Introduction
The inflammatory and invasive rheumatoid arthritis (RA) synovial tissue is characterised by elevated levels of inflammatory T-helper cell 1 (Th1) cytokines such as IL-1β and TNFα (reviewed in [
1]), as well as by lowered oxygen tensions ranging between 2.4 and 4.4% oxygen (18 to 33 mmHg) compared with 8.5 to 13.5% (65 to 103 mmHg) in healthy individuals [
2]. Hypoxia in RA is thought to arise as a consequence of thickening of the synovial lining and infiltration by cells, predominantly circulating T cells, B cells and macrophages. This eventually leads to formation of a thick multilayered granulation tissue, termed pannus, which has propensity for invasion at the interface of cartilage and bone, resulting in progressive joint and soft tissue destruction [
3,
4]. Oxygen delivery becomes progressively compromised with increasing distances between the expanding tissue mass of pannus and existing synovial vasculature, resulting in tissue hypoxia. Inflammation and hypoxia support activation of local blood vessels and ongoing angiogenesis in the synovial membrane, which is an important early step in the pathogenesis of RA [
5]. Counterintuitively, despite attempts at restoring homeostasis through the process of angiogenesis, tissue hypoxia prevails due to the immaturity and dysfunctional nature of the newly formed vessels [
6].
There is considerable evidence to suggest that angiogenesis and chronic inflammation are co-dependent in inflammatory diseases such as RA (reviewed in [
7]). For instance, increased blood vessel formation, and hence an increased surface area of vessels, can maintain the chronic inflammatory state through increased production of cytokines and by allowing inflammatory cells to access the inflamed synovial tissue. Moreover, angiogenesis sustains the supply of nutrients and oxygen to the hyperproliferating inflamed RA tissue. Conversely, inflammatory mediators such as Th1 cytokines and growth factors secreted by the infiltrating inflammatory cells are known to have both direct and indirect angiogenic effects on endothelial cells and resident RA synovial cells, respectively [
8,
9]. These interdependent mechanisms underlying angiogenesis and inflammation in RA explain why biologics targeting inflammatory cytokines also reduce angiogenesis and why induction of angiogenesis is associated with increased synovial inflammation and pannus formation [
6,
10]. Likewise, angiogenesis blockade has been shown to reduce inflammation [
11].
Major transcription factors that may constitute a link between inflammation, hypoxia and angiogenesis are the heterodimeric hypoxia-inducible factor (HIF)-1 and HIF-2. These are best known as the principal mediators of cellular responses to hypoxia, such as angiogenic responses involving the well-known angiogenic factor vascular endothelial growth factor (VEGF) and a plethora of other angiogenic genes [
12]. Both HIF isoforms accumulate in circumstances of oxygen deprivation, which is associated with decreased degradation of the HIF-1α and HIF-2α subunits. This enables the transcription of HIF target genes, characterised by the presence of hypoxia-response elements (HRE). HIF-1α is ubiquitously expressed whereas HIF-2α is expressed in a more limited fashion. It is becoming clear that the two isoforms have overlapping as well as unique roles in hypoxia signalling, which are achieved through differential regulation and specific target gene selection. HIF-1α, HIF-2α and VEGF are all overexpressed in the synovial lining and stromal cells in rheumatoid synovia compared with normal synovia [
13]. Moreover, the number of HIF-1α-positive cells has been shown to correlate strongly with the number of blood vessels in RA synovial tissue and with inflammatory endothelial cell infiltration, cell proliferation and the synovitis score [
14].
