The exploration to identify key molecules in the interaction between endothelial cells and mono-macrophages in atherosclerosis based on the single cell transcriptome Authors: Boyu Li, Ruiyang Zhu, Na Gao, Siwei Li, Na Yang, Hairong Wang, Hong Yu, Zhibing Lu, Augustin Gaétan Julien Sègbo, Fang Zheng Published: 2026 ABSTRACT AbstractCoronary artery disease (CAD) is the leading cause of global cardiovascular deaths, and atherosclerosis (AS) is the characteristic pathological change in CAD. The interactions between monocytes/macrophages and endothelial cells play an essential role in the formation and progression of atherosclerotic plaques. In this study, we used published single-cell RNA sequencing datasets of calcified atherosclerotic core (AC) plaques and patient-matched proximal adjacent (PA) carotid artery tissue to characterize the single-cell profiles of carotid artery cells. We specifically analyzed endothelial cells (ECs) and monocyte-macrophage (MM) populations to identify cell subpopulations and signaling pathways associated with AS. By comparing cell–cell communication (CCC) between AC and PA, we identified signaling pathways and corresponding ligand-receptor pairs significantly altered in AS. For preliminary validation, Transwell co-culture experiments were conducted for 24 h with human umbilical vein endothelial cells (HUVECs) and human aortic endothelial cells (HAECs) co-cultured with M0, M1, or M2 macrophages. Through integrative analyses, we identified 15 cell clusters and 6 cell subpopulations closely related to AS. Among the altered pathways, MK (Midkine), GALECTIN (Galactose lectin), and SPP1 (Secreted phosphoprotein 1) showed significant changes in AC, with the MK pathway being the most prominent. Notably, the MDK–NCL pair was upregulated in the AS-associated EC and MM subpopulations. In vitro assays confirmed that MDK and NCL expression levels were significantly higher in the ECs and M1 macrophage co-culture group than in the M0 or M2 control groups. In conclusion, this work characterizes AS-related cell heterogeneity and identifies potential key signaling pathways and interactions, providing candidate targets for the prevention and treatment of CAD.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12872-026-06031-0.Keywords: Atherosclerosis, Single-cell sequencing, Heterogeneity, MK pathway Introduction Atherosclerosis (AS) is a chronic inflammatory disease and a significant cause of coronary artery disease (CAD) and atherosclerotic cardiovascular and cerebrovascular diseases (ASCVDs) [1, 2]. The pathological process of AS involves lipid buildup and chronic inflammation in the arterial wall [3]. Many key signaling pathways mediated by immune and inflammatory mediators have been identified [4], and these pathological processes are closely linked to interactions among endothelial cells (ECs), smooth muscle cells, and immune cells [5]. ECs, monocytes, macrophages, and macrophage-derived foam cells are found predominantly in the arterial intima. ECs form a continuous monolayer barrier, which involves their immunomodulatory effects on leukocyte recruitment [6]. Macrophages are the primary immune cells in atherosclerotic plaques [7] and are mainly derived from circulating monocytes in the blood [8]. Physiologically, interactions between monocytes and ECs are involved in angiogenesis and vasculogenesis [9]. Monocytes produce platelet-derived growth factor, which induces mitogenic activity in ECs [10]. ECs can selectively recruit monocytes to the endothelium after activation by inflammatory factors, a process that primarily involves monocyte rolling, adhesion, activation, and migration [11]. Thus, the interaction between endothelial cells and monocytes/macrophages plays a vital role in the development of AS. Single-cell RNA sequencing (scRNA-seq) is a powerful research tool that allows the construction of single-cell profiles in tissues, organs, and organisms [12], thus helping to unravel the underlying complex immune mechanisms of human AS, identifying specific pathogenic cell groups and searching for novel ligand-receptor interactions [13], thereby advancing molecularly targeted drug development [14]. Although the single-cell atlas of atherosclerotic plaques has been investigated [15], changes in cell–cell communication (CCC) among different cell groups, especially between ECs and MMs, remain unclear. With the advent of scRNA-seq technology, various computational tools have been developed to explore CCC and have attracted widespread interest [16]. Among them, Cellchat [17], which uses advanced statistical methods to identify CCC from scRNA-seq data, can simultaneously analyze multiple signaling pathways, which is a significant advantage. We speculated about altered signaling pathways and corresponding changes in ligand-receptor expression in ECs and MMs to reveal their heterogeneity in AS. We preliminarily validated the interactions among the obtained cell subpopulations using wet experiments, thereby identifying potential AS prevention and treatment targets. Results scRNA-Seq of human carotid artery To understand changes in CCC across different cell subpopulations in human AS plaques, the GSE159677 dataset from