Articular cartilage and biomechanical properties of the long bones in Frzb-knockout mice




Ligands and antagonists of the WNT pathway are linked to osteoporosis and osteoarthritis. In particular, polymorphisms in the FRZB gene, a secreted WNT antagonist, have been associated with osteoarthritis. The aim of this study was to examine cartilage and bone in Frzb−/− mice.


The Frzb gene in mice was inactivated using a Cre/loxP strategy. Three models of osteoarthritis were used: collagenase, papain, and methylated bovine serum albumin induced. Bone biology was studied using density measurements and microfocal computed tomography. Bone stiffness and mechanical loading–induced bone adaptation were studied by compression of the ulnae.


Targeted deletion of the Frzb gene in mice increased articular cartilage loss during arthritis triggered by instability, enzymatic injury, or inflammation. Cartilage damage in Frzb−/− mice was associated with increased WNT signaling and matrix metalloproteinase 3 (MMP-3) expression and activity. Frzb−/− mice had increased cortical bone thickness and density, resulting in stiffer bones, as demonstrated by stress–strain relationship analyses. Moreover, Frzb−/− mice had an increased periosteal anabolic response to mechanical loading as compared with wild-type mice.


The genetic association between osteoarthritis and FRZB polymorphisms is corroborated by increased cartilage proteoglycan loss in 3 different models of arthritis in Frzb−/− mice. Loss of Frzb may contribute to cartilage damage by increasing the expression and activity of MMPs, in a WNT-dependent and WNT-independent manner. FRZB deficiency also resulted in thicker cortical bone, with increased stiffness and higher cortical appositional bone formation after loading. This may contribute to the development of osteoarthritis by producing increased strain on the articular cartilage during normal locomotion but may protect against osteoporotic fractures.

Osteoarthritis and osteoporosis are common joint and bone diseases that cause significant morbidity and disability in the aging population. Osteoarthritis is primarily characterized by degeneration of the articular cartilage and leads to loss of joint function, and patients often require surgery for placement of a prosthesis to correct it (1). Drugs that convincingly affect the disease process beyond pain relief are not yet available. Osteoporosis is defined by decreased cortical and trabecular bone density and typically results in hip and vertebral fractures (2). Current antiosteoporosis agents inhibit osteoclast-driven bone resorption or stimulate osteoblast-driven bone synthesis, but their long-term use can cause drug safety problems (2). Clinical observations suggest that there is an inverse relationship between osteoarthritis and osteoporosis (3), but this hypothesis remains controversial, particularly since it is not supported by a known molecular mechanism.

A role of ligands and antagonists of the WNT signaling pathway in human bone and joint diseases has been suggested (for review, see refs.4 and5). The WNT family represents a group of at least 19 secreted glycoproteins that modulate cell proliferation, differentiation, and behavior during embryonic development, postnatal growth, homeostasis, and disease (for review, see ref.6). WNT ligands can signal through at least 3 different signaling cascades. The β-catenin–dependent (canonical) pathway relies on the fact that β-catenin is phosphorylated by glycogen synthase kinase 3β (GSK3β), which leads to the degradation of β-catenin by the proteasome complex. Upon WNT signaling, the activity of GSK3β is blocked, leading to the stabilization of β-catenin in the cytoplasm and its subsequent translocation to the nucleus. There, β-catenin forms a complex with transcription factor (TCF), lymphoid enhancer–binding factor (LEF), or other transcription factors, thereby activating specific target genes. WNT signaling can also take place through the phosphatidylinositol/Ca2+ or protein kinase C pathways, or via JNK activation (7–9).

Frizzled-related protein (FRZB; also called secreted Frizzled-related protein 3 [sFRP-3]) was originally isolated from bovine articular cartilage and found to be expressed in developing skeletal elements (10–12). Subsequently, FRZB was identified as an extracellular antagonist of the canonical WNT signaling pathway, preventing axis duplication in a Xenopuslaevis assay (11–13). This inhibitory activity appears to be specific, since FRZB blocks WNT-1 and WNT-8 signaling in both Xlaevis embryos and a mammalian cell line, but does not seem to affect WNT-5 (11–13). Several additional sFRP molecules have been identified, all of which interact with WNTs and modulate WNT signaling (14). Sfrp1−/− mice have a high trabecular bone density, suggesting a specific role for at least some sFRP molecules in bone biology (15). FRZB consist of 2 protein domains, a cysteine-rich domain responsible for WNT interactions and a netrin domain identified as a unique C-terminal basic domain in laminin-related modular proteins called netrins (16). The netrin domain is homologous to the N-terminal domain of tissue inhibitors of metalloproteinases (17), and recently, an interaction between FRZB and matrix metalloproteinase 2 (MMP-2) was demonstrated (18).

