To investigate the transcriptome features of TV that contribute to their higher resistance to calcification, we integrated the publicly available scRNA-seq datasets from TR (accession number HRA010091) and CAVD (PRJNA562645) [13]. A total of 128,817 cells passed quality control (QC) from an initial 138,568 cells across all samples. The detailed patient information and cell counts per sample before and after QC are provided in Supplementary Table. The R package “Harmony” was used to correct for batch effects. After rigorous processing of the raw data, which included the removal of low-quality cells and doublets, a total of 93,916 cells from TR and their TV controls, and 34,901 cells from CAVD and their AV controls, were integrated and clustered (Figures 1(A)-(C)). Seven major cell types were identified using cluster-specific biomarkers, VICs, myofibroblasts, VECs, mast cells, myeloid cells, T/NK cells, and a distinct population of valve-derived stromal cells (VDSCs) [14] in CAVD/AV but not in TR/TV samples (Figures 1(A)-(C)). The VDSC cluster was identified by the expression of inflammatory and mesenchymal activation markers (INHBA, CXCL1, CCL20, ID4, SOX4) [15]. Myofibroblasts were identified by the expression of TAGLN, ACTA2, ITGB1, and THBS2 [16]. The nerve-like VIC population was supported by an enrichment of ECM and neuro-associated genes (VCAN, TTYH1, LGI4, VEGFA) [16]. The VICs and VECs in both CAVD/AV and TR/TV groups were less aggregated than the immune cells (T/NK and myeloid cells) (Figure 1(A)).

Given that the cell atlas of tricuspid valves is less well studied compared to that of aortic valves, we first focused our analysis on various cell types in TR and control TV samples. The TR/TV-derived VICs and myofibroblasts were sub-clustered, and seven distinct VIC clusters were identified, each with unique gene expression profiles and transcriptional regulons (Figure 1(D), Figure 1(E), Figure S1(A), Figure S1(B)). The F5 cluster was characterized as resident VICs due to high expression of TCF21, CD34, OSR1, and the active transcription factor OSR1, which promotes myogenesis (Figure 1(E) and Figure S1(B)) [17][18]. The F1 cluster exhibited high expression and activity of GATA4, potentially involved in myocardial capillarization under pressure overload (Figure 1(E) and Figure S1(B)) [19]. Clusters F2 and F3 were annotated as inflammatory VICs due to their expression of chemokines (CCL2 and CXCL12) and type I interferon (IFN)-regulated genes (Figure 1(E)). The F4 cluster was identified as a structural VIC cluster, enriched in transcriptional regulons related to fibroblast activation and

Figure 1. Cell composition of tricuspid and aortic valve leaflets (A)-(C) and comparison of the transcriptome features of VICs between TV and TR groups (D)-(H).

pathological cardiac fibrosis (MEOX1 and KDM5B) [20][21], and expressed genes involved in ECM production, remodeling, and degradation (Figure 1(E) and Figure S1(B)). The F6 cluster was characterized as a POSTN⁺ myofibroblast cluster with features of healing and scar formation [22], while F7 was designated as a nerve-like VIC cluster based on specific gene expression patterns (LGI4 [23], GPM6B [24], TTYH1 [25] and enriched transcriptional regulons (TEAD [26], SOX9, TBX20, IRF7 [27]-[29]) (Figure 1(E) and Figure S1(B)). Previous studies have identified GUCYA1- and HES4-expressing NO-associated VIC clusters in adult heart-derived non-diseased TVs and an APOE⁺ elastin-VIC cluster in fetal heart-derived valves [12][30]. A few cardiomyocytes with high expression of genes encoding myosin light chains and troponin were also identified, designated as the F8 cluster (Figure 1(D), Figure 1(E)). In our dataset, cells expressing these markers were enriched in inflammatory VICs (F2 and F3), myofibroblasts (F6), and nerve-like VICs (F7) (Figure 1(E)). However, elastogenesis-related genes such as ELN and EMILIN1 were not found in any VIC clusters. Notably, these VIC clusters expressed high levels of MHC class I and IFN-regulated genes, suggesting a potential role in antigen presentation to CD8⁺ T cells.

