Perivascular adipose tissue inflammation in ischemic heart disease

Abstract

Objective:

There is growing recognition that adipose tissue-derived pro-atherogenic mediators contribute to obesity-related cardiovascular disease. We sought to characterize regional differences in perivascular adipose tissue (PVAT) phenotype in relation to atherosclerosis susceptibility.

Approach and Results:

We examined thoracic PVAT samples in 34 subjects (BMI 32±6 kg/m², age 59±11 years) undergoing valvular, aortic or coronary artery bypass graft surgeries, and performed transcriptomic characterization using whole genome expression profiling and quantitative PCR analyses. We identified a highly inflamed region of PVAT surrounding the human aortic root in close proximity to coronary takeoff and adjoining epicardial fat. In subjects undergoing CABG, we found 300 genes significantly up-regulated (FDR q<0.1) in paired samples of PVAT surrounding the aortic root compared to non-atherosclerotic left internal mammary artery (LIMA). Genes encoding proteins mechanistically implicated in atherogenesis were enriched in aortic PVAT consisting of signaling pathways linked to inflammation, WNT signaling, matrix remodeling, coagulation, and angiogenesis. Over-expression of several pro-atherogenic transcripts including IL1β, CCL2 (MCP-1), and IL6 were confirmed by quantitative PCR and significantly bolstered in CAD subjects. Angiographic CAD burden quantified by the Gensini score positively correlated with the expression of inflammatory genes in PVAT. Moreover, peri-aortic adipose inflammation was markedly higher in obese subjects with striking upregulation (~8-fold) of IL-1β expression compared to non-obese individuals.

Conclusions:

Pro-atherogenic mediators that originate from dysfunctional PVAT may contribute to vascular disease mechanisms in human vessels. Moreover, PVAT may adopt detrimental properties under obese conditions that play a key role in the pathophysiology of ischemic heart disease.

Graphical Abstract

Introduction

Obesity is a major cardiovascular risk factor and the accumulation of ectopic and visceral fat, in particular, has been strongly implicated in the pathogenesis of CVD [1, 2]. Associated risk factors such as hypertension, diabetes mellitus, and hyperlipidemia only partly explain disease pathophysiology, and mechanisms by which obesity heightens vascular risk are incompletely understood [3]. Perivascular adipose tissue (PVAT) is increasingly gaining attention as a local modulator of vascular function with endocrine and paracrine functions that regulate vascular biology [4]. While nearly all human blood vessels are surrounded by PVAT, a potential pathogenic role of PVAT dysfunction in obesity and other conditions has received relatively little attention. In particular, inflammatory cross-talk between PVAT and vascular layers has been proposed to contribute to the pathogenesis of atherosclerosis, but the properties of PVAT in human conduit vessels are largely unexplored [5, 6]. The goals of this study were to perform transcriptome based investigations of atherosclerosis-prone and –resistant human blood vessels, and to compare regional differences in perivascular adipose phenotype in subjects with obesity and coronary artery disease.

Methods

Study subjects

Consecutive men and women undergoing elective cardiothoracic surgery at Boston Medical Center were recruited into the study. Surgeries were CABG (n=29), aortic valve replacement (AVR, n=3), mitral valve replacement (MVR, n=1), and ascending aortic aneurysm repair (n=1). Adipose tissue specimens were collected from thoracic (pre-sternal) subcutaneous fat, anterior pericardium and proximal ascending aorta, and subjects undergoing CABG additionally provided distal LIMA samples. Specimens were procured using scissors rather than electrocautery to avoid thermal tissue damage. Subjects undergoing non-CABG surgeries (n=5) all had non-obstructive CAD by angiography. Pregnant individuals were not eligible for surgery and thus excluded. Pre-operatively, clinical characteristics including BMI and cardiovascular risk factors were recorded, and plasma lipid analyses were performed by the Boston Medical Center clinical chemistry laboratory. Left ventricular ejection fraction was derived from clinical echocardiograms. Cross-sectional histology of the LIMA was performed following formalin fixation. After de-paraffinization and rehydration, arterial sections were stained with hematoxylin and eosin. The study was approved by the Boston University Medical Center Institutional Review Board and written consent was obtained from all participants.