Two single-nucleotide polymorphisms (SNPs) in FRZB were shown to be associated with osteoarthritis (19–22), including a differential association between patients requiring hip replacement surgery for osteoarthritis and patients with osteoporotic hip fractures (21). These mutations reduced WNT inhibitory activity of FRZB in vitro, suggesting a function for WNT signaling in osteoarthritis (19).

In the present study, we examined joint and bone development and postnatal biology in Frzb−/− mice. These mice did not display overt developmental abnormalities. Postnatally, however, they showed increased cartilage damage in 3 models of induced osteoarthritis. In addition, the mice had increased cortical bone density in the long bones, as well as altered biomechanical properties and anabolic responses of the long bones. Our results suggest a function for FRZB in bone and cartilage homeostasis and disease, and provide a molecular link in the hypothesized inverse relationship between osteoarthritis and osteoporosis.


Generation of Frzb−/− mice.

Homologous recombination in embryonic stem cells was obtained with a targeting construct in which exon 1 of Frzb was floxed between loxP sites that contained an frt-flanked neomycin resistance gene. Homozygous loxP/loxP/frt/frt mice were bred with EIIa-Cre mice and with mice expressing the Flp transgene under control of the cytomegalovirus promoter. Mice from a mixed background were bred onto the Swiss/CD1 background, which has large litters, thus facilitating developmental studies, or the C57BL/6 background, which is necessary for the induction of inflammatory arthritis for >10 generations. All animal experiments were approved by the Ethical Committee for Animal Research, Katholieke Universiteit Leuven.

Arthritis models.

Osteoarthritis was induced in 8–10-week-old CD1 mice by intraarticular injection of 2 μg/μl of type VII collagenase or 1% papain (both from Sigma-Aldrich, Bornem, Germany), as described elsewhere (23, 24). Mice were killed after 7 and 21 days. Methylated bovine serum albumin (mBSA)–induced arthritis in 8–10-week-old C57BL/6 mice was triggered by intraarticular injection of mBSA on day 1, with subcutaneous injection of interleukin-1 on days 1–3; mice were killed on day 7 (25). Severity of disease was determined according to histologic scores, which were determined as described elsewhere (23–25). Cartilage damage in all models was measured by digital image analysis, using Safranin O staining to quantify the proteoglycan content (26).

Genotyping and quantitative polymerase chain reaction (PCR).

Mice were genotyped by PCR using DNA obtained from tail biopsy tissues. Wild-type and knockout alleles were amplified using forward primer p1 (5′-TGAACTTTGCCCGACCTCTGAG-3′) and reverse primer p2 (5′-GATCGCTCGGATCACTTGTTGG-3′) or using forward primer p3 (5′-CTGATGTCTCTGCCAGAGCGAG-3′) and reverse primer p4 (5′-TGGACGTAAACTCCTCTTCAGACC-3′), respectively. For quantitative PCR, RNA was isolated from different tissues of wild-type and Frzb−/− mice (Nucleobond; Macherey-Nagel, Düren, Germany). After reverse transcription using a RevertAid H Minus complementary DNA (cDNA) synthesis kit (Fermentas, St. Leon-Rot, Germany), cDNA templates were amplified with TaqMan Assays-on-Demand (Applied Biosystems, Lennik, Belgium) primer- probe sets.

Signaling pathway gene expression analysis.

RNA was isolated from the articular cartilage of knees with mBSA-induced arthritis or from contralateral control knees. RNA from 3 different mice was pooled for gene expression analysis. Three different samples from 4 groups were used in the analysis (total of 12 knees). Gene expression was studied using the WNT signaling pathway Oligo GEArray (SuperArray, Frederick, MD) according to the manufacturer's instructions. Signal was detected using chemiluminescence, with equal exposure time for all samples. Data were analyzed using GEArray software (SuperArray). Gene expression levels were normalized using the different available housekeeping genes present on the membranes.

Hematologic and biochemical analyses.

Hematology and biochemistry parameters in wild-type and Frzb−/− mice were determined with a Sysmec XE-2100 analyzer (Sysmec, Kobe, Japan) and a Beckman Coulter CX4 analyzer (Analis, Suarlée, Belgium).

