Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Identification of major hub genes involved in high-fat diet-induced obese visceral adipose tissue based on bioinformatics approach

Identification of major hub genes involved in high-fat diet-induced obese visceral adipose tissue... ADIPOCYTE 2023, VOL. 12, NO. 1, 2169227 https://doi.org/10.1080/21623945.2023.2169227 RESEARCH PAPER Identification of major hub genes involved in high-fat diet-induced obese visceral adipose tissue based on bioinformatics approach a a a b c a a a a Yu Jiang , Rui Zhang , Jia-Qi Guo , Ling-Lin Qian , Jing-Jing Ji , Ya Wu , Zhen-Jun Ji , Zi-Wei Yang , Yao Zhang , d a a Xi Chen , Gen-Shan Ma , and Yu-Yu Yao a b Department of Cardiology, Zhongda Hospital, School of Medicine, Southeast University, Nanjing, P. R. China; Department of Cardiology, Zhejiang Provincial People’s Hospital, Hangzhou, P. R. China; Department of Cardiology, Second Affiliated Hospital, College of Medicine, Zhejiang University, Hangzhou, P. R. China; Department of Cardiology, Anqing First People’s Hospital of Anhui Province, Anqing, P. R. China ABSTRACT ARTICLE HISTORY High-fat diet (HFD) can cause obesity, inducing dysregulation of the visceral adipose tissue (VAT). Received 17 October 2022 This study aimed to explore potential biological pathways and hub genes involved in obese VAT, Revised 1 January 2023 Accepted 11 January 2023 and for that, bioinformatic analysis of multiple datasets was performed. The expression profiles (GSE30247, GSE167311 and GSE79434) were downloaded from Gene Expression Omnibus. KEYWORDS Overlapping differentially expressed genes (ODEGs) between normal diet and HFD groups in Obesity; high-fat diet; GSE30247 and GSE167311 were selected to run protein–protein interaction network, GO and visceral adipose tissue; KEGG analysis. The hub genes in ODEGs were screened by Cytoscape software and further verified bioinformatics analysis; in GSE79434 and obese mouse model. A total of 747 ODEGs (599 up-regulated and 148 down- inflammation; fibrosis; regulated) were screened, and the GO and KEGG analysis showed that the up-regulated ODEGs biomarker were significantly enriched in inflammatory response and extracellular matrix receptor interaction pathways. On the other hand, the down-regulated ODEGs were involved in metabolic pathways; however, there were no significant KEGG pathways. Furthermore, six hub genes, Mki67, Rac2, Itgb2, Emr1, Tyrobp and Csf1r were acquired. These pathways and genes were verified in GSE79434 and VAT of obese mice. This study revealed that HFD induced VAT expansion, inflam- mation and fibrosis, and the hub genes could be used as therapeutic biomarkers in obesity. 1. Introduction can induce a microenvironmental disturbance in an Obesity is a worldwide threatening public health pro- early stage, in visceral adipose tissue (VAT) [6,7]. blem and is a crucial risk factor in several metabolic- Moreover, half a century ago, Jean Vague revealed related diseases [1,2]. High-fat and high-calorie diets that obese individuals who prioritized VAT expansion are common unhealthy lifestyles that contribute to had a higher risk of developing metabolic disorders a significant impact worldwide and affect all ages[3]. than obese individuals with subcutaneous adipose tis- Obesity is characterized by weight gain accompanied sue accumulation[8]. An obese mouse model fed with by excessive fat accumulation, and the expansion of a HFD could be widely used to study the pathophysio- white adipose tissue (WAT) is a mutual result of adi- logical process of obesity-related metabolic diseases. pocyte hyperplasia and hypertrophy[4]. It has been However, the gene networks involved in obese VAT described that WAT is composed of adipocytes and are complex and remain unclear[9]. Therefore, clarify- several non-adipocyte populations, designed as ing the major pathways and genes involved in obese a stromal vascular fraction (SVF), including adipocyte VAT caused by HFD could be relevant to understand precursors, vascular component cells, immune cells, the biological mechanisms associated with obesity. fibroblasts and extracellular matrix (ECM) compo- Furthermore, this study can be valuable to help identify nents[5]. Furthermore, cellular interactions in adipo- therapeutic biomarkers in obesity. cytes and SVF can contribute to a complex WAT Recently, gene expression profile analysis combined microenvironment. Some reports demonstrated that with bioinformatics has become a general and effective a stable WAT microenvironment is essential to main- methodology to investigate vital signalling pathways tain metabolic homoeostasis, and high-fat diet (HFD) and genes involved in various diseases. Therefore, CONTACT Yu-Yu Yao yaoyuyunj@hotmail.com Department of Cardiology, Zhongda Hospital, School of Medicine, Southeast University, 87 Dingjiaqiao, Nanjing 210009, Jiangsu, P. R. China Supplemental data for this article can be accessed online at https://doi.org/10.1080/21623945.2023.2169227 © 2023 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. 2 Y. JIANG ET AL. analysing bioinformatics databases can help to predict, applied to unveil the role of ODEGs. Additionally, diagnose and treat several diseases [10,11]. Although protein–protein interaction (PPI) network was single-cell sequencing has unique advantages and has employed to select the hub genes of ODEGs. become an ideal tool for single-cell research because of Moreover, the major pathways and hub genes found its high accuracy and specificity, and it can sequence in this study were validated using a third dataset and genome, transcriptome and epigenome at the single-cell eWAT of obese mice. level. However, this is precisely the deficiency of single- cell sequencing, as it cannot reflect the comprehensive effects of multiple cells in tissue. Therefore, microarray 2. Results data is a necessary and important analysis source, 2.1. GSE data quality assessment which can guide the research at the tissue level. This study was performed using a multi-step strategy After downloading the raw materials of GSE30247, (Figure 1). Three independent microarray datasets of GSE167311 and GSE79434 from GEO database, the epididymal white adipose tissue (eWAT) in obese mice data quality was assessed, and the boxplot and vioplot were enrolled for research. Then, the overlapping dif- were constructed for data standardization ferentially expressed genes (ODEGs) between normal (Supplementary File 1–3). The results from the boxplot diet (ND) and HFD groups of two datasets were demonstrated that the samples containing median, screened. Gene Ontology (GO) and Kyoto upper, and lower quartiles were similar in GSE30247, Encyclopedia of Genes and Genomes (KEGG) were GSE167311 and GSE79434 (Figure 2a,c). Additionally, Figure 1. A schematic view of the study’s procedure that combining with the analysis between GSE30247 and GSE167311, selecting major pathways and hub genes to validate in GSE79434 and the obese mouse model. eWAT: epididymal white adipose tissue; ODEGs: overlapping differentially expressed genes; MCODE: Molecular Complex Detection; PPI: protein–protein interaction; GO: Gene Ontology; KEGG: Kyoto Encyclopedia of Genes and Genomes; Mki67: monoclonal antibody Ki 67; Rac2: Rac family small GTPase 2; Itgb2: integrin beta 2; Emr1: emerin homolog 1; Tyrobp: TYRO protein tyrosine kinase binding protein; Csf1r: colony-stimulating factor-1 receptor; EMC: extracellular matrix; HFD: high fat diet. ADIPOCYTE 3 Figure 2. Quality assessment of the GSE data. (a-c) Boxplot charts for the standardized data of GSE30247, GSE167311 and GSE79434; (d-f) Vioplot charts for the standardized data of GSE30247, GSE167311 and GSE79434. The X axis contains the name of each sample, and the Y axis showed the gene expression levels for each sample. ND: normal diet. it was possible to observe that the density trends of a Venn diagram was performed to measure the inter- most gene expression levels were consistent in section in these ODEGs. In this study, 747 ODEGs GSE4648, GSE60993 and GSE79434 (Figure 2d,f). composed of 599 up-regulated and 148 down- regulated ODEGs were identified in GSE30247 and GSE167311 (Figure 3a,b). In view of the fact that the 2.2. Screening DEGs feeding environment was not completely consistent, the individual differences of organisms and the limita- In this study, some parameters were defined to screen tion of sample size, the sequencing results conducted DEGs, including |log2 (fold change) | > 1 and FDR < by different researchers were varied, so the number of 0.05. According to these criteria, the up-regulated and ODEGs obtained was usually limited (Supplementary down-regulated genes between HFD and ND groups File 4). Additionally, a volcanic map was constructed in GSE30247 and GSE167311 were evaluated, and 4 Y. JIANG ET AL. Figure 3. Volcanic and heat maps of DEGs in GSE30247 and GSE167311. (a, b) Venn diagram describing the up and down-regulated ODEGs; (c, d) Volcanic maps of all the respective genes in GSE30247 and GSE167311. Red spot: up-regulated; Green spot: down- regulated; Black spot: no significant difference; (e, f) Respective heat maps for top 40 altered genes of DGEs in GSE30247 and GSE167311. DEGs: differentially expressed genes. to identify the DEGs in GSE30247 and GSE167311. 