We next compared the VICs between the TR and control TV groups. At the cellular level, the myofibroblast (F6) cluster was significantly more prevalent in the TR group (Figure 1(F) and Figure S1(C)).

At the transcriptome level, VICs in the TR group exhibited increased expression of genes associated with cell metabolism, IFNα/β signaling, and the matrisome compared to the Ctrl group (Figure 1(G) and Figures S1(D)-(F)). Structural VICs (F4) in the TR group further up-regulated VEGFA targets and their responsiveness to TNF-α and IFN-γ, while myofibroblasts (F6) up-regulated genes involved in endothelial-mesenchymal transition (EMT) and cardiac muscle contraction (Figure 1(G)). Notably, VICs in the TR group revealed significant down-regulation of the elastogenesis regulator APOE, proteostasis-regulator heat shock proteins (HSPs), and oxidative-stress inhibitor metallothioneins (MTs) (Figure S1(E), Figure S1(F)) [30]-[33]. STRING analysis of these DEGs revealed that the up-regulated genes were associated with the network of ECM and IFN responsiveness (BST2, IFITM1, ISG15, IFI27, and HLAs), while down-regulated genes were involved in protein stability and negative regulation of cellular metabolic process (Figure 1(H)). Pseudo-time analysis indicated that the TR group had defects in the upregulation of growth factor MDK (midkine) [34] and the cardioprotective molecule NCL (nucleolin) [35], and down-regulation of the EMT-promoting molecule POSTN (periostin) [36] along the cell differentiation path (Figure S1(G)). Collectively, these results indicate that the TR samples exhibit elevated VIC activation and fibrosis, characterized by upregulation of IFN responsiveness, ECM production, and oxidative stress, enhanced myofibroblast differentiation, and impaired proteostasis.

In the tricuspid valves, VICs accounted for the majority of interactions among all cell types (Figure 2(A)). Compared to the control TV group, the TR group exhibited a marked reduction in cellular communication both among various cell types and within VICs themselves (Figure 2(B)). The most significantly down-regulated ligand-receptor pairs in the TR group were cardioprotective heparin-binding growth factors and their glycosylated protein receptors [37]-[40] (MDK-SDC2, MDK-NCL, PTN-SDC2, PTN-NCL), which are involved in interactions within VICs, between VICs and myofibroblasts, and between VICs and T/NK

Figure 2. General characteristics and comparison of the inferred intercellular ligand-receptor interactions between TV and TR groups.

cells (Figure 2(C)). Additionally, interactions involving the pro-inflammatory cytokine IL-6 and its receptor within VICs, as well as the anti-inflammatory molecule ANGPTL4 [41][42] in VECs and its binding partner HBEGF in VICs, were also decreased in the TR group (Figure 2(C)). Conversely, the TR group showed an upregulation of profibrogenic factors [43]-[45] and their receptors, such as POSTN-ITGAV/ITGB5 and PDGFD-PDGFRB, both within VICs and between VICs and myofibroblasts (Figure 2(D)). Communication between immune cells and VICs, mediated by inflammatory chemokines (CCL2, CCL3, and CCL4) and their receptors, was also heightened in the TR group (Figure 2(D)).

Further analysis of the crosstalk within VIC clusters revealed dominant communication between GATA4⁺ (F1) and two inflammatory VIC clusters (F2 and F3), with the TR group exhibiting increased interaction strength compared to the control TV group (Figure 2(E)). Enhanced communication between these VIC clusters and myofibroblasts (F6) was also observed in TR samples (Figure 2(E)). The patterns of cellular communication involving MDK, PTN, POSTN, and PDGFD and their respective receptors differed between the TR and control TV groups (Figure 2(F)). In the TV group, MDK produced by GATA4⁺ and inflammatory VICs primarily regulated themselves, while GATA4⁺, structural, and resident VICs sent pro-angiogenic factor PTN to GATA4⁺ and inflammatory VICs (Figure 2(F)). In the TR group, this interaction pattern was attenuated, with a substantial increase in communication between inflammatory VICs/myofibroblasts and other VIC clusters (Figure 2(F)). The primary sources of POSTN also shifted from the structural VIC cluster (F4) in the control TV group to GATA4⁺ VIC and myofibroblasts in the TR group (Figure 2(F)). Together, these results indicate that TVs in TR patients are associated with significantly altered cellular communication, characterized by a reduction in cardioprotective signaling but an enhancement in profibrogenic crosstalk within VIC clusters and between VICs and other cell types.