Immunohistochemical analysis.

Dissected knees were fixed in 2% paraformaldehyde/phosphate buffered saline and decalcified using EDTA. Paraffin-embedded sections were stained with goat polyclonal anti-FRZB antibody (SC-7427) or rabbit polyclonal anti–β-catenin antibody (SC-7199) (Santa Cruz Biotechnology, Santa Cruz, CA) and peroxidase-linked donkey anti-goat or anti-rabbit antibody (Jackson ImmunoResearch, Newmarket, UK). Digital image analysis of the β-catenin–positive cells versus the total number of cells was performed using ImageJ software (NIH Image, National Institutes of Health, Bethesda, MD; online at:

MMP activity assay.

MMP-3 bioactivity was measured in vitro using a Biotrack activity assay system (Amersham, Buckinghamshire, UK) according to the manufacturer's instructions. Recombinant human FRZB was incubated for 2 hours with recombinant human proMMP-3 (both from R&D Systems, Abingdon, UK) prior to analysis.

Analyses of bone.

ImageJ software was used to measure the subchondral bone area in paraffin-embedded sections that had been stained with hematoxylin and eosin. The area between the articular cartilage and bone marrow lacunae was normalized to the area between the cartilage and the growth plate. Total body (excluding the head) and lumbar spine bone density and lean and fat body mass were determined by dual x-ray absorptiometry using a Piximus densitometer (Lunar, Madison, WI). Trabecular and cortical bone mineral content were assessed by peripheral quantitative CT using an XCT Research M+ system (Norland Medical Systems, Trumbull, CT). Slices of 0.2 mm in thickness were scanned using a voxel size of 0.07 mm. Three scans were taken 2.4–2.6 mm from the distal end of the femur or 1.4–1.6 mm from the proximal end of the tibia to determine trabecular bone parameters. An additional scan was taken 4 mm from the distal end of the femur or 7 mm from the proximal end of the tibia to determine cortical bone parameters. Cortical bone analysis was further performed by microfocal computed tomography (micro-CT) using an AEA Tomohawk system (Philips, Eindhoven, The Netherlands) at a pixel size of 5.1 μm. The middle 4 mm of the diaphysis of the femur and the middle 2 mm of the ulna were quantitatively analyzed using Mimics software (Materialise, Leuven, Belgium).

Assessments of mechanical loading.

The flexed elbow and carpus of 17-week-old mice were placed in 2 opposing cups in a servo-hydraulic mechanical testing device (5848 MicroTester; Instron, Canton, MA). Cyclic compressive loads were applied to the left limb for 2,400 cycles/day at 4 Hz for 10 days, with the right limb serving as nonloaded control. Ex vivo strains were measured using precision strain gauges (Vishay, Malvern, PA) glued to the midshaft of the lateral side of the ulna. Loads were applied with an amplitude of either 4N, resulting in peak strain rates of 17,500 με/second for the Frzb−/− mice and 47,500 με/second for the wild-type mice, or with an amplitude of 2N, resulting in peak strain rates of 8,700 με/second and 17,500 με/second, respectively.

Statistical analysis.

Data were analyzed with SPSS software version 12.0 for Windows (SPSS, Chicago, IL). Data were compared using 2-tailed independent group tests. In cases where normal distribution could be assumed, Student's t-test was used, and the data are presented as the mean ± SEM. In all other cases, the nonparametric Mann-Whitney U test was used. Wilcoxon's signed rank test was used when bone volume increases in loaded and contralateral paws were compared. These data are then presented as the medians, quartiles, and percentiles (10–90%). For all statistical tests, P values less than 0.05 were considered significant. Comparison of regression lines was performed by analysis of covariance (ANCOVA).


Targeted disruption of the Frzb gene.

Frzb−/− mice were generated using a Cre and flippase recombinase strategy to delete the first exon of the gene encoding the WNT-binding frizzled domain as described in Materials and Methods. The targeting construct, wild-type, and modified alleles are shown in Figure 1A. Genotypes were confirmed by genomic PCR analysis (Figure 1B). The successful genetic deletion in Frzb−/− mice was subsequently confirmed by the absence of Frzb messenger RNA (mRNA) in Frzb−/− mice by reverse transcription–PCR (data not shown) and FRZB protein in normal mouse articular cartilage as demonstrated by immunohistochemistry (Figure 1C).

Figure 1.

