Flow-dependent epigenetic regulation of IGFBP5 expression by H3K27me3 contributes to endothelial anti-inflammatory effects.

Rationale: Atherosclerosis is a chronic inflammatory and epigenetic disease that is influenced by different patterns of blood flow. However, the epigenetic mechanism whereby atheroprotective flow controls endothelial gene programming remains elusive. Here, we investigated the possibility that flow alters endothelial gene expression through epigenetic mechanisms.
Methods: En face staining and western blot were used to detect protein expression. Real-time PCR was used to determine relative gene expression. RNA-sequencing of human umbilical vein endothelial cells treated with siRNA of enhancer of zeste homolog 2 (EZH2) or laminar flow was used for transcriptional profiling.
Results: We found that trimethylation of histone 3 lysine 27 (H3K27me3), a repressive epigenetic mark that orchestrates gene repression, was reduced in laminar flow areas of mouse aorta and flow-treated human endothelial cells. The decrease of H3K27me3 paralleled a reduction in the epigenetic "writer"-EZH2, the catalytic subunit of the polycomb repressive complex 2 (PRC2). Moreover, laminar flow decreased expression of EZH2 via mechanosensitive miR101. Genome-wide transcriptome profiling studies in endothelial cells treated with EZH2 siRNA and flow revealed the upregulation of novel mechanosensitive gene IGFBP5 (insulin-like growth factor-binding protein 5), which is epigenetically silenced by H3K27me3. Functionally, inhibition of H3K27me3 by EZH2 siRNA or GSK126 (a specific EZH2 inhibitor) reduced H3K27me3 levels and monocyte adhesion to endothelial cells. Adenoviral overexpression of IGFBP5 also recapitulated the anti-inflammatory effects of H3K27me3 inhibition. More importantly, we observed EZH2 upregulation, and IGFBP5 downregulation, in advanced atherosclerotic plaques from human patients.
Conclusion: Taken together, our findings reveal that atheroprotective flow reduces H3K27me3 as a chromatin-based mechanism to augment the expression of genes that confer an anti-inflammatory response in the endothelium. Our study exemplifies flow-dependent epigenetic regulation of endothelial gene expression, and also suggests that targeting the EZH2/H3K27me3/IGFBP5 pathway may offer novel therapeutics for inflammatory disorders such as atherosclerosis.

Introduction

Atherosclerosis is a devastating global disease in urgent need of novel theranostics 1-7. Atherosclerotic plaques has a preferential distribution in aortic regions of turbulent/disturbed flow, but regions with steady laminar flow are protected against atherosclerosis development 8-12. The underlying mechanisms of this geographical distribution of atherosclerotic lesions remain largely unclear, but are likely related to hemodynamic forces-induced signal transduction and gene expression. In particular, laminar flow increases the expression of many anti-atherosclerotic genes, such as kruppel-like factor 2 (KLF2), endothelial nitric oxide synthase (eNOS), and nuclear factor erythroid 2-related factor 2 (Nrf2), which critically regulate endothelial homeostasis, but decreases those “pro-atherogenic” genes, such as monocyte chemoattractant protein-1 (MCP1), and endothelin-1 (ET1) 8-14. In contrast, disturbed flow skews endothelial cells towards the direction of endothelial dysfunction 8-12. An understanding of mechanisms that mediate the differential gene expression pattern between laminar flow and disturbed flow regions offers us therapeutic opportunities to intervene in endothelial dysfunction and atherosclerosis.

Atherosclerosis can be regulated by several epigenetic mechanisms, such as DNA methylation/demethylation 15, histone methylation/demethylation and acetylation/deacetylation 16, 17, microRNAs 18 and long non-coding RNAs 19, 20. Research in the past decade has indicated that these epigenetic mechanisms modulate multiple aspects of endothelial function 21, lineage commitment 22, and atherogenesis 4. In the context of epigenetic modifications mediated by flow, recent evidence has shown that flow shear stress regulates endothelial gene expression by DNA methylation 23-25, histone acetylation 26, miRNAs 27-29 and long non-coding RNAs 30 . However, it remains elusive whether flow regulates histone methylation and the expression of novel mechanosensitive genes that regulate endothelial function. Among the epigenetic modifiers, EZH2 in the polycomb repressive complex 2 (PRC2) has gained considerable interest for its regulatory role in maintaining pluripotency, cell differentiation, organ development and cancer 31. PRC2, which directly trimethylates histone 3 lysine 27 (H3K27me3) and causes chromatin compaction and transcriptional silencing, is composed of EED, SUZ12, RbAp46/48, and EZH2 31. In contrast, PRC2-mediated H3K27me3 formation can be reversed by two histone demethylases, named UTX (ubiquitously transcribed X-chromosome tetratricopeptide repeat protein, also known as KDM6A) and JMJD3 (jumonji domain containing-3, also known as KDM6B) 31. Several lines of evidence have implicated EZH2 in the development and progression of atherosclerosis. For example, recent studies observed increased H3K27 dimethylation in endothelial cells from human early and advanced atherosclerotic plaques 16 despite a global decrease in H3K27 methylation level 32, 33. EZH2 is also responsive to fluid shear stress and involved in laminar flow-mediated cell quiescence 34. EZH2 regulates transendothelial migration of leukocytes to microvessels 35 as well as tumor angiogenesis 36. In addition, EZH2 has recently been shown to restrict myocardin-dependent smooth muscle differentiation from the mesothelium 37 and promote the formation of foam cell and atherosclerosis development in hyperlipidemic ApoE-/- mice 38. Moreover, EZH2 was also upregulated in hyperhomocysteinemic ApoE-/- mice and contributed to foam cell formation 39. These evidences suggest that EZH2 could play a pro-atherogenic role in atherosclerosis development.

