Loss of Methyl-CpG–Binding Domain Protein 2 Enhances Endothelial Angiogenesis and Protects Mice Against Hind-Limb Ischemic Injury

Abstract

Background

Despite intensive investigation, how DNA methylation influences endothelial function remains poorly understood. We used methyl-CpG–binding domain protein 2 (MBD2), an interpreter for DNA methylome–encoded information, to dissect the impact of DNA methylation on endothelial function in both physiological and pathophysiological states.

Methods and Results

Human umbilical vein endothelial cells under normal conditions express moderate levels of MBD2, but knockdown of MBD2 by siRNA significantly enhanced angiogenesis and provided protection against H₂O₂-induced apoptosis. Remarkably, Mbd2^−/− mice were protected against hind-limb ischemia evidenced by the significant improvement in perfusion recovery, along with increased capillary and arteriole formation. Loss of MBD2 activated endothelial survival and proangiogenic signals downstream of vascular endothelial growth factor signaling characterized by an increase in endothelial nitric oxide synthase (eNOS) and vascular endothelial growth factor receptor 2 expression, along with enhanced extracellular signal-regulated kinase 1/2 activation and BCL-2 expression. Mechanistic studies confirmed the methylation of CpG elements in the eNOS and vascular endothelial growth factor receptor 2 promoter. MBD2 binds to these methylated CpG elements and suppresses eNOS promoter activity. On ischemic insult, key endothelial genes such as eNOS and vascular endothelial growth factor receptor 2 undergo a DNA methylation turnover, and MBD2 interprets the changes of DNA methylation to suppress their expressions. Moreover, MBD2 modulation of eNOS expression is likely confined to endothelial cells because nonendothelial cells such as splenocytes fail to express eNOS after loss of MBD2.

Conclusions

We provided direct evidence supporting that DNA methylation regulates endothelial function, which forms the molecular basis for understanding how environmental insults (epigenetic factor) affect the genome to modify disease susceptibility. Because MBD2 itself does not affect the methylation of DNA and is dispensable for normal physiology in mice, it could be a viable epigenetic target for modulating endothelial function in disease states.

Endothelial cells (ECs) are the initial line of defense in many vascular diseases, and frequently the first casualty. Therefore, altered endothelial function is associated with numerous human pathologies, such as atherosclerosis, allograft vasculopathy, heart failure, diabetic retinopathy, and scleroderma. Given the essential role that ECs played in angiogenesis, their functional impairment is associated with attenuated angiogenic response that features prominently in patients with poor clinical outcomes after peripheral and/or myocardial ischemia,^1,2 and an epigenetic factor is implicated in the disease process.³

As a major epigenetic mechanism, DNA methylation acts as a footprint, reflecting the consequences of environmental insults in different types of cells.⁴ Therefore, alterations in DNA methylome are commonly seen in animals and patients with vascular diseases,^5,6 which are read by a conserved family of methyl-CpG–binding domain (MBD) proteins (ie, MBD1, MBD2, MBD3, MBD4, and MeCP2).^7-9 The MBD proteins actively implicate in DNA methylation–mediated transcriptional repression and/or heterochromatin formation and are responsible for maintaining and interacting with DNA methylome. They are distinguished by the binding affinity to methylated DNA, and MBD2 binds with the highest affinity, but MBD3 fails to selectively recognize methylated DNA.⁸ Animals deficient in MeCP2, MBD1, MBD2, and MBD4 are viable; deficiency in MBD3 leads to embryonic lethality.⁹ Mice lacking MeCP2 or MBD1 exhibit specific neurological defects,^10,11 whereas loss of MBD4 suppresses CpG mutability and tumorigenesis.^12,13 In contrast, mice deficient in MBD2 are generally normal except for a minor phenotype in maternal behavior.¹⁴

In the present study, we used MBD2 as a model to dissect the effect of DNA methylation on endothelial function. We hypothesized that MBD2 deciphers DNA methylome–encoded information and modulates endothelial gene expression implicated in endothelial apoptosis and angiogenesis. We demonstrated evidence supporting the methylation of CpG elements in the endothelial nitric oxide synthase (eNOS) and vascular endothelial growth factor (VEGF) receptor 2 (VEGF-R2) promoter. We also verified that MBD2 interprets DNA methylome–encoded information by binding to the methylated CpG elements. As a result, knockdown of MBD2 significantly enhanced angiogenesis and protected ECs from H₂O₂-induced apoptosis. Consistently, Mbd2^−/− mice were protected from hind-limb ischemic injury.

Methods

Animals

Mice deficient in MBD2 (Mbd2^−/−) on a C57BL/6 background were kindly provided by Dr Adrian Bird (Edinburgh University, Edinburgh, UK).¹⁴ Eight-week-old male Mbd2^−/− mice and their wild-type (WT) littermates were used in the study. All mice were housed in a standard specific pathogen free (SPF) facility in microisolator cages. All studies were performed in compliance with the Georgia Health Sciences University and Tongji Medical College Animal Care and Use Committee guidelines.

