Functional analysis of two promoters for the human mitochondrial glycerol phosphate dehydrogenase gene.

Mitochondrial glycerol phosphate dehydrogenase (mGPD) is abundant in the normal pancreatic insulin cell, but its level is lowered 50% by diabetes. To evaluate mGPD expression, we cloned and characterized the 5'-flanking region of the human mGPD gene. The gene has two alternative first exons and two promoters. The downstream promoter (B) is 10 times more active than the upstream promoter (A) in insulin-secreting cells (INS-1) and HeLa cells. Promoter B has higher activity in INS-1 than in non-beta cells. Deletion and mutation analysis suggested that a NRF-2 binding site at -94 to -101 and an E2F binding site at -208 to -215 are important regulatory cis elements in promoter B. Gel mobility shift assays indicated that the -94 to -101 region binds the NRF-2 protein. When INS-1 cells were maintained in the presence of high glucose (25 mm) for 7 days, mGPD was the only 1 of 6 enzyme activities lowered (53%). mGPD promoter B activity was reduced by 60% in INS-1 cells by the high glucose, but in HepG2 cells and HeLa cells, promoter B activity was unchanged or slightly increased. Deletion analysis indicated the glucose responsiveness was distributed across the region from -340 to -260 in promoter B. The results indicate that mGPD gene transcription in the beta cell is regulated differently from other cells and that decreased mGPD promoter B transcription is at least in part the cause of the decreased beta cell mGPD levels in diabetes.

Mitochondrial glycerol phosphate dehydrogenase (mGPD) 1 is encoded in the nucleus and located in the outer surface of the inner mitochondrial membrane. mGPD along with the cytosolic NAD-linked glycerol phosphate dehydrogenase forms the glycerol phosphate shuttle that catalyzes the interconversion of glycerol phosphate and dihydroxyacetone phosphate to oxidize cytosolic NADH by transferring reducing equivalents from the cytosol to mitochondria. Since the enzyme activity of mGPD in the pancreatic beta cell is among the highest in the body (1,2) and its mRNA is distinctly higher in the pancreatic islet than in other tissues (3,4), it has been proposed that the enzyme plays a key role in glucose-induced insulin secretion (5). There are numerous reports that the activity of the enzyme is decreased in the pancreatic islet, but not in other tissues, of genetic models of non-insulin-dependent diabetes (NIDDM) (6 -13). The enzyme is also reported to be decreased in the islet in humans with NIDDM (14). However, this decrease appears to be an acquired characteristic because it is present in experimentally induced models of NIDDM (7,8) and because the enzyme level in the islet is restored to a normal level by normalizing the blood sugar with insulin (12).
In view of the various conditions that possibly influence the level of mGPD synthesis at the transcriptional or translational level, we cloned and characterized the 5Ј-flanking region of the human mGPD gene to begin to better understand the transcriptional regulation of mGPD expression. The results indicate that the human mGPD gene possesses two promoters and that the downstream promoter is more active in beta cells than in non-beta cells. The downstream promoter is 10-fold more active than the upstream promoter in transient expression assays. Deletion analysis, site-directed mutagenesis, and gel shift assays identified at least two known cis-acting regulatory elements in the downstream promoter. In the present work we show that the mGPD enzyme activity and mGPD promoter activity is lowered in INS-1 cells by long term exposure to a high concentration of glucose. Activities of other enzymes studied as controls were not lowered. However, the mGPD promoter activity is lower in beta cells (INS-1 cells) but is not lower in non-beta cells, such as human HepG2 hepatoma cells and HeLa cells, after long term exposure to a high concentration of glucose. The results of the current study suggest that transcription of the mGPD gene in the beta cell differs from other cells and that the negative transcriptional response to high glucose at least in part explains the low level of mGPD in hyperglycemic states.

EXPERIMENTAL PROCEDURES
Cell Culture-INS-1 cells were grown in RPMI 1640 medium (Life Technologies, Inc.) supplemented with 10% fetal calf serum, 50 M 2-mercaptoethanol, 1 mM pyruvic acid, and 10 mM HEPES. HeLa cells were grown in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) supplemented with 10% fetal calf serum. HepG2 cells were grown in minimum essential Eagle's medium (Life Technologies, Inc.) supplemented with 10% fetal calf serum. H4-II-E cells were cultured in minimum essential Eagle's medium supplemented with 0.1 mM nonessential amino acids, 10% calf serum, and 10% fetal calf serum. All media contained 50 units/ml penicillin and 50 g/ml streptomycin.
