PDGFRβ signaling regulates local inflammation and synergizes with hypercholesterolemia to promote atherosclerosis

Platelet-derived growth factor (PDGF) is a mitogen and chemoattractant for vascular smooth muscle cells (VSMCs). However, the direct effects of PDGF receptor β (PDGFRβ) activation on VSMCs have not been studied in the context of atherosclerosis. Here, we present a new mouse model of atherosclerosis with an activating mutation in PDGFRβ. Increased PDGFRβ signaling induces chemokine secretion and leads to leukocyte accumulation in the adventitia and media of the aorta. Furthermore, PDGFRβD849V amplifies and accelerates atherosclerosis in hypercholesterolemic ApoE−/− or Ldlr−/− mice. Intriguingly, increased PDGFRβ signaling promotes advanced plaque formation at novel sites in the thoracic aorta and coronary arteries. However, deletion of the PDGFRβ-activated transcription factor STAT1 in VSMCs alleviates inflammation of the arterial wall and reduces plaque burden. These results demonstrate that PDGFRβ pathway activation has a profound effect on vascular disease and support the conclusion that inflammation in the outer arterial layers is a driving process for atherosclerosis.


INTRODUCTION
Atherosclerosis is a disease of the arteries that remains one of the leading causes of death and disability worldwide. It is characterized by the formation of fatty-fibrous plaques at specific arterial sites through a decades-long process driven by lipoprotein deposition in the vessel wall. This stimulus initiates a cascade of events including leukocyte recruitment and foam cell formation, vascular smooth muscle cell (VSMC) proliferation, changes in extracellular matrix, angiogenesis, calcification, and cell death [1][2][3] . Most plaques are asymptomatic unless they obstruct blood flow. Other plaques are vulnerable to rupture, which can trigger thrombosis and a deadly risk of sudden heart attack or stroke. However, the interplay of these disease events is not fully understood and is still being explored mechanistically.
Animal models have been invaluable for understanding disease mechanisms in atherosclerosis. The most widely used models are mice with hypercholesterolemia caused by deficiency for apolipoprotein E (ApoE −/− ) or LDL receptor (Ldlr −/− ). These animals develop atherosclerosis over several months compared to decades in humans. However, they do not form the types of advanced plaques that are clinically relevant in humans, nor do they typically develop coronary plaques or experience spontaneous death from heart attack or stroke [4][5][6] . Therefore, our understanding of advanced plaques is mostly based on descriptive findings from human autopsy studies. To develop a mechanistic understanding of advanced disease processes it is important to generate improved mouse models.
Advanced atherosclerotic plaques develop a morphologically distinct fibrous cap composed of VSMCs that proliferate and migrate from the vessel wall into the plaque [7][8][9] . Plateletderived growth factor (PDGF) was first purified over thirty years ago as a serum-derived factor that stimulates VSMC proliferation and migration 10,11 . PDGF is also a potent inducer of VSMC phenotype switching, or dedifferentiation, wherein VSMCs change from a mature, contractile phenotype to a proliferative, secretory phenotype 8,12,13 . PDGF signaling occurs via PDGF receptors (PDGFRs), which are prototypical of receptor tyrosine kinases. In the absence of PDGF ligand, PDGFR tyrosine kinase activity is inhibited by an autoinhibitory allosteric conformation; ligand binding alters the conformation to allow physiological signaling activity. PDGF signaling via PDGFRβ has been associated with vascular disease and is suspected to play a major role in VSMC migration and proliferation during atherosclerosis 14,15 . In ApoE −/− mice, blockade of the PDGF-PDGFRβ pathway with neutralizing antibodies or chemical inhibitors reduced VSMC involvement in atherosclerosis 16,17 . It seems likely that increased PDGF signaling would stimulate more VSMC involvement in atherosclerosis and potentially drive the formation of more aggressive atherosclerosis in mice. This approach has not been explored previously.
Here, we take advantage of knockin mice having a Cre-inducible D849V point mutation in the endogenous PDGF receptor β gene (PDGFRβ D849V ) 18 , which we activate with a Sm22α-Cre driver to initiate signaling in VSMCs. The D849V mutation destabilizes the autoinhibited conformation of the receptor and thereby increases basal and ligand-mediated PDGFRβ signaling. We show that PDGF signaling increases chemokine signaling from VSMCs and causes inflammation of the aorta, in addition to causing VSMC phenotype switching and architectural changes in the muscle layer of the aorta. Using this model we test the hypothesis that PDGF-driven changes in VSMC phenotype and function synergize with hypercholesterolemia to promote the formation of advanced atherosclerotic plaques. Surprisingly, increased PDGF signaling not only promotes the formation of advanced plaque morphologies but it also enhances the initiation of new plaques at a plaque-resistant site, the thoracic aorta. Cells respond to PDGF signaling by phosphorylating transcription factors of the signal transducer and activator of transcription (STAT) family, and STAT1 in particular is known to regulate chemokine gene expression. Experiments with the kinase inhibitor Ruxolitinib suggest that PDGFRβ phosphorylation of STAT1 requires the activity of JAK1/2 tyrosine kinases. By deletion of a STAT1-floxed allele in VSMCs, we show that PDGF-driven chemokine signaling and inflammation are STAT1-dependent. Finally, in the context of PDGF-driven atherosclerosis, deletion of STAT1 from VSMCs reduces plaque formation, demonstrating that inflammation of the arterial media and adventitia are important mechanisms by which PDGF signaling promotes atherosclerosis.