Cell Culture and Transfection

Human umbilical vein ECs (HUVECs) were cultured in EBM-2 EC basal medium with the SingleQuot kit (Lonza/Cambrex, Walkersville, MD). For siRNA transfection, the cells were seeded in 12-well plates at a density of 4×10⁴ cells per well. On the second day, 50 or 100 nmol/L control or MBD2 siRNA was transfected into the cells with the lipofectamine reagent (Invitrogen, Carlsbad, CA).

Flow Cytometry Analysis of Endothelial Apoptosis

ECs were treated with 0.2 mmol/L H₂O₂ for 24 hours to induce apoptosis, followed by flow cytometric analysis of apoptotic cells.

In Vitro Angiogenesis Assay

Twenty-four hours after transfection, the ECs (2×10⁴ cells) were added into each well of a 96-well plate and cultured for additional 24 hours. Endothelial tubes were then examined under a light microscope every other 4 hours by inspection of the overall tube length and branch points. ³H-thymidine incorporation was used for analysis of endothelial proliferation and presented as counts per minute.

Hind-Limb Ischemic Model and Assessments

After anesthetization, a 1-cm-long incision was made in the skin at the medial thigh to expose the femoral artery. The proximal end of the femoral artery was occluded with double knots. The segment of femoral artery between the distal and proximal knots was then transected. Hind-limb blood flow in mice that had undergone hind-limb ischemia was measured with a laser Doppler imaging system (Perimed, Stockholm, Sweden).

Immunostaining of Methyl-CpG–Binding Domain Protein 2, Capillaries, and Arterioles

A sheep anti-MBD2 antibody and a FITC-conjugated bovine antisheep IgG were used to detect MBD2 expression. Staining of CD31-positive cells was used to define capillary formation, and α-actin–positive cells were used to evaluate arteriole formation.

Western Blotting

An enhanced chemiluminescence Western blotting kit (Millipore, Temecula, CA) was used for analyses of total and phosphorylated protein levels.

Chromatin Immunoprecipitation Assay, Electrophoretic Mobility Shift Assay, and Endothelial Nitric Oxide Synthase Promoter Reporter Assay

Chromatin immunoprecipitation (ChIP) assay was carried out with a ChIP assay kit (Upstate Biotechnology, Lake Placid, NY). The primers used in ChIP assay are summarized in Table I of the online-only Data Supplement. Electrophoretic mobility shift assay was carried out with a LightShift chemiluminescent electrophoretic mobility shift assay kit (Thermo Scientific, Rockford, IL), and a dual luciferase reporter system (Promaga, Madison, WI) was used for eNOS promoter reporter assays.

DNA Bisulfite Sequencing Analysis

Genomic DNA from each preparation was first undergone bisulfite conversion with an EZ DNA Methylation kit, followed by polymerase chain reaction amplification of targeted sequences. The resulting polymerase chain reaction products were directly cloned into a TA vector. The methylation state of each targeted sequence was then analyzed by DNA sequencing.

Statistical Analysis

For pairwise comparisons, the data were analyzed by use of a Student t test. Comparison between multiple experimental groups was accomplished by a Bonferroni test with SPSS 17.0 for Windows. For the blood flow imaging data, repeated measures ANOVA was used. A 3-way repeated measures ANOVA was used to analyze the blood flow of Mbd2^−/− and WT mice with or without N^G-nitro-l-arginine methyl ester (L-NAME) treatment. All data are presented as mean±SEM. In all cases, values of P<0.05 were considered statistically significant.

A detailed version of Methods is available in the online-only Data Supplement.

Results

Knockdown of Methyl-CpG–Binding Domain Protein 2 Promotes Angiogenesis and Protects Endothelial Cells From H₂O₂-Induced Apoptosis

The HUVECs were transfected with either an MBD2 or a control siRNA to examine the potential influence of MBD2 on endothelial function. Moderate levels of MBD2 were detected in HUVECs under control conditions, whereas the MBD2 siRNA dose dependently suppressed MBD2 expression. When HUVECs were transfected with 100 nmol/L MBD2 siRNA, MBD2 was almost undetectable (Figure 1A). We next seeded siRNA-transfected HUVECs into culture plates preconditioned with growth factor–reduced Matrigel (BD Bioscience). Knock-down of MBD2 expression in HUVECs led to a significantly enhanced capacity for tube formation (Figure 1B). Tube formation was observed as early as 4 hours after seeding of HUVECs lacking MBD2, whereas almost no discernable tube formation was observed for control siRNA-transfected cells (data not shown). The average total tube length 24 hours after transfection in each randomly selected field was significantly higher in MBD2 siRNA-transfected cells than control siRNA-transfected cells (Figure 1B).

Figure 1.

