Smad3 regulates smooth muscle cell fate and mediates adverse remodeling and calcification of the atherosclerotic plaque

Abstract

Atherosclerotic plaques consist mostly of smooth muscle cells (SMC), and genes that influence SMC phenotype can modulate coronary artery disease (CAD) risk. Allelic variation at 15q22.33 has been identified by genome-wide association studies to modify the risk of CAD and is associated with the expression of SMAD3 in SMC. However, the mechanism by which this gene modifies CAD risk remains poorly understood. Here we show that SMC-specific deletion of Smad3 in a murine atherosclerosis model resulted in greater plaque burden, more outward remodelling and increased vascular calcification. Single-cell transcriptomic analyses revealed that loss of Smad3 altered SMC transition cell state toward two fates: a SMC phenotype that governs both vascular remodelling and recruitment of inflammatory cells, as well as a chondromyocyte fate. Together, the findings reveal that Smad3 expression in SMC inhibits the emergence of specific SMC phenotypic transition cells that mediate adverse plaque features, including outward remodelling, monocyte recruitment, and vascular calcification.

INTRODUCTION

Decades of research and drug development have led to therapies and interventions that have significantly diminished morbidity and mortality from cardiovascular disease ^1,2. However, coronary artery disease (CAD) remains the leading cause of death in this country and worldwide with a sharp decline in the rate of improvement in mortality observed over the past decade ³. Recent clinical studies targeting well-characterized risk factors such as lipids ^4,5 and molecular targets related to inflammation ^6,7 have had only modest success ⁸, suggesting a continued need to identify additional disease modifiers. Over the past decade, genome wide association studies (GWAS) have identified over 160 loci that contribute to CAD risk ^9,10. Causal variation identified in these loci point primarily to genes and pathways predicted to function in the blood vessel wall to regulate disease risk ^11–13.

Specific features of atherosclerotic plaque have been increasingly recognized to offer significant prognostic value. For instance, it has been noted that cellular composition, such as SMC contribution to the fibrous cap, influences risk of plaque rupture ^14–16. With advances in diagnostic imaging, disease features such as outward remodeling and microcalcification have been found to be highly predictive for myocardial infarctions ^17,18. Although counter-intuitive, studies using both intravascular ultrasound (IVUS), and longitudinal studies using CT coronary angiography ^17,18, have demonstrated that sites of outward remodeling are much more likely to be the culprit site for plaque rupture and myocardial infarction compared with sites with more luminal narrowing. However, little is known regarding the cellular and molecular mechanisms by which these high-risk features are controlled.

A number of post-genomic GWAS follow-up studies in this and other laboratories investigating the genetic disease-related mechanisms of CAD have focused on SMC, and the relationship of SMC cell state changes to disease risk ^19–23. These studies have indicated that a significant portion of the CAD attributable risk is determined by this cell type ¹³. Consistent with this notion, recent lineage tracing studies have demonstrated that the majority of cells inside atherosclerotic plaque, including those expressing some inflammatory markers, are oligo-clonal de-differentiated smooth muscle derivatives ^24–27. Genes that alter SMC behavior are known to influence the composition of atherosclerotic plaque ^{19,22,23,26,28,29}. CAD-associated genes TCF21 and AHR have been linked to these processes, and various genomic data suggests that additional CAD genes might also regulate SMC phenotype ^{21,28,30–33}. While these SMC progeny have been previously lumped together as phenotypically modulated SMC, advances in single cell RNA profiling has demonstrated the presence of subsets of these cells with distinct transcriptomes and cell fates. For example, medial SMC have been shown to give rise to fibroblast-like cells termed fibromyocytes ²³, as well as cells similar to endochondral bone forming cells, chondromyocytes ^19,31, among other bioinformatically defined populations ^19,34. However, the functional significance and relative location of these transcriptionally distinct populations remain to be elucidated.

The TGFβ signaling pathway is central to smooth muscle biology during development and disease, and an important modifier of atherosclerosis ^35,36. Canonical TGFβ signaling is thought to be mediated through Smad family proteins, particularly nuclear signaling factors Smad2 and Smad3. While themselves poor binders to DNA ³⁷, through their interaction with other transcription factors, the Smad factors are central to key transcriptional programs that regulate cell fate in development and disease ^38,39. Multiple GWAS have identified rs17293632 ⁹, a single nucleotide variant that lies within a functional smooth muscle enhancer that regulates SMAD3 expression ²⁰, as an important modifier of risk for myocardial infarction. However, how smooth muscle SMAD3 expression influences risk of myocardial infarction is unclear. In vitro, SMAD3 appears to modify smooth muscle cell differentiation and proliferation through its interaction with other transcription factors critical to SMC biology and risk of CAD ³⁰. However, the exact effects of SMAD3 expression level on SMC plaque biology remain unknown.

Here, we demonstrate that SMC-specific deletion of Smad3 influences the fate of de-differentiated SMC in atherosclerotic plaques in vivo, promoting both a pro-remodeling SMC transition phenotype that expresses remodeling genes such as Mmp3 and inflammatory chemokines such as Cxcl12, as well as an expansion of the SMC-derived chondromyocyte (CMC) population. These cellular changes are associated with increased outward remodeling and plaque calcification that appear to normally be inhibited by Smad3 in conjunction with transcriptional effects of Hox and Sox factors.

RESULTS

Vascular lesion characterization in Smad3^ΔSMC mice

To understand how smooth muscle expression of Smad3 influences atherosclerotic lesions in vivo, we generated a murine model of atherosclerosis with established smooth-muscle specific Cre (Myh11-Cre) crossed with a conditional knockout allele of Smad3 ^40–42 with concurrent lineage tracing provided by conditional tandem dimer tomato (ROSA^Tdt) expression on the ApoE null background (Smad3^ΔSMC) (Fig. 1a). To limit confounding created by the critical role of Smad3 during development, the Smad3 gene was deleted via a tamoxifen inducible Cre only after mice had reached maturity (8-weeks-old), immediately prior to initiation of a Western high-fat diet (HFD, Fig. 1a). The Smad3 conditional knockout mice grew to maturity with no significant change in weight or mortality compared with control (Extended Data Fig. 1a, b). In these studies, evaluation of Smad3 function in atherosclerotic vascular disease was evaluated primarily in the aortic root (Extended Data Fig. 1c), where scRNAseq data revealed highly efficient Cre-mediated deletion of the floxed exons encoding the DNA binding domain of Smad3 (Extended Data Fig. 1d), with corresponding loss of nuclear phospo-Smad3 protein in knockout SMC verified by lesion immunohistochemistry (Extended Data Fig. 1e, f). Investigation of the phenotype of atherosclerotic disease lesions were conducted according to published recommendations on the design, execution, and reporting of animal atherosclerosis studies⁴³.

Figure 1. Smad3^ΔSMC mice have increased lesion burden in the ApoE null model.

