Abstract
Aberrant vascular smooth muscle cell (VSMC) homeostasis and proliferation characterize vascular diseases causing heart attack and stroke. Here we elucidate molecular determinants governing VSMC proliferation by reconstructing gene regulatory networks from single-cell transcriptomics and epigenetic profiling. We detect widespread activation of enhancers at disease-relevant loci in proliferation-predisposed VSMCs. We compared gene regulatory network rewiring between injury-responsive and nonresponsive VSMCs, which suggested shared transcription factors but differing target loci between VSMC states. Through in silico perturbation analysis, we identified and prioritized previously unrecognized regulators of proliferation, including RUNX1 and TIMP1. Moreover, we showed that the pioneer transcription factor RUNX1 increased VSMC responsiveness and that TIMP1 feeds back to promote VSMC proliferation through CD74-mediated STAT3 signaling. Both RUNX1 and the TIMP1–CD74 axis were expressed in human VSMCs, showing low levels in normal arteries and increased expression in disease, suggesting clinical relevance and potential as vascular disease targets.
Similar content being viewed by others
Main
Vascular smooth muscle cell (VSMC) proliferation underlies cell accumulation in atherosclerotic artery disease and vascular restenosis, which occurs following stenting and grafting. While genetic studies point to VSMCs as important hereditary determinants of cardiovascular disease1,2,3, their clinical targeting is presently unexplored. VSMC lesion investment in experimental atherosclerosis results from extensive clonal expansion of a small number of cells4,5. Similar VSMC oligoclonality has been shown in other vascular disease models using genetic VSMC lineage tracing6,7, and clonal VSMC contribution in human disease has been proposed8,9. Modification of VSMC clonality by altered cell–cell communication10 and aging11 suggests that changes in activation frequency could underlie increased vascular disease risk.
Single-cell RNA-sequencing (scRNA-seq) studies in mouse models and human vascular disease have shown the remarkable transcriptional heterogeneity of VSMC-derived cells12. VSMCs can adopt states that range from quiescence, with high expression of contractile genes (such as MYH11 and ACTA2), in healthy arteries, to cells that have induced signatures of other cell types, including fibromyocytes (for example, TNFRSF11B), macrophages (LGALS3) and chondrocytes (RUNX2) in addition to extracellular matrix (ECM) proteins and remodellers that are characteristic of the so-called synthetic state generated by classical VSMC phenotypic switching13,14,15,16.
Functional genomic analyses in lineage-traced animals have shed light on genetic determinants and mechanisms regulating VSMC state changes in disease1,2,17. These experimental approaches have also revealed unexpected regulators of VSMC investment, such as the pluripotency factor OCT4 (ref. 18), and identified a mesenchymal VSMC-derived state marked by expression of vascular cell adhesion protein 1 (VCAM1) and stem cell antigen 1 (SCA1)13,16. SCA1-positive VSMCs are found infrequently in healthy arteries but more abundantly in disease models, and have been linked to VSMC priming and proliferation14,16,19,20. This molecular heterogeneity may explain the apparent distinct effects of VSMC regulators in different contexts21,22. However, the events governing activation of clonal VSMC proliferation are understudied23.
The integration of scRNA-seq data with epigenetic information led to the identification of factors regulating developmental processes and disease24. Here we use this approach to model gene regulatory networks upon acute vascular injury when VSMC proliferation initiates. We find differential use of transcription factors in distinct VSMC transcriptional states along a proliferation-associated trajectory, possibly explaining context-specific effects of VSMC regulators. The analysis identifies known and candidate regulators that are prioritized using in silico simulation analysis. We functionally implicate Runt-related transcription factor 1 (RUNX1) and Tissue Inhibitor of Metalloproteinases-1 (TIMP1) in driving the activation of VSMC proliferation and suggest that these factors also operate in human disease development.
VSMC activation is associated with de novo chromatin opening at distal sites linked to vascular-disease-associated genes
To investigate the molecular regulation of VSMC proliferation, we elicited an injury response by carotid artery ligation, which leads to reproducible VSMC phenotypic switching and proliferation after 5–7 days25. Surgery was performed in Myh11–CreERt2, Rosa26–EYFP (Myh11–EYFP) animals after tamoxifen induction of heritable reporter-EYFP expression in VSMCs to overcome the rapid loss of VSMC marker expression after injury. VSMCs expressing SCA1 have increased proliferative capacity and represent cells with pronounced phenotypic switching at this stage of injury13,16,20. We therefore mapped chromatin accessibility using the assay for transposase-accessible chromatin with sequencing (ATAC-seq) in populations of SCA1+ lineage-traced (EYFP+) VSMCs from animals 7 days after surgery and compared it with the chromatin accessibility of EYFP+ VSMCs from no-injury littermate controls. We also profiled SCA1− cells from injured arteries, which include VSMCs in remodeled areas that show less extensive phenotypic changes and cells in adjacent carotid artery regions without obvious injury responses. Samples from all three conditions (control, injury SCA1+ and injury SCA1−) had a high signal-to-noise ratio and strong replication of peak intensity (Fig. 1a and Extended Data Fig. 1a–c).
Accessibility changes were minor at Myh11 and other contractile genes that are downregulated after injury (Fig. 1a), consistent with the documented retention of the active H3K4me2 marker at contractile genes in synthetic VSMCs26,27. Increased accessibility was found at, for example, Spp1, which is induced in both SCA1+ and SCA1− VSMCs, whereas the chromatin at the SCA1-encoding Ly6a locus became more accessible selectively in SCA1+ cells (Fig. 1a). Of all peaks detected across the three conditions, 96% were accessible in SCA1+ cells (Fig. 1b). This genome-wide de novo chromatin opening in SCA1+ (82,000 peaks) compared with that in control samples (57,000) was consistent with the scRNA-seq transcript levels, which gradually increase across the VSMC population after injury (Fig. 1c).
ATAC-seq signal intensity was also generally higher in SCA1+ samples than in control VSMCs (Fig. 1d), in particular at regions with lower accessibility, which is a characteristic of enhancers. Consistently, the ATAC-seq peak intensity was shifted upward for non-promoter peaks, but not changed for promoters and CpG islands, and the proportion of peaks at distal elements was also higher in SCA1+ cells than in control samples (Fig. 1e,f). A significantly higher signal intensity was detected at 28,303 peaks in SCA1+ samples compared with control VSMCs, whereas only 839 peaks were higher in controls compared with SCA1+ (Fig. 1g). The SCA1− sample showed intermediate results, possibly reflecting the heterogeneity of the cell population and the continuum of VSMC transcriptional states induced by injury20. Genes associated with regions selectively assessible in control cells were enriched for muscle contraction gene ontology (GO) terms (for example, Tcap; Fig. 1h and Extended Data Fig. 1d). By contrast, genes associated with the regions with the biggest chromatin accessibility increase in SCA1+ compared with control samples were enriched for vascular diseases, including atherosclerosis, and biological process ontology terms linked to VSMC proliferation, such as regulation of signal transduction (Mapk6), inflammation (Ccl2) and cell adhesion (Lamb1; Fig. 1h and Extended Data Fig. 1e–g). This analysis shows that injury-induced VSMC activation results in the opening of chromatin at enhancers of genes related to the functions of synthetic VSMCs and vascular disease, particularly in SCA1-expressing VSMCs.
Rewiring of factors shared between injury-induced VSMC states
To identify factors controlling activation of VSMC proliferation, we compared how gene expression is orchestrated in VSMCs that respond to injury versus those that do not. We used scRNA-seq profiles of VSMCs analyzed at the very onset of VSMC proliferation (5 days after the injury), in which we previously identified a proliferation-associated trajectory20. VSMCs were classified based on their position along this trajectory as ‘non-responding’ (‘non-RSP’, defined by persistently high levels of contractile gene expression), a ‘linking’ cell population (‘LNK’), leading on to a pre-proliferative state that includes cells expressing SCA1 (‘PrP’) and, finally, actively cycling cells that express Mki67 and have high G2/S cell cycle phase scores (‘CYC’; Fig. 2a and Extended Data Fig. 2a).
To construct gene regulatory networks (GRNs) for these four VSMC states, we first generated a VSMC-specific transcription factor–target interaction network based on all detected ATAC-seq peaks, which was combined with gene expression profiles. This was done for each of the VSMC states (non-RSP, LNK, PrP and CYC), using CellOracle24, and the resulting four GRNs had comparable centrality score distribution (Extended Data Fig. 2b). We found a substantial overlap between the networks, with 233 out of 597 (37%) shared nodes, including 82 out of 140 (59%) transcription factors (Fig. 2b and Supplementary Table 1a,b). Despite the high number of common nodes, few interactions were present in all GRNs (Fig. 2c, lower panel). Accordingly, abundant changes in the network topology rankings, such as degree centrality, were observed between networks (Fig. 2d and Supplementary Table 1c). GRN modeling using the scRNA-seq profiles of VSMCs analyzed 7 days after injury was conducted in parallel and produced comparable results (Extended Data Fig. 2d).
The network nodes were enriched for processes related to VSMC biology or vascular injury, including ECM organization, actin organization, response to stimuli and regulation of cell proliferation (Fig. 2c and Supplementary Table 1b). Overall, specific regulatory circuits driving biological processes were not detected (Extended Data Fig. 2c). However, the CYC network included a substructure consisting of cell-cycle-associated genes regulated by RAD21, a subunit of the cohesion complex that has roles in mitosis, genome stability and transcriptional regulation (Extended Data Fig. 2b,c). The CYC network, which had the most unique factors (14 transcription factors; Fig. 2b), was dominated by genes driving cell cycle progression, such as Myc, Mybl1, Erg and Rad21 (Fig. 2d). To study processes and factors driving activation, rather than progression, of VSMC proliferation, we focused on comparison to the PrP network (Fig. 2e).
Genes with most network changes in the PrP relative to the non-RSP network were predominantly transcription factors (36 out of 50) and fell into three groups by expression pattern (Fig. 2f). Group 1 increased expression along the injury-associated trajectory, and genes included markers of VSMC phenotypic switching (Spp1, Fn1, Mgp), Vcam1 (that marks activated VSMCs13), Timp1 and transcription factors Scx, Cebpb, Runx1 and Eno1. Group 2 genes, which showed reduced expression along the trajectory, included Mef2c; Thra; Hes; Heyl; Krüppel-like factors (KLF) 9, 13 and 15; and inhibitor of differentiation (Id) genes. Finally, group 3 genes showed little variation in expression between the non-responding and proliferation-associated states. This group contained factors reported to affect VSMC regulation, for example, Klf4, Egr1, Atf3 and AP-1 subunits. We identified significant colocalization events between human VSMC expression quantitative trait loci (eQTLs) and signals in genome-wide association studies for coronary artery disease (CAD) for 13 human orthologs of the 50 genes with the highest rewiring scores (for example, KLF2/4/13/15, HEYL, TCAP, S100A10, VCAM1; Supplementary Table 1d,e), indicating causality of VSMC regulatory rewiring in disease.