Inhibition of methyl-CpG–binding domain protein 2 (MBD2) promotes angiogenesis and protects human umbilical vein endothelial cells (HUVECs) against H₂O₂-induced apoptosis. A, MBD2 siRNA dose dependently suppressed MBD2 expression in HUVECs. The MBD2 levels in siRNA-transfected HUVECs were determined by Western blotting. A 70% reduction was observed when HUVECs were transfected with 100 nmol/L MBD2 siRNA (n=4). B, Knockdown of MBD2 promoted tube formation. The HUVEC tube formation was examined under a light microscope every other 4 hours, and pictures were taken 24 hours after culture. Left, Representative images for tube formation. Right, Average tube length (n=4). Tube length was calculated by measuring the cumulative tube length in 5 randomly selected microscopic fields for each experiment with the Axiovision LE software. C, Inhibition of MBD2 enhanced proliferation. Proliferation of HUVECs was assayed by ³H-labeled thymidine incorporation and presented as mean±SEM (n=3). D, Knockdown of MBD2 protected HUVECs from H₂O₂-induced apoptosis. Twenty-four hours after siRNA transfection, HUVECs were treated with 0.2 mmol/L H₂O₂ to induce apoptosis. Left, Representative flow cytometric data. Right, A graph showing the results from 4 independent experiments. Ctl indicates control. **P<0.001; *P<0.01.

To examine the impact of MBD2 on HUVEC proliferation, siRNA-transfected cells were cultured in the presence of ³H-labeled thymidine, followed by measurement of the in-corporated radioactivity. Knockdown of MBD2 significantly enhanced HUVEC proliferation compared with control siRNA-transfected cells (Figure 1C). To investigate whether MBD2 regulates endothelial apoptosis, the transfected HUVECs were treated with H₂O₂ (0.2 mmol/L) for 24 hours to induce apoptosis. Interestingly, knockdown of MBD2 significantly protected HUVECs from H₂O₂-induced apoptosis (Figure 1D). Together, these results suggest that MBD2 negatively regulates angiogenesis and sensitizes ECs to H₂O₂-induced apoptosis.

Loss of Methyl-CpG–Binding Domain Protein 2 Protects Mice Against Hind-Limb Ischemia

Individuals with cardiovascular disease have an attenuated angiogenic response to ischemia that is associated with poor clinical outcomes.^15,16 Given the observation that MBD2 negatively regulates endothelial angiogenesis in cultured cells, we next examined its impact on perfusion recovery after ischemic injury in vivo. We first confirmed the absence of MBD2 protein in blood vessels of Mbd2^−/− mice (Figure 2A). We next examined MBD2 temporal expression changes in WT mice after femoral artery excision. Under control conditions, MBD2 was very low in the hind limb. However, a steady increase in MBD2 was noticed after ischemic induction; the highest level was observed 4 days after ischemic surgery. After that point, MBD2 underwent a steady decrease and then returned to a relative low level after postsurgery day 14 (Figure 2B). Immunostaining revealed that MBD2 expression (green) in the nonischemic muscle sections was limited to the location of blood vessels and was localized mainly to ECs (CD31 positive; red; Figure 2C, top). In line with vascular neogenesis in response to ischemic insult, significantly more ECs (CD31-positive cells) were found, along with much higher levels of MBD2 expression, in the ischemic sections (Figure 2C, bottom), suggesting that ECs undergoing vascular neogenesis are associated with enhanced MBD2 expression.

Figure 2.

Temporal expression analysis of methyl-CpG–binding domain protein 2 (MBD2) after hind-limb ischemic injury. A, MBD2 was absent in the blood vessels of Mbd2^−/− mice. B, Temporal expression changes of MBD2 in the hind limb after ischemic insult. MBD2 expression was examined by Western blotting with hind-limb lysates (left), and the relative amount of MBD2 was estimated by the ratio to β-actin (right; n=5). C, Coimmunostaining of ECs and MBD2 in the ischemic and nonischemic hind-limb sections. Nuclei were stained with DAPI (a and e; blue); endothelial cells (ECs) were stained with anti-CD31 (b and f; red); and MBD2 was stained green (c and g). Merged images (d and h) were generated by overlay of DAPI, anti-CD31, and MBD2 staining. Inset images with higher resolution in d and h show the colocalization of MBD2 and ECs. The images (×200) are representative of consistent results obtained from 4 mice analyzed in each group.

To determine the role of MBD2 in perfusion recovery, unilateral hind-limb ischemia was induced in Mbd2^−/− and WT male mice. All mice survived the surgical procedure, and blood flow in the ischemic (left side) and nonischemic (right side) limbs was monitored with a laser Doppler imaging system. Blood flow in the ischemic hind limbs was almost undetectable in both Mbd2^−/− and WT mice right after femoral artery excision. Remarkably, Mbd2^−/− mice showed partial blood flow restoration 2 days after induction, and blood flow was restored to ≈50% at day 7 and almost completely recovered at day 14 after the ischemic insult (Figure 3A). However, no visible blood flow restoration was observed in WT mice until day 4, and blood flow was restored to only ≈30% and 60% at days 7 and 14 after ischemic surgery, respectively.

Figure 3.

