Targeted Disruption of the Wnt Regulator Kremen Induces Limb Defects and High Bone Density

Abstract

Kremen1 and Kremen2 (Krm1 and Krm2) are transmembrane coreceptors for Dickkopf1 (Dkk1), an antagonist of Wnt/β-catenin signaling. The physiological relevance of Kremen proteins in mammals as Wnt modulators is unresolved. We generated and characterized Krm mutant mice and found that double mutants show enhanced Wnt signaling accompanied by ectopic postaxial forelimb digits and expanded apical ectodermal ridges. Triple mutant Krm1^−/− Krm2^−/− Dkk1^+/− mice show enhanced growth of ectopic digits, indicating that Dkk1 and Krm genes genetically interact during limb development. Wnt/β-catenin signaling also plays a critical role in bone formation. Single Krm mutants show normal bone formation and bone mass, while double mutants show increased bone volume and bone formation parameters. Our study provides the first genetic evidence for a functional interaction of Kremen proteins with Dkk1 as negative regulators of Wnt/β-catenin signaling and reveals that Kremen proteins are not universally required for Dkk1 function.

Wnt proteins and their receptors play important roles in development, differentiation, and disease, and their activity is regulated by a number of transmembrane and extracellular proteins (10, 12, 31, 39). The Wnt coreceptors low-density lipoprotein receptor-related protein 5 and 6 (LRP5 and LRP6) are essential for signal transmission via the β-catenin pathway (20) and are negatively regulated by Dickkopf1 (Dkk1), a member of a small family of secretory proteins (38). Dkk1 binds to LRP6 and thereby acts as a potent Wnt antagonist. In addition, Dkk1 and LRP6 can form a ternary complex with Kremen1 and Kremen2 (Krm1 and Krm2), which are single transmembrane-spanning proteins that are high-affinity receptors for Dkk1 (8, 29, 30) and which also can directly bind to LRP6 (19). The ternary LRP6/Dkk1/Kremen complex is rapidly endocytosed, leading to the inhibition of Wnt/β-catenin signaling (30).

Dkk1 null mutant mice are embryonic lethal. Besides anterior head truncations, they also show fused vertebrae and limb defects (2, 28, 33), which are consistent with the important role of Wnt signaling in regulating the patterning and growth of the vertebrate limb (11). For example, Wnt3 mutants have a reduced apical ectodermal ridge (AER) (7, 46), a signaling center that controls limb growth. Similarly, the overexpression of Dkk1 in chicks induces limb truncation, and this is accompanied by apoptosis (18, 33). Conversely, overactivated β-catenin induces the expansion of the AER (46). Expanded AER also is observed in both Dkk1 null and hypomorphic Dkk1^d/d mutants, suggesting a genetic interaction with Wnt3 (2, 28, 33). In addition to AER expansion, Dkk1 null and Dkk1^d/d mutants display postaxial polysyndactyly in the forelimbs (28, 33). Consistent with the polydactyly in Dkk1^d/d mice resulting from too-high levels of Wnt signaling, normal digit numbers are restored in Dkk1^d/d Lrp6^+/− mice (28). Polydactyly in Dkk1 mutants can be reversed by the simultaneous loss of Wnt7a (2). Thus, Dkk1 can control different steps involving Wnt3 or Wnt7a signaling during mouse limb development and digit patterning.

By various mechanisms, Wnt/β-catenin signaling also promotes bone formation, including the renewal of stem cells, osteoblast proliferation, the induction of osteoblastogenesis, and the inhibition of osteoblast and osteocyte apoptosis (17, 24). The N-terminal gain-of-function mutant proteins of LRP5 occurring in patients with high bone mass (e.g., G171V) show reduced affinity to Dkk1. Therefore, high bone mass in these patients likely is due to LRP5 derepression (3, 9). Similarly, Lrp5^−/− mice show low bone mass due to the reduced proliferation of precursor cells (23). Conversely, mice overexpressing the G171V LRP5 mutant have high bone mass (5). Heterozygous Dkk1^+/− mice are viable, but they show a strong increase in bone mineral density (32). Thus, Wnt-Lrp5 signaling is under the negative control of Dkk1 to regulate the physiological levels of bone mass.

