Prolyl hydroxylase domain enzyme 2 is the major player in regulating hypoxic responses in rheumatoid arthritis


  • The work was completed at the Kennedy Institute of Rheumatology when it was part of Imperial College, London; the institute became part of the University of Oxford August 1, 2011.



Rheumatoid arthritis (RA) is characterized by hypoxia and the expression of hypoxia-inducible transcription factors (HIFs), which coordinate cellular responses to hypoxia. The objective of this study was to analyze the expression and regulation of prolyl hydroxylase domain (PHD) enzymes and factor-inhibiting HIF-1α (FIH-1), which regulate cellular HIF levels, and to study the roles of these enzymes in RA fibroblast-like synoviocytes (RA FLS).


The expression of PHD and FIH and downstream target genes was assessed by quantitative polymerase chain reaction and Western blotting. A small interfering RNA (siRNA) approach and an in vitro endothelial cell angiogenesis assay were used to analyze the roles of HIF hydroxylases.


In human RA FLS, knockdown of PHD-2, but not knockdown of PHD-1 or FIH-1, dramatically augmented HIF-1α expression, modestly increased HIF-2α protein expression under normoxic conditions, and up-regulated HIF-dependent gene expression. In contrast, silencing of PHD-3 up-regulated HIF-2α but reduced HIF-1α, thereby decreasing the expression of HIF-regulated genes. A similar effect of PHD-2 knockdown was observed in osteoarthritis FLS (OA FLS) but not in nondiseased primary human dermal fibroblasts. These findings correlated with the induction of in vitro angiogenesis by supernatants from RA FLS and OA FLS transfected with siPHD-2 but not by supernatants from nondiseased fibroblasts or from siPHD-3–transfected cells.


Our data suggest that PHD-2 is the major hydroxylase regulating HIF levels and the expression of angiogenic genes in arthritic cells. PHD-2 appears to regulate responses relevant to arthritis via HIF-α, highlighting the major importance of this enzyme in hypoxia- and angiogenesis-dependent inflammatory diseases such as RA.

Rheumatoid arthritis (RA) is associated with low oxygen tension (hypoxia) due to tissue expansion and increased cell metabolism. Fibroblast-like synoviocytes (FLS) represent the most abundant cell type in RA synovium and display altered phenotype characteristics, which are thought to underlie their key role in RA pathogenesis (1, 2). As a consequence of FLS hyperproliferation in the synovium coupled with an influx of inflammatory cells from the blood, the synovial oxygen level decreases from 7–10% O2 in normal synovium to ∼3% O2 in RA synovium (3). It is thought that synovial tissue hypoxia promotes angiogenesis in part via increased expression of proangiogenic factors such as vascular endothelial growth factor (VEGF). VEGF is induced by hypoxia in synovial fibroblasts (4, 5) and is up-regulated in RA serum (6). The newly formed vessels reach cells distant from the vasculature, supplying them with oxygen and nutrients. However, despite expression of proangiogenic factors such as VEGF, immature blood vessels have been observed in RA, which would further perpetuate hypoxia (7, 8).

The effects of hypoxia are mediated via hypoxia-inducible transcription factors (HIFs) to allow cell adaptation and survival. HIF is composed of oxygen-regulated HIF-α and constitutively expressed HIF-β subunits (9). At least 3 isoforms of HIF-α have been described, namely HIF-1α (10), HIF-2α (11, 12), and HIF-3α (13). HIF-α/HIF-β heterodimers are recognized by coactivators and bind to hypoxia-responsive element (HRE) in target genes initiating transcription (14). It is well established that HIF prolyl hydroxylase domain (PHD) enzymes and factor inhibiting HIF-1 (FIH-1), which are governed by O2, 2-oxoglutarate, iron, and ascorbic acid, are the main regulators of HIF-α posttranslational modification and represent the oxygen sensors upstream of HIF (15, 16). PHD enzymes affect HIF-α stabilization by making it recognizable by the von Hippel-Lindau tumor suppressor protein, leading to polyubiquitination and proteolytic destruction by the 26S proteasome (17, 18). Additionally, HIF-α protein is regulated by FIH-1, which blocks the binding of transcriptional coactivators p300/CBP to α-subunits through the hydroxylation of asparaginyl residues in HIF-α (19, 20). Thus, under conditions in which oxygen is limited and hydroxylases are less active or inactive, HIF-α accumulates and activates transcription of HRE-containing genes.

The objective of this study was to investigate the roles of PHD and FIH-1 in the hypoxia-mediated responses of RA FLS, because FLS are likely to be exposed to low oxygen concentrations in RA synovium. We observed that PHD-2 is the major enzyme controlling HIF stabilization and expression of HIF target genes in RA FLS and is implicated in the formation of endothelial cell tubule-like structures as part of the process of angiogenesis. Interestingly, PHD-2 also controlled angiogenic responses in FLS derived from patients with osteoarthritis (OA) but not in normal human dermal fibroblasts. In contrast, the effects of PHD-3 on HIF-1α and HIF target genes appeared to be opposite of the effects of PHD-2. These data suggest that the PHD-2/HIF axis is important in RA progression, and that its modulation could be beneficial in terms of specifically targeting angiogenesis in RA.


Cell culture.

