A20 Regulates Atherogenic Interferon (IFN)-γ Signaling in Vascular Cells by Modulating Basal IFNβ Levels*

Background: IFNγ signaling is a major culprit in occlusive vascular disease. Results: The anti-inflammatory protein A20 negatively regulates IFNγ signaling in vascular cells in vitro and in vivo. Conclusion: A20 impacts IFNγ signaling by modulating basal IFNβ and downstream STAT1 expression. Significance: This novel function of A20 supports its promise as a therapeutic target and prognostic marker for atherosclerotic disease. IFNγ signaling in endothelial (EC) and smooth muscle cells (SMC) is a key culprit of pathologic vascular remodeling. The impact of NF-κB inhibitory protein A20 on IFNγ signaling in vascular cells remains unknown. In gain- and loss-of-function studies, A20 inversely regulated expression of IFNγ-induced atherogenic genes in human EC and SMC by modulating STAT1 transcription. In vivo, inadequate A20 expression in A20 heterozygote mice aggravated intimal hyperplasia following partial carotid artery ligation. This outcome uniquely associated with increased levels of Stat1 and super-induction of Ifnγ-dependent genes. Transcriptome analysis of the aortic media from A20 heterozygote versus wild-type mice revealed increased basal Ifnβ signaling as the likely cause for higher Stat1 transcription. We confirmed higher basal IFNβ levels in A20-silenced human SMC and showed that neutralization or knockdown of IFNβ abrogates heightened STAT1 levels in these cells. Upstream of IFNβ, A20-silenced EC and SMC demonstrated higher levels of phosphorylated/activated TANK-binding kinase-1 (TBK1), a regulator of IFNβ transcription. This suggested that A20 knockdown increased STAT1 transcription by enhancing TBK1 activation and subsequently basal IFNβ levels. Altogether, these results uncover A20 as a key physiologic regulator of atherogenic IFNγ/STAT1 signaling. This novel function of A20 added to its ability to inhibit nuclear factor-κB (NF-κB) activation solidifies its promise as an ideal therapeutic candidate for treatment and prevention of vascular diseases. In light of recently discovered A20/TNFAIP3 (TNFα-induced protein 3) single nucleotide polymorphisms that impart lower A20 expression or function, these results also qualify A20 as a reliable clinical biomarker for vascular risk assessment.


IFN␥ signaling in endothelial (EC) and smooth muscle cells (SMC) is a key culprit of pathologic vascular remodeling. The impact of NF-B inhibitory protein A20 on IFN␥ signaling in vascular cells remains unknown. In gain-and loss-of-function studies, A20 inversely regulated expression of IFN␥-induced atherogenic genes in human EC and SMC by modulating STAT1
transcription. In vivo, inadequate A20 expression in A20 heterozygote mice aggravated intimal hyperplasia following partial carotid artery ligation. This outcome uniquely associated with increased levels of Stat1 and super-induction of Ifn␥-dependent genes. Transcriptome analysis of the aortic media from A20 heterozygote versus wild-type mice revealed increased basal Ifn␤ signaling as the likely cause for higher Stat1 transcription. We confirmed higher basal IFN␤ levels in A20-silenced human SMC and showed that neutralization or knockdown of IFN␤ abrogates heightened STAT1 levels in these cells. Upstream of IFN␤, A20-silenced EC and SMC demonstrated higher levels of phosphorylated/activated TANK-binding kinase-1 (TBK1), a regulator of IFN␤ transcription. This suggested that A20 knockdown increased STAT1 transcription by enhancing TBK1 activation and subsequently basal IFN␤ levels. Altogether, these results uncover A20 as a key physiologic regulator of atherogenic IFN␥/STAT1 signaling. This novel function of A20 added to its ability to inhibit nuclear factor-B (NF-B) activation solidifies its promise as an ideal therapeutic candidate for treatment and prevention of vascular diseases. In light of recently discovered A20/TNFAIP3 (TNF␣-induced pro-tein 3) single nucleotide polymorphisms that impart lower A20 expression or function, these results also qualify A20 as a reliable clinical biomarker for vascular risk assessment.
IFN␥, typically secreted by Th1 and natural killer (NK) 5 cells, is a potent cytokine with well characterized antiviral and immunomodulatory functions (1). In addition to its immunomodulatory effects, IFN␥, which is highly expressed in atherosclerotic lesions of patients and experimental animals, is recognized as a direct driver of pathologic vascular remodeling (2)(3)(4). Sole infusion of exogenous IFN␥ is sufficient to cause transplant vasculopathy in human vascular allografts implanted into severe combined immunodeficiency mice (5), whereas knockdown of IFN␥ receptor (Ifngr) in atherosclerosis-prone apolipoprotein E (ApoE) knock-out mice significantly attenuates atherosclerotic lesions (6). Accordingly, tight control of IFN␥ production and signaling during antiviral and immune responses is necessary to avoid unforeseen pathologic responses, in particular in the vasculature.
The ubiquitin-modifying protein A20 (7) has recognized atheroprotective properties. Those stem from its potent antiinflammatory function in EC and SMC through inhibition of NF-B activation (8,9). They also rely on its NF-B-independent anti-apoptotic, anti-oxidative, and immunomodulatory functions in EC (10 -12) and anti-proliferative and pro-apoptotic (only neointimal SMC) functions in SMC (9,(13)(14)(15). However, the impact of A20 on pro-atherogenic IFN␥ signaling in EC and SMC had never been explored. In this study we uncovered A20 as a physiologic regulator of IFN␥ signaling in human * This work was supported, in whole or in part, by National Institutes of Health EC and SMC cultures and in vivo in a mouse model of pathologic vascular remodeling driven by hemodynamic injury to the vessel wall.
We ascribed the molecular basis of this novel function of A20 to its regulation of the expression of STAT1, the key mediator of IFN␥ signaling (16). Overexpression of A20 significantly decreased and A20 knockdown significantly increased STAT1 expression. Upstream of STAT1, A20 uniquely modulated basal sub-threshold levels of IFN␤, a prime regulator of the cross-talk between type I and type II IFNs (17). By interrupting both IFN␥ and NF-B pro-atherogenic and inflammatory signals in vascular cells, A20 qualifies as an ideal candidate for the treatment of occlusive vascular diseases, the primary cause of death in the Western world (18).

