Involvement of two Sp1 elements in basal endothelial prostaglandin H synthase-1 promoter activity.

Prostaglandin H synthase-1 (PGHS-1) is a constitutively expressed key enzyme in the biosynthesis of physiologically important prostanoids. The promoter of the human PGHS-1 gene lacks a TATA box, has a very GC-rich region, and contains multiple transcription start sites. To identify the elements involved in the constitutive expression of the PGHS-1 gene, we constructed a 2075-base pair fragment (−2095 to −21 relative to the translation start codon) and a series of 5′-deletion mutants into a promoterless luciferase expression vector, which was transfected in HUVEC. Two important regions were identified. DNase I footprinting identified a protected segment, which contains an Sp1 binding site proximal to the transcription start sites. Band shift assays confirmed specific binding of Sp1 to this segment. Band shift assays further revealed specific binding of Sp1 to a distal region containing a canonical Sp1 site. Mutation of either Sp1 binding site significantly reduced the promoter activity. When both sites were mutated, the activity was reduced to 29% of that of the wild type. Mutation of Sp1 sites did not abrogate promoter activity stimulated by phorbol ester. These results indicate that binding of Sp1 or its related proteins to two widely separated Sp1 sites on the promoter region activates the basal PGHS-1 gene transcription.

Prostaglandin H synthase-1 (PGHS-1) is a constitutively expressed key enzyme in the biosynthesis of physiologically important prostanoids. The promoter of the human PGHS-1 gene lacks a TATA box, has a very GCrich region, and contains multiple transcription start sites. To identify the elements involved in the constitutive expression of the PGHS-1 gene, we constructed a 2075-base pair fragment (؊2095 to ؊21 relative to the translation start codon) and a series of 5-deletion mutants into a promoterless luciferase expression vector, which was transfected in HUVEC. Two important regions were identified. DNase I footprinting identified a protected segment, which contains an Sp1 binding site proximal to the transcription start sites. Band shift assays confirmed specific binding of Sp1 to this segment. Band shift assays further revealed specific binding of Sp1 to a distal region containing a canonical Sp1 site. Mutation of either Sp1 binding site significantly reduced the promoter activity. When both sites were mutated, the activity was reduced to 29% of that of the wild type. Mutation of Sp1 sites did not abrogate promoter activity stimulated by phorbol ester. These results indicate that binding of Sp1 or its related proteins to two widely separated Sp1 sites on the promoter region activates the basal PGHS-1 gene transcription.
Prostaglandin H synthase (PGHS, 1 EC 1.14.99.1) is a bifunctional enzyme containing a cyclooxygenase activity that catalyzes the bisoxygenation of arachidonic acid to form the peroxide prostaglandin G 2 (PGG 2 ) and a peroxidase activity that catalyzes the reduction of PGG 2 to PGH 2 (for a review, see Ref. 1). PGH 2 is the common precursor of biologically active prostaglandins, thromboxane and prostacyclin. PGHS, hence, occupies a pivotal position in prostanoid biosynthesis. There are two isoforms of PGHS. The type 1 PGHS (PGHS-1) is constitutively expressed in most mammalian cells, whereas the expression of PGHS-2 is induced by cytokines and mitogenic factors (2). Both isoforms of PGHS are important in catalyzing prostacyclin synthesis in endothelial cells. PGHS-1 catalyzes the synthesis of basal levels of prostacyclin, while prostacyclin productions by inflammatory agents correlate with the induced expression of the PGHS-2 enzyme (3,4). We have been interested in understanding the transcriptional regulation of the constitutive PGHS-1 gene in endothelial cells. The human PGHS-1 gene spans about 22 kb on chromosome 9 containing 11 exons (5,6). The coding region encodes a 599-amino acid protein including a 23-amino acid signal peptide (5). The 5Јflanking region of the human PGHS-1 gene has multiple transcription start sites (TSS), does not possess the canonical TATA or CAAT box, and is GC-rich. These features are consistent with those of a housekeeping gene. The 5Ј-flanking region of PGHS-1 bears several putative binding sites for transcriptional activators (6). However, the promoter activity of the 5Ј-flanking region has not been delineated. Furthermore, the mechanism by which the basal promoter function is activated has not been elucidated. By fusing the 5Ј-flanking region to luciferase cDNA, we have recently demonstrated promoter activity in the 5Јflanking region of the PGHS-1 gene, but the promoter activity was very weak (6). We postulate that other cis-acting elements are important in the basal promoter activation of the PGHS-1 gene. We report here that binding of Sp1 and/or its closely related proteins to two Sp1 binding sites are critical for activating the basal transcription of the PGHS-1 gene.