GEO was analyzed using bioinformatics methods. We obtained 45,091 cells through data preprocessing and stringent quality control. Among which, 13,200 cells were from patient 1, 18,627 from patient 2, and 13,264 from patient 3. After data normalization, clustering, and dimensionality reduction, 40 clusters were identified using a UMAP plot (Figure S1a). Based on established lineage-specific marker genes (Table S2), we assigned a cell type to each cluster and merged clusters with similar gene expression profiles. Finally, we identified a total of 15 cell groups (Fig. 1a), including CD4+ T lymphocyte (CD4+ T), CD8+ T lymphocyte (CD8+ T), Natural killer T cells (NKT), Innate lymphoid cells (ILC), Lymphoid progenitor, B lymphocyte (B), Plasma cells, Neutrophils, Mono-Mac (MM), Conventional DCs (cDC), Plasmacytoid DCs (pDC), Endothelial cells (EC), Adventitial fibroblasts (AF), Vascular smooth muscle cells (VSMC) and Mast cells. Transcriptomic profile of carotid artery from 3 patients: a UMAP plot of GSE159677 dataset by cell type; b UMAP plot of GSE159677 dataset by sample; c Dot plot showing average expressions of classical markers in each cell type; d Line chart showing tissue distributions for each cell type estimated by Ro/e score; e Expressions of classical markers in ECs (up) and MMs (bottom) in each cell type The proportions of cell groups in the patient-derived AS tissues varied, indicating individual differences (Fig. 1b). We identified 2–3 conserved marker genes for each cell type (Fig. 1c), whereas classical marker genes for ECs and MMs were highly expressed in their respective cell groups (Fig. 1e). We explored the relative expression of the 15 marker genes across all cells (Figure S2). These results confirmed the correct cell types. We identified the top 5 marker genes in each cell group relative to the others (sorted by P value) and visualized them with heatmaps (Figure S1c). In addition, the tissue distribution of major cell groups was assessed using the ratio of observed to expected cell numbers (Ro/e) analysis. Among these cell groups, MMs were mainly concentrated in AC, whereas ECs were primarily concentrated in PA (Fig. 1d). The heterogeneity of ECs To explore differences in ECs between the AC and PA groups, we analyzed EC DEGs and performed KEGG-based functional enrichment in AC and PA. The results showed that there were 514 differential genes between AC and PA, including 198 up-regulated genes and 316 down-regulated genes (Fig. 2a), and several signaling pathways were associated with AS (Fig. 2b), including protein processing-related, extracellular matrix (ECM) receptor-related, focal adhesion-related, and PI3K-Akt signaling pathway. The heterogeneity of ECs: a Volcano plot of DEGs in AC and PA; b KEGG analysis of DEGs; c UMAP plot of 4 EC subpopulations; d UMAP plot of AC and PA in ECs; e Line chart showing tissue distributions for each cell subpopulation estimated by Ro/e score; f Dot plot showing the average expression of classical markers in EC1 to EC4; g Heatmap of the top 10 genes of EC1 to EC4 To further analyze the functional differences of ECs and to identify cell subpopulations that may promote AS plaque formation, we extracted and re-clustered the ECs, which yielded 4 EC subpopulations (Fig. 2c) after obtaining 10 subclusters (Figure S3a), namely EC1, EC2, EC3, and EC4. To characterize subpopulations of ECs, marker genes for each subpopulation were identified using heat maps (Fig. 2g), visualized with dot plots (Fig. 2f), and listed in Table S3. The anatomical location of each cell type was then compared (Fig. 2d). The tissue enrichment of the cell subpopulations was quantified by Ro/e analysis (Fig. 2e). We found that among these cell subpopulations, EC1 was almost exclusively distributed in PA, EC4 was almost exclusively distributed in AC, whereas EC2 and EC3 were preferentially distributed in AC. The tissue distribution of EC subpopulations was associated with differences in gene expression. KEGG analysis of each subgroup is shown in Fig. 3a-d. Among them, we inferred that the biological functions of EC1 were mainly associated with autoimmune diseases such as SLE and rheumatoid arthritis, as well as HLA-related pathways, whereas EC2, EC3, and EC4 were all related to inflammatory pathways. Among them, EC2 was enriched in pathways related to HIF-1, cell adhesion, ECM, shear stress, and AGE-RAGE, and EC3 was enriched in inflammatory pathways related to Toll-like receptor, mTOR, NF-κB, cytokine receptors, and chemokine signaling. EC4 was associated with glycolysis and inflammatory pathways related to HIF-1, Rap1, PI3K-Akt, and MAPK. KEGG analyses of EC subpopulations of EC1 a; EC2 b; EC3 c and EC4 d; e Up-regulated genes in AC of EC2; f Up-regulated genes in AC of EC3 Based on these results, the differential gene enrichment of the EC2 and EC3 subpopulations in AC and PA was further analyzed, respectively (Fig. 3e-f). We found that genes related to inflammatory pathways in EC2 and to cytokines and the ECM in EC3 were highly expressed in the AC group. Since EC4 was almost exclusively present in AC, it can be assumed that the EC4 subpopulation was involved in AS development and may be more closely