New findings in osteoarthritis pathogenesis: therapeutic implications

Abstract

This review focuses on the new perspectives which can provide insight into the crucial pathways that drive cartilage-bone physiopathology. In particular, we discuss the critical signaling and effector molecules that can activate cellular and molecular processes in both cartilage and bone cells and which may be relevant in cross talk among joint compartments: growth factors (bone morphogenetic proteins and transforming growth factor), hypoxia-related factors, cell–matrix interactions [discoidin domain receptor 2 (DDR2) and syndecan 4], signaling molecules [WNT, Hedgehog (Hh)]. With the continuous progression of our knowledge on the molecular pathways involved in cartilage and bone changes in osteoarthritis (OA), an increasing number of potentially effective candidates for OA therapy are already under scrutiny in clinical trials to ascertain their possible safe use in an attempt to identify molecules active in slowing or halting OA progression and reducing joint pain. We then review the principal molecules currently under clinical investigation.

Introduction

Osteoarthritis (OA) is the most prevalent chronic joint disease, affecting 30–50% of adults over 65 years [Loeser, 2010]. The burden of disease is growing in relation to aging and the increasing levels of obesity in the world population. OA is a heterogeneous disease for which age and obesity have been recognized as prominent risk factors; however, additional critical factors have been identified, including genetics, mechanical stress, injury and hormonal modifications. All joints may be affected but the most common localizations are the knee, hands and hip.

The disease is characterized by two main features: the progressive damage of articular cartilage and bone remodeling or new bone formation (osteophytes and subchondral bone sclerosis). In addition, variable degrees of synovial inflammation/fibrosis and damage/fibrosis of ligaments, tendons, menisci and capsules are very frequently found (Figure 1). These whole-joint pathological changes progressively lead to severe limitation of physical activity and great impairment of quality of life, which can ultimately lead to articular prosthetic substitution [Loeser, 2010]. In the World Health Report 2002, OA was estimated to be the fourth leading cause of ‘years lived with disability’ (YLDs), accounting for 3% of total global YLDs [Symmons et al. 2002].

Figure 1.

Differences between normal and osteoarthritic (OA) joints. Normal: all relevant tissues are represented; OA: changes occurring in OA joint.

Therapeutic options are currently limited to pain relief and the eventual replacement of damaged joints (hip and knee), as no available treatments are able to substantially modify disease progression at the present time.

The cartilage-centered pathogenic hypothesis focuses on chondrocyte dysregulation as a primary event, triggered by mechanical stress or inflammatory stimuli. Chondrocyte phenotype shifts toward a catabolic phenotype characterized by the increased production and activation of different factors that act as mediators or effectors of progressive cartilage loss, including proinflammatory cytokines [interleukin (IL)-1, tumor necrosis factor (TNF)] [Melchiorri et al. 1998; Goldring, 2000; Kapoor et al. 2011], chemokines, such as growth regulated oncogene α (GROα; CXCL1), IL-8 (CXCL8), monocyte chemotactic protein 1 (MCP1; CCL2), regulated and normal T-cell expressed and secreted (RANTES; CCL5), macrophage inflammatory protein (MIP)-1α (CCL3), MIP1β (CCL4), stromal cell-derived factor 1 (CXCL12) and eotaxin-1 (CCL11) [Borzi et al. 2004; Wei et al. 2010; Chao et al. 2011; Rasheed et al. 2011; Wenke et al. 2011], extracellular matrix (ECM) degrading enzymes including matrix metalloproteinases (MMPs) and aggrecanases [A Disintegrin and Metalloproteinase with Thrombospondin Motifs (ADAMTS)], which act as downstream key players in the inflammatory signal cascade [Cawston and Wilson, 2006; Song et al. 2007]. In addition, a relevant role as mediators of cartilage damage is carried out by nitric oxide (NO) and prostaglandins (PGEs) [Mazzetti et al. 2001; Henrotin et al. 2005], which are actively produced by chondrocytes and appear to be upregulated in osteoarthritic-affected cartilage [Melchiorri et al. 1998; Mazzetti et al. 2001; Masuko-Hongo et al. 2004; Li et al. 2005].

Contribution of these molecules to cartilage degradation stems from their ability to enhance the production and activation of MMPs, to inhibit proteoglycan and collagen biosynthesis and to induce chondrocyte apoptosis [Goldring et al. 1996; Abramson, 1999; Notoya et al. 2000; Henrotin et al. 2005; Abramson, 2008; Attur et al. 2008].

Mechanical stress, which is one of the main risk factors and one of the driving pathogenic forces for OA, has been identified as a relevant inducer of downstream inflammatory and catabolic events in cartilage [Sun, 2010; Leong et al. 2011]. In addition to mechanical stress and inflammatory mediators, catabolic pathway may be activated by toll-like receptor (TLR) engagement. TLRs constitute a phylogenetically conserved family of receptors that recognize different patterns of pathogen-associated antigens [Medzhitov, 2001; Zhang and Schluesener, 2006; Liu-Bryan and Terkeltaub, 2010]. TLR activation can also be triggered by endogenous ligands, including a group of molecules, such as hyaluronan fragment [Termeer et al. 2002; Jiang et al. 2005], heparan sulfate [Johnson et al. 2002], fibronectin extra domain A [Okamura et al. 2001] and high mobility group box chromosomal protein 1 [Liu-Bryan et al. 2010], which may be released following tissue injury/remodeling. Chondrocytes and synovial cell can express TLRs and, in particular, TLR2 and TLR4 appear to be upregulated in damaged areas of OA cartilage [Kim et al. 2006]. TLR activation promotes catabolic response in chondrocytes, by increasing MMPs, NO and PGE2 production and downregulating biosynthesis of matrix macromolecules [Kim et al. 2006; Bobacz et al. 2007; Zhang et al. 2008a; Liu-Bryan et al. 2010].

