Impaired MEK Signaling and SERCA Expression Promote ER Stress and Apoptosis in Insulin-Resistant Macrophages and Are Reversed by Exenatide Treatment

Obese ob/ob, lean^+/+, and ob/+ mice on a C57BL/6J background were obtained from the Jackson Laboratory. Mice deficient in both leptin and Ldlr (ob/ob;Ldlr^−/−) on a C57BL/6J background were generated as previously described (19). All experiments were performed with animals fed regular rodent chow diet, except in the experiments of in vivo exenatide or saline treatment where ob/ob;Ldlr^−/−, Insr^−/−;Ldlr^−/−, and Ldlr^−/− mice were fed a Western-type diet (WTD) (TD88137; Harlan-Teklad). All animal procedures used in this study followed Columbia University’s institutional guidelines and relevant regulatory standards.

Fresh peritoneal macrophages were treated in Dulbecco’s modified Eagle’s medium (DMEM)/10% FBS with LY294002 (10 μmol/L; Biomol), PD98059 (Calbiochem), U0126 (Sigma-Aldrich), rapamycin (2 μmol/L; Calbiochem), GF109203 (2 μmol/L; Sigma-Aldrich), palmitic acid–BSA (0.5 mmol/L; Sigma-Aldrich), thapsigargin (5 μmol/L; Sigma-Aldrich), tunicamycin (2.5 μg/mL; Sigma-Aldrich), 8-Br-cAMP (Sigma-Aldrich), forskolin (Sigma-Aldrich), exendin-4 (California Peptide Research), H-89 (Sigma-Aldrich), Rp-adenosine 3′,5′-cyclic monophosphorothioate (Rp-cAMPS) (Sigma-Aldrich), 6-benzoyl (Bnz)-cAMP (Biolog), 8-pMeOPT-2’-O-Me-cAMP (Biolog), or corresponding vehicle controls, acetylated (Ac)LDL (100 μg/mL; Biomedical Technologies) and compound Sandoz 58035 (10 μg/mL; Sigma-Aldrich), oxidized LDL (100 μg/mL; Biomedical Technologies), or 7-ketocholesterol (40 μg/mL; Sigma-Aldrich) for the times specified in figure legends. Mouse macrophages were infected with adenovirus either harboring lacZ (Ad-LacZ) or constitutively active MEK mutant (Ad-MEK), provided by Dr. Shougang Zhuang (Brown Alpert Medical School, Providence, RI), as described previously (20). After treatments, total RNA or cell lysates were prepared for further analysis. Macrophage total RNA isolation and SYBR Green quantitative PCR (QPCR) analysis were performed as previously described (15). Western analysis was performed using the following primary antibodies: anti–phosphorylated eIF2 (P-eIF2) (Cell Signaling), anti-activating transcription factor (ATF)-6 (Imgenex), anti-LaminB1 (Santa Cruz), anti–sarcoplasmic-ER calcium ATPase 2 (SERCA2) (Santa Cruz), anti–phosphorylated cAMP-responsive element–binding protein (P-CREBP) (Upstate Biotechnology), anti-CREBP (Santa Cruz), anti–glucagon-like peptide 1 (GLP-1) receptor (Santa Cruz Biotechnology), anti–phosphorylated PKR-like ER kinase (P-PERK) (Cell Signaling), anti-AKT and phosphorylated AKT (P-AKT) (Cell Signaling), anti-ERK and phosphorylated ERK (P-ERK) (Cell Signaling), anti-MEK (Cell Signaling), anti–ATF-3 (Santa Cruz), and anti–β-actin antibody (Sigma-Aldrich). Apoptosis assays were performed as described previously (15). Macrophage calcium imaging was performed as described previously (21).

Mouse proximal aortas were serially paraffin sectioned as described previously (15). Serial aortic sections were probed with antibody against P-PERK or ATF-3, or control antibody followed by appropriate Alexa Fluor 488–conjugated secondary antibody (Invitrogen) (22). Specific fluorescence signals in the digitized images of proteins colocalized with macrophage marker Mac-3–positive stains were identified using ImageJ (NIH). Lesional macrophage apoptosis was monitored by the transferase-mediated dUTP nick-end labeling (TUNEL) assays and immunohistochemistry with antibody against active caspase-3 (15). TUNEL- or caspase-3–positive signals were determined to colocalize with Mac-3 staining of lesional macrophages.

