Contributions of Dickkopf‐1 to Obesity‐Induced Bone Loss and Marrow Adiposity

ABSTRACT Low bone strength in overweight individuals is a significant medical problem. One important determinant of mesenchymal stem cell fate into osteoblasts or adipocytes is the Wnt signaling pathway. We recently showed that Dickkopf‐1 (DKK1), a potent Wnt inhibitor, is upregulated in obese mice. In this study, we investigated the role of DKK1 in the pathogenesis of obesity‐induced bone loss using global and tissue‐specific KO mice. Obesity was induced in 8‐week‐old male mice with an inducible global (Rosa26‐CreERT2) or osteoprogenitor‐ (Osx–Cre‐) specific deletion of Dkk1 with a high‐fat diet (HFD) containing 60% fat. After 12 weeks, body weight, bone volume, bone fat mass, and bone turnover were assessed. Dkk1 fl/fl ;Rosa26‐CreERT2 mice experienced a similar increase in body weight and white fat pads as control mice. A HFD significantly reduced trabecular bone mass and the bone formation rate in Cre‐ mice and Dkk1 fl/fl ;Rosa26‐CreERT2 mice. Interestingly, Dkk1 fl/fl ;Rosa26‐CreERT2 mice were protected from HFD‐induced cortical bone loss. Furthermore, a HFD was associated with increased bone marrow fat in the femur, which was less pronounced in Dkk1 fl/fl ;Rosa26‐CreERT2 mice. Mice with an osteoprogenitor‐specific Dkk1 deletion showed similar results as the global knockout, showing a protection against HFD‐induced cortical bone loss and an accumulation of bone marrow fat, but a similar decrease in trabecular bone volume. In summary, DKK1 appears to contribute distinctly to cortical, but not trabecular bone loss in obesity. © 2020 The Authors. JBMR Plus published by Wiley Periodicals, Inc. on behalf of American Society for Bone and Mineral Research.


Introduction
O besity is a worldwide health problem with more than 1.9 billion adults overweight in 2014 and over 650 million people obese. (1) These numbers are predicted to increase to 2.16 billion overweight and 1.12 billion obese adults by 2030. (2) Besides causing a high economic burden, the consequences of overweight and obesity on health are manifold. Overweight is a risk factor for type 2 diabetes mellitus as well as heart and vascular diseases leading to increased mortality. (3) In addition, overweight affects bone metabolism and reduces bone strength. (4) Previously, high-body weight was considered protective against the development of osteoporosis, which is characterized by decreased bone quality and increased fracture risk, (4,5) as BMI is positively correlated with BMD. However, recent studies show that an elevated abdominal fat tissue and BMI are relevant risk factors for osteoporosis in both sexes. (4,(6)(7)(8) Moreover, although fracture risk is frequently reported to be lower at the proximal femur and vertebra in obese adults, it is increased at the proximal humerus, upper leg, and ankles. (9) Therefore, obesity is not per se protective against fractures.
Mechanistically, the positive association of BMI and BMD may stem from increased mechanical loading and enhanced aromatase activity in fat tissue (10,11) as well as the bone-anabolic action of fat-derived hormones. (12,13) However, visceral fat is also associated with inflammation as reflected by the higher secretion of proinflammatory cytokines such as TNFα and IL-6, which negatively affect bone metabolism. (14)(15)(16) Finally, obesity leads to an increase in bone marrow fat, a condition that usually occurs in aged individuals. (17)(18)(19)(20)(21)(22) The increase in bone marrow adiposity may indicate a shift in the fate decision of mesenchymal stem cells (MSCs) from osteoblasts to adipocytes, (23) thereby increasing fat accumulation and reducing osteoblastic activity. The Wnt signaling pathway plays an important role in this process as Wnt activation supports osteoblastogenesis and at the same time, inhibits adipogenesis. (24) Wnt6, Wnt10, and Wnt10b especially contribute to MSC fate decision via β-catenin-dependent mechanisms. (25) Recently, Dickkopf-1 (DKK1) and sclerostin (SOST), two important Wnt inhibitors, have emerged as promising new targets for anabolic therapies. (26) Dkk1 expression was shown to be transiently upregulated during adipogenesis in humans and correlated with an inhibition of the canonical Wnt signaling. (27) Furthermore, Dkk1 overexpression promotes adipogenesis, (27) whereas siRNA-mediated knockdown of Dkk1 inhibits adipogenesis. (28) Both global deletion of SOST and SOST antibody treatment resulted in an increased trabecular bone volume and a decrease in the number of bone marrow adipocytes, as well as a decrease in adipocyte size. (29) Recently, we and others demonstrated that serum and skeletal levels of DKK1, but not SOST, are elevated in obese mice (20,30) and in patients with type 2 diabetes mellitus. (31)(32)(33) As DKK1 is a potent suppressor of bone formation and bone mass, (34)(35)(36) we hypothesized that elevated DKK1 levels may drive obesityinduced bone loss in mice. To test this hypothesis, we fed Dkk1 fl/fl ;Rosa26-ERT2-Cre and Dkk1 fl/fl ;Osx-Cre mice with a highfat diet (HFD) and analyzed bone mass and bone metabolism, as well as bone marrow adiposity. We found that DKK1 plays a site-specific role in obesity-induced bone loss in mice, contributing to cortical, but not trabecular bone loss.