In parallel to the oxygen-dependent pathway, HIF-1α and HIF-2α subunits are also regulated by inflammatory cytokines such as IL-1β and TNFα under normoxic conditions via receptor-mediated signals. In contrast to the specific stabilisation of HIF-α protein occurring under hypoxic conditions, Th1 cytokines appear to act on several regulatory levels and have been reported to stimulate HIF-α mRNA synthesis and stability in macrophages and RA fibroblast-like synoviocytes (FLS) [
15,
16], and to induce changes in HIF-1α levels and/or transcriptional activation in a number of cell types [
17,
18]. HIF-1α stabilisation by Th1 cytokines has been demonstrated to be partially mediated by the NFκB and p38 mitogen-activated protein kinase signalling pathways in human articular chondrocytes [
19], and by phosphatidylinositol 3-kinase/Akt and MEK1/2 inhibitor activation in RA FLS [
16,
20]. Importantly, Th1 cytokines have been demonstrated to synergise with hypoxia to induce HIF-1 protein and activity in HepG2 cells, a human hepatoma cell line [
18], and to synergise with hypoxia and the hypoxia mimetic CoCl
2 to induce HIF-1α mRNA and protein, respectively, in RA FLS [
16,
21]. Inflammatory cytokines and hypoxia were also shown able to act together to augment hypoxia-mediated upregulation of VEGF secretion in RA FLS [
22]. As the composition of the RA synovium includes elevated levels of inflammatory cytokines on a hypoxic background, a strong and continuous presence of HIF transcription factors is favoured through increased HIF-α mRNA levels and protein stabilisation. HIFs thus represent a key convergence point that integrates the cellular response of the RA synovium to low oxygen tension and inflammatory cytokines, and thereby drives synovial inflammation and angiogenesis.
Th1 and T-helper cell 2 (Th2) cytokines were recently demonstrated to have differential effects on HIF isoforms. For instance, in macrophages Th1 cytokines induce HIF-1α and Th2 cytokines induce HIF-2α mRNA, and this differential regulation of HIF-1α versus HIF-2α acts to respectively either increase or suppress nitric oxide synthesis and thus to control overall nitric oxide availability [
15]. In contrast to Th1 cytokines, anti-inflammatory Th2 cytokines are found at very low levels in RA joints and synovial fluid but they may hold an important therapeutic role in RA via HIFs. For instance IL-4, the signature cytokine of CD4
+ Th2 cells, is known to reduce the production of Th1 cytokines by RA synovium [
23] - and IL-4 has been used
in vivo as a treatment for a number of experimental autoimmune diseases in animals, including collagen-induced arthritis (CIA) [
24]. By suppressing Th1 cytokine levels, IL-4 may indirectly lower HIF activation and hence the degree of synovial angiogenesis. Despite these interesting aspects, the downstream effects of stimulating RA synovial cells with Th1 versus Th2 cytokines in a hypoxic environment have not yet been investigated with regard to HIF regulation and downstream angiogenic gene expression.
The use of antibodies to cytokines as pharmacological antagonists has revealed the profound effects of anti-TNFα treatment in reducing inflammation and joint destruction in RA. Despite the clear efficacy of anti-TNFα therapy, the actual mechanisms by which TNF-blocking agents are able to obtain these effects are still incompletely understood. In clinical trials using the TNFα inhibitor infliximab, reduced synovial angiogenesis and vascularity appears to be an effect associated with the neutralisation of TNFα [
10]. As HIFs regulate a plethora of downstream angiogenic factors including VEGF, in response to hypoxia and inflammatory cytokines, the efficacy of anti-TNFα therapy could, at least in part, be due to a reduction in the activation of HIF leading to decreased angiogenesis and less immature synovial vessels in RA patients.
In the present study, we chose to examine the impact of combined Th1 or Th2 cytokines and hypoxia on the angiogenic signature of RA FLS, focusing particularly on the role of HIF isoforms in the expression of downstream angiogenic genes. RA FLS was the chosen study target due to its recognised role in RA pathogenesis, where it contributes to bone and cartilage breakdown through the acquisition of what appears to be a transformed phenotype [
4,
25]. Moreover, as this cell type makes up the bulk of the expanding RA synovial membrane tissue [
26], it is likely to be exposed to various degrees of hypoxia and inflammatory cytokines simultaneously. We investigated the effect of Th2 cytokines on the angiogenic signature of RA FLS in normoxia and hypoxia in an attempt to elucidate the potential that Th2 cytokines could ameliorate autoimmune disease by influencing angiogenesis.