Since EZH2 is a histone methyltransferase that responds to laminar flow 34, it remains unknown whether atheroprotective flow controls EZH2-mediated histone methylation patterns, which in turn alter endothelial gene expression, thus regulating endothelial function. Here, we show that EZH2/H3K27me3 is reduced by laminar flow in endothelial cells both in vivo and in vitro. Through genome-scale RNA sequencing (RNA-seq)-based comparison, chromatin-immunoprecipitation (ChIP) and luciferase assays, we demonstrated that atheroprotective flow upregulates novel mechanosensitive genes, such as IGFBP5 (insulin-like growth factor-binding protein 5) through an H3K27me3-dependent inhibitory mechanism to represses endothelial inflammation. Finally, we observed that EZH2 is upregulated, while IGFBP5 is downregulated in advanced atherosclerotic plaques from patients undergoing carotid endarterectomy, indicating the clinical relevance of EZH2 in regulating atherosclerosis.

Materials and methods

Study population

Human atherosclerotic plaques were obtained from patients undergoing carotid endarterectomy deposited in the Biobank of Department of Vascular and Endovascular Surgery (Klinikum rechts der Isar der Technischen Universität Muenchen, Germany) 16, 33. Atherosclerotic plaques were histologically characterized according to the American Heart Association (AHA) Stary classification 40, 41 and divided into early (types I-III) or advanced (types V-VII) atherosclerosis stages. In this study, carotid arteries (n=10) with advanced stage of atherosclerosis (types VI) were used. Plaque-free healthy carotid arteries (n=10) were obtained from the Department of Trauma Surgery as previously described 16, 33. Details of inclusion and exclusion criteria are provided in Supplementary Material. The demographic data of atherosclerotic patients and healthy donors are summarized in Table . The study was approved by the ethics committee of Technischen Universität Muenchen with patients' written informed consent. The study was performed according to the guidelines of the Declaration of Helsinki.

RNA isolation and quantitative real-time PCR

RNA was extracted from formalin-fixed, paraffin-embedded (FFPE) carotid artery samples using the High Pure RNA Paraffin Kit (#3270289001, Roche, Mannheim, Germany) according to the manufacturer's instructions 16, 33. Two 10 μm sections of each sample were pooled for RNA isolation. Total RNA from cultured cells and mice tissues were isolated with RNeasy Minikit and TRIzol Reagent, respectively. Total RNA was converted into complementary DNA (cDNA) using the High-Capacity cDNA Reverse Transcription Kit (#4374966, Applied Biosystems, Foster City, CA) following the manufacturer's instructions. Then, quantitative real-time PCR (qRT-PCR) was performed using iQ SYBR Green Supermix (#1708886, Bio-Rad) as previously described in detail 42. The comparative cycle threshold (Ct) method (2-ΔΔCt) was used to determine the relative gene expression after normalization to housekeeping gene GAPDH. The Ct values of EZH2 in one control sample and one atherosclerotic patient sample were below the detection limit of qRT-PCR and thus not included for analysis. The sequences of primers are listed in Table .

En face immunofluorescent staining

En face staining was performed as previously described in detail 43, with minor modifications 42, 44. Mouse aortas were collected after perfusion with saline and neutrally buffered 4% PFA. Aortic segments were then permeabilized with 0.1% Triton X-100 in PBS for 10 min and blocked with 10% normal goat serum (Invitrogen) in Tris-buffered saline (TBS) containing 2.5% Tween-20 for 30 min at room temperature. Next, aorta segments were incubated with rat anti-VE-Cadherin (1:100; #555289, BD Bioscicence), rabbit anti-EZH2 (#6263, ProSci) or rabbit anti-H3K27me3 (#39155, Active Motif) antibody overnight at 4 °C. After rinsing with washing buffer 3 times, aortic segments were incubated with Alexa Fluor 488-conjugated goat anti-rat and Alexa Fluor 546-conjugated goat anti-rabbit secondary antibodies (1:1,000; Thermo Fisher Scientific) for 1 h in darkness at room temperature. Finally, aortic segments were carefully whole mounted (with lumen side up) in the ProLong Gold-antifade Mounting Media (Invitrogen) for confocal microscopy (IX81, Olympus) with 60x or 40x oil lens. All animal procedures conformed to the Guideline for the Care and Use of Laboratory Animals published by the U.S. National Institutes of Health and were approved by the Institutional Animal Care and Use Committee of the University of Rochester Medical Center.