Injury-induced GRNs are relevant in atherosclerosis
To assess the relevance of the regulation detected in injury to disease-associated VSMC changes, we used available datasets from VSMC-derived cells in experimental atherosclerosis13. First, we considered genes with higher expression in the PrP compared with the non-RSP state in injury. Expression of this ‘PrP signature’ was increased in VSMC subsets characterized by reduced expression of contractile genes (Fig. 3a,b and Extended Data Fig. 3a). Secondly, we found substantial overlap in transcript regulation between the two disease models for GRN nodes. Differential expression between contractile (cluster 1) and modulated (clusters 0 + 3) VSMC clusters in atherosclerosis was visualized on the PrP network (Fig. 3c), revealing a similarity to the observed changes in injury (Fig. 2e). In general, transcription factors with increased expression in modulated VSMCs in atherosclerosis had positive connections to upregulated target genes in the PrP-GRN (for example, Sox9-Vcam1, Eno1-Tmsb10/Lgals3, Cebpb-Clu). Vice versa, transcription factors with reduced expression in modulated versus contractile VSMCs in atherosclerosis, such as THRA, had positive GRN interactions with genes that were also downregulated in modulated cells (for example, Csrp2, Mfap4, Sost, Myl9; Fig. 3c). This analysis suggests similarity of the VSMC states in atherosclerosis and injury, and indicates that regulatory relationships mapped in injury are also relevant in atherosclerosis.
To compare with human disease, we isolated scRNA-seq profiles of mural cells from a published analysis of human carotid plaques13. The cells were clustered into six groups that were characterized by marker analysis (Fig. 3e and Supplementary Table 2b). Clusters 0–2 had abundant contractile gene expression, cluster 3 cells expressed TNFRSF11B that was identified as a fibromyocyte marker14, VCAM1 was expressed mainly in clusters 3 and 4, and cluster 5 had increased levels of ribosomal proteins (Extended Data Fig. 3d,e). We then visualized the expression profiles for human orthologs of injury-rewired genes in the human cell clusters ordered by contractile gene expression. To capture regulatory relationships, we included the top 10 genes (by GRN rewiring score) and their direct interactors. Genes were visualized in the same order in human plaque and mouse injury cells (hierarchically clustered by their expression along the proliferation-associated trajectory; Fig. 3f). This side-by-side comparison showed similar expression patterns in mouse injury and carotid plaque VSMCs, indicating that the GRN analysis identified regulation that is relevant in human atherosclerosis.
Prioritization of candidate regulators of VSMC activation
Transcription factors changed topology scores between the GRNs (Fig. 4a), prompting a motif enrichment analysis for ATAC-seq peak regions showing differential accessibility between conditions (Fig. 4b). The MEF2 and SRF motifs were enriched for peaks showing higher accessibility in control samples compared with SCA1+ VSMCs from injured arteries. Peaks that gained de novo accessibility after injury were enriched for AP-1, NFκB, CEBP, ETS and RUNX binding sites. Interestingly, RUNX, NFκB and CEBP motifs were also found to colocalize with AP-1 binding sites (Extended Data Fig. 4a), suggesting co-regulatory interaction at some loci.
The generated GRNs enabled computational simulation of how perturbation of transcription factor levels would affect cell state24. KLF4, a known regulator of VSMC phenotypic switching that promotes a macrophage-like state in atherosclerosis28, had high centrality in all networks apart from that of proliferating cells (CYC). In silico simulation of KLF4 depletion predicted a general shift towards the non-responding cell state, concomitantly with promotion of CYC cells (Fig. 4c). Such context-specific effects of KLF4 depletion are consistent with the delayed downregulation of contractile genes, but accelerated neointima formation in VSMC-specific KLF4 knockout after injury animals21.
To systematically predict function, we generated in silico perturbation scores (PSs) for each transcription factor in the networks. Positive scores, reflecting stimulation of progression along the proliferation-associated trajectory20, and negative scores, indicating blocking effects, were calculated for loss of function (knockout, KO) and gain of function (overexpression, OE) separately and showed strong correlation (R2 > 0.9; Extended Data Fig. 4b and Supplementary Table 1g). Transcription factors scoring highly for both positive and negative PSs were predicted to have context-specific effects (Fig. 4d and Extended Data Fig. 4c,d). RUNX1, FOSL1, TWIST1 and PRRX1 were predicted to mainly stimulate trajectory progression (negative KO score < −1, positive KO score < 0.75 (Fig. 4d,e and Extended Data Fig. 4c,d)). The factors predicted to block trajectory progression (positive KO score > 1) were enriched for TGF-beta signaling (KEGG:04350), IL-17 signaling (KEGG:04657) and parathyroid hormone synthesis, secretion and action (KEGG:04928). Interestingly, thyroid hormone receptor alpha (THRA), which has been suggested to impact VSMC cholesterol metabolism29, was among trajectory-blocking factors (Fig. 4d). Thra is downregulated in PrP compared with non-RSP VSMCs and negatively interacts with synthetic genes (Spp1, Vcam1 and Mgp) in the PrP network, while promoting cytoskeletal genes (Actg2, Pdlim4, Cdc42ep3) in the non-RSP GRN (Supplementary Table 1a). Simulation of THRA perturbation also suggested that THRA may safeguard the contractile state (Fig. 4e and Extended Data Fig. 4c–d).
Next, we used the GRN analysis to investigate how targets of this transcriptional rewiring may impact VSMC biology. To this end, we performed pathway analysis of the nodes with high target gene scores (‘in-degree centrality’) in the PrP or CYC GRNs that are also induced in PrP compared with non-RSP cells (Supplementary Table 1c,d). This highlighted GO terms including ECM organization (Spp1, Vcam1, Cxcl2, Postn) and regulation of cell adhesion (Eln, Lum, Fbln2). We also found modulators of proteolysis (Timp1, Serpine1, Thbs1) among the targets of rewired transcription factors, consistent with cellular niche remodeling by activated VSMCs.
Overall, this analysis highlights both known and putative VSMC regulators. To experimentally test factors highlighted by the GRN analysis, we selected RUNX1 as a candidate stimulating transcription factor, based on top ranking in the PS analysis combined with motif enrichment in de novo accessible chromatin, and TIMP1 as an example of a highly rewired target gene that, interestingly, was predicted to be indirectly induced by RUNX1.
RUNX1 promotes VSMC proliferation
The GRN analysis highlighted RUNX1, which belongs to the Runt-related transcription factor family and plays a key role in hematopoiesis, as a potent stimulator of VSMC activation. Consistent with this, Runx1 transcripts were increased in the PrP and CYC states compared with non-responding cells, increased chromatin accessibility was detected at both Runx1 promoters in SCA1+ VSMCs compared with the control samples, and we detected RUNX1 in lineage-labeled VSMCs after injury (Fig. 5a and Extended Data Fig. 5a,b). By contrast, the RUNX family members Runx2 and Runx3 were lowly expressed at this stage of VSMC activation (Extended Data Fig. 2a). The RUNX motif was also highlighted in relation to VSMC regulation by studies in experimental atherosclerosis1, and RUNX1 transcript levels were higher in modulated compared with contractile VSMCs in mouse and human atherosclerosis (Fig. 3d and Extended Data Fig. 3f).
GRN-based simulation of elevated RUNX1 expression predicted that RUNX1 promotes the transition towards VSMC proliferation (Fig. 5b), as expected from the limitation of proliferation observed when simulating Runx1-knockout (Fig. 4e and Extended Data Fig. 4c,d). The PrP GRN predicted that RUNX1 stimulates injury-induced genes, including Vcam1, Lum and Mmp14, and GO analysis of all direct and indirect RUNX1 target genes suggested that RUNX1 could affect cell-substrate adhesion (Fbln2, Mmp14), response to stimulus (Cp, Mt2) and ECM organization (Eln) (Fig. 5c). Interestingly, 4 of the 15 direct RUNX1 target genes were previously associated with activated, SCA1+ VSMCs in healthy arteries16. We also found that expression of the predicted RUNX1 target, matrix metalloproteinase 14 (MMP14), is a hallmark of SCA1+ VSMCs (Extended Data Fig. 5c), further indicating that RUNX1 supports a state that is primed for proliferation.
We experimentally tested how RUNX1 affects gene expression in freshly isolated lineage-traced VSMCs. Lentiviral-induced RUNX1 overexpression significantly increased the expression of predicted direct targets Mmp14, Timp1 and Cebpd. Conversely, the depletion of Runx1 mediated by short interfering RNA (siRNA) significantly reduced Mmp14 and blunted Timp1 expression (Fig. 5d and Extended Data Fig. 5d,e). To test whether RUNX1 affects activation of VSMC proliferation, we adapted an in vitro proliferation assay that reproduces the low-frequency clonal VSMC expansion observed in vascular disease models20. Lentivirus expressing red fluorescent protein (RFP) was introduced in lineage-labeled EYFP+ VSMCs at low titer and expanding RFP/EYFP double-positive clonal VSMC patches identified by repeated live cell imaging over 3 weeks. This showed a significantly increased clonal patch-forming ability of RUNX1-overexpressing versus control cells (Fig. 5e).
In human VSMCs (hVSMCs), RUNX1 depletion decreased the percentage of EdU+ cells and, vice versa, overexpression of RUNX1 increased the percentage of EdU+ cells, relative to their respective controls. This suggests that stimulation of VSMC proliferation by RUNX1 is conserved in humans (Fig. 5f and Extended Data Fig. 5f). To assess RUNX1 functions in human disease, we immunostained arteries without detectable plaques but showing varying degrees of intimal thickening and carotid plaques. RUNX1 protein was detected infrequently in alpha smooth muscle actin (αSMA)+ cells in arteries with a healthy, organized morphology of the medial layer, but a higher proportion of medial RUNX1+αSMA+ cells were observed in arteries with evidence of perturbation (Fig. 5g). For example, in arteries showing intimal thickening, many αSMA+ cells within the intima expressed RUNX1 (Fig. 5h and Extended Data Fig. 5g). Substantial heterogeneity in RUNX1 expression was observed in human atherosclerotic lesions; interestingly, however, αSMA+RUNX1+ cells were also observed in the fibrous cap (Fig. 5i). Jointly, these results show that RUNX1 regulates gene expression in VSMCs and promotes their proliferation, and indicates that RUNX1 is implicated in human disease development.