Generation of Frzb−/− mice. A, Diagram of the Frzb gene, the targeting construct, and the strategy used for the deletion of exon 1 to generate the knockout mice. Primers used for genotyping (p1–4) are also indicated (see Materials and Methods for details). PGK-Neo = neomycin gene under the phosphoglycerin kinase promoter. B, Polymerase chain reaction genotyping of Frzb−/− mice. WT = wild-type. C, Immunohistochemistry for Frizzled-related protein in normal articular cartilage from wild-type and Frzb−/− mice. Bars = 50 μm.

Frzb−/− mice were born with normal Mendelian distribution and were viable without overt phenotype during development and growth. They were indistinguishable from their littermate controls macroscopically as well as histomorphologically upon analysis of different tissues (data not shown). Hematology parameters, liver and kidney function tests, and serum calcium and phosphorus levels were the same in Frzb−/− and wild-type mice, except for the hematocrit values, which were consistently lower in Frzb−/− mice (P = 0.024 by Student's t-test; n = 8 mice per group) (data not shown). The absence of the Frzb gene in Frzb−/− mice did not result in compensatory up-regulation of other Sfrp genes in different tissues expressing Frzb (12, 13), as demonstrated by real-time PCR (data not shown).

Increased cartilage damage in Frzb−/− mice.

Next, we studied postnatal joint homeostasis and disease in the Frzb−/− mice. No differences in sulfated proteoglycan content were found in the cartilage from healthy knees from 10–12-week-old Frzb−/− and wild-type mice of either the CD1 (Figure 2A) or the C57BL/6 (data not shown) background.

Figure 2.

Increased cartilage damage as a result of genetic deletion of Frzb in 3 mouse models of osteoarthritis: papain induced, collagenase induced, and methylated bovine serum albumin (mBSA) induced. There were no differences in the articular cartilage proteoglycan content in cartilage from the healthy knees of 10–12-week-old Frzb−/− and wild-type mice (A); however, Frzb−/− mice had increased cartilage damage at 7 days after papain injection (B), 21 days after collagenase injection (C), and 7 days after mBSA injection (D). The cartilage damage index was calculated by dividing the Safranin O staining intensity of the articular surface by that in the growth plate. Values are the mean and SEM of 12–20 measurements taken in 5–10 mice per genotype per model. = P < 0.05 versus wild-type mice.

We subsequently used 3 models that highlight specific factors that contribute to the pathology of osteoarthritis in humans. In the first model, intraarticular papain injection leads to direct depletion of proteoglycans and cartilage breakdown (23). In the second, intraarticular collagenase administration leads to ligament damage, creating joint instability and accelerated cartilage loss (24). In the third model, acute inflammation leading to destruction is triggered by intraarticular injection of mBSA, which is boosted by subcutaneous injection of interleukin-1 (25). To use this model, mice were backcrossed to the C57BL/6 background necessary for the induction of arthritis.

Loss of sulfated proteoglycans from articular cartilage was increased in Frzb−/− mice compared with wild-type mice 7 days after papain injection (P = 0.0026 by Student's t-test; n = 29 and 23 measurements, respectively) (Figure 2B), 21 days after collagenase injection (P = 0.0185 by Student's t-test; n = 12 and 16 measurements, respectively) (Figure 2C), and 7 days after mBSA injection (P = 0.0361 by Student's t-test; n = 16 and 18 measurements, respectively) (Figure 2D). Scores for inflammatory and fibrotic changes as well as for osteophyte formation were not different between Frzb−/− and wild-type mice in any of the models (data not shown). Also, Frzb−/− mice did not show a higher incidence of spontaneous osteoarthritis of the knee as compared with wild-type mice. Cartilage erosions, synovial lining layer hyperplasia, and small osteophytes were found in 5 of 10 mice in both groups (data not shown).

Association of increased cartilage damage in Frzb−/− mice with active WNT signaling.

Activation of WNT signaling in the articular cartilage of healthy and arthritic mouse joints was studied by immunohistochemistry for β-catenin (Figure 3A). The number of β-catenin–positive cells in healthy cartilage was similar in Frzb−/− and wild-type mice (Figure 3B). Upon injection of collagenase and papain, however, the amount of β-catenin–positive cells increased. Frzb−/− mice showed a higher increase than did wild-type mice, although the difference was not statistically significant for the individual models (for the collagenase model, P = 0.11 [n = 9 and 10 mice, respectively]; for the papain model, P = 0.054 [n = 12 and 11 mice, respectively], by Mann-Whitney U test) (Figure 3B).