Materials and methods
Isolation and culture of cells from human RA tissue
Total RA synovial membrane cells and FLS were derived from synovial membranes of patients at the Royal Free Hospital (London, UK) who met the American College of Rheumatology 1987 criteria for RA [
27]. Full ethical approval was granted for the project (Local Ethics Research Committee EC2003-64). Preoperative informed consent was obtained in all cases.
The RA cell cultures were isolated and cultured as previously published [
28,
29]. The disaggregated total synovial membrane cell cultures were used directly in experiments. Alternatively, after overnight incubation, nonadherent cells were removed to allow overgrowth of FLS. The purity of the FLS culture was confirmed at passage 3 by immunohistochemistry with monoclonal anti-human-Fibroblast-Surface Protein 1 antibody (Abcam, Cambridge, UK) and determined to be >98%. FLS and normal human skin fibroblasts (HSF; Lonza, Walkersville, MD, USA) were cultured in DMEM containing 10% foetal bovine serum, 4.5 g/l glucose and
L-glutamine, supplemented with 100 U/ml penicillin and 100 μg/ml streptomycin (PAA Laboratories, Coelbe, Germany). RA FLS were used between the third and sixth passages.
Experimental setup and siRNA transfection
RA FLS cell cultures were starved in serum-free DMEM overnight (total synovial membrane cell cultures were used directly in complete medium) and subsequently subjected to a hypoxic gas mixture (1% oxygen, 5% CO2, 94% N2) for the desired length of time. Alternatively, 1 mM dimethyloxalylglycine (DMOG; Biomol International, Exeter, UK) or dimethyl sulfoxide (DMSO) (as vehicle) or cytokines IL-4, IL-1β, IFNγ (Peprotech, London, UK), TNFα (R&D Systems, Minneapolis, MN, USA) and IL-13 (Abcam) at 10 ng/ml were used to stimulate cells. Total RNA was isolated and the corresponding cDNA was used for quantitative PCR analysis. The cell supernatants were collected for ELISA and for testing in a functional angiogenesis assay.
Knockdown of genes was performed with siRNA against HIF-1α (5'-(AGCAGGUAGGAAUUGGAACAUU)RNA(tt)DNA-3') and/or HIF-2α (5'-(GCGACAGCUGGAGUAUGAAUU)RNA(tt)DNA-3') at a final concentration of 10 nM (MWG, Ebersberg, Germany) in Opti-MEM I Reduced Serum Medium (Invitrogen, Paisley, UK) using Lipofectamine 2000 (Invitrogen). An siRNA oligonucleotide against luciferase mRNA (siLuc) was used as a negative control, as well as oligonucleotides with the scrambled sequence of siHIF-1α or siHIF-2α (data not shown).
RNA isolation and quantitative PCR
RNA was isolated using the Total RNA E.Z.N.A™ EaZy Nucleic Acid Isolation kit (VWR, Batavia, IL, USA) and was DNase treated (Ambion Ltd, Paisley, UK). First-strand cDNA was synthesised using random primers (Invitrogen) and Moloney Murine Leukaemia Virus reverse transcriptase (Promega, Southampton, UK).
Diluted cDNA was added to SYBR
®Green I JumpStart™ Taq Ready MIX™ (Sigma-Aldrich, Poole, UK), primer mix and nuclease-free water. Exon-spanning PCR primers (MWG) were designed using Primer 3 (Table
1). All primers were validated prior to use. Primers for the housekeeping gene 18S ribosomal RNA and acidic ribosomal protein (data not shown) were used to normalise samples.
