GSE51472 and GSE12644 were downloaded from the GEO database (12, 13). GSE51472 included 5 control, 5 sclerotic, and 5 calcified samples, while GSE12644 included 10 control and 10 calcified samples. In R software, Counts data were converted to FPKM and then log-normalized. Sample boxplots were plotted to assess the degree of normalization.

“edgeR” (14) was used to calculate gene expression changes, and DEGs were screened with the threshold of Log2|FC| ≥ 1 and adjusted P-value <0.05. “msigdbr” (15) was used for gene set enrichment analysis (GSEA) of DEGs, and “enrichplot” was used to plot the top-ranked terms.

The STRING database (16) was used to construct the protein-protein interaction (PPI) network of DEGs, and Cytoscape software (17) was used to visualize the interaction relationships between genes. “clusterProfiler” (18) was used for GO and KEGG pathway analysis using a significance cutoff of P < 0.05, and the SRPLOT platform (19) was used to visualize the relevant enriched terms.

The aortic valve tissues of patients with aortic regurgitation (control) and AS in the hospital from January to April, 2024 were collected. Among them, the organization acquisition method is implemented in accordance with relevant guidelines and regulations, and it is confirmed that all subjects and/or their legal guardians have obtained informed consent. This project was approved by the Xinjiang Uygur Autonomous Region People's Hospital (KY2024030102). Aortic valve tissues were washed with pre-cooled saline within 10 min to remove blood residues. Leaflets were separated, tissues were cut into small pieces (<0.2 cm³), snap-frozen in liquid nitrogen, and stored at −80 °C. After thawing, tissues were soaked in decalcification solution (4 °C, 24–48 h), with fresh solution replaced every 6 h. After decalcification, tissues were ground into powder with liquid nitrogen, and interference was removed by differential centrifugation. Protein expression was detected by liquid chromatography-mass spectrometry (LC-MS/MS), and MaxQuant was used to match mass spectrometry data to the protein database.

The HE staining procedure for aortic valve tissues included: formalin fixation for 24–48 h, dehydration (gradient ethanol treatment), transparency (xylene), paraffin embedding, and sectioning; the staining process included dewaxing and rehydration, hematoxylin nuclear staining for 5–10 min, hydrochloric acid-ethanol differentiation, water reblueing, eosin cytoplasm staining for 1–2 min, followed by gradient ethanol dehydration, xylene transparency, and neutral gum sealing for microscopic observation of cell morphology and collagen fiber structure.

The alizarin red staining procedure for aortic valve tissues was: dewaxing sections to water, staining in alizarin red S solution for 5–10 min, washing with running water to remove floating color; counterstaining nuclei with hematoxylin for 30 s, hydrochloric acid-ethanol differentiation, water reblueing, gradient ethanol dehydration, xylene transparency, and neutral gum sealing.

The MCODE algorithm (20) was used to identify key modules in the PPI network. Five algorithms in Cytohubba (21) were used to detect the top 30 key genes in the PPI network. UpSet (22) was used to visualize the overlap of the five algorithms.

The SRPLOT platform was used to construct ROC curves for proteomic sequencing data. The TRRUST database (23) was used to predict transcription factors of hub genes, and the M6A2Target database (24) was used to predict m6A-modified genes of hub genes.

PCR experiments were performed to detect the mRNA expression levels of MMP9 and PLAU in aortic valve tissues. Specific steps: frozen tissues were ground in liquid nitrogen, lysed using an RNA extraction kit, centrifuged to remove impurities, and total RNA was purified by binding to an RNA adsorption column. Reverse transcription was performed according to the Takara PrimeScript RT Master Mix instructions (42 °C for 15 min, 85 °C for 5 s to inactivate), synthesizing cDNA; qPCR amplification was performed using Takara SYBR Premix Ex Taq (95 °C pre-denaturation for 30 s, 40 cycles: 95 °C for 5 s, 60 °C for 30 s). Melting curves were used to verify product specificity, and the relative expression of target genes was calculated. The primer sequences as shown in Table 1.

Published single-cell transcriptome data (GSE220774) (25) from aortic calcification patients were collected, including single-cell transcriptome data from three regions (fibrosa layer, ventricular layer, and intermediate layer/remaining layer) of five patients. Data preprocessing strictly followed the Seurat official recommended pipeline (26), including filtering low-quality cells and noise genes, data normalization, identification of highly variable genes, principal component analysis for dimensionality reduction, Louvain clustering algorithm for cell subset identification, cell type annotation using “SingleR” and “Cellmarker” (27, 28), and finally “FindMarkers” for DEG analysis between different cell populations.