Microarray Gene Expression Analysis

Affymetrix GeneChip Human Gene 2.0 ST arrays (Thermo-Fisher, Waltham, MA) were used to profile gene expression in total RNA isolated from paired samples of peri-LIMA and peri-aortic fat from 5 subjects with CAD. The probe set mapping interrogates a total of 29,635 human Entrez Genes. Arrays were normalized to produce gene-level expression values using the Robust Multiarray Average (RMA) with the affy-R-package (version 1.58.0) and Entrez-Gene-specific R-packages (version 23.0.0) from the Molecular and Behavioral Neuroscience Institute (Brainarray) at the University of Michigan. Differential expression was assessed using a paired empirical Bayesian (moderated) t test as implemented in the limma R package (version 3.38.3). All analyses were performed using the R environment for statistical computing (version 3.5.1). Correction for multiple hypothesis testing was accomplished using the Benjamini-Hochberg false discovery rate (FDR), performed across the genes that were expressed above the median value of at least 5 arrays. Gene Set Enrichment Analysis (GSEA) (version 2.2.1) was used to perform a pre-ranked analysis (default parameters with random seed 1234), ranking the Entrez Gene identifiers of all genes interrogated by the array according to the paired moderated t statistic, and using the Entrez Gene versions of the Hallmark, c2 (Biocarta, KEGG, and Reactome only), c5 (Gene Ontology), and c3 (transcription factor and microRNA motif) collections of gene sets obtained from the Molecular Signatures Database (MSigDB), version 6.0. All raw and processed data have been made publicly available and deposited into the Gene Expression Omnibus (GEO) repository (Series GSE152326). All microarray procedures and analyses were carried out by the Boston University Microarray and Sequencing Resource (BUMSR).

Gene expression analysis using qPCR

Adipose tissue samples were collected in All Protect Reagent (Qiagen) and stored at −80°C. Total mRNA was isolated using the RNeasy Lipid Tissue Mini Kit (Qiagen), and RNA concentration was measured using a Thermo Fisher Nanodrop 2000 at 260/280 nm. 200ng of RNA were reverse transcribed into cDNA using a High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). TaqMan PreAmp Master Mix and TaqMan gene expression assays (Applied Biosystems) were used for a pre-amplification step of selected target genes, using a T100 Thermal Cycler (Bio-Rad). The PreAmp samples were diluted 1:20 using DNase- and RNase-free water, and stored at −20°C until use. The real-time PCR of GAPDH (glyceraldehyde-3-phosphate dehydrogenase), WNT5A (WNT family member 5A), ROR1 (receptor tyrosine kinase like orphan receptor 1), ROR2 (receptor tyrosine kinase like orphan receptor 2) , VANGL2 (planar cell polarity protein 2), IL1B ( interleukin 1β), CCL2 (C-C motif chemokine ligand 2), CCL8 (C-C motif chemokine ligand 8), IL6 (interleukin 6), MAPK8 (mitogen-activated protein kinase 8), (Table 1) were performed with a Viia7 thermal-cycler (Applied-Biosystems), using TaqMan gene expression assays and TaqMan Gene Expression Master Mix, with GAPDH as the housekeeping gene. Expression data for all target genes was normalized to GAPDH, analyzed using the ΔΔCt method, and expressed as fold difference of aortic compared to LIMA PVAT.

Table 1.

