The identification of LRP5 mutations in bone disorders linked the Wnt/β-catenin pathway to the field of bone biology. Loss-of-function mutations in LRP5 were found in patients with osteoporosis-pseudoglioma (OPPG), a recessive disorder characterized by low bone density and persistent eye/vitreal vasculature [5]. Obligate carriers for the OPPG mutations are also more at risk for developing osteoporosis. The phenotype of Lrp5^−/− knock-out mice was identical to the human OPPG disease and revealed an impairment in osteoblast proliferation [6]. The converse is found in patients with dominant “gain-of-function” mutations in LRP5 which exhibit abnormally high bone density [7, 8]. A transgenic mouse model carrying a human LRP5 high bone mass mutation exhibited significant increases in bone density and mechanical properties [9, 10]. Studies have shown that removing one Lrp6 copy from Lrp5^−/− mice increases the osteoporosis severity, suggesting that both LRP5 and LRP6 are critical determinants of bone mass [11].

Further studies using conditional removal or activation of β-catenin at different times during development have revealed multiple roles that the Wnt pathway plays in osteoblast proliferation and maturation [12]. The absence of β-catenin in early skeletal progenitor cells causes osteoblast precursors to become chondrocytes [13–15]. Inactivation of β-catenin in committed osteoblast progenitors prevented the expression of osteoblast-specific genes, indicating that β-catenin is also essential for osteoblast differentiation [15, 16]. Mice with β-catenin removal in mature osteoblasts displayed severe osteopenia, which surprisingly was due to an increase in osteoclasts [17, 18]. Normally, mature osteoblasts will secrete a RANKL decoy receptor, osteoprotegrin (OPG), to inhibit excess osteoclastogenesis [19]. OPG is a Wnt responsive target gene and was found to be reduced in β-catenin^−/− osteoblasts and upregulated in cells with hyperactive Wnt signaling [17, 18, 20]. Therefore the Wnt/β-catenin pathway directly controls multiple steps of osteoblast development, and indirectly controls osteoclast differentiation through OPG secreted from osteoblasts.

Wnt signaling is modulated by several families of secreted inhibitors, which bind either competitively to the Wnt ligand, such as sFRP, or directly to the LRP5/6 receptor (SOST and Dkks) [21]. Consistent with the removal of a Wnt inhibitor, the mouse knockout for Sfrp1^−/− displays an increase in trabecular bone mass in adult animals [22]. Previously identified as a BMP antagonist, SOST also binds to LRP5 and LRP6 in the extracellular region and antagonizes Wnt signaling [23, 24]. SOST loss-of-function mutations are responsible for Sclerosteosis, a human disorder characterized by increased bone density and excessive bone overgrowth [25, 26]. A subset of SOST mutations are caused by a deletion of a 52-kb downstream regulatory region in patients with Van Buchem disease [27, 28]. Sclerosteosis and Van Buchem disease are highly similar to LRP5 high bone mass disorders suggesting a possible molecular link [4]. Indeed our recent study suggests that SOST binding is diminished in LRP5 high bone mass mutations, resulting in SOST-resistant LRP5 receptors [29].

Dkk1^+/− and Dkk1^+/d animals appear normal and do not display any visible phenotype. Although Dkk1^d/− animals are viable, they consistently display microphthalmia, forelimb polydactyly and vertebral fusions in the caudal vertebrae. Some of the Dkk1^d/− mice have mandibular defects that result in misalignment of the front incisors, however the animals with severe malocclusions were not used in this study. Femora and calvariae were dissected from 8-week old mice and the left femora and parietal bone regions were used for microCT scans. Animal weights were not significantly different between females; however the male Dkk1^d/− mice weighed significantly less than the wild-type males. (Table 1)

Grossly the femora from Dkk1^d/− mice appeared morphologically different. The third trochanter, normally associated as the insertion site for the gluteus maximus, was shifted distally and extended along the lateral edge into the diaphysis. (Fig 2A-D) Additionally all Dkk1^d/− femora exhibited a 0.3–0.5 mm wide fibrous ligament-like structure that ran from the distal, posterior edge of the third trochanter to the lateral sesamoid bone and attached to the lateral condyle of the femur. This accessory ligament was thinner in the Dkk1^d/− females and also in the smaller Dkk1^d/− males by weight. Examination of Dkk1^+/− and Dkk1^+/d femora revealed a thin <0.1 mm fibrous element in a similar location, however this structure was loosely attached and could easily be pulled away from the bone. The femora from Dkk1^+/+ mice did not have this fibrous element. The distally shifted third trochanter and accessory ligament in the Dkk1^d/− femur is similar to the Hoxa10^−/− limb phenotype [45]. However, unlike the severely affected Hoxa10^−/− mutants, the knee joint and surrounding sesamoid bones were not affected in the Dkk1^d/− mice.

