Apoptosis and the loss of chondrocyte survival signals contribute to articular cartilage degradation in osteoarthritis

Abstract

Apoptotic death of articular chondrocytes has been implicated in the pathogenesis of osteoarthritis (OA). Apoptotic pathways in chondrocytes are multi-faceted, although some cascades appear to play a greater in vivo role than others. Various catabolic processes are linked to apoptosis in OA cartilage, contributing to the reduction in cartilage integrity. Recent studies suggest that β1-integrin mediated cell–matrix interactions provide survival signals for chondrocytes. The loss of such interactions and the inability to respond to IGF-1 stimulation may be partly responsible for the hypocellularity and matrix degradation that characterises OA. Here we have reviewed the literature in this area of cartilage cell biology in an effort to consolidate the existing information into a plausible hypothesis regarding the involvement of apoptosis in the pathogenesis of OA. Understanding of the interactions that promote chondrocyte apoptosis and cartilage hypocellularity is essential for developing appropriately targeted therapies for inhibition of chondrocyte apoptosis and the treatment of OA.

Introduction

Apoptosis is an evolutionarily conserved form of programmed cell death (PCD) involved in normal physiological processes concerned with growth, development and homeostasis (Wyllie et al., 1980). It is becoming increasingly apparent that aberrant apoptosis plays a significant role in various disease states, including cancer, acquired immune deficiency syndrome (AIDS) and some autoimmune conditions (Carson and Ribeiro, 1993; Richter et al., 1996). Apoptosis, as distinct from necrosis is characterised by such morphological alterations as chromatin condensation, nuclear fragmentation, cell shrinkage, plasmalemma blebbing and apoptotic body formation (Huppertz et al., 1999). Integral to the execution of apoptosis is the interleukin converting enzyme/Caenorhabditis elegans death gene product-3 (ICE/Ced-3) family of proteases. These proteases share the property of cleaving their substrates at the carboxyl terminal of an aspartate residue using a catalytic cysteine at the active site and are resultantly known as caspases (Bratton et al., 2000). These proteases exist as zymogens requiring activation by proteolytic cleavage, performed by other caspases or autocatalysis. This form of post-translational control allows for the production of an amplification cascade of activation (Bratton et al., 2000). There are currently eleven known human members of the caspase family, divided into two classes: Class I with long pro-domains (“initiator caspases”, e.g., caspase-8 and -9) which are likely to be upstream within the apoptotic pathway and are activated first; Class II with short or absent pro-domains which are likely to be downstream effectors (“effector caspases”, e.g., caspase-3, -6, -7) targeting the majority of cellular substrates (Bratton et al., 2000).

Since most cellular proteins remain intact during apoptosis, targets for cleavage are those vital for cell structure and function and those involved in endergonic processes. These reactions must be shut down to conserve the energy required for PCD. Thus the majority of targets are either cell homeostasis catalysts or structural proteins (Rosen and Casciola-Rosen, 1997). Many of the cell homeostasis catalysts cleaved by caspases, such as poly (ADP-ribose) polymerase (PARP) (Lazebnik et al., 1994), the catalytic subunit of DNA dependent protein kinase (DNA-PK_cs) (Casciola-Rosen et al., 1995), and replication factor C (Ubeda and Habener, 1997), are involved in DNA repair or replication. Others, including several ribonucleoproteins deal with RNA processing and some are signal transduction kinases. Targets among cell structure maintenance proteins include the cytoskeletal proteins laminin and actin (Mashima et al., 1995), and cytoskeletal regulatory proteins such as the actin binding protein Gas-2 (Brancolini et al., 1995; Sun et al., 2001).

Articular cartilage has two main extracellular components. Firstly type II collagen and some other collagens, which confer mechanical stability. These collagen fibres form an interwoven lattice, which together with the second component, the glycosaminoglycans (GAGs), resists deformation by mechanical compression forces. GAGs comprise several different types that share similar components. Those most commonly found in articular cartilage are hyaluronic acid, chondroitin sulphate and keratan sulphate (Muir, 1995). GAGs are covalently bound to a large protein called aggrecan core protein to produce matrix proteoglycans (PGs). This PG principally contains chondroitin sulphate residues with approximately 30% keratan sulphate, all bound to a 350 kDa core protein. Chondroitin sulphate PGs occupy a solution volume 30–50 times their own dry mass and as further hydration is prevented by the type II collagen network a pressure develops within the cartilage that confers its compressive resistance (Gardner, 1992a).

PGs form a vast extended network throughout the collagen lattice, to which they are attached by “link” proteins. Extracellular matrix (ECM) proteins such as collagen, laminin and fibronectin in turn bind integrins in the cell membrane. Integrins themselves bind to numerous intracellular proteins particularly cytoskeletal linker proteins such as fodrin, α-actinin, tensin, paxillin and talin. (Fig. 1A). Chondrocytes are held together by chains of small 47 kDa proteins binding them to the proteoglycans. (Fig. 1B). This model of a fibre–cell–matrix network was proposed following in vitro research into β1-integrin interactions resulting from chondrocyte culture medium withdrawal (Cao et al., 1999). Spent culture medium appears to be essential for the survival of chondrocytes in suspended cultures, since complete change of the spent media induces apoptosis. Thus ECM molecules in the culture medium provide vital chondrocyte–matrix interactions and cells deprived of matrix molecules undergo apoptosis.

Osteoarthritis is the most common non-inflammatory arthropathy of load bearing articulating joints in animals and humans. OA is grossly characterised by hypocellularity, cartilage loss and fibrillation. Among domestic species, OA is particularly prevalent in dogs and horses. This disease, whose incidence increases with age, exists in two forms: primary or idiopathic and secondary. Primary OA occurs without known predisposing causes and is surprisingly frequent. 20% of randomly selected dogs were found to have OA at post mortem examination, and as no predisposing causes could be found for 60% of these (Tirgari and Vaughan, 1975), one can assume hereditary factors in the pathogenesis of OA (Hering, 1999). Secondary OA is the most common cause of clinical lameness in dogs, cats and horses and may result from normal force on abnormal joints or abnormal force on normal joints, with the net effect of accelerated cartilage fibrillation.

