Wnt signalling in the articular cartilage: A matter of balance

Degradation of the articular cartilage is a hallmark of osteoarthritis, a progressive and chronic musculoskeletal condition, affecting millions of people worldwide. The activation of several signalling cascades is altered during disease development: among them, the Wnt signalling plays a pivotal role in the maintenance of tissue homeostasis. Increasing evidence is showing that its activation needs to be maintained within a certain range to avoid the triggering of degenerative mechanisms. In this review, we summarise our current knowledge about how a balanced activation of the Wnt signalling is maintained in the articular cartilage, with a particular focus on receptor‐mediated mechanisms.


| OSTEOARTHRITIS
Osteoarthritis (OA) is a chronic and degenerative musculoskeletal condition, affecting 16% of the worldwide population over the age of 15. 1 From a clinical point of view, OA is characterised by progressive and irreversible degeneration of the articular cartilage (AC), inflammation of the synovial lining and abnormal subchondral bone remodelling. 2 Osteoarthritis is a multifactorial disease: ageing, genetic predisposition, history of trauma and co-morbidities, such as diabetes or hypercholesterolemia, have been associated with the onset of OA or the exacerbation of its symptoms and clinical features. [3][4][5] No pharmacological treatment can revert the progression of the disease, and pain relief is the main therapeutic option for patients. When this does not suffice anymore and joint mobility is completely compromised, the patient is offered a joint replacement surgery. This option is not devoid of potential post-operative complications, especially in the most elderly patients. Furthermore, in 30% of patients, joint replacement with a prosthesis does not resolve chronic pain, thereby only partially improving patients' quality of life. 6

| THE ARTICULAR CARTILAGE
The AC is the connective tissue covering the edges of the bones in the diarthrodial joints. The tissue allows a smooth diarthrosis being elastic and conferring resistance to compression upon the joint. Its main constituents are a thick extracellular matrix (ECM), made of proteoglycans and collagens, and only one cell type, the chondrocyte. Proteoglycans are negatively charged molecules providing the AC with its high resistance to compressive loads by attracting water into the tissue. Collagens are organised in fibres conferring the AC tensile strength. 7 Chondrocytes are responsible for the turnover of ECM components by expressing and secreting both ECM components and the enzymes responsible for their degradation, such as metalloproteases (MMPs). As the AC is nonvascularised and noninnervated, the tissue relies on the subchondral bone and the synovial fluid to receive nutrients and to maintain the necessary tissue lubrication. 8,9 3 | DEGRADATION OF THE

ARTICULAR CARTILAGE IN OSTEOARTHRITIS
During the development of OA, the metabolic balance in the AC skews towards catabolism. The expression and activity of MMPs, such as MMP3 and MMP13, and A Disintegrin And Metalloproteinase with Thrombospondin Motifs (ADAMTs) enzymes, such as ADAMTS4 and ADAMTS5 are strongly upregulated in the AC along disease progression. 10 Conversely, the expression of some tissue inhibitors of metalloproteases (TIMPs) is reduced. 11 Metalloproteases cut and degrade proteoglycans and collagens, leading to tissue erosion and to the onset of the disease.
Increasing evidence suggests that OA is an abnormal recapitulation of osteogenesis. Bones are derived from the cartilage anlagen: chondrocytes in the anlage undergo a coordinated process of proliferation and maturation which culminates in hypertrophic maturation. Hypertrophic chondrocytes (HC) can then indirectly and directly contribute to bone formation. In the first case, HC undergo terminal differentiation, driving the secretion of mineralised cartilage and blood vessels invasion, after which they die through apoptosis. Vascular invasion allows the arrival of blood vessel-associated pericytes, osteoclasts and bone progenitor cells in the anlage, promoting bone formation. Hypertrophic chondrocytes can also transdifferentiate into osteoblasts, therefore directly generating bone tissue. [12][13][14] Differently, synovial joints originate from the condensation of mesenchymal cells in structures called interzones. Interzone cells directly differentiate into articular chondrocytes, which do not undergo hypertrophic differentiation. 15 During the development of OA, deregulation of several signalling cascades can lead articular chondrocytes to restart the differentiation process from where it had been interrupted during joint development and undergo hypertrophic maturation and calcification. 16 These phenomena contribute to the loss of elasticity and resistance to compression of the ECM, leading to tissue destruction and ultimately compromising joint mobility and inducing pain. 16

