Wnt Inhibitory Factor 1 Deficiency Uncouples Cartilage and Bone Destruction in Tumor Necrosis Factor α–Mediated Experimental Arthritis


University of Erlangen–Nuremberg Medical School, Nikolaus-Fiebiger-Zentrum für Molekulare Medizin, Department of Internal Medicine 3, Glückstrasse 6, 91054 Erlangen, Germany. E-mail: mstock@molmed.uni-erlangen.de



Wnt signaling plays a pivotal role in skeletal development and in the control of cartilage and bone turnover. We have recently shown that the secreted Wnt antagonist Wnt inhibitory factor 1 (WIF-1) is mainly expressed in the upper layers of epiphyseal and articular cartilage and, to a lesser extent, in bone. Nevertheless, WIF-1−/− mice develop normally. In light of these findings, we undertook this study to analyze the role of WIF-1 in arthritis.


Expression analyses for WIF-1 were performed by real-time reverse transcription–polymerase chain reaction (RT-PCR). WIF-1−/− and tumor necrosis factor (TNF)–transgenic mice were crossbred, and the progression of arthritis in TNF-transgenic WIF-1−/− mice and littermate controls was evaluated. Structural joint damage was analyzed by histologic staining, histomorphometry, and micro–computed tomography. Wnt/β-catenin signaling was investigated by real-time RT-PCR and immunofluorescence on primary chondrocytes.


WIF-1 expression was repressed by TNFα in chondrocytes and osteoblasts and down-regulated in experimental arthritis and in articular cartilage from patients with rheumatoid arthritis. WIF-1 deficiency partially protected TNF-transgenic mice against bone erosion and loss of trabecular bone, probably as a result of less osteoclast activity. In contrast, arthritis-related cartilage damage was aggravated by WIF-1 deficiency, while overexpression of WIF-1 attenuated cartilage degradation in TNF-transgenic mice. In chondrocytes, TNFα stimulated canonical Wnt signaling, which could be blocked by WIF-1, indicating a direct effect of TNFα and WIF-1 on Wnt signaling in this system.


These data suggest that WIF-1 may take part in the fine-tuning of cartilage and bone turnover, promoting the balance of cartilage versus bone anabolism.

Rheumatoid arthritis (RA) is characterized by chronic joint inflammation leading to the destruction of bone and cartilage. Inflammation of the synovial membrane (synovitis) is considered the essential trigger for perturbing local cartilage and bone homeostasis, which results in a net loss of bone and cartilage tissue in the affected joints ([1]). Mediators derived from inflammatory synovial tissue inhibit essential pathways for chondrocyte and osteoblast differentiation on the one hand, while fostering catabolic pathways such as expression of matrix enzymes and differentiation of bone-resorbing osteoclasts ([2]). Despite several lines of evidence showing that inflammation drives both bone and cartilage destruction in arthritis and the notion that these processes appear to be closely linked to each other, it is unclear how bone and cartilage loss during arthritis are related to each other, and if there are factors, which can differentially affect cartilage and bone during arthritis.

In this context, Wnt proteins and their inhibitors are of particular interest, as they crucially regulate bone–cartilage homeostasis. In most chondrocytes, Wnt proteins induce a catabolic state promoting the expression of proteinases such as matrix metalloproteinases (MMPs) and ADAMTS, which degrade the cartilage matrix ([3-5]). Interestingly, however, Ma et al recently demonstrated that Wnt signaling can reduce MMP expression in human chondrocytes ([5]). Wnt proteins also induce dedifferentiation of chondrocytes, thereby decreasing matrix synthesis ([6, 7]). Thus, Wnt signaling can trigger cartilage damage by increasing degradation and decreasing the synthesis of cartilage matrix. However, complete depletion of canonical Wnt signals results in the loss of articular cartilage, suggesting that Wnt signaling must be tightly regulated to ensure the structural integrity of articular cartilage ([8]).

Wnt proteins are also key mediators of bone formation ([9]). We recently demonstrated that R-spondin 1 (Rspo-1), an amplifier of canonical Wnt signaling, promotes osteoblast differentiation and bone formation in arthritis ([10]). In contrast, Wnt antagonists, such as Dkk-1, inhibit bone formation and enhance bone loss. In RA, for instance, Dkk-1 is overexpressed in synovial fibroblasts and links inflammation to bone destruction ([11]). Furthermore, Wnt signaling in synovial fibroblasts increases the release of proinflammatory cytokines, thereby fostering synovitis and contributing to bone and cartilage damage ([12-14]).

Wnt activity is controlled both by inhibitors binding to the low-density lipoprotein receptor–related protein 5 (LRP-5)/LRP-6 coreceptor, such as Dkk-1, and by inhibitors that bind directly to Wnt proteins, such as Wnt inhibitory factor 1 (WIF-1). WIF-1 is expressed in articular cartilage and bone, and it was first detected in the human retina. Highly conserved homologs have been described in various vertebrates ([15]). We recently provided a detailed overview of WIF-1 expression in developing chicken and mouse skeleton and demonstrated that WIF-1 is predominantly expressed at the superficial layer of epiphyseal and articular cartilage, suggesting a shielding function against Wnt proteins. Furthermore, WIF-1 is also significantly expressed in bone but absent from synovial tissue. We showed that WIF-1 physically interacts with Wnt-4 and Wnt-5a, which are expressed in arthritis ([11, 12, 16]). While we have demonstrated that WIF-1 interferes with Wnt-3a–mediated inhibition of chondrogenesis by effectively blocking the activation of canonical Wnt signaling ([16]), inhibition of noncanonical Wnt signaling has also been reported ([15]).