Primary isolation of mouse endothelial cells, smooth muscle cells and macrophages

Mouse endothelial cells were isolated from adult C57BL/6J mouse lungs using collagenase I digestion (3 mg/mL in serum-free DMEM) at 37 °C in an incubator for 45 min, followed by CD31- and ICAM2-dynabeads-based double positive selection 45, 46. The isolated mouse lung endothelial cells were authenticated by endothelial marker VE-cadherin staining. More than 99% of cultured cells were VE-cadherin positive. The isolation of vascular smooth muscle cells (VSMCs) was performed using the explant method as previously described 47, with minor modifications. VSMCs migrated from thoracic aorta explants at around 8 days, then, sprouted VSMCs were digested with 0.05% Trypsin-EDTA and reseeded for confocal microscopy or RNA isolation. More than 85% of cultured cells were smooth muscle alpha-actin (α-SMA) positive. Mouse peritoneal macrophages were isolated from the peritoneum cavity by injecting 5 mL ice-cold PBS/0.1% BSA 48. Cells were cultured in complete media for 4 h, then, non-adherent cells were removed by washing. Macrophages were stained with CD68, and more than 99% of attached cells were CD68 positive.

Cell culture

Human umbilical vein endothelial cells (HUVECs) were isolated from fresh human umbilical cords and cultured in Medium 200 containing 5% FBS and 1X low serum growth supplement (LSGS) 42, 44. HUVECs were collected in accordance with the University of Rochester human subjects review board procedures that prescribe to the Declaration of Helsinki. Different donors of human coronary artery endothelial cells (HCAECs) were used in this study, which were from Lifeline Cell Technologies (Frederick, MD, #FC-0032, Lot No. 01181, 01354) and Cell Applications Inc. (#300K-05A, Lot No. 1858, San Diego, CA). HCAECs were cultured in MesoEndo Cell Growth Media supplemented with Growth Supplement (#212K-500, Cell Applications Inc.) plus 5% FBS 42. For cell authentication, HUVECs and HCAECs were identified by cobblestone cell morphology, staining of endothelial cell-marker VE-Cadherin (BD Bioscicence, #555661) and functional uptake of DiI-oxidized LDL (10 µg/mL, 4 h; Alfa Aesar, #J64164). HCAECs tested negative for mycoplasma, bacteria, yeast, fungi, Hepatitis B and C, as stated on the Vendors' certificates of analysis on individual lots of cells. HUVECs (confluence 80-90%) and HCAECs (confluence 60-70%) were subcultured on the day before experiments. For flow experiments, ECs cultured in fresh complete media were exposed to laminar flow 42 or disturbed flow 49 in a 60 mm or 100 mm cone and plate viscometer. Due to the fact that prolonged culture of ECs renders ECs less responsive to flow 50, HUVECs and HCAECs were passaged on the day before flow treatment. HUVECs and HCAECs at passage 2 to 5 were used for all the experiments. THP-1 cells (gifted by Y. Cai), a pro-monocytic cell line, were cultured as previously described 42. THP-1 cells were authenticated by the ability to differentiate into macrophages after PMA (100 nM, 48 h) stimulation and form foam cells after loading with oxidized LDL for 24 h. For cell culture studies, all compounds were prepared as stock solution at 10 mM in dimethylsulfoxide (DMSO) and working solutions were freshly prepared before use.

Monocyte adhesion assay

TNFα is a commonly used stimulus in triggering monocyte adhesion/rolling/transmigration 51, 52, by enhancing the production of MCP1 51, 53 and other pro-inflammatory and pro-migratory cytokines (such as IL6) 54. Therefore, TNFα was used in the THP-1 monocyte adhesion assay, which was performed as previously described in detail 55. Quantification of adherent THP-1 monocytes was performed by manual counting of monocytes after washing with pre-warmed HUVEC complete media three times.

siRNA and miRNA transfection

To deplete EZH2, HUVECs at 75-80% confluence were used for transfection. In brief, Lipofectamine 2000 or RNAi Max (6 μL; Thermo Fisher Scientific) 42 was mixed with Opti-MEM (250 μL; Thermo Fisher Scientific), and then two independent siRNA oligos against human EZH2 (20 nM; #SR301494-A, -B, Origene, Rockville, MD) or non-target control siRNA (25 nM; #SR30004, Origene) diluted in 250 μL Opti-MEM was added to the solution, mixed gently, and incubated at room temperature for 20 min. After 3-4 h of transfection, the medium was changed to M200 complete medium and cells were treated 48 h after transfection. Additional Silencer® Select EZH2 siRNA (#s4916, Thermo Fisher Scientific) or Silencer® Select Negative Control #1 siRNA (#S25503, Thermo Fisher Scientific) were used for RNA sequencing. The same method was used to transfect HUVECs with Anti-miRTM negative control (100 nM; #AM17010, Ambion) or Anti-miRTM hsa-miR-101-3p miRNA inhibitor (100 nM; #AM17000, Ambion) using RNAi Max as the transfection reagent.