(a) Representative figure of mouse protocol showing SMC-specific lineage tracing and Smad3 conditional knockout (Smad3^ΔSMC). (b) Smad3^ΔSMC SMC transition cells can contribute to all regions of the lesion, including the lesion cap (arrow), modulated SMC (*), and tunica media (m). Scale bar 250 um. (c) Representative sections from control and Smad3^ΔSMC mice stained for Tagln to highlight cap and tunica media for lesion quantification. Scale bar 250 um. (d) Lesion area between the lumen and internal elastic lamina, two-talied t-test. (e) Lumen area, two-talied t-test. (f) Vessel area quantified as area between lumen and external elastic lamina, across experimental animals, two-talied t-test. (g) Representative sections from control and Smad3^ΔSMC mice revealing tdT fluorescence to show lineage traced cells in the media and plaque. Scale bar 250 um. (h) Quantification of tdT positive area in control and Smad3^ΔSMC mice, two-talied t-test. (i) Representative sections from control and Smad3^ΔSMC mice stained with CD68 antibody. Scale bar 250 um. (j) Quantification of CD68 positive area, two-talied t-test. Each dot on each bar graph represents mean of data measured from a single matched section from each individual animal, error bars represent 95% CI of mean.

Examination of atherosclerotic lesions in the aortic root after 16 weeks of HFD demonstrated that SMC from Smad3^ΔSMC mice were able to migrate into the lesion, expand, and contribute to the formation of atherosclerotic plaque and the fibrous cap (Fig. 1b). Quantification of atherosclerotic lesions revealed a significant increase in plaque volume in Smad3^ΔSMC compared with control animals (Fig. 1c, d). To further characterize the anatomy of diseased vessels, specifically regarding outward remodeling vs luminal narrowing, we evaluated the area encapsulated by the diseased vessel as well as lumen area. The lumen area in these sections showed no significant change (Fig. 1e), but the area circumscribed by the external elastic lamina was significantly increased (Fig. 1f). These findings are consistent with expansion of atherosclerotic plaque volume in conjunction with outward remodeling. To determine the cellular anatomy associated with the increased plaque size, we quantified the area occupied by fluorescent tdTomato SMC lineage traced cells (Fig. 1g, h) as well as CD68 stained monocytes and macrophages in the lesions (Fig. 1i, j). This analysis revealed a statistically significant increase in area for both SMC-derived cells as well as cells of the monocyte-macrophage lineage. Given that the lineage tracing Myh11-Cre transgene is active only in cells emanating from mature SMC, these findings suggest that the increased lesion growth has both a cell autonomous as well as a non-autonomous cellular component, with the latter reflecting an SMC mediated effect on monocyte-macrophage lesion recruitment.

Contents of the plaques were further analyzed via Oil Red O, trichrome, and Ter119 staining. There was an increase in the absolute area stained with Oil Red O in the Smad3^ΔSMC compared with control animals (Extended Data Fig. 2a, b), but there was not a difference noted when the lipid-stained area was corrected for total plaque area (Extended Data Fig. 2c), suggesting the plaques were larger but did not contain proportionally more lipid. There was also an increase in absolute acellular area in Smad3^ΔSMC compared with control animals as determined with trichrome staining, and this difference was sustained when corrected for plaque size (Extended Data Fig. 2d, e, f). This increase in acellular area at the base of the plaque was consistent with the higher risk features seen in outward remodeling.

Vascular lesion single cell RNA sequencing in Smad3^ΔSMC mice

Given the critical role that Smad3 plays in cell fate decisions during development, we hypothesized that alteration in disease-associated SMC phenotype transitions might account for the observed cellular lesion characteristics as well as recruitment of CD68 positive cells and outward remodeling. Thus, to better understand how loss of Smad3 expression produced phenotypic changes in lesion SMC derived cells, and their interactions with other lesion cell types, we performed single cell RNA expression profiling (scRNAseq) of atherosclerotic lesions from Smad3^ΔSMC and control animals. The atherosclerotic tissue was harvested, processed for single cell encapsulation, RNA capture, reverse transcription and amplification with the 10X Genomics Chromium V3 platform, and cDNA libraries were sequenced as previously described ^23,28. Subsequently, the single cell expression data were visualized utilizing Uniform Manifold Approximation and Projection (UMAP) to create a 2-D projection representing the organization of individual cells to each other in clusters, and of the relationship of the organized clusters to each other. After filtering and normalization, a total of 26,219 cells from control and 35,518 cells from Smad3^ΔSMC mice were included in the analysis obtained from 3 and 4 independent captures of pooled tissue from 2 mice in each individual capture. Featureplots were employed to visualize the SMC lineage traced cells (Fig. 2a). SMC traced cells were identified by expression of the tdTomato gene, and the component of this cluster representing mature medial SMC was identified by expression of SMC lineage markers Myh11 and Cnn1. As we have shown previously, a significant portion of the SMC-lineage traced cells did not express these mature SMC markers during disease and thus represented SMC that had undergone phenotypic transition ^22,23,28. We were thus able to determine whether there were alterations in the number of mature differentiated versus transition SMC in Smad3^ΔSMC mice. We investigated whether there was an alteration in the number of differentiated SMC among the lineage traced (tdTomato+) cells, defining “differentiated” cells using classical Myh11 or Cnn1 expression, along with unbiased clustering (SMC vs transition SMC) (Fig. 2a, b). Cells were clustered based on the scRNAseq data using the Lovain algorithm to a resolution where the SMC derived cells are split into two distinct clusters, as we and other have characterized ²³. Regardless of the definition of “mature SMC”, there was a significant decrease in the proportion of mature SMC in the Smad3^ΔSMC compared with control mice (Fig. 2a, b).

Figure 2. Loss of Smad3 alters cell fate decisions of transition SMC in atherosclerotic lesions.

(a) Expression of lineage tracing marker tdTomato, mature SMC markers Myh11, Cnn1, and unbiased clustering identified transition SMC in UMAP depiction of scRNAseq data. (b) Fraction of lineage traced cells that were mature SMC as defined by Myh11, Cnn1 expression or unbiased clustering. Chi square statistic, n=12916 control and 19181 Smad3^ΔSMC cells, from 6 and 8 mice respectively. (c) Expression of key lineage markers of other disease relevant cells captured by our scRNA sequencing, including fibromyocytes (Tnfrsf11b), endothelial cells (Cdh5), macrophages (CD68), and lymphatic endothelial cells (Prox1). (d) Unbiased clustering of all captured cells and respective clustering upon applying the Louvain algorithm, represented in UMAP space, with their respective biological identities as defined by marker genes. (e) Fraction of de-differentiated lineage-traced cells that contribute to fibromyocytes (FMC), chondromyocytes (CMC), and the newly identified cell cluster in control and Smad3^ΔSMC mice. (f) Fraction of lineage traced cells that contribute to CMC in control and Smad3^ΔSMC mice, two tailed t-test, n=916 cells. (g) CMC marker Col2a1 expression by RNAscope, positive area per section in control and Smad3^ΔSMC mice, two-talied t-test. (h) Representative images of colorimetric Col2a1 RNAscope quantification with control and Smad3^ΔSMC mice. Scale Bars 250 um (left), 75 um (right). (i) Von Kossa staining of sections from control and Smad3^ΔSMC mice, scale Bars 250 um (left), 75 um (right). (j) Von Kossa quantification as area of staining, two-talied test. (k) Dot plot representation of expression of cell-cycle markers in transitional SMC as represented by relative expression (color) and fraction of total positive cells (size). (l) Chondrocyte proliferation score of control and Smad3^ΔSMC transitional SMC. Chi square statistic, n=12916 control and 19181 Smad3^ΔSMC cells, from 6 and 8 mice respectively. Each dot on each bar graph represents mean data measured from a single matched section from each mouse, error bars represent 95% CI of mean.