VECs lining the TV serve as the initial cellular responders to shear stress and pressure fluctuations. Our analysis identified five distinct clusters among TR/TV-derived VECs (Figures 3(A)-(C) and Figure S2(A), Figure S2(B)). The predominant cluster, referred to as protective VEC (E1), exhibited high expression levels of COLEC11 [12], ENG, HLA, as well as significant activity of transcription factors FOXC2 [46] and NFATC3 [47][48] (Figure 3(C)). Although not statistically significant, a slight reduction in this cluster was observed in the TR group (Figure S2(C)). Compared to protective VECs (E1) from control TV samples, those from TR samples showed signs of oxidative stress, with upregulation of genes related to oxidative phosphorylation and reactive oxygen species, and downregulation of genes involved in TNF signaling through NF-κB (Figure 3(D)). Cluster E4 was characterized as repairing VECs due to their high expression of WNT, NOS3, TAGLN2, ANGPTL2, integrins, membrane transportation molecules, and repair molecules S100A10 and S100A11 [49], along with high activity levels of EMT-associated transcription factors (SNAI2 (Slug) and TEAD4) (Figure 3(B), Figure 3(C)). However, the reparative function of these cells in the TR group appears to be compromised, as they exhibited upregulation of genes associated with TGFβ signaling and downregulation of genes related to NF-κB, hypoxia, EMT, apoptosis, and cell metabolism (Figure 3(D)). Cluster E5 was designated as shear stress-reactive VECs due to their expression of mechanosensitive channel PIEZO1, pro-angiogenic factor SOX17 [50], and von Willebrand factors (Figure 3(B)). High levels of transcription factor activity, including MEOX1, NFKB2, SMAD1, TEAD2, were also observed in E5 (Figure 3(C)). E5 cells from TR samples displayed a pro-inflammatory phenotype, with increased NF-κB signaling and inflammatory response, but reduced pro-angiogenic HEDGEHOG signaling and EMT (Figure 3(D)). The expression of flow-responsive genes, such as KLF4, ADAMTS [51], was also up-regulated in TR-derived E5 cells (Figure 3(E)). Notably, almost all VECs in TR samples showed an upregulation of the long non-coding

Figure 3. Comparison of the transcriptome features and intercellular interactions of VECs between TV and TR groups.

RNA XIST (X-inactive specific transcript), whose dysregulation has been implicated in cardiovascular diseases like fibrosis and hypertrophy [48] (Figure 3(E), FigureS2(D)). STRING analysis of these DEGs revealed that up-regulated genes were associated with type I IFN and TGF-β responsiveness (PTBP3, CTHRC1, TIMP3) and negative regulation of cell proliferation, while down-regulated genes were linked to metallothioneins and collagen-containing ECM components (TGM2, APOE, TIMP1, ECM1, CLEC3B, SMOC1, and DPT) (Figure 3(F)). These findings suggest that VECs in the TR group are dysregulated, acquiring phenotypes indicative of inflammation and oxidative stress, likely due to shear-stress-induced reprogramming.

We also investigated the intercellular communication between VECs and other cell types. We discovered that the TR group downregulated the anti-fibrogenic and anti-inflammatory VEC-VIC interaction (ANGPTL4-SDC2) [52][53] but up-regulated profibrogenic VEC-myofibroblast interactions (POSTN-ITGAV/ITGB5 and INHBA/SCVR1B [54]) (Figure 3(G), Figure 3(H)). Of note, the ability of VECs to scavenge myofibroblast-produced chemokine CXCL12 via ACKR3 [55] was diminished (Figure 3(G), while their secretion of CXCL12 to recruit T/NK cells via CXCR4 was increased (Figure 3(H)). Consistently, interaction between immune cells and VECs via CCL3/4 and CCR5 was also elevated (Figure 3(H)). These results indicate that VECs in the TR group engage with other cells in a pro-inflammatory and profibrogenic manner.