Adenoviral EZH2 overexpression

EZH2 adenovirus was purchased from Vector Biolab (#1371, Malvern, PA). HUVECs were infected with EZH2 adenovirus as indicated.

Extraction of purified histones from cultured cells

Core histones were purified using a commercial Histone Purification Kit (Active Motif, #40025, Carlsbad, CA, USA) following the manufacturer's instructions, with minor modifications. Briefly, HUVECs were treated with a potent and specific EZH2 inhibitor GSK126 (#A1275, Active Biochem, Hongkong) for the indicated time without changing the media. After washing twice with pre-warmed (37 °C) serum-free M200 media, 0.4 mL ice-cold Extraction Buffer was added to a 60 mm dish. Then, a plastic scraper was used to collect the cell protein extracts. The cells were rotated in Extraction Buffer overnight on a VWR rotating platform at 4 °C to release histones. The following day, cell extracts were transferred to fresh tubes and centrifuged in a microcentrifuge at maximum RCF for 5 min at 4 °C. Crude histones were neutralized with 0.1 mL 5X Neutralization Buffer and purified using columns provided. For Western blots, boiled histone lysates were separated by 12-15% SDS-PAGE. The same membrane was stripped and reprobed with anti-Histone 3 (H3) for equal loading.

Western blot analysis

Cells were harvested in freshly-prepared lysis buffer (20 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% Triton X-100, 1 mM EDTA, 1 mM EGTA, 2.5 mM sodium pyrophosphate, 1 mM β-Glycerolphosphate, 50 mM NaF, 1 mM Na3VO4, and 1X protease inhibitor cocktail) by cell lifters. After clarification at 4 °C, 12, 000 g for 10 min, total cell lysate was collected for SDS-PAGE gel analysis as previously described in detail 56. After transfer, membranes were blocked in PBS-diluted LI-COR blocking buffer at room temperature for 1 h. Then, the blots were incubated with 3% BSA-diluted primary antibodies (listed in Table ) at 4 °C overnight or room temperature for 1 h, followed by incubation with LI-COR IRDye® 680RD goat anti-mouse IgG (H+L) or IRDye® 800CW goat anti-rabbit IgG (H + L) or IRDye® 680RD donkey anti-goat IgG (H+L) (dilution at 1:10,000) at room temperature for 30 min. Images were visualized using an Odyssey Infrared Imaging System (LI-COR). Densitometric analysis of blots was performed using NIH Image J software (http://imagej.nih.gov/ij/).

Chromatin immunoprecipitation coupled with quantitative real-time PCR (ChIP-qPCR)

ChIP assays were performed using EZ-Magna CHIPTM A/G ChIP kits (#17-10086, EMD Millipore), with slight modifications 57. Equal amounts of Bioruptor-sheared chromatin were incubated with control rabbit IgG, or rabbit H3K27me3 antibodies at 4 °C overnight. Chromatin-associated proteins were washed and then digested with 1 µL Proteinase K (10 mg/mL) for 2 h at 62 ºC with gentle shaking and incubated at 95 ºC for 10 min for reverse crosslinking and inactivation of Proteinase K. Chromatin DNA were then purified with spin columns provided with the kit. Additional ChIP DNA purification kits were from Active Motif (#58002, Carlsbad, CA). qRT-PCR was performed using ChIP primers encompassing the proximal IGFBP5 gene promoter (Table ). For each PCR reaction, 1 µL ChIP'ed DNA template was added for each 10 µL reaction. 1% of starting chromatin is used as input. Data are presented as % input using the 2-ΔΔCt method.

Luciferase reporter assay

The predicted IGFBP5 promoter region (-1000 bp to +45 bp relative to transcription start site) was synthesized by GeneScript and subcloned into pGL3-basic plasmid (Promega). Luciferase activity assay was performed as previously described 55. In brief, 100 ng Luciferase plasmid was transfected into HUVECs for 24 h using Lipofectamine 2000 as described above. COS7 cells were co-transfected with 10 ng pRL-TK plasmid (Promega) for normalization of transfection efficiency. The ratio of luciferase units to pRL-TK units was calculated and normalized. Data are presented as fold over control after normalization.

