MDA‐5 activation by cytoplasmic double‐stranded RNA impairs endothelial function and aggravates atherosclerosis

Abstract Recent studies have highlighted the relevance of viral nucleic acid immunorecognition by pattern recognition receptors in atherogenesis. Melanoma differentiation associated gene 5 (MDA‐5) belongs to the intracellular retinoic acid inducible gene‐I like receptors and its activation promotes pro‐inflammatory mechanisms. Here, we studied the effect of MDA‐5 stimulation in vascular biology. To gain insights into MDA‐5 dependent effects on endothelial function, cultured human coronary artery endothelial cells (HCAEC) were transfected with the synthetic MDA‐5 agonist polyIC (long double‐stranded RNA). Human coronary endothelial cell expressed MDA‐5 and reacted with receptor up‐regulation upon stimulation. Reactive oxygen species formation, apoptosis and the release of pro‐inflammatory cytokines was enhanced, whereas migration was significantly reduced in response to MDA‐5 stimulation. To test these effects in vivo, wild‐type mice were transfected with 32.5 μg polyIC/JetPEI or polyA/JetPEI as control every other day for 7 days. In polyIC‐treated wild‐type mice, endothelium‐dependent vasodilation and re‐endothelialization was significantly impaired, vascular oxidative stress significantly increased and circulating endothelial microparticles and circulating endothelial progenitor cells significantly elevated compared to controls. Importantly, these effects could be abrogated by MDA‐5 deficiency in vivo. Finally, chronic MDA‐5 stimulation in Apolipoprotein E/toll‐like receptor 3 (TLR3) double‐deficient (ApoE−/−/TLR3−/−) mice‐enhanced atherosclerotic plaque formation. This study demonstrates that MDA‐5 stimulation leads to endothelial dysfunction, and has the potential to aggravate atherosclerotic plaque burden in murine atherosclerosis. Thus, the spectrum of relevant innate immune receptors in vascular diseases and atherogenesis might not be restricted to TLRs but also encompasses the group of RLRs including MDA‐5.


Introduction
Atherosclerosis is a chronic inflammatory disease of the vasculature [1], and the leading cause of death worldwide [2]. Its main manifestations such as myocardial infarction, ischaemic heart failure, peripheral artery disease and progressive stroke are responsible for significant morbidity and mortality [3]. Atherosclerotic plaque formation is a complex process that is orchestrated by vascular cells including endothelial and smooth muscle cells, but also by infiltrating cells of the innate and adaptive immune system. These cells are stim-ulated by microbial and sterile triggers that act as danger signals inducing innate and acquired inflammatory responses and contribute to atherogenesis [4].
Nucleic acids released from endogenous cells and invading pathogens represent an important danger signal and trigger of inflammation in atherosclerosis. Specific classes and motifs of these nucleic acids bind and activate so called pattern recognition receptors (PRRs) and in turn mediate the production of pro-inflammatory cytokines and growth factors attributed to atherosclerotic plaque formation [4].
Intracellular PRRs comprise cytoplasmatic retinoic acid inducible gene-I (RIG-I) like receptors (RLRs), and include RIG-I and melanoma differentiation associated gene 5 . They consist of 2Nterminal caspase recruitment domains (CARD), an DExD/box helicase domain and a C-terminal repression domain [5]. Retinoic acid inducible gene-I detects RNA with a triphosphate group at the 5 0 end [6,7], whereas MDA-5 is activated by long double-stranded RNA in the cytoplasm [8,11]. Activation by exogenous RNA induces a signalling cascade involving interferon-beta promoter stimulator 1 (IPS-1, also known as VISA, MAVS and CARDIF) [9], and nuclear transcription factors NF-jB and IRF-3, resulting in type I interferon (IFN) production and promotion of an innate immune response. However, details of the underlying cellular interactions and signalling pathways, which may ultimately lead to atherosclerosis, have not been fully deciphered.
Endothelial integrity is indispensable for vascular homeostasis, and the balance of fibrinolysis and coagulation. Furthermore, the endothelial layer directs diapedesis of activated circulating immune cells to sites of inflammation through expression of adhesion molecules, and can be regarded as part of the immune systems itself. In atherosclerosis, endothelial dysfunction is a main feature which persists over all stages [10]. The repair of damaged endothelium is promoted by neighbouring endothelial cells (ECs) and -at least indirectly through paracrine activity -by circulating endothelial progenitor cells (EPCs). Endothelial progenitor cells are mobilized from the bone marrow, and their numbers in peripheral circulation correspond with endothelial health [11,12]. Endothelial cells and EPCs express various PRRs, but their exact role in the innate immune mechanisms involved in atherosclerosis still remains poorly understood.
Recently, we demonstrated that activation of RIG-I in human and murine ECs and EPCs induces pro-inflammatory effects and may therefore play a crucial role in endothelial biology and provoke atherogenesis [13]. Furthermore, double stranded RNA leads endothelial dysfunction via endocytosis and toll-like receptor 3 (TLR3) signalling [14], which is another type of PRR and independent of RLR.
In this study, we sought to examine specific MDA-5 stimulation of vascular cells on endothelial function, gain novel insights on the development of vascular disease and link for the first time MDA-5 to atherosclerosis.