Probes for qPCR

Gene Symbol	Assay ID	Entrez Gene ID
GAPDH	Hs02758991_G1	2597
WNT5A	Hs00998537_m1	7474
ROR1	Hs00938677_m1	4919
ROR2	Hs00896176_m1	4920
VANGL2	Hs00393412_m1	57216
PRICKLE1	Hs01055551_m1	144165
IL1β	Hs01555410_m1	3553
CCL8	Hs04187715_m1	6355
CCL2	Hs00234140_m1	6347
IL6	Hs00174131_m1	3569
MAPK8 (JNK)	Hs01548508_m1	5599

Western Immunoblot Analyses

Proteins were extracted from adipose tissue by homogenization in liquid nitrogen and 1X lysis buffer (Cell Signaling, Danvers, MA) supplemented with protease and phosphatase inhibitor cocktail (Sigma Aldrich, St. Louis, MO). Twenty micrograms of total protein was subjected to electrophoresis using SDS-polyacrylamide gel and blotted to nitrocellulose membranes. The membranes were blocked then washed with TBS and incubated overnight at 4°C with the following primary antibodies: monoclonal mouse anti-human UCP-1 (1:500 dilution, R&D Systems, MN, USA), polyclonal rabbit anti-human CCL-2 (1:500 dilution, LS-BIO, WA, USA), mouse monoclonal antihuman IL-6 (1:500 dilution, LS-BIO, WA, USA), and monoclonal mouse anti-human β-actin (1:5000 dilution; Sigma, St Louis, MO), washed, and incubated on a rotary shaker at room temperature for 90 minutes with appropriate horse radish peroxidase (HRP)-conjugated secondary antibody. After washing, immune complexes were detected by enhanced chemiluminescence (Pierce, Rockford, IL). Protein levels were quantified using Image J software and values were normalized to β-actin.

Adipocyte sizing

Paired samples of peri-LIMA and peri-aortic adipose tissue from 5 subjects with CAD were stained using standard hematoxylin and eosin, and images captured using a color digital camera attached to a microscope at 10-40x. For each tissue section, adipocyte size (diameter, μm) was quantified using Adiposoft version 1.16 (advanced distribution of Image J) by a blinded individual [7].

Gensini quantitative angiography score

Coronary artery disease severity was assessed using quantitative coronary angiography (QCA) software (Philips, Netherlands), using established scoring systems by a single interventional cardiologist blinded to all adipose tissue and clinical data. The modified Gensini score was used to characterize the angiographic burden of CAD [8, 9]. In this model, each identified lesion was assigned a percentage diameter stenosis by QCA, then assigned a score (1 for 0-25% stenosis, 2 for 26-50%, 4 for 51-75%, 8 for 76-90%, 16 for 91-99% and 32 for 100%). If the lesion was ≥99%, a modifier was considered based on collaterals and atherosclerotic burden in the source vessel. A pre-specified multiplier was assigned to each segment depending on location of the lesion (5 for left main coronary artery, 2.5 for proximal left anterior descending (LAD) and proximal left circumflex branches, 1.5 for mid-segment of LAD artery, 0.5 for second diagonal branch and posterolateral branch, and 1 for other branches). The modified Gensini score was reported for each participant as a sum of these weighted scores with higher scores indicating more severe CAD.

Statistical analysis

Data are expressed as the mean ± SD in tables. Student’s paired t test was used to compare within subject differences in adipose tissue gene expression. Non-parametric tests were used to compare group differences in mRNA transcripts and protein expression that were non-normally distributed, as assessed by Kolmogorov-Smirnov methodology. Spearman correlations were used for analyses of Gensini scores with gene expression. Graphic data are presented as mean ± SEM unless otherwise indicated. A value of p <0.05 was considered significant for all analyses.

Results

Clinical Characteristics

A total of 34 subjects were recruited for the study. Clinical characteristics of all subjects are displayed in table 2. Consistent with cardiac surgery populations, there was a predominance of cardiovascular risk factors and obesity. Surgeries consisted of CABG (n=29), AVR (n=3), MVR (n=1) and ascending aortic aneurysm repair (n=1). In these subjects, there was high prevalence of statin use (94%) with mean plasma LDL-C ≤ 90 mg/dl.

Table 2.

