Aberrant Epicardial Adipose Tissue Extracellular Matrix Remodeling in Patients with Severe Ischemic Cardiomyopathy: Insight from Comparative Quantitative Proteomics

There is ample evidence indicating that epicardial adipose tissue (EAT) volume and thickness is positively associated with coronary artery disease (CAD). However, the exact pathological changes in the human EAT after myocardial ischemia remains largely unclear. In the current study, we applied a comparative quantitative proteomics to elucidate the altered biological processes in the EAT of ischemic cardiomyopathy (ICM) patients. A total of 1649 proteins were successfully quantified in our study, among which 165 proteins were significantly changed (ratio <0.8 or >1.2 fold and p < 0.05 in both repetitions) in EAT of ICM individuals. Gene ontology (GO) enrichment analysis revealed that cardiac structure and cellular metabolism were over-represented among these regulated proteins. The hypertrophic cardiomyopathy, adrenergic signaling in cardiomyocytes, extracellular matrix (ECM)-receptor interaction, phagosome, Glycolysis/Gluconeogenesis, and PPAR signaling pathway were highlighted by the KEGG PATHWAY analysis. More importantly, we found that the proteins responsible for extracellular matrix organization were dramatically increased in EAT of ICM patients. In addition, the picrosirius red (PSR) staining results showed that the collagen fiber content was prominently increased, which indicated the EAT of ICM individuals underwent extracellular matrix remodeling and ERK1/2 activation maybe responsible for these pathological changes partially.


Histology analysis of EAT in normal and ICM individuals.
In correlation with the literature reported, aberrant increased EAT and cardiomyocytes loss were observed in the heart of ICM patients (Fig. 1a,c). As indicated by black dotted circle, the EAT located in the atrioventricular groove was isolated in order to perform further experiments (including proteomics, western blots, real-time PCR and immunohistochemical analysis) in this study (Fig. 1a). Coronary angiography showed obvious stenosis of left anterior descending artery and diagonal branch in the presented ICM patient who has higher serum cardiac troponin I (cTnI), creatine kinase-MB (CK-MB), N-terminal-pro-B-type natriuretic peptide (NT-proBNP) concentration ( Fig. 1b and Table S1). H&E staining showed that the size of adipocytes had no significant difference between normal EAT and ICM EAT, neither at atrioventricular groove (Fig. 1d,e) nor left ventricle (Fig. 1f,g). 165 proteins differentially expressed in the EAT of ICM patients. Since EAT volume and thickness were positive correlation with CAD 3,4,22 , it was critical to comprehensively reveal the pathological changes of EAT at protein level. Therefore, comparative quantitative proteomics analysis of EAT obtained from normal donors (heavy labeled with CD 2 O) and ICM patients (light labeled with CH 2 O) was performed by liquid chromatograph-mass spectrometer (LC-MS/MS) (Fig. 2a). Two independent repetition experiments were performed, which included seven normal EATs and eight ICM EATs, and a total of 1649 EAT proteins were quantified in both repetitions (Fig. 2b). Pearson correlation coefficient (PCC) analysis showed good repeatability between two repetitions, as evidenced by r = 0.76 (Fig. 2c). A Gaussian distribution of the shared quantitative data (as log 2 (Ratio)) was further analyzed, showing a reasonable ratio distribution (Fig. 2d). All the quantified proteins were summarized in Table S2. Among the whole quantitative dataset, seventy-two proteins (average ratio < 0.8 fold and p < 0.05 in both repetitions) were found to be significantly down-regulated in ICM EATs, whereas ninety-three proteins (average ratio > 1.2 fold and p < 0.05 in both repetitions) were up-regulated (Table S3). As a way to cross-check the reliability of quantitative proteomics, several proteins expression levels were assessed using conventional western blot on the same specimens. In accordance with the results of proteomics analysis, western blot demonstrated that the protein levels of apolipoprotein A-IV (APOA4), creatine kinase M-type (CKM), carbonic anhydrase 2 (CA2), troponin I, cardiac muscle (TNNI3), and myosin light chain 3 (MYL3) were ectopic expression in ICM EATs compared with normal counterparts (Fig. 2e), while fatty acid synthase (FASN), c-reactive protein (CRP), and glutamine synthetase (GLUL) were obviously down-regulated (Fig. 2f). In this light, we were prudent to conclude that the data of comparative quantitative proteomics analysis was very reliable.
