Blockade of vascular endothelial growth factor receptor 2 inhibits intraplaque haemorrhage by normalization of plaque neovessels

Abstract Background Plaque angiogenesis is associated with atherosclerotic lesion growth, plaque instability and negative clinical outcome. Plaque angiogenesis is a natural occurring process to fulfil the increasing demand of oxygen and nourishment of the vessel wall. However, inadequate formed, immature plaque neovessels are leaky and cause intraplaque haemorrhage. Objective Blockade of VEGFR2 normalizes the unbridled process of plaque neovessel formation and induces maturation of nascent vessels resulting in prevention of intraplaque haemorrhage and influx of inflammatory cells into the plaque and subsequently increases plaque stability. Methods and Results In human carotid and vein graft atherosclerotic lesions, leaky plaque neovessels and intraplaque haemorrhage co‐localize with VEGF/VEGFR2 and angiopoietins. Using hypercholesterolaemic ApoE3*Leiden mice that received a donor caval vein interposition in the carotid artery, we demonstrate that atherosclerotic vein graft lesions at t28 are associated with hypoxia, Hif1α and Sdf1 up‐regulation. Local VEGF administration results in increased plaque angiogenesis. VEGFR2 blockade in this model results in a significant 44% decrease in intraplaque haemorrhage and 80% less extravasated erythrocytes compared to controls. VEGFR2 blockade in vivo results in a 32% of reduction in vein graft size and more stable lesions with significantly reduced macrophage content (30%), and increased collagen (54%) and smooth muscle cell content (123%). Significant decreased VEGF, angiopoietin‐2 and increased Connexin 40 expression levels demonstrate increased plaque neovessel maturation in the vein grafts. VEGFR2 blockade in an aortic ring assay showed increased pericyte coverage of the capillary sprouts. Conclusion Inhibition of intraplaque haemorrhage by controlling neovessels maturation holds promise to improve plaque stability.

Background. Plaque angiogenesis is associated with atherosclerotic lesion growth, plaque instability and negative clinical outcome. Plaque angiogenesis is a natural occurring process to fulfil the increasing demand of oxygen and nourishment of the vessel wall. However, inadequate formed, immature plaque neovessels are leaky and cause intraplaque haemorrhage.
Objective. Blockade of VEGFR2 normalizes the unbridled process of plaque neovessel formation and induces maturation of nascent vessels resulting in prevention of intraplaque haemorrhage and influx of inflammatory cells into the plaque and subsequently increases plaque stability.
Methods and Results. In human carotid and vein graft atherosclerotic lesions, leaky plaque neovessels and intraplaque haemorrhage co-localize with VEGF/ VEGFR2 and angiopoietins. Using hypercholesterolaemic ApoE3*Leiden mice that received a donor caval vein interposition in the carotid artery, we demonstrate that atherosclerotic vein graft lesions at t28 are associated with hypoxia, Hif1a and Sdf1 up-regulation. Local VEGF administration results in increased plaque angiogenesis. VEGFR2 blockade in this model results in a significant 44% decrease in intraplaque haemorrhage and 80% less extravasated erythrocytes compared to controls. VEGFR2 blockade in vivo results in a 32% of reduction in vein graft size and more stable lesions with significantly reduced macrophage content (30%), and increased collagen (54%) and smooth muscle cell content (123%). Significant decreased VEGF, angiopoietin-2 and increased Connexin 40 expression levels demonstrate increased plaque neovessel maturation in the vein grafts. VEGFR2 blockade in an aortic ring assay showed increased pericyte coverage of the capillary sprouts.