JBMR Plus (WOA) water (10 mg/mL in a 3% sucrose solution) ad libitum to repress Cre activity during embryogenesis. Dkk1 fl/fl ;Osx-Cre offspring received doxycycline drinking water until the age of 5 weeks. Respective Cre-negative littermates were used as controls. By suppressing Cre activity during embryogenesis, no effects on bone were observed in Dkk1 +/+ ;Osx-Cre + mice. (36) Breeding of the mouse lines was approved by the institutional animal care committee of the Technische Universität (TU) Dresden and the Landesdirektion Sachsen.

In vivo experiments
All animal procedures were approved by the institutional animal care committee of the TU Dresden and the Landesdirektion Sachsen. All mice were fed a standard diet with water ad libitum and were kept in groups of four animals per cage for the whole experiment. Mice were exposed to a 12-hour light/dark cycle in an airconditioned room at 23 C (no specific pathogen-free room), and housed in cardboard houses with bedding material. Mice were randomly assigned to treatment groups; subsequent analyses were performed in a blinded fashion. Male mice are commonly used for HFD interventions (37)(38)(39) and were therefore chosen for all experiments. To mimic an excess uptake of fat, mice were fed a HFD (60% fat, 20% carbohydrate, and 20% protein; Research diets #12492, Research Diets, Inc., New Brunswick, NJ, USA) at the age of 8 weeks for 12 weeks. Control mice continued to receive the normal diet (ND: 9% fat, 58% carbohydrates, and 33% protein; Sniff #V1534-300, Research Diets, Inc., New Brunswick, NJ, USA). Animal cohort sizes were as follows: Dkk1 fl/fl ;Rosa26-CreERT2: 104 Cre-negative ND: 12, Cre-negative HFD: 10, Cre-positive ND: 8, Cre-positive HFD: 10 and Dkk1 fl/fl ;Osx-Cre: Cre-negative ND: 14, Cre-negative HFD: 12, Cre-positive ND: 12, Cre-positive HFD: 10. Weight and blood glucose were measured every 4 weeks.
Additional methods (glucose tolerance test; assessment of bone mass, microarchitecture, and fat content; histology; histomorphometry; RNA isolation; qRT-PCR; and statistics) are reported in the online Supplemental Material.

HFD increases adiposity regardless of DKK1 expression
To assess the contribution of DKK1 to the pathogenesis of obesity-induced bone loss, we subjected Dkk1 fl/fl ;Rosa26-CreERT2 (global Dkk1 cKO) mice to a HFD for 12 weeks. Dkk1 fl/fl ;Rosa26-CreERT2 and Cre-negative control mice gained a similar amount of weight when fed a HFD (40% to 45%), whereas mice fed a ND only gained 16% to 19% body weight after 12 weeks (Fig. 1A). Furthermore, the HFD decreased glucose tolerance in both genotypes (Fig. 1B).  Data represent the mean AE SD (n = 10 to 14/group). Statistical analysis was performed using two-way ANOVA.
