The cell-specific expression of endothelial nitric-oxide synthase: a role for DNA methylation.

The basis for the endothelial cell-restricted expression of endothelial nitric-oxide synthase (eNOS) is not known. While transgenic promoter/reporter mice demonstrated endothelium cell-specific eNOS expression, we found robust expression of episomal eNOS promoter/reporter constructs in cell types that do not express the native eNOS transcript. To explore the mechanism underlying this differential activity pattern of chromatin-versus episome-based eNOS promoters, we examined the methylation status of 5'-regulatory sequences of the human eNOS gene. DNA methylation differed dramatically between endothelial and nonendothelial cell types, including vascular smooth muscle cells. This same cell type-specific methylation pattern was observed in vivo in endothelial and vascular smooth muscle cells of the mouse aorta at the native murine eNOS promoter. We addressed the functional consequences of methylation on eNOS transcription using transient transfection of in vitro methylated promoter/reporter constructs and found that methylated constructs exhibited a marked decrease in the synergistic action of Sp1, Sp3, and Ets1 on eNOS promoter activity. The addition of methyl-CpG-binding protein 2 further reduced the transcriptional activity of methylated eNOS constructs. Importantly, chromatin immunoprecipitation demonstrated the presence of Sp1, Sp3, and Ets1 at the native eNOS promoter in endothelial cells but not in vascular smooth muscle cells. Finally, robust expression of eNOS mRNA was induced in nonendothelial cell types following inhibition of DNA methyltransferase activity with 5-azacytidine, demonstrating the importance of DNA methylation-mediated repression. This report is the first to show that promoter DNA methylation plays an important role in the cell-specific expression of a constitutively expressed gene in the vascular endothelium.

The importance of endothelium-derived nitric oxide (NO) 1 in governing cardiovascular homeostasis is well established, since it contributes to the atheroprotective, antithrombotic, and vasodilatory properties of the endothelium (1). Regulation of the NO pathway in this system involves post-translational regulation of the enzyme involved in NO synthesis, namely endothelial NO synthase (eNOS). It is now appreciated that control of the relative expression of the eNOS mRNA also plays a prominent role (2). This laboratory and others have previously reported the isolation and characterization of complementary and genomic clones for eNOS. The human eNOS gene has been mapped to chromosome 7q35-36 (3,4). Molecular characterization revealed that the human eNOS gene consists of 26 exons, spanning ϳ21 kb of genomic DNA. Two clustered cis-regulatory regions were identified in the proximal TATA-less eNOS promoter using deletion analysis and linker-scanning mutagenesis: positive regulatory domain I (PRD I) (Ϫ104/Ϫ95 relative to transcription initiation) and PRD II (Ϫ144/Ϫ115). Functional domain studies in Schneider's Drosophila line 2 (SL2) cells revealed examples of positive and negative protein-protein cooperativity involving Sp1, variants of Sp3, Ets1, Elf1, and MAZ (5). Therefore, multiprotein complexes form on the activator recognition sites within this 50-base pair region (PRD I and II) of the human eNOS promoter in vascular endothelium (5). eNOS gene expression has been shown to be relatively restricted to the vascular endothelium. In situ hybridization using antisense cRNA probes in a wide variety of human tissues has revealed that the eNOS mRNA transcript is highly localized to the vascular endothelium of large and medium sized arteries (6,7). eNOS is comparable in its cell specificity with common endothelial markers such as CD31/PECAM and von Willebrand factor. mRNAs known to be even more restricted to the vascular endothelium than eNOS are uncommon but include the endothelial growth factor receptor tyrosine kinases VEGF-R1 (Flt-1), VEGF-R2 (Flk-1/KDR), Tie-1, and Tie2/Tek (8). Elucidating the mechanisms involved in the endothelial cell-specific eNOS would provide valuable insight into the poorly understood basis for endothelium-specific gene regulation.
DNA methylation at CpG dinucleotides is a prominent feature of the vertebrate genome. In eukaryotes, DNA methylation has been implicated in a number of distinct cellular pro-cesses including transcriptional regulation, embryogenesis, regulation of chromatin structure, genomic imprinting, X-inactivation, and cancer pathogenesis (9,10). Evidence accumulated during the past 20 years suggests an inverse correlation between transcriptional activity and methylation density. Methyl-CpG is now recognized as a gene-silencing signal (11). Specific methyl-CpGs in the promoter can prevent the interaction of transcription factors with their cognate sites. Many of the trans-factors known to bind to sequences containing CpG dinucleotides, including AP-2, HIF-1␣, and c-Myc, do not bind when the CpG doublets are methylated (12)(13)(14). However, CpG methylation alone does not importantly affect the binding of other transcription factors, such as Sp1 (15) and NF-1 (16). Evidence indicates that DNA methylation can not only interfere with factor binding but can also directly modulate chromatin structure by modifying the interaction between core histones and DNA (17). Methyl-CpG-binding proteins 1 and 2 (MeCP-1 and MeCP-2) and other methyl binding domain proteins also bind preferentially to 5-methyl-CpG dinucleotides (18 -20) and modulate transcriptional activity in a number of ways. Binding of these proteins can limit access to the recognition site of transcription factors or modulate DNA structure indirectly as a consequence of local binding. MeCP-2 is a broadly expressed protein specific for single symmetrically methylated CpG dinucleotides that also possesses two active transcriptional repression domains (18). MeCP-2 can exist as a complex with histone deacetylase and the transcriptional corepressor Sin3A (19,21), providing a link between DNA methylation and histone deacetylation. MeCP-2 has also recently been demonstrated to associate with histone methyltransferase activity that is specific for lysine 9 of histone H3, a mark of repressive chromatin (22). An intriguing facet of transcriptional regulation by methyl-CpG is the growing appreciation of the specificity of the response (11,23). For example, MeCP-2 interactions with the Hairy2a promoter (25) and a unique brain-derived neurotrophic factor promoter (24) are dynamic and highly regulated.
Although important cis-elements in the promoter of eNOS have been characterized, little is known about the mechanism of the cell-specific expression of this gene (5, 26 -29). There are two components to the present work. To address the transcriptional regulatory mechanisms implicated in eNOS cell-specific expression, a series of transient transfections of eNOS promoter/reporter constructs was performed in a number of endothelial and nonendothelial cell types. These episomal templates displayed unusually high eNOS promoter activities in a variety of nonendothelial cells that did not express appreciable steadystate levels of eNOS mRNA. These results suggested that the absence of eNOS expression in these cell types was not due solely to the lack of the necessary transcriptional machinery, since episomes were able to accurately serve as a template for transcription initiation. Rather, a mechanism that represses native eNOS gene expression in nonexpressing cells must exist. We considered whether cell specificity could be regulated, in part, at the chromatin level through epigenetic contributions. We report here that DNA methylation plays an important role in the control of the transcriptional regulation as well as the endothelial cell-specific expression of eNOS.