Evidence of inflammatory involvement in OA pathogenesis has been very recently strengthened by the findings of a complement dysregulation in human OA [De Seny et al. 2011; Fernandez-Puente et al. 2011; Wang et al. 2011]. Several complement components appear to be differentially expressed in the synovial fluid and serum of patients with OA compared with healthy subjects [De Seny et al. 2011; Fernandez-Puente et al. 2011; Wang et al. 2011]. Furthermore, the role of complement has been supported by animal model studies. In particular, mice who had undergone medial meniscectomy to induce experimental OA and genetically deficient in complement component 5 (C5) (which plays a pivotal role in the activation of the complement cascade) showed a lower degree of cartilage loss, osteophyte formation and synovitis than wild-type mice. Indeed, in C5-deficient mice, chondrocyte expression of proinflammatory and catabolic molecules, such as ECM enzymes (MMPs, ADAMTS4, ADAMTS5), chemokines (MCP1/CCL2, RANTES/CCL5) was lower than the chondrocyte expression in wild-type mice [Wang et al. 2011]. Furthermore, sublytic concentration of the membrane attack complex has been shown to upregulate gene expression of the above-mentioned molecules in cultured human chondrocytes [Wang et al. 2011].

Currently, the pathogenetic concept is moving towards a more complex model of OA as a disease of the whole joint. In particular, in the last few years, a growing body of evidence has highlighted the relevance of bone involvement and its contribution to OA mechanisms, promoting great research interest focused on cartilage bone as a unique functional unit [Lories and Luyten, 2011].

The substantial progress made in identifying the pathways involved in progressive joint destruction has unraveled an ever-increasing complexity of OA molecular pathogenesis [Loeser et al. 2012]. The challenge involved in understanding the primary mechanisms and those responsible for disease progression appears to be extremely relevant for allowing the identification of key pathogenic factors as specific novel therapeutic targets to satisfy the great demand for therapeutic interventions able to effectively modify the course of disease.

This review focuses on the new perspectives which can provide insight into the crucial pathways that drive cartilage-bone pathophysiology. In particular, we discuss the critical signaling and effector molecules that can activate cellular and molecular processes in both cartilage and bone cells and which may be relevant in cross talk among joint compartments: growth factors [bone morphogenetic proteins (BMPs) and transforming growth factor (TGF)], hypoxia-related factors, cell–matrix interactions (DDR2 and syndecan 4), signaling molecules (WNT, Hh) .

Growth factors: transforming growth factor β, bone morphogenetic proteins and fibroblast growth factors

Chondrocyte metabolism, survival, proliferation and differentiation are tightly controlled by several signaling pathways which are also involved in controlling and promoting new bone formation [Onyekwelu et al. 2009; Lories and Luyten, 2011]. In normal adult healthy cartilage, chondrocytes are in a quiescent phase characterized by a fine balance between synthetic activity [synthesis of micro- and macromolecules of the ECM, the most abundant of which are aggrecans and collagen II] and the production of catabolic enzymes, mainly MMPs and ADAMTS [Goldring and Marcu, 2009].

Several studies have demonstrated that the histological process of OA development and progression is associated with the differentiation of a variable number of chondrocytes (mainly located in the deep zone of cartilage) into hypertrophic chondrocytes, which are characterized by the expression and production of collagen X and MMP13 [Von Der Mark et al. 1992; Fuerst et al. 2009; Little et al. 2009; Gelse et al. 2012; Van Der Kraan and Van Den Berg, 2012]. This phenotype modification is called hypertrophic in analogy with the differentiation which takes place during childhood and early adolescence in the growth plate of growing bones [Kronenberg, 2003; Mackie et al. 2008]. In brief, during chondrogenesis and bone formation, mesenchimal chondroprogenitor cells condense and initiate chondrocyte differentiation. Cartilage is formed, but chondrocytes progress into the final differentiation stage: hypertrophy. At this stage, cells begin to produce several MMPs (MMP13 being the most representative), acquiring an ‘autolytic’ phenotype marked by their ability to induce the degradation of surrounding (pericellular) cartilage matrix [Van Der Kraan et al. 2010a, 2012] which has to be replaced by new bone. A similar phenotypic modification occurs in OA (and also ageing) chondrocytes [Bertrand et al. 2010; Van Donkelaar and Wilson, 2011; Van Der Kraan et al. 2012]. We may therefore say that during OA development a proportion of chondrocytes, acquiring an inflammatory-catabolic phenotype, seems to regain the differentiation process and reach the terminal stage.

The differentiation process is controlled by some very stringent signaling pathways, among which the molecules of the TGFβ family play a prominent role. The TGFβ superfamily is comprised of more than 40 members, also including the BMPs [Guo and Wang, 2009]. It is noteworthy that TGFβ1 is one of the major molecules considered to be anabolic for cartilage [Van Beuningen et al. 1993; Serra et al. 1997; Yang et al. 2001], together with insulin-like growth factor 1 [Madry et al. 2005], fibroblast growth factor (FGF)-2 [Chia et al. 2009] and BMP7 [Hunter et al. 2010]. TGFβ signals via its type II receptor which then engages the type I receptors: these receptors are called activin-like kinase (ALK)-1 and ALK5 [Goumans et al. 2003; Konig et al. 2005; Finnson et al. 2008; Blaney Davidson et al. 2009]. ALK1 complex activates the Smad1–5–8 pathway, while ALK5 in turn phosphorylates Smad2–3 [Finnson et al. 2008; Blaney Davidson et al. 2009] (Figure 2). Several pieces of evidence strongly suggest that these two activation pathways are master regulators of chondrocyte phenotypic change and differentiation progression [Blaney Davidson et al. 2009]. This hypothesis is based mainly on animal studies but it is corroborated by confirmation studies on human OA tissue [Serra et al. 1997; Yang et al. 2001; Blaney Davidson et al. 2006; Hellingman et al. 2011].

Figure 2.