Synovial tissue was obtained from patients with RA who were undergoing tenosynovectomy at Royal Free Hospital and who met the American College of Rheumatology/European League Against Rheumatism 2010 criteria for RA (21). Full ethics approval was granted for the project (Local Ethics Research Committee EC2003-64). FLS from 4 patients with OA who underwent hip replacement surgery were a gift from Dr. David Ahern (Local Ethics Research Committee 07/H0706/81). Preoperative informed consent was obtained from all participants.

The tissue was digested as previously described (3, 22). Cells were resuspended in Dulbecco's modified Eagle's medium containing 4.5 gm/liter glucose, 2 mM L-glutamine, 100 units/ml penicillin, 100 μg/ml streptomycin (Cambrex Bioscience), and 10% heat-inactivated fetal bovine serum (Biosera). The purity of the RA FLS culture was confirmed by immunohistochemistry using anti-human fibroblast surface protein 1 antibody (Abcam). Normal human dermal fibroblasts (Lonza) were cultured as described above.

Transfection with small interfering RNA (siRNA).

Knockdown of genes was performed with siRNA at a concentration of 10 nM, unless stated otherwise. The sense sequences of siRNA were as follows: for siRNA targeting PHD-1 (siPHD-1), 5′-(GGACAGAAAGGUGUCCAAGTT)RNA(TT)DNA; for siPHD-2, 5′-(GAAGCUGGGCAGCUACAAAAT)RNA(TT)DNA; for siPHD-3, 5′-(GGAGAGGUCUAAGGCAAUGTT)RNA(TT)DNA; for siFIH-1, 5′-(CAGUUGCGCAGUUAUAGCUUC)RNA(TT)DNA (MWG Biotech); for siLuc, 5′-(CGUACGCGGAAUACUUCGA)RNA(TT)DNA (Dharmacon). Comparable knockdown was obtained using alternative siRNA (data not shown). Transfection was performed using Lipofectamine and Opti-MEM I Reduced Serum Medium (Invitrogen). Cells were subsequently cultured for 24 hours under normoxic (21% oxygen) or hypoxic (1% oxygen) conditions (Galaxy R incubator; Wolf Laboratories). Alternatively, 1 mM dimethyloxaloylglycine (DMOG; Biomol International) was used.

Real-time quantitative polymerase chain reaction (qPCR).

RNA was isolated using E.Z.N.A. EaZy, and complementary DNA (cDNA) was synthesized using 500 ng of RNA, random primers (Invitrogen), and Moloney murine leukemia virus (Promega). Quantitative PCR was performed using SYBR Green I JumpStart (Sigma-Aldrich). The primer sequences were as follows: for 18S ribosomal RNA (18S rRNA), sense 5′-GTAACCCGTTGAACCCCA and antisense 5′-CCATCCAATCGGTAGTAGCG; for PHD-1, sense 5′-CAGCTAGCATCAGGACAGAAA and antisense 5′- CAGAGGCAAAGTCAGAAGCA; for PHD-2, sense 5′-GCACGACACCGGGAAGTT and antisense 5′-CCAGCTTCCCGTTACAGT; for PHD-3, sense 5′-TTGGGATGCCAAGCTACA and antisense 5′-CGTGTGGGTTCCTACGATCT; for FIH-1, sense 5′-CAATGTACTGGTGGCATCACATAG and antisense 5′-GGCCACTTTCTGATGAGCTT; for VEGF, sense 5′-CTTGCCTTGCTGCTCTACCT and antisense 5′-CTGCATGGTGATGTTGGACT; for angiopoietin-like protein family 4 (ANGPTL-4), sense 5′-CCACTTGGGACCAGGATCAC and antisense 5′-CGGAAGTACTGGCCGTTGAG; for glucose transporter 1 (GLUT-1), sense 5′-TGGCATGGCGGGTTGT and antisense 5′-CCAGGGTAGCTGCTCCAGC; for carbonic anhydrase IX (CAIX), sense 5′-GCTGTCACCAGCGTCGCGT and antisense 5′- CCAGTCTCGGCTACCTCTGCT; for Bcl-2/adenovirus E1B 19-kd protein–interacting protein 3 (BNIP-3), sense 5′- GATATGGGATTGGTCAAGTCGG and antisense 5′-CGCTCGTGTTCCTCATGCT; for ephrin A3, sense 5′-CACTCTCCCCCAGTTCACCAT and antisense 5′-CGCTGATGCTCTTCTCAAGCT (MWG Biotech). The 2math image method was used to calculate relative gene expression normalized to the expression of 18S rRNA.

Western blot analysis.

Total cellular protein extracts were prepared using denaturating lysis buffer (8M urea, 1% sodium dodecyl sulfate, 1% glycerol, 10 mM Tris [pH 6.8], 0.5 mM protease inhibitor cocktail, 1 mM dithiothreitol; Sigma-Aldrich). Protein extracts were separated on NuPAGE Tris-Acetate gels (4–12%; Invitrogen) and blotted onto PVDF membranes. The membranes were incubated with mouse anti–HIF-1α monoclonal antibody (Becton Dickinson), anti–endothelial PAS domain–containing protein 1 (anti–EPAS-1)/HIF-2α monoclonal antibody (Santa Cruz Biotechnology), anti–PHD-3 (a kind gift from Prof. Christopher Pugh, Nuffield Department of Clinical Medicine, Oxford, UK), or anti–α-tubulin (Sigma-Aldrich), or with polyclonal rabbit anti–GLUT-1 (Alpha Diagnostic), anti–PHD-1, anti–PHD-2, or anti–FIH-1 (Abcam). Following incubation with polyclonal rabbit anti-mouse or swine anti-rabbit immunoglobulin coupled with horseradish peroxidase (Dako), proteins were visualized using ECL Plus (GE Healthcare).

