To examine the effect of hypoxia on Notch-1 signaling pathway components and angiogenesis in inflammatory arthritis.
To examine the effect of hypoxia on Notch-1 signaling pathway components and angiogenesis in inflammatory arthritis.
The expression and regulation of Notch-1, its ligand delta-like protein 4 (DLL-4) and downstream signaling components (hairy-related transcription factor 1 [HRT-1], HRT-2), and hypoxia-inducible factor 1α (HIF-1α) under normoxic and hypoxic conditions (1–3%) were assessed in synovial tissue specimens from patients with inflammatory arthritis and controls and in human dermal microvascular endothelial cells (HDMECs) by immunohistology, dual immunofluorescence staining (Notch-1/factor VIII), Western blotting, and real-time polymerase chain reaction. In vivo synovial tissue oxygen levels (tissue PO2) were measured under direct visualization at arthroscopy. HDMEC activation under hypoxic conditions in the presence of Notch-1 small interfering RNA (siRNA), the γ-secretase inhibitor DAPT, or dimethyloxalylglycine (DMOG) was assessed by Matrigel tube formation assay, migration assay, invasion assay, and matrix metalloproteinase 2 (MMP-2)/MMP-9 zymography.
Expression of Notch-1, its ligand DLL-4, and HRT-1 was demonstrated in synovial tissue, with the strongest expression localized to perivascular/vascular regions. Localization of Notch-1 to synovial endothelium was confirmed by dual immunofluorescence staining. Notch-1 intracellular domain (NICD) expression was significantly higher in synovial tissue from patients with tissue PO2 of <20 mm Hg (<3% O2) than in those with tissue PO2 of >20 mm Hg (>3% O2). Exposure of HDMECs to 3% hypoxia induced HIF-1α and NICD protein expression and DLL-4, HRT-1, and HRT-2 messenger RNA expression. DMOG directly induced NICD expression, while Notch-1 siRNA inhibited hypoxia-induced HIF-1α expression, suggesting that Notch-1/HIF-1α signaling is bidirectional. Finally, 3% hypoxia–induced angiogenesis, endothelial cell migration, endothelial cell invasion, and proMMP-2 and proMMP-9 activities were inhibited by Notch-1 siRNA and/or the γ-secretase inhibitor DAPT.
Our findings indicate that Notch-1 is expressed in synovial tissue and that increased NICD expression is associated with low in vivo tissue PO2. Furthermore, Notch-1/HIF-1α interactions mediate hypoxia-induced angiogenesis and invasion in inflammatory arthritis.
Inflammatory arthritis is a chronic, progressive disorder that is characterized by synovial tissue proliferation and joint inflammation, leading to degradation of articular cartilage and subchondral bone (1, 2). Angiogenesis, the formation of new capillaries from the pre-existing vasculature, is an early event in inflammation and is closely linked to the initiation and progression of inflammatory arthritis, as evidenced by the fact that the proliferation of vasculature can only be observed in inflamed joint tissue (3). Angiogenesis is dependent on complex and highly conserved processes of endothelial cell activation, migration, and survival and is critical for the formation of the invasive synovial pannus (3). Several studies have demonstrated increased vascularity and elevated levels of proangiogenic molecules in the inflamed joint, including vascular endothelial growth factor (VEGF), angiopoietins, platelet-derived growth factor (PDGF), transforming growth factor β1 (TGFβ1), and hypoxia-inducible factor 1 (HIF-1) (4, 5).
Notch receptor–ligand interaction is a highly conserved mechanism that regulates intercellular communication and directs individual cell fate decisions (6, 7). Notch receptors and ligands are transmembrane proteins; 4 Notch receptors (Notch-1 through Notch-4) and 5 ligands (Jagged 1, Jagged 2, Delta 1, Delta 3, and Delta 4) have been identified in mammalian cells. Studies using constitutively active Notch receptors missing their extracellular domains (Notch intracellular domain) have shown that Notch signaling determines proliferation, differentiation, and apoptosis in several mammalian cell types (6). Following cleavage by the γ-secretase, presenilin, Notch-1 intracellular domain (NICD) is translocated to the nucleus where it interacts with the CSL family of transcription factors (C-promoter binding factor 1/recombination signal sequence binding protein Jκ, suppressor of hairless, and Lag-1) to become a transcription activator that can modulate the expression of Notch target genes that regulate cell fate decisions (8, 9). These include HES and the HES-related transcription factor (HRT) genes (6, 8, 9).
