DNA Methyltransferase 1-Dependent DNA Hypermethylation Constrains Arteriogenesis by Augmenting Shear Stress Set Point.

BACKGROUND: Arteriogenesis is initiated by increased shear stress and is thought to continue until shear stress is returned to its original "set point." However, the molecular mechanism(s) through which shear stress set point is established by endothelial cells (ECs) are largely unstudied. Here, we tested the hypothesis that DNA methyltransferase 1 (DNMT1)-dependent EC DNA methylation affects arteriogenic capacity via adjustments to shear stress set point. METHODS AND
RESULTS: In femoral artery ligation-operated C57BL/6 mice, collateral artery segments exposed to increased shear stress without a change in flow direction (ie, nonreversed flow) exhibited global DNA hypermethylation (increased 5-methylcytosine staining intensity) and constrained arteriogenesis (30% less diameter growth) when compared with segments exposed to both an increase in shear stress and reversed-flow direction. In vitro, ECs exposed to a flow waveform biomimetic of nonreversed collateral segments in vivo exhibited a 40% increase in DNMT1 expression, genome-wide hypermethylation of gene promoters, and a DNMT1-dependent 60% reduction in proarteriogenic monocyte adhesion compared with ECs exposed to a biomimetic reversed-flow waveform. These results led us to test whether DNMT1 regulates arteriogenic capacity in vivo. In femoral artery ligation-operated mice, DNMT1 inhibition rescued arteriogenic capacity and returned shear stress back to its original set point in nonreversed collateral segments.
CONCLUSIONS: Increased shear stress without a change in flow direction initiates arteriogenic growth; however, it also elicits DNMT1-dependent EC DNA hypermethylation. In turn, this diminishes mechanosensing, augments shear stress set point, and constrains the ultimate arteriogenic capacity of the vessel. This epigenetic effect could impact both endogenous collateralization and treatment of arterial occlusive diseases.

Clinical Perspective

What Is New?

Collateral arterioles exposed to an increase in shear stress without a change in flow direction exhibit constrained arteriogenesis, an augmented shear stress set point, and DNA hypermethylation.

In vitro, endothelial cells exposed to an increase in shear stress without a change in flow direction exhibit increased DNA methyltransferase 1 (DNMT1) expression, hypermethylation of gene promoters, and a DNMT1‐dependent reduction in monocyte adhesion.

Pharmacological inhibition of DNMT1 in vivo restores arteriogenic capacity and returns shear stress back to its original set point.

What Are the Clinical Implications?

Modulation of the shear stress set point by DNMT1 inhibition could represent a therapeutic strategy for treating arterial occlusive diseases.

By focusing on the molecular mechanisms regulating the maturation stage of arteriogenesis, as opposed to initiation and growth stages, such an approach could better account for the chronic nature of arterial occlusive diseases in humans.

DNMT1 inhibition may avoid the so‐called Janus phenomenon, which refers to the conundrum created by the fact that proarteriogenic therapies also tend to promote atherosclerosis.

Introduction

Collateral arteriogenesis, the growth of existing arterial vessels to a larger diameter, is a fundamental adaptive response that is often critical for the perfusion and survival of tissues downstream of chronic arterial occlusion(s). Arterial occlusion(s) create steep pressure gradients and increased flow along collateral arterial pathways bypassing the occlusion(s). The resulting increase in shear stress acting on the endothelium initiates a highly coordinated signaling cascade, ultimately resulting in the outward growth of the collateral vessel.1, 2 Outward luminal growth is hypothesized to continue until normalization to the original shear stress level (ie, the shear stress “set point”) has been achieved.2, 3, 4, 5

In addition to shear stress magnitude, however, other hemodynamic factors can influence arteriogenesis. Indeed, we have recently demonstrated that collateral artery segments exposed to both a 2‐fold increase in shear stress magnitude and reversed flow direction (ie, “reversed” flow segments) following femoral artery ligation (FAL; flow direction) in mice exhibit amplified arteriogenesis, whereas segments experiencing just a 2‐fold increase in shear stress magnitude (ie, “nonreversed” flow segments) exhibit a substantially more constrained extent of arteriogenesis.6 Furthermore, this difference in arteriogenic capacity was maintained up to 12 weeks following FAL. Importantly, these collateral artery segments start at the same basal diameter and shear stress magnitude before FAL, suggesting that shear stress set point may be altered because of differential hemodynamics post FAL.

Although critical in determining ultimate arteriogenic capacity, the molecular mechanism(s) involved in establishing and maintaining the shear stress set point remain unknown. Epigenetic mechanisms, such as DNA methylation, histone modifications, and noncoding RNA regulation, could be a way for local hemodynamics to regulate long‐term gene expression changes.7 Of these epigenetic mechanisms, DNA methylation is considered the most stable.8, 9 DNA is methylated at a cysteine base pair, most often at a CpG dinucleotide (CpG site).9 Methylation of CpG sites in the promoter region of a gene is commonly associated with repression of gene expression.10, 11, 12 DNA methylation occurs through the activity of DNA methyltransferases (DNMTs), particularly DNMT1 postdevelopment.13, 14, 15 Recently, DNA methylation has been shown to differentially regulate flow‐mediated endothelial gene expression through DNMT1,16, 17, 18, 19 although the role of DNA methylation in the regulation of arteriogenesis has not been explored. Here, using our previous observations of differential arteriogenic capacity within collateral artery segments in vivo as a basis, we tested the central hypothesis that an augmented shear stress set point constrains arteriogenic capacity via DNMT1‐dependent endothelial cell (EC) DNA hypermethylation.

Materials and Methods

The authors declare that all supporting data are available within the article and its online supplementary files.

Mice

All animal protocols were approved by the Institutional Animal Care and Use Committee at the University of Virginia (protocol 3814) and conformed to all regulations for animal use outlined in the American Heart Association Guidelines for the Use of Animals in Research. Male C57BL/6 mice and Balb/c were purchased from Charles River Laboratory. All animals were housed in the animal facilities at the University of Virginia. C57BL/6 mice were used for all studies unless otherwise noted.

FAL Model

We used a previously detailed FAL scheme6, 20, 21 that produces consistent arteriogenesis in the collateral arteries of the gracilis adductor muscles,6, 22, 23, 24, 25, 26, 27 along with minimal heterogeneity in the baseline collateral structure and known changes in flow direction from baseline. Male mice (10 to 12, 20 to 21, or 70 to 80 weeks of age for C57BL/6, Balb/c, and aged Balb/c studies, respectively) were anesthetized (intraperitoneal [IP] 120 mg/kg ketamine, 12 mg/kg xylazine, and 0.08 mg/kg atropine), depilated, and prepped for aseptic surgery. On the left leg, an incision was made directly above and along the femoral artery, which was gently dissected from the femoral vein and nerve between the bifurcation of the superior epigastric artery and popliteal artery. Two 6.0 silk sutures were placed immediately distal to the epigastric artery, which served as the origin of the muscular branch artery in all mice, and the artery segment between the 2 ligatures was then severed with microdissecting scissors. The surgical site was then closed with 5.0 prolene sutures. A sham surgery, wherein the femoral artery was exposed but not ligated, was performed on the right hindlimb (ie, on the other leg). Animals received 1 injection of buprenorphine for analgesia at the time of surgery and a second dose 8 to 12 hours later.