Macrophages, both resident and monocyte-derived, play a crucial role in the regulation of cardiac remodeling. Our analysis identified nine macrophage clusters and one neutrophil cluster (CSF3R⁺S100A8⁺) based on gene expression profiles and transcriptional regulons (Figures 4(A)-(C), Figures S3(A)-(D)). Four of these clusters (M1, M3, M4, and M5), which constituted 57.91% ± 10.25% of the total macrophages, expressed markers indicative of resident macrophage (RM), such as CD163 and LYVE1, along with the transcriptional regulon (MAFB and MAF) (Figure 4(B), Figure 4(C), FigureS3(D)) [56]-[61]. Among the CD163⁻ macrophages, we identified two MHC-II^hi clusters (M6 and M7) with high levels of pro-angiogenic cytokines CXCL16 and IL18 [62][63], an IL1B-expressing pro-inflammatory cluster (M9), and a DCN-expressing myofibrolast-like macrophage cluster (M2) with high activity of transcription factors that promote valve remodeling and angiogenesis (TWIST1, GATA4, SOX9) [19][28][64]-[66] (Figure 4(B), Figure 4(C), FigureS3(B)). There were no significant differences in the abundance of these subpopulations between the TR and control TV groups (Figure S3(E)).

In comparison to the control TV samples, both CD163⁺ and CD163⁻ macrophages in TR samples were highly activated, as indicated by increased expression of genes related to cell metabolism, matrisome, initial complement triggering, antigen presentation, and chemokine/chemokine receptor interaction (Figure 4(D)). Notably, pro-inflammatory molecules such as CCL2,CCL3,CCL4, S100A8, and S100A9 were significantly up-regulated, while VEGFA and metallothionein genes were down-regulated in multiple TR-derived macrophage clusters (Figure S3F). Neutrophils, though sparse in the tricuspid valves, showed signs of activation in the TR group, with higher expression of matrisome- and oxidative phosphorylation-related genes (Figures 4(A)-(D), FigureS3(D)). STRING analysis

Figure 4. Comparison of the transcriptome features and intercellular communications of myeloid cells between TV and TR groups.

of these DEGs revealed that up-regulated genes were associated with the inflammatory response (LY96, CCL2, CCL4, IL18, LGMN, BAX, CLU, TREM2, C1QC, HLA), while down-regulated genes were linked to metallothioneins and positive regulation of IL-4 and IL-10 production (HSPH1, HSPD1, TLR2, TNFRSF1B, NLRP3, BCL2, CEBPB, RARA, ZBTB16) (Figure 4(E)). These data collectively indicate that myeloid cells in TR exhibit high activation with pro-inflammatory and profibrogenic signatures.

We also observed close interactions between myeloid cells and T/NK cells, as well as between myeloid cells and VICs. Compared to the control TV samples, TR samples showed increased communication between myeloid cells and T/NK/VICs via chemokines CCL2, CCL3, CCL4, and their receptors (Figure 4(F), Figure 4(G)). Conversely, interactions between these cells via C3-C3AR1, PDGFB-PDGFRA, TNF-TNFRSF1B, and NAMPT-integrin were decreased in the TR group (Figure 4(F), Figure 4(G)). These data suggest that TR-derived myeloid cells are integral to the pro-inflammatory chemokine-chemokine receptor interaction pathway, being attracted and activated by these chemokines and further producing them to recruit more T/NK cells and influence VIC functions.