PDGF signaling induces chemokine production from VSMCs
We previously showed that a Cre-inducible gain-of-function knockin mutation in PDGFRβ (PDGFRβ D849V ) induced chemokine mRNA expression in brain pericytes and caused inflammation of the central nervous system 18 . To characterize chemokine expression specifically in VSMCs of the aorta, which was not previously tested, we generated Sm22α-Cre;PDGFRβ +/D849V (PDGFRβ Sm22D849V ) conditional mutant mice and Sm22α-Cre;PDGFRβ +/+ littermate controls (designated as Wt) (Fig. 1a). We then established primary VSMC cultures from mutant and control aortas and examined the expression of chemokines that we previously identified by microarray. Using quantitative RT-PCR (qRT-PCR) we found upregulated mRNA levels for chemokines known to regulate monocyte (CCL2, CCL3, CCL5, CCL6, CCL7, CCL9, CCL12) and T cell recruitment (CXCL9, CXCL10, CXCL11) (Fig. 1b). We also detected increased chemokine secretion in conditioned media from mutant VSMCs (Fig. 1c). Mouse plasma contains much lower levels of chemokines compared to conditioned media, but even with this limitation we detected increased levels of circulating CCL2 and CCL3 in the plasma of PDGFRβ Sm22D849V mice (Fig. 1d). Chemokine expression could be induced by treating control VSMCs with PDGF-BB (Fig. 1e). However, other known inflammatory signals, namely IL-1β, TNF-α, IFN-α, IFN-β and IFN-γ, were not upregulated in mutant VSMCs compared to controls (by qRT-PCR, data not shown). Thus, PDGFRβ signaling in VMSCs induces several chemokines, many of which are known to be involved in atherosclerosis 19 .

PDGF signaling in VSMCs causes inflammation of the aorta
Based on our finding of increased chemokine secretion by PDGFRβ Sm22D849V VSMCs, we sought to assess the resident leukocyte populations in the wall of the aorta. By flow cytometric analysis of enzymatically digested control and mutant thoracic aortas, we found that mutants carried a 7-fold increase in the total number of CD45 + leukocytes (Fig. 2a, b and Supplementary Figure 1). The majority of the leukocytes were found in the adventitia, which is a normal site for small populations of leukocytes to reside in a non-atherosclerotic aorta (Fig. 2c) [20][21][22] . Histological analysis also provided evidence of abundant accumulated leukocytes in the mutant aorta adventitia and media (Fig. 2d). Further flow-cytometric analysis revealed increased numbers of leukocytes expressing CD3, CD19, CD11b, CD11c, and NK1.1 in the mutant aortas (Fig. 2e). These data suggest that PDGF signaling in VSMCs induces signals, likely to be chemokines, which create an inflammatory milieu in the wall of the thoracic aorta. To investigate whether short-term pharmacological inhibition of PDGFRβ could reduce inflammation in the vessel wall, we treated PDGFRβ Sm22D849V mice with Crenolanib, a tyrosine kinase inhibitor specific for class III receptors. After 5 days of treatment, we found a decrease in the number of leukocytes in mutants that received the drug compared to untreated mutants (Fig. 2f, g). This suggests that continual PDGFRβ activation sustains inflammation.

Constitutive dedifferentiation of VSMCs
As suggested previously 18 , PDGFRβ Sm22D849V mutant mice have constitutively dedifferentiated VSMCs and exhibit VSMC hyperplasia (Fig. 3a), decreased expression of contractile proteins SMMHC, SM22α, and αSMA (Fig. 3B), and increased extracellular matrix (ECM) production ( Fig. 3c, d). Further analysis showed a cellular ultrastructure consistent with a phenotypic switch, where instead of contractile bundles there was an abundance of rough endoplasmic reticulum in mutant VMSCs (Fig. 3e). Due to VSMC hyperplasia and stiffening of the vessel wall, the thoracic aorta in PDGFRβ Sm22D849V mutants undergoes adaptive remodeling to become approximately 2-fold larger in diameter by 4 weeks of age, as shown previously 18 . We found no significant difference in blood pressure between adult mutants and controls (mutant: systolic 96.6±2.8/diastolic 73.3±5.8 mmHg; control: systolic 108.3±8.3/diastolic 77.6±4.2 mmHg; mean ± s.e.m., n=6 mice per genotype).

