Lrp4 expression by adipocytes and osteoblasts differentially impacts sclerostin's endocrine effects on body composition and glucose metabolism

Sclerostin exerts profound local control over bone acquisition and also mediates endocrine communication between fat and bone. In bone, sclerostin's anti-osteoanabolic activity is enhanced by low-density lipoprotein receptor–related protein 4 (Lrp4), which facilitates its interaction with the Lrp5 and Lrp6 Wnt co-receptors. To determine whether Lrp4 similarly affects sclerostin's endocrine function, we examined body composition as well as glucose and fatty acid metabolism in mice rendered deficient of Lrp4 in the adipocyte (AdΔLrp4) or the osteoblast (ObΔLrp4). AdΔLrp4 mice exhibit a reduction in adipocyte hypertrophy and improved glucose and lipid homeostasis, marked by increased glucose and insulin tolerance and reduced serum fatty acids, and mirror the effect of sclerostin deficiency on whole-body metabolism. Indeed, epistasis studies place adipocyte-expressed Lrp4 and sclerostin in the same genetic cascade that regulates adipocyte function. Intriguingly, ObΔLrp4 mice, which exhibit dramatic increases in serum sclerostin, accumulate body fat and develop impairments in glucose tolerance and insulin sensitivity despite development of a high bone mass phenotype. These data indicate that expression of Lrp4 by both the adipocyte and osteoblast is required for normal sclerostin endocrine function and that the impact of sclerostin deficiency on adipocyte physiology is distinct from the effect on osteoblast function.

The comorbidity of obesity and osteopenia/osteoporosis illustrates the coordination of adipose and bone metabolism via endocrine communication and by common regulatory mechanisms. Adipose-derived hormones, like leptin and adiponectin, affect bone mass accrual through both direct and indirect effects on the osteoblast (1,2), and recent advances have identified bone-specific factors that affect adipocyte hypertrophy and insulin sensitivity (3)(4)(5). Moreover, the fate specification of bipotential progenitors, present in the bone marrow and adipose stromal vascular fraction, to the osteoblastic and adipocytic lineages are reciprocally regulated by mitogen-activated protein kinase (6), bone morphogenetic (7), and Wnt/␤catenin signaling (8,9).
We recently demonstrated that, in addition to its local actions in bone, sclerostin fulfils an endocrine function that regulates body composition and adipocyte metabolism. Sost Ϫ/Ϫ mice accumulate less body fat and exhibit an increase in insulin sensitivity in association with increased markers of Wnt/␤-catenin signaling activation in adipose tissue depots and alterations in the ratio of anabolic to catabolic metabolism in adipocytes (3). Pharmacological inhibition of sclerostin activity with neutralizing antibodies produces an identical phenotype, whereas ectopic sclerostin overexpression stimulates adipocyte hypertrophy and impairs glucose homeostasis. These data accord with a growing number of studies that document a correlation between serum sclerostin levels and metabolic disease in humans. Serum sclerostin levels are increased in type 2 diabetics (17)(18)(19) and are positively associated with fat mass (20,21) and tissue insulin resistance (22,23).
Within the bone microenvironment, the anti-anabolic actions of sclerostin are facilitated by Lrp4, a member of the same protein family as Lrp5 and Lrp6 (24). Mice globally deficient of this receptor exhibit impairments in embryonic limb development and perinatal lethality, likely the result of Lrp4's function in neuromuscular junction formation (25,26), but point mutations in the extracellular domain of human LRP4 (R1170W and W1186S) are associated with a bone overgrowth phenotype (27) reminiscent of that in sclerosteosis and van Buchem disease patients, who lack normal sclerostin protein production (28 -30). Direct interaction of Lrp4 and sclerostin was confirmed in an MS screen by Leupin et al. (27), who also demonstrated that genetic knockdown of the receptor in vitro ablated sclerostin's inhibitory effect on osteoblast differentiation. Likewise, targeted ablation of Lrp4 expression in the osteoblast lineage (31,32) or pharmacological inhibition of the Lrp4:sclerostin interaction in vivo (31) dramatically increases bone formation and bone mass.
In this study, we explored the contribution of Lrp4 to sclerostin's endocrine function by ablating its expression in adipose tissue and bone. Similar to its role in facilitating sclerostin function in bone, inhibiting Lrp4 function in adipocytes abolished sclerostin's ability to enhance adipogenesis in vitro and resulted in a reduction in adipocyte hypertrophy and improved insulin sensitivity in vivo. Loss of Lrp4 function in osteoblasts leads to sclerostin overexpression and the opposite metabolic phenotype. Collectively, these data indicate that Lrp4 expressed by adipocytes and osteoblasts regulates normal sclerostin endocrine function by affecting its activity and expression, respectively.

