To uncover the mechanism by which hypoxia enhances cartilage matrix synthesis by human articular chondrocytes.
To uncover the mechanism by which hypoxia enhances cartilage matrix synthesis by human articular chondrocytes.
The hypoxic response was investigated by exposing normal (nonarthritic) human articular chondrocyte cultures to 20% oxygen and 1% oxygen. Induction of the differentiated phenotype was confirmed at the gene and protein levels. In its first reported application in human articular chondrocytes, the RNA interference method was used to directly investigate the role of specific transcription factors in this process. Small interfering RNA directed against hypoxia-inducible factor 1α (HIF-1α), HIF-2α, and SOX9 were delivered by lipid-based transfection of primary and passaged human articular chondrocytes. The effect of each knockdown on hypoxic induction of the chondrocyte phenotype was assessed.
Hypoxia enhanced matrix synthesis and SOX9 expression of human articular chondrocytes at both the gene and protein levels. Although HIF-1α knockdown had no effect, depletion of HIF-2α abolished this hypoxic induction. Thus, we provide the first evidence that HIF-2α, but not HIF-1α, is essential for hypoxic induction of the human articular chondrocyte phenotype. In addition, depletion of SOX9 prevented hypoxic induction of matrix genes, indicating that the latter are not direct HIF targets but are up-regulated by hypoxia via SOX9.
Based on our data, we propose a novel mechanism whereby hypoxia promotes cartilage matrix synthesis specifically through HIF-2α–mediated SOX9 induction of key cartilage genes. These findings have potential application for the development of cartilage repair therapies.
Being avascular, cartilage has a low oxygen concentration (1, 2). Chondrocytes, the unique resident cells, are therefore adapted to these hypoxic conditions, e.g., by having lower levels of oxidative phosphorylation (3) and enhanced anaerobic glycolysis (4). Furthermore, recent studies suggest that hypoxia triggers essential positive signals for the chondrocyte phenotype beyond such survival responses. Indeed, we previously reported that reduced oxygen tension increases levels of the cartilage matrix genes COL2A1 and aggrecan and the cartilage transcription factor SOX9 in cultured articular chondrocytes (5, 6). Domm and colleagues also showed a positive effect of hypoxia on Colα1(II) levels in bovine chondrocytes (7). Hypoxia has been shown to promote chondrogenic differentiation of mesenchymal stem cells, and it was recently shown that the mouse Sox9 promoter is activated by hypoxia in mouse mesenchymal cells (8).
In various cell types, hypoxic effects have been shown to be mediated by hypoxia-inducible factor (HIF) transcription factors. These proteins belong to the Per-ARNT-Sim (PAS) subfamily of basic helix-loop-helix (bHLH) transcription factors and consist of an α subunit and a β subunit (9). Under conditions of normoxia, HIF-1α is hydroxylated on specific proline residues, ubiquitinated through interaction with the von Hippel-Lindau tumor suppressor protein, pVHL, and subsequently degraded by the proteasome (10). Conversely, under low oxygen concentrations, the activity of the HIF-targeting prolyl hydroxylase enzymes (PHD1, PHD2, and PHD3) is suppressed, and HIF-1α is not hydroxylated and therefore not targeted for proteasomal degradation. It translocates to the nucleus, heterodimerizes with HIF-1β (aryl hydrocarbon nuclear translocator), and activates transcription by binding hypoxia-responsive elements (HREs) located in the promoter region of hypoxia-inducible genes (11, 12).
More recently, another bHLH-PAS protein, HIF-2α (endothelial PAS domain protein 1), also was implicated in hypoxic responses. HIF-2α and HIF-1α are closely related genes and share 48% amino acid identity and 83% identity in their bHLH domains (13). Both proteins are believed to be subject to identical posttranslational regulation by oxygen. However, HIF-2α does not appear to be redundant, and there is accumulating evidence that the relative importance of HIF-2α and HIF-1α in the response to hypoxia varies among different cell types (14).
Experiments in mice have shown that in the growth plate, HIF-1α is essential for chondrocyte cell survival, through the induction of vascular endothelial growth factor (VEGF) (15). HIF-1α was also implicated in relation to matrix synthesis by isolated murine epiphyseal chondrocytes (16) and human osteoarthritis chondrocytes (17). However, the role of HIF-1α in hypoxic induction of the human articular chondrocyte phenotype remains uncertain, while that of HIF-2α is completely unexplored.