2.3. GO and KEGG functional analysis After analysis, more up-regulated genes were detected To explore the biological functions of the ODEGs, compared to down-regulated genes in the HFD group GO and KEGG analysis was performed using the (Figure 3c,d). Furthermore, a heatmap was con- online websites DAVID and WebGestalt. The results structed with the top 40 significantly altered genes in showed that, for GO terms, the up-regulated ODEGs GSE30247 and GSE167311 to detect DEGs were involved in different biological pathways, (Figure 3e,f’). ADIPOCYTE 5 including inflammatory response, immune system in the structural construction of the extracellular process, positive regulation of tumour necrosis factor matrix, conferring tensile strength and protein bind- production and chemotaxis. Additionally, the up- ing. On the other hand, the results obtained from regulated genes were localized mainly in the extra- KEGG analysis showed that the up-regulated cellular region and cellular membrane. Furthermore, ODEGs were significantly enriched in some obesity- these up-regulated genes were molecularly involved related pathways, such as the chemokine signalling Figure 4. GO and KEGG analysis for the up-regulated ODEGs. (a) GO analysis for the up-regulated ODEGs: biological process (BP), cellular component (CC) and molecular function (MF); (b) KEGG analysis for the up-regulated ODEGs. FDR < 0.05 was considered significant. 6 Y. JIANG ET AL. Figure 5. GO and KEGG analysis for the down-regulated ODEGs. (a) GO analysis for the down-regulated ODEGs; (b) KEGG analysis for the down-regulated ODEGs. FDR < 0.05 was considered significant. pathway, cytokine–cytokine receptor interaction, and identified: monoclonal antibody Ki 67 (Mki67), Rac ECM–receptor interaction (Figure 4a,b). Down- family small GTPase 2 (Rac2), integrin beta 2 (Itgb2), regulated ODEGs were also used for GO and KEGG emerin homolog 1 (Emr1), TYRO protein tyrosine analysis, and the results demonstrated that these kinase binding protein (Tyrobp) and colony- genes were mainly involved in metabolic-related stimulating factor-1 receptor (Csf1r). Finally, a PPI net- pathways. However, there were no significant KEGG work of hub genes was constructed to reflect their pathways enriched in the down-regulated ODEGs interaction (Figure 7a,b). (Figure 5a,b). 2.5. Hub genes and major pathways verification in 2.4. PPI network of ODEGs and hub genes GSE79434 In this study, a PPI network composed of 747 ODEGs The results showed that the expression levels of the six (602 nodes and 5352 edges) was constructed to under- hub genes in GSE79434 were higher in the HFD group stand the interactions of the ODEGs. Then, MCODE than in the ND group (Figure 7c,d). Furthermore, the plug-in was performed to obtain cluster function mod- GO and KEGG analysis of the up-regulated DEGs in ules in the complex PPI networks. The results revealed GSE79434 showed that these genes were mainly that the top 2 modules with the highest score were enriched in inflammatory and chemotaxis-related path- 51.654 (53 nodes and 1343 edges) and 28.412 (35 ways, which were in agreement with the pathways nodes and 483 edges) (Figure 6a,c). Additionally, to found in GSE30247 and GSE167311 (Table 1; Table 2). identify the highly connected genes in this PPI network, cytoHubba plug-in was performed, and the top 20 2.6. Hub genes verification in obese mouse model genes were selected in each calculation method (close- ness, degree and betweenness) (Figure 6d,f). The genes The body weight of mice increased significantly after from the top 2 modules module with the highest score 2 weeks of being submitted to the HFD. Furthermore, were intersected with the top 20 genes for each of the the results showed that the average weight of mice fed three calculation methods, and six hub genes were with a HFD for 8 weeks reached 36.78 ± 0.68 g. On the ADIPOCYTE 7 Figure 6. Using the MCODE and cytoHubba plug-ins to search for hub genes in the PPI network of ODEGs. (a) PPI network of ODEGs containing 602 nodes and 5352 edges. Red circle: up-regulated; Blue circle: down-regulated; (b, c) The top 2 modules with the highest score; (d-f) The top 20 genes for each of the three calculation methods (closeness, degree and betweenness). other hand, mice fed with a ND reached 28.98 ± 0.64 g. exhibited significantly higher serum TG and TC levels Therefore, the body weight in the HFD group was than the ND group. However, no significant differences 26.92% higher than in the ND group (Figure 8a), sug- were detected in the NEFA levels between these two gesting that the obese mouse model was established and groups (Figure 8b). In addition, the validation results of could be used to verify the data. The HFD group the six hub genes showed that the mRNA levels of 8 Y. JIANG ET AL. Figure 7. Verification in GSE79434. (a) Venn diagram describing the genes from the top 2 modules module with the highest score intersecting with the top 20 genes from each of the three calculation methods; (b) The PPI network of hub genes; (c) Heatmap of hub genes expression in GSE79434; (d) Boxplot of hub genes expression in GSE79434. (*p < 0.05, **p < 0.01, ***p < 0.001 vs ND group). GSE167311). The analysis revealed 747 ODEGs com- Mki67, Itgb2, Emr1, Tyrobp and Csf1r in the HFD posed of 599 up-regulated and 148 down-regulated group were significantly increased than in the ND ODEGs. GO and KEGG analysis showed that the up- group. Rac2 mRNA had an upward trend, but no sig- regulated ODEGs were significantly enriched in inflam- nificant differences were observed between the two matory response and fibrosis pathways. On the other mice groups, indicating that a higher number of sam- hand, the down-regulated ODEGs were mainly ples should be used in further studies to confirm this involved in metabolic-related pathways. However, result (Figure 8c). there were no significant KEGG pathways. Besides, in the HFD group, H&E staining showed an Furthermore, six hub genes, Mki67, Rac2, Itgb2, increase in the adipocyte size. Masson’s trichrome Emr1, Tyrobp and Csf1r, were obtained after the PPI staining revealed an increase in the collagen, and the network analysis. In this study, the enriched pathways immunohistochemical staining of F4/80 showed in ODEGs were also confirmed using another dataset, a higher macrophage infiltration with the formation GSE79434, and obese mouse model. The results of a unique histological structure designed as crown- demonstrated that obese eWAT had an increase in like structures (CSLs) (Figure 8d). These results suggest adipocyte size, inflammation and fibrosis. that eWAT expansion in obesity is accompanied by Although these hub genes were independently found fibrosis and inflammation development. to be upregulated in obese VAT in different experi- ments, it was still unknown whether they were all up- 3. Discussion regulated at the early stage of obesity. This study proved that these six hub genes were collectively upre- This study analysed the ODEGs between two microar- gulated at the early stage of obesity (8 weeks of HFD). ray datasets of obese eWAT (GSE30247 and ADIPOCYTE 9 Table 1. GO analysis of DEGs in GSE79434. Category Pathways Biological GO:0006954~ inflammatory response Process GO:0002376~ immune system process GO:0032760~ positive regulation of tumour necrosis factor production GO:0030593~ neutrophil chemotaxis GO:0031663~ lipopolysaccharide-mediated signalling pathway GO:0006935~ chemotaxis GO:0032755~ positive regulation of interleukin-6 production GO:0050766~ positive regulation of phagocytosis GO:0032720~ negative regulation of tumour necrosis factor production GO:0070374~ positive regulation of ERK1 and ERK2 cascade GO:0042590~ antigen processing and presentation of exogenous peptide antigen via MHC class I Cellular Component GO:0016020~ membrane GO:0009897~ external side of plasma membrane GO:0009986~ cell surface GO:0005886~ plasma membrane GO:0016021~ integral component of membrane GO:0045335~ phagocytic vesicle GO:0005925~ focal adhesion GO:0005764~ lysosome GO:0015629~ actin cytoskeleton GO:0005576~ extracellular region GO:0005768~ endosome Molecular Function GO:0004888~ transmembrane signalling receptor activity GO:0031726~ CCR1 chemokine receptor binding GO:0005178~ integrin binding GO:0038023~ signalling receptor activity GO:0048020~ CCR chemokine receptor binding GO:0005515~ protein binding GO:0042056~ chemoattractant activity GO:0008009~ chemokine activity GO:0050839~ cell adhesion molecule binding GO:1990782~ protein tyrosine kinase binding GO:0005114~ type II transforming growth factor beta receptor binding Table 2. KEGG analysis of DEGs in GSE79434. described that Itgb2 is expressed in leukocytes, modu- Category Pathways lating the adhesion of leukocytes to endothelial cells KEGG mmu04380:Osteoclast differentiation mmu04145:Phagosome and promoting the recruitment of leukocytes to inflam- mmu04142:Lysosome matory regions[15]. A previous clinical trial revealed mmu05323:Rheumatoid arthritis mmu05152:Tuberculosis that the Itgb2 expression levels in SVF of obese indivi- mmu04062:Chemokine signalling pathway duals were significantly higher than in non-obese indi- mmu04640:Haematopoietic cell lineage mmu04061:Viral protein interaction with cytokine and viduals[16]. Emr1 also known as F4/80 is a marker of cytokine receptor macrophage. Elevated Emr1 in obese adipose tissue mmu04620:Toll-like receptor signalling pathway mmu05140:Leishmaniasis indicated increased macrophage infiltration[17]. mmu04060:Cytokine-cytokine receptor interaction Tyrobp is a transmembrane adaptor protein expressed in several