and used in Northern blot analysis (28).
Northern Blot Analysis-Total RNA was isolated from 3T3-L1 cells, either before or after the initiation of differentiation, using Trizol reagent (Life Technologies, Inc.). Ten-g aliquots of total RNA were fractionated by electrophoresis on 1% agarose/formaldehyde gels, transferred to nylon membranes and cross-linked with UV light as described (28). [␣-32 P]dCTP-radiolabeled cDNA probes were generated by random priming (Pharmacia Biotech Inc.). Hybridization and quantitation using a PhosphorImager (Molecular Dynamics) were as described (29). Blots were also hybridized to a 36B4 probe (7), analyzed on the phos-phorImager, and the values used to correct for differences in RNA loading in each lane.
Gel Mobility Shift Assay-The double-stranded DNA oligonucleotides used in these studies corresponded to Ϫ192 to Ϫ171 (5Ј-SRE), The 95-bp BglII/SstI restriction fragments of pGPAT-D and GPAT-S contain wild-type binding sites for both SREBP-1a and NF-Y or a wild-type binding site for NF-Y and two mutations in the putative SREBP-1a binding site, respectively.
Crude HeLa nuclear extracts, provided by Dr. M. Carey (UCLA), were separated by size exclusion chromatography on a Superose 12 (Pharmacia) fast protein liquid chromatography column, and the fractions containing enriched NF-Y, as determined by Western blot analysis, were identified as described (15). Recombinant SREBP-1a, containing N-terminal T7 and polyhistidine (His 6 ) tags was purified to homogeneity from Escherichia coli extracts by nickel affinity chromatography as described by Sanchez et al. (17).
Complementary single-stranded DNA, corresponding to the indicated nucleotides of the GPAT promoter and containing either the wild-type sequence or the indicated mutations, were annealed, and the double-stranded probe was isolated from acrylamide gels (28). Radiolabeled probes were generated in the same manner following an initial radiolabeling of a single-stranded oligonucleotide with 32 P (28). The radiolabeled probes (20,000 cpm; 1.5 fmol) were used in gel mobility shift assays in the presence of nonfat milk (2.5 mg/ml) or in the presence of either partially purified NF-Y (5 g) or recombinant SREBP-1a as described previously (13,15). Where indicated, antibody (0.2 g) to NF-YA, NF-YB, CBF-HSP, or the T7 tag (Novagen) on SREBP-1a were added to the incubations on ice for 30 min before gel electrophoresis was performed (13,15).
Transient Transfections and Reporter Gene Assays-Details of the transient transfection of HepG2, CV-1, or Chinese hamster ovary cells with luciferase reporter genes, a plasmid encoding ␤-galactosidase, pCMV-CSA10 (encoding SREBP-1a), and pCMV-CS2 (encoding SREBP-2) appear in previous publications (13,15,29). Cells were then incubated for 20 h in the indicated media and lysed, and the activities of reporter genes were determined as described (13,15). The activity of ␤-galactosidase was used to normalize for transfection efficiencies (13).

RESULTS
GPAT Expression during Adipocyte Differentiation-Analysis of the published nucleotide sequence for the murine GPAT promoter (23) indicated that it contained a number of potential binding sites for SREBP/ADD1 and NF-Y (Fig. 1). Consequently, we hypothesized that transcription of the endogenous GPAT gene might be enhanced under conditions in which the nuclear levels of NF-Y and/or ADD1/SREBP were increased. Previous studies have demonstrated that the differentiation of 3T3-L1 preadipocytes to adipocytes is associated with an increase in the expression of ADD1 mRNA (3,7). Differentiated adipocytes also accumulated large amounts of triglyceride, presumably as a result of increased synthesis of this lipid. Fig. 2 demonstrates that the mRNA levels for GPAT, an enzyme that catalyzes the initial step in triglyceride and phospholipid synthesis (19), increased 7.5-fold during the differentiation of 3T3-L1 preadipocytes to adipocytes. Maximal GPAT mRNA levels were observed 6 -8 days after differentiation was initiated ( Fig. 2 and data not shown).
Effect of Ectopic Expression of either ADD1 or a Dominant Negative Form of ADD1 on GPAT mRNA Levels-3T3-L1 cells undergo many changes in gene expression during differentiation to adipocytes (7). In order to determine whether the increased expression of GPAT mRNA was a direct consequence of the increased expression of ADD1, we infected preadipocytes with a retrovirus that expressed either ADD1 or ADD1-DN, a dominant-negative form of ADD1 that has a point mutation in the DNA-binding domain and represses differentiation of 3T3-L1 cells (7).