Determination of 5Ј-Terminal cDNA Sequence-The 5Ј-end of the human mGPD cDNA was extended and amplified using the 5Ј rapid amplification of cDNA ends (RACE) according to the protocol provided (Life Technologies, Inc.). A mGPD gene-specific antisense primer 5Ј-GGTTGGCACGCTCATAGG-3Ј (bases 593 to 575 in exon 3) was used for reverse transcription of HeLa cell poly(A) ϩ RNA. Polymerase chain reaction amplification of the dC-tailed cDNA was carried out using the Abridged Anchor Primer (sense) provided by Life Technologies, Inc. and a nested antisense primer 5Ј-TCCCTGTTAACTCGTTCTGAAAT-3Ј (bases 270 to 249 in exon 2 of the mGPD cDNA). The amplified products were subcloned into the pCR2.1 vector using the TA cloning system (Invitrogen, San Diego, CA) and sequenced.
Human mGPD Gene Promoter Constructs-A 3.4-kb fragment of exon 1A 5Ј-flanking sequence and a 1.3-kb fragment of exon 1B 5Јflanking sequence from genomic clones were cloned into the pGL3 Basic vector (Promega, Madison, WI) (Fig. 1A). Progressive deletions from the 5Ј end of the 3.4-kb or 1.3-kb DNA fragment were prepared by cutting at restriction sites or by digestion with exonuclease III. In the mGPD-400 to -160B, a 3Ј-end deletion construct was made by removing a StuI-SmaI fragment from the mGPD-400 construct that contains a 400-bp fragment of the 5Ј-flanking region of exon 1B (Fig. 1A). The Altered Site in vitro Mutagenesis System (Promega) was used to prepare the 25-bp and 9-bp internal deletion constructs as well as point mutations in the mGPD promoter B region. Each mutant was verified by DNA sequencing and subcloned into the pGL3 Basic vector. Some constructs in Fig. 6C and D were made by Noaman Hasan.
Transient Transfection and Luciferase Assay-INS-1, HeLa, H4-II-E, or HepG2 cells were seeded in 35-mm plates, and 20 h later 70% confluent cells were transfected with 10 g of LipofectAMINE complexed with 1 g of the luciferase reporter plasmid and 0.1 g of the ␤-galactosidase reference plasmid (pBlue-CMV-␤-galactosidase). The cells were exposed to lipid-DNA complexes for 5 h and cultured in growth medium for a further 42 h. Luciferase activity was measured with a luciferase kit (Promega), and ␤-galactosidase activity was measured with a ␤-galactosidase kit (CLONTECH, Palo Alto, CA). ␤-Galactosidase activity was used to correct for differences in transfection efficiency between plates. Luciferase activity presented in each figure is the mean of two or three independent experiments, with each condition done in triplicate. To study the effect of chronically elevated glucose levels on promoter activity, the cells were cultured in medium containing 5 or 25 mM glucose for 4 days before transfection. Glucose concentrations were maintained through the 3-day transfection procedure.
Gel Mobility Shift Assay-Nuclear extracts were prepared from HeLa cells and INS-1 cells for mobility shift DNA-binding protein assays (16,17). A 30-mer wild type oligonucleotide 5Ј-AGCGGGAG-GAGGAAGTCGGGAAGAGGGAAG-3Ј, which corresponds to the sequence from Ϫ110 to Ϫ81 in promoter B, and five oligonucleotides with single point mutations (G-100T, A-98C, G-96T, A-90C, and G-88T), which are indicated by bold letters in the wild type sequence, were synthesized. After end-labeling with [␥-32 P]ATP and T4 kinase, the probe was annealed with an unlabeled complementary strand. Fifteen micrograms of crude nuclear extract was preincubated with poly(dI⅐dC) for 20 min at room temperature in binding buffer (8 mM Tris-HCl, pH 7.6, 8 mM KCl, 1 mM dithiothreitol, 0.8 mM EDTA, and 4% glycerol), and then 300 fmol of 32 P end-labeled double-stranded oligonucleotide probe was added. After incubation for 30 min at 25°C, the reaction mixtures were loaded onto a 4% polyacrylamide gel in electrophoresis buffer (10 mM Tris, 10 mM HEPES, pH 8.0, 1 mM EDTA). The shifted DNA bands were visualized by autoradiography. Competition analyses were performed by mixing 32 P end-labeled double-stranded oligonucleotide probe with 20-or 100-fold molar excess mutated double-stranded oligo before adding to nuclear extracts in binding buffer.