OA is the clinical phenotype resulting from a number of possible abnormalities of connective tissue function combined with aberrant cell behaviour and an overwhelming of the cartilage’s reparative abilities. OA is especially relevant to dogs since some breeds are prone to initiating causes such as hip or elbow dysplasia. Cranial cruciate ligament rupture is very common in the dog and in large animals as a cause of OA (Gardner, 1992b), and experimental sectioning of the canine and lapine anterior cruciate ligament has been used as an excellent model for the study of OA (Dunham et al., 1985; Cox et al., 1985; Adams et al., 1987; Hashimoto et al., 1998c).

The molecular changes involved in OA pathology appear to follow a pattern involving distinct catabolic and reparative events. In OA mechanical stress initiates cartilage lesions by altering chondrocyte-matrix interaction and metabolic responses in the chondrocytes (Goldring, 2000). There are initial increases in the amounts of water and proteoglycans associated with the observed transient chondrocyte proliferation of early OA. Proliferating chondrocytes appear in clusters or islands and are accompanied by a change in cellular organisation, indicating their undifferentiated nature. Changes in the expression of proteoglycans in OA have also been described, e.g., abnormal retrospective expression of versican, a large chondroitin sulphate proteoglycan normally expressed in prechondrogenic areas of chick embryo limb buds (Nishida et al., 1994), whilst in contrast, collagen X, normally only produced by terminally differentiated hypertrophic chondrocytes, was shown surrounding de-differentiated chondrocyte clusters in osteoarthritic cartilage (von der Mark et al., 1992).

Chondrocyte proliferation is an attempt to counteract the cartilage degradation occurring but disease progression indicates that this is generally unsuccessful. The short-lived hypercellularity (chondrocyte “cloning”) is followed by hypocellularity. Catabolic events responsible for cartilage matrix degradation comprise the release of catabolic cytokines such as interleukin-1β (IL-1β) and tumour necrosis factor-α (TNF-α) inducing matrix degrading enzymes such as matrix metalloproteinases (MMPs) by chondrocytes and by synovial cells in early OA (Martel-Pelletier, 1998; Blanco, 1999; Goldring, 2000). Imbalance between MMPs and tissue inhibitors of MMPs occurs and may be mainly responsible for cartilage matrix degradation (Blanco, 1999), although IL-1β also contributes to the depletion of cartilage matrix by decreasing synthesis of cartilage specific proteoglycans and type II collagen (Taskiran et al., 1994; Studer et al., 1999; Blanco, 1999; Robbins et al., 2000; Goldring, 2000). As the result of matrix degradation and synthesis of an inappropriate matrix, cartilage cannot withstand mechanical load and cartilage fibrillation and breakdown occurs by the focal formation of vertical, oblique and tangential clefts into the ECM preferentially localised in areas of proteoglycan depletion (Gardner, 1992c).

Apoptosis appears to be one reason for the increase in cell loss observed in ageing osteoarthritic cartilage (Stockwell, 1991; Blanco et al., 1998; Adams and Horton, 1998) and ageing is an important predisposing factor for OA. Calcified areas that can be seen in osteoarthritic cartilage may be induced by apoptotic bodies (Blanco, 1999; Hashimoto et al., 1998b). These remnants remaining after the apoptotic death of chondrocytes probably become matrix vesicles and thus participate in the mineralization that occurs during OA particularly in humans (Hashimoto et al., 1998b).

There now exists an overwhelming body of evidence (e.g., Blanco et al., 1998; Hashimoto et al., 1998a), linking chondrocyte apoptosis to the characteristic cartilage degradation of OA although the mechanisms involved are yet to be fully elucidated. There are numerous mediators that will induce apoptosis in chondrocyte cultures, but it is likely that far fewer could function in vivo (Mobasheri, 2002). Those systems implicated in vivo appear to converge on a single united pathway, “the point of commitment” involving the caspase family of proteases, and include:

• The free radical nitric oxide (NO) (Blanco et al., 1995; Studer et al., 1999).

• Fas (APO-1, CD95) and its ligand FasL (CD95L) (Hashimoto et al., 1997; Kim et al., 2000; Kühn et al., 2000).

• TNF-α (Rath and Aggarwal, 1999; Westacott et al., 2000; Aizawa et al., 2001).

• The Bcl-2/Bax proto-oncogene family (Feng et al., 1998).

A considerable amount of work has been done on the role of telomeres and telomerase in the control of cellular senescence, and it is likely that telomerase plays an important role in neoplastic proliferation and is linked to the prevention of apoptosis. Telomerase is a ribonucleoprotein that synthesises 5^′-TTAGGG-3^′ repeats to create telomeres (Morin, 1989). These tandemly repeated sequences located at the ends of chromosomes, serve to protect them from end-to-end fusion and unstable structure formation (Blackburn, 1991). Overexpression of telomerase appears to depress the incidence of apoptosis in both endothelial cells and fibroblasts (Yang et al., 1999; Ren et al., 2001, respectively), while inhibition of the enzyme induces apoptosis in multiple myeloma, Burkitt lymphoma and epidermoid tumour cell lines (Akiyama et al., 2001; Jin et al., 1999; Zhang et al., 1999, respectively). Thus telomerase clearly helps to protect neoplastic cells from PCD, whilst immortality and protection from apoptotic death may be conferred by the overexpression of telomerase. However there is no published evidence that telomerase contributes to the pathogenesis of OA.

NO is a multifunctional gas that is freely permeable across biological membranes. Therefore NO functions as a transcellular messenger and is involved in physiological (e.g., post-translational modification of proteins, protein phosphorylation) and pathological conditions (Chung et al., 2001). There is increasing evidence, that excess NO production is important in the aetiopathogenesis of OA (Amin and Abramson, 1998; Clancy, 1999; Studer et al., 1999). Significantly raised (pathological) levels of NO appear to precede OA symptoms, nitrate, nitrite and iNOS expression, confirmed by experimental induction of OA by anterior cruciate ligament transection (ACLT) (Hashimoto et al., 1998c), although this may not be accompanied by the exhibition of clinical signs.

