Integrated bioinformatics and network pharmacology to identify and validate macrophage polarization related hub genes in the treatment of osteoarthritis with Astragalus membranaceus

Abstract

Background: Macrophage polarization exacerbates the pathological processes of osteoarthritis (OA). Astragalus membranaceus (AM) can repair chondrocytes and serve as a protective agent for OA. Therefore, the study intended to identify macrophage polarization-related genes (MPRGs) in the treatment of OA with AM.

Methods: We utilized data from GSE57218 as training set, while GSE117999 serves as a validation set, all obtained from Gene Expression Omnibus(GEO). The MPRGs were exported from the Molecular Signatures Database. Target genes of AM were obtained by network pharmacology. Differentially expressed genes (DEGs) were identified in OA vs. control groups. Then key module genes were acquired through weighted gene co-expression network analysis (WGCNA) and intersected with DEGs and target genes of AM to obtain candidate genes. Subsequently, the candidate genes were further screened for hub genes by machine learning, receiver operating characteristic (ROC) curve analysis, and expression validation. Further, reverse transcription-quantitative real-time polymerase chain reaction (RT-qPCR) was applied to verify the mRNA expression levels of hub genes. In addition, the mechanism of these hub genes was investigated through enrichment analysis, immune microenvironment analysis, regulatory network construction, and molecular docking.

Results: Ultimately, 1,430 DEGs, 4,577 key module genes, and 486 target genes of AM were intersected to derive 28 candidate genes. After machine learning, ROC curve analysis and expression validation, CREBBP and PIM3 were identified. The mRNA expression of tissue CREBBP and PIM3 was significantly decreased in OA compared with the control group. Furthermore, the enrichment analysis indicated that eight pathways, including oxidative phosphorylation, were simultaneously enriched by two hub genes. Microenvironment analysis revealed negative correlations between both hub genes and 11 differential immune cells. We identified that CREBBP and PIM3 were regulated by 6 miRNAs (e.g., hsa-mir-942-5p) and 79 transcription factors (TFs) (e.g., IRF1). Molecular docking experiments indicated that isoflavone strongly bound to CREBBP, while (3R)-3-(2-hydroxy-3,4-dimethoxyphenyl) chroman-7-ol exhibited significant binding affinity for PIM3, suggesting that these two active ingredients were core components of AM in treating OA via hub genes.

Conclusion: This study identified CREBBP and PIM3 as potential focal points for the treatment of OA with AM, providing valuable clues to help treat and predict OA.

Clinical trial number: Not applicable.

Keywords: Astragalus membranaceus; Bioinformatics; Macrophage polarization; Network pharmacology; Osteoarthritis.

Fig. 1

Identification of DEGs by WGCNA and selecting key module genes. (A) The volcano diagram shows the differentially expressed genes with the threshold set at|log2FC|>0.5 and P.Val < 0.05. (B) Heatmap of differentially expressed genes. The previous section shows the expression density heatmap of differentially expressed genes in the samples, while the following section presents the expression heatmap of these genes. (C) Box plot of the difference of MPRGs between OA samples and control samples. (D) Box plot of ssGSEA scores between OA samples and control samples. (E) Sample-level clusterin g about ssGSEA scored. (F) Soft threshold filtering(left is scale free Topology Model Fit to determining the optimal soft threshold; right is Network connectivity under different soft thresholds. (G) Co-expression module identification. The upper section presents the hierarchical clustering dendrogram of genes, while the lower section displays the gene modules. (H) Correlation heatmap between modules and traits

Fig. 2

Functional enrichment analysis of candidate genes. (A) Venn diagram of candidate targets. (B) Protein interaction network diagram of intersection genes.(C) GO enrichment map of candidate targets. (D) KEGG enrichment map of candidate targets

Fig. 3

Verifying CREBBP and PIM3. (A) Support Vector Machine Model Accuracy. (B) Lasso model diagram. (C) Identification of candidate key target 1. (D) ROC curve of candidate key target 1 in the training and validation sets. (E) The difference in expression of candidate key target 2 between the disease and the control groups (Left: training set; Right: validation set). (F)-(G) RT-qPCR validation of CREBBP and PIM3

Fig. 4

Tests the potential of CREBBP and PIM3 in diagnosing OA. (A) Key gene nomogram construction in training GSE57218. (B)The bias-corrected of actual OA and nomogram-predicted probability of OA. (C)The ROC curve of false positive rate and true positive rate. (D) The decision Analysis Curve of model and hub genes

Fig. 5

GSEA analysis of CREBBP and PIM3. (A) Significantly enriched pathways for CREBBP. (B) Enriched considerably pathways for PIM3

Fig. 6

Immune microenvironment analysis of CREBBP and PIM3. (A) Heat map of 28 immune cells expression in disease group and control group. (B) Differences in expression of 11 immune cells between the disease group and the control group. (C) Heat map of correlation between key targets and differential immune cells

Fig. 7

Molecular docking. (A) miRNA-key target regulatory network. (B) TF-key target network diagram. (C) CREBBP-isoflavanone molecular docking