EXPERIMENTAL PROCEDURES
Materials-Reagents for cell cultures were obtained from Sigma. Lipofectin, Optimen I, and kits for the ␤-galactosidase assay were obtained from Life Technologies, Inc. Restriction enzymes and Klenow polymerase were obtained from New England Biolabs. pSV-␤-gal plasmids, DNase I, Taq polymerase, purified Sp1, and kits for the luciferase assay and DNase I protection assay were obtained from Promega. Radiolabeled nucleotides were obtained from Amersham Corp. Consensus oligonucleotides containing binding sites for Sp1 (ATTC-GATCGGGGCGGGGCGAGC), AP-2 (GATCGAACTGACCGCCCGCG-GCCCGT), and NF-B (AGTTGAGGGGACTTTCCCAGGC) were obtained from Promega.
Isolation and Sequencing of PGHS-1 5Ј-Flanking Region-A bacteriophage EMBL-3 human genomic library constructed from placental DNA (Clontech) was screened with a 32 P-labeled 0.7-kb fragment at the 5Ј-flanking region of human PGHS-1 genomic DNA, the sequence of which had been previously reported (6). One positive clone containing a 16-kb insert was isolated from 3 ϫ 10 6 plaques. The positive clone was plaque-purified and mapped. This clone included the first eight exons and a 2.5-kb 5Ј-flanking region. A 3.5-kb XhoI/EcoRI fragment of this clone containing the 5Ј-flanking region, the first two exons, and part of the second intron was subcloned into pGEM7-Zf (Promega) for sequencing. Nucleotide sequences were determined by the chain termination method using specific primers of PGHS-1 genomic DNA.
Construction of 5Ј-Deletion and Site-directed Mutants-5Ј-Deletion constructs of the PGHS-1 promoter were generated by polymerase chain reaction (PCR), using the 3.5-kb XhoI/EcoRI fragment (described above) as the template. All of the upstream PCR amplification primers contained either a BamHI or an XhoI restriction recognition site, and the downstream PCR primers were linked to a HindIII site. Primers were synthesized according to the sequence relative to the ATG codon (A as ϩ1): the upstream primers, Ϫ2095 to Ϫ2071, Ϫ1261 to Ϫ1237, Ϫ916 to Ϫ899, Ϫ744 to Ϫ728, Ϫ565 to Ϫ547, Ϫ257 to Ϫ240, and Ϫ137 to Ϫ117; the downstream primers, Ϫ41 to Ϫ21; Ϫ143 to Ϫ126; and Ϫ761 to Ϫ744. The PCR products were purified from agarose gel, digested, and cloned into a promoterless luciferase expression vector, pXP1 (7). The constructs were designated according to their positions relative to the ATG codon as shown in Fig. 2. To generate the constructs of Ϫ916/Ϫ21, where the two Sp1 sites were changed, a PCR-mediated site-directed mutagenesis was employed. The primers used for mutations were as follows (from 5Ј to 3Ј, with mutated bases underlined): primer A, GGGCTGGCTCTGAAACCTGAAGCCA; primer AЈ, TGGCT-TCAGGTTTCAGAGCCAGCCC; primer B, GGAGGAGCGGTTTTAG-AGCCCGGGGG; primer BЈ, CCCCCGGGCTCTAAAACCGCTCCTCC. The mutant Ϫ610/Ϫ604 (see Fig. 5, construct b) was obtained by first amplifying the template Ϫ916/Ϫ21 with the primers A and Ϫ41/Ϫ21 to generate a 0.6-kb fragment. In a separate tube, a 0.3-kb fragment was generated by 30 cycles of PCR using primers AЈ and Ϫ916/Ϫ899. Both 0.6-and 0.3-kb fragments were gel-purified and combined for one cycle of PCR. Subsequently, primers Ϫ916/Ϫ899 and Ϫ41/Ϫ21 were added for another 30 cycles of PCR. The amplified fragment was digested and subcloned into pXP1 vector. Each PCR cycle was 94°C for 1 min, 52°C for 1 min, and 72°C for 1 min. The mutant Ϫ111/Ϫ105 (see Fig. 5, construct c) was prepared by identical methods using primers B and BЈ. The double mutant (construct d) was obtained using the mutant Ϫ610/ Ϫ604 as the template and primers B and BЈ as mutagenized primers. The mutants were confirmed by nucleotide sequencing.
