Involvement of plasminogen activator inhibitor-1 and its related molecules in atrial fibrosis in patients with atrial fibrillation

Microarray data

The Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/geo, RRID:CVCL_VR24) is a public repository that provides access to high-throughput genomic datasets (Wang et al., 2019). We downloaded the GSE2240 datasets (Affymetrix GPL 96 platform. Affymetrix Human Genome U133A Array) from the GEO database to use in our analysis. Furthermore, the probes were converted into corresponding gene symbols based on the platform annotation information from the published data. The GSE2240 datasets contained 10 AF tissue samples and 20 SR tissue samples.

Principal component analyses

All related genes were normalized to their median, and the data quality was assessed by performing a principal component analysis using NetworkAnalyst (https://www.networkanalyst.ca, RRID:SCR_016909) (Tong et al., 2011; Ye et al., 2019).

Identification of DEGs

DEGs common to the AF and SR samples were identified using NetworkAnalyst. Probe sets without corresponding gene symbols or genes that matched more than one probe set were discarded or averaged, respectively. The absolute value of logFC > 0.5 (FC, fold change) and P-value (adjusted) < 0.05 were regarded as statistically significant.

Gene Ontology and Kyoto Encyclopedia of Genes and Genomes Analysis of DEGs

We performed a Gene Ontology (GO) analysis and a Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis of the DEGs using clusterProfiler package in R (v3.6.0) (Yu et al., 2012). The GO functional categories contained the terms biological process (BP), cellular component (CC), and molecular function (MF). Enrichment significance thresholds were set at P-value (adjusted) < 0.05.

Gene set enrichment analysis of DEGs

We performed a gene set enrichment analysis (GSEA) of the DEGs utilizing the Reactome Pathway Database and the web-based Gene Set Analysis Toolkit version 2.0 (Liao et al., 2019; Wang et al., 2017) (http://bioinfo.vanderbilt.edu/webgestalt/, RRID:SCR_006786). —NES— > 1 and FDR < 0.25 were considered significant (NES, normalized enrichment score; FDR, false discovery rate).

Gene cluster identification

Next, the candidate DEGs were analyzed using the ClueGO app (Cytoscape; http://apps.cytoscape.org/apps/cluego, RRID:SCR_005748), a functional classification tool that uses the latest precompiled annotation files, including GO and functional pathways enrichment analysis (Bindea et al., 2009; Cao et al., 2018). Significance was defined as a P value < 0.05 (Shannon et al., 2003).

Protein-protein interaction network analysis

Protein-protein interaction (PPI) networks were established according to the STRING analysis (https://string-db.org, RRID:Addgene_36407), and an interacting activity with a combined score >0.4 was deemed statistically significant. Hub modules in the PPI network were then selected using the Molecular Complex Detection (MCODE) plug-in for Cytoscape (Yuan et al., 2018). To screen highly connected hub genes in the PPI network, the Cytoscape cytoHubba plugin was also used (Niemira et al., 2020).

Identifying differentially expressed mRNAs in the public database

The transcriptional profile data from the highly connected hub genes from the AF and SR samples were downloaded from the GSE2240 datasets. Then, the “ggpubr” R packages was used to measure the mRNA expression level differences between the two groups (AF vs. SR).

Gene Set Enrichment Analysis

Another GSEA was performed using the GSEA v4.0.3 software (http://software.broadinstitute.org/gsea/index.jsp, RRID:SCR_003199) (Subramanian et al., 2005). We analyzed the SERPINE1/PAI-1 and DCN mRNA level changes in genomic information in GO, BIOCARTA, and KEGG pathways using the GSEA tool. The 30 samples in the GSE2240 datasets (20 SR samples and 10 AF samples) were classified into the low and high expression groups using the SERPINE1 (PAI-1) expression median level as a cut-off point. The potential function of SERPINE1 (PAI-1) (high vs. low) was screened using a GSEA to determine whether defined lists (or sets) of genes exhibited a statistically significant bias in their distribution within a ranked gene list (Tebbji et al., 2014). The expression value of SERPINE1 (high vs. low) was used as the phenotypical label, the number of permutations was set to 1000, and c2.all.v7.1.symbols.gmt from the Molecular Signatures Database (MSigDB) (http://software.broadinstitute.org/gsea/msigdb) was defined as the functional gene set (Liberzon et al., 2015). All other parameters were set to default values. the absolute value of NES > 1 and FDR < 0.25 were chosen as the significance cut-off criteria (D’Haene et al., 2016). The GSEA of DCN was performed using the same method.