WIF-1−/− mice develop normally without overt alterations in the skeleton ([16, 17]), suggesting that WIF-1 might have a function beyond development. The high expression of WIF-1 in cartilage and bone and its Wnt-modifying activity prompted us to raise the question of whether WIF-1 has a role in the modulation of cartilage–bone homeostasis during inflammatory arthritis.


Primary skeletal cells and treatment

Primary mouse chondrocytes were isolated from the epiphyseal cartilage or rib cages of newborn mice, as previously reported ([18, 19]). Chondrocytes were incubated for 16 hours at 37°C in Dulbecco's modified Eagle's medium/Ham's F-12 medium with 10% fetal calf serum (FCS) prior to stimulation with recombinant human tumor necrosis factor α (TNFα) (Life Technologies) or recombinant WIF-1 as indicated. Primary murine osteoblastic cells were isolated from the calvariae of mice at age 4–8 days, as outlined previously ([20, 21]). Osteoblastic cells were plated at a density of 6,000/cm2 in α-minimum essential medium with 10% FCS. After reaching confluence, cells were treated with TNFα or WIF-1 as indicated. Recombinant murine WIF-1 was prepared as previously reported ([16]).


WIF-1−/− mice (C57BL/6) were generated as reported previously ([17]). To induce experimental arthritis in WIF-1−/− mice, these mice were crossbred with human TNFα–transgenic mice (Tg197-transgenic C57BL/6 mice) to obtain TNF-transgenic WIF-1−/− mice. Clinical signs of arthritis (i.e., body weight, loss of grip strength, and paw swelling) were assessed twice a week starting at age 7 weeks in a blinded manner as previously described ([22]). One knee per mouse was used for RNA extraction, while the other knee was used for histologic analyses. One cohort of mice (n = 6 per group) was subjected to cartilage and osteoclast evaluation on paraffin-embedded sections of decalcified knee joints. A second cohort (n = 5 or 6 per group) was used for the analyses of calcified tissues. Hind paws from the mice in both cohorts were used to prepare paraffin sections to evaluate inflammation. Wild-type (WT) and WIF-1+/– mice were used as healthy controls.

Hepatic overexpression of WIF-1 by hydrodynamic transfection of mice

Complementary DNA (cDNA) encoding full-length WIF-1 including the signal peptide was cloned into a vector for overexpression in the liver ([23]). Plasmid DNA was purified using an EndoFree Plasmid Maxi Kit (Qiagen) and a MiraCLEAN Endotoxin Removal Kit (Mirus Bio). Mice were transfected via hydrodynamic tail vein injection with 10–20 μg of DNA in Krebs-Ringer solution as described previously ([24]).

Human samples

RNA samples from articular cartilage were derived from arthroplasty specimens from 3 RA patients fulfilling the American College of Rheumatology/European League Against Rheumatism classification criteria ([25]) and from 3 osteoarthritis (OA) patients undergoing total knee joint arthroplasty at the University Hospital Erlangen. Each patient gave informed consent prior to surgery, and the institutional ethics committee approved the study protocol.

Gene expression analyses

Total RNA was extracted from cells or from murine knee joints using an RNeasy Fibrous Tissue Mini Kit (Qiagen). Complementary DNA was synthesized using a SuperScript II reverse transcriptase system (Invitrogen), and quantitative determination of messenger RNA expression was performed by real-time reverse transcription–polymerase chain reaction (RT-PCR), as previously reported ([18]). Cyclophilin A (murine samples) or GAPDH (human samples) was used as a reference gene and typically varied by ∼0.25–1 Ct between different cDNA samples within 1 experiment. All real-time RT-PCR reactions were run in triplicate. Primer sequences are available at http://www.medizin3.uk-erlangen.de/e1846/e741/e750/e980/inhalt981/Suppl-StockA&R2013.pdf.

Protein expression analyses

For immunodot blotting, 5 μl of serum was spotted onto a nitrocellulose membrane. Western blotting and immunodetection of WIF-1 were performed as described previously ([16]). Intracellular localization of β-catenin by immunofluorescence was performed as reported previously using anti–β-catenin antibody (1:300; Santa Cruz Biotechnology) and Cy3-labeled anti-rabbit IgG antibody (1:800; GE Healthcare) ([16]).

Conventional histology

Knee joints and hind paws were fixed overnight in 4.0% paraformaldehyde and then decalcified in 0.5M EDTA (pH 8.0) until bones were pliable. Serial paraffin sections (6 μm) were stained with hematoxylin and eosin for assessment of synovial inflammation and bone erosion. Proteoglycan content in articular cartilage was analyzed by Safranin O staining of paraffin sections as described previously ([26]). Osteoclasts were detected by tartrate-resistant acid phosphatase (TRAP) staining of tibial paraffin sections using a leukocyte acid phosphatase staining kit (Sigma-Aldrich).

