VASCULAR BIOLOGY A mechanistic role for DNA methylation in endothelial cell (EC)-enriched gene expression: relationship with DNA replication timing

Proximal promoter DNA methylation has been shown to be important for regulating gene expression. However, its relative contribution to the cell-specific expression of EC-enriched genes has not been defined. We used methyl-DNA immunoprecipitation (MeDIP) and bisulfite conversion to analyze the DNA methylation profile of EC-enriched genes in ECs versus non-expressing cell types, both in vitro and in vivo . We show that prototypic EC-enriched genes exhibit functional, differential patterns of DNA methylation in proximal promoter regions of most (eg. CD31, vWF, VE-cadherin and ICAM-2), but not all (eg. VEGFR-1 and VEGFR-2) EC-enriched genes. Comparable findings were evident in cultured ECs, human blood origin ECs and murine aortic ECs. Promoter-reporter episomal transfection assays for eNOS, VE-cadherin and vWF indicated functional promoter activity in cell types where the native gene was not active. Inhibition of DNA methyltransferase activity indicated important functional relevance. Importantly, profiling DNA replication timing patterns indicated that EC-enriched gene promoters with differentially methylated regions replicate early in S-phase in both expressing and non-expressing cell types. Chromatin-based mechanisms are critical for the transcriptional regulation of EC-enriched genes both in vitro and in vivo . Collectively, these studies highlight the functional importance of promoter DNA methylation in controlling vascular endothelial cell gene expression.


Introduction
The functional identity of an endothelial cell (EC) is dictated, in part, by its unique gene expression profile. The application of microarray profiling has reinforced the view that the concept of EC-enriched genes is valid and functionally relevant with respect to cellular phenotype [1][2][3] . In this regard, epigenetic processes are now appreciated to play a key role in regulating gene transcription. However, the relative contribution in ECs is not well understood 4,5 .
Decreased expression of constitutively active genes in ECs is a key component of endothelial cell dysfunction, such as is observed with endothelial nitric oxide synthase (eNOS) in atherosclerosis 6 . Defining whether perturbations of the DNA methylation status of key ECenriched genes contributes to changes in gene expression and cellular phenotype requires a firm understanding of DNA methylation profiles of these genes under normal conditions. We previously identified a differentially methylated region (DMR) at the proximal promoter of the eNOS/NOS3 gene 7 . While non-EC types showed high levels of methylation, a repressive mark associated with transcriptional silencing, ECs lacked DNA methylation in this region. These and other studies have suggested that epigenetics plays an important role in the regulation of gene expression in vascular ECs (reviewed in 8,9 ). DNA methylation also needs to be transmitted faithfully to nascent DNA subsequent to the replication of the genetic code. Generally, early timing of DNA replication in the cell cycle correlates with global gene expression 10 , though less is known about whether this paradigm applies to cell-restricted genes, especially within the vascular endothelium.
The contribution of DNA methylation to EC gene regulation remains to be fully explored.
We therefore wished to determine whether the epigenetic mechanisms first characterized for eNOS in ECs is a unique feature for eNOS expression, or is also applicable to a repertoire of EC-enriched genes. Discerning the epigenetic state of unique cell types is a key goal of the International Human Epigenomic Consortium (IHEC). Similarly, the Encyclopedia of DNA For personal use only. on August 29, 2017. by guest www.bloodjournal.org From Elements (ENCODE) project also aims to delineate functional elements and chromatin signatures for specific cell types 11 . These genome-based approaches are, using currently existing methodologies, examining these changes in low-resolution. High-resolution data is especially needed. In this present study, we describe both in vitro and in vivo studies of differential DNA methylation in the proximal promoter regions of the EC-enriched genes CD31/PECAM1, Endoglin/ENG, ICAM-2, P-selectin/SELP, Tie-2/TEK, VE-Cadherin/CDH5 and vWF in terminally differentiated ECs versus non-ECs, in both humans and mice, and provide evidence that epigenetic modifications are functionally important for EC gene expression. Furthermore, DNA replication timing studies show that EC-enriched genes with DMRs replicate in early S-phase in both ECs and non-ECs. This was an exciting, yet unexpected finding. The aim of the study was to determine the functional relevance of promoter DNA methylation for EC gene expression. Here we describe important features regulating EC-enriched gene expression involving both chromatin-based and cell cycle pathways.