Results were expressed as average ± SE (n is noted in the figure legends or figures), and statistical significance of differences was evaluated with Student t test.

In this study, we used peritoneal macrophages isolated from either Insr^−/− or ob/ob mice to investigate the mechanisms responsible for an increased UPR in insulin-resistant cells. Three distinct branches of the UPR have been identified involving signaling by PERK, ATF-6, and X-box binding protein 1 (XBP1) (23). To determine whether these were involved in ER stress–induced apoptosis, we monitored the expression of P-eIF2, ATF-6 cleavage, and XBP1 splicing. Basal insulin-resistant macrophages showed an increase in phosphorylation of PERK target eIF2, ATF-6 cleavage, and XBP1 mRNA splicing compared with controls (Fig. 1A and B). Time-course analysis of the response to free cholesterol loading revealed a more prominent transcriptional induction of UPR-responsive genes Chop (PERK branch of UPR) and Grp78 (an ATF-6 target), and increased Xbp1 splicing in Insr^−/− cells with a similar time course, suggesting parallel induction of all three major branches of the UPR in insulin-resistant macrophages (Fig. 1C). Similar to Insr^−/− cells (15), freshly isolated macrophages from ob/ob mice showed increased PERK phosphorylation and increased apoptosis when challenged with free cholesterol, 7-ketocholesterol, or oxidized LDL loading (Supplementary Fig. 1A and B). More pronounced expression of some ER stress markers such as CHOP was also found in free cholesterol–loaded ob/ob macrophages (Supplementary Fig. 1C). An elevation in UPR, however, did not significantly reduce global protein translation (Supplementary Fig. 1A, bottom) or increase oxidative stress in ob/ob versus wild-type (WT) cells, as assessed using reactive oxygen species detection reagents (data not shown).

The phosphorylation of PERK is a hallmark of UPR activation and ER stress (23). We next analyzed the signaling pathways downstream of the insulin receptor to determine which branch of the insulin signaling pathway was responsible for PERK activation in ob/ob and Insr^−/− macrophages. WT mouse macrophages cultured in regular DMEM medium containing 10% FBS were treated with different inhibitors, and P-PERK expression was examined by Western analysis. Interestingly, inhibition of macrophage MEK/ERK activity by PD98059 resulted in a robust induction of P-PERK, similar to ER stress inducers thapsigargin or tunicamycin, whereas treatment with phosphatidylinositol 3-kinase (PI3K), mammalian target of rapamycin, or protein kinase C inhibitors did not result in comparable phosphorylation of PERK (Fig. 2A). In addition, because ob/ob mice have elevated plasma free fatty acid levels, we treated macrophages with palmitate/BSA complexes but did not find any significant induction of P-PERK (Fig. 2A). The induction of P-PERK by MEK inhibition was dose dependent and could be triggered by different classes of MEK inhibitors (Fig. 2B). In addition, various UPR target genes representing involvement of all three branches of the UPR were induced after MEK inhibition in WT macrophages (Supplementary Fig. 2A), resembling the response in Insr^−/− and ob/ob cells (Fig. 1 and Supplementary Fig. 1C).

To determine whether MEK inhibitor studies were relevant to the defects associated with insulin-resistant macrophages, we next examined the role of MEK/ERK signaling in apoptotic responses of Insr^−/− macrophages. MEK activity, as indicated by the levels of P-ERK, was markedly reduced in Insr-deficient macrophages with or without free cholesterol loading, concomitant with decreased signaling of PI3K and AKT (Fig. 2C). To study the contribution of MEK defects to macrophage apoptosis during ER stress, WT cells pretreated with different inhibitors were incubated with free cholesterol or 7-ketocholesterol, and annexin V apoptosis assays were performed. Inhibition of macrophage MEK activity significantly promoted free cholesterol– or 7-ketocholesterol–induced apoptosis compared with untreated controls (Fig. 2D), indicating enhanced sensitivity of MEK-defective macrophages to ER stress–associated apoptotic death. Conversely, adenoviral expression of constitutively active MEK in insulin-resistant macrophages (Fig. 2E) reduced the levels of P-PERK compared with cells transduced with control lacZ adenovirus (Fig. 2F) and significantly rescued these mutant cells from free cholesterol–triggered cell death (Fig. 2G), indicating a major role of defective MEK signaling in enhanced apoptosis of insulin-resistant macrophages.