EXPERIMENTAL PROCEDURES
Cell Culture-Bovine aortic endothelial cells (BAEC) and human umbilical vein endothelial cells (HUVEC) were isolated and characterized as described previously (6,30). Cell cultures were utilized for experiments at passages 3-5. Pooled human dermal microvascular endothelial cells (HuDermMVEC), human iliac artery endothelial cells, human pulmonary artery microvascular endothelial cells, human aortic vascular smooth muscle cells (HuAoVSMC), human saphenous vein smooth muscle cells, and human pulmonary artery smooth muscle cells were obtained from Clonetics (Cambrex; East Rutherford, NJ) and maintained as recommended. Primary human hepatocytes were obtained from BD Gentest (Woburn, MA) and were maintained in Hepato-STIM hepatocyte culture medium (BD Gentest). TGW (human neuroblastoma cell line) was a generous gift from Dr. Toshio Kuroki (Department of Cancer Cell Research, Institute of Medical Sciences, University of Tokyo, Japan), and cells were maintained in ␣-minimal essential medium supplemented with 15% fetal bovine serum. KP-N-RT (KPN), also a human neuroblastoma cell line, was a generous gift from Dr. Tohru Sugimoto (Department of Pediatrics, Miyazaki Medical College, Japan) and was maintained in RPMI 1640 supplemented with 10% fetal bovine serum. SL2 was obtained from the American Type Culture Collection (ATCC, Manassas, VA), propagated in Schneider's Drosophila medium supplemented with 10% fetal bovine serum, and maintained at 23°C with atmospheric CO 2 . HepG2 (human hepatocellular carcinoma cell line), HeLa (human cervical squamous cell carcinoma cell line), NIH 3T3 (murine fibroblast cell line), NT2 (human teratocarcinoma cell line), C2C12 (murine muscle myoblasts), CHO (Chinese hamster ovary cell line), JEG-3 (human choriocarcinoma cell line), DAMI (human megakaryocytic cell line), F9 and P19 (murine embryonal carcinoma cell lines), and Jurkat (human mature T cell line) were obtained from ATCC and maintained at 37°C with 5% CO 2 . Cell culture reagents were from Invitrogen.
Drosophila Eukaryotic Expression Constructs-Expression cassettes for Sp1, Sp3, Ets1, and MeCP-2 were based upon pPacUO, a transient episomal vector that contains the 2.6-kb Drosophila actin 5C promoter, a 0.7-kb 5Ј-untranslated region (UTR) ultrabithorax internal ribosome entry site, the first eight codons of the ultrabithorax open reading frame, and 1.1 kb of 3Ј-UTR from the actin 5C gene (polyadenylation signal sequences). pPacUO and pPacUSp1 were provided by R. Tjian (Howard Hughes Medical Institute, Berkeley, IL) and have been described previously (5). pPacUSp3 was provided by G. Suske (Institut fur Molekularbiologie und Tumorfurchung, Marburg, Germany). pPac-UEts1 was constructed as previously described (5). The rat MeCP-2 cDNA was kindly provided by A. Bird (Institute of Cell and Molecular Biology, University of Edinburgh, UK) and has been described previously (31). For construction of pPacUMeCP-2, rat MeCP-2 was cleaved with a partial SacI digestion, and the linearized plasmid was digested to completion with EcoRI. The resulting 1.8-kb fragment was bluntended with Klenow (Invitrogen) and subcloned into the BamHI sites of pPacUO.
Methylation of Promoter/Reporter Constructs-The eNOS promoter/ luciferase reporter construct (pGL2Ϫ1193/ϩ109) was in vitro methylated by various methylases including M. HpaII (C m CGG), M. MspI ( m CCGG), and M. SssI ( m CG) as previously described (16). Mock methylation reactions did not contain any methylase. High and low density methylation constructs were prepared by varying the amount of M. SssI added (i.e. 5 times more M. SssI was added for high density versus low density methylation). The methylation status of the constructs was verified using HpaII and MspI. Methylated and mock-methylated constructs were phenol/chloroform-purified and precipitated in ethanol prior to transient transfection experiments. To assess the role of eNOS promoter-specific methylation in the absence of reporter methylation, methylated or mock-methylated Ϫ1193/ϩ109 and Ϫ265/ϩ109 eNOS sequences were ligated into unmethylated pGL2-Basic using the restriction enzymes SacI/HindIII and PstI/HindIII, respectively. Ligation reactions representing a total of 5 g of DNA and equimolar ratios of insert to vector were phenol/chloroform-extracted, ethanol-precipitated, and directly transfected into cells without DNA preparation and purification in bacteria.
Transient Transfection Assays-All transient transfections were carried out using Lipofectin reagent (Invitrogen). BAEC cultures were plated at 3.3 ϫ 10 4 cells/ml (3.5 ml) and grown on 60-mm dishes 48 h prior to transfection. Transfection conditions were optimized using the SV40 promoter/enhancer luciferase control plasmid, pGL2-Control, to confirm that increasing amounts of templates resulted in proportional increases in reporter activity. Various endothelial and nonendothelial cell types were co-transfected with 1.0 g of promoter/reporter construct, 0.5 g of pRSV-␤-gal, and 1.5 g of pBluescript II SK(Ϫ) DNA. ␤-Galactosidase activity was used to control for transfection efficiency, and pBluescript II SK(Ϫ) DNA was used to optimize DNA/Lipofectin ratios and hence transfection efficiency. DNA-Lipofectin complexes (2:1 (mass/mass) ratio) were incubated for 60 min at 22°C and then added to cells at 37°C in serum-free Opti-MEM I. The transfection mix was replaced at 5 h with RPMI 1640 supplemented with 15% calf serum. Each transfection experiment was performed in triplicate and repeated a minimum of three times. pGL2-Control vector was used as a positive control. The pGL2-Basic vector, lacking both an eukaryotic promoter and enhancer sequences, was used as a negative control. For Drosophila SL2 studies, cells were co-transfected with 1 g of promoter/luciferase construct, the indicated amount of expression plasmids, and 0.5 g of pADH-␤-gal (a gift from R. Tjian) (32). The total amount of DNA transfected was kept constant (2 g) by the addition of pPacUO. SL2 cells were seeded at a density of 2-3 ϫ 10 6 cells/ml at the time of transfection. Cells were harvested 48 h post-transfection with 300 l of lysis buffer (0.1 M potassium phosphate buffer (pH 7.8), 1% Triton X-100, 1 mM dithiothreitol, 2 mM EDTA). Protein extracts were centrifuged at 10,000 ϫ g for 2 min to pellet residual cellular debris and stored at Ϫ80°C for subsequent assay. Luciferase activity was evaluated using the MonoLight 2010C luminometer (Analytical Luminescence Laboratory, Ann Arbor, MI) as described previously (5). Data in raw luciferase units was corrected for nonspecific background of mocktransfected cells, which represented Յ0.5% of experimental luciferase activities, and was then normalized for ␤-galactosidase activity, as previously described (5). For studies involving methylated or mockmethylated ligation reactions, the efficiency of ligation reactions was modest and resulted in luciferase activities for mock-methylated controls that averaged 5-10% of intact nonligated control vectors.
Methylation-sensitive Isoschizomer Mapping-High molecular weight genomic DNA was isolated from various human cell types including HUVEC, HuAoVSMC, HeLa, HepG2, JEG-3, and TGW as described previously (33). Genomic DNA (10 -20 g) was digested with SacI (New England Biolabs, Beverly, MA) at 37°C for 16 h to produce a 2.4-kb fragment encompassing the transcription start site of the human eNOS gene (Ϫ1193 to ϩ1207 bp). This fragment was further digested with either MspI (methylation-insensitive) or HpaII (methylation-sensitive) 5Ј-CCGG-3Ј (New England Biolabs). Digests were carried out for 16 h at 37°C with excess enzyme to ensure completion of the reaction. To increase the efficiency of HpaII digestion, a HpaII doublestranded oligonucleotide (5Ј-TAT AGC CGG CTA TA-3Ј) was added in 10-fold molar excess of genomic DNA (34). DNA was subjected to agarose gel electrophoresis, downward transferred to nitrocellulose membranes (Scheicher and Schuell), and UV-cross-linked. For Southern hybridization, a 986-bp SacI/PstI fragment (from Ϫ1193 to Ϫ207 bp of the human eNOS promoter) was labeled with [␣-32 P]dCTP using a nick translation kit (Amersham Biosciences) to a specific activity of 0.5 to 1.0 ϫ 10 9 cpm/g. Membranes were washed at high stringency and subjected to STORM PhosphorImager analysis. Experiments were performed with multiple independent DNA isolations and repeated a minimum of three times.