Introduction of ADD1 into 3T3-L1 preadipocytes via retrovi-ral infection resulted in a 6.7-fold increase in GPAT mRNA levels (Fig. 3, lane 3 versus lane 1) under conditions in which preadipocytes do not differentiate into adipocytes (7). The mRNA levels for PPAR␥, adipsin, and aP2 did not change under these conditions (Fig. 3). Furthermore, GPAT, PPAR␥, adipsin and aP2 mRNA levels were unaffected, or decreased slightly, following the infection of the cells with the vector containing ADD1-DN (Fig. 3, lane 2 versus lane 1). Six days after cell differentiation was initiated, GPAT mRNA levels were increased 2.3-fold in cells infected with the vector alone (Fig. 3, lane 4 versus lane 1). This increase in GPAT mRNA levels was attenuated in cells that ectopically expressed the dominant-negative form of ADD1 (Fig. 3, lane 5 versus lane  4). Expression of the dominant-negative form of ADD1 also resulted in reduced mRNA levels for PPAR␥, adipsin, and aP2 but had no effect on the expression of either ribosomal RNA or an mRNA identified by probe 36B4 (Fig. 3, data not shown). In contrast, GPAT mRNA levels increased 11.3-fold in cells that had differentiated for 6 days in the presence of ectopically FIG. 1. Schematic representation of the proximal promoter of the murine GPAT gene. The nucleotide sequence of this promoter has previously been reported (23). The numbering is given relative to the transcription start site that is designated as ϩ1. The sequences that correspond to putative binding sites for SREBP/ADD1 (1, 2, and 4) or NF-Y (3) (ATTGG) within the region Ϫ200 to ϩ17 are indicated.
FIG. 2. Induction of GPAT mRNA during the differentiation of 3T3-L1 preadipocytes to adipocytes. 3T3-L1 preadipocytes were grown to confluence. On day 0 the medium was changed to the differentiating medium as described under "Experimental Procedures." At the times indicated, total RNA was isolated, and 10-g aliquots were separated by electrophoresis on a denaturing agarose gel, transferred to a nylon membrane, and hybridized initially with the probe for GPAT followed by the control 36B4 probe, as described under "Experimental Procedures." The radioactivity associated with the different mRNAs was quantitated on a PhosphorImager screen, and the levels of GPAT mRNA were normalized for differences in RNA loading. The GPAT mRNA levels are given compared with the level obtained at day 0. 3T3-L1 preadipocytes were infected with an empty retroviral vector (vector), or a vector containing either ADD1 or a dominant negative form of ADD1 (ADD1-DN) as described under "Experimental Procedures." Total RNA was isolated at confluence (day 0) or 6 days after the addition of the differentiating medium (day 6). RNA (10 g/lane) was separated by electrophoresis, transferred to a nylon membrane, and hybridized with 32 P-labeled probes corresponding to the indicated cDNAs or a control 36B4 cDNA. The relative GPAT mRNA levels were determined after quantitation on a PhosphorImager and normalization using the control 36B4 mRNA; the relative values in lanes 1-6 were 1.0, 0.3, 6.7, 2.3, 1.8, and 11.3, respectively. In a second experiment, the corresponding GPAT mRNA values were 1.0, 0.8, 7.3, 16.1, 8.8, and 45.6, respectively. expressed ADD1 (Fig. 3, lane 6 versus lane 1), consistent with a stimulatory role for ADD1 in the expression of GPAT mRNA.
Taken together, these results demonstrate that GPAT mRNA levels are induced during adipocyte differentiation. In addition, ectopic expression of ADD1, either in preadipocytes or during adipocyte differentiation, results in increased levels of endogenous GPAT mRNA.
