Dickkopf-3 is upregulated in osteoarthritis and has a chondroprotective role

Summary Objective Dickkopf-3 (Dkk3) is a non-canonical member of the Dkk family of Wnt antagonists and its upregulation has been reported in microarray analysis of cartilage from mouse models of osteoarthritis (OA). In this study we assessed Dkk3 expression in human OA cartilage to ascertain its potential role in chondrocyte signaling and cartilage maintenance. Methods Dkk3 expression was analysed in human adult OA cartilage and synovial tissues and during chondrogenesis of ATDC5 and human mesenchymal stem cells. The role of Dkk3 in cartilage maintenance was analysed by incubation of bovine and human cartilage explants with interleukin-1β (IL1β) and oncostatin-M (OSM). Dkk3 gene expression was measured in cartilage following murine hip avulsion. Whether Dkk3 influenced Wnt, TGFβ and activin cell signaling was assessed in primary human chondrocytes and SW1353 chondrosarcoma cells using qRT-PCR and luminescence assays. Results Increased gene and protein levels of Dkk3 were detected in human OA cartilage, synovial tissue and synovial fluid. DKK3 gene expression was decreased during chondrogenesis of both ATDC5 cells and humans MSCs. Dkk3 inhibited IL1β and OSM-mediated proteoglycan loss from human and bovine cartilage explants and collagen loss from bovine cartilage explants. Cartilage DKK3 expression was decreased following hip avulsion injury. TGFβ signaling was enhanced by Dkk3 whilst Wnt3a and activin signaling were inhibited. Conclusions We provide evidence that Dkk3 is upregulated in OA and may have a protective effect on cartilage integrity by preventing proteoglycan loss and helping to restore OA-relevant signaling pathway activity. Targeting Dkk3 may be a novel approach in the treatment of OA.


Introduction
Osteoarthritis (OA) is characterized by loss of articular cartilage, joint pain and instability. The mechanisms regulating disease pathogenesis remain elusive with a combination of genetic, inflammatory, mechanical and metabolic factors implicated 1e3 .
Chondrocytes from OA cartilage exhibit a disrupted phenotype, hallmarks of which include; altered synthesis of extracellular matrix (ECM) and ECM-degrading enzymes, altered cell signaling activity and increased proliferation 4 . Dysregulation of cell signaling pathways likely contributes to OA pathogenesis by reducing the chondrocyte's ability to maintain cartilage integrity, leading to or exacerbating the phenotypic shift associated with OA. The Wnt and transforming growth factor beta (TGFb) signaling pathways have been strongly implicated in OA pathogenesis 5,6 .
Dickkopf-3 (Dkk3) is a structurally and functionally divergent member of the Dkk family of Wnt antagonists. Dkk3 activates or inhibits Wnt signaling in a tissue-dependent manner and its impact on cartilage Wnt signaling is unknown 7e9 . Dkk3 is a tumour suppressor that inhibits proliferation of cancer cells and is downregulated in several types of human cancer 8e10 . It can modulate inflammatory cell activity, maintain tissue organisation via TGFb signaling and can protect against myocardial infarction-induced fibrosis 11e14 .
The function of Dkk3 in other tissues suggests it could be an important mediator of chondrocyte homeostasis and maintenance of cartilage integrity. Several studies using animal models of OA have reported increased Dkk3 in diseased cartilage 15e17 . However Dkk3 expression has not been well characterized in human OA tissue nor has its role in chondrocyte biology been explored. Our aim was to assess whether Dkk3 shows aberrant expression in human OA and to establish whether it can regulate chondrocyte behaviour and OA-associated cartilage degradation in vitro.

