We acquired the single-cell RNA-seq gene expression dataset from GEO under accession number GSE207651. Mice aged 6-8 weeks were divided into three groups: sham, CLP.24, and CLP.48. Each group consisted of three mice. The CLP.24 group and CLP.48 group underwent cecal ligation and puncture (CLP) surgery. The sham group and CLP.24 group were sacrificed after 24 hours, while the CLP.48 group was sacrificed after 48 hours. Lung tissue samples from each group were mixed for single-cell RNA sequencing. Magnetic-activated cell sorting (MACS) was used to achieve a ratio of 1:1 for CD45+ cells to CD45- cells.

The single-cell library was prepared following the protocol of the 10×Genomics Chromium Next GEM Single Cell 3′ Reagent Kits v3.1. Processed scRNA-seq data were required in the form of raw unique molecular identifier (UMI) count matrix by Cell Ranger (version 5.0.0). We used the Seurat R toolkit (version 4.1.1) for further analyses. The number of cells in each group required from the original data were 8744 cells in the sham group, 12159 cells in the CLP.24 group and 10631 cells in the CLP.48 group. Cells containing fewer than 300 genes or more than 7,500 genes were excluded. Genes expressed in fewer than five cells and those with mitochondrial UMI counts exceeding 10% were excluded. The QC metrics plots can be found in the article/ Supplementary Material . A total of 29,723 cells remained after filtering for further analysis including 7984 cells in the sham group, 11589 cells in the CLP.24 group and 10150 cells in the CLP.48 group. The data were normalized using Seurat’s NormalizeData function. We selected 2,000 variable genes using the FindVariableGenes function and then applied scaling with the ScaleData function, including the regression of the cell cycle. Principal component analysis based on the variable genes was performed using the RunPCA function. Batch effects were mitigated using the RunHarmony function. The cells were clustered into 36 clusters using FindNeighbors and FindClusters functions with a resolution parameter of 1.2. To re-cluster the endothelial cells (3,801 cells), we used the FindClusters function with a resolution of 0.25, resulting in seven clusters. We employed UMAP to visualize the cluster results.

We employed a double-checking strategy to identify the marker genes. Initially, we identified data-derived marker genes using the FindAllMarkers function. Subsequently, we compared these data-derived marker genes to those reported in the CellMarker database. Additionally, marker genes derived from other studies have been identified (11, 12). Finally, we annotated the cell clusters by manually inspecting the cell types based on both data- and study-derived marker genes. Consequently, we categorized the cells into 7 immune cells and 4 stromal cells. Immune cells were categorized using canonical marker genes, including Cd19, Cd79a, and Cd79b for B cells (n=263 cells); Trbc2 for T cells (n=1317 cells); Gzma for NK cells (n=797 cells); Flt3 and Ccl17 for DC cells (n=516 cells); Mrc1, Marco, and Cd44 for Macrophages (n=1550 cells); Ccr2, F13a1, and Plac8 for Monocytes (n=3514 cells); Retnlg and S100a8 for Granulocytes (n=5846 cells). Epithelial cells (n=796 cells) were identified by Sftpa1, Ager, and Cbr2. Smooth muscle cells (2381 cells) and Fibroblasts (8942 cells) were identified by Tagln, Myh11, Acta2, and Dcn, Gsn, Col1a2, respectively. Endothelial cells (n=3801 cells) were distinguished by the expression of Cdh5, Cldn5, Lyve1, and Calcr1, and were further sub-clustered. Arteries (200 cells), veins (243 cells), and lymphatics (315 cells) clusters were annotated using marker genes (Mgp2 for arteries; Slc6a2 and Bst1 for veins; Fxyd6 and Mmrn1 for lymphatics). Capillaries (2233 cells) were annotated using the marker genes Car4, Lpl, and Fibin, and divided into three types of capillaries (Capillary-1, 939 cells; Capillary-2, 686 cells; Capillary-3, 608 cells).

GO enrichment analyses were conducted using Metascape, specifically focusing on “biological process” annotations (http://metascape.org/; v3.5; updated date: 2020-09-16). We used “M. musculus” as the species, and the results were filtered using a p-value cutoff of <0.01 (13). The results were visualized using ggplot2 (version 3.3.6).

Initially, we downloaded the MSigDB Gene Sets for “Mus musculus,” specifying “curated gene sets” and “KEGG” pathways using the “msigdbr” function. Subsequently, we performed a gene set variation analysis using the GSVA package (version 1.42.0).

We conducted differential expression analysis using the “FindMarkers” function. Additionally, we downloaded “c2.cp.kegg.v7.5.1. symbols.gmt” which used as the database. Finally, we performed GSEA analysis by the GSEA function in the clusterProfiler package (version 4.2.2). The results were presented using the gseaplot2 function in the enrichplot package (version 1.14.2) or interpreted based on the normalized enrichment score (NES) and adjusted p-value.

Differentially expressed genes were identified using the “FindMarkers” function, applying criteria of “|avg_log2FC| > 1.2” and “p_val_adj < 0.05” to define the significant DEGs.