Muscular Branch Ligation Model

Male mice were anesthetized (IP 120 mg/kg ketamine, 12 mg/kg xylazine, and 0.08 mg/kg atropine), depilated, and prepped for aseptic surgery. On the left leg, an incision was made directly above and along the femoral artery. The muscular branch artery was gently dissected from the paired vein and ligated with one 6.0 silk suture just distal to the epigastric artery. The surgical site was then closed with 5.0 prolene sutures. A sham surgery, wherein the muscular branch artery was exposed but not ligated, was performed on the right hindlimb (ie, on the other leg). Animals received 1 injection of buprenorphine for analgesia at the time of surgery and a second dose 8 to 12 hours later.

5AZA Treatment

5‐Aza‐2′‐deoxycytidine (5AZA) (#A3656; Sigma) was reconstituted in dimethyl sulfoxide (DMSO) to a stock concentration of 0.25 mg/μL. Each day immediately before use, stock 5AZA was diluted to 1.25 μg/μL and DMSO to 0.01% DMSO in sterile saline. Both solutions were then passed through a 0.22‐μm sterile syringe filter. Mice were treated daily with an IP injection of 0.1 mg/kg 5AZA in sterile saline or 0.01% DMSO in a total volume of 100 μL.

Laser Doppler Perfusion Imaging

For monitoring blood flow recovery and postsurgical ischemia, mice were anesthetized via 1.5% isofluorane under constant oxygen. Mice were placed in a prone position and the soles of the feet were scanned (PeriCam PSI, PeriMed). Mean foot perfusion was used to calculate relative perfusion ratio (ligated/unligated).

Quantification of Global DNA Methylation by HRM

Methylation of genomic repeat elements, such as long interspersed nuclear element‐1 (LINE1), have been used as markers of global genomic DNA (gDNA) methylation.8, 28, 29 LINE1 methylation in the peripheral blood was therefore used as an indicator of the efficacy of our 5AZA treatment protocol on DNA methylation in vivo. Peripheral blood (100–150 μL) was collected retro‐orbitally with heparinized capillary tubes from mice 7 days after beginning daily 5AZA or DMSO IP injections. gDNA was immediately extracted using the Quick‐gDNA MiniPrep Kit (#D3006; Zymo Research) and gDNA (80 ng) underwent bisulfite conversion using the EZ‐DNA Methylation‐Gold kit (#D5005; Zymo Research) according to manufacturer's instructions. Polymerase chain reaction (PCR) and high‐resolution melting (HRM) analysis were adapted from a previously determined protocol.28 Briefly, modified DNA was diluted to 10 ng/μL with nuclease‐free water. A 20‐μL reaction mix of 20 ng bisulfite modified DNA and a final concentration of 1x EpiTect HRM Master Mix (#59445; Qiagen), 0.75 μmol/L LINE1 forward primer (5′‐GTTGAGGTAGTATTTTGTGTGGGTT‐3′), 0.75 μmol/L LINE1 reverse primer (5′‐TCCAAAAACTATCAAATTCTCTAACAC‐3′), and nuclease‐free water. PCR cycling conditions for LINE1 were 95°C for 5 minutes followed by 40 cycles of 95°C for 20 seconds then 55°C for 30 seconds then 72°C for 20 seconds. Melt analysis occurred from 60°C to 90°C rising by 0.1°C/5 s. PCR and HRM were performed on a CFX96 Real‐Time Detection System (Bio‐Rad) and each sample was run in duplicate.

Differences in DNA methylation as detected by HRM were quantified by the net temperature shift as previously calculated.28 Briefly, the net temperature shift was calculated as average distance between the normalized melt curves of experimental samples from a universal methylated positive control (#D5012; Zymo Research) where a more negative net temperature shift indicates a less methylated sample. For LINE1, the 2 normalization regions used were between 72°C and 73°C and 82°C and 83°C.

Tissue Harvesting for Whole‐Mount Vascular Casting and Cross‐Sectional Analysis

For analysis of luminal diameters in the gracilis collateral arteries and to enable sectioning at specific regions, vascular casting was performed using an opaque polymer that allows for accurate luminal diameter measurements.24 At the determined point of harvest after FAL, mice were anesthetized (IP 120 mg/kg ketamine, 12 mg/kg xylazine, and 0.08 mg/kg atropine) and euthanized via an overdose of pentobarbital sodium and phenytoin sodium (Euthansol, Virbac), and then the abdominal aorta was cannulated. The lower body was then perfused with 7 mL of 2% heparinized saline with 2 mmol/L adenosine (16404; Fisher Scientific) and 0.1 mmol/L papaverine (P3510; Sigma‐Aldrich) to clear and vasodilate the downstream vasculature at a constant rate of 1 mL/min (PHD2000; Harvard Apparatus). Once perfused, we waited 5 minutes to enable vasodilation. Tissues were then perfused with 3 mL of 4% paraformaldehyde solution (19943; Affymetrix) at 1 mL/min and allowed to fix for 10 minutes. The lower body was then perfused with 0.8 mL of Microfil casting agent (FlowTech, Inc) at a constant speed of 0.15 mL/min. Viscosity of Microfil was adjusted to minimize transport across capillaries. After curing for 1.5 hours at room temperature, gracilis muscles were dissected free and then cleared in 50% glycerol in PBS overnight. Cleared tissues were mounted between 2 coverslips using 500‐μm thick spacers (645501; Grace Bio‐Labs Inc) to keep constant thickness between muscles. Muscles were imaged using transmitted light at 4× magnification on a Nikon TE200 inverted microscope with a charge‐coupled device camera (Quantifier, Optronics Inc). Individual fields of view were montaged together (Photoshop CS2; Adobe Systems Inc).

For analysis of luminal diameters from intact gracilis collateral whole mounts (ie, vascular casting), collateral entrance regions were defined according to the following method. A cropped portion (560 μm×560 μm) of the montaged image (previously randomized and deidentified) was taken of the collateral artery at the first visible branch point of a terminal arteriole from the primary collateral as it extended from either the muscular branch or saphenous artery as previously described.6 After each cropped image region was taken, all images were randomized and deidentified. The mean diameter was then taken from 4 or 5 separate diameter measurements along the length of the cropped portion of the collateral artery.

After imaging, muscles were rehydrated, cut, and then paraffin embedded for cross‐sectional analysis at the muscular branch and saphenous artery entrance regions to the collateral arteries. Resulting cross‐sections were rehydrated and immunolabeled for 5‐methylcytosine (84 days post FAL) or hematoxylin‐eosin stained for collateral artery structure analysis (day 28 post FAL).

Immunofluorescence Labeling of 5‐Methylcytosine

Sections (5‐μm thickness) of paraffin‐embedded muscle from the muscular and saphenous regions were rehydrated and subjected to heat‐mediated antigen retrieval for 20 minutes in a citrate‐based antigen retrieval buffer (H‐3300; Vector Laboratories). After cooling, tissues were encircled with a hydrophobic barrier pen and blocked with PBS+0.1% saponin+ 2% bovine serum albumin (Jackson Immunoresearch) for 1 hour at room temperature. Tissues were then incubated overnight at 4°C with anti‐5‐methylcytosine (1:100; BI‐MECY, Eurogentec) and rat anti‐CD31 (1:75; SZ31, Dianova). Following primary antibody incubation, slides were washed in PBS then incubated with DRAQ5 (a nuclear marker), a donkey‐anti‐mouse Cy3 Fab (1:200; Jackson Immunoresearch), and a goat‐anti‐rat‐488 secondary antibody (1:200; Jackson Immunoresearch) for 1 hour at room temperature. Following incubation, slides were washed again in PBS, mounted with Prolong Gold (Life Technologies) to minimize photobleaching, allowed to cure overnight, and imaged using a Nikon TE2000 C1 laser scanning confocal microscope with a 20× oil objective. All settings were held constant throughout imaging. Cropped fields of view (200 μm×200 μm) encompassing the collaterals in each region were randomized and deidentified. The collateral diameter, nuclear area, raw integrated density, and nuclear raw integrated density of 5‐methylcytosine (5‐mC) within an individual cross‐section were determined in Fiji.30 For each mouse, mean collateral diameter, nuclear 5‐mC raw integrated density per nuclear area and 5‐mC positive area per total nuclear area was calculated from the average of the 2 primary gracilis collateral arteries, with 2 immunolabeled sections per collateral artery for a total of 4 images per mouse.

