ADAMTS7 promotes smooth muscle foam cell expansion in atherosclerosis

Coronary artery disease (CAD) is the leading cause of death in the United States despite the widespread availability of multiple highly effective lipid-lowering therapies (1). Thus, there remains a need for novel, non-lipid-lowering therapeutic approaches for the treatment of CAD. Human genome-wide association studies (GWAS) have proven an effective, unbiased approach to discovering novel genomic loci associated with CAD, potentially uncovering therapeutic targets.

GWAS have repeatedly identified the 15q25 locus as a region of interest for CAD (2–5). This locus harbors the gene a disintegrin and metalloproteinase with thrombospondin motifs 7 (ADAMTS7), which encodes a secreted metalloproteinase previously implicated in osteoarthritis (6). After its identification by GWAS, subsequent mouse studies demonstrated that ADAMTS7 is proatherogenic, because mice with whole-body Adamts7 KO had reduced atherosclerosis (7). Notably, this reduction in atherosclerosis occurred without any changes in plasma lipoprotein levels, suggesting that ADAMTS7 has a lipid-independent role in atherogenesis (8). Further mouse models of Adamts7 have confirmed this proatherogenicity while demonstrating that ADAMTS7 catalytic activity is required for this effect (9, 10). Collectively, these findings identify ADAMTS7 as an attractive lipid-independent therapeutic target for CAD, although the molecular mechanisms linking ADAMTS7 to atherogenesis remain incompletely understood.

There has recently been an increased appreciation for the role of smooth muscle cells (SMCs) in atherosclerosis (11). Although historically viewed primarily as contributors to fibrous cap formation, single-cell RNA-seq (scRNA-seq) studies in mice and humans have revealed that SMCs are highly plastic and capable of undergoing phenotypic modulation, acquiring features of multiple cell types, including macrophage-like characteristics (12, 13). Consistent with this plasticity, studies in the Apoe^–/– model of atherosclerosis demonstrated that most lipid-laden aortic foam cells are not macrophages but cells of SMC origin, highlighting that SMC foam cells may play a more significant role in atherogenesis than previously understood (14). Mechanistic studies of ADAMTS7 have largely focused on its effects on SMC function, because both Adamts7 KO and overexpression alter neointima formation after vascular injury and modulate SMC migration in ex vivo assays (7, 15). Because ADAMTS7 is a secreted protein, it may influence multiple vascular cell types through both cell-autonomous and paracrine mechanisms (16). Adamts7 is not constitutively expressed in vascular cells; rather, its expression is induced in response to vascular injury and cytokine signaling (7, 9). Low basal expression of Adamts7 in vivo, together with its transient induction, has hindered efforts to identify the cell types that express Adamts7 and to define its function during atherosclerosis (7). Overall, ADAMTS7 has consistently been linked to SMC function, yet the mechanistic basis by which this association promotes atherosclerosis remains unknown (8).

In this study, we identified ADAMTS7 expression across multiple human vascular cell types, including SMCs and endothelial cells (ECs). We additionally generated several novel tissue-specific Adamts7 mouse models, enabling mechanistic interrogation of ADAMTS7 under conditions of sustained induction or targeted deletion. Using these models, we demonstrate that Adamts7 expression in either SMCs or ECs is individually sufficient to increase atherosclerosis and promote SMC foam cell formation, whereas deletion in either cell type alone does not significantly affect lesion burden. Mechanistically, we found that ADAMTS7 alters the chromatin landscape of SMCs, inducing expression of the myeloid transcription factor (TF) PU.1, increasing Cd36 expression, and enhancing SMC oxidized low-density lipoprotein (oxLDL) uptake. Together, these findings demonstrate that ADAMTS7 drives SMC foam cell expansion and, ultimately, atherosclerosis.

ADAMTS7 is induced within multiple vascular cell types during lesion formation. Prior literature indicates that Adamts7 expression is induced within the vasculature and promotes SMC migration while inhibiting re-endothelialization (7, 17). ADAMTS7 is secreted (18), and as such, the cell type that responds to it does not necessarily have to produce it. Thus, to determine the cellular origins of ADAMTS7, we interrogated the largest human atherosclerotic carotid artery scRNA-seq dataset generated to date (19) and found that multiple vascular cell types express ADAMTS7 in mature plaques. The highest ADAMTS7 expression was present in ECs, whereas SMCs, fibroblasts, and mast cells displayed ADAMTS7 expression at a lower level (Figure 1A). Interestingly, the highest level of ADAMTS7 expression occurred in ECs annotated as inflammatory (19), whereas fibrotic ECs did not display as high a level of expression.

Figure 1

Identifying endogenous ADAMTS7 expression. (A) scRNA-seq of human carotid atherosclerosis and cell clustering identities. Cluster identities are reproduced with permission (19). (B) RNAscope of the BCA in mice after WTD feeding. Nuclei are stained with DAPI. Scale bar: 100 μm. RNAscope of the BCA of Ldlr^KO mice after 7 (C) and 10 (D) weeks of WTD feeding and probed against Adamts7 and Pecam1. White arrows highlight EC expression of Adamts7. Yellow arrows indicate non-EC expression of Adamts7. Scale bar: 50 μm.

We next sought to confirm this observation in mouse models of atherosclerosis. Previous work in Apoe^–/– mice has shown that Adamts7 expression is induced after 4 weeks of a Western-type diet (WTD) feeding but is absent by terminal time points (7). To determine when Adamts7 is expressed during atherosclerosis, we performed RNAscope within the brachiocephalic artery (BCA) of hyperlipidemic LDL receptor KO (Ldlr^KO) mice at 4, 6, and 9 weeks of WTD. Adamts7 expression appears during early atheroma formation and is especially prominent at 9 weeks when established lesions are present (Figure 1B). Across multiple experiments, Adamts7 expression was localized to the developing atherosclerotic neointima and to multiple other locations in the BCA, consistent with its expression in SMCs and other cells (Figure 1, B–D). To confirm EC expression as seen in the human scRNA-seq data, we performed RNAscope in the BCA of Ldlr^KO mice after 7 and 10 weeks of WTD (Figure 1, C and D). At the early lesion time point of 7 weeks, we detected colocalization of the Adamts7 transcript with Pecam1, an EC marker, whereas at the 10-week time point, we observed Adamts7 expression in both ECs and non-ECs. These data support the notion that multiple vascular cell types produce ADAMTS7 during atherosclerosis in mice and humans.

SMC- or EC-specific KO of Adamts7 does not reduce atherosclerosis. Given Adamts7 expression in multiple vascular cell types, we next asked if the KO of Adamts7 in any single vascular cell type could reduce atherosclerosis. We generated a conditional KO model of Adamts7 in which exons 5 and 6 are floxed (Supplemental Figure 1A; supplemental material available online with this article; https://doi.org/10.1172/JCI187451DS1). Given the existing prior literature on ADAMTS7 regulation of SMC function (7, 15), we crossed this mouse to the Myh11-CreER^T2 and Ldlr^KO to generate a hyperlipidemic SMC conditional KO of Adamts7 (Adamts7^fl/fl; Ldlr^KO; Myh11-CreER^T2, referred to as Adamts7^SMCKO Ldlr^KO) and subsequently induced KO through intraperitoneal tamoxifen injections (Figure 2A).

