Rheumatoid arthritis (RA) is an autoimmune disease that causes chronic inflammation of synovial joints, eventually resulting in the destruction of cartilage and bone [1]. During RA, the synovial tissue becomes infiltrated by inflammatory cells and increases greatly in mass due to the tumour-like proliferation of activated synoviocytes. In response to the growing metabolic demand caused by synoviocyte proliferation, blood vessels develop, which nourish and oxygenate this synovial pannus, promoting it to invade and degrade adjacent cartilage and bone. Neovessels exacerbate inflammation by further facilitating the ingress of inflammatory cells and mediators into the joint [2, 3]. Targeting the synovial vasculature has therefore been proposed as a possible therapeutic strategy in RA, especially since the approval of angiogenesis inhibitors for certain cancers.

In RA, a luxuriant vasculature is an early feature of the arthritic synovium , and the number of synovial blood vessels correlates with hyperplasia, mononuclear cell infiltration and indices of joint tenderness [4]. The vascular turnover in the arthritic synovium is increased, and synovial endothelial cells express markers of proliferation [5]. Although the hyperplasic RA synovium is highly vascularised, paradoxically the tissue environment is chronically hypoxic [6]. Synovial fluids from RA joints have been shown to promote endothelial cell migration and proliferation, and to induce vessel formation in an angiogenesis assay [7, 8], which reflects an active, pro-angiogenic phenotype of the arthritic synovium. Indeed, a number of angiogenic factors, expression of which is triggered by the hypoxic and inflammatory environment within the arthritic joint [9, 10], are abundant in RA synovial tissue, including vascular endothelial growth factor (VEGF) [8, 11], angiopoietins [10, 12], hepatocyte growth factor (HGF) [13] and fibroblast growth factor (FGF)-2 [14]. Although new vessel formation is a highly coordinated process, VEGF is generally agreed to be a crucial regulator of angiogenesis in RA [3]. Increased amounts of VEGF can be detected in the synovial tissue and fluid as well as in the circulation of RA patients [9, 11, 15]. Serum levels of VEGF correlate with markers of inflammation, disease activity and radiographic progression [15, 16]. During RA, VEGF seems to mediate its effects through its two tyrosine kinase receptors fms-like tyrosine kinase (FLT)-1 and kinase insert domain receptor (KDR; mouse homolog is fetal liver kinase (FLK)-1), and neuropilin (NRP)-1 [11, 17].

However, although the importance of angiogenesis in arthritis progression is well recognized, there is little information about the function of the synovial vasculature as well as the molecular mechanisms implicated in arthritis-associated angiogenesis. Furthermore, the concomitant presence of hypoxia and angiogenesis is a conundrum. The model of collagen-induced arthritis (CIA) resembles many pathological features of RA, and although, it does not perfectly duplicate the human disease, it has helped to validate TNFα as a therapeutic target for RA [18]. In the mouse model of CIA, extensive synovial neovascularisation is a prominent histological feature of arthritic joints [19–21], and disease onset is associated with a reduction in synovial oxygen tension [19]. Synovial tissue isolated from arthritic paws of CIA mice induced a strong angiogenic response in a vascular window model, which was in part mediated through TIE-2 receptor signalling [20]. VEGF and its receptors, FLT-1 and FLK-1, are expressed within the inflamed synovial tissue of CIA joints, and local VEGF levels correlate with the degree of neovascularisation and disease severity [21]. Further, blocking VEGF function by administration of a neutralising VEGF antibody [22] or a soluble VEGF receptor (sFLT-1) [23] to mice with established CIA, suppressed disease progression, whereas exogenous VEGF exacerbated synovial inflammation and joint destruction [24].