Genome-wide RNA sequencing study

RNA sequencing (RNA-seq) is a useful approach for probing the pathological mechanisms of vascular diseases 58. RNA-seq was performed at the University of Rochester Genomics Research Center (URGRC) as previously described 57. For flow RNA-seq, HUVECs were subjected to atheroprotective laminar flow for 48 h using cells under static conditions as the control (n=2). For RNA-seq in HUVECs depleted of EZH2, cells were transfected with control siRNA (20 nM) or EZH2 siRNA (20 nM) for 48 h before RNA was prepared for RNA-seq (n=3). DESeq2-1.10.1 was used to perform data normalization and differential expression analysis with an adjusted p-value threshold of 0.05. The significance test used in DeSeq2 analysis was the Wald test, and a log2 fold change between groups is presented. Adjusted p-values were derived using the Benjamini and Hochberg multiple test correction.

Accession number

The RNA-seq data reported in this paper were deposited in NCBI Gene Omnibus (http://www.ncbi.nlm.nih.gov/geo/) with the accession number GEO: GSE87534. In Figure , published datasets were mined from the GEO database with the accession number of GEO: GDS5083 59.

Statistics

Data are expressed as mean ± SEM. Statistical analysis was performed using GraphPad Prism software 5.0 (GraphPad, La Jolla, CA). Student's t test and one-way analysis of variance (ANOVA) with post hoc Bonferroni tests were used for comparisons between two groups and multiple comparisons, respectively. A p value less that 0.05 was considered statistically significant.

Results

Laminar flow specifically reduces repressive epigenetic mark H3K27me3 in endothelial cells in vitro and in vivo

PRC2 is an evolutionarily conserved multi-protein complex that leads to gene silencing through H3K27me3 31. To examine the potential involvement of PRC2/H3K27me3 in laminar flow-mediated endothelial gene transcription programming, we first performed a RNA-seq study on human endothelial cells exposed to long-term atheroprotective laminar flow (12 dyne/cm2, 48 h), with an aim to assess the effect of long-term flow on the endothelial transcriptome. We focused on the genes in the PRC2 (Figure ), including SUZ12, EED, RbAp46 (RBBP7), RbAp48 (RBBP4), and EZH2. Our data showed that treatment with laminar flow decreased gene expression of EZH2 and RpAp46, without influencing the other three components of PRC2 (Table ). Since histone methylation is dynamically controlled by the equilibrium of histone methylation and demethylation 60, we also looked at the effect of laminar flow on the expression of the H3K27me3 demethylases including KDM6A (also known as UTX) and KDM6B (also known as JMJD3) 60. Our RNA-seq data suggest that laminar flow does not significantly alter KDM6A, or KDM6B gene expression (Table ), implicating that laminar flow could potentially influence net PRC2-dependent H3K27me3 modification. We also observed that, similar to the effect of EZH2 enzymatic inhibitor GSK126 61, laminar flow treatment significantly downregulated the global level of H3K27me3 in endothelial cells (Figure ), indicating that laminar flow influences EZH2-dependent H3K27me3 modification. However, laminar flow does not significantly affect global protein expression of histone 3 lysine 9 trimethylation (H3K9me3) or acetylation (H3K9ac) (Figure ), suggesting the specific inhibitory effects of laminar flow on H3K27me3. To address whether our in vitro observations can be translated into in vivo conditions, we compared the differential expression of H3K27me3 in two different regions of mouse aorta: one is the inner curvature of the aortic arch exposed to disturbed flow, and the other is thoracic aorta generally exposed to laminar flow. We observed a marked decrease in H3K27me3 staining in the laminar flow area of mouse aorta (Figure ). We also observed that the expression pattern of H3K27me3 in vascular endothelium is heterogeneous.

Laminar flow decreases the expression of chromatin modifier EZH2 in vitro and in vivo

Since EZH2 is the histone methyltransferase that imposes H3K27me3 epigenetic mark, we next asked whether the decrease of H3K27me3 is due to the decrease of EZH2 expression. We first determined the tissue and cell distribution of EZH2. Consistent with the human GTEx database (https://www.gtexportal.org/home/gene/EZH2), we observed that EZH2 was mostly expressed in the spleen, consolidating the key role of EZH2 in immunity 62, 63. Other organs enriched in EZH2 include lung, brain and aorta, which are endothelial cells-enriched organs (Figure ). We next compared EZH2 expression in vascular cells, using primary isolated mouse lung endothelial cells, aortic vascular smooth muscle cells and peritoneal macrophages. We observed that EZH2 was prominently expressed in endothelial cells (Figure ). This expression pattern suggests EZH2 may have an important role in regulating immunity and endothelial functions. We then evaluated the time course of laminar flow-induced EZH2 downregulation. We observed that laminar flow significantly decreased EZH2 protein expression starting from 6 h, with maximal reduction at 24 h (Figure ). The flow-dependent reduction of EZH2 was recapitulated in HCAECs exposed to laminar flow (Figure ). In agreement with the RNA-seq data, we observed that laminar flow does not significantly affect the protein expression of PRC2 components EED and SUZ12 (Figure ). To examine whether disturbed flow increases EZH2 expression, we compared the protein expression of EZH2 under disturbed flow and laminar flow conditions, using cells under static conditions as a control. We observed that, compared with static conditions, disturbed flow has no significant effect on EZH2 mRNA and protein expression (Figure ). To corroborate our in vitro study, we analyzed mRNA and protein expression of EZH2 in mouse aorta regions with disturbed flow (inner curvature of aortic arch) and laminar flow (thoracic aorta) (Figure ). Interestingly, we observed a significant reduction of EZH2 mRNA and protein expression in arterial regions with laminar flow compared with that of regions with disturbed flow (Figure ).