Introduction
Plaque angiogenesis and intraplaque haemorrhage are critical determinants of plaque instability [1]. Plaque angiogenesis or neovessel formation correlates with lesion progression, plaque inflammation and negative clinical outcome after cardiovascular events [2,3]. Fragile atherosclerotic plaques do not only cause plaque instability in native atherosclerosis but also in postinterventional lesions such as in vein grafts and in in-stent neoatherosclerosis [4,5].
Hypoxia in atherosclerotic lesions is a driver of plaque instability [6]. Furthermore, it can induce lesion growth and affect vascular remodelling [7,8]. Angiogenesis, a natural occurring process induced by hypoxia, fulfils the increasing demand of oxygen and nourishment of the vessel wall. Neovessel formation is stimulated by hypoxiainduced up-regulation of vascular endothelial growth factor (VEGF) [9,10]. VEGF binds to and mediates its activity primarily through VEGF receptor 2 (VEGFR2). Plaque neovessels are frequently found dysfunctional, especially immature plaque neovessels. These neovessels are characterized by increased permeability caused by underdeveloped interendothelial junctions, incomplete basement membranes and partial pericyte coverage [11]. As a result, neovessels leak blood components into the lesions, that is intraplaque haemorrhage. Erythrocytes in the plaque become phagocytosed, and their cholesterol-rich membranes contribute to the free cholesterol content of the plaque [12][13][14]. Leaky neovessels are clearly associated with inflammatory cells [1]. Recently, it was shown by some of the coauthors that especially haemoglobin-haptoglobin receptor CD163+ macrophages interact with plaque neovessels and induce vascular permeability resulting in the propagation of the instable character of lesions [15].
Anti-angiogenic therapies are used in cancer and eye diseases. However, these therapies are not always found beneficial [16]. Normalization of the neovasculature, that is creating healthy mature neovessels, is a relatively new strategy to target neovascularization [17]. Generation of a basement membrane and recruitment of pericytes are crucial steps in vessel maturation. These processes are regulated by VEGF-VEGFR2 and the tightly balanced angiopoietin-Tie2 system [18]. High levels of VEGF increase vessel permeability, whereas low levels of VEGF are necessary for a stable vessel [19]. Angiopoietin (Ang)-1 mediates pericyte-endothelial cell adhesion, and Ang-2 induces vessel permeability and acts as an antagonist to Ang-1, resulting in pericyte loss [19].
In preclinical models, it has been demonstrated that pro-angiogenic strategies augment atherosclerotic plaque growth and vascular inflammation, whereas anti-angiogenic strategies inhibit atherosclerosis [20][21][22]. Previously, we have shown that lesions induced by vein grafting in atherosclerosis-prone mice display profound plaque neovessels and intraplaque haemorrhage [23]. These plaque neovessels frequently lack pericyte coverage classifying them as immature [23].
We hypothesized that improving the maturation state of plaque neovessels reduces the extent of vascular 'leakiness', which results in reduced intraplaque haemorrhage and lesion progression. Since low levels of VEGF are necessary for vessel homeostasis, we investigated the impact of the VEGFR2-blocking antibody (DC101) on plaque angiogenesis, maturation status, and atherosclerotic lesion size and composition in murine vein grafts.

Human tissue specimens
Human coronary artery vein graft specimens (n = 12) were available from the CVPath Institute. A detailed patient description can be found in Table S1. The severity of the vein graft lesions was scored as early, intermediate or late as described previously [4]. Anonymous carotid endarterectomy (n = 12) specimens obtained at the LUMC in accordance with guidelines set out by the 'Code for Proper Secondary Use of Human Tissue' of the Dutch Federation of Biomedical Scientific Societies (Federa) and conform with the principles outlined in the Declaration of Helsinki. The carotid endarterectomy specimen phenotype was scored based on the Athero Express Biobank classification [2]. Unstable plaques were selected based on relative necrotic core size, foam cell and inflammatory cell infiltration score, and the presence of neovascularization. Specimens were formalin fixed, embedded in paraffin, sectioned and stained as described below.