Enzyme Assays-INS-1 cells were cultured in 60-mm dishes in medium containing 5 or 25 mM glucose for 4 or 7 days. Then the cells were washed three times with ice-cold phosphate-buffered saline and homogenized in 400 l of potassium-HEPES buffer (5 mM, pH 7.5) containing 220 mM mannitol, 70 mM sucrose, and 1 mM dithiothreitol. Enzyme activities were assayed under V max conditions by standard methods, as described previously (2,18). The activity of all enzymes except mGPD was estimated in cytosol (20,000 ϫ g ϫ 10 min supernatant fraction) in continuous spectrophotometric assays at 37°C. mGPD activity was estimated in homogenates of whole cells in a timed and stopped assay as described previously (2).

Analysis of the 5Ј-Flanking Region of the mGPD Gene and
Identification of Transcription Start Sites-The first exon (exon 1B) of human HeLa cell cDNA reported by our laboratory (19,20) differs from an expressed sequence tag (EST clone R15742) from human brain that contains 358 nucleotides in front of exon 2 of the mGPD gene. This suggests alternative splicing of the first exon and the possibility of more than one promoter in the mGPD gene. To understand the origin of these heterogeneous mRNAs and to begin to study the regulation of the mGPD gene at the transcriptional level, the 5Ј-flanking region of the human mGPD gene was cloned, and its sequence was analyzed. Two first exons that we have named exon 1A and exon 1B were identified. The 1A region is located 605 nucleotides 5Ј of the 1B region (Fig. 1A). Sequencing of the 5Ј-flanking regions containing the two putative mGPD promoters indicated that promoter A (5Ј flanking region of exon 1A) contains sequences homologous to human repetitive elements. Promoter B (5Ј-flanking region of exon 1B) is GC-rich and contains no TATA box but has several putative binding sites for transcription factors such as Sp1, AP1, AP2, E2F, and NRF-2, the human homologue of mouse GA-binding protein (GABP) (Fig. 1B). We obtained information about the 5Ј end of human mGPD cDNAs by analyzing clones we had isolated from a cDNA library and from the published sequence of EST clones R15742, T23788, and Z38279. Attempts at determining the transcriptional start sites of the mGPD gene by the primer extension method and RNase protection assay were not successful because of the low abundance of mGPD mRNAs. Therefore, we used 5Ј RACE to determine the transcriptional start sites of the two transcripts. HeLa cell poly(A) ϩ RNA was used as the template. Eighteen clones of various lengths were characterized. Seven of them contained sequences from exon 1A, and 11 of them contained sequences from exon 1B. Three transcriptional start sites were detected for exon 1A and four for 1B (Fig. 1B). The longest clones containing the sequence from exon 1A correspond to the last 159 nucleotides of 358 bp of exon 1A of the EST clone R15742 from human brain. The difference in length between the transcripts from HeLa cells and the EST, which was from brain, may be due to a failure of reverse transcriptase to read through regions of secondary structure in the RNA due to its high G/C content, or this difference in length could be due to differences in transcription start sites between the two tissues. The 5Ј end from four other clones containing the sequence of 1B corresponds to position of ϩ1 of the previously published cDNA sequence by our laboratory (20), and we have designated this position as ϩ1 (Fig. 1B). The 5Ј end of the longest clonecontaining sequence from exon 1B corresponds to the position Ϫ48. In addition, we identified three clones with 5Ј ends at position of ϩ33 of exon 1B and one clone with a 5Ј end at position of Ϫ29 of exon 1B.
Mapping of the mGPD Promoter A Region-To map promoter activity of the 5Ј-flanking region of exon 1A, various 5Ј progressive deletions starting from Ϫ3400 were prepared, and their activities were measured. Fig. 2A shows the relative luciferase activities of these constructs in INS-1 and HeLa cells. The activities of the constructs in the forward orientation were 7-15

FIG. 2. Mapping of the human mGPD gene promoter A and B. A, the relative luciferase activities of the 5Ј deletion constructs of the promoter A region in INS-1 (upper panel) and HeLa cells (lower panel).
All constructs contain variable lengths of the 5Ј-flanking region of exon 1A plus 190 bp of exon 1A. The name of each construct indicates the number of nucleotides of 5Ј-flanking sequence of exon 1A included in the construct. R3400 represents the entire 3400-bp segment from the promoter A region inserted into the pGL3 Basic vector in the reverse orientation. Promoter activity is presented as a percentage of the mGPD-460A, which was set at 100%. B, the relative luciferase activities of 5Ј deletion constructs of promoter B in INS-1 cells (upper panel) or HeLa cells (lower panel). The names along the abscissa indicate the number of nucleotides of the 5Ј-flanking sequence of exon 1B present in the constructs. The sequence from ϩ6 to Ϫ140 was deleted in the 3Ј deletion construct Ϫ400 to Ϫ140B. R400B represents the mGPD promoter region inserted in the reverse orientation. Promoter activity is presented as a percentage of the activity of the mGPD-600B, which was set at 100%.