p Values from ND versus HFD. The HFD-induced gain in body weight in both genotypes was accompanied by a significant increase in their percentage of body subcutaneous and epigonadal fat mass compared with mice fed a ND (Fig. 1C,D). Gene expression analysis revealed an increased fatty acid binding protein (Fabp4) and reduced adiponectin (AdipoQ) expression after a HFD in subcutaneous and epigonadal adipose tissue of Dkk1 fl/fl ;Rosa26-CreERT2 and Crenegative control mice, whereas Pparγ and Cebpα expression was not altered (Fig. 1E,F). As obesity also causes low-grade chronic inflammation, (40,41) we investigated the expression of the inflammation markers Tnfα, Il-1β, and Il-6 in subcutaneous and epigonadal fat. Tnfα and Il-6 expression was similarly increased in global as well as control mice fed a HFD, whereas Il-1β was not altered (Fig. 1G,H).

DKK1 may contribute to the HFD-induced increase in bone marrow adiposity
Obesity is associated with an accumulation of fat in the bone marrow. Thus, we counted adipocyte ghosts on tissue sections of the tibia and performed osmium staining on the femur. A HFD increased the adipocyte number (+70-77%) and adipocyte area (+59-77%) in both genotypes ( Fig. 2A and Table 1). However, analyzing the entire bone marrow fat content using μCT, bone marrow fat tissue in global cKO mice was increased to a smaller extent than in their WT littermate controls (Fig. 2B). Although Fabp4 and AdipoQ showed a similar regulation after HFD in both genotypes, the reduction of Cebpα expression after HFD was less pronounced in global cKO mice (Fig. 2C). Pparγ expression was not affected by a HFD. Similar to fat tissue, a HFD increased the expression of the inflammation markers Tnfα and Il-6 in both groups (Fig. 2D). Furthermore, global cKO mice fed a ND exhibited a lower Tnfα expression, when compared with littermate controls. Global DKK1 deletion does not protect against obesityinduced cancellous bone loss, but protects cortical bone loss Dkk1-proficient mice showed a significant reduction of vertebral trabecular bone volume (−14%) and trabecular number (−10%) after a HFD, whereas trabecular thickness was not affected (Fig. 3A-C). Furthermore, Cre-negative mice showed a significant increase in trabecular separation (+10%; Fig. 3D). Similarly, global Dkk1 cKO mice lost vertebral bone structure, even though to a smaller extent (Fig. 3A-D). Similar results were found in the femur (Table 1) HFD mice exhibited a significantly reduced cortical thickness (−7%) and BMD (−3%), whereas global cKO mice were protected from bone loss (Fig. 3E,F).
Global DKK1 deletion tends to reduce the HFD-mediated reduction of bone formation The HFD in control mice led to a reduction of P1NP serum levels (−28%), a reduced mineral surface per bone surface (−21%), a reduced number of osteoblasts (−32%), and a decreased bone formation rate (−31%) in the appendicular skeleton, whereas the mineral apposition rate (MAR) was not altered (Fig. 4A,B and Supplemental Table S2). All these parameters were reduced to a lower extent in global Dkk1 cKO mice, yet the interaction effect did not show a significant difference (Fig. 4A,B and Supplemental Table S2). Similar results were found in the tibias of the mice (Table 1). Local Dkk1 expression was significantly increased in Cre-negative control mice fed a HFD when compared with a ND, and was not detectable in mice with a global Dkk1 deletion (Fig. 4C). SOST expression was increased under HFD conditions and was further increased in global cKO bones (Fig. 4C). Runtrelated transcription factor 2 (RUNX2), alkaline phosphatase (Alp), and osteocalcin (Ocn) expression was significantly reduced after a HFD in both genotypes.
Furthermore, CTx serum levels were significantly increased (+27%) after HFD in control mice, but not in Dkk1 cKO mice (Fig. 4D). The number of osteoclasts was not altered by a HFD (Fig. 4E).
Osteoprogenitor-specific DKK1 deletion does not protect against HFD-induced adiposity As the effects of HFD on bone were mitigated in the cortical compartment of global DKK1-deficient mice, we further elucidated (I) Serum Dickkopf-1 (DKK1) serum levels were assessed using commercially available ELISAs. Gene expression levels were normalized to Rpl26. Data represent the mean AE SD (n = 10 to 14/group). Statistical analysis was performed by two-way ANOVA for the effect of genotype and HFD and the interaction. For weight and GTT area under the curve was determined. *p < 0.05, **p < 0.01, ***p < 0.001 for ND versus HFD. #p < 0.05, ##p < 0.01, ###p < 0.001 versus respective Cre-negative control.
whether osteoprogenitor-derived DKK1 contributes to a HFDinduced bone loss. Cre-negative as well as Dkk1 fl/fl ;Osx-Cre mice gained a similar amount of weight over time when fed a ND or a HFD (Fig. 5A). Also, a HFD decreased glucose tolerance in Dkk1 fl/fl ;Osx-Cre and control mice (Fig. 5B).