related to AS. Combining the results of tissue distribution and KEGG analysis, we inferred that EC2, EC3, and EC4 may be more closely associated with the development of AS. The heterogeneity of MMs To explore differences in MMs between AC and PA, we used KEGG analysis to identify differentially expressed genes in AC and PA (Fig. 4a). The results showed 199 differentially expressed genes between AC and PA, including 80 upregulated and 119 downregulated genes. The results indicated several signaling pathways associated with AS (Fig. 4b), including glycolysis, cholesterol metabolism, inflammation, and other metabolic pathways. The heterogeneity of MMs: a Volcano plot of DEGs in AC and PA; b KEGG analysis of DEGs; c UMAP plot of 6 MM subpopulations; d UMAP plot of AC and PA in MMs; e, f Line chart showing tissue distributions for each cell subpopulation estimated by Ro/e score; g Dot plot showing average expression of classical markers in each cell subpopulation; h Heatmap of the top 10 genes of MM1 to MM6 To further analyze the functional differences in MMs and identify subpopulations that might promote AS plaque formation, the MMs group was extracted individually and re-clustered, yielding six subpopulations (Fig. 4c) from eight subclusters (Figure S3b). Similarly, marker genes for each cell subpopulation were identified using a heatmap (Fig. 4h), visualized with dot plots (Fig. 4g), and listed in Table S4. The functional enrichment analysis of each subpopulation is shown in Fig. 5. Among them, the biological functions of MM1 were similar to those of EC1, which were mainly focused on pathways related to transplant rejection and autoimmune diseases, MM2, MM4, and MM5 were primarily enriched in pathways related to inflammation and cytokines, MM3 was enriched in pathways related to apoptosis, glucose and lipid metabolism, and inflammation, whereas MM6 was enriched in pathways related to inflammation, cell adhesion, and ECM receptors. By comparing the anatomical location of each cell subpopulation (Fig. 4d) and quantifying tissue distribution using Ro/e analysis (Fig. 4e-f), MM2 and MM5 were predominantly distributed in the PA. In contrast, the remaining subpopulations were preferentially distributed in AC, and MM3 was almost exclusively present in AC. KEGG analyses MM subpopulations of MM1 a; MM2 b; MM3 c; MM4 d; MM5 e and MM6 f Several subtypes of monocytes and macrophages have been reported, each with a unique function. Human monocytes are generally classified into three subtypes: classical (CD14+ CD16−) monocytes, intermediate (CD14+ CD16+) monocytes, and non-classical (CD14dim CD16+) monocytes [18]. Macrophages can be divided into two subtypes according to their functions: classically activated macrophages, or M1, and alternatively activated macrophages, or M2. M1 macrophages are associated with a pro-inflammatory phenotype, whereas M2 macrophages are associated with an anti-inflammatory phenotype [19, 20]. Therefore, we proposed exploring the relationship between the previously obtained MM1 to MM6 and the known traditional monocyte-macrophage types. First, we investigated the joint density distributions of 2 classical markers for each conventional monocyte-macrophage type using the Nebulosa package (Fig. 6). We inferred the relationship between MM subpopulations and known monocyte-macrophage types by comparing Fig. 6a, b, d, e, f with Fig. 6c. Among them, classical monocytes mainly overlapped with MM1 and MM2 distributions; intermediate monocytes and non-classical monocytes mainly overlapped with MM1 distributions; M1 macrophages mainly overlapped with MM3; and M2 macrophages mainly overlapped with MM5. Moreover, MM2, MM3, and MM5 correlated well with conventional monocyte-macrophage types, but the remaining three subpopulations did not. Therefore, we used violin plots to analyze the expression levels of cell markers for each MM subpopulation (Fig. 7) and combined the two methods to explore the potential functions of each subpopulation. We found that MM1 expressed three markers of 3 conventional monocyte types. In contrast, classical monocytes may be mainly concentrated in MM2. For macrophages, MM3 favored M1 macrophages, while MM5 favored M2 macrophages. Therefore, based on the results of the heterogeneity analysis and the literature, MM3, MM4, and MM6 may promote the development of AS. Joint density distribution of classical markers of conventional types: a Classical monocyte; b Intermediate monocyte; c UMAP plot of MM subpopulations; d Non-classical monocyte; e M1 macrophage; f M2 macrophage The expression of classical markers of monocytes and macrophages in MM subpopulations of Classical monocyte a; Intermediate monocyte b; Non-classical monocyte c; M1 macrophage d; and M2 macrophage e Cell–cell communication in all cell subpopulations The CCC patterns were analyzed using the CellChat package. These subpopulations included 4 EC and 6 MM subpopulations. The analysis showed that AC's cell interactions were much higher than in PA. Still, the intensity was only slightly higher than that of PA (Fig. 8a). Both incoming and outgoing signals of MM3, MM4, MM6, and EC4 were enhanced in AC compared with PA. To identify the pathways altered