Cross‐Sectional Analysis of Collateral Artery Structure

Sections (5‐μm thickness) of paraffin‐embedded muscle from the muscular and saphenous regions were labeled for hematoxylin and eosin. Individual fields of view encompassing the collateral vessels were imaged with a 40× water objective on a Zeiss inverted microscope (Zeiss Axioskop) with a charge‐coupled device camera (Quantifier, Optronics Inc). All images were randomized and deidentified before analysis. Luminal diameter, wall area, and wall thickness were determined using Fiji30

Human Umbilical Vein Endothelial Cell Culture

Human umbilical vein ECs (HUVECs) purchased from VEC Technologies Inc. were thawed and maintained on 0.1% gelatin‐coated flasks in M‐199 medium (Lonza), supplemented with 10% fetal bovine serum (Life Technologies Inc), 100 U/mL penicillin‐G + 100 μg/mL streptomyocin (Life Technologies Inc), 2 mmol/L L‐glutamine (Life Technologies Inc), 5 μg/mL EC growth supplement (Biomedical Technologies), and 10 μg/mL heparin (Sigma‐Aldrich). HUVECs are both phenotypically consistent and one of the most extensively used cell culture models for study of flow‐mediated EC signaling, including numerous studies of flow responses of the arterial system17, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41 (eg, atheroprone versus atheroprotective waveforms) and DNMT1‐dependent DNA methylation.17, 18 For each set of experimental comparisons, cells were used from the same cell line between subculture passages 2 to 3.

In Vitro Exposure of Endothelial Cells to Biomimetic Shear Stress Waveforms

HUVECs were plated on cell culture–grade plastic dishes coated with 0.1% gelatin and grown to confluence. A cone and plate flow apparatus,39 which maintains cells at 5% CO2 and 37°C, was used to induce a shear stress protocol. The applied shear stress protocol consisted of a 24‐hour preconditioning period at a steady shear stress of 15 dyne/cm2, which was then either increased to 30 dynes/cm2 (nonreversed flow) or increased to 30 dynes/cm2 and reversed in direction (reversed flow) to simulate relative hemodynamics previously quantified in our in vivo FAL model.42 Fresh culture media consisting of M199 with 4% dextran from Leuconostoc spp (Sigma‐Aldrich, Mr ≈500 000), 2% fetal bovine serum, 100 U/mL penicillin‐G + 100 μg/mL streptomyocin, 2 mmol/L L‐glutamine, 5 μg/mL EC growth supplement, and 10 μg/mL heparin was added to cells before exposure to shear stress and was continuously exchanged throughout the duration in the cone and plate apparatus.

HUVEC RNA Isolation and Quantitative Reverse Transcriptase PCR

Total RNA was extracted with the PureLink total RNA purification system using the on‐column DNase protocol (Life Technologies Inc) according to manufacturer's instructions. RNA concentration and purity were determined with a NanoDrop spectrophotometer (Thermo Fisher Scientific) in duplicate. For quantitative reverse transcriptase PCR, 500 ng of total RNA was reverse transcribed using the iScript cNDA synthesis kit (Bio‐Rad). A reaction mixture of 12.5 ng of reverse‐transcribed cDNA, DNMT1 forward primer (TGCCAGCTGAGCGTGGTGGT), DNMT1 reverse primer (GCATGCGGGCAGCCACCAAT), and FastStart SYBR Green (Roche Applied Sciences) underwent quantitative reverse transcriptase PCR on a CFX96 Real‐Time Detection System (Bio‐Rad). Expression was normalized to β2‐microglobulin (forward 5′‐AGCATTCGGGCCGAGATGTCT‐3′, reverse 5′‐CTGCTGGATGACGTGAGTAAACCT‐3′), which is endogenously expressed and is not altered by many stimuli including shear stress.36 Normalized expression was quantified using the comparative 2ΔΔCt method.

RRBS and mRNA‐Seq

Total gDNA and total RNA were extracted from flow‐exposed HUVECs using the Quick‐gDNA MiniPrep Kit (#D3006; Zymo Research) and the Quick‐RNA MiniPrep Kit (#R1054; Zymo Research) according to manufacturer's instructions. Total gDNA and total RNA concentration and purity were determined with a NanoDrop spectrophotometer in duplicate. Both gDNA and RNA were isolated from the same plate of cells for each condition within an experiment. Purified gDNA and purified total RNA isolated from flow‐exposed HUVECs were pooled from 2 independent flow experiments. Pooled gDNA samples were sent to Zymo Research where DNA fragmentation, library preparation, bisulfite conversion, next‐generation sequencing, and bioinformatics were performed. Pooled RNA samples were also sent to Zymo Research where they performed mRNA sequencing (mRNA‐Seq). HiSeq 50 bp singleton reads from RNA‐Seq were first adaptor trimmed and then analyzed using TopHat and Cufflinks software. TopHat (version 2.2.0) was used for alignment of short read to the human genome hg19. Cufflinks (version 2.2.0) was used to transcript assembly and differential expression. CommeRbund (version 2.0.0) was used for visualization of differential analysis. Default parameters were used in all instances.

RRBS Analysis

Reduced representation bisulfite sequencing (RRBS) next‐generation sequencing reads were mapped to the “Feb. 2009 (GRCh37/hg19)” genome assembly by Zymo Research. The %CpG methylation was calculated as the percent of methylated CpG sites per total CpG sites in a given differentially methylated region (DMR) with ≥10x CpG coverage in a given DMR. Significance was determined using a Fisher exact test then applying a Benjamini‐Hochberg procedure to find false discovery rate. DMRs with ≥10× CpG coverage in their promoter regions (transcription start site ±1 kb) were considered significant if the false discovery rate was <0.1 and the absolute value of %CpG methylation of a DMR in N minus the %CpG methylation of a DMR in R was ≥10%. We then compared this list of significant DMRs with our mRNA‐Seq data set to determine genes with relative gene expression changes that correspond to their methylation status between shear stress conditions, ie, identify genes upregulated in N conditions that also have a significantly hypomethylated promoter region as well as genes that are downregulated in N conditions that have hypermethylated promoter regions compared with in R conditions. Only genes demonstrating this relative gene expression‐methylation correlation were used for gene ontology analysis using MSigDB43 from the Broad Institute.

