Oxygen is essential to life for all higher organisms. Molecular oxygen is required as an electron acceptor in the generation of cellular energy (ATP) through the process of oxidative phosphorylation, and it is also used as a substrate in various enzymatic reactions [1]. Oxygen homeostasis is, therefore, a basic requirement and complex systems have evolved to maintain this at the cell, tissue and whole organism levels. These include increased reliance on anaerobic glycolysis in the formation of ATP within the cell; increased angiogenesis and blood supply (through vasodilation) to affected organs; and systemic changes such as enhanced erythropoiesis and increased ventilation [2, 3].

Cartilage develops in a hypoxic environment [4], and indeed proximity to a blood supply appears to be a determining factor in the formation of bone over cartilage [5, 6]. In addition, due to the absence of vasculature, articular cartilage (unlike most tissues) is maintained and functions in a low oxygen environment throughout life [7–10]. The resident cells, the chondrocytes, are the only cell type present in the tissue and appear to have developed specific mechanisms to promote tissue function in response to this chronic hypoxia, for example, by inducing increased expression of cartilage matrix components [11–13], and through the inhibition of angiogenesis [14]. In addition to mediating the ubiquitous hypoxia responses, hypoxia-inducible factors (HIFs) also appear to be critical to these tissue-specific responses in chondrocytes.

In the mid-1990s a major breakthrough was made in our understanding of the molecular mechanisms mediating cellular responses to hypoxia with the discovery of HIF-1 [15]. The stability and function of HIF is regulated post-translationally by hydroxylation of specific amino acid residues. In the presence of sufficient molecular oxygen, HIF is degraded almost as soon as it is made due to hydroxylation of specific proline residues that target the HIF-α subunit for Von Hippel-Lindau tumour suppressor protein (pVHL)-mediated proteosomal degradation. Conversely, when oxygen levels are limiting (typically <5%), hydroxylation is inhibited and HIF-α escapes degradation, and is free to heterodimerise with the constitutively expressed HIF-β subunit (also called Aryl hydrocarbon nuclear translocator (ARNT)). This complex translocates to the nucleus, binding specific consensus sequences (-RCGTG-) within the promoter of its target genes and thus activating their transcription (Figure 1).

Other HIF-α members were subsequently discovered, namely HIF-2α, which is structurally similar to HIF-1α, and more recently HIF-3α. The latter was shown to produce at least six different isoforms following alternative splicing [16]. HIF-1α and HIF-2α have the same fundamental protein structure, a basic-helix-loop-helix (bHLH) domain at the amino terminus, an intermediate PER-ARNT-SIM (PAS) domain, and a trans-activation domain (TAD). HIF3-α lacks the last of these, and it has been suggested that it could act as a dominant negative for HIF-1α and HIF-2α [16, 17].

Data have emerged in recent years highlighting the importance of HIF-1α in the developing growth plate in the mouse [18]. Schipani and colleagues [4] first demonstrated that the developmental growth plate was hypoxic, and deletion of HIF-1α led to chondrocyte death coupled with decreased expression of the CDK inhibitor p57, thus strongly suggesting that HIF-1α is essential for chondrocyte survival and growth arrest. More recent data have highlighted HIF-1α's role in regulation of differentiation of the limb bud mesenchyme and in joint development [18]. Hypoxia was also shown to increase matrix synthesis of isolated epiphyseal chondrocytes in a HIF-1α-dependent manner [19]. HIF-2α was shown to be elevated during chondrocyte differentiation and to be present in the articular cartilage in a study by Stewart and colleagues [20]. Deletion of VHL (which results in overexpression of HIF-1α and HIF-2α) increases matrix deposition by chondrocytes during growth plate development [21]. The role of HIFs in the permanent articular cartilage has been little studied. However, a recent study has reported induction of osteoarthritis in BALB/c mice after intra-articular injection of the anti-angiogenic compound 2-methoxyoestradiol [22]. Although promising, 2-methoxyoestradiol is not a specific HIF-targeting compound and its mechanism of action is not clear, although it is thought to be related to disruption of microtubule assembly in the cell [23]. In addition, HIF-2α was not investigated in this study, but presumably was also affected by 2-methoxyoestradiol treatment in a manner similar to HIF-1α.