Primary tissue
Primary human OA cartilage and synovium were obtained from age-matched individuals undergoing hip replacement for OA and control cartilage and synovium obtained upon hip replacement for neck-of-femur fracture (NOF); cartilage OA n ¼ 13, NOF n ¼ 12, OA synovium n ¼ 8; NOF synovium n ¼ 11. Anteromedial OA (AMG) specimens were obtained from patients undergoing unicompartmental knee replacement (UKR) for OA. Primary human chondrocytes (HAC) were obtained from macroscopically normal regions of the tibial plateau of OA patients undergoing total knee replacement (TKR) and collagenase digested following standard protocols. Explants of cartilage were used for proteoglycan and collagen release assays (DMMB and hydroxyproline respectively). Synovial fluid was collected from individuals undergoing TKR (n ¼ 3), UKR (n ¼ 3), arthroscopy for cartilage lesions (n ¼ 5), matrixinduced autologous chondrocyte implantation (MACI, n ¼ 7) or control patients (n ¼ 3) with no cartilage lesion but meniscal tears.
Ethical approval (09/H0606/11 and 2005ORTHO7L) was granted by Oxfordshire Research Ethics Committee and East Norfolk and Waveney Research Governance Committee. Informed consent was obtained from all patients.

Cartilage explant assays
Bovine nasal septum and human articular cartilage were dissected and 2 mm cartilage discs explanted and equilibrated for 24 h before treatment with interleukin-1b (IL1b) (0.5 ng/ml), oncostatin-M (OSM) (5 ng/ml) plus Dkk3 (50, 125 and 250 ng/ml). Treatments were refreshed every 2e3 days and collected forglycosaminoglycan (GAG) and collagen release assays using dimethylmethylene blue (DMMB) and hydroxyproline assays respectively. Remaining cartilage was harvested at 14 days for papain digestion and DMMB and hydroxyproline assays 21 . Control and IL1/OSM-treated explants were collected throughout the time course for RNA extraction (Trizol, Invitrogen, UK), subsequent cDNA synthesis (Superscript, Invitrogen UK) according to manufacturer's instructions prior to quantitative real time PCR (qRT-PCR). Three intra-experimental replicates were carried out for each treatment condition.

Hip avulsion assay
The hip joint from 5 to 6 week old C57BL/6J mice was dislocated at the femur and the femoral cap avulsed using forceps as previously described 22 . Hip joint cartilage was cultured for 1e48 h in serum-free medium before RNA extraction using Trizol (Invitrogen, UK). cDNA synthesis using Superscript (Invitrogen, UK) was performed prior to qRT-PCR.

Enzyme-linked immunosorbent assay (ELISA)
Dkk3 level in synovial fluid was measured using Dkk3 ELISA (R&D Systems, UK) according to manufacturer's instructions.

Cytokine treatments
Cells were serum starved overnight and treated with recombinant IL1b (5 ng/ml) and/or OSM (10 ng/ml) for 24 h or pre-treated for 1 h with recombinant Dkk3 (250 ng/ml unless otherwise stated) or carrier alone before addition of recombinant Wnt3a (100 ng/ml, 10 h), activin (20 ng/ml, 6 h) or TGFb1 (4 ng/ml, 6 h). All cytokines from R&D Systems. At least three intra-experimental replicates were carried out per cytokine treatment.
Following cytokine treatment cDNA was synthesized using MMLV from DNase-treated cell lysates harvested in Cells-to-cDNA lysis buffer (Ambion) according to manufacturer's instructions.

qRT-PCR
Expression of genes was measured by qRT-PCR on a ViiA7 (Applied Biosystems). Relative quantification is expressed as 2 ÀDCt , where DC t is C t (gene of interest) À C t (18S rRNA). Samples which gave a C t reading þ 1.5C t greater or less than the median for 18S were excluded from further analyses.

Statistical analysis
Analyses were carried out using Graphpad Prism 6.0. Student's t test was used to test differences between two samples whilst analysis of variance (ANOVA) with either Dunnett's or Tukey posttest was used for multiple samples. Normality was tested using the ShapiroeWilk test. P < 0.05 was considered statistically significant.  * 0.05, ** 0.01, *** 0.001. Graphs show mean ± 95% confidence intervals of biological (patient or cell) replicates.