Chondrocytes produce large quantities of nitric oxide (NO), synthesised via L-arginine oxidation by nitric oxide synthase (NOS), when stimulated by IL-1 or lipopolysaccharide (Stadler et al., 1991). There are at least two isoforms of this enzyme, classed as Ca²⁺ dependent and constitutive—cNOS (represented by endothelial and neuronal NOS) or Ca²⁺ independent and inducible—iNOS (Chung et al., 2001). iNOS is expressed by a variety of cells including chondrocytes following exposure to cytokines such as IL-1 and TNF-α and is a high output enzyme capable of generating elevated, sustained levels of NO (Blanco, 1999). The signalling pathways for the induction of iNOS by IL-1 have recently been demonstrated. There is, it seems, a single pathway activated by IL-1 which rapidly diverges to produce its varied effects. The pathway involves stimulation of protein tyrosine kinases (PTK) and induces iNOS expression and activation of the anti-apoptotic nuclear factor kappa B (NF-κB). The downstream pathway involves the p38 group of mitogen activated protein kinases (MAPK/Erk) but diverges prior to their induction to produce activation of NF-κB (Mendes et al., 2002).

At this stage it is important to clarify the distinction between the concepts of endogenous and exogenous NO. In the context of this review endogenous NO is defined as NO produced intracellularly by chondrocytes, such as that produced in response to IL-1 stimulation. Exogenous NO refers to that produced extracellularly with respect to chondrocytes, including that produced by synoviocytes in vivo and that administered by NO donors such as sodium nitroprusside (SNP) or S-nitroso-N-acetylpenicillamine (SNAP) in vitro.

Experiments in the mid 1990’s were able to link NO to chondrocyte apoptosis (Blanco et al., 1995), but it appears that only exogenous nitric oxide from in vivo sources such as synovial cells, (Amin and Abramson, 1998) or from NO donors such as SNP in vitro will induce chondrocyte apoptosis (Blanco et al., 1995; Notoya et al., 2000; see Fig. 2). Endogenous NO induced by IL-1 treatment fails to induce apoptosis (Blanco et al., 1995; Studer et al., 1999), unless oxygen radical scavengers are also introduced and it was surmised that oxygen radicals were preventing endogenous NO from inducing apoptosis. Blanco et al. (1995), suggested that a regulation system is at work, whereby NO prevents oxygen radical induced necrosis and oxygen radicals prevent NO induced apoptosis. Thus the relative amounts of NO and oxygen radicals determines whether a cell will survive, die by apoptosis or die by necrosis.

The distribution of iNOS expressing cells may demonstrate how NO mediates aspects of OA pathology. iNOS expression occurs in synoviocytes and in cartilage, is significantly greater in cells in the superficial zones of articular cartilage and is strongest in the synovial lining and subsynovium (Amin and Abramson, 1998). The expression of iNOS by the synovial lining is an exciting discovery since chondrocytes do not undergo apoptosis in response to endogenous NO, but exogenous NO will induce apoptosis. The synovial lining expresses iNOS, and thus might release NO into the synovial fluid in response to cytokine stimulation. Since chondrocytes receive their nutrition from the synovium by diffusion through the synovial fluid, it is conceivable that NO might also diffuse across to the chondrocytes and induce apoptosis. Alternatively or coincidentally OA cartilage itself may be a source of exogenous NO. OA cartilage spontaneously produces NO and expresses iNOS (Amin and Abramson, 1998) and so chondrocytes might act as autocrine or paracrine stimulators of apoptosis in themselves or neighbouring cells.

The ongoing production of NO observed in OA cartilage must have a biochemical basis, e.g., cytokine stimulation. The initial induction of iNOS, may however be the result of biomechanical stress. Several studies into the effect of shear stress and injurious mechanical compression on bovine cartilage explants have suggested that NO mediates the effects observed (Loening et al., 2000; D’lima et al., 2001). Apoptosis was induced by peak stresses as low as 4.5MPa concomitant with increased nitrite levels (Loening et al., 2000), while Das et al. (1997), found that NO release by chondrocytes increased in response to both duration and magnitude of fluid shear. More recent data relating to human OA chondrocytes again demonstrates that NO production is induced by shear stress and importantly links this stress with decreased expression of both aggrecan and type II collagen mRNA (Lee et al., 2002). This likely contributes to the reduction of cartilage integrity that characterises OA.

Generally, NO may be characterised as a “proapoptotic modulator” since it affects the expression of apoptosis-associated proteins including p53 and those of the Bcl-2 family and induces ceramide formation (Chung et al., 2001). Unfortunately there is currently no evidence for NO enhancing ceramide production in chondrocytes since this would demonstrate a profound and intriguing link between the NO and Fas mediated apoptotic pathways.

In many cell types including macrophages and endothelial cells NO can directly cause mitochondrial dysfunction, cytochrome release into the cytosol and caspase-3 activation effecting apoptosis, via activation of the Jun-N-terminal kinase/stress activated protein kinase (JNK/SAPK) and p38 group of MAPK’s (Chung et al., 2001). However until recently there was no evidence of these effects in chondrocytes (Kim et al., 2002). Notoya et al. (2000), showed that in chondrocytes NO initially activates extracellular signal-regulated kinases 1/2 (Erk-MAPK) and p38 kinases and induces a cyclo-oxygenase-2 (COX-2) catalysed increase in prostaglandin $\text{E}{}_2\text{(}\text{PGE}{}_2$) production. It now appears that the activation of Erk 1/2 and p38 kinases acts as a regulation system for NO induced effects. NO generated by SNP inhibited PG and type II collagen synthesis and caused apoptosis by the activation of caspase-3. Both Erk 1/2 and p38 kinases were activated following SNP treatment. Regulation appears to be via the differential activation of these two systems, since inhibition of Erk 1/2 leads to enhanced apoptosis, whilst inhibition of p38 kinase leads to blockage of apoptosis. It was suggested that the stimulus for activation of one system over the other was the differentiation status of the chondrocytes. NO induced p38 kinase activity in dedifferentiated chondrocytes, (those not expressing type II collagen or synthesizing PGs), was low when compared with those expressing a differentiated phenotype (Kim et al., 2002).