Sodium Bisulfite Genomic Sequencing-Genomic DNA (5 g) was subjected to sodium bisulfite treatment, as described previously (35). Given that the nascent strand of replicating DNA is hemimethylated immediately following DNA replication, only quiescent postconfluent cells were utilized for DNA isolation. BamHI-digested DNA was denatured with 0.3 M NaOH for 15 min at 37°C in a volume of 20 l, treated with 3.1 M sodium bisulfite and 0.5 mM hydroquinone, overlaid with mineral oil, and incubated at 55°C for 16 h in the absence of light. Free bisulfite was removed using a Promega Wizard DNA clean-up desalting column, followed by an incubation with 0.3 M NaOH for 15 min at 37°C to denature and remove the -SO 3 adduct from the uracil bases prior to the PCR. Sample DNA was then neutralized with NH 4 OAc, precipitated, and stored at Ϫ20°C. An aliquot of the bisulfite-treated DNA (25-50 ng) was subjected to 35 cycles of PCR amplification, followed by another 35 cycles of nested PCR amplification. PCR primers in the regions of interest were designed for the sodium bisulfite-modified template. Oligonucleotides were generated using a Beckman Oligo 1000 DNA Synthesizer (Beckman Coulter, Fullerton, CA). Primer sets used in sodium bisulfite genomic sequencing (Table I) were designed to amplify the 5Ј-flanking region and the GC-rich region of exon 24 of the eNOS gene according to criteria previously described (35) using Oligo 4.0 Primer Analysis Software (Molecular Biology Insights, Cascade, CO). The final PCR products were either directly sequenced using an automated ABI Prism 377 DNA Sequencer (Applied Biosystems, Foster City, CA) or subcloned using the TA cloning kit (Invitrogen) to yield 10 -20 individual plasmid clones for sequencing.
Laser Capture Microdissection-Tissue samples were obtained from adult C57/Bl6 mice that were sublethally anesthetized with a ketamine (10 mg/ml) and xylazine (2 mg/ml) mixture by intraperitoneal injection. Frozen tissue sections, 8 m thick, were collected onto noncharged SuperFrost slides (Fisher). Tissue sections were fixed, stained, and dehydrated prior to laser capture microdissection using the PixCell II System (Arcturus, Mountain View, CA). Endothelial cells and vascular smooth muscle cells were microdissected from the greater curvature of the aortic arch. Genomic DNA isolation from captured cells was performed using the Arcturus DNA PicoPure isolation kit. Genomic DNA was purified using a standard DNA precipitation protocol with 50 ng of salmon sperm DNA added as a carrier prior to sodium bisulfite treatment. Following nested PCR of the mouse eNOS promoter (primers listed in Table I), 10 -20 subcloned PCR products were sequenced. Laser capture microdissection and sodium bisulfite sequencing were performed on three separate mice. Similar approaches were used to isolate endothelial cells from hemizygous insertional Ϫ5200/ϩ28 Mu eNOS nls LacZ transgenic adult mice (36). Three lines containing one, five, or 10 copies of the murine eNOS promoter/␤-galactosidase reporter cassette were studied. Primers for bisulfite analysis spanned the eNOS promoter/reporter junction and therefore could distinguish insertional genomic events from the native murine eNOS promoter.
Chromatin Immunoprecipitation-ChIP was performed using the ChIP Assay Kit as recommended by the manufacturer (Upstate Biotechnology, Inc., Lake Placid, NY). Briefly, ϳ3 ϫ 10 6 cells were used per ChIP assay. Formaldehyde was added to a final concentration of 1% and incubated at 37°C for 10 min. Sonication was performed on ice using a Sonics and Materials Vibra-Cell sonicator with a 3-mm tip set at 30% maximum power using 5 ϫ 10-s pulses with a 10-s interval between sonications to achieve chromatin fragments ranging between 200 and 400 bp in size. Samples were diluted 10-fold in ChIP dilution buffer, and a 20-l aliquot (1% of total) was removed to serve as an input sample. Chromatin was precleared with 80 l of a mixture of salmon sperm DNA/Protein A/Protein G at 4°C with rotation for 2 h, followed by the addition of 2 g of Sp1, Sp3, or Ets1 antibody or 5 g of RNA Pol II antibody, 2 g of control IgG, or a no antibody control. Antibodies used for ChIP analysis were as follows: Sp1 (catalog no. 07-124; Upstate Biotechnology), Sp3 (catalog no. 07-107; Upstate Biotechnology), Ets1 (catalog no. SC-350; Santa Cruz Biotechnology, Inc., Santa Cruz, CA), and RNA polymerase II (catalog no. SC-899; Santa Cruz Biotechnology). Immunoprecipitations were performed at 4°C overnight with rotation. To collect immune complexes, 60 l of a mixture of salmon sperm DNA/Protein A/Protein G was added and incubated at 4°C with rotation for 2 h. After washing immune complexes, formaldehyde crosslinks were reversed in immunoprecipitated samples and the input chromatin sample by the addition of 20 l of 5 M or 2 l of NaCl, respectively, and incubation at 65°C for 4 h. Following proteinase K treatment, phenol/chloroform extraction, and ethanol precipitation, DNA was resuspended in 25 l of water. tRNA was used to aid in precipitation. ChIP analyses were performed at least three times on three independent lots of HUVEC or HuAoVSMC. Similar results were achieved using a separate set of antibodies for Sp1 (catalog no. SC-59X; Santa Cruz Biotechnology), Sp3 (catalog no. SC-644X; Santa Cruz Biotechnology), and Ets1 (catalog no. SC-111X; Santa Cruz Biotechnology). Cellular levels of Sp1, Sp3, and Ets-1 were assessed using Western blots, as described previously (37). Total cellular protein was extracted from confluent plates of HUVEC and HuAoVSMC, and 20 g of protein was size-fractionated on a 7.5% polyacrylamide gel and transferred overnight onto a polyvinylidene difluoride membrane. Polyclonal antibodies used were as follows: Sp1 (2 g/ml; Upstate Biotechnology), Sp3 (2 g/ml; Upstate Biotechnology), and Ets1 (1 g/ml; catalog no. SC-350; Santa Cruz Biotechnology) and a horseradish peroxidase-linked secondary antibody (Amersham Biosciences). Detection was performed using the Supersignal West Pico chemiluminescent substrate (Pierce) and the Fluor-S Max Multimager (model 170-7720; Bio-Rad).