Transillumination Laser Speckle Imaging and Shear Stress Analysis

Transillumination laser speckle imaging was performed as previously described.42 Briefly, 28 days after FAL, mice were anesthetized (IP 120 mg/kg ketamine, 12 mg/kg xylazine, and 0.08 mg/kg atropine), depilated, and prepped for aseptic surgery. On the left leg (ligated leg), an incision was made above and along the femoral artery such that a window of skin was dissected free and retracted directly above the superficial adductor muscles. Exposed tissue was superfused throughout the procedure and during imaging with a warmed solution of Tris‐CaCl2 (0.1 g/L CaCl2) with 2 mmol/L adenosine (16404; Fisher Scientific) and 0.1 mmol/L papaverine (P3510; Sigma‐Aldrich). To image the gracilis muscle, the mouse was placed supine on an intravital microscope stage (Zeiss Axioskop). A 30mW, 658‐nm laser diode (LPM658‐30; Newport Corporation) was coupled to a fiber optic cable and placed beneath the mouse in a transillumination orientation. A cooled, monochrome charge‐coupled device camera (Optonics Quantifier) was used to acquire the raw speckle images using a 4× air objective (Zeiss Acroplan LD NA=0.1). The objective and camera were chosen to ensure satisfaction of the Nyquist sampling criteria of at least 2 pixels per individual speckle.27 An objective mounted fiber optic light guide allowed for brightfield imaging to enable luminal diameter measurements (A08650; Schott Inc). For each field of view, a sequence of 20 12‐bit raw speckle images was acquired with a 5‐ms exposure time to capture average velocity over multiple cardiac cycles.

All processing of raw speckle images was performed using Fiji30 as previously described.42 Briefly, raw speckle images were converted to laser speckle flow index maps, removing any images with excessive motion artifact. To then account for the influence of whole background tissue variations, the processed flow images were normalized to median background intensity. Individual flow images were then merged into larger 2‐dimensional maps using Adobe Photoshop (CS2, Adobe Systems Inc). Finally, to allow for comparison of velocity change across experiments, vessel speckle intensity was normalized to the background tissue according to equation 1 to obtain the normalized speckle index.

Blood velocity analysis of laser speckle images was limited to defined muscular branch and saphenous collateral artery regions and assumed Poiseuille flow. The mean speckle shear rate in each region was calculated using the normalized speckle index and vessel diameter according to equation 2:

Small interfering RNA Transfection in HUVECs

Twenty‐four hours before exposure of HUVECs to flow conditions, HUVECs were plated without antibiotics on 0.1% gelatin‐coated plates in serum‐free M199 (Life Technologies) supplemented with 10% fetal bovine serum, 2 mmol/L L‐glutamine, 5 μg/mL EC growth supplement (Biomedical Technologies), and 10 μg/mL heparin (Sigma‐Aldrich). After cells were allowed to adhere for 2 hours after plating, cells were transfected with either 120 pmol of ON‐TARGETplus SMARTpool human DNMT1 small interfering RNA (L‐004605‐00‐0005; GE Dharmacon) or 120 pmol of ON‐TARGETplus nontargeting small interfering RNA (D‐001810‐10‐05; GE Dharmacon) in 52 μL of Oligofectamine transfection reagent (Life Technologies) and 6.8 mL Opti‐MEM media (Life Technologies) for 5 hours at 37°C. After 5 hours, plates were flooded with 8 mL of M199 media without antibiotics supplemented with 10% fetal bovine serum + L‐glutamine + 5 μg/mL EC growth supplement (Biomedical Technologies) and 10 μg/mL heparin (Sigma‐Aldrich). Twenty‐four hours posttransfection this solution was aspirated off and normal flow media was applied. Validation of transfection was performed on HUVEC plates (54 hours posttransfection) via DNMT1 Western blotting.

Western Blot Analysis

HUVECs were lysed in radioimmunoprecipitation assay buffer (Sigma‐Aldrich, #R0278) with protease inhibitor (Sigma, 1:50, #P8340). Samples were then cleared for 30 minutes at 4°C under constant agitation. Samples were centrifuged for 1 minute at 10 000g, the supernatant was collected, and a Pierce BCA assay (ThermoFisher Scientific, #23225) was used to determine total protein concentration. Samples were diluted 1:1 in 2× Laemmli sample buffer (Bio‐Rad, #1610737) with β‐mercaptoethanol (1:200) and boiled for 10 minutes. Equal protein was loaded onto a 10% SDS‐PAGE gel and blotted on a nitrocellulose membrane. After transfer, membranes were blocked for 1 hour at room temperature with Odyssey Blocking Buffer (LICOR, #927‐40000) and then incubated with primary antibodies overnight at 4°C. Western blots were performed by using primary antibodies directed against DNMT1 (Abcam, 1:1000, ab92314) and GAPDH (EMD Millipore, 0.0625 μg/mL, #AB2302). Secondary antibodies were purchased from LICOR and used at a 1:10 000 dilution. A LICOR Odyssey imager was used for blot image acquisition and densitometry analysis.

Monocyte Adhesion Functional Assay

Human‐derived monocytes (THP‐1 cell line) were purchased from the ATCC. Monocytes were unthawed and maintained in RPMI 1640 (Life Technologies, #11875‐093) + 10% fetal bovine serum (Life Technologies Inc) + 0.05 mmol/L β‐mercaptoethanol per ATCC culture instructions. Monocytes subcultured once cell density approached 800 000 cells/mL. Cells were used between passages 2 to 6.

Before the adhesion assay, cells were counted to obtain 3 000 000 cells per plate of HUVECs. Cells were pelleted, washed with PBS, pelleted, and then resuspended in serum‐free RPMI media at 1 000 000 cells/mL. Thawed calcein AM was added at 1 μg/mL and incubated with cells for 15 minutes at 37°C. After 15 minutes, the reaction was stopped by adding excess serum‐free RPMI to the cell solution then pelleted. Cells were washed once with serum‐free M199 media, pelleted, and then resuspended in serum‐free M199 at 500 000 cells/mL. Immediately following completion of flow exposure to HUVECs, flow media was removed by aspiration. HUVECs were quickly washed with serum‐free M199 media. This media was then aspirated off and 6 mL of serum‐free M199+monocytes (3 000 000 cells per plate) were added to and incubated with HUVECs for 30 minutes at 37°C. Following the 30 minutes, cells were washed twice with PBS to remove unbound monocytes. Adhered monocytes and HUVECs were fixed with 4% PFA for 10 minutes followed by 2 washes with PBS. Coverslips were mounted with Prolong Gold (Life Technologies). Plates were then imaged using a Nikon TE2000 C1 laser scanning confocal microscope. Randomly selected fields of view (8‐9) per condition for 3 independent experiments were obtained. Images were then deidentified and randomized in MATLAB. Images were converted to 8‐bit images, set to an equivalent threshold, and bound monocytes were quantified using Fiji's Analyze Particles tool (20 μm2 minimum particle size). Results were centered on the mean of all conditions within each independent experiment.

Immunofluorescence Labeling of Pericollateral Mac3+ Cells

Sections (5‐μm thickness) of paraffin‐embedded muscle from the muscular and saphenous regions were rehydrated and subjected to heat‐mediated antigen retrieval for 10 minutes in a citrate‐based antigen retrieval buffer (Vector Laboratories; H‐3300). Slides were then quenched of endogenous peroxidase activity with a 30‐minute incubation in 3% H2O2, blocked, and labeled with rat‐anti‐Mac3 (1:100, M3/4 clone, 550292; BD Biosciences) overnight at 4°C. Slides were washed and incubated with a biotinylated sheep‐anti‐rat secondary antibody (Abcam, ab6851, 1:500) for 1 hour at room temperature. Slides were washed and incubated with an avidin‐biotin complex (Vectastain ABC solution, Vector Laboratories) for 30 minutes at room temperature. Slides were washed and incubated with a Tyramide Signal Amplification reagent (Perkin Elmer; 1:50) for 10 minutes at room temperature. Slides were washed and incubated with streptavidin‐488 (Life Technologies Inc; 1:500), Cy3‐anti‐SMA (1A4 clone, Sigma; 1:1000) and DRAQ5 (Thermo Scientific Inc; 1:1000). Slides were then mounted with Prolong Gold (Life Technologies Inc) to minimize photobleaching, allowed to cure overnight, and imaged using a Nikon TE2000 C1 laser scanning confocal microscope with a 20× oil objective. Cropped fields of view (512×512 pixels) encompassing the collaterals in each region were randomized and deidentified. The pericollateral region was outlined (25 microns around the vessel) and pericollateral Mac3+ nuclei were counted in Fiji.30

Statistical Analyses

All results are reported as mean±SEM, unless otherwise noted. All data for group comparisons were first tested for normality and equal variance; no significant deviations from these assumptions were found. Statistical significance was then assessed by Student t test or a 2‐way ANOVA followed by a Holm‐Sidak multiple comparisons test, unless otherwise noted (SigmaStat 3.5, Systat Inc). Significance was assessed at P<0.05.