The calcium pump SERCA plays an important role in maintaining ER calcium stores, which are required for proper protein folding (24). Accumulation of free cholesterol in ER membranes was previously shown to inhibit SERCA2b activity and deplete ER calcium stores (25), thereby contributing to free cholesterol–associated apoptosis. In addition, insulin receptor substrate 1 deficiency in mouse pancreatic islets is accompanied by reduced SERCA2b expression and altered glucose-stimulated calcium signaling (26). Therefore, we next sought to determine whether SERCA2b activity is decreased in insulin-resistant macrophages. Similar to the findings in insulin receptor substrate 1–deficient islets, SERCA2b mRNA and proteins were significantly reduced in ob/ob and Insr^−/− macrophages (Fig. 3A). The decrease in SERCA expression was accompanied by depletion of ER calcium pools in Insr^−/− (Fig. 3B) and ob/ob (data not shown) macrophages, as indicated by the measurement of releasable ER-stored calcium (9). Moreover, inhibition of MEK activity in WT macrophages led to a significant decrease in Fluo3 fluorescence ratio (Fig. 3B), indicating depletion of ER calcium by decreased MEK signaling. Expression of constitutively active MEK in insulin-resistant macrophages induced SERCA mRNA expression (Fig. 3C) and subsequently replenished ER calcium stores (Fig. 3D). These results suggest a direct role of defective MEK/SERCA activity in increased ER stress–associated apoptosis in insulin-resistant macrophages (Figs. 2 and 3).

The transcription factor CREBP has been reported to support the transcription of SERCA (27,28). To determine whether decreased mRNA expression of SERCA is related to CREBP in insulin-resistant macrophages, we examined the levels of phosphorylation of CREBP at Ser-133 by Western analysis. As shown in Fig. 4A, nuclear-phosphorylated CREBP was profoundly reduced in Insr^−/− macrophages. Accordingly, mRNA levels of known CREBP target genes c-fos and cyclin D were decreased in MEK-inhibited WT and in Insr^−/− and ob/ob macrophages (data not shown). MEK inhibition by U0126 and PD98059 compounds in WT macrophages led to downregulation of CREBP phosphorylation (see below). Moreover, chromatin immunoprecipitation assays indicated less P-CREBP binding to the SERCA2 promoter in Insr^−/− than in control macrophages (Supplementary Fig. 2B). In addition, inhibition of macrophage MEK activity resulted in a reduction in P-CREBP binding (Supplementary Fig. 2C). Our results suggest that reduced MEK/ERK/nuclear CREBP activity might be responsible for decreased mRNA expression of SERCA and increased ER stress, and contribute to increased apoptosis in insulin-resistant macrophages. To examine whether activation of macrophage CREBP via protein kinase A (PKA) could increase CREBP activity and rescue MEK-defective cells from free cholesterol–induced apoptosis, insulin-resistant ob/ob cells were incubated with free cholesterol with or without PKA activator 8-bromo-cyclic AMP (Br-cAMP). Br-cAMP stimulated SERCA2b mRNA expression and significantly reduced apoptosis induced by free cholesterol loading in ob/ob macrophages (Fig. 4B and C) and Insr^−/− cells (data not shown). Adenylyl cyclase activator forskolin exerted a similar effect on apoptosis in these macrophages (data not shown). To further explore whether, in addition to PKA, PKA-independent cAMP effector Epac (exchange nucleotide protein directly activated by cAMP) is involved in cAMP-mediated macrophage protection, we used different cAMP analogs 6-Bnz-cAMP (a PKA activator but poor Epac agonist) and 8-pMeOPT-2’-O-Me-cAMP (an Epac activator but poor PKA agonist) to examine the role of Epac signaling in cAMP action. Both 6-Bnz-cAMP and Br-cAMP significantly protected macrophages from free cholesterol–induced apoptosis, whereas the reduction of apoptosis by Epac agonist 8-pMeOPT-2’-O-Me-cAMP was less significant compared with PKA activators (Fig. 4D).