Methods
All animal studies were performed in accordance with the guidelines of the Canadian Council on Animal Care and were approved by the University of Toronto Animal Care Committee. This study was conducted in accordance with the Declaration of Helsinki.
Isolation and culturing of BOECs was conducted as previously reported and assayed at passage 3-5, after 6-8 weeks in culture.  11 and is available on the University of California, Santa Cruz Genome Browser. RepeatMasker data was generated by Arian Smit's RepeatMasker program (http://www.repeatmasker.org).

BrdU Labeling and Flow Activated Cell Sorting (FACS)
Asynchronous populations of HUVEC and HuAoVSMC were pulse labeled with 5-bromo-2'deoxyuridine (BrdU) as previously described 15 . Briefly, BrdU (Sigma) was added to cells (50 μ M) and incubated for 1 h at 37°C. Equal numbers of cells (50,000) were sorted into cell cycle fractions based on DNA content (G1, S1, S2, S3, S4, G2). Sorted cells were collected directly into lysis buffer and DNA was extracted using the DNeasy Blood and Tissue Kit (Qiagen). 20 μ g of salmon sperm DNA (Invitrogen) and 100,000 copies of BrdU-labeled E. coli DNA was added to each fraction prior to extraction to control for DNA recovery and immunoprecipitation efficiency. DNA was eluted in 500 μ L TE. BrdU-labeled cells and extracted DNA was manipulated in the dark to prevent photolysis of BrdU-incorporated DNA.
For personal use only. on August 29, 2017. by guest www.bloodjournal.org From

Sodium Bisulfite Genomic Sequencing and Pyrosequencing
Genomic DNA from human cultured cells (5 μg) was subjected to sodium bisulfite treatment as described previously 7,16,17 . 500 ng of murine genomic DNA was bisulfite converted using EZ DNA Methylation-Direct™ kit (Zymo Research, Irvine, CA) and subjected to nested PCR amplification (Table S3). For pyrosequencing analysis, 10 pmol of primers per reaction and a biotinylated reverse primer was used (Tables S2-S3, EpigenDx Inc.).

Statistics
Unless otherwise stated, all experiments were performed a minimum of three times, and data represent the mean ± S.E.M. Statistical analyses were performed using a Student's t test or analysis of variance, as appropriate. A p value less than 0.05 was considered to be statistically significant.

Types: Low-Resolution and High-Resolution Mapping
We first analyzed promoter DNA methylation for key EC-enriched genes using data generated by the ENCODE consortium 11 . We utilized Methyl 450K Bead Array data to define regions of methylation at proximal promoters of key EC-enriched genes in ECs and non-ECs.
The region surrounding the start of transcription displayed no methylation in HUVEC, while the same region was partially methylated or methylated in AoSMC or hepatocytes, for the genes NOS3/eNOS and CDH5/VE-Cadherin ( Figure 1). Regions of differential methylation were not We used high resolution assays to examine DNA methylation of promoter regions around the TSS of EC-enriched genes using methyl-DNA immunoprecipitation (MeDIP), followed by qPCR, a robust method for assessing DNA methylation at specific loci 18 . The genomic regions analyzed and citations used to determine the TSS are summarized in Table S4  Using gold standard, high-resolution sodium bisulfite sequencing and single strand DNA analysis or quantitative pyrosequencing in EC and non-EC types, we confirmed our screen of promoter methylation of EC-enriched genes. Consistent with MeDIP assays, bisulfite analyses identified robust differences in DNA methylation between HUVEC and HuAoVSMC, at proximal promoter regions of EC-enriched genes. Pyrosequencing analyses for CD31 in the EC types HUVEC, HMVEC and BOEC, and in non-EC types HuAoVSMC, human saphenous vein vascular smooth muscle cells (HuSVVSMC), keratinocytes and hepatocytes was conducted. CD31 displayed very low levels of DNA methylation in all EC types, while high levels of methylation were observed in non-ECs ( Figure 3A). The findings from quantitative pyrosequencing assay are confirmed by single strand analysis results ( Figure S2).
Genomic regions corresponding to exon 1 of vWF are known to be functionally important.
The 4 CpG sites downstream of the vWF TSS in exon 1 were assessed and low levels of methylation in HUVEC and HMVEC were found, in comparison to dense methylation in the non-EC HuAoVSMC, HuSVVSMC, keratinocytes and hepatocytes ( Figure 3B). The eNOS promoter also showed dense methylation in both hepatocytes and keratinocytes ( Figure S3A).
Methylation patterns in the ICAM2 promoter was similar to other EC-enriched genes, as HUVEC For personal use only. on August 29, 2017. by guest www.bloodjournal.org From were 0% methylated and HuAoVSMC were densely methylated, apart from two CpG sites at -24 and -12 ( Figure S3C). The EC-enriched genes VE-Cadherin and Tie2 showed no methylation in HUVEC, whereas high levels of DNA methylation were observed in HuAoVSMC ( Figure 3C-D).
Differential methylation of VE-Cadherin was restricted to a region downstream of the TSS, with methylation in HuAoVSMC observed at positions +103 to +229 ( Figure 3C). Though there were low levels of DNA methylation seen in the proximal promoter of the Tie2 gene, clear differences in DNA methylation across cell types was evident ( Figure 3D).
In contrast to the other EC-enriched genes examined where genomic regions surrounding the TSS evidenced no DNA methylation, P-selectin displayed low but evident levels of methylation in HUVEC. In contrast, dense methylation was seen in HuAoVSMC ( Figure S3D).