Animals
All animal experiments were performed in compliance with Dutch government guidelines and the Directive 2010/63/EU of the European Parliament. Male ApoE3*Leiden mice, crossbred in our own colony on a C57BL/6 background for at least 18 generations, 10-16 weeks old, were fed a diet (AB diets) containing 1% cholesterol and 0.05% cholate (VEGF experiment) or 0.5% cholate (time courses and DC101 experiment) from 3 weeks prior to surgery until sacrifice. The mice were housed on regular bedding and nesting material; water and diet were provided at libitum. Mice were randomized based on their plasma cholesterol levels (inclusion criteria of cholesterol level > 8 mol L À1 ; kit 1489437; Roche Diagnostics, Basel, Switzerland) and body weight. Mice were anesthetized with midazolam (5 mg kg À1 ; Roche Diagnostics), medetomidine (0.5 mg kg À1 ; Orion, Espoo, Finland) and fentanyl (0.05 mg kg À1 ; Janssen Pharmaceutical, Beerse, Belgium). After the surgery, the anaesthesia of the mice was antagonized with atipamezol (2.5 mg kg À1 , Orion) and fluminasenil (0.5 mg kg À1 ; Fresenius Kabi, Bad Homburg vor der Höhe, Germany). Buprenorphine (0.1 mg kg À1 ; MSD Animal Health, Keniworth, NJ, USA) was given after surgery to relieve pain.

Vein grafts
Vein graft surgery was performed by a donor caval vein interposition in the carotid artery of recipient mice as described before [23,24]. At sacrifice, patency of the vein grafts was visually checked for pulsations and blood flow, and occluded vein grafts were excluded from the study. Animals underwent 3 minutes of in vivo perfusion fixation with PBS and formalin under anaesthesia. Vein grafts were harvested, formalin fixed, dehydrated and paraffinembedded for histology.

In vivo detection of hypoxia
One hour prior to sacrifice mice (n = 6) received an intraperitoneal injection with the hypoxia marker pimonidazole hydrochloride (100 mg kg À1 ; hypoxyprobe Omni kit; Hypoxyprobe Inc., Burlington, MA, USA). Pimonidazole was detected with the polyclonal antibody (clone 2627) that is included in the kit.

Histological and immunohistochemical assessment of vein grafts
Cross sections were routinely stained with haematoxylin-phloxine-saffron (HPS) or Movat's pentachrome staining. Picrosirius red was used to detect collagen. The following antibodies were used for immunohistochemistry: endothelial cell CD31 Images of the human lesions were obtained with the Ultrafast Digital Pathology Slide Scanner and associated software (Phillips, Eindhoven, the Netherlands). Bright-field photographs were obtained with a Zeiss microscope and associated software. Fluorescent double and triple staining were acquired with the fluorescent slide scanner (3DHistech, Budapest, Hungary) and panoramic viewer software (3DHistech).

Morphometric analysis of vein grafts
Image analysis software (Qwin, Leica, Wetzlar, Germany) was used for morphometric analysis. For each mouse, eight (150 lm spaced) cross sections were used to determine lesion size and occurrence of intraplaque haemorrhage over a total vein graft length of 1050 lm. Since elastic laminas are nonexistent in these venous grafts, we analysed the putative vessel wall area (or lesion area) by measuring total vessel area (area within the adventitia) and the lumen area. The lesion area was calculated as total vessel area minus lumen area. Immuno-positive areas in vein grafts are expressed as total area or percentage of the lesion area.

Morphologic analyses of intraplaque haemorrhage
Intraplaque haemorrhage was analysed using CD31/Ly76 double-stained sections. Lesions where erythrocytes were found extravascular, adjacent to neovessels, were regarded as lesions with intraplaque haemorrhage. Using image analysis software (Qwin, Leica), the extravasated erythrocyte content was evaluated by measuring the total erythrocyte area in the lesion, followed by subtraction of the area of erythrocytes within the CD31stained neovessels.
VEGFR2 experiment: Total RNA was isolated from 10 (20-lm-thick) paraffin sections of vein grafts (n = 6/group). RNA was isolated according to manufacturers protocol (FFPE RNA isolation kit; Qiagen, Venlo, the Netherlands). RNA for q-PCR was reverse transcribed using a High Capacity RNA-to-cDNA kit (Applied Biosystems). Commercially available Taq-Man gene expression assays for the housekeeping gene hypoxanthine phosphoribosyltransferase (HPRT1) and selected genes were used (Applied Biosystems, Foster City, CA, USA); Vegfa (Mm 00437306_m1), Hif1-a (Mm 0468869_m1), Sdf-1 and Il6 (Mm00441242_m1)). q-PCR products were performed on the ABI 7500 Fast system (Applied Biosystems). The 2-DDCt method was used to analyse the relative changes in gene expression.