Responsiveness of GPAT Promoter-reporter Genes to Co-expressed SREBP-1-In order to determine whether transcription of the GPAT gene was responsive to SREBP/ADD1, we generated six different murine GPAT promoter constructs by polymerase chain reaction and linked them to a luciferase reporter gene. The GPAT promoter-reporter genes are shown schematically in Fig. 4B. The numbering of the nucleotides is taken from the published sequence of the murine gene (23). The putative binding sites for SREBP/ADD1 and NF-Y in the mu-rine proximal promoter are indicated (Figs. 1 and 4B). Each promoter-reporter gene was transiently transfected into HepG2 cells together with a plasmid encoding ␤-galactosidase under the control of the cytomegalovirus promoter and in the presence or absence of pCMV-CSA10. The latter plasmid encodes the transcriptionally active, mature form of SREBP-1a and for simplicity is referred to as pSREBP-1a below. After 20 h, the luciferase activity was determined in cell lysates, and the values were normalized for any differences in transfection efficiency, as determined from the assay for ␤-galactosidase activity (13,15).
The results demonstrate that the expression of pGPAT-A, a construct that contains three potential SREs within the GPAT proximal promoter, was increased in proportion to the amount of co-transfected pSREBP-1a (Fig. 4A). Subsequent transfection studies utilized 20 ng of pSREBP-1a, since stimulation of the GPAT promoter-reporter genes was not maximal under these conditions (Fig. 4A). Fig. 4B demonstrates that co-transfection of pSREBP-1a to- gether with pGPAT-A, pGPAT-B, or pGPAT-C resulted in a 12-14-fold increase in luciferase activity. Under these same conditions, pSREBP-1a stimulated the luciferase activity 4.4fold in cells transfected with pGPAT-D (Fig. 4B). Thus, deletion of the two potential 5Ј-SREs reduced, but did not block, the induction of the reporter genes by co-expressed SREBP-1a. In contrast, luciferase activity was not increased when cells were co-transfected with pSREBP-1a together with either pGPAT-S or pGPAT-Y (Fig. 4B). These latter two promoter-reporter constructs contain mutations in the putative binding sites for either SREBP-1a (CTCAGCCTAG) or NF-Y (ATTGG), respectively ( Figs. 1 and 4). Basal expression of pGPAT-S and pG-PAT-Y approximated 10 -15% of that of the pGPAT-A construct (Fig. 4). Qualitatively and quantitatively similar results were obtained when experiments, identical to those shown in Fig. 4, were carried out with either CV-1 or Chinese hamster ovary cells (data not shown).
The results indicate that 1) murine GPAT promoter-luciferase genes are activated in response to increased expression of SREBP-1a, 2) a putative binding site for NF-Y is important for SREBP-1a-regulated expression of GPAT promoter-reporter genes, and 3) pGPAT-D may contain a novel binding site for SREBP-1a, since it does not contain the functionally important TCAC sequence that is found in the core of SRE-1, SRE-3, and E-box motifs. Instead, pGPAT-D contains the sequence TCAG, that when mutated to TAAA impairs the induction of the reporter gene by co-expressed SREBP-1a. Nucleotide sequences surrounding the TCAG have only 40 -60% identity with SRE-1, SRE-3, or E-box motifs and thus appear to represent a novel binding sequence for SREBP-1a.
SREBP Activation of GPAT-reporter Genes Requires NF-Y-We have previously shown that sterol-regulated transcription of FPP synthase and HMG-CoA synthase promoter-reporter genes requires both SREBP and NF-Y (13,15,16). In addition, electromobility shift assays were used to demonstrate that NF-Y synergistically enhanced the binding of SREBP-1a to SRE-3 in the FPP synthase proximal promoter (16).
In the current studies, mutation of the putative NF-Y binding site in the GPAT promoter produced a reporter gene (pGPAT-Y) that was unresponsive to co-expressed SREBP-1a (Fig. 4B). This result suggested that NF-Y might also play a functional role in the SREBP-dependent induction of GPAT promoter-reporter genes.