Cell Culture, Transient Transfection, and Luciferase Assays-Human umbilical vein endothelial cells (HUVECs) were cultured as described previously (8). To ensure consistent results, passage 1 cultured cells were used throughout the studies unless otherwise indicated. One day before transfection, cells were seeded at 30 -40% confluence in a six-well dish. Liposome-mediated transient transfection was performed as described (9). Briefly, HUVECs were transfected with a Lipofectin/ DNA mixture containing 12 g of Lipofectin (Life Technologies, Inc.) and 2 g of the promoter constructs with or without 0.3 g of pSV-␤-gal (Promega) in 1.2 ml of Optimen I for 4 h. Lipofectin and DNA plasmids were subsequently removed and replaced with complete medium. Cells were harvested and lysed with 200 l of reporter lysis buffer (Promega). Cell extracts were centrifuged in a microcentrifuge for 5 s to remove debris. 50 l of the supernatant was removed for luciferase assay in a luminometer (Monolight, model 2010) according to the manufacturer's procedures. The protein content was determined by the BCA protein assay kit (Pierce) using bovine serum albumin as a standard. ␤-Galactosidase activity was assayed by chemiluminescence (Clontech) as described (10).
Preparation of Nuclear Extracts-HUVEC nuclear extracts were prepared as described previously (11) with the following modifications. HUVECs were harvested by scraping, washed in cold phosphate-buffered saline, and incubated in two packed cell volumes of buffer A (10 mM HEPES, pH 7.9, 1.5 mM MgCl 2 , 300 mM sucrose, 0.5% Nonidet P-40, 10 mM KCl, 0.5 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, 1 g/ml each leupeptin and aprotinin) for 10 min on ice. The crude nuclei released by lysis were collected by microcentrifugation (9500 rpm, 20 s), rinsed once in buffer A, and resuspended in 2 ⁄3 packed cell volume of buffer B (20 mM HEPES, pH 7.9, 1.5 mM MgCl 2 , 420 mM NaCl, 0.2 mM EDTA, 25% glycerol, 0.5 mM phenylmethylsulfonyl fluoride, 1.0 mM dithiothreitol, and 1.0 g/ml each leupeptin and aprotinin). Nuclei were disrupted by passing through a 23-gauge syringe 10 times. The homogenate was gently stirred on ice for 30 min, and the debris was removed by microcentrifugation for 2 min. The resulting supernatant was diluted 1:1 with buffer C (20 mM HEPES, pH 7.9, 100 mM KCl, 0.2 mM EDTA, 20% glycerol, 1 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, and 1 g/ml each leupeptin and aprotinin). Nuclear extracts were frozen on dry ice and stored at Ϫ80°C. The final protein concentration of the nuclear extracts ranged from 5 to 8 mg/ml.
DNase I Footprint Assay-DNase I footprinting was performed according to a method previously described (11). Labeled probe (Ϫ476 to Ϫ21) was prepared by PstI/HindIII digestion of construct pXP1/(Ϫ744 to Ϫ21), and the digested probe was isolated and labeled with [␣-32 P]dATP by Klenow DNA polymerase. The labeled probe (2.5 ϫ 10 4 cpm) was incubated with either purified Sp1 (Promega) or HUVEC nuclear extracts (30 g) at room temperature for 15 min in 50 l of binding buffer containing 50 g/ml bovine serum albumin, 10 g/ml poly(dI:dC), and 0.03% Nonidet P-40 to allow binding and then digested with 0.15 g/ml DNase I at room temperature for 1 min. The samples were analyzed on a sequencing gel.