Aortic ring assay
Three separate experiments were conducted using three mice per experiment. C57BL/6 mice, age between 8 and 12 weeks, were anesthetized (as described above), and the aorta was dissected and stored in the medium. Each aorta was cut in 1-mm rings and serum-starved in Opti-MEM + Glutamax (Gibco, Gaithersburg, MD, USA) overnight at 37°C and 5% CO 2 . The next day, each ring was mounted in a well of a 96-well plate in 70 lL of 1.0 mg mL À1 acid-solubilized rat tail collagen I (Millipore, Burlington, MA, USA) in DMEM. After collagen polymerization (60 min at 37°C and 5% CO 2 ), Opti-MEM supplemented with 2.5% FCS and 30 ng mL À1 VEGF (R&D systems, Minneapolis, MI, USA) was added with or without DC101 or control antibodies (30 lg mL À1 ). The rings were cultured for 7 days, and pictures were taken (Zeiss, Oberkochen, Germany). The number of sprouts was counted manually.

Statistical analysis
Results are expressed as mean AE SEM. A twotailed Student's t-test was used to compare individual groups. Non-Gaussian distributed data were analysed using a Mann-Whitney U-test using GraphPad Prism version 6.00 for Windows (Graph-Pad Software, La Jolla, CA, USA). Probability values < 0.05 were regarded as significant.

Leaky neovessels in human vein graft and carotid lesions
Both vein graft specimens (Fig. 1, panel 1) and carotid atherosclerotic lesions (Fig. 1, panel 2) show features of classical atherosclerotic lesions with, foam cells, calcification and necrotic cores. Neovessels were found throughout the lesions in both vein grafts and carotid specimen, with a preference for the media and at inflammatory regions around necrotic cores, Fig. 1(b) panels 1 and 2. Frequently, these neovessels were leaky as demonstrated by the presence of erythrocytes (Glycophorin A-expressing cells) outside the neovessels, Fig. 1(c) panels 1 and 2. Both Ang-1 [ Fig. 1(d) panels 1 and 2] and Ang-2 [ Fig. 1(e) panel 1 and 2] were localized around the neovessels, although not all neovessels were found positive. Most neovessels, also in regions of intraplaque haemorrhage, did express VEGF, Fig. 1(f) panel 1 and 2. VEGFR2 staining was present around the neovessels but not as strong as VEGF expression, Fig. 1(g) panels 1 and 2.

Hypoxia drives plaque angiogenesis in vein grafts
In a time-course experiment of murine vein grafts, the expression of Hif1-a mRNA rapidly and significantly increased in vein grafts at all timepoints until day 28 (t28), when compared to native caval veins, with the highest level at t7, Fig. 2(a). Hif1-a protein expression was clearly visible at t28, Fig. 2(b). Sdf-1 mRNA was significantly up-regulated from t7 to t28 when compared to caval veins, Fig. 2(c). At the latter time-point, Sdf-1 protein expression could be detected especially in SMCs, Fig. 2(d). Interestingly, while we could not detect an increase in Vegf-a mRNA during the time course, Fig. 2(e), positive VEGF staining could be seen at t28, especially in plaque neovessels, Fig. 2(f). In vivo, we determined hypoxia by injecting the hypoxia probe pimonidazole (n = 6). Hypoxia was evident in all layers of the vein graft (t28), especially in macrophages scattered throughout the vein graft, Fig. 2(g,h).
A histological time course of vein grafts was used to study the timeframe in which the first plaque neovessels appear. From t14 (n = 4), the first plaque neovessels were detectable. These neovessels were primarily in the outer region of the vein grafts, suggesting sprouting from the vasa vasorum, Fig. 2(i). At t28, CD31+ plaque neovessels could be detected throughout all layers of the vein graft (n = 4), Fig. 2(j). The majority of these plaque neovessels have an activated endothelium, demonstrated by the expression of ICAM1 [ Fig. 2(k)] and VCAM1, Fig. 2(l). Up-regulation of ICAM1 and VCAM1 can lead to increased interactions with inflammatory cells. Therefore, 28-dayold vein grafts were stained with a combination of CD31 and CD163, an exclusive marker for neovessel-associated macrophages [15] and CD31 and CD3+T cells. CD163+ macrophages can be abundantly found throughout the vein graft lesion

