Comprehensive bulk and single-cell transcriptome profiling give useful insights into the characteristics of osteoarthritis associated synovial macrophages

Abstract

Background: Osteoarthritis (OA) is a common chronic joint disease, but the association between molecular and cellular events and the pathogenic process of OA remains unclear.

Objective: The study aimed to identify key molecular and cellular events in the processes of immune infiltration of the synovium in OA and to provide potential diagnostic and therapeutic targets.

Methods: To identify the common differential expression genes and function analysis in OA, we compared the expression between normal and OA samples and analyzed the protein-protein interaction (PPI). Additionally, immune infiltration analysis was used to explore the differences in common immune cell types, and Gene Set Variation Analysis (GSVA) analysis was applied to analyze the status of pathways between OA and normal groups. Furthermore, the optimal diagnostic biomarkers for OA were identified by least absolute shrinkage and selection operator (LASSO) models. Finally, the key role of biomarkers in OA synovitis microenvironment was discussed through single cell and Scissor analysis.

Results: A total of 172 DEGs (differentially expressed genes) associated with osteoarticular synovitis were identified, and these genes mainly enriched eight functional categories. In addition, immune infiltration analysis found that four immune cell types, including Macrophage, B cell memory, B cell, and Mast cell were significantly correlated with OA, and LASSO analysis showed that Macrophage were the best diagnostic biomarkers of immune infiltration in OA. Furthermore, using scRNA-seq dataset, we also analyzed the cell communication patterns of Macrophage in the OA synovial inflammatory microenvironment and found that CCL, MIF, and TNF signaling pathways were the mainly cellular communication pathways. Finally, Scissor analysis identified a population of M2-like Macrophages with high expression of CD163 and LYVE1, which has strong anti-inflammatory ability and showed that the TNF gene may play an important role in the synovial microenvironment of OA.

Conclusion: Overall, Macrophage is the best diagnostic marker of immune infiltration in osteoarticular synovitis, and it can communicate with other cells mainly through CCL, TNF, and MIF signaling pathways in microenvironment. In addition, TNF gene may play an important role in the development of synovitis.

Keywords: Scissor analysis; immune infiltration; macrophage; osteoarthritis; synovium.

Figure 1

Differential expression genes and function analysis. (A, B) Volcano plot showed DEGs in GSE1919 and GSE55235, respectively. Blue dots represented downregulated DEGs, red dots represented upregulated DEGs, and gray dots represented the rest of the no significant differential expressed. (C, D) Heatmap showed the potential top 15 DEGs in GSE1919 and GSE55235, respectively. (E) Venn plot of the common DEGs of GSE1919 and GSE55235. (F) KEGG function analysis of the 172 DEGs. (G) Metascape function clustering analysis of the 172 DEGs.

Figure 2

Protein–protein interactions analysis with 172 DEGs. (A) PPI network of DEGs. Red nodes labeled subNet1 indicate the cellular response of unfolded proteins; blue nodes labeled subNet2 indicate the regulation of cell cycle and kinase activity; green nodes labeled subnet3 indicate chemokine signaling pathway; purple nodes labeled subnet4 indicate the metabolic processes; orange nodes labeled subnet5 indicate the immune system signaling; yellow nodes and brown nodes labeled subnet6 and subnet7 indicate the signal translocation and inflammation during biofilm processes; pink nodes labeled subnet8 indicate antigen the presentation and T-cell differentiation. (B) The key subnetworks screened after using MCODE tool.

Figure 3

Immune infiltration analysis and GSVA analysis of OA samples. (A) The immune score of GSE55235 and GSE1919 samples. (B) The immune infiltration of GSE55235 samples. (C) The immune infiltration of GSE1919 samples. (D) The immune score of GSE32317 and GSE46750 samples. (E) The immune infiltration of GSE32317 samples. (F) The immune infiltration of GSE46750 samples. (G) Heatmap of Hallmark gsva scores in GSE55235 dataset. (H) Heatmap of Hallmark gsva scores in GSE1919 dataset. (I) Heatmap showed that the spearman correlation between gsva score and immune infiltration cell types. (*p < 0.05; **p < 0.01).

Figure 4

LASSO analysis for immune infiltration cell types. (A) CV statistical graph during the construction of the LASSO regression model, which shows that the minimum lambda at model construction is 0.1614863. (B) The model regression coefficient diagram shows the change trend of the coefficient corresponding to each immune infiltrating cells variable with the change of lambda value. (C) ROC curve predicts the identification effect of the above models in different datasets. The closer AUC value is to 1, the better of prediction effect on the model. The figure is shown that the AUC in the training set is 0.8995 and that, in the test, set is 0.8262, indicating that the model has a robust prediction accuracy. In the validation datasets, the AUC is 0.7575, which shows that the models constructed by the Macrophage and B cell can also have good accuracy in different types of datasets. (D, E) The immune score of GSE89408 and GSE143514 datasets. The figure shows that immune sore is significantly increased in the two datasets. (F, G) The immune infiltration analysis of six cell types in GSE89408 and GSE143514 datasets.

Figure 5

Cellular communication of Macrophage in OA. (A) Circle plot of the significant connections among ten interacting immune cell types (Macrophage, DC, EC, SMC, Mast cell, SIF, SSF, ProIC, T cell, and B cell). Different colors represent different cell groups. (B) Circle plot of the significant connections of Macrophage with other cells. (C) Pattern recognition of the immune cell types; the graph shows the interpretation of intercellular communication networks by incoming and outcoming communication patterns. (D) Bubble plot of the ligand-receptor pairs in Macrophage cell types. Colors in the bubble plot are the proportional of the communication probability, where the blue and red colors correspond to the smallest and largest values. (E) Circular plots of the communication among the 10 immune cell subtypes according to the three major signaling pathways (CCL, TNF, and MIF). Different colors represent different immune cell types. (F) Cellular communication strength on a two-dimensional manifold according to incoming and outcoming communication patterns; each dot represents one cell type. (G, H) Communication probability by ligand-receptor pairs according to CCL, TNF, and MIF signaling pathway, each dot represents the communication network of one immune cell types. Line size is proportional to the overall communication probability.

Figure 6

Scissor identification results on Macrophage. (A) The UMAP visualization of scRNA-seq datasets, scisoor1 represents that cell identified to be positively associated with transcriptome synovial inflammation, scissor0 were the background cells. (B) The bar plot shows the proportion of scissor cells in the three scRNA-seq samples. (C) The volcano plot of differential gene expression in scissor1 Macrophages (M_Scissor1) versus scissor0 Macrophages (M_Scissor0). (D) The violin plots show the several significant upregulated genes (RNASE1, CQB, S100A9, LYZ, and CD163) in M_Scissor1 group. (E) GSVA enrichment analysis of the hallmarker signaling pathways in Macrophage scissor group. (F) The Box plot shows the significant signaling pathways in gsva analysis. (G) The TNF gene expression in Macrophage scissor group and specific express in M_Scissor1 group. (H) The violin plot showing TNF gene expression in each cell type. (I) The cell communication number in Macrophage scissor group cells. (J) TNF signaling pathway usage in each cell clusters. (K) The ligand-receptor pairs usage of TNF signaling pathway in each cell clusters.