Interplay of various signaling pathways acting upon the two main chondrocyte protagonist in osteoarthritis (OA) cartilage derangement and bone formation: fibrogenic chondrocyte and hypertrophic chondrocyte. ADAMTS, a disintegrin and metalloproteinase with thrombospondin motifs; ALK, activin-like kinase; BMP, bone morphogenetic protein; Coll, collagen; DDR2, discoidin domain receptor 2; HIF, hypoxic inducible factor; IL, interleukin; MMP13, matrix metalloproteinase 13; RUNX2, Runt-related transcription factor 2; SMAD, small mother against decapentaplegic; SOX9, SRY-related high-mobility- group (HMG) box transcription factor 9; TGFβ, transforming growth factor β; TNF, tumor necrosis factor; VEGF, vascular endothelial growth factor.

ALK5 activation by TGFβ engagement and subsequent signaling via Smad2–3 contributes to the maintenance of the stable quiescent phase of chondrocytes and aggrecan and collagen II production. Smad2 and 3 exert an inhibitory effect on chondrocyte hypertrophy [Yang et al. 2001; Li et al. 2006] (Smad3 appearing to have a more prominent role than Smad2) [Alvarez and Serra, 2004]. In addition to this protective role, Smad2–3 are involved in the well-known action of TGFβ in promoting chondrophyte and then osteophyte formation [Yang et al. 2001; Blaney Davidson et al. 2006] and synovial fibrosis [Scharstuhl et al. 2003; Blaney Davidson et al. 2006]. BMPs cooperate in the final stage (endochondral ossification) of osteophyte formation [Van Beuningen et al. 1998; Blaney Davidson et al. 2007; Itasaki and Hoppler, 2010].

Phosphorylated Smad1–5–8 cooperate with the transcription factor Runt-related transcription factor 2 (RUNX2) to stimulate hypertrophic differentiation with the consequent production of collagen X, MMP13, osteopontin, alkaline phosphatase, osteocalcin and vascular endothelial growth factor (VEGF) by chondrocytes [Van Der Kraan et al. 2010b; Hellingman et al. 2011]. The release of MMP13, ADAMTS and HtrA1 (an enzyme involved in the degradation of pericellular or territorial matrix) causes the focal breakdown of aggrecan, facilitating the direct exposure of collagen II to chondrocyte surface receptors. Subsequent DDR2 receptor engagement by collagen II further stimulates the hypertrophy process (see page 8).

Recent elegant studies by the Radboud University group in Nijmegen have demonstrated a shift in ALK1/ALK5 ratio occurring in ageing and during OA both in humans and in mice [Blaney Davidson et al. 2009]. In ageing and in OA, a loss of the TGFβ receptor ALK5 causes reduced phosphorylation of Smad2–3. Conversely, only a small reduction in ALK1 expression is documented [Blaney Davidson et al. 2009], therefore a relative predominance of Smad1–5–8 signaling is operating in ageing and OA cartilage, thus promoting the hypertrophic differentiation (Figure 2).

Another series of papers stress the role of TGFβ signaling through ALK5/Smad2–3 in the transition of chondrocytes and chondroprogenitor cells to a fibrogenic (dedifferentiated) phenotype, in turn responsible for many of the degradative processes of OA [Frazier et al. 2007; Miyaki et al. 2009; Pais et al. 2010; Plaas et al. 2011]. This hypothesis arises from the results obtained by gene expression analysis, histological observation and immunohistochemical characterization performed in OA cartilage [Matthews et al. 2004; Yuan et al. 2004; Aigner et al. 2006; Barley et al. 2010; Brew et al. 2010; Wei et al. 2010]. Indeed, different molecular biology studies highlighted a significant increase in collagen type I or III gene expression, but no relevant elevation has been found for collagen type X gene expression [Aigner et al. 2006; Brew et al. 2010; Wei et al. 2010], supporting the hypothesis that OA chondrocytes may acquire a ‘fibrogenic’ phenotype. Further evidence has been reported by histological observations concerning the presence of foci and plaques of a new pannus-like fibrous tissue over the OA cartilage surface, frequently also described as ‘reparative tissue’ or ‘fibrocartilage’, particularly in post-injury OA [Shibakawa et al. 2003; Yuan et al. 2004; Barley et al. 2010]. Imunohistochemical characterization has shown that OA pannus cells express both collagen type I and type II [Shibakawa et al. 2003; Yuan et al. 2004; Barley et al. 2010]. Pannus-like tissue in OA appears to be different from invasive and aggressive synovial tissue, defined as ‘pannus’, observed on the articular surface in rheumatoid arthritis (RA). Indeed, OA pannus-like tissue is characterized by focal distribution, a limited cell density, absence of lymphocyte follicles and presence of a very limited number of macrophages [Shibakawa et al. 2003; Yuan et al. 2004; Barley et al. 2010]. Similarly to RA, pannus-like tissue in OA may acquire catabolic property, particularly in late-stage of the disease, being able to produce IL-1β and MMPs (MMP1, -3, -13) [Shibakawa et al. 2003; Yuan et al. 2004], thus contributing to OA cartilage degradation. OA is a focal disorder, hence populations of metabolically different chondrocytes may coexist in different zones or areas of the involved cartilage. Thus, hypertrophic chondrocytes and fibrogenic ones may operate at the same time (Figure 2).