Table 1
Primers used in the study
ANGPTL-4 | 5´-CCACTTGGGACCAGGATCAC-3´ | 5´-CGGAAGTACTGGCCGTTGAG-3´ |
Leptin | 5´-GGCTTTGGCCCTATCTTTTC-3´ | 5´-GGAATGAAGTCCAAACCGGTG-3´ |
VEGF | 5´-CTTGCCTTGCTGCTCTACCT-3´ | 5´-CTGCATGGTGATGTTGGACT-3´ |
EFNA3 | 5´-CACTCTCCCCCAGTTCACCAT-3´ | 5´-CGCTGATGCTCTTCTCAAGCT-3´ |
HIF-α | 5´-CACCTCTGGACTTGCCTTTC-3´ | 5´-GGCTGCATCTCGAGACTTTT-3´ |
HIF-2α | 5´-CCTTCAAGACAAGGTCTGCA-3´ | 5´-TTCATCCGTTTCCACATCAA-3´ |
18S ribosomal RNA | 5´-GTAACCCGTTGAACCCCA-3´ | 5´-CCATCCAATCGGTAGTAGCG-3´ |
Quantitative PCR was performed with the following programme: pre-incubation at 50°C for 2 minutes, initial denaturation at 95°C for 5 minutes, 40 cycles of denaturation at 95°C for 10 seconds, annealing at 60°C for 30 seconds, elongation at 72°C for 30 seconds, and a 20-second final extension step. Rotor-Gene Software version 6.0 was used to analyse the data (Qiagen, Crawley, UK). The comparative cycle threshold (Ct) method (2
-ΔΔCt model) was used to calculate relative fold-changes in gene expression [
30].
PCR array
Human Angiogenesis RT2 Profiler™ PCR Arrays (TebuBio, Peterborough, UK) were used to screen cDNA from human RA FLS exposed to 21% or 1% oxygen and to DMSO or 1 mM DMOG. The PCR array was used according to manufacturers' protocol on an ABI 7700 sequence detector (Applied Biosystems, Foster City, CA, USA) using SYBR green technology (TebuBio). Every sample was run on duplicate PCR array plates and the relative quantification of mRNA was performed using the 2-ΔΔCt model and normalised to the average of two or more housekeeping genes on the PCR array.
Protein measurement by ELISA
Leptin and ANGPTL-4 released into the medium by RA FLS were measured using ELISA duosets (R&D Systems). VEGF was measured using reagents from Becton Dickinson (Oxford, UK) and antibodies from R&D Systems.
Western blotting
Total protein extracts were separated on NuPAGE Novex Tris-Acetate pre-cast gels 3 to 8% (Invitrogen) under reducing conditions and proteins were blotted onto polyvinylidene fluoride membranes (Perkin Elmer, Waltham, MA, USA). The membranes were blocked with 0.01% Tween 20, PBS and 5% nonfat milk for 1 hour at room temperature, washed and incubated for 2 hours with either anti-human HIF-1α mouse mAb at 1:250 (BD Transduction Laboratories, Oxford, UK) or anti-human α-tubulin mouse mAb at 1:10,000 (Sigma-Aldrich). The membranes were washed and incubated for 1 hour at room temperature with horseradish peroxidase-coupled rabbit anti-mouse IgG at 1:5,000 (Dakocytomation, Glostrup, Denmark), and developed using ECL plus and hyper-film ECL (GE Healthcare, Chalfont St Giles, UK).
Nuclear extraction and measurement of HIF-1 DNA binding activity
HIF-1 DNA binding activity was determined in RA FLS nuclear extracts. Nuclear extraction was performed according to the manufacturer's protocol (Active Motif, Carlsbad, CA, USA) and the protein concentration was determined using the BCA method. To measure HIF-1 DNA binding activity following stimulation of cells with low oxygen levels and/or cytokines, the nuclear extracts were tested using the ELISA-based TransAM HIF-1 Transcription Factor Assay (Active Motif). A secondary horseradish peroxidase-conjugated antibody and enzyme substrate included in the assay kit were used before HIF-1 binding to HRE was measured by absorbance at 450 nm. Each sample to be assayed for HIF-1 DNA binding was tested in duplicate. The specificity of HIF-1 binding to the HRE-coated wells was confirmed by competition experiments where either wild type or mutated HRE oligonucleotides (Active Motif) were added to the wells along with the nuclear extracts to be tested (data not shown).