immune cells, including T cells, B cells and macrophages, and plays an essential role activating these cells[18]. It has been described that Tyrobp was Mki67, one of the hub genes found in this study, is up-regulated in obese adipose tissue and decreased its a marker of cell proliferation, and it has been found to expression levels after bariatric surgery[19]. have higher expression levels in SVF during obesity in Additionally, Csf1r is a class III receptor tyrosine kinase response to adipose tissue expansion[12]. Furthermore, that belongs to the platelet-derived growth factor recep- the other identified five hub genes were involved in tor family. CSF1R can be found in mononuclear pha- inflammation response pathways. For example, Rac2 gocytic cells and can modulate the development, belongs to a subfamily of Ras homology (Rho) small differentiation and activation of these types of cells GTPases. A previous study demonstrated that Rac2 [20]. A previous report demonstrated that Csf1r activa- activation is essential in adjusting chemotaxis and reac- tion was positively correlated with macrophage infiltra- tive oxygen species production in phagocytic cells, tion and Csf1r inhibition can reduce adipocyte including neutrophils and macrophages, by regulating hypertrophy in HFD mice[21]. Overall, these six hub NADPH oxidase[13]. Therefore, Rac inhibition has genes were associated with WAT expansion and been associated with better outcomes in metabolic syn- immune cell infiltration. drome in obese mice[14]. Additionally, it has been 10 Y. JIANG ET AL. Figure 8. Establishment of the obese mouse model fed with the HFD for 8 weeks. (a) Weekly body weight; (b) Serum TG, TC and NEFA; (c) Validation of six hub genes in obese eWAT using qRT-PCR; (d) Sections of eWAT with H&E staining, immunohistochemistry F4/80 staining (Red arrow: CSLs) and Masson’s trichrome staining (Blue-purple: collagenous connective tissue fibres). TG: triglyceride; TC: total cholesterol; NEFA: non-esterified fatty acid; CSLs: crown-like structures. All data are expressed as mean ± SD. (*p < 0.05, **p < 0.01, ***p < 0.001 vs ND group, n = 5). [23]. Several reports have revealed that the chemokine WAT remodelling occurs during obesity and is ligand and its receptors are essential for the recruit- usually characterized by adipocyte hypertrophy, ment of macrophages during WAT, mainly the che- immune cell infiltration, fibrosis and angiogenesis mokine C–C motif ligand 2 (CCL2)/C–C chemokine [22]. Chronic low-grade inflammation of WAT caused receptor 2 (CCR2) axis [24,25]. Adipocyte death due by obesity is the origin of chronic inflammation in to metabolic stress is a universal biological phenom- metabolic syndrome. It has been described that sub- enon usually during obesity development. stantial immune cells are involved in WAT inflamma- Furthermore, previous studies demonstrated that tion, and macrophages are the most abundant and toxic lipids released by dead adipocytes could activate characteristic immune cells in this metabolic disorder ADIPOCYTE 11 macrophages. These macrophages can phagocytize expression database: Gene Expression Omnibus (GEO, dead or dying adipocytes, forming CSLs that continu- https://www.ncbi.nlm.nih.gov/geo). These databases ously release inflammatory cytokines and conse- were applied to analyse the gene expression matrixes quently induce WAT inflammation and adipocyte in eWAT of C57BL/6 mice fed the ND or HFD for death [26,27]. Therefore, the immune response and 8 weeks. inflammation are constantly induced in adipocytes and macrophages. 4.2. ODEGs analysis WAT fibrosis is characterized by excessive accumu- lation of EMC. The ECM is a non-cellular component Firstly, using the Origin 2021 software, a boxplot and mainly composed of proteoglycans and the fibrous vioplot analysis was performed to evaluate the quality proteins, the most abundant of which is collagen. of each microarray data. Then, differentially expressed Most collagen is expressed by fibroblasts, myofibro- genes (DEGs) between the ND and HFD groups in blasts, adipocyte progenitors and adipocytes. Other GSE30247 and GSE167311 were selected using the cell types, such as macrophages, also contribute to GEO2R online data analysis tool (https://www.ncbi. ECM production[28]. It has been described that under nlm.nih.gov/geo/geo2r/) which was based on physiological conditions, the ECM could maintain the GeoQuery and Limma R packages. DEGs were structural integrity of the adipocytes and modulate cell– obtained using the formula |log2 (fold change) | > 1. cell communication. However, in obesity-induced The false discovery rate (FDR) was adjusted, and WAT fibrosis, there is the accumulation of ECM, lead- a p-value < 0.05 was used as cut-off criteria. After ing to the formation of collagen bundles that stiffens deleting probes matching multiple genes, the DGEs WAT[29]. Moreover, WAT fibrosis can limit the exces- were plotted in volcanic maps, and the top 40 signifi- sive expansion of adipocytes, exerting mechanical stress cant DGEs were used to construct the corresponding on adipocytes, inhibiting the fat storage capacity of heatmap graphics. Additionally, a Venn diagram WAT, and finally leading to ectopic lipid deposition defined the intersection of up-regulated or down- [30]. WAT inflammation and fibrosis are dependent regulated ODGEs between GSE30247 and processes, and they can interact through several biolo- GSE167311(https://bioinfogp.cnb.csic.es/tools/venny/ gical mechanisms, eventually leading to WAT dysfunc- index.html). tion and systemic metabolic disorders[29]. However, the sequence and causality of these biological events 4.3. GO and KEGG analysis remain unclear. GO and KEGG analyses were performed for functional gene annotation. The GO terms are mainly composed 4. Methods of three biological categories: biological process (BP), cellular component (CC) and molecular function (MF). 4.1. Data origin and collection In this study, an online website tool designed as GSE is a dataset, which can be obtained by entering WebGestalt (http://www.webgestalt.org/) was applied keywords in a public high-throughput gene expression to analyse up-regulated and down-regulated ODEGs. database: Gene Expression Omnibus (GEO, https:// Finally, significant enrichment was determined when www.ncbi.nlm.nih.gov/geo). A GSE can contain one FDR was lower than 0.05. or more GSM samples. Each GSE represents the data of an independent sample and each sample has a GSE 4.4. PPI network ID. In this study, each GSE dataset contained GSE samples of ND and GSE samples of HFD. Using HFD An online website designed as STRING (https://string- and adipose tissue as search keywords, we screened db.org/) was used to establish a PPI network to analyse three datasets about HFD for 8 weeks feeding with the internal connection between the ODEGs. Some 60% calories from fat. The basic information of the parameters were defined to construct this PPI network: datasets were shown in Supplementary Table S1. the minimum interaction score was greater than 0.4, Microarray data of GSE30247 (https://www.ncbi.nlm. and the unconnected nodes were removed from the nih.gov/geo/query/acc.cgi?acc=GSE30247), GSE167311 analysis. Subsequently, the obtained PPI network was (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc= imported to the Cytoscape 3.9.1 software. The hub GSE167311) and GSE79434 (https://www.ncbi.nlm.nih. genes were selected by two Cytoscape plugins: gov/geo/query/acc.cgi?acc=GSE79434) were acquired Molecular Complex Detection (MCODE) and from GEO database a public high-throughput gene CytoHubba. The top 20 genes from three approaches 12 Y. JIANG ET AL. Table 3. Sequences of qRT-PCR primers. Gene Forward Sequence Reverse Sequence Species Csf1r CCGCCTGCCTGTAAAGTGGATG CCAGAGGAGGATGCCGTAGGAC Mouse Emr1 TTCCTGCTGTGTCGTGCTGTTC GCCGTCTGGTTGTCAGTCTTGTC Mouse Itgb2 AGGTCGGCAAGCAACTGATTTCC CACCAGCAGCCTCGTGACATTG Mouse Mki67 GCCTGCCCGACCCTACAAAATG CTCATCTGCTGCTGCTTCTCCTTC Mouse Rac2 ATACCGCAGGTCAGGAGGACTATG AACCACTTGGCACGGACATTCTC Mouse Tyrobp TGACACTTTCCCAAGATGCGACTG ATCAGCAGAGTCAACACCAAGTCAC Mouse non-esterified fatty acid (NEFA) levels were determined using the cytoHubba plug-in were extracted: between- by enzymatic colorimetric kits (Jian Cheng ness, degree and closeness. In this study, the hub genes Bioengineering, Nanjing, China). were obtained by intersecting the genes from three approaches with genes in the top 2 modules module with the highest MCODE score. Finally, a Venn dia- 4.6.3 Histochemistry analysis gram was constructed to analyse the data. The histochemistry analysis of haematoxylin-eosin (H&E), Masson’s trichrome and immunohistochemistry stainings was performed in this study according to 4.5. Verification of major pathways and hub genes a previous methodology[31]. After sacrificing the mice, in GSE79434 the eWAT was removed and fixed with 4% paraformal- dehyde at 4°C overnight, and then eWAT was embedded As previously described, up-regulated DEGs between in paraffin, sectioned, and stained with the correspond- the ND and HFD groups in GSE79434 were screened ing dyes (Beyotime, Shanghai, China) and the F4/80 and performed GO and KEGG analysis. Furthermore, antibody (ab6640; Abcam; 1:200). the expression levels of hub genes in GSE79434 were investigated using heatmap and boxplot drawing. 