Electrophoretic Mobility Shift Assay-The shift assay was performed by a previously described procedure (12). The binding mixture (20 l) contained 5 ϫ 10 4 cpm of 32 P-labeled Sp1 consensus oligonucleotide or DNA probe, 10 g of HUVEC nuclear extracts, or 5 ng of purified Sp1 and 2.5 g of poly(dI:dC) in a binding buffer (10 mM HEPES, pH 7.9, 50 mM KCl, 0.1 mM EDTA, 10% glycerol, 0.5 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, 0.5 g/ml each of leupeptin and aprotinin). After a 15-min incubation on ice, the samples were incubated at room temperature for 20 min. Band shift patterns were resolved by electrophoresis. In competition experiments, nuclear extracts or purified Sp1 were incubated for 5 min with the unlabeled oligonucleotide or DNA fragment in a 50 -150-fold molar excess prior to the addition of the labeled probe. The gel supershift assay was performed by adding 2 g of rabbit polyclonal anti-Sp1 antibody (PEP2, Santa Cruz Biotechnology) to the DNA/protein mixture for 30 min on ice, and the band formation was analyzed on gel electrophoresis as described above. An unrelated rabbit polyclonal anti-PGHS-1 antibody was included as negative control.
Primer Extension Analysis-Primer extension using HUVEC PGHS-1 mRNA as the template was performed by a procedure previously described (6). Primer extension using the PGHS-1 promoter/ luciferase construct was done similarly. The primer used in the extension analysis is shown in Fig. 6. The extension products were analyzed on a 6% polyacrylamide sequencing gel.
Rapid Amplification of cDNA Ends (RACE)-The 5Ј-end RACE for determining the 5Ј-end of the PGHS-1 transcript was based on the procedure previously described (13). 3 g of cellular RNA prepared from HUVECs was reverse transcribed to cDNA using an antisense PGHS-1 oligonucleotide (ϩ83/ϩ107 relative to the translation start site). The aliquoted cDNA products were used for amplification by PCR using the following primers: primer a, 5Ј-GACTCGAGTCGACATCGATTTTTTT-TTTTTTTTT-3Ј, which contains an XhoI digestion site as underlined, and primer b, 5Ј-CCCTCGAGCAGGACGGGGAGCGGC-3Ј, which also contains an XhoI digestion site. This primer contains an antisense sequence corresponding to PGHS-1 nucleotides ϩ70 to ϩ51. RACE reaction products were digested with XhoI and cloned into the XhoI sites of pGEM7. The transformed cells were screened with PGHS-1 genomic probe (Ϫ744/ϩ163).

Functional Analysis of the 5Ј-Untranslated Region of the
Human PGHS-1 Gene-A genomic clone containing a 2.5-kb fragment of the 5Ј-flanking region of the human PGHS-1 gene was isolated and sequenced. The sequence reveals that this region bears several putative binding sites for transcriptional activators (Fig. 1). The adenine residue of the ATG is designated as ϩ1. A GC box containing a canonical Sp1 binding site (GGGGCGG) is located at nucleotides Ϫ610 to Ϫ604. Three shear stress-responsive elements (SSRE) are localized near nucleotides Ϫ395, Ϫ625, and Ϫ1810, respectively. In addition, two Sp1 sites are located at nucleotides Ϫ83 to Ϫ89 and Ϫ105 to Ϫ111 proximal to the ATG site. To characterize cis-acting elements in the human PGHS-1 promoter region, we constructed a 2075-bp fragment (Ϫ2095/Ϫ21) and a series of 5Јdeletion mutants into a promoterless luciferase expression vector, pXP1. These constructs were transiently transfected into the cultured HUVECs by lipofection according to the methods described under "Experimental Procedures." The parental construct containing nucleotides from Ϫ2095 to Ϫ21 of