While the role of Dkk1 in development has been well characterized, the role of its coreceptors Krm1 and Krm2 is less certain. Kremen genes are evolutionarily conserved in vertebrates and are differentially expressed during mouse and frog development (14, 35). Krm1 is required for thymus epithelium formation in mice by acting as a Wnt inhibitor (40). Antisense morpholino knockdown experiments in Xenopus laevis embryos showed that Krm1 and Krm2 function synergistically with Dkk1 in inhibiting Wnt/β-catenin signaling in embryonic head formation (14). Recently, we found that Krm2 can function independently from Dkk proteins during neural crest induction in Xenopus. In the absence of Dkk1, Krm1 and Krm2 promote Wnt/β-catenin signaling by binding to LRP6 and enhancing LRP6 protein levels at the plasma membrane (13, 19).

It remains unclear what roles Kremen proteins have during mouse development. A particularly important question is whether there is an absolute requirement for Kremen proteins in Dkk1-mediated Wnt inhibition or whether this interaction is physiologically relevant only in certain tissues. Transgenic Dkk1 misexpression now is widely used to probe the role of Wnt/β-catenin signaling in the mouse, and a universal requirement for Krm would have important consequences for the design and interpretation of such experiments. Furthermore, recently it has been reported that Krm1 acts as a receptor for R-spondins, a family of extracellular Wnt modulators (8). This is a controversial claim, since previous reports indicated that R-spondin acts via LRP6 and Frizzled (37, 49).

To address these questions, we have generated and characterized Krm1 Krm2 double mutant mice. Our results indicate that Kremen genes are negative regulators of Wnt/β-catenin signaling that interact with Dkk1 during limb development and that are required for normal bone formation. However, Kremen genes are not universally required for Dkk1 function. Furthermore, our results do not support the hypothesis that Kremen proteins mediate R-spondin signaling.

MATERIALS AND METHODS

Animals.

Mice were kept according to international standard conditions, and all animal experiments complied with local and international guidelines for the use of experimental animals. Kremen1 knockout mice were generated by Deltagen, Inc. (San Mateo, CA), using a targeting construct that leads to the deletion of 70 nucleotides in exon 2 of Krm1 and the insertion of a 6.9-kb internal ribosomal entry site-lacZ reporter and neomycin resistance cassette. Kremen2 null mice were generated by a conventional gene-targeting strategy by replacing sequences of exons 1 to 6 of Krm2 with a selection cassette. Krm1 Krm2 double-knockout mice were generated from double heterozygous mice by interbreeding. Triple mutant mice (Krm1^−/− Krm2^−/− Dkk1^+/−) were obtained after mating Krm1^−/− Krm2^−/− mice with Dkk1^+/− mice and interbreeding triple heterozygous mice. All mouse lines were maintained and analyzed in CD1 outbred and C57BL/6 congenic genetic backgrounds.

Genotyping.

Adults, newborns, and embryos were genotyped by gene-specific triplex PCRs using the following primers (lengths of wild-type-mouse- and knockout mouse-specific PCR products, respectively, are in parentheses). For Krm1, primers were K1-p1, 5′-GGCCGCCCAAGATCTAGCAAAACAT-3′; K1-p2, 5′-CCAGAACAGACATGGCTTCCCACCT-3′; and K1-p3, 5′-AGACAAACGCACACCGGCCTTATTC-3′ (214 bp and 377 bp); for Krm2, the primers were K2-p1, 5′-AAACCTGGGTGAGGGAGTCT-3′; K2-p2, 5′-AATGGGAAGAGGAGGAGGAA-3′; and K2-p3, 5′-GTTTTCCCAGTCACGACGTT-3′ (370 bp and 452 bp); for Dkk1, the primers were D1-p1, 5′-AGAACTAACCCAGCCCCACAGCAGA-3′; D1-p2, 5′-CTCCTCAGGGAAGACAACAAAGCCG-3′; and D1-p3, 5′-GGGTCAACTCCAAGTTCCACCCAAA-3′ (287 bp and 423 bp). Detailed PCR conditions are available upon request.