A HFD also was associated with an increased percentage of subcutaneous (+68-71%) and epigonadal (+75-79%) fat mass in Dkk1 fl/fl ;Osx-Cre and control mice (Fig. 5C,D). The HFD upregulated Fabp4 expression in both subcutaneous and epigonadal fat tissue, whereas it reduced AdipoQ expression (Fig. 5E,F). Except for subcutaneous fat tissue of Dkk1 fl/fl ;Osx-Cre mice, Cebpα expression was significantly downregulated after a HFD, whereas Pparγ expression was not altered. Osteoprogenitor-specific Dkk1 deletion did not influence their gene expression when compared with Cre-negative controls fed a ND. Again, the HFD was associated with an increased expression of Tnfα and Il-6 in subcutaneous and epigonadal fat tissue, whereas Il-1β expression was not affected (Fig. 5G,H).
Furthermore, a HFD was associated with increased DKK1 serum levels in Cre-negative controls, but not in osteoprogenitor-specific Dkk1 cKO mice, in which DKK1 serum levels are dramatically reduced (Fig. 6I). (34) As seen before in global cKO mice, HFD had no impact on SOST serum level (Cre-negative: ND: 155 AE 13, HFD: 149 AE 17, Cre-positive: ND: 183 AE 13, HFD: 176 AE 143. However, osteoprogenitor-specific DKK1 deletion is associated with an increased SOST serum level. Osteoprogenitor-derived DKK1 has a minor influence on bone marrow adiposity Dkk1 fl/fl ;Osx-Cre and control mice showed both a significant increase in the adipocyte number (+83-91%), adipocyte area (+65-77%), and bone marrow fat mass (+36-48%) after a HFD, albeit increased bone marrow fat was less pronounced in Dkk1 fl/fl ;Osx-Cre mice (Fig. 6A,B and Table 2). In both groups, a HFD resulted in an increased expression of Fabp4, whereas Cebpα and AdipoQ expression were reduced in the bone tissue (Fig. 6C). Pparγ expression was not altered. Tnfα and Il-6 expression were increased in both genotypes after a HFD, albeit osteoprogenitor-specific Dkk1 cKO mice fed a ND exhibited an overall reduced Tnfα expression (Fig. 6D).
Osteoprogenitor-specific DKK1 deletion protects against obesity-induced cortical bone loss, but not trabecular bone loss bone after a HFD (−17-23%; Fig. 7A). Furthermore, both showed a significant reduction of trabecular number (−9%), no changes in trabecular thickness, whereas trabecular separation was significantly increased (+7-13%) after a HFD (Fig. 7B-D). Similar results were found in the femurs of these mice (Table 2). However, only Cre-negative HFD mice exhibited a significantly reduced cortical thickness (−6%), whereas cortical bone was not affected in Dkk1 fl/fl ;Osx-Cre mice (Fig. 7E). Both groups showed no changes in cortical BMD (Fig. 7F). Even though only Cre-negative mice exhibited significantly reduced P1NP serum levels after a HFD, both groups showed a reduced bone formation rate (−24-54%) and mineral surface per bone surface (−21%), whereas the MAR was not altered (Fig. 8A,B and Supplemental Fig. S2). The HFD increased local Dkk1 expression only in the bones of Cre-negative mice (Fig. 8C). Sost expression was significantly increased in both HFD groups, with Dkk1 fl/fl ;Osx-Cre mice showing an overall higher Sost expression compared with their Cre-negative controls (Fig. 8C). RUNX2 expression was only significantly decreased in control mice fed HFD, whereas Alp and Ocn were reduced in both groups. The Dkk1 fl/fl ;Osx-Cre mice showed a higher Alp and Ocn expression compared with littermate controls. A HFD increased CTx serum levels in Dkk1 fl/fl ;Osx-Cre and Cre-negative control mice, whereas the number of osteoclasts was not altered (Fig. 8C,D).