Patients

This investigation was reviewed and approved by the research ethics committee at Guangdong General Hospital, Guangdong Academy of Medical Sciences (Approval No. GDREC20160128H). Left atrial appendages (LAAs) were obtained from SR or AF patients undergoing cardiac thoracotomy. Patients with diabetes, hypertension, pneumonia, or other infectious diseases were excluded from the study. All patients or their legally authorized representative signed the necessary informed consent forms. After surgical excisions were performed, the specimens were immediately snap-frozen in liquid nitrogen and stored at −80 °C before the experiments were performed. Patient data, including age, sex distribution, type of valve disease, and left ventricular function are shown in Table S1.

Isolation and culture of human atrial fibroblasts

Human atrial fibroblasts were isolated from the LAAs using direct adherent culture methods. In brief, the tissue specimens (approximately 200 mg) were cut into approximately 1 mm³ pieces and suspended in 500 µL of fetal bovine serum and then placed in T-25 flasks. Two hours later, the flasks were flooded with 5 mL of complete media (Fibroblast Basal Medium [Lonza, CC-3131], 10% fetal bovine serum [Gibco], Insulin [Lonza, CC-4021WW], rhFGF [Lonza, CC-4065WW], GA-1000 [Lonza, CC-4081WW]). The tissues were cultured in a humidified incubator at 37 °C and 5% CO₂. After approximately 5 days, new cells migrated from the edge of the tissues and when they reached 80% confluence, cells were passaged. Cells were plated at densities of 80,000 cells/well in 6-well plates for the p53 knockdown experiment.

A mixture of three p53 siRNA chain (hTP53 si-1 sense GCG CAC AGA GGA AGA GAA UTT; hTP53 si-2 sense CCA CUG GAU GGA GAA UAU UTT; hTP53 si-3 sense CCA UCC ACU ACA ACU ACA UTT; Jikai Biotechnology Co. Shanghai, China) was used to knock down protein expression of p53. When the cells reached 70–80% confluence, transfection was performed according to the following method. First, 6 µL of Lipofectamine 3,000 Reagent (Thermo Fisher) and 5 µL p53 siRNA/well (final concentration: 50 nM) were diluted in 250 µL of Opti-MEM reduced serum medium (Gibco) respectively, and then mixed them well. After incubation for 10–15 min at room temperature, the complexes were added directly to cells in 1.5 mL fresh complete culture medium. The cells were then continuously cultured for 48 h. Finally, they were harvested, and protein expression was detected by Western blot.

Western blot analysis

Detailed experiment procedure was as previously reported method (Li et al., 2020). Antibodies used for these experiments are listed as follows: (anti-Collagen I antibody [1:2000, Abcam Cat^# ab34710, RRID:AB_731684]; anti-Collagen III antibody [1:2000, Abcam Cat^# ab7778, RRID:AB_306066]; anti-p53 antibody [1:1000, Cell Signaling Technology Cat^# 2524, RRID:AB_331743]; anti-PAI-1 antibody [1:1000, Cell Signaling Technology Cat^# 11907, RRID:AB_2797763]; anti-p21 antibody [1:1000, Proteintech 27296-1-AP, RRID:AB_2880834]; anti-TGF-β antibody [1:1000, Cell Signaling Technology Cat^# 3711, RRID:AB_2063354]. Anti-GAPDH antibody [1:5000, Proteintech Cat^# 60004-1-lg, RRID:AB_2107436].

Statistical analysis

All values are expressed as means ± standard error of mean (SEM), and Student’s t-test was used to determine pairwise statistical significance of the differences between two group means. One-way analysis of variance was used for comparisons of multiple groups. A P-value < 0.05 was considered statistically significant.

Identification of DE-mRNAs

We performed a principal component analysis (PCA) of the samples (Fig. 1A), and the results showed good separation of AF from the normal samples. Pre- and post-normalization of data was described in Fig. S1A. Furthermore, —log₂ FC— > 0.5 and p-value (adjusted) < 0.05 were considered as the criteria to identify the DEGs. A total of 148 differentially expressed mRNAs were selected from the 30 samples, among which 105 genes were downregulated and 43 were upregulated in the AF group. The distribution of the DEGs is represented by a volcano map (Fig. 1B). DEGs were ordered according to the p-value (adjusted) and the top 50 genes were selected for plotting in heatmap (Fig. S1B). The 148 genes showing differential expression are listed in Table 2.