Lrp4 is required for sclerostin to enhance in vitro adipocyte differentiation
The ability of osteocyte-produced sclerostin to inhibit osteoanabolic Wnt signaling is facilitated by its interaction with the Lrp4 receptor expressed by osteoblasts and their progenitors (27,32). Because Lrp4 is expressed in adipose tissue (Fig. 1, A and B) and by adipocytes induced to differentiation in vitro (Fig.  1C), we predicted that Lrp4 also contributes to the metabolic actions of sclerostin. As a first step in testing this hypothesis, stromal vascular cells were isolated from the inguinal fat pads of mice containing Lrp4 flox alleles and infected with an adenovirus encoding Cre recombinase to abolish Lrp4 expression (⌬Lrp4, Fig. 1D) or GFP as a control. Cells were then cultured under adipogenic conditions in the presence of vehicle or recombinant mouse sclerostin (rScl). 3 Consistent with our previous report (3), rScl treatment inhibited the expression of the Wnt target gene Axin2 (33) (Fig.  1D) and enhanced the adipogenic differentiation of control cells, as indexed by histochemical staining for accumulated lipids (Fig. 1F) and qPCR analysis of Cebpa, Fabp4, Lpl, and Pparg mRNA levels (Fig. 1G). ⌬Lrp4 adipocytes exhibited increases in Axin2 mRNA (Fig. 1D) and ␤-catenin protein levels (Fig. 1E) and impairment in differentiation and were strikingly resistant to rScl treatment, which was unable to enhance lipid accumulation or the expression of adipocyte markers. A similar effect of Lrp4 loss of function was observed when radiolabeled tracers were used to assess adipocyte metabolism in vitro. Although rScl increased de novo fatty acid synthesis (Fig. 1H), suppressed fatty acid oxidation (Fig. 1J), and induced coordinate changes in anabolic and catabolic gene expression (Fig. 1, I and K) in control adipocytes, ⌬Lrp4 adipocytes were unaffected by rScl treatment. Importantly, loss of Lrp4 function did not affect the expression Dkk1, Sostdc1, and Agrin or the ability of Dkk1 to enhance adipocyte differentiation (Fig. S1). Thus, Lrp4 is required for sclerostin to enhance adipocyte differentiation in vitro.