4.6.4 Real-time quantitative PCR (qRT-PCR) As previously described, the total RNA of eWAT was 4.6. Animal research extracted using TRIzol reagent (Thermo Fisher Scientific, Waltham, MA, USA). The cDNA was reverse 4.6.1 Animal transcribed, and a qRT-PCR amplification was per- The animal procedures were performed accordingly to formed using the Vazyme PCR kit (Nanjing, China) the China National Standards and Guidelines for according to the manufacturer’s instructions[32]. The Laboratory Animal Management and were approved sequences of primers are listed in Table 3. In this study, by the Animal Care and Use Committee of Southeast β-actin was used as internal control, and the relative University (No.: 20200326003). In this study, ten mRNA expression levels of the target genes were nor- 7-week-old male C57BL/6 mice were purchased from malized to the β-actin levels. The results were analysed the Changzhou Cavens Laboratory Animal Ltd. by the 2-ΔΔCt method of relative quantification. (Changzhou, China). These animals were housed at room temperature and exposed to a 12-hour light/ dark cycle and free access to food and water. After 4.7. Statistical analysis a week of ND feeding adaptation, half of these mice The data were presented as the mean ± standard devia- (n = 5/group) were submitted to a HFD containing 60% tion (SD) and analysed using SPSS 23.0 and GraphPad kcal of fat from Xietong Pharmaceutical Bio- engineering Co., Ltd. (Jiangsu, China) for 8 weeks. Prism 8.0 software. The significant differences between groups were calculated by unpaired two-tailed Student’s Additionally, the body weight of the two groups was t-test or one-way analysis of variance. A p-value < 0.05 measured regularly every week. was considered to be statistically different. 4.6.2 Serum lipids determination After 8 weeks of exposure to these diets, the mice 5. Conclusions experienced fasting overnight and were anesthetized by intraperitoneal pentobarbital (50 mg/kg) injection, In summary, in this study, six hub genes, including and whole blood was collected via the retro-orbital Mki67, Rac2, Itgb2, Emr1, Tyrobp and Csf1r were sinus plexus. This blood was then centrifuged to obtain identified which could be used as therapeutic biomar- serum. Finally, after obtaining the serum from these kers for obesity. In addition, the up-regulated ODEGs mice, the total cholesterol (TC), triglyceride (TG) and identified in this study were related to inflammatory ADIPOCYTE 13 response, chemokine signalling pathway and ECM– [6] Ibrahim MM. Subcutaneous and visceral adipose tis- sue: structural and functional differences. Obes Rev. receptor interaction. Overall, these results suggest that 2010;11(1):11–18. inflammation and fibrosis could be important patholo- [7] Jeffery E, Wing A, Holtrup B, Sebo Z, Kaplan JL, gical alterations associated with the development of Saavedra-Pena R, et al. The adipose tissue microenvir- obese eWAT. onment regulates depot-specific adipogenesis in obesity. Cell Metab. 2016;24(1):142–150. [8] Vague J. The degree of masculine differentiation of Acknowledgments obesities: a factor determining predisposition to dia- betes, atherosclerosis, gout, and uric calculous disease. We would like to thank MogoEdit (https://www.mogoedit. Am J Clin Nutr. 1956;4(1):20–34. com) for its English editing during the preparation of this [9] Unamuno X, Gomez-Ambrosi J, Rodriguez A, et al. manuscript. Adipokine dysregulation and adipose tissue inflamma- tion in human obesity. Eur J Clin Invest. 2018;48(9): e12997. Disclosure statement [10] Sarafidis M, Lambrou GI, Zoumpourlis V, et al. An integrated bioinformatics analysis towards the identifi- No potential conflict of interest was reported by the cation of diagnostic, prognostic, and predictive key author(s). biomarkers for urinary bladder cancer. Cancers (Basel). 2022;15(1):14. Funding [11] Sai Swaroop R, Akhil PS, Sai Sanwid P, et al. Integrated multi-omic data analysis and validation with yeast model This research was supported by the National Natural Science show oxidative phosphorylation modulates protein aggre- Foundation of China [grant number NSFC 81770452]. gation in amyotrophic lateral sclerosis. J Biomol Struct Dyn. 2022;1–20. doi: 10.1080/07391102.2022.2090441. [12] Haim Y, Bluher M, Slutsky N, et al. Elevated autophagy Institutional review board statement gene expression in adipose tissue of obese humans: a potential non-cell-cycle-dependent function of The animal procedures were conducted following the China E2F1. Autophagy. 2015;11(11):2074–2088. National Standards and Guidelines for Laboratory Animal [13] Aflaki E, Balenga NA, Luschnig-Schratl P, et al. Management and were approved by the Animal Care and Impaired Rho GTPase activation abrogates cell polar- Use Committee of Southeast University (No.: 20200326003). ization and migration in macrophages with defective lipolysis. Cell Mol Life Sci. 2011;68(23):3933–3947. [14] Ko SH, Lee JK, Lee HJ, et al. 8-Oxo-2’-deoxyguanosine Data availability statement ameliorates features of metabolic syndrome in obese mice. All of the data are presented in the manuscript. Biochem Biophys Res Commun. 2014;443(2):610–616. [15] Sun H, Hu L, Fan Z. β2 integrin activation and signal transduction in leukocyte recruitment. Am J Physiol ORCID Cell Physiol. 2021;321(2):C308–C16. [16] Nair S, Lee YH, Rousseau E, et al. Increased expression Yu-Yu Yao http://orcid.org/0000-0001-7841-288X of inflammation-related genes in cultured preadipo- cytes/stromal vascular cells from obese compared with non-obese Pima Indians. Diabetologia. 2005;48 References (9):1784–1788. [17] Chen Q, Ruedl C. Obesity retunes turnover kinetics of [1] Collaboration NCDRF. Trends in adult body-mass tissue-resident macrophages in fat. J Leukoc Biol. index in 200 countries from 1975 to 2014: a pooled 2020;107(5):773–782. analysis of 1698 population-based measurement studies [18] Haure-Mirande JV, Audrain M, Ehrlich ME, et al. with 19.2 million participants. Lancet. 2016;387 TYROBP/DAP12 in Alzheimer’s disease: transduction (10026):1377–1396. of physiological and pathological signals across [2] Ahmed B, Sultana R, Greene MW. Adipose tissue and TREM2. Mol Neurodegener. 2022;17(1):55. insulin resistance in obese. Biomed Pharmacother. [19] Yan K, Zhang P, Jin J, et al. Integrative analyses of hub 2021;137:111315. genes and their association with immune infiltration in [3] Jayedi A, Khan TA, Aune D, et al. Body fat and risk of adipose tissue, liver tissue and skeletal muscle of obese all-cause mortality: a systematic review and patients after bariatric surgery. Adipocyte. 2022;11 dose-response meta-analysis of prospective cohort (1):190–201. studies. Int J Obes (Lond). 2022;46(9):1573–1581. [20] Sletta KY, Castells O, Gjertsen BT. Colony stimulating [4] Gugliucci A. Biomarkers of dysfunctional visceral fat. factor 1 receptor in acute myeloid leukemia. Front Adv Clin Chem. 2022;109:1–30. Oncol. 2021;11:654817. [5] Sun Y, Chen S, Zhang X, et al. Significance of cellular [21] Cox N, Crozet L, Holtman IR, et al. Diet-regulated cross-talk in stromal vascular fraction of adipose tissue production of PDGFcc by macrophages controls in neovascularization. Arterioscler Thromb Vasc Biol. energy storage. Science. 2021;373(6553):373. 2019;39(6):1034–1044. 14 Y. JIANG ET AL. [22] Lecoutre S, Lambert M, Drygalski K, et al. Importance adipose tissue of obese mice and humans. J Lipid Res. of the Microenvironment and Mechanosensing in 2005;46(11):2347–2355. Adipose Tissue Biology. Cells. 2022;12(1):11. [28] Datta R, Podolsky MJ, Atabai K. Fat fibrosis: friend or [23] Hachiya R, Tanaka M, Itoh M, et al. Molecular foe?. JCI Insight. 2018;3(19). doi:10.1172/jci.insight. mechanism of crosstalk between immune and meta- 122289. bolic systems in metabolic syndrome. Inflamm Regen. [29] Marcelin G, Gautier EL, Clement K. Adipose tissue 2022;42(1):13. fibrosis in obesity: etiology and challenges. Annu Rev [24] Kanda H, Tateya S, Tamori Y, et al. MCP-1 contributes Physiol. 2022;84(1):135–155. to macrophage infiltration into adipose tissue, insulin [30] Marcelin G, Clement K. The multifaceted progenitor resistance, and hepatic steatosis in obesity. J Clin fates in healthy or unhealthy adipose tissue during Invest. 2006;116(6):1494–1505. obesity. Rev Endocr Metab Disord. 2021;22 [25] Weisberg SP, Hunter D, Huber R, et al. CCR2 mod- (4):1111–1119. ulates inflammatory and metabolic effects of high-fat [31] Ji JJ, Qian LL, Zhu Y, et al. Kallistatin/Serpina3c feeding. J Clin Invest. 2006;116(1):115–124. inhibits cardiac fibrosis after myocardial infarction [26] Lindhorst A, Raulien N, Wieghofer P, et al. Adipocyte by regulating glycolysis via Nr4a1 activation. death triggers a pro-inflammatory response and Biochim Biophys Acta Mol Basis Dis. 2022;1868 induces metabolic activation of resident macrophages. (9):166441. Cell Death Dis. 2021;12(6):579. [32] Guo J, Qian L, Ji J, et al. Serpina3c regulates adipose [27] Cinti S, Mitchell G, Barbatelli G, et al. Adipocyte death differentiation via the Wnt/beta-catenin-PPARgamma defines macrophage localization and function in pathway. Cell Signal. 2022;93:110299. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Adipocyte Taylor & Francis

Identification of major hub genes involved in high-fat diet-induced obese visceral adipose tissue based on bioinformatics approach

Loading next page...
 
/lp/taylor-francis/identification-of-major-hub-genes-involved-in-high-fat-diet-induced-DQkMQ3FEO5

References (36)

Publisher
Taylor & Francis
Copyright
© 2023 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group.