Mbd2^−/− mice show a significantly improved perfusion recovery after hind-limb ischemia. A, Blood flow restoration after hind-limb ischemia. Blood flow was measured with a laser Doppler imaging system and presented as changes in the laser frequency manifested by different color pixels (left). The mean hind-limb blood flow was calculated as the ratio of ischemic (left) side to nonischemic (right) side (right; n=6). B, Mbd2^−/− mice showed significant higher capillary formation at day 7 of ischemic injury. The amount of endothelial cells (CD31-positive; red) in hind-limb sections (×200) served as an indicator of capillary formation (left). The average capillary density for each group was shown in a bar graph (right; n=5). C, Arteriole formation at day 7 of ischemic injury. Similarly, the amount of smooth muscle cells (α-actin positive; red) in hind-limb sections (×200) was used as an indicator for arteriole formation (left), and the mean arteriole density was shown in a bar graph (right; n=5). **P<0.001; *P<0.01.

To investigate differences in neovascularization, we examined capillary (CD31-positive cells) and arteriole (α-actin-positive cells) formation. The numbers of CD31-positive cells (ECs; Figure 3B) and α-actin positive cells (smooth muscle cells; Figure 3C) were significantly higher in the ischemic sections of Mbd2^−/− mice than in WT mice. Taken together, our results demonstrate that loss of MBD2 promoted both angiogenesis and arteriogenesis and was associated with significantly improved perfusion after hind-limb ischemia.

Suppression of Methyl-CpG–Binding Domain Protein 2 Activates Endothelial Cell Survival and Proangiogenic Signals

To dissect the mechanisms by which MBD2 regulates endothelial survival, we examined 2 major endothelial survival signals, the extracellular signal-regulated kinase 1/2 (ERK1/2) and the serine-threonine kinase Akt, in siRNA-transfected HUVECs. Interestingly, MBD2 siRNA did not show a significant impact on Akt expression and activation (Figure I in the online-only Data Supplement). However, MBD2 siRNA promoted ERK1/2 activation, as revealed by the presence of higher phosphorylated (p-) ERK1/2, although total ERK1/2 did not change (Figure 4A). In line with the role of p-ERK1/2 in BCL-2 stabilization,¹⁷ higher BCL-2 expression was noticed in MBD2 siRNA-transfected cells on H₂O₂ treatment (Figure 4B). Surprisingly, no perceptible difference was detected after knockdown of MBD2 for Mn superoxide dismutase and Cu/Zn superoxide dismutase, the 2 enzymes detoxifying reactive oxygen species (Figure II in the online-only Data Supplement). Our results suggest that MBD2 regulates endothelial apoptosis probably by modulating survival signals relevant to ERK1/2 and BCL-2 signaling.

Figure 4.

Loss of methyl-CpG–binding domain protein 2 (MBD2) activates endothelial survival and proangiogenic signals. A, Knock-down of MBD2 enhanced extracellular signal-regulated kinase 1/2 (ERK1/2) activity. Left, A representative Western blot results for total and phosphorylated ERK1/2 (p-ERK1/2). Right, Quantitative analysis of p-ERK1/2 in transfected human umbilical vein endothelial cells (HUVECs; n=3). B, MBD2 siRNA-transfected HUVECs showed increased BCL-2 levels. A 1.2-fold higher BCL-2 was observed in MBD2 siRNA-transfected cells. C, Inhibition of MBD2 in HUVECs enhanced endothelial nitric oxide synthase (eNOS) and vascular endothelial growth factor receptor 2 (VEGF-R2) expression. MBD2 siRNA transfected HUVECs showed much higher levels of total eNOS and VEGF-R2. The increased eNOS expression was associated with higher Ser-1177 phosphorylated (p-) eNOS (left). The graph shows the results of quantitative analysis for total eNOS and VEGF-R2 (n=3). D, Analysis of blood vessels derived from Mbd2^−/− mice showed results similar to those of HUVECs after knockdown of MBD2 (n=3). E, Suppression of MBD2 in HUVECs promoted p38 activation. A 1.3-fold higher p-p38 was observed by quantitative analysis (n=3). Twenty-four hours after transfection, HUVECs were treated with 0.2 mmol/L H₂O₂ for 24 hours; then, cell lysates were used for Western blot analysis of ERK1/2 and BCL-2. In contrast, lysates from HUVECs 24 hours after transfection were directly used for Western blot analysis of eNOS, VEGF-R2, and p38. Ctl indicates control. **P<0.001; *P<0.01.

To study the mechanism underlying MBD2 regulation of angiogenesis, we analyzed the activity of eNOS and VEGF-R2, the 2 key proangiogenesis signals.^18,19 Remarkably, knockdown of MBD2 increased both total and activated eNOS (p-eNOS) and VEGF-R2 compared with control siRNA-transfected HUVECs (Figure 4C). Importantly, blood vessels isolated from Mbd2^−/− mice also showed enhanced eNOS and VEGF-R2 expression compared with their control counterparts (Figure 4D); a similar trend was noticed in ischemic tissues of Mbd2^−/− mice (Figure III in the online-only Data Supplement). In contrast, MBD2 did not show a significant impact on total p38, but higher p-p38 was detected in MBD2 siRNA-transfected HUVECs (Figure 4E), and studies in Mbd2^−/− blood vessels showed similar results (Figure IV in the online-only Data Supplement). Together, loss of MBD2 facilitates the activation of cell survival and proangiogenic signals to enhance endothelial survival and to promote angiogenesis.