To determine an optimum biologically relevant clustering resolution, we implemented an empirical approach, aimed at determining the minimum clustering resolution required to separate known biologically distinct populations of endothelial cells (EC) (Fig. 2c, d), i.e., blood vessel EC (VE-cadherin expressing, Prox1 negative) vs lymphatic EC (Prox1 expressing) endothelial cells ^44,45. The identities of the cellular subpopulations that were created with this approach were then determined based on specifically expressed canonical “marker genes” in each cellular cluster (Fig. 2c). Previous work in SMC lineage tracing in atherosclerosis has identified three distinct clusters of SMC derived cells, including mature SMC, fibromyocytes (FMC), and pro-calcific chondromyocytes (CMC) ^19,28,34,46. Using the cluster resolution identified above, similar subsets of lineage traced SMC were identified in these data (Fig. 2d), and in addition two distinct additional clusters of Myh11-Cre lineage labeled populations were also identified. One cluster was composed of pericytes, and an additional small population of uncharacterized disease-associated SMC-derived cells clustered by itself as a unique transcriptomic phenotype, that we named “remodeling-SMC” (R-SMC), to be studied in more detail below.

To validate this clustering resolution of the SMC transition cells identified, we conducted a sensitivity analysis by performing clustering of the scRNAseq data at different resolutions. At the lowest resolution, lesion transition SMC segregated from mature SMC, and with increasing resolution the transition cells separated into the FMC and CMC disease related phenotypes. Further increase in resolution of clustering witnessed the appearance of R-SMC and pericyte clusters. An additional increase in resolution resulted in partitioning of the mature SMC into two separate clusters that we could not easily spatially discern as distinct populations in the lesion, suggesting possible over-clustering. The robustness of this clustering was further confirmed by the ability to produce similar clusters using different clustering algorithms (Louvain and SLM (smart local moving)), with both all lesion cells and SMC-derived cells alone. Taken together, these various approaches demonstrate the appropriateness of the SMC transition clusters that are expected to reflect relevant biological phenotypes.

To determine if loss of Smad3 expression alters the cell fate transitions of disease-associated SMC lineage cells, we ascertained the fraction of de-differentiated cells that contribute to clusters for each of the two known and the R-SMC transition phenotypes. Smad3^ΔSMC transition SMC showed an increase in relative proportion of CMC along with the newly defined population at the expense of the intermediate FMC population (Fig. 2e, f). The increase in fraction of CMC among disease-associated Smad3^ΔSMC cells corresponded with a significant increase in lesion area expressing chondromyocyte marker Col2a1 (Fig. 2g, h), suggesting an absolute increase in the CMC population. Functionally, this increase in Col2a1 expressing cells correlated with an increase in lesion calcified area as assayed by von Kossa staining of atherosclerotic lesions (Fig. 2i, j). These cells appeared to result from an increase in proliferation of de-differentiated SMC, as evidenced by a higher level and increased proportion of cells expressing proliferation markers such as Mki67, Ccnd1, Ccnb1, and Myc, with limited alteration in cell-cycle inhibitor genes Cdkn2a/b (Fig. 2k). To further assess this possibility, we generated a “chondrocyte proliferation score” that represents a scaled-average expression of all genes associated with GO category “promote chondrocyte proliferation” with each individual cells’ transcriptome ⁴⁷. Consistent with higher expression of specific cell-cycle regulators, Smad3^ΔSMC transition SMC had a significantly higher chondrocyte proliferation score than control cells (Fig. 2l). Applying the analysis utilizing a “mesenchymal proliferation score” based on GO categories of “promote mesenchymal proliferation” produced equally significant results (Extended Data Fig. 3a), further suggesting that the increase in number of tdTomato positive cells in the lesions likely reflects proliferation of SMC derivative cells. The increase in cell number did not result from alterations in apoptosis, since there was no difference in TUNEL staining of control vs Smad3^ΔSMC sections (Extended Data Fig. 3b).

Disease SMC promote remodeling and leukocyte recruitment

In addition to increased CMC in the Smad3^ΔSMC animals, there was also an increase in the proportion of transition SMC that contributed to the newly identified cell cluster (Fig. 2d, 3a), that we refer to as remodeling-SMC (R-SMC) due to their high expression of genes involved in extracellular matrix remodeling. These cells were identified at the junction of CMC, fibroblast and FMC clusters. The vast majority of cells within this cluster were lineage traced at a proportion similar to that seen with CMC and FMC (Fig. 3b) confirming they were SMC derived. In vascular lesions, they constituted 6% of SMC transition cells in Smad3^ΔSMC mice, while they represented only 2% of this SMC population in control animals. To investigate the ontogenic relationship of this group of cells to SMC, FMC, and CMC, we ported the scRNAseq data to Slingshot ⁴⁸, a lineage inference tool designed to map trajectories involving multiple branching lineages. Applying this algorithm suggested that SMC give rise to FMC which in turn serve as a source for both CMC and R-SMC (Fig. 3c). Although the R-SMC gene expression pattern is most similar to that identified in CMC as shown by their juxtaposition in UMAP space and by hierarchical clustering, they were easily distinguished from CMC at the transcriptional level (Fig. 3d).

Figure 3. A cluster of Mmp3-expressing R-SMC transition cells promote remodeling and inflammation.

a) UMAP representation and location of the unique cluster of cells with high Mmp3 expression. (b) Percent of tdTomato positive cells representing different cellular clusters. (c) Pseudotemporal alignment of cells with Slingshot lineage inference (solid black line) identified a likely relationship for SMC, FMC, CMC, and R-SMC. (e) Heatmap of differentially regulated genes among distinct lineage-traced SMC in the lesion. Differentially expressed genes between CMC and R-SMC are indicated by yellow boxes. (e) Normalized Mmp3 expression of all the cells in the lesion. (f) Immunohistochemistry of Mmp3 in control and Smad3^ΔSMC mice. Scale bar 75 um. (g) RNAscope of Mmp3 (red) and Col2a1 (blue) expressing cells in atherosclerotic lesions showing restricted and distinct localization and clustering. Scale bars 75 um (left), 50 um (right). (h) RNAscope for MMP3 expression in human coronary artery specimens. Scale bars 250 um (left), 25 um (right). (i) Relative MMP3 expression in human coronary artery smooth muscle cells (HCASMC) evaluated by quantitative PCR in the presence and absence of TGFβ and SMAD3, two-tailed t-test, n=3. (j) Biological processes enriched in genes that are preferentially expressed by R-SMC compared with transitional SMC. Chi-square analysis as per the GREAT algorithm. (k) Expression of chemoattractants (Cxcl12, Saa3, Ccl2) that are specific to R-SMC with their respective receptors (Cxcr4, Tlr1, Ccr2) showing expression restricted to the monocyte cluster. Error bars represent 95% CI of mean.