Real Time Polymerase Chain Reaction-RNA was extracted from vehicle and 5-azacytidine-treated HUVEC, HuAoVSMC, and HeLa as previously described (38). 5 g of RNA was reverse transcribed with random hexamers using the Superscript II kit from Invitrogen according to the manufacturer's recommendations. cDNA was diluted to a final volume of 50 l. The amount of target cDNA was quantified on an ABI 7900HT Sequence Detection System using the Taqman methodology with the following eNOS-specific primers and probe: heNOS1488F (5Ј-GGC ATC ACC AGG AAG AAC ACC-3Ј), heNOS1522R (5Ј-TCA CTC GCT TCG CCA TCA C-3Ј), and heNOS1522T (5Ј-FAM-CCA ACG CCG TGA AGA TCT CCG C-3Ј). Similarly, the following GAPDH-specific primer set and probe were utilized: GAPDHforward (5Ј-GAA GGT GAA GGT CGG AGT C-3Ј), GAPDHreverse (5Ј-GAA GAT GGT GAT GGG ATT TC-3Ј), and huGAPDH (5Ј-VIC-CAA GCT TCC CGT TCT CAG CC-3Ј). Measurements were performed in triplicate on 2 l of cDNA in a 25-l volume using the Taqman universal master mix with the following cycling parameters: 95°C for 10 min, followed by 40 cycles of 95°C for 15 s and 60°C for 1 min. Transcript copy number was quantified by comparison with a calibration curve using known amounts of amplicon-containing plasmid. Copies of the eNOS transcript/g of total RNA were assessed as raw copy number and were normalized to GAPDH. For ChIP experiments, real time PCR was performed similarly using the following primers directed against the eNOS proximal promoter (Ϫ166 to Ϫ26): heNOSprom5Ј (5Ј-GTG GAG CTG AGG CTT TAG AGC-3Ј) and heNOSprom3Ј (5Ј-TTT CCT TAG GAA GAG GGA GGG-3Ј) with probe eNOSpromTQMprobe (5Ј-FAM-CCA GCC GGG CTT GTT CCT GTC C-TAMRA-3Ј). Determinations were performed in triplicate on 2 l of bound chromatin, 2 l of a no antibody control or control IgG immunoprecipitation, and 2 l of a 10-fold dilution of input chromatin in a 25-l reaction. The number of copies of target sequence was determined by comparison with a standard curve using known amounts of an eNOS promoter-containing plasmid. The number of target DNA molecules immunoprecipitated by the specific antibody was divided by the number of copies nonspecifically immunoprecipitated in the no antibody control or IgG sample. Results are expressed as mean Ϯ S.E. of at least three independent experiments.
RNase Protection Assays-Total RNA was extracted from various human cell types (HUVEC, HepG2, HeLa, JEG-3, and TGW) as described previously (3). 50 Ci of [␣-32 P]CTP (800 Ci/mmol) was used to generate an internally labeled antisense riboprobe in an in vitro transcription reaction using T7 RNA polymerase (Promega, Madison, WI) and 1 g of EcoRI (New England Biolabs) linearized template DNA spanning the 292 terminal nucleotides of the human eNOS cDNA. Specific activity of the probe averaged 1.0 ϫ 10 9 cpm/g. A labeled RNA molecular size standard was generated using the Century Marker template (Ambion, Austin, TX). Hybridizations and RNase digestions were performed as previously described (33). Samples were electrophoresed through a denaturing 6% acrylamide, 8 M urea gel, followed by autoradiography and STORM Phos-phorImager analysis (Amersham Biosciences).
Data Analysis-Unless otherwise indicated, data are expressed as the mean Ϯ S.E. obtained from at least three independent experiments, each done in triplicate. Comparisons were made with an unpaired Student's t test or analysis of variance, followed by the Student-Newman-Keuls test, where appropriate. The level of statistical significance was defined as p Ͻ 0.05.

eNOS Promoter Activity in Endothelial and Nonendothelial
Cell Types-We assessed the cell specificity of human eNOS promoter/reporter constructs by transiently transfecting the plasmids pGL2Ϫ1193/ϩ109 or pGL2Ϫ133/ϩ109 into a number of endothelial and nonendothelial cell types. These plasmids contained 1193 or 133 bp of the human eNOS promoter and 109 bp of exon 1 upstream of a luciferase reporter. In prior studies, pGL2-1193/ϩ109 was found to be maximally active in endothelial cells derived from a variety of sources (5). To provide a relative assessment of the endothelial specificity of the eNOS promoter, the functional activity of eNOS constructs was compared with a strong heterologous promoter (pGL2-Control, SV40 promoter/enhancer). Shown in Table II are luciferase   activities (in relative light units) of the pGL2-1193/ϩ109 and pGL2-control constructs (mean Ϯ S.E.; n ϭ 3-4). All reporter gene activities were normalized with respect to ␤-galactosidase activity. We observed that the pGL2-1193/ϩ109 was strongly active in both BAEC and HUVEC (11,300 Ϯ 1140 raw luciferase units and 2180 Ϯ 170 raw luciferase units, respectively). When expressed as a percentage of pGL2-Control luciferase activity (to normalize for transfection efficiency across cell types), BAEC and HUVEC exhibited similar levels of transcriptional activation of the pGL2-1193/ϩ109 eNOS promoter/reporter construct (9.8 Ϯ 0.4 and 11.5 Ϯ 0.9%, respectively). Surprisingly, many of the nonendothelial cell types also demonstrated considerable transcriptional activity. These included KPN (220 Ϯ 44% of pGL2-Control), TGW (420 Ϯ 27%), C2C12 (30.5 Ϯ 2%), P19 (48.0 Ϯ 8.6%), F9 (34.3 Ϯ 3.0%), DAMI (10.8 Ϯ 0.2%), and HepG2 (7.0 Ϯ 0.4%). Importantly, these cell types did not express appreciable levels of steady-state eNOS mRNA transcripts, as assessed by RNase protection assays ( Fig. 1) and Northern blot analysis (data not shown). Larger regions of the eNOS promoter (Ϫ3500/ϩ109) were also expressed robustly in endothelial and nonendothelial cell types (data not shown), and the core promoter (Ϫ133/ϩ109) demonstrated activity that averaged ϳ25% the activity of the Ϫ1193/ϩ109 construct in the varied cell types (Table II). Therefore, the core promoter (Ϫ133/ ϩ109) or larger portions of the eNOS promoter (Ϫ1193/ϩ109 or Ϫ3500/ϩ109) did not confer endothelial cell specificity in transient transfection analyses. To analyze the mechanisms of transcriptional regulation important in the human eNOS gene in these nonendothelial cell types, we used a series of eNOS 5Јto 3Ј-deletion promoter/reporter constructs to assess functional activity profiles in two representative nonendothelial cell types and compared this activity to activity profiles in BAEC. Constructs had a variable 5Ј-end but a common 3Ј-end, terminating at ϩ109 relative to the start site of transcription. Shown in Fig.  2, eNOS constructs had a similar profile of transcriptional activity in BAEC, HepG2, and C2C12 cells. We previously demonstrated functionally important contributions to eNOS transcription in endothelial cell types from two positive regulatory domains (PRD I and II) in the proximal promoter of the human eNOS gene (5). In all of the cell types examined, the Ϫ151/ϩ109 construct, which contains PRD I and II, exhibited more activity than the Ϫ133/ϩ109 construct, which only contains PRD I. Removing PRD I (Ϫ92/ϩ109) further reduced the activity of transfected constructs.
Prior studies from us and others have indicated that transgenic mice containing eNOS promoter/reporter constructs exhibit endothelial cell-restricted reporter expression (36,39). Although native eNOS mRNA and chromosomally integrated eNOS promoter activity are relatively endothelium-restricted, the in vitro activity of episomal eNOS promoter/reporter constructs are not.