In order to test this hypothesis, cells were transiently trans-fected with pGPAT-A together with increasing amounts of pSREBP-1a in the absence or presence of a constant amount of a dominant negative form of NF-YA, pNF-YA29. The latter plasmid encodes a mutant form of NF-YA that is unable to bind DNA but can bind and sequester two separate proteins (NF-YB and NF-YC) that comprise transcriptionally active, heterotrimeric NF-Y (30). Fig. 5 shows that co-expression of NF-YA29 attenuated by over 90% the normal increase in luciferase activity observed when SREBP-1a was co-transfected with pGPAT-A. This inhibitory effect of pNF-YA29 was not the result of a general effect on transcription, since pNF-YA29 had only a slight inhibitory effect on the luciferase activity in cells that had been transiently transfected with pSREBP-1a and a reporter gene under the control of the LDL receptor promoter (Fig. 5). This latter result is consistent with a previous study (15) and with the findings that the LDL receptor gene is transcriptionally regulated by SREBP and Sp1 (17, 18) but not NF-Y. Regulation of GPAT Promoter-reporter Genes by Sterols-Transcription of a number of genes involved in either cellular cholesterol homeostasis (12)(13)(14) or fatty acid synthesis (8,31) are known to be regulated by changes in the sterol content of the cell. The sterol-regulated transcription of these genes results from changes in the nuclear levels of SREBPs. Fig. 6 shows the results obtained when HepG2 cells were transiently transfected with pGPAT-A and then incubated for 48 h in media containing 10% lipoprotein-deficient fetal calf serum and compactin either in the absence or presence of 25-hydroxycholesterol and cholesterol. The results indicate that the luciferase activity was 4.4-fold higher in cells deprived of sterols as compared with the sterol-treated cells (Fig. 6). Under these same conditions, an FPP synthase promoter-reporter gene was regulated 4.6-fold (Fig. 6), consistent with previous studies (24).
Differential Stimulation of Reporter Genes by SREBP-1a and SREBP-2-The functional consequences of activation of SREBP-1 versus SREBP-2 are not known. In cell culture, SREBP-responsive promoter-reporter gene constructs are known to be stimulated by both SREBP-1 and SREBP-2. However, in vivo data indicate that hypolipidemic drugs result in increased levels of mature SREBP-2 but decreased levels of mature SREBP-1 in hepatic nuclei (32). Thus, SREBP-responsive genes may not all respond equally to different forms of this family of transcription factors. Table I shows the results obtained after HepG2, Chinese TABLE I Differential stimulation of reporter genes by co-expressed SREBP-1a or SREBP-2 Duplicate plates of the indicated cell type were transiently transfected with a plasmid encoding ␤-galactosidase and a luciferase reporter gene under the control of the indicated promoter in the absence or presence of 20 ng of either pSREBP-1a (pCMV-CSA10, encoding amino acids 1-490 of SREBP-1a) or pSREBP-2 (pCMV-CS2, encoding amino acids 1-485 of SREBP-2). The cells were incubated for 20 h in medium containing 10% fetal bovine serum and lysed, and enzyme assays were performed as described under "Experimental Procedures." The luciferase activities in relative light units (RLU) are given after normalization for variations in transfection efficiency. The corrected luciferase values obtained from duplicate dishes varied by less than 10%. The numbers in parentheses indicate the -fold increase in luciferase activity relative to cells that were not transfected with plasmids encoding either SREBP-1a or SREBP-2. Similar results were obtained in duplicate experiments. The reporter genes were under the control of promoters derived from FPP synthase (pFPPS-0.247), LDL receptor (pLDL SRE) or HMG-CoA synthase (pSYN SRE) or GPAT (pGPAT-A) genes. hamster ovary, or CV-1 cells were transiently transfected with equivalent amounts of different promoter-luciferase reporter genes together with 20 ng of a plasmid encoding either SREBP-1a or SREBP-2. As expected, luciferase activities increased when SREBP-1a or SREBP-2 were co-expressed with reporter genes under the control of promoters derived from the LDL receptor, FPP synthase, HMG-CoA synthase, or GPAT genes (Table I). The results in Table I also demonstrate that, compared with the other reporter genes, pGPAT-A was 2-13fold more sensitive to SREBP-1a than SREBP-2. Similar results were obtained when the amount of the co-transfected SREBP plasmids was decreased to 5 ng (data not shown). Electromobility Shift Assays-Gel shift and supershift assays demonstrated that purified, recombinant SREBP-1a formed a complex with double-stranded DNA corresponding to the sequences encompassing the 5Ј and central putative SREs (Fig. 1) in the proximal GPAT promoter (Fig. 7A). The addition of antibody to the epitope-tagged SREBP resulted, in each case, in a super-shifted complex (Fig. 7A). Recombinant SREBP-1a also bound to a probe that contained the 3Ј putative SRE but not to a probe in which the TCAG had been mutated to TAAA (Fig. 7B). A preliminary DNase I footprinting analysis with the murine GPAT promoter (Ϫ78 to ϩ17) and recombinant SREBP-1a indicated that this protein bound to sequences that included the TCAG at nucleotides Ϫ63 to Ϫ60 (data not shown).