Despite the above-mentioned important findings in the mouse, extending these data to humans is fraught with difficulties. A major concern with regard to hypoxia is the different thickness between human and mouse cartilage. For example, being merely a few cells in thickness, appreciable oxygen diffusion is possible in mouse knee articular cartilage; whereas the equivalent site in humans is several millimetres thick. As a consequence, the oxygen concentration in human articular cartilage may be significantly lower than that in the mouse [7–10, 24]. In addition, the mechanical loads experienced by mouse and human knees are obviously hugely different [25]. Hence, although extremely useful for developmental studies, for the understanding of adult articular cartilage in humans, the mouse model is limited.

It has long been known that the chondrocyte phenotype is unstable in culture [26–28]. Moreover, chondrocyte phenotypic alterations are observed in cartilage pathology, such as osteoarthritis [29]. Controlling the chondrocyte phenotype remains, therefore, a major challenge for cartilage repair strategies. Being the only cell type within the tissue, the chondrocytes are solely responsible for secreting the specialised extracellular matrix that gives the tissue its biomechanical function. Articular cartilage is under two permanent stresses, mechanical and hypoxic. Although it is widely accepted that loading and compression applied to cartilage are potent regulators of chondrocyte physiology [30–33], the role of hypoxia on chondrocyte function is less well established. A general response of articular chondrocytes to their hypoxic environment is their reliance on anaerobic metabolism to generate cellular energy (ATP), and oxygen consumption of the tissue is accordingly low [34]. In addition, hypoxia has specifically been shown to promote tissue function by upregulating expression of cartilage matrix genes in isolated bovine [13, 35] and human articular chondrocytes (HACs) [36]. Similar results have been reported for human meniscal cells [37]. Applying the technique of RNA interference, we subsequently demonstrated that HIF-2α, but not HIF-1α, was critical for this hypoxic induction of cartilage matrix synthesis in HACs [11]. Furthermore, the main matrix genes, such as those encoding Col-2a1, aggrecan and Col-9, are not direct HIF targets, but are upregulated by hypoxia through cartilage-specific transcription factor SOX9. Whether HIF-2α directly targets SOX9 in HACs remains unknown. However, mouse stromal cells (ST2) transfected with a Sox9 promoter construct showed upregulation under hypoxia [38], and when putative hypoxia response element sequences (located within the first 500 bp) were mutated, hypoxic induction was abolished. These results have been supported more recently in micromass culture experiments, which showed, using chromatin immunoprecipitation, recruitment of HIF-1α to the Sox9 promoter precisely on the same hypoxia response element-containing site [39].

The ability of mesenchymal stem cells (MSCs) to differentiate into chondrocytes (in vitro and in vivo) and to be readily expanded in tissue culture without loss of multilineage potential has made them very attractive candidates for cell-based articular cartilage repair. In addition, unlike articular chondrocytes, the use of MSCs is not hindered by the availability of suitable healthy tissue since MSCs can be isolated from a variety of tissues [40–42]. Implantation of MSCs in an animal model of osteoarthritis has resulted in engraftment of the cells in the meniscus, fat pad, and synovium, with regeneration of the medial meniscus [43]. In addition, degeneration of the articular cartilage and osteophytic remodelling were reduced in MSC implanted joints compared with control joints. Similar results have been reported in the treatment for focal defects in articular cartilage [44]. In a clinical trial MSCs were transplanted using hydroxyapatite ceramic scaffolds to treat severe osteochondral damage after septic arthritis of the knee [45]. Successful cartilage-like tissue regeneration was observed by a second athroscopy.