Therefore, we addressed these important issues in the present study. In the first reported application of RNA interference (RNAi) in human chondrocytes, we specifically investigated the roles of HIF-1α, HIF-2α, and SOX9 transcription factors. Our results provide the first direct evidence of the essential role of SOX9 in hypoxic induction of the differentiated chondrocyte phenotype, and, strikingly, the dependence of this specifically on HIF-2α rather than HIF-1α.
Healthy articular cartilage was obtained from patients after they provided informed consent, and following local ethics committee guidelines. Cartilage was harvested from the femoral condyle and tibial plateau following amputation due to osteosarcoma or soft tissue sarcoma with no involvement of the cartilage. Tissue was obtained from 9 donors (7 male and 2 female donors; age 8–50 years [mean age 27 years]). Cartilage specimens were collected on the day of surgery and cut into small pieces (1–2 mm3). Diced cartilage was placed in 1.5 mg/ml collagenase 2 (Worthington, Freehold, NH) with Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal calf serum (FCS) (Biosera, East Sussex, England, UK) and incubated at 37°C for 18 hours with shaking. Isolated cells were then passed through a cell strainer, pelleted, and washed twice with medium. Cells were seeded at a density of 8 × 103 cells/cm2 in DMEM with 10% FCS. Cultures were passaged at the time of confluency (∼7 days) and subsequently were seeded at 5 × 103 cells/cm2. Both primary (unpassaged) and passaged cells were used in the experiments, as indicated.
Cells (both freshly isolated and passaged chondrocytes) were seeded at 5–8 × 104 cells/dish in 3.5-cm dishes. Cultures were transfected at ∼50% confluency (typically 1–2 days after seeding). Transfection with siRNA was carried out at a final concentration of 10 nM, using Lipofectamine 2000 (Invitrogen, San Diego, CA), for 4 hours in serum-free OptiMEM I. Small interfering RNA against HIF-1α, HIF-2α, and SOX9 were used (MWG Biotech, Ebersberg, Germany) (Table 1). As a nontargeting control, siRNA against luciferase (Dharmacon, Lafayette, CO) were transfected in parallel. Following transfection, medium was changed with pre-equilibrated DMEM (in 20% or 1% oxygen) containing 10% FCS, and cells were incubated in each oxygen environment for 3 days, in a Galaxy triple gas incubator (CRS Biotech, Ayrshire, Scotland, UK).
RNA was extracted and prepared using the RNeasy Mini Kit (Qiagen, Crawley, England, UK). Complementary DNA (cDNA) was generated using a reverse transcription kit (Promega, Southampton, England, UK) and random hexamers from 0.5 μg of total RNA. Four percent of this cDNA was then used for real-time PCR, using TaqMan technology. The delta delta threshold cycle (ΔΔCt) method of relative quantitation was used to calculate relative messenger RNA (mRNA) levels for each transcript examined. Pre-developed primer/probe sets for the following genes were purchased from Applied Biosystems (Foster City, CA): COL1A2, COL2A1, COL9A1, aggrecan, SOX9, HIF-1α, HIF-2α, and the housekeeping gene RPLP0.
To investigate the effect of hypoxia on SOX9 mRNA stability, chondrocytes were exposed to normoxic or hypoxic conditions (as described above) for 48 hours, after which 1 μM actinomycin D (Sigma, Gillingham, England, UK) was added, and cells were incubated for up to 5 hours. To exclude any effects attributable to reoxygenation, hypoxic cells were maintained in 1% oxygen until the time of RNA extraction. RNA was extracted at 0, 1.5, 3, and 5 hours after the addition of actinomycin D. Real-time RT-PCR was performed as described above. Relative levels of remaining SOX9 mRNA (normalized to RPLP0) were reported for each time point.
Cells were seeded in LabTek chamber slides (Nalge Nunc International, Naperville, IL) at 2.5 × 104 cells/cm2. Immunolabeling was performed after 2 washes and fixation with methanol. Cells were incubated with Colα1(II) antibody (MAB8887; Chemicon, Southampton, England, UK) at a 1:200 dilution for 1 hour, followed by incubation with secondary horseradish peroxidase antibody.