Specific bone histology

For bone analyses of calcified tissue, tibiae were embedded in methylmethacrylate without decalcifying, and serial plastic sections were used for evaluation of bone loss and osteoblast counts. Thereafter, von Kossa's and Goldner's trichrome stainings were performed using standard procedures. Synovial inflammation, bone erosions, and osteoclast and osteoblast numbers were quantified with the use of an Axioskop 2 microscope (Carl Zeiss) equipped with a bone histomorphometry analysis system (OsteoMeasure; OsteoMetrics) as previously described ([10]). The area evaluated by bone histomorphometry encompassed the top 300 μm of the secondary spongiosa. Cartilage destruction was determined analogously. For measurement of cartilage volume, the epiphysis volume was used as total volume.

Micro–computed tomography (micro-CT) imaging and analysis

Micro-CT image acquisition of tibiae was performed using a laboratory cone-beam micro-CT scanner for ultra-high resolution imaging (tube voltage 50 kV, current 200 μA, exposure time 3,000 msec/projection, 800 projections). The tomographic data sets had an isotropic voxel size of 10 μm. A dedicated bone mineral density (BMD) calibration phantom (QRM) was scanned together with each specimen to calibrate the measured x-ray absorption values to BMD. For image analysis, we used the software MIAF-μCT, developed at the Institute of Medical Physics (University of Erlangen–Nuremberg). The structure of the trabecular network was quantified in a cylindrical volume of interest (VOI) positioned manually by the operator to encompass most of the metaphysis (further information is available at http://www.medizin3.uk-erlangen.de/e1846/e741/e750/e980/inhalt981/Suppl-StockA&R2013.pdf).

For segmentation a multistep process was used, as recently reported ([27]). First, 2 global thresholds were determined automatically from the histogram of the gray values within the cylindrical VOI. All voxels with gray values outside the range defined by the 2 thresholds were directly categorized as bone or marrow. For voxels with gray values inside this range, a 3-dimensional volume-growing approach, which preserves the connectivity of the trabecular network, was used in combination with local adaptive thresholding. The local adaptive thresholding algorithm categorized voxels as bone or marrow based on their gray value relative to the gray value distribution of the local neighborhood. Finally, trabecular number, spacing, and thickness were measured ([28]). All analyses were performed in a blinded manner.

Statistical analysis

Data are presented as the mean ± SEM, as indicated. The Mann-Whitney U test, one-way analysis of variance (ANOVA), Dunnett's T3 test, and 95% confidence intervals (95% CIs) of mean expression levels used to evaluate statistical significance were calculated using GraphPad Prism and SPSS software.


Down-regulation of WIF-1 and other secreted Wnt antagonists by TNFα.

We first tested whether TNFα—the key proinflammatory cytokine in arthritis—modulates the expression of Wnt antagonists in osteoblasts and chondrocytes. We stimulated murine chondrocytes and calvarial (pre)osteoblasts with TNFα. As expected, TNFα induced Dkk-1 expression in both cell types (Figures 1Aa and Ba). In contrast, as shown in Figures 1Ab and Bb, TNFα repressed WIF-1 expression (95% CIs in chondrocytes: control 0.951–1.049, 5 ng/ml TNFα 0.2412–0.6337, 25 ng/ml TNFα 0.2208–0.3384) (95% CIs in osteoblasts: control 0.8894–1.111, 5 ng/ml TNFα 0.3059–0.5823, 25 ng/ml TNFα 0.166–0.3116). Messenger RNA expression for secreted Frizzled-related proteins (sFRP) 1, 2, and 3 was also significantly down-regulated by TNFα (Figures 1Ac–e and Bc–e). Interestingly, arthritis-related Wnt ligands Wnt-4, Wnt-5a, and Wnt-10b were only mildly affected by TNFα in either cell system (further information is available at http://www.medizin3.uk-erlangen.de/e1846/e741/e750/e980/inhalt981/Suppl-StockA&R2013.pdf).

Figure 1.

Wnt inhibitory factor 1 (WIF-1) is repressed by tumor necrosis factor α (TNFα) in chondrocytes and osteoblastic cells and down-regulated in inflammatory arthritis. A and B, Primary murine chondrocytes (A) or calvarial cells (preosteoblasts) (B) were cultivated for 1 day in the absence or presence of TNFα. RNA was isolated, and relative expression levels of Wnt antagonists were quantified by real-time reverse transcription–polymerase chain reaction (RT-PCR). Cyclophilin A mRNA levels were assessed for standardization of relative mRNA expression. Bars show the mean ± SEM relative expression from 3 individual experiments. As expected, TNFα induced mRNA expression of the canonical Wnt signaling inhibitor Dkk-1 (Aa and Ba). In contrast, mRNA expression of WIF-1 (Ab and Bb) and other pan–Wnt antagonists (secreted Frizzled-related proteins [sFRP] 1, 2, and 3) (Ace and Bce) was significantly down-regulated by TNFα. C, RNA was isolated from the knee joints of 11-week-old wild-type (WT) and TNF-transgenic (TNFtg) mice (n = 6 each). WIF-1 mRNA expression levels were determined by real-time RT-PCR and standardized against cyclophilin A mRNA expression. D, Relative WIF-1 mRNA expression in human articular cartilage was determined in patients with osteoarthritis (OA; n = 3) and patients with rheumatoid arthritis (RA; n = 3). Expression was standardized against GAPDH mRNA expression. Bars show the mean ± SEM. Statistical significance was calculated by one-way analysis of variance and subsequent Dunnett's test (A and B) or by Mann-Whitney U test (C and D).