Immunoprecipitation and Western blotting.

Protein extracts were prepared by homogenizing freshly isolated tissue in lysis buffer containing 150 mM NaCl, 20 mM Tris-HCl, pH 7.4, 1% Triton X-100, 5 mM MgCl₂, and protease inhibitors (Roche). Lysates were cleared by centrifugation, and equal amounts of total protein from wild-type and knockout mouse tissue were subjected to immunoprecipitation using anti-mouse Krm1 antibody (R&D Systems) and protein G-agarose (Roche) overnight at 4°C. Bound proteins were eluted in Laemmli buffer and analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and Western blotting.

Cell culture.

All cell lines were maintained in Dulbecco's modified essential medium supplemented with 1% l-glutamine, 1% penicillin-streptomycin, 10% fetal calf serum and were grown in 5% CO₂. Mouse Wnt3a-conditioned medium was produced as previously described (15). Xenopus Dkk1- and human Rspo3-conditioned media were produced by the transfection of HEK293T cells with pCS2+-Xdkk1 or pCS2+-ΔC-hRspo3-flag and the harvesting of medium fractions. Primary mouse embryo fibroblasts (MEFs) were isolated and cultured as previously described (34). For the preparation of primary cells from embryonic day 10.5 (E10.5) whole embryos or adult mouse kidneys, tissues were manually dissociated and incubated in 0.5% trypsin for 5 or 30 min at 37°C. Primary cells were plated in 24-well plates and stimulated for 4 or 6 h with conditioned medium 1 day after plating.

Quantitative real-time PCR.

RNA from cells was isolated and transcribed into cDNA using TRIzol and Superscript II reverse transcriptase according to the manufacturer's protocol (Invitrogen). Quantitative real-time PCR was performed with Sybr green and the Roche LightCycler 480 system. Primer sequences for Axin2 were the following: forward, 5′-GCAGCAGATCCGGGAGGATGAA-3′; and reverse, 5′-GATTGACAGCCGGGGGTCTTGA-3′. Primers for Actb were the following: forward, 5′-GACCCAGATCATGTTTGAGACCT-3′; and reverse, 5′-AGGTCCAGACGCAGGATG-3′. Relative Axin2 expression (Axin2/Actb) was normalized to that of unstimulated control samples.

Whole-mount in situ hybridization.

Whole-mount in situ hybridization was performed according to a protocol adapted from Wilkinson (50). For AER markers, the treatment of the embryos with proteinase K was omitted or reduced to 4 min at 5 μg/ml. RNA probes were transcribed from plasmids containing full-length open reading frames. Where indicated, whole-mount in situ-stained limb buds were paraffin embedded, sectioned (10 μm), and counterstained with eosin according to standard protocols (34).

Skeletal preparations.

Adult mouse limbs and E17.5 embryos were skinned, eviscerated, and fixed in 95% ethanol for 3 to 5 days. Cartilage was stained with 0.15 mg/ml Alcian blue (Polyscience) in 80% ethanol, 20% acetic acid for 24 h. After two washes in 95% ethanol, samples were cleared in 1% (wt/vol) potassium hydroxide and then stained for bone with 0.05 mg/ml Alizarin red S (Sigma) in 2% (wt/vol) potassium hydroxide for 2 h. Samples were rinsed in water, further cleared in 2% (wt/vol) potassium hydroxide, and transferred to 100% glycerol.

Bone phenotyping.

The bone phenotype was determined for 12-week-old Dkk1^+/−, Krm1^−/−, Krm2^−/−, Krm1^−/− Krm2^−/−, Krm1^−/− Krm2^−/− Dkk1^+/−, and wild-type littermate mice. To label bone mineralization fronts, mice were given calcein (20 mg/kg; Sigma, St. Louis, MO) by subcutaneous injections 10 and 3 days before sacrifice. For osteocalcin level measurements, blood was collected on the day of sacrifice. Urinary deoxypyridinoline cross-link levels (D-Pyr) were measured using a colorimetric assay from Pacific Biometrics Inc. (Tampa, FL) and were normalized to the creatinine concentration (measured by a Metra creatinine assay kit [Quidel, San Diego, CA]) to correct for water excretion as described previously (45). The intra-assay coefficient of variation was 3.1 to 4.8%, and the interassay coefficient of variation was 4.3 to 8.4%.