Hypoxia is a key driving force in angiogenesis and is recognized as an important event in the perpetuation of joint destruction in inflammatory arthritis (10, 11). Previous studies have demonstrated that oxygen levels in the synovium of patients with inflammatory arthritis are reduced compared to those in healthy controls (10–12). Furthermore, low in vivo oxygen levels in the inflamed joint are inversely related to increased macroscopic vascularity, synovial inflammation, and oxidative damage (4, 13, 14), levels of which improve in response to tumor necrosis factor (TNF) inhibitors (15). In addition, studies have demonstrated HIF-1α expression in synovial tissue and, in synovial cell cultures, have shown that hypoxia induces key angiogenic growth factors (VEGF and angiopoietins), chemokines (monocyte chemotactic protein 1, interleukin-8 [IL-8], and CCL20), and matrix metalloproteinases (MMPs) but down-regulates IL-10 (4, 11, 12).
More recently, the interactions between the Notch signaling pathway and hypoxia have been described. Hypoxia-induced expression of Notch receptors and ligands has been demonstrated in murine endothelial progenitor and HeLa cell lines (16, 17). Hypoxia requires Notch signaling to inhibit myogenic cell line differentiation, and Notch signaling mediates hypoxia-induced tumor cell angiogenesis (18). Blockade of Notch signaling with γ-secretase inhibitors significantly decreases chondrocyte proliferation (19). Previous studies have also shown that tolerance to severe hypoxia is lost following small interfering RNA (siRNA)–transfected Notch knockdown in Drosophila melanogaster populations (20).
This study was undertaken to determine the effect of hypoxia on the Notch-1 signaling pathway components in rheumatoid arthritis (RA) synovium and human microvascular endothelial cells, including the effect of Notch-1 inhibition on HIF-1α activation and on hypoxia-induced angiogenesis in vitro.
Synovial tissue specimens from patients with inflammatory arthritis were obtained at arthroscopy under local anesthetic using a Wolf 2.7-mm needle arthroscope (Storz) as previously described (13). In vivo synovial tissue PO2 levels were measured using a Licox probe as previously described (14). All patients had clinically active disease that included at least 1 inflamed knee joint. Four patients undergoing interventional arthroscopy for ligament tears were also recruited. All research was performed in accordance with the Declaration of Helsinki and were approved by the St. Vincent's University Hospital medical research and ethics committee. Written informed consent was obtained from all patients. Biopsy specimens were either embedded in OCT compound (Tissue-Tek) or snap-frozen in liquid nitrogen for immunohistochemical and protein analysis.
Human dermal microvascular endothelial cells (HDMECs; Lonza) was grown in endothelial basal medium (EBM) supplemented with 5% fetal calf serum (FCS), 0.5 ml human epidermal growth factor, 0.5 ml hydrocortisone, 0.5 ml gentamicin, and 0.5 ml bovine brain extract (Lonza) and were used for experiments between passages 3 and 8. Previous studies have shown that HDMECs have responses that are similar to those of primary synovial endothelial cells (21). RA synovial fibroblasts (RASFs) were isolated from synovial biopsy specimens using 1 mg/ml type I collagenase (Worthington Biochemical) and were grown in RPMI 1640 (BRL Life Technologies) containing 10% FCS (BRL Life Technologies), penicillin (100 units/ml; Bio-Sciences), streptomycin (100 units/ml; Bio-Sciences), and Fungizone (0.25 μg/ml; Bio-Sciences). RASFs were used between passages 4 and 8. HDMECs and RASFs were cultured for 24 hours under normoxic conditions and 3% O2 (reflecting the joint environment in vivo). To determine whether HIF-1α activation directly induces NICD, HDMECs were cultured for 24 hours with 1 mM dimethyloxalylglycine (DMOG) (Sigma-Aldrich), and NICD was assessed by Western blotting.