Multiple biological processes were affected in EAT of ICM patients. To ascertain the functional representation of these differentially expressed proteins, the data were further analyzed based on the gene ontology (GO) term enrichment. The results showed that the proteins up-regulated in the EAT of ICM patients were found significantly enriched among the inflammation and cardiac structure related the GO terms including regulation of biological quality, response to inorganic substance, muscle contraction, wound healing, acute inflammatory response, positive regulation of blood coagulation, and striated muscle tissue development (Fig. 3a), while Scientific RepoRts | 7:43787 | DOI: 10.1038/srep43787 down-regulated proteins were mainly enriched in metabolic process, such as organic acid metabolic process, cellular ketone metabolic process, oxoacid metabolic process, carboxylic acid metabolic process, monocarboxylic acid metabolic process, as well as translational elongation (Fig. 3b).
To understand the intracellular pathways involvement in these differentially expressed proteins, the KEGG PATHWAY database mapping analysis was performed. Our results revealed that a total of 22 pathways were found to be involved (Fig. 4), and among them proteins involved in the cardiac muscle function related pathways like dilated cardiomyopathy (p = 1.17E-3), the hypertrophic cardiomyopathy (p = 2.94E-15), adrenergic signaling in cardiomyocytes (p = 1.30E-11), and cardiac muscle contraction(p = 4.06E-3) were mainly increased, while those involved in tricarboxylic acid (TCA) cycle (p = 6.93E-03), Ribosome (p > 0.05), pyruvate metabolism (p = 4.59E-03) and AMPK signaling pathways (p = 1.62E-03) were mainly decreased. The p-value here represents the protein enrichment false discovery rates (FDRs) of the pathway terms (Fig. 4).
The proteins related to ECM were ectopic expression in EAT of ICM patients. To further investigate the possible regulated signaling network among altered EAT proteome of ICM patients systematically, we established the protein interaction network based on all those differentially expressed proteins using the STRING database and the Uniprot functional annotation. Our data indicated that protein interaction network related to the three categories: cytoskeleton organization, cellular metabolism, and immune response were significantly altered in the EAT of ICM individuals (Fig. 5a-c). Most of the altered proteins involved in lipid homeostasis, amino acids metabolism, glycolysis/gluconeogenesis, as well as transcription were prominently reduced, but the majority of changed proteins participated in cellular ion homeostasis and immune response were obviously increased in ICM EATs (Fig. 5b,c). As for cytoskeleton organization, most of changed proteins associated with membrane organization were down-regulated (e.g. CLTC, ARCN1, COPA, and AP2M1) in EAT of ICM patients, while opposite situation was observed in the relevant proteins of cardiac muscle contraction, actin cytoskeleton organization, and ECM organization (Fig. 5a). All the proteins involved in the network analysis were summarized in Table S4. Consistent with the results of proteomics, the protein or mRNA levels of A2M, PLG, Col3A1, BGN, Col1A1, Col6A2, Col6A3, Col12A1 and LUM were significantly increased in the atrioventricular groove EAT of ICM individuals (Fig. 6a-c). More importantly, concomitant changes with the ECM related proteins, abundant ECM was detected by picrosirius red (PSR) staining in both atrioventricular groove EAT and left ventricle EAT of ICM patients (Fig. 6d,e). Therefore, all the aforementioned data indicated that severe ECM remodeling was taken place in the EAT of ICM patients.
The ERK1/2 activation may be partially responsible for the ECM remodeling of EAT in ICM patients. To investigate the mechanism promoting ECM accumulation in the EAT of ICM patients, the mitogen-activated protein kinase (MAPK) signaling pathway, which is closely related to cardiac fibrosis 23 , was detected in the EAT samples. There are three classic MAPK subfamilies, including ERK1/2, JNK1/2 and P38 24 . Our results demonstrated that the phosphorylation levels of MEK1/2, JNK1/2 and P38 showed no significant difference between normal and ICM EAT groups (Fig. 7a-c). However, the phosphorylated ERK1/2 was  Nine terms including translation elongation and various metabolic processes were over represented (p < 0.00001) in the biological process category of down-regulated proteins. Terms are depicted as nodes connected by arrows that represent hierarchies and relationships between terms. Node size is proportional to the number of proteins assigned to a given ontology term, whereas node color represents the corrected p-value (Benjamin Hochberg false discovery rate correction) corresponding to enrichment of the term. dramatically increased in the ICM EATs compared with normal EATs (Fig. 7a-c). These results indicated that ERK1/2, but not JNK1/2 and P38 activation may be involved in ECM remodeling in the EAT of ICM patients.