EXPERIMENTAL PROCEDURES
Cell Culture-Primary human coronary artery EC and SMC were purchased from Lonza (Allendale, NJ) and cultured in 5% CO 2 -humidified atmosphere using EGM TM 2-MV (EC) and SmGM TM (SMC) media supplemented with insulin, 5% fetal calf serum, and respective growth factors, as recommended. In all experiments, cells between passage 5 and 7 were serumstarved 24 h before stimulation. Human recombinant IFN␥ was purchased from PeproTech (Rocky Hill, NJ). Neutralizing sheep anti-IFN␣ and IFN␤ antisera were obtained from PBL InterferonSource (Piscataway, NJ). The monocytic cell line U937 was cultured in Dulbecco's modified essential medium (DMEM; Sigma) supplemented with 10% FCS, penicillin, and streptomycin.
Recombinant Adenoviruses-A20 cDNA, a kind gift of Dr. Vishva Dixit (Genentech, previously University of Michigan), was cloned into pAC.CMV-pLpASR ϩ to generate a recombinant adenovirus (rAd.) (11). rAd.Inhibitor of Nf-B␣ (IB␣), encoding porcine IB␣, was generated by C. J. Wrighton (19), and control rAd.␤gal was a kind gift of Dr. Robert Gerard (University of Texas, Southwestern Medical Center). Human SMC cultures were transduced at a multiplicity of infection of 500, which achieves expression of the transgene in Ͼ95% of the cells without causing toxicity (9).
siRNA Transfection-EC (A20) and SMC (A20, STAT1, and IFN␤) were transfected with specific FlexiTube siRNA (Qiagen, Valencia, CA) at a final concentration of 20 M, using HiPerfect transfection reagent (Qiagen) in serum-free Opti-MEM medium (Invitrogen). Four hours after transfection, the medium was replaced with regular culture medium, and cells were maintained in culture for 24 h prior to being used in selected experiments. EC and SMC transfected with AllStars negative control siRNA (Qiagen) served as controls.
Quantitative Real Time-Polymerase Chain Reaction-RNA from cells and tissue samples was extracted using RNeasy mini kit (Qiagen) and reverse-transcribed using iScript cDNA synthesis kit (Bio-Rad). cDNA was subsequently analyzed by quantitative PCR using gene-specific primers, as detailed in Table 1, and iTaq Universal SYBR Green Supermix (Bio-Rad) or TaqMan Mm00627280_m1 for A20 (Applied Biosystems, Foster City, CA) on a 7500 Fast Real Time PCR System (Applied Biosystems). Gene expression was quantified using the Pfaffl method (20), and results were normalized using cyclophilin A (CYPA, cell cultures) and TATA box-binding protein (Tbp) or 28 S (tissue samples), previously shown to be adequate housekeeping genes (15).
Chromatin Immunoprecipitation-For chromatin immunoprecipitation (ChIP), chromatin of nontreated and IFN␥treated SMC was cross-linked by adding formaldehyde to a final concentration of 10%. After cell lysis and shearing by sonication, chromatin was immunoprecipitated with anti-polymerase II antibody (Millipore, Billerica, MA). DNA was subsequently purified using the EZ-Chip kit (Millipore) according to the manufacturer's instructions and then analyzed by PCR using primers covering the proximal region of STAT1 transcription start codon (Ϫ18 to ϩ81), i.e. forward primer 5Ј-AACAGCC-GCGTCTAATTG-3Ј and reverse primer 5Ј-ACTACCCG-GCAGGAGAAAAG-3Ј.
Migration Assay-Supernatants of IFN␥-treated SMC were added to the lower chamber of a 96-transwell plate (5 m pore), whereas 1 ϫ 10 5 U937 cells were seeded in the upper chamber. Six hours later, migration of U937 cells to the lower chamber was analyzed using the CytoSelect Cell Migration Assay (Cell Biolabs, Inc., San Diego) according to the manufacturer's instructions.
Mouse Model of Focal Arterial Stenosis-A20 heterozygote (HET) mice (21), a kind gift of Dr. Averil Ma (University of California in San Francisco) and WT littermate mice, fed standard chow diet, were used in a model of focal arterial stenosis, as achieved by partial carotid artery ligation (CAL). Prior to surgery, mice were anesthetized by an intraperitoneal injection of a mixture of ketamine (50 mg kg Ϫ1 ) and xylazine (10 mg kg Ϫ1 ). After preparing mice with betadine and alcohol and shaving the neck from thorax to jaw, we performed a midline incision from mandible to thorax. Using blunt end forceps, carotid sheath structures were dissected to mobilize the common carotid artery. A blunt 35-gauge needle (World Precision Instruments, Inc., Sarasota, FL) was placed parallel to the artery, and a 9-0 nylon suture was placed 2.5 mm proximal to the bifurcation and tied with a surgeon's knot around both artery and needle. The needle, which served as mandrel, was then carefully removed to restore blood flow (Fig. 6A). Sham mice received identical treatment but without vessel ligation. The neck was closed with a 5-0 nylon suture. All surgical procedures were conducted aseptically, and animal body temperature was maintained at 37°C on a heated water pad. For pain control, Meloxicam (5 mg kg Ϫ1 ) was injected subcutaneously up to 2 days post-surgery. Sutures were removed 10 days after surgery.