Results

Collateral Artery Segments Exposed to an Increase in Shear Stress, Without a Change in Flow Direction, Exhibit DNA Hypermethylation, and Constrained Arteriogenic Capacity

Using an FAL model identical to that employed previously by our group to demonstrate differential arteriogenesis at either end of gracilis collateral arteries (Figure 1A), we first sought to determine whether these two collateral artery regions displayed differential EC DNA methylation. To this end, we immunolabeled collateral artery cross‐sections for 5‐mC, a marker of DNA methylation, along with EC (CD31) and nuclear (DRAQ5) counterlabels (Figure 1B). These cross‐sections confirmed our previous findings6 that muscular region (nonreversed flow) collateral artery segments grow to a smaller diameter when compared with collateral segments in the saphenous (reversed flow) region at 12 weeks post FAL (Figure 1C). Nuclear 5‐mC staining intensity (Figure 1D) and total 5‐mC+ staining area per nuclear area (Figure 1E) were significantly increased in muscular (nonreversed flow) segments compared with both saphenous (reversed flow) segments and unligated controls.

Figure 1

Gracilis collateral arteries exposed to a nonreversed increase in shear stress magnitude exhibit limited arteriogenic capacity and hypermethylated DNA. A, Schematic of the primary gracilis adductor collateral flow pathways after femoral artery ligation (FAL). Arrows indicate the direction and magnitude of blood flow pre (yellow) and post (white) FAL. The femoral artery is ligated just distal to the epigastric artery such that some collateral segments (“muscular”) experience a 2‐fold increase in shear stress magnitude (“nonreversed” flow), while other segments (“saphenous”) are exposed to both a 2‐fold increase in shear stress magnitude and reversed flow direction (“reversed” flow). B, Representative cross‐sections of gracilis collateral regions in C57BL/6 mice 12 weeks post FAL immunolabeled with 5‐methylcytodine (5‐mC, green), CD31 (endothelial cells, magenta), and DRAQ5 (nuclei, blue) (scale bar=50 μm). Yellow line indicates vessel wall. C–E, Bar graphs of luminal diameter, nuclear 5‐mC raw integrated density per total nuclear area, and fraction of 5‐mC+ area per total nuclear area for both muscular and saphenous regions 12 weeks post FAL (n=6–7). *P<0.05, 2‐way ANOVA followed by a Holm‐Sidak multiple comparisons test. Data are mean±SEM. n.s. indicates not significant.

ECs Exposed to a Nonreversed Increase in Shear Stress Magnitude Exhibit Augmented DNMT1 Expression

To further investigate the influence of these FAL‐elicited hemodynamic changes on EC DNA methylation, HUVECs were exposed to flow waveforms biomimetic of those experienced by collateral arteries following FAL in vivo42 (Figure 2A). Briefly, ECs were preconditioned for 24 hours at 15 dynes/cm2 to establish basal EC alignment and planar cell polarity, thereby mimicking the in vivo baseline state. An FAL was then simulated by a step‐wise 100% increase in shear stress, in either the same direction or in the opposite direction, to mimic shear stress changes occurring in the muscular branch (nonreversed flow) and saphenous artery (reversed flow) entrance regions, respectively (Figure 2A). We examined DNMT1 mRNA expression by quantitative reverse transcriptase PCR 1 hour and 6 hours after our simulated FAL, determining it was transiently increased by ≈25% after 1 hour in HUVECs exposed to nonreversed flow, but was unchanged in reversed flow conditions (Figure 2B). DNMT1 mRNA expression returned to basal level by 6‐hours after simulated FAL (Figure 2B).

Figure 2

Endothelial cells exposed to a nonreversed increase in shear stress magnitude exhibit augmented DNA methyltransferase 1 (DNMT1) expression and altered genome‐wide DNA methylation patterns in vitro. A, Schematic depicting biomimetic shear stress conditions applied to human umbilical vein endothelial cells to simulate nonreversed/muscular (N, blue) and reversed/saphenous (R, orange) regions. B, Bar graph of DNMT1 mRNA expression in each flow condition 1 hour or 6 hours after simulated femoral artery ligation (FAL) (n=8 for 1 hour and n=6 for 6 hours). *P<0.05 between reversed and nonreversed within a time point, Student t test. Data are mean±SEM. C through D, Scatter plot and Volcano plot of all differentially methylated regions in a gene promoter region (17 227 total). Significant differentially methylated regions were designated as hypomethylated (blue, 213 total) or hypermethylated (red, 603 total) with respect to the nonreversed condition (Table S3). E through F, Using our mRNA‐sequencing data set, we further filtered this list of significantly hypermethylated and hypomethylated genes (816 total genes) to contain only those with gene expression changes between nonreversed and reversed conditions in the expected direction based on their change in promoter methylation, ie, genes that were downregulated and had a hypermethylated promoter (red, 250 genes) or were upregulated and had a hypomethylated promoter (blue, 127 genes) in nonreversed vs reversed conditions (Tables S4 and S5). FDR indicates false discovery rate.

Genome‐Wide DNA Methylation Patterns Are Altered in ECs Exposed to Biomimetic Arteriogenic Flow Waveforms In Vitro

To then determine how these biomimetic waveforms affect global DNA methylation patterns, we exposed HUVECs to these same flow waveforms. Six hours after simulated FAL, we isolated both gDNA and total RNA and performed both RRBS and mRNA‐Seq on these samples, respectively. Both data sets were mapped to the hg19 human genome assembly (GRCh37/hg19, NCBI, Feb. 2009) and showed a similar degree of coverage between our nonreversed and reversed data sets (Tables S1 and S2). In addition, there was a similar degree of total CpG and promoter CpG coverage in both nonreversed and reversed data sets (Figure S1).

To characterize global DNA methylation changes between nonreversed (N) and reversed (R) data sets in genomic regions, we analyzed the CpG methylation in promoter (transcription start site ±1 kb), exon, and intron regions. From our RRBS analysis, only read regions with at least 10x CpG read coverage (henceforth termed DMRs) and mRNA expression in both nonreversed and reversed data sets were selected for further analysis. Intron regions displayed a higher degree of CpG methylation compared with exon or promoter regions; however, exon regions displayed the highest mean methylation density within a DMR, consistent with a previous study19 (Figure S2A and S2B). Average methylation across gene regions was similar for both nonreversed and reversed data sets. However, when we considered only significantly different (false discovery rate <0.1) DMRs, we observed global hypermethylation in nonreversed conditions compared with reversed conditions, across all gene regions (Figure S2C and S2D). Expectedly, we observed that the degree of promoter methylation inversely correlated with raw gene expression levels on a global scale (Figure S3).

