Cardiac epigenetic changes in VEGF signaling genes associate with myocardial microvascular rarefaction in experimental chronic kidney disease

Abstract

Chronic kidney disease (CKD) is common in patients with heart failure and often results in left ventricular diastolic dysfunction (LVDD). However, the mechanisms responsible for cardiac damage in CKD-LVDD remain to be elucidated. Epigenetic alterations may impose long-lasting effects on cellular transcription and function, but their exact role in CKD-LVDD is unknown. We investigate whether changes in cardiac site-specific DNA methylation profiles might be implicated in cardiac abnormalities in CKD-LVDD. CKD-LVDD and normal control pigs (n = 6 each) were studied for 14 wk. Renal and cardiac hemodynamics were quantified by multidetector CT and echocardiography. In randomly selected pigs (n = 3/group), cardiac site-specific 5-methylcytosine (5mC) immunoprecipitation (MeDIP)- and mRNA-sequencing (seq) were performed, followed by integrated (MeDiP-seq/mRNA-seq analysis), and confirmatory ex vivo studies. MeDIP-seq analysis revealed 261 genes with higher (fold change > 1.4; P < 0.05) and 162 genes with lower (fold change < 0.7; P < 0.05) 5mC levels in CKD-LVDD versus normal pigs, which were primarily implicated in vascular endothelial growth factor (VEGF)-related signaling and angiogenesis. Integrated MeDiP-seq/mRNA-seq analysis identified a select group of VEGF-related genes in which 5mC levels were higher, but mRNA expression was lower in CKD-LVDD versus normal pigs. Cardiac VEGF signaling gene and VEGF protein expression were blunted in CKD-LVDD compared with controls and were associated with decreased subendocardial microvascular density. Cardiac epigenetic changes in VEGF-related genes are associated with impaired angiogenesis and cardiac microvascular rarefaction in swine CKD-LVDD. These observations may assist in developing novel therapies to ameliorate cardiac damage in CKD-LVDD.

NEW & NOTEWORTHY Chronic kidney disease (CKD) often leads to left ventricular diastolic dysfunction (LVDD) and heart failure. Using a novel translational swine model of CKD-LVDD, we characterize the cardiac epigenetic landscape, identifying site-specific 5-methylcytosine changes in vascular endothelial growth factor (VEGF)-related genes associated with impaired angiogenesis and cardiac microvascular rarefaction. These observations shed light on the mechanisms of cardiac microvascular damage in CKD-LVDD and may assist in developing novel therapies for these patients.

INTRODUCTION

Chronic kidney disease (CKD) is growing relentlessly in the United States and worldwide (1). This chronic disease represents an enormous burden to healthcare systems and is a hub for numerous cardiovascular complications, which increases the complexity of this disorder and in turn makes the development of comprehensive therapies challenging. Thus, understanding the pathophysiology of CKD and its consequences is still a central need to move the field forward.

CKD is an independent cardiovascular risk factor, which is mirrored by the growing prevalence of cardiovascular disease (CVD) and more than 50% of deaths of patients with CKD are due to cardiovascular diseases (1–3). Prior work shows that CKD pairs with left ventricular hypertrophy (LVH) and impaired relaxation (4, 5), resulting in LV diastolic dysfunction (LVDD) and a higher risk to develop heart failure (6–9). A robust and dynamic interaction between the kidney and the heart has been recently discussed (2), describing a fertile ground for studies to dissect and increase the pathophysiological knowledge of the precise biological mechanisms leading to cardiac abnormalities in CKD, which remains unclear.

Significant hurdles for the development of treatments are also nurtured by the need for translational models to increase mechanistic understanding, to identify new targets, and to support the development of new therapies (10). We recently developed a swine model of CKD with a significant loss of renal blood flow (RBF) and glomerular filtration rate (GFR) compatible with CKD stage 3, accompanied by albuminuria, fibrosis, and cortical and medullary microvascular rarefaction (11–13). The model also displays a characteristic cardiac phenotype of an early stage of HF, disclosed by LVH, abnormal LV strain, LVDD grade I (14–16) with preserved ejection fraction, and mild fibrosis (11, 17). This translational model mimics human CKD and LVDD and serves as a suitable platform to study potential mechanisms underlying cardiac pathophysiology in chronic renal disease.

Epigenetic changes, defined as modifications in gene expression that do not change the DNA sequence, have been implicated in the pathogenesis of renal and cardiac disease (18–20). Promoter and enhancer activities, which are central regulatory elements of gene expression, can be influenced by epigenetic changes such as methylation of the carbon-5 of cytosine (5mC) on DNA, a repressive epigenetic mark that deters the binding of transcription factors and may facilitate recruitment of corepressor complexes to methylated target promoters (21–23). Possibly, an altered cardiac microenvironment may favor hyper- or hypomethylation of cytosine and contribute to cardiac remodeling and abnormalities in CKD, which remains unknown.