Perivascular VEGF increases plaque neovessel density
To examine whether in the vein graft model we could target plaque angiogenesis, we applied a pluronic gel containing 250 ng VEGF in the perivascular region of the vein grafts, directly after surgery. Local treatment with VEGF did not affect cholesterol levels or bodyweight, Figure S1A,B. After 28 days, we observed an increase in the number of neovessels in the VEGF-treated group compared to controls, Fig. 3(a). Quantification of the plaque neovessel density per section revealed a significant 60% increase in neovessels in the VEGF group (P = 0.014), Fig. 3(b). However, local application of VEGF did not result in a significant effect on the vessel wall area (P = 0.628), Fig. 3(c). In both the control group and VEGF-treated group 1 out of 6 mice, intraplaque haemorrhage was observed.

Angiopoietin expression is augmented in intraplaque haemorrhage regions
Neovessels associated with intraplaque haemorrhage are characterized by reduced pericyte coverage. Mature neovessels are covered by SMC actinpositive pericytes, Fig. 4(a). In regions of intraplaque haemorrhage (characterized by perivascular erythrocytes), neovessels were partly devoid of pericyte coverage, Fig. 4(b). Tie2, the main receptor of the angiopoietins, was found to be specifically expressed by endothelial cells of plaque neovessels. The expression of Tie2 did not differ between mature neovessels [ Fig. 4(c)] or neovessels associated with intraplaque haemorrhage, Fig. 4(d). Increased staining of both Ang-1 and Ang-2 could be observed in areas of intraplaque haemorrhage. Ang-1 was predominantly expressed in intraplaque haemorrhage regions, whereas no staining around mature neovessels could be observed, Fig. 4(e). Ang-2 showed increased expression in lesions with intraplaque haemorrhage in contrast to regions of the lesions without intraplaque haemorrhage, Fig. 4(f).

VEGFR2-blocking antibodies inhibit intraplaque haemorrhage and erythrocyte extravasation
To interfere in the process of vessel integrity, we treated ApoE3*Leiden receiving a vein graft with the VEGFR2-blocking antibody (DC101). Treatment with DC101 did not change cholesterol levels or bodyweight in comparison with the control group, Figure S1C,D. Intraplaque haemorrhage was less frequently observed in mice treated with DC101 (7 out of 14 mice, 50%) in comparison with control animals (10 out of 12 mice, 83%). In the DC101 group, a smaller segment of the vein grafts (242 lm, 26% of the vein graft length) was affected by intraplaque haemorrhage in comparison with the control group (620 lm, 59% of the vein graft length, P = 0.037), Fig. 5(a). In addition, 80% less extravasated erythrocytes were observed in the DC101 group than in the control group (P = 0.049), Fig. 5(b). These extravasated erythrocytes were predominantly observed in the regions near the adventitia and in the mid-portion of the vein graft lesions, Fig. 5(b).