Down-regulation of WIF-1 expression in murine inflammatory arthritis and human RA

TNFα-dependent repression of WIF-1 expression prompted us to investigate whether WIF-1 is also down-regulated in TNFα-mediated arthritis in mice and in human RA. We assessed WIF-1 expression in knee joint tissue from TNF-transgenic mice and WT littermates at age 11 weeks and observed that WIF-1 was down-regulated during TNFα-mediated arthritis (95% CI in TNF-transgenic mice 0.1901–0.469, 95% CI in WT mice 0.7112–1.276) (Figure 1C). Moreover, in comparison to OA cartilage (95% CI 0.8971–1.067), RA cartilage (95% CI 0.2707–0.6985) exhibited dramatically decreased WIF-1 expression levels (Figure 1D). These data indicate that WIF-1 expression is decreased under inflammatory conditions, and that this repression may be mediated by TNFα.

WIF-1 −/− mice develop normal levels of inflammation in experimental arthritis

To analyze the role of WIF-1 in cartilage and bone homeostasis in vivo, we crossbred WIF-1−/− mice with arthritic TNF-transgenic mice. To exclude an effect of WIF-1 on inflammation, we carefully monitored the clinical signs of arthritis twice a week starting at age 7 weeks. In the following 4 weeks, TNF-transgenic mice and TNF-transgenic WIF-1−/− mice developed a similar extent of severe inflammatory arthritis, as characterized by loss of grip strength in conjunction with joint swelling in the front and hind paws (further information is available at http://www.medizin3.uk-erlangen.de/e1846/e741/e750/e980/inhalt981/Suppl-StockA&R2013.pdf). Likewise, there was no difference in body weight (further information is available at http://www.medizin3.uk-erlangen.de/e1846/e741/e750/e980/inhalt981/Suppl-StockA&R2013.pdf). Furthermore, histomorphometric analysis of hind paws from mice at age 11 weeks confirmed that WIF-1 deficiency did not significantly alter inflammation in arthritic mice (further information is available at http://www.medizin3.uk-erlangen.de/e1846/e741/e750/e980/inhalt981/Suppl-StockA&R2013.pdf).

WIF-1 deficiency provides partial protection against bone loss during arthritis

Our previous findings that (re)activating Wnt signaling in murine experimental arthritis by anti–Dkk-1 antibody or Rspo-1 attenuated arthritis-related bone loss prompted us to investigate whether similar effects could be observed during arthritis in WIF-1−/− mice ([10, 11]). We first confirmed earlier findings that WIF-1−/− and WT mice (also age 11 weeks) do not differ in their cartilage and bone phenotypes (results not shown) ([16, 17]). We then compared the skeletal phenotypes of TNF-transgenic WIF+/– and TNF-transgenic WIF+/+ mice at age 11 weeks (results not shown). Since these were indistinguishable, both genotypes were grouped together as TNF-transgenic mice. We performed micro-CT analyses to assess bone structure in the tibiae of WT, TNF-transgenic, and TNF-transgenic WIF-1−/− mice (Figures 2A–F). These analyses revealed that arthritis-mediated loss of trabecular bone was significantly lower in TNF-transgenic WIF-1−/− mice compared to TNF-transgenic mice (Figures 2A and B). While trabecular thickness remained unchanged (Figure 2E), TNF-transgenic WIF-1−/− mice exhibited a strong trend toward higher numbers (Figure 2C) and decreased spacing (Figure 2D) of bone trabeculae compared to TNF-transgenic littermates.

Figure 2.

WIF-1 deficiency partially protects against arthritis-related bone damage in TNF-transgenic mice. Bone density of healthy control mice (n = 5), TNF-transgenic mice (n = 6), and TNF-transgenic WIF-1−/− mice (n = 5) at age 11 weeks was evaluated by micro–computed tomography (micro-CT) imaging and analysis. A, Representative virtual sections of the tibiae are shown. B, Improved bone density (bone volume/total volume [BV/TV]) was observed in TNF-transgenic WIF-1−/− mice compared to TNF-transgenic mice. CE, Bone density improvement appeared to be a consequence of an increased number of trabeculae (Trab. No.) (C), with a consequent decrease in trabecular spacing (Trab. Sp.) (D), rather than resulting from a change in trabecular thickness (Trab. Th.) (E). F, CT rendering shows the tibia, fibula, and femur of a control mouse, a TNF-transgenic mouse, and a TNF-transgenic WIF-1−/− mouse. Note the less-eroded bone surface in the TNF-transgenic WIF-1−/− mouse compared to the TNF-transgenic mouse. A and GI, By micro-CT (A and H) and on von Kossa–stained methylmethacrylate sections from tibiae (G), large cortical erosions located directly below the growth plate (arrows in A, G, and H; dashed line in G) were detected primarily in TNF-transgenic mice (5 of 6), but only rarely in TNF-transgenic WIF-1−/− mice (1 of 5) (I). J, Shown is histomorphometric analysis of erosion diameters. Bars in BE and J show the mean ± SEM. Statistical significance was calculated using the Mann-Whitney U test. See Figure 1 for other definitions.