Mouse tibias were recovered from 12-week-old mice following sacrifice and were used for tomodensitometric and histomorphometric analysis. For tomodensitometry, right tibias were fixed overnight in 3.7% formaldehyde in phosphate-buffered saline (PBS), washed in PBS, and then stored in 70% ethanol. Micro-computed tomography (μCT) scans of the metaphyseal region were performed at an isotropic resolution of 9 μm to obtain trabecular bone structural parameters. Using a two- and three-dimensional (2D and 3D) model and a semiautomatic contouring algorithm, we determined three-dimensional bone volume, bone surface, and trabecular thickness. Three-dimensional images were obtained on a Scanco Medical micro-CT scanner (μCT 20; Scanco Medical AG, Bassersdorf, Switzerland). A total of 450 images were obtained from each bone sample using a 512 by 512 matrix, resulting in an isotropic voxel resolution of 18 by 18 by 18 μm³. Measurements were stored in 3D image arrays with an isotropic voxel size of 9 μm. A constrained 3D Gaussian filter was used to partly suppress the noise in the volumes. The bone tissue was segmented from marrow using a global thresholding procedure. In addition to the visual assessment of structural images, morphometric indices were determined from the microtomographic data sets. Cortical and trabecular bones were separated using a semiautomated contour-tracking algorithm to detect the outer and inner boundaries of the cortex. In trabecular bone, basic structural metrics were measured using direct 3D morphometry (22, 42). For histomorphometry, the fixed samples then were embedded in methylmethacrylate as previously described (43). Four-micrometer sections were stained with von Kossa stain to quantify the structural parameters (i.e., bone volume, trabecular number, and thickness) and with toluidine blue to measure the cellular parameters (i.e., osteoblast and osteoclast numbers). Unstained sections were used to assess the dynamic parameters (i.e., bone formation rate and mineral apposition rate). A standard histomorphometric analysis of the tibial metaphysis was performed (41) using the Osteomeasure system (Osteometrics, Inc., Decatur, GA). The measurements were performed in a 1.12-mm² area starting 0.3 mm from the proximal growth plate. All histomorphometric measurements were performed in a double-blind fashion, with respect to the treatment regimens, by one individual.

RESULTS AND DISCUSSION

Kremen mutant mice are viable and fertile.

We generated constitutive Krm1 and Krm2 knockout mice using conventional knockout approaches. In Krm1 knockout mice, a 70-bp fragment of exon 2 was replaced by a lacZ reporter and neomycin resistance cassette (Fig. 1A). In Krm2 knockout mice, the majority of the protein-coding region was deleted and replaced by a selection cassette (Fig. 1B). To confirm that the targeted alleles are functional null mutations that completely abolish the production of active protein, we performed an immunoprecipitation of Krm1. Western blot analysis showed that a 68-kDa band corresponding to Krm1 is absent in Krm1^−/− mouse tissue lysates (Fig. 1C). The absence of Krm2 transcripts in Krm2^−/− tissue was verified by whole-mount in situ hybridization (Fig. 1D) and real-time PCR (not shown).

FIG. 1.

Targeted mutagenesis indicates that Kremen proteins are not essential mediators of Dkk1 and R-spondin function. (A and B) Diagrams of the Krm1 (A) and Krm2 (B) loci, targeting constructs, and knockout alleles after homologous recombination. Targeted exons and homologous sequences are indicated. Small arrows mark positions of primers for triplex PCR genotyping. Diagrams are not drawn to scale. Dta, diphtheria toxin α-subunit; EII, exon 2; EI-VI, exons 1 to 6. (C) Immunoprecipitation and Western blot analysis show the presence of Krm1 protein in wild-type mouse tissue lysates and its absence in Krm1^−/− mouse tissue lysates. A cross-reactive band is shown as a loading control. (D) Whole-mount in situ hybridization shows the absence of Krm2 transcripts in Krm2^−/− embryos. (E) Representative genotypic frequencies of 400 mice obtained from crossing mice that were double heterozygous for Krm1 and Krm2. (F and G) Real-time PCR analysis of Axin2 expression in MEFs (F) or adult kidney cells (G) from wild-type and mutant mice that were treated with the indicated secretion factors (*, P < 0.05).