FIG. 3. Comparison of mGPD activities of promoters A and B.
Activities of the two most active constructs of promoter A, which contain 3400-bp (3400A) or 460-bp (460A) nucleotides, are compared with the activities of constructs of promoter B, which have the highest activity (600B and 400B) and contain 600-bp and 400-bp nucleotides, respectively. times higher than the activity of the construct possessing the reverse orientation (R3400). The construct containing 460 nucleotides (460A) had the highest activity ( Fig. 2A).
Mapping of the mGPD Gene Promoter B-To map promoter B of the mGPD gene, progressive deletions from the 5Ј end as well as a 3Ј deletion were prepared, and their activities were measured. Fig. 2B shows the relative luciferase activities of these constructs after transient transfection into INS-1 cells or HeLa cells. All constructs contain 6 bp of the 5Ј end of exon 1B and the promoter B region that lies immediately upstream of exon 1 B. The activity produced by the 600B construct, which contains the entire region of the promoter B, was comparable with that of the SV40/enhancer control vector when expressed in INS-1 cells. The mGPD promoter B inserted in the reverse orientation in the pGL3 vector was used as a control. Its activity was 9.6% and 3.2% that of the mGPD-600B construct in INS-1 cells and HeLa cells, respectively. The promoterless pGL3-Basic vector was also used as a control, and its activity was less than 5% of mGPD-600B in both INS-1 and HeLa cells.
Progressive 5Ј deletions from Ϫ600 to Ϫ400 (400B) still retained 85 and 80% activity in INS-1 and HeLa cells, respectively. Further deletions resulted in a sharp stepwise decrease in activity. Deleting bases ϩ6 to Ϫ140 and retaining bases Ϫ141 to Ϫ400 (Ϫ400 to Ϫ140B) eliminated about 90% of promoter activity, indicating that the proximal 140 bases contains strong positive elements.
Comparison of mGPD Promoter A and B Activities-The activities of two promoter A constructs (3400A and 460A) that have the highest activity were compared with the activities of two promoter B constructs that have the highest activity (600B and 400B). The promoter A constructs were 10% as active as the promoter B constructs in both INS-1 cells and HeLa cells (Fig. 3).
mGPD-400B Promoter Activity in INS-1 and Non-beta Cells-Since the activity of promoter B is much higher than that of promoter A, we further characterized promoter B. First we compared promoter B activity in non-beta cells including human HeLa cervical adenocarcinoma, human HepG2 hepatoblastoma, and rat H4-II-E hepatoma cells to that in INS-1 cells. The last cell line was derived from a rat pancreatic insulinoma (15) and expresses a high level of mGPD enzyme activity comparable with that of normal pancreatic islets. 2 We compared the activity of pGL3-mGPD-400B with the activity of the pGL3 2 M. J. MacDonald, unpublished data. control vector, which contains the luciferase gene under control of the SV40 promoter in INS-1, HeLa, HepG2, and H4-II-E cells. Fig. 4A shows that the luciferase activity driven by mGPD-400B was 1.8-fold higher than that driven by the SV40 promoter in INS-1 cells, whereas the activities driven by mGPD-400B were 3, 5, and 43% that driven by SV40 in HeLa, HepG2, and H4-II-E cells, respectively. If the absolute light units of luciferase activity driven by mGPD-400B was normalized for the protein concentration of the cell lysates, the activity of the 400B construct expressed in INS-1 cells was 3-and 29-fold higher than that in HeLa and H4-II-E cells, respectively, and was slightly higher than in HepG2 cells (Fig. 4B).
Internal Deletion Analysis of Promoter B-Since the construct containing ϩ1 to Ϫ400 nucleotides possesses 80% activity of the entire 600 bp of promoter B, a series of 25-nucleotide internal deletions from this region were prepared to attempt to identify important regulatory regions. Fig. 5 shows that removal of nucleotides Ϫ76 to Ϫ100 eliminated 50 and 70% of the promoter activity in INS-1 and HeLa cells, respectively, and that removal of sequence Ϫ201 to Ϫ225 eliminated 47 and 46% of the promoter activity in INS-1 and HeLa cells, respectively, indicating that these regions contain strong positive regulatory elements. The region from Ϫ252 to Ϫ276 also might contain important element(s) because removal of the region resulted in a 35 and 55% decrease in activity in INS-1 and HeLa cells, respectively. Examination of the sequence in this region revealed the presence of a Sp1 binding site between Ϫ280 to Ϫ275, suggesting this Sp1 may also contribute to promoter B activity, especially in HeLa cells.