In this study, we performed an unbiased analysis to compare the genomic-wide mapping of site-specific 5mC patterns in the hearts of normal and CKD-LVDD pigs using methylated DNA immunoprecipitation combined with deep sequencing (MeDIP-seq). We further investigated whether changes in cardiac methylation were associated with changes in the expression of genes implicated in the development of cardiac abnormalities in CKD-LVDD.

METHODS

The Institutional Animal Care and Use Committee at the University of Mississippi Medical Center approved all the studies. Twelve juvenile (3 mo old) domestic pigs (average weight of 30–35 kg) (Sus Scrofa Domesticus) of both sexes were studied for a total of 14 wk (Fig. 1). In six pigs (3 females, 3 males), anesthesia was induced with 0.25 g of intramuscular tiletamine hydrochloride/zolazepam hydrochloride (Telazol) and 0.5 g of xylazine and maintained with intravenous ketamine (0.2 mg/kg/min) and xylazine (0.03 mg/kg/min). CKD was induced by bilateral renal artery stenosis (24–26) combined with a 2% high-cholesterol atherogenic diet (24, 25) initiated on the day of induction of renal artery stenosis and maintained for 14 wk, as shown in previous studies (11, 17). Blood pressure was continuously measured by telemetry, as previously described (24–26). The remaining six pigs (3 females, 3 males) were used as normal time controls. Rimadyl 4 mg/kg was administered subcutaneously (pre-op) and then for 72 h (post-op, PO) for analgesia.

Figure 1.

Experimental design. Flow chart showing induction of CKD-LVDD and in vivo quantification of renal (multidetector CT, MDCT) and cardiac (echocardiography) hemodynamics after 14 wk of observation, as well as ex vivo studies after euthanasia to evaluate cardiac microcirculation and vascular endothelial growth factor (VEGF) signaling. CT, computed tomography; CVD, chronic kidney disease; LVDD, left ventricular diastolic dysfunction. n = 6/group; 3 females, and 3 males.

Fourteen weeks after induction of CKD, all animals were similarly anesthetized, intubated, and mechanically ventilated, and prepared for echocardiographic studies, as previously described (11, 17). Using a clinically available portable echocardiography machine (Philips EPIQ CVx Ultrasound system), we quantified LV ejection fraction (EF); LV wall thickness, LV mass index, and global longitudinal strain; left atrial diameter (LAD) and volume; and mitral E and A velocities, E/A and tissue doppler velocities, and E/e′ ratios to evaluate diastolic function, as shown previously (17, 27).

Immediately after completion of echocardiographic studies, the chest area was cleaned again using aseptic techniques and prepared for surgical access needed for multidetector computed tomography (MDCT)-derived quantification of renal function studies. Briefly, a vascular cut-down was performed to place 9-Fr vascular sheaths in the carotid artery and external jugular vein, respectively. Then, using fluoroscopy, an arterial guide was advanced to the renal arteries and the degree of renal artery stenosis was quantified using renal angiography, as shown previously (24–26). The arterial guide was then used for direct blood pressure recording, as described previously (17, 24–26). In addition, a 5-Fr pigtail catheter was placed in the right atrium for administration of a non-ionic low-osmolar contrast media using a power injector (20-mL bolus in 2 s) to perform in vivo helical MDCT quantification of RBF and GFR, as previously shown and validated (24–26, 28). Rimadyl (4 mg/kg) was administered subcutaneously (pre-op) and then for 72 h (post-op, PO) for analgesia.

A 3-day recovery period was selected to allow sufficient time for the animals to wash out the contrast agents administered during in vivo studies (MDCT imaging). Pigs were then euthanized (sodium pentobarbital, 100 mg/kg iv), and hearts were harvested for ex vivo studies to investigate cardiac microvascular density (micro-CT), and mRNA and protein expression of VEGF and signaling-related factors. Furthermore, in female normal and CKD-LVDD pigs (n = 3, each), unbiased cardiac MeDIP-seq and mRNA-seq analysis were performed and then followed by integrated (MeDiP-seq/mRNA-seq analysis) analysis.