Endothelial Nitric Oxide Synthase Synergizes With Vascular Endothelial Growth Factor Receptor 2 to Promote Endothelial Survival and Angiogenesis

To demonstrate the connection of the signals characterized above, we treated HUVECs with VEGF in the presence of blockades for either VEGF-R2 or eNOS. Stimulation with VEGF did not affect total VEGF-R2, eNOS, and ERK1/2 expression but promoted VEGF-R2, eNOS, and ERK1/2 activation, along with enhanced BCL-2 expression (Figure 5A). As expected, VEGF-R2–neutralizing antibody or inhibitor (SU1498) suppressed VEGF-induced VEGF-R2 activation and diminished VEGF-induced ERK1/2 activation and BCL-2 upregulation. Blockade of VEGF-R2 also led to a modest decrease in p-eNOS (Figure 5A). Notably, blockade of eNOS by L-NAME not only inhibited eNOS activation (p-eNOS) but also suppressed VEGF-induced p–VEGF-R2, along with a significant decrease in p-ERK1/2 and BCL-2 expression (Figure 5A). These results suggest that ERK1/2 and BCL-2 are downstream molecules of eNOS and VEGF-R2 signaling. Our data also suggest crosstalk between eNOS and VEGF-R2 signaling in which eNOS plays a predominant role by synergizing with VEGF-R2 to enhance endothelial survival and to promote angiogenesis.

Figure 5.

Endothelial nitric oxide synthase (eNOS) synergizes with vascular endothelial growth factor (VEGF) receptor 2 (VEGF-R2) to promote endothelial survival and angiogenesis. A, Blocking assays identified a network downstream of VEGF signaling. Human umbilical vein endothelial cells (HUVECs) were stimulated with VEGF in the presence of blockades for either VEGF-R2 (blocking antibody or SU1498) or eNOS (with N^G-nitro-l-arginine methyl ester [L-NAME]). Extracellular signal-regulated kinase 1/2 activation and BCL-2 expression were shown to be regulated by eNOS and VEGF-R2 signaling, whereas eNOS and VEGF-R2 were found to be downstream molecules of VEGF signaling. It was also found that eNOS synergizes with VEGF-R2 to promote endothelial survival and angiogenesis. B, Blockade of eNOS by L-NAME abolished the improvement of perfusion recovery in Mbd2^−/− mice. Mice after hind-limb ischemic surgery were provided drinking water containing L-NAME (1 mg/mL) to inhibit eNOS signaling, and mice provided with regular drinking water served as controls. Blood flow was monitored as described earlier. **P<0.001.

To confirm the above conclusion, we next treated Mbd2^−/− and control mice after hind-limb ischemic surgery with the eNOS inhibitor L-NAME. Administration of L-NAME significantly impaired blood flow restoration in both Mbd2^−/− and control mice (Figure V in the online-only Data Supplement). More significantly, L-NAME treatment completely abolished the protective effect seen in Mbd2^−/− mice as manifested by the similar extent of impaired perfusion after ischemic injury (Figure 5B).

Methyl-CpG–Binding Domain Protein 2 Binds to the Methylated CpG Elements in the Endothelial Nitric Oxide Synthase and Vascular Endothelial Growth Factor Receptor 2 Promoter

Previous studies suggested evidence supporting a role for DNA methylation regulating eNOS and VEGF-R2 transcription.^3,20,21 We therefore hypothesized that MBD2 represses eNOS and VEGF-R2 expression by directly binding to the methylated CpG elements in their promoter. Although no typical CpG island exists in the eNOS promoter, a region containing CpG elements is found in the 5′-flanking region (Figure 6A). On the contrary, bioinformatic analysis characterized 2 putative CpG islands in the VEGF-R2 promoter (Figure 6B). To test our hypothesis, we first treated HUVECs with 5-aza-2′-deoxycytidine (5-azadC), a DNA methylation inhibitor. Remarkably, knockdown of MBD2 by siRNA had no perceptible effect on 5-azadC–treated HUVECs (Figure VI in the online-only Data Supplement), indicating that the effect of MBD2 involves DNA methylation. Then, ChIP was used to pull down the MBD2/DNA complexes. Primers flanking the entire eNOS promoter/5′-flanking region and the putative CpG islands of VEGF-R2 were used to amplify the MBD2-targeted DNA with the resultant precipitates. No amplification was observed for all primers in the eNOS promoter, but primers (F9/R9 and F10/R10) in the 5′-flanking region yielded positive results (Figure 6C). Surprisingly, among the 3 primer pairs used to flank the putative CpG islands for VEGF-R2, only primers (F3/R3) covering the region between −64 and 166 yielded positive results (Figure 6D).

Figure 6.