We utilized the R package AUCell to assess the pathway activities in individual cells. Initially, we computed gene expression rankings for individual cell using “AUCell_buildRankings” function. The canonical KEGG pathway database “c2.cp.kegg.v7.5.1. symbols.gmt” was obtained from the Broad Institute’s website. Area-under-the-curve (AUC) values were determined by utilizing the “AUCell_calcAUC” function, which relies on gene expression rankings.

The Cellchat package was used to investigate intracellular communication within all endothelial cell types and between T cells or granulocytes and endothelial cell subpopulations. We utilized the mouse database, which includes categories such as “Secreted Signaling,” “Cell-Cell Contact,” and “ECM-Receptor” from CellchatDB. Communication probabilities were analyzed across the three conditions.

To monitor alterations in pulmonary cell composition during endotoxemia, we conducted scRNA-seq analysis of CD45+ and CD45- cells extracted from mouse lung tissue at three time points. We defined four cell types in CD45- cells (15920 cells) ( Figure 1A ): Fibroblasts (Col1a2, Dcn, Gsn, Mfap4), Endothelial cells (Cdh5, Cldn5, Lyve1, Calcrl), Smooth muscle cells (Tagln, Myh11, Tpm2, Acta2), and Epithelial cells (Sftpa1, Ager, Cbr2, Lamp3, Cxcl15) ( Figure 1B ). We identified seven cell types among CD45+ cells (13803 cells) ( Figure 1C ): B cells (Cd19, Cd79a, Cd79b), DC cells (Ccl17,Flt3), Granulocytes (S100a8, Retnlg), Macrophages (Mrc1, Marco, Cd44, Ear2), Monocytes (Plac8, Ccr2, F13a1), NK cells (Gzma, Nkg7, Ccl5), and T cells (Trbc2, Trbc1) ( Figure 1D ). Compared with those in the sham group, there were significant alterations in CD45+ cells. Granulocytes expanded, whereas B cells, T cells, and NK cells shrank in an endotoxemia time-dependent manner. Macrophages and monocytes increased in the CLP.24 group, but remained in the CLP.48 group ( Figures 2A, B ). Among CD45- cells, the number of epithelial cells decreased slightly in an endotoxemia time-dependent manner. The number of endothelial cells and smooth Muscle cells decreased at 24 h in the CLP group but showed a rebound at 48 h ( Figures 2C–E ).

Given the critical role of endothelial cells in ALI. We further subdivided endothelial cells into seven clusters ( Figure 3A ). Three clusters among them were annotated as lymphatic (Fxyd6, Mmrn1) cells ( Figure 3B ), vein (Slc6a2, Bst1) ( Figure 3C ) and artery (Mgp) ( Figure 3D ). There were no significant marker genes in one cluster; therefore, we identified this as an unknown cell type. The other three clusters showed capillary marker genes (Car4, Lpl, and Fibin); therefore, we identified them as Capillary-1, Capillary-2, and Capillary-3 ( Figure 3E ). The proportions of the three capillary clusters exhibited significant changes during endotoxemia ( Figure 3F ). In the sham group, Capillary-3 was the main cell type, whereas in the CLP group at 24 h, Capillary-1 was the main cell type and Capillary-3 decreased to zero. Capillary-2 expanded in an endotoxemia time-dependent manner, peaking at 48 h ( Figures 3G, H ). The change in the total number of endothelial cells between groups was not obvious, while the alteration of capillary subsets was evident. This indicates that the change in capillary subsets is crucial during endotoxemia.

Endothelial cell subtypes showed significant changes in capillary phenotypes, highlighting the crucial role of pulmonary capillaries in endotoxemia. We identified marker genes for the three different capillary clusters in endothelial cells. Inhbb, Plat, Npr3, and Spry4 were highly expressed in Capillary-1 cells ( Figure 4A ). Clec1a, Atp8a1, Lpin2, and Arap2 were highly expressed in Capillary-2 ( Figure 4B ). Capillary-3 cells expressing high levels of Cd74, H2-Ab1, Gbp4 and Cyp1a1 ( Figure 4C ). Biological processes were examined to investigate capillary subtype function. Regulation of sprouting angiogenesis, lipid storage, regulation of blood coagulation, and the Microtubule-Associated Protein Kinase (MAPK) cascade were enriched in Capillary-1 cells ( Figure 4D ). Capillary-2 showed vascular endothelial growth factor receptor signaling and lipid transport enrichment ( Figure 4E ). Antigen processing and presentation, and response to interferon were enriched in Capillary-3 ( Figure 4F ). Moreover, we explored highly expressed genes and their functions in each capillary subset. We found that Plat, Plaur, Lpl and Flt1 were enriched in capillary-1 cells, while Flt1 and CD36 were enriched in capillary-2 cells. Among them, Plat and Plaur are related to blood coagulation (14). Flt1 encodes the receptor for the pro-angiogenic family and plays important roles in angiogenesis (15). Lpl encodes lipoprotein lipase which is a crucial enzyme in lipid metabolism (16). CD36, as an adipocyte receptor, optimizes tissue fatty acid uptake in endothelial cells (17). Capillary-3 was involved in adaptive immune response with high expression of Cd74, Stat, Cd274 ( Figures 5A, B ). Interactions within endothelial cells suggested stronger interactions among capillary subtypes ( Figure 5C ).

