Eight-week-old mature male Sprague-Dawley (SD, Hsd, Harlan) rats from the Animal Core Facility of Nanjing Medical University (Nanjing, China) were used in our experiment. Rats were housed in a temperature-controlled room with a 12 h light/dark cycle and free access to fresh water and food. All animal procedures were approved by the Experimental Animal Care and Use Committee of Nanjing Medical University.

The rat carotid artery balloon injury model is the most common in vivo model widely used to study ISR. This approach consists of isolating a segment from the right common carotid artery in SD rats under general anesthesia (pentobarbital sodium, 60 mg/kg, i.p.), creating an arteriotomy incision in the external carotid branch followed by inserting a balloon catheter (Fogarty, 12A0602F, 0.67 mm, Edwards Lifesciences) into the common carotid artery, repeated balloon inflation and pulling back five times to imitate percutaneous transluminal coronary angioplasty (PTCA), and finally removal of the catheter with external carotid ligation (10). The rats were sacrificed with an overdose of pentobarbital sodium (200 mg/kg, iv) at 28 days after the procedure, and their left (control group) and right (case group) carotid arteries (Supplementary Figure 1A) were subsequently collected for further study. We used HE staining to identify the success of the model (Supplementary Figure 1B).

To avoid data variation incurred by sex differences, only two male SD rats were selected for the study. To capture single cells, the common carotid arteries were washed with PBS twice and stored in MACS tissue storage solution (Cat#: 130-100-008). Tissues were digested by 0.25% Trypsin-EDTA (1^*, Gibco) and 0.1% collagenase (1^*, Gibco). Single-cell RNA sequencing with the 10x Genomics platform was performed by a commercial service (Shanghai OE Biotech Co., Ltd., China; Figure 1A). Briefly, it uses microfluidic technology to wrap the beads and single cells with Cell Barcodes in droplets, lyses the cells in the droplets to connect the mRNA in the cells to the Cell Barcodes on beads, and finally forms single-cell GEMs. Reverse transcription was performed with the droplets to construct a cDNA library. The sample source of the target sequence is distinguished by the sample index on the library sequence.

We first concentrated the normal carotid artery (control group), which has never been mapped previously. The two whole left carotid arteries were enzymatically digested, and then scRNA-seq libraries were built with the 10x Genomics platform.

A total of 2,445 cells were captured by quality control and visualized in t-SNE dimensionality reduction plots within 11 clusters (Supplementary Table 1, Supplementary Figure 2, Figure 1B). To classify these 11 clusters into known cell types, we filtered out their highly expressed genes and labeled them based on known marker genes (Figure 1C). We also examined recognized cell-type markers, such as VSMCs (Acta2, Tagln, Cald1, Myl9), fibroblasts (Col1a1, Col3a1, Dcn, Pdfra), ECs (Cdh5, Cldn5, Sox17), and macrophages (Cd68, Ncf4) (Figure 1D). The differentially expressed genes of cluster 11 (only 19 cells) were not specific and could not match the existing recognized cell types; thus, we performed our analysis after excluding cluster 11. A new t-SNE dimension reduction plot of 5 cell types showed the different proportions of cell types, including ECs, VSMCs, fibroblasts, macrophages and unknown cells (Figure 1E). Notably, VSMCs and fibroblasts were the major cell types in the normal carotid artery (Figure 1F).

We found that cells in cluster 7 highly expressed marker genes from both VSMCs and fibroblasts at the same time (Figure 1D), but these cells in cluster 7 also had their own specific markers (Adm, Olr1, Thbd, Gfp2, Figures 2A,B). Previous studies have reported that the VSMC phenotype might switch to a fibroblast phenotype in some situations (11, 12). Therefore, we tried to investigate the roles of cluster 7 during the process of VSMC phenotype switching via the current model.

We performed pseudotime analysis on VSMCs, fibroblasts, and unknown cells consisting of clusters 1-7 and 10. According to the degree of differentiation, cells were classified into three trajectories (states 1, 2, and 3) in pseudotime analysis (Figures 2C–F). Obviously, VSMCs were mainly distributed in state 1 and state 2, while fibroblasts were distributed in state 2. Previous evidence demonstrated that VSMCs had two important phenotypes in the vessel wall, including a contractile phenotype and a synthetic phenotype (13, 14). Fully differentiated/contractile VSMCs, responsible for vascular tone regulation, could be transformed into dedifferentiated/synthetic VSMCs to acquire proliferation, migration and synthesis abilities after vascular injury (13). According to the marker gene analysis, VSMCs in state 2 were considered contractile VSMCs, and VSMCs in state 1 could be regarded as synthetic VSMCs. The unknown cells in cluster 7 were scattered in states 2 and 3 by pseudotime analysis, indicating that these cells could play crucial roles in phenotype switching between contractile VSMCs and fibroblasts. Finally, cells in cluster 7 were termed transitional cells due to the co-expression of cell-type markers of VSMCs and fibroblasts. Moreover, we selected and analyzed the top 10 genes of VSMCs, fibroblasts, and transitional cells through pseudotime analysis, as summarized in Figure 2G.