DNA binding assay.

A TransAM kit (Active Motif) was used to isolate nuclear extracts and quantify HIF binding to HRE. Nuclear extracts (10 μg) were added to the wells containing immobilized HRE oligonucleotides and were detected using mouse anti–HIF-1α (Active Motif) or anti–EPAS-1/HIF-2α monoclonal antibody (Santa Cruz Biotechnology).

PCR array.

Human Angiogenesis RT2 ProfilerPCR Arrays (SABiosciences) were used to screen cDNA from cells transfected with siRNA. The array was performed according to manufacturer's protocol, using an ABI 7700HT Sequence Detection System (Applied Biosystems). The data were expressed as scatter plots of 2math image values of siLuc versus siPHD-2. Data were analyzed using the 2math image model normalized to the expression of housekeeping genes (B2M, HPRT1, RPL13A, ACTB). Statistical analyses were performed by comparing ΔCt values.

Angiogenesis assay.

A commercially available AngioKit assay (TCS CellWorks) was used to measure tubule formation in vitro. Wells with pre-seeded human endothelial cells were treated with or without human recombinant VEGF (5 ng/ml) or with supernatants from siRNA-transfected cell cultures on days 1, 4, 7, and 9. On day 11, the expression of CD31 (platelet endothelial cell adhesion molecule 1) was visualized by staining with mouse anti-human CD31 antibody followed by goat anti-mouse IgG alkaline phosphate and nitrophenol phosphate, and absorbance was measured. Subsequently, to visualize the tubules, insoluble substrate BCIP/nitroblue tetrazolium was added. Each condition was examined in triplicate, and 3 images were obtained from each well. The results of the angiogenesis assay were scored using AngioSys Image Analysis Software, which provides quantitative measurement of tubule development by analyzing total tubule number, total tubule length, mean tubule length, number of junctions, and the percentage of field area covered by CD31-positive tubules.

Statistical analysis.

Data were analyzed using Student's paired t-test or one-way analysis of variance with Bonferroni adjustment for multiple comparisons, as indicated.


Expression of HIF hydroxylases in RA FLS.

To study regulation of the HIF pathway, the expression of the upstream HIF hydroxylase enzymes PHD-1, PHD-2, PHD-3, and FIH-1 was measured in RA FLS. Quantitative PCR analysis of the Ct values was performed to study the abundance of PHD and FIH1 at a constant Ct value of 0.05 and showed that under normoxic conditions, the mean ± SEM Ct values for PHD1, PHD2, PHD3, and FIH1 were 21.6 ± 0.2, 19.3 ± 0.2, 25.9 ± 0.3, and 20.2 ± 0.3, respectively (where higher Ct value indicates less abundance of RNA/cDNA when the amplification efficiencies of the PCR primers are equal). Thus, PHD2 was the most abundant of the PHD isoenzymes, and messenger RNA (mRNA) levels further increased ∼2-fold under hypoxic conditions and ∼7-fold in cells treated with the hypoxia mimetic DMOG (Figure 1A).

Figure 1.

Differential effects of prolyl hydroxylase domain 2 (PHD-2) and PHD-3 knockdown on hypoxia-inducible transcription factors (HIFs) in rheumatoid arthritis fibroblast-like synoviocytes (RA FLS). A, RA FLS were exposed to normoxia (21% oxygen), hypoxia (1% oxygen), or 1 mM dimethyloxaloylglycine (DMOG) for 24 hours. Messenger RNA levels of PHD genes and factor inhibiting HIF-1 (FIH1) were assayed by quantitative polymerase chain reaction. Expression was normalized to the expression of cells cultured under normoxic conditions. Data were analyzed by one-way analysis of variance (ANOVA). Values are the mean ± SEM (n = 13 individual experiments). ∗∗ = P < 0.01; ∗∗∗ = P < 0.001 versus normoxia. B, PHD-1, PHD-2, PHD-3, and FIH-1 protein levels were evaluated by Western blotting, using α-tubulin as a loading control (arrow indicates the predicted size of PHD-1). C and D, RA FLS were transfected using small interfering RNA (siRNA) targeting PHD-1, PHD-2, PHD-3, FIH-1, and luciferase (Luc) and cultured under normoxic conditions for 24 hours. C, Protein levels of PHD-2, HIF-1α, HIF-2α, and α-tubulin were assessed by Western blotting. D, HIF-1 and HIF-2 binding to hypoxia-responsive element (HRE) was measured after PHD-2 and PHD-3 knockdown. Bars show the mean ± SEM of 3 individual experiments. The broken line shows HIF binding to HRE under normoxic mock control conditions. Responses induced by 1 mM DMOG and hypoxia for 24 hours are shown for comparison. Data were analyzed by one-way ANOVA. ∗ = P < 0.05; ∗∗ = P < 0.01; ∗∗∗ = P < 0.001 versus siLuc.