Figure 2

SMC or EC KO of Adamts7 does not affect atherosclerosis. (A) Schematic outlining experimental design and mouse comparisons for SMC KO of Adamts7. Created with BioRender.com. (B) Representative ORO staining of en face aortas and enumeration. n = 12–14. (C) Representative images of H&E-stained aortic root sections, quantification of plaque areas (n = 5–7). Scale bar: 500 μm. (D) Schematic outlining experimental design and mouse comparisons for EC KO of Adamts7. Created with BioRender.com. (E) Representative ORO staining of en face aortas and enumeration. n = 7. (F) Representative images of H&E-stained aortic root sections, quantification of plaque areas (n = 6–7). Scale bar: 500 μm. Statistics were analyzed using a 2-tailed Student’s t test.

Although Adamts7 expression is typically low in uninjured vessels, SMC explants from Adamts7^SMCKO Ldlr^KO mice had an approximately 50% reduction in Adamts7 expression compared with WT controls Adamts7^SMCWT Ldlr^KO (Supplemental Figure 1B). After 16 weeks of WTD feeding, we observed no changes in body weight or total cholesterol level (Supplemental Figure 1, D and E). Surprisingly, Adamts7^SMCKO Ldlr^KO did not alter the lesion area in either en face Oil Red O (ORO) analysis (Figure 2B) or aortic root lesion area (Figure 2C).

Because ADAMTS7 expression is also prominent in ECs, on the basis of human scRNA-seq data (Figure 1A), we next generated an Adamts7 EC-specific conditional KO by crossing the hyperlipidemic conditional KO mouse to the Cdh5-CreER^T2 (Adamts7^fl/fl; Ldlr^KO; Cdh5-CreER^T2, referred to as Adamts7^ECKO Ldlr^KO). To confirm EC-specific KO, we isolated CD31⁺ ECs from aortas and observed markedly reduced Adamts7 expression in the CD31⁺ fraction from Adamts7^ECKO Ldlr^KO mice compared with WT mice (Adamts7^ECWT Ldlr^KO), whereas no difference was observed in the CD31^– (non-EC) fraction (Supplemental Figure 1C). After 16 weeks of WTD feeding, we assessed atherosclerosis in Adamts7^ECKO Ldlr^KO compared with Adamts7^ECWT Ldlr^KO (Figure 2D). Again, we observed no changes in body weight or total cholesterol levels (Supplemental Figure 1, F and G). Similar to Adamts7^SMCKO Ldlr^KO mice, EC KO of Adamts7 also did not demonstrate any changes in the atherosclerotic burden (Figure 2, E and F). These data strongly suggest that individual KO of Adamts7 in ECs or SMCs alone is not adequate to alter atherogenesis in mice.

Transgenic overexpression of Adamts7 in mouse SMCs or ECs exacerbates atherosclerosis. Given that KO of Adamts7 in either SMCs or ECs alone did not appear to reduce atherogenesis in our model, we next asked whether sustained induction of Adamts7 expression in a single cell type can increase atherogenesis. Because ADAMTS7 is induced in multiple vascular cell types in human atherosclerosis (Figure 1A), we generated a transgenic mouse model with conditional Adamts7 overexpression to mimic its induction and promote its function in multiple cell types. We inserted murine Adamts7 into the Rosa26 locus with a lox-stop-lox (LSL) cassette preceding the gene, allowing for tissue type–specific overexpression of Adamts7. The transgenic overexpression model was subsequently bred to the Ldlr^KO background as well as to either the Myh11-CreER^T2 (Adamts7^SMCTG) or the Cdh5-CreER^T2 (Adamts7^ECTG) to allow for SMC (Figure 3A) or EC (Figure 3G) specific overexpression of Adamts7, respectively. We verified Adamts7 overexpression in the SMC transgenic model via qPCR (Supplemental Figure 2A) and Western blot (Supplemental Figure 2B) from isolated aortic RNA and protein. We note that Adamts7^SMCTG Ldlr^KO exhibited a large-fold change in Adamts7 RNA, but this was partly due to the near-complete absence of baseline expression of Adamts7. Primary SMC explants from Adamts7^SMCTG displayed increased migration (Supplemental Figure 2C), consistent with prior findings (7). Validation of Adamts7^ECTG was achieved by examining Adamts7 expression in isolated CD31⁺ vascular ECs as well as mouse lung ECs (Supplemental Figure 2D); Adamts7 expression was elevated in ECs, whereas CD31^– cells showed no difference in Adamts7 expression.

Figure 3

SMC and EC transgenic Adamts7 increases aortic atherosclerosis. (A) Schematic outlining experimental design and the generation of the SMC transgenic ADAMTS7 mouse. Created with BioRender.com. (B) ORO staining of the en face aorta and quantification of lesion area. n = 9. (C) Representative images of H&E-stained aortic root sections, quantification of plaque areas (n = 6). Scale bar: 500 μm. (D) Representative images of H&E-stained aortic root sections with necrotic core outlined in dotted line (n = 6 mice). Scale bar: 300 μm. (E) Representative images of aortic root sections stained against α-SMA (Cy3; red), MAC2 (Alexa Fluor 488 green), and cell nuclei (DAPI blue) with their subsequent quantification relative to lesion area (n = 6). Scale bar: 300 μm. (F) Representative Picrosirius red staining of aortic root lesions with bars indicating fibrous cap thickness and quantification of fibrous cap thickness (n = 6). Scale bar: 300 μm. (G) Schematic outlining experimental design and the generation of the EC transgenic ADAMTS7 mouse. Created with BioRender.com. (H) ORO staining of the en face aorta and quantification of lesion area. n = 3–4. (I) Representative images of H&E-stained aortic root sections and quantification of plaque areas (n = 5–6). Scale bar: 500 μm. ****P < 0.0001, *P < 0.05. Statistics were analyzed using a 2-tailed Student’s t test.

To investigate the role of sustained cellular Adamts7 in atherosclerosis development, Adamts7 overexpression was induced in all animal models via tamoxifen injection, and mice were fed a WTD for 16 weeks. There were no plasma cholesterol or body weight changes in any animal models (Supplemental Figure 2, E–H). Adamts7^SMCTG Ldlr^KO mice had a stark 3.5-fold increase in atherosclerosis as shown by en face ORO staining as compared with control Adamts7^SMCCTL Ldlr^KO mice (Figure 3B) but no change in the aortic root lesion area as assessed by H&E staining (Figure 3C), consistent with findings in other Adamts7 mouse models (9).

Given the dramatic increase in aortic atherosclerosis, we asked whether shorter durations of WTD feeding would confer an atherosclerosis phenotype. Indeed, we found increased atherosclerosis at 3, 6, and 9 weeks of WTD feeding (Supplemental Figure 3A).

During atherosclerosis, SMCs play an essential role in plaque stabilization (20). As such, we further characterized the aortic root plaque morphology in Adamts7^SMCTG Ldlr^KO mice. We found no changes in necrotic core area (Figure 3D) or macrophage content via MAC2 staining (Figure 3E). However, Adamts7^SMCTG Ldlr^KO lesions had reduced SMC content as determined by α-SMA staining (Figure 3E), especially in the fibrous cap region. The reduction in α-SMA also coincided with a 48% reduction in fibrous cap thickness as assessed with Picrosirius red staining (Figure 3F). The lack of change in MAC2 staining and the reduction in cap thickness indicate that increased ADAMTS7 reduces indices associated with human plaque stability in these mouse models.