Cells were lysed in urea lysis buffer (8M urea, 10% glycerol, 1% sodium dodecyl sulfate [SDS], 5 mM dithiothreitol, 10 mM Tris HCl) for HIF analysis or otherwise in radioimmunoprecipitation assay buffer. A protease inhibitor cocktail (Sigma) was added just prior to cell lysis. Lysates were cleared by centrifugation for 10 minutes at 14,000 revolutions per minute. Twenty micrograms of protein was separated by SDS–polyacrylamide gel electrophoresis, transferred onto polyvinylidene difluoride membranes for Western blotting, and finally visualized using the enhanced chemiluminescence method. Primary antibodies used were rabbit anti-SOX9 (AB5535; 1:1,000 dilution) (Chemicon), mouse anti–HIF-1α (clone 54; 1:250 dilution) (BD Transduction Laboratories, Lexington, KY), mouse anti–HIF-1β (clone 29; 1:1,000 dilution) (BD Transduction Laboratories), mouse anti–HIF-2α (Sc-13596; 1:250 dilution) (Santa Cruz Biotechnology, Santa Cruz, CA), mouse anti–α-tubulin monoclonal antibody (Sigma), and rabbit anti–total ERK (developed in-house). For matrix protein analysis, 0.5 ml of conditioned medium was deglycosylated with chondroitinase and keratinase for 3 hours at 37°C and then precipitated with trichloracetic acid. Aggrecan antibody (1:1,000 dilution) was a kind gift from Prof. Timothy Hardingham; Colα1(II) antibody (MAB8887; 1:1,000 dilution) was purchased from Chemicon.
Data were compared by one-way analysis of variance with Bonferroni post test, using GraphPad Prism 4 software (GraphPad Software, San Diego, CA). Results are expressed as the mean ± SEM from 3–4 independent experiments (i.e., using cells from 3–4 individual patients). P values less than 0.05 were considered significant.
To test whether hypoxia is able to promote the chondrocyte phenotype, human articular chondrocytes were incubated in 20% oxygen (normoxia) and 1% oxygen (hypoxia) for up to 7 days, and real-time PCR analysis was used to measure transcript levels of COL2A1 and aggrecan. The expression of both genes was significantly increased in 1% oxygen (compared with normoxia) over the 7-day period (Figure 1A). Hypoxic induction of matrix genes was also observed at the protein level, as assessed by Western blotting of Colα1(II) and aggrecan from the secreted protein fraction (Figure 1B) and by immunostaining of the cell layer for Colα1(II) (Figure 1C).
Steady-state SOX9 mRNA levels were similarly analyzed after incubation of human articular chondrocytes in hypoxia or normoxia for 4 or 7 days. Real-time PCR quantification revealed a significant increase in SOX9 levels in hypoxia (Figure 2A). Hypoxia had no effect on SOX9 mRNA stability, as assessed by actinomycin D chase experiments (Figure 2B). From the decay curves, the half-life of SOX9 mRNA was calculated to be 1.67 hours in 20% oxygen and 1.50 hours in 1% oxygen. The increased steady-state mRNA levels thus appear to be attributable to increased transcription of SOX9 in hypoxic conditions rather than changes in message stability. Western blot analysis showed a strong up-regulation of SOX9 protein after exposure of human articular chondrocytes to hypoxia (Figure 2C), and this response was observed in both primary and passaged chondrocytes (Figure 2D). Taken together, these data show that hypoxia acts as a strong positive regulator of the chondrocyte phenotype.
In order to determine whether up-regulation of chondrocyte markers was mediated by the HIF pathway, we depleted HIF-1α and HIF-2α in human articular chondrocytes using siRNA. Messenger RNA analysis by real-time PCR revealed that HIF-1α message was depleted by 83%, on average (compared with luciferase controls), and that HIF-2α message was depleted by ∼93% (Figures 3A and B). Western blotting was performed to confirm the depletion of HIF-α isoforms at the protein level. As shown in Figure 3C, a clear depletion of each HIF-α protein was achieved in both passaged and primary human chondrocytes. The mRNA and protein data demonstrate the gene specificity of each siRNA. These siRNA (Table 1) were chosen after preliminary tests of 2–3 different oligonucleotides targeting each HIF-α isoform. Finally, to assess the functional effect of the knockdowns, the expression levels of known HIF target genes, glucose transporter 1 (GLUT1) and vascular endothelial growth factor (VEGF), were analyzed. Depletion of HIF-1α significantly reduced hypoxic induction of GLUT1 (Figure 3D), while the VEGF response to hypoxia was significantly reduced by HIF-2α depletion (Figure 3E).