PDGF signaling amplifies atherosclerosis
Inflammation, VSMC dedifferentiation, and aortic enlargement in PDGFRβ Sm22D849V mice provided clues that they might be more susceptible to atherosclerosis than normal mice. In particular, vascular inflammation is an early event in atherosclerosis, and VSMC dedifferentiation is characteristic of later disease. However, the most important risk factor for atherosclerosis is high cholesterol and PDGFRβ Sm22D849V mice had normal cholesterol and triglyceride levels compared to control counterparts (Table I). Accordingly, we did not detect atherosclerotic plaques up to 1 year of age whether the mutants were fed a normal chow or Western diet (WD). Therefore, to produce hypercholesterolemia we backcrossed PDGFRβ Sm22D849V mice onto an ApoE-deficient genetic background to generate ApoE −/− ;Sm22α-Cre;PDGFRβ +/D849V (ApoE,PDGFRβ Sm22D849V ) mutant mice and ApoE −/− ;Sm22α-Cre;PDGFRβ +/+ (ApoE) littermate controls. Thus, serum lipid profiles increased as expected (Table I).
After 16 weeks of WD we assessed atherosclerotic plaque formation by two common methods: Oil red O staining for plaques on the luminal surface of flat-mounted aortas, and cross sectional analysis of plaques in the aortic root. Both approaches revealed a dramatic increase in plaque formation in ApoE,PDGFRβ Sm22D849V mice. Flat-mounted ApoE aortas exhibited plaques in the aortic arch, the brachiocephalic artery (BCA), and occasionally at branch points to the renal arteries, which are typical sites of plaque formation. In contrast, ApoE,PDGFRβ Sm22D849V aortas exhibited confluent plaques throughout the aortic tree, including the typically plaque-resistant thoracic aorta (Fig. 4a). Similarly, plaques in the aortic root were larger in the ApoE,PDGFRβ Sm22D849V mice and they exhibited an advanced morphology with well-defined fibrous caps overlying acellular necrotic cores (Fig. 4b).
Necrotic core area was much larger and fibrous cap thickness was reduced in ApoE,PDGFRβ Sm22D849V aortic root plaques compared to ApoE (Supplementary Figure 2b, c). ApoE,PDGFRβ Sm22D849V aortic root plaques were enriched for macrophages as a percent of plaque area (Supplementary Figure 2a). Gene expression analysis suggested that macrophages in ApoE,PDGFRβ Sm22D849V plaques were more of the M1 (pro-inflammatory) phenotype than M2, in comparison to ApoE (Supplementary Figure 3).
To carefully track plaque progression over time we quantified cross-sectional plaque area after 6, 8, 16, and 24 weeks of WD, measuring plaques in the aortic root, aortic arch, BCA, and thoracic aorta (Fig. 4c). At all time points and all sites examined, plaques were significantly larger in ApoE,PDGFRβ Sm22D849V aortas. We also performed quantification of cross-sectional plaque area in a cohort of mice with chow diet feeding, where moderate hypercholesterolemia results in slower plaque development. We examined time points between 6 and 12 months of age, and found that ApoE,PDGFRβ Sm22D849V aortas still developed distinctive plaques in the thoracic aorta and larger plaques at other sites compared to ApoE controls (Fig. 4d).
We wanted to ensure that the synergistic effect of PDGF signaling and hypercholesterolemia was not specific to the ApoE mutant. Therefore, we crossed PDGFRβ Sm22D849V mice with mice deficient for low-density lipoprotein receptor (LDLR), which also causes hypercholesterolemia. We compared Ldlr and Ldlr,PDGFRβ Sm22D849V mutants after 16 weeks of WD feeding and again found a striking increase in atherosclerotic plaque area in Ldlr,PDGFRβ Sm22D849V aortas (Fig. 4e, f). This finding is important because ApoE itself can influence PDGFRβ activity by interacting with LDL receptor-related protein-1 (LRP1) to attenuate PDGFRβ function 23,24 . The finding that PDGFRβ D849V synergizes with Ldlr argues that the effects shown here are not attributable to lack of inhibition from the ApoE-LRP1 complex. Taken together, the results from ApoE/WD, ApoE/chow, and Ldlr/WD studies clearly demonstrate that increased PDGF signaling in VSMCs strongly amplifies atherosclerosis under a variety of hypercholesterolemic conditions.