Clinical characteristics of the study population

Parameter	N=34
Age, (years)	59 ± 11
Sex (M, %)	76
Weight (kg)	98 ± 22
BMI, kg/m2	32 ± 6
Obese (%)	65
Total cholesterol (mg/dl)	168 ± 48
LDL cholesterol (mg/dl)	90 ± 41
HDL cholesterol (mg/dl)	43 ± 20
Triglycerides (mg/dl)	142 ± 72
Statin medication use (%)	94
Hypertension (%)	85
Coronary artery disease (%)	85
Diabetes mellitus (%)	35
Angiographic Gensini score	67 ± 32
Left ventricular ejection fraction (%)	50 ± 14

Data are mean ± SD

Adipose tissue sampling

We collected adipose tissue samples from thoracic subcutaneous, anterior pericardial, proximal peri-aortic, and distal LIMA regions. Exploratory gene expression analyses demonstrated the largest differences between the latter two sample sites, which were thus used for subsequent analyses. A representative image of the surgical field where peri-aortic specimens were taken is displayed in figure 1. We used histology to examine cross-sections of the distal LIMA (near touchdown site) and confirmed absence of atherosclerotic plaques as displayed in figures 2A-B. Therefore, LIMA PVAT specimens were used as control samples in subsequent gene expression analyses. We performed adipocyte sizing of paired PVAT samples in five subjects with CAD and observed no difference in adipocyte diameter between LIMA (figure 2C) compared to aortic (figure 2D) PVAT (54±9 μm vs. 69±19 μm, respectively, p=0.14)

Figure 1:

Representative operative field during coronary artery bypass graft surgery in an obese subject demonstrating extensive adipose tissue accumulation surrounding cardiovascular territories. SVG= Saphenous venous graft.

Figure 2:

Cross-sectional hematoxylin staining of the LIMA demonstrating no visible plaques. (A. upper panel 4X, B. lower panel 10X magnification). Representative histology of paired PVAT samples from LIMA (C) and aortic (D) specimens (20x magnification). There was no significant difference in adipocyte morphology or size between the two depots.

Differential microarray expression analysis between aortic and LIMA PVAT

We used whole-genome microarrays to profile gene expression in total RNA obtained from PVAT of five male subjects with CAD (age 60±5 years, 80% obese, mean LDL-C 89±31 mg/dl, 100% statin use). Each subject provided paired PVAT samples from both the aortic root and LIMA simultaneously, thus avoiding confounding influence from systemic variables and sex. A total of 324 genes were significantly differentially expressed between depots (FDR q < 0.1), with 300 genes upregulated and 24 genes downregulated in aortic PVAT compared to that from LIMA. The relative transcriptional signature and gene patterning was remarkably consistent for each subject. Genes that were up-regulated in aortic PVAT included transcripts associated with inflammasomes (NRLP3), TNFα signaling (TNFAIP2, TNFAIP3, TNFSF8, TNFRSF1A, and TNFRSF10D), interleukin-1 signaling (IL1R1, IRAK3), interleukin-6 signaling (IL6), NFκB signaling (NFKB1), leukocyte adhesion (VCAM1), macrophage chemotaxis (CCL2/MCP-1), and chemokine signaling (CXCL16, CCL21, CCL13). The complete set of statistical analysis results is provided in Supplementary File I.

We then performed Gene Set Enrichment Analysis (GSEA) to identify pathways and biological processes that were coordinately upregulated in aortic PVAT compared to the LIMA. A high number of gene sets were significantly coordinately upregulated (FDR q < 0.01), including those corresponding to the Gene Ontology biological process terms "positive regulation of inflammatory response" (GO:0050729, FDR q < 0.001), "cellular response to interleukin-1" (GO:0071347, FDR q < 0.001), "positive regulation of interleukin-1 production" (GO:0032732, FDR q = 0.0057), "cellular response to interleukin-6" (GO:0071354, FDR q < 0.001); "leukocyte chemotaxis" (GO:0030595, FDR q < 0.001), and "non-canonical Wnt signaling pathway" (GO:0035567, FDR q = 0.0057); the BioCarta IL1R pathway (FDR q < 0.001), and the KEGG adipocytokine signaling pathway (hsa04920, FDR q = 0.0048). A heatmap of the expression of the combined set of "leading edge" genes of these results (i.e., those genes that contributed most to the significance of each result) is shown in Figure 3. A complete set of GSEA results is provided in Supplementary File II.