TIMP1 promotes VSMC proliferation in an MMP-independent manner
TIMP1 was among the most central nodes and was selected among the targets of regulatory rewiring as increased TIMP1 protein levels have been associated with cardiovascular disease and severity30,31. In the PrP GRN, TIMP1 was predicted to be induced by multiple factors including Cebpd, Myc, Prrx1 and Scx (Figs. 2e and 6a), and an indirect RUNX1 target (see above). TIMP1 transcripts were also upregulated by VSMCs in human and mouse atherosclerosis scRNA-seq datasets (Fig. 3d, Extended Data Fig. 3f and refs. 1,13,16).
TIMP1 protein was expressed in clonally expanded VSMCs in both mouse atherosclerotic and injury-induced lesions, and also detected in αSMA+ cells located in the medial layer of non-plaque human arteries (Fig. 6b,c and Extended Data Fig. 6a,b). This suggested that TIMP1 could affect VSMCs at an early disease stage. To investigate the potential impact of TIMP1 protein on VSMC function, we performed bulk RNA-seq in TIMP1-treated versus control hVSMCs. This revealed a significant induction of cell cycle genes (E2F targets and G2/M checkpoint), fatty acid metabolism and oxidative phosphorylation gene sets in TIMP1-treated cells (Extended Data Fig. 6c), indicating an impact on metabolism and cell proliferation. We therefore tested the impact of TIMP1 on VSMC proliferation, as the role of TIMP1 is controversial32,33,34. We found a small, but significant, increase in the percentage of EdU+ cells in TIMP1-treated hVSMCs versus controls (Fig. 6d), similar to a recent study34. TIMP1 treatment also stimulated the clonal expansion of quiescent VSMCs isolated directly from aortas of lineage-labeled Myh11–Confetti animals (Fig. 6e,f). TIMP1 increased the frequency of VSMC clone formation to a similar degree as PDGF-BB treatment (Fig. 6f). However, TIMP1 did not change the size of clonal patches, in contrast to platelet-derived growth factor (PDGF)-BB (Fig. 6f, lower panel), suggesting that the establishment and growth of VSMC clones are regulated independently.
TIMP1 functions both as an MMP inhibitor and as a cytokine35,36. We found no effect of a broad-spectrum MMP inhibitor (GM6001) on clone dynamics, neither in control cells nor in TIMP1-treated cells (Fig. 6g). This suggested that the ability of TIMP1 to promote clonal VSMC proliferation is not linked to the inhibition of proteolysis, similar to established hVSMC cultures32. Taking all these results together, we conclude that TIMP1 stimulates clonal VSMC proliferation in an MMP-independent manner.
TIMP1 induces STAT3 phosphorylation to induce VSMC proliferation
The signaling pathways activated by TIMP1 show context dependency34,36,37, so we screened for TIMP1-mediated phosphokinase activation in hVSMCs. This suggested that signal transducer and activator of transcription (STAT3) phosphorylation at serine 727 (S727) is strongly increased in TIMP1-treated cells, which was confirmed by western blotting in independent hVSMC isolates (Fig. 7a,b and Extended Data Fig. 7a). Increased phosphorylation was observed at both S727 and Tyr705, while total STAT3 protein levels were unchanged (Fig. 7b,c and Extended Data Fig. 7b,c). The rapid and transient induction of STAT3 phosphorylation suggests a direct response to TIMP1 (Fig. 7b,c). We also analyzed the effect of TIMP1 on signal transducers known to regulate VSMC growth. Similar to STAT3, phosphorylation of AKT (Ak strain transforming) was induced by TIMP1, whereas phosphorylation levels of p38 mitogen-activated protein kinase (p38) were variable between hVSMCs from different individuals (Fig. 7b,c and Extended Data Fig. 7b,c).
The role of these signaling pathways in TIMP1-mediated stimulation of clonal VSMC proliferation was investigated by addition of small-molecule inhibitors in the clonal VSMC proliferation assay. Inhibition of STAT3 using TT-101 abolished the increased clone number in response to TIMP1, whereas the impact of AKT and p38 inhibition was less obvious (Fig. 7d). To further test the specificity, we depleted STAT3 using siRNA, which also prevented the TIMP1-mediated increase in clone formation (Fig. 7e). Phosphorylation of STAT3 leads to dimerization, nuclear import and binding at target gene promoters38. Consistent with its activation, STAT3 chromatin immunoprecipitation showed increased binding in TIMP1-treated cells, specifically at promoters of canonical STAT3 target genes (JUNB, TWIST1), compared with control cells (Fig. 7f).
VSMC-derived cells in atherosclerotic plaques and neointimal lesions expressed high levels of pSTAT3 compared with medial cells (Fig. 7g and Extended Data Fig. 7d), indicating its involvement in clonal VSMC expansion. To further assess whether STAT3 phosphorylation is linked with VSMC proliferation, we co-stained hVSMCs for KI67 and pSTAT3 (S727) after TIMP1 treatment. Overall, KI67+ cells had higher nuclear pSTAT3 levels compared with KI67− cells and there was significantly increased pSTAT3 expression 15 min after TIMP1 stimulation (Fig. 7h,i). There was no effect of TIMP1 on hVSMC proliferation within this time frame (Extended Data Fig. 7e). We next analyzed VSMCs isolated from aortas of lineage-labeled mice and cultured to mimic phenotypic switching and activation of proliferation. STAT3 phosphorylation was also induced by TIMP1 after 4 days of culture, but this effect was less pronounced in cells analyzed at day 7 (Extended Data Fig. 7f). Interestingly, baseline pSTAT3 intensity was higher in cells cultured for 4 days after isolation, compared with day 7 cells, suggesting the existence of a ‘window of opportunity’ with higher responsiveness to TIMP1. Overall, these data link TIMP1-mediated STAT3 phosphorylation to the activation of VSMC proliferation.
TIMP1 signals through CD74 to induce VSMC proliferation
Signaling by TIMP1 can be executed via several surface receptors that bind to different TIMP1 protein domains37. We found that the ability to induce clonal VSMC proliferation was retained by the N-terminal part of TIMP1 (Fig. 8a), which binds the HLA class II histocompatibility antigen gamma chain, CD74 (ref. 39), that has also been detected in modulated or ‘macrophage-like’ VSMCs in experimental atherosclerosis40.
Immunostaining revealed that Myh11-lineage-labeled VSMCs also express CD74 protein after vascular injury in vivo (Fig. 8b and Extended Data Fig. 8a), providing evidence that TIMP1 could function via CD74 in VSMCs, as suggested34. CD74 was also detected in cultured lineage labeled mouse VSMCs (mVSMCs) (Fig. 8c); we found an association between high pSTAT3 and greater CD74 expression (Fig. 8c,d), indicating that CD74 signaling results in phosphorylation of STAT3. To directly test whether TIMP1 acts via CD74 in VSMCs, we used a CD74-blocking antibody, which abolished TIMP1-augmented VSMC clone formation (Fig. 8e). Interestingly, anti-CD74 also reduced the frequency of clone formation in untreated cells, which could be due to blocking the effect of TIMP1 secreted by VSMCs after culture-induced phenotypic switching, or other CD74 ligands such as MIF. Blocking CD74 in VSMCs overexpressing RUNX1 also abolished the stimulation of proliferation (Extended Data Fig. 8b), suggesting that the observed increase in Timp1 expression could be important for the RUNX1-OE phenotype.
To assess whether the TIMP1–CD74–pSTAT3 axis is active in atherosclerosis, we investigated lineage-labeled Apoe−/− animals after a 4-week exposure to a high-fat diet, at a timepoint before robust VSMC contribution in lesions10,20. Imaging flow cytometry showed a low frequency of KI67+ lineage-labeled VSMCs, as expected, and this varied between animals. We found a striking enrichment of CD74 expression in KI67+ cells (Fig. 8f), and the levels of phospho-STAT3 were higher in proliferating (KI67+) and in CD74+ VSMCs, compared with all lineage-labeled cells (Fig. 8g). This suggests that CD74 and phosphorylation of STAT3 are linked to VSMC proliferation. TIMP1 treatment further increased pSTAT3 levels in CD74+, but not in CD74−, proliferating VSMCs (Fig. 8h and Extended Data Fig. 8c), providing in vivo evidence linking this signaling pathway to the initiation of VSMC proliferation in atherosclerosis.
Next, we explored the TIMP1–CD74–pSTAT3 axis in humans. As shown in Fig. 8i, TIMP1-induced STAT3 phosphorylation was reduced in cells co-treated with the anti-CD74 antibody and a similar effect was observed using a CD74-blocking peptide41, showing that TIMP1-mediated induction of STAT3 phosphorylation is downstream of CD74 (Fig. 8i,j). We found that CD74 was detected by immunostaining in αSMA+ cells in non-plaque arteries from human organ donors and in samples taken after carotid endarterectomies (Fig. 8k and Extended Data Fig. 8d). To quantify expression, we performed multi-probe in situ hybridization (Fig. 8l,m). TIMP1 and CD74 transcripts were detected in ACTA2+ VSMCs in non-plaque arteries, but were observed at low frequency. Compared with medial cells in arteries without lesions, the frequency of CD74 and TIMP1 expressing cells was elevated in samples with obvious plaques. Quantification of positive cells revealed substantial co-expression of TIMP1 and CD74 with ACTA2 (12–43% of ACTA2+ plaque cells expressed TIMP1; 1–41% for CD74). Altogether, these findings suggest that the induced expression of TIMP1 upon VSMC activation could stimulate VSMC proliferation by binding to CD74 via activation of STAT3 phosphorylation.