miR101 is involved in laminar flow-induced EZH2 downregulation

To dissect the mechanism whereby laminar flow reduces EZH2 expression, we first evaluated the effect of laminar flow on EZH2 mRNA expression, and we observed that EZH2 mRNA was significantly decreased in endothelial cells by laminar flow treatment (Figure ). Non-coding RNAs, such as microRNAs (miRNAs) are important post-transcriptional mechanisms that induce gene silencing. Mounting studies have shown that several different miRNAs, including miR101, miR26a, miR214, and miR124, can directly target EZH2 expression 64-66. EZH2 can also be potentially targeted by other miRNAs (Table ) as predicted by TargetScan 67. Among these miRNAs, miR101 is the top miRNA that has been reported to be mechanosensitive to date 29, 68. We then asked whether miRNA 101 might mediate laminar flow-dependent reduction of EZH2 expression. We pre-treated HUVECs with miR101-anti-miR, and found that laminar flow-induced EZH2 downregulation was reversed by miR101-anti-miR treatment (Figure ), suggesting that miR101 is involved in laminar flow-induced EZH2 downregulation.

Inhibition of EZH2 expression/activity reduces monocyte adhesion to endothelial cells

To understand the biological function of EZH2 in endothelial cells, we used two independent siRNAs to deplete EZH2 expression and associated H3K27me3 (Figure ), then performed monocyte adhesion assays to evaluate the adhesiveness of monocytes to inflamed endothelial cells. Our data showed that EZH2 depletion dramatically inhibited monocyte adhesion to endothelial cells (Figure ). To confirm this finding, we also incubated cells with the EZH2 inhibitor GSK126 61; we observed that GSK126 also inhibited monocyte adhesion to endothelial cells (Figure ), suggesting that diminishing H3K27me3 by EZH2 siRNA or pharmacological inhibitors could reduce endothelial inflammation.

Genome-wide transcriptomic profiling identified IGFBP5 as a flow-sensitive and EZH2/H3K27me3-target gene in endothelial cells

To gain mechanistic insights into the anti-inflammatory effects conferred by inhibiting H3K27me3, we first compared the common transcriptome between laminar flow-treated and EZH2-depleted HUVECs (Figure ). Since H3K27me3 is a repressive epigenetic mark that leads to gene silencing, we mapped out the upregulated common genes between both treatments (Table ).

By comparing the common sets of upregulated genes, we identified many anti-atherosclerotic genes, including HMOX1 (heme oxygenase 1), NQO1 (NAD(P)H quinone dehydrogenase 1), GPX3 (glutathione peroxidase 3), KLF4, PLAT (plasminogen activator, tissue type) and IGFBP5 (Table ). Among genes of interest, we selected a less-characterized new gene IGFBP5 for further studies. Our real-time PCR and western blot analysis indicated that IGFBP5 is a flow-responsive gene that can be upregulated by laminar flow at mRNA and protein levels (Figure ). To validate whether IGFBP5 is an H3K27me3 target gene, we performed ChIP-qPCR analysis in the promoter region of IGFBP5. We observed enrichment of H3K27me3 at IGFBP5 proximal gene promoter, and binding of H3K27me3 to IGFBP5 gene promoter was reduced by laminar flow treatment (Figure ). To confirm whether IGFBP5 is an EZH2/H3K27me3 target gene, we used four approaches: (1) EZH2 gain-of-function by infection with adenoviral EZH2, (2) EZH2 loss-of-function by EZH2 siRNA, (3) pharmacological inhibition of EZH2/H3K27me3 activity by GSK126, and (4) dual-luciferase reporter assay. We observed that EZH2 overexpression decreased, while EZH2 depletion by siRNA or H3K27me3 inhibition by GSK126 increased IGFBP5 expression (Figure ). Luciferase reporter assay also showed that EZH2 overexpression decreased IGFBP5 promoter activity (Figure ), suggesting that IGFBP5 is a bona fide methylation target of H3K27me3. Altogether, our data demonstrate that EZH2 epigenetically silences flow-responsive IGFBP5 expression in endothelial cells.

IGFBP5 inhibits endothelial inflammation

IGFBP5 is an important IGF binding protein that specifically binds to endothelial cell monolayers 69. Previous studies have shown that IGFBP5 is induced by VEGF and IGF1 treatments 70, and mediated radiation-induced endothelial senescence 71. However, the effect of IGFBP5 on endothelial inflammation is unknown. Due to the anti-inflammatory profile in laminar flow-treated endothelial cells, we asked whether IGFBP5 could be an anti-inflammatory molecule in endothelial cells. To this end, we performed a monocyte adhesion assay. Our data showed that adenoviral IGFBP5 overexpression attenuated monocyte adhesion to endothelial cells in a dose-dependent manner (Figure ). IGFBP5 overexpression also decreased the protein expression of pro-adhesive molecules ICAM1 and VCAM1 in TNFα-inflamed endothelial cells (Figure ).