For tissue harvesting, animals were sacrificed 10 days (n ϭ 5-6 per group) or 4 weeks (n ϭ 5-6 per group) after surgery, and their carotid arteries were recovered and either frozen in liquid nitrogen for RNA isolation (10 days) or embedded in tissue freezing medium (Triangle Biomedical Science, Durham, NC) for immunohistochemistry (4 weeks). All animal experiments were approved by the Institutional Committee for Use and Care of Laboratory Animals and were in accordance with the United States Department of Health and Human Services "Guide for the Care and Use of Laboratory Animals." Histology-For morphometric analysis, serial 6-m tissue sections were collected up to 1500 m proximal to the stenosis site and stained with H&E. After careful delineation of the external and internal elastic laminae, we measured media, intima, and lumen surface areas using the ImageJ software, as described (9). By immunohistochemistry, carotid sections were stained for Cd3 ϩ cells using a specific antibody for mouse Cd3 (Abcam), and for NK cells using the Nkp46-specific antibody (R&D Systems, Minneapolis, MN) followed by HRP-conjugated secondary antibody (Invitrogen). Stat1 expression in vascular tissue sections was evaluated by immunofluorescence using a polyclonal anti-mouse Stat1 primary antibody (Santa Cruz Biotechnology), followed by Alexa Fluor 594-conjugated secondary antibodies (Invitrogen). Nuclear counterstain was done using 4Ј,6-diamidino-2-phenylindole (DAPI, 1 g/ml, Sigma).
Laser Capture Microdissection-Mice were anesthetized and prepared as described above. The belly was shaved and harvested by using aseptic technique and RNase-free surgical instruments. The thoracic aortae were retrieved and rinsed with heparin (10 units/ml). Subsequently, aortic tissue was embedded in tissue freezing medium (Triangle Biomedical Science) and frozen in liquid nitrogen. For whole aortic lysates, tissue was snap-frozen in liquid nitrogen. Serial 6-m sections were cut using a cryotome E (Fisher), captured on RNase-free slides, and coated with RNAlater (Ambion, Austin, TX). Slides were then dehydrated sequentially in ethanol and xylene. By laser capture microdissection (LCM), performed on Arcturus XT LCM system (Applied Biosystems), we dissected out the endothelium, then dissected the medial layer off the adventitia, and captured it on a specimen cap.
Microarray Analysis-Medial layer samples from LCM were lysed in Arcturus PicoPure RNA lysis buffer, and RNA was isolated according to the manufacturer's protocol (Applied Biosystems). Samples with RNA integrity numbers of Ͼ6.0, as assessed by QC bioanalyzer (Agilent, Santa Clara, CA), were amplified and labeled using NuGEN Ovation Pico WTA System Version 2 and Encore TM biotin module (Fisher Scientific), respectively. After final purification using MinElute Reaction Cleanup Kit (Qiagen), hybridization to Affymetrix Mouse Gene ST2.0 microarrays (Affymetrix, Santa Clara, CA) was performed at the Microarray Core Facility of Children's Hospital, Boston. Normalization and analysis of microarray data were carried out using R/Bioconductor statistical software packages. Canonical pathway enrichment analysis was performed using Ingenuity Pathway Analysis (IPA) tools.
Statistics-Differences between groups were analyzed by ANOVA followed by post hoc Bonferroni's or Tukey's multiple comparison tests using GraphPad Prism. Student's t test was used when comparing differences between mRNA expression levels in WT versus HET aortae. p Ͻ 0.05 was considered significant.

A20 Knockdown Increases and A20 Overexpression Decreases IFN␥-mediated Up-regulation of Atherogenic Genes in Human
Endothelial and Smooth Muscle Cells-To investigate whether A20 impacts pathologic IFN␥ signaling in vascular cells, we evaluated the response of A20-deficient and A20-overexpressing human coronary artery EC and SMC cultures to IFN␥. In particular, we probed for mRNA levels of bona fide IFN␥ atherogenic genes, including intercellular adhesion molecule-1 (ICAM-1) (22), the chemoattractant molecules IP-10, monocyte chemoattractant protein-1 (MCP-1), and interferon-inducible T cell ␣-chemoattractant (I-TAC) (23), the transcription factor interferon regulatory factor-1 (IRF1) (24), and the metabolic and immunoregulatory enzyme IDO (25). In loss-offunction studies, we silenced A20 by siRNA transfection, which decreased A20 mRNA levels by 70 -80% in EC and SMC (Fig. 1,  A and B). This corresponded to total blunting of the A20 protein in SMC (Fig. 1C). IFN␥-mediated up-regulation of all tested interferon-stimulated genes (ISG) was significantly greater in A20-silenced versus control (nontransfected and transfected with AllStars siRNA) EC and SMC (Fig. 1, A and B). ICAM-1 and MCP-1 mRNA levels, already significantly higher at baseline in A20-silenced versus control SMC, were further super-induced in these cells following IFN␥ treatment. This contrasted with their negligible increase in control cells upon protein levels in control, A20 siRNA, and control siRNA-transfected SMC, 6 and 24 h after IFN␥ treatment. C, immunoblotting for A20 verified knockdown, whereas immunoblotting for GAPDH corrected for loading and enabled semi-quantitative evaluation of IDO by densitometry (ImageJ). Graphs show mean Ϯ S.D. of 3-7 independent experiments. EC and SMC derived from three different donors were used in all experiments. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001. N.D., not detectable. treatment with IFN␥ ( Fig. 1, A and B). At the protein level, we verified by Western blot and ELISA that A20 knockdown significantly increased IFN␥-mediated up-regulation of IDO (Fig.  1C) and IP-10 ( Fig. 1D).
In gain-of-function studies, we overexpressed A20 by means of rAd.-mediated transduction, which achieves expression of the transgene in Ͼ95% of cultured cells (9). Overexpression of A20 in SMC significantly blunted IFN␥-mediated up-regulation of ICAM-1, IP-10, MCP-1, I-TAC, IRF1, and IDO mRNA ( Fig. 2A). Remarkably, mRNA levels of ICAM-1, MCP-1, and I-TAC were not significantly induced by IFN␥ treatment in A20-overexpressing cells. We confirmed this outcome for IDO and IP-10 at the protein level (Fig. 2, B and C). These data uncover a novel function for A20 as a physiologic regulator and a potential therapeutic inhibitor of atherogenic IFN␥ signaling in vascular cells.