Ad⌬Lrp4 mice have small adipocytes and improved glucose metabolism
To determine the role of Lrp4 in adipocytes in vivo, we crossed Lrp4 flox mice with AdipoQ-Cre mice (34) to generate mice in which Lrp4 expression was ablated specifically in white and brown adipose tissue (Ad⌬Lrp4, Figs. 1B and 2A). Ad⌬Lrp4 mice were born at the expected Mendelian ratios (data not shown) and had normal body weight (Fig. 2B), serum sclerostin levels (Fig. 2C), and bone volume (Fig. S2). In agreement with Lrp4's role as a facilitator of Wnt signaling antagonism in bone, Axin2 expression (Fig. 2D) and ␤-catenin protein levels ( Fig.  2E) were increased in the white adipose tissue of the mutant mice relative to controls.
We expected that the increase in Wnt signaling in white adipocytes of Ad⌬Lrp4 mice would lead to the development of a lean phenotype similar to that observed in Sost Ϫ/Ϫ mice (3), but whole-body fat mass and the weights of individual fat pads in Ad⌬Lrp4 were comparable with control littermates (Fig. 2, F-H). However, histological analysis of adipocyte morphology in both the gonadal and inguinal fat pads revealed significant reductions in adipocyte size in Ad⌬Lrp4 mice relative to controls that led to an increase in adipocyte numbers per field (Fig.  2, I-L). The modest down-regulation of genes associated with adipocyte differentiation (Fig. 2M) together with alteration of genes involved in fatty acid synthesis and lipid catabolism (Fig.  2, N and O) suggest that the reduced adipocyte size is at least partially due to an inhibition of adipocyte hypertrophy and changes in the ratio of anabolic to catabolic metabolism. The maintenance of normal fat pad weights in Lrp4 mutant mice is likely due to an increase in proliferation and the accumulation of small adipocytes, as Ki67 staining in the stromal vascular fraction (Fig. 2P), fat pad DNA content (Fig. 2Q), and Ccnd1 expression were increased (Fig. 2R), whereas the mRNA levels of cyclin-dependent kinase inhibitors were reduced in adipose tissue relative to controls. The expression of Dkk1 and Sostdc1 was not affected by loss of Lrp4 function in white adipose tissue (Fig. S1), and the morphology of brown adipose tissue and liver was identical in control and mutant mice (Fig. S2).
Further analysis of Ad⌬Lrp4 mice revealed an improvement in glucose metabolism. Although the blood glucose levels of randomly fed mice were normal (Fig. 3A), mutant mice exhibited a significant reduction in serum insulin (Fig. 3B), which is usually indicative of an increase in insulin sensitivity. In line with this idea, the mutants performed better in glucose tolerance tests (Fig. 3, C and D) and insulin tolerance tests (Fig. 3, E and F). The increase in insulin sensitivity is likely to be driven, at least in part, by changes in adipose tissue sensitivity, as insulinstimulated Akt phosphorylation was increased in the inguinal fat pads of the mutants (Fig. 3, G and H), and serum free fatty acid levels were reduced (Fig. 3J) without a change in serum triglyceride levels (Fig. 3I). Serum levels of leptin, adiponectin, and osteocalcin were similar in control and mutant mice (Table  S1). The metabolic phenotype of female Ad⌬Lrp4 mice mirrored that of male mutants (Fig. S3). Therefore, adipocyte-spe-cific Lrp4 mutants partially phenocopy the metabolic effects of sclerostin deficiency (3) and raise the possibility that Lrp4 mediates at least some of sclerostin's endocrine function in white adipose tissue.

Lrp4 enables sclerostin endocrine actions Epistasis reveals a genetic interaction between adipocyteexpressed Lrp4 and sclerostin
As a genetic test of the interaction between sclerostin and Lrp4 in adipocytes, we performed an epistasis study by crossing Ad⌬Lrp4 mice with Sost ϩ/Ϫ mice to generate cohorts of compound heterozygous mice lacking one allele of the Sost gene globally and one allele of Lrp4 in adipocytes (Ad⌬Lrp4/ ϩ ; Sost ϩ/Ϫ ) as well as the appropriate heterozygous controls (Sost ϩ/Ϫ and Ad⌬Lrp4/ ϩ ; see "Experimental procedures" for the breeding strategy). In support of the notion that sclerostin and adipocyte-expressed Lrp4 work in concert to regulate adipose tissue metabolism, compound heterozygous mice exhibited an increase in Axin2 expression ( Fig. 4A) even though serum sclerostin levels were reduced by only 30% (Fig. 4B). More importantly, the compound mutants developed reductions in both whole-body fat mass and the weights of the gonadal and inguinal fat pads, whereas the body weight remained normal (Fig. 4, C-F). Within white adipose tissue depots, adipocyte size (Fig. 4, G and H) was reduced in compound heterozygotes relative to controls and was accompanied by changes in metabolic gene expression that were similar to the genetic ablation of both Lrp4 alleles in adipocytes (Fig. 4I). An increase in the abundance of multilocular adipocytes was also evident in compound heterozygotes, raising the possibility of an increase in the browning of this tissue. In all cases, Sost ϩ/Ϫ and Ad⌬Lrp4/ ϩ mice were similar to controls.
The compound heterozygous mice also mirrored the improvements in glucose metabolism evident here in Ad⌬Lrp4 mice (Fig. 3) and previously in Sost Ϫ/Ϫ mice (3). Blood glucose levels of randomly fed mice were comparable with controls, but serum insulin levels were significantly reduced in compound heterozygotes (Fig. 4, J and K). Similarly, glucose tolerance (Fig.  4, L and M) and insulin tolerance (Fig. 4, N and O) were increased in compound heterozygous mice, which also had significantly lower serum free fatty acid levels ( Fig. 4, P and Q). Sost ϩ/Ϫ and Ad⌬Lrp4/ ϩ mice were indistinguishable from controls in all of these measures. Taken together, these results   16-week-old male control and Ad⌬Lrp4 mice (n ϭ 10 mice/group). B, insulin levels in randomly fed mice (n ϭ 8 mice/group). C and D, glucose tolerance test (GTT) and area under the curve (AUC) analysis (n ϭ 8 -9 mice/group). E and F, insulin tolerance test (ITT) and area under the curve analysis (n ϭ 7-8 mice/group). G and H, Western blot analysis and quantification of Akt phosphorylation in inguinal adipose tissue before and after insulin stimulation (n ϭ 6 mice/group). I and J, serum triglyceride and free fatty acid levels in randomly fed control and Ad⌬Lrp4 mice (n ϭ 10 mice/group). All data are represented as mean Ϯ S.E. *, p Ͻ 0.05.