Cell pseudotime analysis maps single-cell transcriptome data to a low-dimensional space, constructs developmental or differentiation trajectories between cells, and infers dynamic changes in cell states. Monocle3 (29) was used to analyze the differentiation trajectories of monocytes and macrophages, which assigns a “pseudotime” value to each cell, identifies differential gene modules along the trajectory, reveals differentiation-driving genes and branching events, and finally visualizes time-dependent gene expression patterns through trajectory plots. The specific steps include using DDRTree to reduce dimensionality, sort and map cells, and the built-in Branched expression analysis modeling (BEAM) is used to assist in branch judgment.

After gene annotation and normalization of the GSE51472 dataset (Figure 1A), comparison of gene expression between the control group and aortic sclerosis group showed that aortic sclerosis had minimal impact on gene expression (Figure 1B), but activated proinflammatory signals (IL-6-STAT3, TNF-α, and IL-2-STAT5), hypoxic signals, and fibrosis (epithelial-mesenchymal transition) (Figure 1C). Compared with the control group, aortic calcification patients had significantly upregulated collagen molecules (COL1A1 and SPP1), proinflammatory molecules (CXCL13, TNFRSF17, and S100A8), and matrix metalloproteinase (MMP) family genes (Figure 1D). Meanwhile, GSEA results for aortic calcification and sclerosis were consistent (Figures 1C,E), and the PPI network of upregulated genes showed that integrins and proinflammatory factors played important roles (Figure 1F). Two key modules were identified by the MCODE algorithm (Figure 1G), both related to inflammatory responses such as cytokine production, Toll-like receptor pathway, cell chemotaxis, and NF-κB signaling pathway (Figures 1H,I). This suggests that the accumulation of extracellular matrix (ECM) and the local inflammatory microenvironment may jointly promote the formation of aortic calcification.

Normalization of the GSE12644 dataset (Figure 2A) showed that aortic calcification significantly increased genes such as MMP9, MMP12, and SPP1 compared with the control group (Figure 2B), which are involved in inflammatory response and fibrosis progression (Figure 2C). Construction of the PPI network revealed that upregulated MMP9, SPP1, and COL3A1 were in central positions (Figure 2D). Similar to the GSE51472 dataset, these DEGs were mainly related to extracellular matrix formation, cell chemotaxis, and cytokine production (Figure 2E). Intersection analysis of the two datasets identified 31 genes significantly upregulated in aortic calcification (Figure 2F), with MMP9, CXCL8, SPP1, and PLAU ranking among the top (Figures 2G,H). GO and KEGG enrichment analyses showed that intersection genes were related to cell chemotaxis and ECM formation, consistent with the pathological changes in the overall valve tissue (Figure 2I).

Eight aortic valves from patients with aortic regurgitation (control) and eight from AS patients were collected, with basic information listed in Supplementary Table S1. HE staining of valve tissues showed that collagen fibers (red) in the control group were neatly arranged at 100× magnification (Figure 3A), while those in AS patients showed disorganized collagen fibers with extensive blue-violet calcium salt deposition (Figure 3B). Alizarin red staining showed that normal valve tissues had almost no red staining and aggregation (Figure 3C), while AS valves had abundant red complexes with minimal adhesion at junctions (Figure 3D). Aortic calcification is accompanied by the activation of tissue fibrosis, yet its driving factors remain to be comprehensively evaluated.

Since proteins are the primary executors of cellular functions, proteins were collected and subjected to proteomic sequencing in this study. After proteomic sequencing of 16 samples, gene annotation and normalization were performed (Figure 4A). DEGs showed significant upregulation of proteins such as COL10A1, THBS2, and S100A8 (Figure 4B). Heatmaps showed stable high expression of COL10A1, S100P, and ITGA2B in AS (Figure 4C), with these DEGs involved in inflammatory response, hypoxia, and fibrosis (Figure 4D), consistent with transcriptomic data. GO enrichment analysis showed these genes were related to ECM formation, interleukin and chemokine production (Figure 4E), as well as pathways such as complement and coagulation cascades, and ECM-receptor interaction (Figure 4F). The PPI network showed that dysregulated genes were primarily upregulated (Figure 4G), participating in processes such as wound healing response, ECM formation, and cell chemotaxis (Figure 4H). This further confirms that immune cell activation and fibrosis are risk factors for aortic calcification.