To examine the expression of Notch-1, delta-like protein 4 (DLL-4), and HRT-1 and their localized distribution in the synovium, immunohistochemical analysis was performed in RA (n = 7) and normal (n = 4) synovial tissue samples. OCT compound sections (7 μm) were placed on glass slides coated with 2% 3-aminopropyltriethoxysilane (Sigma-Aldrich) and dried overnight at room temperature. Tissue sections were fixed in acetone for 10 minutes and air-dried. Nonspecific binding and endogenous peroxidase activity were blocked using 10% casein and 0.3% H2O2, respectively. A routine 3-stage immunoperoxidase labeling technique incorporating avidin–biotin–immunoperoxidase complex (Dako) was used. Separate sections were incubated with rabbit polyclonal anti–Notch-1 (Millipore), goat polyclonal anti–DLL-4 (Santa Cruz Biotechnology), and goat polyclonal anti–Hrt-1 (Santa Cruz Biotechnology) at room temperature for 1 hour. Sections were also incubated with an appropriate isotype-matched rabbit polyclonal or goat polyclonal antibody (Santa Cruz Biotechnology) as negative controls. Color was developed in solution containing diaminobenzidine tetrahydrochloride (Sigma-Aldrich), 0.5% H2O2 in phosphate buffered saline (PBS) buffer (pH 7.6). Slides were counterstained with hematoxylin (BDH Laboratories) and mounted. Images were captured using an Olympus DP50 light microscope and Analysis software (Soft Imaging System). Slides were analyzed using a well-established semiquantitative scoring method, and the perivascular/vascular, lining layer, and sublining regions were scored separately (15).
To examine colocalization of Notch-1 to synovial endothelial cells, dual immunofluorescence staining for factor VIII and Notch-1 was performed. Briefly, slides were coincubated with mouse monoclonal anti–factor VIII (Dako) and rabbit polyclonal anti–Notch-1 (Millipore) for 1 hour in a humidified chamber. Immunofluorescence sections were incubated with Cy2–Cy3–conjugated secondary antibodies (Jackson ImmunoResearch) for 30 minutes and counterstained with DAPI nuclear stain (1:1,000; Sigma-Aldrich) for 10 minutes. Sections were mounted with Antifade (Molecular Probes) and assessed by immunofluorescence microscopy. Fluorescent images were captured using an Olympus BX51 fluorescence microscope.
Synovial tissue biopsy specimens were obtained from a total of 13 patients (7 with RA and 6 with psoriatic arthritis [PsA]) in whom in vivo tissue PO2 levels were obtained at arthroscopy. Biopsy specimens were powdered using a Mikro-dismembrator U (B Braun Biotech); HDMECs and RASFs were trypsinized, and pellets were collected after centrifugation. The resulting powder of biopsy specimens or cell pellets was lysed, and protein was extracted in ice-cold radioimmunoprecipitation assay buffer (Sigma-Aldrich) containing 10 μg/ml phosphatase inhibitor cocktail and 10 μg/ml protease inhibitor cocktail (Sigma-Aldrich). Nuclear extraction of HDMECs was performed using a Nuclear Extract kit according to the recommendations of the manufacturer (Active Motif). The protein concentration was measured using a bicinchoninic acid assay (Pierce).