This up-regulation in PHD2 mRNA was mirrored at the protein level (Figure 1B). PHD3 was least abundant when compared with other hydroxylases but was nonetheless extremely sensitive to hypoxia, with mRNA levels increasing 19-fold under hypoxic conditions and 341-fold after treatment with DMOG (Figure 1A). However, PHD-3 protein was detectable only in DMOG-treated cells (Figure 1B). In contrast, PHD-1 and FIH-1 were not affected by hypoxia or DMOG at either the mRNA level or the protein level (Figures 1A and B). Although our studies were performed under conditions of 1% O2, we also assessed the effect of 3–10% O2 and observed significant up-regulation of PHD2 and PHD3 mRNA under conditions of 3% O2 and, to a lesser extent, under conditions of 5% O2. This corresponds to stabilization of HIF-1 and HIF-2 (data not shown).

Differential stabilization of HIF-α isoforms by silencing of PHD-2 and PHD-3.

To examine the function of individual hydroxylases in the HIF pathway in RA FLS, PHD and FIH-1 were silenced using siRNA. The specificity of knockdown was confirmed at the mRNA level, with reductions of individual PHD and FIH1 genes under normoxic conditions of 56% (for PHD1), 65% (for PHD2), 71% (for PHD3), and 85% (for FIH1) (data not shown). In terms of the effects of PHD/FIH knockdown on other HIF hydroxylases, the only effect observed was increased expression (∼4-fold) of PHD-3 mRNA following treatment with siPHD-2 (data not shown). The specificity of PHD-2 knockdown was confirmed at the protein level (Figure 1C).

Under normoxic conditions, selective knockdown of PHD-2 stabilized both HIF-α proteins (Figure 1C) and significantly up-regulated HRE binding of HIF-1 and HIF-2, to levels comparable with those observed under hypoxic conditions (Figure 1D). Silencing of PHD-3 significantly increased HIF-2 binding to HRE, by 2-fold (Figure 1D). This effect was relatively difficult to detect by Western blotting due to the scarcity of both PHD-3 and HIF-2α under normoxic conditions. PHD-3 knockdown either did not increase or only modestly reduced HIF-1α protein levels (Figure 1C) and HIF-1 binding to HRE (Figure 1D). Furthermore, enzyme-linked immunosorbent assay revealed a decrease in HIF-1α protein from 1,412 pg/ml to 516 pg/ml after PHD-3 knockdown, whereas PHD-2 knockdown increased the level of HIF-1α protein to 2,047 pg/ml (data not shown).

HIF target gene induction by siPHD-2.

In RA FLS, qPCR analysis revealed increased expression of ANGPTL4, BNIP3, CAIX, EFNA3, GLUT1, and VEGF following selective knockdown of PHD-2, to levels resembling those observed under hypoxic conditions (Figure 2A). HIF1A and HIF2A mRNA levels were not affected (results not shown), further confirming that PHD-2 regulates the levels of HIF protein rather than mRNA. In addition, single knockdown of PHD-1 or FIH-1 did not affect the expression of GLUT-1 (Figure 2C) or the expression of other HIF-dependent genes such as VEGF and ANGPTL4 (results not shown). Double knockdown (siPHD-2 plus either siPHD-1 or siPHD-3) or even triple knockdown did not further affect HIF-α stability when the responses were compared with the effect of siPHD-2 alone (Figure 2B) and had no additional effect on GLUT1 mRNA (Figure 2C) and protein (Figure 2D) or on mRNA levels of ANGPTL4, CAIX, BNIP3, EFNA3, and VEGF (results not shown).

Figure 2.

PHD-2 knockdown increases HIF-dependent gene expression in RA FLS. RA FLS were transfected with siRNA targeting luciferase or PHD-2 and subsequently cultured under normoxic conditions (21% oxygen) for 24 hours. Hypoxia-treated cells were used for comparison. A, Gene expression of angiopoietin-like protein family 4 (ANGPTL4), Bcl-2/adenovirus E1B 19-kd protein–interacting protein 3 (BNIP3), carbonic anhydrase IX (CAIX), ephrin A3 (EFNA3), glucose transporter 1 (GLUT1), and vascular endothelial growth factor (VEGF) was assayed by quantitative polymerase chain reaction (qPCR). Data were analyzed by one-way ANOVA. Bars show the mean ± SEM of 7–14 experiments. B–D, RA FLS were transfected using siRNA targeting luciferase, PHDs, and FIH-1, individually or in different combinations, and subsequently cultured under normoxic conditions (21% oxygen) for 24 hours. DMOG-treated cells were used for comparison. B, Protein levels of HIF-1α and HIF-2α were assessed by Western blotting, with α-tubulin as a loading control. C, The expression of GLUT1 mRNA (normalized to the expression of normoxic mock control) was assayed by qPCR. Data were analyzed by one-way ANOVA. Bars show the mean ± SEM of one representative experiment. D, Protein levels of GLUT-1 were assessed by Western blot analysis, with α-tubulin as a loading control. ∗ = P < 0.05; ∗∗ = P < 0.01; ∗∗∗ = P < 0.001 versus siLuc. See Figure 1 for other definitions.