Since Krm1^−/− and Krm2^−/− mice were viable and fertile and did not exhibit any obvious abnormalities, we produced Krm1 Krm2 double knockouts by intercrossing Krm1 and Krm2 single mutants. Homozygous Krm1^−/− Krm2^−/− mice also were viable and fertile and did not exhibit unusual premature mortality. All genotypes produced by intercrossing double heterozygous mice were recovered at the expected Mendelian ratio (Fig. 1E). Also, after intercrossing homozygous mice, normal litter sizes and the normal implantation of embryos in the uterus were observed. These results do not support a recent claim, based on antisense inhibition, that Krm1 functions in embryo implantation and uterine receptivity (25). Furthermore, the requirement of Krm1 and Krm2 in Xenopus neural crest development (19) does not seem to be evolutionarily conserved, since no changes in neural crest markers were observed in Krm double mutant mice (not shown).

Kremen proteins are not essential mediators of Dkk1 and R-spondin function.

We isolated primary cells to analyze if the knockout of Krm1 and Krm2 affects the activation of Wnt/β-catenin signaling by Wnt3a or its inhibition by Dkk1. MEFs stimulated with Wnt3a-conditioned medium upregulate the expression of the direct Wnt target gene Axin2 (Fig. 1F) (21). In two independent Krm1^−/− Krm2^−/− MEF lines, the Wnt3a-induced upregulation of Axin2 was elevated compared to that of wild-type cells, supporting the idea that Kremen proteins are negative regulators of Wnt/β-catenin signaling (Fig. 1F and G). Elevated Wnt signaling was not restricted to MEFs but was also found in adult kidney (Fig. 1G) and whole E10.5 embryos (see Fig. S1 in the supplemental material) of mutant mice. However, the treatment of Wnt3a-stimulated MEFs with Dkk1-conditioned medium shows that Dkk1 can efficiently inhibit Wnt/β-catenin signaling despite the absence of Krm1 and Krm2 (Fig. 1F).

To test if Kremen proteins are essential mediators for R-spondin/Wnt signaling, we tested primary kidney cells from wild-type and mutant mice. Axin2 was upregulated by stimulation with either Wnt3a- or Rspo3-conditioned medium in wild-type as well as mutant cells (Fig. 1G), and the treatment of cells with a combination of Rspo3 and Wnt3a further enhanced Axin2 expression in both genotypes. Our results indicate that the R-spondin-mediated activation of Wnt/β-catenin signaling does not require Kremen proteins. This conclusion also is supported by the fact that R-spondin knockout mouse phenotypes (4, 36) are much more severe than the Kremen double mutant phenotypes (see below). These results do not support the notion that Kremen proteins are essential mediators of R-spondin function (8).

Kremen proteins and Dkk1 interact during limb development.

While Krm1 Krm2 double knockout mice were viable, 74% (n = 140) showed ectopic postaxial forelimb digits (Fig. 2A and B). The size of these ectopic skeletal elements was variable, with approximately half of them being smaller (n = 45/140) and the other half larger (n = 59/140) than the distal phalanx of adjacent digit V. The ectopic digits were outgrowths of digit V rather than originating from the carpus, as regular digits do.

FIG. 2.

Krm1 Krm2 and Dkk1 functionally interact during forelimb formation. (A) Morphology of embryonic (E17.5) and adult forelimbs in the ventral view. Cartilage and bone were stained with Alcian blue and Alizarin red. Double and triple mutants exhibit postaxial polydactyly (arrows). (B) Statistics of the extra-digit phenotype. (C) Whole-mount in situ hybridization of the indicated genes. Shown are (from left to right) dorsal, frontal, and posterior views of E11.5 forelimbs.