Methyl-CpG–binding domain protein 2 (MBD2) binds to the methylated CpG-elements in endothelial nitric oxide synthase (eNOS) 5′-flanking region and vascular endothelial growth factor receptor 2 (VEGF-R2) promoter. A, The eNOS 5′-flanking region (−135 to 500 bp; transcriptional starting site as position 1) enriches with CpG elements. Primer pairs flanking the entire eNOS promoter and 5′-flanking region were indicated. B, Two putative CpG islands were characterized in the VEGF-R2 promoter, and 3 primer pairs flanking this region were indicated. C, Chromatin immunoprecipitation (ChIP) results for eNOS. Polymerase chain reaction amplifications were performed with purified total input DNA (input), immunoprecipitates resulting from a specific antibody (MBD2 antibody) or an unrelated antibody (β-actin antibody), and human umbilical vein endothelial cell (HUVEC) genomic DNA (positive control), respectively. Only primer pairs F9/R9 (31 to 198) and F10/R10 (275 to 396) yielded positive results. D, ChIP results for VEGF-R2. Polymerase chain reaction amplifications were carried out as above, but only primer pair F3/R3 (−64 to 166) produced positive results. E, The eNOS 5′-flanking region underwent a DNA methylation turnover after ischemic insult in HUVECs. Bisulfite sequencing analysis was carried out for the region between −135 and 419 of the eNOS 5′-flanking region, which contains 13 CpG elements. Analysis of 10 randomly selected clones revealed a 1-fold increase in the average methylation rate of these CpG elements in the ischemic condition compared with the physiological condition.

To confirm the above ChIP assay results, we specifically examined the methylation states of 13 CpG elements in the eNOS 5′-flaking region by bisulfite DNA sequencing. Indeed, we detected CpG methylation in both the physiological and ischemic condition (Figure VII in the online-only Data Supplement). We also characterized a significant DNA methylation turnover on ischemic insult (hypoxia plus serum starvation) in HUVECs as evidenced by a 1-fold increase in the methylation rate of these CpG elements (Figure 6E). Electrophoretic mobility shift assay was then carried out to further confirm that MBD2 binds to the methylated CpG elements within this region. Biotin-labeled polymerase chain reaction products flanking 31 to 197 were first methylated by SssI methylase and then used as the probe to detect MBD2 binding activity. It was found that MBD2 bound to those methylated polymerase chain reaction products with high affinity (Figure VIII in the online-only Data Supplement). To address that MBD2 represses eNOS transcription by binding to those methylated CpG elements, eNOS promoter (−1700 to 350) was subcloned into a pGL-2 vector (pGL-eNOS). A mutated eNOS promoter reporter (pGL-eNOSm), in which cytosines in all CpG elements between −200 and 300 were mutated to adenosine, was also constructed. Luciferase assays were then performed with those in vitro methylated reporters along with a pcDNA-MBD2 plasmid in HUVECs, respectively. Consistently, only those methylated reporters showed differences for luciferase activities, in which loss of CpG elements resulted in a 1-fold increase in the reporter activities. In contrast, both pGL-eNOS and pGL-eNOSm showed similar reporter activities in the unmethylated condition (Figure IX in the online-only Data Supplement). Taken together, these results suggest that MBD2 represses eNOS and VEGF-R2 transcription by directly binding to the methylated CpG elements in their promoter region.

Methyl-CpG–Binding Domain Protein 2 Repression of Endothelial Nitric Oxide Synthase Transcription Involves Chromatin Remodeling and Is Confined to Endothelial Cells

It has been suggested that MBD2-mediated transcriptional repression is associated with chromatin remodeling that involves histone deacetylases (HDACs).^22,23 To check whether MBD2 repressing eNOS transcription also involves chromatin remodeling, we exposed cells to trichostatin A (TSA), an HDAC inhibitor that inhibits eNOS transcription,²⁴ into HUVECs transfected with MBD2 or control siRNA. Surprisingly, the effect of MBD2 siRNA on eNOS expression was completely abolished by TSA treatment, as revealed by the same levels of eNOS expression in both MBD2 and control siRNA-transfected HUVECs (Figure 7A).

Figure 7.

Methyl-CpG–binding domain protein 2 (MBD2) repression of eNOS expression is confined to endothelial cells (ECs). A, Trichostatin A (TSA) treatment abolished the effect of MBD2 siRNA on endothelial nitric oxide synthase (eNOS) expression in human umbilical vein ECs (HUVECs). Transfected HUVECs were treated with TSA or control vehicle for 48 hours and then subjected to Western blot analysis. Similar levels of eNOS were detected in MBD2 and control siRNA-transfected HUVECs on TSA treatment. B, Knockdown of MBD2 failed to induce eNOS expression in HeLa cells. No detectable eNOS was observed in HeLa cells transfected with either MBD2 or control siRNA; however, the addition of TSA induced eNOS expression in both transfected cells. C, Splenocytes derived from Mbd2^−/− mice did not express eNOS, but TSA induced eNOS expression in both Mbd2^−/− and wild-type (WT) splenocytes. Splenocytes derived from both Mbd2^−/− and WT mice were treated with TSA for 48 hours before Western blot analysis.