Murine EC-enriched Genes are Differentially Methylated in EC and non-EC Types
We examined murine promoters in vivo, namely the EC-enriched genes eNOS, CD31, VEcadherin and vWF, in descending thoracic aortic EC and descending thoracic AoVSMC. We used pyrosequencing analysis of bisulfite-converted DNA to assess the methylation status and found that eNOS, VE-Cadherin, CD31 and vWF gene promoters were hypomethylated in EC and heavily methylated in AoVSMC (Figure 4), consistent with our findings in human ECs and non-ECs. Therefore, the DMRs of EC-enriched genes identified in cultured human cells are in agreement with our in vivo methylation analyses in the mouse.

EC-enriched Gene Promoters are Not Methylated in Human Blood Outgrowth ECs
We wished to determine the methylation in a cell type capable of initiating stable outgrowth ECs. Blood outgrowth endothelial cells (BOEC) are an EC type, derived from cultured peripheral blood and are capable of establishing mature outgrowth ECs 14 . We used well-validated methods to isolate and characterize these cells, as previously described 14 . Since eNOS, as well as the EC-enriched genes CD31 and vWF, are fundamental to the proper functioning of BOEC, we were interested in determining the DNA methylation status of these EC genes along this EC differentiation cascade. Using quantitative pyrosequencing of bisulfite-converted DNA from blood outgrowth EC, the promoters of eNOS, vWF and CD31 were analyzed. All EC-enriched genes displayed hypomethylation in BOEC ( Figure 3A-B and Figure S3A,C-D), displaying a methylation pattern similar to mature ECs.

Promoter Activity of EC-enriched Genes in Non-EC Types
We have previously demonstrated robust expression of episomal eNOS promoter/reporter constructs in cell types in which eNOS is not normally expressed 7 . Surprisingly, these same genomic regions exhibited exquisite cell-specificity when stably integrated into the genome in insertional transgene murine promoter/reporter studies 19,20 . This implied that eNOS episomal vectors or naked DNA did not faithfully recapitulate expression of the native gene, which highlighted functional relevance of chromatin-based mechanisms. We used episomal constructs containing the regions -487/+247 and -1912/+1, of human vWF and VE-cadherin, respectively, given that the same regions are very EC-enriched in insertional promoter transgenes 21,22 . These episomal constructs were transiently transfected into ECs and non-EC types that do, or do not express eNOS, vWF or VE-cadherin, respectively (Table 1; Figure S1). These data indicate that VE-cadherin and vWF promoter reporter espisomal constructs are transcriptionally active even when the native chromatin-based genes are not, as we previously noted for eNOS 7 .

Functional Role of Methylation in Repressing EC-enriched Genes in non-EC Types
The presence of DNA methylation at proximal promoter regions of these EC-enriched genes suggested that this chromatin mark may silence the expression of these genes in non-

VEGFR-1 and VEGFR-2 Promoters Are Not Differentially Methylated in ECs Versus Non-EC
Types Surprisingly, not all EC-enriched genes examined were differentially methylated between HUVEC and HuAoVSMC. The promoter region of VEGFR2 did not display an enrichment of methylation in HuAoVSMC, as determined by MeDIP analysis ( Figure 2E). The absence of proximal promoter methylation was confirmed by examining Methyl 450K Bead Array data at the promoters of VEGFR1 and VEGFR2 in ECs and non-ECs, in addition to single strand DNA analysis and pyrosequencing methods ( Figure S4A-D). Bisulfite analysis demonstrated that the VEGFR2 promoter was also not methylated in keratinocytes or hepatocytes ( Figure S4D). Importantly, VEGFR2 did not show any changes in expression when treated either with 5azacytidine alone, or in combination with TSA ( Figure 5E). These findings suggest that DNA methylation is not important for the transcriptional regulation of VEGFR1 and VEGFR2.