Based on these observations, we utilised the mouse CIA model to investigate the functional capillary density in the inflamed synovium and to analyse genes that are involved in synovial angiogenesis during arthritis. Gene expression profiling of murine CIA has been conducted before [25, 26]. However, these reports were based on a whole genome approach, and many angiogenesis-related genes, such as VEGF and its receptors, have not been assessed. Given that relatively little information is available on the kinetics of angiogenesis-relevant gene expression in CIA, the present study applied quantitative reverse transcription (RT)-PCR to investigate the expression of selected genes at onset, peak and declining phases of arthritis in DBA/1 mice. The hypoxic microenvironment in RA, coupled with the paradoxical feature of increased synovial vascularity, led us to speculate that the function of the synovial vasculature is disturbed and that abnormal angiogenic activity might contribute to the progression of CIA. We thus hypothesised that angiogenesis-related genes, including growth factors and receptors, would correlate with arthritis progression and angiogenic activity, particularly in the established phases of arthritis, allowing us to identify potential molecular targets for therapeutic intervention.

In RA and the relevant mouse CIA model, angiogenesis is considered to promote and exacerbate synovial inflammation and cartilage/bone erosion by facilitating the recruitment of inflammatory cells, and by providing oxygen and nutrients to the hyperplasic tissue [2, 3]. Although the importance of angiogenesis in the pathology of arthritis is well recognised, there is little information about the function of the blood vasculature. In the present study, we demonstrated for the first time that despite an increased vascular turnover, the functional capillary density is decreased in synovial tissue of arthritic joints during CIA. This reduction will further exacerbate the insufficient supply of nutrients and oxygen to the synovium, which could explain the observed hypoxic environment found in arthritic joints of mice [19]. Additionally, we observed an active pro-angiogenic microenvironment in arthritic paws, which was also reflected in an increased number of CD31-positive vessels and endothelial cells. These findings suggest that an imbalance of molecular factors regulating vessel formation and maturation contributes to abnormal synovial neovessel function.

To further elucidate the molecular mechanisms underlying this dysregulated synovial angiogenesis and to identify potential targets for therapeutic intervention, we characterised the expression of angiogenesis-related genes during the progression of murine CIA applying quantitative real-time RT-PCR. Our results show that the expression of hypoxia inducible factor 1alpha (HIF-1α), a key mediator of hypoxic responses, is elevated during the course of CIA in arthritic paws compared to healthy paws. In RA, HIF-1α induces the transcription of matrix metalloproteinases (MMPs) and pro-angiogenic growth factors, including VEGF, thus activating the angiogenic cascade as well as matrix degradation [35]. Not surprisingly, the most notable alterations occurred in the expression of genes involved in inflammation (for example, pro-inflammatory cytokines Il-1β, Il-6, Tnfα) and tissue remodelling (for example, Mmp-2, Mmp-9, Mmp-19). TNFα and IL-1β, as well as playing roles in the inflammatory component of RA, modulate synovial angiogenesis by inducing the production of angiogenic mediators like VEGF, ANG-1, ANG-2 and TIE-2 by RA synoviocytes [9, 10, 20]. MMPs participate in angiogenesis by degrading and remodelling the extracellular matrix and basement membranes, allowing activated endothelial cells to proliferate and migrate, as well as releasing extracellular matrix-bound growth factors such as FGF-2, VEGF or IGF-1 [36].

Our results further indicate that a hypoxic and pro-inflammatory microenvironment induces the transcriptional activation of angiogenic growth factors in arthritic joints of CIA mice, including midkine and Hgf. Elevated levels of midkine have been detected in the serum and synovial fluid of RA patients [37, 38], and HGF levels within the arthritic joint correlate with disease activity [39] and synovial microvessel density [13]. On the other hand, EGF, which has been reported as elevated in RA synovial fluids [40], was down-regulated during CIA. Further, we detected a significant decrease of leptin mRNA levels in arthritic paws, which is consistent with published gene expression data in murine CIA [41]. Leptin, apart from the regulation of food intake, has been implicated in the regulation of immune responses and angiogenesis [42].