Figure 3:

Heatmap of the union set of leading edge genes from eight selected gene sets with significant coordinate upregulation (FDR q < 0.01) in aortic vs. LIMA PVAT. The membership of each gene within each gene set is indicated with a colored box at the top of the heatmap: a) GO term "positive regulation of inflammatory response"; b) GO term "cellular response to interleukin-1"; c) BioCarta IL1R pathway; d) GO term "positive regulation of interleukin-1 production"; e) GO term "cellular response to interleukin-6"; f) GO term "leukocyte chemotaxis"; g) GO term "non-canonical Wnt signaling pathway"; h) KEGG adipocytokine signaling pathway. Rows and columns correspond to samples and genes, respectively. Columns are sorted in left to right first by gene set and then by moderated t statistic. Gene expression values were z-score-normalized to a mean of zero and SD of 1 across all samples in each column, with blue, white and red indicating final z scores of ≤ 2, 0, and ≥ 2, respectively. Genes selected for subsequent qPCR analyses are indicated with asterisks next to each gene symbol below the heatmap and black boxes above the heatmap.

Quantitative PCR and Western immunoblot analyses

We then selected several genes of interest from the microarray analysis for confirmatory quantitative PCR in a larger cohort of subjects. These candidates included genes that were relevant to atherosclerosis and significantly upregulated in aortic PVAT (CCL2, FDR q = 0.064; CCL8, FDR q = 0.12; IL1B, FDR q = 0.12; IL6, FDR q = 0.06; ROR1, FDR q = 0.19), as well as others that were not nominally significant (p > 0.05) but were included in the leading edge of at least one gene set with a significant GSEA result (MAPK8/JNK, WNT5A; see Figure 3) or have been previously implicated in obesity and cardiovascular disease (ROR2) [10, 11]. As shown in Figure 4, these genes were significantly upregulated in aortic compared to LIMA PVAT. For protein confirmation, we then performed Western immunoblot analyses for IL-6 and CCL-2 and demonstrated significant upregulation in aortic PVAT (Figure 5) firmly mirroring gene expression data. In contrast, uncoupling protein-1 (UCP-1) expression as a marker for distinguishing white/beige/brown fat phenotypes [12, 13] was not differentially expressed between these two depots (p=NS). Furthermore, we observed significant upregulation of pro-inflammatory transcripts (IL1B, CCL2, CCL8, IL6, MAPK8/JNK) and genes involved in non-canonical Wnt signaling (ROR2, VANGL2, PRICKLE1) in the aortic PVAT of obese subjects (BMI ≥ 30kg/m²) relative to non-obese subjects (Figure 6) and in inflammatory genes in subjects with CAD versus those without CAD (Figure 7). Additional sub-analyses excluding all subjects without CAD still showed significant up-regulated expression of IL1B, PRICKLE1, CCL2, and IL6 (p<0.05 for all) in the aortic PVAT of obese compared to non-obese subjects. Finally, we found that the expression of both CCL2 (MCP-1) and IL6 in aortic PVAT significantly correlated with angiographic Gensini score, a measure of the extent of CAD (Figure 8), in this group of individuals who had a high prevalence of statin use and relatively good LDL-C control (≤ 90 mg/dl).

Figure 4:

Expression of WNT5A signaling and pro-inflammatory genes is significantly upregulated in aortic compared to LIMA PVAT. Data are expressed as fold difference of aortic compared to LIMA (set to 1) depots, performed in triplicates by qPCR and normalized to GAPDH gene expression. Paired t tests were used for statistical analysis (n=20). Data are mean ±SEM. *p<0.05, **p<0.01, ***p<0.001.