The eNOS Promoter Region Is Differentially Methylated in Human Endothelial and Nonendothelial Cell Types-The methylation status of the proximal promoter of the human eNOS gene was compared between endothelial cells (HUVEC) and nonendothelial cells (HuAoVSMC, TGW, HeLa, HepG2, and JEG-3) using Southern hybridization with methylationsensitive HpaII and methylation-insensitive MspI restriction  endonucleases. Genomic DNA isolated from the various cell types was digested with SacI to generate a 2.4-kb restriction fragment (Ϫ1193 to ϩ1207 bp, relative to the start site of transcription), representing the proximal promoter and exon 1 regions (Fig. 3A). SacI-digested genomic DNA was further cleaved with MspI or HpaII. 10 MspI/HpaII sites are located in this genomic region. The hybridization probe consisted of a 986-bp SacI/PstI fragment corresponding to a region from Ϫ1193 to Ϫ207 bp and contained two MspI/HpaII sites. Digestion of genomic DNA from a variety of nonendothelial cell types (HuAoVSMC, TGW, HeLa, and JEG-3) with HpaII gave rise to Southern hybridization bands of higher molecular weight than those obtained with MspI digestion (Fig. 3B). These findings suggest that the eNOS proximal promoter is methylated in a variety of nonendothelial cell types. In contrast, HpaII tiny fragments were clearly evident in HpaII-digested HUVEC genomic DNA, indicating that the proximal promoter of the eNOS gene is relatively hypomethylated in this cell type. In contrast to the other nonendothelial cells examined, HepG2 demonstrated a pattern of HpaII tiny fragments similar to HUVEC. These results demonstrate that this genomic region is methylated in several transformed and freshly isolated nonendothelial cell types but is unmethylated in HUVEC and HepG2.
The methylation pattern of all CpG doublets within the eNOS promoter between nucleotide positions Ϫ1186 and Ϫ51 (Supplemental Fig. S1) in two types of endothelial cells (HUVEC and HuDermMVEC) and two types of vascular smooth muscle cells (human aortic VSM cells (HuAoVSMC) and human saphenous vein smooth muscle cells (HuSV-VSMC)) was determined by high resolution sodium bisulfite genomic sequencing (Fig. 4A). 15-20 subclones for each amplicon from each cell type were analyzed. A high density of methylation was present in vascular smooth muscle cell types at core promoter regions, but little or no methylation was present in endothelial cell types. DNA methylation extended up to the Ϫ361 CpG dinucleotide in VSM cells, with low levels of methylation detected upstream of this site. The differentially methylated region spanning Ϫ361 to ϩ3 includes the well characterized positive regulatory domains I (Ϫ104/Ϫ95) and II (Ϫ144/Ϫ115). CpG sites further upstream (Ϫ4912 to Ϫ4587) (Fig. S1), encom-

FIG. 2. 5 to 3 deletion analysis of eNOS promoter/reporters in endothelial and nonendothelial cell types.
Various lengths of the eNOS promoter fused to a luciferase reporter were transfected into endothelial (BAEC) and nonendothelial (HepG2 and C2C12) cell types. pGL2Ϫ1193/ϩ109, pGL2Ϫ743/ϩ109 and pGL2Ϫ151/ϩ109 contain positive regulatory domains I and II (PRD I and II), pGL2Ϫ133/ϩ109 contains all of PRD I and a portion of PRD II, and pGL2Ϫ92/ϩ109 contains neither PRD. Data are expressed as percentage of luciferase activity relative to pGL2Ϫ1193/ϩ109 and represent the mean Ϯ S.E. (three or four independent experiments, normalized to ␤-galactosidase). Where error bars are not evident, the values are below the figure resolution.
passing an upstream enhancer (27), were also assessed (Fig.  4A). In contrast to the core promoter, this region was only lightly methylated in all cell types and did not display important differences in methylation pattern between cell types. Shown in Fig. 4B and in Supplemental Fig. 2 (Fig. S2) are results from a variety of other human primary cell types and cultured cell lines for the proximal promoter region (Ϫ209 to Ϫ51). This comprehensive methylation analysis revealed a stark contrast between endothelial and both primary and transformed nonendothelial cell types. The nine CpG doublets in the proximal promoter (Ϫ209 to Ϫ51) were completely unmethylated in HUVEC, very lightly methylated in HuDerm-MVEC and human iliac artery endothelial cells, and lightly methylated (ϳ25%) in human pulmonary artery microvascular endothelial cells. In contrast, the proximal promoter was almost completely methylated in all of the vascular smooth muscle cell types studied, as well as in peripheral blood leukocytes, primary human hepatocytes, and TGW, cell types known not to express eNOS mRNA. In HeLa and HepG2, we observed a region of especially low methylation density, also known as a hypomethylation footprint, surrounded by upstream adjacent sites that were methylated to a much higher extent (Fig. 4B). Specifically, in HepG2, the Ϫ95 and Ϫ100 sites were only 10% methylated; however, only 40 bp downstream, the Ϫ51 site was found to be 60% methylated, whereas the Ϫ167 site was found to be 90% methylated. It should be noted that isoschizomer mapping of the eNOS promoter in HepG2 cells produced a restriction fragment profile similar to that of HUVEC (Fig. 3B). This pattern is consistent with the relative hypomethylation at the Ϫ137 5Ј-CCGG-3Ј MspI/HpaII site in HepG2 (Fig. 4B). In HeLa, the Ϫ95 and Ϫ100 sites were 20 and 10% methylated, respectively. This contrasted with the Ϫ137 site, located only 40 bp upstream, which was found to be 80% methylated (Fig.  4B). Analysis of the methylation status of subclones from all of the nonendothelial cell types revealed that none of the individual strands was completely unmethylated.
The genomic sodium bisulfite sequencing method may, for a given primer pair, give rise to an underrepresentation of the extent of methylated DNA (40). Therefore, we analyzed the methylation status of the CpG dinucleotides on the opposite strand (data not shown). Sequencing of PCR products and individual subcloned products indicated that the antisense strand was methylated at exactly the same CpG doublets as the sense strand, suggesting that the DNA was symmetrically methylated on both strands in the cell types evaluated; TGW, HepG2, and HeLa (data not shown). Conversely, examination of the antisense strand in HUVEC did not reveal any methylated cytosines. Since the sequence CpNpG can also be a target of methylation (41), the CpNpG sequence present on the antisense strand in PRD I within the high affinity Sp1 site was assessed with respect to methylation status. Examination of both strands of the proximal promoter region of the eNOS gene did not reveal any methylated sequence outside the context of a CpG dinucleotide. Taken together, these data confirmed the results of methylation-sensitive isoschizomer restriction enzyme mapping. We conclude that there is a significant difference between the methylation status of the eNOS proximal promoter in endothelial cells and nonendothelial cell types.