the PGHS-1 gene conferred strong constitutive luciferase expression. The promoter activity of this parental construct was about 22% of that of the pSV2-luc construct, which utilizes the SV40 early promoter and enhancer to drive luciferase expression (Fig. 2). The promoter activity remained unchanged by progressive 5Ј-deletion until reaching nucleotide Ϫ565, where the activity of construct Ϫ565/Ϫ21 dropped to about 50% of the parental construct activity (Fig. 2). There was a small gradual decline in the promoter activity with further progressive deletion. However, even a short fragment Ϫ137/Ϫ21 still conferred about 100-fold higher activity than the promoterless vector, pXP1. To demonstrate that the basal promoter activity resides between Ϫ137 and Ϫ21, we constructed two 3Ј-deletion mutants, Ϫ916/Ϫ126 and Ϫ257/Ϫ126, into the luciferase reporter gene and carried out transient transfection experiments. As shown in Fig. 2, both constructs almost entirely lost the basal promoter activity. To normalize the variation of transfection efficiency among these constructs, we also performed co-transfection experiments. The ␤-galactosidase reporter gene driven by the SV40 early promoter was co-transfected with each construct described above into the HUVECs. The promoter activities of these constructs were compared by determining the ratios of the luciferase activity to the ␤-galactosidase activity. The results of the co-transfection experiments are consistent with those of the experiments with luciferase alone (Fig. 2). These results strongly suggest that two regions located between nucleotides Ϫ744 and Ϫ565 and between Ϫ137 and Ϫ21 are critical for PGHS-1 promoter activity.
Binding of HUVEC nuclear proteins to these two regions were investigated. Fig. 3A shows band shift when nuclear extracts of HUVECs were incubated with a Ϫ744/Ϫ569 probe containing the distal activator region. Two bands were noted with this probe. Since this region comprises putative binding sites for Sp1 and AP-2, a 150-fold molar excess of consensus oligonucleotide containing Sp1 or AP-2 sequence was preincu-bated with nuclear extracts prior to the addition of the probe. Both bands were specifically competed by Sp1 or unlabeled probes but not by AP-2 oligonucleotide. Furthermore, a 50-fold molar excess of Ϫ744/Ϫ569 in which the canonical Sp1 site,  The right lanes show the supershift of nuclear extract and the Ϫ119/ Ϫ94 complex. The arrows denote the two bands formed between nuclear extract and the probe, which were supershifted in the presence of anti-Sp1 antibodies.
Ϫ610 GGGCGG Ϫ604 was mutated to GTTTCTG failed to inhibit the formation of these two DNA-protein complexes (Fig. 3A,  lane 5). These results indicate that Sp1 and/or closely related proteins bind to the canonical Sp1 site. This is further confirmed by the formation of two bands when labeled consensus Sp1 oligonucleotides were incubated with HUVEC nuclear extracts (Fig. 3B). Both bands were competed by unlabeled Sp1 oligonucleotides (lane 2) and Ϫ744/Ϫ569 fragment (lane 5) but not by AP-2 or NF-B oligonucleotides (lanes 3 and 4). In all of the experiments, nuclear extracts formed two bands with the canonical Sp1 site. Since both bands are specifically competed by Sp1 sequences but not by mutated Sp1 sequence, both bands are complexes formed between Sp1 and/or Sp1-related protein and the Sp1 sequence. Our results are in keeping with several reports of the formation of two distinct complexes between nuclear extracts and Sp1 binding sites (14 -17).