Whilst endogenously produced nitric oxide may not induce apoptosis, it does appear to mediate various catabolic effects observed in cartilage following IL-1 treatment. These reduce ECM production, disrupt chondrocyte function and damage the structural integrity of cartilage and include inhibition of chondrocyte proliferation (Blanco and Lotz, 1995), type II collagen/proteoglycan synthesis and integrin adhesion to the ECM (Taskiran et al., 1994; Cao et al., 1997).

The inhibition of chondrocyte proliferation by IL-1 is NO dependent and occurs via PGE₂, a known chondrocyte growth inhibitor (Blanco and Lotz, 1995). It appears that both NO and PGE₂ are downstream mediators of the anti-proliferative effects of IL-1, and that production of PGE₂ requires NO. This effect can be prevented by treating with iNOS inhibitors, demonstrating its primary dependence on NO. Prostaglandins including PGE₂ are produced from arachidonic acid in reactions catalysed by prostaglandin synthase, which contains the cyclo-oxygenase component COX-2. PGE₂ and COX-2 are both over-expressed by chondrocytes following SNP administration (Pelletier et al., 2000), suggesting a potential signalling pathway whereby IL-1 induces iNOS to catalyse NO production. In turn this NO increases PGE₂ production from arachidonic acid.

An alternative pathway has, however been suggested. Notoya et al. (2000), showed that COX-2 inhibition dose-dependently reduces SNP induced apoptosis. It was hypothesised that the p38 kinase pathway might link the administration of SNP to the production of PGE₂. This raises an important question; if both IL-1 and SNP induce PGE₂ production, why does IL-1 lead to inhibition of chondrocyte proliferation while SNP produces cellular changes consistent with apoptosis? It appears that this is not due to a difference in induced levels of PGE₂ since the data for both IL-1 and SNP are similar. An observation by Notoya et al. (2000), may provide an explanation, since it was recorded that the induction of PGE₂ production following SNP administration was accompanied by a down-regulation of the anti-apoptotic protein Bcl-2. This observation suggests that exogenous NO may sensitise chondrocytes to the apoptotic effects of PGE₂ by retarding the cells anti-apoptotic defence. This explanation is corroborated by the signalling pathway elucidated for the SNP induction of apoptosis since it involves Erk’s. These translocate into the nucleus to regulate transcription factor phosphorylation. It follows that p38 kinase activation by NO may lead to the down-regulation of Bcl-2 by interfering with gene transcription. However, it ought to be noted that there is considerable criticism of the use of NO donors such as SNP, since these chemicals are generally quite cytotoxic, and likely activate various deleterious pathways, possibly leading to conflicting results.

It is clear that PGE₂ controls the proliferation of articular chondrocytes in pathological states and may be involved in SNP mediated apoptosis, but additional research suggests further involvement of PGE₂ in cartilage degradation (Miwa et al., 2000). Exogenous PGE₂ administered to bovine chondrocytes in vitro results in cellular changes associated with apoptosis, and is accompanied by a dose-dependent increase in intracellular cyclo-adenosine monophosphate (cAMP) levels. It may be concluded that intracellularly, sub-apoptotic levels of PGE₂ mediate proliferation inhibition (Blanco and Lotz, 1995), but PGE₂ may induce apoptosis in its own right when administered exogenously, independent of p38 kinase, by means of a cAMP dependent pathway (Miwa et al., 2000).

The reduction in type II collagen synthesis seen in IL-1 treated chondrocytes (Cao et al., 1997; Richardson and Dodge, 2000) has both a NO dependent and a NO independent component. This explains the only partial restoration of type II collagen synthesis following iNOS inhibition (Cao et al., 1997). IL-1, independently of NO, reduces the intracellular quantities of type II collagen (Col2A1) mRNA, by a yet unknown mechanism (Cao et al., 1997; Robbins et al., 2000; Richardson and Dodge, 2000), although Robbins et al. (2000), reported that p38 MAPK activation is one of the signals required for IL-1β induced inhibition of Col2A1 gene expression. The NO dependent inhibition may occur at the post-translational level, where collagen is subject to several processing stages including hydroxylation, carried out by prolyl hydroxylase (Cao et al., 1997). The activity of this enzyme (which usually correlates well with the rate of collagen synthesis) is decreased following IL-1/SNP treatment (Cao et al., 1997). Prolyl hydroxylase requires Fe²⁺ and ascorbate and generates free radicals, all of which interact with NO. Additionally it is inhibited by poly (ADP) ribose whose synthesis is stimulated by NO.

The initial observation (Stadler et al., 1991), that IL-1 treated chondrocytes show a suppression of proteoglycan synthesis has since been duplicated (Taskiran et al., 1994; Stefanovic-Racic et al., 1995) and it is now clear that NO is again the mediator of this catabolic effect which is relieved by iNOS inhibition. Both endogenous NO and that exogenously provided by SNP addition produce this suppression (Taskiran et al., 1994).

Research published in 1998 concerning aggrecan cleavage in response to IL-1 treatment suggests a second pathway to explain the effect of IL-1 on proteoglycan depletion. Cleavage of the principal proteoglycan in cartilage, aggrecan is carried out by aggrecanase, whose activity is upregulated by TNF-α and IL-1 (Arner et al., 1998). Aggrecanase catabolises proteoglycans and is therefore directly involved in matrix degradation. This effect is not mediated by NO but by the lipid messenger ceramide. Produced by the TNF-α/IL-1 induced breakdown of sphingomyelin, ceramide stimulates the expression of MMP enzymes, such as collagenase-1 at the mRNA level and hence augments their production and activity. (Sabatini et al., 2000) (Fig. 3). As previously mentioned it has been reported that NO-donors increase cellular ceramide levels in HL-60 (human leukaemic) and renal mesangial cells (Chung et al., 2001) suggesting the possibility of a subtler causative link. However the lack of research into this area of chondrocyte biology means that the likelihood of this link cannot yet be evaluated. Ceramide is also the mediator of the effects of TNF-α/IL-1 on matrix related genes observed by Richardson and Dodge (2000), whereby these cytokines increased amounts of MMP 1, 3, and 13 mRNA (collagenase 1, stromelysin 1 and collagenase 3 respectively) by up to 100 times while decreasing amounts of aggrecan mRNA. Importantly this is in contrast to the discovery of Sasaki et al. (1998), that NO mediates the expression of MMP genes following IL-1 stimulation of cultured rabbit chondrocytes, suggesting a bimodal induction of MMP expression.