To investigate whether downstream regulation of HIFs by PHD-2 is specific for RA FLS, fibroblasts from patients with OA were treated in a manner analogous to that used in previous experiments. Under normoxic conditions, OA fibroblasts responded to PHD-2 reduction in a manner similar to that of RA FLS, with stabilized expression of HIF-1α and HIF-2α and increased expression of GLUT-1 protein (Figure 3A). Knockdown of PHD-2 led to a significant increase in the expression of a panel of hypoxia-dependent genes in OA FLS under normoxic conditions (Figure 3B). The same set of experiments was carried out using commercially available normal human dermal fibroblasts, and the results showed that following PHD-2 knockdown, these nondiseased cells did not stabilize HIF-α, and that GLUT-1 protein levels were unchanged under normoxic conditions, even though up-regulation of HIF-α, PHD-2, and GLUT-1 protein was observed when cells were exposed to hypoxia or DMOG (Figure 3A). Therefore, although PHD2 mRNA was effectively silenced under normoxic conditions (85% for OA FLS and 75% for normal human dermal fibroblasts; data not shown), normoxia had no effect on hypoxia-regulated genes (Figure 3B).

Figure 3.

PHD-2 knockdown in osteoarthritis (OA) FLS, but not in normal human dermal fibroblasts (NHDFs), elicits a response similar to that observed in RA FLS. OA FLS and normal human dermal fibroblasts were transfected using siRNA targeting luciferase or PHD-2 and subsequently cultured under normoxic conditions (21% oxygen) for 24 hours. A, Protein levels of PHD-2, HIF-1α, HIF-2α, glucose transporter 1 (GLUT-1), and α-tubulin were assessed by Western blot analysis. Hypoxia (1% oxygen)– and DMOG-treated cells were used for comparison. B, The effect of PHD-2 knockdown on the expression of ANGPTL4, BNIP3, CAIX, EFNA3, GLUT1, and VEGF was assessed at the mRNA level by quantitative polymerase chain reaction in OA FLS (n = 4) and normal human dermal fibroblasts (n = 3). The broken line represents the value for normoxic mock control. Data were analyzed by one-way ANOVA. ∗ = P < 0.05; ∗∗ = P < 0.01; ∗∗∗ = P < 0.001 versus siLuc. See Figure 1 for other definitions.

Regulation of angiogenic genes by PHD-2.

An angiogenesis PCR array was used to further investigate the effect of PHD-2. Using these arrays, we previously demonstrated that hypoxia increases the expression of several genes in RA FLS, including the ANGPTL-4, ephrin A3, and leptin genes (23). Silencing of PHD-2 resulted in altered expression of 6 genes in RA FLS. In a representative patient, PHD-2 knockdown increased transcripts for LEP (17.5-fold), EFNA3 (4.3-fold), ANGPTL4 (3.6-fold), EGF (2.7-fold), and VEGFA (2.2-fold) and also decreased PGF (−2.5-fold) (Figure 4A). When results from 4 experiments in RA FLS were pooled, it was clear that PHD-2 knockdown significantly up-regulated the mean expression of 5 genes, namely ANGPTL4 (3.0-fold), EFNA3 (3.6-fold), EGF (2.7-fold), LEP (10.0-fold), and VEGFA (2.0-fold) and significantly decreased the expression of PGF (−2.5-fold) (Figure 4D).

Figure 4.

Silencing PHD-2 under normoxic conditions in osteoarthritis (OA) FLS, but not in normal human dermal fibroblasts (NHDFs), regulates angiogenic genes similar to those in RA FLS. A–C, RA FLS (A), OA FLS (B), and normal human dermal fibroblasts (C) were transfected using siRNA against luciferase or PHD-2 and cultured under normoxic conditions for 24 hours. The scatter plots represent results from a representative experiment. Only genes that were altered by ≥2-fold are shown. D, The fold change in genes altered by siPHD-2 in RA FLS (n = 4), OA FLS (n = 2), and normal human dermal fibroblasts (n = 1) was determined. Data were analyzed using t-tests comparing ΔCt values (gene of interest minus housekeeping gene). Bars show the mean ± SEM. ∗ = P < 0.05; ∗∗ = P < 0.01 versus siLuc. See Figure 1 for other definitions.

PCR array analysis revealed that in OA FLS from a representative patient, PHD-2 knockdown increased the expression of several genes, including LEP (3.6-fold), ANGPTL4 (3.4-fold), EFNA3 (2.8-fold), and VEGFA (2.5-fold) plus an extra gene, CXCL9 (2.6-fold). In addition, the expression of 6 genes was significantly decreased by PHD-2 knockdown, namely CXCL3 (−8-fold), HAND2 (−3.8-fold), FIGF (−2.7-fold), ANGPT1 (−2.3-fold), IL8 (−2.3-fold), and STAB1 (−2.0-fold) (Figure 4B). PHD-2 silencing was repeated in a second patient with OA. The expression of ANGPTL4, EFNA3, LEP, and VEGFA was increased in RA FLS and OA FLS but not in normal human dermal fibroblasts (Figure 4D). Silencing of PHD-2 in normal human dermal fibroblasts revealed that although none of the genes that changed in RA FLS or OA FLS were altered in these cells, the expression of 4 other genes was altered: KDR was up-regulated 2.2-fold, COL4A3 was up-regulated 2.0-fold, THBS1 was down-regulated 3.7-fold, and PLAU was down-regulated 2.3-fold (Figure 4C).