Figure 3.

Active WNT signaling in arthritis. A, Immunohistochemical analysis of joint sections from a wild-type mouse and from a Frzb−/− mouse with collagenase-induced arthritis. There are increased numbers of β-catenin–positive cell nuclei in articular chondrocytes in the section from the Frzb−/− mouse. Bars = 50 μm. B, Percentage of β-catenin–positive articular chondrocytes in wild-type and Frzb−/− mice at baseline, at 7 days after papain injection, and at 21 days after collagenase injection. Values are the mean ± SEM of 13 and 10, 12 and 12, and 8 and 8 measurements per genotype, respectively, per model. C, Gene expression patterns in healthy and arthritic wild-type and Frzb−/− mice, as determined using multigene arrays specific for the WNT signaling pathway. The healthy cartilage from Frzb−/− mice and arthritic cartilage from wild-type mice show similarities. Gene names enclosed in boxes indicate that up-regulation or down-regulation of the gene was present in all 3 of the samples studied. Lines indicate up-regulation or down-regulation between groups. Values are the mean of 3 mice per group.

Next, we analyzed gene expression patterns in the articular cartilage using multigene membrane arrays specific for the WNT signaling pathway. In these experiments, we used C57BL/6 mice (either healthy or with mBSA-induced arthritis). By using an inbred strain and a model with low variability, we aimed to limit variations in gene expression that would hinder further analysis.

In healthy cartilage from Frzb−/− mice, 4 genes were consistently expressed at lower levels than in the wild-type mice (Figure 3C): β-catenin (Ctnnb1), C-terminal binding protein (Ctbp1), Fos-like antigen 1 (Fosl1), and mycelomatosis oncogene (Myc). Protein phosphatase 2a (Ppp2ca), Ras homolog (Rhou), WNT inhibitory factor 1 (Wif1), and casein kinase α1 (Csnk1a1) showed a lower expression in 2 of 3 samples from Frzb−/− versus wild-type mice. In contrast, Wnt8b expression was up-regulated in 2 of 3 samples of healthy cartilage from Frzb−/− versus wild-type mice and in all samples in affected versus healthy wild-type cartilage. Ctbp1, Fosl1, and β-catenin–interacting protein (Catnbip1) were down-regulated in all samples from arthritic wild-type mice versus healthy mice (Figure 3C). Csnk1a1 was down-regulated in one-third of the samples from arthritic versus wild-type mice.

These data suggest that the preinduction gene expression profile of the Frzb−/− mice shows similarities to that of arthritic cartilage from the wild-type mice. In addition, Wnt8b appears to be down-regulated in arthritic versus healthy joints of Frzb−/− mice, which suggests a distinct regulation of Wnt ligand expression in the genetic model as compared with the wild-type mice (Figure 3C).

Association of loss of the Frzb gene with increased Mmp3 expression and activity.

Expression of tissue-destructive enzymes in the cartilage of mice with mBSA-induced arthritis was further studied by real-time PCR. On day 7, Mmp3 mRNA was up-regulated in arthritic Frzb−/− mice compared with wild-type mice (range 19.81–94.57-fold increase versus 9.73–16.23-fold increase, respectively, over the level in healthy wild-type controls; n = 3 groups of 3 mice per genotype) (Figure 4A). In contrast, no up-regulation of mRNA for Mmp9, Mmp13, Adamts4, or Adamts5 was detected. We further studied the binding of FRZB to MMP-3 using an in vitro bioactivity assay. In this system, recombinant FRZB dose-dependently inhibited recombinant MMP-3 activity, indicating a direct interaction between the two proteins (Figure 4B), thus providing an additional mechanism for the observed phenotype of increased cartilage proteoglycan loss in the Frzb−/− mice.

Figure 4.

Increased Mmp3 expression and activity in the absence of the Frzb gene. A, Up-regulation of Mmp3 mRNA levels in arthritic Frzb–/– mice as compared with wild-type and healthy (normal) mice. Data points represent RNA pooled from 3 different mice. B, Inhibition of recombinant human matrix metalloproteinase 3 (MMP-3) activity in vitro by recombinant human Frizzled-related protein (FRZB), as measured by a colorimetric in vitro activity assay.

Bone density in Frzb−/− mice.