Figure 5:

Western immunoblot analyses demonstrate significantly increased protein expression of IL-6 and CCL-2 in aortic compared to LIMA PVAT confirming gene expression data. In contrast, no inter-depot PVAT difference in UCP-1 protein expression was observed (p=NS). Data are expressed as relative protein expression using optical density (OD), normalized to β-actin protein. Wilcoxon rank-sum testing was used for statistical analyses (n=5). Data are mean ±SEM,*p<0.05.

Figure 6:

Peri-aortic PVAT expression of pro-inflammatory transcripts is significantly upregulated in obese compared to non-obese subjects. Data are expressed as relative gene expression (a.u.), performed in triplicates by qPCR and normalized to GAPDH gene expression. Mann Whitney testing was used for statistical analyses (n=17 obese, n=7 non-obese). Data are mean ±SEM. *p<0.05, **p<0.01.

Figure 7:

Peri-aortic PVAT expression of pro-inflammatory transcripts is significantly upregulated in subjects with CAD (n=22) compared to individuals without CAD (n=5). Data are expressed as relative gene expression (a.u.), performed in triplicates by qPCR and normalized to GAPDH gene expression. Mann Whitney testing was used for statistical analyses. Data are mean ±SEM. *p<0.05.

Figure 8:

Spearman correlations demonstrating significant positive correlations peri-aortic adipose tissue gene expression of CCL2 (left panel) and IL6 (right panel) with cumulative CAD burden assessed by quantitative angiography Gensini scores (n=19).

Discussion

In the present study, we demonstrate transcriptional upregulation of a broad range of pro-atherogenic mediators in the perivascular adipose tissue of subjects with coronary artery disease. Even with a small sample size, microarray analyses showed strong concordance across subjects with significant upregulation of at least 300 genes, many of which are mechanistically implicated in cardiovascular disease. We also describe for the first time, to our knowledge, that the expression of specific pro-inflammatory transcripts was significantly increased in the PVAT of subjects with obesity or CAD, and correlated positively with angiographic atherosclerotic burden. These findings were observed in a group with relatively well-controlled LDL-C levels and high prevalence of statin use, evoking the notion of residual inflammatory cardiovascular risk in many individuals[5]. Collectively, these data provide basis for a potential role of PVAT dysregulation contributing to the pathogenesis of atherosclerosis.

Vascular inflammation plays a key role in all phases of atherosclerosis [14] and there is growing recognition that PVAT can participate in this process [3]. Nearly all human conduit vessels are surrounded by adipose tissue coatings that contain a mixture of cell types including adipocytes, stem cells, immune cells, and nervous tissue that juxtapose with the adventitial vasa vasorum and communicate bi-directionally with all vascular layers. While PVAT functions are largely homeostatic [6, 15-18], perturbations can develop in obesity and other conditions that support pathological rather than protective roles for PVAT [3, 19]. Traditional concepts of vascular inflammation have evolved beyond the classic "inside-out" monocyte adhesion and lipid oxidation paradigms that begin at the endothelial surface. In fact, there are growing animal [20] and human [21, 22] data supporting a parallel "outside-in" inflammatory process initiated externally that can propagate inward [23]. In this regard, the adventitia elaborates a range of chemokines and cytokines that regulate neovascularization, vascular tone, and angiogenesis [17, 24], and can function as major sites of immune cell accumulation that influence plaque pathobiology [20, 23, 25-27]. Direct causal evidence of a pro-atherogenic role for PVAT comes from experimental animal transplant studies where relocation of pathogenic fat onto recipient carotid vessels promotes atherosclerosis in apoE knock out mice [28], and transposition of aortic PVAT from obese mice accelerates neointimal hyperplasia and adventitial macrophage infiltration [29]. Human observational studies report that PVAT co-localizes with atherosclerotic plaques [30] and associates with advanced lesions [31]. Several studies focusing on epicardial adipose tissue (EAT) in CAD patients describe a more pro-inflammatory profile compared to subcutaneous fat [32-34] and higher secretion of adipocytokines and monocyte chemotaxis near stenotic coronary segments [21, 35]. More compelling data come from recent computer tomography (CT) vascular imaging studies which show that quantitative measures of coronary perivascular fat inflammation predict cardiovascular events independent of CAD stenosis and even prior to the evolution of discernable plaques [36, 37].