To more explicitly identify potential regulatory elements in the region between Ϫ76 to Ϫ100 in promoter B, a series of 9-nucleotide internal deletions were prepared. The results suggest that the interval Ϫ94 to Ϫ102 contains an important regulatory element for the mGPD promoter B because removal of this sequence results in a 50% decrease in activity in INS-1 cells (Fig. 6A) and a 60% decrease in activity in HeLa cells (Fig. 6B). A similar study was also performed with the region Ϫ201 to Ϫ225. The deletion from Ϫ204 to Ϫ221 led to a decrease in relative activity of about 50% in both INS-1 cells (Fig. 6C) and HeLa cells (Fig. 6D). These data indicated that both the Ϫ94 to Ϫ102 and Ϫ204 to Ϫ221 regions contain regulatory sequences that significantly contribute to the activity of mGPD promoter B.
Site-directed Mutagenesis of mGPD Promoter B Elements-Examination of the DNA sequence of promoter B revealed the presence of a putative E2F (TTCCGCGT) binding site and a putative Sp1 (CCGCCC) binding site between Ϫ204 to Ϫ221 and a putative NRF-2 binding site (AGGAAG) between Ϫ94 and Ϫ102. To assess the contribution of these sites to promoter activity, the binding sites were altered by site-directed mutagenesis, and the activities of the mutant constructs were estimated. Fig. 7, A and B, shows that when the Sp1 site, CCGCCC, located at Ϫ220 to Ϫ225, was replaced by CAAACC, the promoter construct still retained 90 and 85% activity in INS-1 and HeLa cells, respectively. However, when the E2F site, TTCCGCGT, located at Ϫ208 to Ϫ215 was mutated to TTCAAAGT, the promoter activity was decreased by 48 and 43% in INS-1 and HeLa cells, respectively. To study the role of the NRF-2 binding site in the Ϫ94 to Ϫ102 region, two point mutations (G-96T and A-98C) were created, and their activities were measured (Fig. 7, C and D). Adjacent to this NRF-2 site there is an NRF-2 like sequence (GGGAAG) at Ϫ93 to Ϫ88 that differs from the consensus sequence only in the first nucleotide (G instead of A). Two point mutations (G-88T and A-90C) were created in this region to serve as a control for the Ϫ94 to Ϫ102 region. The G-96T mutation resulted in 47 and 60% decreases in activity of INS-1 cells and HeLa cells, respectively, whereas the A-98C mutation resulted in 32 and 42% decreases in activity in INS-1 cells and HeLa cells, respectively, as compared with the wild type (400B) construct. The mutations G-88T and A-90C did not dramatically alter the promoter activity. In fact the G-88T mutation increased the activity to twice that of the wild type construct. These results suggest that the E2F consensus sequences in the Ϫ204 to Ϫ221 region and NRF-2 in Ϫ94 to Ϫ102 region significantly contribute to the activity of promoter B, but the Sp1 and NRF-2-like sequences in these regions do not.
NRF-2 Protein Binds at the Ϫ102 to Ϫ94 Region in Promoter B-To determine if the sequences at Ϫ102 to Ϫ94 and Ϫ93 to Ϫ88 are capable of binding proteins, we performed gel shift assays (Fig. 8). The wild type and mutated DNA sequences covering region Ϫ81 to Ϫ110 were used as probes and competitors in mobility gel shift studies as described under "Experimental Procedures." As shown in Fig. 8, the wild type oligonucleotide caused retardation of two bands in INS-1 cells and three bands in HeLa cells, which were competed away by an excess of unlabeled oligonucleotide. The oligonucleotides that carried point mutations within the NRF-2 motif (G96T, A98C, and G100T) did not compete out the retarded bands in INS-1 or HeLa cell nuclear extracts. In contrast, the retarded bands were competed away by the oligonucleotides containing mutations G88T and A90C, which were at the NRF-2-like site. Taken together with the data from functional analysis of these point mutations in luciferase assays, these results suggest that the NRF-2 motif at Ϫ94 to Ϫ102 in mGPD promoter B plays an important role in the expression of the mGPD gene.