Although the mRNA levels of VEGF isoforms and Plgf were only modestly (below 2-fold) altered during CIA, the VEGF signalling pathway was clearly affected, since the expression of its tyrosine kinase receptors Flt-1 and Flk-1 as well as its co-receptors Nrp-1 and Nrp-2 was significantly increased during CIA. This relatively small increase in Vegf expression was unexpected, especially in view of data showing high VEGF secretion by mouse CIA synovial membrane cells [23] or the effectiveness of anti-VEGF treatment in experimental arthritis [21–23]. However, the difference in VEGF mRNA levels [43] or protein [44] levels between non-arthritic and arthritic tissue was also reported to be not more than 1.5- to 1.6-fold. Taken together, these findings indicate that in the CIA model, VEGF expression does not change markedly, but seems nonetheless sufficient to induce pathological angiogenesis, and suggest that VEGF-mediated angiogenesis is largely modulated through an increase in surface expression of VEGF receptors on target cells, thereby increasing their responsiveness towards the angiogenic stimuli.

Another pathway involved in vessel formation and maturation is the ANG/TIE system. In a previous study we have shown that the expression of Ang-1, Ang-2, Tie-1 and Tie-2 mRNA is significantly higher in arthritic tissue of CIA mice, and that the application of a splice variant of TIE-1 ameliorates disease severity, reduces synovial vascularisation and bone destruction [33]. Our angiogenesis array identified Ang-1 as up-regulated by more than 2-fold, whereas Ang-2 was increased by only 1.6-fold on day 8 of CIA. Co-expression of angiopoietins and their receptors is observed in rheumatoid synovial tissue [10, 12]. ANG-1 signalling through TIE-2 maintains vessel integrity through the recruitment of pericytes, while ANG-2 blocks ANG-1/TIE-2 signalling, loosening the vascular structure and exposing the endothelium to inducers of angiogenesis such as VEGF [45]. In addition to increased transcription of angiogenic factors, the mRNA levels of some anti-angiogenic molecules (thrombospondins, endostatin) were also elevated, although their expression might not be sufficient enough to counteract the angiogenic response, since arthritic paw homogenates were able to induce endothelial cell chemotaxis in vitro as well as angiogenesis in vivo.

Our results reveal the existence of angiogenesis-related gene expression signatures in arthritic tissue and clearly indicate an active pro-angiogenic microenvironment in arthritic paws, which was also reflected in the increased number of CD31 positive vessels and endothelial cells. However, although the vascular turnover was increased within the inflamed synovium during CIA, the functional capillary density in arthritic synovial tissue was decreased. A previous IVM study revealed a trend towards a reduced functional capillary density in the synovium of pre-arthritic knee joints already before the onset of CIA [26]. Further, the diameter of the synovial capillaries in pre-arthritic mice was enlarged. These findings are in agreement with our observations in arthritic mice. It may seem counterintuitive that the arthritic synovium exhibits a lower functional microvessel density than healthy synovium, despite exhibiting a pro-angiogenic microenvironment and histologically an increased vascularisation. The human RA synovium contains regions of ongoing angiogenesis, indicated by endothelial cell proliferation, and regions of vascular regression, evident by markers of DNA damage and endothelial cell death [5]. However, angiogenesis not only depends on endothelial cell invasion and proliferation, but also requires coverage of vascular sprouts by peri-endothelial cells for vessel stabilization. Similar to tumour vasculature, a significant fraction of the synovial vasculature in arthritic joints of RA patients has not yet recruited peri-endothelial cells, which are essential for functional vascular patterning, diameter regulation, and vessel stabilisation [46, 47].The vascular network is thus dysfunctional in RA, and the presence of unstable vessels is associated with hypoxia, incomplete endothelial cell-pericyte interactions, increased DNA damage, disease progression and increased lymphocyte infiltration [46, 47]. Over-expression of angiogenic factors and receptors, as demonstrated in our present study, is likely to drive and maintain angiogenesis, as evidenced by the pro-angiogenic activities of arthritic paw extracts, but if uncontrolled, might actually promote further hypoxia and hence angiogenesis, while not allowing vessel maturation. For example, VEGF-mediated vessel formation has been associated with an imbalance between endothelial cell tube formation and the parallel development of pericytes [48]. Although the maturation status of vessels within the CIA synovium needs further investigation, vessel immaturity as well as an increased endothelial cell turnover could be an important factor contributing to the hypoxic environment found in the arthritic synovium of mice [19] and humans [6, 49], thereby aggravating synovial inflammation. Thus, despite apparent high vascularity, the arthritic synovium remains a rather inhospitable environment, with marked hypoxia and acidosis [6, 50], suggesting that blood flow is impaired and insufficient to meet the metabolic demand of inflamed synovial tissue. It might therefore be speculated that just like in cancer, vessel normalisation as a therapeutic intervention could improve the efficacy of immunotherapy in RA.