Lrp4 enables sclerostin endocrine actions
provide strong genetic evidence that Lrp4 facilitates sclerostin's function in fat in a manner analogous to that in bone.

Lrp4 deficiency in adipocytes alters Bmp signaling
Because alterations in sclerostin function in adipocytes produce coordinate changes in Bmp signaling, and administration of noggin inhibits the effect of rScl on adipocyte differentiation in vitro (3), we predicted that the loss of Lrp4 expression would decrease the activation of this adipogenic pathway (35). Consistent with our hypothesis, Bmp4 mRNA expression levels and the phosphorylation of Smad1/5/9 were reduced in the inguinal fat pads of Ad⌬Lrp4 mice (Fig. 5, A and B). Likewise, although rScl increased the expression of Bmp4 in in vitro cultures of control adipocytes, expression was reduced in ⌬Lrp4 cells and unresponsive to exogenous sclerostin (Fig. 5C).
To ensure that regulation of Bmp4 was indeed downstream of Lrp4 in the control of adipogenesis, we next treated differentiating control and ⌬Lrp4 adipocytes with recombinant Bmp4. Bmp4 treatment enhanced in vitro adipogenesis in control cultures, which was marked by an increase in Oil Red O staining and an increase in adipogenic gene expression (Fig. 5, D and E). Importantly, ⌬Lrp4 cultures were fully responsive to Bmp4 treatment, which was able to rescue the defect in adipogenesis associated with Lrp4 loss of function. Similarly, Bmp4 increased fatty acid synthesis (Fig. 5F) and affected the expression of metabolic genes in both control and ⌬Lrp4 cultures (Fig.  5, G and H) and, in both cases, reversed the changes induced by Lrp4 deficiency. Therefore, alterations in Bmp4 expression and activity similar to those evident in Sost Ϫ/Ϫ mice (3) appear to underlie the differentiation and metabolic defects associated with Lrp4 deficiency in adipocytes.

Ob⌬Lrp4 mice overproduce sclerostin and accumulate adipose tissue
Although the loss of Lrp4 function in the osteoblast lineage increases bone mass, it also leads to the overexpression of Sost via an unknown compensatory mechanism (31,32). We questioned whether this deregulation of sclerostin expression would, in turn, lead to alterations in body composition and metabolism by generating cohorts of male osteoblast/osteocyte-specific Lrp4 mutants (Lrp4 flox ; Ocn-Cre, Ob⌬Lrp4) and control littermates.
As reported previously (31,32), ablation of Lrp4 expression in osteoblasts and osteocytes (Fig. 6A) increased bone volume (Fig. 6, C and D) and induced a more than 4-fold increase in serum sclerostin levels (Fig. 6B). Analysis of Sost mRNA levels in the femur indicated that the increase in serum abundance is at least partially due to increased expression (Fig. 6A). Ob⌬Lrp4 mice maintained normal body weight (Fig. 6E), but, consistent with the idea that sclerostin favors adipose tissue accumulation via an endocrine mechanism, the mutants exhibited significant increases in whole-body fat mass (Fig. 6F) and the weight of white adipose tissue depots (Fig. 6H) and a reduction in lean tissue mass (Fig. 6G). Additionally, osteoblast-specific mutants exhibited a significant increase in adipocyte hypertrophy in the gonadal and inguinal fat pads (Fig. 6, I and J), coincident with a reduction in Wnt target gene expression and increased expression of genes involved in de novo fatty acid synthesis (Fig. 6K).
Glucose and lipid metabolism were also negatively affected by Lrp4 deficiency in osteoblasts and osteocytes and the resulting overproduction of sclerostin. Randomly fed Ob⌬Lrp4 mice were hyperglycemic (Fig. 6L) and hyperinsulinemic (Fig. 6M) compared with control littermates and performed poorly in standard glucose tolerance (Fig. 6, N and O) and insulin tolerance tests (Fig. 6, P and Q). Serum triglyceride levels (Fig. 6, R and S) were also elevated in osteoblast/osteocyte-specific mutants, suggesting development of dyslipidemia, but serum leptin, adiponectin, and undercarboxylated osteocalcin were similar in control and mutant mice (Table S1). The increase in total osteocalcin levels evident in osteoblast-specific mutants likely reflects the increase in bone formation. Therefore, Lrp4 expression by osteoblasts and osteocytes also appears to be essential for normal sclerostin endocrine action, likely as a result of the requirement for the receptor to maintain sclerostin expression levels within the normal range.