As numerous studies have shown that DNA methylation in the promoter region regulates transcription44, we focused on DNA methylation differences within gene promoters. We found that 4.74% (816/17 227) of DMRs in promoter regions have a ≥|10%|difference in CpG methylation and a false discovery rate <0.1 between nonreversed and reversed conditions (Table S3). Of these 816 genes, 73.9% (603/816) were hypermethylated in nonreversed compared with reversed conditions (Figure 2C and 2D, red). To determine which mechanosensitive genes demonstrate a correlation between relative gene expression and promoter DNA methylation status, we further filtered this list of 816 genes to contain only genes with expression changes in the expected direction based on their change in promoter methylation (ie, genes that were downregulated and had a hypermethylated promoter in nonreversed versus reversed conditions and vice versa; Figure S4 and Table S4). We found that 66.3% of these genes (250/377) were hypermethylated and downregulated compared with 33.7% (127/377) that were upregulated and hypomethylated in nonreversed versus reversed conditions (Figure 2E and 2F, Tables S4 and S5). We then performed gene ontology analysis on these 377 genes to identify overrepresented pathways (Figure S5 and Table S6). Cellular metabolism, nucleic acid metabolism, and transcription processes were among the most significantly regulated pathways. However, a number of additional pathways were overrepresented, including protein metabolism, MAPK signaling, apoptosis, cellular localization and transport, and signal transduction.

DNMT1 Regulates the Adhesion of Monocytes to Endothelial Cells Exposed to a Nonreversed Increase in Shear Stress Magnitude

We next sought to determine whether DNMT1 regulates monocyte adhesion to ECs, which is a required step in the arteriogenesis cascade.45, 46, 47, 48, 49, 50 HUVECs were transfected with DNMT1 small interfering RNA or nontargeting control (siC) and subjected to the biomimetic flow waveforms. Western blot analysis showed a dominant band at the expected full length molecular weight ≈183 kDa for DNMT1 as well as a weaker, lower molecular weight band. This weaker band is approximately the same molecular weight (≈144 kDa) as the spliced isoform of DNMT1, which is missing amino acids 1‐336. DNMT1 expression was increased >40% in HUVECs exposed to the nonreversed flow waveform when compared with HUVECs exposed to the reversed flow waveform in siC‐treated conditions (Figure 3A). This corresponded to a 60% reduction in monocyte adhesion (P=0.023) to HUVECs exposed to the nonreversed flow waveform (Figure 3B and 3C). Knockdown of DNMT1 significantly (P=0.002) increased monocyte adhesion to HUVECs exposed to the nonreversed waveform compared with siC, while there was no significant effect on HUVECs exposed to the reversed waveform (Figure 3B and 3C).

Figure 3

DNA methyltransferase 1 (DNMT1) regulates the adhesion of monocytes to endothelial cells exposed to a nonreversed increase in shear stress magnitude in vitro. A, Relative DNMT1 protein expression, normalized to GAPDH, in human umbilical vein endothelial cells transfected with either DNMT1 small interfering RNA (siDNMT1) or nontargeting control (siC) and exposed to the nonreversed (N) or reversed (R) biomimetic shear stress waveforms in Figure 2, 6 hours after simulated femoral artery ligation (n=4). *P<0.05, 2‐way ANOVA followed by a Holm‐Sidak multiple comparisons test. B, Representative confocal microscopy images of fluorescently labeled THP‐1 monocytes (white) adhered to flow‐exposed human umbilical vein endothelial cells (scale bar=100 μm). Insets are the magnified 300 μm×300 μm regions outlined by white boxes. C, Bar graph of the relative number of adhered monocytes in each condition (n=3). *P<0.05, 2‐way ANOVA followed by a Holm‐Sidak multiple comparisons test. Data are mean±SEM. n.s. indicates not significant.

DNMT1 Inhibition Restores the Arteriogenic Capacity of Nonreversed Flow Collateral Artery Segments and Improves Perfusion Recovery in Aged Mice

Our observations led us to test the hypothesis that arteriogenic capacity can be rescued in nonreversed collateral artery segments by reversing DNA hypermethylation through DNMT1 inhibition. As outlined in Figure 4A, we performed FALs on C57BL/6 mice and allowed them to recover for 2 weeks, which is sufficient time for collaterals to achieve steady‐state diameters in this model.6 We then treated mice with daily IP injections of 0.1 mg/kg 5AZA or 0.1% DMSO vehicle control for an additional 2 weeks (Figure 4A). 5AZA is a nucleoside analog that preferentially targets DNMT1 via ubiquitin‐dependent proteasomal degradation.51 5AZA treatment was shown to be effective in reducing global DNA methylation after only 1 week by HRM (Figure 4B).

Figure 4

The arteriogenic capacity of nonreversed collateral artery segments is restored by DNA methyltransferase 1 inhibition. A, Experimental treatment time course. Femoral artery ligation (FAL) is performed on day 0. On day 14, mice begin receiving daily intraperitoneal injections of dimethyl sulfoxide (DMSO) or 5‐aza‐2′‐deoxycytidine (5AZA) until day 28. B, Bar graph of normalized temperature shift (NTS) determined by high‐resolution melting (HRM) of LINE1 repeat elements after 7 days of treatment (n=6). *P<0.05 between DMSO and 5AZA, Student t test. C, Representative vascular cast images from muscular (nonreversed) and saphenous (reversed) collateral artery regions 28 days post FAL from C57BL/6 mice treated according to (B) (scale bar=50 μm). D, Bar graph of regional luminal diameter in DMSO‐ or 5AZA‐treated mice. *P<0.05; # P<0.01 between ligated and unligated within the given region. Two‐way ANOVA followed by a Holm‐Sidak multiple comparisons test. E, Representative hematoxylin‐eosin–stained cross‐sections of gracilis collateral artery regions in DMSO‐ and 5AZA‐treated mice at day 28 post FAL (scale bar=25 μm). Microfil casting material (M) is evident in the artery lumen. F, Bar graph of wall area ratio of saphenous:muscular collateral regions. # P<0.01 between ligated and unligated for a given treatment; *P<0.05 between DMSO‐ and 5AZA‐treated mice; 2‐way ANOVA followed by a Holm‐Sidak multiple comparisons test. Data are mean±SEM. n.s. indicates not significant.

Vascular casting was used to determine collateral artery diameter in both muscular (nonreversed flow) and saphenous (reversed flow) regions 28 days post FAL. Consistent with previous results (Figure 1C and Heuslein et al6), we observed limited arteriogenesis in muscular (nonreversed flow) compared with saphenous (reversed flow) collateral artery segments in DMSO‐treated vehicle control mice. However, DNMT1 inhibition increased the arteriogenic capacity of nonreversed flow collateral segments by >40%, while there was no significant (P=0.33) effect on reversed flow segments (Figure 4C and 4D). Cross‐sections were used to determine collateral wall area, a metric that further indicated that the differential arteriogenic capacity along the collateral length in DMSO‐treated mice was normalized by DNMT1 inhibition (Figure 4E and 4F). Of note, we observed similar results in FAL‐treated Balb/c mice (Figure S6). Here, DNMT1 inhibition increased arteriogenic capacity by ≈44% in nonreversed segments, while there was no effect on reversed flow segments (P=0.163), indicating that this response is not limited to the C57BL/6 strain. We also sought to determine whether this increased arteriogenic capacity in nonreversed flow segments following DNMT1 inhibition corresponded to altered macrophage recruitment, a necessary component of collateral artery growth45, 46, 47, 48, 49, 50, 52 in vivo. Nonreversed flow collateral segments exhibited a trend (P=0.057) toward increased pericollateral Mac3+ macrophages in 5AZA‐treated mice (day 17 post FAL, 3 days of 5AZA treatment), corresponding to our previous in vitro results (Figure S7). There was no difference in macrophage recruitment in reversed segments between 5AZA‐ and DMSO‐treated mice (Figure S7).