in AC, we identified the signaling networks with the most significant differences based on functional similarity. We calculated and visualized the distance of overlapped signaling pathways between AC and PA. We found that molecules such as CALCR, CXCL, and MK showed substantial differences in functional similarity, indicating that these signaling pathways exhibited more pronounced differences in the expression levels of their ligands and receptors (Fig. 8d). In addition, we compared the overall information flow of signaling pathways between PA and AC. The results showed that several signals, such as MK, CALCR, and GALECTIN, were more potent in AC than in PA. Seven signals, such as APELIN, ANGPT(Angiopoietin), and SPP1 (Secreted phosphoprotein 1), were unique to the AC (Fig. 8b). For further refinement, the differences in the numbers or intensities of incoming and outgoing signals among cell subpopulations were compared (Fig. 8c). The results showed a significant increase in signaling from MM3 to MM1, MM2, MM4, and MM6, and a slight rise in signaling to EC2, EC3, and EC4. In addition, the outgoing signals from EC4, MM4, and MM6 were also enhanced. In comparison, the incoming and outgoing signals of EC1 and MM5 were weakened. This further suggested the critical roles of EC2, EC3, EC4, MM3, MM4, and MM6 in the development of AS. Comparison of cell–cell communication in PA and AC: a Number and strength of interactions; b The overall information flow of signaling pathways; c Heatmap of differential interaction strength; d The pathway distance in the learned joint manifold The altered signals in AS plaque To further analyze changes in signaling across cell subpopulations, we compared the strength of each signaling pathway between PA and AC (Fig. 9a). Multiple signaling pathways involving EC2 and EC3 were enhanced in AC, including CXCL and VEGF. Multiple MM2 signals, such as IL-1 and OSM, were attenuated in AC. MM3, MM4, MM6, and EC4 signaling pathways were almost exclusively present in AC. The ANGPT, ANGPTL (Angiopoietin-Like Proteins), APELIN, and GDF signals were only from EC4 in AC (Figure S4a). ANGPT is a second endothelium-specific ligand-receptor signaling system that regulates postnatal angiogenesis, vascular permeability, remodeling, and inflammation to maintain vascular homeostasis. Whereas secreted ANGPT/ANGPTL proteins play multiple roles in angiogenesis, inflammation, lipid and glucose metabolism, hematopoietic stem cell activity, and cancer, 7 of 11 ANGPT/ANGPTL genes were linked to AS [21]. APELIN is highly expressed in vascular endothelial cells, induces EC proliferation and migration, stimulates angiogenesis, and possesses a broad spectrum of anti-inflammatory activity that can modulate various diseases associated with inflammation and oxidative stress [22]. In our study, this signal was sent from EC4 and only to EC3 (Figure S4a, b). The GDF-15 is a stress-responsive cytokine that inhibits endothelial cell proliferation both in vitro and in vivo. Patients with cardiovascular disease have high overall GDF-15 levels and an increased risk of AS [23]. Multiple signals in EC1 and MM5 were present only in PA, such as TGF-β and IGF. TGF-β is only sent to MM5 (Figure S4b), but this signal has both pro-inflammatory [24] and anti-inflammatory roles, and its role in AS remains unclear [25]. IGF-1 is associated with cell growth, differentiation, and anabolism, and macrophage-specific IGF-1 overexpression reduced CXCL12 chemokine levels in ApoE-deficient mice, thereby inhibiting AS development [26]. Figure 9b summarizes the major signaling pathways involved in the interactions between key cell subpopulations. The altered signals in AS plaques: a The overall signal strength of each signaling pathway in cell subpopulations; b The main signaling pathways between key cell subpopulations are represented by the arrow color, which means the type of signaling pathway, and the tip, which represents the direction of the signaling pathway Based on the results of heterogeneity analysis and a preliminary inter-subpopulation communication analysis, we predicted that the EC subpopulations of EC2, EC3, and EC4, and the MM subpopulations of MM3, MM4, and MM6, may be closely related to the development of AS. Therefore, we focused on comparing the outgoing signals from EC2, EC3, and EC4 and the incoming signals to MM3, MM4, and MM6 between PA and AC (Fig. 10a). Among them, signaling pathways such as GALECTIN, MK, and SPP1 changed more than others in the above cell subpopulations. The altered signals in AS plaques: a Changes of outgoing signaling from ECs and incoming signaling to MMs; b Circle plot of GALECTIN and MK pathway in PA and AC; c Communication probabilities of ligand-receptor pairs of signaling pathways sent from ECs to MMs In the GALECTIN signaling pathway, Gal-9 is involved in immunomodulation and cell differentiation, regulating multiple inflammatory mechanisms. Although Gal-9 may have a protective effect against AS [27], new findings suggest that Gal-9 can induce the release of inflammatory mediators in macrophages and activate monocytes, thereby promoting AS [28]. The changes in signaling pathways among cell subpopulations were