Pulmonary disease with imbalance of cell ratioSingle-cell transcriptome data of 86 samples were gained from three data sets including Normal, COPD, COVID-19 and IPF, and a total of 358,217 cell expression profiles were obtained by using the “Harmony” function of Seurat package to remove the batch effect (Fig. 1A–C). After comparing the Acquired single-cell atlas with the human cell atlas database, nine cell types were identified, including endothelial cells, epithelial cells, smooth muscle cells, B cells, T cells, macrophages and monocytes (Fig. 1D). However, the proportion of these cells is seriously out of balance in different disease types (Fig. 1E and F). For example, the proportion of lung macrophages in COVID-19 patients decreased and the proportion of smooth muscle cells increased, while macrophages in IPF patients, T cells in COPD patients and epithelial cells in ILD patients proliferated. These results seem to indicate that four different diseases have different pathological mechanisms.Fig. 1Integrated single cell atlas, cell type and cell ratio of four diseases. UMAP diagram before (A) and after (B) batch removal by integrating the single cell atlas; UMAP diagram of cell grouping (C) and annotation (D); The proportion of all cell types (E) and the proportion of immune cells (F).Abnormal cell matrix adhesion and ubiquitination of endothelial cellsVisualization and differential gene analysis of 42,821 endothelial cells from four kinds of patients showed that the endothelial cells showed abnormal gene expression (Fig. 2A and B), but this abnormality did not show the same functional abnormality, such as oxidative phosphorylation activation of endothelial cells in COPD, ILD and IPF, while COVID-19 showed inhibition (Fig. 2C–F). Notably, endothelial cells in COPD and ILD showed similar inflammatory signal activation and increased collagen formation. GO enrichment analysis showed that protein ubiquitination, oxidative stress, integrin and abnormal cell matrix adhesion (Supplementary Fig. S1A) occurred in lung disease, but ubiquitination hydrolysis was up-regulated in COVID-19 and down-regulated in others (Supplementary Fig. S1B). For explore the key genes that drive these changes, hub gene were screened for DEGs (Supplementary Fig. S1C–F), and found that Ubiquitin-ribosomal protein eL40 fusion protein (UBA52), Polyubiquitin-C (UBC), Polyubiquitin-B (UBB) and other genes participated in the ubiquitination protein ligand binding (Fig. 2G,H and Supplementary Fig. S1G and H) by constructing a gene co-expression network. In terms of oxidative stress of endothelial cells, IPF and ILD showed abnormal expression of Epidermal growth factor receptor (EGFR), Hypoxia-inducible factor 1-alpha (HIF1A) and RACK1, while COVID-19 showed abnormal expression of Cytochrome c (CYCS), JUN and IL-6. Briefly, the endothelial cells in four diseases have abnormal processes of ubiquitination, cell matrix formation and oxidative stress.Fig. 2DEGs in endothelial cells and functional analysis. UMAP diagram of endothelial cell grouping (A). Compared with the normal group, the volcano map of DEGs in different disease groups (B). GSEA analysis (C–F) of DEGs in COPD, COVID-19, ILD and IPF, co-expression network and enrichment analysis of hub genes of COPD (G) and IPF (H).Epithelial oxidative stress and abnormal protein ubiquitinationDEGs analysis of 52,572 epithelial cells showed that chemokine family and ribosome family genes were down-regulated in four diseases (Fig. 3A and B). Interestingly, epithelial cells in COPD, COVID-19 and IPF showed multi-signal down-regulation, while ILD showed interferon activation and increased collagen synthesis (Fig. 3C–F). In addition, epithelial cells also show increased apoptosis and oxidative stress, and abnormal cell matrix adhesion and ubiquitination (Supplementary Fig. S2A and B). Through the screening of hub genes (Supplementary Fig. S2C–F), we found that epithelial cells showed obvious ubiquitination abnormalities (UBA52, UBC, UBB). In addition, EGFR, JUN, RACK1 participated in the oxidative stress process (Fig. 3G,H and Supplementary Fig. S2G and H) of COVID-19 epithelial cells.Fig. 3DEGs in epithelial cells and functional analysis. UMAP diagram of epithelial cell grouping (A). Compared with the normal group, the volcano map of DEGs in different disease groups (B). GSEA analysis (C–F) of DEGs in COPD, COVID-19, ILD and IPF, co-expression network and enrichment analysis of hub genes of COPD (G) and IPF (H).Protein ubiquitination and oxidative stress disorder of smooth muscle cellsBy analyzing the gene expression of 55,783 smooth muscle cells from four diseases, it was found that the expression of various chemokines was down-regulated and collagen-related molecules were up-regulated (Fig. 4A and B). GSEA showed that only ILD showed extensive inflammatory signal activation (Fig. 4C–F), which was consistent with KEGG results (Supplementary Fig. S3B). At the same time, smooth muscle cells also showed the disorder of protein ubiquitination, oxidative stress, cell matrix adhesion and collagen synthesis (Supplementary Fig. S3A) in different disease types. Through the screening and enrichment analysis of hub genes, it was found that UBA52, heat shock protein family A member 8 (HSPA8), UBC, UBB, heat shock protein 90 alpha family class A member 1 (HSP90AA1) and JUN participated in the protein ubiquitination process of different diseases (Fig. 4G,H and Supplementary Fig. S3G and H).Fig. 4DEGs and functional analysis in smooth muscle cells. UMAP diagram of smooth muscle cell grouping (A). Compared with the normal group, the volcano map of DEGs in different disease groups (B). GSEA analysis (C–F) of DEGs in COPD, COVID-19, ILD and IPF, co-expression network and enrichment analysis of hub genes of COPD (G) and IPF (H).Abnormal interferon signal activation and ubiquitination signals of B cells and T cellsComparative analysis of 10,083 B cells (Fig. 5A and B) showed that hypoxia of COVID-19 and IPF was inhibited, while ILD was activated (Fig. 5C–F). In addition, COVID-19, ILD and IPF showed abnormal ubiquitination signals, but only COVID-19 upregulated endoplasmic reticulum protein synthesis (Supplementary Fig. S4A and B). The enrichment analysis of hub genes (Supplementary Fig. S4C–F) showed that UBA52, UBB, UBC and HSPA8 were involved in Toll-like receptor signal, interferon signal and ubiquitination abnormality of B cells in various diseases (Fig. 5G and H and Supplementary Fig. S4G and H). DEGs of 31,507 T cells showed that the ribosome gene was abnormally expressed in COVID-19, which seemed to be related to the decrease of oxidative phosphorylation of T cells. (Fig. 6A–F). Ubiquitin of protein and abnormal T cells seem to be the common features of T cells in four diseases, but only COVID-19 is in the activated state (Supplementary Fig. S5A and B). The hub gene screening (Supplementary Fig. S5C–F) and functional analysis showed similar phenomena to those of B cells (Fig. 6G and H and Supplementary Fig. S5G and H).Fig. 5DEGs in B cells and functional analysis. UMAP diagram of B cell grouping (A). Compared with the normal group, the volcano map of DEGs in different disease groups (B). GSEA analysis (C–F) of DEGs in COPD, COVID-19, ILD and IPF, co-expression network and enrichment analysis of hub genes of COPD (G) and IPF (H).Fig. 6DEGs in T cells and functional analysis. UMAP diagram of T cell grouping (A). Compared with the normal group, the volcano map of DEGs in different disease groups (B). GSEA analysis (C–F) of DEGs in COPD, COVID-19, ILD and IPF, co-expression network and enrichment analysis of hub genes of COPD (G) and IPF (H).Increased chemotaxis in monocyte and macrophageIn COVID-19 and ILD, 130,273 macrophages showed a large number of DEGs, but their signals activation and inhibition were very difference (Fig. 7A–F). GO enrichment analysis showed abnormal differentiation and chemotaxis of macrophages, which may be related to the activation of HIF-1 signal and chemotaxis signal (Supplementary Fig. S6A and B). Different from others, there is no abnormality of protein ubiquitination in macrophages of IPF. The hub genes (Supplementary Fig. S6C–F), such as UBB, UBC, HSPA8 and JUN, are involved in the Toll-like receptor signal and ubiquitination process of macrophages in COPD, COVID-19 and ILD (Fig. 7G and H and Supplementary Fig. S6G and H). The 25,194 monocytes showed that the cells were abnormally active when the disease occurred, which was related to the abnormal activation of cellular interferon signal (Fig. 8A–F). Different from others, there was no abnormal protein ubiquitination in ILD monocytes (Supplementary Fig. S7A), but COVID-19 showed the activation of multiple immune pathways (Supplementary Fig. S7B). The hub gene screening (Supplementary Fig. S7C–F) showed that UBA52, UBC and HSP90AA1 were involved in the process of interferon and protein ubiquitination (Fig. 8G and H and Supplementary Fig. S7G and H) of COPD and COVID-19.Fig. 7DEGs and functional analysis of macrophages. UMAP diagram of macrophage grouping (A). Compared with the normal group, the volcano map of DEGs in different disease groups (B). GSEA analysis (C–F) of DEGs in COPD, COVID-19, ILD and IPF, co-expression network and enrichment analysis of hub genes of COPD (G) and IPF (H).Fig. 8DEGs and functional analysis of monocytes. UMAP diagram of monocyte grouping (A). Compared with the normal group, the volcano map of DEGs in different disease groups (B). GSEA analysis (C–F) of DEGs in COPD, COVID-19, ILD and IPF, co-expression network and enrichment analysis of hub genes of COPD (G) and IPF (H).Analysis of expression proportion and abundance of hub genesTo validate the expression stability