A20 Modulates IFN␥ Signaling in SMC by Regulating Expression of STAT1 in a Non-NF-B-dependent Manner-This significantly impacts monocyte chemoattractant properties of IFN␥-stimulated SMC. Having established A20 as a physiologic regulator of IFN␥ signaling in vascular cells, we probed for the molecular target(s) of A20 within the IFN␥ signaling cascade that could account for this effect. After ruling out any impact of A20 on surface expression of IFNGR1 and -2 in SMC (data not shown), we determined that A20 silencing significantly increased, whereas A20 overexpression significantly decreased basal mRNA and protein levels STAT1, the key transducer of IFN␥ signals, in SMC (Fig. 3, A and D). We confirmed STAT1 as the principal mediator of IFN␥ signals in SMC by showing that STAT1 silencing, akin to A20 overexpression, inhibits IFN␥mediated up-regulation of all atherogenic genes analyzed above (Fig. 3E). We obtained similar results in EC. Indeed, A20 silenc-   ). B, representative Western blot analysis of phospho-Tyr-701 and total STAT1 in control, A20 siRNA, and control (C) siRNA-transfected SMC, before and 5 min after 100 units/ml IFN␥. Immunoblotting for GAPDH corrected for loading and enabled semi-quantitative evaluation of STAT1 by densitometry using ImageJ. C, qRT-PCR analysis of basal STAT1 mRNA in nontransduced and rAd.A20-or rAd.␤gal-transduced SMC. D, representative Western blot analysis of phospho-Tyr-701 and total STAT1 in nontransduced, rAd.A20-, and rAd.␤gal-transduced SMC, before and 5 min after treatment with 400 units/ml IFN␥. E, STAT1 and downstream IFN␥-stimulated genes (ISG) ICAM-1, IP-10, MCP-1, I-TAC, IRF1, and IDO mRNA levels in nontransfected (NT), and STAT1 siRNA or control (C) siRNAtransfected SMC 6 h following IFN␥ treatment (100 units/ml), as measured by qRT-PCR. F, representative Western blot analysis of total STAT1 in control, and A20 siRNA or C siRNA-transfected EC. Immunoblotting for the housekeeping protein GAPDH corrected for loading and enabled semi-quantitative evaluation of STAT1 by densitometry using ImageJ. G, migration of U937 monocytic cells in response to conditioned media, applied in the lower chamber of a 5-m transwell plate. Conditioned medium was recovered from IFN␥-treated nontransduced/nontransfected (Ctrl), control (C siRNA), A20, or STAT1 siRNA-transfected, and rAd.A20-or rAd.␤gal-transduced SMC. Migration of oU937 cells into the lower chamber was determined by fluorescence using a 485/538-nm filter after labeling with CyQuantா GR dye. Results are reported as relative light units. Graphs represent mean Ϯ S.D. of 3-6 independent experiments using EC and SMC derived from three different donors. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001. NS, not significant.
ing resulted in a significant increase in STAT1 protein levels in these cells (Fig. 3F). Paralleling STAT1 protein levels, levels of Tyr-701 phospho-STAT1 (a proxy for STAT1 activation) following IFN␥ treatment were significantly higher in A20-silenced and significantly lower in A20-overexpressing SMC than in control cells (Fig. 3, B and D).
The biologic importance of A20-dependent modulation of STAT1 expression to contain IFN␥ signaling is highlighted by our data showing a significantly higher migration rate of monocytic U937 lymphoma cells in response to supernatants from IFN␥-treated (24 h) A20-silenced SMC versus control cells (Fig.  3G). Conversely, the migration rate of U937 cells was significantly lower than controls when using supernatants recovered from A20 overexpressing or STAT1-silenced SMC treated with IFN␥ for 24 h (Fig. 3G).
Despite well described synergy between STAT1 and NF-B pathways (26,27), overexpression of IB␣ in SMC failed to alter basal STAT1 mRNA and protein levels or IFN␥-mediated upregulation of STAT1 (protein), ICAM-1 (mRNA), and IDO (mRNA and protein) (Fig. 4, A and B). This result indirectly rules out the contribution of A20's NF-B inhibitory function to its modulation of STAT1 and selected ISG expression in vascular cells. The inability of IB␣ overexpression to reduce heightened STAT1 levels in A20-silenced SMC further supports this conclusion (Fig. 4C) Notably, overexpression of IB␣ in SMC still significantly reduced IFN␥-induced up-regulation of IP-10 (Fig. 4, A and D), which implied that certain ISG still required NF-B activity to be expressed in SMC (28).
Aggravated Pathologic Remodeling Following Focal Carotid Artery Stenosis in A20 Heterozygote Mice Associates with Increased Expression of Atherogenic Genes, in Particular a Unique Subset of IFN␥-dependent Genes-To validate in vivo this novel function of A20 as a regulator of IFN␥ signaling in vascular cells and to gauge its implication in modulating the severity of occlusive vascular disease, we subjected A20 HET to a model of hemodynamic injury of the carotid artery that causes significant intimal hyperplasia (IH). In this model, vascular injury is achieved by partial CAL, which by altering blood flow and shear stress results in florid IH and lumen reduction within 4 weeks of the procedure (Fig. 5A). We confirmed this outcome in A20-competent WT mouse arteries (Fig. 5B). H&E-stained tissue sections of WT carotid arteries showed significant lumen narrowing, as evaluated by the ratio of lumen area over lumen and neointima areas, at stenosis site and up to 300 m proximal to it. Narrowing of arterial lumen lessened by 600 and 1000 m and was totally absent by 1500 m proximal to the stenosis (Fig.  5C). Paralleling luminal narrowing, intima over media (I/M) ratio was highest at the site of stenosis and receded by 1500 m proximal to the stenosis (Fig. 5, B and D). Although lumen narrowing was similar at the stenosis site in HET and WT carotid

. A20-mediated decrease in basal STAT1 expression and subsequent inhibition of IFN␥ signaling in SMC cannot be recapitulated by overexpression of the standard NF-B inhibitor IB␣. A, STAT1
and select ISG (ICAM-1, IDO, and IP-10) mRNA levels before and 6 h after IFN␥ treatment, respectively, in nontransduced (Ctrl), rAd.IB␣-, and rAd.␤gal-transduced SMC, as determined by qRT-PCR and normalized by mRNA levels of the housekeeping gene cyclophilin A (CYPA). B, STAT1 and IDO protein expression in nontransduced, rAd.A20-, rAd.IB␣-, or rAd.␤gal-transduced SMC before and after 24 h stimulation with IFN␥ (100 units/ml). C, STAT1 protein expression in IB␣ (rAd.IB␣) and control ␤-galactosidase (rAd.␤gal) overexpressing SMC that were transfected with A20 siRNA or control (C) siRNA. In both B and C, IB␣ and ␤gal transgene expression was confirmed by immunoblotting. Also, GAPDH immunoblots were used to correct for loading and enable quantitative evaluation of STAT1 and IDO by densitometry using the ImageJ software. D, supernatants of SMC were recovered 24 h after stimulation with IFN␥ (100 units/ml) for ELISA measurements of IP-10. Bars represent mean Ϯ S.D. of 3-4 independent experiments using SMC derived from three different donors. (*, p Ͻ 0.05; **, p Ͻ 0.01).