illustrated by network diagrams (Fig. 10b), which revealed that the GALECTIN signaling pathway in AC was significantly increased from EC3 to MM subpopulations and from MM3 to other MM subpopulations, suggesting a possible pro-inflammatory role. The MDK can regulate the expression of pro-inflammatory cytokines, which play a vital role in chronic inflammatory diseases [29]. In AC, the MK pathway is predominantly sent to the MMs from EC2 and EC4 (Fig. 10b). In SPP1 + macrophages, lipids and cholesterol accumulated abnormally, possibly due to an imbalance in cholesterol metabolism, which in turn induced endoplasmic reticulum stress, leading to the formation of necrotic cores and subsequent cell death, thereby exacerbating the progression of AS [30]. In the circle plot, the SPP1 pathway was mainly sent from MM3 (Fig. 9b), predominantly M1 macrophages, consistent with previous results. Subsequently, we analyzed the communication probabilities of ligand-receptor pairs in signaling pathways from ECs to MMs. The results showed that several ligand-receptor pairs of the MK pathway were upregulated in AC, and the MDK-NCL pair was significantly increased (Fig. 10c). However, we found that differential expression of the MDK-NCL pair was observed primarily in patients 1 and 2 (Figure S5), indicating possible patient-to-patient variability. The expression of MDK-NCL in donor and recipient cells To further compare differences in the MK signaling pathway between PA and AC, we analyzed the gene expression of MK ligands and receptors in each subpopulation (Fig. 11a). In AC, the ligand MDK was highly expressed in EC2 and EC4 compared with PA. In contrast, the receptor NCL was significantly higher in MM4 and MM6 in PC, and possibly slightly higher in MM3. The expression of ligand-receptor pairs in ECs and MMs, respectively: a Violin plots of the expression levels of ligand-receptor pairs of MK in ECs and MMs subpopulations. Cell subpopulations closely associated with AS were labeled red at the top of the graph. In the violin plot, the wider the graph was in the horizontal direction, the higher the probability that the expression level of the gene was distributed here; b The expression levels of macrophage markers. Left: M1, right: M2; c The expression of MDK (left) and NCL (right) of MK signaling in Transwell co-culture system (n = 3 independent biological replicates). *P <.05, **P <.01, ***P <.001. Data are presented as the mean ± SD. Unpaired t-test was used for comparison in b. One-way ANOVA followed by Tukey’s post-hoc test was used for comparison c Based on these results, a co-culture system of endothelial cells and M0/M1/M2 macrophages was constructed for simulating cell–cell interactions (CCIs) under AS in vivo, and macrophage markers were detected. The results showed that the cell model was successfully constructed, as M1 and M2 macrophages highly expressed the corresponding polarization markers (Fig. 11b). Human umbilical vein endothelial cells (HUVECs) and human aortic endothelial cells (HAECs) were co-cultured with M1 and M2 macrophages using Transwell chambers with M0 macrophages as a control. Macrophages were plated in the lower chamber, and endothelial cells in the upper chamber. After 24 h of co-culture, M0/M1/M2 macrophages and endothelial cells were collected, and RNA was extracted for fluorescent quantitative PCR. The MDK and NCL expression levels were detected in endothelial cells and macrophages, respectively. We found that MDK in endothelial cells and NCL in macrophages were significantly higher in the co-culture system of endothelial cells and M1 macrophages than in the single group of endothelial cells and M0/M2 (Fig. 11c). These results suggested that the interaction of MDK-NCL pairs was enhanced between AS inflammatory macrophages and inflammatory endothelial cells. Discussion During AS progression, circulating monocytes are continuously recruited to the lesion, where they differentiate into macrophages and take up cholesterol, eventually forming foam cells [31, 32]. Endothelial cells play an essential role in monocyte migration, and activated endothelial cells secrete adhesion molecules, chemokines, and other inflammatory mediators to promote monocyte recruitment [33]. Our study used scRNA-seq data of carotid AS samples to generate a single-cell transcriptome profile containing 15 cell populations, including endothelial cells, fibroblasts, macrophages, and smooth muscle cells. These cell populations included both immune and non-immune (Fig. 12). We focused on the heterogeneity of ECs and MMs to identify their subpopulations and analyze changes in signaling pathways between them. Immune and non-immune cell groups in a single-cell transcriptome profile The ECs group was mainly distributed in PA. We categorized the ECs into four subpopulations, and the MM cell group was predominantly distributed in AC and was categorized into six subpopulations. The profiles of MM subpopulations are summarized in Table S5. The EC subpopulations EC2 and EC3, as well as the MM subpopulations MM3, MM4, and MM6, were more strongly associated with the development of AS. There were stark differences between PA and AC in MK, GALECTIN, and SPP1. In