We next studied the response to ER stress–induced apoptosis of PKA inactivation in macrophages using specific PKA inhibitors H-89 or Rp-cAMPS. Inhibition of macrophage PKA led to an increase in apoptosis of free cholesterol–loaded macrophages (Fig. 4E), suggesting an antiapoptotic role of PKA during ER stress. Cyclic AMP and PKA were recently shown to enhance ERK signaling via AKAP-Lbc-KSR-1 (29). In macrophages, PKA activator 6-Bnz-cAMP, but not Epac agonist 8-pMeOPT-2′-O-Me-cAMP, stimulated phosphorylation of ERK (Fig. 4F). More importantly, specific PKA inhibitor H-89, like MEK inhibitor U0126, decreased the expression of both P-CREBP and SERCA2b in macrophages (Fig. 4G). Our results thus indicated that reduced expression of SERCA2b as a result of defective ERK and PKA signaling led to a significant increase in apoptosis in macrophages loaded with free cholesterol (Fig. 2D and Fig. 4E–G).

The antidiabetic drug exenatide (exendin 4) increases GLP-1 receptor signaling and has been shown to promote survival of pancreatic β-cells during ER stress (30). Exenatide or GLP-1 treatment leads to activation of PKA, PI3K/AKT, and possibly ERK in these cells (31,32). To investigate whether GLP-1 signaling could rescue insulin-resistant macrophages from ER stress–induced apoptotic death, we examined the responses of primary macrophages to exenatide treatment in vitro. The GLP-1 receptor was readily detected in mouse macrophages, similar to that found in MIN6 mouse pancreatic β-cells, by Western analysis (Supplementary Fig. 3A) and was expressed similarly in control and insulin-resistant macrophages (Fig. 5A). Both 5′ and 3′ termini of GLP-1 receptor mRNA transcripts could be found in these cells by RT-PCR (data not shown). Protein expression of macrophage GLP-1 receptor could be reduced specifically by small interfering RNA–induced GLP-1 receptor mRNA knockdown (Supplementary Fig. 3B). Treatment with GLP-1 receptor agonist exenatide–induced macrophage AKT and ERK phosphorylation (Fig. 5B). Moreover, inactivation of macrophage PKA activity by its inhibitor H-89 abolished the increase of ERK phosphorylation by exenatide (Fig. 5C), suggesting an involvement of PKA in ERK stimulation. H-89 exerted a similar effect on P-ERK expression in forskolin-treated macrophages (data not shown). Importantly, exenatide significantly increased SERCA2 expression and reduced free cholesterol–induced P-PERK, CHOP, and GADD34 expression and apoptotic death in Insr^−/− and WT macrophages (Fig. 5D–F). A time-course study showed that the decrease in CHOP mRNA expression could be detected as early as 6 h after free cholesterol loading (Supplementary Fig. 3C). Exenatide treatment led to a similar significant reduction in apoptosis triggered by free cholesterol or oxysterol in ob/ob macrophages (Fig. 5G and H). It is noteworthy that exenatide treatment almost completely reversed the increase in apoptosis associated with cellular insulin resistance in either Insr^−/− or ob/ob macrophages (Fig. 5F and G), in association with improvements in both ERK and AKT signaling (Fig. 5B). Conversely, treatment of insulin-resistant macrophages with PKA inhibitor H-89, thereby decreasing the expression of P-ERK and SERCA2b (Fig. 4G and Fig. 5C), abolished the exenatide-mediated beneficial effect on cell survival (Fig. 5I). Similar results were obtained in cells incubated with another PKA inhibitor, Rp-cAMPS (data not shown).

To examine whether administration of exenatide ameliorates the UPR and apoptosis in macrophages in mouse models of insulin resistance and type 2 diabetes, we used WTD-fed ob/ob;Ldlr^−/− or control Ldlr^−/− mice. After 3 months on WTD, mice were injected intraperitoneally daily with either PBS (saline) or exenatide (20 ng/g body weight) for 2 weeks. GLP-1 receptor agonists, including exenatide, are known to reduce food intake and body weight and improve hyperglycemia, peripheral insulin sensitivity, and plasma lipid profiles in mouse models of type 2 diabetes or WTD-fed WT mice (33,34). We also showed that, despite pair feeding and similar changes in body weight, exenatide reduced VLDL/LDL in ob/ob;Ldlr^−/− mice (35). In an attempt to avoid the potentially confounding effects of altered body weight and plasma lipid levels by exenatide, we pair-fed the two groups of mice of the respective genotype during the period of treatment. Exenatide-treated ob/ob mice had similar body weight and plasma lipid profiles to saline controls after 2 weeks of treatment (Supplementary Fig. 4A and B). Systemic insulin sensitivity was also not significantly changed by exenatide treatment, as assessed by intraperitoneal glucose tolerance test (Supplementary Fig. 4C). A similar response was also found in exenatide- versus PBS-treated Ldlr^−/− mice (data not shown). To determine whether macrophage UPR was reduced in vivo by exenatide, we harvested peritoneal macrophages after treatment. In the basal state, primary ob/ob macrophages from exenatide-treated mice showed a significant elevation of SERCA levels and normalization of ER calcium stores accompanied by a significant reduction in expression of UPR markers P-PERK, CHOP, spliced Xbp1, and 78-kDa glucose-regulated protein (immunoglobulin-binding protein) (Fig. 6A and Supplementary Fig. 4D and E), indicating in vivo exenatide-mediated attenuation of macrophage UPR independent of plasma lipid parameters or systemic insulin sensitivity. Moreover, these cells from exenatide-treated mice were significantly less susceptible to apoptosis induced by free cholesterol or 7-ketocholesterol (Fig. 6B).