Interestingly, the R-SMC cell population was marked by a particularly high expression of matrix metalloproteinase-3 (Mmp3) (Fig. 3e, f), an enzyme required for outward remodeling of atherosclerotic vessels ⁴⁹. Mmp3 was in fact among the most upregulated genes in this population compared with other SMC transition groups. Given our observation of increased outward remodeling identified in Smad3^ΔSMC mice, this finding was further investigated. Single-cell transcriptomic data suggested a higher number of Mmp3-expressing cells and an overall higher level of Mmp3 expression in Smad3^ΔSMC than control cells (Sup. Fig. 4a, c). In addition to increased mRNA levels, an in vitro functional study utilizing a florescent based Mmp3 activity assay demonstrated that homogenized aortic tissue from Smad3^ΔSMC mice had more Mmp3 activity than control tissue (Extended Data Fig. 4b).

In situ hybridization with RNAscope showed Mmp3 expressing R-SMC to be a distinct population from CMC, which were marked by Col2a1 expression (Fig. 3g), indicating that these R-SMC transition cells are juxtaposed but not overlapping in the lesion plaque and further supporting their distinct phenotype. Consistent with the established role for Mmp3 in outward remodeling, its expression was most prominent at the base of the atherosclerotic lesion in cells juxtaposed to the elastic lamina. Strikingly, these Mmp3 expressing cells were commonly associated with areas of disrupted elastic lamina, consistent with their possible invasion through this structure (Fig. 3g, Extended Data Fig. 4d). Importantly, the same rare population of MMP3 expressing cells were also found in human coronary arteries, where they were also associated with regions of disrupted elastic lamina (Fig. 3h). There were more cells expressing Mmp3 in Smad3^ΔSMC compared with controls as per the scRNAseq data and more prominent staining in Smad3^ΔSMC vs control cells (Fig. 3F, Extended Data Fig. 4a, c). Mmp3 expression was altered only in the SMC derived cells, and not significantly different in the non-SMC derived population, suggesting that altered Mmp3 levels by Smad3 happens in a cell-autonomous manner in SMC lineage transition cells (Extended Data Fig. 4a). By performing TGF-β stimulation of human coronary artery smooth muscle cells (HCASMC), we were able to show that MMP3 is suppressed by TGF-β (Fig. 3i). SMAD3 knockdown did not significantly increase the expression of MMP3 at baseline, but did negate the TGF-β dependent suppression of expression, suggesting SMAD3 is required for the TGF-β dependent regulation of MMP3 in atherosclerotic lesions.

To better understand the vascular function of this cell population, we performed pathway analyses of all significantly differentially expressed genes in the R-SMC as compared with other de-differentiated SMC phenotypes (Extended Data Table 1). PANTHER analysis revealed the top biological processes enriched in this list of genes are related to remodeling of ECM followed by regulation of chemotaxis and inflammation (Fig. 3j). Analysis of genes contributing to the enrichment of these pathways reveal that this population of cells express higher levels of multiple chemoattractants whose receptors are primarily restricted to macrophages (Fig. 3k). For example, this population is the main SMC source of Cxcl12 in the lesion, with the Cxcr4 receptor for this ligand being expressed exclusively by lesion macrophages. In addition, other R-SMC differentially expressed cytokine genes, including Saa3, and Ccl2, all have well-established roles in monocyte recruitment, with their respective receptor genes Tlr1 and Ccr2 expressed mostly by inflammatory cells in the lesion (Fig. 3k). There is high overlap between the Mmp3 high cells and cytokine expressing cells. In fact, if a marker-based definition of a pro-remodeling and recruitment population were defined as cells with elevation of Mmp3 and Cxcl12, one can recapitulate the unbiased clustering results (Extended Data Fig. 4e). Consistent with the hypothesis that SMAD3 suppresses the SMC chemotaxis signal to inflammatory cells, in vitro transwell migration studies demonstrated augmentation of HCASMC ability to recruit human monocytes with SMAD3 knockdown (Extended Data Fig. 4f). Taken together, these findings suggest that beyond its role in outward remodeling, the R-SMC population of cells may orchestrate monocyte recruitment and underlie the increase in CD68 positive cells in lesions.

Smad3 binds other transcription factors to regulate disease

Given that Smad3 is likely affecting all SMC lineage phenotypes beyond just the CMC and R-SMC populations described above, we characterized the alteration in gene expression of all de-differentiated SMC derived cells in the atherosclerotic lesions. To rigorously identify differentially regulated genes, we employed a Wilcoxon rank sum-based analysis comparing Smad3^ΔSMC versus control mouse data for FMC plus CMC and R-SMC cellular clusters and 83 genes were identified (Extended Data Table 2). Pathway analysis performed using DAVID demonstrated significant enrichment in TGFβ signaling, which is consistent with Smad3’s role in the TGFβ pathway (Fig. 4a). In addition, biological processes broadly related to ECM remodeling, SMC differentiation, and chemotaxis were also highly enriched. Utilizing GREAT ⁵⁰, we compared the list of 83 identified Smad3^ΔSMC marker genes to the top 1000 genes expressed by de-differentiated SMC in diseased murine vessels. The analysis revealed enrichment for genes linked to human atherosclerosis and arterial dilatation (Fig. 4b). In addition to Mmp3, and consistent with the observed phenotype of increased outward remodeling in Smad3^ΔSMC mice, Lox and Mfap5 were also among the top significantly down-regulated genes in de-differentiated Smad3^ΔSMC SMC (Extended Data Table 2). Pathogenic loss of function in both of these genes has been linked to human vascular syndromes including aortic aneurysms ^51–53. To investigate whether these genes are directly regulated by TGFβ, we stimulated HCASMC with TGFβ in the presence and absence of SMAD3. A subset of genes, such as LOX, appeared to be directly regulated by TGFβ in HCASMC, and this effect was abrogated by SMAD3 silencing (Fig. 4c), similar to the regulation of MMP3. However, a subset of genes, such as MFAP5, did not appear to be TGFβ responsive, but were sensitive to loss of SMAD3 (Fig. 4c). These findings suggested involvement of additional TGFβ-independent co-regulatory factors.

Figure 4. Smad3 regulates a transcriptional program associated with vascular outward expansion in conjunction with Sox9 and HoxB2.

(a) Significantly enriched GO biological processes represented by Smad3 differentially regulated genes. (b) Enrichment of human disease terms associated with differentially regulated genes. (c) Relative expression of LOX and MFAP5 in HCASMC in the presence and absence of TGFβ, in cells transfected with siRNA against SMAD3 (SMAD3-KD) or non-targeting control, two tailed t-test. (d) HOMER analysis results for motifs enriched in promoter regions of genes differentially regulated in Smad3^ΔSMC mice. (e) Expression of MMP3 and MFAP5 in control and SOX9 or HOXB2 knockdown HCASMC, two tailed t-test. (f) Co-immunoprecipitation of His-tagged HOXB2 (lane 4/5) or Flag-tagged SOX9 (lane 6) with endogenous SMAD3 compared with IgG or control vector (lanes 2/3). (g) Proximity ligation assay demonstrates close proximity of HOXB2-SMAD3 and SOX9-SMAD3 in the nucleus of HCASMC. (h) Relative luciferase activity of MFAP5-Luc in HEK-293 cells transfected with HOXB2 and SOX9 expression vectors in the absence (left) or presence (right) of SMAD3 siRNA (SMAD3-KD) or control siRNA (control), two-tailed t-test. All error bars represent 95% CI of mean. HEK-293 authenticity was validated upon receipt from the ATCC, and assessed periodically during the course of this study by cell morphometry and PCR evaluation of lineage markers. Abd, abdominal.