Enzyme in the pancreatic islets of several genetic models of NIDDM has been reported (6 -13). However, in our laboratory the mGPD level in pancreatic islets cultured for 24 h at various of concentration of glucose did not change (21). To examine whether a longer term exposure of insulin cells to a high concentration of glucose causes a decrease in mGPD activity, we cultured the INS-1 cells in medium at 5 or 25 mM glucose for 4 or 7 days. Table I shows that mGPD activity in the cells exposed to 25 mM glucose for 4 days and for 7 days was 81 and 47%, respectively, that found in the cells cultured in 5 mM glucose for the same lengths of time. To discern whether this decrease in enzyme activity was a general phenomenon, activities of other enzymes were measured (Table I). The activities of isocitrate dehydrogenase, glucose-6-phosphate dehydrogenase, and glucokinase in INS-1 cells maintained in 25 mM glucose were similar to those in the cells maintained in 5 mM glucose, and the activity of cytosolic GPD in the cells cultured in 25 mM glucose was actually 3-and 5-fold higher than that in the cells cultured in 5 mM glucose for 4 and 7 days, respectively. The activity of malic enzyme was increased 1.5-and 2-fold after the cells were exposed to 25 mM glucose for 4 and 7 days, respectively.
Effect of Long Term Exposure of Cells to Elevated Levels of Glucose on mGPD Promoter Activity-The decrease in enzyme activity of mGPD in INS-1 cells after long term exposure to a high concentration of glucose may involve down-regulation of transcription of mGPD. To test this idea, we transiently transfected mGPD promoter constructs into INS-1 cells maintained in 5 or 25 mM glucose for 5 days and continued the exposure of the transfected cells to 5 or 25 mM glucose for another 2 days. Fig. 9A shows that culturing cells in 25 mM glucose decreased mGPD promoter B activity to about 40% that found in the cells cultured in 5 mM glucose, whereas the activities of mGPD promoter A and the SV40 promoter were not changed. This suggests that the glucose-induced down-regulation of mGPD activity may be a consequence of its influence on promoter B. The degree of decrease in luciferase activity showed a good correlation with the decrease in the enzyme activity in INS-1 cells treated with the high concentration of glucose. To investigate whether the effect of elevated glucose on the activity of mGPD promoter B was insulin cell-specific, non-beta cells, HepG2, and HeLa were transfected and cultured as described above. Fig. 9B shows that no decrease in the activity of mGPD promoter A or B was detected with either HeLa cells or HepG2 cells. In fact, the high concentration of glucose slightly increased the promoter activity of the 600B construct in both cell lines.
To define the region of promoter B involved in the glucose down-regulation of mGPD expression in INS-1 cells, the activities of a series of constructs carrying 5Ј end deletions and one 3Ј end deletion of promoter B were tested in INS-1 cells treated with 5 or 25 mM glucose (Fig. 10A). Deletions from the 5Ј end to position Ϫ340 and from the 3Ј end (position ϩ6) to Ϫ140 had no significant effect upon the down-regulation of promoter B activity by glucose. Removal of the 5Ј end to Ϫ260 completely abolished the down-regulation by high glucose. However, the activity of this construct in the presence of 5 mM glucose was already reduced to 30% that of the construct 400B. These results suggest that the sequence between Ϫ340 and Ϫ260 contains elements that may play a role in the down-regulation by high glucose of mGPD promoter B activity in beta cells.
To further map promoter B for glucose responsiveness, the promoter activity of the constructs with a series of 25-bp internal deletions from Ϫ376 to Ϫ201 were tested in INS-1 cells. Fig.  10B shows that deletions of single 25-bp fragments in this region did not abolish the response of the mGPD promoter B to a high concentration of glucose. It is possible that the regulation of mGPD expression in INS-1 cells by glucose is determined by multiple cis elements. Deletion of one or a few of them may not be sufficient to abolish the influence of glucose on promoter B activity. Since the NRF-2 element in the region Ϫ94 to Ϫ101 is a potential candidate for glucose responsiveness, we also assayed the construct with the deletion from Ϫ94 to Ϫ102, where the NRF-2 motif is located. Fig. 10B shows that deletion of the NRF-2 binding site did not abolish the responsiveness of the promoter B to high glucose. The construct with point mutation at NRF-2 site also did not abolish the response (data not shown). These results suggest that NRF-2 may not be involved in down-regulation of mGPD promoter B activity by glucose, although it is an important element for the basal activity of promoter B.