This assay was performed in a 96-well plate with 50 µl Growth Factor Reduced Matrigel per well (VWR, Lutterworth, UK) that was left to gel for 45 minutes at 37°C. Human microvascular endothelial cell (HMEC)-1 (Center for Disease Control and Prevention, Atlanta, GA, USA) was cultured in RPMI containing 5% foetal bovine serum and supplemented with 100 U/ml penicillin and 100 μg/ml streptomycin (PAA Laboratories). The cells were used for matrigel assays when 80% confluent without prior starvation. Cells were dispersed with trypsin and 15,000 cells were loaded into each well in 100 µl full-growth medium. Then 100 µl conditioned medium from RA FLS stimulated with cytokines and/or subjected to hypoxia for 24 hours were added per well and the assay was incubated at 37°C for 4 to 6 hours until sufficient tube formation was visible with a microscope. Direct effects of recombinant TNFα and IL-4 were tested by adding 100 μl of 10 ng/ml cytokine in fibroblast medium per well in triplicate. The tubules were fixed with 4% para-formaldehyde (Sigma-Aldrich), washed in PBS and stained for 5 minutes with 50 µl Gram's Crystal Violet (Sigma-Aldrich), and the washing step was then repeated.
Images were captured with a camera (QICAM FAST; QImaging,Surrey, BC, Canada) attached to a microscope (CKX41 Olympus,Southend-on-Sea, UK) and the matrigel assay was scored using AngioSys Image Analysis software (TCS Cell Works, Buckingham, UK). This program provides quantitative measurement of tubule development in images captured from the 96-well plate, analysing the three parameters of total tubule number, number of tubule junctions and percentage of field area covered by tubules. Each experimental condition was tested in triplicate wells with a single picture captured from the centre of each well. The images were blinded prior to analysis.
Statistical analysis
Data were analysed using Prism software (Graph Pad Software, San Diego, CA, USA). P values were determined using a paired two-tailed t test assuming unequal variances or where indicated a one-way analysis of variance with Bonferroni's multiple comparison test.
Discussion
Angiogenesis in RA is an important hallmark of disease, and the degree of angiogenesis and immature blood vessel formation correlates with inflammation and disease activity. Interestingly, therapies targeting Th1 cytokines (for example, TNFα and IL-1β) have been shown to reduce angiogenesis in addition to decreasing inflammation, whereas therapies using Th2 cytokines ameliorate RA disease through less well described mechanisms [
6,
24]. The aim of the present work was therefore to elucidate how Th1 and Th2 cytokines affect angiogenesis in RA, alone and in combination with hypoxia, a major driving force for angiogenesis in RA. Such research can contribute to the understanding of how therapies directed against Th1 cytokines directly and indirectly affect angiogenesis, and also whether Th2 cytokines reduce disease severity through an effect on angiogenesis. Moreover, understanding how angiogenic factors are affected by inflammatory and anti-inflammatory cytokines against a background of synovial hypoxia may pave the way for future optimisation of existing RA therapies to better target the dual aspect of RA disease, namely angiogenesis and inflammation.
We specifically chose to focus on the differential effects of Th1 versus Th2 cytokines and hypoxia on HIFs, the master regulators of angiogenesis, and on the expression of downstream angiogenic genes by RA FLS. FLS are well-described contributors to inflammation and joint degradation in RA [
4,
25]. As hypoxia and proinflammatory cytokines typically co-exist within the RA synovium, we were particularly interested in examining the angiogenic effect of combined stimuli on RA FLS given that both types of stimuli are capable of inducing HIF expression [
16]. Although we demonstrated by PCR arrays that several other novel angiogenic genes were significantly altered in RA FLS exposed to hypoxia and the HIF activator DMOG, we focused on the genes EFNA3, leptin, ANGPTL-4 and VEGF as surrogate markers of HIF-induced genes and angiogenesis because they demonstrated consistent and maximal induction on human angiogenesis PCR arrays following hypoxia and DMOG stimulation. The first gene, EFNA3, is a component of the Eph/ephrin tyrosine kinase system and this receptor/ligand system is associated with various signalling pathways related to cell growth and viability, cytoskeletal organisation, cell migration, and apoptosis (reviewed in [
37]). In adult life, ephrin upregulation - particularly that of ephrin B - has been correlated to vascular invasion, blood vessel formation and sprouting by tumours, and soluble Eph A receptors have been shown to inhibit tumour angiogenesis [
38]. EFNA3 has not previously been associated with RA.