The predominant lymphocyte populations in the tricuspid valves were CREM⁺ CD4⁺ helper T (Th) cells and GZMB^loGZMA^hi CD8⁺ T cells (Figures 5(A)-(C)). In comparison to the control TV group, T cells from the TR group exhibited minimal upregulation of genes, but there was a significant downregulation in the expression of multiple HSP genes, suggesting impaired proteostasis within the resident T cells (Figure 5(D)). Additionally, we identified an FCGR3A^loGZMA⁺ NK cluster that expressed high levels of chemokines CCL4 and XCL1, as well as the reparative molecule AREG, but low levels of cytotoxicity-related molecules such as GZMB and PRF1 (Figure 5(A), Figure 5(B)). Notably, these NK cells in the TR group showed increased expression of pro-inflammatory molecules CCL3, CCL4, and IFITM2, and decreased expression of the reparative molecule AREG compared to those in the control TV group (Figure 5(D)).

We observed a decrease in VIC-T/NK communication via the MDK/PTN-NCL/integrin pathway and in T/NK-myeloid cell crosstalk via the anti-inflammatory ANXA1/FPR1 [67] axis in the TR group (Figure 5(E)). Conversely, TR samples demonstrated increased two-way communication between T/NK cells and myeloid cells via the pro-inflammatory CCL3/4 and CCR5/CCR1 pairs (Figure 5(F)). T/NK cells also contributed higher levels of CCL4 to VICs in TR samples (Figure 5(F)). The ligand-receptor pair CXCL12 and CXCR4, which mediate communication between VECs and T cells, was up-regulated in TR samples (Figure 5(F)). These results indicate that in TR, chemokines such as CCL3 and CCL4 recruit and regulate T/NK cells and macrophages, contributing to a pro-inflammatory environment that affects the differentiation and function of VICs. This also highlights close communication between GZMA⁺ T/NK cells and activated macrophages within the context of tricuspid valves in the TR group.

Mast cells serve as sentinels in the heart, swiftly responding to metabolic and immune

Figure 5. Comparison of the transcriptome features and cell interactions of T/NK and mast cells between TV and TR groups.

changes within their microenvironment. They have been reported to colocalize with macrophages, dendritic cells, and lymphocytes in stenotic human aortic valves [68]. In our study of tricuspid valves, we identified three distinct mast cell clusters, including FAU⁺CD69⁺ resident mast cells (M1), CCL2⁺IFITM3⁺ pro-inflammatory mast cells (M2), and CCN1⁺ITGB1⁺ profibrotic mast cells (M3) (Figure 5(G)). Given the small abundance of mast cells and the absence of significant differences in the distribution of mast cell clusters between the two groups, we analyzed the transcriptome of these cells as a single population. Mast cells in the TR group exhibited significantly higher expression of fibrosis-associated genes and lower expression of NF-κB signaling genes compared to those in the control TV group (Figure 5(H)). We also observed that mast cells, which receive MDK and PTN from VICs and send various soluble factors to VICs, VECs, and immune cells, showed decreased activity in the TR samples (Figure 5(I)). Conversely, TR-derived mast cells received substantially increased levels of pro-inflammatory chemokines CCL3 and CCL4 from myeloid cells, and INHBA from myofibroblast (Figure 5(J)). These findings suggest that in TR, mast cells become dysregulated and adopt a profibrotic phenotype, likely due to the elevated levels of pro-inflammatory chemokines produced by activated myeloid cells.

We next assessed the differences between TR and CAVD. Compared to VICs derived from TR, those derived from CAVD exhibited lower expression of genes related to ECM organization, TGFβ response, and cell-substrate adhesion, but higher expression of genes associated with myeloid leukocyte migration, ossification, oxidative stress response, and IL-1 cellular response (Figure 6(A)). Similarly, CAVD-derived VECs demonstrated significantly higher expression of genes involved in NF-κB signal transduction, TNF response, lipopolysaccharide (LPS) response, and myeloid leukocyte migration (Figure 6(B)). Notably, VICs and VECs from AV control samples exhibited up-regulated NF-κB signaling and responses to LPS and TNF compared to TV controls (Figures S4(A)-(E)). As shown in Figure S4, the unique cluster of VDSC in aortic valves also showed increased expression of genes associated with these pathways (Figures S4(C)-(E)). Immune cells in CAVD further exhibited increased LPS response and migration compared to those in TR (Figure 6(C), Figure 6(D)).