| THE WNT SIGNALLING
Deregulation of the Wnt signalling plays a major role in the development and progression of OA. Wnts are a family of 19 glycosylated and lipid-modified ligands activating a complex network of signalling pathways. To activate any of these pathways, Wnts engage with the Frizzled (FZD) family of receptors. The Wnt/FZD complexes associate with different co-receptors, leading to the initiation of specific intracellular signalling. The association of FZDs with the Low-density Lipoprotein-Related Protein 5 or 6 (LRP5/6) receptors leads to the destabilisation of a multiprotein 'destruction complex' which includes among its members Axin2, Adenomatous polyposis coli (APC), Glycogen Synthase Kinase 3 (GSK3), Casein Kinase 1 (CK1), protein phosphatase 1A (PP2A) and the E3-ubiquitinin ligase beta-transducin repeat-containing protein (β-TrCp). In the absence of an activating stimulus, the complex traps βcatenin within the cytoplasm and addresses it for ubiquitin-mediated degradation. 17 The disaggregation of the destruction complex promotes the accumulation of βcatenin which can then translocate to the nucleus where it binds to components of the T-cell factor/lymphoid enhancer factor (TCF/LEF) family of proteins, promoting the transcription of cell and context-specific genes. 18 Among these, Axin2 seems to be a well-preserved target across cell types and species. 19 Frizzled, however, have also been reported to associate with other co-receptors, such as the Receptor Tyrosine Kinase-like Orphan Receptor 2 (ROR2) and the Related to Receptor Tyrosine Kinase (RYK), 20,21 promoting the activation of less characterised signalling cascades. These have been grouped under the name of βcatenin-independent pathways. Their triggering does not promote the accumulation of βcatenin but leads instead to the activation of other downstream targets such as Calcium Calmodulin Kinase II (CaMKII), 22,23 Protein Kinase A (PKA) and Protein Kinase C (PKC). 24,25

| THE WNT SIGNALLING IN OA
The Wnt/β-catenin-dependent signalling is the most characterised of the Wnt pathways at molecular level. The activation of this pathway is required and sufficient for synovial joint formation; 26 however, it needs to be tightly regulated in a time-and space-dependent manner to allow the integrity of the tissue to be maintained in the adult life. 27,28 Overactivation of the Wnt/β-catenin signalling has indeed been associated with degradation of the AC in OA. Polymorphisms in genes expressing components and/or modulators of this pathway have been linked to a higher risk of developing OA. The Arg200Trp and the Arg324Gly substitutions in the secreted Frizzled-Related Protein 3 (sFRP3), a soluble scavenger of Wnt ligands, lead to lossof-function mutations which have been reported to be more frequent in patients with clinical signs of OA. [29][30][31] sFRP3-deficient mice developed more severe cartilage destruction both in a model of inflammatory arthritis and in a surgically induced OA model. 32,33 In addition, the expression of Dikkopf1 (Dkk1), an inhibitor of the Wnt/ βcatenin pathway, is lower in the AC of OA patients in comparison to healthy donors. 34 Increased expression of βcatenin has been detected in the AC of OA patients and of mice in which OA was surgically induced. [35][36][37] However, overactivation of the Wnt/β-catenin pathway in the tissue both through genetic manipulation or pharmacological modulation did not conclusively determine a cause-effect link between the activation of this branch of the Wnt signalling and tissue degeneration. Genetic activation of the βcatenin gene in the AC promoted OA development in mice. This was associated with increased expression of MMP13 and ADAMTS5, upregulation of hypertrophic markers such as Collagen Type X (ColX) and increased cell death, both in the temporomandibular joint (TMJ) and in the knee joint. 38,39 An additional study from Cai et al. showed that mechanical stress can promote TMJ OA via overactivation of the Wnt/β-catenin signalling. 40 However, loss of βcatenin transcriptional activity was also shown to disrupt tissue homeostasis and promote the development of OA-like lesions in the AC. Overexpression of the inhibitor of βcatenin and TCF-4 (ICAT) led to reduced chondrocyte proliferation, increased apoptosis and reduced skeletal grown after birth in mice. 41 These mice also developed spontaneous OA with ageing. 42 Downregulation of βcatenin in the superficial cells of the AC also induced downregulated expression of lubricin (PRG4), an important joint lubricant, and OA-like cartilage degeneration. Indeed, modulation of PRG4 expression at mRNA level has been shown to be linked to the activation of the Wnt/β-catenin signalling in isolated chondrocytes. 43 These data suggest that a tight regulation of the Wnt/ βcatenin signalling in space and time within the AC is required to maintain tissue homeostasis; however, the mechanisms regulating this balanced activation remain largely uncharacterised.