DISCUSSION
We cloned and sequenced the 5Ј-flanking region of the human mGPD gene. When this sequence and the sequence of the HeLa cDNA clone isolated by our laboratory (19) were compared with GenBank TM , a match was found to EST clones R15742, T23788, and Z38279. A region of 358 bp of EST clone R15742 corresponded to a region in the mGPD gene located 605 bp upstream of the first exon of the HeLa cDNA clone (Fig. 1A). In the EST clone, the 358-bp region was spliced directly to exon 2 of the gene (20), indicating that this upstream region was an alternative first exon. The evidence for two alternative exons has been strengthened by a report from another group (4) of a human pancreatic islet mGPD cDNA containing this alternative first exon (1A) and more recently by a report of the sequence of another EST clone (AA169176) possessing the sequence identical to the portion of the region we designated exon 1B. We have designated the 5Ј-flanking region of exon 1A as promoter A and the 5Ј flanking region of exon 1B as promoter B. No consensus TATA box or CAAT box were present in the first 100 bp upstream from the transcription initiation sites in either promoter A or promoter B. Promoters lacking a TATA box but containing GC boxes are frequently found in housekeeping genes, which usually have several transcription start sites (22). However, studies have demonstrated that some highly regulated genes, including immediate early genes, developmentally regulated genes, and tissue-specific genes also have promoters that lack TATA boxes (23). Sequence analysis of the immediate upstream of exon 1A of mGPD gene shows the presence of an E-box, but this region lacks the GC-box, AP1, AP2, E2F, and NRF-2 transcription factor binding sites that are present in promoter B (Fig. 1B). This together with the presence of human repeat elements in the more upstream region of promoter A may explain the lower activity of promoter A compared with promoter B. There is some evidence to suggest that Alu repeat elements function as either positive or negative regulatory elements to influence the expression of nearby genes (24 -26).
Although the human mGPD gene uses two different promoters, the rat mGPD gene probably utilizes three promoters since the 5Ј end of the rat mGPD gene contains three alternative first exons. These exons are transcribed from different promoters, and each of these is spliced to the common second exon (27). The sequence of rat exons 1a and 1b show homology to human exons 1A and 1B, respectively, whereas rat exon 1c does not show homology to the human transcripts (27,28). 3 The presence of this third promoter probably explains why the rat is the only animal known to express a high level of mGPD in testis (29,30). In the rat, promoter A is utilized primarily in brain, promoter B is used ubiquitously, and promoter C is used only in the testis (27). Multiple and tissue-specific promoters have been reported for several other genes including pyruvate carboxylase, acyl-CoA synthetase, the glucocorticoid receptor, and macrophage colony-stimulating factor receptor (31)(32)(33)(34)(35). The enzyme activity of mGPD shows marked tissue differences. Highest activity is found in the pancreatic islet (1-3, 36 -40), testis (rat only (30)), and brown adipose tissue (3); intermediate activity is found in brain and skeletal muscle (29); and low activity is found 3 L. J. Brown, unpublished data.  in other tissues (29). The existence of the alternate promoters in the mGPD gene is of potential interest because it could provide an explanation for the tissue-specific regulation of mGPD.
That previous studies (1-3, 36 -40) showed mGPD enzyme activity to be extremely high in rodent and human insulin cells suggests the enzyme is important for insulin secretion. In the present work, we compared promoter B activity in the insuli-noma cell line, INS-1, to that in non-beta cell lines, including human HeLa cervical adenocarcinoma, human HepG2 hepatoblastoma, and rat H4-II-E hepatoma cells, using the SV40 promoter as a control and found that the activity of promoter B is 1.8-fold higher than that of the SV40 promoter expressed in INS-1 cells, whereas in other cells the activity of the mGPD promoter B is 3 to 43% that of SV40 (Fig. 4A). Ferrer et al. (4) showed the presence of exons 1A and 1B in RNA from human pancreatic islets, insulinomas, and other tissues by the reverse transcription-polymerase chain reaction. Since the reverse transcription-polymerase chain reaction was not performed quantitatively, the proportions of the two transcripts within a single tissue and the relative level of these two transcripts in different tissues are not certain. However, the evidence of the usage of exon 1B in human pancreatic islets and insulinomas and the evidence of the high promoter activity in INS-1 cells may explain the differences in the enzyme activity of mGPD between the beta cells and non-beta cells.