We next sought to determine whether DNMT1 inhibition could improve perfusion recovery. As young C57BL/6 mice do not exhibit a significant long‐term perfusion deficit (ie, they fully recover in ≈7 days following FAL),6 we chose to used aged (10 to 11 months old) Balb/c mice instead. We hypothesized that these mice would have a poorer baseline perfusion recovery, thus enabling the necessary dynamic range to test whether DNMT1 inhibition altered perfusion recovery. Mice were subjected to FAL, allowed to recover for 2 weeks and then treated daily with either 5AZA or DMSO vehicle control for an additional 2 weeks, according to Figure 4A. As expected, FAL‐operated limbs of DMSO‐treated mice exhibited impaired perfusion recovery, only reaching ≈73% of unligated limb blood flow (Figure S8). There was no difference in perfusion between day 14 and day 28 post FAL in control‐treated mice (P=0.441), indicating that perfusion had reached steady state by day 14. Despite beginning at a similar degree of foot perfusion at day 14 (ie, start of 5AZA treatment), DNMT1 inhibition via 5AZA significantly (P=0.04) improved perfusion recovery (≈90% of unligated limb) compared with DMSO‐treated mice (Figure S8).

Shear Stress Set Point is Reestablished in Nonreversed Flow Collateral Artery Segments by DNMT1 Inhibition

Finally, we sought to determine whether DNMT1‐dependent DNA hypermethylation alters nonreversed collateral artery shear stress set point. Mice were treated with 5AZA or DMSO according to Figure 4A. Relative changes in collateral artery shear rates were then determined by transillumination laser speckle flowmetry42 28 days post FAL. Interestingly, in DMSO‐treated vehicle control mice, shear stress remained elevated (≈2.5‐fold) in muscular (nonreversed flow) collateral segments 28 days after FAL, while it was restored to pre‐FAL levels in saphenous (reversed flow) segments (Figure 5). DNMT1 inhibition restored shear stress in nonreversed flow segments to pre‐FAL levels, whereas there was no significant effect on reversed flow segments (Figure 5).

Figure 5

Shear stress set point in nonreversed flow collateral artery segments is restored by DNA methyltransferase 1 inhibition. A, Representative relative speckle velocity maps across the primary gracilis muscle collateral pathways 28 days post femoral artery ligation (FAL) in C57BL/6 mice treated with dimethyl sulfoxide (DMSO) and 5‐aza‐2′‐deoxycytidine (5AZA) according to Figure 4A. (Scale bar=500 μm). B, Greyscale images of the 300×200 pixel boxed regions shown in (A) for both the muscular (nonreversed, left) and saphenous (reversed, right) collateral regions. (Scale bar=100 μm). Dotted line indicates collateral artery region of interest used for analysis. C, Bar graph of regional speckle shear rate in DMSO‐ and 5AZA‐treated mice (n=6 for DMSO, n=5 for 5AZA). *P<0.05, n.s. (not significant). Student t test. Data are mean±SEM.

Discussion

In this study, we tested the hypothesis that DNMT1‐dependent EC DNA methylation regulates arteriogenic capacity via adjustments to shear stress set point. Previously, we demonstrated that collateral artery segments exposed to an increase in shear stress magnitude, without a change in flow direction, display limited arteriogenic capacity when compared with segments exposed to both increased shear stress magnitude and reversed flow direction. Here, we first determined that these nonreversed flow collateral segments exhibit global DNA hypermethylation in vivo. We then applied flow waveforms, biomimetic of those leading to either amplified arteriogenic capacity (ie, reversed flow) or constrained arteriogenic capacity (ie, nonreversed flow) in vivo, to ECs in vitro, and performed both RRBS and mRNA‐Seq. ECs exposed to the nonreversed waveform exhibited increased DNMT1 expression, genome‐wide hypermethylation of significantly regulated gene promoters, and a DNMT1‐dependent reduction in proarteriogenic monocyte adhesion. Together, this led us to next test whether DNMT1 regulates arteriogenic capacity in vivo. We ascertained that, in nonreversed flow collateral artery segments, DNMT1 inhibition rescued arteriogenic capacity and returned the elevated shear stress back to its original set point. Collectively, these results demonstrate that DNMT1‐dependent DNA hypermethylation constrains arteriogenesis by dampening EC mechanosensing, which effectively augments shear stress set point. The epigenetic regulation of shear stress set point may therefore have critical impact on both endogenous and therapeutic arteriogenesis in patients with arterial occlusive disease.

Mapping EC Mechanosensitive DNA Methylation to Differential Arteriogenic Capacity

The significance of epigenetics in vascular biology, with roles as regulators of molecular signaling known to drive physiology and as potential therapeutic targets to treat disease, is now well recognized.9, 15, 53 Both histone modifications and microRNAs regulate flow‐mediated EC gene expression54, 55, 56, 57, 58, 59, 60 and arteriogenesis61, 62, 63, 64, 65; however, DNA methylation has only recently been shown to regulate flow‐mediated EC gene expression in any context.16, 17, 18, 19 Moreover, to our knowledge, the role of DNMT1‐mediated DNA methylation in arteriogenesis has not been previously studied.

Our study directly maps EC mechanosensitive DNA methylation to differential, sustained arteriogenesis responses. In addition, by using both RRBS and mRNA‐Seq, we discovered a set of mechanosensitive genes whose expression correlates to gene promoter DNA methylation status. Gene ontology analysis of these genes identified a number of pathways crucial for EC mechanotransduction and arteriogenesis, including several metabolism, transcription, MAPK signaling, and cell transport pathways.31, 66 Of note, SIRT4 was involved in a number of these significantly overrepresented pathways (Figure S5). SIRT4 has been shown to disrupt the nuclear factor κB pathway, whereby overexpression of SIRT4 in ECs abrogates nuclear factor κB nuclear translocation and decreases expression of proinflammatory cytokines (interleukin 1β, interleukin 6, and interleukin 8), matrix metallopeptidase 9, and intercellular adhesion molecule 1.67 As we have previously reported, ECs exposed to nonreversed flow waveforms exhibit decreased nuclear factor κB–intercellular adhesion molecule‐1 activity.6 Thus, because the nuclear factor κB–intercellular adhesion molecule‐1 pathway is crucial for arteriogenesis,6, 45, 68 the flow‐dependent regulation of SIRT4 could be of particular interest.

In addition, studies examining flow‐mediated EC DNA methylation have identified Homeobox transcription factors (eg, HOXA5) as being differentially regulated in atheroprone conditions.17, 19 HOX transcription factors are considered “master regulators” as they regulate EC proliferation, migration, differentiation, morphogenesis, and permeability during development and vascular remodeling.69 Interestingly, we found HOXB3 to be among the genes downregulated (decreased 25%) and hypermethylated (20% versus 0.1% methylation) in nonreversed compared with reversed flow conditions (Table S4). As HOXB3 regulates EC activation and promotes angiogenesis,70 our results are consistent with the hypothesis that hypermethylation of the HOXB3 promoter decreases its expression, thereby limiting EC activation and arteriogenic potential of collateral artery segments.