Neovessel density is not reduced by VEGFR2-blocking antibodies in vivo
The anti-angiogenic effect of suppression of VEGFsignalling in vivo was analysed by quantifying the neovessel density in the vein graft lesions. In the DC101 group, an average of 63 AE 25 neovessels per vein graft section was observed, whereas in the control IgG-treated group, 52 AE 19 neovessels per vein graft section were found (P = 0.327), Fig. 5(c). The vein graft model is characterized by the denudation of the luminal endothelium in the early days after engraftment, which is restored later in time [5]. In the both DC101 group and control group, the endothelium was completely restored at 28 days after surgery (P = 0.639), Fig. 5(d).

DC101 prevents vein graft thickening and results in a more stable lesion composition
VEGFR2 blockade resulted in a reduction in the lesion size compared to the control group, Fig. 6(a). Quantification of these lesions showed that the DC101-treated group had a significant reduction of 32% in vein graft thickening compared to the control IgG-treated group (P = 0.044), Fig. 6(a). A decrease in outward remodeling as measured by the total vessel area was detected in the DC101treated group (33%, P = 0.05), Figure S2A. The luminal area, however, was not significantly affected by DC101 treatment (P = 0.369), Figure S2B. Next, the effect of DC101 treatment on vein graft lesion composition was assessed. In the DC101 group, an increased collagen content was observed in comparison with the control group (46%, P = 0.066), Figure S2C. When corrected for the differences in vein graft thickening, the relative percentage of collagen was significantly increased in the DC101-treated group (54% P = 0.047), Fig. 6(b). In addition, a substantial increase in the SMCA area was observed (118% P = 0.003) in the DC101 group as well as a significant increase in the percentage of SMCA (123% P = 0.0005), Fig. 6(c) and Figure S2D. Plaque macrophages were significantly reduced after DC101 treatment with 30% (P = 0.001), Fig. 6(d), whereas the total macrophage area was reduced by 42% (P = 0.018), Figure S2E.

VEGFR2 blockade stimulates expression of genes associated with a more mature neovessel phenotype
To investigate the local inflammatory response, we measured the gene expression levels of pro-inflammatory genes Ccl2, Il6 and Icam1 in the vein grafts; no differences in expression levels could be detected between the groups, Fig. 7(a-c). Also, the expression of VEGF/VEGFR mRNA in the vein graft wall was analysed. Interestingly, the expression of both Vegfa [ Fig. 7(d)] and Vegfr1 [ Fig. 7(e)] was significantly reduced upon DC101 treatment [24% (P = 0.014) and 32% (P = 0.048), respectively], whereas the expression of Vegfr2 was not affected, Fig. 7(f). Furthermore, the angiopoietin receptor Tie2 [ Fig. 7(g)] was not differently expressed between the groups, nor was the vessel stabilizing factor Ang1, Fig. 7(h). The vessel destabilizing factor Ang2  was significantly decreased (P = 0.039) after DC101 treatment, Fig. 7(i). As a measure for proper endothelial function, we measured Connexin (Cx43, Cx37 and Cx40) expression. DC101 treatment showed no effect on Cx43 [ Fig. 7(j)] and Cx37 [ Fig. 7(k)] expression levels, but remarkably, significantly increased (P = 0.047) levels of Cx40 were observed pointing towards increased interendothelial cell connections, Fig. 7(l).

VEGFR2-blocking antibodies induce concentration-dependent vessel maturation
The effects of VEGFR2 blockade on vessel maturation were further studied in an aortic ring assay. Of the two concentrations DC101 (10 and 30 lg mL À1 ) tested, only the highest concentration resulted in a significant reduction (66% P = 0.003) in sprout formation when compared to no treatment, Fig. 8(a). The pericyte coverage of the sprout in the 30 lg mL À1 DC101 group was not significantly different than the control. Interestingly, the 10 lg mL À1 DC101 concentration induced a significant increase in SMCA+ pericyte coverage of the CD31+ sprouts (20%, P = 0.005), Fig. 8(b,c).