Another mechanism of ALK1/ALK5 balance regulation has recently been clarified by a team from Rush University in Chicago [Li et al. 2011]. In a mouse model of OA, the authors found that cartilage damage was associated with the overgrowth of fibrous tissue coming from synovium, periosteum and meniscal attachments. This disease feature was abrogated in ADAMTS5 knockout mice. The authors suggest a predominant role of ADAMTS5 (in mice) in shifting the ALK1/ALK5 balance to the ALK5/Smad2–3 dependent fibrogenic phenotype. This hypothesis has been supported by the finding obtained by a subsequent study on mechanism of dermal repair in ADAMTS5 knockout mice [Velasco et al. 2011]. In this work, a model for regulation of fibroblast response to TGFβ signal has been provided. The model highlights the relevant role of cellular/ECM interactions in driving TGFβ cellular response. Indeed, secreted aggrecan and hyaluronan form an aggregated complex that associates with hyaluronan cell-surface receptor (CD44). In the presence of ADAMTS5, aggrecan removal alters the pericellular matrix environment and this modification may promote preferential TGFβ signaling through ALK5/Smad2–3 activation, favoring a shift towards a ‘fibrogenic response’. In the absence of ADAMTS5, aggrecan-rich pericellular matrix prevents ALK5 binding by TGFβ, which associates instead with ALK1, inducing Smad1–5–8 phosphorylation [Velasco et al. 2011]. The key role of cell–matrix interaction in TGFβ signaling has been further highlighted by additional findings obtained in CD44/ADAMTS5 double knockout mice [Velasco et al. 2011] and in primary culture of murine chondrocytes treated with retinoic acid, in order to degrade pericellular matrix component (D. Gorki and A. Plaas, unpublished data described by Plaas and colleagues) [Plaas et al. 2011]. These conditions, affecting CD44–hyaluronan–aggrecan complex, lead to a restoration of fibrogenic signaling via the ALK5/Smad2–3 pathway. The above-mentioned studies underline the complexity of the various actions of TGFβ in cartilage homeostasis and OA development. Much of the data from mouse studies can be applied to human pathology with caution. Nonetheless, during OA development (and in ageing) chondrocytes are under the influence of simultaneous series of stimuli which probably also have reciprocal opposing effects, the net sum of which determine the final metabolic response (Figure 2).

BMPs, together with TGFβ and several other signaling pathways, are involved in both chondrogenesis and cartilage formation and in osteoblastogenesis and bone formation (reviewed by Nishimura and colleagues) [Nishimura et al. 2012]. These activities are carried out by regulating Smad1–5–8 and Smad4, which are critical transcription regulators. During endochondral ossification, which also takes place in OA osteophyte formation, BMP2 controls chondrogenesis through the transcriptional factor SRY-related high-mobility group (HMG) box transcription factor 9 (SOX-9) and subsequently osteogenesis through the transcriptional factor RUNX2 (hypertrophic differentiation of chondrocytes) [Lee et al. 2000] and Osterix (calcification of cartilage and bone formation) [Matsubara et al. 2008]. Negative feedback regulation of osteoblast differentiation is carried out by Smad6 [Nishimura et al. 2012]. Among BMPs synthesized by human chondrocytes, BMP7 has been shown to have both anabolic and anticatabolic effect on cartilage, being able to induce ECM synthesis by chondrocytes and to counteract the catabolic effect induced on chondrocytes by IL-1, IL-6 and fibronectin fragments [Flechtenmacher et al. 1996; Huch et al. 1997; Koepp et al. 1999; Fan et al. 2004; Nishida et al. 2004]. BMP7 is already marketed as an agent able to speed up bone healing after fracture [Kanakaris et al. 2008].

FGFs constitute another family of relevant growth factors for cartilage development and homeostasis [Ellman et al. 2008]. In particular, two members of this family, FGF-2 and FGF-18, have been implicated in the regulation of cartilage remodeling [Ellman et al. 2008; Fortier et al. 2011]. Studies on FGF-2 have showed conflicting results concerning its effects on chondrocytes. Different animal model studies have shown a regenerative effect on cartilage defect treated with FGF-2 [Hiraide et al. 2005; Inoue et al. 2006; Kaul et al. 2006; Deng et al. 2007]. These findings might result from the strong mitogenic effect exerted on chondrocytes by this growth factor [Stewart et al. 2007], which appear to be ineffective in inducing a successful regeneration of cartilage ECM [Stewart et al. 2007]. Consistent with the antianabolic and catabolic effect of FGF-2, studies performed on human chondrocytes [Loeser et al. 2005; Im et al. 2007, 2008; Muddasani et al. 2007] showed the ability of FGF-2 to upregulate MMP13, ADAMTS4, -5, to inhibit cartilage matrix production, to antagonize activity of anabolic factors (e.g. BMP7) and to stimulate proinflammatory cytokines, such as IL-1 and TNF.

In contrast to FGF-2, the anabolic effect of FGF-18 in cartilage has been well established [Ellsworth et al. 2002; Liu et al. 2002; Davidson et al. 2005; Moore et al. 2005]. In porcine and human cartilage, FGF-18 has been shown to act as an inducer of cell proliferation and ECM synthesis [Ellsworth et al. 2002]. Furthermore, in a rat OA model, FGF-18 intra-articular injections induce increasing cartilage formation and reduce cartilage degeneration scores [Moore et al. 2005].

Hypoxia-inducible factors

Healthy articular cartilage is typically an avascular tissue. Chondrocyte survival in this hypoxic condition requires activation of adaptive strategies that allow functionality to be maintained. Hypoxic response is mainly mediated by hypoxic inducible factor (HIF) family members (HIF-1α, 2α, 3α). The first member of this transcription factor family to be identified was HIF-1α [Wang et al. 1995]. In the presence of normal levels of oxygen, HIF-1α is hydroxilated and degradated. Conversely, under hypoxic conditions, hydroxylation appears to be inhibited, leaving HIF-1α free to heterodimerize with the constitutive HIF-1β unit. This complex binds specific target gene consensus sequences and promotes their transcription [Semenza, 2000].

In cartilage, HIF-1α has been shown to be essential for chondrocyte growth arrest and survival [Schipani et al. 2001] and appears to be involved in the regulation of angiogenetic factor expression, mainly VEGF [Forsythe et al. 1996; Schipani et al. 2001; Murata et al. 2008], thus modulating an essential step in endochondral bone formation. Furthermore, HIF-1α contributes to the maintenance of ECM homeostasis, inducing the gene expression of its two main components: collagen II and aggrecan [Duval et al. 2009].

In OA cartilage, HIF-1α expression has been reported by several studies [Aigner et al. 2001; Stokes et al. 2002; Coimbra et al. 2004; Pfander et al. 2005; Yudoh et al. 2005]. Aigner and coworkers have shown an increased transcription of HIF-1α in OA cartilage compared with normal samples, particularly in the late stage of the disease [Aigner et al. 2001]. Consistent with this evidence, subsequent studies reported a growing number of HIF-1α-positive chondrocytes during OA progression [Pfander et al. 2005] and a higher expression of HIF-1α mRNA in degenerated cartilage compared with uninjured cartilage [Yudoh et al. 2005].