Discussion
We here used GRN modeling after acute vascular injury to identify factors governing the activation of VSMC proliferation, which could be important to reduce vascular disease susceptibility and for disease prevention strategies. Our analysis indicates that differential use of shared transcription factors plays an important role in this process. This also suggests that VSMC state-dependent mechanisms may underlie observed context-specific functions, such as the dichotomous function of monocyte chemoattractant protein-1 (MCP-1), encoded by Ccl2 (ref. 22). Using in silico simulation experiments, we prioritize both factors known to control VSMCs in disease and candidate regulators proposed to safeguard the contractile state (for example, THRA) or promote cell activation (for example, RUNX1, CEBP/D, ENO1). We experimentally validate RUNX1 and the TIMP–STAT3–CD74 axis in vivo and in vitro as VSMC regulators with relevance to human diseases. Genetic evidence linking these factors to vascular disease includes a variant in the RUNX1 locus associated with stroke (rs116262092-A) and colocalization of a STAT3 VSMC eQTL with a CAD GWAS signal (Supplementary Table 1g). Furthermore, vascular abnormalities and increased aneurysm risk have been identified in patients with missense STAT3 variants causally implicating this pathway in vascular regulation42,43.
RUNX1 has well-known functions in hematopoiesis and also impacts differentiation and proliferation in other contexts, including cardiovascular cells44,45. We here show that RUNX1 increases the frequency of VSMCs that start proliferating, possibly by directly inducing the expression of VSMC ‘transition state’ genes, such as MMP14. It is therefore tempting to speculate that this pioneer transcription factor could drive the activation of alternative transcriptional programs in VSMCs. The expression of RUNX1 protein in human arteries with intimal thickening is higher compared with those with a healthy morphology, suggesting that RUNX1 may indeed play a role in vascular changes predisposing to disease.
We show that TIMP1, a target of transcriptional rewiring, promotes the establishment of clonal VSMC proliferation itself, suggesting a positive feedback loop that might impact vascular disease susceptibility. TIMP1 also stimulates proliferation in hVSMC cultures34, and higher TIMP1 serum levels have been associated with elevated cardiovascular disease risk in humans31. Future studies are needed to assess whether increased TIMP1 in serum is a cause or consequence of activated VSMCs within diseased arteries. In experimental atherosclerosis, Apoe−/−, Timp1−/− animals had smaller plaques with reduced VSMC content compared with Apoe−/− controls30,46. By contrast, femoral artery injury yielded larger neointimal lesions and increased MMP activity in Timp1−/− animals47. As VSMC lineage tracing was not performed, the impact on VSMC proliferation was not assessed in these studies. Interpreting the effects of global TIMP1 manipulation on VSMC function is further complicated by, firstly, the dual effect of TIMP1 on MMP activity and cell signaling and, secondly, the ability of TIMP1 signaling to impact multiple cell types in vascular disease34. Whereas TIMP1 depends on AKT in macrophages34, CD74 signaling via STAT3 is required for TIMP1 stimulation of VSMC clonal proliferation in vitro, and this axis is also active in vivo.
The regulatory interactions observed here are consistent with gene expression changes in atherosclerosis and include factors impacting on cultured VSMCs and VSMC-derived plaque cells. For example, enriched motifs in single cell ATAC-seq analysis of human CAD patients include AP-1, RUNX and CEBP, and also highlighted the importance of STAT3 in human vascular disease2. Similarly, AP-1 and STAT motifs were highlighted by analysis of lineage-traced VSMCs that adopt a more fibroblastic phenotype during mouse atherogenesis1. We also detected RUNX1 and CD74 within αSMA+ cells in developed human atherosclerotic lesions. However, whether enrichment of these motifs in atherosclerosis is caused by ongoing VSMC activation in disease or implies that these factors also function at later disease states remains to be tested. Interestingly, studies in zebrafish heart regeneration showed that RUNX1 promotes expression of smooth muscle cell genes in mesenchymal cells45, suggesting that in addition to regulating VSMC proliferation, it could also impact fibrous cap formation. The potential links between mechanisms regulating VSMC proliferation and VSMC-derived atherosclerotic plaque cell phenotypes are an exciting topic for future research.
Methods
Detailed methods are available in Supplementary Information; antibodies are listed in Supplementary Table 4 and primer sequences in Supplementary Table 5.
Animals and procedures
All experiments were done according to UK Home Office regulations (project licenses P452C9545 and PP7513347) and were approved by the Cambridge Animal Welfare and Ethical Review Body. Mice (C57Bl/6) were housed on a 12 h dark and light cycle at 19–21 °C and 45–65% humidity. Alleles were previously generated; Myh11–CreERt2 (RRID:IMSR_JAX:019079) is a Y-linked transgene that confers expression of a tamoxifen-inducible Cre recombinase in smooth muscle cells, Rosa26–Confetti (RRID:IMSR_JAX:013731) and Rosa26–EYFP (RRID:IMSR_JAX:006148) are Cre-recombination reporter alleles, KI67–RFP (RRID:IMSR_JAX:029802) is an insertion in the Mki67 locus resulting in expression of a KI67–RFP fusion protein, and the mutant Apoe allele (RRID:IMSR_JAX:002052) sensitizes mice to atherosclerosis induced by a high-fat diet (HFD). VSMC lineage labeling was achieved by administration of tamoxifen (10 intraperitoneal injections of 1 mg ml−1 over 2 weeks). Note that Confetti–GFP induction occurs at low frequency4. Myh11–CreERt2 is Y linked, so males were used for VSMC lineage tracing. The animals were allowed to rest for >1 week after the last injection to allow tamoxifen washout before tissue harvest, HFD feeding (21% fat and 0.2% cholesterol, Special Diets Services) and vascular injury, or randomized into groups receiving intraperitoneal injections of recombinant mouse TIMP1 protein (200 μg kg−1, N = 5) or vehicle (phosphate-buffered saline (PBS), N = 4) daily for 9 days at the end of a 4 week HFD protocol. Pre-operative analgesic was given subcutaneously (~0.1 mg per kg body weight, buprenorphine) and the left carotid artery was tied off with a silk suture just under the bifurcation point under isoflurane anesthesia (inhalation, 2.5–3%; 1.5 l min−1 induction, maintained at 1.5%). The animals were euthanized (cervical dislocation or CO2 asphyxiation) 5–28 days after surgery and perfused with cold PBS before tissue removal.
Human tissue
Human arteries were collected after informed consent and approval by the Cambridgeshire 1 or East of England—Cambridge South Research Ethics Committee (REC ref 15/EE/1052, H00/514) were obtained. Non-plaque aortas were obtained from organ donors via the Cambridge Collaborative Biorepository for Translational Medicine. Carotid endartectomy samples were obtained from the Royal Papworth Hospital research tissue bank. Cell isolates from a total of 13 donors (age (years) and sex: 27, male; 20, male; 70, male; 75, female; 45, female; 65, male; 62, female; 51, male; 68, female; 69, male; 72, female; 68, male; 61, male) and histological analysis of 9 carotid artery samples (72, female; 82, male; 80, female; 67, male; 67, female; 58, female; 74, male; 54, male; and 50, female) and 10 aortas (60, male; 24, male; 84, female; 65, male; 49, male; 35, male; 41, male; 75, female; 18, male; 59, female) were included. Both male and female tissue and cell isolates were included in all analyses. The experiments with human samples conformed to the principles outlined in the Declaration of Helsinki.
VSMC isolation, culture and treatment
The hVSMCs were cultured from the aortas of patients undergoing cardiac transplantation or aortic valve replacement in an hVSMC-specific medium (Promocell, SMC-GM2, C22062) supplemented with antibiotics. The hVSMCs were serum starved in media supplemented with 0.1% BSA before TIMP1 treatment. Single-cell suspensions of mVSMCs were generated from freshly isolated aortas of wild-type or VSMC lineage-labeled animals (Myh11–Confetti or Myh11–EYFP) and cultured in DMEM supplemented with 10% FCS and antibiotics (complete medium).
The RUNX1 overexpression lentiviral vector was generated by inserting full-length mCherry and Runx1 cDNA (Runx1–202) linked by T2A into a pLentiGFP backbone (Addgene, catalog number 17448). VSMCs were transduced with lentivirus in media containing 10 µg ml−1 protamine sulfate (Sigma, P3369). Transfection with siRNA (50 nM ON-TARGETplus SMART Pool, Dharmacon, targeting human RUNX1 (L-003926-00-0005), mouse Runx1 (L-048982-00-0005), mouse Stat3 (L-040794-01-0005) or non-targeting control siRNA, Control Pool, D-001810-10-05) was done using Lipofectamine RNAiMAX transfection reagent (13778030, Invitrogen).
Cells were treated with recombinant (r) mouse (m) or human (h) protein, peptide or small-molecule inhibitors as detailed in the figure legends and Supplementary Table 3. Transfection with siRNA was performed the day before TIMP1 addition, and cells were pre-treated for 1 h (CD74 antibody, GM6001, STAT3i, AKTi, p38i1, p38i2, p38 control i) or 6 h (CD74 peptide) before TIMP1 treatment.
Clonal VSMC proliferation assay
Cells (EYFP+ VSMCs from aortas of Myh11–EYFP isolated by flow-cytometry-assisted cell sorting or medial cells from Myh11–Confetti animals mixed with medial cells from wild-type animals in a 1:3 ratio) were seeded in 96-well imaging plates (5,000 cells per well, CellCarrier-96 Ultra, Perkin Elmer) in complete medium. Low-titer lentiviral transduction with RUNX1 and control virus was performed 3 days after seeding, or cells were treated as indicated in complete medium from day 4 after seeding. Medium with fresh reagents was added twice weekly. Cells were imaged 4, 7, 14 and 21 days after plating using an Opera Phenix high-content screening system (Perkin Elmer). Image analysis was done using Harmony software v5 (Perkin Elmer), and quantification was performed in Fiji v2.15.1. Patches were defined as three or more contiguous EYFP+RFP+ (RUNX1 experiment) or same-color lineage-traced Confetti+ cells (TIMP1 assays).
ATAC-seq
Single-cell suspensions of carotid arteries from 4 to 5 lineage-labeled Myh11–EYFP/Mki67–RFP animals (11 weeks old) were generated 7 days after carotid ligation surgery, stained for SCA1 and EYFP+SCA1+ or EYFP+SCA1− cells isolated by cell sorting (BD FACSAria III, BD Bioscience). In parallel, lineage-traced (EYFP+) VSMCs were isolated from non-ligated carotid arteries of a non-ligated littermate. The Omni-ATAC protocol48 was used to process 5,000 cells from each sample type in parallel (two independent replicates generated on separate days), using 10–13 PCR amplification cycles and libraries sequenced with a 50 bp paired-end run cycle (Illumina HiSeq or HiSeq2500-RapidRun, 40–60 million reads per sample).