While the mechanisms by which PVAT promotes atherosclerosis progression and instability in humans remain an open question, our comparative data may provide a number of important clues. It has been long recognized that the internal mammary artery (IMA), which represents a thoracic vessel of similar size and histology as the coronaries and is exposed to the same systemic risk factors rarely develops atherosclerosis, for largely unknown reasons, and is thus widely used as a bypass conduit. Autopsy studies that compared IMA to coronary specimens have identified the importance of local anatomical factors as determinants of lesion development [38]. One such variable may relate to regional PVAT heterogeneity, yet few comparative studies have specifically examined the cellular composition or adipokine expression of IMA PVAT, and these were mainly cadaveric [21, 39, 40], limited to males [41] or non-obese subjects [42]. PVAT surrounding the IMA exerts vasodilatory effects [43] and endothelial cells express higher eNOS and atheroprotective genes [44, 45]. Moreover, IMA grafts pass this protection onto downstream native target vessels which strongly suggests paracrine signaling [46]. Our global analyses demonstrated a more quiescent IMA phenotype with regard to inflammation, macrophage chemotaxis, cytokine reactome, matrix turnover, and coagulation between depots, raising the possibility that lesser adipose tissue dysfunction around the IMA may be one mechanism explaining the relative protection of this vessel against atherosclerosis. Understanding mechanisms of IMA resistance to atherogenesis may generate new strategies to treat or prevent CHD.

A novel aspect of our study is our discovery of a regional hub of highly inflamed PVAT surrounding the aortic root in close proximity to coronary takeoff and epicardial fat. This anatomical region frequently referred to as Rindfleisch’s folds [47] has received no attention in clinical investigation, yet shares vascular networks contiguous with epicardial and myocardial territories as no distinct anatomical boundaries separate these regions. Moreover, very few studies have compared the characteristics of PVAT in subjects with and without CAD, and published studies with non-obstructive coronaries are extremely rare [48, 49]. Reports in non-obese CAD patients showed lower expression of the vasculoprotective factor adiponectin [50] and higher inflammation in EAT [49], while other studies utilizing microarrays showed few differences [51] and no differential UCP-1 expression [52]. In CAD patients, we identified significant enrichment in aortic PVAT genes encoding for numerous proteins implicated in atherogenesis including CCL2/MCP-1, CCL8, IL1β, and MAPK8/JNK, as well as components of non-canonical Wnt pathway signaling [53] not previously highlighted.

Another novel aspect of our study was the ability to examine the influence of obesity on PVAT phenotype and to our knowledge no prior study has directly compared gene expression in the aortic PVAT of obese vs. non-obese subjects. A prior swine study reported a greater constrictive effect of coronary PVAT in obese animals [54], and higher EAT inflammation was described in CAD subjects with BMI>27 kg/m² [48]. Perivascular fat can adopt detrimental properties under obese conditions that contribute to the pathophysiology of cardiovascular disease [55, 56] and severe obesity can fully encase cardiac structures in fat ≥ 2 cm thick (“cor adipe plane tectum”) [57]. In the Framingham Heart study, measures of peri-aortic fat were associated with aortic stiffness and cardiovascular risk [58, 59]. In contrast, coronaries lacking adipose coatings in segments bridged by myocardial tissue are protected from atherosclerosis [60]. Thus far, mechanistic evidence for obesity-induced changes in PVAT have largely come from animals. For instance, high-fat-fed mice exhibit impaired peri-aortic adipocyte differentiation [39] and endothelial vasodilator dysfunction via mechanisms linked to adipose macrophage chemotaxis and oxidative stress [61]. High-fat feeding triggers microRNA conversion that evokes PVAT inflammatory responses and arterial remodeling [62]. In the present study, we observed significant inflammasome upregulation in PVAT associated with obesity, directionally similar to that seen in CAD patients. In this regard, a chronic state of low-grade inflammation in obesity driven in part by adipose tissue inflammation has been well recognized, and we have previously described the contribution of non-canonical Wnt pro-inflammatory signaling to mechanisms of insulin resistance and vascular dysfunction [2, 11, 63, 64]. Experimental data suggest that PVAT responds to external factors highlighting the plasticity of this tissue and its potential for therapeutic modulation. Macrophage ablation restores anti-contractile functions of PVAT [65], and nitric oxide-dependent vasodilation of the microvasculature is restored with weight loss [66]. Statins modify PVAT in hypertensive rat aortas [67], however nothing is known with regard to therapeutic modulation in human disease. Presently, we are only at the beginning of understanding PVAT biology, and future studies are required to characterize its pathogenic role and treatment implications in obesity-related cardiovascular diseases.