To quantify the geometric differences in the Dkk1 mutants, femur length and third trochanter position were measured from the microCT scans. (Table 1) The average femur length was significantly lower in the Dkk1^d/− males, consistent with the decreased weight in this group. To determine the third trochanter position relative to the total femur length, the ratio of femur length to the vertical distance of the third trochanter was calculated. This revealed significant differences between the wild type and Dkk1 mutant femora, particularly in the male mice.

To examine the properties of the cortical bone, a 3 mm region of the femur mid-diaphysis was analyzed using a previously described method [44]. Compared to the wild-type controls, cortical thickness and cross sectional area were significantly increased in the Dkk1^+/− and even more so in the Dkk1^d/− samples. (Table 2) The calculated moment of inertia for the medial-lateral axis (Iyy) was used as an estimate of the mechanical properties of the femur and significant increases were found in the female Dkk1^d/− and the male Dkk1^+/− and Dkk1^d/− samples. No changes were detected in tissue mineral density, suggesting that the degree of mineralization of cortical bone within the area analyzed was equivalent between the sex matched groups. Cross section alpha blends of the mid-diaphysis illustrate the increase in voxel density in the Dkk1^+/− and Dkk1^d/− mice. (Fig. 2E–J) Trabeculation within the marrow space was commonly observed in the Dkk1^d/− male femora.

To evaluate the alterations in trabecular bone from the Dkk1 mutant mice, a region of the distal femur was used for the trabecular analysis. (Fig. 3A-H) Measurements of bone volume fraction show significant increases in all Dkk1 mutant mice compared to wild-type controls. (Table 3) Furthermore the incremental increases in bone volume fraction within the geneotypes (+/+ < +/d < +/− < d/−) indicate an inversely proportional relationship to Dkk1 expression levels. Significant increases in trabecular thickness were found in the Dkk1^+/− and even more so in Dkk1^d/− mice, compared to sex matched wild-type controls. Trabecular number was also increased and trabecular spacing decreased, further supporting the measured overall increase in trabecular mass with progressively lower levels of Dkk1 expression. Tissue mineral density was unchanged between most of the groups, with the exception of the Dkk1^+/− males which displayed a decrease in mineralization. Taken together with the increase in trabecular bone mass for the Dkk1^+/− males, it is likely that a higher proportion of more recently formed trabecular bone is present within this group. Measurements from the trabecular analysis plotted using the Dkk1 expression levels of d/− = 25%, +/− = 50%, +/d = 75%, +/+ = 100% graphically demonstrate the correlation between Dkk1 dosage and trabecular bone properties. (Fig. 3I–L)

To date there have been no reported human mutations in DKK1, although null mutations would most likely be incompatible with life due to the early requirement of DKK1 during embryonic head induction. Nonetheless, our findings suggest that even a 25% reduction of DKK1 expression will produce significant increases in trabecular bone mass. It is possible that reduced DKK1 levels (or activity) will resemble Van Buchem disease, a high bone mass disorder caused by reduced SOST expression. Both DKK1 and SOST act through LRP5/6 to inhibit Wnt/β-catenin signaling and both are expressed in mature osteoblasts and osteocytes suggesting a similar biological role in bone homeostasis [24]. In addition, the hypomorphic Dkk1^d allele bears a resemblance to the Van Buchem mutation that affects SOST expression. The Dkk1^d mutation is a 60 kb deletion at the transgene insertion site which is located over 150 kb downstream from the Dkk1 gene and the Van Buchem disease mutation results from a 52-kb deletion of a regulatory region that is 35 kb downstream from SOST [46]. It is unknown if the reduced expression from the Dkk1^d allele is due to transgene induced silencing or the deletion of a regulatory element [37]. However, our study implies that a human high-bone mass disorder originating from hypomorphic mutations at the DKK1 locus or within the surrounding genomic region is probable.