arteries, it was significantly worse in HET carotids at all other analyzed sites and remained substantial by 1500 m proximal to the stenosis (Fig. 5, B and C). I/M ratios were higher in HET versus WT carotid arteries at all analyzed sites, and this difference was significant at 300 and 600 m proximal to the stenosis (Fig. 5, B and D). Altogether, these data indicated that partial loss of A20 aggravates IH following focal stenosis by increasing lesion size and length. We confirmed A20 knockdown by demonstrating ϳ30% lower A20 mRNA in carotid arteries of HET versus WT mice before stenosis (Fig. 5E). HET carotids also failed to adequately up-regulate A20 mRNA 10 days after CAL, which resulted in 50% lower A20 mRNA in HET versus WT carotid arteries at this time point (Fig. 5E).
In exploring the molecular signature of aggravated vascular remodeling in HET carotid arteries, we checked for Ifn␥ mRNA levels 10 days following CAL. Ifn␥ mRNA levels were substantially increased in carotid arteries following CAL, confirming IFN␥ as part of the molecular response to hemodynamic injury (Fig. 6A). However, Ifn␥ levels were not significantly different between HET and WT. We next investigated the presence of cytotoxic Th1 and NK cells, the main source of Ifn␥ (29), and showed heightened Cd3 ϩ T cell infiltration (but not NK cells, data not shown) in media and neointima of HET and WT carotid arteries (Fig. 6B). This suggested that T cells were the likely source of Ifn␥ in this model. We next evaluated mRNA levels of the Ifn␥ inducible proatherogenic genes previously screened in vitro. Agreeing with heightened Ifn␥ levels in carotid arteries following CAL, Icam-1, Ip-10, and Mcp-1 mRNA levels were significantly increased in both WT and HET versus Sham treated vessels 10 days after the procedure (Fig. 6C). However, CAL-induced increase in Icam-1 and Ip-10 mRNA levels was significantly greater in HET than in WT carotid arteries (Fig. 6C). Interestingly, I-Tac, Irf1, and Ido mRNA levels also significantly increased in A20 HET but not in WT carotid arteries 10 days following CAL (Fig. 6C). Together, these results indicated that aggravated IH in HET carotid arteries associates with both quantitative and qualitative differences in expression of atherogenic ISG when compared with WT vessels. Remarkably, expression of ISG whose up-regulation either requires synergy between IFN␥ and NF-B (IP-10 and I-TAC) or only depends on IFN␥ (IRF1 and IDO) occurred only in HET and not in WT FIGURE 5. Partial A20 knockdown aggravates vascular remodeling after CAL. A, carotid artery focal stenosis model. The left carotid artery was ligated 2.5 m proximal to the bifurcation using a 35-gauge needle as mandrel. The mandrel was then removed to restore blood flow. B, 4 weeks after CAL, the vessel was retrieved and sections at 300, 600, 1000, and 1500 m proximal to CAL were analyzed for neointima formation by H&E staining. Arrows on representative photomicrographs indicate media (M) and intima (I). The degree of lumen occlusion ratio (1 Ϫ(lumen)/(lumen ϩ neointima)) (C) and I/M ratios at different sites from the stenosis (D) were determined by morphometric analysis of vascular sections using ImageJ software (n ϭ 6 -7 mice/group). E, A20 mRNA levels in carotid arteries 10 days after CAL were determined by qRT-PCR. n ϭ 5-6. (∨∨, p Ͻ 0.01; ∨∨∨, p Ͻ 0.001 versus 1500-m section in each given group, and *, p Ͻ 0.05; **, p Ͻ 0.01.) WT, wild-type mice; HET, A20 heterozygote mice. SHAM-treated carotid arteries served as controls.
vessels. We verified that none of the four aforementioned genes could be induced in human SMC by the sole NF-B activators TNF and IL-1␤ (data not shown).
Upstream of ISG, Stat1 mRNA levels were both significantly up-regulated and significantly higher in HET versus WT mouse carotid arteries 10 days after CAL (Fig. 6D). By immunofluorescence, we also noted higher number of Stat1-positive nuclei/ high power field in neointima of HET versus WT carotid arteries, 4 weeks after CAL. Because nuclear localization of Stat1 is an indication of its activation, these data indicated that higher Stat1 expression corresponded with higher levels of its activated form in HET vessels (Fig. 6E).

Microarray Analysis of Aortic Medial SMC from A20 HET Versus WT Mice Implicates Increased Ifn␤ Levels and Signaling in Promoting Higher Stat1 Expression Levels in HET Vessels-
To check the mechanism(s) involved in A20-mediated regulation of STAT1, we evaluated the rate of STAT1 mRNA decay in A20 overexpressing SMC treated with actinomycin D, an inhibitor of mRNA synthesis, and we showed it was comparable with control cells (data not shown). This ruled out any impact of A20 on degradation rate or half-life of STAT1 mRNA. Also, we excluded any impact of A20 on STAT1 proteasomal degradation (30,31) by showing that addition of the proteasome inhibitor MG132 fails to increase STAT1 protein levels in A20 overexpressing SMC (data not shown). Rather, we showed that A20 regulates STAT1 expression by influencing its transcription. Indeed, we demonstrated by ChIP assay that polymerase II was recruited less to the STAT1 transcriptional start site in A20overexpressing versus control SMC (Fig. 7A).