Upon comparing cellular communication between the CAVD and TR groups, we found that, similar to TR samples, CAVD samples exhibited decreased crosstalk strength compared to AV controls (Figure S5(A)). In contrast to TR samples, VDSCs, rather than VICs, accounted for the majority of interactions among all cell types in CAVD (Figure S5(B)). Compared to TR samples, CAVD samples displayed increased expression of the pro-inflammatory chemokine CCL2 in VECs and VDSCs and enhanced interaction via CCL2-CCR2 and NAMPT-integrin among all cell types, including VDSCs (Figure 6(E), Figure 6(F), FigureS5(C)). The interaction mediated by MIF and its receptor between myeloid cells and other cells, such as VICs, VECs, T/NK cells, was also up-regulated (Figure S5(D), Figure S5(E)). Conversely, downregulated communication in CAVD samples included anti-inflammatory pairs such as MDK, PTN, ANGPTL4, and their receptors within VICs, and between VECs and VICs (Figure 6(E), Figure 6(F)). Additionally, interactions such as VEGFB-FLT1 between VICs and VECs, GZMA-F2R between T/NK and VECs or VICs, and CXCL12-CXCR4 between VICs and immune

Figure 6. Comparison of the transcriptome features and intercellular communications between tricuspid valves obtained from TR and aortic valves obtained from CAVD.

cells were decreased (Figure 6(E), Figure 6(F), Figure S5(D), Figure S5(E)). These results collectively suggest that aortic valves, relative to tricuspid valves, are more susceptible to bacteria-induced inflammation, characterized by TNF response, NF-κB activation, and CCL2- and nicotinamide adenine dinucleotide (NAD⁺)-mediated cellular communication. This highlights the distinct inflammatory signatures and cellular interactions between tricuspid valves and aortic valves, which may underlie their differential resistance to calcification and inflammation.

The scRNA-seq data in FASTQ files were processed by Cell Ranger software (Version 6.1.2, 10× genomics). The sequencing reads were demultiplexed, mapped to the GRCh38 human reference genome, and counted by unique molecular identifier (UMI). The UMI count matrix was then analyzed using the Seurat package (v4.4.0) in R software (v4.3.1). The cells expressing hemoglobin genes (HBA1, HBA2, HBB, HBD, HBE1, HBG1, HBG2, HBM, HBQ1, HBZ) were removed from the data set, as they likely represented erythrocytes. Only the cells with 200~6000 detected genes and <10% mitochondrial UMIs were considered valid cells and used for downstream analysis. After the quality control, there were 91,018 cells and 26,179 genes for downstream analysis.

After normalizing and scaling, the “RunPCA” function in Seurat was used for Dimension reduction. The R package “Harmony” was used to remove batch effects. The “RunUMAP” function was performed to visualize each cell. The differentially expressed genes (DEGs) of the cluster were calculated by the “FindAllMarkers” function. Cell types were annotated based on differentially expressed genes (DEGs) of each subcluster.

The Seurat function CellCycleScoring was used to predict the cell cycle stage of each individual cell in a given cluster. Previously well-defined S and G2/M phase marker genes were used to calculate the S score and G2/M score, respectively.

The DEGs between Ctrl and TR groups were calculated by the “FindMarkers” function. Gene Set Enrichment Analysis (GSEA) was performed using the R package cluster-Profiler (Version 3.18.1) based on the hallmark gene sets or curated gene sets described in the MSigDB database. Gene Ontology (GO) Analysis was performed using the R package cluster-Profiler (Version 3.18.1) based on the DEGs. Statistical note: The “FindMarkers” function used a Wilcoxon rank-sum test at the single-cell level, treating each cell as an independent observation. This approach does not explicitly account for donor-to-donor biological variation, which may lead to inflated statistical significance.