Moreover, the surface of subchondral bone in TNF-transgenic WIF-1−/− mice showed less erosive changes than in TNF-transgenic controls (Figure 2F). The tibiae of most TNF-transgenic mice further exhibited large erosions, penetrating the entire cortical bone shell (Figures 2A, G, and H), while TNF-transgenic WIF-1−/− mice were partially protected against these lesions (Figures 2A, G, and I). The diameter of these lesions was also larger in TNF-transgenic mice than in TNF-transgenic WIF-1−/− mice (Figure 2J). Reduced arthritis-related bone destruction in WIF-1−/− mice was also confirmed by bone histomorphometry in the hind paws and tibiae of TNF-transgenic WIF-1−/− mice and TNF-transgenic mice (further information is available at http://www.medizin3.uk-erlangen.de/e1846/e741/e750/e980/inhalt981/Suppl-StockA&R2013.pdf).

WIF-1 deficiency reduces arthritis-related increase in osteoclast numbers

To elucidate whether the protective effect of WIF-1 deficiency on arthritis-related bone loss is mediated by lower bone resorption or a higher bone formation rate, bone cells were analyzed in tibiae from TNF-transgenic WIF-1−/− mice and TNF-transgenic mice. TRAP staining (Figure 3A) revealed an increase in the number of metaphyseal osteoclasts (Figure 3B) and in the amount of bone surface covered by osteoclasts (Figure 3C) in arthritic TNF-transgenic mice compared to healthy control mice. These parameters were normalized in TNF-transgenic WIF-1−/− mice.

Figure 3.

WIF-1 deficiency impedes the increase in osteoclast activity associated with inflammatory arthritis. AC, Metaphyseal osteoclast numbers were decreased in TNF-transgenic WIF-1−/− mice in comparison to TNF-transgenic mice. Sections of tibiae from 11-week-old healthy control mice, TNF-transgenic mice, and TNF-transgenic WIF-1−/− mice were stained for tartrate-resistant acid phosphatase (A), and osteoclast numbers per bone perimeter (N.Oc./B.Pm) (B) and osteoclast surface per bone surface (Oc.S/BS) (C) were determined by histomorphometry in the secondary spongiosa. D, E, and H, Metaphyseal osteoblast numbers in the secondary spongiosa were not significantly altered in TNF-transgenic WIF-1−/− mice as compared to TNF-transgenic mice. For quantification of osteoblast numbers, methylmethacrylate-embedded tibial sections were stained with Goldner's trichrome stain (H). Osteoblast numbers per bone perimeter (N.Ob./B.Pm) (D) and bone surface covered by osteoblasts (Ob.S/BS) (E) were then evaluated by bone histomorphometry in the secondary spongiosa. F and G, Expression levels of RANKL (F) and osteoprotegerin (OPG) (G) in the knee joints was determined by real-time RT-PCR, indicating a decreased RANKL:OPG ratio in TNF-transgenic WIF-1−/− mice compared to TNF-transgenic mice. Bars show the mean ± SEM (n = 6 mice per group). Statistical significance was calculated using the Mann-Whitney U test. NS = not significant (see Figure 1 for other definitions).

This prompted us to investigate RANKL and osteoprotegerin (OPG) expression levels in these mice. Although the 95% CIs of mean RANKL expression levels in TNF-transgenic mice (1.194–2.663) and TNF-transgenic WIF-1−/− mice (0.5365–1.409) overlapped, both a Mann-Whitney U test and a one-way ANOVA with subsequent Dunnett's test indicated a statistically significant difference in expression levels, indicating at least a very strong trend toward reduced RANKL levels in TNF-transgenic WIF-1−/− mice compared to TNF-transgenic littermates (Figure 3F). These analyses were supported by the finding that WIF-1 promotes RANKL expression in chondrocytes and osteoblasts, both in the presence and in the absence of TNFα (further information is available at http://www.medizin3.uk-erlangen.de/e1846/e741/e750/e980/inhalt981/Suppl-StockA&R2013.pdf). In contrast, TNF-transgenic WIF-1−/− mice and TNF-transgenic mice were indistinguishable in OPG expression (Figure 3G), trabecular osteoblast numbers (Figures 3D and H), and bone surface covered by osteoblasts (Figure 3E). These data indicate that WIF-1 promotes the activity of bone-resorbing osteoclasts and thereby enhances bone loss in arthritis.