Next, we investigated in vivo effects of exenatide treatment compared with saline on macrophage UPR and apoptosis in atherosclerotic lesions in ob/ob.Ldlr^−/− versus Ldlr^−/− mice. ATF-3 and P-PERK, among other UPR markers, were previously used in atherosclerosis studies. Like P-PERK, ATF-3 is induced by MEK inhibitors in peritoneal macrophages (Supplementary Fig. 4E). Its expression is increased in ob/ob versus WT macrophages and is normalized in vivo by exenatide (Supplementary Fig. 4E). Double immunofluorescence with antibodies against ATF-3 or P-PERK and against macrophage marker Mac-3 was carried out. As shown in Supplementary Fig. 4F and G, ATF-3 and P-PERK staining is increased in atherosclerotic lesions of ob/ob.Ldlr^−/− compared with Ldlr^−/− mice treated with saline. ATF-3–positive lesional macrophages in ob/ob.Ldlr^−/− mice were significantly reduced in percentage by exenatide treatment (Fig. 6C). A similar response to exenatide of macrophage P-PERK in the lesions of obese mice was also found (Fig. 6C). Macrophage apoptosis in atherosclerotic plaques was evaluated by active caspase-3 expression and TUNEL assay (15). In vivo exenatide significantly attenuated lesional macrophage apoptosis in ob/ob.Ldlr^−/−, but not in Ldlr^−/−, mice (Fig. 6D). Both macrophage ATF-3 and caspase-3 expression in the lesions of Insr^−/−.Ldlr^−/− mice was also reduced by exenatide (Supplementary Fig. 4H). Preliminary results revealed that exenatide during short periods of treatment did not significantly alter the sizes of atherosclerotic lesions and necrotic cores in ob/ob.Ldlr^−/− and Ldlr^−/− mice (data not shown).

Defective insulin signaling in macrophage foam cells promotes ER stress–associated apoptotic death in atherosclerotic lesions (15,16). In the current study, we show defects of macrophage MEK signaling and SERCA activity as a key mechanism underlying the impaired survival of stressed macrophage foam cells. Perturbation of calcium homeostasis in insulin-resistant cells as a result of reduced SERCA activity and subsequent release of ER calcium stores triggers the UPR and increases susceptibility to apoptotic death in response to “second hits” with additional adverse stimuli, such as overload with free cholesterol, oxidized lipids such as 7-oxysterols, oxidized phospholipids present in oxidized LDL, or apoptotic cells (Fig. 7).