GSEA analysis of the DEGs

To gain further insight into the biological functions of the DEGs, we performed GSEA using the WebGestalt Gene Set Analysis Toolkit, version 2.0. The four most significantly enriched Reactome terms are illustrated in Figs. 1C–1F, including positive enrichment in the integrin cell surface interaction, degradation of the extracellular matrix, collagen formation, and extracellular matrix (ECM) proteoglycans. The top 10 GSEA terms are displayed as a bar chart in Fig. 1G.

Enrichment analysis of the DEGs

GO and KEGG annotations were performed utilizing the clusterProfiler package for R (v3.6.0). The DEGs were divided into three main groups: CCs, MFs, and BPs. The top 10 GO items were significantly enriched (Fig. 2A). The GO term analysis showed that the ECM organizational process was the most significant enrichment in BPs, followed by the extracellular structure organizational process. Under the CC, the DEGs were significantly enriched for the collagen-containing ECM, endoplasmic reticulum lumen, and collagen trimer. In the MF category, the most abundant term was the ECM structural constituent. To further visualize these results, we generated heat maps to depict fold changes in the expression levels of genes associated with several GO terms, including ECM organization (Fig. 2C). Dot plots of the enriched KEGG pathways are shown in Fig. 2B. The PI3K-Akt signaling pathway, human papillomavirus infection, and focal adhesion were significantly enriched. Furthermore, the heatmap showed DEGs related to the enriched signaling pathways (Fig. 2D).

ClueGO analysis

GO, KEGG, and Reactome analyses were performed using ClueGO and CluePedia in the Cytoscape software to explore the key functions of the target genes. The ClueGO analysis indicated that the collagen-activated tyrosine kinase receptor signaling pathway, regulation of endothelial cell migration, regulation of cardiac muscle tissue development, and organ growth were significantly enriched GO terms (Fig. 3A). Furthermore, the results of our pathway enrichment analysis showed significant associations between ECM organization, HIF-1 signaling pathway, Tie2 signaling pathway, platelet activation and signaling by PDGF (platelet derived growth factor) pathway, and the 148 DEGs (Fig. 3B).

PPI network analysis

We used Cytoscape (https://cytoscape.org/) (version 3.7.2) to perform a visual analysis of the protein-protein association networks for predicting the interactions of the proteins encoded by the DEGs (Yu et al., 2020). The PPI network was constructed as shown in Fig. 3C. The hub modules of the PPI network were selected using MCODE. The most significant module was screened out with a cutoff MCODE score of > 5 (Fig. 3D). Furthermore, the top 10 hub genes were identified using cytoHubba (Fig. 3E). SERPINE1, DCN, and TIMP3 were included in this analysis.

GO analysis of the hub genes and the relative mRNA expression of the three hub genes in the public database

To systematically assess the potential biological functions of the candidate genes, we performed GO enrichment analysis using clusterProfiler for R 3.6.0. The GO analysis indicated that these hub genes were significantly associated with the term ECM organizational process (Fig. 4A). Moreover, the three hub genes, SERPINE1 (PAI-1), DCN, and TIMP3, were differentially expressed. SERPINE1 was highly expressed in AF, and the two other genes exhibited opposite expression levels (Figs. 4B–4D).

GSEA associated with SERPINE1(PAI-1) and DCN expression

Figures 5A–5C show the significantly enriched KEGG pathways of SERPINE1 (PAI-1): p53 signaling pathway, G2 pathway, and ERK pathway. Figures 5D–5F show the main enriched GO of DCN: negative regulation of cellular response to growth factor stimulus, positive regulation of interleukin 10 production, and endocardial cushion formation.

PAI-1 was highly expressed in AF patients

To investigate whether p53/PAI-1 plays a vital role in the pathogenesis of AF, we addressed the PAI-1, p53, p21, Col1a1/3a1 and TGF-β protein expression levels in LAAs from patients with AF and the SR controls. We found that the protein levels of PAI-1, p53, p21, Col1a1/3a1 and TGF-β were higher in the AF samples than in the SR controls (Figs. 6A–6G). The PAI-1 serum levels were also significantly higher in patients with AF (Fig. 6H) (the patient characteristics are shown in Table S3). These results indicate that p53/PAI-1 might be involved in the pathophysiology of atrial fibrosis.

Knocking down p53 reduces the expression level of PAI-1 in human atrial fibroblasts

To determine whether p53 is involved in the regulation of PAI-1, we performed p53 knockdown using siRNA in human atrial fibroblasts. As shown in Figs. 7A–7F, knocking down p53 significantly inhibited PAI-1, p21 and Col 1a1/3a1 protein expression. As such, p53 might regulate the expression of PAI-1 and fibrosis in human atrial fibroblasts, thus leading to the structural remodeling of the atrium and, therefore, AF.