Plaque progression to fibroatheroma and hemorrhage
We performed careful morphological assessment of the thoracic aorta plaques in ApoE,PDGFRβ Sm22D849V mice because of their striking size and unusual location. At 6 weeks of WD the plaques initiated as accumulations of foam cell macrophages within the intima (Fig. 5a, Supplementary Figure 4). By 8 weeks of WD, smooth muscle cells were seen invading the foam cell-rich plaques and fibrous cap structures began to form at this time ( Fig. 5b, black arrows). By 16 and 24 weeks of WD, the plaques had reached a fibrous cap atheroma stage 4 , exhibiting well-formed collagen-rich fibrous caps and large necrotic cores (Fig. 5c, d, e). An additional feature seen at 16 weeks WD and later was areas of hemorrhage within advanced fibroatheroma plaques (seen in 7 of 10 mutants, Fig. 5f).
Cross-sectioning through such areas showed that hemorrhages were contained within the plaque without evidence of a ruptured fibrous cap (Fig. 5g). Iron deposition at hemorrhage sites was evident by Prussian blue stain (Fig. 5h). In advanced human atherosclerosis, neovascularization of plaques is associated with increased inflammation, intraplaque hemorrhage, and increasing plaque vulnerability 25,26 . Accordingly, we observed PECAM + microvessels in advanced plaques of ApoE,PDGFRβ Sm22D849V mice (Fig. 5i). Together, these data demonstrate that ApoE,PDGFRβ Sm22D849V mice readily develop plaques with advanced morphologies much like those seen in humans. Fibroatheromas with these features are extremely rare (or nonexistent) in standard ApoE and Ldlr-deficient mouse models.

Coronary atherosclerosis
Coronary artery plaques are not typical in ApoE single mutant mice unless they are fed a WD for 10 months 27,28 . We observed coronary atherosclerosis in some of the PDGFRβ D849V cohorts. This was observed in 70% (7 of 10) ApoE,PDGFRβ Sm22D849V mice that were fed a chow diet for 10-12 months (Fig. 6a, b) and in 71% (5 of 7) Ldlr,PDGFRβ Sm22D849V mice that were fed WD for 16 weeks (Fig. 6c, d). Some of these plaques appeared to be fibroatheromas (Fig. 6b). Sudden deaths occurred in some ApoE,PDGFRβ Sm22D849V mice as early as 12 weeks of WD, and 50% died by 19 weeks of WD (n = 26) (Supplementary Figure 5). It is possible that coronary plaques contributed to the reduced viability of WD-fed mutants, but this needs to be investigated further in a study specifically designed to identify causes of death.

PDGF signaling activates STAT1 in VSMCs
The striking amplification of atherosclerosis raised mechanistic questions about which PDGF-regulated processes actually drive the atherosclerotic phenotype and which processes are bystanders. Clearly, inflammation is seen throughout atherogenesis. On the other hand, VSMC hyperplasia, phenotype switching, and increased ECM synthesis seemed important for helping early plaques to progress to the fibroatheroma morphology. Therefore, we hypothesized that PDGF-induced chemokine secretion and inflammation of the vessel wall (see Fig. 1 and Fig. 2) were the processes driving plaque initiation. PDGFRβ regulates several downstream signaling cascades including Ras/MAP kinase, PI3 kinase/AKT, PLCγ/ PKC, and STATs. We chose to focus on STAT1 as a potential mediator of pro-inflammatory PDGFRβ signaling because this transcription factor is a direct activator chemokine gene expression in the context of interferon signaling 29 .
First, PDGFRβ +/+ (Wt) control VSMCs were incubated with or without interferon gamma (IFN-γ) and STAT1 phosphorylation was assessed. IFN-γ was required for STAT1 phosphorylation in control VSMCs, but STAT1 was constitutively phosphorylated in PDGFRβ Sm22D849V cells (Fig. 7a). This was specific to STAT1, as STAT3 and STAT5 were not phosphorylated in mutant cells. Treatment of control VSMCs with PDGF resulted in STAT1 phosphorylation (Fig. 7b), and treatment of PDGFRβ Sm22D849V cells with the PDGFR inhibitor AG-1295 abolished STAT1 phosphorylation (Fig. 7c, d), indicating that constitutive STAT1 phosphorylation depends on PDGF pathway activity. AG-1295 had no effect on STAT1 phosphorylation in response to IFN-γ treatment (Fig. 7e, f, fourth lane), indicating that PDGFRβ is not required for this response. The constitutively phosphorylated STAT1 in PDGFRβ Sm22D849V VSMCs could be further phosphorylated by the addition of exogenous IFN-γ ( Fig. 7e, f, seventh lane), suggesting that PDGFRβ Sm22D849V VSMCs remain capable of sensing interferon. However, as discussed previously, we did not detect increased interferon (IFN-α, β, or γ) mRNA expression in PDGFRβ Sm22D849V VSMCs compared to control cells, consistent with the model that PDGFRβ signaling is responsible for the phosphorylation of STAT1.
In the IFN-γ signaling cascade, activated interferon receptors recruit Janus-activated kinase (JAK)-1 and -2, which are cytoplasmic tyrosine kinases that subsequently phosphorylate STAT1. Previous studies found that PDGFRβ associates with and phosphorylates JAK1, JAK2, and the related kinase Tyk2, but suggested that, unlike the interferon pathway, JAKs may be dispensable for PDGFRβ-induced STAT1 activation 30,31 . To explore this question in VSMCs we treated control or mutant cells and the JAK1/2-selective inhibitor Ruxolitinib (Rux). As expected, pretreatment of VSMCs with Rux blocked STAT1 phosphorylation in response to IFN-γ (Fig. 7g, h, fourth lane). Interestingly, Rux also blocked STAT1 phosphorylation in control VSMCs treated with PDGF-BB, and in unstimulated PDGFRβ Sm22D849V VSMCs (Fig. 7g, h, sixth and eighth lanes). This points to the conclusion that JAK1 and/or JAK2 are required for STAT1 phosphorylation in the PDGF signaling pathway. As a control, Rux had no effect on PDGFRβ-induced PLCγ1 phosphorylation, indicating that Rux does not impair PDGFRβ tyrosine kinase activity.