The details of ATAC-seq data analysis are provided in Supplementary Information. Condition-specific peak lists include peaks overlapping at least 50% between biological replicates, and the pan-VSMC peak list is the union of condition-specific lists. Peaks were associated with genomic features using ChIPseeker v.1.24.0. Differential accessibility was scored in LIMMA (SeqMonk v.1.47.2) using a >2-fold change threshold and a Benjamini–Hochberg adjusted P < 0.01. Peaks were annotated to genes and GO term enrichment analysis performed using GREAT v.4.0.4 (ref. 49; http://great.stanford.edu/public/html/index.php) with default settings and all genes as a background. For the SCA1+ condition, the top 4,000 peaks ranked by fold change were used. Motif enrichment analysis was done on 500 bp genomic sequences centered on ATAC-seq peak summits using MEME-ChIP v.5.4.1 (ref. 50) for differentially accessible peaks using HOCOMOCO mouse v.11.
GRN analysis
The scRNA-seq profiles of VSMCs isolated from mouse carotid arteries 5 days after carotid ligation surgery (GSE162167) and the bulk ATAC-seq data generated here (accession number below) were used for GRN modeling using CellOracle v.0.10.5 (ref. 24). A ‘BaseGRN’, including a list of all potential transcription factor-target gene interactions, was constructed using pan-VSMC ATAC-seq peaks. Transcriptional start site (TSS) annotation and motif scan were performed with this union peak data using the default settings of CellOracle. Cells from the day 5 carotid ligation injury scRNA-seq data, clustered as described20, were annotated with VSMC states (cluster 9 as CYC; clusters 4 and 3 as PrP; cluster 2 as LNK; clusters 0, 5, 7 and 8 as non-RSP; cluster 1 as stress; and clusters 6 and 10 as path 2 (Fig. 2a and Extended Data Fig. 2a)) and processed using Scanpy v.1.9.1 (ref. 51). For GRN modeling, the top 4,000 highly variable genes were supplemented with transcription factors showing pseudotime-dependent expression (Eno1, Id3, Mef2c, Hif1a, Nfia, Hmga1, Ebf1, Rora, Klf9, Jund, Foxp1, Thra, Prrx1, Nfix), and Stat3, Twist1 and Runx2.
GRNs were constructed with the top 2,000 interactions for all VSMC states. Topological analyses of GRNs were performed with CellOracle, and GRNs were visualized with Cytoscape v.3.7.2 (ref. 52). GRNs for the stress and path 2 states (colored gray in Fig. 2a) were not investigated further. Venn and Euler diagrams were created using CRAN R package eulerr v.7.0.0. GO term analysis and visualization were performed with BiNGO v.3.0.4 and GOlorize v.1.0.0.beta1 with the default settings, hypergeometric test, Benjamini–Hochberg false discovery rate correction and a significance level of 0.05. Communities were detected using clusterMaker2 v.1.3.1 (ref. 53) with the MCODE algorithm54 (fluff option on).
A rewiring score was calculated for each node as the sum of absolute change in connectivity score between the PrP and non-RSP networks in R v.4.2.2 using igraph v.1.4.2 (https://igraph.org) functions to obtain adjacency matrices. Differential gene expression values for cells in PrP versus non-RSP cell states were calculated using the Wilcoxon rank-sum test (with Bonferroni correction as the method for multiple testing correction), and gene expression values were plotted using Seurat v.4.3.0 (ref. 55) based on sctransform-normalized values20.
In silico simulations were performed with the default settings and top 10,000 interactions in all GRNs with an expression value of zero for the knockout simulations or of 1.5 times the maximum gene expression value for the overexpression simulations. Systematic knockout and overexpression simulations were performed for all TFs present in the GRNs with the same parameters along the proliferation-associated trajectory including cells from the non-RSP, LNK, PrP and CYC populations. TFs were scored based on the sum of positive or negative PSs, which were calculated as the inner product of developmental flow and simulation vectors as described24. Root cell selection for pseudotime calculation was based on Myh11 expression and cell clustering.
Before GRN modeling with the post-carotid ligation day 7 dataset, these data were integrated with the day 5 dataset with sctransform-based normalization using Seurat v.4.3.0 (ref. 55) and further clustered (24 principal components and 1.9 resolution) to identify cell clusters corresponding to equivalent cell states. This information was then used for the GRN modeling.
Analysis of published scRNA-seq datasets
Filtered scRNA-seq profiles (GSE155513) of VSMC-lineage labeled cells from mouse atherosclerotic arteries from 3 time points (8, 16 and 22 weeks) were integrated with sctransform-based normalization (Seurat v.4.3.0)55 and clustered (20 principal components at 0.3 resolution). Differential expression testing for each cell cluster versus all other VSMCs, or for modulated (clusters 3 and 0) versus contractile (cluster 1) cells, was done with the Wilcoxon rank-sum test (adjusted P-value (Padj ) < 0.05, Bonferroni corrected).
Gene set signatures were assessed using UCell v.2.2.0 (ref. 56). PrP-state signature genes include those showing differential expression between PrP and non-RSP cells (log2 fold change (FC) > 0.5, Padj < 0.05; Supplementary Table 1e). ‘Positive regulator’ signatures were defined as TFs with the top 10 highest positive PSs for overexpression simulation or negative PSs for knockout simulation plus the target genes of these TFs in the PrP GRN (edge connectivity > 0.1). ‘Negative regulator’ signatures contained TFs with the top 10 highest negative PSs for overexpression simulation or positive PSs for knockout simulation plus the target genes of these TFs in the PrP GRN (edge connectivity > 0.1).
The scRNA-seq profiles (GSE155512) of human atherosclerotic carotid plaques13 were filtered to include only cells with a minimum of 200 and a maximum of 4,000 genes, total counts >500 and mitochondrial reads <10%, and data from three different patients integrated. Data integration and clustering were repeated after subsetting VSMCs (17 principal components; Seurat v.4.3.0)55. Differential expression testing was performed for each cell cluster against all other VSMCs as above, with log2 fold change (FC) > 0.25 threshold. Heat maps were created with the CRAN R package pheatmap v.1.0.12 with complete linkage and correlation as the distance metric.
Colocalization analysis
Colocalization analyses were done for human orthologs of genes ranking in the top 50 for the rewiring score or top 10 for a positive or a negative PS, plus factors associated with TIMP1 signaling (STAT3, CD74). Summary statistics for CAD GWAS57 were converted from hg19 to hg38 build with CrossMap58 and colocalized with VSMC cis-eQTL from smooth muscle cells isolated from human umbilical cord (N = 1,499)59 and human aortic smooth muscle cell cis-eQTL for quiescent (N = 139) and proliferative (N = 145) cells60. For eCAVIAR, variants with P < 0.001 within 1 mb of the top cis-eQTL association for each gene were overlapped with corresponding GWAS variants (P < 0.001) in the same locus. An eCAVIAR colocalization test was performed for genes that had five or more overlapping variants with GWAS summary data. Linkage between overlapping variants was estimated from individuals of European ancestry in the 1000 Genomes Phase 3 using PLINK61,62. Analysis based on Summary-based Mendelian Randomization (SMR) was done with SMR v1.3.1 (ref. 63) with default settings and peqtl-smr set to 0.001. European samples in the 1000 Genomes project served as reference for linkage disequilibrium estimation. Colocalization events with eCAVIAR colocalization posterior probability (CLPP) > 0.01 or PSMR < 0.05 was considered significant.
Bulk gene expression
RNA was extracted using an RNeasy Mini kit (Qiagen, 74104), and cDNA was generated using QuantiTect Reverse Transcriptase (Qiagen, 205311). Bulk RNA-seq was conducted on RNA isolated from of six different hVSMC isolates following 3 h of serum starvation in serum-free media containing 0.1% BSA, followed by 6 h of TIMP1 (or vehicle) treatment in serum-free media containing 0.1% BSA. Libraries were prepared from oligo-dT-purified mRNA and sequenced (150 bp paired-end reads, Illumina Novaseq 6000). Raw data reads were trimmed with Trim Galore v.0.6.7 and aligned to the human genome (GRCh38) with Kallisto v.0.46.2. Gene set enrichment analysis64 was performed after trimmed mean of M values (TMM) normalization. Quantitative real-time PCR was performed using SsoAdvanced Universal SYBR Green Supermix (Biorad, 1725270).
Protein detection
Whole-cell protein lysates were prepared in RIPA buffer freshly supplemented with proteinase inhibitors (Millipore) and phosphatase inhibitors (Millipore) and separated on gradient (4–12%) polyacrylamide gels. Protein concentration was determined using the BCA method (23227, Pierce BCA protein assay kit, Thermo Fisher). Primary antibodies were detected by HRP-labeled secondary antibodies using chemiluminescence detection (Amersham ECL detection reagent, GE Healthcare) and exposure to photographic film. Phosphokinase array (R&D Systems, ARY003C) was incubated with cell lysates according to the manufacturer’s instructions. Densitometry values of spots were calculated in ImageJ and expressed as arbitrary units.
Immunostaining, EdU incorporation and imaging
Cryosections (14 μm) of ligated left carotid arteries from lineage-labeled Myh11–Confetti animals and plaque-containing arteries from Myh11–Confetti/Apoe animals after high-fat feeding were permeabilized, stained and mounted in RapiClear 1.52 (Sunjin Lab). Confocal imaging was done with a Leica SP8 scanning laser microscope (Leica) using a ×20 lens, and image analysis was done in Imaris v9.2.
Human arteries were formaldehyde fixed and paraffin embedded, and sections (4 μm) were dewaxed and processed for antigen retrieval; sequential sections were either H&E stained, or co-stained for αSMA and either RUNX1, TIMP1 or CD74.
Cells cultured in 96-well plates and treated as indicated in figure legends were fixed and stained. Cells seeded in 96-well imaging plates (10,000 per well) were incubated with EdU (10 µM) for 16 h, and EdU incorporation was detected using the Click-iT EdU kit (C10340, Thermo Fisher Scientific), stained with DAPI for 10 min. Cells were imaged using an Opera Phenix high-content screening system, with a ×20 water objective (Perkin Elmer). Image analysis was done using Harmony software (Perkin Elmer), quantification based on intensity values with nuclei defined by DAPI staining, and for mouse VSMCs from Myh11–EYFP animals, within cell borders (EYFP). Quantification of EdU+ cells was done by thresholding intensity values measured in detected nuclei.