Proteins from synovial tissue lysates, HDMECs, and RASFs were resolved on sodium dodecyl sulfate–polyacrylamide gel electrophoresis (10% resolving, 5% stacking) prior to being transferred onto nitrocellulose membranes (Amersham Biosciences). Membranes were blocked for 2 hours at room temperature in wash buffer containing 5% nonfat dry milk with gentle agitation. Following 3 15-minute washes in wash buffer, membranes were incubated with rabbit polyclonal anti–Notch-1 (1:500; Millipore) or mouse monoclonal anti–HIF-1α (1:500; BD Biosciences) diluted in PBS containing 0.05% Tween 20 and 2.5% nonfat dry milk at 4°C overnight with gentle agitation. We used β-actin (1:5,000; Sigma-Aldrich) as a loading control. After 3 additional 15-minute washes, membranes were incubated in a 1:1,000 dilution of horseradish peroxidase–conjugated anti-rabbit IgG or anti-mouse IgG (Amersham Biosciences) for 2 hours. Following 3 final 15-minute washes, the ECL detection reagent (Amersham Biosciences) was placed on the membranes for 5 minutes before they were exposed to Hyperfilm ECL. The signal intensity of the appropriate bands on the autoradiogram was calculated using the Kodak EDAS 120 System.
Total RNA was isolated from cells using an RNeasy Mini kit (Qiagen). The purity of the RNA was measured by spectrophotometry. Samples with a ratio of absorbance at 260 nm and 280 nm of >1.8 were used, and total RNA (1 μg) was reverse transcribed to complementary DNA. Gene expression was analyzed by relative quantification with preoptimized conditions using LightCycler PCR technology (Roche Diagnostics). Specific primers for Notch-1 (Hs00413187_ml), HRT-1 (Hs01114113_ml), HRT-2 (Hs00232622_ml), and DLL-4 (Hs00184092_ml) were acquired from Applied Biosystems. Primers for 18S ribosomal RNA (Hs99999901_sl) were used as an endogenous control. All primers and sequences were obtained from Applied Biosystems.
For each 25-cm2 flask of HDMECs to be transfected, 5 μl of 20 pmole gene-specific siRNA duplexes (Notch-1 or scrambled) and 5 μl of Lipofectamine 2000 reagent (Invitrogen) was gently mixed with 0.99 ml serum/antibiotic-free Opti-MEM (Invitrogen) and incubated at room temperature for 20–30 minutes in the dark. The combination was mixed with full (5% FCS) EBM, added to the cells, and incubated overnight. The siRNA duplexes for Notch-1 (5459NM-017617) and scrambled control (a nonsense siRNA of the target sequence) were from Sigma. To examine the effect of Notch-1 siRNA on HIF-1α activation, 75-cm2 flasks of HDMECs were transfected for 24 hours and subsequently exposed to 3% hypoxia for 24 hours. Nuclear protein was extracted, and HIF-1α was assessed by Western blotting.
HDMECs were plated at 1 × 106 cells/well in 48-well plates. Notch-1 inhibition was achieved by incubating the cells with the γ-secretase inhibitor DAPT (Sigma-Aldrich) at 50 μM or 10 μM, as appropriate (22), or with DMSO vehicle control.
HDMECs that had been transiently transfected with Notch-1 siRNA or scrambled siRNA were plated in 96-well plates for 24 hours under normoxic or hypoxic conditions (3% O2). Supernatants were collected, and the expression of VEGF in the conditioned media was determined by ELISA according to the recommendations of the manufacturer (R&D Systems). The standard curve for the ELISA ranged from 31.25 pg/ml to 2,000 pg/ml, and the minimal detectable level of VEGF was 14.8 pg/ml. Absorbance was measured at 405 nm.
Matrigel Transwell invasion chambers (BD Biosciences) were used to investigate HDMEC migration under conditions of normoxia or 3% hypoxia. HDMECs (3 × 104) that had been transiently transfected with Notch-1 siRNA or scrambled control siRNA were added to the invasion chambers on membranes precoated with Matrigel containing EBM supplemented with 2.5% FCS, and EBM containing 5% FCS was placed in the lower wells. Cells were exposed to normoxic or hypoxic conditions (3% O2) for 24 hours. Noninvading cells were gently scrubbed off the upper surface. Migrated cells adhering to the lower membrane were fixed in 1% glutaraldehyde (VWR) and stained using 0.1% crystal violet solution (Pro-Lab). The average number of migrated cells was quantified by counting cells in 5 randomly selected high-power fields.