Discussion
Since EAT volume and thickness were closely related to CAD 9,10 , it's important to ascertain the changed biological processes of EAT in CAD or ICM patients. In the present study, we identified 165 differentially expressed proteins in the EAT of ICM patients, which were associated with cytoskeleton organization, cellular metabolism, and immune response respectively. Notably, compared with normal EAT, the ECM proteins were aberrant expression in EAT of ICM, which may partially regulate by ERK1/2 activation.
EAT, a kind of adipose tissue, has the general characteristics of adipose tissue, including energy storage and metabolism, endocrine/paracrine cytokines, thermal regulation, and protecting organs from mechanical damage 25 . Though we did not yet know the exact mechanisms on how increased EAT affects the pathological changes of CAD or even ICM, considerable advances have been made in recent years. Uchida et al. demonstrated that pericoronary adipose tissue as storage and possible supply site for oxidized low-density lipoprotein (oxLDL) and high-density lipoprotein (HDL) in coronary arteries, and CD68(+ )-macrophages or vasa vasorum were responsible for oxLDL reaching to coronary intima, while HDL was supplied to coronary intima only by vasa vasorum 26 . Excessive oxLDL would accelerate formation of atheromatous plaque and CAD development. Compared with non-CAD patients, the polarization of M1/M2 macrophages was shifted toward a pro-inflammatory state in the EAT of CAD patients 18 . Multiple pro-inflammatory and atherogenic cytokines (e.g. MCP-1, IL-1β , IL-6, TNFα, PAI-1) were secreted by EAT in excess into the adjacent myocardium and coronary arteries 3 . Pericardial fluid, containing many kinds of peptide hormones and growth factors, is directly contact with the EAT and the myocardium 27,28 . Angiotensin-converting enzyme (ACE), bradykinin (BK) were detected in the pericardial fluid of coronary artery disease (CAD) patients 29 , which played a function on adjacent coronary artery or myocardium. However, the source of ACE and BK was unclear, and EAT may contribute to their accumulation in the pericardial fluid. Higher levels of reactive oxygen species were observed in the EAT of CAD patients 21 . In severe CAD subjects, the transcriptional level of gene sets related to intracellular trafficking, proliferation/transcription regulation, protein catabolism, innate immunity/lectin pathway, and ER stress were downregulated 30 . In addition, the research of EAT on atrial cardiomyopathy and atrial fibrillation had made some progress recently. Venteclef et al. demonstrated that the secretome from EAT promotes fibrosis of the atrial myocardium through the secretion of adipo-fibrokines such as Activin A 31 . In our current study, our results indicated that, a hitherto unrecognized finding, the ECM was significantly increased in the EAT of ICM patients and ERK1/2 activation (Figs 5-7) may participate in this biology process. These findings indicated that EAT affected atria and ventricles in a similar fashion, at least partially.
Previous studies indicated that ECM, mainly composed of structural proteins (e.g. collagens, osteopontin), adhesion proteins, and proteoglycans, was crucial for the structural integrity of adipocytes and its function 32 . Henegar et al. demonstrated that a higher amount of fibrosis was observed in the obese subcutaneous white adipose tissue (WAT) than that of lean human beings 33 . Similarly, increased collagens were detected in the epididymal fat pads of db/db mice (a well-known genetic model of obesity) 34 . The changes, such as components or content, happened in the ECM could affect intracellular signaling pathways of adipose and adjacent cells through its receptors (e.g. integrins and CD44), and finally trigger the necrosis of adipocytes, physical restriction on adipose tissue expansion, and adipose inflammation in WAT 32,35 . These research achievements on WAT may provide clues for EAT study. Ample evidence demonstrated that CAD severity was positively associated with EAT volume or thickness 4,10,22,36 . However, it remains unclear whether the ECM in EAT of ICM patients altered. Our present study demonstrated that compared with normal EAT, the ECM proteins, such as Col3A1, Col1A1, Col6A1, Col6A2, Col6A3, Col12A1, TGFBI, LUM, A2M, BGN, were remarkably increased in the EAT of ICM patients (Figs 5a and 6a-c). Furthermore, a large number of collagen fibers deposition was observed in the EAT of ICM patients via PSR staining (Fig. 6d,e). However, in pathologic conditions, more research is nevertheless Since fibrosis not only impaired adipocyte functionality, but also limited adipocyte hypertrophy 37 , we assessed the adipocyte size in EAT. As was recognized that the EAT volume and thickness were obviously