MeDIP-Sequencing

MeDIP-seq was performed as previously described (29, 30). DNA was extracted from cardiac tissues using the DNeasy Blood & Tissue Kits (Qiagen, Cat. No. 69504) with RNase treatment following the manufacturer’s instructions. DNA was quantitated by Nano-drop instrument and diluted into a concentration of 100 ng/μL with TE buffer. The aliquot (100 μL) of diluted gDNA was sonicated using the Bioruptor Pico (Diagenode, Seraing, Belgium) for 7–10 cycles of 30 s on and 30 s off. The size of fragmented DNA was analyzed by the Fragment analyzer (Advanced Analytical Technologies, Ankeny, IA) using the High Sensitivity NGS Fragment Analysis Kit (Cat. No. DNF-486). Fragmented DNA with an average size of 200 bp was denatured at 95°C for 10 min. Then, 2.5–5 µg of DNA in 1× DIP buffer (10 mM sodium phosphate, pH 7.0, 140 mM NaCl, 0.05% Triton X-100) was incubated with 1 µg of anti-5mC antibody (Diagenode, Cat. No. C15200081, clone 33D3) generated from the hybridoma clone EDL HMC 1 A (Millipore, Cat. No. MABE1093) for 3 h at 4°C on a rotator. Protein G Dynabeads (Thermo Fisher, Cat. No. 10003 D) were added, and reactions were further incubated at 4°C on a rotator overnight. Beads-antibody-DNA complexes were extensively washed by DIP buffer and TE buffer, and enriched DNA fragments eluted from the beads, purified with the ssDNA/RNA Clean & Concentrator Kit (Zymo Research, Cat. No. D7010), and quantified using the Qubit ssDNA High Sensitivity Assay (Thermo Scientific, Cat. No. Q10212). Libraries were prepared (ACCEL-NGS 1S Plus DNA Library kit, Swift Bioscience, Cat. No. 10024) (29) following the manufacturer’s instructions and sequenced to 51 base pairs from both ends on an Illumina HiSeq4000 instrument in the Mayo Clinic Medical Genomics Facility.

Bioinformatic analysis was performed by aligning paired-end sequenced FASTQ files to the pig reference genome using bowtie2 2.3.3.1 (29). Duplicates were removed (PICARD 1.67, MarkDuplicates) and peaks were identified using MACS2 (31). Differential peak analysis was performed to determine sites of differential 5mC coverage using the DiffBind 2.14.0 application package (32) and the HOMER 4.10 (33) peak annotation tool. Subsequent 5mC coverage analysis applied per-base coverage of regions of interest (Bedtools 2.20.0, genomeCoverageBed) and sequence read values for the overall exonic 5mC coverage per gene calculated (htseq-count 0.9.1) (34). Faux-RNA counts of 5mC coverage were processed (edgeR 3.28.1) (35) to determine differences in read frequencies for genomic 5mC coverage at exons analogous to differential expression (DE) analysis of RNA reads. A volcano plot and heat maps of genes higher (fold change ≥ 1.4 and P ≤ 0.05) and lower (fold change ≤ 0.7 and P ≤ 0.05) 5mC levels in CKD-LVDD versus normal were generated using Excel (Microsoft) and Morpheus (https://software.broadinstitute.org/morpheus/), respectively. Gene ontology (GO) analysis of cellular components, molecular function, biological processes, and biological pathways of genes with higher or lower 5mC levels in CKD-LVDD versus normal hearts was performed using Functional Enrichment analysis tool (FunRich 3.1.4) (36) and categories were ranked based on the number of genes in overlap. In addition, 5mC profiles of specific candidate genes (XBP1 and XDH) were visualized using Integrative Genomics Viewer (IGV) (37).

Integrated MeDIP-Seq/mRNA-Seq Analysis

To explore whether CKD-LVDD-induced changes in VEGF-related genes were associated with changes in their mRNA expression, we performed an integrated (MeDIP-seq/mRNA-seq) analysis using our mRNA-seq data published previously (27).

RNA sequencing analysis was performed as previously described (38, 39). RNA libraries were prepared according to the manufacturer’s instructions (TruSeq RNA Sample Prep Kit v2, Illumina). The raw RNA sequencing paired-end reads for the samples were processed through the Mayo Clinic’s RNA-seq bioinformatics pipeline, MAP-RSeq version 3.1.4 (40). Briefly, MAP-RSeq employs a very fast, accurate, and splice-aware aligner, STAR (41), that aligns reads to the reference genome build hg38. The aligned reads were then processed through a variety of modules in a parallel fashion. Gene and exon expression quantification were performed using the Subread (42) package to obtain both raw and normalized by average Counts per Million mapped reads (CPM). STAR fusion algorithms (41) were used to identify and report any expressed gene fusions in the samples. Likewise, expressed single nucleotide variants (SNVs) and small insertions-deletions (indels) were detected using a combination of bioinformatics tools such as GATK (43), Haplotype caller (43), and RVBoost (44). Known and novel gene isoforms were assembled and quantified using StringTie (45) to enable the detection of alternatively spliced isoforms. In addition, differential exon usage was also evaluated using DEXSeq (46) to enable comparison across conditions for alternative splicing at the exon level. Finally, comprehensive analyses were performed on the aligned reads to assess quality of the sequenced libraries. Results from all modules described earlier were linked through a single document and reported by MAP-RSeq. Differential expression analysis was performed using the raw gene counts report from MAP-RSeq. Differentially expressed genes were identified using the bioinformatics package edgeR (35), reported along with their magnitude of change (log2 scale) and level of significance (FDR <5%).