In vitro experiments
Approval for the use of the human buffy coats was granted by the university ethics review board and in accordance with the Declaration of Helsinki.

Human coronary artery cells
Human coronary artery endothelial cells (HCAEC; Lonza, Basel, Switzerland) were cultured at 37°C and 5% CO 2 atmospheric concentration. Endothelial cell growth medium (MU, Promo Cell, Heidelberg, Germany) was renewed every 2 days and cells grown to 80% confluence. For stimulation cells were incubated with polyIC or vehicle polyA with 1 lg/ ml. For transfection, Lipofectamine 2000 (Invitrogen, Darmstadt, Germany) was used according to manufacturer's instructions. For TLR3 inhibition, TLR3/Double Strand RNA Complex Inhibitor (Calbiochem/ Merck Millipore, Darmstadt, Germany) was added to cell culture as indicated in the Figures S1-S3.

Endothelial progenitor cells
Mononuclear cells were isolated from human buffy coats utilizing density Ficoll gradients and circulating angiogenic cells (CAC) were cultured using endothelium cell basal medium-2 (EBM-2; Clonetics, Lonza, Basel, Switzerland), as previously described [15]. Experiments were performed on day seven after isolation. For stimulation cells were incubated with polyIC or vehicle polyA in concentrations of 1 lg/ml. For transfection Lipofectamine 2000 (Invitrogen) was used according to manufactures instructions. For analysis of early (CFU-Hill) and late EPCs, mononuclear cells were isolated from murine spleen and seeded into Fibronectin (Sigma Aldrich, Taufkirchen, Germany) pre-coated 24 well plates at 1 9 10 7 cells/ml. After 2 days, 1 9 10 6 non-adherent cells were cultured for another 7 days. CFU-Hill were counted using light microscopy. For analysis of late EPCs, the adherent cells were cultured changing EBM-2 medium every other day. On day 21 cells were regarded late EPCs and quantified using AxioVision version 4.5.0 software (Zeiss, Oberkochen, Germany) and a Zeiss Axiovert 200 M microscope.

Immunohistochemistry
Cells were grown in 24‚ well plates on glass slides. For detection of MDA-5 expression, immunohistochemical stains were performed. Briefly, cells were fixed with acetone at À20°C. They were then washed twice in PBS plus 0.025% Triton X-100 for 5 min. and subsequently incubated in 1% bovine serum albumin (BSA) with PBS for 2 hrs at room temperature. Next, the primary antibody (Rabbit polyclonal MDA-5 antibody, ab69983; Abcam, Cambridge, UK) was diluted 1:1000 in 1% BSA with PBS and the slides allowed to incubated overnight at 4°C. Immunodetection was accomplished using an TRITC-conjugated goat anti-rabbit secondary antibody (1:1000 dilution). The nuclei were stained with DAPI mounting medium.

Measurement of reactive oxygen species in vitro
For analysis of cellular reactive oxygen species (ROS) production HCAEC and EPC were cultured as described above. Experiments were

Apoptosis
To determine HCAEC and EPC apoptosis, the Cell Death Detection ELISA Kit (Roche Diagnostics, Mannheim, Germany) was used according to manufacturer's instructions. The results are shown in relative proportion to the control group.

Proliferation
Human coronary artery endothelial cells proliferation was assessed by detection of bromodeoxyuridine (BrdU) incorporation after a 24 hrs incubation with polyIC and BrdU. Bromodeoxyuridine positive nuclei were determined using an anti-BrdU antibody (Abcam) and the total number of cells using DAPI staining.

Migration
Endothelial scratch assays were performed as previously described [13]. Briefly, ECs were grown to confluence and incubated with polyIC or vehicle polyA with 1 lg/ml cell culture medium. For transfection Lipofectamine 2000 (Invitrogen) was used according to manufactures instructions. The scratch was performed with a sterile pipette and then a marked position was photographed every 2 hrs for 10 hrs by using AxioVision version 4.5.0 software (Zeiss) and a Zeiss Axiovert 200 M microscope. The remaining cell free area was measured and then correlated with the initially scratched area (in percent).