Finally, in tandem, we performed a similar atherosclerosis study on the Adamts7^ECTG Ldlr^KO mice and observed a 1.5-fold increase in aortic atherosclerosis in Adamts7^ECTG Ldlr^KO mice as compared with controls (Figure 3H). We also observed the same lack of changes in aortic root lesion area between groups (Figure 3I).

Adamts7 promotes SMC foam cell formation. The significant increase in ORO staining of the en face aorta indicated an increased accumulation of neutral lipids within vascular cells, presumably due to an expansion in foam cell formation. Thus, we next assessed aortic foam cell content in our Adamts7 transgenic mice using a neutral lipid dye, LipidTOX. After excluding debris, doublets, and dead cells (Supplemental Figure 4A), we observed that the Adamts7^SMCTG Ldlr^KO had a 3-fold increase in foam cell content, mirroring the increase in aortic atherosclerosis (Figure 4A). Notably, greater than 80% of these foam cells lacked CD11B and CD64, indicating that non-leukocytes were accumulating lipids and becoming foamy (Figure 4B) (14).

Figure 4

ADAMTS7 promotes foam cell expansion. (A) Flow cytometry–based quantification of foam cells through LipidTOX staining of the aorta. A normolipidemic aorta was used to establish the LipidTOX high gate. n = 3. (B) Breakdown of foam cells stained positively for CD11B and CD64. n = 3. (C) Confirmation of efficient ZsGreen labeling of SMCs with nuclei counterstained with DAPI. Scale bar: 500 μm. (D) Foam cell analysis with the SMC lineage tracer ZsGreen and leukocyte marker CD45 after 12 weeks of WTD. Counts were normalized to 100,000 live cells. n = 4–6 mice. Statistics were analyzed by 2-way ANOVA with Šidák’s multiple comparison test. (E) Flow cytometry–based quantification of foam cells through LipidTOX staining of the aorta. A normolipidemic aorta was used to establish the LipidTOX high gate. n = 3–4. (F) Normalized foam cell counts out of 100,000 live cells after 16 weeks of WTD. Leukocytes were identified as CD45⁺. SMCs were identified as CD31^– CD45^– CD200⁺. n = 10–11 mice. Statistics were analyzed by 2-way ANOVA with Šidák’s multiple comparison test. (G) In vitro foam cell analysis with primary cells from transgenic mice treated with 10 μg/mL DiI-oxLDL for 24 hours. n = 4. (H) In vitro foam cell analysis with explanted primary SMCs from whole-body Adamts7 KO mice treated with 10 μg/mL DiI-oxLDL for 24 hours. n = 3. ****P < 0.0001, ***P < 0.001, ** P < 0.01, *P < 0.05 Between-sample comparisons were analyzed by a 2-tailed Student’s t test.

Given that increased ORO staining of the aorta was evident even at early time points, we next examined foam cell composition after 3 weeks of WTD. Despite the short feeding period, Adamts7^SMCTG Ldlr^KO mice already exhibited a significant increase in foam cell formation, driven primarily by an expansion of SMC-derived foam cells as identified by CD200 expression (Supplemental Figure 3B).

To further confirm that the increase in foam cells was predominantly of SMC origin, the Adamts7^SMCCTL Ldlr^KO and Adamts7^SMCTG Ldlr^KO mouse model was subsequently bred to an LSL-ZsGreen reporter (Adamts7^SMCCTL Ldlr^KO ZsGreen^SMC+ and Adamts7^SMCTG Ldlr^KO ZsGreen^SMC+) to allow for the lineage tracing of SMCs and their derived cells (12). Efficient ZsGreen labeling of SMCs was achieved through microscopy (Figure 4C). After 12 weeks of WTD feeding, ZsGreen status and concurrent staining with CD45 revealed that greater than 65% of lipid-laden foam cells were ZsGreen⁺ and CD45^- (Supplemental Figure 4B), indicating again that Adamts7 was predominately increasing SMC foam cell formation (Figure 4D). We repeated the foam cell analysis within Adamts7^ECCTL Ldlr^KO and Adamts7^ECTG Ldlr^KO and observed that the Adamts7^ECTG Ldlr^KO mice contained an increase in foam cell content as well (Figure 4E). Further assessment of these foam cells within the aorta revealed that the majority were CD45^– and CD31^– while being CD200⁺ (21), indicating again that SMCs are also becoming foamy in this model (Figure 4F).

To test if ADAMTS7 promotes SMC foam cell formation by enhancing lipid uptake, we generated primary SMC explants from the Adamts7^SMCCTL and Adamts7^SMCTG mice with WT Ldlr and treated them with fluorescently labeled oxLDL (DiI-oxLDL). We chose oxLDL because it is a pathophysiologically relevant modified lipoprotein that accumulates within atherosclerotic lesions (22). Adamts7^SMCTG primary SMCs had a 38% increase in DiI-oxLDL uptake compared with Adamts7^SMCCTL as measured by median fluorescence intensity (Figure 4G). Both the proportion of DiI-oxLDL–positive cells and the intensity of staining per cell were elevated in the Adamts7^SMCTG group.

To further validate the role of Adamts7 in SMC lipid uptake, we examined primary SMCs isolated from our previously described global Adamts7 KO mice (7). After TNF-α stimulation for 72 hours to induce Adamts7 expression, the SMCs were loaded with DiI-oxLDL. KO of Adamts7 reduced the percentage of cells accumulating lipids and the amount of lipid uptake (Figure 4H), complementing the transgenic phenotype.

Adamts7 leads to an increase in lipid uptake gene expression in SMCs. To identify how ADAMTS7 mediates an increase in SMC lipid uptake, we performed bulk RNA-seq on SMCs explanted from Adamts7^SMCCTL and Adamts7^SMCTG mouse aortas 10 days after transgene induction (Figure 5A). These cells were from mice with WT Ldlr expression and that were fed a chow diet, ensuring that any changes in gene expression were due specifically to increased Adamts7 expression. Differential gene expression analysis revealed that Adamts7^SMCTG cells had increased expression of lipid uptake genes typically associated with macrophages, including Cd36, Fabp5, and Trem2, as well as the macrophage-associated marker Adgre1. Importantly, this represents a transcriptional elevation; these cells are not bona fide macrophages. These findings were validated in vivo by qRT-PCR of whole-aorta RNA from Adamts7^SMCTG mice (Figure 5B). Although the increase in macrophage-like gene expression indicates SMC phenotypic modulation, we did not observe decreases in canonical SMC contractile markers, either by RNA-seq or by qRT-PCR (Supplemental Figure 5A), even with oxLDL treatment (Supplemental Figure 5B). Similarly, the expression of Klf4, a marker typically associated with modulated SMCs, remained unchanged even with oxLDL treatment (13). Ingenuity pathway analysis (IPA) of the differentially expressed gene set identified pathways enriched within the Adamts7^SMCTG SMCs that are typically ascribed to macrophages, such as phagosome formation and immune signaling (Figure 5C). Because 1 of the hallmark functions of macrophages during atherosclerosis is perpetuating inflammation, we asked whether our Adamts7^SMCTG SMCs were more inflammatory. Cytokine analysis of conditioned media revealed no differences in IL-1β, IL-6, or TNF-α secretion (Supplemental Figure 5C). These findings indicate that Adamts7^SMCTG SMCs are not inflammatory and any atherosclerosis phenotype is likely independent of canonical inflammatory pathways. Collectively, the RNA-seq and qRT-PCR data indicate that Adamts7^SMCTG SMCs acquire a gene signature typically associated with macrophages, with an enrichment of phagocytic and lipid-uptake genes, consistent with prior reports that modulated SMCs can upregulate macrophage processes (12, 23).