In addition to hypoxia, HIF-1α expression can be upregulated by other factors, including inflammatory cytokines (IL-1 and TNF-β), reactive oxygen species and mechanical loading [Hellwig-Burgel et al. 1999; Haddad and Land, 2001; Haddad, 2002; Chang et al. 2003; Jung et al. 2003; Petersen et al. 2004; Pufe et al. 2004; Yudoh et al. 2005], which are all recognized as key players in OA joint damage.

Since HIF-1α has a pivotal role in chondrocyte survival and in supporting cartilage homeostasis, these findings have led to consider HIF-1α as a stress-inducible responder, rather than solely as a HIF, which potentially acts to maintain the chondroprotective functions challenged by the detrimental conditions occurring in the OA joint environment. This hypothesis has been supported by further experimental evidence that underlines the central importance of HIF-1 in supporting the following: chondrocyte energy generation via increased glucose uptake and by regulating the activity of glycolytic enzymes [Mobasheri et al. 2005; Pfander et al. 2005; Yudoh et al. 2005]; cartilage matrix synthesis [Duval et al. 2009]; and the activation of a protective mechanism enabling the prevention of chondrocyte cell death induced by IL-1β [Yudoh et al. 2005].

Overall, these findings characterize HIF-1α as a key factor for chondrocyte survival promoting compensatory mechanisms in response to catabolic modifications of OA cartilage.

Another member of the HIF family, HIF-2α, has subsequently been recognized. HIF-2α and HIF-1α have extensive structural homology and both are regulated by similar mechanisms [Ema et al. 1997]. Despite these similarities, these transcriptional factors have been shown to have specific and distinct functions under physiological and pathological conditions [Wang et al. 2005; Ratcliffe, 2007; Patel and Simon, 2008].

HIF-2α appears to be highly expressed in OA cartilage and is strongly implicated in OA by inducing mechanisms leading to cartilage breakdown and endochondral bone formation. The role of HIF-2α as catabolic inducer of OA cartilage destruction has been clearly demonstrated by the studies of Yang and colleagues [Yang et al. 2010]. They showed that the adenoviral overexpression of HIF-2α causes progressive cartilage damage directly upregulating the expression of a set of degradative enzymes, including MMP1, MMP3, MMP12, MMP13, ADAMTS4. Further recent evidence has broadened the involvement of HIF-2α in OA by identifying this transcription factor as an extensive regulator of endochondral bone formation [Saito et al. 2010].

Endochondral ossification is a fundamental pathway in OA progression leading to osteophyte formation, considered one of the typical OA outcomes [Kawaguchi, 2008] (Figure 1). This process is marked by sequential steps, including chondrocyte hypertrophic differentiation (characterized by secretion of collagen X), cartilage degradation (via proteinases, mainly MMP13) and vascular invasion depending on angiogenic stimuli, such as VEGF [Kronenberg, 2003; Mackie et al. 2008]. HIF-2α appears to be a central regulator of all these steps, directly regulating collagen X, MMP13 and VEGF expression.

In addition, further possible transcription targets of HIF-2α related to endochondral ossification have been identified, namely RUNX2 and Indian Hh proteins [Tamiya et al. 2008; Saito et al. 2010] (see page 10).

The relevance of HIF-2 in OA has been strengthened by additional evidence from human knee joint samples showing increasing expression related to OA development, which reaches a maximum at the early and progressive stage, yet appears to decrease in the late stage. HIFs are also implicated in regulating chondrocyte autophagy. Autophagy is a primary process that degrades and removes damaged and dysfunctional intracellular organelles and molecules, thus protecting cells during stress responses [Mizushima and Klionsky, 2007; Uchiyama et al. 2008; Mizushima, 2009]. Dysfunctions of this mechanism have been found to contribute to the development of aging-related diseases [Austin, 2009; Salminen et al. 2009]. This process appears to be compromised in OA cartilage. The role of HIF-1α and HIF-2α in controlling chondrocyte autophagy has been shown by studies performed by Bohensky and coworkers [Bohensky et al. 2009], demonstrating the ability of HIF-2α to counterbalance the capacity of HIF-1α to accelerate chondrocyte autophagy functions. Seeing that HIF-2α expression decreases in the late stage of OA, as previously reported, the downregulation of this factor may enhance autophagy of OA chondrocytes [Bohensky et al. 2009]. Consequently, the control of cartilage homeostasis is potentially related to balance of HIF-1α/HIF-2α activities, as HIF-1α function drives to maintain cartilage integrity, whereas HIF-2α strongly favors cartilage degradation and endochondral ossification. The shift from HIF-1α to HIF-2α expression may be a critical step in OA pathogenesis [Husa et al. 2010] and has therefore been considered as a potential candidate for a novel therapeutic strategy (Figure 2).

Role of cell–matrix interactions

Chondrocytes express several receptors, including those that sense ECM molecules, enabling the activation of cross-talking pathways that may greatly affect chondrocyte behavior. Xu and coworkers showed that increased chondrocyte expression of the DDR2 represents a key event in OA pathogenesis [Xu et al. 2007]. DDR2 preferentially binds type II collagen, but this interaction in physiological conditions is prevented by an intact/uninjured proteoglycan network of chondrocyte pericellular matrix (territorial region).

In OA, early cartilage damage results in depletion of proteoglycans, which leads to exposure of collagen II and allows the interaction with chondrocytes via DDR2 binding. DDR2 activation induces synthesis of MMP13 and upregulation of DDR2 itself, thus contributing to amplification of OA cartilage loss [Xu et al. 2005, 2011]. The relevance of this mechanism in OA pathology has been demonstrated by studies carried out using both animal models and human cartilage [Xu et al. 2005, 2007, 2010].

Additional recent findings have strengthened the role of DDR2 in the OA process. DDR2 appears to mediate the collagen-II-dependent release of IL-6, another catabolic cytokine, in primary human chondrocytes [Klatt et al. 2009]. Furthermore, a role of DDR2 in osteoblast differentiation and chondrocyte maturation via modulation of RUNX2 activation has been demonstrated in in vitro studies [Zhang et al. 2011].