Because TSA has been shown to induce eNOS expression in non-EC types,²⁴ we wondered whether loss of MBD2 would lead to eNOS expression in non-ECs. We then treated siRNA-transfected HeLa cells with either TSA or dimethyl sulfoxide (vehicle). Surprisingly, knockdown of MBD2 failed to induce eNOS expression in HeLa cells, but eNOS was detected in the same cells after TSA treatment (Figure 7B). Studies in HEK293 cells yielded similar results (data not shown). Next, we analyzed eNOS expression in splenocytes originated from Mbd2^−/− and control mice and found that eNOS was absent in both Mbd2^−/− and WT splenocytes. However, eNOS was induced in all cells after TSA treatment (Figure 7C). Together, our results indicate that MBD2 repressing eNOS expression involves chromatin remodeling and that this effect is likely confined to ECs.

Discussion

Tissue- or organ-specific gene expression patterns after embryonic development are maintained and controlled by epigenetic changes instead of DNA sequence alterations.²⁵ DNA methylation as a major epigenetic mechanism of the genome provides a new perspective on transcriptional control paradigms in ECs. Despite intensive investigation, how DNA methylation influences endothelial function remains poorly understood. In the present report, we used MBD2, an epigenetic interpreter, to decipher DNA methylome–encoded information in the regulation of endothelial function in both physiological and disease states. For the first time, our studies provided direct evidence that DNA methylation regulates endothelial survival and proangiogenic signals, which constitutes the molecular basis for understanding how an epigenetic factor affects the genome to modify disease susceptibility.

An interesting finding was that MBD2 was dispensable for the normal physiology of ECs and animals. Endothelial cells under physiological condition expressed only moderate levels of MBD2, but loss of MBD2 significantly enhanced angiogenesis and protected ECs from H₂O₂-induced apoptosis (Figure 1). Hind-limb ischemia was then used to address the importance of MBD2 in disease pathogenesis. In the hind-limb ischemic model, ischemic injury was positively correlated with MBD2 expression (Figure 2). As a result, loss of MBD2 afforded significantly improved recovery of perfusion after ischemic surgery, along with increased capillary and arteriole formation (Figure 3). Given the fact that MBD2 itself does not affect the methylation of DNA and is dispensable for routine physiological activities,¹⁴ our results high-light the potential of MBD2 to be an epigenetic therapeutic target relevant to disease states.

Survival signals conducted by ERK1/2 and Akt signaling have been increasingly recognized to be critical for endothelial viability.²⁶ Interestingly, MBD2 only selectively influences the activation of ERK1/2, as evidenced by higher levels of p-ERK1/2 (Figure 4A) but no perceptible changes in p-Akt. Because ERK1/2 has been found to phosphorylate BCL-2 and to prevent its ubiquitin-dependent degradation,¹⁷ we also detected significantly higher levels of BCL-2 after knockdown of MBD2 (Figure 4B). These results suggest that ECs with reduced MBD2 expression exhibit enhanced ERK1/2 activity, which antagonizes proapoptotic stimuli by increasing BCL-2 expression.

Studies in MBD2 siRNA-transfected HUVECs and Mbd2^−/− mice consistently demonstrated that MBD2 negatively regulates eNOS and VEGF-R2 expression (Figure 4C and 4D). The importance of eNOS in endothelial function is underscored by the observation that eNOS^−/− animals develop systemic and pulmonary hypertension, altered vascular remodeling, abnormal angiogenesis, impaired wound healing, and defective mobilization of stem and progenitor cells.^27,28 Similar to eNOS, VEGF-R2 transduces most of the VEGF-mediated mitogenic, survival, and vascular permeability signals critical for angiogenesis.^29,30 We first demonstrated evidence suggesting that MBD2 regulates angiogenesis by affecting VEGF downstream signaling because VEGF stimulation of HUVECs activated signals similar to those of knockdown of MBD2 expression (Figure 5A). Interestingly, blocking either eNOS or VEGF-R2 signaling inhibited VEGF-induced ERK1/2 activation and BCL-2 expression, indicating that ERK1/2 and BCL-2 are downstream molecules of the eNOS and VEGF-R2 signaling. Given the fact that MBD2 did not show a perceptible impact on ERK1/2 expression, this result suggests that the enhanced ERK1/2 activation and BCL-2 expression after knockdown of MBD2 could be indirect effects resulting from increased eNOS and VEGF-R2 expression. Of important note, studies in VEGF-stimulated HUVECs also suggested evidence supporting crosstalk between eNOS and VEGF-R2 signaling in which eNOS synergizes with VEGF-R2 to enhance endothelial survival and angiogenesis; therefore, blockade of eNOS by L-NAME completely abolished the protective effect against hind-limb ischemia seen in Mbd2^−/− mice (Figure 5B).

We also noticed higher p38 activation after knockdown of MBD2 expression (Figure 4E). The exact role for p38 in endothelial function, however, remains controversial.^31,32 It is likely that the activation of p38 may serve a number of different functions, depending on the cellular context or rather the activation state of other signaling pathways. The role of p38 in the epigenetic pathways affected by MBD2 remains unclear in the present study; therefore, dissecting the role of p38 signaling in MBD2 deficiency–related phenotypes would be a critical focus of future studies.