ISSN
2162-3945
eISSN
2162-397X
DOI
10.1080/21623945.2023.2169227
Publisher site
See Article on Publisher Site

Abstract

ADIPOCYTE 2023, VOL. 12, NO. 1, 2169227 https://doi.org/10.1080/21623945.2023.2169227 RESEARCH PAPER Identification of major hub genes involved in high-fat diet-induced obese visceral adipose tissue based on bioinformatics approach a a a b c a a a a Yu Jiang , Rui Zhang , Jia-Qi Guo , Ling-Lin Qian , Jing-Jing Ji , Ya Wu , Zhen-Jun Ji , Zi-Wei Yang , Yao Zhang , d a a Xi Chen , Gen-Shan Ma , and Yu-Yu Yao a b Department of Cardiology, Zhongda Hospital, School of Medicine, Southeast University, Nanjing, P. R. China; Department of Cardiology, Zhejiang Provincial People’s Hospital, Hangzhou, P. R. China; Department of Cardiology, Second Affiliated Hospital, College of Medicine, Zhejiang University, Hangzhou, P. R. China; Department of Cardiology, Anqing First People’s Hospital of Anhui Province, Anqing, P. R. China ABSTRACT ARTICLE HISTORY High-fat diet (HFD) can cause obesity, inducing dysregulation of the visceral adipose tissue (VAT). Received 17 October 2022 This study aimed to explore potential biological pathways and hub genes involved in obese VAT, Revised 1 January 2023 Accepted 11 January 2023 and for that, bioinformatic analysis of multiple datasets was performed. The expression profiles (GSE30247, GSE167311 and GSE79434) were downloaded from Gene Expression Omnibus. KEYWORDS Overlapping differentially expressed genes (ODEGs) between normal diet and HFD groups in Obesity; high-fat diet; GSE30247 and GSE167311 were selected to run protein–protein interaction network, GO and visceral adipose tissue; KEGG analysis. The hub genes in ODEGs were screened by Cytoscape software and further verified bioinformatics analysis; in GSE79434 and obese mouse model. A total of 747 ODEGs (599 up-regulated and 148 down- inflammation; fibrosis; regulated) were screened, and the GO and KEGG analysis showed that the up-regulated ODEGs biomarker were significantly enriched in inflammatory response and extracellular matrix receptor interaction pathways. On the other hand, the down-regulated ODEGs were involved in metabolic pathways; however, there were no significant KEGG pathways. Furthermore, six hub genes, Mki67, Rac2, Itgb2, Emr1, Tyrobp and Csf1r were acquired. These pathways and genes were verified in GSE79434 and VAT of obese mice. This study revealed that HFD induced VAT expansion, inflam- mation and fibrosis, and the hub genes could be used as therapeutic biomarkers in obesity. 1. Introduction can induce a microenvironmental disturbance in an Obesity is a worldwide threatening public health pro- early stage, in visceral adipose tissue (VAT) [6,7]. blem and is a crucial risk factor in several metabolic- Moreover, half a century ago, Jean Vague revealed related diseases [1,2]. High-fat and high-calorie diets that obese individuals who prioritized VAT expansion are common unhealthy lifestyles that contribute to had a higher risk of developing metabolic disorders a significant impact worldwide and affect all ages[3]. than obese individuals with subcutaneous adipose tis- Obesity is characterized by weight gain accompanied sue accumulation[8]. An obese mouse model fed with by excessive fat accumulation, and the expansion of a HFD could be widely used to study the pathophysio- white adipose tissue (WAT) is a mutual result of adi- logical process of obesity-related metabolic diseases. pocyte hyperplasia and hypertrophy[4]. It has been However, the gene networks involved in obese VAT described that WAT is composed of adipocytes and are complex and remain unclear[9]. Therefore, clarify- several non-adipocyte populations, designed as ing the major pathways and genes involved in obese a stromal vascular fraction (SVF), including adipocyte VAT caused by HFD could be relevant to understand precursors, vascular component cells, immune cells, the biological mechanisms associated with obesity. fibroblasts and extracellular matrix (ECM) compo- Furthermore, this study can be valuable to help identify nents[5]. Furthermore, cellular interactions in adipo- therapeutic biomarkers in obesity. cytes and SVF can contribute to a complex WAT Recently, gene expression profile analysis combined microenvironment. Some reports demonstrated that with bioinformatics has become a general and effective a stable WAT microenvironment is essential to main- methodology to investigate vital signalling pathways tain metabolic homoeostasis, and high-fat diet (HFD) and genes involved in various diseases. Therefore, CONTACT Yu-Yu Yao yaoyuyunj@hotmail.com Department of Cardiology, Zhongda Hospital, School of Medicine, Southeast University, 87 Dingjiaqiao, Nanjing 210009, Jiangsu, P. R. China Supplemental data for this article can be accessed online at https://doi.org/10.1080/21623945.2023.2169227 © 2023 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. 2 Y. JIANG ET AL. analysing bioinformatics databases can help to predict, applied to unveil the role of ODEGs. Additionally, diagnose and treat several diseases [10,11]. Although protein–protein interaction (PPI) network was single-cell sequencing has unique advantages and has employed to select the hub genes of ODEGs. become an ideal tool for single-cell research because of Moreover, the major pathways and hub genes found its high accuracy and specificity, and it can sequence in this study were validated using a third dataset and genome, transcriptome and epigenome at the single-cell eWAT of obese mice. level. However, this is precisely the deficiency of single- cell sequencing, as it cannot reflect the comprehensive effects of multiple cells in tissue. Therefore, microarray 2. Results data is a necessary and important analysis source, 2.1. GSE data quality assessment which can guide the research at the tissue level. This study was performed using a multi-step strategy After downloading the raw materials of GSE30247, (Figure 1). Three independent microarray datasets of GSE167311 and GSE79434 from GEO database, the epididymal white adipose tissue (eWAT) in obese mice data quality was assessed, and the boxplot and vioplot were enrolled for research. Then, the overlapping dif- were constructed for data standardization ferentially expressed genes (ODEGs) between normal (Supplementary File 1–3). The results from the boxplot diet (ND) and HFD groups of two datasets were demonstrated that the samples containing median, screened. Gene Ontology (GO) and Kyoto upper, and lower quartiles were similar in GSE30247, Encyclopedia of Genes and Genomes (KEGG) were GSE167311 and GSE79434 (Figure 2a,c). Additionally, Figure 1. A schematic view of the study’s procedure that combining with the analysis between GSE30247 and GSE167311, selecting major pathways and hub genes to validate in GSE79434 and the obese mouse model. eWAT: epididymal white adipose tissue; ODEGs: overlapping differentially expressed genes; MCODE: Molecular Complex Detection; PPI: protein–protein interaction; GO: Gene Ontology; KEGG: Kyoto Encyclopedia of Genes and Genomes; Mki67: monoclonal antibody Ki 67; Rac2: Rac family small GTPase 2; Itgb2: integrin beta 2; Emr1: emerin homolog 1; Tyrobp: TYRO protein tyrosine kinase binding protein; Csf1r: colony-stimulating factor-1 receptor; EMC: extracellular matrix; HFD: high fat diet. ADIPOCYTE 3 Figure 2. Quality assessment of the GSE data. (a-c) Boxplot charts for the standardized data of GSE30247, GSE167311 and GSE79434; (d-f) Vioplot charts for the standardized data of GSE30247, GSE167311 and GSE79434. The X axis contains the name of each sample, and the Y axis showed the gene expression levels for each sample. ND: normal diet. it was possible to observe that the density trends of a Venn diagram was performed to measure the inter- most gene expression levels were consistent in section in these ODEGs. In this study, 747 ODEGs GSE4648, GSE60993 and GSE79434 (Figure 2d,f). composed of 599 up-regulated and 148 down- regulated ODEGs were identified in GSE30247 and GSE167311 (Figure 3a,b). In view of the fact that the 2.2. Screening DEGs feeding environment was not completely consistent, the individual differences of organisms and the limita- In this study, some parameters were defined to screen tion of sample size, the sequencing results conducted DEGs, including |log2 (fold change) | > 1 and FDR < by different researchers were varied, so the number of 0.05. According to these criteria, the up-regulated and ODEGs obtained was usually limited (Supplementary down-regulated genes between HFD and ND groups File 4). Additionally, a volcanic map was constructed in GSE30247 and GSE167311 were evaluated, and 4 Y. JIANG ET AL. Figure 3. Volcanic and heat maps of DEGs in GSE30247 and GSE167311. (a, b) Venn diagram describing the up and down-regulated ODEGs; (c, d) Volcanic maps of all the respective genes in GSE30247 and GSE167311. Red spot: up-regulated; Green spot: down- regulated; Black spot: no significant difference; (e, f) Respective heat maps for top 40 altered genes of DGEs in GSE30247 and GSE167311. DEGs: differentially expressed genes. to identify the DEGs in GSE30247 and GSE167311. 