Identification of key genes in atrial fibrillation using bioinformatics analysis

The data of gene expression profiles were acquired by logging in the Gene Expression Omnibus (GEO) database affiliated to National Center for Biotechnology Information (NCBI) (https://​www.​ncbi.​nlm.​nih.​gov) and editing key word “atrial fibrillation”. Three hundred eight results were shown including 42 series, 3 datasets, 2 platforms and 261 samples with detailed information. Among the 42 series, 29 series belonged to Homo sapiens, 9 series belonged to Mus musculus, 2 series belonged to Rattus norvegicus, 1 series belonged to Canis lupus familiaris and 1 series belonged to Sus scrofa. We divided the 29 series which belonged to Homo sapiens into 2 groups: expression profiling by high throughput sequencing and expression profiling by array and chose the latter group for further research. Microarray data which included RNA content of left atrial appendage (LAA) were selected. According to the criteria mentioned above, five microarray data (GSE79768, GSE31821, GSE115574, GSE14975 and GSE41177) were selected and downloaded. In GSE79768, there were 7 cases of samples from patients with AF and 6 cases of samples from patients without AF. In GSE31821, the number of cases from AF patients and SR patients was 2 and 2, respectively. In GSE115574, the number of patients with AF was 14 and the number of patients without AF was 15. GSE14975 included 5 atrial tissue samples from patients with AF and 5 from patients without AF. Sixteen samples from AF patients and 3 samples from SR patients were included in GSE41177. More detailed information is shown in Table 1.

First, GEO Sample (GSM) which stands for LAA was selected. Then the downloaded platform files and series matrix files were transformed by Perl language software 5.28.1. The probe ID in the matrix files was converted into gene symbol by Perl language and saved as a TXT file. After that, five datasets were merged by Perl language and batch normalization was conducted using the “sva” package of the R software. We deleted duplicated genes and values lacking specific gene symbols. DEGs analysis was conducted using the “limma” package of R software. DEGs were selected by using cut-off values of P < 0.05 and |log₂FC| > 0.5, where FC = fold change. Finally, volcano plot was drawn according to the data above and the “pheatmap” package of R software was applied to construct heat map.

On the basis of the DEGs from the merged five datasets, gene ontology (GO) enrichment analysis and pathway enrichment analysis were performed using the “clusterProfiler” package of R software and reactome [10, 11] (https://​reactome.​org/​), respectively. GO analysis is a widely used bioinformatics tool to investigate the annotation of genes and proteins. It can be used to integrate annotation data and provide tools access to all the data provided by the project. Reactome is an open-source, open access, manually curated and peer-reviewed pathway database. The goal is to provide intuitive bioinformatics tools for the visualization, interpretation and analysis of pathway knowledge to support basic and clinical research, genome analysis, modeling, systems biology and education. A q value< 0.05 was identified as significant difference in GO analysis; false discovery rate (FDR) < 0.05 was identified as significant difference in pathway analysis.

We applied Search Tool for the Retrieval of Interacting Genes (STRING) [12] 11.0 (https://​string-db.​org/​) to construct a protein-protein interaction (PPI) network based on the DEGs detected above. Interactions of proteins in the STRING database were screened by using an interaction score criterion. PPI pairs with an interaction score > 0.4 were considered significant in this analysis.

From 2017.6 to 2018.2, LAA tissues were collected from 20 AF patients who underwent mitral valve repair and maze procedure and 20 SR patients with left atrial thrombus who underwent mitral valve repair, LAA resection and thrombectomy. The study was approved by the Ethics Committee of the First Affiliated Hospital of China Medical University (NO.2017-69-2) and was carried out in accordance with Declaration of Helsinki. Signed informed consents were obtained from all patients prior to tissue collection.

We selected three key genes for validation using reverse transcription-quantitative PCR (RT-qPCR). Total RNA was extracted with TRIzol Reagent (Invitrogen, USA) according to the manufacturer’s instructions. RNA was converted to cDNA using the PrimeScript™ RT reagent Kit (TaKaRa, Japan). RT-qPCR was performed using TB Green™ Premix Ex Taq™ II (TaKaRa, Japan) and the data were analyzed with LightCycler® 480 Instrument (Roche, Switzerland) and the relative changes were calculated using the 2^-△△Ct method. SPP1 primers were: forward 5′- CTGTGTTG GTGGAGGATGTCTGC − 3′, and reverse 5′- GTCGGCGTTTGGCTGAGAAGG -3′. COL5A1 primers were: forward 5′- CGTATGATGACCTCACCTATGG − 3′, and reverse 5′- CGTAGTAGTTCTCGTCAAGGTT-3′. VCAN primers were: forward 5′- ACTGAAACTTCCTACGTATGCA − 3′, and reverse 5′- CTCACAAAGTG CACCAACATAA − 3′. β-actin primers were: forward 5′-CCTGGCACCCAGCA CAAT-3′, and reverse 5′- GGGCCGGACTCGTCATAC-3′.

A total of 106 DEGs were obtained in the merged dataset. Among these DEGs, 74 genes were up-regulated and 32 genes down-regulated. The top 10 DEGs are shown in Table 2. The detailed information about the 106 DEGs is shown in Additional file 1. Volcano plot is shown in Fig. 1 and heat map shown in Fig. 2.

GO analysis of genes includes biological process, cell composition and molecular function. The results are shown in Fig. 3. GO biological process analysis found that DEGs were mainly enriched in extracellular matrix organization, protein activation cascade, synapse pruning, extracellular structure organization, complement activation and collagen fibril organization. In the cell composition part, the DEGs were involved in collagen-containing extracellular matrix, extracellular matrix, collagen trimer, endoplasmic reticulum lumen, basement membrane and extracellular matrix component. In the molecular function section, the genes participated mainly in extracellular matrix structure constituent, extracellular matrix structural constituent conferring tensile strength, extracellular matrix structural constituent conferring compression resistance and collagen binding. The results suggested that DEGs were mostly involved in extracellular matrix structure. We also performed our reactome pathway analysis, which showed DEGs were mainly enriched in Extracellular matrix organization and Collagen formation. All the pathways involved are shown in Table 3.

We used the STRING database to investigate PPI network, including 74 up-regulated genes and 32 down-regulated genes. Four major proteins were discovered in the PPI network analysis, with secreted phosphoprotein 1(SPP1) connecting 12 nodes, with collagen type V alpha 1 chain (COL5A1) connecting 11 nodes, with versican (VCAN) connecting 11 nodes and TYRO protein tyrosine kinase binding protein (TYROBP) connecting 11 nodes. A PPI network of DEGs and the key proteins were performed as shown in Figs. 4 and 5, respectively. Disconnected nodes in the network were hidden.

We performed RT-qPCR in LAA tissues to further verify the expression of three key genes. As illustrated in Fig. 6, compared with SR group, the transcription levels of SPP1, COL5A1 and VCAN in AF group were significantly increased (P < 0.05), which is consistent with our bioinformatics analysis results.