Enhanced cartilage destruction in WIF-1−/− arthritic mice

The abundant expression of WIF-1 at the cartilage surface warrants the assumption that loss of WIF-1 expression during arthritis may affect articular cartilage. Therefore, we investigated cartilage degeneration in 11-week-old TNF-transgenic WIF-1−/− mice and TNF-transgenic mice. Quantitative histomorphometric analyses of articular cartilage volume in tibiae revealed a significant decrease in articular cartilage in TNF-transgenic WIF-1−/− mice as compared to TNF-transgenic mice (Figure 4A). Moreover, proteoglycan loss was also significantly higher in TNF-transgenic WIF-1−/− mice than in TNF-transgenic mice. Thus, Safranin O staining for proteoglycans was virtually absent in tibial articular cartilage from TNF-transgenic WIF-1−/− mice. In contrast, in TNF-transgenic mice, only ∼60% of the articular cartilage showed evidence of proteoglycan loss. Healthy control mice did not exhibit significant levels of proteoglycan loss (Figures 4B and C). Taken together, these results suggest that WIF-1 has a protective role in articular cartilage and retards proteoglycan loss and cartilage destruction during arthritis.

Figure 4.

WIF-1 deficiency aggravates arthritis-related cartilage damage in TNF-transgenic mice. Cartilage destruction in 11-week-old TNF-transgenic mice and TNF-transgenic WIF-1−/− mice was assessed histomorphometrically at the proximal tibial epiphysis and compared to articular cartilage in healthy control mice of the same age. A, Articular cartilage volume/total volume (CV/TV) was assessed. B and C, Proteoglycan content in articular cartilage was analyzed by Safranin O staining, with proteoglycan loss measured as Safranin O–negative cartilage. Arrows in C indicate Safranin O–positive or Safranin O–negative articular cartilage. Bars show the mean ± SEM (n = 6 mice per group). Statistical significance was calculated using the Mann-Whitney U test. Original magnification × 100. See Figure 1 for other definitions.

Systemic overexpression of WIF-1 partially protects against arthritis-related cartilage loss

To test whether systemic administration of WIF-1 may consequently protect articular cartilage against arthritis-triggered degradation, we established a transient overexpression system based on hydrodynamic transfection of WIF-1. A hepatocyte-specific WIF-1 expression vector was transferred to WT and TNF-transgenic mice at age 6 weeks. Efficient transgene expression and secretion from the liver to the circulation were confirmed by detection of WIF-1 protein in the serum 2 days after transfection (Figure 5A). Eighteen days after transfection, WIF-1 overexpression was still detectable in the liver by real-time RT-PCR and Western blotting, although serum levels were too low to detect WIF-1 by immunoblotting at that time (Figures 5B and C).

Figure 5.

Hepatic overexpression of WIF-1 attenuates cartilage damage in TNF-transgenic mice. Hepatic overexpression of WIF-1 and resulting systemic administration of recombinant WIF-1 via the circulation was achieved by hydrodynamic transfection. AC, Overexpression of WIF-1 was confirmed by immunodot blotting of serum 2 days after transfection (A) and by real-time RT-PCR (B) and Western blotting (C) in livers from randomly chosen mice 18 days after transfection. Lanes 1–3, Mice subjected to mock transfection; lanes 4–8, mice subjected to hydrodynamic transfection with WIF-1. D, There was significantly less proteoglycan loss (measured as Safranin O–negative cartilage) in tibial articular cartilage from TNF-transgenic mice transfected with WIF-1 than in tibial articular cartilage from mock-transfected TNF-transgenic mice, revealing a protective effect of WIF-1 overexpression against arthritis-related cartilage damage (n = 7 mice per group). E, Articular cartilage volume/total volume (CV/TV) of tibiae from TNF-transgenic mice transfected with WIF-1 was higher than that in mock-transfected TNF-transgenic mice (n = 7 mice per group). Bars show the mean ± SEM. Statistical significance was calculated using the Mann-Whitney U test. See Figure 1 for other definitions.

In order to investigate the effect of WIF-1 overexpression on arthritis-related cartilage degradation, mice were killed at age 9 weeks (18 days after transfection). Cartilage degradation was analyzed by histomorphometry of Safranin O–stained knee sections, showing that WIF-1 overexpression in TNF-transgenic mice led to a substantial decrease in arthritis-related proteoglycan loss (Figure 5D) and loss of articular cartilage (Figure 5E). As expected, WIF-1 overexpression did not affect clinical signs of arthritis that are related to inflammation, such as body weight, loss of grip strength, and paw swelling (further information is available at http://www.medizin3.uk-erlangen.de/e1846/e741/e750/e980/inhalt981/Suppl-StockA&R2013.pdf).

Canonical Wnt signaling in chondrocytes induced by TNFα and blocked by WIF-1.

Our findings on the protective effect of WIF-1 on articular cartilage in arthritis prompted us to investigate whether TNFα may induce canonical Wnt signaling in chondrocytes, which in turn may at least in part mediate the cartilage degradation observed in TNFα-mediated inflammatory arthritis. Murine chondrocytes were treated with or without TNFα (5 ng/ml) and/or WIF-1 at a dose similar to that used for blockade of Wnt-3a signaling (5 μg/ml) ([16]). Real-time RT-PCR analysis revealed TNFα-mediated induction of the Wnt target gene axis inhibition protein 2 (95% CI: control 0.8631–1.137, TNFα 1.433–2.229). This up-regulation was efficiently blocked by the addition of WIF-1 (95% CI 0.431–1.158) (Figure 6A). Moreover, β-catenin accumulated in the cells and translocated into the nucleus when chondrocytes were treated with TNFα. Again, WIF-1 interfered with this TNFα-mediated effect (Figure 6C), suggesting that TNFα induced canonical Wnt signaling, which was blocked by WIF-1 in chondrocytes. Such a concept is supported by the finding that TNFα-mediated repression of Col2a1 expression (Figure 6B) was also antagonized by WIF-1.