HDMEC tube formation was assessed using Matrigel basement membrane matrix (Becton Dickinson). Matrigel (50 μl) was added to 96-well culture plates and allowed to polymerize at 37°C for 1 hour before the cells were plated. HDMECs (1 × 104) that had been transiently transfected with Notch-1 siRNA or scrambled Notch-1 control siRNA were then plated in 250 μl EBM/well onto the surface of the Matrigel and incubated under conditions of normoxia or 3% hypoxia for 24 hours. HDMEC tubes were photographed using phase-contrast microscopy at 40× magnification. A connecting branch between 2 discrete endothelial cells was counted as 1 tube if it had a consistent intensity and thickness. Tube formation was assessed by 2 observers (WG and UF) in a blinded manner and was determined from 5 random fields per duplicate well, focusing on the surface of the Matrigel.
To perform scratch assays, HDMECs that had been transiently transfected with Notch-1 siRNA or scrambled control siRNA were plated in 24-well cell culture plates, and maintained in EBM for 4 hours prior to being serum starved with EBM (1% FCS) overnight. A scratch was made across the cell layer using a sterile pipette tip. After washing twice with serum-free medium, full EBM (5% FCS) was added and the cells were incubated under conditions of normoxia or 3% hypoxia for 8 hours. Cells were fixed and stained with 1% glutaraldehyde and 0.1% crystal violet solutions, respectively. Plates were photographed, and the extent of migration was assessed by 2 observers (WG and UF) in a blinded manner. All experiments were performed at least 3 times in triplicate.
The activity of proMMP-2 and proMMP-9 secreted by HDMECs into culture medium was determined by gelatin zymography. HDMECs (1 × 104) were seeded and grown to confluence in 48-well plates in full EBM. Cells were then incubated with 50 μM DAPT, 10 μM DAPT, or DMSO vehicle control for 24 hours. Following incubation, supernatants were harvested, and 10 μl of supernatants was loaded into appropriate gel lanes. Zymogram gels consisted of 7.5% polyacrylamide gels polymerized together with gelatin (1 mg/ml). Following electrophoresis, gels were washed with 2.5% Triton X-100 and incubated with substrate buffer (50 mM Tris, 5 mM CaCl2, pH 7.5) at 37°C for 24 hours. Gels were then stained with Coomassie brilliant blue R250 and destained with water. Bands were identified using gelatinase standards (Millipore).
The SPSS version 15 system for Windows was used for statistical analysis. The nonparametric Mann-Whitney U test or Wilcoxon's signed rank test was performed for the analysis of RA synovial tissue and synovial cell culture data, which were not normally distributed. Parametric Student's t-tests were performed for the analysis of HDMEC data, which were normally distributed. P values less than 0.05 were considered significant.
To examine the localized expression of Notch-1, DLL-4, and HRT-1, immunohistochemical analysis was performed in synovial tissue sections from patients with RA (n = 7) and healthy controls (n = 4). Notch-1 receptor, DLL-4, and the target gene HRT-1 were detected in the perivascular/vascular and synovial sublining regions (Figure 1). Figures 1a–d show Notch-1 expression in samples from 2 different RA patients at low and high magnifications. Notch-1 was strongly expressed around blood vessels but was also expressed in the lining layer (Figures 1c and d). Minimal expression of Notch-1 was observed in healthy control tissue (Figures 1i and j). DLL-4 and HRT-1 were also localized to the blood vessel (Figures 1e and f), sublining, and lining layer regions (Figures 1g and h).