To identify the driving factors that drive aortic calcification, five algorithms were used to calculate the top 30 genes in the DEG network, with overlapping genes defined as hub genes. Sixteen hub genes were obtained (Figure 5A), significantly enriched in processes such as ECM-receptor interaction, damage response, and leukocyte migration (Figure 5B). Analysis of hub gene expression in proteomic data showed upregulation in AS (Figure 5C), while in datasets GSE12644 and GSE51472, only MMP9, PLAU, THBS2, and SERPINE1 had significantly increased mRNA expression in calcified aortic valves (P < 0.05, Figures 5D,E).

To clarify the important value of the identified genes, 16 hub genes were predicted AS in proteomic data. 13 hub genes including ITGA2B, THBS2, and MMP9 had ROC values >0.8, indicating good diagnostic efficacy in distinguishing AS (Figure 6A). To clarify the regulation of hub genes, transcription factors were predicted, identifying 33 TFs with regulatory relationships to hub genes (Figure 6B), but these TFs had no impact on the high expression of hub genes (Figure 6C). m6A modification prediction showed that hub genes such as MMP9 and PLAU were regulated by 29 m6A enzymes (Figure 6D), with significantly reduced protein expression levels of RBMX, YTHDF1, and HNRNPC (P < 0.05, Figure 6E). Intersection of aortic calcification intersection genes (transcriptome) and hub genes (proteome) yielded two genes, PLAU and MMP9 (Figure 6F). qPCR results showed significantly higher mRNA expression of PLAU and MMP9 in AS compared with controls (Figures 6G,H).

To clarify the molecular mechanism of PLAU and MMP9 in aortic calcification, published patient single-cell transcriptome data (GSE220774) were collected and characterized. In sequencing data, the number of RNAs showed no significant correlation with mitochondrial proportion (Figure 7A) but a high correlation with RNA features (Figure 7B), indicating high data quality. Cell annotation identified endothelial cells, macrophages, monocytes, smooth muscle cells, and T cells (Figure 7C), distributed across different fibro-calcification (FC) scores (Figure 7D). In total smooth muscle cells, PLAU and MMP9 expression had no obvious correlation with FC scores, with PLAU mainly highly expressed in ventricular-side smooth muscle cells of highly calcified regions (Figure 7E). In both overall and region-specific T cells, the two genes were mainly expressed in T cells of moderately calcified regions (Figure 7F). In endothelial cells, PLAU was highly expressed in ventricular-side endothelial cells of calcified regions (Figure 7G). In both overall and regional analyses, PLAU and MMP9 were significantly highly expressed in macrophages of highly calcified regions (Figure 7H), and monocytes, similar to endothelial cells, had PLAU highly expressed in ventricular-side monocytes of calcified regions (Figure 7I). This suggests that the high expression of the two genes may be associated with the immune cell activation identified in the bulk-RNA data.

To clarify the impact of fibrosis on different cells, the study conducted a systematic analysis of 5 cell types in different regions separately. In smooth muscle cells, higher FC scores were associated with significant changes in cardiac valve morphology, glycolysis, damage repair, TGF-β signaling pathway, and HIF-1 signaling pathway compared with lower FC scores (Figure 8A). In T cells with higher FC scores, cytokine production, T cell receptor signaling pathway, NF-κB signaling pathway, and HIF-1 signaling pathway were significantly altered compared with lower scores (Figure 8B). In different regions of the aortic valve, highly calcified endothelial cells showed significant upregulation of TNF signaling pathway, MAPK signaling pathway, endothelial cell development, and response to oxidative stress (Figure 8C). In macrophages, the highly calcified fibrosa layer showed stronger proinflammatory signals (Figure 8D). Similar to macrophages, monocytes in the highly calcified fibrosa layer also showed enhanced proinflammatory signals (Figure 8E), possibly an important factor in their differentiation into macrophages.

Given the high expression of PLAU and MMP9 in macrophages and monocytes of highly calcified regions, pseudotime analysis was used to clarify their roles. In monocytes, initial positions of the differentiation trajectory were mainly cells not expressing PLAU and MMP9, while terminal positions were cells expressing both genes (Figures 9A,B). Similar to monocytes, initial-position cells in macrophage differentiation trajectories did not express PLAU and MMP9, while terminal cells mainly expressed these genes (Figures 9C,D). Compared with cells not expressing PLAU, PLAU-expressing cells had enhanced chemotaxis, migration, and inflammatory responses (Figure 9E). Compared with cells not expressing MMP9, MMP9-expressing cells showed enhanced lipid transport and small-molecule metabolism (Figure 9E). Therefore, these two genes may play different roles in the functions of monocytes and macrophages.