Figure 5

Bulk RNA-seq of Adamts7^SMCTG primary SMCs reveals an increase in lipid-handling genes. (A) Volcano plot of the most upregulated and downregulated genes by P value adjusted for multiple comparisons. Passage 2 cells were used for RNA-seq. Adamts7 (x = 3.65, y = 122) was excluded from the plot for visualization purposes because it fell completely off scale. (B) qRT-PCR confirmation of upregulated lipid uptake and macrophage-like genes (n = 4). Normalization was performed relative to Hprt1. (C) IPA of differentially expressed genes with Ingenuity Canonical Pathways highlighted. Shown pathways are predicted to be upregulated within Adamts7^SMCTG primary SMCs. (D) Confirmation of enhanced SMC CD36 in vivo. Flow cytometry analysis of lineage traced mice after 12 weeks of WTD feeding. n = 4–6 mice. ****P < 0.0001, **P < 0.01. Statistics were analyzed using a 2-tailed Student’s t test.

Knockdown of Cd36 ameliorates enhanced oxLDL uptake conferred by Adamts7. Given the known role of CD36 as a receptor for oxLDL (24), we next investigated whether CD36 mediates the increased lipid uptake observed in Adamts7^SMCTG cells. Flow cytometric analysis was used on Adamts7^SMCTG Ldlr^KO ZsGreen^SMC+ aortas after 12 weeks of WTD feeding to examine CD36 levels in vivo. ZsGreen⁺ cells from Adamts7^SMCTG Ldlr^KO ZsGreen^SMC+ mice had a 40% increase in cells expressing Cd36 and a 2.3-fold increase in median fluorescence intensity of CD36 signal (Figure 5D), consistent with the in vitro increase in CD36 expression in Adamts7^SMCTG cells. To determine whether CD36 is functionally required for ADAMTS7-induced lipid uptake, we performed siRNA-mediated knockdown of Cd36 in primary SMCs isolated from Adamts7^SMCCTL and Adamts7^SMCTG mice. siRNA-mediated knockdown of Cd36 in primary SMCs from Adamts7^SMCTG mice reduced Cd36 expression to that of nontargeting control–treated (NTC-treated) Adamts7^SMCCTL cells (Figure 6A). Consistent with this, siCd36 treatment normalized oxLDL uptake in Adamts7^SMCTG cells to levels comparable to that of NTC-treated controls (Figure 6B). Together, these data demonstrate that ADAMTS7 promotes SMC lipid uptake through upregulation of Cd36, identifying CD36 as a key mediator of the enhanced oxLDL uptake and foam cell formation driven by ADAMTS7 during atherogenesis.

Figure 6

CD36 and PU.1 mediate ADAMTS7-conferred foam cell expansion. (A) Confirmation of knockdown of Cd36 through siRNA. Primary SMCs were treated with 10 nM of either an NTC or Cd36 siRNA. Knockdown was assessed 48 hours after transfection. n = 3–5 mice. (B) Flow analysis of oxLDL lipid uptake. 24 hours after siRNA transfection, cells were treated with 10 μg/mL of DiI-oxLDL, and uptake was assessed 24 hours after oxLDL treatment. n = 3. (C) Knockdown of Spi1 in primary SMCs. Cells were treated with 10 nM of siRNA, and 48 hours after transfection, cells were harvested for qRT-PCR. (D) Flow analysis of lipid uptake. At 24 hours after Spi1 siRNA transfection, cells were treated with 10 μg/mL DiI-oxLDL, and uptake was assessed 24 hours after oxLDL treatment. (E) Expression of SPI1 in human carotid atherosclerosis. ****P < 0.0001, ***P < 0.001, ** P < 0.01, *P < 0.05. Statistics were analyzed by 2-way ANOVA with Šidák’s multiple comparison test.

Knockdown of Spi1 in Adamts7 overexpressing SMCs attenuates the expression of macrophage-like and lipid-uptake genes. We next used IPA to identify upstream transcriptional regulators that could mediate the observed differential gene expression (Table 1) in Adamts7^SMCTG cells. The top predicted activated TF was Spi1, which encodes the protein PU.1. Spi1 itself also exhibited a 3.16-fold increase in expression in Adamts7^SMCTG cells. To test whether PU.1 is required for the Adamts7-mediated increase in lipid uptake genes, we used siRNA to achieve a 66% knockdown of Spi1 in primary SMCs. Knockdown of Spi1 abrogated the increases in gene expression of Cd36, Fabp5, Trem2, Adgre1, and Cd68 (Figure 6C). We also performed siRNA knockdown of other predicted master-regulator TFs identified by IPA and found that knockdown of Cebpa, Smarca4, Bhlhe40, and Klf6 did not ameliorate the increases in lipid uptake gene expression (Supplemental Figure 6). DiI-oxLDL loading of primary Adamts7^SMCTG SMCs with Spi1 knockdown demonstrated that loss of Spi1 attenuated the increased oxLDL uptake in Adamts7^SMCTG cells while having no effect in Adamts7^SMCCTL primary SMCs (Figure 6D). This finding suggests increased Spi1 levels resulting from Adamts7 expression are responsible for the increased oxLDL uptake observed in Adamts7^SMCTG SMCs.

Table 1

IPA upstream regulator analysis to identify candidate TFs.

Finally, we examined the level of SPI1 expression within our human single-cell data and found clear SPI1 expression in the modulated SMC populations (Figure 6E), mirroring our observations in mice and suggesting that SPI1 may contribute similarly to SMC phenotypic modulation in humans. Within these human SMCs, phenotypic modulation was also accompanied by increased expression of lipid-handling genes CD36 (19), FABP5, and TREM2, as well as the macrophage marker ADGRE1 (Supplemental Figure 7).

ADAMTS7 promotes an AP-1–dependent chromatin remodeling program in SMCs. RNA-seq analysis revealed widespread transcriptional reprogramming in Adamts7^SMCTG SMCs (Figure 5), leading to increased oxLDL uptake. Given that ADAMTS7 is a secreted extracellular protein, these changes must be driven by outside-in signaling that alters TF activity. To identify changes in TF activity, we conducted an assay for transposase-accessible chromatin using sequencing (ATAC-seq) on primary SMCs explanted from Adamts7^SMCCTL and Adamts7^SMCTG mice to map regions with altered chromatin accessibility. Differential ATAC-seq peak analysis revealed that Adamts7 overexpression broadly remodeled the chromatin landscape in SMCs, with 47,452 peaks (19.5% of all ATAC peaks) showing differential accessibility between groups (Figure 7A). TF binding motif analysis of the differential peaks identified enrichment of binding sites for FOS and JUN, members of the AP-1 TF complex (Figure 7B). Integration of ATAC-seq with RNA-seq data (Figure 5) further demonstrated that 85.8% of genes differentially expressed in Adamts7^SMCTG SMCs were associated with a nearby differential ATAC peak (±100 kb from the transcription start site), indicating a close coupling between chromatin accessibility and transcriptional reprogramming by ADAMTS7 (Figure 7C).