Syndecan 4 is a transmembrane sulfate proteoglycan that interacts with several ECM molecules, growth factors and cytokines [Tkachenko et al. 2005]. Over the last few years, increasing evidence has been accumulating concerning the function of syndecans in modulating cellular activities by controlling the interaction of ECM components and soluble ligands with the cell surface [Tkachenko et al. 2005]. Syndecan expression appears to be modulated during development and cellular differentiation, as well as in association with pathological tissue changes [Tkachenko et al. 2005]. In cartilage, syndecan 4 appears to be specifically induced in hypertrophic chondrocytes and its expression is elevated in OA cartilage in human and animal models, and in OA animal models [Barre et al. 2000; Echtermeyer et al. 2009]. In particular, syndecan 4 has been shown to be crucial for ADAMTS5 activity [Echtermeyer et al. 2009], by controlling its activation through either direct or indirect interactions [Echtermeyer et al. 2009]. These results were further underlined by the evidence that syndecan 4 knockout mice exhibited reduced cartilage loss in a surgical OA model [Echtermeyer et al. 2009] and that mice treated by local intra-articular injections of syndecan-4-specific antibodies were similarly protected against ADAMTS5 mediated cartilage loss [Echtermeyer et al. 2009] (Figure 2).

These data highlight the critical role of chondrocyte–ECM interaction in maintaining cartilage integrity, stressing the importance of early structural alteration of ECM articular cartilage in inducing mechanisms sustaining OA progression.

Wnt/frizzled receptor/β-catenin pathway

Wnts constitute a large family of 19 secreted glycoproteins implicated in several processes in the development, growth and homeostasis of different tissues and organs, including joints, bone and cartilage. Cellular response to Wnt proteins occurs via their binding to frizzled (FZD) receptors, which can trigger multiple signal cascades, the best characterized of which is the β-catenin-dependent canonical Wnt pathway. In this pathway, in the absence of FZD engagement, β catenin is phosphorylated and sequestered in a destruction complex [Logan and Nusse, 2004; Monroe et al. 2012]. Conversely, when Wnts bind FZD and lipoprotein receptor-related protein 5/6 (LRP5/6) (that act as coreceptors), the destruction complex is prevented and β catenin accumulates in the cell, migrates to the nucleus and modulates the transcription of target genes.

The Wnt/β-catenin pathway is tightly regulated by several natural extracellular Wnt antagonists, namely secreted frizzled related proteins (sFRPs), Wnt inhibitory factors (Wifs), Dickkopf (Dkk) factors (1–4) and sclerostin. Signaling cascade inhibition essentially occurs through two different mechanisms. sFRPs and Wifs directly bind Wnts, thus interfering with the receptor interactions. Conversely, Dkk and sclerostin bind to the Wnt coreceptor, LRP5/6, preventing Wnt/receptor interaction [Logan and Nusse, 2004; Monroe et al. 2012].

Wnts have been extensively recognized as key regulators of bone and cartilage homeostasis [Chun et al. 2008; Nalesso et al. 2011; Yasuhara et al. 2011; Monroe et al. 2012] and this evidence has driven the researchers to investigate the potential association between Wnt modifications and OA. Enhanced activation of canonical Wnt pathways in OA human cartilage [Dell’accio et al. 2008] and in injured cartilage [Dell’accio et al. 2006, 2008; Eltawil et al. 2009] has been reported. These findings have been confirmed by further studies performed on animal models that report a relationship between β-catenin signaling activation and OA-like phenotype [Zhu et al. 2009; Miclea et al. 2011; Lodewyckx et al. 2012].

Furthermore, genetic studies have identified some components of the Wnt cascade as candidate genes associated with OA. Loughlin and colleagues showed that a single nucleotide polymorphism in the sFRP3 gene, a Wnt antagonist, appears to be linked to an increased risk for hip OA [Loughlin et al. 2004], while an association between a single LRP5 gene polymorphism and spine OA has been reported [Urano et al. 2007].

The relevance of Dkk-1 as key regulator of bone remodeling has been elegantly demonstrated by Diarra and coworkers [Diarra et al. 2007]. Blockade of Dkk-1, abrogating Dkk-1-mediated Wnt suppression, reverts from the bone-destructive pattern in a mouse model of RA and induces bone formation and osteophyte growth, thus resembling the bone-forming pattern of OA [Diarra et al. 2007].

Wnt signaling antagonists are also considered as potential biomarkers of OA progression. Elevated Dkk-1 serum levels have been shown to be associated with reduced progression of hip OA in a white population [Lane et al. 2007]. Similarly, in patients with knee OA, a study performed by Honsawek and coworkers reported an inverse correlation between Dkk-1 levels in plasma and synovial fluid and the radiographic severity [Honsawek et al. 2010]. In addition, high serum concentrations of sFRP-3 showed a tendency to be associated with lower risk of developing hip OA [Lane et al. 2007].

Extensive studies on Wnt signaling related to bone and cartilage physiology and pathology have undoubtedly led to accumulation of evidence consistent with a critical role in OA pathogenesis. Nevertheless, considering the complexity of Wnt downstream effects and the multiple interplay with several other pathways, further research is needed to better characterize the specific role and relevance of single Wnt agents/antagonists in bone, cartilage and the osteochondral junction. In the new perspective of the role of complement in OA, the very recent demonstration that complement C1q activates canonical Wnt signaling is noteworthy [Naito et al. 2012].