A total of 4,674 cells in the stenotic carotid artery (case group) were acquired after removing unqualified cells from 5,656 cells by quality control. Specific genes from a total of 11 clusters were identified and used to construct a heatmap (Supplementary Figure 2, Figures 3A,B). We classified all of these cells into five cell types according to recognized markers of VSMCs, fibroblasts, ECs and macrophages (Figure 3C). We used the abovementioned strategy to verify the transitional cells and found a similar result. Obviously, the proportion of different cell types changed considerably between the normal artery and stenotic artery. The proportion of fibroblasts, ECs, and macrophages increased in the stenotic artery compared to the normal artery, while the proportion of VSMCs and transitional cells decreased in the stenotic artery (Figures 3D,E).

Pseudotime analysis of VSMCs, fibroblasts and transitional cells (Supplementary Figures 3A,B) showed that these cells could be classified into 5 states: state 1, 2, and 3 fibroblasts; state 4 synthetic VSMCs; and state 5 contractile VSMCs (Supplementary Figures 3C,D, 4). Interestingly, transitional cells in the stenotic artery were distributed in state 4 (synthetic VSMCs), which was different from their co-expression with markers of VSMCs and fibroblasts in the normal artery. This novel finding indicated that contractile VSMCs could switch to both a fibroblast-like phenotype and a synthetic VSMC phenotype in the stenotic artery after balloon injury. Pseudotime analysis and heatmaps of differentially expressed genes in VSMCs, fibroblasts and transitional cells are shown in Supplementary Figures 5–11.

Contractile VSMCs, the major cell type in the artery (Figure 3E), are important for vessel contraction/dilation and regulating hemodynamics (14, 15). Thus, the differences in VSMCs between normal and stenotic arteries were analyzed in depth to explore the mechanism of ISR. VSMCs were isolated and then divided into 6 distinct clusters (Figures 4A,B) based on gene enrichment analysis. From Figures 4C,D, we found that the proportion of VSMCs in these 6 clusters varied greatly, and then several differentially expressed genes were selected to identify these 6 different clusters. Markers of cluster 1 (Hes1, Ccn1), cluster 2 (Notch1, Tmem140), cluster 3 (Rpa3, Ap3s2), cluster 4 (Pycard, Ep400), cluster 5 (Igfbp4, Gpx3), and cluster 6 (Tnfaip6, Mt-cyb) are summarized as dot plots (Figures 4E,F). GO and KEGG analyses were also performed to dig deeper into the functions and interactions of VSMCs (Supplementary Figures 12–19).

Contractile VSMCs could switch into a synthetic phenotype through dedifferentiation in response to vascular injury, and they could also regain the contractile property via differentiation under certain conditions (13, 14). We assigned each cluster to two phenotypes based on recognized contractile (Acta2, Tagln, Figure 5A) or synthetic VSMC markers (S100A4, Figure 5B). We also discovered an extra novel gene, N-myristoyltransferase 1 (Nmt1), after comparing our data with classical markers, which was previously reported to participate in the development of cancers (16). Higher expression of Nmt1 was found in the stenotic artery, indicating that it might be a new marker in our ISR model (Figure 5C) but also in general synthetic VSMCs. To verify the findings above, we used platelet-derived growth factor BB [PDGF-BB, a classical cytokine promoting VSMCs dedifferentiated into the synthetic phenotype (17)] to treat mouse aortic vascular smooth muscle cells (MOVAS) at 20 ng/ml and found that MOVAS expressed a higher level of NMT1 after PDGF-BB treatment (Figure 5D).

Pseudotime analysis found that VSMCs in different clusters were separated and distributed in various routes, which were linked via three nodes (Figures 5E,F). Six clusters of VSMCs (Figure 5G) indicated that they were distributed at different stages: clusters 4, 5, and 6 might pause at advanced differentiation stages; clusters 1, 2, and 3 might pause at early stages. Meanwhile, we found that VSMCs in the case group showed a more extreme distribution in pseudotime routes but a disseminated distribution in the control group (Figure 5H). Finally, we identified cluster 4, 5, and 6 cells as contractile VSMCs and cluster 1, 2, and 3 cells as synthetic VSMCs (Figure 5J).

It has been reported that the Notch-Hes1 pathway, a pro-differentiation pathway, plays an essential role in the differentiation of many cell types (18). Notch1 increases after vascular injury, and inactivation of Notch1 might reduce neointimal hyperplasia (19). In our study, Notch1 and Hes1 were highly expressed in synthetic clusters 1 and 2 (Figure 4E), indicating that these two genes might be synthetic VSMC-related genes.