EZH2 and IGFBP5 expression in human atherosclerotic plaques

Since our data pointed to the concept that EZH2 depletion and ensuing IGFBP5 upregulation is anti-inflammatory and probably atheroprotective in endothelial cells, we investigated the expression levels of EZH2 and IGFBP5 in atherosclerotic plaques from patients undergoing carotid endarterectomy. We first performed a datamining study on published GEO datasets examining differential gene expression patterns between atherosclerotic plaques and adjacent disease-free regions 59. We observed a significant upregulation of EZH2 and downregulation of IGFBP5 in plaque areas compared with adjacent plaque-free areas (n=32, Figure ). To confirm this finding, we also performed real-time PCR analysis using the FFPE samples from healthy controls and atherosclerotic patients. H&E staining showed minimal lesions in control samples; however, extensive plaque hemorrhage was observed in advanced plaques from patients (Figure ). Our data also indicated an increase in EZH2 expression in diseased vessels compared with healthy control carotid artery (Figure ). Consistent with the proposition that IGFBP5 is epigenetically silenced by EZH2, we observed a significant decrease in IGFBP5 expression in human atherosclerotic plaques (Figure ). These observations collectively suggest that the clinical relevance of EZH2 in regulating IGFBP5 expression in human atherosclerotic plaques.

Discussion

Atherosclerosis is a focal disease that preferentially develops in regions of disturbed flow, but less in regions of atheroprotective laminar flow 8-12. Several landmark studies 23-25 have recently shown that atheroprone disturbed flow alters endothelial gene expression via DNA methyltransferases (DNMTs)-mediated DNA methylation of key genes involved in endothelial function, such as KLF3, KLF4, and homeobox protein hox-A5 (HoxA5). Disturbed flow also promotes endothelial DNA methylation by repressing the expression of Tet methylcytosine dioxygenase 2 (Tet2) 72. However, the mechanisms by which atheroprotective laminar flow prevents atherosclerosis at the epigenetic level, especially at the histone methylation level, remain poorly defined. In this study, we uncover that the EZH2-H3K27me3-IGFBP5 pathway contributes to laminar flow-dependent atheroprotective effects.

Through genome-wide RNA-seq, we first identified that laminar flow decreased the expression of two PRC2 components, EZH2 and RpAp46, without affecting the expression of H3K27me3 demethylases-UTX and JMJD3. The net outcome of laminar flow treatment could potentially lead to reduced PRC2 activity, evidenced by decreased global H3K27me3 levels. Recent studies have shown that miR101 is mechanosensitive and EZH2 is a well-defined target of miR101 29, 68. By using a miR101 inhibitor treatment of endothelial cells, we observed that miR101 was involved in flow induced EZH2 downregulation, which reproduced a recent study using a different fluid shear system (an ibidi fluidic unit) 34. In light of the negative feedback loop between miRNA 101 and EZH2 65, it is plausible that flow sustains low levels of EZH2 through this regulatory loop. We also showed downregulation of EZH2 mRNA and protein expression in laminar flow areas (thoracic aorta) of mouse aorta, compared with disturbed flow areas (inner curvature of aortic arch). The biological function of EZH2 is to impose a repressive epigenetic mark of H3K27me3 onto target gene promoters, thereby leading to gene silencing 31. Consistent with the proposition that laminar flow exerts strong anti-inflammatory effects; we observed that both EZH2 depletion by siRNA and inhibition by a specific inhibitor GSK126 reduced monocyte adhesion to endothelial cells, suggesting EZH2 and the associated H3K27me3 modifier are pro-inflammatory in endothelial cells. The cellular effects of EZH2 could possibly be mediated by modulating H3K27me3 level and the binding of this repressive epigenetic mark to target gene promoters. To evaluate the effect of atheroprotective flow on the EZH2/H3K27me3-dependent genome-wide transcriptional profile, we performed an RNA-seq transcriptional profiling on laminar flow- and EZH2 siRNA-treated human endothelial cells. A Venn diagram was used to compare the common genes regulated by both laminar flow and EZH2 depletion. We found that atheroprotective flow and EZH2 depletion alter the endothelial gene landscape, which includes upregulating atheroprotective genes (such as IGFBP5) while downregulating pro-atherosclerotic genes. Through studying the overlapping genes between flow and EZH2 depletion, we observed that both treatments upregulate several genes maintaining endothelial homeostasis (HMOX1, NQO1, GPX3, KLF4, and PLAT). Our study extends previous observations showing that laminar flow decreases EZH2 expression in cultured endothelial cells and EZH2 downregulation is involved in flow-induced cell quiescence 34.