The adipokines, leptin and ANGPTL-4, are secreted mainly by the liver and adipocytes, and their original role as primary energy regulators in cell metabolism has progressively encompassed other functions including modulation of immune and inflammatory processes as well as angiogenesis [
39,
40]. Serum leptin levels have been reported to be elevated in RA patients and to correlate to accelerated atherosclerosis, thus potentially accounting for increased incidence of cardiovascular disease afflicting RA patients [
41,
42]. ANGPTL-4 has likewise been linked with arthritis because it was identified in a gene expression profiling analysis as one of the most highly expressed genes in early CIA, a widely used mouse model of RA. Moreover, ANGPTL-4 transcript has been reported to localise to stromal fibroblast-like cells adjacent to blood vessels in the mouse arthritic tissue, confirming that - like human RA FLS - murine RA FLS also secrete ANGPTL-4 [
43].
Although ANGPTL-4 and leptin have previously been reported to be upregulated by hypoxia in RA FLS, the study by Del Rey and colleagues did not investigate the involvement of HIFs or cytokines in adipokine expression by RA FLS [
31]. We demonstrated for the first time that these hypoxia-regulated genes were differentially regulated by HIF isoforms under hypoxic conditions, with leptin being primarily HIF-2 dependent, whereas ANGPTL-4 and VEGF were both HIF-1 and HIF-2 dependent, and EFNA3 was induced mainly by HIF-1. We have recently shown that these genes are similarly regulated in osteoarthritis and RA FLS by HIFs and prolyl hydroxylase-2, but not in normal HSF, suggesting that this is a pathological feature of synoviocytes from both diseases [
44]. Once we had established the relative contribution of HIF isoforms to the hypoxic induction of the adipokines and EFNA3 in RA FLS, we proceeded to investigate the regulation of HIFs by cytokines alone and in combination with hypoxia, as has been done before for HIF-1 and Th1 cytokines [
21] but not for Th2 cytokines and HIF-2. Th1 and Th2 cytokines were previously reported to have differential effects on HIF isoforms in macrophages, with Th1 cytokines inducing HIF-1 and Th2 cytokines inducing HIF-2 [
15], but a similar relationship has not been established in RA FLS. We demonstrated that HIF-1 was activated by Th1 cytokines, but not when RA FLS were stimulated with Th2 cytokines. In contrast, HIF-2 expression was unaffected by cytokine stimulation of either kind, thus highlighting a redundant role of HIF-2 in mediating cytokine-induced angiogenic responses in RA FLS.
Interestingly, unlike the induction by Th1 cytokines, we found that prolonged hypoxia reduced HIF-1α mRNA levels although HIF-1α protein was concomitantly upregulated. The downregulation of HIF-1α in hypoxia has been described before in a human lung epithelial cell line (A549) and is thought to be due to message destabilisation by a naturally occurring antisense to HIF-1α, aHIF [
29,
45]. aHIF is therefore possibly responsible for hypoxia-mediated downregulation of HIF-1α in RA FLS. HIF-2α mRNA was also downregulated by hypoxia in our study; however, as aHIF is complementary only to the 3'-UTR of HIF-1α but not to any part of HIF-2α, aHIF may not explain the hypoxia-mediated inhibition of HIF-2α observed in our study. We are currently investigating the role of aHIF in hypoxic and cytokine-stimulated RA FLS.