Figure 7

ADAMTS7 drives AP-1–dependent chromatin remodeling and transcriptional reprogramming in SMCs. (A) Differential ATAC-seq peaks in primary SMCs from Adamts7^SMCCTL and Adamts7^SMCTG mice reveal widespread chromatin remodeling. (B) Motif enrichment analysis of differential peaks identifies significant overrepresentation of FOS and JUN (AP-1) binding sites. (C) Integration of ATAC-seq and RNA-seq data show that 85.8% of genes upregulated in Adamts7^SMCTG SMCs are associated with nearby differential ATAC peaks (±100 kb from TSS. (D) Genome browser tracks at the Spi1 locus indicate increased chromatin accessibility in Adamts7^SMCTG SMCs at a predicted cis-regulatory element containing an AP-1 motif. (E) Proliferation as determined by CellTiter of primary SMCs of Adamts7 overexpression (n = 4) and Adamts7 KO cells (n = 8). (F) qRT-PCR of ERK1/2 inhibition with ulixertinib suppresses ADAMTS7-induced upregulation of Spi1, lipid-handling genes (Cd36, Trem2). (G) Ulixertinib treatment and subsequent DiI-oxLDL uptake in primary SMCs. n = 3. Statistics were analyzed using a 2-tailed Student’s t test or 2-way ANOVA with Šidák’s multiple comparison test. ****P < 0.0001, **P < 0.01, *P < 0.05.

We next focused on the Spi1 genomic locus in greater detail. Chromatin accessibility increased markedly within ±200 kb of the Spi1 transcriptional start site (TSS) (Supplemental Figure 8). Notably, a region of differential accessibility within a Spi1 intron overlaps both a predicted cis-regulatory element and AP-1 binding site, consistent with AP-1–dependent regulation of Spi1 transcription in Adamts7^SMCTG SMCs (Figure 7D). AP-1 transcriptional activity is canonically regulated by MAPK/ERK-dependent phosphorylation of FOS and JUN, which enhances their DNA-binding and transactivation potential (25). Consistent with this model, Adamts7^SMCTG SMCs displayed increased proliferation, a hallmark of MAPK/ERK pathway activation, whereas SMCs isolated from global Adamts7-KO mice exhibited slower proliferation, confirming that ADAMTS7 contributes to SMC proliferative capacity (Figure 7E).

We next tested whether inhibiting MAPK/ERK signaling could block ADAMTS7-induced phenotypes. Treatment with ulixertinib, a selective ERK1/2 inhibitor, significantly suppressed Spi1 expression (Figure 7F), reduced the induction of Cd36 and Trem2, and blunted oxLDL lipid accumulation (Figure 7G). Together, these results are consistent with a model in which ADAMTS7 promotes AP-1 activity, which, in turn, upregulates Spi1 and downstream lipid-handling genes, driving both transcriptional and phenotypic reprogramming of SMCs.

Given that ADAMTS7 is a metalloproteinase, we next investigated whether previously identified ADAMTS7 extracellular matrix substrates (THBS1, cartilage oligomeric matrix protein [COMP], LTBP4, and TIMP1) mediate the observed increases in lipid-handling gene expression in Adamts7^SMCTG SMCs. To test this, we performed siRNA knockdowns of putative substrates in primary SMCs from Adamts7^SMCCTL and Adamts7^SMCTG mice (Supplemental Figure 9A). Among these, Thbs1 knockdown most strongly attenuated the ADAMTS7-induced upregulation of Cd36, Spi1, and Trem2, whereas Fabp5 expression remained unaffected. Although it did not fully restore gene expression to baseline, this partial rescue indicated Thbs1 contributes, in part, to ADAMTS7-dependent gene activation. In contrast, Comp silencing modestly reduced the expression of select lipid-related genes with minor to no effects on Cd36, and knockdown of Ltbp4 or Timp1 produced minimal effects.

Finally, we assessed THBS1 and COMP protein levels by Western blot of whole-aorta lysates after WTD feeding. This analysis revealed the presence of cleaved THBS1 fragments in Adamts7^SMCTG lysates with no reduction in total THBS1 protein, whereas total COMP protein levels were noticeably reduced (Supplemental Figure 9B). Collectively, these results demonstrate ADAMTS7 promotes PU.1-dependent lipid uptake and transcriptional reprogramming in SMCs through a combination of extracellular matrix substrate–mediated signaling and AP-1–dependent chromatin remodeling.

GWAS have revealed more than 300 genomic regions associated with CAD (2, 26). Yet functional studies elucidating the precise mechanisms by which these loci influence disease are still limited, hampering the clinical translation of these genetic findings. In 2011, our group and others reported ADAMTS7 as a GWAS gene for CAD (4). Subsequent GWAS have replicated this signal multiple times and across different ethnic groups (2, 26), highlighting the importance of ADAMTS7 in CAD pathogenesis. Experimental studies in mice demonstrated that ADAMTS7 is proatherogenic and that its atherogenicity is conferred from its catalytic function (7, 9). Despite the sizable body of prior work on ADAMTS7, we still do not understand how ADAMTS7 participates in atherogenesis. Additionally, there have not been any cell-specific in vivo ADAMTS7 atherosclerosis studies prior to this one, to our knowledge, leaving the question of which vascular cell type(s) is critical for ADAMTS7 function. Here, we identified prominent expression of ADAMTS7 in human and mouse stromal cells and ECs and developed multiple novel mouse models in which Adamts7 can be genetically overexpressed or conditionally knocked out in a cell-specific manner. Through these models, we demonstrate that both SMC- and EC-derived expression of Adamts7 is proatherogenic, whereas KO of Adamts7 in either cell type alone is not sufficient to alter atherogenesis. These findings suggest ADAMTS7 exerts redundant proatherogenic effects across multiple cell types through a shared mechanism in which it promotes SMC foam cell formation by enhancing oxLDL uptake via upregulation of Spi1 and its downstream target Cd36.

Adamts7 has very low overall expression across all murine vascular cell types in the basal state, and its expression is induced in response to vascular injury, such as endothelial denudation or prolonged hyperlipidemia (7, 15). In primary mouse aortic SMCs, Adamts7 induction is required to confer a detectable migration phenotype (7), confirming that this induction is critical for ADAMTS7 function. At terminal time points in mouse atherosclerosis studies, ADAMTS7 is again undetectable in plaques (7, 9), suggesting that Adamts7 induction is transient and ADAMTS7 is mechanistically active in mice during plaque maturation. The transient induction of Adamts7 has made studying its exact function in vascular cells difficult. The novel Adamts7 mouse models presented here overcome this limitation, demonstrating that sustained Adamts7 expression in either SMCs or ECs is sufficient to drive atherosclerosis progression.