The eNOS Promoter Is Differentially Methylated in Vivo-Since bisulfite analysis in various human cell types revealed a major difference in the methylation status of the eNOS promoter in cultured endothelial versus nonendothelial cells, we sought to determine whether this pattern recapitulated the cell-specific methylation status of the eNOS gene in vivo. Laser capture microdissection was used to isolate murine aortic endothelial cells and underlying aortic vascular smooth muscle cells from the greater curvature of the aortic arch. The methylation status of the murine proximal eNOS promoter was examined using sodium bisulfite sequencing with primers designed to the top (sense) converted strand (Table I). Importantly, the location of the majority of the CpG sites was conserved between human and mouse eNOS. Subcloning the PCR  HeLa, HepG2, and JEG-3). C, sodium bisulfite analysis of the native murine eNOS locus in laser capture microdissection-extracted endothelial and vascular smooth muscle cells from mouse aorta. D, schematic representation of the murine eNOS promoter/␤-galactosidase reporter transgene. Shown above is the location of the amplicon used for sodium bisulfite analysis, with forward primers located in the eNOS promoter and reverse primers located in the ␤-galactosidase reporter. Sodium bisulfite analysis of laser capture microdissection-extracted endothelial cells from the aortas of transgenic mice containing 10 copies of the eNOS/␤-galactosidase transgene is shown to the right. products and sequencing of the subclones yielded the methylation status of nine doublets from the Ϫ199 to Ϫ4 eNOS promoter region. The nine CpG doublets examined were all completely unmethylated in endothelial cells, whereas they were completely methylated in aortic vascular smooth muscle cells (Fig. 4C). Thus, DNA methylation differences observed in varied cultured cells are indicative of methylation differences observed in vivo in mice. The methylation status of the eNOS promoter was also examined in endothelial cells isolated by laser capture microdissection from hemizygous insertional Ϫ5200/ϩ28 Mu eNOS nls LacZ transgenic adult mice (36). Fig.  4D indicates that the promoter of the transgene exhibits very low levels of DNA methylation in vascular endothelial cells. These cells express the ␤-galactosidase reporter. Similar findings were observed in three distinct lines of hemizygous mice. Each line is known to represent a separate insertional genomic event (36) and contains 1, 5, or 10 copies of the murine eNOS promoter/␤-galactosidase reporter cassette. A

GC-rich Region in Exon 24 of the Human eNOS Gene Is Methylated in Both Endothelial and Nonendothelial Cell
Types-An assessment of the distribution of CpG dinucleotides in the eNOS gene sequence from Ϫ1323 to ϩ21,720 bp, a region that encompasses the promoter, all 26 exons, and the 3Ј-UTR, revealed that the CpG content (598 CpGs) is depleted for the gene as a whole when compared with GpC content (1675 GpCs). The observed number of CpG dinucleotides (598) is significantly lower than the expected number of CpG dinucleotides (1935), whereas the observed (1675) and expected (1935) numbers of GpC dinucleotides are comparable (Geneworks). A region near the 3Ј-end of the gene (exon 24 to the start of the 3Ј-UTR) was found to be especially GC-rich. In addition, CpG analysis using the Wisconsin GCG program demonstrated that this region satisfied the criteria for a CpG island (42). To study the methylation pattern of this region of the human eNOS gene, primers were designed to amplify exon 24 and a portion of intron 24 for sodium bisulfite genomic sequencing (Table I). Direct sequencing of PCR products from bisulfite-modified DNA was used to determine the methylation status of this region in endothelial (HUVEC) and nonendothelial (HeLa, HepG2, JEG-3, TGW, and peripheral blood leukocyte) cell types. 10 CpG dinucleotides within exon 24 and 15 CpGs in intron 24 were analyzed, and sequencing results indicated that this genomic region was methylated in both endothelial and nonendothelial cell types (data not shown). The lack of DNA methylation differences in this region contrasts with the methylation differences observed between endothelial and nonendothelial cells at the proximal promoter of eNOS.
DNA Methylation Inhibits eNOS Promoter Activity in BAEC-To evaluate the functional consequences of CpG methylation of the eNOS promoter, promoter/reporter constructs containing the 5Ј-flanking region of the human eNOS gene were in vitro methylated using various methylases and transiently transfected into BAEC. The maximally active promoter/ reporter construct, pGL2Ϫ1193/ϩ109, was methylated in vitro using M. SssI, which specifically methylates cytosines of CpG dinucleotides (43). Methylation reactions used varying concentrations of M. SssI. M. HpaII (which methylates 5Ј-C m CGG-3Ј) and M. MspI (which methylates 5Ј-m CCGG-3Ј) were also used to methylate promoter/reporter constructs. In each case, the extent of methylation was verified by comparing the patterns of MspI and HpaII restriction digests (data not shown). M. SssI high density methylation had no effect on the restriction endonuclease MspI but completely inhibited HpaII digestion. M. SssI low methylation density also had no effect on MspI digestion and allowed only partial digestion by HpaII. Based upon isoschizomer digestion of promoter/reporter constructs, we es-timated that ϳ20% of all CpG sites were methylated at the lower concentrations of M. SssI. M. HpaII methylation completely abolished HpaII cutting. Although the methylation density may be similar between the M. SssI low and M. HpaII methylated constructs, the specific sites of methylation must have differed, since only 29 of the 310 CpG dinucleotides represent 5Ј-CCGG-3Ј sites.
Methylation of the pGL2Ϫ1193/ϩ109 promoter/reporter construct had a dramatic effect on eNOS promoter activity in BAEC (Fig. 5) (n ϭ 3, triplicate determinations). eNOS promoter activity was completely abolished by high density methylation (M. SssI high). The two constructs with low density methylation (M. SssI low and M. HpaII) yielded different results. The eNOS promoter activity of the M. HpaII construct was merely 1% that of the mock-methylated construct, whereas the activity of the M. SssI low density methylation construct was ϳ20%. The difference in the promoter activity between the two low density methylation constructs suggests that specific sites, rather than mere density of methylation, may be critical. M. MspI-methylation did not have any significant effect on eNOS promoter activity in BAEC, demonstrating that methylation of the outer C in 5Ј-CCGG-3Ј sites does not have a repressive effect.
It could be argued that methylation of promoter/reporter constructs decreases transfection efficiency. As reported by others (16,44), we assessed the effects of co-transfecting pBluescript plasmid that was mock-methylated or methylated with M. SssI. Co-transfection with methylated carrier DNA partially rescued the transcriptional repression observed when methylated eNOS promoter/reporter constructs were transiently transfected into BAEC (data not shown). The finding that transfecting an excess of methylated DNA increased eNOS FIG. 5. Effects of DNA methylation on eNOS promoter activity in BAEC. Luciferase expression constructs driven by the eNOS promoter (Ϫ1193/ϩ109) were methylated or mock-methylated and transiently transfected into BAEC cells. SssI high density methylation indicates that the promoter/reporter construct was methylated with SssI methylase ( m CG) at a high density. SssI low density methylation indicates that ϳ20% of the CpG sites were methylated using a lower concentration of enzyme. HpaII methylation represents the construct that was methylated with the HpaII methylase (C m CGG). MspI-methylation represents the construct that was methylated with the MspI methylase ( m CCGG). Promoter activity as assessed by luciferase readout was normalized for ␤-galactosidase activity. The data represent the mean Ϯ S.E. Shown is a representative experiment of three (triplicate determinations). *, p Ͻ 0.05. transcriptional activity argues against reduced transfection efficiency of methylated promoter/reporters as a trivial explanation for our functional findings. Also, methylation of Ϫ1193/ ϩ109 and Ϫ265/ϩ109 eNOS promoter regions with M. SssI in the absence of reporter plasmid sequence methylation, accomplished by methylating promoter sequences prior to ligation to unmethylated vector, yielded comparable findings with activity for both representing Ͻ5% of mock-methylated and ligated promoter/reporter constructs (data not shown). These findings argue against an important repressive role for methylation of the reporter open reading frame as an explanation for our findings. Taken together, these results imply that DNA methylation of the eNOS promoter plays an important role in the transcriptional repression of the human eNOS gene.