Dkk3 expression is upregulated in OA tissue
Expression of DKK3 mRNA was increased >10-fold (P < 0.0001) in OA cartilage compared to NOF control [ Fig. 1(A)]. Analysis of synovium from OA patients and NOF controls showed a 3.2-fold (P ¼ 0.0235) increase in DKK3 mRNA in diseased tissue. DKK3 mRNA expression [ Fig. 1(B)] was 2.1-fold (P ¼ 0.019) higher in damaged cartilage from patients with AMG. Our previous work shows reduced matrix metalloproteinase (MMP) and mRNA expression in damaged compared to undamaged cartilage 25 . Immunohistochemistry in AMG patients also showed significant Dkk3 staining in the superficial zone of damaged but not undamaged cartilage [ Fig. 1(C)]. Dkk3 protein [ Fig. 1(D)] in synovial fluid was 2.1-fold higher (P ¼ 0.0002) in patients undergoing TKR for OA compared to control individuals, those with cartilage lesions (4.33fold, P < 0.0001) or patients undergoing UKR (2.83-fold, P ¼ 0.0016). Matrix-induced autologous chondrocyte implantation (MACI) is performed 4e6 weeks following initial assessment of cartilage lesions by arthroscopy. Dkk3 levels at the time of MACI were significantly higher than at arthroscopy (i.e., lesion) (2.3-fold,

DKK3 expression is downregulated following cartilage injury and during chondrogenesis
The OA phenotype includes reinitiation of development 26 , thus establishing Dkk3 regulation in chondrogenesis is important. ATDC5 differentiation is an established model of chondrogenesis. Following chondrogenic differentiation, microarray analysis showed Dkk3 expression decreased relative to non-induced control cultures [ Fig. 2(A)]. Expression of chondrogenic markers collagen, type II, alpha I (Col2a1) and aggrecan (Acan) (data not shown) were increased across these time points 23 . Human MSCs in high density transwell cultures also showed a significant 1.3e21-fold reduction (P < 0.01) in DKK3 expression throughout chondrogenic differentiation into cartilage discs [ Fig. 2(B)], with increases in COL2A1 and ACAN across the time course 18 .
Joint injury is associated with secondary OA therefore Dkk3 regulation during injury or in response to inflammatory mediators of injury was investigated. Dkk3 expression in murine cartilage was decreased 1.

Dkk3 inhibits Wnt signaling
Dkk3 is a non-canonical member of the Dkk family of Wnt antagonists with tissue-dependent effects on Wnt signaling activity. To determine whether Dkk3 did regulate Wnt signaling in cartilage we treated HAC with Dkk3 and Wnt3a. The Wnt3a-induced increase of the Wnt target gene axis inhibition protein 2 (AXIN2) [Fig. 4(A)] was decreased in HAC by co-incubation with Wnt3a and 125, 250 or 500 ng/ml Dkk3 (1.6-, 2.2-and 2.5-fold, P ¼ 0.0050, <0.0001, <0.0001 respectively) compared to Wnt3a alone. Furthermore the activity of the Wnt-responsive TOPFlash reporter was reduced by the addition of Dkk3 (1.7-fold, P ¼ 0.0010) [ Fig. 4(B)] compared to Wnt3a alone. Knockdown of Dkk3 in HAC increased Wnt3a-induced AXIN2 expression compared to a non-targeting siRNA control [ Fig. 4(C)]. Micromass cultures of HAC show significant reduction in proteoglycan production following Wnt3a treatment for 4 days [ Fig. 4(D)]. Proteoglycan levels were restored by addition of Dkk3 demonstrating inhibition of Wnt3a-mediated effects on proteoglycan synthesis.
Activin is a member of the TGFb superfamily that also signals via Smad2/3. To assess whether Dkk3 impacted other Smad2/3-related signaling pathways, HAC and SW1353 were treated with activin ± Dkk3. Activin-induced TIMP3 expression and (CAGA) 12 -luc activity whilst co-incubation with Dkk3 caused a dose-dependent reduction in both of these outputs [ Fig. 6(A and B)]. Knockdown of Dkk3 enhanced activin-induced TIMP3 expression and CAGA-luc activity suggesting endogenous Dkk3 may act to reduce cellular activin-induced responses [ Fig. 6(C and D)]. There was no repression of HAC TIMP3 expression when p38 MAPK activity was inhibited [ Fig. 6(E)]. Activin-induced PAI1 expression followed the same trends as TIMP3 [ Supplementary Fig. 3(AeC)].