Limitations: First, the number of subjects was limited but expected with this type of invasive study involving cardiac surgeries. We point out that we utilized freshly isolated samples from living subjects rather than rely on autopsy specimens. Even with a small sample size, the microarray results were strongly significant and we were able to verify the expression of specific genes by qPCR. Second, we did not have LIMA samples from non-CAD patients since this region is not accessed by the surgeon. Third, unmeasured clinical confounders could influence inter-group differences, however, this was avoided in our microarray study as paired samples were obtained from the same subject. Fourth, the study is observational and does not prove causal relationships, however, our data firmly build upon the scant human literature in the field. Fifth, multi detector-row CT angiography was not available in all subjects to quantify degree of aortic atherosclerosis. Lastly, from a conceptual standpoint, it is possible that PVAT inflammation stems partly from the diseased vessel that extends to surrounding adipose tissue, and while imaging studies suggest this may be otherwise [36], this concept warrants further mechanistic investigation.

In conclusion, human thoracic PVAT displays marked territorial heterogeneity with distinct over-expression of pro-atherogenic transcriptional profiles in subjects with obesity and CAD. Local actions of PVAT may contribute to vascular pathologies in conditions associated with obesity-induced adipose tissue dysfunction.

Highlights.

• We describe peri-vascular adipose tissue (PVAT) regional heterogeneity in pro-atherogenic gene expression of atherosclerosis-prone and -resistant human thoracic blood vessels.

• Pro-inflammatory transcripts encoding for proteins mechanistically implicated in atherogenesis were markedly up-regulated in the PVAT of individuals with obesity and CAD.

• Local actions of PVAT may contribute to vascular pathologies in conditions associated with obesity-related adipose tissue dysfunction and ischemic heart disease.

Abbreviations:

a.u.

arbitrary units

BMI

body mass index

CABG

coronary artery bypass graft

CAD

coronary artery disease

CVD

cardiovascular disease

CCL2

C-C motif chemokine ligand 2

CCL8

C-C motif chemokine ligand 8

EAT

epicardial adipose tissue

FDR

false discovery rate

GAPDH

glyceraldehyde-3-phosphate dehydrogenase

GEO

Gene Expression Omnibus

HDL-C

high-density lipoprotein cholesterol

IL1β

interleukin-1β

IL6

interleukin-6

JNK

c-jun-N terminal kinase

LDL-C

low-density lipoprotein cholesterol

LIMA

left internal mammary artery

MAPK8

mitogen-activated protein kinase 8

MCP-1

monocyte chemoattractant protein-1

PVAT

perivascular adipose tissue

PRICKLE1

prickle planar cell polarity protein 1

QCA

quantitative coronary angiography

qPCR

quantitative polymerase chain reaction

ROR1

receptor tyrosine kinase like orphan receptor 1

ROR2

receptor tyrosine kinase like orphan receptor 2

VANGL2

planar cell polarity protein 2

VCAM1

vascular cell adhesion molecule-1

WNT5A

Wnt family member 5A