Increased thickness and density of the subchondral and cortical bone may contribute to osteoarthritis, since these features are associated with loss of the ability to absorb shock within the joint and with increased cartilage damage (3, 27). We analyzed subchondral and cortical bone properties of the femur in Frzb−/− and wild-type mice. No differences in bone area, mineral content, or density were observed in the subchondral bone (data not shown).

In contrast, peripheral quantitative CT of the femur revealed a higher cortical mineral content and thickness in Frzb−/− mice at weeks 10, 16, and 20 in both males and females (Table 1). Increased cortical bone mineral density was also confirmed in the tibia of male and female Frzb−/− mice at weeks 16 and 20 (data not shown). In older mice, the differences were not significant since the variability increased. Cortical bone analysis by micro-CT further corroborated these findings. The midfemoral bone surface area was significantly higher in Frzb−/− mice than in wild-type mice, both in 10-week-old males (mean ± SEM 28.14 ± 1.14 mm2 versus 23.48 ± 0.65 mm2; P = 0.0022 by Student's t-test [n = 10 mice per group]) and in 10-week-old females (22.47 ± 0.34 mm2 versus 20.63 ± 0.47 mm2; P = 0.0061 by Student's t-test [n = 10 mice per group]). In contrast, no consistent differences in trabecular bone in the femur or spine were seen using peripheral quantitative CT or dual x-ray absorptiometry (data not shown). Ovariectomy and orchiectomy did not differentially affect cortical or trabecular bone in Frzb−/− and wild-type mice (data not shown).

Table 1. Properties of cortical bone in the femur of male and female Frzb−/− and wild-type mice*
Parameter, age groupMale miceFemale mice
  • *

    Parameters were determined by peripheral quantitative computed tomography. Numbers of mice per group were as follows: at age 10 weeks, n = 11 mice per group for all groups; at age 11 weeks, n = 10 Frzb−/− and n = 12 wild-type female mice and n = 13 Frzb−/− and n = 14 wild-type male mice; and at age 20 weeks, n = 10 Frzb−/− and n = 12 wild-type female mice and n = 13 Frzb−/− and n = 13 wild-type male mice. Values are the mean ± SD. P values were determined by Student's 2-sided t-test. NS = not significant.

Cortical bone mineral content, mg/mm      
 10 weeks1.56 ± 0.231.37 ± 0.180.0431.41 ± 0.151.18 ± 0.120.002
 16 weeks1.71 ± 0.141.57 ± 0.120.0141.52 ± 0.191.36 ± 0.160.044
 20 weeks1.56 ± 0.161.50 ± 0.12NS1.67 ± 0.211.51 ± 0.100.036
Cortical bone mineral density, mg/cm2      
 10 weeks1,089.9 ± 24.31,051.6 ± 25.70.0021,124.5 ± 17.31,102.8 ± 23.70.026
 16 weeks1,196.5 ± 23.71,162.6 ± 45.20.0301,199.9 ± 36.51,164.7 ± 51.0NS
 20 weeks1,099.9 ± 19.51,071.8 ± 18.70.0011,186.1 ± 33.61,163.5 ± 17.8NS
Cortical thickness, mm      
 10 weeks0.250 ± 0.0210.225 ± 0.0180.0080.245 ± 0.0200.219 ± 0.0200.006
 16 weeks0.277 ± 0.0200.255 ± 0.0170.0080.270 ± 0.0230.240 ± 0.0160.004
 20 weeks0.236 ± 0.0160.230 ± 0.010NS0.264 ± 0.0230.242 ± 0.0120.011

Altered biomechanical properties in the long bones of Frzb−/− mice.

We tested whether increased cortical thickness affects the biomechanical properties of the long bones using a model of axial dynamic compression (28). Strain gauge measurements demonstrated a statistically significant difference in the ex vivo stress–strain relationship in 17-week-old Frzb−/− mice as compared with wild-type controls (P = 0.001 by ANCOVA comparison of linear regression lines; n = 3 mice per genotype) (Figure 5A). Similar loads resulted in lower strains in the Frzb−/− mice, suggesting that Frzb−/− bones are stiffer. This can be explained by the increased cortical bone thickness, which was also detected in the ulna (mean ± SEM midulnar bone volume in Frzb−/− versus wild-type mice, 0.52 ± 0.01 mm3 versus 0.45 ± 0.01 mm3; P < 0.05 by Student's t-test [n = 11 and 22 mice per group]).

Figure 5.