eNOS is Not Hydroxymethylated in ECs and non-ECs
The recent discovery of 5-hydroxymethylcytosine (5hmC) as a modified base in mammalian DNA, catalyzed by the TET family of proteins, has led to newer insight into the epigenetic regulation of genes 26 . The modified cytosine species 5hmC and 5-methylcytosine (5mC) are indistinguishable upon sodium bisulfite conversion and importantly, DNA containing 5hmC is not efficiently amplified by PCR following bisulfite conversion 27 . As such, using an antibody directed against 5-hydroxymethylcytosine in a technique similar to MeDIP, or hydroxymethyl-DNA immunoprecipitation (OH-MeDIP), we analyzed the promoter region of the EC-enriched genes eNOS, CD31, VE-cadherin and vWF in HUVEC ( Figure S5). In contrast to robust levels of methylation at the promoter of eNOS in HuAoVSMC, we failed to detect appreciable steadystate levels of hydroxymethylation at the eNOS promoter in this same cell type ( Figure S5A).
The same was true for the promoters of CD31, VE-cadherin and vWF ( Figure S5B-D). We can therefore be confident the observed bisulfite findings addresses 5-methylcytosine and not 5hydroxymethylcytosine.

Differentially Methylated EC-enriched Genes Replicate in Early S-phase in ECs and non-ECs
Though DNMT1 is thought to localize to the replication fork during S-phase 28 , how methylation patterns on newly replicated DNA at cell-specific genes is poorly understood. It is generally accepted that transcriptionally active genes replicate early, while inactive genes replicate late, during S-phase of the cell cycle. This principle was reaffirmed in recent whole genome analyses 29   To investigate the DNA replicating timing patterns of EC-enriched genes, we used a BrdUpulse labeling and qPCR-based approach 29 . Cyclophilin A, a housekeeping gene expressed in both ECs and non-ECs, exhibited early S-phase replication, while β -globin, which is expressed specifically in erythrocytes and not transcriptionally active in ECs and non-ECs, exhibited late Sphase replication ( Figure S6C-D), as shown by others 30  Again, these findings were unexpected, since generally, early DNA replication timing correlates with gene expression 10 . Interestingly, the EC-enriched genes VEGFR-1 and VEGFR-2 displayed early S-phase (S1) replication in HUVEC, and late S-phase (S2-S3) replication in a non-expressing cell type ( Figure 6F-G). As stated, these were the only EC-enriched genes that did not display differential DNA methylation between ECs and non-ECs ( Figure S4).

The Newly Replicated eNOS Promoter has Low Levels of DNA Methylation in Early S-phase
Since we determined that EC-enriched genes replicate in early S-phase in ECs and non-ECs, we were motivated to determine when in S-phase the promoter region of eNOS becomes methylated in HuAoVSMC. Site-specific methylation analysis at the proximal promoter of eNOS displayed an average 40% methylation in G1 and S1 phases ( Figure S7A). In S2, methylation levels were observed to be double that of S1. For the remainder of S-phase, methylation levels at eNOS were comparable to the methylation level of a whole dish of cells ( Figure S7A). The levels of eNOS methylation that is observed in terminally differentiated cells ( Figure S3A) does not occur until late S-phase. These results indicate that while the replication timing of eNOS For personal use only. on August 29, 2017. by guest www.bloodjournal.org From occurs in S1 phase, there is a measurable lag in re-methylation of the nascent strand of the hemimethylated DNA duplex. To address a possible cause or effect relationship, we treated AoVSMC with 5-azacytidine for 24 hr, and profiled the timing of eNOS replication. It was evident that the timing of eNOS replication remained in early S-phase ( Figure S7B). Thus, inhibition of DNA methylation at the eNOS promoter did not affect the replication timing of eNOS. Not all cells in S-phase are dynamically proceeding through DNA replication. We then asked whether AoVSMC arrested in S-phase have a distinct DNA methylation profile. As shown in Figure S8B, the eNOS DNA methylation profile did not differ between cells proceeding or arrested in Sphase.