PDGF/STAT1-regulated inflammation amplifies atherosclerosis
PDGFRβ Sm22D849V aortas exhibit inflammation and dedifferentiated VSMCs, and deletion of STAT1 in VSMCs reduces inflammation but does not affect the VSMC phenotype (Fig.  8). This suggested that STAT1 deletion in ApoE,PDGFRβ Sm22D849V mice would be an approach to evaluate the importance of inflammation vs. VSMC dedifferentiation as proatherosclerotic mechanisms. To test this we generated ApoE,PDGFRβ Sm22D849V single conditional mutants with hypercholesterolemia and ApoE −/− ;STAT1 flox/flox ;Sm22α-Cre;PDGFRβ +/D849V (ApoE,STAT1 Sm22cKO PDGFRβ Sm22D849V ) double conditional mutants with hypercholesterolemia. After feeding WD for 12 weeks we screened for atherosclerosis. Compared to ApoE,PDGFRβ Sm22D849V mutants, a much smaller area of the aorta was covered by plaques in ApoE,STAT1 Sm22cKO PDGFRβ Sm22D849V mutants that were lacking STAT1 in VSMCs (Fig. 9a, b). Moreover, plaques in the aortic root of the double mutants were significantly smaller than those of the single mutants (Fig. 9c, d), appearing similar to the size of plaques in ApoE −/− aortic roots at 8-16 weeks of WD (Fig. 4c). Next, we examined plaque composition in the thoracic aorta. After 12 weeks of WD the less abundant plaques in ApoE,STAT1 Sm22cKO PDGFRβ Sm22D849V mutants were clearly less advanced with poorly developed fibrous caps and small necrotic cores compared to ApoE,PDGFRβ Sm22D849V mutants (Fig. 9e). Therefore, PDGFRβ-STAT1 signaling in VMSCs creates an inflammatory milieu in the aortic adventitia and media that strongly predisposes arteries to form atherosclerotic plaques in synergy with hypercholesterolemia.