Semiquantification demonstrated increased expression of Notch-1 in RA tissue versus healthy control tissue in the perivascular/vascular regions (mean ± SEM 2.75 ± 45 versus 0.25 ± 0.25), sublining (1.5 ± 0.53 versus 0 ± 0), and lining layer (1.25 ± 0.45 versus 0 ± 0). DLL-4 expression was increased in RA versus healthy control tissue in the perivascular/vascular regions (mean ± SEM 2.25 ± 0.45 versus 0 ± 0) and sublining (0.375 ± 0.183 versus 0 ± 0), as was expression of HRT-1 in the perivascular/vascular regions (mean ± SEM 2.4 ± 0.67 versus 0 ± 0) and sublining (0.6 ± 0.24 versus 0 ± 0). In contrast, DLL-4 and HRT-1 expression were similar in RA tissue and healthy control tissue in the lining layer (mean ± SEM 0.875 ± 0.22 versus 1.5 ± 0.866 and 1.6 ± 0.14 versus 1.75 ± 0.75, respectively). No expression of IgG control was observed. Figure 2 shows representative immunofluorescence images for the endothelial cell marker factor VIII, Notch-1, and colocalization of Notch-1 with factor VIII.
To examine whether NICD protein expression was associated with in vivo oxygen levels in the joint, protein was extracted from synovial tissue biopsy specimens obtained from patients with RA or PsA in whom in vivo synovial tissue PO2 levels were measured under direct visualization at arthroscopy. Expression was quantified in synovial tissue biopsy specimens (n = 13) categorized into 2 subgroups based on in vivo measurements of tissue PO2, as previously described (14). One group had PO2 <20 mm Hg, and the other group had PO2 >20 mm Hg. NICD expression was significantly higher in patients with tissue PO2 <20 mm Hg (<3% O2; n = 7) than in patients with tissue PO2 >20 mm Hg (>3% O2; n = 6) (P < 0.05) (Figure 3A), suggesting a clinically relevant association between Notch-1 and hypoxia in vivo. No difference was observed between tissue PO2 levels in RA patients and those in PsA patients.
Since Notch-1 was localized to the blood vessels and lining layer in RA synovial tissue (Figures 1a and c), we investigated whether hypoxia directly induced Notch-1 expression in primary RASFs and endothelial cells. In RASFs, 1% hypoxia (PO2 6.7 mm Hg) and 3% hypoxia (PO2 20 mm Hg) induced NICD protein expression, with 3% hypoxia inducing maximum expression (Figure 3B), reflecting the in vivo synovial tissue PO2 levels (10). In HDMECs, HIF-1α protein expression was undetectable under conditions of normoxia but was induced following exposure to 1% and 3% hypoxia (Figure 3C). Furthermore, NICD expression was induced under conditions of 1% and 3% hypoxia in HDMECs, with maximal expression observed at 3% oxygen. Finally, 3% hypoxia significantly increased the expression of mRNA for Notch-1, HRT-1, HRT-2, and DLL-4 compared to normoxia (Figure 3D). Thus, hypoxia is associated with synovial tissue NICD expression in vivo and induces NICD expression in vitro.
To investigate whether Notch-1 mediates hypoxia-induced angiogenesis in vitro, HDMECs were transiently transfected with Notch-1 siRNA or scrambled control siRNA for 24 hours, and NICD expression was examined by Western blotting. We found that NICD expression was significantly inhibited by Notch-1 siRNA as compared to scrambled control (Figure 4A). VEGF protein expression was significantly induced under conditions of 3% hypoxia as compared to normoxia (P < 0.01), an effect that was significantly inhibited by Notch-1 siRNA (P < 0.05) (Figure 4B). DMOG induced NICD expression, demonstrating that this effect is mediated through HIF-1α activation (Figure 4C). Furthermore, while HIF-1α activation was not detected under normoxic conditions, 3% hypoxia–induced HIF-1α was inhibited by Notch-1 siRNA (Figure 4D), with no effect observed for scrambled peptide, thereby providing further evidence of bidirectional interactions between Notch-1 and HIF-1α.
To further assess the potential role of Notch-1 in mediating hypoxia-induced endothelial cell function, HDMECs were exposed to 3% hypoxia in the presence of Notch-1 siRNA (or scrambled control), and angiogenesis, cell invasion, and migration were assessed. Compared to conditions of normoxia, 3% hypoxia significantly induced HDMEC invasion, as measured using Matrigel Transwell invasion chambers (P < 0.01) (Figure 5A) and endothelial cell tube formation, as measured using a Matrigel-based assay (P < 0.01) (Figure 5B); both effects were significantly blocked by Notch-1 siRNA (P < 0.05). Representative images of endothelial cell invasion and tube formation are available from the author upon request.