2.3. GO and KEGG functional analysis After analysis, more up-regulated genes were detected To explore the biological functions of the ODEGs, compared to down-regulated genes in the HFD group GO and KEGG analysis was performed using the (Figure 3c,d). Furthermore, a heatmap was con- online websites DAVID and WebGestalt. The results structed with the top 40 significantly altered genes in showed that, for GO terms, the up-regulated ODEGs GSE30247 and GSE167311 to detect DEGs were involved in different biological pathways, (Figure 3e,f’). ADIPOCYTE 5 including inflammatory response, immune system in the structural construction of the extracellular process, positive regulation of tumour necrosis factor matrix, conferring tensile strength and protein bind- production and chemotaxis. Additionally, the up- ing. On the other hand, the results obtained from regulated genes were localized mainly in the extra- KEGG analysis showed that the up-regulated cellular region and cellular membrane. Furthermore, ODEGs were significantly enriched in some obesity- these up-regulated genes were molecularly involved related pathways, such as the chemokine signalling Figure 4. GO and KEGG analysis for the up-regulated ODEGs. (a) GO analysis for the up-regulated ODEGs: biological process (BP), cellular component (CC) and molecular function (MF); (b) KEGG analysis for the up-regulated ODEGs. FDR < 0.05 was considered significant. 6 Y. JIANG ET AL. Figure 5. GO and KEGG analysis for the down-regulated ODEGs. (a) GO analysis for the down-regulated ODEGs; (b) KEGG analysis for the down-regulated ODEGs. FDR < 0.05 was considered significant. pathway, cytokine–cytokine receptor interaction, and identified: monoclonal antibody Ki 67 (Mki67), Rac ECM–receptor interaction (Figure 4a,b). Down- family small GTPase 2 (Rac2), integrin beta 2 (Itgb2), regulated ODEGs were also used for GO and KEGG emerin homolog 1 (Emr1), TYRO protein tyrosine analysis, and the results demonstrated that these kinase binding protein (Tyrobp) and colony- genes were mainly involved in metabolic-related stimulating factor-1 receptor (Csf1r). Finally, a PPI net- pathways. However, there were no significant KEGG work of hub genes was constructed to reflect their pathways enriched in the down-regulated ODEGs interaction (Figure 7a,b). (Figure 5a,b). 2.5. Hub genes and major pathways verification in 2.4. PPI network of ODEGs and hub genes GSE79434 In this study, a PPI network composed of 747 ODEGs The results showed that the expression levels of the six (602 nodes and 5352 edges) was constructed to under- hub genes in GSE79434 were higher in the HFD group stand the interactions of the ODEGs. Then, MCODE than in the ND group (Figure 7c,d). Furthermore, the plug-in was performed to obtain cluster function mod- GO and KEGG analysis of the up-regulated DEGs in ules in the complex PPI networks. The results revealed GSE79434 showed that these genes were mainly that the top 2 modules with the highest score were enriched in inflammatory and chemotaxis-related path- 51.654 (53 nodes and 1343 edges) and 28.412 (35 ways, which were in agreement with the pathways nodes and 483 edges) (Figure 6a,c). Additionally, to found in GSE30247 and GSE167311 (Table 1; Table 2). identify the highly connected genes in this PPI network, cytoHubba plug-in was performed, and the top 20 2.6. Hub genes verification in obese mouse model genes were selected in each calculation method (close- ness, degree and betweenness) (Figure 6d,f). The genes The body weight of mice increased significantly after from the top 2 modules module with the highest score 2 weeks of being submitted to the HFD. Furthermore, were intersected with the top 20 genes for each of the the results showed that the average weight of mice fed three calculation methods, and six hub genes were with a HFD for 8 weeks reached 36.78 ± 0.68 g. On the ADIPOCYTE 7 Figure 6. Using the MCODE and cytoHubba plug-ins to search for hub genes in the PPI network of ODEGs. (a) PPI network of ODEGs containing 602 nodes and 5352 edges. Red circle: up-regulated; Blue circle: down-regulated; (b, c) The top 2 modules with the highest score; (d-f) The top 20 genes for each of the three calculation methods (closeness, degree and betweenness). other hand, mice fed with a ND reached 28.98 ± 0.64 g. exhibited significantly higher serum TG and TC levels Therefore, the body weight in the HFD group was than the ND group. However, no significant differences 26.92% higher than in the ND group (Figure 8a), sug- were detected in the NEFA levels between these two gesting that the obese mouse model was established and groups (Figure 8b). In addition, the validation results of could be used to verify the data. The HFD group the six hub genes showed that the mRNA levels of 8 Y. JIANG ET AL. Figure 7. Verification in GSE79434. (a) Venn diagram describing the genes from the top 2 modules module with the highest score intersecting with the top 20 genes from each of the three calculation methods; (b) The PPI network of hub genes; (c) Heatmap of hub genes expression in GSE79434; (d) Boxplot of hub genes expression in GSE79434. (*p < 0.05, **p < 0.01, ***p < 0.001 vs ND group). GSE167311). The analysis revealed 747 ODEGs com- Mki67, Itgb2, Emr1, Tyrobp and Csf1r in the HFD posed of 599 up-regulated and 148 down-regulated group were significantly increased than in the ND ODEGs. GO and KEGG analysis showed that the up- group. Rac2 mRNA had an upward trend, but no sig- regulated ODEGs were significantly enriched in inflam- nificant differences were observed between the two matory response and fibrosis pathways. On the other mice groups, indicating that a higher number of sam- hand, the down-regulated ODEGs were mainly ples should be used in further studies to confirm this involved in metabolic-related pathways. However, result (Figure 8c). there were no significant KEGG pathways. Besides, in the HFD group, H&E staining showed an Furthermore, six hub genes, Mki67, Rac2, Itgb2, increase in the adipocyte size. Masson’s trichrome Emr1, Tyrobp and Csf1r, were obtained after the PPI staining revealed an increase in the collagen, and the network analysis. In this study, the enriched pathways immunohistochemical staining of F4/80 showed in ODEGs were also confirmed using another dataset, a higher macrophage infiltration with the formation GSE79434, and obese mouse model. The results of a unique histological structure designed as crown- demonstrated that obese eWAT had an increase in like structures (CSLs) (Figure 8d). These results suggest adipocyte size, inflammation and fibrosis. that eWAT expansion in obesity is accompanied by Although these hub genes were independently found fibrosis and inflammation development. to be upregulated in obese VAT in different experi- ments, it was still unknown whether they were all up- 3. Discussion regulated at the early stage of obesity. This study proved that these six hub genes were collectively upre- This study analysed the ODEGs between two microar- gulated at the early stage of obesity (8 weeks of HFD). ray datasets of obese eWAT (GSE30247 and ADIPOCYTE 9 Table 1. GO analysis of DEGs in GSE79434. Category Pathways Biological GO:0006954~ inflammatory response Process GO:0002376~ immune system process GO:0032760~ positive regulation of tumour necrosis factor production GO:0030593~ neutrophil chemotaxis GO:0031663~ lipopolysaccharide-mediated signalling pathway GO:0006935~ chemotaxis GO:0032755~ positive regulation of interleukin-6 production GO:0050766~ positive regulation of phagocytosis GO:0032720~ negative regulation of tumour necrosis factor production GO:0070374~ positive regulation of ERK1 and ERK2 cascade GO:0042590~ antigen processing and presentation of exogenous peptide antigen via MHC class I Cellular Component GO:0016020~ membrane GO:0009897~ external side of plasma membrane GO:0009986~ cell surface GO:0005886~ plasma membrane GO:0016021~ integral component of membrane GO:0045335~ phagocytic vesicle GO:0005925~ focal adhesion GO:0005764~ lysosome GO:0015629~ actin cytoskeleton GO:0005576~ extracellular region GO:0005768~ endosome Molecular Function GO:0004888~ transmembrane signalling receptor activity GO:0031726~ CCR1 chemokine receptor binding GO:0005178~ integrin binding GO:0038023~ signalling receptor activity GO:0048020~ CCR chemokine receptor binding GO:0005515~ protein binding GO:0042056~ chemoattractant activity GO:0008009~ chemokine activity GO:0050839~ cell adhesion molecule binding GO:1990782~ protein tyrosine kinase binding GO:0005114~ type II transforming growth factor beta receptor binding Table 2. KEGG analysis of DEGs in GSE79434. described that Itgb2 is expressed in leukocytes, modu- Category Pathways lating the adhesion of leukocytes to endothelial cells KEGG mmu04380:Osteoclast differentiation mmu04145:Phagosome and promoting the recruitment of leukocytes to inflam- mmu04142:Lysosome matory regions[15]. A previous clinical trial revealed mmu05323:Rheumatoid arthritis mmu05152:Tuberculosis that the Itgb2 expression levels in SVF of obese indivi- mmu04062:Chemokine signalling pathway duals were significantly higher than in non-obese indi- mmu04640:Haematopoietic cell lineage mmu04061:Viral protein interaction with cytokine and viduals[16]. Emr1 also known as F4/80 is a marker of cytokine receptor macrophage. Elevated Emr1 in obese adipose tissue mmu04620:Toll-like receptor signalling pathway mmu05140:Leishmaniasis indicated increased macrophage infiltration[17]. mmu04060:Cytokine-cytokine receptor interaction Tyrobp is a transmembrane