Excess NO, in addition to mediating several of the effects of IL-1, also damages cartilage integrity by interfering with the adhesion of chondrocytes to the ECM. In so doing NO may detrimentally affect proteoglycan synthesis since synthesis and release of proteoglycans appears to require anchorage signals from integrin stimulation (Clancy et al., 1997; Clancy, 1999).

Chondrocytes bind to ECM proteins via cell surface receptors such as α5β1-integrin whose binding to the ECM protein fibronectin leads to the assembly of a subplasmalemmal Focal Adhesion Complex (FAC). The FAC contains various cytoskeletal proteins, linker proteins and kinases and is a likely signal transduction mechanism, relaying the adhesion-anchorage signals (Clancy et al., 1997). NO prevents FAC assembly without interfering with the ligation of fibronectin to α5β1-integrin. A cyclic-guanosine monophosphate (cGMP) agonist mimicked NO inhibition of FAC assembly, suggesting that NO may act at a cGMP receptor site (Clancy et al., 1997). Fibronectin binding by chondrocytes doubles the rate of proteoglycan synthesis and release, an effect inhibited by NO, probably due to its disruption of FAC assembly (Clancy et al., 1997).

Integrin signalling was also implicated in studies into the effects of ECM anchorage signals on cell survival (Frisch and Francis, 1994; Cao et al., 1999). Apoptosis can be induced by the disruption of interactions between epithelial cells (i.e. kidney derived MDCK cells) and the ECM (Frisch and Francis, 1994). This phenomenon, was termed “anoikis” by Frisch and Francis and has since been observed in articular chondrocytes (Cardone et al., 1997; Cao et al., 1999), is put forward here as an additional pathway for exogenous NO to induce apoptosis.

Integrins function both as cell adhesion receptors and as intracellular signalling receptors (Hynes, 1992) (Fig. 4). Among the downstream factors in these signalling pathways is the docking protein Src-homology collagen (Shc), involved in transduction between receptor tyrosine kinase (RTK) and Ras proteins. The intracellular signalling of β1-integrin ligation by such ECM components as type II collagen involves the Ras-MAPK pathway (Shakibaei et al., 1999). This ubiquitous system regulates such essential cellular functions as growth and differentiation through the phosphorylation and consequent activation of a cascade of proteins. Ras is a membrane anchored intracellular guanosine triphosphate (GTP) binding protein which on binding of its transcellular RTK activates the Ser/Thr kinase Raf. Raf in turn phosphorylates MEK—a mitogen activated protein kinase kinase (MAPKK). The activated MAPKs migrate from the cytosol to the nucleus (Lenormand et al., 1993) where they regulate the activity of transcription factors such as Jun, Fos, and Myc (Denhardt, 1996).

These factors regulate gene transcription and thus orchestrate the cellular processes that may occur. Jun, one of the transcription factors phosphorylated as a result of Ras-MAPK activation, may play a role in the regulation of anoikis, since the JNK pathway is activated following the loss of cell-matrix contact (anoikis), promoting apoptosis in response (Cardone et al., 1997).

The importance of the Ras-MAPK pathway to cell survival was clearly demonstrated by the induction of chondrocyte apoptosis following the inhibition of MEK (Shakibaei et al., 2001). MEK was blocked using a selective inhibitor, interrupting the phosphorylation cascade and preventing the activation of Erk1/Erk2. As mentioned these kinases would normally translocate into the nucleus to regulate the phosphorylation status of transcription factors (Lenormand et al., 1993). Interruption of the pathway produces apoptosis, possibly as the result of increased expression of proapoptotic proteins or repression of anti-apoptotic proteins. Alternatively the inhibition of Erk1/Erk2 may prevent inactivation of proapoptotic factors (Shakibaei et al., 2001).

In addition to mediating cell matrix interactions, integrins are also involved in regulating inflammatory mediators in osteoarthritic cartilage. It was observed that fibronectin and osteopontin mRNA were upregulated in osteoarthritic cartilage compared to normal cartilage and that integrins (such as α5β1 and αVβ3) (Fig. 5) are involved in the up- or down-regulation of several inflammatory mediators including IL-1β via their receptor interactions (Attur et al., 2000). Increased β1-integrin expression was observed in osteoarthritic cartilage, compared with normal cartilage (Loeser et al., 1995), although Lapadula et al. (1998) found that only in minimally damaged osteoarthritic cartilage zones was there a pronounced increase in β1-integrin expression when compared to zones with more pronounced damage. It was finally concluded that β1-integrin-mediated chondrocyte-ECM interaction decreases in osteoarthritic cartilage because of perturbations of chondrocyte matrix signalling. Perhaps this initial increase in β1-integrin expression in early OA represents an attempt at reparation linked to the proliferation of chondrocytes in early OA, that is, α5β1-integrin interactions may mediate early proliferation of chondrocytes (Hering, 1999).

In addition to activating the Ras-MAPK cascade, it is apparent that interactions exist between integrins and insulin-like growth factor-1 receptors. Chondrocytes cultured with type II collagen and exposed to insulin-like growth factor-1 (IGF-1) showed a significantly greater density of chondrocyte adhesions when compared to cultures without IGF-1 (Shakibaei et al., 1999). β1-integrin ligation produces increased phosphorylation of tyrosine residues when IGF-1 is present and the docking signal transduction protein Shc forms a complex with growth factor receptor-bound protein-2 (Grb-2) inducing Ras activation (Shakibaei et al., 1999). These results concur with a number of reports indicating that the deleterious imbalance of matrix catabolism to synthesis may be related to an inability of IGF-1 to exert its physiological anabolic effect on chondrocytes. In healthy cartilage IGF-1 dose-dependently promotes cellular functions including DNA and protein synthesis and the production of ECM components including PGs, aggrecan and type II collagen (Yaeger et al., 1997; Messai et al., 2000). Messai et al. (2000), also noted an age related decline in the chondrocyte response to IGF-1 may be related to inadequate signal transduction indicated by decreased IGF-1 induced cAMP levels.