Both the Sites of DNA Methylation and the Density of Methylation Affect the Synergistic Activation of eNOS Promoter by Sp1, Sp3, and Ets1 in Drosophila Cells-Drosophila melanogaster belongs to the group of eukaryotic organisms that exhibit only modest genomic methylation compared with mammals (45). As well, Drosophila do not express Sp1 or related proteins (46), allowing dissection of the potential role of Sp1/ Sp3 with methylated DNA. We have previously utilized Drosophila SL2 cells to explore the transcriptional properties of the eNOS promoter (5). Transfection of SL2 cells with in vitro methylated Ϫ1193/ϩ109 eNOS promoter/reporter constructs together with an expression vector for Sp1 (pPacUSp1, 40 ng) demonstrated that high density methylation had a modest repressive effect on eNOS promoter activity (activity ϳ60% of mock-methylated constructs) (data not shown). Clearly, the repression observed in this system was modest compared with the marked effects of methylation when constructs were expressed in mammalian cells. We previously proposed a model in which Sp1, Sp3, and Ets1 synergistically activate the human eNOS gene (5). Mutating any of the activator recognition sites for these factors in PRD I (Ϫ104/Ϫ95) or PRD II (Ϫ144/Ϫ115) resulted in a marked decrease in functional promoter activity (5). In order to examine the effects of methylation and methylation density on the synergistic activation of the eNOS promoter by Sp1, Sp3, and Ets1, in vitro methylated promoter/ reporter constructs were transiently transfected into Drosophila Schneider cells along with limiting amounts (10 ng) of each of Sp1 (pPacUSp1), Sp3 (pPacUSp3), and Ets1 (pPac-UEts1) expression vectors (Fig. 6) (n ϭ 3, triplicate determinants). We have previously reported that transient transfection of this concentration of each of these expression vectors on their own has no effect on transactivation (5). The effects of methylation on Sp1/Sp3/Ets1-mediated eNOS transactivation were marked compared with the modest effects of methylation on the effect of Sp1 alone. Results in Fig. 6 indicate that there was a significant decrease in promoter activity compared with the mock-methylated constructs. The M. SssI high density methylation construct exhibited less than 10% of the mockmethylated promoter activity. Methylation at lower densities (M. SssI low density methylation and M. HpaII methylation) did not affect the synergistic transactivation by the three factors to the same extent as high density methylation, with the M. SssI low density methylated construct demonstrating ϳ30% promoter activity, compared with the mock-methylated construct. Taken together, these results suggest that the synergistic action of Sp1, Sp3, and Ets1 on activation of the eNOS promoter is markedly affected by DNA methylation.
MeCP-2 Further Represses Methylation-mediated Transcriptional Repression-In addition to its direct effect on trans-factor binding and chromatin structure, methylation also represses transcription through the binding of methyl-CpG-binding proteins. Drosophila SL2 cells represent a use-ful model for assessing the functional effect of heterologous expression of methyl-CpG-binding proteins (47). To dissect the role of methyl-CpG-binding proteins in methylationmediated transcriptional repression of eNOS, a heterologous eukaryotic expression vector encoding MeCP-2 (pPacUMeCP-2) was used in the Drosophila SL2 system. The addition of 100 ng of MeCP-2 further repressed the already minimal promoter activity of the M. SssI high density methylated construct in the presence of Sp1, Sp3, and Ets1, abolishing the residual activity, whereas the promoter activity of the M. SssI low density methylated construct was reduced by approximately half (Fig. 6) (n ϭ 3, triplicate determinants). A substantial 5-fold reduction in promoter activity was observed in the M. HpaII-methylated construct, suggesting that the binding of MeCP-2 to methylated HpaII site(s) markedly decreased the ability of Sp1, Sp3, and Ets1 to synergistically transactivate the eNOS promoter. This indicates that unique HpaII site(s) are situated at critical activator regions in the eNOS promoter. An obvious candidate is the HpaII site within PRD II at Ϫ137, where these three trans-factors have been shown to form multiprotein complexes (5).
DNA Methylation Is Associated with Decreased Binding of Sp1, Sp3, and Ets1 to the eNOS Proximal Promoter-Since methylation of eNOS promoter/reporter constructs resulted in a marked reduction in the synergistic trans-activation by Sp1/ Sp3/Ets1, we hypothesized that methylation at the eNOS promoter may prevent the association of these trans-factors with native eNOS promoter sequences. Since these factors are thought to recruit RNA polymerase II to the eNOS promoter, RNA polymerase II recruitment would also be expected to be diminished in methylated cell types. To address this hypothesis, we used ChIP combined with real-time PCR to assess the binding of RNA Pol II, Sp1, Sp3, and Ets1 to the eNOS proximal promoter (Ϫ166 to Ϫ26). As shown in Fig. 7A, RNA Pol II was highly enriched at the proximal promoter of eNOS in HUVEC compared with HuAoVSMC. Likewise, Sp1, Sp3, and Ets-1 were relatively enriched at the eNOS promoter in HUVEC versus HuAoVSMC (Fig. 7, B-D). The differences in ChIP results for Sp1, Sp3 and Ets1 were not due to the absence of these trans-factors in vascular smooth muscle cells, as demonstrated by Western blotting (Fig. 7E).
Effects of DNA Demethylation on eNOS Gene Expression in Human Cell Lines-The preceding studies established a strong association between hypomethylation of the eNOS promoter and an increased level of gene expression. In order to define the molecular mechanisms implicated, hypomethylation of genomic DNA was induced by 5-azacytidine in endothelial (HUVEC) and nonendothelial cells (HuAoVSMC, JEG-3, HepG2, and HeLa) (5 M, 7 days, n ϭ 3). Bisulfite genomic sequencing confirmed that 5-azacytidine could lead to demethylation of the eNOS promoter in nonendothelial cell types (data not shown). Real time RT-PCR was used to assess the effects of 5-azacytidine treatment on steady-state eNOS mRNA expression. Shown in Fig. 8 is a representative experiment demonstrating that 5-azacytidine treatment increased eNOS mRNA levels in HuAoVSMC and HeLa. RNase protection assay also indicated that 5-azacytidine treatment increased eNOS mRNA levels in other nonendothelial cell types (JEG-3, HepG2, and HeLa), some from previously undetectable levels (data not shown). In contrast, the expression of eNOS steady-state mRNA levels in HUVEC was not increased by treatment with 5-azacytidine. Taken together, these results demonstrate that eNOS mRNA steady-state levels increase in nonendothelial cell lines following inhibition of DNA methyltransferase activity. DISCUSSION An intriguing aspect of eNOS gene expression is its relative restriction to the vascular endothelium. Murine eNOS promoter/␤-galactosidase (lacZ) reporter transgenic animals, reported by us and others, are known to recapitulate the known expression profile of eNOS mRNA and protein (36,39). Paradoxically, we report here that eNOS promoter/luciferase reporter episomal constructs are active in transient transfections of both endothelial and nonendothelial cell types, although comparable promoter fragments are endothelium-restricted in transgenic mice. We considered whether the cell-specific expression of eNOS is regulated, in part, through epigenetic mechanisms. Specifically, we proposed that eNOS 5Ј-flanking regions are differentially methylated in expressing and nonexpressing cell types and that this is an important mechanism operative in the cell-restricted expression of eNOS mRNA. Cell type-specific patterns of DNA methylation were proposed in the past as a potential mechanism implemented in cell-specific transcrip- FIG. 7. Binding of Sp1, Sp3, and Ets1 to the eNOS promoter in vivo. A-D, chromatin immunoprecipitation was used to assess the binding of RNA polymerase II, Sp1, Sp3, and Ets1, respectively, to the eNOS proximal promoter (Ϫ166 to Ϫ26) in HUVEC and HuAoVSMC. Results are presented as the number of copies of the eNOS promoter in immunoprecipitated samples divided by the number of copies present in parallel samples in the absence of antibody (no antibody control (NAC)). Shown is the mean Ϯ S.E. of at least three independent experiments. Results were similar if compared with control IgG and were similar using a second set of antibodies directed to Sp1, Sp3, and Ets1. *, p Ͻ 0.05, relative to HuAoVSMC. E, cellular protein from HUVEC and HuAoVSMC was assessed for the presence of Sp1, Sp3, and Ets1 by Western blotting. Sizes of protein bands are indicated to the right. Shown is a representative experiment of three. tional control pathways. Well characterized examples have only recently been provided. For example, CpG methylation was shown to be important in the cell-restricted expression of the maspin gene (SERPINB5) (48) as well as the human erythropoietin gene (49). Importantly, the relevance of this pathway in the cardiovascular system, especially vascular endothelium, has not been addressed.