Cellchat (v1.6.1) was applied for ligand-receptor analysis. The raw counts and cell type annotation for each cell were imputed into CellChat to determine the potential ligand-receptor pairs. Only receptors and ligands whose expression was detected in more than 10 cells were included in this analysis. Pairs with a P-value > 0.05 were filtered out from further analysis. The number and weight of intercellular communication in fibroblasts and total cells were compared between Ctrl and TR groups, as well as the cellular communication among AV, CAVD, Ctrl, and TR groups. The “rankNet” function was used to calculate the comparison of pathway information flow based on the Wilcoxon test. The “identifyOverExpressedGenes” function was used to identify overexpressed signaling genes associated with each cell group. Upregulation was defined as log₂-transformed expression > 0.2 for ligand genes, while downregulation was defined as log₂-transformed expression < −0.1 for ligand genes and <−0.1 for receptor genes.

The SCENIC (v1.3.1) package was used for transcription factor activity analysis. The co-expression modules were based on cisTarget databases (hg38_refseq-r80_10kb_up_and_down_tss, hg38_refseq-r80_500bp_up_and_100bp_down_tss). A total of 6832 fibroblasts and myofibroblasts (10% of the total 68320 fibroblasts and myofibroblasts) were randomly selected and used as input at the network inference step. The AUCell scores were then calculated using all 68320 fibroblasts and myofibroblasts. For other cell types, all cells were used as input at the network inference step and to calculate AUCell scores.

The STRING analysis (v12.0) was applied to study protein-protein-interaction (PPI) between DEGs. Initially, we identified the top 50 DEGs in fibroblasts, endothelial cells, and macrophages using the “FindMarkers” function mentioned earlier. These DEGs were up-regulated or down-regulated in the TR group compared to the Ctrl group. The STRING analysis provided high-confidence PPI interactions based on the neighborhood, gene fusion, co-occurrence, co-expression, experiments, text-mining, and so on. In this study, only interactions for homo sapiens were considered, and a confidence score > 0.4 was requested.

Monocle2 (v2.28.0) was used for constructing pseudo-time trajectories. Fifty thousand cells from fibroblasts were selected according to the proportion of each cluster for pseudotime trajectory construction. The cell information was extracted, and an R object was constructed using the “newCellDataSet” function with lowerDetectionLimit = 0.5 and the expressionFamily parameter set as “negbinomial.size ()”. Then, the size factor and dispersion were estimated using the “estimateSizeFactors” and “estimateDispersions” functions, and low-quality cells with an average expression level > 0.1 and genes expressed in at least 10 cells were filtered out. The top 400 genes were selected as sorting genes based on q-value using the dpFeature method. Finally, the dimensionality reduction and construction of the pseudo-time trajectories were completed.

R (version 4.3.1) was used for the statistical analysis. The Wilcoxon Rank Sum test by the Seurat (version 4.4.0) FindAllMarkers function was used to identify differentially expressed genes (DEGs) between the cell clusters. The cell ratios of each type were compared using the Wilcoxon rank sum test. A P-value of <0.05 was considered statistically significant.

The TR scRNA-seq datasets were downloaded from the Genome Sequence Archive (Genomics, Proteomics & Bioinformatics 2021) in the National Genomics Data Center, China National Center for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Sciences, under accession code HRA010091 at https://ngdc.cncb.ac.cn/gsa-human. The CAVD scRNA-seq datasets were downloaded from PRJNA562645 at https://ncbi.nlm.nih.gov/bioproject. A total of 16 samples were included (TR: n = 5, control: n = 5; CAVD: n = 4, control: n = 2). Datasets were processed separately and integrated using Harmony to correct for batch effects across samples while preserving biological differences. All analyses were performed using R (v4.3.1). The processed single-cell RNA sequencing data, including expression matrices, cell annotations, and metadata, have been deposited in the Zenodo repository and are publicly available at https://doi.org/10.5281/zenodo.19568228. Patient information is recorded in Supplementary Table. The aortic valve data were obtained from a public dataset comprising 6 individuals (2 healthy controls and 4 CAVD patients), and healthy control aortic valve tissues were harvested from patients undergoing aortic valve replacement during repair of aortic dissection. The tricuspid valve data were obtained from another public dataset comprising 10 individuals (5 controls and 5 with moderate-to-severe functional regurgitation), and control tricuspid valves were collected from transplanted hearts with normal tricuspid structure and function.