To better understand the context in which Smad3 regulates this complex transcriptional program and identify possible co-regulatory factors, we analyzed the 5’ regulatory elements of the 83 Smad3 differentially regulated genes to look for enriched transcription factor motifs. Motif analysis conducted with HOMER revealed enrichment for Hox/homeobox motifs as well as Sox9/10 related motifs (Fig. 4d). Hox and Sox genes were not dysregulated in the Smad3^ΔSMC SMC (Extended Data Fig. 5i), suggesting that these factors are likely directly interacting with Smad3 to regulate a joint transcriptional program. Consistent with this hypothesis, Sox9 is known to be a critical transcription factor regulating calcification, and has been shown in chondrosarcoma cells to selectively interact with Smad3, but not Smad2, to modulate an endochondral ossification program ⁵⁴. Since Hoxb2 and Sox9 were found to be expressed in transition SMC (Extended Data Fig. 5g, h) we further investigated the possible interaction of these factors with SMAD3. Knocking down SOX9 and HOXB2, the most highly expressed HOX gene in human coronary SMC, recapitulated the changes in expression of key vascular remodeling genes MMP3 and MFAP5 observed in vivo in mouse (Fig. 4e). Thus, we hypothesize that HOX factors, whose motifs are also enriched, can directly interact with SMAD3. To test this hypothesis, we performed nuclear co-immunoprecipitation experiments to test their interaction. In cells over-expressing His-HOXB2 or Flag-SOX9, SMAD3 protein was co-immunoprecipitated with anti- His or Flag antibody respectively (Fig 4f, Extended Data Fig. 6). Furthermore, proximity ligation assays demonstrated that endogenous SMAD3 is localized in proximity (<40nm) to HOXB2 and SOX9 to suggest their involvement in multi-protein complexes in the nucleus of HCASMC (Fig. 4g). Taken together, these data suggest HOX family proteins such as HOXB2, as well as SOX family member SOX9, physically interact in the nucleus to regulate transcription of targeted genes.

To test the functional significance and epistatic relationship of these findings, we cloned an evolutionarily conserved regulatory region near the 5’ end of the human MFAP5 gene, which contains conserved putative HOX, SOX, and SMAD binding elements, into a luciferase vector and tested its ability to respond to SOX9 and HOXB2 binding. SOX9 and HOXB2 efficiently activated this enhancer element but not the control luciferase vector, further establishing a role in this transcriptional program (Fig. 4h, Extended Data Fig. 5j). Knocking down SMAD3 diminished the SOX9 and HOXB2-dependent activation of the luciferase construct, suggesting interaction with SMAD3 is required for regulation of MFAP5 (Fig. 4h).

DISCUSSION

High resolution scRNAseq data has allowed us to describe a small but distinct subset of SMC transition cells that expresses pro-remodeling enzymes. Mmp3, a metalloprotease required for outward remodeling, is expressed most prominently in the R-SMC that are located in the basal plaque where they appear to migrate into the media towards the adventitia, potentially related to the previously described SMC lineage traced population in the adventitia ⁵⁵. Strikingly, in human coronary arteries, this MMP3 expressing pro-remodeling population resides in the same regions of the artery with observed disruption of the nearby elastic lamina, consistent with a role promoting outward remodeling in human coronary disease. These findings suggest that expansion of this population may be a significant contributing factor to outward remodeling and other adverse features. The importance of this MMP3-expressing SMC transition population in modifying plaque rupture risk may explain the seemingly paradoxical observation that a SNP associated with lower-expression of MMP3 is associated with greater luminal coronary artery stenosis on cardiac catheterization, but the higher-expressing variant is associated with more myocardial infarctions ⁵⁶. These genetic observations provide further evidence that modulation of R-SMC alters plaque features and CAD risk in human. While Mmp3 was previously thought to be expressed by macrophages in the lesions ⁵⁶, we found no evidence that Mmp3 was expressed by cells in the macrophage cluster. Previous work has shown that IL1 drives outward remodeling of plaques and this function was completely reversed by deletion of Mmp3 ⁴⁹. This suggests that SMC are the primary mechanism for the high-risk plaque feature of outward remodeling, as seen in the Smad3^ΔSMC mice, and may account in part for the beneficial effect of IL1 blockade on the risk of plaque rupture in humans ⁶. Mmp3 also appears differentially regulated among restricted populations of SMC progeny in carotid artery plaque ¹⁹, suggesting R-SMC exist in other atherosclerotic beds.

Interestingly, this population of R-SMC also expresses a number of chemokines, including Cxcl12, Ccl2, Saa3, whose main receptors are expressed solely on the monocyte-macrophage lineage. This suggests that the R-SMC population likely plays a role in regulating the inflammatory response to the lesion, contributing to the increase in observed monocyte-macrophage population detected in the lesions. The combination of remodeling and inflammatory cell recruitment, both factors that determine plaque stability, highlights the critical role that this specific sup-population of cells may play in modulating human disease risk. This contributes to existing literature ⁵⁷ suggesting that plaque SMC play a central role in regulating inflammatory cell recruitment and retention in atherosclerotic plaque and identifies a specific sub-population of transition SMC critical for high-risk plaque features.

There was also an increase in CMC in Smad3^ΔSMC mice suggesting Smad3 actively inhibits differentiation to this phenotype or inhibits their proliferation. The transcriptomic, topological, and lineage inference data presented here indicate that they are a distinct population from Mmp3-expressing R-SMC. The CMC exhibit a chondrogenic transcriptomic program ²⁸, expressing Col2a1, Acan, and Sox9, with similarities to chondrogenic progenitors in endochondral bone formation and repair. Smad3 has been shown to regulate Sox factor transcriptional activity in a TGFβ-independent manner through physical interactions ^54,58. The observed expansion of the CMC also provides an interesting parallel to established findings of accelerated bone and wound healing in Smad3 knockout mice ^59,60. The concomitant increase in Col2a1 expressing cells in the plaque and increased vascular calcification suggests that this cell type is at least partially responsible for coronary calcifications seen in human coronary artery disease. It remains to be determined whether increased calcification is harmful or protective in terms of plaque rupture risk, since conflicting observational data exists in humans. While increased coronary calcification is correlated with increased risk of myocardial infarction ⁶¹, local calcification appears to be protective against plaque rupture ^62–64 and interventions that lower risk of plaque rupture increase calcification ^65,66. Given the multiple populations of SMC-derived transition cells observed in our studies, their relative ratios could possibly determine the quality of calcification as well, which is also considered to confer differential risk of plaque stability^67,68.