of hub genes, the data sets of human and mice were integrated. Firstly, the human single-cell atlas (Supplementary Fig. S8A–D) with 400,555 cells including COVID-19, ILD and IPF from external data set was integrated, and then the mouse single-cell atlas (Supplementary Fig. S8E–H) with 281,880 cells from influenza virus mouse model and pulmonary fibrosis model was integrated. Analysis of the expression ratio and abundance of hub genes in the single-cell atlas shows that UBA52 and ribosomal protein S27a (RPS27A) are highly expressed in the endothelial cells of IPF, while UBB, UBC, HSP90AB1 and RACK1 are highly expressed in the endothelial cells of ILD (Fig. 9A). In the another human single atlas, UBC, UBA52, UBB and HSP90AB1 are highly expressed in IPF and ILD (Fig. 9D). However, in the mouse single-cell atlas, the traditional pulmonary fibrosis model cannot simulate the abnormal protein ubiquitination of endothelial cells (Supplementary Fig. S9A). In the two human single-cell atlases, HSP90AB1, JUN, UBA52, UBB, HSP90AA1 and other hub genes are highly expressed in the epithelial cells of IPF and ILD (Fig. 9B–E). Different from COVID-19, UBB, UBC and HSP90AB1 are highly expressed in the influenza animal model and the proportion of cells increases (Supplementary Fig. S9B). The high expression of hub gene and the increased cell proportion can be observed in human smooth muscle cells with two lung diseases (Fig. 9C–F). In addition, the hub genes were significantly up-regulated in B cells, T cells, macrophages and monocytes of IPF and ILD patients in the two single-cell atlases, but the opposite was found in COVID-19 (Fig. 10). In the pulmonary fibrosis model, the high expression of genes and the increased cell proportion can only be observed in T cells, which indicates that the pulmonary fibrosis model induced by bleomycin needs to be further optimized (Supplementary Fig. S9C–F).Fig. 9Expression proportion and abundance of hub genes in two atlases. The expression of hub genes in endothelial cells, epithelial cells and smooth muscle cells in first atlas (A–C) and second atlas (D–F). The circle represents the proportion of genes in cell type, and the color represents the average expression of genes.Fig. 10Expression proportion and abundance of hub genes in two atlases. The expression of hub genes of B cells, T cells, macrophages and monocytes in first atlas (A–D) and second atlas (E–H). The circle represents the proportion of genes in cell type, and the color represents the average expression of genes.Bulk-RNAseq data integration and validationTo further observe the differences of hub genes, we try to verify them on bulk RNA-seq. Firstly, 396 samples from 7 data sets were collected and integrated by “sva” algorithm. Sample distribution boxplot and TSNE analysis before and after integration showed the removal of batch effect (Supplementary Fig. S10). In view of the difference in DEGs of immune cells displayed by single-cell atlas, we found that the infiltration of immune cells in lung tissue was generally out of balance. For example, in plasma cells differentiated from B cells, the infiltration of COVID, ILD and IPF patients increased, while that of CD4 T cells and monocytes decreased (Fig. 11A). On the other hand, the differences of hub genes such as UBA52, UBB, UBC, CD74 and CDC42 between groups can be clearly observed (Fig. 11B). When using hub genes to construct diagnosis models respectively for different diseases, the performance of the hub gene is poor (Fig. S11), while the hub genes shows good diagnostic performance when distinguish normal and diseases (Fig. 11C). In COVID-19, ILD and IPF lung tissues, UBA52, UBB and UBC had a good expression correlation with cell matrix related genes annexin A2 (ANXA2), S100A10 and COL family genes (Fig. 11D), which indicated that protein ubiquitination modification might be related to tissue fibrosis.Fig. 11Bulk RNA-seq data integration analysis of COVID-19, ILD and IPF. Immune infiltration analysis (A), hub gene expression analysis (B), gene diagnosis model (C) and ECM-related genes correlation analysis (D) of three diseases.Drug screening based on hub genesL1000FWD database was used to predict the potential regulatory drugs of hub genes, and we screened out 10 small molecular compounds that are expected to be used for treatment (Table 2). Due to the good binding scores of these compounds to hub genes, molecular docking was performed to explore the possibility of these drugs interacting with ubiquitination-related proteins UBA52, UBB and UBC. It was found that molecules, such as salermide and SSR-69071, have low binding energy with hub genes and have good therapeutic potential.
Table 2 Drug Screening Based on Hub Genes.
Comprehensive analysis of scRNA-Seq and bulk RNA-Seq reveals ubiquitin promotes pulmonary fibrosis in chronic pulmonary diseases