Hedgehog/smoothened pathway

Another signaling pathway strongly involved in both chondrogenesis and chondrocyte proliferation and differentiation in the growth plate during development [Lanske et al. 1996; Vortkamp et al. 1996; Chung et al. 2001; Kobayashi et al. 2005; Maeda et al. 2007; Mak et al. 2008] and in OA pathogenesis [Lin et al. 2009; Ruiz-Heiland et al. 2012] is the Hh pathway. The ligands of this pathway [Indian Hh (Ihh) and Sonic Hh] are expressed by chondrocytes in response to mechanical stress (Ihh is a mechanoresponsive gene) [Ng et al. 2006]. These ligands engage Patched 1 receptor (Ptch1) terminating, and in this way, the inhibitory action of Ptch1 on another cell surface receptor called Smoothened (Smo) [Wilson and Chuang, 2010]. Hh ligand binding to Ptch1 causes Smo localization and accumulation in the chondrocyte cilia [Corbit et al. 2005] where it activates the effector proteins of the Hh signaling pathway: the final results being the expression of RUNX2 (a master regulator of chondrocyte hypertrophic differentiation) and, indirectly, the expression of the aggrecanase ADAMTS5 [Lin et al. 2009]. Mechanical stress on chondrocytes also elicits the expression of parathyroid hormone-like hormone, which inhibits the Hh–Smo signaling pathway [Chen et al. 2008].

It is of note that Hh signaling targets (Ptch1, ADAMTS5, MMP13) are highly expressed both in human and experimental OA cartilage [Lin et al. 2009], and the pharmacological blockade of Smo specifically inhibits osteophyte formation in mouse models of OA without having any effect on inflammation and without stimulating bone destruction [Lin et al. 2009; Ruiz-Heiland et al. 2012].

Therapeutic implications: future treatment possibilities

Present-day therapy of OA is directed at pain relief [painkillers and nonsteroidal anti-inflammatory drugs (NSAIDs)] and surgery (the possibility of failed joint replacement) [Zhang et al. 2008b]. While over the last 10–12 years several effective biotechnological drugs have been approved and successfully marketed for the treatment of inflammatory arthritis, no approved disease-modifying OA drugs (DMOADs) are currently available on the market [Le Graverand-Gastineau, 2010].

There are several underlying reasons at the basis of this discrepancy. Here we would like to underline that whilst investigation into RA pathogenesis has brought to light how several inflammatory molecules (mainly cytokines) exert a pathogenetic role without playing a relevant role in the physiology of joint tissues, many of the aforementioned molecular pathways involved in ageing and OA-related joint changes are also essential for cartilage and bone homeostasis in the healthy functioning of the joints. Therefore, the consequences of molecular pathway inhibition brought about by therapeutic manipulation may have either no significant effect or disappointing side effects.

Therefore, as OA pathogenic mechanisms are more complicated and intertwined and less well understood than those of inflammatory arthritis, the difficulty in identifying safe and relevant targets for molecular OA therapies persists.

Here we review the principal molecules currently under clinical investigation: most of the trials are in phase II or III and involve knee and, to lesser extent, hand OA. At the time of publication, there were more than 500 ongoing trials worldwide (http://www.clinicaltrials.gov/). Many of them are devoted to the evaluation of painkillers/NSAIDs or nutraceuticals: we will only concentrate on studies concerning the target molecules involved in pathogenic pathways.

Inflammatory cytokines and related signaling pathways

Since the role of proinflammatory cytokines has been recognized in promoting chondrocyte catabolic phenotype and, furthermore, given the success of anticytokine therapy in inflammatory arthritis, several studies have been carried out and a number of trials are still in progress in patients with OA which use antagonists of the above-mentioned cytokines. IL-1β exerts its various actions by engaging type I IL-1 receptor (while type II receptor is ineffective in signal transduction and therefore is a ‘decoy’ receptor) [Ricote et al. 2004] then the IL-1/IL-1R complex recruits the IL-1 receptor accessory protein (IL-1RAcP) to form a high-affinity complex, which is able to initiate intracellular signaling cascade [Wesche et al. 1997]. Thus IL-1 action can be blocked by targeting different molecules in different ways: using IL-1 receptor antagonist (IL-1Ra), antibodies to IL-1 and to IL-1Rs or the soluble (membrane shedded) type I IL-1R/IL-1RAcP complex.

Anakinra

Investigations into the effect of anakinra (IL-1Ra: a registered drug for RA) using intra-articular injections have produced conflicting results in patients with knee OA (in whom a single articular injection was administered) [Chevalier et al. 2009], while some good preliminary results have been obtained (regarding pain and global disability) with daily subcutaneous injections for 3 months in patients with erosive hand OA [Bacconnier et al. 2009].

Rilonacept

Rilonacept, a drug that combines the extracellular domains of both the type I IL-1R and IL-1RAcP (registered for some hereditary chronic inflammatory disorders) has failed to show substantial benefit in a phase II clinical trial in RA. No data are available for patients with OA [McDermott, 2009; Geyer and Muller-Ladner, 2010].

Canakinumab

Canakinumab, a fully human monoclonal antibody to IL-1β, has been registered for the treatment of cryopyrin-associated periodic syndromes and is under evaluation for RA and other inflammatory/autoimmune disorders [Geyer and Muller-Ladner, 2010]. Again no data are available in patients with OA.

AMG 108

A different human monoclonal antibody (AMG 108) that binds type I IL-1R (thus blocking both IL-1α and IL-1β) is under evaluation in patients with knee OA (phase II clinical trial terminated) [Cohen et al. 2011].

TNF inhibitors

TNF inhibitors are the most successful ‘biological’ treatment in arthritides. It is of note that they exert a convincing and robust protection against structural damage in RA. Therefore, they have been evaluated in the inflammatory/erosive subset of hand OA. In this subset, severe structural damage of the distal and proximal interphalangeal (DIP, PIP) joints is characterized by central bone erosion with cartilage thinning and eventual ankylosis or deviation. In addition, inflammatory features of DIP and PIP joints can be detected clinically and by ultrasound and magnetic resonance imaging [Kloppenburg and Kwok, 2012].

Unfortunately, a small open-label 3-month therapy study with fully humanized anti-TNF antibody (adalimumab) did not demonstrate significant clinical improvement [Magnano et al. 2007]. Preliminary data from a phase II clinical trial showed a trend toward reduced progression of erosive damage [Verbruggen et al. 2012]. A phase III trial is under way [ClinicalTrials.gov identifier: NCT00597623].