To identify VEGF-related genes dysregulated at both epigenetic (MeDIP-seq) and expression (mRNA-seq) levels, mRNA expression of VEGF-related genes (KEGG_VEGF_SIGNALING_PATHWAY) with higher or lower 5mC levels in CKD-LVDD compared with normal were visualized in a volcano plot. VEGF-related genes with CPM > 0.1, fold change (CKD-LVDD/normal) ≥1.4, and P ≤ 0.05 were classified as upregulated, and those with CPM > 0.1, fold change (CKD-LVDD/normal) ≤0.7 and P ≤ 0.05 were classified as downregulated in CKD-LVDD versus controls. Interactions between VEGF signaling-related proteins were visualized using the Search Tool for the Retrieval of Interacting Genes (STRING) version 9.1 (http://string-db.org/). In addition, integrated (MeDIP-seq/mRNA-seq) analysis was performed to identify changes in fibrotic, extracellular matrix, calcium signaling, angiogenesis inhibitors, and pericyte genes [Kyoto Encyclopedia of Genes and Genomes (KEGG) gene lists] as described earlier.

Cardiac micro-CT Studies: Micro-CT Procedure

An intravascular silicon-polymer contrast agent (Microfil, MV-122, Flow Tech, Inc.) that remains in the intravascular compartment was used to perfuse the heart of CKD-LVDD and normal pigs through the cannulated left anterior descending coronary artery at a flow rate of 2 mL/min and physiological pressure. A transmural portion of the left ventricular myocardium (about 2 × 2 × 1 cm) was then sectioned, prepared, and scanned as described previously (47). Images were digitized for reconstruction of three-dimensional (3-D) volume images, which consisted of cubic voxels of 20 µm on a side and were displayed at 40-µm cubic voxels for subsequent analysis. Images analysis was performed with the Analyze software package (Biomedical Imaging Resource, Mayo Clinic, Rochester, MN). The myocardium was tomographically divided into 10 slices obtained at equal intervals. The outer portion of the myocardium was considered subepicardium, and the inner portion (about half) was considered subendocardium. Using the “Object Counter” function of Analyze, as shown (47), the spatial density of subepicardial and subendocardial microvessels (diameters: 0–100 and 0–500 µm) was calculated in each region.

Cardiac mRNA and Protein Expression Studies

To investigate the cardiac mRNA expression of VEGFA and genes with key roles in VEGF-related angiogenic signaling (KDR, NOS3, and AKT1), real-time qPCR was performed, as previously described (27). Briefly, cDNA was produced from cardiac tissue lysate according to the manufacturer’s instructions (PureLink RNA Mini Kit, Thermo Fisher Scientific) and PCR was performed to quantify cardiac expression of VEGFA and related genes (Thermo Fisher Scientific, assay ID Hs00900055_m1). Expression of mRNA was calculated relative to GADPH, β-actin, and TATA-binding protein (TBP) using the 2^−ΔΔCt method (13, 47). Results were expressed relative to normal controls. In addition, cardiac protein expression of VEGF was measured by Western blotting. In brief, frozen LV tissue was pulverized, homogenized in chilled protein extraction buffer, and the homogenate incubated in buffer. Homogenized lysates were then centrifuged, the supernatant removed, and protein concentration determined by spectrophotometry. The lysate was then diluted in polyacrylamide gel electrophoresis sample buffer, sonicated, heated, loaded onto a gel, and run using standard Western blotting protocols with antibodies against VEGF (No. MA1-16626, 1:200, Invitrogen, MA). The membrane was exposed to a chemiluminescence developing system and an X-ray film, and intensities of the protein bands were determined by densitometry and quantified in a blinded manner using ImageJ, as described previously (13, 47).

Statistical Analysis

Power calculations were performed and indicate that six animals per group were needed to detect biologically meaningful differences in renal hemodynamics (RBF), function (GFR), as well as echocardiographic and ex vivo (micro-CT) studies while controlling type I error rates at 5% using a two-sided Mann–Whitney–Wilcoxon test. Results are expressed as means ± SE. Comparisons between the groups were performed using paired Student’s t test with Bonferroni correction to counteract inflated type I errors while engaging in multiple pairwise comparisons. Statistical significance was defined as P ≤ 0.05.

RESULTS

General Parameters of CKD-LVDD and Normal Controls

Body weight was similar between the groups after 14 wk of observation, but blood pressure was higher in CKD-LVDD compared with controls (Table 1). All CKD-LVDD pigs developed significant stenosis. Cholesterol levels were higher in CKD-LVDD compared with normal pigs, whereas triglycerides, glucose, and insulin levels were similar between the groups. RBF and GFR were lower in CKD-LVDD compared with controls, accompanied by LVH, abnormal diastolic relaxation, and LV strain with preserved EF, underscoring successful development of CKD-LVDD. There were no differences in any of the parameters listed in Table 1 between male and female pigs (all P > 0.05).

Table 1.