Limb defects similar to those in Krm knockouts are found in Dkk1 mutants (33). We therefore tested for a genetic interaction between Dkk1 and Krm1 Krm2 mice by generating triple mutants. While Dkk1^+/− mice have normal limbs (Fig. 2B) (33), heterozygosity for Dkk1 in triple mutant Krm1^−/− Krm2^−/− Dkk1^+/− mice enhanced the frequency and size of ectopic digits. Furthermore, 40% of the extra digits originated at the carpus, like regular digits (Fig. 2A and B), which was never observed in Krm knockouts. In contrast to Dkk1^−/− embryos, which exhibit polysyndactyly (33), webbing of bone elements was found in neither Krm1^−/− Krm2^−/− nor Krm1^−/− Krm2^−/− Dkk1^+/− mice.

Kremen proteins and Dkk1 attenuate Wnt signaling in the developing limb to allow normal limb patterning.

Consistent with their role in limb development, Krm1, Krm2, and Dkk1 are coexpressed in the AER and also are present at slightly deeper levels in the mesenchyme underlying the AER (Fig. 2C; also see Fig. S2 in the supplemental material). Interestingly, Dkk1 expression is differentially distributed from distal to more proximal portions of the limb bud. Distally, Dkk1 expression is restricted to the AER and the adjacent mesenchyme. Proximally, Dkk1 expression is mesenchymal and vanishes in the ectoderm (see Fig. S2 in the supplemental material). We therefore examined if the knockout of Krm1 Krm2 affects Wnt signaling in the AER. The expression of Axin2, a marker for Wnt/β-catenin signaling (21), is concentrated in the distal AER but is absent from the anterior and posterior margins of the AER in wild-type embryos (Fig. 3A). In Krm1^−/− Krm2^−/− and Krm1^−/− Krm2^−/− Dkk1^+/− embryos, Axin2 expression in the forelimb buds clearly was expanded toward more proximal parts of the AER (Fig. 3A). Axin2 expression was increased in the AER of mutant limbs but not significantly enhanced in the mesenchyme (see Fig. S3A in the supplemental material), consistent with derepressed Wnt/β-catenin signaling in the AER, in which Kremen genes and Dkk1 are coexpressed.

FIG. 3.

Kremen genes are required for AER patterning. Whole-mount in situ hybridization of marker genes in mutant embryos is indicated. (A) Shown are dorsal and posterior views of E10.5 (Axin2 and Shh) or E11.5 (Fgf8) forelimb buds. Red arrowheads in Axin2 and Shh indicate gene expression boundaries and a widening expression gap, respectively. (B) In the upper part, sections from E10.5 forelimbs show the restriction of Lmx1b expression to the dorsal side; in the lower part, dorsal, posterior, and frontal views of E10.5 forelimb buds (Msx1) are shown. A green arrowhead indicates expanded posterior Msx1 expression. A, anterior; P, posterior; D, dorsal; V, ventral; d, distal; and p, proximal.

To characterize the molecular consequences of increased Wnt signaling, markers for limb patterning were analyzed. The expression of Fgf8, the earliest AER marker and an important regulator of limb growth, clearly was expanded in the posterior limb bud of mutant embryos, the region in which extra digits develop, coinciding with expanded Axin2 expression (Fig. 3A). Shh, which is expressed in the posterior mesenchyme complementary to Fgf8, was reduced in mutant limbs in the region underlying the expanded AER, leading to a bifurcation of the Shh expression domains (Fig. 3A; also see Fig. S3B in the supplemental material).

Unlike with Axin2, the expression of Lmx1b, a downstream target of Wnt7a signaling, was not altered in the developing limbs of mutant embryos (Fig. 3B). Consistent with this finding, defects in dorsoventral patterning, which is mediated by Wnt7a signaling, were not detected in mutant limbs, suggesting that Krm and Dkk1 specifically regulate limb patterning on the level of Wnt3-mediated AER induction.