Paradoxically OA cartilage shows a greater than normal expression of IGF-1 receptors (Doré et al., 1994), although it appears there is no upregulation of insulin-like growth factor receptor 1 (IGFR-1) mRNA in OA chondrocytes (Tardif et al., 1996). OA chondrocytes secrete and express higher than normal levels of insulin-like growth factor binding proteins (IGFBPs)-2, -3, and -4 (Tardif et al., 1996), and it is likely this upregulation is due to cytokine stimulation by IL-1 or TNF-α (Olney et al., 1995). Thus overall, chondrocytes are hyporesponsive to IGF-1 stimulation in OA cartilage probably because of these increased levels of IGF-binding proteins reducing the bioavailabilty of IGF-1 (see reviews by Martel-Pelletier et al., 1998; Loeser et al., 2000). This may also explain the paradoxic increase in IGFR-1 expression. This may result from increased post-transcriptional processing, of the normal pre-existing levels of IGFR-1 mRNA due to the reduction in bioavailability of IGF-1.

Furthermore it appears that NO, whether administered by S-nitroso-N-acetylpenicillamine (SNAP), adenovirus iNOS transfection or IL-1 is partly responsible for the observed cartilage hyposensitivity to IGF-1 illustrated by a reduction in PG synthesis (Studer et al., 2000). This relationship is dose-dependent and it was suggested that NO acts by reducing tyrosine phosphorylation thereby preventing IGF-1 signal transduction. This is in contrast to Clancy et al. (1997), who suggested that NO inhibits the assembly of a sub-plasmalemmal FAC thereby rendering chondrocytes IGF-1 insensitive. These two conflicting conclusions may be due to differing experimental techniques since Clancy et al. (1997), cultured chondrocytes on fibronectin coated plates in contrast to Studer et al. (2000), who used uncoated plastic plates. Although Studer et al. (2000), did concede that the cells might have deposited a matrix which played a part in the effects of NO on IGF-1 responses, they suggested that NO may act in addition on other signal transduction pathways including the phosphorylation status of tyrosine.

Importantly Loeser and Shanker (2000), concluded that autocrine production of IGF-1 helps to maintain chondrocyte survival in vitro, because blocking the IGF-1 receptor with antibodies resulted in apoptotic cell death in chondrocytes, an observation consistent with findings that IGF-1 has anti-apoptotic properties in intervertebral disc cells (Gruber et al., 2000).

At this stage one might justifiably ask how these various catabolic processes are linked to apoptosis in OA cartilage. The answer is, that they all contribute to the reduction of cartilage integrity. Chondrocytes are dependent on adherence to the ECM for their survival, since adherence provides an essential survival signal (Shakibaei et al., 1999). Chondrocytes require cell-cell interactions and cell-ECM interactions, including those with collagen fibres, proteoglycan components and ECM proteins (Shakibaei et al., 1997; Shakibaei, 1998). Thus by the interactions described above depleting the ECM of these components the potential for interactions with chondrocytes is reduced, and may represent an accelerating degradative process (Lotz et al., 1999):

Decreased ECM production and secretion, increases the rate of apoptosis directly due to lack of anchorage survival signals but may also leave chondrocytes vulnerable to apoptosis when challenged by inducers such as Fas or Bax. An increased rate of apoptosis reduces cartilage cellularity, and hence the potential for cell-cell interactions and the amount of ECM components produced. There are resultantly fewer interactions to generate survival signals further increasing the rate of apoptosis. Localised ECM loss, hypocellularity and cartilage degradation leads to the development of the lesional areas observed in OA cartilage (Fig. 6).

In addition to the cellularity reductions observed in aged and OA cartilage there is a >40% incidence of abnormal calcium crystal deposition in the articular cartilage of human patients over the age of 60 (Ohira and Ishikawa, 1987). The increased rates of chondrocyte apoptosis associated with ageing and OA may explain this abnormal calcification, since chondrocyte apoptosis produces apoptotic bodies. Chondrocytes are embedded in ECM, articular cartilage is avascular, and contains no mononuclear phagocytes. Thus there is no apparent mechanism for the removal of apoptotic bodies. These closely resemble the matrix vesicles associated with hydroxyapatite deposition in epiphyseal growth plates of a range of species including chicken, mouse, rabbit, rat and fetal bovines (Wuthier et al., 1985; Wikstrom et al., 1984; Caswell et al., 1987; Murphree et al., 1982; Hsu and Anderson, 1982, respectively). Apoptotic bodies themselves are able to precipitate calcium, since both matrix vesicles and chondrocyte-derived apoptotic bodies contain alkaline phosphatase and nucleoside triphosphate pyrophosphohydrolase (NTPPH) a pyrophosphate forming enzyme. Pyrophosphate is a ubiquitous physiological intermediate in sources of high-energy phosphate, and so its role in OA may not appear obvious. Excess pyrophosphate sometimes accumulates and crystallises as crystals of calcium pyrophosphate dihydrate (CPPD), causing pain, debilitation and other clinical signs in humans (Gardner, 1992c). Calcium deposits may result in increased rates of fibrillation and promote chondrocyte protease synthesis through expression of synoviocyte IL-1. These findings have been replicated in non-human primates (Roberts et al., 1984), laboratory rabbits (Fam et al., 1995), and in in vitro models (Mandel et al., 1984). To conclude, it is clear that chondrocyte-derived apoptotic bodies are capable of contributing to the pathological calcification of cartilage observed in OA (Hashimoto et al., 1998b, Hashimoto et al., 1998c) both in humans and in other mammalian species.