Discussion
Immature intraplaque neovessels have been characterized as the main contributors to intraplaque haemorrhage. Intraplaque haemorrhage occurs in native atherosclerosis but also in accelerated atherosclerosis after vein grafting or stenting [4]. However, most of the evidence is descriptive in nature [9,13]. In the present study, we used an intervention to show that VEGFR2 blockade reduces intraplaque haemorrhage and increases plaque stability by enhancing neovessels maturation in vein graft atherosclerosis.
We observed that neovessels in human carotid and vein graft specimen are associated with VEGF/ VEGFR2 and angiopoietins. In both types of atherosclerotic lesions, numerous regions with intraplaque haemorrhage and leaky vessels were observed. We demonstrated that plaque neovessels in the vein grafts originate primarily from the adventitia. This is also the general idea for native atherosclerotic lesion; however, luminal angiogenesis cannot be excluded [1]. VEGFR2 is involved in this process as the main receptor. VEGFR2 is involved in tip-cell-stalk-cell differentiation in the early phase of angiogenesis and mediates the permeability-enhancing effects of VEGF in adult endothelial cells as well as neovessel maturation [16]. We have previously shown that the majority of plaque neovessels in vein grafts express a basement membrane and that pericyte coverage is heterogeneous [23]. Here, we demonstrate that incomplete pericyte coverage of murine plaque neovessels is angiopoietin-related. Incomplete pericyte coverage in regions of intraplaque haemorrhage is also observed in human instable atherosclerotic plaques [11].
A modest induction in Vegf mRNA expression between 3 and 7 days but no further regulation between other time-points was observed. Interestingly, Hamdan et al. [27] showed comparable absent induction of Vegf mRNA in a canine vein graft model between native vein and 4 weeks after surgery, but did see a significant induction after 48 h. It seems that Vegf mRNA expression is only induced for a short period and is not the main driver of the remodelling response after vein graft surgery. This early induction of Vegf mRNA expression can be a result of the hypoxic period during surgery.
Atherosclerotic plaque angiogenesis can be manipulated as we show here by intervening in the VEGF pathway: locally applied VEGF-enhanced neovessel density. We found that low concentration of VEGFR2-blocking antibodies induced pericyte coverage in the aortic ring assay. This is comparable to the observation of increased pericyte coverage in murine and human tumours after VEGF signal blockade [28]. Blockade of VEGFR2 has been shown to facilitate the recruitment of pericytes to tumour vessels by enhancing Ang-1 expression and increasing perivascular matrix metalloproteases activity [29]. Ang-1 decreases endothelial cell permeability and increases vascular stabilization via enhancing endothelial cell interactions with the surrounding matrix and recruitment of pericytes to growing blood vessels. Ang-2 functions as a competitive Ang-1 antagonist in a VEGFdependent manner and mediates angiogenic sprouting and vascular regression [19]. This concurs with our finding that in regions of intraplaque haemorrhage, the expression of both Ang-1 and Ang-2 is increased. VEGFR2 blockade by DC101 treatment reduced intraplaque haemorrhage, reduced Ang-2 expression and improving gap junctions as shown by the increased Cx40 expression, pointing towards more mature neovessels. Post et al. [30] showed that in plaques with high neovessel density, the local balance between Ang-1 and Ang-2 is in favour of Ang-2. Unfortunately, vascular maturation and intraplaque haemorrhage were not studied in this context.
Recently, it was shown that treatment with axitinib (inhibitor of VEGFR1, 2 and 3) attenuated plaque angiogenesis [31]. Treating vein grafts with VEGFR2-blocking antibodies in vivo did not result in a reduction in neovessel density in comparison with control IgG-treated animals. Interestingly in a model for breast cancer, tumour vascular density was also not affected with this dose (10 mg kg À1 DC101) but was significantly decreased with a four times higher dose [32]. Furthermore, these authors observed that low-dose but not high-dose VEGFR2blocking antibodies treatment resulted in improved vascular maturation. In the aortic ring assay, we observed that the high-dose DC101 resulted in reduced sprouting, whereas the low dose did not reduce sprouting but did increased pericyte coverage.