Chromatin immunoprecipitation
Chromatin immunoprecipitation (ChIP) was performed on 3 × 106 hVSMCs using a SimpleChIP kit according to the manufacturer’s instructions (Cell Signaling Technology, 56383). Cells were cross-linked with formaldehyde (1% final concentration, 15 min). Chromatin (5 μg) sheared using a Diagenode bioruptor (15 pulses of 30 s with a 30 s rest) was immunoprecipitated with 10 μl STAT3 antibody (Clone D3Z2G, Cell Signaling Technology, 12640) or an equal amount of control IgG (Cell Signaling Technology, 2729). Antibody-bound protein–DNA complexes were captured using protein-G-coated magnetic beads, reversed cross-linked and digested with proteinase K. DNA was purified and quantified using quantitative (q) PCR.
Flow cytometry
Single-cell suspensions were incubated with 5 µg ml−1 TruStain FcX anti-mouse CD16/32 antibody (Biolegend) in FACS buffer (0.5% (w/v) BSA in PBS) for 15 min on ice, stained with primary antibody for 30 min at room temperature and washed twice (FACS buffer). Samples were then incubated with secondary antibody (in FACS buffer, 30 min at room temperature) and washed twice (FACS buffer), for non-conjugated antibodies. Intracellular targets were stained using the Foxp3 staining buffer set (eBioscience). Cells were analyzed using an Imagestream system (Amnis ImageStream XMk II, Luminex) using IDEAS v6.2 or sorted (BD FACSAria III, BD Bioscience). Gating was done using cells stained with control IgG or non-expressing cell populations.
RNA scope
RNA in situ hybridization was performed using RNA Scope Multiplex Fluorescent v2 kits, according to the manufacturer’s instructions (ACD). Experiments used Hs-TIMP1-C2 coupled to opal 570, Hs-CD74-C3 coupled to opal 620 and Hs-ACTA2-C1 coupled to opal 690 (ACD). Imaging was done using a ZEISS Axioscan slide scanner. Analysis was performed in QuPath using the subcellular detection feature.
Statistical analysis
Group size and replicate numbers were decided on following power calculations based on observed or expected variation using a power of 90% and a 0.05% type I error rate. Where technical repeats were conducted of the same cell line or animal, these have been averaged and biological repeats used for statistical analysis. Data visualization and analysis were conducted using R, the IGV browser and GraphPad Prism V8. Statistical testing information is provided in Supplementary Table 6. Data meeting normality (P > 0.05, Shapiro–Wilk test) and equal variance criteria were subjected to two-sided Student’s t-tests (with F-tests for equal variance) or ANOVA (Brown Forsythe and Bartlett testing for equal variances), unless otherwise stated. When these assumptions were not met, appropriate nonparametric tests were chosen. ANOVA was followed by post hoc testing (Dunnett’s test for comparisons against the control; Tukey’s test for all pairwise comparisons) to calculate multiplicity-adjusted P values. For the analysis of clonal patch formation over time, generalized linear modeling (Poisson regression) was used. Model suitability was assessed by ensuring residual deviance was smaller than the degrees of freedom and checking for over-dispersion, where the variance of the response variable is not greater than the mean as assumed by the Poisson distribution. Diagnostic checks, including examination of deviance residuals and consideration of alternative models, were performed to assess goodness of fit and model adequacy. Clonal patch sizes were analyzed using linear modeling. In cases where data were not normally distributed, log transformation was applied to achieve normality. Normality of residuals was assessed using quantile-quantile (QQ) plotting and the Shapiro–Wilk test. Differential expression and accessibility analyses were done using default tests in LIMMA (two-sided modified t-test, Benjamini–Hochberg correction for multiple testing) and Seurat v.4.3.0 (two-sided nonparametric Wilcoxon rank-sum test, Bonferroni correction). UCell is based on the Mann–Whitney U statistic.
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
Data availability
The ATAC-seq datasets (GSE246646) and bulk RNA-seq data from TIMP1-treated hVSMCs and controls (GSE246647) have been deposited to the NCBI gene expression omnibus (GEO). The scRNA-seq datasets of VSMCs from mouse arteries after injury (GSE162167 (ref. 20)), mouse atherosclerosis (GSE155513 (ref. 13)) and human carotid plaque cells (GSE155512 (ref. 13)) are available from GEO. CAD GWAS57 was downloaded from the GWAS catalog. The cis-eQTL data are VSMC cis-eQTL from smooth muscle cells isolated from human umbilical cord (N = 1,499)59 and human aortic smooth muscle cell cis-eQTL for quiescent (N = 139) and proliferative (N = 145) cells60. Source data are provided with this paper.
Code availability
Data analysis was done using available packages, as listed in Methods. Scripts used for GRN inference, rewiring calculation and analysis of published scRNA-seq datasets are available in GitHub (https://github.com/hfj22/Lambert_Oc_et_al_2024_NCVR).
References
Örd, T. et al. Single-cell epigenomics and functional fine-mapping of atherosclerosis GWAS loci. Circ. Res. 129, 240–258 (2021).
Turner, A. W. et al. Single-nucleus chromatin accessibility profiling highlights regulatory mechanisms of coronary artery disease risk. Nat. Genet. 54, 804–816 (2022).
Liu, B. et al. Genetic regulatory mechanisms of smooth muscle cells map to coronary artery disease risk loci. Am. J. Hum. Genet. 103, 377–388 (2018).
Chappell, J. et al. Extensive proliferation of a subset of differentiated, yet plastic, medial vascular smooth muscle cells contributes to neointimal formation in mouse injury and atherosclerosis models. Circ. Res. 119, 1313–1323 (2016).
Feil, S. et al. Transdifferentiation of vascular smooth muscle cells to macrophage-like cells during atherogenesis. Circ. Res. 115, 662–667 (2014).
Clément, M. et al. Vascular smooth muscle cell plasticity and autophagy in dissecting aortic aneurysms. Arterioscler. Thromb. Vasc. Biol. 39, 1149–1159 (2019).
Chen, P.-Y. et al. Endothelial TGF-β signalling drives vascular inflammation and atherosclerosis. Nat Metab. https://doi.org/10.1038/s42255-019-0102-3 (2019).
Steffensen, L. B. et al. Somatic mutations reveal clonal cell populations in atherosclerotic plaques. Preprint at medRxiv https://doi.org/10.1101/2022.05.16.22275001 (2022).
Lin, A. et al. Clonal expansion in cardiovascular pathology. JACC Basic Transl. Sci. https://doi.org/10.1016/j.jacbts.2023.04.008 (2023).
Misra, A. et al. Integrin beta3 regulates clonality and fate of smooth muscle-derived atherosclerotic plaque cells. Nat. Commun. 9, 2073 (2018).
Sheikh, A. Q., Saddouk, F. Z., Ntokou, A., Mazurek, R. & Greif, D. M. Cell autonomous and non-cell autonomous regulation of SMC progenitors in pulmonary hypertension. Cell Rep. 23, 1152–1165 (2018).
Winther et al. Translational opportunities of single-cell biology in atherosclerosis. Eur. Heart J. 44, 1216–1230 (2022).
Pan, H. et al. Single-cell genomics reveals a novel cell state during smooth muscle cell phenotypic switching and potential therapeutic targets for atherosclerosis in mouse and human. Circulation 142, 2060–2075 (2020).
Wirka, R. C. et al. Atheroprotective roles of smooth muscle cell phenotypic modulation and the TCF21 disease gene as revealed by single-cell analysis. Nat. Med. 25, 1280–1289 (2019).
Conklin, A. C. et al. Meta-analysis of smooth muscle lineage transcriptomes in atherosclerosis and their relationships to in vitro models. Immunometabolism 3, e210022 (2021).
Dobnikar, L. et al. Disease-relevant transcriptional signatures identified in individual smooth muscle cells from healthy mouse vessels. Nat. Commun. 9, 4567 (2018).
Wang, Y. et al. Dynamic changes in chromatin accessibility are associated with the atherogenic transitioning of vascular smooth muscle cells. Cardiovasc. Res. 118, 2792–2804 (2021).
Cherepanova, O. A. et al. Activation of the pluripotency factor OCT4 in smooth muscle cells is atheroprotective. Nat. Med. 22, 657–665 (2016).
Wang, Y. et al. Clonally expanding smooth muscle cells promote atherosclerosis by escaping efferocytosis and activating the complement cascade. Proc. Natl Acad. Sci. USA https://doi.org/10.1073/pnas.2006348117 (2020).
Worssam, M. D. et al. Cellular mechanisms of oligoclonal vascular smooth muscle cell expansion in cardiovascular disease. Cardiovasc. Res. 119, 1279–1294 (2022).
Yoshida, T., Kaestner, K. H. & Owens, G. K. Conditional deletion of Krüppel-like factor 4 delays downregulation of smooth muscle cell differentiation markers but accelerates neointimal formation following vascular injury. Circ. Res. 102, 1548–1557 (2008).
Owsiany, K. M., Deaton, R. A., Soohoo, K. G., Nguyen, A. T. & Owens, G. K. Dichotomous roles of smooth muscle cell-derived MCP1 (Monocyte Chemoattractant Protein 1) in development of atherosclerosis. Arterioscler. Thromb. Vasc. Biol. 42, 942–956 (2022).
Worssam, M. D. & Jørgensen, H. F. Mechanisms of vascular smooth muscle cell investment and phenotypic diversification in vascular diseases. Biochem. Soc. Trans. 49, 2101–2111 (2021).
Kamimoto, K. et al. Dissecting cell identity via network inference and in silico gene perturbation. Nature 614, 742–751 (2023).
Kumar, A. & Lindner, V. Remodeling with neointima formation in the mouse carotid artery after cessation of blood flow. Arterioscler. Thromb. Vasc. Biol. 17, 2238–2244 (1997).
McDonald, O. G., Wamhoff, B. R., Hoofnagle, M. H. & Owens, G. K. Control of SRF binding to CArG box chromatin regulates smooth muscle gene expression in vivo. J. Clin. Invest. 116, 36–48 (2005).
Gomez, D., Shankman, L. S., Nguyen, A. T. & Owens, G. K. Detection of histone modifications at specific gene loci in single cells in histological sections. Nat. Methods 10, 171–177 (2013).
Shankman, L. S. et al. KLF4-dependent phenotypic modulation of smooth muscle cells has a key role in atherosclerotic plaque pathogenesis. Nat. Med. 21, 628–637 (2015).
Neggazi, S. et al. Thyroid hormone receptor alpha deletion in ApoE–/– mice alters the arterial renin–angiotensin system and vascular smooth muscular cell cholesterol metabolism. J. Vasc. Res. 55, 224–234 (2018).
Silence, J., Collen, D. & Lijnen, H. R. Reduced atherosclerotic plaque but enhanced aneurysm formation in mice with inactivation of the tissue inhibitor of metalloproteinase-1 (TIMP-1) gene. Circ. Res. 90, 897–903 (2002).