Cytokine quantification
Human coronary artery endothelial cells were incubated with polyIC for 24 hrs, and the concentration of IL-6 and IFN-c-induced protein 10 (IP-10) was then quantified by ELISA in the supernates. Commercially available kits for human IL-6 and IP-10 (R&D Systems, Wiesbaden, Germany) were used according to the manufactures protocols. The experimental setup was part of a large screening assay and data from the control group have been published previously [13].

Animal studies Animals and surgical procedures
For this study, eight to twelve weeks old wild-type mice (C57BL/6J; Charles River, Sulzfeld, Germany), MDA-5 À/À , and ApoE À/À TLR-3 À/À mice (C57Bl6 background) were used. The experiments were performed conform to the guidelines from Directive 2010/63/EU, in accordance with institutional guidelines and the German animal protection law. All animals were kept at 22°C in a 12 hrs day/night rhythm and received food and water ad libitum. For acute injury experiments wild-type and MDA-5 -/-mice were injected intravenously with either polyIC or polyA, respectively, at doses of 32.5 lg suspended in 200 ml 0.9% NaCl solution every other day for 7 days. Transfection was accomplished with in vivo JetPei agent (Polyplus Transfection, Illkirch, France). For chronic injury experiments ApoE À/À /TLR-3 À/À mice were fed a high-fat, cholesterol-rich diet, that contained 21% fat, 19.5 casein and 1.25% choles-terol for a total of 7 weeks. Mice were injected with polyIC and polyA as described above once a week from week 1 to 5 and then every other day during the last 2 weeks of diet. Aortic segments, hearts, spleen and blood were collected immediately after sacrifice.
For carotid artery injury experiments, wild-type mice were exposed intravenously to polyIC and polyA every other day during 10 days. The operation was performed on day 5 and animals killed on day 10.

Measurement of reactive oxygen species
Reactive oxygen species release in intact aortic segments was determined by L-012 chemiluminescence, as previously described [16]. Aortic segments were carefully excised and placed in chilled, modified Krebs-HEPES buffer. Connective tissue was removed and aortas were cut into 2 mm segments. Chemiluminescence of aortic segments was assessed in scintillation vials containing Krebs-HEPES buffer with 100 lmol/l L-012 over 15 min. in a scintillation counter (Lumat LB 9501; Berthold, Bad Wildbad, Germany) in 1 min. intervals. The vessel segments were then dried and dry weight was determined. Reactive oxygen species release is calculated as relative chemiluminescence per mg aortic tissue and as percent of control.

Flow cytometry
Murine blood samples were analysed as described previously [17]. Following red cell lysis, the viable lymphocyte population was analysed for Sca-1 (Becton Dickinson) and flk-1 (Becton Dickinson, Heidelberg, Germany) to measure EPC and AnnexinV (Becton Dickinson) and CD 31 (Becton Dickinson) to measure endothelial microparticles (EMP). Isotype identical antibodies and unstained samples served as controls in every experiment (Becton Dickinson). Cell fluorescence was measured immediately after staining using a FACS Calibur instrument (Becton Dickinson). Data were analysed using CellQuest software (Becton Dickinson). Units of all measured components are specific events obtained after measuring 50,000 events (EPC) and 20,000 events (EMP) in a pre-specified gate during FACS analysis.

Aortic ring preparations and tension recording
Vasodilation and vasoconstriction of isolated aortic ring preparations were determined in organ baths filled with oxygenated modified Tyrode buffer (37°C), as previously described [16,18]. Adventitial tissue was carefully removed, and 3 mm segments of the thoracic aorta were investigated. A resting tension of 10 mN was maintained throughout the experiment. Drugs were added in increasing concentrations in order to obtain cumulative concentration-response curves: KCl 20 and 40 mmol/ l, phenylephrine 1 nmol/l to 10 lmol/l, carbachol 10 nmol/l to 100 lmol/l (assessment of endothelium-dependent vasodilation after pre-contraction with phenylephrine), and nitroglycerin 1 nmol/l to 10 lmol/l (assessment of endothelium-independent vasodilation after pre-contraction with phenylephrine). The drug concentration was increased when vasoconstriction or -relaxation was completed. Drugs were washed out before the next substance was added.