Discussion
Altered expression of cytokines and consequent disruption of cell signaling is associated with OA pathogenesis. Dkk3 is a noncanonical member of the Dkk family of Wnt antagonists that has not been explored in cartilage biology despite numerous studies noting its increased expression in models of OA. In this study we demonstrate that Dkk3 is upregulated in adult human OA cartilage and synovial tissue but is decreased during chondrogenesis. Dkk3 protects against in vitro cartilage degradation and its expression is regulated by both injury and inflammatory cytokines. Wnt and activin signaling are both inhibited by Dkk3 whilst TGFb signaling is enhanced. The upregulation of Dkk3 in OA may be a protective mechanism to limit cartilage damage and to regulate aberrant cell signaling associated with disease.
OA is a complex disease affecting multiple joint tissues, with a unique combination of factors likely to regulate pathogenesis within each tissue and across different joint locations. We show that Dkk3 is upregulated in both hip and knee OA and in both synovial tissue and cartilage from diseased joints. Dkk3 upregulation is also reported in OA subchondral bone from patients undergoing TKR 29 . This suggests Dkk3 is relevant to whole joint biology in two common sites of disease. The increased Dkk3 in synovial fluid of patients with tricompartmental OA may implicate Dkk3 as a biomarker distinguishing end-stage disease. Further studies of Dkk3 as a circulating biomarker are warranted.
Dysregulation of Wnt and TGFb family members has been strongly implicated in experimental and human OA 5,6 . An imbalance in Wnt signaling leads to OA development in murine models, and Wnt antagonists DKK1 and FRZB have been reported as downregulated in human OA 30e32 . Wnts and activin are also released following cartilage injury 33,34 . TGFb signaling and responsiveness decrease with age and OA development, whilst increased activin has been detected in OA tissues 34,35 . Dkk3 has both agonistic and antagonistic effects on the Wnt pathway dependent on tissue of expression and thus investigation of its impact on Wnt signaling in cartilage was investigated in our study 7e9 . Opposing regulatory roles of Dkk3 on TGFb signaling in Xenopus and prostate cancer 13,28 have been reported but its function in musculoskeletal tissue has not been studied.
In adult HAC we have shown that Dkk3 antagonized Wnt signaling and protected against Wnt-induced proteoglycan reduction. Dkk3 enhanced TGFb signaling in chondrocytes and interestingly was necessary for TGFb-induced reduction of MMP13 expression. Dkk3 may mediate protective effects on cartilage partially through upregulation of TGFb signaling and inhibition of Wnt signaling. Surprisingly, Dkk3 inhibited activin signaling in cartilage despite both activin and TGFb commonly signaling through Smad2/3. Inhibition of p38 MAPK signaling abrogated the effects of Dkk3 on both TGFb and activin signaling which shows Dkk3 action here is p38 MAPK dependent. A previous study demonstrated Dkk3-dependent Smad4-stabilization by p38 MAPK and this requires further investigation in chondrocytes 36 . Our data may indicate that Dkk3 effects on TGFb require p38 MAPK for stabilization of Smad4. The effect of Dkk3 on activin signaling is also p38 MAPK dependent but may operate through a pathway that does not use Smad4. The mechanism by which differential (250 ng/ml)-induced reduction in TIMP3 expression following Activin treatment (n ¼ 4). (AeD) ANOVA with Dunnett's post-test, significance shown for comparison between Activin and Activin þ Dkk3 (A, B) and between Activin þ siDkk3 to Control þ siDkk3 (C, D). (E) ANOVA with Tukey post-test, significance shown for comparison between Activin alone and Activin þ Dkk3 in the absence and presence of SB202190. n represents biological replicates (the mean of three technical replicates per condition for luciferase assays and four technical replicates per condition for gene expression assays). All statistical analysis carried out on biological replicates. regulation of activin and TGFb can occur is currently unknown and beyond the scope of this study.