To assess the role of Notch signaling in hypoxia-induced endothelial cell migration, HDMEC monolayers transfected with scrambled or Notch-1 siRNA were scratched and exposed to normoxia or 3% hypoxia. Hypoxia-induced endothelial cell migration across the wound was inhibited in the presence of Notch-1 siRNA–transfected cells compared to hypoxia control (Figure 6A). Endothelial cell migration was further assessed by examining the expression of proMMP-2 and proMMP-9 in vitro (Figure 6B). Hypoxia induced proMMP-2 and proMMP-9 activities (Figure 6B, lane 5) as compared to normoxia (Figure 6B, lane 1). Under hypoxic conditions, DAPT clearly inhibited MMP-2 and MMP-9 expression (Figure 6B, lane 8) compared to hypoxic control (Figure 6B, lane 5), with no effect observed for DMSO vehicle control (Figure 6B, lane 6). While DAPT inhibited hypoxia-induced MMP-2 and MMP-9, there was no inhibitory effect on MMP-2 and MMP-9 under normoxic conditions, suggesting that the Notch signaling pathways may respond differently under normoxic control conditions versus hypoxic conditions.
In this study, we demonstrated that Notch signaling pathway components are expressed in inflamed synovial tissue and are localized predominantly to perivascular/vascular regions and, to a lesser extent, to the sublining and lining layer regions. Localization of Notch-1 to synovial endothelial cells was confirmed by dual immunofluorescence staining. NICD expression was higher in patients with synovial tissue PO2 <20 mm Hg (<3% O2) than in those with tissue PO2 >20 mm Hg (>3% O2). We demonstrated that hypoxia induces HIF-1α and Notch signaling pathway components, with maximal expression observed under conditions of 3% oxygen (reflecting the level in the joint in vivo). Furthermore, we showed that 3% hypoxia induces angiogenic tube formation, endothelial cell invasion, migration, and proMMP-2 and proMMP-9 activity, an effect that was inhibited by deletion of Notch-1 using siRNA or DAPT. Finally, we demonstrated that hypoxia-induced HIF-1α activation is inhibited by Notch-1 siRNA, suggesting bidirectional signaling interactions.
Angiogenesis is one of the primary events in the pathogenesis of inflammatory arthritis. Previous studies have demonstrated increased angiogenesis in the synovial membrane in inflammatory arthritis, which is associated with distinct vascular morphology (23–25) and elevated levels of proangiogenic molecules, such as VEGF, TGFβ1, MMP-9, PDGF-B, and angiopoietins (4, 5, 24). The majority of the blood vessels in the joint have recruited pericytes; however, a small but significant proportion of vessels lack pericytes (16), suggesting that synovial blood vessels are in a constant state of remodeling, undergoing simultaneous angiogenesis, pericyte recruitment, and vessel destabilization/stability (4). Most antiangiogenic therapeutic strategies are largely focused on inhibiting neovessel growth (26, 27); however, recent studies suggest that these strategies may actually fuel tumor invasiveness and metastasis (28, 29), and further suggest that therapies should target normalization of vessels (30).
Several studies have established that hypoxia is a key regulator of angiogenesis in vitro in many cell types (11–13, 17, 31). We have recently demonstrated in vivo that the inflamed synovial joint is hypoxic, with mean oxygen levels of ∼3%. Tissue PO2 levels inversely correlated with vascularity, synovitis, and oxidative damage (4, 13, 14), effects that were reversed in responders to TNF inhibitors (15). Consistent with the findings of those studies, it has been shown that hypoxia can induce angiogenic tube formation, invasion, migration, and matrix breakdown in vitro (12, 32, 33). HIF-1α is highly expressed in the inflamed synovium and mediates inflammatory processes in synovial cells (4, 13, 14).