Carotid artery injury
Electrical carotid artery denudation was performed on day 5 of polyIC treatment prior to the third injection. A small incision from the cranial apex of the sternum to just below the mandibule was made. After careful preparation of an approximately 6-mm-long segment proximal of the bifurcation, the common carotid artery was electrically denuded applying two serial 5 sec. bursts of 2 Watt using a 2-mm-wide forceps resulting in a 4 mm lesion. Afterwards the skin was sutured and all mice allowed to recover in individual cages before returning to their littermates. On day 10 of intravenous polyIC treatment, 50 ll Evans Blue were injected intravenously and allowed to circulate for 5 min., mice killed and both common carotid arteries fully excised. The vessels were rinsed in 0.9% NaCl solution and the residual connective tissue carefully removed. Photographs were taken and the remaining denuded area (stained blue) correlated with the total lesion area (4 mm) to measure re-endothelialization. Images were taken by using AxioVision version 4.5.0 software and Axiovert 200 M microscope (Zeiss).

Atherosclerotic plaques, Oil red O and CD68 staining
Hearts with ascending aorta were collected and embedded at À80°C. Samples were sectioned on a Leica cryostat (7 lm) starting at the apex and progressing through the aortic valve area into the ascending aorta and the aortic arch, and placed on slides. Aortic cryosections were then fixed with 3.7% formaldehyde for 1 hr, rinsed with deionized water, stained with oil red = working solution (0.5%) for 30 min. and rinsed again. Haematoxylin staining was performed according to standard protocols. For CD68 staining, rat anti-mouse CD68 (Acris Antibodies; dilution 1:100) was used as primary antibody, and a Cy3-conjugated goat anti-rat secondary antibody (dilution 1:200). Cell nuclei were stained with DAPI. All sections were examined using a Zeiss Axiovert 200 M microscope and AxioVision version 4.5.0 software. To quantify atherosclerotic plaque formation in the aortic root, total area and lipid stained area of serial histological sections were measured. Atheroscle-rotic data are expressed as lipid-stained area correlated with total surface area in percent.

Statistical analysis
Data are presented as mean AE S.E.M. For statistical analysis, twotailed, unpaired Student's t-test and ANOVA for multiple comparisons were employed where applicable. P < 0.05 indicates statistical significance.

MDA-5 activation induces endothelial dysfunction in wild-type mice
On the basis of our in vitro data, we next investigated whether MDA-5 activation influences endothelial biology in wild-type mice. PolyIC or control RNA (polyA) were supplemented with JetPEI for cytoplasmic transfection, and C57Bl6J mice injected intravenously every other day for 7 days. Similar to packaging polyIC with Lipofectamine (Invitrogen) for transfection in vitro, conjugation of polyIC with jetPEI (Polypus) mediates activation MDA-5 in vivo through its transfection to the cytoplasm.
First, regulation of vascular tension by aortic rings, as a measurement of endothelial function, was investigated in organ chamber experiments. Endothelium-dependent vasodilation was significantly impaired in mice transfected with polyIC, which marks endothelial dysfunction (Fig. 3A: polyIC 37.42 AE 8.96% versus polyA 18.47 AE 7.14% of maximal contraction, n = 5, P < 0.05). On the contrary, phenylephrine-induced vasoconstriction (data not shown) and endothelium-independent vasodilation through nitroglycerin ( Fig. 3B: polyIC 23.91 AE 7.84% versus polyA 20.41 AE 11.88% of maximal contraction, P > 0.05, n = 5) showed no differences between the two groups, suggesting that vascular smooth muscle cells were fully functional and intact after polyIC treatment.  The number of circulating EMPs which are released by dysfunctional ECs and therefore a biomarker for endothelial dysfunction, was significantly increased compared to control mice ( Fig. 3C: polyIC 0.73 AE 0.21% versus polyA 0.17 AE 0.05% gated events, n = 5, P < 0.05). Reactive oxygen species production of isolated aortic rings was measured by L-012 chemiluminescence. Melanoma differentiation associated gene 5 stimulation increased vascular oxidative stress compared to control mice ( Fig. 3D: polyIC 323.3 AE 49.45% versus polyA 100.0 AE 14.52%, n = 5, P < 0.05).

Endothelium-dependent Vasodilation
To assess endothelial regeneration capacity in vivo, wild-type mice were injected intravenously with polyIC or control RNA every other day for a total of 10 days. On day 5 electrical carotid artery injury was performed and a 4 mm segment proximal to the bifurcation was denuded. The rate of reendothelialization was assessed after staining with Evans Blue on day 10. Systemic MDA-5 stimulation resulted in significantly reduced reendothelialization compared to control (Fig. 4: polyIC 67.62 AE 2.53% versus polyA 42.11 AE 1.78% of total lesion area, P < 0.05, n = 5-7).