General characteristics and cardiac parameters of normal and CKD-LVDD pigs

Parameter	Normal	CKD-LVDD
Body weight, kg	61.2 ± 6.5	68.5 ± 3.7
Mean arterial pressure, mmHg	101.3 ± 5.4	131.9 ± 5.8*
Degree of stenosis, %	0.0 ± 0.0	76.8 ± 10.3*
Cholesterol, mg/dL	92.2 ± 17.9	498.1 ± 47.7*
Triglycerides, mg/dL	7.8 ± 2.2	9.6 ± 8.5
Glucose, mg/dL	124.5 ± 5.4	122.0 ± 17.2
Insulin, pg/mL	0.08 ± 0.04	0.06 ± 0.02
RBF, mL/min	868.2 ± 51.3	491.8 ± 19.8*
GFR, mL/min	128.8 ± 1.8	71.1 ± 2.1*
LV wall thickness, cm	0.70 ± 0.11	1.34 ± 0.6*
Left atrial diameter, cm	3.58 ± 0.4	6.64 ± 1.23*
Ejection fraction, %	>50	>50
E/e′	6.20 ± 0.40	8.38 ± 1.26*
E/A	1.38 ± 0.05	1.0 ± 0.52*
LV longitudinal strain	−17.8 ± 0.08	−6.03 ± 3.59*

Values are means ± SE; n = 6 animals/group; 3 females, and 3 males. CVD, chronic kidney disease; GFR, glomerular filtration rate; LVDD, left ventricular diastolic dysfunction; RBF, renal blood flow. *P < 0.05 vs. normal.

Pigs with CKD-LVDD Show Altered Cardiac 5mC Levels

Unbiased MeDIP-seq analysis identified 261 genes with higher (fold change > 1.4, P < 0.05) and 162 genes with lower (fold change < 0.7, P < 0.05) 5mC levels in CKD-LVDD versus normal pigs (Fig. 2).

Figure 2.

Pigs with CKD-LVDD show altered cardiac 5mC levels. Volcano plot and heatmaps of genes with significant changes in cardiac 5mC levels in CKD-LVDD compared with normal controls. The y-axis corresponds to the mean expression value of −log 2 (P value), whereas the x-axis displays the log2 fold change (CKD-LVDD/normal) value. Genes with higher and lower 5mC levels are indicated with red and blue dots, respectively. A P < 0.05 and fold changes ≥ 1.4 and ≤ 0.7 are indicated by gray dashed lines. CVD, chronic kidney disease; LVDD, left ventricular diastolic dysfunction. n = 3 females/group.

Genes with Higher 5mC Levels in CKD-LVDD Are Primarily Implicated in Processes Associated with VEGF-Mediated Angiogenic Signaling

Genes with higher 5mC levels in CKD-LVDD encode for nuclear, cytoplasmic, and plasma membrane proteins with binding and transcription factor receptor activity, implicated in signal transduction and cell communication (Fig. 3). Biological pathways analysis revealed that these genes encode for proteins implicated in VEGF and VEGF-related signaling and angiogenesis, such as platelet-derived growth factor receptors (PDGFRs).

Figure 3.

Genes with higher 5mC levels in CKD-LVDD are primarily implicated in processes associated with VEGF-mediated angiogenic signaling. Genes were classified by cellular component, molecular function, biological processes, and biological pathways (FunRich, %). CVD, chronic kidney disease; LVDD, left ventricular diastolic dysfunction; VEGF, vascular endothelial growth factor. n = 3 females/group.

Genes with Lower 5mC Levels in CKD-LVDD Encode Primarily for Kinases Implicated in Several Cellular Signaling Processes

Genes with lower 5mC levels in CKD-LVDD also encode for cytoplasmic, nuclear, and plasma membrane proteins with transcription factor, receptor, and DNA activity, implicated in signal transduction, cell communication, and regulation of nuclear acid and protein metabolism (Fig. 4). Biological pathways analysis revealed that these genes also encode for proteins implicated in angiogenesis, including ErbB protein family, phosphoinositide 3 kinase (PI3K), and mammalian target of rapamycin (mTOR) signaling.

Figure 4.

Genes with lower 5mC levels in CKD-LVDD encode primarily for kinases implicated in several cellular signaling processes. Genes were classified by cellular component, molecular function, biological processes, and biological pathways (FunRich, %). CVD, chronic kidney disease; LVDD, left ventricular diastolic dysfunction. n = 3 females/group.

Epigenetic Changes in Vascular Endothelial Growth Factor-Related Genes Are Associated with Changes in Their mRNA Expression

Integrated MeDiP-seq/mRNA-seq analysis identified three VEGF-related genes (XDH, PTGS2, and MT3) with higher 5mC and lower mRNA levels in CKD-LVDD compared with normal pigs (Fig. 5; Table 2). Contrarily XBP1, DCN, and DDL4 exhibited lower 5mC and higher mRNA levels in CKD-LVDD versus normal pigs. Interaction analysis showed that proteins encoded by these genes have 71 known or predicted interactions with other proteins implicated in VEGF-related signaling (Fig. 6) with an average combined score of 0.59 ± 0.18, including interactions with VEGFA, FLT1, KDR, PDGFB, ANGPT1, and ANGPT2.