Identification of the Proximal Sp1 Element-The proximal activating region (nucleotides Ϫ137 to Ϫ21) is GC-rich and contains at least two Sp1 sites (Fig. 1). The DNase I footprinting assay revealed a protected area from nucleotide Ϫ114 to Ϫ98 when labeled probes were incubated with Sp1 (Fig. 4A) or HUVEC nuclear extracts (data not shown). This protected area bears an Sp1 site (GGGGTGG) (Fig. 4A). When labeled probes (Ϫ137 to Ϫ21) containing the protected region were incubated with HUVEC nuclear extracts, two bands were formed (Fig. 4B,  lane 2), and both bands were competitively inhibited by unlabeled probes, Sp1 consensus sequence, and/or an Sp1-containing oligonucleotide (Fig. 4B, lanes 3, 5, and 6, respectively). More importantly, both bands were not competed by a 50-fold molar excess of the parental probe where only the Ϫ111 GGGGTGG Ϫ105 site had been mutated (Fig. 4B, lane 4). The results clearly demonstrated the binding of Sp1 and/or its related proteins to the Ϫ111/Ϫ105 site but not to the Ϫ89/Ϫ83 site. A single band shift was noted when purified Sp1 proteins were incubated with labeled probe containing the protected area (Fig. 4C, lane 2), and this band was competitively inhibited by Sp1 oligonucleotides (lane 5). This band was super-shifted with specific antibody directed against Sp1 (lane 3) but not with unrelated antibody such as anti-PGHS-1 antibody (lane 4). Incubation of HUVEC nuclear extracts with this probe resulted in the formation of two complexes (Fig. 4C, lane 7). Both bands were competitively inhibited by Sp1 oligonucleotides (Fig. 4C, lane 7 versus lane 8) and supershifted with anti-Sp1 antibodies (lane 6). Hence, Sp1 and/or Sp1-related proteins bind to a Sp1 site at Ϫ111 to Ϫ105.
Effects of Mutation of Sp1 Binding Sites on PGHS-1 Promoter Activity-To ascertain that these two separate Sp1 binding sites are functionally important in enhancing the PGHS-1 basal promoter activity, one or both Sp1 binding sites in the 5Ј-flanking promoter (Ϫ916/Ϫ21) of the PGHS-1 gene were altered by site-directed mutagenesis. These mutants were constructed in pXP1 luciferase expression vectors and transfected in HUVECs. Alteration of the distal Sp1 ( Ϫ610 GGGGCGG Ϫ604 to GTTTCAG) reduced the promoter activity to 53% of that of the wild-type promoter (Fig. 5, b versus a). Alteration of the proximal Sp1 binding site from Ϫ111 GGGGTGG Ϫ105 to GTTT-TAG reduced the promoter activity to 38% of the wild-type promoter (Fig. 5, c versus a). When both Sp1 binding sites were simultaneously mutated, the promoter activity was reduced to 29% of that of the wild-type promoter. This reduction is statistically significantly larger (p Ͻ 0.05) than the reduction caused by individual mutation of the proximal or distal site.
The extent of promoter activity reduction caused by the distal Sp1 site mutation is comparable with that of the 5Јdeletion mutants in which the distal Sp1 site was deleted (Ϫ565/Ϫ21 in Fig. 2 versus Fig. 5b). To determine whether mutation of the Sp1 site in the construct (Ϫ257/Ϫ21) would lead to a comparable reduction in the promoter activity, we mutated the proximal Sp1 site located in the Ϫ257/Ϫ21 fragment and expressed it in HUVECs. The promoter activity expressed by this was only 27% of that of the wild-type promoter (Ϫ916/Ϫ21) (Fig. 5f). This value was essentially identical to that of the double Sp1 site mutations in Ϫ916/Ϫ21, confirming that these two Sp1 binding sites contribute to 73% of the basal

PGHS-1 promoter activity.
Experiments were then carried out to determine the impact of Sp1 site mutations on PMA-stimulated promoter activity conferred by the Ϫ916/Ϫ21 region. PMA treatment (50 nM, 4-h incubation) increased the promoter activity of the wild-type Ϫ916/Ϫ21 by 1.8-fold (Fig. 5a). This result was comparable with that previously reported (18). Mutations of either or both Sp1 sites were accompanied by a marked decrease in the basal promoter activity as described above. However, the level of stimulation by PMA treatment was not significantly altered by the mutations (Fig. 5, b-f). These results suggest that PMA stimulation of PGHS-1 promoter activity depends on additional activators.
Mapping the TSS-Multiple TSS for PGHS-1 were identified by primer extension and S1 nuclease mapping in our previous study (6). A major TSS was identified as adenine Ϫ135, relative to the ATG translation start codon. TSS were situated upstream from the proximal Sp1 cognate site (Ϫ111/Ϫ105). However, primers used in those experiments were corresponding to nucleotide sequence Ϫ22 to Ϫ48, which might mask TSS downstream from the proximal Sp1 site. Additional primer extension experiments were, therefore, carried out to reevaluate the TSS. A P2 primer corresponding to nucleotides ϩ115 to ϩ137 (Fig. 6A) was used as an antisense primer in extension experiments. The results from one experiment are shown in Fig. 7A.