DISCUSSION
The mechanisms that control the initiation and progression of atherosclerosis are incompletely understood. In this study, we investigated the effects of increased PDGFRβ signaling on atherosclerosis. We first showed that conditional activation of PDGFRβ induced the expression of chemokines in VSMCs and caused inflammation of the aortic media and adventitia in PDGFRβ Sm22D849V mutant mice. Taken together with VSMC dedifferentiation, hyperplasia, and increased ECM synthesis, these phenotypes suggested that the activating mutation in PDGFRβ would be pro-atherogenic. Analysis of ApoE,PDGFRβ Sm22D849V mice provided clear evidence that PDGF signaling strongly amplified hypercholesterolemia-induced atherosclerosis. Intriguingly, increased PDGF signaling expanded the sites of plaque initiation, allowing atherosclerosis to occur in the thoracic aorta and coronary arteries that are typically free of plaques in ApoE or Ldlr single mutant mice. In ApoE,PDGFRβ Sm22D849V mice, the plaques consistently developed into fibroatheromas. Moreover, the aortic plaques exhibited neovascularization and intraplaque hemorrhage, which are common in advanced atherosclerotic lesions in humans. Neovascularization is thought to contribute to disease by providing a route for macrophages and other leukocytes to enter the plaques 33 . Intraplaque hemorrhage worsens atherosclerosis by depositing large amounts of free cholesterol from erythrocyte membranes 25,26 . In addition, iron deposits in atheromas may worsen atherosclerosis by increasing the formation of reactive oxygen species and oxidized LDL 34 . ApoE −/− mice do not typically develop coronary atherosclerosis without long-term (10 months) of WD feeding 27,28 . ApoE −/− Akt1 −/− double mutants develop coronary atherosclerosis after 14 weeks of high cholesterol diet through mechanisms involving increased endothelial cell and macrophage apoptosis 35 . More dramatically, scavenger receptor B1 (SR-BI)-deficient mice on an ApoE −/− background develop severe coronary atherosclerosis by 5 weeks of age due to profound alterations in lipoprotein metabolism 36 . Thus, the PDGFRβ Sm22D849V mice combined with ApoE or Ldlr mutations provide a distinct model of atherosclerosis that may help define the role of VSMCs in plaque advancement and coronary heart disease.
We provided genetic evidence that inflammation directly contributes to novel plaque formation in the thoracic aorta. Conditional deletion of STAT1 in ApoE +/+ VSMCs was sufficient to normalize much of the chemokine overexpression and inflammation caused by PDGF signaling. However, VSMC phenotype switching and adaptive remodeling (dilation) of the aorta persisted in STAT1 Sm22cKO PDGFRβ Sm22D849V double-mutant mice, indicating that these aspects of the phenotype are not dependent on STAT1. Even though aortic dilation persisted in double mutants, far fewer plaques occurred in the thoracic aorta. Thus, our data point to an inflammatory pathway as an important mechanism by which PDGF signaling amplifies atherosclerosis. We cannot completely exclude possible hemodynamic changes secondary to remodeling of the aorta as a contributing factor towards enhanced atherosclerosis, but our genetic evidence demonstrates that STAT1-linked inflammation is required. This inflammatory pathway involves PDGFRβ activation in VSMCs, activation of JAKs, tyrosine phosphorylation of STAT1, secretion of chemokines from VSMCs, and accumulation of leukocytes in the adventitia and media of the aortic wall. Leukocyte accumulations were apparent in histological sections of PDGFRβ Sm22D849V adventitia (Fig.  2d). Intriguingly, recent studies highlight the potential of VSMCs to act as lymphoid tissue organizers by secreting chemokines (CXCL13, CCL21, and others) that promote lymphoid tissue neogenesis 37,38 . Furthermore, aged ApoE −/− mice develop tertiary lymphoid tissue in the aorta adventitia 39 . As we did not find well-organized lymphoid tissue in the adventitia, the relevance of these observations to our model is unknown. It is possible that our mice were too young at the time of analysis, or that the chemokines involved are not sufficient for the formation of organized lymphoid tissue. Specific chemokines we identified as STAT1dependent PDGF targets have been previously shown to have pro-atherogenic functions. CCL2/MCP-1 and CCL5/RANTES are important chemoattractants for monocytes and macrophages in the atherosclerotic setting [40][41][42][43] . CXCL10/IP-10 modulates the balance of effector and regulatory T cells in atherosclerosis by activating the receptor CXCR3 44,45 . There is evidence to suggest that the related chemokines CXCL9/MIG and CXCL11/ITAC also regulate T cell trafficking in atherosclerosis [45][46][47] . Together, these data suggest that chemokines released by VSMCs work together to inflame the aorta, which may create new niches for plaque initiation when combined with hypercholesterolemia.
Arterial VSMCs have long been viewed as one of the major cell types involved in atherosclerosis 48 . Most recent work on plaque initiation has focused on the importance of inflammatory cells and endothelial cells, while VSMCs are more often discussed in the context of advanced atherosclerosis when VSMCs switch from a contractile to a secretory phenotype, proliferate, migrate into plaques, and create fibrous cap structures. However, from human autopsy studies it is known that VSMCs predominate in the thickened arterial intima at atherosclerosis-prone sites even prior to the onset of plaque formation [49][50][51] . Therefore, it has been suggested that VSMCs might help initiate plaques at atherosclerosis-prone sites by transdifferentiating into foam cells 52,53 , by producing proteoglycans that trap lipoproteins 54 , by expressing adhesion molecules that recruit inflammatory cells 55 , or by enhancing the activity or survival of monocytes and macrophages 56 . Further investigation may reveal whether any of these disease processes are exacerbated in PDGFRβ Sm22D849V mice and whether they are modulated by the loss of STAT1.
The D849V mutation in PDGFRβ increases basal and ligand-mediated signaling activity of PDGFRβ, which leads to constitutive phosphorylation of STAT1 in a JAK1/2-dependent manner. Phosphorylation of STAT1 in PDGF-treated cells was demonstrated over 20 years ago 57 , and our study now shows an in vivo role for the PDGFRβ-STAT1 signaling pathway in regulating the VSMC inflammatory phenotype. STAT1 knockout mice do not exhibit vascular phenotypes under basal conditions 58,59 . However, STAT1 may operate in VSMCs in a context-dependent manner, for instance during the response to injury and inflammation. STAT1 may be especially important under conditions where IFN-γ and PDGF are both present, as previous studies suggested interactions between the two signaling pathways [60][61][62][63] . Further work will be needed to understand the importance of VSMC-expressed STAT1 for atherosclerosis in standard ApoE and Ldlr-deficient mice.
This study shows that increased PDGFRβ signaling causes inflammation and VSMC dedifferentiation in vivo. In the setting of hypercholesterolemia, PDGFRβ signaling promotes new plaque initiation at atherosclerosis-resistant sites and promotes the formation of fibroatheroma plaques with neovascularization and plaque rupture. A novel aspect of this study is the observation that initiation of plaques in the subendothelium of thoracic aorta requires inflammation of the adventitia and media. The conventional model of atherosclerosis has focused on the inner layers of the vessel, guided by the paradigm of endothelial cell activation, subendothelial inflammation, monocyte recruitment to the subendothelium, and foam cell formation in the subendothelium. This view emphasizes inflammation beginning at the luminal side of the vessel wall and progressing outward, with the outer layers of the vessel playing a passive role. However, the present study contributes experimental evidence pointing towards the importance of outside-in mechanisms where inflammation of the adventitia and media creates new niches for atherosclerotic plaque initiation.