Discussion
This report defines that the majority of endothelial cell (EC)-enriched genes show evidence of differential DNA methylation of proximal promoter regions. These genomic regions, which encompass the start sites of transcription, are unmethylated in endothelial cells that express these mRNAs, and are methylated in non-expressing cell types, such as HuAoVSMC, keratinocytes and hepatocytes. We identified differentially methylated regions (DMRs) in the EC-enriched genes PECAM1/CD31, vWF, VE-cadherin, ICAM-2, P-selectin, Endoglin and Tie2. Importantly, our in vivo studies of murine EC-enriched gene promoters recapitulated our findings. Use of DNA methylation inhibitors indicated that DNA methylation is functionally relevant. Inhibition of global DNA methylation, either alone or combined with inhibitors of HDAC activity, led to increases in expression of select mRNAs in cell types that do not basally express these mRNAs. We infer that DNA methylation actively represses transcription of these transcripts in non-expressing cell types. This is an important concept given that most models of We find that for the majority of EC-enriched genes that contain DMRs in their proximal promoter, DNA replication timing occurs in early S-phase in both ECs and non-ECs. This finding is not consistent with the general paradigm of replication timing patterns and gene expression. Importantly, the majority of studies are based on global approaches of replication timing analyses, and focus on developmentally controlled and tissue-specific genes that show differential DNA replication timing profiles between cell types 31,32 . We examined publicly available data on replication timing in embryonic stem (ES) cells, neural progenitor cells and lymphoblasts, and noted early replication timing for eNOS 33 (www.replicationdomain.org). We found that eNOS is replicated early in the cell cycle and there is a small, but discernible, delay in We demonstrated that VEGFR1 and VEGFR2 did not contain DMRs in their proximal promoters. In disease settings, these genomic regions can be methylated. For instance, the For personal use only. on August 29, 2017. by guest www.bloodjournal.org From VEGFR2 promoter and exon 1 was reported to be methylated in prostate cancer and upon ischemic injury 34,35 . Interestingly, VEGFR1 and VEGFR2 are among the earliest EC-enriched genes to be expressed in vascular development. Expression of VEGFR1 and VEGFR2 is first detected at ED6.5 36  We found that VEGFR1 and VEGFR2 replicated early in expressing cell types and late in non-expressing cell types ( Figure 6). These two EC-enriched genes also did not contain differentially methylated regions in their proximal promoter regions. Studies have shown that genes with high absolute CpG-density at the promoter replicate early in mammalian cells 40

. High
CpG-density regions can also represent CpG islands if the observed/expected ration is high. In this regard, the promoter regions of VEGFR1 and VEGFR2 represent CpG islands.
Interestingly, when the areas around the TSS of all EC-enriched genes were analyzed for the presence of CpG islands, only VEGFR1 and VEGFR2 showed an observed/expected ratio of a strong CpG island (>0.75), whereas the other EC-enriched genes were classified as poor CpG islands (Table S4). The relationship between the presence or absence of DNA methylation and timing of onset of gene expression in development, DNA replication timing in the mitotic cell cycle and the presence of CpG islands requires further study.
We assayed the methylation status of EC-enriched gene promoters in blood outgrowth endothelial cells and found comparable findings to results in HUVEC and HMVEC. This is an important finding, as BOEC have proved to be valuable for several therapeutic applications, including the ability to home to sites within the tumour neovasculature for gene delivery 41,42 .
Interestingly, recent studies by others have shown that in early proangiogenic cells, the eNOS        HMVEC (yellow) and BOEC (red) are compared to non-EC types AoVSMC (light green), SVVSMC (dark green), keratinocytes (blue) and hepatocytes (purple). BOEC exhibit a cobblestone-shape, have high proliferative capacity, take up acetylated low density lipoprotein and are uniformly positive for several endothelial markers 14 . BOEC are expanded from small numbers of cells after long-term culture, and are distinct from the early outgrowth colonies obtained after 4-7 days in culture, as reviewed by others 47,48 . Quantitative pyrosequencing (A and B) and single strand analysis (n ≥15) (C and D) of bisulfite-converted DNA was used to assess methylation. Extensive mixing studies of in vitro methylated or mock-methylated templates revealed that pyrosequencing a PCR product cannot distinguish 0% from 5% methylation, or 100% from 95% methylation. Single strand plasmid clone analysis indicates that CpG sites are not methylated in ECs for these genes.