Multiple bands corresponding to A Ϫ31 , G Ϫ33 , G Ϫ37 , and A Ϫ135 were noted on the primer extension gel. Of these four TSS, A Ϫ135 is in accord with that detected by using nucleotides Ϫ22/Ϫ48 as the primer. Results from two other experiments revealed four bands corresponding to A Ϫ31 , G Ϫ33 , G Ϫ37 , G Ϫ111 , or A Ϫ135 . Hence, TSS at A Ϫ31 , G Ϫ33 , and G Ϫ37 are consistent in all three experiments. A Ϫ135 and G Ϫ111 are, on the other hand, alternative TSS. We observed a similar alternative extension between these two TSS when nucleotides Ϫ22 to Ϫ48 were used as the primer. In all three experiments, A Ϫ31 and G Ϫ33 had the highest density and were considered to be major TSS. Transcription start sites for the PGHS-1 promoter-luciferase construct were determined using PLuc primer corresponding to nucleotides ϩ44 to ϩ70 downstream from the luciferase start codon (Fig. 6B). Multiple bands were detected (Fig. 7B). Four bands with higher densities are mapped to G Ϫ37 , A Ϫ43 , G Ϫ78 , and G Ϫ111 of the PGHS-1 promoter region. Several less dense bands are scattered between A Ϫ43 and G Ϫ111 (Fig. 7B). Judging by the density of the primer extension bands, the major TSS for native PGHS-1 transcript resides in the region at nucleotides A Ϫ31 and G Ϫ33 , whereas the major TSS for luciferase fusion transcript resides in the region at A Ϫ43 . Another region that serves as TSS from both transcripts resides at G Ϫ111 . 5Ј-End RACE experiments were carried out to further determine TSS. The 3Ј-primer used in the 5Ј-end RACE experiments corresponds to the nucleotide sequence from ϩ51 to ϩ70. Two TSS were identified by the RACE procedure, G Ϫ33 and A Ϫ17 . Taken together, the results indicate that a promoter region proximal to the Sp1 enhancer elements contains multiple TSS including A Ϫ43 , G Ϫ37 , G Ϫ33 , A Ϫ31 , and A Ϫ17 . G Ϫ111 is a TSS identified by primer extension of native and luciferase fusion transcripts but is not detected as a TSS by 5Ј-end RACE. A Ϫ135 was identified as an alternative TSS to G Ϫ 111 only when primer extension was performed on native PGHS-1 transcript. G Ϫ78 and several less dense bands were noted only in luciferase fusion transcripts. The importance of G Ϫ111 , A Ϫ135 , or G Ϫ78 as TSS for the PGHS-1 gene in vivo is unclear. It is intriguing to note that these nucleotides are localized upstream from the proximal Sp1 recognition site. A similar spatial relationship has been reported for the promoter of the human cyclin-dependent kinase-2 gene (19), in which among multiple TSS identified, one TSS was located upstream from an Sp1 site functionally important in basal transcription for this gene. DISCUSSION Results from this study demonstrate that two Sp1 elements are essential for basal transcription of the human PGHS-1 gene. Evidence to support this consists of specific binding of purified Sp1 and nuclear extract proteins to these two regions and a marked loss of promoter activity by Sp1 site mutations. Sp1 is a sequence-specific, ubiquitously expressed nuclear factor essential for basal expression of a variety of eukaryotic genes (for review, see Ref. 20). It confers transcriptional activation by interacting with transcription-associated factors, thereby facilitating the assembly of the basal transcription machinery (21). Reported data suggest that for activation of constitutive transcription of mammalian TATA-less gene, Sp1 is required to bind to a spatially defined region, within 100 base pairs upstream from the TSS (22)(23)(24)(25)(26). Hence, Sp1 involvement in basal transcription of the PGHS-1 gene is different from that of other reported mammalian housekeeping genes in that two Sp1 sites required for basal transcription are separated by about 500 bp on the 5Ј-flanking region. To our knowledge, this is the first instance of critical involvement of two distantly located Sp1 sites in the activation of basal mammalian housekeeping gene transcription.