Figure 6.

TNFα stimulates canonical Wnt signaling in murine chondrocytes, which can be blocked by WIF-1. Primary mouse epiphyseal chondrocytes were stimulated for 16 hours with or without 5 ng/ml TNFα and/or 5 μg/ml WIF-1. A, TNFα-mediated induction of Wnt signaling was investigated by expression analysis (real-time RT-PCR) of the Wnt target gene axis inhibition protein 2 (Axin-2). B, TNFα-mediated repression of anabolic chondrocyte gene expression (Col2a1) was demonstrated by real-time RT-PCR. Bars show the mean ± SEM from 3 individual experiments. Statistical significance was calculated by one-way analysis of variance and subsequent Dunnett's test. C, Induction of canonical Wnt signaling was also demonstrated by visualizing accumulation and nuclear translocation of β-catenin by immunofluorescence. Results shown are representative of 2 individual experiments. Addition of WIF-1 partially blocked the effects of TNFα on gene expression (A and B) and β-catenin accumulation (C). Original magnification × 630. See Figure 1 for definitions.


In this study we provide evidence for a role of the secreted Wnt antagonist WIF-1 in bone and cartilage homeostasis during arthritis. We demonstrate that while WIF-1 promotes bone erosion, it also protects against cartilage damage in experimental arthritis. Moreover, we found that, in contrast to the Wnt inhibitor Dkk-1, WIF-1 is down-regulated in arthritis, probably through TNFα signaling. Finally, we show that WIF-1 interferes with some of the TNFα-triggered effects on cartilage by a cell-autonomous mechanism, probably by attenuating TNFα-induced canonical Wnt signaling.

In RA and murine experimental arthritis, the canonical Wnt antagonist Dkk-1 has been shown to be overexpressed ([11]). The major source of elevated Dkk-1 levels in RA and experimental arthritis is the synovium; yet, considerable Dkk-1 expression was also shown in articular chondrocytes. Since the pan–Wnt antagonist WIF-1 is predominantly expressed in articular cartilage and to a lesser extent in bone, we raised the question of whether WIF-1 is also up-regulated in arthritis. Most surprisingly, however, our studies revealed that WIF-1 is repressed in chondrocytes and preosteoblasts upon treatment with TNFα. Moreover, other pan–Wnt antagonists such as sFRP-1, sFRP-2, and sFRP-3 were also down-regulated by TNFα, whereas, as expected, Dkk-1 expression was promoted. Furthermore, WIF-1 was down-regulated in articular cartilage from RA patients as compared to OA cartilage. Interestingly, sFRP-1 has been shown to be down-regulated by TNFα in synovial fibroblasts, and synovial tissue from RA patients exhibits lower levels of sFRP-1, sFRP-2, sFRP-3, and sFRP-4 than that from OA patients ([29, 30]). It remains to be elucidated whether down-regulation of WIF-1 and other pan–Wnt antagonists may reflect a negative feedback mechanism to compensate for Dkk-1 overexpression in arthritis.

We previously provided evidence that dysregulation of canonical Wnt signaling is involved in unbalanced bone turnover in arthritis ([10, 11, 31]). Thus, fostering canonical Wnt signaling can switch bone turnover from an erosive into an anabolic phenotype ([10, 11, 32]). Therefore, we hypothesized that WIF-1 deficiency might also affect bone loss during experimental arthritis. Indeed, WIF-1–knockout mice were partially protected against arthritis-triggered bone erosions and loss of bone mass, thus mimicking the effect of anti–Dkk-1 antibody or Rspo-1 administration. However, in contrast to anti–Dkk-1 or Rspo-1 treatment, WIF-1 deficiency was not sufficient to induce the formation of osteophytes in arthritic mice. It remains to be elucidated whether this difference is a result of varying quantities in the different systems or whether there is a qualitative difference between modulators specific for canonical Wnt signaling (i.e., Dkk-1, Rspo-1) and those regulating all Wnt signaling pathways (i.e., WIF-1). However, sFRP-3–deficient mice also do not exhibit alterations in osteophyte formation in 3 models of experimental OA when compared to WT mice ([33]), which supports the latter hypothesis. Furthermore, sclerostin, a specific inhibitor of canonical Wnt signaling, is down-regulated in joint diseases in which osteophytes develop, such as OA and ankylosing spondylitis ([31, 34]).

