We collected single-cell RNA sequencing data in matrix format from the GEO database and subsequently imported the gene expression matrix into R software (version 4.0.0). For downstream analysis, we utilized the Seurat package (version 4.0.0). The gene expression matrix underwent normalization and scaling procedures for each sample. Genes expressed in more than 10 cells were retained for further analysis. Cells with a high proportion (≥10%) of unique molecular identifiers (UMIs) mapping to mitochondrial genes were excluded, as well as cells with fewer than 200 detected genes. We employed the DoubletFinder function for doublet detection and filtering of cells. Following data integration, the expression levels of highly variable genes were scaled and centered.

Clusters were identified using the FindClusters function in Seurat, with a resolution parameter set to 0.1. Differential expression analysis between the identified clusters was performed using the FindAllMarkers function.

Marker genes were identified through the FindAllMarkers function, and cell types were annotated based on well-established classical markers. A second round of clustering was conducted using the same parameters to identify subclusters within the major cell types.

Trajectory analyses were carried out using Monocle2 to predict differentiation pathways among endothelial subclusters.

The CellChat R package (version 1.1.3) was utilized to infer the intercellular communication network between clusters. A CellChat object was created with the normalized expression matrix, and visualization was performed using functions such as netVisual_aggregate and netAnalysis_contribution.

All experimental procedures were approved by the Institutional Animal Care and Use Committee of Anhui University of Science and Technology (Approval No. 20240301) and conducted in accordance with the ARRIVE guidelines. Male C57BL/6 mice (6–8 weeks old, 20–22 g) were purchased from Cavens Laboratory Animal Co., Ltd (Suzhou, China) and housed under specific pathogen-free conditions with a 12 h light/dark cycle, ambient temperature (22 ± 1°C), and ad libitum access to food and water.

The silicosis mouse model was established following the protocol described by Li et al. with modifications. Briefly, mice were randomly divided into control (n=20) and silica-exposed groups (n=20). After anesthesia with 3% isoflurane, animals received a single intranasal instillation of 50 μl sterile silica suspension (Sigma-Aldrich; Sigma, 14808-60-7; SiO₂ particles, 1–5 μm diameter, 100 mg/ml in saline). Control mice received equal-volume saline instillation. To investigate dynamic pathological changes, subgroups of silica-exposed mice (n=10 per timepoint) were euthanized at 7 days (acute inflammation phase) and 56 days (fibrotic phase) post-exposure via intraperitoneal pentobarbital overdose (150 mg/kg).

Lung tissues were fixed via tracheal cannulation with 4% paraformaldehyde under room temperature for 24 h, followed by graded ethanol dehydration (70% to 100%), xylene clearing, and paraffin infiltration (60°C, two cycles). Paraffin-embedded blocks were sectioned at 5 μm thickness using a rotary microtome (Leica RM2235).

Paraffin-embedded lung tissue sections (4-μm thickness) were deparaffinized in xylene and rehydrated through a graded ethanol series (100%, 95%, 70%). Antigen retrieval was performed by boiling slides in 10 mM sodium citrate buffer (pH 6.0) for 20 min using a microwave. After cooling to room temperature (RT), sections were permeabilized with 0.3% Triton X-100 for 15 min and blocked with 5% normal serum (species-matched to secondary antibodies) containing 1% BSA for 1 h at RT. Tissues were incubated overnight at 4°C with primary antibodies against VEGF-a (Servicebio, Cat. No. GB15165-100; 1:200; green fluorescence) and HIF-1α (36169S; 1:200; red fluorescence), followed by species-specific Alexa Fluor-conjugated secondary antibodies (Servicebio, Cat. No. G1213-100UL; 1:500; 1 h at RT). Nuclei were counterstained with DAPI (0.1 μg/mL, 10 min). Slides were mounted with antifade medium (ProLong Gold) and imaged using a confocal laser scanning microscope (FV2000, Olympus) with a 20× or 40× oil-immersion objective. Co-localization analysis of VEGF-a and HIF-1α signals was performed using ImageJ (Fiji plugin) with threshold-based region-of-interest (ROI) selection.

Quantitative data for VEGF-a and HIF-1α fluorescence intensity (Panels G-H) are expressed as mean ± standard deviation (SD). Statistical significance between groups was assessed using unpaired two-tailed Student’s t-test (for two-group comparisons) or one-way ANOVA with Tukey’s post hoc test (for multi-group comparisons). Normality and variance homogeneity were verified using Shapiro-Wilk and Levene’s tests, respectively. Sample size (n) represents biologically independent lung tissue sections (technical replicates: 3–5 fields per section). Significance thresholds were defined as *p < 0.05 and **p < 0.01; ns indicates non-significant differences (p ≥ 0.05). All analyses were conducted using GraphPad Prism v9.0 ( Figure 1 ).