Ex vivo and in vivo loading of the ulna of Frzb−/− and wild-type mice. A, Strain gauge measurements of ulnae performed ex vivo. The ex vivo stress–strain relationship in ulnae from Frzb−/− mice was significantly different from that in ulnae from wild-type controls (P = 0.001 by analysis of covariance comparing linear regression lines; n = 3 mice per group). B, Immunohistochemical localization of Frizzled-related protein (FRZB) in the periosteum of wild-type and Frzb−/− mice. Bars = 50 μm. C, Anabolic response of cortical bone to cyclic mechanical loading in vivo. Application of a cyclic load with an amplitude of 4N resulted in a larger increase in periosteal bone formation in Frzb−/− than in wild-type mice (P < 0.05 by Mann-Whitney U test; n = 8 mice per group). Moreover, ulnae from Frzb−/− mice responded to strain rates that did not elicit an anabolic response in wild-type mice. Each symbol represents an individual mouse: open symbols = Frzb−/− mice; solid symbols = wild-type mice; circles = peak strain rate of ∼7,000 με/second; triangles = peak strain rate of ∼17,500 με/second; diamonds = peak strain rate of ∼47,500 με/second. D, Localization of reactive bone formation in the Frzb−/− model, as determined by microfocal computed tomography (micro-CT). The anabolic response to cyclic loading is mainly periosteal, as illustrated by the overlay of 2-dimensional representative cross-sectional micro-CT images of a nonloaded (gray bone images) and a loaded (bone contours shown as overlay lines) ulna from a Frzb−/− mouse (knockout [KO]) and a wild-type (WT) mouse, respectively.

Differences in cortical bone may result from adaptive processes in growing or mature skeletal elements. We tested the anabolic response of cortical bone to cyclic mechanical loading in the ulnae of mice in vivo, since this process is associated with active WNT signaling (29) and since FRZB is present in the periosteum (Figure 5B). Application of a cyclic load with an amplitude of 4N (2,400 cycles per day for 10 days at 4 Hz) to the ulnae of 17-week-old animals resulted in a significant increase in periosteal bone volume in the wild-type mice (P = 0.01 by Wilcoxon's signed rank test; n = 8 mice per group). This anabolic effect of loading was more pronounced in Frzb−/− mice (P = 0.01 versus the contralateral paw by Wilcoxon's signed rank test [n = 8 mice per group] and P = 0.005 versus wild-type mice by Mann-Whitney U test [n = 8 mice per group]) (Figure 5C).

Since the strain rate, rather than the peak load, determines cortical bone apposition (30), we loaded the ulnae of wild-type and Frzb−/− mice at comparable peak strain rates. The ulnae of the wild-type mice did not respond to a strain rate of ∼17,500 με/second, while the ulnae of the Frzb−/− mice responded with a significant increase in bone volume at this strain rate (P = 0.004 by Wilcoxon's signed rank test; n = 8 and 5 mice per group, respectively) (Figure 5C). Reactive bone formation in this model is predominantly periosteal, with little change at the endosteal side of the cortical bone, as revealed by micro-CT (Figure 5D).


FRZB is a secreted protein that consists of a Frizzled-like cysteine-rich WNT-binding domain and a netrin domain. Its original isolation from a chondrogenic extract of articular cartilage and its expression in the developing bones (10) prompted us to generate Frzb−/− mice in order to study skeletal development, homeostasis, and disease. The Frzb−/− mice showed no overt developmental skeletal abnormalities but were more prone to cartilage damage in different models of osteoarthritis. In addition, the long bones of Frzb−/− mice were stiffer due to increased cortical bone thickness and density, and they showed enhanced anabolic responses to loading. These, in turn, may contribute to cartilage damage, since stress distribution within the cartilage–bone unit of Frzb−/− mice will be different from that of wild-type animals. The increased bone stiffness might, on the other hand, protect against fractures.

The absence of a discernible developmental phenotype was unexpected, given the extent of Frzb expression during morphogenesis and organogenesis (10, 31–33). While it is surprising, targeted inactivation of another member of the Frzb family, Sfrp1, also did not show a developmental phenotype (15, 34), whereas only 3% of Sfrp2−/− mice showed hind limb syndactyly (34). In contrast, deficiency of both Sfrp1 and Sfrp2 was embryologically lethal, with severe shortening of the thoracic region due to problems with anteroposterior axis elongation and somitogenesis (34). These double-knockout mice also have short limbs and syndactyly. These data support the existence of a functional redundancy between different WNT antagonists, at least during development. In contrast, targeted disruption of the dickkopf homolog 1 (Dkk1) gene, which inhibits low-density lipoprotein receptor–related protein (Lrp) WNT coreceptors, is lethal (35). Of interest, some behavioral abnormalities using specific tests have been suggested in another strain of Frzb−/− mice ( However, we did not notice any abnormal behavior of our knockout mice.