Incubation of a 32 P-end-labeled probe (Ϫ85 to Ϫ65), which encompassed the inverted CCAAT box (ATTGG) within the GPAT promoter, with a nuclear fraction enriched in NF-Y resulted in the formation of a major specific complex (Fig. 7C). The addition of anti-NF-YA or anti-NF-YB, but not anti-CBF-HSP, to the assay resulted in a supershifted band, consistent with the in vitro interaction of NF-Y with the DNA fragment (Fig. 7C).

DISCUSSION
The current experiments demonstrate that the gene encoding mitochondrial GPAT is regulated during the differentiation of 3T3-L1 preadipocytes to adipocytes by a process that involves ADD1/SREBP. This conclusion is based on the following observations: 1) endogenous ADD1 levels increased during this differentiation process (7), 2) the increase in GPAT mRNA levels that occurred during differentiation was enhanced significantly when ADD1 was ectopically expressed in either preadipocyte or adipocyte cells (Fig. 3), 3) the increase in GPAT mRNA levels that occurred during 3T3-L1 cell differentiation was attenuated when the cells were infected with a retrovirus that expressed a dominant negative form of ADD1 (Fig. 3), and 4) GPAT promoter-reporter genes were regulated by sterols or following the co-expression of SREBP-1a or SREBP-2 (Figs. 4 -6, Table I). Thus, these studies expand the repertoire of lipid metabolic pathways that are responsive to changes in ADD1/ SREBP levels; these pathways include cholesterol biosynthesis (12)(13)(14), low density lipoprotein receptor-mediated endocytosis (12), fatty acid synthesis (3,7,8,31), and, in the current report, glycerolipid synthesis.
Jerkins et al. (23) previously reported that luciferase reporter genes that were controlled by nucleotides Ϫ86 to ϩ102 of the murine mitochondrial GPAT promoter were expressed at 2-fold higher levels in adipocytes than preadipocytes. In addition, these investigators demonstrated that transient expression of luciferase reporter genes in both preadipocytes and adipocytes was decreased 9-fold when the 5Ј-end of the promoter was further deleted from nucleotide Ϫ86 to Ϫ55 (23). The current studies are in agreement with these earlier findings, since we now demonstrate that important binding sites for NF-Y and SREBP/ADD1 are located between nucleotides Ϫ78 and Ϫ55 (Fig. 1).
Previous reports demonstrated that the sterol-regulated changes in the transcription of genes encoding FPP synthase and HMG-CoA synthase were dependent on the binding of SREBP and NF-Y to the proximal promoters of these genes (13,15,16). More recently, we demonstrated that the binding of NF-Y to an inverted CCAAT box in the promoter of the FPP synthase promoter synergistically stimulated the binding of SREBP-1a to a cis-element, termed SRE-3, that lies 20 bp 3Ј of the NF-Y binding site (16). In the current studies, we investigated whether the expression of GPAT promoter-reporter genes was regulated by conditions that are known to regulate the expression of FPP synthase and HMG-CoA synthase promoter-reporter genes in order to determine whether a similar mechanism is involved for all three genes. The results (Figs. 4 -6, Table I) demonstrate that GPAT promoter-reporter genes are regulated by sterols or by co-expression of SREBPs in a manner that parallels that observed with FPP synthase or HMG-CoA synthase promoter-reporter genes (2,12,13,15). In addition, the SREBP-1a-induced increase in GPAT promoterreporter gene activity was attenuated by co-expression of the dominant negative NF-YA29 (Fig. 5). We have previously demonstrated that expression of NF-YA29 attenuates the induction of FPP synthase and HMG-CoA synthase promoter-reporter genes in cells that are sterol-deprived (15). Finally, we showed that, consistent with these in vivo studies, the GPAT proximal promoter contains recognition sites for SREBP-1a and NF-Y (Fig. 7). Taken together, these results demonstrate that luciferase reporter genes under the control of promoters derived from GPAT, FPP synthase, or HMG-CoA synthase genes are all regulated by sterols, SREBP, and NF-Y. The transcription of the squalene synthase gene is also regulated by sterols (14). The finding that the proximal promoter of the squalene synthase gene also contains cis-elements that correspond to an inverted CCAAT box (the binding site for NF-Y), SRE-1 and SRE-3, and Sp1-binding sites (13) may indicate that this represents an additional gene that requires both SREBP/ADD1 and NF-Y or Sp1 for sterol-mediated regulation of transcription.