The ER stress response is associated with transient activation of both MEK/ERK and AKT signaling pathways, and both have antiapoptotic activity in this context (36). Macrophage and vascular AKT have been shown to be protective of lesional apoptosis and the development of atherosclerosis (15,37). Our previous studies showed decreased AKT and increased nuclear FoxO1 in insulin-resistant macrophages during the ER stress response, resulting in induction of IkBε and an attenuated nuclear factor-κB/p65 response with proapoptotic and anti-inflammatory consequences during the ER stress response (20) (Fig. 7). However, increased macrophage FoxO1 activity only led to increased apoptosis in the presence of an ER stressor, such as free cholesterol loading or thapsigargin, and active FoxOs1/3 did not induce ER stress per se (20). Thus, decreased AKT/increased FoxO1 could not explain the enhanced ER stress response seen in insulin-resistant cells (Figs. 1 and 2 and Supplementary Fig. 1). The current study shows that decreased MEK/ERK signaling is primarily responsible for more prominent activation of the UPR in insulin-resistant macrophages. MEK signaling is involved in the regulation of various cellular activities, including apoptosis. During ER stress induced by thapsigargin, tunicamycin, or free cholesterol, MEK/ERK signaling is transiently activated (16). Inhibition of MEK activity has been shown to increase apoptosis in a variety of cell lines in response to ER stress (38,39), but the relevant mechanisms have not been well defined. Our findings indicate a link between macrophage MEK signaling and ER-resident SERCA activity. Attenuation of the MEK/ERK signaling cascade and subsequent reduction of SERCA activity and depletion of ER calcium stores sensitized insulin-resistant cells to ER stress–associated apoptotic stimuli more than normal controls (Figs. 2 and 3). Expression of constitutively active MEK rescued the defect in SERCA and P-PERK expression and partly decreased the enhanced apoptotic response of insulin-resistant cells. The fact that rescue of apoptosis was incomplete was expected since reconstitution of MEK/ERK activity did not restore defective AKT signaling in these cells. In contrast, treatment with agents that increased cellular cAMP, such as exenatide and forskolin, restored both ERK and AKT signaling pathways in insulin-resistant macrophages, and reversed their increased susceptibility to apoptotic death during free cholesterol or oxysterol loading (Figs. 4–6).

Our results indicate that macrophage insulin resistance promotes ER stress–associated apoptotic death partly through defects in SERCA and calcium mobilization. The calcium pump SERCA plays a major role in the regulation of cellular calcium homeostasis. The generation of a concentration gradient of calcium across the ER membrane requires SERCA activity. ER luminal calcium is essential for proper processing, folding, and maturation of proteins in the ER (24). Perturbation of the ER calcium flux as a result of decreased SERCA activity, such as in insulin-resistant macrophages, activates the UPR. The increase in cytosolic calcium caused by decreased SERCA might also be expected to directly promote the apoptotic response (21,40). A reduction in SERCA activity, rather than an increase in the activity of ER-resident inositol trisphosphate receptors and ryanodine receptors, was shown to trigger major ER stress–associated apoptosis in cultured β-cells (41,42). Consistent with our findings in insulin-resistant macrophages, SERCA expression or activity is reduced in the islets, liver, heart, and platelets in type 2 diabetes and insulin resistance (26,43–45), suggesting a perturbed mobilization of intracellular calcium in these tissues. The contribution of impaired SERCA to ER stress and insulin resistance found in ob/ob liver (44,45) illustrates an important role of SERCA in the development of metabolic derangements in type 2 diabetes and insulin resistance.

Our in vitro and in vivo findings (summarized in Fig. 7) suggest potential atheroprotective benefits of the antidiabetic drug exenatide in type 2 diabetes. We showed that exenatide alleviates cellular ER stress and reduces apoptotic death in lesional macrophages of mouse models of type 2 diabetes and insulin resistance (Fig. 6 and Supplementary Fig. 4), likely involving the activation of both PKA or MEK/SERCA and AKT prosurvival signaling (Figs. 5 and 6). In stressed pancreatic β-cells treated with the SERCA inhibitor thapsigargin, Drucker and colleagues (30) showed that exenatide ameliorates ER stress–induced apoptosis by magnifying feedback attenuation of P-eIF2α by GADD34-associated phosphatase. In WTD-fed ob/ob;Ldlr^−/− mice, we showed increased SERCA expression/activity and reduced ER stress responses as well as apoptosis of macrophages, without changes in plasma lipids or insulin resistance (Fig. 6). Thus, our findings offer new insights into the mechanisms of protection against apoptotic cell death by exenatide, and suggest that exenatide or similar therapies may offer a way to improve atherosclerosis in insulin resistance and type 2 diabetes by reversing the adverse effects of enhanced apoptosis during lesion development.

This study was supported by National Institutes of Health grants HL-087123 and HL-075662.

No potential conflicts of interest relevant to this article were reported.

C.-P.L. researched and evaluated data and wrote the manuscript. S.H. and G.L. researched data and reviewed the manuscript. I.T. and A.R.T. evaluated data and reviewed and edited the manuscript. C.-P.L. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

The authors thank Dr. Carrie Welch and Sandra Abramowicz (Columbia University) for preparation of paraffin sections of mouse proximal aortas.