Figure 5.

Epigenetic changes in vascular endothelial growth factor (VEGF)-related genes are associated with changes in their mRNA expression. A: volcano plot showing mRNA expression of VEGF-related genes with higher or lower 5mC levels in CKD-LVDD compared with normal controls (downregulated genes, blue; upregulated genes, red). The y-axis corresponds to the mean expression value of −log 2 (P value), whereas the x-axis displays the log2 fold change (CKD-LVDD/normal) value. B: representative reads (Integrative Genomics Viewer) for XBP1 and XDH showing lower and higher 5mC levels in the hearts of CKD-LVDD vs. normal pigs. CVD, chronic kidney disease; LVDD, left ventricular diastolic dysfunction. n = 3 females/group.

Table 2.

Gene symbol, name, and function of VEGF-related genes with epigenetic and mRNA changes in CKD-LVDD hearts

Symbol	Name	Function
XBP1	X-Box binding protein 1	Regulates VEGF-mediated cardiac angiogenesis (PMC4933664).
DCN	Decorin	Downregulates VEGF expression (PMID: 24016866) and interferes with nitric oxide release (PMID: 18373940). Antagonist ligand of VEGFR-2 (PMC5417242).
DLL4	Delta-like canonical notch ligand 4	Negative regulator of VEGF-mediated angiogenic sprouting (PMC1805530, PMC4869683).
XDH	Xanthine dehydrogenase	Mediates VEGF-stimulated Akt phosphorylation and VEGF-induced endothelial cell survival (PMID: 18386220).
PTGS2	Prostaglandin-endoperoxide synthase 2	Promotes VEGF mRNA and protein production (PMC1505025, PMID: 14769824).
MT3	Metallothionein 3	Contributes to induction of vascular endothelial growth factor (PMID: 23962989).

CVD, chronic kidney disease; LVDD, left ventricular diastolic dysfunction; VEGF, vascular endothelial growth factor.

Figure 6.

Interactions between VEGF-related genes. Nodes represent genes and color lines the interactions according to the functional association network from the STRING database. Red and blue circles indicate genes (with higher or lower 5mC levels) upregulated and downregulated in CKD-LVDD compared with normal controls. CVD, chronic kidney disease; LVDD, left ventricular diastolic dysfunction; STRING, Search Tool for the Retrieval of Interacting Genes; VEGF, vascular endothelial growth factor.

Unlike VEGF-related genes, our integrated (MeDIP-seq/mRNA-seq) analysis identified that only few fibrotic, extracellular matrix, calcium signaling, angiogenesis inhibitors, and pericyte genes were dysregulated in CKD-LVDD compared with normal hearts (Supplemental Tables S1 and S2).

Impaired Cardiac Angiogenesis and Microvascular Rarefaction in CKD-LVDD

Cardiac expression of the VEGF-related genes VEGFA, KDR, NOS3, and AKT1 and VEGF protein expression were lower in CKD-LVDD compared with normal controls (Fig. 7A, Supplemental Figs. S1 and S2). Spatial density of subepicardial and, mostly, subendocardial microvessels (diameters: 0–100 and 0–500 µm) was lower in CKD-LVDD compared with normal pigs (Fig. 7B). There were no differences in cardiac angiogenic signaling or microvascular density between male and female pigs (all P > 0.05).

Figure 7.

Impaired angiogenesis and cardiac microvascular rarefaction in CKD-LVDD. A: cardiac VEGFA, KDR, NOS3, and AKT1 gene and VEGF protein levels (relative to GAPDH) were lower in CKD-LVDD compared with normal pigs. B: 3-D micro-CT images of the LV and quantification of subendocardial microvascular density showing decreased number of small (0–100 µm/diameter) and all-size (0–500 µm/diameter) microvessels in CKD-LVDD pigs compared with normal controls [n = 6/group; 3 females (red), and 3 males (blue)]. Comparisons between the groups were performed using paired Student’s t test with Bonferroni correction. CT, computed tomography; CVD, chronic kidney disease; LVDD, left ventricular diastolic dysfunction; 3-D, three-dimensional; VEGF, vascular endothelial growth factor.

DISCUSSION

Our study builds upon our early work that describes the cardiac phenotype (11, 17) and transcriptomic landscape (48) of a novel translational model of CKD and LVDD. The current study extends our previous observations and demonstrates that CKD-LVDD also exerts epigenetic (pretranscriptional) regulation of gene expression. Using MeDIP-seq and a subsequent integrated MeDIP-/mRNA-seq analysis, the current study shows that cardiac epigenetic changes in VEGF-related and angiogenic signaling genes associate with cardiac abnormalities in experimental CKD-LVDD. Specifically, accompanying LV remodeling and abnormal LV relaxation, we observed defective angiogenic signaling and significant subepicardial and, mostly, subendocardial microvascular rarefaction in the heart of CKD-LVDD pigs. Therefore, our work identified a novel potential mechanistic component that associates with LVDD in CKD and in turn may guide for the development of targeted interventions and studies to further elucidate cardiac pathophysiology in CKD-LVDD.