Sundström, J. et al. Relations of plasma total TIMP-1 levels to cardiovascular risk factors and echocardiographic measures: the Framingham heart study. Eur. Heart J. 25, 1509–1516 (2004).
Akahane, T., Akahane, M., Shah, A. & Thorgeirsson, U. P. TIMP-1 stimulates proliferation of human aortic smooth muscle cells and Ras effector pathways. Biochem. Biophys. Res. Commun. 324, 440–445 (2004).
Forough, R. et al. Overexpression of tissue inhibitor of matrix metalloproteinase-1 inhibits vascular smooth muscle cell functions in vitro and in vivo. Circ. Res. 79, 812–820 (1996).
Ebert, S. et al. Tissue inhibitor of metalloproteinases-1 interacts with CD74 to promote AKT signaling, monocyte recruitment responses, and vascular smooth muscle cell proliferation. Cells 12, 1899 (2023).
Song, G. et al. TIMP1 is a prognostic marker for the progression and metastasis of colon cancer through FAK-PI3K/AKT and MAPK pathway. J. Exp. Clin. Cancer Res. 35, 148 (2016).
Schoeps, B., Frädrich, J. & Krüger, A. Cut loose TIMP-1: an emerging cytokine in inflammation. Trends Cell Biol. 33, 413–426 (2023).
Grünwald, B., Schoeps, B. & Krüger, A. Recognizing the molecular multifunctionality and interactome of TIMP-1. Trends Cell Biol. 29, 6–19 (2019).
Tesoriere, A., Dinarello, A. & Argenton, F. The roles of post-translational modifications in STAT3 biological activities and functions. Biomedicines 9, 956 (2021).
Schoeps, B. et al. Identification of invariant chain CD74 as a functional receptor of tissue inhibitor of metalloproteinases-1 (TIMP-1). J. Biol. Chem. 297, 101072 (2021).
Xue, Y. et al. Macrophages regulate vascular smooth muscle cell function during atherosclerosis progression through IL-1β/STAT3 signaling. Commun. Biol. 5, 1316 (2022).
Figueiredo, C. R. et al. Blockade of MIF–CD74 signalling on macrophages and dendritic cells restores the antitumour immune response against metastatic melanoma. Front. Immunol. 9, 1132 (2018).
Chandesris, M.-O. et al. A neuropathological study of cerebrovascular abnormalities in a signal transducer and activator of transcription 3-deficient patient. J. Allergy Clin. Immunol. 136, 1418–1421.e5 (2015).
Chandesris, M.-O. et al. Frequent and widespread vascular abnormalities in human signal transducer and activator of transcription 3 deficiency. Circ. Cardiovasc. Genet. 5, 25–34 (2012).
Riddell, A. et al. RUNX1: an emerging therapeutic target for cardiovascular disease. Cardiovasc. Res. 116, 1410–1423 (2020).
Koth, J. et al. Runx1 promotes scar deposition and inhibits myocardial proliferation and survival during zebrafish heart regeneration. Development 147, dev186569 (2020).
Gregoli, K. D., George, S. J., Jackson, C. L., Newby, A. C. & Johnson, J. L. Differential effects of tissue inhibitor of metalloproteinase (TIMP)-1 and TIMP-2 on atherosclerosis and monocyte/macrophage invasion. Cardiovasc. Res. 109, 318–330 (2016).
Lijnen, H. R., Soloway, P. & Collen, D. Tissue inhibitor of matrix metalloproteinases-1 impairs arterial neointima formation after vascular injury in mice. Circ. Res. 85, 1186–1191 (1999).
Corces, M. R. et al. An improved ATAC-seq protocol reduces background and enables interrogation of frozen tissues. Nat. Methods 14, 959–962 (2017).
McLean, C. Y. et al. GREAT improves functional interpretation of cis-regulatory regions. Nat. Biotechnol. 28, 495–501 (2010).
Ma, W., Noble, W. S. & Bailey, T. L. Motif-based analysis of large nucleotide data sets using MEME-ChIP. Nat. Protoc. 9, 1428–1450 (2014).
Wolf, F. A., Angerer, P. & Theis, F. J. SCANPY: large-scale single-cell gene expression data analysis. Genome Biol. 19, 15 (2018).
Shannon, P. et al. Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res. 13, 2498–2504 (2003).
Morris, J. H. et al. clusterMaker: a multi-algorithm clustering plugin for Cytoscape. BMC Bioinformatics 12, 436 (2011).
Bader, G. D. & Hogue, C. W. An automated method for finding molecular complexes in large protein interaction networks. BMC Bioinformatics 4, 2 (2003).
Hao, Y. et al. Integrated analysis of multimodal single-cell data. Cell 184, 3573–3587.e29 (2021).
Andreatta, M. & Carmona, S. J. UCell: robust and scalable single-cell gene signature scoring. Comput. Struct. Biotechnol. J. 19, 3796–3798 (2021).
Aragam, K. G. et al. Discovery and systematic characterization of risk variants and genes for coronary artery disease in over a million participants. Nat. Genet. 54, 1803–1815 (2022).
Zhao, H. et al. CrossMap: a versatile tool for coordinate conversion between genome assemblies. Bioinformatics 30, 1006–1007 (2014).
Solomon, C. U. et al. Effects of coronary artery disease-associated variants on vascular smooth muscle cells. Circulation 146, 917–929 (2022).
Aherrahrou, R. et al. Genetic regulation of SMC gene expression and splicing predict causal CAD genes. Circ. Res. 132, 323–338 (2023).
Auton, A. et al. A global reference for human genetic variation. Nature 526, 68–74 (2015).
Chang, C. C. et al. Second-generation PLINK: rising to the challenge of larger and richer datasets. Gigascience 4, 16 (2015).
Zhu, Z. et al. Integration of summary data from GWAS and eQTL studies predicts complex trait gene targets. Nat. Genet. 48, 481–487 (2016).
Subramanian, A. et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc. Natl Acad. Sci. USA 102, 15545–15550 (2005).
Acknowledgements
We acknowledge the National Institute for Health Research Cambridge Biomedical Research Centre Cell Phenotyping Hub for cell sorting, D. Clements at the MRC/Wellcome Stem Cell Institute Advanced Imaging Facility for assistance with confocal imaging, J. Arnold and J. Jones at the CRUK for RNA scope hybridization, the Babraham Institute Sequencing Facility (UKRI-BBSRC Core Capability Grant) and S. Andrews and F. Krueger (Babraham Institute’s Bioinformatics) for assistance with ATAC-seq data processing, J. Palmer and A. Krishnakumar for help with image interpretation, and P. Johnson and L. Kitt for help with animal models. We thank the donors, their families, the specialist nurses on organ donation (SNOD) and the Papworth Tissue Bank team for human tissue samples. GRN analyses were done using the high-performance computer clusters at the University of Cambridge and Imperial College Research Computing Service (https://doi.org/10.14469/hpc/2232). This work was supported by the British Heart Foundation (PG/19/6/34153, FS/15/62/32032, FS/15/38/31516, RM/13/3/30159, RE/13/ 6/30180, RE/18/1/34212, CH/2000003/12800, PG/23/11350, M.R.B., H.F.J., M.S.; RG/16/13/32609, RG/19/9/34655, PG/16/9/31995, PG/18/73/34059, SP/19/2/344612, S.Y., T.R.W.; AA/18/3/34220, C.U.S.); the Cambridge NIHR Biomedical Research Centre; the Deutsche Forschungsgemeinschaft, Bonn, Germany (KR2047/8-2, KR2047/14-1, KR2047/15-1, KR2047/16-1, A.K.); the Medical Research Council of the UK (MC-A652-5QA20, M.S., core-funded investigator); Singapore’s National Medical Research Council (CIRG22jul‐0002, S.Y.); National University of Singapore and National University Health System (NUHSRO/2022/004/Startup/01, S.Y.); the National Institute for Health Research Leicester Biomedical Research Centre; and the Chan Zuckerberg Initiative (2018-190766/RG98793; the Collaborative Bioresource for Translational Medicine, K.T.M.).
Author information
Authors and Affiliations
Contributions
J.L. and S.O. designed and performed experiments and data analysis, and wrote the paper. M.D.W. generated and analyzed the ATAC-seq profiling data and edited the paper. D.H. and A.K. generated material, designed experiments and edited the paper. R.B. performed experiments. M.I. and J.C.K.T. performed data analysis. N.L.F., K.F. and A.F. performed experiments supervised by M.R.B. K.T.M. retrieved human tissue samples. C.U.S. performed colocalization analyses designed by T.R.W. and supervised by S.Y. M.S. designed and supervised the network analysis and wrote the paper. H.F.J. conceived the study, supervised and performed experiments and data analysis, and wrote the paper.
Corresponding author
Ethics declarations
Competing interests
M.S. is a shareholder of Enhanc3D Genomics Ltd. H.F.J. is a key opinion leader for Novo Nordisk A/S. The other authors declare no competing interests.
Peer review
Peer review information
Nature Cardiovascular Research thanks Muredach Reilly and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Extended data
Extended Data Fig. 1 Quality assessment of ATAC-seq data and gene ontology analysis of differentially accessible genes.
a, Scatter plot for two independent biological replicates of ATAC-seq profiling for SCA1+ cells showing high correlation between replicates. Colors show point density on a blue (low)-to-red (high) scale. b-c, Hierarchical clustering (b) and principal component analysis (c) of ATAC-seq samples, confirming similarity of replicate samples for all conditions. d, Significantly enriched gene ontology (GO) biological process terms for genes associated with peaks that are more accessible in the control samples from healthy animals, compared to SCA1+ cells isolated from injured arteries. e-g, Significantly enriched GO biological process (e), disease (f) and signaling pathway terms (g) for genes associated with peaks that are more accessible in SCA1+ cells isolated from injured arteries compared to the control samples from healthy animals. Significance was assessed with binomial testing (binom) in the GREAT package (http://great.stanford.edu/public/html/).
Extended Data Fig. 2 GRN analysis supplement.
a, Force-directed graph of scRNA-seq profiles of VSMCs analyzed 5 days after injury showing the cell clusters and normalized, log-transformed expression of markers characteristic of different VSMC states and selected genes. b, Degree distributions for constructed GRNs. c, Community detection in the union of all four networks, coloured by gene ontology terms as in Fig. 2c. d, Nodes with top (30) degree centrality scores for the equivalent GRNs constructed with scRNA-seq profiles of VSMCs analyzed 7 days after injury. Supplementary Fig. 6 shows panel d at enlarged size.