To gain further insights into the molecular basis of A20-mediated regulation of STAT1 transcription in vascular cells, we isolated aortic medial SMC by LCM from WT and A20 HET mice and performed mRNA expression analysis using Affymetrix mouse gene 2.0 ST array (Ͼ24,000 coding transcripts). Canonical pathway enrichment using IPA tools showed significant enrichment in type II and unexpectedly type I (predominantly IFN␤) IFN-associated genes in the medial SMC of HET versus WT aortae. All 19 differentially expressed Ifn␤-associated genes, using a cutoff of 1.5-fold, were higher in HET versus WT aortae (Fig. 7B). We validated by quantitative PCR higher mRNA levels of two of these genes in HET versus WT aortae as follows, mitogen-activated protein kinase kinase kinase 7 (Map3k7) and Stat2, respectively, implicated in increasing IFN␤ transcription and signaling (Fig. 7C) (16,32). We also validated that A20 HET aortae had significantly higher mRNA levels of Stat1 and Irf1 (Fig. 7D).
To evaluate whether heightened IFN␤ signaling in HET media contributed to increased STAT1 expression, we checked whether antibody-mediated neutralization of IFN␤or siRNA-induced knockdown of IFN␤ reduces STAT1 levels in A20-silenced SMC. Anti-IFN␤ but not anti-IFN␣ antisera significantly decreased STAT1 mRNA in A20-silenced SMC to levels of control cells (Fig. 7E). A similar decrease in STAT1 mRNA levels occurred upon silencing IFN␤ in A20-silenced SMC (Fig. 7F). Treatment of A20-silenced SMC cultures with neutralizing anti-IFN␤ antiserum, or co-silencing IFN␤ in these cells also significantly lowered IFN␥-induced up-regulation of STAT1-dependent genes ICAM-1, IP-10, and IDO (Fig.  7, E and F). These results highly suggested that A20 regulates STAT1 expression and subsequently IFN␥-triggered signal transduction in vascular cells by modulating basal IFN␤ levels. Despite the technical difficulties to measure basal IFN␤ protein levels, we demonstrated using a hypersensitive IFN␤ ELISA that A20 knockdown substantially increased basal IFN␤ levels in supernatants of SMC (Fig. 7G). Basal IFN␤ levels remained undetectable (i.e. sub-threshold) in control cells.

A20 Modulates Sub-threshold IFN␤ Levels in SMC by Regulating Expression and Activation Status of Its Transcriptional
Activators IRF3 and IRF7-IFN␤ transcription relies on activation of IRFs, namely IRF3 and IRF7. Hence, we measured IRF3 FIGURE 6. STAT1/IFN␥-dependent, pro-atherogenic gene expression is increased in HET versus WT carotid arteries post-CAL. A, Ifn␥ mRNA levels in carotid arteries from WT and A20 HET mice retrieved 10 days after CAL (gray histograms) were analyzed by qRT-PCR (n ϭ 5-6 mice/group). SHAM-treated carotid arteries served as controls (white histograms). B, representative photomicrographs of Cd3 ϩ T cells immunostaining in mouse carotid arteries 4 weeks after CAL; dotted squares indicate amplified areas of the photomicrograph. Icam-1, Ip-10, Mcp-1, I-Tac, Irf1, and Ido (C) and STAT1 mRNA levels in carotid arteries from WT A20 HET mice (D) retrieved 10 days after CAL were analyzed by qRT-PCR (n ϭ 5-6). SHAM-treated carotid arteries served as controls. E, immunofluorescence staining for Stat1 (red) expression in WT and A20 HET carotid arteries 4 weeks after CAL. Nuclei were stained in blue using DAPI; n ϭ 3 (*, p Ͻ 0.05; **, p Ͻ 0.01). N.D., not detectable; NS, not significant. NOVEMBER 7, 2014 • VOLUME 289 • NUMBER 45 and IRF7 mRNA levels in WT and HET aortae. Although IRF3 is constitutively expressed in most cell types, including SMC, IRF7 is transcriptionally induced by type I IFN signaling and is possibly part of a positive feedback loop aimed at enhancing IFN␣/␤ expression (33). Whereas IRF3 mRNA levels were comparable in HET and WT aortae, IRF7 mRNA levels were significantly higher in HET versus WT aortae (Fig. 8A). We obtained concordant results in SMC cultures and showed that A20 silencing significantly increased, whereas A20 overexpression significantly decreased IRF, mRNA levels as compared with control cells, but again IRF3 mRNA levels were not affected by A20 knockdown or overexpression (Fig. 8, B and C). Beyond mRNA levels, transcriptional activity of IRF3 and IRF7 ultimately depends on their phosphorylation by activated noncanonical kinases TBK-1 and IB-kinase-⑀ (IKK⑀) (34). Interestingly, we detected a significant increase in basal levels of Ser-172 phosphorylated (not total) TBK1 protein in A20-silenced versus control SMC (Fig. 8D), precluding an increase of IRF3 and IRF7 activation, and subsequently (as previously verified) to an increase of basal transcription of IFN␤. Similar to SMC, A20-silenced EC also had higher basal levels of Ser-172phosphorylated TBK1 than control EC (Fig. 8D).