analyzing the communication probability of ligand-receptor pairs, the MDK-NCL pair showed a significant increase in AC, and we preliminarily verified higher expression levels of MDK and NCL in the co-culture group of HUVECs/HAECs and M1 macrophages. MDK is a heparin-binding growth factor that binds to LDL receptor-related proteins, as well as to integrins α4β1 and α6β1 [34], and is involved in a variety of biological processes, including angiogenesis and inflammatory cell migration. Some mouse studies have shown that MDK can promote AS plaque formation through pro-inflammatory, angiogenic, and anti-apoptotic effects [35]. Still, the mechanism of its interaction with NCL has been little studied. Under hypoxic conditions, MDK promotes pulmonary vascular remodeling by interacting with surface NCL, prompting the translocation of NCL from the nucleus to the cell surface of pulmonary artery smooth muscle cells (PASMCs) and inducing PASMC proliferation and migration [36]. Thus, we inferred that endothelial cell-expressed MDK is associated with increased NCL on macrophage surfaces, and that MDK-NCL binding may promote macrophage migration and polarization, thereby inducing increased inflammation and plaque formation. Our study compared adventitial and plaque-associated regions within the same patients. Previous scRNA-seq studies of atherosclerosis have typically analyzed plaques as bulk tissue or focused solely on the intimal plaque core. By contrast, our study design allowed us to directly interrogate the transcriptional gradients between a relatively healthy tissue microenvironment and the diseased microenvironment. In addition, our study prioritized specific intercellular communication pathways rather than merely cataloging cell types. While MDK signaling has been implicated in angiogenesis and inflammatory cell migration, its role in atherosclerosis has received little attention. The MDK-NCL axis, in particular, has not been previously described in the context of plaque endothelial-macrophage crosstalk. By applying CellChat to our spatially resolved dataset, we inferred this pathway as a top-ranked signal in the AC region. Finally, combining CellChat predictions with experimental results enhanced the credibility of our findings. In a word, we present the direct comparison of matched AC and PA tissues, identify the MDK–NCL EC–MM signaling axis, and integrate CellChat predictions with experimental results. However, several limitations of this study must be acknowledged. First, our discovery dataset is derived from only three donors; although the use of matched PA samples effectively controls for inter-individual genetic variability, the small cohort size may constrain the statistical power and generalizability of the findings. Second, the comparison with PA samples may overlook potential "field effects"-subclinical inflammatory changes in per-plaque tissues that represent a pre-disease state rather than truly healthy vasculature. Future studies incorporating external healthy control datasets would more rigorously isolate plaque-specific signatures. Third, regarding our experimental validation, we recognize the inherent gaps between in vitro models and the complex biology of plaque. While HUVECs are a standard vascular model, their venous origin is a limitation; thus, we further validated the MDK–NCL axis in HAECs to provide a more physiologically relevant arterial context. Nonetheless, THP-1-derived macrophages and static Transwell systems cannot fully capture the phenotypic complexity or the spatial proximity of primary cells and lipid-loaded foam cells found in vivo. Crucially, while our co-culture experiments demonstrate a clear correlation between MDK and NCL expression, they do not establish a definitive causal hierarchy. Therefore, these foundational findings should be viewed as hypothesis-generating. Future research employing functional perturbation assays, such as siRNA-mediated knockdown in complex co-culture systems or in vivo neutralization studies, is required to confirm the mechanistic role of the MDK-NCL signaling axis as a driver of atherosclerosis. Materials and Methods The scRNA-seq data acquisition, preprocessing, and analysis We downloaded the scRNA-seq dataset of human carotid atherosclerotic plaques from the NCBI GEO data repository (GEO accession no. GSE159677), which contains calcified atherosclerotic core (AC) plaques and patient-matched proximal adjacent (PA) portions of carotid artery tissue from 3 patients undergoing carotid endarterectomy [15]. The raw gene-expression matrix was transformed into a Seurat object using the Seurat package (v4.3.0) of R (v4.3.1) [37]. To exclude low-quality cells, cells with more than 25,000 unique molecular identifiers (UMIs) and mitochondrial count ratios not less than 10% were removed, cells expressing between 200 and 4,000 genes were selected, and genes expressed in at least ten cells were chosen for further analysis, and identical quality control thresholds were applied in AC and PA to ensure comparability. Data normalization, data scaling, and identification of highly variable genes (HGVs) were performed using the sctransform package (v 0.3.5) [38], and the top 3,000 