In contrast, sterol regulation of the LDL receptor, acetyl-CoA carboxylase, and fatty acid synthase genes requires SREBPs/ ADD1 and the ubiquitous transcription factor Sp1 (8,17,31). SREBP-1, SREBP-2, or ADD1 have been shown to bind to SREs in all three promoters. However, the binding of SREBP-1/ ADD1 to an E-box motif (CANNTG) in the fatty acid synthase or S 14 promoters is required for normal regulation (3,6,8). The ability of SREBP-1/ADD1 to bind to both E-box and SRE-1 motifs results, in part, from the presence of a tyrosine residue in a position that is normally occupied by an arginine in other basic-helix-loop-helix family members (6).
The most 3Ј-SRE in the GPAT proximal promoter (Ϫ64 to Ϫ55) is critical for induction of reporter genes by co-expressed SREBP-1a (Fig. 4A) and is bound by recombinant SREBP-1a in gel mobility shift assays (Fig. 7). This SREBP-1a binding sequence (CTCAGCCTAG) has 70% identity with SRE-3 (CTCA-CACGAG) (13) and 60% identity with SRE-1 (ATCACCCCAC) (11) but does not contain the E-box motif (CANNTG). Thus, mature SREBP-1a appears to be able to bind to a number of distinct nucleotide sequences that have now been functionally identified within various promoters. The GPAT proximal promoter contains two additional sequences approximately 100 bp 5Ј of the inverted CCAAT box (Fig. 1) that show 80% identity to the SRE-1 found in the LDL receptor promoter. Deletion of these sequences from promoter-reporter genes did not impair the induction of the reporter gene by co-expressed SREBP-1a (Fig. 4B). Thus, although both of these sequences are bound by recombinant SREBP-1a avidly in electromobility shift assays, they appear to be of little importance in the regulated expression of the promoter-reporter genes used in the current studies.
The current studies demonstrate the importance of ADD1/ SREBP in the regulated expression of GPAT during adipocyte differentiation. However, other as yet unknown factors, in addition to ADD1/SREBP and NF-Y, appear to be required for the increased expression of GPAT mRNA. This latter conclusion is based on the observation that GPAT mRNA levels change less that 2-fold in the livers of rats or mice fed diets supplemented with either cholesterol or cholestyramine and mevinolin (data not shown). The mRNA levels for fatty acid synthase in these livers also change less than 2-fold (data not shown). In contrast, the mRNAs for FPP synthase and HMG-CoA synthase are regulated 10 -35-fold in these same livers (33,34). Sheng et al. (32) demonstrated that when rodents were fed diets supplemented with cholestyramine and mevinolin, the hepatic mRNA levels for SREBP-2 increased, while the mRNA levels for SREBP-1 decreased. These investigators also demonstrated that, under conditions of hepatic cholesterol deprivation, there was an increase in the cleavage and release of mature SREBP-2 but a decrease in cleavage of SREBP-1 (32). This suggests that SREBP-2 and -1 have different regulatory functions and that SREBP-2 may be responsible for the induction of genes, such as HMG-CoA synthase, FPP synthase, and the LDL receptor, that are involved in hepatic cholesterol homeostasis.
The observation that, of four reporter genes studied, GPAT promoter-reporter constructs were more sensitive to SREBP-1a than SREBP-2 (Table I) may be important in understanding the increased expression of genes during adipocyte differentiation. The mRNA for ADD1, the rat homologue of human SREBP-1, increased during the differentiation of preadipocytes to adipocytes (3, 7), a period of increased synthesis of fatty acids and triglycerides (3). Ectopic expression of the dominant negative form of ADD1 prevents both the differentiation of preadipocytes to adipocytes and the enhanced expression of a number of genes (PPAR␥, adipsin, aP2, and C/EBP␣) that normally occurs during this process (7) (Fig. 3). Extrapolating these findings to the in vivo situation suggests that ADD1/SREBP-1 may be more important than SREBP-2 in the stimulation of GPAT expression during adipocyte differentiation. Further studies, perhaps utilizing transgenic mice that overexpress different SREBPs in adipose tissue, will be necessary to test this hypothesis.