It has been reported that the activity of the thymidine kinase promoter of the herpes simplex virus and an artificially constructed promoter is enhanced by interaction of two spatially widely separated Sp1 proteins to form Sp1 multimers (27)(28)(29). Detailed analysis of the promoter region by electron microscopy revealed DNA looping whereby the separated Sp1 proteins were brought into contact to form a tetramer followed by multiple tetramer formation (30). It is conceivable that basal PGHS-1 transcription regulated by two Sp1 sites may be mediated by a similar mechanism.
It is estimated from the mutation experiments that these two Sp1 enhancer elements contribute to about 70 -75% of the promoter activity conferred by the 2.0-kb promoter/enhancer fragment of PGHS-1 gene. Full basal promoter activity may require the involvement of additional enhancer element(s). Comparison of the promoter activity conferred by the construct shown in Fig. 5d versus Fig. 5f suggests that the region between nucleotides Ϫ257 and Ϫ21 bears additional enhancer elements important in full PGHS-1 gene expression. In this region, besides the functionally active Sp1 site at Ϫ105 to Ϫ111, there is another Sp1 site located at Ϫ83 to Ϫ89. The 3Ј-mutant fragment Ϫ257/Ϫ126 (Fig. 2) in which both Sp1 sites are removed exhibited almost no promoter activity. This result implies that the Sp1 site at Ϫ83 to Ϫ89 may be important in PGHS-1 basal transcription in vivo. However, this 3Ј-deletion mutant is probably devoid of the binding site for the transcription initiation complex and consequently is expected to confer minimal (if any) promoter activity, even when enhancer elements are present distally such as the Ϫ916/Ϫ126 mutant (Fig.  2). Since Sp1 did not bind to the Sp1 site at Ϫ83 to Ϫ89 by DNase I footprinting, it would be highly unlikely that this Sp1 site is functionally active. Other sites in this region including a putative PEA3 binding site ( Ϫ155 AGGAAG Ϫ150 ) may be the potential enhancer element for a full PGHS-1 promoter activity. This is now being investigated.
It has recently been reported in different types of cells that PGHS-1 gene expression is stimulated by serum, cytokines, or growth factors (for a review, see Ref. 2). We have shown that endothelial PGHS-1 expression is stimulated approximately 2-fold over the basal level by PMA and interleukin-1␤ (18). In this study, our results indicate that PMA increased the promoter activity conferred by the 5Ј-flanking promoter/enhancer of the PGHS-1 gene in HUVECs. Stimulation of the promoter activity by PMA is not entirely dependent on the two Sp1 sites but requires additional elements. Sp1 is involved in stimulation of gene expression by interacting with other transcriptional activators such as GATA, NF-B, Egr-1, YY1, and Rb (31)(32)(33)(34)(35)(36)(37). It is possible that PMA stimulates PGHS-1 transcription by a similar mechanism. Further studies are needed to elucidate the mechanism by which PMA stimulates PGHS-1 promoter activity.
Nuclear extracts of HUVECs form two distinct bands with distal or proximal probes and with consensus Sp1 recognition sequences. Both bands are specifically competed by unlabeled Sp1 cognate oligonucleotides but not by Sp1 mutant, AP-2, or NF-B sequences. These two bands are complexes of DNA with Sp1 and/or Sp1-related proteins. Purified Sp1, on the other hand, forms only a single DNA⅐Sp1 complex with fragment Ϫ119 to Ϫ94 (Fig. 4C). These results are similar to those reported in several recent studies (13)(14)(15)(16). The reason for the double band formation with nuclear extracts is unclear. It has been attributed to binding of Sp1 and a closely related protein to Sp1 recognition sites (38). Three Sp1-related proteins, Sp2, Sp3, and Sp4, have been identified (39 -42). These isoforms of Sp1 bind to Sp1 recognition sites and are antigenically very close to Sp1. Hence, despite specific inhibition of binding by Sp1 cognate sequences and supershift by specific Sp1 antibodies, the additional band is probably formed as a result of binding of a Sp1-related protein to the Sp1 binding site.