Identification of the R-SMC transition phenotype adds to the complexity of SMC cell state changes that are associated with vascular disease development. We have previously identified fibroblast-like transition phenotype cells that we have termed fibromyocytes (FMC) ²³, as well as the noted chondromyocytes (CMC). Through our studies of the mechanisms by which CAD GWAS genes modulate disease risk, we have shown that transition of SMC to FMC is regulated by Tcf21 ²³ and Zeb2²⁹, and the transition to CMC is regulated by Ahr ²⁸. We have now also implicated Smad3 in this transition. Pseudotime assisted trajectory analyses have suggested that FMC may represent an intermediate transition phenotype that gives rise to the terminal CMC phenotype, and this possibility has been validated by other groups ¹⁹. Trajectory analyses presented here suggest that R-SMC may represent a distinct terminal phenotype. Published studies by other groups investigating the cell state changes by SMC in the disease setting in murine models have also identified FMC and CMC phenotype cells ^19,34,46. At least one publication has identified fibromyocytes in human carotid tissue samples ⁶⁹. Another murine study has clearly identified groups of cells expressing Mmp3 and Cxcl12 as independent transition clusters, but did not study the biology of these cells ¹⁹. Finally, it is worth noting that single cell transcriptomic data has been interpreted to show trans-differentiation of SMC to the macrophage lineage³⁴. These findings are not observed in our study or other recently published work^28,29,70.

Recent work by Chen et al. ⁴⁶ has employed single cell studies to investigate the role of Tgfβ signaling in vascular disease, employing a combined Marfan II/Loeys-Dietz and hypercholesterolemia mouse model. A key finding by these investigators was evidence for an SMC derived mesenchymal stem cell that gives rise to numerous cell types, including adipocytes, osteoblasts/chondrocytes (CMC) and macrophages, in the context of Tgfbr2 knockout and high fat diet. In their study increased plaque inflammation was due in large part to increased SMC-derived macrophage number in the vessel wall through this process. In studies reported here, we did not find evidence for an SMC-derived stem cell that mediates this effect and no evidence that SMC can transition into adipocytes or macrophages in control or Smad3^ΔSMC mice. By contrast, we found that control SMC give rise to fibromyocyte phenotype cells that subsequently give rise to CMC, and with Smad3 knockout the unique cluster of R-SMC derived cells. The observed differences between the Tgfbr2 knockout and Smad3 knockout suggest a fundamentally different mechanism by which SMC transition in response to signaling through these two different molecules.

Beyond the changes in proportions of the different SMC derivatives, loss of Smad3 also resulted in alterations in SMC transition phenotype transcriptomes as a whole. Gene knockout down-regulated several important ECM genes, including Lox, Mfap5 and Eln, whose loss of function mutations have been associated with Mendelian aortopathies. These findings suggest that global transcriptomic changes associated with Smad3 loss weaken the vascular wall and thus further promote outward vascular remodeling. These finding may also have implications in non-atherosclerotic vasculopathies, such as Marfan and Loeys-Dietz syndromes. Aortopathies such as Marfan’s syndrome have been shown to produce aberrant SMC derived populations that contribute to pathogenesis ⁷¹. In fact, our previous scRNAseq studies of a murine Marfan model also demonstrated an increase in Mmp3 expressing SMC progeny and lower Mfap5 expression, suggesting our findings here may extend beyond atherosclerotic disease.

Human genetics data has previously suggested the lead CAD-associated SNP rs56062135 at 15q22 is in linkage disequilibrium (LD) with SNP rs17293632 that appears to promote AP-1 binding and increase SMAD3 expression in vitro and in vivo ²⁰, suggesting higher SMAD3 expression may be associated with risk of myocardial infarction ^9,20,30. This is contradictory to our finding that complete loss of Smad3 increases plaque size in our murine atherosclerosis model, and there are several possible explanations for this disparity. It is possible that alternate SNPs in LD with rs56062135 have the opposite effect on SMAD3 expression in the context of certain types of cellular stimulation, i.e., serve as response QTLs. In this case, the response QTLs may have a greater effect on SMAD3 expression and the integrative effect of the entire haploblock on SMAD3 expression would be opposite and greater than the effect of rs17294632. Alternatively, it is possible that cell-fate changes identified in Smad3^ΔSMC mice overall stabilize the human lesion and thereby protect it from plaque rupture, despite there being larger lesion size and plaque burden. This paradoxical effect has previously been observed. For example, the IL1β blocking antibody canakinumab decreased the risk of myocardial infarction in human trials but Il1r blockade/knockout increased the lumen obstruction and plaque size in mouse models^{49 72 73}. Importantly, Il1r1 loss drastically changed plaque composition, suggesting SMC cell fate in plaques may be a stronger determinant for plaque rupture than plaque size alone. Indeed, it has been observed that the largest plaques seen on coronary angiogram are usually not the ones that rupture and cause myocardial infarction ⁷⁴. Recent human epidemiological and clinical data also suggest that the quality of calcification is critical, and that some types of more calcified plaques are less likely to rupture ⁶⁸. It is possible that the increased calcification seen in Smad3^ΔSMC mice is correlated with lesions in humans that are protected against myocardial infarction, which then contributes to the protective genetic signal.

METHODS

Mouse strains

SMC-specific lineage tracing and Smad3 knockout was generated by a well-characterized BAC transgene that expresses a tamoxifen-inducible Cre recombinase driven by the SMC-specific Myh11 promoter (Tg^{Myh11-CreERT2}; 019079; JAX). These mice were bred with a floxed-stop-flox tdTomato fluorescent reporter line (B6.Cg-Gt(ROSA)26Sor^{tm14(CAGtdTomato)Hze}/J; 007914; JAX) to allow SMC-specific lineage tracing. Smad3 conditional knockout were obtained from Matzuk lab from Univ Texas SW ^40,41 with LoxP sites flanking exons 2 and 3 which contains Smad3 DNA binding domain and creates a non-functioning frame-shift mutation after deletion⁴². All mice were back-crossed onto the C56BL/6 ApoE^−/− background. As the Cre-expressing BAC was integrated into the Y chromosome, all lineage-tracing mice in the study were male. The animal study protocol was approved by the Administrative Panel on Laboratory Animal Care at Stanford University.

Induction of lineage marker and Smad3 knockout by Cre recombinase

For all experiments, tamoxifen gavage schedule was as follows: two doses of tamoxifen, at 0.2 mg g⁻¹ bodyweight, were administered by oral gavage at 7–8 weeks of age, with each dose separated by 72–96 hrs. HFD was started (101511; Dyets; 21% anhydrous milk fat, 19% casein and 0.15% cholesterol) after the second gavage.