Recent studies by Yang et al. 39 and Lv et al. 38 have shown that EZH2 expression is upregulated in mouse models of hyperhomocysteinemia and hyperlipidemia, respectively, unveiling a new functional connection between EZH2 and atherosclerosis. In addition, lentivirus-mediated EZH2 overexpression exacerbates plaque development in ApoE-/- mice fed a high fat diet 38, suggesting that EZH2 may play a pro-atherogenic role in atherosclerosis. However, to date, there are limited EZH2 target genes identified in endothelial cells. Interestingly, our study identifies IGFBP5 as a novel mechanosensitive gene regulated by EZH2-dependent H3K27me3 that may serve as a critical regulator of flow-elicited responses. IGFBPs are important in vascular biology as they regulate gene expression and extracellular matrix accumulation 73. Our study shows that IGFBP5 is a flow-responsive but EZH2-H3K27me3-repressive gene that confers anti-inflammatory effects. It is possible that laminar flow-induced secretion of IGFBP5 could exert anti-inflammatory effects by modulating IGF-dependent or -independent effects 73. These data, together with two recent findings showing that atheroprotective drugs, including statins 74 and omega 3-polyunsaturated fatty acids 75, reduce EZH2 expression in cancer cells, suggesting EZH2 is a promising therapeutic target for pharmacological intervention. More importantly, our data-mining study and the analysis of the expression level of EZH2 and IGFBP5 in human atherosclerotic plaques demonstrated an association of EZH2 and IGFBP5 with human atherosclerotic plaques, thus increasing the translational potential of our study.

In the current study, we mainly focused on the role of EZH2/H3K27me3 in endothelial function. To ascertain the anti-inflammatory profile conferred by laminar flow, it will be important to address in future studies whether genetic or pharmacological inhibition of H3K27me3 demethylase activity by JMJD3 and UTX will reverse the effects of laminar flow or EZH2 deficiency on endothelial function. In light of the important role of phosphorylation at Ser21 76, Thr345 77 and Thr487 77 in regulating EZH2 expression/methyltransferase activity, it will be important to evaluate the temporal phosphorylation profile of EZH2 in endothelial cells in response to flow, and the phosphorylation of Ser21 in particular, which can be activated by flow-responsive kinase Akt 78. Although the precise role of EZH2 in the development of atherosclerosis still requires in-depth investigation using endothelium-specific knockout mice, results presented here indicate that fine-tuned regulation of the levels of EZH2 and H3K27me3 is critical for maintaining endothelial homeostasis. Furthermore, in addition to transcriptional silencing, EZH2 has a non-canonical role in transactivating target genes 31. This aspect has been well defined in cancer cell biology, showing that EZH2 activates NF-κB target genes 79. We are currently investigating the role of EZH2 in transactivating endothelial cell-specific genes. We understand that flow-dependent regulation of gene transcriptional programming is a result of the complex interplay between multiple chromatin modifiers (DNA and histone methyltransferases for example 80). This is particularly intriguing since a recent study has shown that EZH2-dependent PRC2 is associated with multiple long non-coding RNAs 81, thereby participating in gene regulation. We also recognize that laminar shear stress elicits profound changes in endothelial transcriptional programming, which include IGFBP5 and other atherorelevant genes. The study focused on the effects of flow on epigenetic regulation of IGFBP5 expression by H3K27me3 in endothelial cells. All these mechanisms were only confirmed by in vitro studies. It is well established that the cell components are complicated in the advanced plaques, including endothelial cells, macrophages, foam cells and smooth muscle cells. Although EZH2 show higher expression in endothelial cells, compared with other smooth muscle cells and macrophages, it cannot be totally ruled out that EZH2/H3K27me3 could also regulate functions of smooth muscle cells and macrophages, since IGFBP5 is a secreted anti-inflammatory protein, which could possibly impact endothelial cells-smooth muscle cell/macrophage communication. Determining the expression of EZH2/H3K27me3 with different markers of individual vascular cell type in the advanced plaque, will provide further evidence supporting the epigenetic modulation of IGFBP5 in plaque endothelial cells. The elucidation of these novel EZH2-dependent genes that regulate endothelial function will provide us a new direction of cardiovascular research.

In summary, the present study reveals EZH2 as a mechanosensitive and potentially atherogenic chromatin modifier that serves as a molecular switch to determine atheroprotective or atheropromoting outcomes in the endothelium based on distinctive H3K27me3 marks. Defining the link between hemodynamic forces and epigenetic regulation of endothelial gene transcription could deepen our understanding of mechanisms underlying blood flow-dependent gene expression and atherosclerosis, and finally will translate into new remedies for atherosclerosis.

Supplementary methods, figures and tables S1-S4.

Click here for additional data file.

Table S5.

Click here for additional data file.

Table S6.

Click here for additional data file.

Table S7.

Click here for additional data file.