Extended Data Fig. 3 Gene expression changes of GRN nodes in human and murine atherosclerosis scRNA-seq datasets.
a-d, Analysis of scRNA-seq profiles of VSMC lineage-traced cells from a murine atherosclerosis model (GSE155513). a, UMAP showing cluster annotation for cells isolated after 8, 16 or 22 weeks. b, Dot plot showing expression across VSMC clusters for contractile genes (Myh11, Acta2), markers of ‘minor SMCs’ (Rgs5, Pan et al., 2020), genes expressed by synthetic VSMCs (Spp1, Col8a1, Tnfrsf11b), a VSMC transition state (Vcam1, Ly6a) and fibroblast-like (F3), fibrochondrocyte (Sox9) or macrophage-like (C1qc) VSMCs. Scaled expression levels are shown on a gray-to-blue scale and dot size represent the percentage of expressing cells. c-d, UMAP showing normalized, log-transformed expression of indicated genes (c) or the UCell signature enrichment scores for the 10 top trajectory-stimulating (Positive regulators) or trajectory-blocking (Negative regulators) transcription factors (ranked by perturbation score), and associated direct target genes (positive interactions, connectivity score>0.1, d). e-f, Analysis of VSMC profiles from human carotid plaque scRNA-seq dataset (GSE155512). e, Dot plot showing gene expression across VSMC subclusters for contractile VSMC markers (MYH11, ACTA2, TAGLN, CNN1), genes associated with phenotypically modulated VSMCs (COL8A1, TNFRSF11B, MMP2, MGP, VCAM1) and genes associated with different VSMC clusters (CCL2, RPS6, RPS12). Scaled expression levels are shown on a gray to blue scale and dot size represent the percentage of expressing cells. f, Normalized, log-transformed expression of selected genes is shown on a gray-blue scale on a UMAP.
Extended Data Fig. 4 In silico network simulation results.
a, The RUNX, CEBP and NFκB motifs were detected adjacent to the AP-1 motif in peaks showing increased accessibility in SCA1+ cells after injury compared to control VSMCs from no-surgery animals. b, Schematic explaining how perturbation scores were generated. Left panels show vector fields for inferred cell state trajectory (i) and results of in silico perturbation of transcription factor levels (ii) on a force directed graph representation. Right panels show the computed cell perturbation score on a force directed graph representation (iii) or as a function of pseudotime (iv). Negative (pink) and positive scores (green) were summarized separately for cells in the Non-RSP, LNK, PrP and CYC states. (c, d) Force-directed graph of VSMCs isolated 5 days after injury, showing perturbation scores from in silico knockout (KO, c) or overexpression (OE, d) simulation for the indicated transcription factors with the color-coding described in b.
Extended Data Fig. 5 Analysis of RUNX1 as a candidate VSMC regulator.
a, ATAC-seq data tracks for the Runx1 locus, annotated as described in Fig. 1a. b, RUNX1 immunostaining in cryosections from the injured left carotid artery of a Myh11-Confetti animal, Scale bar= 50 µm. Representative of 2 animals. Signals for RUNX1 (magenta), nuclear DAPI (white) and confetti lineage-labeling are shown as indicated. c, Imagestream analysis of lineage-labeled EYFP+ (green) VSMCs from Myh11-EYFP animals stained for SCA1 (red) and MMP14 (magenta). Left columns show brightfield images. Representative of 3 animals. d, Runx1 and Cebpd transcript levels detected by quantitative RT-PCR in mVSMCs transfected with non-targeting (NTC) or Runx1-targeting siRNA (Runx1-siRNA), or transduced with an empty vector (CTRL-EV) or RUNX1-overexpressing (OE) lentivirus. Dots show values for cells from independent animals (N = 5), lines mean, error bars SEM, p-value: two-tailed MWU (Runx1) or two-tailed t-test (Cebpd). e-f, Immunofluorescence staining of RUNX1 in mVSMCs (e) and hVSMCs (f) transfected with non-targeting (NTC) or Runx1-targeting siRNA (Runx1-siRNA), or transduced with an empty vector (CTRL-EV) or RUNX1-overexpressing (OE) lentivirus. Scale=100 µM. Graph shows quantification of nuclear RUNX1 signals. Dots show values for cells from individual animals (E, N = 4, p-value: two-sided MWU) or hVSMC isolates from different donors (F, N = 3, p-value: two-sided t-test), lines mean, error bars SEM, p-value: two-tailed t-test. g, Representative images (N = 5) of non-plaque aorta with intimal thickening showing sequential sections after H&E (left) or immunostaining for αSMA (blue) and RUNX1 (brown). Scale bars = 100 µm. Lumen, intima, media and adventitia are indicated.
Extended Data Fig. 6 Impact of TIMP1 on VSMCs.
a, Example immunohistochemistry images for α-smooth muscle actin (αSMA, blue) and TIMP1 (brown) in non-plaque human aorta. Scale bar=500 µm. N = 7 donors. b, Representative images of TIMP1 immunostaining in cryosections of an atherosclerotic plaque from Myh11-Confetti/Apoe animals fed a HFD (right, N = 3) or an injured Myh11-Confetti carotid artery (left, N = 3 animals). Signals for TIMP1 (magenta), nuclear DAPI (white) and confetti lineage-labeling (CFP: blue, RFP: red, YFP: yellow, GFP: green) are shown as indicated. White box indicate region shown in Fig. 6b. Scale bar = 50 μm (left), 20 μm (right). c, Enrichment plots from gene set enrichment analysis (GSEA) for E2F targets (padj = 0.005), G2M checkpoint genes (padj= 0.04), adipogenesis (padj = 0.023) bile acid metabolism (padj = 0.00376), fatty acid metabolism padj = 0.00337) and oxidative phosphorylation (padj = 0.022) in bulk RNA-seq data from hVSMCs treated with 500 ng/mL recombinant human TIMP1 for 6 hours versus control cells (N = 6 independent human VSMC isolates/donors).
Extended Data Fig. 7 TIMP1 induces phosphorylation of STAT3 in VSMCs.
a, Heatmap showing relative spot intensity of phosphokinase array analysis of hVSMCs following 15 minutes 500 ng/mL rhTIMP1 or vehicle control treatment on a black (low) to red (high) scale [AU]. N = 1. b, Western blots of hVSMCs from 4 different donors showing total, pS727 and pTyr705 STAT3, total and phospho AKT, total and phospho-p38 (p-p38), and GAPDH in serum-starved cells after 0, 5 10, 15 or 30 minutes rhTIMP1 treatment. c, Quantification of relative band intensity in (b) for total STAT3, total AKT and total p38 protein levels, normalized to GAPDH. Points show independent hVSMC isolates (N = 4 donors), bars mean, error bars SEM. d, pSTAT3 S727 immunostaining (magenta) in cryosections from injured carotid artery of Myh11-Confetti (top, N = 3 animals) or atherosclerotic plaque from Myh11-Confetti/Apoe animals fed a HFD (lower panel, N = 3). Signals for nuclear DAPI (white) and confetti lineage-label are also shown. White box indicate region shown in Fig. 6g. Scale bar= 50 μm. e, %KI67+ hVSMCs in samples treated with 500 ng/mL rhTIMP1 following serum-starvation for indicated timepoint. Symbols show average values for independent hVSMC isolates (N = 3, analyzed in triplicate), lines the mean, error bars SEM. f, Nuclear pSTAT3 (S727) intensity in FACS-isolated EYFP+ mVSMCs without (CTRL, dots) or with 500 ng/mL rmTIMP1 (squares) either 4 (left) or 7 days post seeding (right). Symbols show average values from independent (N = 4 Myh11-EYFP animals analyzed in quadruplicate), lines mean, error bars SEM, p-value: two-tailed t-test.
Extended Data Fig. 8 CD74 expression and function in VSMCs.
a, Example images of immunostaining for CD74 (magenta) in cryosections of Myh11-Confetti lineage-labeled carotid arteries, analyzed 28 days (left) or 10 days post carotid ligation (right). Arrows mark CD74+ lineage labeled VSMCs. Scale bars: 100 µm (overview), 30 µm (zoomed). N = 5 animals. b, Bar graph showing the percentage of EdU+ lineage-labeled VSMCs in samples infected with lentivirus carrying an empty vector (EV) or overexpressing RUNX1 (OE) in the presence of a control (IgG) or a CD74 blocking antibody (CD74). VSMC-lineage labeled medial cells were isolated from mouse aorta. Points show values for different animals (N = 8), bars the mean, error bars SEM, p-value: one-way ANOVA. c, Example imagestream analysis of lineage-labeled EYFP+ (green) VSMCs from Myh11-EYFP/Apoe animals stained for KI67 (orange), pSTAT3 S727 (magenta) and CD74 (red). Left columns show brightfield images. d, Example images of CD74 (brown) and αSMA (blue) immunostaining in non-plaque human aorta. Scale bar=500 µm. N = 7 tissue samples from different donors.
Supplementary information
Supplementary Information
Supplementary Discussion, Methods, References (also included in the main text) and Figs. 1–6.
Supplementary Tables
Supplementary Tables 1–6.
Supplementary Data 1
List of sequences used for filtering ATAC-seq data.
Source data
Source Data Fig. 1
Zipped folder with peak information and statistical source data.
Source Data Fig. 2
Panels 2d+e+f (font size).
Source Data Fig. 3
Panels 3c+f (font size).
Source Data Fig. 4
Zipped folder with motif analysis output.
Source Data Fig. 5
Zipped folder with statistical source data and Fig. 5c (font size).
Source Data Fig. 6
Statistical source data.
Source Data Fig. 7
Zipped folder with statistical source data and unprocessed western blots.
Source Data Fig. 8
Zipped folder with statistical source data and unprocessed western blots.
Source Data Extended Data Fig. 1
GO analysis results (GREAT).
Source Data Extended Data Fig. 2
Panel ExtData 2d (font size).
Source Data Extended Data Fig. 5
Statistical source data.
Source Data Extended Data Fig. 7
Statistical source data.
Source Data Extended Data Fig. 8
Statistical source data.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Lambert, J., Oc, S., Worssam, M.D. et al. Network-based prioritization and validation of regulators of vascular smooth muscle cell proliferation in disease. Nat Cardiovasc Res 3, 714–733 (2024). https://doi.org/10.1038/s44161-024-00474-4
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1038/s44161-024-00474-4
- Springer Nature Limited