Epigenetic modifications are important for cell homeostasis by regulating dynamic expression of essential genes and consequently their protein products, and by repressing those that are not needed (18, 19). Epigenetic changes play an important role in the development and progression of CKD. Studies have shown that the renal cell epigenome might be altered by environmental modifiers, and may drive the onset and progression of renal diseases (18, 20). Similarly, prior work demonstrated that cardiac lineage commitment, cardiovascular disease, and development of cardiac fibrosis are tightly regulated by epigenetic mechanisms, with a prominent role in DNA methylation (49, 50), a powerful means for transcriptional silencing and inactivation of transposable elements. In line with these concepts, our research in CKD-LVDD identified a robust population of genes with significant changes in cardiac 5mC levels compared with controls, supporting the notion that epigenetic modifications might be associated with the development of cardiac abnormalities in our model. Interestingly, our unbiased gene ontology analysis indicates that among all genes mapped in our MeDIP-seq database, genes with higher 5mC levels in CKD-LVDD are primarily implicated in processes associated with VEGF-mediated angiogenic signaling, suggesting a repressing force on the VEGF pathway.

VEGF (mainly VEGF-A) is a prominent angiogenic cytokine with critical roles in proliferation, maintenance, and repair of microvascular networks in virtually every organ. In the heart, cardiomyocytes are both targets and sources of VEGF, and this growth factor plays several necessary roles that include vascular proliferation and repair, myocardial contractility, and wound repair (51), underscoring the importance of cardiac VEGF homeostasis. Furthermore, prior research shows that modulation of cardiac VEGF is important in the pathophysiology of HF and in compensatory mechanisms. Indeed, sequestration of VEGF in the heart has a direct impact on capillary density and development of fibrosis, whereas restoring cardiac VEGF stimulates cardiomyocyte regeneration and repair of the damaged heart (52–54).

Our study also found that genes with lower 5mC levels in CKD-LVDD encode primarily for kinases involved in diverse signaling processes, including ErbB2 signaling network genes, which have been implicated in regulation of angiogenesis and as VEGF promoters (55). Likewise, genes with lower 5mC levels are implicated in PI3K and mTOR signaling, which play an important role in regulating angiogenesis. VEGF via VEGF receptor (VEGFR)-1 stimulates the PI3K axis to promote vasculogenesis (56) Furthermore, PI3K acts in conjunction with the mTOR pathway to increase VEGF expression by upregulating hypoxia-inducible factor (HIF)-1α (57), creating a positive feedback mechanism to sustain VEGF expression. It is possible that apparent or potential unrepressed ErbB2, PI3K, and mTOR signaling in the heart of CKD-LVDD animals may serve as compensatory mechanisms for the repressed VEGF signaling and consequent development of cardiac abnormalities.

To determine whether CKD-LVDD-induced epigenetic changes were associated with altered gene expression, we performed an integrated MeDIP-/mRNA-seq analysis and identified three VEGF-related genes (XDH, PTGS2, and MT3) that exhibited higher 5mC and lower mRNA expression levels in CKD-LVDD versus normal hearts. XDH encodes for xanthine dehydrogenase, a molybdenum-containing hydroxylase that mediates VEGF-stimulated Akt phosphorylation and VEGF-induced endothelial cell survival (58). Similarly, proteins encoded by PTGS2, and MT3 contribute to induction of VEGF by promoting its mRNA and protein production (59–61), suggesting that DNA hypermethylation imposed by CKD-LVDD could have repressed the expression of these VEGF-related proangiogenic genes.

In addition, we found three VEGF-related genes (XBP1, DCN, and DDL4) that displayed lower 5mC and higher mRNA levels in CKD-LVDD versus normal pigs. Although XBP1 encodes a transcription factor that regulates VEGF-mediated cardiac angiogenesis (62) and preserves endothelial cell integrity (62), both DCN and DDL4 exert negative effects on VEGF signaling. DCN encodes a member of the small leucine-rich proteoglycan family of proteins that downregulates VEGF expression (63) and interferes with nitric oxide (NO) release (64). Furthermore, this protein is an antagonist ligand of VEGFR-2 (65), the primary functional VEGFR expressed by vascular endothelial cells. Likewise, DDL4 is induced by VEGF as a negative feedback regulator that prevents VEGF-mediated angiogenic sprouting (66, 67), suggesting that DNA hypomethylation could have partly accounted for the increased expression of anti-angiogenic genes in CKD-LVDD hearts.