The results indicate that Kremen proteins and Dkk1 cooperate as inhibitors of Wnt3/β-catenin signaling during AER induction and maintenance, thereby fine-tuning downstream target gene expression to maintain the normal structure and function of the AER. However, Dkk1 has additional Kremen-independent functions during limb patterning and development. In Dkk1^−/− mice, the webbing of digits and the downregulation of apoptosis mediating Msx1 is observed (18, 33). In contrast, no syndactyly (Fig. 2A) or decreased Msx1 expression (Fig. 3B) was observed in limbs of Krm1^−/− Krm2^−/− or Krm1^−/− Krm2^−/− Dkk1^+/− mice compared to that of the limbs of wild-type mice. Conversely, we observed a slight expansion of the posterior Msx1 domain in Krm1^−/− Krm2^−/− limb buds (Fig. 3B). As AER ectoderm-mesenchyme interactions are known to regulate Msx1 in the murine limb bud (48), this is a likely consequence of posterior AER expansion.

Taken together, our results suggest that Kremen proteins and Dkk1, by inhibiting Wnt signaling, delimit the posterior boundary of the AER. Krm loss leads to increased Wnt3/β-catenin signaling and AER expansion. It is well established that AER signaling regulates mesenchymal cell numbers in the nascent limb bud (47) and, ultimately, the number of digit condensations that form. Various limb mutants with expanded AER display polydactyly phenotypes (1, 26, 27), and conversely, when AER and mesenchymal mass are reduced, fewer digits develop. Therefore, the ectopic digit formation in Krm mutants is fully in line with these findings.

Kremen proteins regulate bone formation.

Human genetic studies have pointed out the critical roles of Wnt/β-catenin signaling in bone metabolism and bone formation (for a review, see reference 6). This motivated us to investigate the bone phenotype of Kremen mutants, including Krm1^−/−, Krm2^−/−, and Krm1^−/− Krm2^−/− animals, which we compared to the bone phenotype of Dkk1 heterozygous animals (Dkk1^+/−). Furthermore, we assessed whether the deletion of both Krm1 and Krm2 in Dkk1^+/− (Krm1^−/− Krm2^−/− Dkk1^+/−) mice further enhances the high bone mass observed in Dkk1^+/− mice (32).

Tibias from 12-week-old animals were analyzed by μCT for bone volume. An analysis of bones from Krm1^−/− or Krm2^−/− mice showed no difference in bone volume compared to those from wild-type littermates (data not shown), suggesting redundant functions between these two Dkk1 coreceptors in bone tissue. In contrast, the bone volume in Krm1^−/− Krm2^−/− double mutants was significantly higher than that of wild-type littermates and in both genders (Fig. 4A). As described previously (32), Dkk1^+/− mice displayed increased bone volume that was comparable to that observed in Krm1^−/− Krm2^−/− mice. In addition, the deletion of one allele of Dkk1 in Krm1^−/− Krm2^−/− mice (Krm1^−/− Krm2^−/− Dkk1^+/− animals) did not result in a further increase of bone volume (Fig. 4A). Collectively, these data indicate that deleting both Krm1 and Krm2 is equipotent to deleting Dkk1 in increasing bone mass.

FIG. 4.

Knockout of Kremen induces high bone mass. (A) Tomodensitometry analysis of metaphysis tibial bone volume of 12-week-old mice. (B) Photomicrographs of coronal sections through proximal tibias of wild-type and Krm1^−/− Krm2^−/− 12-week-old mice stained with von Kossa stain. Note that the number of trabeculae is strongly increased in Krm1^−/− Krm2^−/− animals of both genders. The numbers represent the trabecular bone volume of each genotype (*, P < 0.05; **, P < 0.005). BV, bone volume; TV, total volume.

To further confirm the increase in bone mass observed in Krm1^−/− Krm2^−/− mice and to investigate the underlying mechanism, we performed a histomorphometric analysis of the secondary spongiosa in the tibial metaphysis. As shown in Fig. 4B and Table 1, the bone mass of Krm1^−/− Krm2^−/− mice was markedly increased compared to that of wild-type mice (Fig. 4B). Trabecular bone volume was significantly elevated in both males and females. While trabecular thickness was unchanged, the number of trabeculae was significantly increased, and trabecular separation was decreased in Krm1^−/− Krm2^−/− mice (Table 1). Bone formation parameters all were very significantly elevated in Krm1^−/− Krm2^−/− mice, including those for the osteoblast surface, the number of osteoblasts, and the osteoid surface. The dynamic histomorphometry confirmed this marked increase in bone formation, with the mineral apposition rate and the mineralized surface each being increased, leading to a pronounced increase in the bone formation rates (Table 1). The changes observed in both males and females were very similar overall (data not shown).