Mice
Sm22α-Cre;PDGFRβ +/D849V (PDGFRβ Sm22D849V ) conditional knockin mice 18 were crossed to ApoE or Ldlr-deficient mice obtained from the Jackson Laboratory. STAT1 flox/flox mice were provided by L. Hennighausen (NIDDK, NIH, Bethesda, MD) 32 . Collagen1a1-EGFP transgenic reporter mice were provided by D. Brenner (UCSD, La Jolla, CA) 64 . All mice were maintained on a mixed C57BL/6;129 background and were fed standard chow or Western diet (diet#TD88137, Harlan Teklad). Some mice were injected with 15mg kg −1 Crenolanib (Selleck Chemicals) for 5 consecutive days. Mutant mice were always compared to age and sex-matched controls that carried the appropriate Cre driver, and littermates were used whenever possible. We did not observe any vessel aneurism development (defined as focal dilation/weakening of the vessel wall), aortic dissection, kidney failure, or tumor formation in any of the PDGFRβ Sm22D849V mutants we examined during the period of our study. All animal experiments were performed under approval from the Institutional Animal Care and Use Committee of the Oklahoma Medical Research Foundation.

Isolation and culture of primary VSMCs
Mouse VSMCs were isolated by the explant method from aortas of 6-8 weeks old mice 65 . VSMCs were cultured in DMEM medium plus 10% FBS and antibiotics. All assays were performed at passages 2 to 5. To induce chemokine expression, wild type VSMCs were serum starved in DMEM + 0.5% FBS for 36 hrs, followed by stimulation with 10ng ml −1 PDGF-BB (R&D Systems). For STAT phosphorylation experiments, 10ng ml −1 mouse IFNγ (EMD Millipore) served as a positive control. To block PDGFRβ activity, VSMCs were treated with 400ng ml −1 Crenolanib (Selleck Chemicals) for 2 hrs or 10 μM AG-1295 (Calbiochem) for 16 hrs. To block JAK1/JAK2 kinase activity, VSMCs were treated with 10 μM Ruxolitinib (Selleck Chemicals) for 16 hrs. STAT1 siRNA and nonsilencing control siRNA were obtained from Santa Cruz Biotechnology. Mouse VSMCs were transfected with siRNA for 48 hrs using Lipofectamine RNAiMAX (Invitrogen) according to the manufacturer's instructions.

RNA isolation and quantitative RT-PCR
Total RNA was extracted from cultured VSMCs using RNeasy Mini Kit (Qiagen). Reverse transcription to cDNA was performed using random primers and Superscript III RT (Invitrogen). Quantitative real-time PCR was performed using a Bio-Rad iCycler with iQ TM SYBR green master mix. Results were normalized relative to GAPDH unless indicated otherwise.

Quantification of mouse chemokine proteins
Mouse VSMCs were cultured in 6-well plates. Chemokine protein levels in harvested conditioned media and mouse plasmas were determined using a multi-analyte ELISA kit (Qiagen MEM-009A) according to the manufacturer's instructions.