INDEPENDENT SIGNALLING NETWORK IN THE AC
The role of the Wnt/β-catenin-independent pathways in the AC and OA is far less characterised. Our current knowledge is mainly derived from developmental studies.
The Wnt5a/ROR2 pathway has an important homeostatic function in the musculoskeletal system. Patients affected by Robinow Syndrome, a condition characterised by several dysmorphic features associated with skeletal dysplasia, bear mutations in either Wnt5a or ROR2 genes. [44][45][46] Studies in vivo showed that Wnt5a drives limb elongation in a concentration-dependent manner during development by activating the Planar Cell Polarity (PCP) pathway in chondrocytes. 47 As for the Wnt/β-catenin pathway, the ROR2 pathway also seems to get re-activated in the AC in OA: the expression of Wnt5a and of the ROR2 targets Yes-associated protein (YAP) and connective growth factor (CTGF) are upregulated in the AC of OA patients. ROR2 blockade supported an anabolic response in the AC in an in vivo model of OA, although through Wnt-independent mechanisms. 48 The Wnt/CaMKII pathway has also been shown to have a pivotal role in bone and cartilage development. Studies in chicken and mice showed that activation of CaMKII promotes cartilage hypertrophic differentiation during limb formation. 49,50 Several studies showed that CaMKII is activated in OA, both in human and in animal models of the disease. 51,52 We have recently shown that pharmacological blockade of CaMKII can exacerbate cartilage damage in a murine model of OA. 51 This is in keeping with a recent in vitro study showing that inhibition of CaMKII can decrease the anabolic activity of bone morphogenetic proteins 2 and 4 on the synthesis of ECM components in isolated chondrocytes 53 but it is in contrast with previous work from the Saito's group showing that CaMKII can promote OA development in mice via activation of the Hes1 transcription factor. 52