Discussion
Obesity is becoming a global epidemic, which is associated with chronic low-grade inflammation and metabolic dysregulation including alterations in bone metabolism and strength. (3,4) The underlying mechanisms, however, remain poorly understood. Recently, we detected an increased amount of DKK1-expressing osteoblasts and elevated DKK1 serum levels in obese mice. (20) Thus, in this study we tested the hypothesis that DKK1 plays a role in obesity-induced negative effects on bone.
First, our data support the notion that HFD induces bone loss in mice, which is controversially discussed in the literature. Although most studies also report bone loss in young, adolescent, and mature obese rodents on a HFD, (18,(42)(43)(44) some find no alterations (17,45,46) or even increased bone mass. (47) These discrepancies may be provoked by different age and sex, varying intervention times (6 to 12 weeks), and diet compositions (45% to 60% fat in chow), highlighting the importance of detailed investigations and standardized procedures.
Our key finding was that global and osteogenic deletion of Dkk1 protected mice from obesity-induced cortical, but not trabecular bone loss. Distinct effects of some components of the Wnt signaling pathway on cortical versus trabecular bone have been described previously. For example, Wnt16 has distinct effects only on cortical, but not trabecular bone, suggesting a differential homeostatic regulation between the cortical and trabecular bone compartments. (48,49) Furthermore, it should be mentioned that although a HFD has detrimental effects on cancellous bone mass, the effects on cortical bone mass in rodents are quite diverse. Although some studies have reported increases (47,50) or decreases in cortical bone mass, (37,51) the majority of studies report no difference between normal and HFD. (52)(53)(54) In our study, cortical bone loss was not profound, but consistent among the experiments. Thus, though under homeostatic conditions Dkk1 is critical for both the cortical and trabecular compartments, (34) in obesity it drives cortical bone loss.
As suppression of bone formation is one of the main mechanisms of HFD-induced bone loss, we speculated that osteogenic deletion of DKK1 would ameliorate bone loss. However, a HFD still reduced trabecular bone formation parameters in Dkk1 fl/fl ; Osx-Cre mice, which, as previously published, were generally higher in Dkk1 fl/fl ;Osx-Cre mice compared with Cre-negative controls. (34) This indicates that blocking osteogenic DKK1 is not sufficient to protect from HFD-induced bone loss and raises the question, which cell type may then be involved. DKK1 is  produced by numerous cells, including osteogenic cells, platelets, adipocytes, fibroblasts, and vascular cells. (27,(55)(56)(57) As obesity is associated with an expansion of adipose tissue and adipocytes express DKK1, it may be envisaged that DKK1 produced from this cell type may play a major role. In fact, a previous study by Gao and colleagues found an increased expression of Dkk1 in epididymal adipose tissue of HFD-fed mice. (30) Furthermore, the natural compound embelin efficiently inhibited adipogenesis and lipogenesis in vitro, which was likely caused by an activation of canonical Wnt signaling, as embelin treatment attenuated the induction of Dkk1 in adipose tissue in HFD-fed mice.