Fas is a type I membrane protein belonging to the TNF/NGF (Nerve Growth Factor) receptor family. Fas is a receptor ligated by FasL, a cytokine of the TNF family and a type II membrane protein. Binding of Fas by agonistic antibodies or FasL triggers apoptosis in the expressing cell. The first extensive study of chondrocyte Fas expression produced exciting results (Hashimoto et al., 1997). Cartilage from patients with OA showed comparable levels of Fas expression to that of normal joints, implying that expression of Fas is not up-regulated in OA and suggesting that increased abundance of FasL or sensitisation of Fas pathways may be responsible for the increase in apoptosis observed in OA (Hashimoto et al., 1997). It was found that cell density is an important modulator of Fas/FasL expression in chondrocytes and thus influences susceptibility to Fas mediated apoptosis (Kühn et al., 1999).

Fas positive cells are normally located only in superficial and middle zones, although due to fibrillation in OA cartilage Fas expressing cells may also be detected in deeper layers, and especially in lesional areas (Hashimoto et al., 1997; Kim et al., 2000). Fas ligation induces apoptosis in articular chondrocytes and importantly the process is NO independent, demonstrating the existence of another pathway to apoptosis in chondrocytes (Hashimoto et al., 1997).

Apoptosis induced by Fas ligation is mediated by ceramide, which acts via numerous mechanisms, including direct caspase activation, down-regulation of Bcl-2 family cell-survival proteins and by inducing cytoskeletal protein breakdown (Rudel and Bokoch, 1997). Ceramide also induces the caspase dependent synthesis of ganglioside GD3. This membrane constituent may have deleterious effects on inner mitochondrial membrane potential (Δψ_m) (De Maria et al., 1997; Fernandez-Checa et al., 2000). Δψ_m may be involved in both FasL and Bax induced apoptosis (Richter et al., 1996), and the collapse of Δψ_m is an early characteristic of cells undergoing apoptosis (Petit et al., 1996). The release of mitochondrial proteins into the cytosol was proposed as an explanation, and led to the discovery of apoptosis inducing factor (AIF). This pro-apoptotic mitochondrial protein is possibly an unidentified member of the caspase family since its activity can be reduced by the general caspase inhibitor zVADfmk (benzyloxycarbonyl-Val-Ala-Asp-fluoromethylketone) (Susin et al., 1997).

There is no evidence that chondrocytes express FasL and thus it is probable that if Fas-induced apoptosis occurs in OA then the ligand is exogenous to the cartilage. Since it has been demonstrated that synovial cells express FasL (Hashimoto et al., 1997), the expression of Fas by chondrocytes in articular cartilage would provide a pathway for the induction of apoptosis. Synovial cells are the most likely source of exogenous FasL, in the form of soluble FasL (sFasL) as detected in the synovial fluid of OA patients (Hashimoto et al., 1998).

Soluble FasL is produced from FasL by the action of MMP enzymes. It has been shown that cytokines dramatically increase the levels of MMP activity and mRNA expression in chondrocytes (Richardson and Dodge, 2000). This conversion of FasL represents a link between the Fas pathway and the many catabolic processes induced by IL-1/TNF-α. In the experiments conducted by Hashimoto et al. (1998) “naturally processed” sFasL was unable to induce apoptosis although recombinant sFasL was, but despite this apparent breakdown in the evidence, sFasL from synovial cells is still the most credible explanation for Fas induced apoptosis in chondrocytes.

Several recent publications have however questioned the extent of FasL induced apoptosis in chondrocytes. Masuko-Hongo et al. (2000), measured Fas and FasL expression by chondrocytes and induced apoptosis with agonistic antibodies. They found that FasL expression was minimal and despite a high level of Fas expression the agonistic antibodies failed to induce apoptosis at a comparable frequency. Coupled with these observations Kühn et al. (2000), showed that cytokine IL-1β protects human chondrocytes from Fas induced apoptosis and it was suggested that an NF-κB mediated gene activation is involved in this anti-apoptotic mechanism, accompanied by upregulation of Bcl-2 expression. Following this Kühn and Lotz (2001), examined the roles of NF-κB and caspases 3, 8 and 9 following FasL induced apoptosis. Fas binding caused an increase in NF-κB activity, which was chondroprotective. Apoptosis was augmented by proteasome inhibitors, which interfere with NF-κB translocation indicating the protective influence of this factor. Caspases 3 and 8 were upregulated following Fas ligation, but procaspase 9 processing was not detected. It is likely that NF-κB acts to prevent caspase 3 activation at or downstream of caspase 8 and that caspase 9 is not a downstream mediator of FasL induced apoptosis. Additionally the low expression of caspase 8 by chondrocytes demonstrated by Kühn and Lotz (2001), allows only slow activation of caspase 3. This may possibly explain the low levels of FasL-induced apoptosis in chondrocytes, relative to other death-signalling systems, and since IL-1β initiates many of the catabolic events in OA it may be assumed that Fas-mediated apoptosis does not play a pivotal role in this disease.

In osteoarthritic cartilage the expression of Bcl-2 is lower than in healthy cartilage (Kim et al., 2000). This mammalian proto-oncogene first discovered during studies on B-cell lymphomas is homologous to the product of apoptosis suppressing ced-9 gene found in C. elegans. Since the initial discovery in the mid 1980’s at least 15 members of this family have been discovered in mammalian cells. As an oncogene, Bcl-2 is unusual, in that it doesn’t enhance cell proliferation per se but instead has the rather unique activity of suppressing apoptosis. Detailed characterisation of the Bcl-2 family members revealed several key observations (Adams and Cory, 1998; Adams and Cory, 2001; Bratton et al., 2000; Oltvai et al., 1993):

• Conserved regions in Bcl-2 family members, four Bcl-2 homology domains and the membrane spanning domain.

• Multiple types of interaction between Bcl-2 family members (hetero-dimerization).

• There are anti-apoptotic members such as Bcl-2, Bcl-x_L, and Bcl-w.

• There are pro-apoptotic members such as Bax, Bak, Bad, Bid, and Bik (Kelekar et al., 1997).

Bcl-2 protects cells from apoptotic death in response to numerous stimuli, such as growth factor withdrawal, ionising radiation, heat/cold shock and oxidative stress. There are several Bcl-2 insensitive pathways to apoptosis such as ligation of Fas or binding at TNF-R1 sites (Adams and Cory, 2001). These bindings occur at the cell membrane, the site of caspase activation by these stimuli. Bcl-2 is located intracellularly, not at the plasma membrane and so it is logical that Bcl-2 is unable to block these pathways.