Although neither Adamts7^SMCKO Ldlr^KO nor Adamts7^ECKO Ldlr^KO affected atherosclerosis, prior research on whole-body KO models indicates Adamts7 is proatherogenic. These prior studies include 3 unique models of whole-body Adamts7 inactivation. Two models used genetic perturbation to inactivate ADAMTS7 by either exon deletion or mutation of the ADAMTS7 catalytic site (7, 9); the third approach was to vaccinate against ADAMTS7 (10). Although prior findings clearly demonstrated that Adamts7 is expressed in cultured SMCs (7, 9, 15, 27), our RNAscope studies of Adamts7 expression revealed its induction in multiple vascular cell types during murine atherosclerosis, yet only a small subset of vascular cells express Adamts7 at any given time. Although we cannot exclude the existence of a yet-unidentified dominant producer cell type, the reproducible multicellular distribution across independent datasets (19, 28) supports the view that Adamts7 is broadly expressed and that its proatherogenic effects likely arise through cooperative or redundant mechanisms among vascular cells. Consistent with this, our data suggest that loss of Adamts7 in a single cell type is insufficient to blunt atherogenesis, whereas broad inactivation across multiple compartments as in whole-body KO reduces lesion burden (7, 9). Determining which combination of cell type–specific ADAMTS7 expression is required to drive atherosclerosis will necessitate future double-KO models using multiple Cre drivers.

Exactly how these findings translate to human atherosclerosis is another open question. Our analysis of the most extensive carotid-plaque single-cell dataset to date highlights that ECs and SMCs both express ADAMTS7, consistent with our murine findings and supporting the concept of multicellular expression in advanced disease. This scRNA-seq data also show that mast cells and fibroblasts have detectable ADAMTS7 expression, highlighting these cell types for further ADAMTS7 studies. Notably, the scRNA-seq data were from mature atherosclerotic plaques, whereas our mouse data suggest the window for ADAMTS7 activity is earlier in plaque progression. Thus, whether this human expression pattern reflects relevant ADAMTS7 activity remains to be determined. Moreover, the therapeutic potential of targeting ADAMTS7 in established lesions is unknown. Nonetheless, the human data indicate ADAMTS7 is active in multiple vascular cell types, and our mouse data show that sustained ADAMTS7 activity in individual cell types can drive atherogenesis.

From a clinical perspective, much of the mechanistic understanding of ADAMTS7 function in atherosclerosis has been derived from murine models. Future studies examining functionally relevant ADAMTS7 genetic variation, including the previously reported S214P common variant (27), as well as yet-unidentified loss-of-function variants that impair catalytic activity, may help bridge mechanistic insights such as those described here with human disease pathology.

Our SMC-specific and EC-specific Adamts7 overexpression mouse models have increased atherosclerotic burden and aortic foam cell formation, and these foam cells are SMC derived. Although macrophages have historically been thought to be the source of foam cells in atherosclerotic lesions, it is increasingly appreciated that SMCs constitute most foam cells in atherosclerosis (14). Elegant fate mapping and integrative genomic studies have revealed that SMCs can transition into diverse phenotypic states, including inflammatory, fibroblast-like, and progenitor-like cells, through distinct transcriptional programs regulated by TFs such as KLF4, OCT4, TCF21, and retinoid signaling (11). Our findings raise the possibility that PU.1 is another TF directing SMCs specifically toward the foam cell fate. Consistent with this, when Adamts7 is induced in the vasculature, SMCs increase their expression of genes involved in lipid uptake, inflammation, and phagocytosis, a gene expression profile reminiscent of a macrophage. Recent studies have definitively shown that SMCs do not contribute to the pool of mature macrophages in atherosclerosis (29); therefore, we posit that these Adamts7^SMCTG SMCs are modulated SMCs with increased lipid-uptake capacity and foam cell features. Although the specific role of SMC foam cells in atherosclerosis remains unclear, it has been reported that SMC foam cells and true macrophages differ in their intracellular lipid metabolism (30, 31). Lipid-laden foamy macrophages tend to be less inflammatory (32), and by becoming foamy, these macrophages may no longer contribute to the inflammatory cycle associated with atherosclerosis, raising the possibility that foamy macrophages are beneficial in atherosclerosis. Conversely, we observed reduced fibrous cap thickness with Adamts7 overexpression. Because 1 well-described role of SMCs is forming the fibrous cap, it is possible that foamy SMCs cannot stabilize lesions, raising an intriguing possibility that macrophage foam cells are beneficial, whereas SMC foam cells are detrimental. Sharifi et al. showed that ADAMTS7 degrades TIMP1, thus increasing MMP2 and MMP9 activity (33). Consistent with that model, ADAMTS7 whole-body KO increased collagen content and plaque stabilization, and ADAMTS7 expression was higher in caps of unstable human carotid plaques (33). These prior observations and those presented here support the adverse role of ADAMTS7 in plaque stability and suggest multiple mechanistic explanations for ADAMTS7-mediated plaque instability.

Mechanistically, we chose to pursue the function of SMC Adamts7, given the large body of existing data on Adamts7 modulation of SMC function. We find that constitutive Adamts7 SMC expression markedly alters lipid uptake gene programs, resulting in enhanced oxLDL accumulation. Among these lipid uptake genes, 1 of the most significantly upregulated genes was Cd36. CD36 is a well-described receptor for oxLDL, and our studies demonstrate that CD36 is mainly responsible for the SMC foam cell phenotype in Adamts7^SMCTG mice (24). In addition, upstream regulator analysis of primary SMCs identified Spi1 as a potential mediator of expression changes (34), and knockdown of Spi1 in Adamts7^SMCTG SMCs ameliorated the increase in lipid-handling gene expression and macrophage markers. Spi1, which encodes the TF PU.1, is well recognized as a critical mediator of hematopoietic lineage commitment. These observations suggest PU.1 may also be a crucial factor in driving the in vivo production of foamy SMCs. Thus, future cell-specific studies of SMC PU.1 may help reveal disease-relevant consequences of foamy SMCs.

Using ATAC-seq, we observed that ADAMTS7 enhances chromatin accessibility and activates AP-1–dependent transcription. AP-1 activity is regulated by MAPK signaling via ERK1/2, and pharmacologic inhibition of ERK1/2 reduced ADAMTS7-induced expression of lipid-handling genes. These findings support a model in which extracellular ADAMTS7 activates MAPK-ERK1/2 signaling, modulating AP-1 activity and altering SMC gene expression profiles. Importantly, this suggests ADAMTS7-mediated cleavage of an extracellular substrate triggers MAPK signaling cascades. ADAMTS7 substrates have been proposed in prior studies (8, 33, 35, 36), but none have been tested in the context of SMC foam cell formation. To explore this, we examined several previously proposed substrates and found that Thbs1 knockdown partially attenuated ADAMTS7-associated lipid gene expression. Western blot analysis of aortic lysates revealed cleaved THBS1 fragments without a reduction in total THBS1 protein, while COMP levels were modestly reduced, suggesting differential substrate processing in vivo. Notably, Comp silencing also suppressed Spi1 expression, raising the possibility that COMP may contribute to ADAMTS7-mediated oxLDL uptake. Together, these findings support a model in which ADAMTS7-dependent substrate processing, potentially involving both THBS1 and COMP, contributes to downstream signaling. The identification of ADAMTS7 cleavage substrates remains an active area of investigation. Although THBS1 and COMP have been reported as ADAMTS7 substrates (17, 37), not all studies have reproduced these findings (35, 36), and the relative contribution of these substrates remains undefined. Definitive evaluation of these substrates is necessary because identifying biologically meaningful cleavage products may reveal functional therapeutic biomarkers and inform the development of targeted ADAMTS7-directed therapies.