Importantly, our protein interaction analysis identified numerous functional protein association networks among VEGF-related genes, including several interactions with proteins encoded by genes with significantly different 5mC and mRNA expression levels in CKD-LVDD versus normal hearts. These observations suggest that epigenetic and gene expression changes imposed by CKD-LVDD could have a direct impact on overall cardiac VEGF-related signaling. In line with this, we found that expression of VEGFA and KDR, which encode for VEGF and VEGFR2, respectively, were blunted in CKD-LVDD hearts, associated with decreased expression of NOS3 and AKT1. Endothelial nitric oxide synthase (eNOS) produces NO, a key mediator of VEGF-induced angiogenesis, whereas the protein product of AKT1 is a member of the human Akt serine-threonine protein kinase family that regulates angiogenesis by acting in concert with other cell signaling pathways, such as PI3K and mTOR (56, 57).

Finally, we found that protein expression of VEGF was lower in CKD-LVDD compared with normal hearts. We have previously shown that renal bioavailability of VEGF in CKD is reduced and that VEGF therapy protects renal microcirculation and recovers renal function (13). Our current data extend these findings and demonstrate that cardiac expression of VEGF is also blunted in CKD-LVDD, associated with decreased spatial density of subepicardial, and mostly, subendocardial microvessels. Thus, future studies applying VEGF therapy in the heart or targeting VEGF-related factors may dissect the role of blunted angiogenesis or vascular repair as novel mechanisms of LVDD pathophysiology and as potential new targets for treatment.

Limitations and Opportunities

Although we are aware that additional functional characterization (e.g., impaired exercise tolerance) and consideration for other models (e.g., cardiometabolic, pressure overload, and multihit models) might be needed to further define phenotypic traits of LVDD, the current data supports our model as a translational platform to study cardiac angiogenic mechanisms possibly implicated in the development of LVDD in CKD. We did not observe sex-based differences, which concurred with our recent studies (68), but our study might be limited by the relatively small number of animals and the fact that pigs were juvenile and more resilient to stressors. The number of samples for MeDIP- and mRNA-seq studies was modest (n = 3 each), as often used in seq studies (69–71), due to the costs associated with these techniques. For the MeDiP-seq and mRNA-seq analyses, only female normal and CKD-LVDD were selected. It is well known that cardiovascular disease develops more rapidly and becomes more severe in male than in female animals in several models of the disease (72, 73). Therefore, in the current study, we opted for using female pigs to test whether the deleterious effects of CKD-LVDD on myocardial epigenetic and gene expression outweigh this sex-specific protection. Future studies at more advanced ages, more severe CKD, and eventually larger sample sizes, will help to offset those limitations. Finally, additional experiments in vivo and likely ex vivo using other experimental platforms that allow manipulations of cardiac angiogenic and vascular remodeling pathways discussed in this manuscript might be needed in future studies to establish a cause-effect relationship between cardiac epigenetics and phenotype of this model, which in turn may help to develop novel therapeutic interventions to protect the heart in CKD.

Conclusions

In summary, our unbiased analysis compared the cardiac genomic-wide mapping of site-specific 5mC patterns between CKD-LVDD and normal pigs and found that cardiac epigenetic changes in VEGF-related genes are associated with impaired angiogenesis and cardiac microvascular rarefaction in swine CKD-LVDD. The current study extends our previous observations that CKD-LVDD induced posttranscriptional (miRNA) regulation of genes primarily implicated in processes associated with cardiac remodeling (27), demonstrating that CKD-LVDD also exerts pretranscriptional (epigenetic) regulation of gene expression. Unlike posttranscriptional regulation, epigenetic regulation was primarily directed toward genes implicated in VEGF-related angiogenic signaling. Therefore, our observations suggest that at least two distinct regulatory mechanisms of gene expression might contribute to cardiac damage in CKD-LVDD. These observations shed light on the mechanisms of cardiac microvascular damage in CKD-LVDD and may assist in developing novel therapies for these patients.

DATA AVAILABILITY

The raw mRNA- and MeDIP-seq data for this paper is available at: https://doi.org/10.6084/m9.figshare.20146703.v1. https://doi.org/10.6084/m9.figshare.20496786.v1.

SUPPLEMENTAL DATA

Supplemental Table S1: https://doi.org/10.6084/m9.figshare.21346530.v1.

Supplemental Table S2: https://doi.org/10.6084/m9.figshare.21346536.v1.

Supplemental Fig. S1: https://doi.org/10.6084/m9.figshare.21346482.v1.

Supplemental Fig. S1 Legend: https://doi.org/10.6084/m9.figshare.21346515.v1.

Supplemental Fig. S2: https://doi.org/10.6084/m9.figshare.21514275.v1.

Supplemental Fig. S2 Legend: https://doi.org/10.6084/m9.figshare.21514281.v1.

GRANTS

This work was supported by National Institutes of Health Grants R01HL095638, P20GM104357, DK129240, and R01HL158691 and Regenerative Medicine Minnesota Grant RMM 091620 DS 004.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

A.E. and A.R.C. conceived and designed research; performed experiments; analyzed data; interpreted results of experiments; prepared figures; drafted manuscript; edited and revised manuscript; approved final version of manuscript.