increased in the CAD and ICM patients 4,10 , this conclusion was verified in our study (Fig. 1a). Generally, the increased fat volume, thickness or weight was resulted from increased adipocyte size (hypertrophy) and/or number (hyperplasia) 38 . Adipocyte hypertrophy, leading to WAT dysfunction (e.g. abnormal lipid storage and adipogenesis, impaired secretion of adipokines, exacerbated fibrosis deposition and insulin resistance), was a hallmark of WAT enlargement in obesity 38 . However, in the present study, our data showed that the adipocyte size in EAT had no significant difference between normal donors and ICM patients (Fig. 1e,g). These data indicated that the increased EAT volume and thickness in ICM individuals may not due to adipocyte hypertrophy. Previously published literature were demonstrated that fibrosis might affect the expandability of WAT by physically limiting adipocyte hypertrophy 39 . The adipocytes in WAT exhibited hypertrophy in ob/ob mice lacking collagen VI which were absence of fibrotic deposits 34 . In addition, the adipocyte size was profoundly affected by genetic ablation of collagens or remodelling enzymes 38 . Given that, the increased EAT volume or thickness of ICM patients maybe the result of hyperplasia. However, further research is needed to elucidate whether hyperplasia is the reason for increased EAT volume or thickness in ICM patients in the future.
Most of currently published studies were using subcutaneous or omental adipose tissue as control group, but not normal EAT 18,20,21 . As we known, compared with EAT, subcutaneous or omental adipose tissue have a different gene profile and function in normal conditions. These differentially expressed genes may not contribute to the pathological processes of diseases. Therefore, in our present study, the ICM and normal EAT were applied to comparative quantitative proteomics to elucidate the events occurred in the EAT of ICM patients at protein level. However, because of the limitation of sensitivity of mass spectrometry, lots of low abundance proteins may have been missed in our study. The post-translational modification of proteins, such as activation by phosphorylation, was not studied in the present study. In addition, a relatively small number of samples were included in the present study. Therefore, more research and biological methods were needed to make great progress in the field of EAT.

Methods and Materials
Patients and tissue samples. Approval was got from Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology Review Board in Wuhan, China. All the procedures involving human samples conformed to the principles outlined in the Declaration of Helsinki. Participation was voluntary and informed consent was obtained in all cases. EAT samples of ICM were collected from patients undergoing heart transplantation due to irreversible heart failure caused by myocardial infarction. Normal EAT samples were obtained from the normal heart donors who died in accidents but whose hearts were not suitable for transplantation for noncardiac reasons [40][41][42] . Samples were flash frozen and processed for proteomic analysis, western blot and real-time PCR.
Extraction and digestion of epicardial adipose proteins. EAT biopsies from eight patients with ICM and seven healthy controls were used for proteomics analysis. Tissues were washed in PBS immediately after removal and cut into small pieces, then directly frozen in liquid nitrogen until needed. Homogenization of EAT were processed (a pooled approach to minimize the biological variations) using the HNTG lysis buffer (50 mM HEPES, 150 mM NaCl, 10% glycerol, 1% Triton X-100) with a glass-Teflon homogenizer. The resulted samples were then vortex in cold room for 30 min and centrifuge at 12000 g, 4 °C for 20 min. the fat would form a layer at the top and the cell debris would form a pellet. Avoiding the cell debris layer and pipetting going under the fat to pull off the clear/reddish brown lysate, repeat this last step twice. The epicardial adipose proteins were collected and divided into aliquots for two independent experimental replicates to be conducted.
The in-solution digestion of EAT proteins was performed as described in a previous study 43 . Proteins were precipitated by mixing with 50% acetone/50% methanol/0.1% acetic acid, followed by centrifugation at 2000 g for 20 min. The protein pellets were re-suspended with 8 M Urea/4 mM CaCl 2 /0.2 M Tris-HCl, pH 8.0, and reduced with 10 mM DTT at 50 °C for 30 min and then alkylated with 40 mM iodoacetamide in the dark for 30 min. After measuring protein concentrations via Bradford assay, the proteins were digested with trypsin at a ratio of 1:50 (trypsin/protein w/w). The digested peptides were desalted using a SepPak C18 cartridge (Waters) and dried with a SpeedVac.