adaptor protein expressed in several immune cells, including T cells, B cells and macrophages, and plays an essential role activating these cells[18]. It has been described that Tyrobp was Mki67, one of the hub genes found in this study, is up-regulated in obese adipose tissue and decreased its a marker of cell proliferation, and it has been found to expression levels after bariatric surgery[19]. have higher expression levels in SVF during obesity in Additionally, Csf1r is a class III receptor tyrosine kinase response to adipose tissue expansion[12]. Furthermore, that belongs to the platelet-derived growth factor recep- the other identified five hub genes were involved in tor family. CSF1R can be found in mononuclear pha- inflammation response pathways. For example, Rac2 gocytic cells and can modulate the development, belongs to a subfamily of Ras homology (Rho) small differentiation and activation of these types of cells GTPases. A previous study demonstrated that Rac2 [20]. A previous report demonstrated that Csf1r activa- activation is essential in adjusting chemotaxis and reac- tion was positively correlated with macrophage infiltra- tive oxygen species production in phagocytic cells, tion and Csf1r inhibition can reduce adipocyte including neutrophils and macrophages, by regulating hypertrophy in HFD mice[21]. Overall, these six hub NADPH oxidase[13]. Therefore, Rac inhibition has genes were associated with WAT expansion and been associated with better outcomes in metabolic syn- immune cell infiltration. drome in obese mice[14]. Additionally, it has been 10 Y. JIANG ET AL. Figure 8. Establishment of the obese mouse model fed with the HFD for 8 weeks. (a) Weekly body weight; (b) Serum TG, TC and NEFA; (c) Validation of six hub genes in obese eWAT using qRT-PCR; (d) Sections of eWAT with H&E staining, immunohistochemistry F4/80 staining (Red arrow: CSLs) and Masson’s trichrome staining (Blue-purple: collagenous connective tissue fibres). TG: triglyceride; TC: total cholesterol; NEFA: non-esterified fatty acid; CSLs: crown-like structures. All data are expressed as mean ± SD. (*p < 0.05, **p < 0.01, ***p < 0.001 vs ND group, n = 5). [23]. Several reports have revealed that the chemokine WAT remodelling occurs during obesity and is ligand and its receptors are essential for the recruit- usually characterized by adipocyte hypertrophy, ment of macrophages during WAT, mainly the che- immune cell infiltration, fibrosis and angiogenesis mokine C–C motif ligand 2 (CCL2)/C–C chemokine [22]. Chronic low-grade inflammation of WAT caused receptor 2 (CCR2) axis [24,25]. Adipocyte death due by obesity is the origin of chronic inflammation in to metabolic stress is a universal biological phenom- metabolic syndrome. It has been described that sub- enon usually during obesity development. stantial immune cells are involved in WAT inflamma- Furthermore, previous studies demonstrated that tion, and macrophages are the most abundant and toxic lipids released by dead adipocytes could activate characteristic immune cells in this metabolic disorder ADIPOCYTE 11 macrophages. These macrophages can phagocytize expression database: Gene Expression Omnibus (GEO, dead or dying adipocytes, forming CSLs that continu- https://www.ncbi.nlm.nih.gov/geo). These databases ously release inflammatory cytokines and conse- were applied to analyse the gene expression matrixes quently induce WAT inflammation and adipocyte in eWAT of C57BL/6 mice fed the ND or HFD for death [26,27]. Therefore, the immune response and 8 weeks. inflammation are constantly induced in adipocytes and macrophages. 4.2. ODEGs analysis WAT fibrosis is characterized by excessive accumu- lation of EMC. The ECM is a non-cellular component Firstly, using the Origin 2021 software, a boxplot and mainly composed of proteoglycans and the fibrous vioplot analysis was performed to evaluate the quality proteins, the most abundant of which is collagen. of each microarray data. Then, differentially expressed Most collagen is expressed by fibroblasts, myofibro- genes (DEGs) between the ND and HFD groups in blasts, adipocyte progenitors and adipocytes. Other GSE30247 and GSE167311 were selected using the cell types, such as macrophages, also contribute to GEO2R online data analysis tool (https://www.ncbi. ECM production[28]. It has been described that under nlm.nih.gov/geo/geo2r/) which was based on physiological conditions, the ECM could maintain the GeoQuery and Limma R packages. DEGs were structural integrity of the adipocytes and modulate cell– obtained using the formula |log2 (fold change) | > 1. cell communication. However, in obesity-induced The false discovery rate (FDR) was adjusted, and WAT fibrosis, there is the accumulation of ECM, lead- a p-value < 0.05 was used as cut-off criteria. After ing to the formation of collagen bundles that stiffens deleting probes matching multiple genes, the DGEs WAT[29]. Moreover, WAT fibrosis can limit the exces- were plotted in volcanic maps, and the top 40 signifi- sive expansion of adipocytes, exerting mechanical stress cant DGEs were used to construct the corresponding on adipocytes, inhibiting the fat storage capacity of heatmap graphics. Additionally, a Venn diagram WAT, and finally leading to ectopic lipid deposition defined the intersection of up-regulated or down- [30]. WAT inflammation and fibrosis are dependent regulated ODGEs between GSE30247 and processes, and they can interact through several biolo- GSE167311(https://bioinfogp.cnb.csic.es/tools/venny/ gical mechanisms, eventually leading to WAT dysfunc- index.html). tion and systemic metabolic disorders[29]. However, the sequence and causality of these biological events 4.3. GO and KEGG analysis remain unclear. GO and KEGG analyses were performed for functional gene annotation. The GO terms are mainly composed 4. Methods of three biological categories: biological process (BP), cellular component (CC) and molecular function (MF). 4.1. Data origin and collection In this study, an online website tool designed as GSE is a dataset, which can be obtained by entering WebGestalt (http://www.webgestalt.org/) was applied keywords in a public high-throughput gene expression to analyse up-regulated and down-regulated ODEGs. database: Gene Expression Omnibus (GEO, https:// Finally, significant enrichment was determined when www.ncbi.nlm.nih.gov/geo). A GSE can contain one FDR was lower than 0.05. or more GSM samples. Each GSE represents the data of an independent sample and each sample has a GSE 4.4. PPI network ID. In this study, each GSE dataset contained GSE samples of ND and GSE samples of HFD. Using HFD An online website designed as STRING (https://string- and adipose tissue as search keywords, we screened db.org/) was used to establish a PPI network to analyse three datasets about HFD for 8 weeks feeding with the internal connection between the ODEGs. Some 60% calories from fat. The basic information of the parameters were defined to construct this PPI network: datasets were shown in Supplementary Table S1. the minimum interaction score was greater than 0.4, Microarray data of GSE30247 (https://www.ncbi.nlm. and the unconnected nodes were removed from the nih.gov/geo/query/acc.cgi?acc=GSE30247), GSE167311 analysis. Subsequently, the obtained PPI network was (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc= imported to the Cytoscape 3.9.1 software. The hub GSE167311) and GSE79434 (https://www.ncbi.nlm.nih. genes were selected by two Cytoscape plugins: gov/geo/query/acc.cgi?acc=GSE79434) were acquired Molecular Complex Detection (MCODE) and from GEO database a public high-throughput gene CytoHubba. The top 20 genes from three approaches 12 Y. JIANG ET AL. Table 3. Sequences of qRT-PCR primers. Gene Forward Sequence Reverse Sequence Species Csf1r CCGCCTGCCTGTAAAGTGGATG CCAGAGGAGGATGCCGTAGGAC Mouse Emr1 TTCCTGCTGTGTCGTGCTGTTC GCCGTCTGGTTGTCAGTCTTGTC Mouse Itgb2 AGGTCGGCAAGCAACTGATTTCC CACCAGCAGCCTCGTGACATTG Mouse Mki67 GCCTGCCCGACCCTACAAAATG CTCATCTGCTGCTGCTTCTCCTTC Mouse Rac2 ATACCGCAGGTCAGGAGGACTATG AACCACTTGGCACGGACATTCTC Mouse Tyrobp TGACACTTTCCCAAGATGCGACTG ATCAGCAGAGTCAACACCAAGTCAC Mouse non-esterified fatty acid (NEFA) levels were determined using the cytoHubba plug-in were extracted: between- by enzymatic colorimetric kits (Jian Cheng ness, degree and closeness. In this study, the hub genes Bioengineering, Nanjing, China). were obtained by intersecting the genes from three approaches with genes in the top 2 modules module with the highest MCODE score. Finally, a Venn dia- 4.6.3 Histochemistry analysis gram was constructed to analyse the data. The histochemistry analysis of haematoxylin-eosin (H&E), Masson’s trichrome and immunohistochemistry stainings was performed in this study according to 4.5. Verification of major pathways and hub genes a previous methodology[31]. After sacrificing the mice, in GSE79434 the eWAT was removed and fixed with 4% paraformal- dehyde at 4°C overnight, and then eWAT was embedded As previously described, up-regulated DEGs between in paraffin, sectioned, and stained with the correspond- the ND and HFD groups in GSE79434 were screened ing dyes (Beyotime, Shanghai, China) and the F4/80 and performed GO and KEGG analysis. Furthermore, antibody (ab6640; Abcam; 1:200). the expression levels of hub genes in GSE79434 were investigated using heatmap and boxplot drawing. 