Our study thus confirms previous studies reporting that HIF-1 represents a convergence point for inflammatory and hypoxic signalling in RA FLS, since we found that both Th1 cytokines and hypoxia induced HIF-1 protein as previously reported and this effect was additive when cells were co-stimulated [
21]. We extended existing studies by showing additive effects of Th1 cytokines and hypoxia at the HIF-1 DNA binding activity level. Surprisingly, this cytokine-mediated increase in HIF-1 activity did not necessarily lead to induction of the four chosen downstream HIF-1 target genes in either normoxic or hypoxic RA FLS. In agreement with previous work [
18,
22], we found that VEGF was induced in a HIF-1-dependent manner by both of the Th1 cytokines tested in an additive (TNFα) and synergistic (IL-1β) fashion with hypoxia, as would be expected if both stimuli converge on HIF-1 with subsequent downstream target gene induction. Similarly, EFNA3 was induced by Th1 cytokines, albeit modestly, in a HIF-1-dependent manner, with IL-1β exerting an additive effect on EFNA3 expression in hypoxia. In contrast however, when TNFα were added to supernatants of RA FLS cultured under normoxic or hypoxic conditions, a strong negative effect on RA FLS-mediated expression of ANGPTL-4 was observed, despite the ability of Th1 cytokines to induce large amounts of active HIF-1 protein. Expression of the leptin gene, which was predominantly regulated by HIF-2, was also inhibited by TNFα in hypoxic RA FLS. These negative effects on adipokine expression were not HIF mediated, as we demonstrated with siRNA oligonucleotides against HIF isoforms. A complex picture therefore emerges from our study, in which angiogenic HIF target gene expression does not necessarily correlate positively with the level of HIF activity as is the case for VEGF in RA FLS.
The suppressive effects of combined hypoxia and TNFα on angiogenesis were further reflected in our findings that supernatants from RA FLS subjected to both stimuli did not induce tubule formation by HMEC-1 cells in a matrigel assay above controls. This observation could be due to the downregulation of potent mediators of synovial angiogenesis; for instance, adipokines. In contrast, supernatants from RA FLS stimulated with either TNFα or hypoxia as single stimuli induced angiogenesis above that of unstimulated cells, with hypoxia having the greatest effect on angiogenesis. These data suggest that the development of tubules is induced in areas of the RA synovial tissue where either inflammation or hypoxia dominates but is suppressed where inflammation and hypoxia co-exist, in spite of significant amounts of TNFα-induced HIF-1 and VEGF.
Our study confirmed that IL-4 stimulated RA FLS express VEGF in concordance with previous literature [
46], but also highlighted a novel finding that IL-4 strongly induced ANGPTL-4 expression in a HIF-independent manner under both normoxia and hypoxia. In contrast, we found that IL-4 could completely abrogate hypoxia-induced leptin expression by RA FLS. In agreement with a study where IL-4 was shown to be proangiogenic in murine lungs
in vivo under hypoxic conditions [
47], we found supernatants from RA FLS co-stimulated with IL-4 and hypoxia to have even stronger functional angiogenic activity than supernatants from cells stimulated with IL-4 alone. These supernatants contained much less VEGF protein than supernatants from TNFα and hypoxia-stimulated cells.
In contrast to VEGF, the levels of ANGPTL-4 present in the RA FLS supernatants correlated with the degree of tubule formation by HMEC-1. ANGPTL-4 has previously been shown to stimulate tubule formation in a human umbilical vein endothelial cell-based matrigel assay in the absence of additional growth factors [
43] and to induce anti-apoptotic activity in human vascular endothelial cells [
48]. Moreover, ANGPTL-4 can induce a proangiogenic response in the chicken chorio-allantoic membrane, an effect that was shown not to require the presence of VEGF [
49]. The presence of high levels of ANGPTL-4 might therefore contribute to the proangiogenic effects observed in the supernatants from IL-4-stimulated RA FLS. Similarly, the observed reduction in angiogenesis with hypoxia and TNFα-treated cell supernatants, when compared with effects of hypoxic supernatants, may be a consequence of adipokine downregulation by TNFα.
