Integrated bioinformatics and validation reveal TMEM45A in systemic lupus erythematosus regulating atrial fibrosis in atrial fibrillation

Abstract

Background: Accumulative evidence has shown that systemic lupus erythematosus (SLE) increases the risk of various cardiovascular diseases including atrial fibrillation (AF). The study aimed to screen potential key genes underlying co-pathogenesis between SLE and AF, and to discover therapeutic targets for AF.

Methods: Differentially expressed genes (DEGs) were identified, and co-expressed gene modules were obtained through weighted gene co-expression network analysis (WGCNA) based on the AF and SLE expression profiles from the GEO database. Subsequently, machine learning algorithms including LASSO regression and support vector machine (SVM) method were employed to identify the candidate therapeutic target for SLE-related AF. Furthermore, the therapeutic role of TMEM45A was validated both in vivo and vitro.

Results: Totally, 26 DEGs were identified in SLE and AF. The PPI network combined with WGCNA identified 51 key genes in SLE and AF. Ultimately, Machine learning-based methods screened three hub genes in SLE combined with AF, including TMEM45A, ITGB2 and NFKBIA. The cMAP analysis exposed KI-8751 and YM-155 as potential drugs for AF treatment. Regarding TMEM45A, the aberrant expression was validated in blood of SLE patients. Additionally, TMEM45A expression was up-regulated in the atrial tissue of patients with AF. Furthermore, TMEM45A knockdown alleviated AF occurrence and atrial fibrosis in vivo and Ang II-induced NRCFs fibrosis in vitro.

Conclusion: The crosstalk genes underlying co-pathogenesis between SLE and AF were unraveled. Furthermore, the pro-fibrotic role of TMEM45A was validated in vivo and vitro, highlighting its potential as a therapeutic target for AF.

Keywords: Atrial fibrillation; Bioinformatics; Fibrosis; Systemic lupus erythematosus; TMEM45A.

Fig. 1

DEGs identification from AF and SLE. A The heatmap of AF DEGs analysis results based on merged datasets including GSE14975, GSE31821, GSE41177 and GSE79768. B The volcano plot of AF DEGs analysis results based on merged datasets including GSE14975, GSE31821, GSE41177 and GSE79768. C The heatmap of SLE DEGs analysis results based on GSE50772. D The volcano plot of SLE DEGs analysis results based on merged datasets including GSE50772. E Identification of 26 overlapping genes between DEGs of AF and SLE

Fig. 2

Weighted co-expressed network analysis and related key gene modules construction. A, B Analysis of scale-free topology model to identify the appropriate soft threshold based on the average connectivity and scale independence in AF and SLE. C, D The network heatmap of gene dendrogram and module eigengenes in AF and SLE. E, F The heatmap of correlation between module eigengenes and status of AF and SLE respectively. The correlation and p value were presented. G Number of intersecting genes of each key module of AF and SLE

Fig. 3

Key module genes identification based on the PPI network analysis in AF and SLE. A The PPI network composed of genes in SLE-MEblue module and AF-MEturquoise module. B The PPI network was composed of genes in SLE-MEturquoise module and AF-MEturquoise module. C Screening 31 key module genes in the PPI network of SLE-MEblue module and AF-MEturquoise module based on Stress, Betweenness, Closeness, and Degree. D Screening 29 key module genes in the PPI network of SLE-MEturqoise module and AF-MEturquoise module based on Stress, Betweenness, Closeness, and Degree

Fig. 4

GO and KEGG pathway analysis for core genes merged from 26 DEGs and 51 key modules genes of PPI network. A The first 20 significantly enriched GO annotation of BP. B The 20 first significantly enriched GO annotation of MF. C The first 20 significantly enriched GO annotation of CC. D The first 20 significantly enriched KEGG pathways

Fig. 5

Screening potential small-molecular therapeutic compounds for AF with SLE through cMAP analysis. A The heatmap of the top 10 small-molecular compounds with the most significant negative enrichment scores in 10 cell lines. B The description of the targeted pathways of the top 10 small-molecular compounds. C The chemical structures of the top 10 small-molecular compounds