COL2A1 expression was significantly increased by hypoxia. HIF-1α siRNA had no effect, whereas HIF-2α siRNA abolished this hypoxic induction (Figure 4A). COL9A1 mRNA levels were likewise up-regulated in hypoxia (∼3.5 fold). Again, HIF-1α siRNA did not affect expression, whereas HIF-2α siRNA significantly reduced hypoxic induction of COL9A1 (Figure 4B).
SOX9 expression was then analyzed in the different knockdowns, and as shown in Figure 4C, hypoxic induction of SOX9 mRNA was not significantly affected when HIF-1α was depleted but was significantly reduced by HIF-2α depletion (P < 0.05). Supporting this finding, Western blot analysis of SOX9 from passaged and primary chondrocytes after HIF-α knockdowns clearly demonstrated that SOX9 hypoxic induction was dependent on HIF-2α, but not on HIF-1α (Figures 4D and E).
SOX9 was depleted using 2 different siRNA (Figures 5A and B). The first (siSOX9 #1) gave a reproducibly stronger knockdown, depleting mRNA levels by 77%, on average (compared with relevant luciferase controls). When SOX9 was depleted in 20% oxygen, COL2A1 and COL9A1 mRNA levels were likewise very similarly down-regulated (Figures 5C and D), whereas expression of COL1A2 (a non-SOX9 target gene) was not affected (Figure 5E). This demonstrates that SOX9 is necessary for basal expression of matrix genes (COL2A1 and COL9A1).
COL2A1 and COL9A1 were up-regulated ∼4-fold and ∼3-fold, respectively, by hypoxia. SOX9 depletion prevented this induction (Figures 5C and D). Thus, these results show that the main cartilage matrix genes are not directly regulated by hypoxia, but require SOX9 as an intermediary.
In the present study, we provide clear evidence of up-regulation by hypoxia of key cartilage matrix genes, at both the mRNA and protein levels, in healthy (nondiseased) human articular chondrocytes. We then directly investigated the role of HIF transcription factors in this response, using RNAi to specifically deplete both HIF-1α and HIF-2α. Although Lin and colleagues have used RNAi on a chondrosarcoma cell line (18), to our knowledge, the present study is the first application of this technique in human articular chondrocytes, and successful depletion of both HIF-α isoforms was achieved in both passaged and primary cells. Our results demonstrate that HIF-2α, but not HIF-1α, is essential for hypoxic induction of the key cartilage matrix genes and of the cartilage transcription factor SOX9. In parallel knockdown experiments performed in normoxia, we observed the same HIF-2α–specific dependence of SOX9 induction in the presence of the hypoxia mimetic desferrioxamine, which stabilizes both HIF-1α and HIF-2α (data not shown).
Pfander and colleagues cultured HIF-1α–null and wild-type murine epiphyseal chondrocytes in normoxia and in hypoxia, in order to investigate the contribution of HIF-1α to cell metabolism (16). Phosphoglycerate kinase (PGK), GLUT1, and VEGF were all clearly up-regulated by hypoxia in wild-type cells but not in HIF-1α–null cells, indicating the importance of HIF-1α for hypoxic induction of these genes. However, the case was less clear for cartilage matrix genes. Although increased type II collagen protein levels were reported under hypoxic conditions, there was no hypoxic induction of COL2A1 or aggrecan mRNA or of matrix proteoglycan levels in wild-type cells. The lack of a hypoxic induction of cartilage matrix genes is different from our observations, and might be explained by the fact that human cells from permanent articular cartilage were used in the present study, whereas Pfander and colleagues used chondrocytes from the epiphyseal cartilage of newborn mice. In addition, unlike the present study, the study by Pfander et al did not investigate HIF-2α. However, mouse articular chondrocytes have been shown to express HIF-2α protein (19); moreover, HIF-2α−/− mice (which die within days after birth) are very small in size, suggesting abnormal endochondral bone growth (20). Based on our findings of the importance of HIF-2α in human articular chondrocytes, a cartilage-specific and inducible HIF-2α knockout is being developed to investigate its contribution to maintenance of articular cartilage in an in vivo setting.