DISCUSSION
Despite the initial controversy (35,36), IFN␥, highly expressed in atherosclerotic lesions of patients and experimental animals, is now a recognized culprit of pathologic vascular remodeling (2)(3)(4). Infusion of exogenous IFN␥ induces atherosclerotic lesions in human vascular allografts implanted in mice with severe combined immunodeficiency (5), and Ifngr knock- FIGURE 7. A20 knockdown activates basal Ifn␤ signaling in mouse medial aortic SMC, total mouse aortic lysates, and human coronary artery SMC. A, ChIP using the anti-polymerase II (Pol II) antibody, in nontransduced, rAd.A20-, or rAd.␤gal-transduced SMC before and 1 h after IFN␥ treatment (100 units/ml). Data shown are representative of two independent experiments. B, heat map of differentially expressed Ifn␤ pathway-related genes in media of A20 HET versus WT mouse aortae, after isolation by LCM. Columns represent samples, and rows represent genes. Gene expression is shown with a pseudo-color scale (Ϫ2 to 2), with red indicating increased and green indicating decreased expression. Relative mRNA expression of Stat2 and Map3k7 (C) and Stat1 and Irf1 (D) in the aortae of WT versus HET mice was analyzed by qRT-PCR (n ϭ 9 -11). Basal STAT1-and IFN␥ (100 units/ml, 6 h)-induced ICAM-1, IP-10, and IDO mRNA levels were evaluated by qRT-PCR in nontransfected (Ctrl), A20 siRNA, and control (C) siRNA-transfected SMC treated with neutralizing anti-IFN␣ or anti-IFN␤ antiserum for 24 h (E), and in SMC co-transfected with a combination of C and A20 siRNA or A20 and IFN␤ siRNA (F). SMC co-transfected with a double dose of C siRNA served as control. Graphs represent mean Ϯ S.D. of three independent experiments. G, basal IFN␤ protein levels were determined in supernatants of nontransfected (Ctrl), A20 siRNA, and control (C) siRNA-transfected SMC by ELISA. Graphs depict mean Ϯ S.D. of three independent experiments using SMC derived from three different donors. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001. NS, not significant; N.D., not detectable. FIGURE 8. Increased STAT1 expression and sensitivity toward IFN␥ signaling after A20 knockdown is dependent on basal IFN␤ expression. A, mRNA expression of Irf3 and Irf7 in aortae of WT versus A20 HET mice were analyzed by qRT-PCR (n ϭ 9 -11). Relative mRNA levels of basal IRF3 and IRF7 mRNA in nontransfected (Ctrl), A20 siRNA, and control (C) siRNA-transfected SMC (B) and control, rAd.A20-, and rAd.␤gal-transduced SMC, as evaluated by qRT-PCR (C). Graphs represent mean Ϯ S.D. of 4 -6 independent experiments. D, representative Western blot analysis of basal phospho-Ser-172 and total TBK1 in control and A20 siRNA or C siRNA-transfected SMC and EC. Immunoblotting for GAPDH was corrected for loading and enabled semi-quantitative evaluation of phospho-Ser-172-TBK1 by densitometry, using ImageJ (n ϭ 3). **, p Ͻ 0.01; ***, p Ͻ 0.001. down attenuates atherosclerotic lesions in atherosclerosisprone ApoE knock-out mice (6). In this study, we noted significant super-induction of ISG in A20-deficient human EC and SMC in response to exogenous IFN␥. Conversely, overexpression of A20 in SMC significantly inhibited IFN␥-induced up-regulation of these ISG, highlighting the translational promise of A20-based therapies to curb IFN␥-driven vascular pathologies.
Interestingly, A20 impacted IFN␥ signaling by uniquely modulating expression levels of its signaling transducer STAT1 (17). A20 knockdown significantly increased STAT1 mRNA and protein levels, both in vitro (EC and SMC) and in vivo (HET mouse aorta). This transpired in higher levels of phosphorylated/active STAT1 levels, paralleling total STAT1 protein levels, in A20-silenced versus control SMC after IFN␥ treatment. Conversely, overexpression of A20 in SMC significantly inhibited STAT1 mRNA and protein levels. These results are in keeping with recent data demonstrating that A20-deficient astrocytes also have increased Stat1 expression and signaling (37). However, although increased Stat1 expression in A20 KO astrocytes was linked to heightened NF-B activation, in these cells, we showed that it was independent from A20's NF-B inhibitory function in vascular cells. Indeed, inhibition of Nf-B by overexpression of IB␣ (rather than A20) totally failed to decrease STAT1 in SMC. Furthermore, overexpression of IB␣ was unable to decrease heightened STAT1 levels in A20-silenced SMC. The discrepancy in mechanism(s) engaged by A20 to impact STAT1 expression in astrocytes versus SMC may relate to cell type specificity and/or potentially result from unforeseen effects of the chemical IKK inhibitor used in astrocytes, although we used genetic means to inhibit NF-B (37). ChIP assays and proteasome inhibition experiments supported transcriptional regulation as the main mechanism of A20-dependent modulation of STAT1 in SMC.
Using monocyte migration in response to the chemoattractant properties of IFN␥-treated SMC as an in vitro biologic surrogate for this cytokine's pro-inflammatory/pro-atherogenic effects, we demonstrated that STAT1 silencing in SMC was as protective as A20 overexpression in limiting monocyte migration. This implied that STAT1 targeting by A20 was key to its ability to harness pathologic vascular remodeling (26).
We confirmed this in vivo by showing that a mere partial loss of A20 in carotid arteries of A20 HET mice significantly increased Stat1 expression following CAL, a model that approximates hemodynamic perturbations associated with vascular stenosis in patients. HET carotid arteries, which start with 30 -50% lower baseline levels of A20 and fail to up-regulate it following CAL, displayed a pro-atherogenic profile characterized by the remarkable up-regulation of strict IFN␥-dependent ISG (I-Tac, Irf1, and Ido) (38) together with amplified inflammation (Icam-1 and Ip-10), all of which correlated with aggravated IH. These results not only implied a pathogenic role for Ifn␥ in CAL-induced pathologic vascular remodeling but also confirmed A20 as a novel physiologic modulator and potential therapeutic target of atherogenic Ifn␥/Stat1 signaling in the vessel wall.
In seeking the molecular basis for A20-mediated regulation of Stat1 transcription, we uncovered, thanks to transcriptional profiling of HET versus WT medial aortic SMC, that A20 knockdown also promoted Ifn␤ signaling and expression. Even in the absence of a viral insult, sub-threshold concentrations of Ifn␤ accumulate in tissues, providing local immune surveillance by engaging type I IFN signaling to support basal Stat1 expression. Basal expression of Stat1 is prerequisite for enabling secondary Ifn␥ and full-blown Ifn␣/␤ signaling, typically in response to viral infections (39). Accordingly, basal Stat1 levels are reduced in macrophages and mouse embryonic fibroblasts of IFN␤ and IFN␣ receptor (Ifnar)-1 KO mice (17,40). Also, defective Ifn␥ antiviral activity in Ifn␤1 KO mice recovers upon restoring adequate Stat1 levels.