Mouse aortic root/ascending aorta cell dissociation

Immediately after sacrifice, mice were perfused with phosphate buffered saline (PBS). The aortic root and ascending aorta were excised, up to the level of the brachiocephalic artery. Tissue was washed three times in PBS, placed into an enzymatic dissociation cocktail (2 U ml⁻¹ Liberase (5401127001; Sigma–Aldrich) and 2 U ml⁻¹ elastase (LS002279; Worthington) in Hank’s Balanced Salt Solution (HBSS)) and minced. After incubation at 37 °C for 1 h, the cell suspension was strained and then pelleted by centrifugation at 500g for 5 min. The enzyme solution was then discarded, and cells were resuspended in fresh HBSS. To increase biological replication, multiple mice were used to obtain single-cell suspensions at each time point. For each scRNA capture, 2 mice were used. 4 separate pairs of isolation were performed for control and Smad3^ΔSMC, but one control 10X capture unexpectedly failed resulting in a final of 3 captures of control and 4 captures from conditional KO that was included in the analysis. Cells were sorted FACS sorted based off tdTomato expression. tdT⁺ cells (considered to be of SMC lineage) and tdT⁻ cells were then captured on separate but parallel runs of the same scRNA-Seq workflow (gating strategy and threshold identical to those published in previous work by Wirka et al²⁴), and datasets were later combined for all subsequent analyses.

Single cell capture and library preparation and sequencing

All single cell capture and library preparation was performed at the Stanford Functional Genomics Facility and Stanford Genomic Sequencing Service Center. Cells were loaded into a 10x Genomics microfluidics chip and encapsulated with barcoded oligo-dT-containing gel beads using the 10x Genomics Chromium controller according to the manufacturer’s instructions. Single-cell libraries were then constructed according to the manufacturer’s instructions (Illumina). Libraries from individual samples were multiplexed into one lane before sequencing on an Illumina platforms with targeted depth of 50,000 reads per cell.

Human coronary artery cell acquisition

Human coronary arteries used in this study were dissected from explanted hearts of transplant recipients and were obtained from the Human Biorepository Tissue Research Bank under the Department of Cardiothoracic Surgery from consenting patients with approval from the Stanford University Institutional Review Board as previously described.

Generation and studies of aortic root sections

Immediately after sacrifice, mice were perfused with 0.4% paraformaldehyde (PFA). The mouse aortic root and proximal ascending aorta, along with the base of the heart, was excised and immersed in 4% PFA at 4 °C for 24 hrs. After passing through a sucrose gradient, tissue was frozen in optimal cutting temperature compound (OCT) to make blocks. Blocks were cut into 7-μm-thick sections for further analysis.

Immunohistochemistry (IHC) was performed according to standard protocol. Primary antibodies: Anti-SM22alpha rabbit polyclonal primary antibody (1:300 dilution; ab14106; Abcam), a Mmp3 Rabbit monoclonal antibody (1:200 dilution; Abcam 52915 ) or a CD68 rabbit polyclonal antibody (1:400 dilution; ab125212; Abcam). Secondary: Rabbit-on-Rodent HRP Polymer (RMR622; Biocare Medical). The processed sections were visualized using a Leica DM5500 microscope objective magnification, and images were obtained using Leica Application Suite X software. Von Kossa stain was performed using Abcam 150687 kit with manufacturer’s recommended protocol with 90-minute development time. All biological replicates for each staining were performed simultaneously on position-matched aortic root sections to limit intra-experimental variance. Folded sections that were uninterpretable after processing were removed.

Sections obtained at equal distance measured from the superior margin of the aortic sinus were used for comparison of lesion features. Lesion size was defined as the area encompassing the luminal edge of the lesion to the border of Tagln positive intimal-medial junction, i.e., area circumscribed by intimal border of Tagln staining minus the lumen area. Vessel media area was defined as that circumscribed by the outer edge of Tagln staining, i.e., the external elastic lamina, minus the area circumscribed by inner edge of the Tagln staining, i.e., the internal elastic lamina. Areas of interest were quantified in a blinded fashion using ImageJ (National Institutes of Health) software and compared using a two-sided t-test.

RNAscope assays

Slides were processed according to the manufacturer’s instructions, and all reagents were obtained from ACD Bio. In short, slides were washed once in PBS, then immersed in 1× Target Retrieval reagent at 100 °C for 5 min. Slides were washed twice in deionized water, immersed in 100% ethanol and air dried, and sections were encircled with a liquid-blocking pen. Sections were incubated with Protease III reagent for 30 min at 40 °C, then washed twice with deionized water. Sections were incubated with commercially available probes against mouse Mmp3, Col2a1, and human MMP3 or a negative control probe for 2 hrs at 40 °C. Colorimetric assays were performed per the manufacturer’s instructions.

Analysis of scRNA-Seq data

Fastq files from each experimental time point and mouse genotype were aligned to the reference genome (mm10) individually using CellRanger Software (10x Genomics). Individual datasets were aggregated using the CellRanger aggr command without subsampling normalization. The aggregated dataset was then analyzed using the R package Seurat ⁷⁵. The dataset was trimmed of cells expressing fewer than 750 genes, and genes expressed in fewer than 50 cells. The number of genes, number of unique molecular identifiers and percentage of mitochondrial genes were examined to identify outliers. As an unusually high number of genes can result from a ‘doublet’ event, in which two different cell types are captured together with the same barcoded bead, cells with >6000 genes were discarded. Cells containing >7.5% mitochondrial genes were presumed to be of poor quality and were also discarded. The gene expression values then underwent library-size normalization and normalized using established Single-Cell-Transform function in Seurat. Principal component analysis was used for dimensionality reduction, followed by clustering in principal component analysis space using a graph-based clustering approach via Louvain algorithm. UMAP was then used for two-dimensional visualization of the resulting clusters. Lineage inference was performed using Slingshot with available Slingshot software in R using converted Seurat object into singlecellexperiment objects. Analysis, visualization and quantification of gene expression and generation of gene module scores were performed using Seurat’s built-in function such as “FeaturePlot”, “VlnPlot”, “AddModuleScore”, and “FindMarker.” Lists of genes associated with each GO category were obtained from Geneontology.org. Panther / DAVID / GO / GREAT analysis was performed using web-based platform at Geneontology.org, Great.Stanford.Edu, and David.ncifcrf.gov. Top 1000 genes expressed in modulated SMC was defined by the highest expressing 1000 transcripts (based on average expression) from scRNA data in all de-differentiated lineage traced cells. Promoter/5’-Regulatory region of genes were extracted utilizing UCSC table browser based off 1kb upstream of TSS of transcripts. Motif analysis was performed using freely available HOMER software ⁷⁶ with findMotifGenome function. The regulatory region of the top 1000 gene was used as the background as bases for motif enrichment.

HCASMC culture/experiments

Human coronary artery smooth muscle cells were obtained from Lonza, and cultured in smooth muscle growth medium (Lonza; catalog number: CC-3182) supplemented with human epidermal growth factor, insulin, human basic fibroblast growth factor and 5% FBS, according to the manufacturer’s instructions. All HCASMC lines were used at passages 4–8. siRNA knockdowns were performed using Lipofectamine RNAiMax (Life Technologies) using manufacturer’s recommended protocol at 50pg siRNA / 100,000 cells. Cells were allowed to recover in SMC growth medium (with or without additional growth factor) for 36 hours prior to RNA harvest. Recombinant TGFB (PeproTech 100–21) concentration used in stimulation was 10ng/ml. Phenotype of the HCASMC was surveyed by checking expression of lineage-specific markers, e.t., Myh11, Tagln, Cnn1.