Previous studies have shown COL2A1 and aggrecan expression to be correlated with SOX9 during mouse cartilage development (21, 22). Using mouse Sox9−/− embryonic stem cells to generate chimeras, Bi and colleagues (23) elegantly showed that SOX9 is required at the early mesenchymal condensation stage of chondrocyte differentiation during cartilage formation. Ex vivo studies have demonstrated that SOX9 binds to enhancer elements in the COL2A1 and COL11A2 genes and can activate these enhancers in nonchondrocytic cells (23, 24). Furthermore, Tew and colleagues reported increased expression of COL2A1 after retroviral transduction of SOX9 in human osteoarthritis chondrocytes (25).
Thus, in the present study, we directly investigated whether SOX9 is required for hypoxic induction of cartilage matrix genes, by depleting the gene using RNAi. Our results clearly showed that SOX9 is essential for hypoxic induction of cartilage matrix genes in human articular chondrocytes. Indeed, when SOX9 expression was depleted, levels of COL2A1 and COL9A1 were similarly down-regulated, showing tight regulation of their expression by SOX9. This is in strong agreement with an in vivo study that used a Cre/loxP recombination system to ablate SOX9 after the chondrogenic mesenchymal condensation stage in developing mouse limbs (26). Using in situ hybridization, those investigators demonstrated severe down-regulation of the matrix genes COL2A1 and aggrecan following deletion of SOX9.
Although SOX9 is considered the key transcription factor in cartilage, very little is known about its regulation in articular chondrocytes. Fibroblast growth factors can stimulate SOX9 in mouse primary chondrocytes as well as in undifferentiated mesenchymal cells (27). In the growth plate, SOX9 is a target of parathyroid hormone–related protein signaling in prehypertrophic chondrocytes and helps to maintain the chondrocyte phenotype of cells in the prehypertrophic zone (28). This study is the first to provide direct evidence that hypoxia promotes the articular chondrocyte phenotype through SOX9 and, furthermore, that this is mediated specifically by HIF-2α. Robins et al previously suggested that HIF could be involved in SOX9 up-regulation by hypoxia in mouse stromal cells (8). Those investigators identified HREs (sequences known to bind HIF proteins) in the mouse Sox9 promoter. Mutations of these consensus sites abrogated the hypoxic response, as assessed in a luciferase activity assay.
Although these experiments suggest HIF participation in SOX9 regulation by hypoxia in mouse stromal cells, they are not informative about which HIF isoform is involved. Based on our findings, it would be of interest to determine whether SOX9 is a direct target of HIF-2α in human articular chondrocytes. However, analysis of recruited factors on the human SOX9 promoter is difficult because of the particular complexity of this promoter (29–32), with important regulatory elements dispersed over a large region upstream of SOX9 (32). Furthermore, unlike the murine promoter, no HREs have been reported in the human SOX9 promoter (33). Indeed, given the progressive increase in SOX9 levels observed in the present study, it may be that although HIF-2α is essential, other factors are also involved in hypoxic induction of SOX9. However, further studies are needed to address these issues.
In a recent study, the p38 MAPK pathway was shown to be involved in SOX9 mRNA stability (34). However, no change in SOX9 mRNA stability by hypoxia was detected in the present study. Thus, it appears that hypoxic induction of SOX9 is attributable to increased transcription (via HIF-2α) rather than changes in mRNA stability.
The present study represents the first detailed analysis of HIF-1α/HIF-2α in human articular chondrocytes. Based on our data, we propose a novel mechanism by which hypoxia promotes the chondrocyte phenotype through up-regulation of cartilage matrix genes, and specifically via HIF-2α–mediated SOX9 induction (as depicted in Figure 6). Because HIF-α protein stability and activity are highly dependent on hydroxylation status, it will be interesting in future studies to uncover which specific hydroxylases inhibit HIF-2α function in human articular chondrocytes. Indeed, it may be possible to enhance cartilage matrix deposition by directly inactivating these enzymes (thereby stabilizing and/or activating HIF-2α and increasing SOX9 and matrix protein levels). Such an approach may help identify potential points of intervention for cartilage repair therapies through, for example, the use of small molecule inhibitors (or RNAi) against specific hydroxylases in cartilage tissue.
Dr. Murphy 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 design. Lafont, Murphy.
Acquisition of data. Lafont, Talma, Murphy.
Analysis and interpretation of data. Lafont, Murphy.
Manuscript preparation. Lafont, Murphy.
Statistical analysis. Lafont, Murphy.
We thank Jeremy Saklatvala, Jon Dean, and Andrew Clark for helpful discussions.