This is not the first time that negative regulation of angiogenic mediators by Th1 cytokines has been described in RA FLS. Recent work demonstrates a negative effect of IL-1β on matrix metalloproteinase-13 expression in hypoxic RA FLS [
50]. In contrast to the HIF-independent downregulation of adipokines that we observed in hypoxic RA FLS, the work by Lee and colleagues demonstrated the involvement of HIF-1 in hypoxia-mediated downregulation of matrix metalloproteinase-13. Our data suggest that other overriding signalling pathways are induced in RA FLS, which can circumvent the strong induction of HIF-1 following stimulation with combined Th1 cytokines and hypoxia. This could possibly involve PPAR-α and PPAR-γ, established regulators of ANGPTL-4 in adipose tissue [
51,
52]. In addition to peroxisome proliferator-activated receptor response elements in the promoter region of ANGPTL-4, transforming growth factor β has been shown to regulate ANGPTL-4 via an enhancer element located ∼8 kb upstream of the transcriptional start site involving SMAD3, ETS1, RUNX, and AP-1 transcription factors [
53,
54]. We are presently investigating the signalling routes by which TNFα exerts its negative effects on adipokines in RA FLS.
The ability of TNFα to inhibit ANGPTL-4 expression was specific for RA FLS because normal HSF stimulated with the same concentration of TNFα exhibited a strong induction of ANGPTL-4. In contrast, the ability by IL-4 to upregulate ANGPTL-4 was shared by both RA FLS and HSF. These observations suggest that the inhibitory effect of TNFα is mediated via a decline in ANGPTL-4 gene expression, and moreover that it may be a disease-specific effect. Decreased expression of ANGPTL-4 might contribute to the transformed phenotype that characterises RA FLS [
4,
25]. Although ANGPTL-4 was identified in a gene expression profiling analysis as one of the most highly expressed genes in early CIA (day 28), this expression subsided with time (day 49) [
43] - perhaps suggesting that RA FLS lose the ability to express ANGPTL-4 as disease progresses, with possible impact on blood vessel formation. The RA FLS we used are from patients with established RA of long duration.
The process of blood vessel maturation from immature vessels involves recruitment of perivascular cells, assembly of the basement membrane and establishment of tight and adherens junctions. Failure to form functional mature vessels is known to contribute to oedema formation, swelling and inflammation in RA joints (reviewed in [
55]). Interestingly, recent work in
angptl-4-deficient mice has shown that ANGPTL-4 is important in vessel maturation because mice lacking ANGPTL-4 exhibited disruption of endothelial adherens junctions and pericyte coverage, with impaired angiogenesis and vascular leakage as a result [
56,
57]. RA synovium contains a significant fraction of neoangiogenic, immature and leaky blood vessels that may be observed from the early stages of RA [
6]. The presence or density of immature vessels is significantly increased in patients with longer disease duration, higher activity and severity, and stronger inflammatory cell infiltration [
6]. Interestingly, immature vessels are depleted in response to anti-TNF therapy, highlighting the co-dependency of angiogenesis and inflammation [
6]. Based on our findings we speculate that in areas of the synovium where hypoxia and Th1-driven inflammation co-exist there would be an excessive, pathological HIF-1-mediated response with elevated VEGF production, yet suppression of functional angiogenesis with immature vessel formation, a well-known effect of elevated VEGF [
58]. If there is suppression of expression of angiogenic factors necessary for vessel maturation such as ANGPTL-4, the final outcome may thus be leaky and immature vessels. This is supported by a study showing that ANGPTL-4 decreases VEGF-induced vascular leakage in the Miles assay, which measures extravasation of Evans Blue dye from vessels in mouse back skin [
58]. In contrast, Th2 cytokine therapy might ameliorate inflammation in CIA in hypoxic areas of the synovium, through the formation of mature vessels by overriding the local anti-angiogenic effects of hypoxia combined with Th1 cytokines and by inducing factors such as ANGPTL-4 resulting in vessel maturation. Elucidating the difference in the angiogenic gene profile induced by Th1 and Th2 cytokines in hypoxia is thus of great importance, because such differences may account for the pathological ratio of mature to immature vessels in RA depending on the type of angiogenic factors they induce or inhibit. We are currently investigating the significance of TNFα-induced inhibition of adipokines in RA FLS-mediated angiogenesis, with specific interest in defining the function of ANGPTL-4.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
HL, EMP and MF designed the study. HL and EMP wrote the manuscript. BM contributed to the siRNA knockdown studies. TLK assisted with generating the ELISA data. All authors read and approved the final manuscript.