Discussion
Down-regulation of Wnt/␤-catenin signaling is a key step in the commitment of progenitor cells to the adipogenic lineage and attainment of a mature adipocyte phenotype (8,36,37). In our previous work (3), we demonstrated that bone-derived sclerostin, which had been viewed previously only as a local inhibitor of bone formation (28,38), acts as an endocrine factor that promotes adipogenesis and adipocyte hypertrophy in association with the inhibition of Wnt signaling in white adipose tissue. In this study, we used a series of in vitro and in vivo approaches to determine the contribution of Lrp4 to sclerostin's endocrine actions in fat. As noted above, this receptor directly interacts with sclerostin in the bone microenvironment and thereby facilitates sclerostin's ability to antagonize the recognition of Wnt ligands by Lrp5/6 in the osteoblast and inhibit bone formation (27,31,32).
Lrp4 is well-expressed by white adipocytes, with mRNA levels commensurate with those found in bone, and it appears to function in a manner analogous to that described in osteoblasts. Our in vitro experiments demonstrated that ablating Lrp4 expression produces a defect in the adipogenic differentiation of stromal vascular cells (which could be due to the presence of sclerostin in the serum used during in vitro culture) and renders  16-week-old control, Sost ϩ/Ϫ , Ad⌬Lrp4/ϩ, and Ad⌬Lrp4/ϩ; Sost ϩ/Ϫ mice (n ϭ 5-6 mice/group). B, serum sclerostin levels (n ϭ 7-10 mice/ group). C, body weights at 16 weeks of age (n ϭ 7-10 mice/group). D and E, qNMR analysis of fat mass and lean mass (n ϭ 7-10 mice/group). F, adipose depot weights of 16-week-old male mice (n ϭ 7-10 mice/group). G and H, representative histological sections (ϫ10 magnification) and frequency distribution of adipocyte size in inguinal fat pads (n ϭ 7-10 mice/group). I, qPCR analysis of genes involved in de novo fatty acid synthesis and fatty acid catabolism (n ϭ 5-6 mice/group). J, glucose levels in randomly fed mice (n ϭ 7-10 mice/group). K, insulin levels in randomly fed mice (n ϭ 7-10 mice/group). L and M, GTT and area under the curve (AUC) analysis (n ϭ 7-10 mice/group). N and O, ITT and area under the curve analysis (n ϭ 7-10 mice/group). P and Q, serum triglyceride and free fatty acid levels in randomly fed control and Ad⌬Lrp4 mice (n ϭ 7-10 mice/group). All data are represented as mean Ϯ S.E. *, p Ͻ 0.05. Scale bar ϭ 200 m.

Lrp4 enables sclerostin endocrine actions
adipocytes resistant to the effects of recombinant sclerostin on both adipogenesis and the ratio of catabolic to anabolic metabolism. Importantly, the adipogenic effects of Dkk1, which may also interact with Lrp4 (39), remained intact, and the mRNA levels of other interacting proteins remained largely normal. The metabolic phenotype of mice lacking Lrp4 in adipocytes partially mirrors that of Sost Ϫ/Ϫ mice (3). Both mutants exhibit an increase in markers of Wnt signaling in white adipose tissue, reduction in adipocyte size, and improvements in insulin sen-sitivity that were at least partially driven by an increase in the sensitivity of the adipocyte. The development of these phenotypes also appears to be secondary to modifications in the same pathway, as both Ad⌬Lrp4 and Sost Ϫ/Ϫ mice exhibited reductions in Bmp4 expression and Smad phosphorylation, which has been reported previously to induce adipocyte hypertrophy by stimulating fatty acid synthesis (35). Further, our genetic epistasis experiment places Lrp4 and sclerostin in the same genetic pathway that regulates adipocyte physiology.