variable genes from all cells were used for principal component analysis (PCA). Subsequently, all cells were subjected to nonlinear dimensionality reduction using the uniform manifold approximation and projection (UMAP) method based on the first 40 principal components estimated from variable genes and elbow plots. Cell clustering was performed using the “FindNeighbor” and “FindClusters” functions in the Seurat package, with the “resolution” parameter set to 1.4 in the first clustering and set to 0.2 and 0.3 for the ECs and MMs cell populations, respectively. The "FindAllMarkers" function with its default parameters was used to list markers for all cell clusters. We manually assigned the cell clusters to cell types based on cell markers reported in the literature. For the subclustering of the ECs and MMs, the same procedures of finding HGVs, dimensionality reduction, and clustering were repeated. Tissue distribution of clusters To understand the differential distribution and enrichment of different cell clusters and subpopulations of ECs and MMs between AC and PA, the ratio of observed cell number over the expected cell number (Ro/e) obtained by the bilateral chi-square test was calculated for each cell cluster in different tissues [39]. If the Ro/e of a cell cluster in a tissue > 1, it suggests that the number of cells in the cluster is higher than the expected cell number in that tissue, and the cell cluster tends to be enriched in that tissue [40]. The Ro/e was calculated according to the following formula [40]: where denotes the number of cells observed in the tissue by the cell cluster, denotes the expected number of cells obtained by the chi-square test. Differential gene and functional-enrichment analysis To explore the functions of different cell types, we used the “FindAllMarkers” function to analyze cell type-specific and subcluster-specific markers for ECs or MMs. Before differential gene analysis, we performed pseudobulk analysis to reduce pseudoreplication, inflated significance, and false positives. The “FindMarkers” function was used to identify differentially expressed genes (DEGs) between specified cell types or subclusters, with parameters set to their defaults. Genes with |logFC|> 0.25 were considered DEGs. All enrichment analyses were performed using the clusterProfiler package (v4.8.3) [41]. P-values were adjusted by the Benjamini–Hochberg method. KEGG terms with a corrected P < 0.05 were considered significantly enriched. Joint density distribution of monocyte/macrophage markers in MMs The Nebulosa package provides an enhanced visualization of single-cell data using kernel density estimation [42]. It aims to recover the signal from dropped-out features by incorporating cell similarity, enabling a “convolution” of cell features. In addition, characterizing cell populations usually relies on multiple markers, whereas the Nebulosa package allows visualization of the joint density of multiple genes in a single plot. In our study, we used the Nebulosa package (v1.0.1) to combine 2 classical markers for each monocyte/macrophage type, then visualized the joint density of the cell markers, thus exploring the potential links between MMs and conventional types of monocytes/macrophages. Cell–cell communication analysis We performed CCC analysis between cell subpopulations of ECs and MMs using CellChat (v1.6.1) with default settings [17]. The CCC analysis of a single dataset was performed first. Two CellChat objects, cellchat_PA, and cellchat_AC, were created by taking the Seurat object as input, setting the species of the built-in ligand-receptor interaction database CellChatDB (v2) to human, and setting CCIs method of CellChatDB to “Secreted Signaling”. The “computeCommunProb” function from CellChat was used to calculate the communication probability. We set the minimum number of cells required in each cell group for cell–cell communication to 10. We then compared the signaling pathways between the PA and AC groups. The cellchat_PA and cellchat_AC objects for which CCC analysis had been performed were merged into a cellchat_all object to compare the total number and strength of CCIs and to identify conserved and context-specific signaling pathways, as well as upregulated and downregulated ligand-receptor pairs. The signaling pathways were visualized using the hierarchy plot, circle plot, or chord diagram, and the communication patterns among target cells were visualized using dot plots. All parameters were set by default. Statistical significance was set at P < 0.05. Cell culture HUVECs, HAECs, and the human monocyte cell line THP-1 were purchased from the China Center for Type Culture Collection (CCTCC). HUVECs and HAECs were cultured in Endothelial Cell Medium (ECM). THP-1 cells were maintained in RPMI-1640 medium (Gibco, USA) supplemented with 10% fetal bovine serum (FBS; Solarbio) and 1% penicillin/streptomycin (P/S; Hyclone, USA). All cells were incubated at 37 °C with 5% CO2, and the culture medium was refreshed daily. THP-1 cells were seeded in 6-well plates (1 × 105 cells/mL). To differentiate THP-1 into M0 macrophages, the cells at the log phase of growth were cultured with 100 ng/mL phorbol myristate acetate (PMA; Solarbio) for 24 h. M0 were differentiated into M1 by adding 100 ng/mL lipopolysaccharide (LPS; Solarbio) and 60 ng/mL interferon-γ (IFN-γ; Solarbio) for 24 h, and M2 cells with 60 ng/mL interleukin-4 (IL-4; PeproTech) for 24 h. Transwell co-culture For the Transwell co-culture experiments, M0, M1, or M2 macrophages (1 × 105 cells/mL, 1 mL) were seeded into the lower chamber (Corning, USA) in RPMI-1640 medium. Simultaneously, HUVECs or HAECs (5 × 104 cells/mL, 200 µL) were seeded into the upper chamber in ECM. The co-culture systems were incubated at 37 ℃ with 5% CO2 for 24 h before further analysis. RNA extraction and qPCR Total RNA from HUVECs, HAECs, THP-1, and M0/M1/M2 cells was extracted using TRIzol reagent (Life Technologies, USA) and reversely transcribed into complementary DNA (cDNA) using PrimeScript RT reagent kit (Takara, Japan). Quantitative polymerase chain reaction (qPCR) was performed on a Bio-Rad CFX96 (BioRad Laboratories, USA) with UltraSYBR mixture (CWBIO, China). We used glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as the endogenous reference gene, and all relative mRNA expression levels were analyzed using the 2−ΔΔCt method [43]. The primers used are listed in Table S1. Statistical analysis For bioinformatical analysis, statistical analyses were performed using the default methods in the Seurat (v4.3.0) and CellChat (v1.6.1) packages by R (v4.3.1). The P-values for gene expression analysis across different groups were estimated using the Wilcoxon Rank Sum test, and a P-value < 0.05 was used to determine significant interactions. All images were generated using R Studio. For wet experiments, all experiments were performed at least three independent biological replicates, and the results are expressed as the means ± SDs. Statistical analyses and illustrations were performed using the GraphPad Prism (v8.0.2). Statistical differences in the expression of macrophage markers between the two groups were analyzed using an unpaired two-tailed Student’s t-test. For comparisons involving three or more groups, ordinary one-way analysis of variance (ANOVA) followed by Tukey’s post-hoc test for multiple comparisons was employed. P-values < 0.05 were considered statistically significant. Statistical significance in figures is indicated as follows: NS, not significant; *P < 0.05; **P < 0.01; ***P < 0.001; ***P < 0.0001. Conclusions Our study provides a comprehensive single-cell transcriptomic profile of human atherosclerotic plaques. By characterizing the heterogeneity of endothelial cells and MMs, we identified several AS-associated cell subpopulations and signaling pathways. Notably, the MK pathway emerged as a potential key signaling axis in AS. Furthermore, the enhanced interaction between MDK and NCL was validated in both HUVECs and HAECs co-culture models, suggesting its involvement in macrophage-endothelial communication and macrophage polarization. These findings provide a valuable foundation for future mechanistic studies and potential therapeutic strategies in AS. Supplementary Information Supplementary Material 1. Acknowledgements Not applicable. Abbreviations Atherosclerotic core Angiopoietin Angiopoietin-Like Proteins Atherosclerosis Atherosclerotic cardiovascular and cerebrovascular diseases Biglycan Calcitonin receptor Coronary artery disease Cell-cell communication Cell-cell interactions CXC chemokine ligand Decorin Endothelial cells Endothelial-to-mesenchymal transition Fibronectin-1 Galactose lectin Growth differentiation factor Human aortic endothelial cells Human umbilical vein endothelial cells Insulin-like growth factor Matrix gla protein Midkine Mono-macs Oncostatin M Proximal adjacent Single-cell RNA sequencing Secreted phosphoprotein 1 Transforming Growth Factor-β Vascular endothelial growth factor Authors’ contributions Author Contributions: Conceptualization, B.L.; methodology, B.L., R.Z. and F.Z.; software, B.L.; validation, B.L. and R.Z.; formal analysis, B.L., R.Z. and N.G.; investigation, B.L. and R.Z.; resources, B.L., R.Z., S.L., N.Y., H.W., H.Y., Z.L., A.S. and F.Z.; data curation, B.L. and R.Z.; writing—original draft preparation, B.L.; writing—review and editing, B.L., R.Z. and F.Z.; visualization, B.L.; supervision, F.Z.; project administration, F.Z.; funding acquisition, F.Z. All authors have read and agreed to the published version of the manuscript. Funding This research was funded by the Noncommunicable Chronic Diseases National Science and Technology Major Project (2024ZD0533000) and National Key Research and Development Program of China (2024YFC3607500). Data availability The scRNA‑seq data were accessed from the Gene Expression Omnibus database under accession number GSE159677 and are available at the following URL: https://identifiers.org/geo:GSE159677. Declarations Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Boyu Li and Ruiyang Zhu contributed equally to this study. References Associated Data Supplementary Materials Supplementary Material 1. Data Availability Statement The scRNA‑seq data were accessed from the Gene Expression Omnibus database under accession number GSE159677 and are available at the following URL: https://identifiers.org/geo:GSE159677.