To further characterize the cis elements involved in the transcriptional regulation of the mGPD gene, we performed deletion and mutation analysis. We determined that E2F binding site regions from Ϫ204 to Ϫ221 is important for promoter B activity in both INS-1 and HeLa cells. An alignment analysis of the human mGPD promoter with the rat promoter shows the FIG. 9. Effect on mGPD promoter activity of long term exposure of cells to elevated glucose. Promoter constructs were transfected into cell lines that were treated with 5 or 25 mM glucose for 7 days as described under "Experimental Procedures." Luciferase activities were measured and presented as a percentage of the activity of mGPD-600B expressed in the cells treated with 5 mM glucose. A, two promoter B constructs (600B and 400B), two promoter A constructs (3400A and 460A), and a pGL3-control (SV40) construct were expressed in INS-1 cells. The inset shows activity of promoter A constructs. B, Promoter A and promoter B constructs (same as panel A) were expressed in HeLa and HepG2 cells. E2F site to be conserved. A NRF-2 sequence in the region between Ϫ96 to Ϫ101 also appears to be essential for promoter B activity. When a single nucleotide was mutated in this consensus NRF-2 sequence, the promoter activity was reduced 50% (Fig. 7, C and D). Gel mobility shift assays (Fig. 8) suggested that the NRF-2 binding site at the region between Ϫ94 and Ϫ101 of promoter B is capable of binding proteins. The two retarded bands detected with the INS-1 nuclear extract and three with the HeLa nuclear extract may arise through heteromultimers of NRF-2 subunits. A similar gel shift pattern has been reported by Virbasius et al. (41) when studying the promoter of cytochrome c oxidase subunit IV with HeLa cell nuclear extract. Human NRF-2 is composed of five subunits, ␣, ␤1, ␤2, ␥1, and ␥2. Only the ␣ subunit has intrinsic DNA binding ability. The other subunits associate with the ␣ subunit and participate in the formation of heteromeric complexes with distinct binding properties and regulate NRF-2-mediated transcription by competing with each other (41,42). NRF-2 belongs to the Ets family of transcription factors (41,43) that bind to the consensus sequence (C/A)GGA(A/T)(A/G). NRF-2 is the human homologue of mouse GABP, and recent reports suggest that NRF-2 (GABP) regulates the expression of nuclear-encoded mitochondrial proteins, including cytochrome c oxidase subunits IV and Vb as well as the expression of mitochondrial transcription factor 1 (41,(43)(44)(45)(46)(47)(48). There is evidence that GABP binding activity is regulated by the redox state of the cell and provides redox regulation of its target genes (49,50). Thus GABP acts as a regulatory link between mitochondrial function and nuclear gene expression. The results of the current study as well as the fact that the NRF-2 site is conserved in promoter B of the rat mGPD gene (27) suggest that NRF-2 is a major determinant of mGPD gene transcription. The idea that NRF-2 activity can be regulated by the redox state of the cell is particularly interesting in respect to regulation of mGPD synthesis because mGPD is the key enzyme of the glycerol phosphate shuttle, which is one of the two shuttles that control the cytosolic NAD/NADH ratio.
Pancreatic islet mGPD levels are low in genetic models of NIDDM, such as the db/db mouse (9), GK rat (6), and Zucker diabetic fatty (ZDF) rat (13), and in models of NIDDM in which diabetes is experimentally induced by the neonatal administration of streptozotocin (7,8). The present study demonstrates a reduced level of mGPD enzyme activity in INS-1 cells after long term exposure to 25 mM glucose. This reduction seems to be specific to mGPD, since the activity of several enzymes relevant to glucose metabolism or the cellular redox state, including cytosolic GPD, malic enzyme, isocitrate dehydrogenase, glucose-6-phosphate dehydrogenase, and glucokinase, remain unchanged or are even increased after the long term exposure to high glucose (Table I). That the decrease of mGPD enzyme activity in animals with NIDDM is caused by a decreased net synthesis of the enzyme rather than by the decreased activity of a normal amount of enzyme (12,13) suggests the down-regulation of expression occurs at the transcriptional or/and the translational level. The region of promoter B that confers sensitivity to glucose is between Ϫ340 and Ϫ260. Apparently regulation of promoter activity is contributed by multiple cis elements because single 25-bp deletions extending across this region did not localize the glucose effect. Similar to the in vivo studies (6 -11), the down-regulation of promoter B by a high concentration of glucose occurs in the insulin-secreting cells (INS-1 cells) but not in non-beta cells (Fig. 9B). These observations suggest that expression of mGPD is regulated differently in different tissues. The varied expression of mGPD among various tissues may be a reflection of different transacting factors from tissue to tissue. In addition to mGPD, levels of glucose transporter 2 and pyruvate carboxylase are also decreased in the pancreatic islet of rodent models of NIDDM (12,(51)(52)(53)(54)(55)(56). The coordinate decrease in a number of these key proteins of glucose metabolism could be the consequence of an adaptive decrease in the activities of transcription factors that influence genes that encode these proteins (57). This would have the effect of moderating the adverse effects of hyperglycemia on beta cell metabolism (13).