The Role of DNMT1 in Flow‐Mediated Endothelial Inflammation is Dependent on Hemodynamic Context

Monocyte adhesion to an activated endothelium is required for collateral artery growth.45, 46, 47, 48, 49, 50, 52 Here, we employed a monocyte adhesion assay, which has been previously used to examine flow‐mediated EC function,6, 17, 40, 71 to determine the role of endothelial DNMT1 expression in regulating this essential step in arteriogenesis. Our results showed increased DNMT1 expression and limited monocyte adhesion to ECs exposed to the nonreversed flow waveform. Upon DNMT‐1 inhibition, monocyte adhesion was increased >2‐fold to ECs exposed to the nonreversed flow waveform. Corresponding to these in vitro results, DNMT1 inhibition in vivo via 5‐AZA increased pericollateral Mac3+ macrophages in nonreversed flow collateral segments (Figure S7). We observed no changes in monocyte adhesion or pericollateral macrophage accumulation in reversed flow conditions, indicating an anti‐inflammatory role for DNMT1 that is dependent on hemodynamic context. In contrast, DNMT1 has been shown to promote EC inflammation in HUVECs exposed to atheroprone flow conditions, as demonstrated by a DNMT1‐dependent increase in monocyte adhesion.17 However, our proarteriogenic flow conditions, which include a laminar flow preconditioning phase, are different from the oscillatory, atheroprone conditions of previous studies,17 further supporting the idea that DNMT1's role in endothelial inflammation is dependent on hemodynamic context. Furthermore, exposure to atheroprone conditions led to a chronic increase in DNMT1 expression,17 whereas our results suggest a transient increase in DNMT1 expression. This adaptive, instead of chronic, response may contribute to a context‐dependent role of DNMT1 on monocyte adhesion to ECs.

Molecular Regulation of Collateral Artery Shear Stress Set Point

Finally, we have determined that DNMT1‐dependent DNA methylation regulates, at least in part, long‐term arteriogenic capacity and shear stress set point. The concept of an arterial homeostatic wall shear stress magnitude (ie, shear stress set point) at which vessels maintain a steady‐state luminal diameter72 arises from Murray's principle of minimum work.73 Outward collateral artery growth is therefore hypothesized to stop once normalization to the shear stress set point has been achieved.2, 3, 4, 5 Premature normalization to the shear stress set point has been a predominant rationalization for the failure of collateral arteries to realize full arteriogenic capacity, frequently reaching only 30% to 40% of the maximal conductance.74 However, our results indicate that shear stress actually remains elevated in collateral artery segments exhibiting limited arteriogenic capacity. In essence, EC DNA hypermethylation prevents these collaterals from continuing to increase in diameter; therefore, shear stress remains chronically elevated. Yet, when DNMT1‐dependent DNA methylation is inhibited, these collaterals become resensitized to their elevated shear stress and are able to resume the arteriogenic process until the original set point is achieved (Figure 6). This response appears to require a basal level of DNA methylation, as there was no effect of DNMT1 inhibition on collateral artery segments that were not hypermethylated.

Figure 6

Summary of DNA methyltransferase 1 (DNMT1)–dependent regulation of arteriogenic capacity and shear stress set point in gracilis adductor collateral arteries following femoral artery ligation.

There are several other studies that have reported an altered set point following arterial adaptation. To this end, a left‐right carotid anastomosis was used to induce an acute increase in blood flow in the common carotid artery of mature and weanling rabbits. Two months later, shear stress remained augmented in mature rabbits caused by a lack of compensatory arterial enlargement, whereas weanling rabbits exhibited significant diameter enlargement, enabling for shear stress to normalize to the set point.75 This age‐dependent remodeling was also observed in rats in which ligation of the left internal and external carotid arteries increased right carotid blood flow by 46%. After 4 weeks, the right carotid outer diameter increased and shear stress returned to initial values in juvenile but not adult rats.76 The constrained arterial remodeling and augmented shear stress set point of aged animals is strikingly similar to the phenotype we observed in nonreversed flow collateral segments. Given that age alters DNA methylation,77 together, these results would be consistent with the hypothesis that acutely increased shear stress yields incomplete arterial remodeling and augmented shear stress set point caused by DNA hypermethylation.

Potential Clinical Implications

Given our results, modulation of the shear stress set point by DNMT1 inhibition could represent a therapeutic strategy for treating arterial occlusive diseases. By focusing on the molecular mechanisms regulating the maturation stage of arteriogenesis, as opposed to initiation and growth stages, such an approach could better account for the chronic nature of arterial occlusive diseases in humans. To this end, we did not begin DNMT1 inhibition until 2 weeks after FAL. We observed an increase in nonreversed collateral artery diameter in both C57BL/6 mice and Balb/c mice. Moreover, DNMT1 inhibition trended toward improved perfusion recovery in aged (10 to 11 months old) Balb/c mice (Figure S8). Although a few previous studies have demonstrated increased arteriogenic capacity after such a delayed treatment,46, 78, 79 ours is the first to demonstrate an epigenetic mechanism. DNMT1 inhibition may also be clinically advantageous because it appears to avoid the so‐called Janus phenomenon, which refers to the conundrum created by the fact that proarteriogenic therapies also tend to promote atherosclerosis.80 To this point, DNMT1 inhibition with 5AZA reverses DNA hypermethylation induced by atheroprone shear stress17, 18 and reduces atherosclerotic plaque size.17

Conclusions

Ultimately, because shear stress–induced changes in DNMT1 expression markedly affect both atherosclerosis and arteriogenic capacity, DNMT1 may represent an interesting target for peripheral arterial disease therapy.

Sources of Funding

This work was supported by National Institutes of Health R03 EB017927 and R01 EB020147. JLH was supported by a National Science Foundation Graduate Research Fellowship Program grant No. NSF DGE‐1315231.

Disclosures

None.

Table S1. Summary of Total Number of Reads, Mapping Ratio, and CpG Coverage in RRBS Data Sets

Table S2. Summary of Total Number of Reads, Mapped Reads, and Mapping Ratio for mRNA‐Sequencing Data Sets

Table S3. All DMRs Corresponding to the Promoter Regions of Genes With a Methylation Ratio Difference ≥|0.10| and FDR <0.1 Between Nonreversed and Reversed Conditions

Table S4. All Significantly Hypermethylated Promoter Regions Corresponding to Genes That are Downregulated in Nonreversed Conditions Compared With Reversed Conditions

Table S5. All Significantly Hypomethylated Promoter Regions Corresponding to Genes That are Upregulated in Nonreversed Conditions Compared With Reversed Conditions

Table S6. Top 50 Significantly Overrepresented Biological Process GO Terms for Genes With Expression Patterns Corresponding to Differential Promoter Methylation Under Arteriogenic Shear Stress Waveforms

Figure S1. Similar degree of methylation coverage in human umbilical vein endothelial cells (HUVECs) exposed to biomimetic shear stress waveforms.

Figure S2. DNA methylation across different gene regions.

Figure S3. DNA methylation negatively correlates to mRNA expression in gene promoter regions.

Figure S4. Genome‐wide mRNA expression patterns in human umbilical vein endothelial cells (HUVECs) exposed to arteriogenesis biomimetic shear stress waveforms.

Figure S5. Top 50 most significantly overrepresented gene ontology (GO) biological processes.

Figure S6. DNA methyltransferase 1 (DNMT1) inhibition also improves the arteriogenic capacity of nonreversed collateral artery segments in Balb/c mice.

Figure S7. Pericollateral Mac3+ macrophages increased in nonreversed collateral segments with inhibition of DNA methyltransferase 1 (DNMT1).

Figure S8. DNA methyltransferase 1 (DNMT1) inhibition leads to improved perfusion recovery in aged Balb/c mice.

Click here for additional data file.