Discussion
The exact mechanisms for endothelial dysfunction and consecutive vascular inflammation in atherosclerosis are not fully understood. Our results show for the first time that specific stimulation of MDA-5 induces endothelial dysfunction and can significantly aggravate atherosclerosis, most likely through the enhanced expression of adhesion molecules on EC with subsequent plaque infiltration of macrophages, which is accompanied by oxidative stress. These findings are in line with recent studies that linked RIG-I [13], and TLR2, 3, 4, 5 and 9 to endothelial deterioration [14,[20][21][22] and the orchestration of vascular inflammatory responses. Thus, damaged ECs seem to be not only a byproduct of vascular inflammation, but actively participate in the inflammatory process through activation innate immune sensors such as MDA-5. The RLRs MDA-5 and RIG-I are mainly known as a virus sensing system that induce the production of type I IFNs by dendritic cells and macrophages. Viral fragments have been found in atherosclerotic lesions [23,24], and there is a growing body of evidence that recurrent or chronic viral infections worsen atherosclerosis. However, viral infections alone most likely do not cause the development, but rather are a contributing factor for the course of atherosclerosis. A study by Paik and colleagues demonstrated that LDLR À/À mice with norovirus infection had a significantly higher plaque burden [25]. Admittedly, the investigators found more macrophage activation in the infected mice, and did not analyse endothelial health or activity. As norovirus is a natural ligand for MDA-5, this interaction might be similar to our stimulation of ApoE À/À mice with polyIC. According to our study results, endothelial dysfunction could therefore be a complementary mechanism for the detrimental effects of viruses on atherosclerosis through the production of pro-inflammatory cytokines such as IP-10, IL6 and IL-8, and expression of cell adhesion molecules VCAM-1 and ICAM-1 upon MDA-5 stimulation. In humans, viral infections such as influenza are strongly associated with the occurrence of cardiovascular events during the flu season, and matched vaccination proved to have protective effects in patients with coronary artery disease [26].
Nevertheless, recent studies indicate that RLRs are not only involved in antiviral immune response but also contribute to chronic inflammatory disease. Retinoic acid inducible gene-I activation for instance has been associated with rheumatoid arthritis and lupus nephritis [27], and is highly expressed in intimal macrophages of atherosclerotic lesions [28]. Melanoma differentiation associated gene 5 is associated with increased risk of developing type I diabetes [5,29], and has been described as an auto-antigen for humoural immune response in myositis [30]. However, MDA-5 had not yet been associated with vascular biology and atherogenesis.
Although specific ligands for most PRR have been identified, numerous PRRs are not restricted to a single exogenous danger signal but rather show cross reactivity to several endogenous molecules such as IFNs, heat shock proteins or oxidized cholesterol [4,27]. One may speculate that nucleic acids from apoptotic or necrotic vascular cells could stimulate RLRs of neighbouring cell and provoke a proinflammatory vascular state that serves as a fertile ground to accelerate vascular inflammation in viral infections.
Retinoic acid inducible gene-I and MDA-5 share morphologic homologies and both trigger an innate immune response through a common intracellular signalling cascade. Both stimulate IPS-1 which in turn activates NF-jB and IRF-3 resulting in the production of type I IFN-b [27]. However, some differences in endothelial reaction after specific receptor stimulation can be noted. Melanoma differentiation associated gene 5 activation in vivo increased EPCs and in vitro induced HCAEC apoptosis, whereas RIG-I stimulation had no such effects. Slight differences in individual effects seem not surprising given that RIG-I and MDA-5 share only~25% homology within the CARD domain regions and 40% within the helicase domain [5].
LGP2 (Laboratory of Genetics and Physiology 2) is a third RLR that has been considered incomplete because of its lack of a CARD domain. Recent data have shown a sensitizing effect of LGP2 on MDA-5, but not on RIG-I [31].
The inhibition of PRR mediated vascular inflammation and endothelial dysfunction could benefit future anti-atherosclerotic therapies. Preventing the multimerization of the RIG-I CARD domains through the anti-aging protein Klotho, for example, diminished the release of the pro-inflammatory cytokines IL-6 and IL-8 [27].
In summary, our study demonstrates that stimulation of the cytosolic RLR MDA-5 induces endothelial dysfunction and promotes atherosclerosis. Approaches to inhibit these vascular innate immune processes should be investigated for potential beneficially effects on endothelial dysfunction or even atherogenesis.  Video S1 Representative video of endothelial cell migration scratch assay. The video is a composition of sequential photographs taken every 3 min. after physically scratching a cell-free strip. Vehicle (left) and polyIC (right) stimulated cells.