Despite the apparent insensitivity of the Fas pathway to Bcl-2 anti-apoptotic effects, there does appear to be a link between the two pathways. The link is downstream, and it may thus be argued that since the point of commitment has already been reached, links at this stage are irrelevant. The activation of caspases by ligation of Fas induces the cleavage of Bcl-2 in the loop domain of the molecule. This cleavage not only inactivates the anti-apoptotic activity of Bcl-2 but also produces a pro-apoptotic fragment of some potency (Cheng et al., 1997). Whether this breakdown of Bcl-2 represents the removal of a block on apoptosis or a step in a positive feedback cycle remains to be seen, although the latter appears more substantiated since overall Fas-induced apoptosis is Bcl-2 insensitive.

Bcl-2 is involved in the determination of which cells in the epiphyseal cartilage survive during development, via the suppression of apoptosis (Amling et al., 1997). It was realised that it might play a role in chondrocyte apoptosis in articular cartilage (Feng et al., 1998) and in ageing and pathological states in particular. Experiments to determine the patterns of Bcl-2 expression in rats, revealed both the tissue-wide and cellular distributions of Bcl-2 expression (Wang et al., 1997). Bcl-2 immunoreactivity was confined to the cells, i.e. no Bcl-2 was found in the ECM. Bcl-2 is found in the cytoplasm, but not in nuclei, their subcellular location was pinpointed to the mitochondria (Wang et al., 1997). The levels of Bcl-2 expression in cytoplasm decrease as cells mature, i.e. proliferative cells express more than mature cells, which in turn express more than hypertrophic cells (Wang et al., 1997).

In chondrocytes of transgenic mice lacking collagen type II, decreased levels of Bcl-2 protein were found suggesting that distinct matrix components are important for chondrocyte survival (Yang et al., 1997). Independent of its control of apoptosis a novel role of Bcl-2 was proposed in regulating differentiated phenotype of chondrocytes and the expression of the differentiation specific aggrecan gene (Feng et al., 1999).

The precise mechanism by which Bcl-2 and similar proteins actually prevent apoptosis is not known. Numerous methods have been proposed but most likely is that directly or indirectly, Bcl-2 prevents caspase activation. Bcl-2 may indirectly prevent caspase activation by preventing the release of cytochrome C and AIF from mitochondria by stabilising the membrane and averting collapse of Δψ_m. This latter mechanism is given credence by the intracellular location of Bcl-2, as Bcl-2, Bcl-xL and Bax predominantly reside on nuclear, endoplasmic reticulum and mitochondrial membranes (Conus et al., 2000).

The overexpression of Bcl-2 has been shown to modulate the activity of telomerase in human cervical carcinoma HeLa cells, human colorectal carcinoma DiFi cells and the mouse cytotoxic T-cell line CTLL-2 (Mandal and Kumar, 1997), although not in Jurkat T cells (Johnson et al., 1999). These results suggest a link between apoptosis and telomerase activity. Telomerase has been linked to the acquisition of immortality by a wide range of neoplastic cells (Shay and Bacchetti, 1997). Since telomere length appears to assist in the regulation of normal cell proliferation and senescence (Harley et al., 1992; Holt and Shay, 1999) it is possible that telomerase may play a role in OA or chondrocytes in vivo but no experimental evidence has been put forward in support of this. Interestingly, a recent publication suggests a potential therapeutic use for telomerase in OA. The exogenous expression of telomerase by chondrocytes following retroviral transfection augmented cell proliferation whilst maintaining the differentiated phenotype. This might represent a method for the restoration of the articular cartilage defects that characterise OA (Piera-Velazquez et al., 2002).

Bcl-2 may have an additional influence on chondrocyte survival, independent of its direct anti-apoptotic role. Research into the relationship between matrix protein expression and that of Bcl-2 has revealed that constitutive expression of Bcl-2 will maintain levels of aggrecan expression in conditions of serum withdrawal (Feng et al., 1999). This allows for Bcl-2 to prevent apoptosis by maintaining the integrity of the ECM, thus indirectly preventing apoptosis due to a lack of anchorage signals.

Further links between components of the ECM, anchorage signalling and Bcl-2 have been suggested, since the expression of Bcl-2 is down-regulated in conditions of serum withdrawal and in mutant mice lacking the ability to express type II collagen (Horton et al., 1998; Yang et al., 1997). These studies suggest that ECM components are required for trophic support and survival of chondrocytes. The Ras-MAPK signalling pathway (see Fig. 4) elucidated in chondrocytes by Shakibaei et al. (1999) provides an explanation for both aforementioned observations, and again demonstrates the links between anchorage signals from integrin interactions and apoptotic systems (Shakibaei et al., 2001).

Bax, a pro-apoptotic member of the Bcl-2 family, accelerates apoptosis, by forming heterodimers with Bcl-2 (Oltvai et al., 1993; Yin et al., 1994), although this heterodimerisation is not required for the cell survival function of Bcl-2 (Cheng et al., 1997). In unstimulated cells Bax resides in the cytosol but a cell death signal such as caspase activation causes translocation of Bax to the mitochondria where it integrates in the membrane by its transmembrane domain and mediates and amplifies membrane dysfunctions (Goping et al., 1998).

The ratio of anti-apoptotic proteins to pro-apoptotic proteins appears to determine the fate of cells (Oltvai et al., 1993). An excess of anti-apoptotic proteins such as Bcl-2 will allow the cell to survive while an excess of pro-apoptotic proteins such as Bax will lead to programmed cell death. Bcl-2 expression declines with age, but it was observed that Bax expression, located similarly to that of Bcl-2 remained constant as the cells matured (Wang et al., 1997). Thus, as chondrocytes mature, the ratio of pro-apoptotic Bax to anti-apoptotic Bcl-2 shifts in favour of apoptosis. This may explain the age-related hypocellularity and consequent cartilage degeneration observed in ageing (Vignon et al., 1976; Adams and Horton, 1998).