Infliximab is the first partially humanized antibody to be registered for RA treatment: a 12-month pilot study has shown significant effects on pain and radiological score progression in patients with erosive hand OA [Fioravanti et al. 2009]. More interestingly, the data from a multicenter, randomized clinical trial designed to evaluate different treatment modalities in patients with RA has recently been analyzed for radiological hand OA scores. Patients treated with infliximab had significantly reduced incidence of secondary OA in PIP joints [Guler-Yuksel et al. 2010].

Finally, despite the probable pathogenic role of IL-6 in OA as well as in RA, no data about the efficacy of IL-6 inhibitors in patients with OA are available.

Inflammatory cytokine intracellular signaling cascades

Several inflammatory cytokine intracellular signaling cascades operate in the various joint tissues. Four of them have a prominent role: the mitogen-activated protein kinase (MAPK) family, the nuclear factor κB (NFκB) cascade, the extracellular signal-regulated protein kinases (ERKs), the c-Jun N-terminal kinases or stress-activated protein kinases (JNK/SAPK) and the p38 family [Goldring and Otero, 2011].

A p38 MAPK inhibitor (PH-797804) is in a phase II clinical trial attempting to evaluate its efficacy in the treatment of pain in patients with a knee OA flare up [ClinicalTrials.gov identifier: NCT01102660].

The NFκB cascade is being recognized as a central pathway involved in several pathophysiological processes in OA: proinflammatory responses to cytokines and cytokine synthesis in chondrocytes, osteoblasts, synovial fibroblasts, the regulation of chondrocyte differentiation process to hypertrophic-like stage. The main problem associated with the therapeutic manipulation of NFκB and other signaling cascades is the necessity to avoid any undesirable effects since these cascades are involved in a variety of physiological intracellular signaling pathways [Marcu et al. 2010].

A phase II clinical trial evaluating the efficacy of A3 adenosine receptor (A3aR) agonist in patients with knee OA is underway [ClinicalTrials.gov identifier: NCT00837291]. A3aR engagement is able to induce anti-inflammatory and antiapoptotic effects. Many of the actions are secondary to a dysregulation of the NFκB pathway and the secondary downregulation of TNFα expression [Bar-Yehuda et al. 2009].

IKKα and IKKβ are essential kinases that activate NFκB transcription factors. A phase I IKK inhibitor study has been completed [ClinicalTrials.gov identifier: NCT01463488].

Protease inhibitors and nitric oxide

Protease inhibitor development caused great excitement and there were high hopes that an effective molecular DMOAD treatment had finally been found. Several chemical families of protease inhibitors have been synthesized, yet results from the first clinical trials have highlighted either musculoskeletal (joint stiffness and joint fibrosis) and systemic (anemia, liver toxicity) side effects or lack of efficacy [Krzeski et al. 2007]. These disappointing results are probably due to the above-mentioned role played by MMPs and ADAMTSs in the healthy functioning of the joints.

The semisynthetic homologues of the antibiotic, tetracycline, exert a potent MMP inhibitory action. The results of a randomized double-blind trial, which utilized doxycycline in the treatment of patients with knee OA, documented a reduction in the rate of joint space narrowing without joint pain improvement [Brandt and Mazzuca, 2006].

Recently, new potent selective inhibitors of MMP13 (the most potent protease for collagen II and proteoglycans) and ADAMTS4 and -5 (aggrecanases) have been identified and studies utilizing these selective inhibitors are eagerly awaited. A phase I trial using an ADAMTS4/-5 inhibitor has been completed [ClinicalTrials.gov identifier: NCT00454298]. The findings recently reported concerning the key role of syndecan 4 (a cartilage ECM molecule produced by OA chondrocytes) in controlling ADAMTS5 activation (as already described in the section on the role of cell–matrix interactions) indicate syndican 4 as a promising candidate for a potential alternative therapeutic approach targeting ADAMTS5 inhibition [Echtermeyer et al. 2009].

NO and its metabolites are a double-edge sword in the regulation of metabolic processes: constitutive NO production is essential for many physiological processes but its production secondary to inducible NO synthase (iNOS) is toxic to cells and damages ECM. iNOS is expressed after cell activation by cytokines and other inflammatory signals. OA chondrocytes express and produce a greater amount of iNOS and NO than normal or rheumatoid chondrocytes [Melchiorri et al. 1998; Mazzetti et al. 2001]. NO has various deleterious effects on chondrocyte metabolism, MMP and cytokine expression and ECM macromolecule production [Henrotin et al. 2005]. A phase III trial with an inducible NO synthase inhibitor has been completed [ClinicalTrials.gov identifier: NCT00565812].

Cartilage repair and bone formation

Cartilage and bone growth factor use in therapy, whilst offering promising chances in tissue repair, also reveals a possible setback due to the short half life of most molecules and the progressive age-related unresponsiveness of chondrocytes to growth factor stimulation.

At present, two growth factors are being evaluated in phase I and II clinical trials in patients with knee OA: BMP7 also known as osteogenic protein 1 [Hunter et al. 2010] and FGF-18 [ClinicalTrials.gov identifier: NCT00911469 and NCT01033994].

Very recently, using a stem cell and image-based, high-throughput screen approach, a new small molecule called kartogenin was found able to induce multipotent stem cell differentiation into chondrocytes for cartilage regeneration. Testing the molecule in two different mouse models of OA, it was shown that kartogenin was able to initiate the process in cytosol leading to the translocation of core-binding factor β into the nucleus where it was able to dimerize with the DNA transcription factor RUNX1. This dimer was able to activate chondrogenesis genes (aggrecan, type II collagen, tissue inhibitors of metalloproteinases), thus providing a cartilage protective effect in OA mouse models [Johnson et al. 2012]

Conclusion

With the continuous progression of our knowledge on the molecular pathways involved in cartilage and bone changes in OA, an increasing number of potentially effective target candidates for OA therapy are entering the first phase of evaluation: in vitro cell culture systems probing and subsequent in vivo animal experiments. A wealth of networking molecules (and their agonists or inhibitors) are already under scrutiny in clinical trials to ascertain their possible safe use in an attempt to identify molecules active in slowing or halting OA progression and reducing joint pain.