4.6.4 Real-time quantitative PCR (qRT-PCR) As previously described, the total RNA of eWAT was 4.6. Animal research extracted using TRIzol reagent (Thermo Fisher Scientific, Waltham, MA, USA). The cDNA was reverse 4.6.1 Animal transcribed, and a qRT-PCR amplification was per- The animal procedures were performed accordingly to formed using the Vazyme PCR kit (Nanjing, China) the China National Standards and Guidelines for according to the manufacturer’s instructions[32]. The Laboratory Animal Management and were approved sequences of primers are listed in Table 3. In this study, by the Animal Care and Use Committee of Southeast β-actin was used as internal control, and the relative University (No.: 20200326003). In this study, ten mRNA expression levels of the target genes were nor- 7-week-old male C57BL/6 mice were purchased from malized to the β-actin levels. The results were analysed the Changzhou Cavens Laboratory Animal Ltd. by the 2-ΔΔCt method of relative quantification. (Changzhou, China). These animals were housed at room temperature and exposed to a 12-hour light/ dark cycle and free access to food and water. After 4.7. Statistical analysis a week of ND feeding adaptation, half of these mice The data were presented as the mean ± standard devia- (n = 5/group) were submitted to a HFD containing 60% tion (SD) and analysed using SPSS 23.0 and GraphPad kcal of fat from Xietong Pharmaceutical Bio- engineering Co., Ltd. (Jiangsu, China) for 8 weeks. Prism 8.0 software. The significant differences between groups were calculated by unpaired two-tailed Student’s Additionally, the body weight of the two groups was t-test or one-way analysis of variance. A p-value < 0.05 measured regularly every week. was considered to be statistically different. 4.6.2 Serum lipids determination After 8 weeks of exposure to these diets, the mice 5. Conclusions experienced fasting overnight and were anesthetized by intraperitoneal pentobarbital (50 mg/kg) injection, In summary, in this study, six hub genes, including and whole blood was collected via the retro-orbital Mki67, Rac2, Itgb2, Emr1, Tyrobp and Csf1r were sinus plexus. This blood was then centrifuged to obtain identified which could be used as therapeutic biomar- serum. Finally, after obtaining the serum from these kers for obesity. In addition, the up-regulated ODEGs mice, the total cholesterol (TC), triglyceride (TG) and identified in this study were related to inflammatory ADIPOCYTE 13 response, chemokine signalling pathway and ECM– [6] Ibrahim MM. Subcutaneous and visceral adipose tis- sue: structural and functional differences. Obes Rev. receptor interaction. Overall, these results suggest that 2010;11(1):11–18. inflammation and fibrosis could be important patholo- [7] Jeffery E, Wing A, Holtrup B, Sebo Z, Kaplan JL, gical alterations associated with the development of Saavedra-Pena R, et al. The adipose tissue microenvir- obese eWAT. onment regulates depot-specific adipogenesis in obesity. Cell Metab. 2016;24(1):142–150. [8] Vague J. The degree of masculine differentiation of Acknowledgments obesities: a factor determining predisposition to dia- betes, atherosclerosis, gout, and uric calculous disease. We would like to thank MogoEdit (https://www.mogoedit. Am J Clin Nutr. 1956;4(1):20–34. com) for its English editing during the preparation of this [9] Unamuno X, Gomez-Ambrosi J, Rodriguez A, et al. manuscript. Adipokine dysregulation and adipose tissue inflamma- tion in human obesity. Eur J Clin Invest. 2018;48(9): e12997. Disclosure statement [10] Sarafidis M, Lambrou GI, Zoumpourlis V, et al. An integrated bioinformatics analysis towards the identifi- No potential conflict of interest was reported by the cation of diagnostic, prognostic, and predictive key author(s). biomarkers for urinary bladder cancer. Cancers (Basel). 2022;15(1):14. Funding [11] Sai Swaroop R, Akhil PS, Sai Sanwid P, et al. Integrated multi-omic data analysis and validation with yeast model This research was supported by the National Natural Science show oxidative phosphorylation modulates protein aggre- Foundation of China [grant number NSFC 81770452]. gation in amyotrophic lateral sclerosis. J Biomol Struct Dyn. 2022;1–20. doi: 10.1080/07391102.2022.2090441. [12] Haim Y, Bluher M, Slutsky N, et al. Elevated autophagy Institutional review board statement gene expression in adipose tissue of obese humans: a potential non-cell-cycle-dependent function of The animal procedures were conducted following the China E2F1. Autophagy. 2015;11(11):2074–2088. National Standards and Guidelines for Laboratory Animal [13] Aflaki E, Balenga NA, Luschnig-Schratl P, et al. Management and were approved by the Animal Care and Impaired Rho GTPase activation abrogates cell polar- Use Committee of Southeast University (No.: 20200326003). ization and migration in macrophages with defective lipolysis. Cell Mol Life Sci. 2011;68(23):3933–3947. [14] Ko SH, Lee JK, Lee HJ, et al. 8-Oxo-2’-deoxyguanosine Data availability statement ameliorates features of metabolic syndrome in obese mice. All of the data are presented in the manuscript. Biochem Biophys Res Commun. 2014;443(2):610–616. [15] Sun H, Hu L, Fan Z. β2 integrin activation and signal transduction in leukocyte recruitment. Am J Physiol ORCID Cell Physiol. 2021;321(2):C308–C16. [16] Nair S, Lee YH, Rousseau E, et al. Increased expression Yu-Yu Yao http://orcid.org/0000-0001-7841-288X of inflammation-related genes in cultured preadipo- cytes/stromal vascular cells from obese compared with non-obese Pima Indians. Diabetologia. 2005;48 References (9):1784–1788. [17] Chen Q, Ruedl C. Obesity retunes turnover kinetics of [1] Collaboration NCDRF. Trends in adult body-mass tissue-resident macrophages in fat. J Leukoc Biol. index in 200 countries from 1975 to 2014: a pooled 2020;107(5):773–782. analysis of 1698 population-based measurement studies [18] Haure-Mirande JV, Audrain M, Ehrlich ME, et al. with 19.2 million participants. Lancet. 2016;387 TYROBP/DAP12 in Alzheimer’s disease: transduction (10026):1377–1396. of physiological and pathological signals across [2] Ahmed B, Sultana R, Greene MW. Adipose tissue and TREM2. Mol Neurodegener. 2022;17(1):55. insulin resistance in obese. Biomed Pharmacother. [19] Yan K, Zhang P, Jin J, et al. Integrative analyses of hub 2021;137:111315. genes and their association with immune infiltration in [3] Jayedi A, Khan TA, Aune D, et al. Body fat and risk of adipose tissue, liver tissue and skeletal muscle of obese all-cause mortality: a systematic review and patients after bariatric surgery. Adipocyte. 2022;11 dose-response meta-analysis of prospective cohort (1):190–201. studies. Int J Obes (Lond). 2022;46(9):1573–1581. [20] Sletta KY, Castells O, Gjertsen BT. Colony stimulating [4] Gugliucci A. Biomarkers of dysfunctional visceral fat. factor 1 receptor in acute myeloid leukemia. Front Adv Clin Chem. 2022;109:1–30. Oncol. 2021;11:654817. [5] Sun Y, Chen S, Zhang X, et al. Significance of cellular [21] Cox N, Crozet L, Holtman IR, et al. Diet-regulated cross-talk in stromal vascular fraction of adipose tissue production of PDGFcc by macrophages controls in neovascularization. Arterioscler Thromb Vasc Biol. energy storage. Science. 2021;373(6553):373. 2019;39(6):1034–1044. 14 Y. JIANG ET AL. [22] Lecoutre S, Lambert M, Drygalski K, et al. Importance adipose tissue of obese mice and humans. J Lipid Res. of the Microenvironment and Mechanosensing in 2005;46(11):2347–2355. Adipose Tissue Biology. Cells. 2022;12(1):11. [28] Datta R, Podolsky MJ, Atabai K. Fat fibrosis: friend or [23] Hachiya R, Tanaka M, Itoh M, et al. Molecular foe?. JCI Insight. 2018;3(19). doi:10.1172/jci.insight. mechanism of crosstalk between immune and meta- 122289. bolic systems in metabolic syndrome. Inflamm Regen. [29] Marcelin G, Gautier EL, Clement K. Adipose tissue 2022;42(1):13. fibrosis in obesity: etiology and challenges. Annu Rev [24] Kanda H, Tateya S, Tamori Y, et al. MCP-1 contributes Physiol. 2022;84(1):135–155. to macrophage infiltration into adipose tissue, insulin [30] Marcelin G, Clement K. The multifaceted progenitor resistance, and hepatic steatosis in obesity. J Clin fates in healthy or unhealthy adipose tissue during Invest. 2006;116(6):1494–1505. obesity. Rev Endocr Metab Disord. 2021;22 [25] Weisberg SP, Hunter D, Huber R, et al. CCR2 mod- (4):1111–1119. ulates inflammatory and metabolic effects of high-fat [31] Ji JJ, Qian LL, Zhu Y, et al. Kallistatin/Serpina3c feeding. J Clin Invest. 2006;116(1):115–124. inhibits cardiac fibrosis after myocardial infarction [26] Lindhorst A, Raulien N, Wieghofer P, et al. Adipocyte by regulating glycolysis via Nr4a1 activation. death triggers a pro-inflammatory response and Biochim Biophys Acta Mol Basis Dis. 2022;1868 induces metabolic activation of resident macrophages. (9):166441. Cell Death Dis. 2021;12(6):579. [32] Guo J, Qian L, Ji J, et al. Serpina3c regulates adipose [27] Cinti S, Mitchell G, Barbatelli G, et al. Adipocyte death differentiation via the Wnt/beta-catenin-PPARgamma defines macrophage localization and function in pathway. Cell Signal. 2022;93:110299.

Journal

AdipocyteTaylor & Francis

Published: Dec 31, 2023

Keywords: Obesity; high-fat diet; visceral adipose tissue; bioinformatics analysis; inflammation; fibrosis; biomarker

There are no references for this article.