Injury to the joint commonly leads to OA development. To model cartilage injury ex vivo the murine hip was avulsed and Dkk3 levels found to be decreased within 1 h. Decreased Dkk3 protein was also shown in pilot data from an ex vivo porcine explant model 37 following cutting injury (data not shown). Treatment with IL1b/OSM in our study led to a reduction in DKK3 expression that was partially p38 MAPK dependent. In contrast, previous reports on murine OA 15e17 and our data in human tissue show an increase in Dkk3 expression in established disease. Dkk3 may be regulated in a temporal manner during disease pathogenesis. This is supported by our BNC data that shows an initial decrease in DKK3 expression followed by an increase as cartilage degradation occurs. It is also of note that levels of Dkk3 protein in synovial fluid were lower at the time of arthroscopy than 4e6 weeks later when MACI was performed. This may indicate that injury to the joint capsule leads to significant Dkk3 release from other joint tissues that overcomes any decrease due to cartilage injury. The sources of Dkk3 in the joint require further investigation. The initial injury response leading to decreased Dkk3 may be completed at MACI with Dkk3 levels consequently increased in the ensuing, later stage repair attempt.
Paralleling the potential roles of the Wnt and TGFb pathways in OA pathogenesis, chondrogenesis and articular cartilage development require TGFb signaling as well as regulation of Wnt signaling 5,38 . Given the reversion of OA chondrocytes to a developmental-like phenotype 39 our data showing decreased DKK3 during chondrogenesis, shows a potential role for Dkk3 in chondrogenesis, and also suggests that the immediate downregulation of DKK3 in injury may be an early repair response.
Strikingly, Dkk3 protected against IL1b/OSM-stimulated cartilage degradation. The increase in Dkk3 in OA may be a protective mechanism to minimize cartilage degradation and the OAassociated shift in chondrocyte phenotype. This is supported by the reduction in cartilage-degrading MMP13 expression by Dkk3 in the presence of IL1b/OSM. Microarray analysis of HAC treated with siRNA against Dkk3 did not reveal pathways of Dkk3 action on unstimulated cells (data not shown), thus future analysis will use cytokine-stimulated cells. However siRNA treatment did increase MMP13 expression in TGFb-treated cells suggesting that Dkk3 may limit cartilage damage partially through reduction of both IL1b/ OSM and TGFb-effects on MMP13.
Overall Dkk3 upregulation in disease may be a defence mechanism to counteract disease-related dysregulation of cell signaling pathways; inhibiting inflammatory cytokine effects on cartilage degradation and enhancing TGFb signaling whilst maintaining regulation of Wnt signaling in an attempt to counteract diseaseassociated changes in these pathways. Supplementation with Dkk3 at an early stage of disease or post-injury may therefore be therapeutically beneficial.
Further investigation of Dkk3 in murine models of OA is necessary to ascertain its contribution to cartilage homeostasis and disease pathogenesis. Although the Dkk3 null mouse 40 does not have an overt musculoskeletal phenotype our preliminary analysis suggests increased knee OA in 6-month old animals, we are currently investigating injury-models of OA. Dkk3 gene therapy is in clinical trial for prostate cancer with promising results 41 , but further preclinical evaluation is necessary alongside more detailed investigation of the role of Dkk3 in other tissues of the healthy and OA joint.
In summary we have demonstrated that Dkk3 is upregulated in human OA and reduces cartilage degradation. These findings may have clinical implications as treatment with Dkk3 may prevent cartilage degeneration in OA and early intervention with Dkk3based therapy may slow OA progression.

Contributors
SJBS and IMC designed the study. SJBS, RKD, TES, MJB, KC and LL carried out data acquisition. AJC and AP provided patient samples and assisted with data interpretation. SJBS and IMC carried out data analysis and interpretation. All authors helped prepared the manuscript and approved the manuscript for submission.