Dkk1^d/− mice display forelimb and hindlimb polydactyly, but since no skeletal abnormalities were observed in the stylopod structures (humerus or femur) of newborn mice, we chose femora for the microCT study. However, we now report that the adult femora of Dkk1^d/− mice exhibit a distal shift in the third trochanter and an accessory ligament that runs along the lateral edge connecting the third trochanter to the lateral condyle of the femur. These irregular structures were less pronounced in the smaller female and male Dkk1^d/− mice suggesting that either size or weight influences this phenotype. Additionally, slight changes in third trochanter location in Dkk1^+/− and Dkk1^+/d mice, and a thin fibrous element along the lateral femur, indicate that phenotype severity is related to Dkk1 dosage. Curiously, a distally shifted third trochanter and ectopic ligament is also found in Hoxa10^−/− mice, and the Dkk1^d/− femora are remarkably similar to the mildly affected Hoxa10^−/− mice on an inbred 129/Sv background [45]. The paralogous Hox10 genes are responsible for patterning the stylopod and stylopod/zeugopod junction (elbow and knee joints), particularly in the hindlimb [47]. Favier et al speculate that the ectopic ligament in Hoxa10^−/− originates from a thickening of the septum intermuscularis femoris lateralis, and the abnormal thickening of this structure may be due to the antagonistic tensions between the gastocnemius and the gluteus superficialis muscles [45]. As a result, the third trochanter is subject to an alteration in forces generated by the gastrocnemius which could cause the remodeling and distal displacement of this tuberosity. In contrast to the Dkk1 and Hoxa10 mutants, mice deficient for myostatin (Gdf8^−/−), a negative regulator of muscle mass, display a proximal displacement and enlargement of the third trochanter due to the increased forces from an enlarged gluteus maximus [48]. This further suggests that the third trochanter is subject to remodeling in the adult mouse, and displacement of the third trochanter in the Dkk1 mutant femora is a mechano-response to loads imparted from a distal muscle of the hindlimb. However, we would like to note that in the Dkk1^d/− mutants, the distal edge of the third trochanter extends into the mid-diaphyseal region used to analyze the cortical bone traits. This adds further complexity to the cortical bone geometry for these animals and may affect our conclusions regarding the Dkk1^d/− cortical bone phenotype.

The Dkk1 mutant mice used in the study were maintained on a C3H background. During the original characterization of the doubleridge mouse (Dkk1^d/d), we observed a higher incidence of anterior head malformations in mice from the C57BL/6 (B6) strain compared with C3H Dkk1 mutants [37]. Furthermore, Dkk1^d/− mice are not viable on the B6 background (BTM unpublished observation) which necessitated a different strain background in order to study the adult bone phenotype. Recently the high bone mass phenotype of the Dkk1^+/− heterozygote was described using mice on the B6 background [42]. Bone density varies greatly among the mouse inbred strains, particularly between the C3H and B6 strains. Total density is approximately 50% higher in C3H femora compared to B6, and cortical bone properties are consistently higher in C3H mice [49, 50]. Osteoblasts from C3H mice exhibit higher alkaline phosphatase activity and greater mineralization compared to cultured B6 osteoblasts, which contributes to the increased bone formation rate observed in the C3H inbred strain [51]. Nevertheless, our study concludes that even on a higher bone density mouse strain like C3H, reduction of Dkk1 is able to significantly increase bone mass. In addition, we also observed sex-specific differences in our 8-week old bone samples on the C3H background. Measurements of cortical bone were higher in males, whereas trabecular BVF and trabecular thickness were higher in females. Yet compared to sex-matched wild-type controls, the relative changes in bone properties were consistent between females and males in the Dkk1 mutant mice (Fig. 2,​,3).3). Morvan et al examined the tibiae from 12 week old B6 mice and did not find any sex differences in their samples; and in their study male and female samples were analyzed together [42]. However the sex-specific differences may be smaller in the low bone density B6 background. Due to the differences in mouse strain background, it is difficult to directly compare the relative increases in bone properties from the heterozygous Dkk1^+/− mice between our studies. However in our study we have uncovered the unique bone phenotype of the Dkk1^d/− hypomorph/null and our examination of the Dkk1^+/d mice expressing 75% of the normal Dkk1 levels reveals the minimum Dkk1 reduction needed to produce significant changes in trabecular bone mass.

Due to high variation in the calvaria measurements, we are unable to determine what effect Dkk1 reduction has on intramembranous bone with our current sample sizes. In addition to the bone apposition differences between endochondral and intramembranous bone, the parietal bones are exposed to very different and less dynamic mechanical loading patterns than the femora [51]. A recent study has shown that osteoblasts from Lrp5^−/− mice are defective for mechanotransduction and fail to synthesize bone matrix after mechanical loading [52]. Therefore, the reduction of a Wnt inhibitor such as DKK1 should result in an increased LRP5-mediated osteogenic response to mechanical stress. This may explain why dramatic changes in bone properties were only found in the limb bones that are subjected to large variations in repetitive mechanical loading. Presently, the lack of strong evidence of increased calvaria BMD suggests that Dkk1 deficiency may not result in the severe craniosynsystosis present in the SOST loss of function diseases.