To explore the alterations in signaling pathways during endotoxemia in endothelial cells, we performed GSEA and GSVA ( Figure 6A ). The results showed that antigen processing and presentation, and oxidative phosphorylation were decreased in the CLP group at 24 h compared to those in the sham group, whereas complement and coagulation cascades, and leukocyte transendothelial migration were increased ( Figures 6B, C ). The AUCell results indicated that Capillary-3 primarily played a role in antigen processing and presentation, as well as oxidative phosphorylation, while Capillary-1 was associated with complement and coagulation cascades ( Figure 6D ). Compared to the CLP group at 24 h, Jak-Stat, cytokine receptor interaction, Nod-like receptors, cell adhesion molecules, and Toll-like receptor signaling pathways were alleviated in the CLP group at 48 h, while the peroxisome proliferator activated receptor (PPAR) signaling pathway was stronger ( Figures 7A, B ). The AUcell results showed that the Jak-Stat signaling pathway and cytokine receptor interactions were enriched in Capillary-1 cells ( Figure 7C ).

Differential genes were identified using the FindMarkers function. The expression of genes, including Plat, Lpl, Neat1, Npr3, Igfbp4, Entpd1, Apod, Thbs1, and Selp, was upregulated in the CLP group at 24 h compared to that in the sham group ( Figure 8A ). Among these, the protein encoded by Plat can lead to hyperfibrinolysis. Lipoprotein lipase encoded by Lpl is an essential enzyme in lipid metabolism that can increase the production of fatty acids. Neat1 is a long noncoding RNA that is considered an oncogene. Neat1 aggravates sepsis-induced lung injury. Igfbp4, Entpd1, and Apod regulate angiogenesis. Thbs1 is an adhesive glycoprotein that mediates cell-cell or cell-matrix interactions. Selp encodes selectin P, which mediates endothelial-leukocyte interactions. The expression of genes related to antigen presentation, including Gbp4, Gbp7, Cd74, H2-Ab1, H2-Eb1, and Stat1, significantly decreased in the CLP group at 24 h. S100a8, Jcad, Cxcl12, S100a6, Cavin2, Pltp, Nrp1, and Ptp4a3 were upregulated in the CLP group at 48 h compared with those in the CLP group at 24 h ( Figure 8B ). Some genes were differentially expressed in the CLP.24 group compared to the sham group but showed a reverse pattern in the CLP.48 group, including Plat, Spry4, Npr3, Igfbp4, Cavin2, Pltp, Nrp1, and Ptp4a3.

Since the alteration of endothelial cells during endotoxemia will result in a change in communication between endothelial cells and other cell types, furthermore, immune cells are the most involved in endotoxemia. Therefore, we investigated the communication between endothelial cells and certain immune cells. Neutrophils serve as the initial defense during acute inflammation and enter the circulation in large numbers. Their migration from circulation into lung tissues involves a trans-endothelial process. GSEA results showed that leukocyte transendothelial migration was elevated in the CLP group at 24 h compared to that in the sham group. Cellchat was used to examine communication between endothelial cells and neutrophils ( Figure 9A ). In the sham group, capillary-1 was excluded because it contained only three cells. In the CLP.24 group, capillary-3 was excluded because it contained no cells. The results revealed that communication among granulocytes was strengthened in the CLP.24 group, whereas the changes in communication between granulocytes and endothelial cells were not significant ( Figure 9B ). In the CLP.48 group, communication between granulocytes and capillary-2 increased significantly, whereas communication between granulocytes and capillary-1 decreased significantly ( Figure 9C ). Ligand-receptor analysis indicated that intercellular cell adhesion molecule-1 (ICAM-1) found in endothelial cells interacted with granulocytes, with a heightened interaction observed in the CLP group. In the CLP.24 group, capillary-1 primarily participated in the interaction, whereas capillary-2 played a more prominent role in this interaction in the CLP.48 group. In contrast to ICAM-1, intercellular cell adhesion molecule-2 (ICAM-2) was decreased in the CLP.24 group compared to that in the sham group. As a ligand, the strength of the interaction between granulocytes and endothelial cells changed depending on the receptor ( Figure 9D ).

As the antigen processing and presentation signaling pathways changed significantly in the capillaries during endotoxemia, we hypothesized that capillaries might be involved in communication with T cells. The results indicated that the interaction between T cells and capillary-1 was significantly weaker than that between T cells and capillary-3. As the ratio of capillary-3 to capillary-1 decreased, the interaction strength also decreased in the CLP.24 group in comparison to that in the sham group ( Figures 10A–C ). Ligand-receptor analysis also suggested that capillary-T cell communication was noticeably weakened when CD8 acted as a receptor during endotoxemia, indicating a decrease in the MHC-I signaling pathway. There were no notable alterations in the interaction between CLP groups at 24 and 48 h ( Figure 10D ).