VEGF is known to induce re-endothelialization and has been shown to inhibit intimal hyperplasia after vascular injury [33]. Application of VEGF directly after surgery in a rabbit vein graft model showed attenuation of the vessel wall size [34]. We show that local delivery of VEGF directly after surgery results in a nonsignificant trend towards reduction in intimal hyperplasia, whereas blockade of VEGFR2 resulted in significant attenuation of lesion growth. In the VEGFR2 blockade experiment, treatment with DC101 was started at day 14 after surgery to specifically study the effects on plaque neovessel formation which starts from this time-point on as demonstrated in Fig. 2(i). An important mechanism of action of VEGF is enhancing the re-endothelialization of the luminal endothelium which occurs primarily in the early period after surgery [35]. The late treatment with DC101 does not interfere with the re-endothelialization process. This was confirmed by the observation that at sacrifice (t = 28 days) both the control and DC101 group showed full luminal endothelial coverage. Vein graft lesion formation is largely driven by inflammation [36]. The positive effect of the VEGFR2 blockade on this process most likely overrules the VEGF-induced attenuation of lesion growth.
It has been shown that VEGFR2 activation can activate and degrade VE-cadherin resulting in vascular permeability [37]. Guo et al. [15] showed that CD163+ macrophages promote endothelial permeability via VEGF/VEGFR2 interaction with VE-cadherin. These CD163+ macrophages are clearly present, localized in areas of plaque neovascularization, in the murine vein grafts (Fig. 2). Blockade of VEGFR2 could reverse the VE-cadherin induced vascular permeability and induce the observed plaque neovessel maturation and reduced intraplaque haemorrhage.
Phagocytosis of intraplaque erythrocytes and erythrocyte-derived cholesterol by macrophages results in lipid core and plaque expansion, and promotion of plaque instability [12,38]. Systematic VEGFR2 blockade led to a reduction in intraplaque haemorrhage, lesion size and a reduction in lesion macrophages. Binding of VEGF to VEGFR2 can result in NF-jB-induced activation of VCAM-1 and ICAM-1 leading to increased adherence of leucocytes [39]. In various experimental models, inhibition of vascular leakage and NF-jB-dependent macrophage influx by DC101 was demonstrated [40,41]. Although at t28 no effect on inflammatory gene expression could be seen in the vein grafts, blockade of the binding of VEGF to VEGFR2 inhibited macrophage influx and subsequent effects on plaque composition including increased collagen and smooth muscle cell content. The NF-jB signalling cascade is an obvious route, since NF-jB-induced inflammation has been previously reported to be a critical pathway to stimulate macrophage influx and plaque instability in vein grafts [24,36,42].
In this study, we used a vein graft model in hypercholesterolaemic mice to study the role of plaque neovessel maturation. This model shows large atherosclerotic lesions with abundant plaque angiogenesis [23]. Vein graft atherosclerosis differs from native atherosclerosis since the onset (surgery) is acute with endothelial denudation and hypoxia resulting in the accelerated form. The lesions formed are concentric and highly dispersed with inflammatory cells and foam cells [36]. Local processes regarding plaque neovessel maturation in vein grafts show high similarities with native atherosclerosis as demonstrated in Fig. 1. The findings in this study can be, with cause, extrapolated to other cardiovascular diseases.
In summary, VEGFR2-blocking antibodies inhibit intraplaque haemorrhage and erythrocyte extravasation, resulting in more stable plaque neovascularization, decreased lesion development and increased plaque stabilization in a vein graft model, due to the maturation of the plaque neovessels. Our study indicates that vascular maturation (and more specifically VEGFR2) stands as an attractive target to stabilize atherosclerotic (vein graft) disease.

Author contributions
MdV, MJG and PQ designed the experiments and interpreted data. MdV, LP, EP, AS, LG and AF performed experiments and analysed data. JH, JWJ, CKO and RV provided intellectual contributions throughout the project. MdV and PQ wrote the manuscript and were responsible for the overall supervision of the manuscript. All authors discussed the results and commented on the manuscript.

Supporting Information
Additional supporting information may be found online in the Supporting Information section at the end of the article.