Increased cartilage degeneration and osteoarthritis-like changes in Frzb−/− mice and in humans with reduced FRZB activity can be explained by several contributing mechanisms. We hypothesized that increased WNT signaling in the articular cartilage in the absence of Frzb could explain the findings in the knockout mice. WNT signaling through β-catenin stabilization and its nuclear translocation has been associated with negative regulation of early chondrogenesis and stimulation of chondrocyte hypertrophy during development (36, 37). FRZB antagonizes WNT signaling during such developmental processes (38). In vitro data suggest that WNT signaling stimulates both the expression and the activity of tissue-destructive enzymes, including MMPs (37). Furthermore, active β-catenin signaling has been demonstrated in a surgically induced mouse model of osteoarthritis (39). These observations are further supported by the gene expression profile of the Wnt pathway observed in the Frzb−/− and wild-type mice described herein. Similar genes were up- or down-regulated in healthy Frzb−/− cartilage and in arthritic wild-type cartilage as compared with those in normal wild-type cartilage. The precise function of the identified pathway components in cartilage homeostasis and disease remains to be clarified. In this context, up-regulation of Wnt8b expression in arthritic cartilage is of particular interest, since the antagonistic interaction between Frzb and Wnt8 has been clearly demonstrated in Xlaevis and in chick chondrocyte cultures (11–13, 38).

Interaction of FRZB with MMPs, as demonstrated here for MMP-3, may provide a local protective mechanism within the articular cartilage where FRZB is expressed. The sFRP molecules, including FRZB, consist not only of a WNT-binding domain, but also of a netrin domain (12, 13). Interactions between this domain and MMPs have been hypothesized (17, 18). Both SNPs associated with osteoarthritis (19) are localized within the netrin domain. However, these SNPs also result in reduced WNT inhibition in vitro, suggesting that interactions between both domains of the protein are important for its proper function (19).

The articular cartilage forms a biomechanical unit with the subchondral and cortical bone in order to attenuate forces through joints, particularly following impact loading (40). Frzb deficiency in mice resulted in a thicker cortical bone, with increased stiffness and higher cortical appositional bone formation after mechanical loading within physiologic ranges. This could contribute to the development of osteoarthritis by increasing strain on the articular cartilage during normal locomotion. These anabolic responses can be explained by increased canonical WNT signaling as recently demonstrated (29, 41). It is noteworthy that genetic models of FRZB and sFRP-1 deficiency both result in increased bone formation postnatally. However, Frzb−/− mice only show increased cortical thickness without changes in the trabecular bone, whereas the opposite is true for Sfrp1−/− mice (15). These data further highlight the complexity of the WNT signaling in bone biology (4).

The consequences of the targeted inactivation of Frzb in mice corroborate the reported genetic association of the gene with osteoarthritis in humans. Moreover, the increased cortical density of the long bones in Frzb−/− mice supports our earlier observation that polymorphisms in the human FRZB gene are differentially associated with hip osteoarthritis and osteoporotic hip fractures in selected cohorts (21). Therefore, the regulatory effects of FRZB on cartilage and bone may provide a molecular basis for the longstanding hypothesis that osteoarthritis and osteoporosis show an inverse clinical relationship. Consequently, systemic modulation of FRZB levels may not be recommended in the treatment of these disorders, since it appears to have inverse effects on cartilage and bone. However, local modulation of FRZB levels may be of more value in these diseases, particularly in osteoarthritis, since this is a process that mostly affects a single or a few load-bearing joints, such as the knee or hip.


Drs. Lories and Luyten had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Lories, Peeters, Bakker, Tylzanowski, Derese, Schrooten, Thomas, Luyten.

Acquisition of data. Lories, Peeters, Bakker, Tylzanowski, Derese, Schrooten, Thomas, Luyten.

Analysis and interpretation of data. Lories, Peeters, Bakker, Tylzanowski, Derese, Schrooten, Thomas, Luyten.

Manuscript preparation. Lories, Bakker, Luyten.

Statistical analysis. Lories.


We thank K. Reekmans, B. Weynants, E. Vanherck, D. Vanderschueren, G. Kerckhofs, and S. Marcelis for their technical assistance with this study.