| β-CATENIN-DEPENDENT AND -INDEPENDENT PATHWAYS: A MATTER OF BALANCE
What is then the role of the Wnt signalling in the AC and OA? Can it be considered as a therapeutic target for this disease? An answer to this question comes from the recent discovery of the first molecule classified as a diseasemodifying drug for the treatment of OA, SM04690, now known under the commercial name of Lorecivivint. Lorecivivint, which is currently in Phase III clinical trial, is an inhibitor of Cdc2-like kinase (CLK1), an enzymeregulating splicing events, and blocks the activation of the Wnt signalling at transcriptional level. Nonetheless, the drug also inhibits the Dual-specificity tyrosine phosphorylation-regulated kinase 1A, which modulates the inflammatory response. 54 The discovery of this drug highlights how our understanding of the molecular events, intrinsic and extrinsic to the pathway, is still relatively poor and that the lack of this information is potentially dampening the development of additional therapeutics. Integrative approaches, aimed at investigating the simultaneous modulation of multiple molecules, will therefore be required to understand how tissue homeostasis is maintained and could be re-established once disrupted.
Our recent work contributed to unravel the complexity of the regulation of the Wnt signalling in the AC. We showed that a single ligand, Wnt3a, could promote both intracellular accumulation of βcatenin and intracellular activation of CaMKII in the AC, thus simultaneously activating βcatenin-dependent and -independent branches of the Wnt signalling. Critically, while the first increases proportionally to Wnt3a concentration and drives proliferation, the second response is higher when chondrocytes are stimulated with low concentrations of Wnt3a and drives phenotypic changes. 23 These two physiological outcomes are in a steady-state equilibrium under normal conditions and are reciprocally inhibitory, leading to the hypothesis that signalling pathway interactions must be considered to understand how homeostasis is maintained in the tissue. This concept has recently been extrapolated to describe the interaction among the separate branches of the signalling in mathematical terms. 55 Our more recent data further validated this hypothesis by showing that while the Wnt/CaMKII pathway is also upregulated in OA in vivo, its pharmacological inhibition exacerbated cartilage degradation in a murine model of the disease. 51 Wnt3a has been shown to mediate the simultaneous activation of βcatenin-dependent and Ca 2+ -dependent pathways in other biological systems 24,56,57 which have been shown to be reciprocally regulatory. 58 Separately, we showed that Wnt16 also shares the ability of mediating multiple and contrasting signals in the cartilage. 43 Despite promoting anabolic effects through the activation of the Wnt/β-catenin pathway, Wnt16 antagonises the activation of this signalling cascade when induced by Wnt3a, therefore avoiding the activation of the pathway over a certain threshold. 43 Taken together, these data strengthen the argument that a fine-tuned balance in the activation of the Wnt signalling is required to maintain cartilage homeostasis and that buffering mechanisms -for example competitive antagonism of ligands for the same receptors or epigenetic modifications -are in place to avoid prolonged perturbation of this equilibrium, as when this occurs, it can lead to tissue degeneration and OA development (Figure 1).