Proximity Ligation Assay

Human coronary artery smooth muscle cells were cultured on tissue-culture slides in serum containing media for 24 hours. The cells were then fixed with 4% PFA for 30 minutes at room temperature. Proximity ligation assays were performed on these slides using a Sigma DuoLINK kit (DUO92101) with rabbit anti-SMAD3 antibody (Cell Signaling 9523S (1:200)), mouse anti-HOXB2 monoclonal antibody (DSHB: PCRP-HOXB2–1C9 (1:50 (hybridoma supernatant)), or mouse anti-SOX9 monoclonal antibody (DSHB PCRP-SOX9–1A2 (1:50 hybridoma supernatant)).

Co-immunoprecipitation Experiment

Myc-Flag-tagged SOX9 (PS100016 Origene) and 6x-HIS-tagged-HOXB2 (Addgene 8522) were obtained from commercial vendors and cloned into pCMV6 vector and transfected into HEK cells. The cells were allowed to recover for 36 hours after media change and nuclear complex co-immunoprecipitation was performed using commercially available Nuclear-Complex Co-immunoprecipitation kit from ActiveMotif(54001) with manufacturer’s recommended protocol using mouse Anti-Flag (Sigma F3165) or mouse Anti-His (Abcam 18184) antibody for immunoprecipitation, followed by blotting using Rabbit anti-Smad3 antibody (Cell Signaling 9523S), followed by Anti-Rabbit HRP (Cell Signaling 7074S) and detected via Luminata Forte Western HRP substrate (Millipore).

Luciferase experiments

For Luciferase experiments, an evolutionary conserved region of human MFAP5 regulatory region (chr12:8815212–8815569) was cloned from human genomic DNA and placed into pLuc-MCS vector, whereas inert/scramble similar length spacer was cloned into baseline pLuc-MCS as control. pCMV6-empty, and cloned pCMV6-Flag-SOX9 or pCMV6-his-HOXB2 were transfected into cells via lipofectamine 2000 along with respective luciferase and Renilla vector. Media was changed after 6 h, and dual-luciferase activity (Promega) was recorded after 24 h using a SpectraMax L luminometer (Molecular Devices). Relative luciferase activity (firefly/Renilla luciferase ratio) is expressed as the fold change over control conditions.

Mmp3 Activity Assay

Mmp3 activity was measured via Abcam MMP-3 Activity Assay Kit (ab118972) following their tissue-based activity measurement protocol. 0.5 cm of dissected thoracic aorta from identical locations of control and Smad3^ΔSMC mice were placed in the tissue homogenizer for 10 seconds on ice in chilled assay buffer. After centrifugation, the supernatants were then directedly assayed. Mmp3 activity was measured at 10 minutes and 30 minutes after initiation of the reaction, using a SpectraMax luminometer at Ex/Em=325/393 nm, with exposure of 600ms. Two biological replicates with 3 separate segments of thoracic aorta were used for the assay.

Transwell Assay

HCASMC were grown in 24 well plates at low density at 10,000cells/well. SMAD3 knockdown was performed 24 hours prior to initiation of the migration experiment. HCASMC were washed with PBS and cultured in serum free HCASMC media. 8um cell culture inserts (Corning 353097) were placed into the wells. THP-1 cells (ATCC) were grown in standard culture conditions. then spun down, and resuspended in serum-free HCASMC media, and placed in the top chamber for 3 hours at 37C. After 3 hours, THP1 cells in the bottom chamber in suspension were quantified.

TUNEL Assay

TUNEL assay was performed on cryopreserved aortic root sections using commercially available chromogenic TUNEL assay kits. Quantification was performed on 40X magnification at one random point on each cusp and TUNEL+ nuclei was counted manually in a blinded manner.

Statistical methods

Differentially expressed genes in the scRNA-Seq data were identified using a Wilcoxon rank-sum test. Distribution of cells within defined-populations was tested via X-square test. Significance determination of histological measurements, luciferase studies, qPCR results, and composite gene-scores were done via two-tailed T-test, which was performed after verifying normal distribution of the data using the D’Agostino-Pearson test, performed using Prism software. Multiple comparisons were corrected via Bonferroni correction when necessary.

Extended Data

Extended Data Fig. 1.

Experimental model validation. (A) Measured weight of experimental mice used in sections at time of sacrifice. (B) Percent of total experimental mice cohort that survived to final time point at 24 weeks. (C) Schematic indicating the location of aortic root section used in the study (dashed line). (D) Aligned captured mRNA sequence of tdTomato positive cells in control (top) and Smad3^ΔSMC tdTomato cells, showing absence of reads mapping to exon 2–3 which are flanked by LoxP sites. Atherosclerotic plaque in control (E) and Smad3^ΔSMC (F) mice stained for phospho-Smad3 (brown), demonstrating loss of Smad3 in SMC lineage cells in the plaque. Scale bar: 50um.

Extended Data Fig. 2.

Vascular lesion analysis. (A) Representative image of Oil Red O-stained aortic root in control and Smad3^ΔSMC mice, (B) quantified total Oil Red O positive area, and (C) quantified fraction of plaque Oil Red O staining (total Oil Red-O positive area over plaque area). (D) Representative image of trichrome stained aortic root in control and Smad3^ΔSMC, with (E) quantified total acellular area, and (F) quantified normalized plaque acellular fraction (total acellular area over plaque area). Scale bar: 50um.

Extended Data Fig. 3:

SMC cell state changes. (A) Mesenchymal proliferation score of de-differentiated SMC in control and SMC specific Smad3^ΔSMC mice. (B) Number of TUNEL labeled apoptotic cells per high magnification field (HPF) in control and Smad3^ΔSMC mice.

Extended Data Fig. 4:

R-SMC gene expresson. (A) Expression of Mmp3 in lineage labeled (Cre positive) and non-lineage labelled (Cre negative) cells based on scRNAseq data. (B) Measured Mmp3 activity detected in isolated aortic tissue from wild type (brown) vs Smad3^ΔSMC mice (orange). (C) Featureplot of lineage-traced cells expressing Mmp3 in control and Smad3^ΔSMC aortic root. (D) High magnification of a section of atherosclerotic plaque with region of broken elastic lamina stained for Mmp3 expression (orange-brown color). Scale bar: 75um. (E) Fraction of transition SMC in control and Smad3^ΔSMC lesions with R-SMC fate as defined by unbiased clustering (left) and by concurrent high Mmp3 and CxCl12 expression (right). (F) Number of migrated THP-1 cells after 3 hours of transwell-incubation with no cells, control HCASMC, and SMAD3-deficient HCASMC.

Extended Data Fig. 5:

(A - H) FeaturePlot of Smad3, Smad2, Col2a1, Mki67, Lox, Mfap5, HoxB2, and Sox9 expression in SMC in atherosclerotic lesions in the aortic root. (I) Individual cell expression of HoxB2, HoxB3, HoxB4, and Sox9 in lineage labeled cells in control and Smad3^ΔSMC aortic root. (J) Control reporter luciferase activity in response to SOX9 and HOXB2 overexpression.

Extended Data Fig 6:

Replicate of Flag-SOX9 immunoprecipitation with endogenous SMAD3 immunodetection, with additional controls for anti-flag antibody.