Effect of silencing PHD-3 on angiogenic genes.

PHD-3 protein was not detected under normoxic or hypoxic conditions; therefore, knockdown of this enzyme was confirmed by DMOG treatment, as shown in Figure 5A. The effect of PHD-3 knockdown under normoxic conditions was examined at the mRNA level of the target genes previously shown to be up-regulated by siPHD-2. The expression of these genes was modestly but significantly down-regulated, with the exception of EFNA3 (Figure 5B). Reduced expression of GLUT-1 after silencing of PHD-3 was also observed (Figures 2C and D). In contrast to PHD-2 knockdown, silencing of PHD-3 appeared to decrease expression of the hypoxia-induced proteins BNIP-3 and GLUT-1 (Figure 5C).

Figure 5.

PHD-2 and PHD-3 regulate genes in contrasting ways in RA FLS. RA FLS were transfected using siRNA (10 nM or 25 nM) against luciferase, PHD-2, or PHD-3. A, PHD-3 knockdown (10 nM) was demonstrated by Western blotting, using α-tubulin as loading control, in cells treated with DMOG for 24 hours. B–D, Following PHD knockdown, cells were cultured under normoxic conditions (21% oxygen) for 24 hours. B, Gene expression was assayed by quantitative polymerase chain reaction (PCR) and was normalized to the expression of normoxic mock control (broken line). Data were analyzed by one-way ANOVA. Bars show the mean ± SEM of 5–8 experiments. C, Bcl-2/adenovirus E1B 19-kd protein–interacting protein 3 (BNIP-3) and glucose transporter 1 (GLUT-1) protein levels were assessed by Western blotting, with α-tubulin as a loading control. Hypoxia (1% oxygen) is shown as a positive control. D, The fold change in genes altered by siPHD-2 and siPHD-3 in RA FLS on the PCR array was determined, using a 2-fold induction or reduction in expression as the cutoff. Data were analyzed using t-tests comparing ΔCt values (gene of interest minus housekeeping gene). Bars show the mean ± SEM of 4 individual experiments. ∗ = P < 0.05; ∗∗ = P < 0.01; ∗∗∗ = P < 0.001 versus siLuc. NS = not significant (see Figure 1 for other definitions).

To compare the effect of PHD-3 knockdown with that of PHD-2 knockdown on genes involved in modulating angiogenesis in RA FLS, we used an angiogenesis PCR array. Silencing PHD-3 under normoxic conditions led to significant up-regulation of 6 genes, IL8 (5.8-fold), IL1B (3.9-fold), EREG (3.3-fold), FGFR3 (3.0-fold), EFNA1 (2.4-fold), and CCL11 (2.3-fold), and down-regulation of 1 gene, CDH5 (−5.9-fold) (Figure 5D). Genes that were up-regulated by siPHD-2 were either unchanged by siPHD-3 (EFNA3, EGF, LEP, and PGF) or were down-regulated by siPHD-3 (ANGPTL4 and VEGFA). Three genes that were up-regulated on the array after treatment with siPHD-3 were further assessed by qPCR, revealing that the expression of IL1B, IL8, and FGFR3 was indeed up-regulated by siPHD-3 but down-regulated by siPHD-2 (data not shown).

Induction of proangiogenic activity by supernatants from siPHD-2–treated arthritic fibroblasts.

An in vitro angiogenesis assay was performed to investigate the functional roles of PHD-2 and PHD-3. Supernatants from RA FLS, OA FLS, and normal human dermal fibroblasts transfected with siPHD-2, siPHD-3, or siLuc were added, followed by staining with CD31 antibody. After silencing of PHD-2, a marked proangiogenic effect was observed in RA FLS and OA FLS (Figure 6). Tubule formation was increased after the addition of supernatants from siPHD-2–treated RA FLS and OA FLS and covered 13.1% and 10.2% of the total field area, respectively, compared with 5.8% coverage with supernatants from siLuc-transfected cells. In the case of normal human dermal fibroblast supernatants, tubule formation was comparable in all conditions (Figure 6A). Similarly, the mean number of intertubule junctions increased from 48 (RA FLS) and 45 (OA FLS) in the presence of supernatants from siLuc-transfected cells to 137 and 90 in supernatants from siPHD-2–transfected RA FLS and OA FLS, respectively (Figure 6B). In addition, the mean number of tubules formed in the presence of supernatants from siPHD-2–transfected RA FLS and OA FLS and the total tubule length were increased, while normal human dermal fibroblast supernatants had no significant effect (Figures 6C and D). Knockdown of PHD-3 had no significant proangiogenic effect in any of the cell types tested. Figures and images depicting formation of the tubules are available from the corresponding author.

Figure 6.