A key question is how MBD2 regulates eNOS and VEGF-R2 expression. We assumed that MBD2 executes this function by binding to the methylated CpG elements in their promoter region. Although a previous study suggested evidence for DNA methylation in restricting eNOS expression to the vascular endothelium,²¹ direct evidence indicating the methylation of CpG elements in its promoter is lacking. Treatment of MBD2 siRNA-transfected HUVECs with 5-azadC yielded the first line of evidence supporting that MBD2 regulation of eNOS and VEGF-R2 expression implicates the methylation of CpG elements (Figure VI in the online-only Data Supplement). Indeed, bisulfite sequencing analysis of the eNOS 5′-flanking region demonstrated direct evidence confirming the methylation of these CpG elements (Figure VII in the online-only Data Supplement). The eNOS 5′-flanking region (−135 to 419) contains 13 CpG elements. Under the ischemic condition, the average methylation rate of these CpG elements increased by 1-fold compared with that in the physiological condition (Figure 6E). Our subsequent ChIP, electrophoretic mobility shift assay, and eNOS promoter reporter assays provided convincing evidence that MBD2 directly binds to the methylated CpG elements in the eNOS 5′-flanking region and suppresses its transcription (Figure 6B and 6D and Figures VIII and IX in the online-only Data Supplement). Collectively, our data suggest that ECs in the ischemic condition undergo a DNA methylation turnover that encodes the information in favor of disease pathologies and that MBD2 deciphers the change of DNA methylome by binding to the methylated CpG elements of targeted genes.

Previous studies have revealed that MBD2 responds to methylated DNA by recruiting HDACs and other transcription repression factors to the chromatin.²³ Unexpectedly, administration of TSA, a specific inhibitor of HDACs, completely abolished the effect of MBD2 siRNA on eNOS expression in HUVECs (Figure 7). Strikingly, unlike TSA, which induces eNOS expression in non-ECs,³³ the effect of MBD2 silencing on eNOS expression was observed only in ECs. For example, HeLa cells or HEK293 cells with reduced MBD2 failed to express eNOS, and eNOS was undetectable in splenocytes derived from Mbd2^−/− mice. On the contrary, TSA treatment induced eNOS expression in both WT and MBD2-deficient splenocytes. Together, our results suggest that MBD2 regulation of eNOS expression is probably confined to ECs. In contrast to this conclusion, we also noticed an enhanced arteriogenesis in Mbd2^−/− mice after ischemic insult as evidenced by the increase in α-actin–positive cells (Figure 3C). However, this phenotype is probably an indirect effect resulting from enhanced eNOS expression because eNOS also plays an essential role in inducing arteriogenesis.^34,35

Conclusions

The present report demonstrates evidence that MBD2 deciphers DNA methylome–encoded information to modulate endothelial function. On environmental insults, ECs undergo a DNA methylation turnover that encodes the information in favor of disease pathologies. MBD2 deciphers the change in DNA methylome by binding to the methylated CpG elements of targeted genes. As a result, inhibition of MBD2 enhances angiogenesis and protects ECs from H₂O₂-induced apoptosis. Mice deficient in MBD2 also show significantly improved perfusion after hind-limb ischemic injury. Given the fact that MBD2 itself does not modify the patterns of DNA methylation and appears to be dispensable for normal physiology, our results suggest that MBD2 could be a viable epigenetic target to modulate endothelial function in disease states such as targeting MBD2 to prevent endothelial dysfunction during the course of atherosclerosis and diabetic nephropathy.

CLINICAL PERSPECTIVE.

Functional impairment for endothelial cells is associated with an attenuated angiogenic response that features prominently in patients with poor clinical outcomes after peripheral and/or myocardial ischemia. Although DNA methylation has long been implicated in endothelial pathologies, the underlying mechanisms remain poorly understood. The present article reports the impact of methyl-CpG–binding domain protein 2 (MBD2), an interpreter for DNA methylome-encoded information, on endothelial function in both physiological and disease states. Moderate levels of DNA methylation were characterized for some key endothelial genes, such as endothelial nitric oxide synthase and vascular endothelial growth factor receptor 2 in the physiological condition. However, these genes undergo a significant DNA methylation turnover when endothelial cells are under ischemic condition. MBD2 deciphers this methylation turnover–encoded information by directly binding to the methylated CpG elements, leading to the suppression of their transcription in favor of endothelial pathologies. Therefore, suppression of MBD2 expression enhances endothelial nitric oxide synthase and vascular endothelial growth factor receptor 2 transcription, which then promote extracellular signal-regulated kinase 1/2 activation and BCL-2 expression. Consistent with these results, knockdown of MBD2 by siRNA enhanced endothelial angiogenesis and provided protection against H₂O₂-induced apoptosis. Mice deficient in MBD2 showed remarkably improved perfusion recovery after hind-limb ischemic surgery. Furthermore, MBD2 regulation of endothelial nitric oxide synthase expression is likely confined to endothelial cells because other types of cells fail to express endothelial nitric oxide synthase after loss of MBD2 such as splenocytes. These discoveries are important for understanding how an epigenetic factor affects the genome to modify disease susceptibility. Because MBD2 itself does not affect DNA methylation and is dispensable for normal physiology in mice, it could be a viable epigenetic target for modulating endothelial function in disease states.