Our data demonstrated that fibroblasts, mainly distributed in the adventitia of blood vessels, could proliferate greatly during the progression of neointimal hyperplasia (Supplementary Figure 20). A heatmap could help us identify specific markers of these proliferous fibroblasts and provide a new strategy to distinguish fibroblasts in the normal position or the neointima (Supplementary Figure 21). Pseudotime analysis demonstrated that cells in cluster 3 might be proliferous fibroblasts in the case group (Supplementary Figures 22–25). In addition, GO terms and KEGG terms reminded us of the relationship between differentially expressed and specific functions (Supplementary Figures 26–33).

Originally, our model indicated damaged endothelial cells and VSMCs cause intimal hyperplasia. The increase in ECs might be caused by the appearance of EC-like VSMCs or re-endothelialization (20). Total ECs were clustered, indicating that cells in cluster 2 acted as EC-like VSMCs and that cells in cluster 1 could be related to re-endothelialization (Supplementary Figures 34–37). GO terms and KEGG terms also reminded us of the functional changes between the case and control groups (Supplementary Figures 38–45).

To verify the similarity between our model and real-world disease, we compared four sets of independent human disease susceptibility genes from GeneCards (https://www.genecards.org/) (ISR-related genes, VSMC phenotype switching-related genes, atherosclerosis-related genes, neointima proliferation-related genes) with our data to determine the expression abundance of previously recognized disease susceptibility genes (Figures 6A–D).

When analyzing ISR-related genes, the Ace, Selp, Nos3, Serpine1, and Sele genes were highly expressed in ECs; the Plat, Il6, Vegfa, F3, and Cdkn2a genes were highly expressed in fibroblasts; the P2ry12, Ccl2, Cdkn1a, and Hmox1 genes were highly expressed in macrophages; the Spp1 gene was highly expressed in transitional cells; and the Crp and Spp1 genes were highly expressed in VSMCs (Figure 6A). When analyzing atherosclerosis-related genes, we screened out high expression of the Olr1 and Sele genes in ECs; the Lpl, Ccl2, and Vegfa genes in fibroblasts; the Apoe, Abca1, Pparg, and Ccl2 genes in macrophages; the Ldlr and Eln genes in transitional cells; and the Ldlr, Lmna, Apoa2 and Crp genes in VSMCs (Figure 6B). Afterwards, we focused on VSMCs and found that the Lmna, Myh11, Mylk, Cacna1c, and Ptpn11 genes were intensely related to VSMC phenotype switching; the Pdgfb and Mtor genes were closely related to neointimal proliferation.

The matrix metallopeptidase (Mmp) gene family related to ISR was highly expressed in the stenotic artery but expressed at low levels in the control group (Mmp3 in fibroblasts; Mmp9 in macrophages; Mmp3 in transitional cells; Figure 6A). Meanwhile, several genes related to ISR exhibited high expression in the case group and low expression in the control group, such as Serpinc1 in ECs, Crp in fibroblasts, and Serpinc1 and Serpind1 in transitional cells (Figure 6A). When compared with the control group, several genes related to atherosclerosis were differentially expressed in the case group (Up: Apob in fibroblasts; Cyp7a1 in VSMCs; Down: Pcsk9, Apoa2, Crp, Cyp7a1 in fibroblasts; Apoa1 in macrophages; Alb in transitional cells, Figure 6C). We also found that the expression of several VSMC genes related to neointimal proliferation changed considerably with the progression of disease (Up: Pcna; Down: Akt1, Cdk4, and Rb1).

By comparing the susceptibility gene databases with our sequencing data, we obtained some results similar to previous studies and verified them in an in vivo study. Moreover, we discovered several novel genes based on the current model, such as Cyp7a1 and Cdk4, which should be further validated.

We performed receptor ligand analysis in the normal artery and stenotic artery (Figure 7A, Supplementary Figure 46). In the normal artery, Fn1, especially Fn1/α8β1 integrin, was enriched in ECs. Col1a1 and Col3a1, whose ligands are α1β1 integrin and α11β1 integrin, were enriched in fibroblasts. Moreover, macrophages were associated with the high expression of Cd74, Cd44/Copa, App, and Hbegf. Col1a1/α1β1 integrin and Col1a1/α11β1 integrin were also enriched in transitional cells. However, the gene expression of ligands in VSMCs was low.

Of note, the expression levels of receptors and ligands in the stenotic artery changed considerably compared with those in the normal artery (Supplementary Figure 15). Fn1/α8β1 integrin, Fn1/α5β1 integrin, Fn1/αVβ1 integrin, and Fn1/αVβ5 integrin were all enriched in ECs. Fn1, Col3a1, and Cd44 (ligands) and αVβ1, αVβ5, α8β1, α5β1, α1β1 integrin, SPP1, and Hbegf (receptors) were enriched in fibroblasts. Nrp2/Vegfa, Cd74/Copa, Cd74/App, and Cd74/Hbegf were enriched in macrophages. Moreover, transitional cells seemed to have high enrichment levels of Col18a1/α1β1 integrin and Cd44/Hbegf. The ligand and receptor circuit diagram suggested decreased communication between VSMCs and ECs (Figures 7B,C). In addition, ligands from fibroblasts decreased relative to the control group.