RA FLS and osteoarthritis (OA) FLS, but not normal human dermal fibroblasts (NHDFs), transfected with siPHD-2 exert increased proangiogenic activity. RA FLS, OA FLS, and NHDFs were transfected with siRNA targeting luciferase, PHD-2, or PHD-3 and cultured under normoxic conditions (21% oxygen) for 24 hours. Supernatants from these cultures were applied on pre-seeded human endothelial cells for 11 days, after which endothelial cells were stained for CD31. For comparison, cells were left untreated or were treated with vascular endothelial growth factor (VEGF; 5 ng/ml). Quantification of tubule formation was assessed using AngioSys software. A, Percentage of field area covered by CD31-positive tubules. B, Number of junctions. C, Number of tubules. D, Total tubule length. Bars show the mean ± SEM of at least 6 determinations per experiment for RA FLS (n = 4 experiments), OA FLS (n = 2 experiments), and NHDFs (n = 1 experiment). Data were analyzed by one-way ANOVA. ∗ = P < 0.05; ∗∗∗ = P < 0.001 versus siLuc. NS = not significant; U = unstimulated (see Figure 1 for other definitions).


Hypoxia and the HIF pathway are implicated in the pathogenesis of several diseases, including RA. RA synovium exhibits a significant decrease in oxygen tension due to an altered proliferative response of the residing cells (including FLS), and this is paralleled by abnormal vascularization (3, 24). Induction of HIF and downstream HIF target genes facilitates adaptation to these stressful conditions. In order to understand and treat diseases in which HIF is involved, scrutinizing the upstream and downstream events is crucial. Oxygen-dependent regulation of HIF is under the tight control of a family of dioxygenases, including PHD, the activity of which results in binding of HIF to the von Hippel-Lindau ubiquitylation complex and HIF degradation (16, 25), and FIH-1, which controls HIF–CBP/p300 interaction and HIF transcriptional activity (19, 20, 26). Apart from a study showing PHD and FIH-1 expression in nonimmortalized and mouse embryonic fibroblasts (27, 28), there are, to our knowledge, no studies investigating the expression of the enzymes in RA FLS. In addition, the functions of PHD and FIH-1 in RA have not yet been investigated.

In this study, the expression of PHDs and FIH-1 in RA FLS was evaluated at both the mRNA and protein levels. Quantitative PCR analysis of the relative amount of these hydroxylases under normoxic conditions demonstrated that PHD1, PHD2, and FIH1 mRNA levels were high and comparable, whereas PHD-3 was the least abundant. Protein analysis illustrated that endogenous PHD-2 expression was the highest, which is consistent with observations in other cell types (29–31). PHD-1 and FIH-1 were constitutively expressed in RA FLS, but to a lesser degree, while PHD-3 protein was undetectable under normoxic conditions. Although PHD3 mRNA was extremely sensitive to hypoxia, the protein was still undetectable under hypoxic conditions. This could be explained by a lesser abundance of PHD-3 compared with other hydroxylases and possibly by the fact that hypoxia increases expression of the E3 ligase Siah2, which is responsible at least partially for proteasomal degradation of PHD-3 (32). The expression of PHD-2 and PHD-3 in RA FLS was up-regulated by hypoxia and was further up-regulated by the hypoxia-mimetic DMOG, which, by virtue of being a 2-oxoglutarate analog, inhibits the activity of the HIF hydroxylases.

To investigate the role of HIF hydroxylases in RA, all of them were silenced in RA FLS, resulting in an unambiguous reduction in the expression of these enzymes. No effect on HIF-α was observed following knockdown of PHD-1 or FIH-1, suggesting that these HIF hydroxylases do not actively take part in regulation of HIF levels in RA FLS. Interestingly, although PHD-2 itself was regulated only by HIF-1α (data not shown), silencing of PHD-2 stabilized the expression of both HIF-1α and, to a lesser extent, HIF-2α protein, which was mirrored by enhanced HIF binding to HRE and increased transcription of hypoxia-regulated genes. In addition, double and triple knockdown of PHD in different combinations confirmed the selective stabilizing effect of siPHD-2 on HIF-α proteins and HIF targets. This leads to the conclusion that PHD-2 is indeed the major player in regulating the HIF system in RA FLS. In addition, PHD-2 is postulated to be the most important prolyl hydroxylase, in that global knockout of PHD-2 was lethal during mouse embryogenesis, resulting in severe placental and heart defects (33).

Although PHD-3 itself was regulated by HIF-α (data not shown), silencing of PHD-3 in RA FLS under normoxic conditions resulted in a clear increase in the binding of HIF-2 (but not HIF-1) to HRE. Further analysis revealed that PHD-3 knockdown under normoxic conditions actually modestly decreased the expression of HIF-1α protein. One explanation for the differences observed after silencing of PHD-2 and silencing of PHD-3 could be individual preferences of these hydroxylases toward HIF-α (34). It was reported previously that PHD-2 had the highest activity toward HIF-1α, whereas PHD-3 had the highest activity toward HIF-2α (29, 35). These preferences were verified in several immortalized and primary cell lines (27, 36–40) and were corroborated by in vivo studies, in which PHD-2 knockout in adult mice led to HIF-1α stabilization (41), whereas PHD-3 knockout led to HIF-2α accumulation (42).