To investigate the potential mechanisms of endothelial cells during the progression of silicosis, we analyzed scRNA-Seq data from lung tissues of SiO2-7d and SiO2-56d silica-induced mouse models (6). Transcriptomic data from all groups were integrated and normalized, followed by non-linear clustering using UMAP. Cell clusters were annotated using classical cell markers from the CellMarker database, identifying nine major cell types: macrophages, neutrophils, T cells, B cells, endothelial cells, epithelial cells, fibroblasts, cycling cells, and neuronal cells, which were visualized on the UMAP plot ( Figures 2A–C ). To explore the cellular composition across samples, we calculated the proportion of each cell type in each group ( Figure 2D , Supplementary Table S1 ). Compared to control mice (NS-7d), silica-exposed mice at 7 days (SiO₂-7d) exhibited a significantly increased macrophage proportion (from approximately 30% to nearly 50%), consistent with Zhou et al. who highlighted the predominant role of macrophages in early inflammation. Notably, the proportion of endothelial cells was also significantly increased at both 7 days and 56 days post-exposure (NS-56d: 3%; SiO₂-56d: 6%), indicating sustained endothelial activation throughout silicosis progression.

We performed differential gene expression (DEG) and Gene Ontology (GO) enrichment analyses at 7 and 56 days to investigate the molecular processes underlying endothelial changes. GO analysis revealed significant enrichment of endothelial cell migration, vascular endothelial growth factor production, and negative regulation of endothelial proliferation ( Figures 3A, B ). Heatmap and dot plot analyses further demonstrated enhanced expression of inflammatory and pro-angiogenic genes such as Vegfa, Hif-1α, Tnf, II1a, and Apoe, predominantly within macrophages and endothelial cells ( Figures 3C, E ). Feature plots corroborated these findings, illustrating macrophage-driven regulation of endothelial biological processes, including migration and angiogenesis ( Figure 3D ).

Immunofluorescence staining validated these bioinformatics results, showing significantly increased co-localization of VEGF-a and HIF-1α proteins in endothelial cells following silica exposure ( Figure 3F ). Quantitative analysis indicated significantly higher VEGF-a expression in endothelial cells at 7 days post-silica exposure compared to controls (p < 0.05), whereas no significant difference was observed at 56 days ( Figure 3G ). In contrast, HIF-1α expression was markedly elevated in endothelial cells at both 7 days (p < 0.001) and 56 days (p < 0.01) post-exposure compared to controls ( Figure 3H ). These findings confirm a key role for macrophage-mediated VEGF-a and HIF-1α signaling in endothelial cells during silica-induced lung injury.

To further dissect endothelial heterogeneity, we subclustered endothelial cells into five distinct subpopulations (C0–C4) ( Figures 4A, B ). The composition of endothelial subpopulations shifted notably following silica exposure; specifically, the proportion of subpopulation C1 significantly increased (from 21.8% to 32.8%) at 56 days, whereas subpopulations C0 and C2 decreased, suggesting divergent differentiation trajectories associated with silicosis progression ( Figure 4C ). Pseudotime trajectory analyses supported this bifurcation, revealing two primary differentiation paths from progenitor-like (C0) cells towards pro-inflammatory (C1) and angiogenesis-related (C2) states ( Figures 4D, E ).

Marker gene analysis further clarified subpopulation characteristics. The progenitor-like C0 subpopulation, characterized by Kit expression, showed enrichment for endothelial development and migration pathways. Subpopulation C1 predominantly expressed immune-related genes involved in neutrophil chemotaxis (e.g., Il1b, Vcam1), whereas C2 exhibited gene signatures associated with angiogenesis and tissue repair ( Figures 5A–C ).

To elucidate the interaction between endothelial cells and immune cells during silicosis, we utilized CellChat to analyze cell-cell communication. At 7 days post-exposure, C1 endothelial cells predominantly communicated through signaling interactions mediated by ESAM and CCL family proteins, and SPP1 pathways with neutrophils and macrophages, enhancing neutrophil recruitment ( Figures 6C, D ). Conversely, the C0 and C2 subpopulations were regulated primarily through TNF signaling from macrophages, indicating macrophage-mediated modulation of endothelial cell function ( Figures 6A, B ).

By day 56, C1 endothelial cells sustained pro-inflammatory signaling via SPP1 and CCL pathways, actively regulating macrophage and neutrophil activities, reinforcing persistent inflammation ( Figures 7B–D ). In contrast, C0 and C2 subpopulations showed unique activation of GDF (growth differentiation factor) and GAS (growth arrest-specific protein) pathways, indicating attempts at repair and regeneration ( Figures 7A, B ). Thus, endothelial subpopulations orchestrate complex responses during silicosis, simultaneously contributing to inflammation and attempting tissue repair.

Collectively, these findings indicate that endothelial cells significantly contribute to silicosis progression, predominantly through promoting sustained inflammation while concurrently attempting tissue repair. These insights have important implications for the development of targeted therapeutic strategies.