Lrp4 enables sclerostin endocrine actions
We expected that Ad⌬Lrp4 mice would exhibit the decrease in adipose tissue mass evident in Sost Ϫ/Ϫ mice, but assessment of body composition by qNMR analysis and at necropsy revealed that the fat mass is equivalent in Lrp4 mutants and their control littermates. Our histological experiments suggest that the mutants are able to maintain normal fat mass by accu-mulating small adipocytes, likely through a compensatory increase in proliferation in the stromal vascular compartment and, potentially, subsequent commitment to the adipocyte lineage. Additional studies, including lineage tracing, will be required to more fully evaluate the impact of sclerostin and Lrp4 on the proliferation and differentiation of cells in this week old control and Ob⌬Lrp4 mice (n ϭ 8 mice/group). B, serum sclerostin levels (n ϭ 8 mice/group). C and D, representative Micro-computed tomography images and quantification of trabecular bone volume per tissue volume (BV/TV) in the distal femur (n ϭ 5 mice/group). Scale bar ϭ 2 mm. E, body weight was assessed weekly (n ϭ 10 -11 mice/group). F and G, qNMR analysis of fat mass and lean mass (n ϭ 9 -13 mice/group). H, adipose depot weights at 12weeks of age (n ϭ 11-12 mice/group). I, representative histological sections of the gonadal and inguinal fat pads; ϫ10 magnification. Scale bar ϭ 200 m. J, frequency distribution of adipocyte size in inguinal fat pads (n ϭ 5 mice). K, qPCR analysis of gene expression in inguinal fat pads (n ϭ 6 mice/group). L, glucose levels in randomly fed mice (n ϭ 12-13 mice/group). M, insulin levels in randomly fed mice (n ϭ 8 mice/group). N and O, GTT and area under the curve (AUC) analysis (n ϭ 12-13 mice/group). P and Q, ITT and area under the curve analysis (n ϭ 9 -10 mice/group). R and S, serum triglyceride and free fatty acid levels in randomly fed control and Ad⌬Lrp4 mice (n ϭ 8 mice/group). All data are represented as mean Ϯ S.E. *, p Ͻ 0.05.

Lrp4 enables sclerostin endocrine actions
compartment. Because Lrp4 expression increases during the differentiation process, these findings clearly indicate that Lrp4 facilitates the effects of sclerostin on adipocyte hypertrophy. However, we cannot rule out a possible role of sclerostin and Lrp4 in adipocyte specification because the AdipoQ-Cre transgene we utilized primarily drives recombination in mature adipocytes and does not impact the stromal vascular compartment that contains progenitor cells (40,41). Examining the interaction of sclerostin and Lrp4 at this earlier stage of differentiation will require additional studies using stage-specific Cre lines and lineage tracing.
A surprising finding in this study is the lack of an effect of genetic ablation of Lrp4 in adipocytes on serum sclerostin levels. Most endocrine interactions are associated with feedback loops wherein the level of signaling in the target tissue influences the production and release of the endocrine factor. According to this paradigm, loss of Lrp4 function in adipocytes would be expected to increase sclerostin production and, perhaps, indirectly impact skeletal homeostasis. Our data indicate that, in mice fed a chow diet, signals downstream of the sclerostin:Lrp4 interaction do not signal back to bone. It remains to be seen whether signaling from adipocytes regulates sclerostin production under obesogenic conditions, but the recent observation that PPAR␥ contributes to the regulation of sclerostin expression in osteocytes could represent a potential mechanism (42).
The increase in sclerostin serum levels following genetic ablation of Lrp4 in the osteoblast lineage is consistent with previous studies (31,32) suggesting a sclerostin chaperone action for Lrp4 that retains the protein within the bone microenvironment because an increase in expression was not observed. Our study suggests that the sclerostin:Lrp4 interaction may also induce feedback inhibition of sclerostin expression in the osteocyte, as Sost mRNA levels were increased in the femoral bone of osteoblast-specific mutants. The discrepancy between our findings and those of Chang et al. (31) could be due to the age or genetic background of the mice. Nonetheless, loss of Lrp4 function in the osteoblast is sufficient to alter body composition and phenocopy the effects of Adeno-associated virusmediated Sost overproduction (3), which include accumulation of white adipose tissue and development of impaired glucose homeostasis. Thus, expression of Lrp4 by two cell types, the osteoblast and the adipocyte, is required for normal sclerostin endocrine actions.
The phenotypes of these mutant mice are also informative regarding the interactions of sclerostin's effects on metabolism and bone mass. A number of recent studies, including some from our own laboratory, indicate that Wnt signaling regulates the intermediary metabolism of the osteoblast and thereby influences metabolic homeostasis (43)(44)(45)(46). The coexistence of a high bone mass phenotype because of increased Wnt signaling and increased fat mass in Ob⌬Lrp4 mice indicates that the effect of sclerostin on adipose tissue mass and metabolism is distinct from its effect on bone metabolism.
In summary, this study elaborates the mechanisms by which sclerostin influences whole-body metabolism and the mechanisms of cross-talk between adipose tissue and bone. We expect that it will have important implications for the use of sclerostin-neutralizing therapeutic agents intended to treat osteopenia/ osteoporosis but may also be useful in the treatment of obesity and metabolic disease.
All mice were maintained on a C57BL/6 background. Mice were housed on ventilated racks on a 14-h light/10-h dark cycle and fed ad libitum with a standard chow diet (Extruded Global Rodent Diet, Harlan Laboratories).