Previously, we detected increased DKK1 serum levels in obese mice, whereas the serum concentration of another important Wnt inhibitor SOST was unaffected. (20) Other studies, however, have shown increased skeletal SOST levels under HFD conditions, which we also confirm in this study. Moreover, because of the well-documented negative feedback loop between DKK1 and SOST, (58,59) skeletal expression of SOST was also higher in Cre-positive mice. Nonetheless, the increase in skeletal SOST levels was not sufficient to balance the lack of Dkk1 expression and therefore did not alter bone mass. Interestingly, serum SOST levels were found to be positively correlated with abdominal and gonadal fat, as well as with biochemical markers related to metabolic disease in postmenopausal women. (60) Additionally, SOST knockout or antibody treatment results in reduced adipose tissue accumulation, which is associated with increased insulin sensitivity and improved metabolic parameters. (61) This indicates that Wnt inhibitors are influenced by obesity and may have endocrine functions between adipose tissue and the skeleton. A human study showed that DKK1 serum levels are not correlated with insulin levels or insulin resistance, but to BMI, waist circumference, and total-body adiposity. (62) Therefore, DKK1 is discussed as a potential novel marker for adiposity. (62) However, we did not previously find any differences between serum DKK1 levels in lean, overweight, and obese individuals. (20) It should be mentioned that the investigated human cohort did not include patients diagnosed with type 2 diabetes, whereas the HFD model used in this study is known to result in prediabetes, including an impaired glucose tolerance as well as insulin resistance. (63) We therefore assume that DKK1 could be influenced by insulin sensitivity as obese mice with impaired glucose tolerance and insulin sensitivity showed increased DKK1 serum levels in the current and previous study. (20) Deletion of β-catenin in pancreatic progenitors led to a decreased β-cell mass and impaired glucose tolerance. (64) Furthermore, this loss of β-catenin made the mice resistant to HFD-induced obesity and insulin resistance, implicating an important role of β-catenin in the regulation of metabolism and energy homeostasis. Thus, Wnt signaling may modulate the susceptibility to diabetes by acting on different tissues. Feeding mice a HFD resulted in the accumulation of subcutaneous and epigonadal fat depots as well as bone marrow adiposity in Cre-negative mice, which validates the existing literature. (17,20,65) The amount of body fat also increased to a similar extent in Cre-positive Dkk1 fl/fl ;Osx-Cre mice, along with similar dysregulations in adipogenic markers, which are typically associated with the development of insulin resistance. (66)(67)(68) As obesity leads to an increased secretion of proinflammatory cytokines such as TNFα and IL-6, we also investigated them in this study. (69,70) These cytokines are known to promote the development of insulin resistance, type 2 diabetes mellitus, and elevate osteoclastogenesis. (69,(71)(72)(73) The expression of Tnfα in bone was less pronounced in bone of ND-fed Cre-positive mice of both Dickkopf-1 (Dkk1), sclerostin (SOST), runt-related transcription factor 2 (RUNX2), alkaline phosphatase (Alp), and osteocalcin (Ocn) mRNA expression in ulna bone tissue was analyzed using real-time PCR analysis. Gene expression levels were normalized to β-actin. (D) Serum CTx was measured using ELISA. (E) Tartrate-resistant acid phosphatase (TRAP) staining was used to determine the number of osteoclasts/bone parameter (N.Oc/b.pm) in tibias. Data represent the mean AE SD (n = 10 to 14/group). Statistical analysis was performed by two-way ANOVA for the effect of genotype and HFD and the interaction. *p < 0.05, **p < 0.01, ***p < 0.001 for ND versus HFD. #p < 0.05, ##p < 0.01, ###p < 0.001 versus respective Cre-negative control. genotypes compared with Cre-negative controls, whereas a HFD increased its expression level to the same amount than in the Cre-negative HFD group.
Our study has potential limitations. First, a tamoxifeninducible Cre line was used to delete Dkk1 postnatally, even though tamoxifen has been shown to have direct effects on bone homeostasis. (74,75) However, as global deletion of Dkk1 is lethal, an inducible transgenic mouse model represents the currently best available option. Moreover, both Cre-negative and Cre-positive mice received tamoxifen, thus exposing both experimental groups to the same bias. Also, the Osx-Cre has known limitations including targeting other cell types than just osteoprogenitors and an intrinsic bone phenotype when not suppressed during embryogenesis. (76) However, we suppressed Cre activity during embryogenesis and during early life using doxycycline and found no difference in the bone phenotype using this protocol. (34) Also, according to our previous study, (34) we expected stronger effects of Dkk1 on trabecular bone, and thus sectioned the bones in such a way that it is not possible to obtain data on bone formation in cortical bone. Such data might have provided further insights into how DKK1 protects from obesity-induced cortical bone loss. Finally, a HFD induces bone loss in mice, which is not observed in humans. (20) Thus, this model does not fully recapitulate impaired bone strength in obese individuals. Nonetheless, the impact of high glucose and insulin resistance on bone can be studied well in this model, but requires further validation in humans.
In summary, we showed that Dkk1 from osteogenic cells has no effect on metabolic parameters, but may contribute to the expansion of bone marrow fat in obesity. More importantly, osteogenic Dkk1 appears to drive cortical, but not trabecular bone loss caused by obesity.

Disclosures
All authors state that they have no conflicts of interest.