Messenger RNA levels of Stat2, which form together with Stat1 and Irf9 the type I IFN transcriptional complex interferon-stimulated gene factor (ISGF3) (16), were also significantly increased in HET versus WT aortae. This result agrees with data showing increased Stat1 and Stat2 levels in cells with chronic elevation of IFN␤ levels (41). Interestingly, nonphosphorylated STAT1 and STAT2 can still bind IRF9 to form ISGF3 and hence drive, in a feedback loop, their own transcription (41). This maintains cells in a primed state, ready for fullscale antiviral responses. Our data showing that antibodymediated neutralization of IFN␤ or siRNA-induced IFN␤ knockdown decreased basal STAT1 and blunted IFN␥-mediated hyperinduction of ICAM-1, IP-10, and IDO mRNA levels in A20-silenced SMC confirm the link between IFN␤ and STAT1 in our system. To our knowledge these data are the first showing that loss of A20 modulates STAT1 and STAT2 expression in vascular cells by enhancing IFN␤ levels.
The IFN␤ promoter consists of four positive regulatory domains (PRD-I to PRD-IV), representing binding sites for IRFs, p65, and c-JUN (42,43). p65 binds to PRD-II, and its activity was reported to be crucial for basal Ifn␤ expression in mouse embryo fibroblasts (44). However, there is also strong evidence that IRFs can outweigh p65 deficiency and enable adequate Ifn␤ transcription in p65 knock-out mouse embryo fibroblasts (45). In a similar vein, our data strongly argue that A20's impact on the expression and function IRF supersedes its effect on NF-B in modulating basal IFN␤ and hence STAT1 levels in SMC. Indeed, as discussed previously, overexpression of IB␣ to retain p65 in SMC cytosol and therefore block NF-B activation did not affect STAT1 expression nor downstream STAT1/IFN␥-mediated expression of ISG. However, levels of the key transcriptional regulator of IFN␤, IRF7, were significantly higher in HET versus WT mouse aortae and in A20silenced versus control SMC. Conversely, IRF7 mRNA levels were significantly lower in A20 overexpressing SMC. Because IRF7 transcription depends on type I IFN signaling, it engages in a feed-forward loop that would amplify Ifn␤ transcription and subsequently type I IFN responses, including its own upregulation (33,39).
IRF3, the other transcriptional activator of IFN␤, is constitutively expressed, and its activation, together with that of IRF7, is mostly regulated by IKK⑀/TBK1-mediated phosphorylation (46). TNF receptor-associated factor-3 (TRAF3)-dependent Lys-63-linked polyubiquitination of TBK1 is required for its dimerization and autophosphorylating activation at Ser-172 (47). Our data revealed that A20 knockdown in SMC signifi- NOVEMBER 7, 2014 • VOLUME 289 • NUMBER 45 cantly increased Ser-172 basal phosphorylation of TBK1, and it was therefore likely to enhance IRF3 and IRF7 activation. This result resonates with earlier work showing that myeloid-specific A20 knock-out mice exhibit enhanced IRF3 activation, which protected them from influenza A viral infection (48). It is also in keeping with gain-of-function studies demonstrating that A20 precluded virus-induced up-regulation of type I IFN in HEK293 cells by inhibiting IRF3/7 activation (49 -52).

Loss of A20 Aggravates Pathologic Vascular IFN␥ Signaling
The mechanism(s) by which A20 modulates TBK1 phosphorylation in vascular cells remain(s) to be explored. We surmise that it may relate, as in mouse embryo fibroblasts and 293T cells, to the disruption of the TRAF3 and IKK⑀-TBK1 complex by A20 and its partner T cell leukemia virus type I-binding protein 1 (TAX1BP1) (53). This would preclude ubiquitination and secondary phosphorylation of IKK⑀-TBK1 and hence the ability of these kinases to activate IRF3/IRF7, upstream of IFN␤ transcription (53). Alternatively, A20 may modulate expression or activity of signaling molecules upstream of TBK1. One such molecule could be the kinase MAP3K7/TAK1 (32), whose mRNA levels were significantly higher in HET versus WT aortae. Additional experiments are needed to address these hypotheses. Regardless of the mechanism(s), our data strongly support an atherogenic (54) rather than an atheroprotective (55) role for IFN␤.
In summary, our study uncovers a novel atheroprotective function of A20 in vascular cells through modulation of IFN␥/ STAT1 signaling in an IFN␤-dependent, but NF-B-independent, manner. Partial loss of A20 increases basal IFN␤ levels and signaling, and subsequently STAT1 transcription, thereby amplifying IFN␥ signaling. In contrast, overexpression of A20 reduces basal IFN␤ signaling and hence STAT1 expression in EC and SMC, thereby hampering IFN␥ signaling. We ascribed heightened IFN␤ levels in A20-silenced vascular cells to increased phosphorylation of the noncanonical IKK, TBK1, upstream of IRF3 and IRF7, the direct transcriptional activators of IFN␤ (Fig. 9).
Clinically, tag polymorphisms at the A20/TNFAIP3 locus that associate with 30 -45% decrease in A20 mRNA levels elicit 2-3-fold higher risk for coronary artery disease in diabetic patients carrying minor versus major alleles (56). Whether these alleles also associate with heightened levels of basal IFN␤ and STAT1 and with amplified IFN␥ responses in the vessels of these patients needs to be determined. Such single nucleotide polymorphisms at the A20/TNFAIP3 locus may prove highly informative of patients' risk for vascular disease.
From a therapeutic standpoint, our data strongly support the promise of A20-based therapies for the prevention/treatment of atherosclerotic vascular diseases, based on its ability to not only intercept the NF-B pathway but also another major atherogenic pathway, i.e. the IFN␥/STAT1 pathway. To that end, our discovery that A20 modulates TBK1 phosphorylation and activation in vascular cells could represent a "druggable" target that could recapitulate the ability of A20 to interrupt IFN␥ -driven atherogenesis (57). Future studies will explore this hypothesis.