BALANCE: THE ROLE OF RECEPTORS AND CO -RECEPTORS
Ligands and receptors are important gatekeepers of the activation of the different Wnt pathways. Beyond Wnt3a, Wnt5a and Wnt7a have been reported to activate βcatenin-dependent and -independent pathways 59-61 in a context-specific manner.
Engagement of Wnt ligands with different isoforms of FZDs is linked to the activation of both βcatenindependent and -independent pathways. Interaction of FZDs with different co-receptors is also considered pivotal in determining which branch of the network is going to be activated. Nonetheless, a detailed characterisation of the specificity of binding of different ligands for different isoforms, as well as how the availability in the extracellular environment of other Wnts can influence their affinity and specificity of binding, remains unclear for most tissues and primary cells.
The generation of fluorescence and luminescencebased biosensors has allowed some progresses in this direction. 62 Bourhis and colleagues demonstrated that in insect cells, individual Wnts can simultaneously bind to different domains of LRP6. They also showed that Wnt5a, Wnt5b and Wnt3a can all bind to FZD8. While simultaneous interactions for the same receptors are possible, the affinity of the different ligands for the different isoforms and binding sites can vary. Therefore, the binding scenario and the activation of different intracellular signalling can indeed be different depending on the availability of different ligands in the extracellular space. 63 Furthermore, also the availability of FZD on the cell surface can be influenced by external factors: R-spondins, Wnt agonists binding to leucine-rich repeat-containing G-protein-coupled receptors (LGRs), decrease the ubiquitination of FZD receptors, reducing their turnover and therefore potentiating the activation of the Wnt/β-catenin pathway. 64 Finally, the activity and availability of different coreceptors such as LRP-5 and 6 can also add up to the complexity of this signalling network. Studies in vivo showed that deficiency of LRP5 can exacerbate or decrease degeneration of the AC in murine models of osteoarthritis, in a model-dependent way. 65,66 This suggests that external factors, such as biomechanics and inflammation, can also alter the signalling response, through vastly uncharacterised mechanisms.
Frizzled are a family of seven-span transmembrane Gcoupled receptors. 67 G protein-coupled-receptor are not solely limited to activating and initiating their own downstream signalling pathways; they are also able to communicate with each other, influencing and mediating the activation and subsequent downstream signalling of one another. This can occur through synergistic crosstalk interactions between either different signalling transduction pathways, through either the same/different classes of GPCRs or through competitive antagonism, whereby they are able to communicate and prevent their own signal transduction pathway. 68 As a result, this can influence the potency and efficacy of the signalling, which, at times, can resemble behaviour that is typically observed with allosteric interactions.
Crosstalk can occur between GPCRs coupled to different G proteins, enhancing the cellular response of one of the pathways. This could be interactions between G i -and G q -coupled receptors or G s -and G q -coupled receptors. In the case of smooth muscle contraction, crosstalk occurs between G i -and G q -coupled receptors. 68 When there is no crosstalk among GPCRs and their downstream signalling, noradrenaline activates the G q -coupled αadrenergic receptors, resulting in smooth muscle contractions. However, in the presence of Neuropeptide Y (NPY), a cotransmitter that activates the G i -coupled Neuropeptide Y receptors, crosstalk initiates with αadrenergic receptors, enhancing the smooth muscle contractions. An advantage to this augmented response is that lower concentrations of noradrenaline are thus required to elicit smooth muscle contraction when co-administered with NPY. 68 Synergistic interactions can also occur between different classes of GPCRs. The Smoothened Receptor (SMO) F I G U R E 1 Schematic representation illustrating the importance of three branches of the Wnt-signalling network in the maintenance of cartilage homeostasis. is a Class F GPCR that, upon activation, mediates Sonic Hedgehog (SHH) signalling. Pusapati et al showed, using CRISPR experiments, that the sensitivity of NIH/3 T3 fibroblasts and spinal neural progenitors to SHH is increased when there is a loss of GPR161, an orphan Class A GPCR (Pusapati et al., 2018). It is important to note that the signalling still depends on the SMO receptor, however, through increasing sensitivity to the morphogen SHH. This shows the unique manner in which GPR161 is able to interact with the SMO receptor, influencing its subsequent signalling. 69 Finally, Civciristov and colleagues demonstrated that GPCRs can also respond differently to high and low concentrations of the same ligands. The response to low concentrations of the ligands was spatially and temporally distinct and resulted in different intracellular proteomic profiles. This was due to the preassembly of GPCR highorder complexes to the plasma membrane. 70 The characterisation of how the activation of the Wnt signalling is mediated at receptor level in the AC and in musculoskeletal tissues remains vastly unknown. While this presents multiple challenges, it has, however, the potential to allow the development of new therapeutic strategies to re-establish tissue homeostasis and halt OA progression.

| CONCLUSIONS
Several intracellular mechanisms have been shown to be key in maintaining the right threshold of activation of the different branches of the Wnt signalling network. 71,72 Some of these have been reviewed recently by us and others in other reviews [73][74][75][76] and will also be discussed in future publications.
The overall message of this review is the consideration that a change of mindset in the way we consider signalling pathways as a pharmacological target for OA and other diseases is needed. As already highlighted by Amerongen and Nusse in 2009, 77 we should stop considering the Wnt signalling as a group of individual and linear signalling cascades but rather start considering them as a network, whose branches interact with each other and reciprocally influence their activation or repression state. It is also clear that we should not define ligand/receptor interaction as absolute concepts, but rather we should validate them within specific biological contexts taking in account and, when possible, mimicking, the extracellular environment characterising different tissues in physiology and disease. System biology approaches and mathematical models will become essential tools to be used to clarify these complicated interactions.
Clarifying how the balance of this intricate network is maintained will be pivotal for the discovery of new therapeutic targets, where the treatment will probably not focus anymore on individual molecules but on the reestablishment of the overall signalling homeostasis.

ACKNOWLEDGEMENT
This work was funded by the Medical Research Council (grant reference MR/S008608/1).