Stable isotope dimethyl labeling and strong cation exchange (SCX) fractionation. Desalted peptides were re-suspended in 0.1 M sodium acetate, pH 6.0. Next, 4% formaldehyde (CH 2 O, which serves as "light labeled") was added to the peptides from ICM EAT, and 4% deuterated formaldehyde (CD 2 O, which serves as "heavy labeled") was added to the peptides extracted from normal EAT. After mixing, 0.6 M sodium cyanoborohydride (NaBH 3 CN) was added, and the mixtures were incubated at room temperature (20 ± 2 °C) for 1 hour. The samples were then quenched by adding 1% ammonium hydroxide, followed by the addition of 5% formic acid. After labeling, the peptides were mixed at a ratio of 1:1 and desalted again prior to separation via SCX chromatography.

LC-MS/MS and data processing. All ESI-based LC-MS/MS experiments were performed on a TripleTOF
5600+ System coupled with an Ultra 1D Plus nano-liquid chromatography device (SCIEX, USA). Dried peptides were dissolved in 0.1% formic acid/2% acetonitrile/98% H 2 O, loaded onto a C18 trap column (5 μ m, 5 × 0.3 mm, Agilent Technologies) at a flow rate of 5 μ L/min, and subsequently eluted from the trap column over the C18 analytic column (75 μ m × 150 mm, 3 μ m particle size, 100 Å pore size, Eksigent) at a flow rate of 300 nL/min in a 100 min gradient. The mobile phase consisted of two components: component A was 3% DMSO/97% H 2 O with 0.1% formic acid, and component B was 3% DMSO/97% acetonitrile with 0.1% formic acid. The information dependent acquisition (IDA) mode was used to acquire MS/MS data. Survey scans were acquired in 250 ms and 40 product ion scans were collected at 50 ms/per scan. The precursor ion range was set from m/z 350 to m/z 1500, and the product ion range was set from m/z 100 to m/z 1500.
The generated raw MS spectra was analyzed with Proteinpilot 4.5 software (SCIEX, USA) by using the Paragon algorithm. The Uniprot_Human_201510 database (downloaded from UniProt) was used. The data analysis parameters were as follows: Sample type: Dimethyl + 0, + 4, + 8 (Peptide Labeled); Cys Alkylation: Iodoacetamide; Digestion: Trypsin; Instrument: TripleTOF 5600; Special Factors: Urea denaturation; ID Focus: Biological modifications, Amino acid substitution; Search Effort: Thorough ID; Detected Protein Threshold [Unused ProtScore (Conf)]: 1.3 (95.0%). The false discovery rates (FDRs) of the peptide-spectra matches determined by a decoy database search were set to 1.0%. Proteins were considered to be successfully identified when at least two correct assigned peptide (95% confidence) was obtained.
Two independent experimental replicates were generated and analyzed. The Pearson's correlation coefficiency for protein ratios between two replicates was measured. The protein ratio values used in the bioinformatics analysis described below were the means of the two experimental replicates.
Bioinformatics. Biological Networks Gene Ontology (BiNGO) 3.03 was used to calculate the gene ontology (GO) term enrichment of significantly up-or down-regulated proteins (defined as quantitative ratio > 1.2 or < 0.8 and p-value < 0.05 in both replicates) and determine significantly under-and over-represented functional GO categories. The Cytoscape network visualization platform (http://www.cytoscape.org/) implementing the latest release of the BiNGO plug-in was used to identify proteins that were annotated on the basis of biological process categories. The analysis was conducted using the default BiNGO Homo sapiens database. Statistical significance was determined by means of hypergeometric analysis, followed by Benjamini and Hochberg's false discovery rate correction (p < 0.00001) 44 .
The intracellular pathway analysis was performed by using the KEGG PATHWAY database via the KEGG automatic annotation server (http://www.genome.jp/kegg/) 45 . The differentially expressed proteins matched in the KEGG PATHWAY database were counted and processed by Microsoft Office Excel.
For protein interaction network analysis of regulated proteins involved in various pathways, the Uniprot functional annotations (http://www.uniprot.org/uniprot) were used to classify the proteins into several clusters. Based on the quantified MS results, proteins matched in any clusters were extracted and submitted to STRING 9.0 (the Search Tool for the Retrieval of Interacting Genes/Proteins) to qualify the physical and functional interactions of these proteins. The proteins and their interactions were then uploaded to Cytoscape (version 2.8.3) for data visualization.