Aorta flow cytometry analysis
Single-cell suspensions were prepared from the mouse aortas as described 20 . Briefly, mice were anesthetized and perfused through the heart with PBS containing 10 U ml −1 heparin to remove blood from all vessels. After careful removal of the perivascular adipose tissue, the aortas were minced into small pieces and digested with 125 U ml −1 collagenase type XI, 60 U ml −1 hyaluronidase type I-s, 60 U ml −1 DNase1, and 450 U ml −1 collagenase type I (all enzymes were obtained from Sigma-Aldrich) in PBS containing 20 mM Hepes at 37°C for 3 hrs. For separating adventitia and media, the aortas were predigested in enzyme cocktail (300U ml −1 Collagenase I and 10U ml −1 Elastase in PBS) at 37°C for 15 minutes and then the adventitia was carefully peeled away from the aorta. After filtering through a 70 μm filter, cells were resuspended in FACS buffer, and incubated with Fc-blocking antibody (eBioscience) for 15 min on ice before being stained with specific antibodies. Antibodies used were as follows: FITC-CD45.2, PE-CD11c, APC-CD3, APC-Cy7-CD19, PE-Cy7-NK1.1, PerCy5.5-CD11b (all antibodies were obtained from Biolegend and used at 1:100 dilution). Cells were also stained with propidium iodide. After washing, immunofluorescence was detected by an LSR II (BD Biosciences). Data were analyzed using FlowJo (Tree Star Inc.) software.

BrdU incorporation and labeling index
Mice were injected with 50mg kg −1 BrdU and then sacrificed 24 hrs later. Aortas were perfused with PBS/heparin, perfusion fixed with 4% paraformaldehyde (PFA), paraffin embedded, and sectioned longitudinally. Cells were identified as VSMCs based on their position between the internal and external elastic lamina of the tunica media. Proliferating VSMCs (BrdU + ) were detected with anti-BrdU antibody (Sigma) at 1:20 dilution. Hematoxylin was used as a counterstain for nuclei.

Blood pressure measurements
Blood pressure was measured by the carotid catheter method. 8-10 weeks old mice were anesthetized with a ketamine and xylazine mixture (70:6 mg kg −1 , intramuscular injection) and placed under warm light (37°C). The left common carotid artery was carefully exposed via a 0.5-to 1.0-cm midline incision in the ventral neck region. The tip of the artery toward the head was ligated with a suture (5-0 silk), and the tip toward heart was occluded with a microclip (no. 18055-03; Fine Science Tool, Foster City, CA). A small cut was then made in the vessel wall using microscissors (no. 15000-08, Fine Science Tool). A catheter (PE10 tubing) containing a sterile 10% heparin-90% saline solution was inserted into the common carotid. Blood was directed to a pressure transducer through the catheter to obtain computerized blood pressure measurements (AD instruments) after removing the clip. (mean ± s.e.m. n = 6 mice per group).

Serum lipid profiles
Wild type or PDGFRβ Sm22D849V mice from both sexes were fed chow diet for 2-3 months or WD for 24 weeks. ApoE and ApoE,PDGFRβ Sm22D849V mice were fed chow diet for 8-10 months or WD for 16 weeks. Blood was drawn from the facial vein with light anesthesia and serum was separated by centrifugation. Cholesterol and triglyceride levels were determined by commercially available kits (BioVision and ThermoScientific) followed by spectrophotometric quantification.

Quantitative analysis of plaques
Mice were perfused by cardiac puncture with PBS containing 10U ml −1 of heparin and then 4% PFA, followed by overnight post-fix with 10% formalin. For en face staining, the adventitia was thoroughly cleaned under dissecting microscope, and the aorta was cut open longitudinally and pinned on parafilm. Plaques were stained with Oil red O and photographed with a digital camera. For cross-sectional quantification of plaque progression, 5μm serial paraffin sections from the indicated sites of the aorta were stained with hematoxylin and eosin. Plaques were imaged with a Nikon Eclipse 80i microscope connected to a digital camera. Plaque area was then quantified with NIS.Elements D3.2 imaging software (Nikon Inc.) to obtain the maximum cross-sectional plaque area for each site. Fibrous cap thickness was quantified by choosing the largest necrotic core in a section and measuring the thinnest part of the fibrous cap. Representative images were finally processed with Photoshop (Adobe).

Transmission Electron Microscopy
Two week-old mice were perfusion fixed through the heart, then thoracic aorta segments were isolated and immersion fixed with 1.5% glutaraldehyde and 2.5% paraformaldehyde in 0.1M HCl-Na cacodylate buffer for 2 hours, then post-fixed for 90 minutes at 40°C with 1% osmium tetraoxide in the same buffer, and then incubated with 0.5% tannic acid for 30 minutes at room temperature. After dehydration in graded ethanol, and propylene oxide treatment, specimens were embedded in epoxy resin. Sections were cut with a diamond knife, stained with lead citrate and uranyl acetate and examined with a Hitachi H-7600 electron microscope.

Statistical analysis
Data are expressed as mean ± the standard error of the mean (s.e.m.). Statistical difference was analyzed with the Student's t test. A value of P < 0.05 was considered statistically significant.