In this study, we have demonstrated that DNA methylation plays an important role in the endothelium-specific expression of eNOS. Whereas the eNOS proximal promoter was either devoid or very lightly methylated in a wide range of human endothelial cell types, the promoter was heavily methylated in nonendothelial cells including multiple vascular smooth muscle cell types. Laser-capture microdissection was used to isolate endothelial and vascular smooth muscle cells from the mouse aorta and revealed a similar pattern of differential methylation at the eNOS proximal promoter. To our knowledge, this is the first reported application of laser capture microdissection to the study of cell-specific DNA methylation patterns in normal tissues. Given that cultured cell lines, human primary cell types, and laser capture microdissected murine cells demonstrated comparable findings, we infer that our observations of cell type-specific differentially methylated regions in the proximal promoter of the eNOS gene are relevant to both the in vitro and in vivo setting. Moreover, the differences in eNOS promoter methylation are conserved across species.
The differentially methylated region localized to the proximal promoter (Ϫ361 to the start site of transcription), a region that encompasses cis-DNA elements known to be necessary for eNOS transcription (e.g. PRD I and PRD II). Methylation of this region probably impedes transactivation of the gene given that the juxtaposition of a methylated patch of DNA and the transcription initiation site is functionally important (50). We found that genomic regions located ϳ4.9 kb upstream of the proximal promoter, which were suggested to represent an eNOS enhancer (27), were lightly methylated in all cell types. We failed to find cell type-specific methylation patterns in this upstream region. We have also begun to study genomic regions downstream of transcription initiation. The eNOS promoter is not located in a CpG island. However, a genomic region encompassing exon 24 to the start of the 3Ј-UTR is especially GC-rich and fits the criteria of a CpG island. 3Ј-CpG islands have been associated with varied tissue-specific genes (42). Our studies indicated that this region was methylated in both endothelial and nonendothelial cell types. Although CpG islands were historically believed to be free of methylation, exceptions are well known. For example, CpG islands can be methylated on the inactive X chromosome, at sites of repetitive DNA elements, at imprinted regions, and in the promoter regions of tumor suppressor genes in cancer (51). Of note, repetitive elements are known to be absent from this region of the eNOS gene (3). CpG islands can also be methylated within transcribed regions of genes (52), although the biological implications are not well understood. The functional relevance of methylation at this CpG island, located near the end of the eNOS gene, remains to be determined.
Genomic regions that are subject to differential methylation should be distinguished conceptually from those genomic regions that are functionally important in specifying epigenetic states, such as cell-specific methylation patterns. This is a newer area of investigation. To begin to address the basis for the differentially methylated region at the eNOS proximal promoter, we were interested in defining the methylation status of the promoter in eNOS/␤-galactosidase insertional transgenic mice. The promoter/reporter transgene Ϫ5200/ϩ28 Mu eNOS nls LacZ, numbered with respect to the transcription start site, contained 5200 bp of the murine eNOS promoter directing expression of the lacZ gene. We have previously reported that the expression pattern of the ␤-galactosidase reporter mirrored that of the endogenous eNOS mRNA. Interestingly, expression was uniform across multiple founders and was not significantly affected by the genomic integration site. This can be contrasted with the variability in expression observed across founders with eNOS reporter transgenes containing smaller 5Ј-flanking sequences (39,53). We report here that the methylation status of our transgene mirrors the methylation pattern observed in vascular endothelial cells. This is an unanticipated finding, given that insertional DNA events can be subject to epigenetic silencing (54). The identification of genomic regions that functionally account for eNOS promoter methylation, whether differentially methylated regions are regulated in cis or trans, and when methylation patterns are established during development is the focus of current research in our laboratory.
Methylation of eNOS promoter/reporter regions markedly impaired functional activity in mammalian cells. Using transient transfection studies, we noted that the density of DNA methylation and the specific sites of DNA methylation were critically important. To begin to define the mechanisms underlying the functional effects of DNA methylation on eNOS transcription, we used the Drosophila SL2 system. Sp1 is essential for eNOS transcription, and binding sites are present in both PRD I and II (5). DNA methylation of eNOS promoter/reporters in Drosophila cells co-transfected with Sp1 expression constructs revealed that Sp1-mediated transactivation was impaired, but only modestly. The lack of a marked effect is consistent with prior findings indicating that direct Sp1 binding is unaffected by m CG methylation but sensitive to m C m CG methylation (16,55). Because m C m CG sites are absent in this system, these results indicate that m CG methylation alters Sp1 transactivation perhaps via limiting accessibility of cis-regulatory regions through the condensation of chromatin or altering protein-protein interactions (17,56,57). The hypomethylation footprints identified at the high affinity Sp1 sites in PRD I in HepG2 and HeLa was of interest. It has been proposed that Sp1 helps maintain regions of DNA methylation-free (55,58). The basis for hypomethylation footprints is not clear but is gaining increased attention (59). The relevance of changes in Sp1 function to the regulation of eNOS transcription in mammalian cells warrants further study.
The modest effects of DNA methylation on Sp1-mediated eNOS transactivation can be contrasted with the markedly repressive effects of methylation on the transactivation observed with the combined addition of Sp1, Sp3, and Ets1 in SL2 cells. The extreme sensitivity of the Sp1/Sp3/Ets1 synergy to methylation-induced repression may be due to the requirement for precise stereo-specificity in the formation of highly structured nucleoprotein complexes (5). Thus, HpaII methylase, which methylates a CpG site in PRD II, where these factors are thought to bind cooperatively, greatly decreased transcriptional activity. The addition of MeCP-2 further suppressed the expression of methylated eNOS promoter/reporters in SL2 cells. Importantly, ChIP assays of the native eNOS proximal promoter in endothelial and vascular smooth muscle cells demonstrated that DNA methylation was associated with impaired binding of Sp1, Sp3, and Ets1. All of these factors were present in both cell types, arguing against the endothelium-specific expression of these factors. The promiscuous activity of episomal promoter/reporter constructs in cell types that do not express the native gene undoubtedly reflects both the presence and activity of these trans-factors in the nuclei of these cell types. Methylation per se would not be expected to completely abrogate the binding of these factors to the eNOS promoter.
Other mechanisms, such as the binding of methyl-binding proteins (e.g. MeCP-2) or alterations to chromatin structure (e.g. post-translational histone modifications), may also be important determinants of the ability of nucleoprotein complexes to form on the native promoter. Because RNA polymerase II recruitment was diminished at chromatin-based methylated eNOS sequences, we conclude that DNA methylation is responsible, in part, for the disruption of transcriptional initiation complex formation at the eNOS promoter. We found that eNOS expression could be induced upon demethylation of the eNOS promoter in a variety of nonendothelial cell types. Thus, promoter methylation was both inversely correlated and mechanistically relevant to the initiation of eNOS transcripts in nonendothelial cells.
Taken together, the results presented here indicate that DNA methylation plays an important role in both the transcriptional regulation and endothelial cell-specific expression of the human eNOS gene. The relevance of these findings to perturbations in eNOS expression in disease as well as the broader applicability of these findings to other endothelium cell-specific genes will need to be established in future studies. eNOS is the first constitutively expressed gene in the vascular endothelium whose expression has been shown to be regulated by DNA methylation. It is of great interest that the methylation status of genomic DNA of the human eNOS gene differs so markedly between cell types that are juxtaposed in the blood vessel wall, namely vascular endothelium and vascular smooth muscle cells. Understanding the fundamental principles that govern endothelial cell-specific gene expression is critical to the development of gene transfer techniques targeting exogenous genes to the vascular wall. The studies described may be relevant to our understanding of gene regulation in endothelial cells as well as in the cardiovascular system in general.