In conclusion, our study establishes that ADAMTS7 promotes SMC foam cell formation and atherosclerosis through PU.1 and CD36 and that its expression is not confined to a single vascular cell type. By integrating human single-cell data with new cell-specific genetic models, we show that ADAMTS7 is distributed across diverse vascular populations and that its sustained activation, whether from SMCs or ECs, drives disease progression. These findings reveal a unifying mechanism linking human GWAS data with cellular pathophysiology, reinforcing ADAMTS7 inhibition as a potential therapeutic target for atherosclerotic cardiovascular disease.

Sex as a biological variable. Our human data contained both male and female samples. SMC mouse studies were limited to male mice, due to the integration of the Myh11-CreER^T2 onto the Y-chromosome.

Animals. C57BL/6J (catalog 000664), Ldlr^KO (catalog 002207), ZsGreen (catalog 007906), and Myh11-CreER^T2 (catalog 019079) were purchased from The Jackson Laboratory and bred in our laboratory. Adamts7 KO mice were previously described (7). The Cdh5-CreER^T2 mouse was a gift from Carol Troy (Columbia University, New York City, New York, USA) and developed by the laboratory of Ralf Adams (Max Planck Institute for Molecular Biomedicine, Münster, Germany), and has been previously described (38). To generate a floxed Adamts7, the Adamts7 KO mouse (MMRRC, 046487) was bred to an FLP recombinase mouse. Adamts7 transgenic mice were created by cloning the mouse allele of Adamts7 into the pR26 CAG/GFP Asc plasmid (Addgene plasmid 74285). Subsequently, FL19 embryonic stem cells were used to generate chimeras, which were then backcrossed to the C57BL/6J background. For CreER^T2 activation, tamoxifen (MilliporeSigma, T5648) was dissolved in 90% corn oil and 10% ethanol; subsequently, mice at 7 weeks of age were injected with tamoxifen for 5 days at a dose of 40 mg/kg body weight per day. At 8 weeks of age, the mice were fed a WTD consisting of 40% kilocalories by fat and 0.15% cholesterol (Research Diets, D12079Bi). For blood collection, mice were fasted for 4 hours and subsequently sedated with isoflurane, then blood was collected by retro-orbital puncture with heparinized microcapillary tubes. Plasma was isolated by centrifugation at 1,500 relative centrifugal force (RCF) for 10 minutes at 4°C. Total cholesterol was measured using the Infinity Cholesterol Reagent (Thermo Fisher Scientific, TR13421).

Atherosclerosis analysis. For atherosclerosis studies, mice were sacrificed by cervical dislocation and perfused with PBS. The aorta was dissected and subsequently fixed in 10% formalin. Whole aortas were stained with ORO. Hearts containing the aortic root were fixed in 4% paraformaldehyde for 24 hours, transferred to 70% ethanol, and embedded in paraffin wax. Sectioning was performed using a Leica microtome at 8 μm thickness. The heart was sectioned from the inferior heart toward the aortic root. We collected 25 slides; each slide contained 2 sections. We aimed to have the breakage of the leaflet occur on slide 13. Slides 3, 8, 13, 18, and 23 were used for H&E staining to allow for lesion size quantification using ImageJ (NIH). H&E and Picrosirius red staining was performed at Columbia’s Molecular Pathology Shared Resource.

Immunofluorescence. Immunofluorescence was performed on paraffin sections of the aortic root. Deparaffinization was performed using xylene. Antigen retrieval was performed using a citrate-based agent. Sections were subsequently blocked using 10% normal goat serum for at least 1 hour. The sections then were stained using the following Abs and dilutions: 1:1,000 α-SMA-Cy3 (MilliporeSigma, C6198) and 1:500 MAC2 (Cedarlane, CL8942AP). IgG controls were purchased from Jackson ImmunoResearch. The sections were washed with PBS containing 0.01% Tween 20. For the secondary Ab, a donkey anti–rabbit Alexa Fluor 647 (Invitrogen, Thermo Fisher Scientific, A-31573) and goat anti–mouse Alexa Fluor 546 (Invitrogen, Thermo Fisher Scientific, A-11030) were used, both at a 1:1,000 dilution in 1% goat serum. The slides were washed with DAPI and mounted with ProLong Glass Antifade Mountant with NucBlue Stain. Imaging was conducted on a Nikon Eclipse Ti-S microscope, and image processing was performed using ImageJ. ZsGreen fluorescence was confirmed separately on unfixed frozen sections without additional Ab staining to verify reporter expression.

Vascular SMC isolation for tissue culture. SMCs were isolated as previously described (39). The aorta was dissected from the ascending arch to the diaphragm and predigested for 10 minutes at 37°C with 175 U/mL collagenase II and 1.25 U/mL elastase in HBSS to allow for the removal of the adventitia. After removing the adventitia, the media layer was further digested in 400 U/mL collagenase II, 2.5 U/mL elastase, and 0.2 mg/mL soybean trypsin inhibitor at 37°C for 1 hour in a rotating incubator. The cells were cultured in DMEM supplemented with 20% FBS, 1% l-glutamine, 1% sodium pyruvate, and 1% penicillin-streptomycin. Cells between passages 2 and 7 were used for experiments.

Vascular SMC isolation for flow analysis. Mice were sacrificed and perfused with PBS. The ascending aorta to the bifurcation at the common iliac artery was dissected and digested with a cocktail consisting of 4 U/mL Liberase (MilliporeSigma, 5401127001), 60 U/mL hyaluronidase (MilliporeSigma, H3506), and 60 U/mL DNase I (Worthington Biochemical, LS006333) in RPMI-1640. The digestion enzymes were neutralized with 10% FBS in RPMI and washed once in FACS buffer (2% FBS, 5 mM EDTA, 20 mM HEPES, and 1 mM sodium pyruvate in Dulbecco’s PBS). Cells were then blocked with TruStain FcX PLUS anti-mouse CD16/32 (BioLegend, 156603) for 10 minutes at 4°C. Staining was performed in FACS buffer at the following dilutions: 1:1,000 LipidTOX Red (Thermo Fisher Scientific, H34476), 1:100 CD36 Super Bright 600 (Invitrogen, Thermo Fisher Scientific, 63-0362-82), and 1:200 CD45-Alexa Fluor 700 (BioLegend, 157209). Cell viability was assessed with DAPI. To establish the LipidTOX-high population and thus foamy cells, a C57BL/6J mouse with normal lipidemia was used as previously described (14).

In vitro cell culture experiments. For in vitro foam cell analysis, cells were treated with DiI-oxLDL (Invitrogen, Thermo Fisher Scientific, L34358) at a concentration of 10 μg/mL for 24 hours. Subsequently, they were washed with PBS and detached using 0.25% trypsin EDTA. Neutralization of trypsin was performed using FACS buffer. The cells were counterstained with DAPI to assess live cells, and all analyses were done using either a Novocyte Quanteon or Penteon. For siRNA experiments, all siRNAs were purchased from Integrated DNA Technologies (Supplemental Table 2). Transfection was performed using RNAiMAX according to the manufacturer’s protocol. All siRNAs were used at a concentration of 10 nM.