Recent studies have reported that hypoxia enhances chondrogenic differentiation of MSCs (in comparison to control cultures at ambient oxygen tension) [38, 46]. In addition, Lennon and colleagues have also reported that low oxygen enhanced in vivo chondrogenesis of rat MSCs [47]. Our laboratory investigated the effects of oxygen tension (20%, 5%, and 1% O₂) on the chondrogenic differentiation of both murine and human MSCs when cultured in the presence of 5 ng/ml transforming growth factor-β3. Chondrogenic differentiation took 3 to 4 days in reduced oxygen tensions, and 1 week in cultures exposed to 20% oxygen. Furthermore, low oxygen tension significantly enhanced the number of chondrogenic nodules formed, as well as the intensity of cartilage-specific type II collagen staining, in comparison to 20% cultures (Figure 2). Interestingly, when cultures were exposed to reduced transforming growth factor-β3 levels (1 ng/ml), chondrogenic nodules formed in only 5% and 1% O₂ (that is, not 20%; unpublished data), suggesting hypoxia plays an essential role in the onset of chondrocyte differentiation from MSCs. However, a fundamental problem with the use of MSCs for cartilage repair is that they readily become terminally differentiated, with production of type × collagen [48]. Obviously, for the permanent articular cartilage this is undesirable. Since type × collagen production occurs with a concomitant decrease in SOX9 expression in joint development [49], it is tempting to speculate that hypoxia may delay or inhibit terminal differentiation of MSCs through induction and maintenance of SOX9 levels.

The specific role of HIFs in this hypoxic induction of chondrogenesis from MSCs deserves further exploration and, interestingly, Hardingham and colleagues [50] have recently shown that human MSCs isolated from the infrapatellar fat pad showed enhanced chondrogenic differentiation in hypoxia and, furthermore, that HIF-2α, but not HIF-1α, was upregulated in these cultures. This supports findings in our laboratory that specifically HIF-2α promotes the differentiated HAC phenotype [11].

The direct oxygen sensors are not the HIFs, but the hydroxylases targeting them since the latter are enzymes that require oxygen as a co-factor. Hydroxylation of HIF proline residues occurs on the amino-terminal end of the trans-activation domain (on Pro402 and Pro564 of human HIF-1α) [51]. Three prolyl hydroxylases, prolyl hydroxylase domain enzymes 1 to 3 (PHD-1 to PHD-3) have been shown to act in this way [52]. An asparaginyl residue located in the carboxy-terminal domain (on Asn803 of human HIF-1α) is also hydroxylated by a specific enzyme called Factor inhibiting HIF (FIH). Hydroxylation by FIH inhibits transcriptional activity of HIF by preventing recruitment of the transcriptional co-activator p300/CBP [53, 54].

The HIF targeting hydroxylases (PHD1/2/3 and FIH) belong to a family of iron- and oxoglutarate-dependent hydroxylases, and dimethyloxaloylglycine (DMOG; being an analogue of oxoglutarate) can inhibit all family members. We have shown that DMOG, like hypoxia, stabilises HIF-1α and HIF-2α in human articular chondrocytes, and subsequently induces SOX9 in a specifically HIF-2α-dependent manner (Figure 3a, b). Although a useful experimental tool, DMOG is not appropriate for therapeutic application since, being a non-specific hydroxylase inhibitor, it also inhibits procollagen hydroxylases such as prolyl and lysyl hydroxylases, which are critical for post-transcriptional processing and triple helical formation of cartilage collagens [55]. In fact, these collagen prolyl hydroxylases are themselves upregulated by hypoxia [56], and we have observed similar hypoxic induction of lylsyl hydroxylase (PLOD2) in human articular chondrocytes [12]. Interestingly, Gelse and colleagues [22] have recently reported that DMOG injection into murine joints led to increased intra-cellular accumulation of collagen molecules, presumably due to defective collagen processing and hence impaired secretion. Thus, there is a critical need for development of HIF-specific hydroxylase inhibitors if these deleterious effects on collagen processing are to be avoided.

All three HIF-targeting prolyl hydroxylases (PHD1/2/3) have been detected in the maturing zone of the mouse growth plate [57]. PHD2 was shown by Pouyssegur and colleagues to be dominant hydroxylase regulating HIF-1α [58], at least in non-chondrocytic cell lines. Such PHD selectivity for HIF-1α has also been shown by Applehoff and co-workers [52]. Since HIF-2α and not HIF-1α is involved in the control of the human chondrocyte phenotype [11], it is now important to uncover if PHDs show selectivity for HIF-2α in human articular chondrocytes. Interestingly, in recent microarray experiments on HACs, we have observed a very pronounced hypoxic induction of PHD3 mRNA [12], although PHD2 message was the most abundant both in hypoxia and normoxia (Figure 3c). Nevertheless, the relative contribution of each hydroxylase may be dependent on the prevailing oxygen tension.