The present study showed that increased NICD levels are associated with low tissue PO2 levels in vivo in the inflamed joint. We demonstrated that hypoxia induces NICD protein expression in RASFs and HDMECs in vitro, with maximal levels observed at 3% hypoxia (4). Also, we showed that induction of NICD was mediated through HIF-1α activation following stimulation with DMOG. In addition, 3% hypoxia increased DLL-4, HRT-1, and HRT-2 mRNA expression in HDMECs. These findings are consistent with those of studies showing that HIF-1α binding domains and a hypoxia response element are present in HRT-1, HRT-2, and DLL-4 promoters (17). Furthermore, we showed that Notch blockade inhibited hypoxia-induced endothelial cell function. Finally, we demonstrated for the first time that Notch-1 signaling can alter HIF-1α activation. We showed that Notch-1 siRNA inhibited hypoxia-induced HIF-1α expression, suggesting that hypoxia–HIF–Notch interactions are not unidirectional. Indeed, Gustafsson et al have reported a protein–protein interaction between HIF-1α and NICD, which results in increased stability under hypoxic conditions in P19 and COS-7 cells (16).
Moreover, factor-inhibiting HIF (FIH), a negative regulator of HIF, can hydroxylate the ankyrin repeat domain of Notch without affecting its function. Additionally, FIH displays a higher affinity for Notch, suggesting that Notch may monopolize FIH activity, thereby regulating HIF-1α expression (34, 35). A negative feedback loop between Notch and HIF-1α has been proposed, as the findings of one study suggest that HRT-1 and HRT-2 are capable of repressing HIF-1α gene induction (17). Previous studies have implicated STATs as potential mediators of HIF-1α/Notch signaling, specifically STAT-3 (36, 37). Taken together, these data indicate that Notch plays a potentially important role in hypoxia-mediated induction of angiogenesis.
One limitation of this study is that dermal microvascular cells were used as opposed to synovial microvascular cells, which are difficult to culture, particularly after the first passage. Microvessel function can vary between different endothelial cell beds, with pericyte–endothelial cell contact and vessel maturity/stability constantly changing during the course of the disease. However, studies have shown similar pericyte recruitment and increased expression of growth factors, adhesion molecules, MMPs, and HIF-1α in both the synovium and skin vascular regions, levels of which are altered by TNF inhibitors, suggesting a similar pathogenesis (24, 15, 38–42). Furthermore, previous studies using early passage synovial microvascular endothelial cells showed that these cells had responses similar to those of dermal microvascular endothelial cells (21).
The role of the Notch signaling pathway in embryonic vasculogenesis and angiogenesis is well established (43). Targeted disruption of components of this pathway can result in embryonic death due to defects in vasculogenesis, angiogenesis, and endothelial cell migration (16, 43, 44). Deletion of Notch-1 in mice, for example, resulted in defects in angiogenic vascular remodeling (45). Alterations in Notch-1 signaling produce abnormalities in vessel structure, branching, and patterning of the vasculature (46). The differentiation between vascular tip versus stalk cells is also dependent on interactions between ligand DLL-4 on endothelial tip cells and NICD on the neighboring stalk cell (47, 48). This process also involves induction of cytoskeletal pathways in the tip cell via NF-κB–dependent signaling (49, 50).
In summary, our data provide evidence of the importance of the Notch signaling pathway in hypoxia-induced angiogenesis in inflammatory arthritis. This study demonstrated synovial expression of Notch signaling pathway components and, for the first time, demonstrated an inverse relationship between low synovial tissue PO2 levels and NICD expression in vivo. We showed that Notch inhibition significantly attenuates hypoxia-induced angiogenesis and endothelial cell function in vitro. Finally, our results demonstrate that complex signaling interactions mediate these processes. Future studies are warranted to fully delineate this pathway, which may have important therapeutic implications.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Fearon had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Gao, Sweeney, Connolly, Kennedy, Ng, McCormick, Veale, Fearon.
Acquisition of data. Gao, Sweeney, Connolly, Kennedy, Ng, McCormick, Veale, Fearon.
Analysis and interpretation of data. Gao, Sweeney, Connolly, Kennedy, Ng, McCormick, Veale, Fearon.