RNA isolation and qRT-PCR. For tissue samples, specimens were collected in TRIzol, and homogenization was performed using a TissueLyser (Qiagen). The RNA was extracted using the Direct-zol RNA MiniPrep Kit (Zymo Research) according to the manufacturer’s instructions. For cells, RNA was collected using the Quick-RNA Microprep Kit (Zymo Research). Complementary DNA was synthesized using a High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems). All qPCR was performed using TaqMan probes (Supplemental Table 2) and performed on the Applied Biosystems QuantStudio 7 Flex Real-Time PCR System. Analysis was performed using the comparative Ct method.

RNAscope. RNAscope was performed according to the manufacturer’s instructions. RNAscope was performed on sections of mouse BCAs after specified durations of WTD feeding. The BCAs were freshly frozen and sectioned at 8 μm thickness. The sections were first fixed in 10% neutral buffered formalin. Subsequently, the sections were dehydrated and treated with hydrogen peroxide and protease. Hybridization with Adamts7 (Advanced Cell Diagnostics, 533341) and Pecam1 (Advanced Cell Diagnostics, 316721-C3) probes was performed. After probe hybridization, hybridization with AMPs 1-3 was performed. Signal development was conducted using Akoya Biosciences Opal Reagent Kits, and slides were mounted using Prolong Glass Antifade Mountant (Thermo Fisher Scientific, P36984). Imaging was conducted on a Nikon Eclipse Ti-S microscope, and image processing was performed using ImageJ.

Western blot. Samples were lysed in RIPA buffer (Thermo Fisher Scientific, 89900) supplemented with 1% Halt Protease and Phosphatase Inhibitor Cocktail. Tissue samples were lysed with the Qiagen TissueLyser and samples were subsequently cleared by centrifugation at 20,000 RCF for 10 minutes at 4°C. Samples were denatured, resolved on a 4%–12% Bis-Tris gel, and transferred to either PVDF or a nitrocellulose membrane. Blocking was performed with 5% milk, and primary Abs were diluted 1:1,000 in 5% milk and incubated overnight at 4°C. The following primary Abs were used: β-actin (Cell Signaling Technology, 5125S), thrombospondin 1 (D7E5F) rabbit mAb (Cell Signaling Technology, 37879S), GAPDH (14C10) rabbit mAb, HRP-conjugated (Cell Signaling Technology, 3683S), anti-COMP (Abcam, ab42225), and rabbit Ab ADAMTS7 (custom-made by 21st Century Biochemicals). The custom polyclonal rabbit anti-ADAMTS7 Ab was generated using standard immunization protocols with recombinant mouse ADAMTS7 as the antigen. Membranes were washed with 0.1% TBST and probed with the appropriate secondary Ab for 1 hour at room temperature. Membranes were visualized with ECL using the Amersham Imager 600 gel imager.

RNA-seq of primary mouse vascular SMCs. Primary SMCs were isolated as described above and cultured for 2 passages spaced 1 week apart. Total RNA was extracted using the Zymo Microprep Kit, and RNA integrity was assessed using TapeStation. No cell sorting was performed. Polyadenylated mRNA was purified from total RNA using a poly(A) pull-down and reverse-transcribed into cDNA, with the final PCR step performed using KAPA HiFi HotStart Ready Mix. Libraries were sequenced on the Element AVITI sequencer with 75 bp paired-end reads, targeting approximately 40 million reads. Transcript abundance was quantified using kallisto with the GRCm38 mouse transcriptome, and differential expression analysis was performed using DEseq2 via DEBrowser. Up- and downregulated genes are listed in Supplemental Table 1.

ATAC-seq. ATAC-seq was performed as previously described (40) with minor modifications. Briefly, 100,000 passage 3 primary mouse SMCs with greater than 90% viability were washed with PBS and pelleted at 500g for 5 minutes at 4°C. Cells were lysed in cold lysis buffer (10 mM Tris-HCl pH 7.4, 10 mM NaCl, 3 mM MgCl₂, 0.1% NP-40, 0.1% Tween-20, and 0.01% digitonin). Nuclei were washed with 1 mL of ATAC lysis buffer without NP-40 or digitonin and centrifuged at 500g for 10 minutes at 4°C. The nuclei pellet was resuspended in transposition mix containing 25 μL of 2× TD buffer (Illumina Nextera DNA Library Prep Kit), 2.5 μL of Tn5 transposase, 16.5 μL of PBS, 0.5 μL of 1% digitonin, 0.5 μL of 10% Tween-20, and 5 μL of nuclease-free water. Transposition was carried out at 37°C for 30 minutes in a thermomixer. DNA was purified using the Zymo DNA Clean & Concentrator-5 Kit. Transposed DNA was amplified using NEBNext High-Fidelity 2× PCR Master Mix (New England Biolabs, M0541) with Nextera indexing primers, and libraries were purified with AMPure XP magnetic beads (Beckman Coulter, A63880). Library size distribution and quality were assessed using an Agilent Bioanalyzer, and DNA concentration was measured with a Qubit dsDNA HS Assay Kit (Thermo Fisher Scientific). Paired-end 150 bp sequencing was performed on an Illumina platform by Azenta/Genewiz. Sequencing reads were quality-trimmed with Trimmomatic, version 0.38, aligned to the mm10 reference genome using Bowtie2, and filtered for uniquely mapped, high-quality (MAPQ ≥30) nonduplicate reads with SAMtools, version 1.9, and Picard, version 2.18.26. Reads mapping to mitochondrial DNA and unplaced contigs were excluded. Peak calling was performed using MACS2, version 2.1.2, with peaks overlapping ENCODE blacklisted regions removed. Peaks detected in at least 66% of biological replicates per group were retained for downstream analysis. Differential chromatin accessibility was determined using DiffBind (R package). Sequencing yielded a total of 528 million paired-end reads (mean Q score = 36.9; 85% bases ≥Q30). After alignment (mean alignment rate = 98%) and filtering, approximately 60–100 million high-quality read pairs per sample were retained for downstream analyses.

Statistics. All statistical tests are indicated in figure legends. P values were calculated using a 2-tailed Student’s t test in GraphPad Prism 10 (GraphPad Software). P < 0.05 was considered significant. We used 2-way ANOVA analysis with multiple corrections for comparison between groups with post-analysis as indicated in the figure legends. Aortic root histology, qPCR, and ELISA results are presented as the mean ± SEM. All other data are presented as mean ± SD.

Study approval. All mouse experiments were performed according to procedures approved by Columbia University’s IACUC under protocol AABU8650. The IRB of Columbia University approved all human studies under IRB approvals AAAJ2765 and AAAR6796, with written informed consent provided by all participants.

Data availability. Human carotid artery scRNA-seq data were previously published (19). ATAC and RNA-seq data have been deposited in the NCBI’s Gene Expression Omnibus (GEO) database (GSE316135). All individual values represented in graphs are provided in the Supporting Data Values file.