Fig. 6

Screening hub genes based on the machine learning methods in AF and SLE. A Coefficient profiles of variables in LASSO regression model in AF. B Ten fold cross-validation for turning parameter. C The optimum root mean squared error (RMSE) of SVM-based method based on 13 characteristic genes in AF. D Coefficient profiles of variables in LASSO regression model in SLE. E Ten fold cross-validation for turning parameter. F The optimum root mean squared error (RMSE) of SVM-based method based on 65 characteristic genes in SLE. G Venn diagram showed that 5 common genes were selected by LASSO regression model and 12 common genes were screened by SYM-based method. And 3 hub genes in AF with SLE were finally identified after intersection

Fig. 7

The GSEA analysis of TMEM45A in AF. A The GSEA analysis showed the KEGG pathways enrichment of TMEM45A in AF. B–F The specific enrichment pathway including “KEGG N GLYCAN BIOSYNTHESIS”, “KEGG RIBOSOME”, “KEGG SPHINGOLIPID METABOLISM”, “KEGG FC GAMMA R MEDIATED PHAGOCYTOSIS” and “KEGG TGF BETA SIGNALING PATHWAY”

Fig. 8

The validation of dynamic TMEM45A expression in AF and SLE. A The HE staining of the left atrium from normal and AF patients. B Representative Masson images and relative densitometric analysis of left atrium from normal (n = 10) and AF (n = 10) patients. C Representative immunohistochemistry images and relative densitometric analysis of TMEM45A in left atrium from normal (n = 10) and AF patients (n = 10). D The HE staining of the left atrium from normal and AF model constructed in rats. E Representative Masson images and relative densitometric analysis of left atrium from normal and AF model constructed in rats. F Representative immunohistochemistry images and relative densitometric analysis of TMEM45A in left atrium from normal and AF model constructed in rats. G Representative electrocardiogram from normal and AF model constructed in rats. H The aberrant expression of TMEM45A in SLE patients’ blood. **p < 0.01

Fig. 9

Knockdown of TMEM45A reduces AF inducibility. The (A) Representative immunofluorescence images and relative densitometric analysis of TMEM45A protein in rat atrial tissues. B Representative electrocardiogram from rats in different groups. C, D AF inducibility and AF durations in AF rats. *p < 0.05 and **p < 0.01

Fig. 10

Knockdown of TMEM45A alleviated atrial fibrosis and collagen deposition in AF. A The HE staining of atrial tissues in rats. B Representative Masson images and relative densitometric analysis of atrial tissues in rats. C Representative immunofluorescence images and relative densitometric analysis of α-SMA protein in rat atrial tissues. D Representative immunofluorescence images and relative densitometric analysis of collagenase I protein in rat atrial tissues. E Representative immunofluorescence images and relative densitometric analysis of collagenase III protein in rat atrial tissues. F Representative western blot images and relative densitometric analysis of TMEM45A, collagen I, collagen III and α-SMA protein of atrial tissues in rats (n = 3). *p < 0.05 and **p < 0.01

Fig. 11

Knockdown of TMEM45A inhibited NRCFs activation through TGF-β/smad2/3 pathway. A Representative western blot images and relative densitometric analysis of TMEM45A protein in NRCFs after exposure to Ang II for different durations. B Representative western blot images and relative densitometric analysis of TMEM45A, collagen I, collagen III, α-SMA, TGF-β and p-smad2/3 protein in NRCFs transfected with si-TMEM45A under Ang II stimulation. β-Actin was used as a loading control. C Cell proliferation ability was measured by CCK8 assay. D EdU incorporation in VSMCs, the EdU-positive signal (purple) was merged with nuclei (blue) stained with Hoechst 33342. EdU-positive cells were quantified by Image-Pro Plus. E VSMC migration ability was measured by Transwell assays (n = 3), and migrated cells were quantified by Image-Pro Plus. *p < 0.05 and **p < 0.01