TABLE 1.

Bone structural, cellular, and dynamic parameters of wild-type and Krm1^−/− Krm2^−/− male mice

Parametera	Value for indicated mice
	Wild type	Krm1−/−Krm2−/−b
Structural
BV/TV (%)	10.44 ± 3.41	24.11 ± 5.73**
Trabecular thickness (μm)	46.46 ± 10.17	50.65 ± 9.31
Trabecular separation (μm)	428.23 ± 136.65	163.12 ± 33.98**
No. of trabeculae/mm	2.24 ± 0.62	4.76 ± 0.63**
Cellular
Osteoid surface/BS (%)	0.90 ± 0.69	4.24 ± 2.23*
Osteoblast surface/BS (%)	6.77 ± 1.32	25.14 ± 6.86**
No. of osteoblasts/μm of BS	2.94 ± 1.91	13.73 ± 3.57**
Eroded surface/BS (%)	1.22 ± 0.55	0.78 ± 0.31
Osteoclast surface/BS (%)	0.70 ± 0.31	0.54 ± 0.20
No. of osteoclasts/mm of BS	0.57 ± 0.21	0.43 ± 0.19
Dynamic
MS/BS (%)	17.75 ± 4.29	26.23 ± 3.43*
MAR (μm/day)	1.24 ± 0.45	1.81 ± 0.10*
BFR/BS (μm3/μm2/day)	83.25 ± 38.62	170.76 ± 25.80**
BFR/BV (%/year)	378.34 ± 177.89	693.50 ± 175.90*
BFR/TV (%/year)	39.97 ± 25.02	159.82 ± 7.78**
D-Pyr (nM)/creatinine (mM)	18.11 ± 1.5	17.08 ± 2.6

^aAbbreviations: BV, bone volume; TV, total volume; BS, bone surface; MS, mineralized surface; MAR, mineral apposition rate; and BFR, bone formation rate.

^b*, P < 0.05; **, P < 0.005.

In contrast to bone formation, there was no significant modification in bone resorption parameters, such as those for the osteoclast surface, the number of osteoclasts, and the eroded surface in Krm1^−/− Krm2^−/− male mice (Table 1). In addition, urinary D-Pyr levels were similar in Krm1^−/− Krm2^−/− animals and their wild-type littermates, confirming that the loss of Kremen proteins has no impact on bone resorption (Table 1). There was, however, a trend toward a decrease in the histological parameters of bone resorption in males that became statistically significant in females (data not shown). Based on these results, we propose that Kremen proteins and Dkk1 promote bone formation more efficiently than they inhibit bone resorption and osteoclast activity via β-catenin (16).

The results indicate that Kremen proteins are required to negatively regulate bone formation, consistent with their function as Wnt inhibitors. While no enhanced phenotype was observed with Krm1^−/− Krm2^−/− Dkk1^+/− triple mutants, the similarity between the Krm1 Krm2 null mutant and Dkk1 heterozygous mutant phenotypes strongly suggests that Kremen proteins function together with Dkk1 in Lrp5 modulation in bone. Our finding that Kremen proteins negatively regulate bone formation may provide new opportunities for therapeutic intervention in, e.g., osteoporosis (6).

In conclusion, Krm1 and Krm2 show functional redundancy and are not universally required for Dkk1 function. Dkk1 mutants display embryonic lethality, head truncations, and vertebral and limb defects. In contrast, Kremen mutants are viable and share only limb defects with Dkk1 homozygous mice and high bone density with Dkk1 heterozygous mice. Thus, in many cells, the ability of Dkk1 to prevent Wnt-Lrp6 interaction (44) may be sufficient for effective Wnt antagonism.