Culture of primary adipocytes
Primary adipocyte precursors were harvested from the inguinal fat pads by collagenase digestion as described previously (48) and cultured in DMEM (Gibco) containing 10% FBS (Sigma-Aldrich). Lrp4 gene deletion in vitro was induced by infection with an adenovirus directing the expression of Cre recombinase (multiplicity of infection 100, Vector Biolabs). An adenovirus directing the expression of GFP was used as a control. Upon reaching confluence, cells were induced to differentiate by treatment with 0.5 mM 3-isobutyl-1-methylxanthin, 1 M dexamethasone, and 167 nM insulin for 2 days, followed by continued culture in 167 nM insulin, which was changed every 2 days thereafter until analysis. Recombinant mouse sclerostin (R&D Systems) and Bmp4 (PeproTech) were replaced with each medium change. Oil Red O staining was carried out according to a standard technique.

Fatty acid metabolism
Fatty acid oxidation by cultured adipocytes was measured in flasks with stoppers equipped with center wells as described previously (49). Reactions were incubated at 37°C in medium containing 0.5 mM L-carnitine, 0.2% BSA, and [ 14 C]oleate (PerkinElmer Life Sciences). Expired 14 CO 2 was captured and counted by addition of 1 N perchloric acid to the reaction mixture and 1 M NaOH to the center well containing Whatman filter paper. To measure de novo lipogenesis, cellular lipids were

Metabolic phenotyping and bioassays
Whole-body fat and lean mass were assessed by qNMR (echo MRI). Plasma triglycerides and free fatty acids were measured colorimetrically (Sigma) in plasma collected 3 h after initiation of the light cycle. Serum sclerostin and insulin were measured in plasma by ELISA (Alpco). Glucose levels were measured using a Bayer Contour handheld glucose monitor. For glucose tolerance testing, glucose (2 g/kg of BW) was injected i.p. after a 6-h fast. For insulin tolerance testing, mice were fasted for 4 h and then injected i.p. with insulin (0.25 units/kg of BW). Insulin signaling in WAT was assessed by injection of insulin (1 units/kg of BW) into the portal vein before excising tissue and snap-freezing for immunoblot analysis. Tissue for histological analysis was collected at necropsy, weighed, and then fixed in 4% paraformaldehyde before embedding and sectioning. Adipocyte size was assessed using ImageJ. DNA was isolated from WAT using the Quick-DNA Universal Kit (Zymo Research) following proteinase K digestion. Ki67 (Abcam, ab15580) staining was performed on frozen adipose tissue sections according to standard immunohistochemical techniques.

Imaging
High-resolution images of the mouse femur were acquired using a desktop microtomographic imaging system (Skyscan 1275, Bruker) in accordance with the recommendations of the American Society for Bone and Mineral Research (50). Bones were scanned at 65 keV and 153 A using a 1.0-mm aluminum filter with an isotropic voxel size of 10 m. Trabecular bone parameters were assessed 500 m proximal to the growth plate and extending for 2 mm (200 CT slices).

Statistical analysis
All results are expressed as mean Ϯ S.E. Statistical analyses were performed using unpaired, two-tailed Student's t test or analysis of variance followed by post hoc tests. A p value of less than 0.05 was considered significant (*, p Յ 0.05 in all figures).