The protection against arthritis-related bone loss appeared to be predominantly a consequence of lower numbers of osteoclasts, whereas osteoblast numbers remained rather stable. Lower osteoclast numbers in arthritic WIF-1−/− mice may be explained by enhanced Wnt signaling in osteoblasts and/or chondrocytes due to the ablation of WIF-1. This notion is consistent with previous studies demonstrating that canonical Wnt signaling in osteoblasts represses osteoclastogenesis by inhibiting the expression of RANKL and inducing OPG expression ([35, 36]). Supporting this hypothesis, we found that RANKL expression in TNF-transgenic WIF-1−/− mice was reduced in comparison to that in TNF-transgenic mice, and we could show that WIF-1 enhanced RANKL expression in chondrocytes and osteoblasts. Although the major source of RANKL is believed to be osteoblasts and bone marrow stromal cells, emerging evidence suggests an important role for chondrocytes in RANKL production and the control of osteoclastogenesis ([37-40]). Our experiments imply that WIF-1 affects the production of both osteoblast- and chondrocyte-borne RANKL. Taken together, these findings suggest that protection against arthritis-related bone loss in WIF-1−/− mice may be at least partially a consequence of modulating the RANKL:OPG ratio, resulting in a decrease in osteoclastogenesis.

Under normal conditions, WIF-1−/− mice develop normally without any obvious alterations in the skeletal phenotype ([16, 17]). Moreover, cortical and trabecular bone structure is not significantly altered in WIF-1−/− mice ([17]). Therefore, the observed milder bone damage in WIF-1−/− mice is not a consequence of generally higher bone mass in WIF-1−/− mice. Instead, the bone phenotype of arthritic WIF-1−/− mice recapitulates to a large extent the anabolic effect of anti–Dkk-1 antibody or Rspo-1 on bone turnover in TNF-transgenic mice, including less osteoclast activity and bone erosion. Synovitis was unaffected in arthritic WIF-1−/− mice, and they exhibited normal development of paw swelling and loss of grip strength, which is also consistent with the finding that modulating Wnt activity in experimental arthritis via anti–Dkk-1 antibody or Rspo-1 does not affect synovial inflammation ([10, 11]). Thus, the data presented here support our earlier notion that (re)activation of Wnt signaling protects against arthritis-triggered bone destruction without affecting synovitis.

While the connection between synovial inflammation and bone erosion in arthritis is quite well established, the mechanisms resulting in cartilage damage in RA are less well understood. However, several lines of evidence indicate that abnormal Wnt signaling as a consequence of synovial inflammation may also contribute to cartilage degradation. Thus, Wnt signaling is involved in the control of cartilage matrix synthesis and degradation, chondrocyte dedifferentiation, and apoptosis, and excessive Wnt signaling leads to cartilage damage ([41]). In cartilage, β-catenin–dependent Wnt signaling has been shown to induce the expression of catabolic enzymes such as MMPs and ADAMTS while repressing gene expression of cartilage matrix proteins and promoting dedifferentiation of chondrocytes ([3, 6]). Moreover, Wnt-5a–mediated noncanonical Wnt signaling also has been shown to trigger catabolic effects on articular cartilage. Thus, Wnt-5a signaling induces the expression of MMPs and mediates the inhibitory effects of interleukin-1β on cartilage matrix synthesis in articular chondrocytes ([4, 42]).

Our findings indicate that WIF-1–knockout mice develop greater cartilage damage in TNFα-mediated arthritis, which is consistent with the notion that excessive Wnt signaling promotes cartilage degradation. These data are supported by our finding that overexpression of WIF-1 reduced cartilage damage in arthritic TNF-transgenic mice. This concept is also consistent with the finding that sFRP-3–knockout mice develop greater cartilage damage in experimental OA ([33]). In the current study we also demonstrated that TNFα represses WIF-1 expression and that of other pan–Wnt antagonists (sFRP-1, sFRP-2, and sFRP-3) in chondrocytes. This may reflect a pathologic effect of synovial inflammation, reducing the capacity to control Wnt activity in the cartilage compartment, thereby priming the joint for cartilage damage. The most plausible source of Wnt ligands involved in cartilage degradation is the inflamed synovium, where Wnt ligands including Wnt-1, Wnt-4, Wnt-5a, Wnt-10b, Wnt-11, and Wnt-13 are abundantly expressed ([11, 12]). These ligands could possibly elicit both canonical and noncanonical pathways in the chondrocytes of articular cartilage. However, our results indicate that there might also be a cell-intrinsic mechanism that leads to the activation of canonical Wnt signaling in chondrocytes in response to TNFα. Recombinant WIF-1 blocked this activation, pointing to an autocrine mechanism.

Taken together, our results demonstrate a pivotal role for WIF-1 in the balance of bone and cartilage in inflammatory arthritis. Our data clearly demonstrate that WIF-1 promotes local bone erosion and systemic bone loss but protects against arthritis-triggered cartilage degradation, indicating a shielding function of WIF-1 to rescue the articular surface from excessive Wnt signals. Therefore, we introduce WIF-1 as a molecule that uncouples bone damage from cartilage damage in inflammatory arthritis.


All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Stock had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Stock, van den Berg, Schett.

Acquisition of data. Stock, Böhm, Scholtysek, Klinger, Gelse, Gayetskyy, Engelke, Billmeier, Wirtz, van den Berg.

Analysis and interpretation of data. Stock, Englbrecht, Fürnrohr.


We gratefully acknowledge Maria Gesslein and Herbert Rohrmüller for excellent technical assistance. We especially want to thank Igor Dawid (Laboratory of Molecular Genetics, National Institute of Child Health and Human Development, National Institutes of Health, Bethesda, MD) for sharing the WIF-1−/− mouse strain.