Due to the fact that PHD-2 depletion protected HIF-1α and HIF-2α from degradation in RA FLS, the downstream effect observed for HIF target genes, which are positively regulated by HIF-1α and HIF-2α, was as anticipated. Indeed, PHD-2 knockdown promoted expression of the angiogenic factors ANGPTL-4, ephrin A3, and VEGF, together with a marker of anaerobic metabolism active in hypoxic conditions, GLUT-1, as well as the survival markers BNIP-3 and CAIX. These responses closely resembled hypoxic induction of these genes, implying that reduced or absent PHD-2 can mimic the hypoxic milieu. It was anticipated that PHD-3 knockdown would be able to regulate HIF-2α–dependent genes, specifically ANGPTL4 and VEGF (23), because siPHD-3 increased HIF-2 binding to HRE. However, silencing of PHD-3 did not increase the expression of these, most likely because induction of HIF-2 alone is not sufficient to induce their expression.

It is believed that both PHD-2 and PHD-3 are implicated in the negative feedback loop controlling the HIF pathway, because they contain HRE and thus are themselves hypoxia-induced and HIF-dependent in certain cells (27, 43, 44). Silencing of PHD-2 accounted for a significant increase in PHD3 mRNA as an HIF target; however, silencing of PHD-3 did not have the same effect on the PHD2 mRNA level, most likely because PHD-3 knockdown did not protect HIF-1α from degradation and even reduced protein expression. The reduction in HIF-1α expression could be attributable to the fact that PHD-2 is present at very high levels and, without any competing PHD-3 (in siPHD-3–treated cells), can more efficiently maintain low HIF-1α levels. On the basis of the findings that reduction of PHD-2 and reduction of PHD-3 do not stabilize HIFs to the same extent and act differently on each other, we suggest that these isoenzymes may have divergent roles in RA.

Subsequent screening of 84 putative gene candidates by PCR array in RA FLS confirmed that silencing of PHD-2 induced the expression of several proangiogenic genes, including ANGPTL4, EFNA3, and VEGF, and also revealed induction of EGF and LEP (which we and other investigators have previously shown to be hypoxia-inducible [23]). Leptin was up-regulated to the greatest extent in this condition. Silencing of PHD-2 mainly increased angiogenic factors; however, silencing of PHD-3 decreased the expression of both ANGPTL4 and VEGF. PHD-3 knockdown decreased the expression of CDH5 and increased the expression of IL1B, IL8, FGFR3, EREG, and EFNA1. Thus, according to the PCR array and qPCR analysis, PHD-2 and PHD-3 may have contrasting roles during neovascularization, because they regulate different groups of genes, and the common ones are regulated in the opposite manner.

To investigate whether the selective central role of PHD-2 is specific for diseased or nondiseased cells, the study was extended to arthritic fibroblasts isolated from OA tissue and nonarthritic normal human dermal fibroblasts. Comparing protein levels of genes involved in the HIF pathway revealed that all 3 cell types exposed to hypoxia and DMOG responded in a comparable manner, up-regulating HIF-α proteins, their targets, and their regulators PHD-2 and PHD-3. However, the substantial differences between the cells were visible by creating hypoxia-mimetic conditions after silencing PHD-2 under normoxic conditions. In normal human dermal fibroblasts, HIF-α proteins were not stabilized, and HIF target genes were not altered. OA FLS, however, responded in a manner similar to that of RA FLS but to a lesser extent, inducing ANGPTL4, EFNA3, VEGF, and LEP. It was previously reported that similar levels of HIF-1α and HIF-2α are expressed in OA and RA (45). In addition, CXCL9 was induced by siPHD-2 in OA FLS, and a group of transcripts were also reduced in these cells, including CXCL3, HAND2, FlGF, ANGPT1, IL8, and STAB1, indicating differences in the transcriptional response of these 2 types of diseased cells.

Consequently, the functional assay testing the ability of conditioned media from cells after silencing PHD-2 or PHD-3 to induce angiogenesis was used to test the hypothesis that PHD-2 is involved in angiogenesis. Supernatants from siPHD-2–treated RA FLS and OA FLS, but not from normal human dermal fibroblasts transfected with siPHD-2, were able to exert a proangiogenic effect on human endothelial cells in these formed tubule-like structures, most likely via increased HIF target gene expression. Conditioned media from siPHD-3–treated cells did not significantly increase angiogenesis in all 3 cell types, possibly due to opposite effects on powerfully proangiogenic transcripts when compared with PHD-2.

Collectively, our data demonstrate that PHD-2 depletion in RA may reproduce hypoxic conditions, thereby promoting synovial angiogenesis. The aforementioned results suggest that in vivo PHD-2 may regulate antiangiogenic responses in RA and OA by controlling HIF-α–dependent responses under normoxic conditions. Regulating a specific master molecule such as PHD-2, which controls HIF and multiple angiogenic growth factors, could be a better and more efficient choice for gene therapy targeting angiogenesis. In addition, these results suggest that targeting PHD-2 could be specific for arthritic/diseased cells.


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. Paleolog 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. Muz, Larsen, Paleolog.

Acquisition of data. Muz, Larsen.

Analysis and interpretation of data. Muz, Larsen, Madden, Kiriakidis, Paleolog.


We are grateful to Drs. Jerome Lafont and Brendan Thoms for designing and providing the siRNA.