Using IVM, we detected a significant increase in the diameter of synovial capillaries in arthritic knee synovium. Vessel dilation probably indicates ongoing angiogenesis, since it is one of the first steps preceding vessel sprouting. Capillaries might be in a plastic state, responsive to pro-angiogenic stimuli (for example, VEGF), resulting in increased capillary diameter, remodelling and sprouting of new capillaries. This finding might also suggest that during CIA, the synovium tries to adapt to hypoxia by both angiogenesis and dilation of microvessels. However, increased permeability of capillaries allows the escape of plasma and plasma proteins in addition to the accumulation of leukocytes in the synovium, exacerbating joint inflammation.

The requirement of several angiogenic growth factors for synovial angiogenesis and inflammation during CIA has been shown previously, for example, by neutralising VEGF function [21–23], depletion of midkine [37], or by blocking angiopoietin signalling [33]. Our gene expression results suggest that VEGF-mediated angiogenesis and inflammation during CIA is largely modulated through increased expression of VEGF receptors and co-receptors. NRP-1 not only modulates the function of VEGF during angiogenesis by enhancing the binding of VEGF165 isoform to VEGF-R2 [51], but also interacts directly with VEGF121 [34] and other heparin-binding growth factors, such as HGF, VEGF/B or PLGF [52]. In the rheumatoid synovium, NRP-1 expression can be found on synoviocytes, infiltrating leukocytes and on the endothelium [17, 53] and concomitant expression of VEGF165, KDR, and NRP-1 is associated with a high vascular density and increased inflammation [17]. These observations, together with our findings showing increased Nrp-1 expression during CIA, suggested that blocking NRP-1 function may affect disease progression by directly reducing synovial angiogenesis and leukocyte infiltration. We applied a monoclonal antibody that targets the VEGF binding site on NRP-1 without interfering with its semaphorin signalling, and which has been shown to reduce neovascularisation in tumour models [28]. Our study revealed that anti-NRP1^B is able to inhibit endothelial cell migration induced by arthritic paw extracts, confirming that arthritic tissue contains factors that signal through NRP-1. Treatment with anti-NRP1^B significantly attenuated the progression of CIA as assessed by paw swelling and clinical score. Importantly, histological examination of arthritic hind paws demonstrated that this clinical improvement was associated with a reduction in synovial inflammation, pannus formation, as well as destruction of cartilage and bone, supporting the concept of a key role for this ligand-receptor interaction in the pathogenesis of arthritis. The role of NRP-1 in experimental arthritis has been studied before [54]. The prophylactic administration of an anti-NRP-1 peptide diminished disease severity in mouse CIA. Although prophylactic studies offer valuable insight into the molecular mechanisms underlying a disease, care must be taken in translating such results into potential therapeutic application. In contrast, our study employed a therapeutic approach, in which only mice with clinically evident arthritis received anti-NRP1^B. Our data indicate that NRP-1 signalling is required for the maintenance and progression of established inflammatory arthritis. However, further investigations are necessary to elucidate the exact mechanism by which anti-NRP1^B therapy improves inflammatory arthritis.

While a number of gene expression studies in CIA have been described, the present study was the first to focus on genes associated with angiogenesis, which is a key feature of RA. Our data are compatible with existing theories of angiogenesis in RA, and suggest that synovial angiogenesis results in the formation of dysfunctional vessels primarily caused by an imbalance of pro- and anti-angiogenic factors. Our results confirm NRP-1 as a key player in the pathogenesis of CIA, and support the VEGF/VEGF receptor pathway as a potential therapeutic target in RA.