This study was approved by the Ethics Committee of Peking Union Medical College Hospital. C57BL/6J mice were obtained from the Vital River Laboratory Animal Technology (Beijing, China). Lrp5^fl/fl conditional-ready mice and Prrx1-cre transgenic mice were purchased from the Shanghai Model Organisms Center (Shanghai, China). To generate Lrp5 CKO mice, Lrp5^fl/fl mice were crossed with Prrx1-cre mice to get Lrp5^fl/fl; Prrx1-cre genotype. Mouse genome DNA was extracted from tail of mice at postnatal day 10 (P10), and the amplification of target genes (Lrp5 and Prrx1-cre) was done following a described protocol (Truett et al., 2000; Cui et al., 2011). A 2.5% Agarose-TAE gel containing Gel-Green (D0143, Beyotime, Beijing, China) was used for electrophoresis of PCR products.

Murine BMSCs were obtained using a previous described protocol with minor modifications (Grassel et al., 2012). In brief, murine femur and humerus were dissected from Lrp5^fl/fl or wild-type C57BL/6J mice at P14 after euthanasia and washed three times in cold phosphate buffered saline (PBS; Cytiva) with 1% penicillin-streptomycin (Sigma-Aldrich). The isolated long bones were then cut with sterile scissors, and the bone marrow was extracted by centrifuging at 500 g for 30 s. Pellet was resuspended in PBS and filtered with 70-μm cell strainer (BD Falcon). Cells were then resuspended and seeded in 10-cm dishes (Corning) in Dulbecco’s modified Eagle’s medium (DMEM) high glucose (Hyclone), 1% penicillin-streptomycin (Sigma-Aldrich), and 10% fetal bovine serum (FBS; Gibco). BMSCs were cultured at 37°C and 5% CO₂, and the cell medium was changed every 48 h. After 80–90% cell confluence was reached, the adherent cells were washed twice in PBS, trypsinized using 0.25% Trypsin-EDTA (Thermo Fisher Scientific), centrifuged at 1,000 g for 3 min, and then seeded in 10-cm dish (Corning) at the next passage.

Cell viability was determined by the methylthiazolyldiphenyl-tetrazolium bromide (MTT) assay as described previously (Riss et al., 2004). In brief, MTT solution was added to each well 7 days after osteogenic induction with mediums of different lipid treatment. Three hours later, dimethyl sulfoxide (DMSO) was added and the absorbance of the solution was measured at 570 nm with an automatic microplate reader (CFX Connect, Bio-Rad). The survival rate of cells in percentage was calculated using DMSO-treated cells as a standard. One-way ANOVA was conducted with GraphPad Prism software.

The osteogenic differentiation of BMSCs was induced as described previously (Grassel et al., 2012). The third-passage BMSCs were seeded in 0.1% gelatin (Sigma-Aldrich) pretreated 6- or 12-well plates (Corning) at 5,000 cells/100 μl. The osteogenic induction was initiated at 70–80% cell confluence by using an osteogenic medium composed of DMEM high glucose (Hyclone), 100 nM dexamethasone (Sigma-Aldrich), 0.05 mM ascorbate 2-phosphate (Sigma-Aldrich), 10 mM sodium beta-glycerophosphate (Sigma-Aldrich), and 1% penicillin-streptomycin (Sigma-Aldrich). FBS (Gibco), lipid-reduced serum (Biowest), oleate (Sigma-Aldrich), and recombinant Wnt3a protein (5036-WN, R&D Systems) were added into the medium according to the experiment design.

After 5/7/11 days of osteogenic induction, the osteogenic differentiation was assessed by alkaline phosphatase (ALP) staining kit (VectorLab) in the 12-well plate. After 30 min staining, the BMSCs were fixed in 4% paraformaldehyde (P1110, Solarbio, Beijing, China), washed in PBS (Cytiva). Then, the 12-well plate was scanned and photographed under × 4 magnification using EVOS Cell Imaging System (EVOS M7000, Thermo Fisher Scientific). The ALP activity of each well was quantified with ImageJ Software (¹ National Institutes of Health) as previously described (Dupont et al., 2011). In essence, the ALP + area was determined with ImageJ as the number of blue pixels across the picture, and then this value was normalized to the total number of pixels for each picture. For each well, views of four different zones were adopted for calculations. Multiple t tests were conducted with GraphPad Prism software using the Holm–Sidak test.

Bone marrow mesenchymal stromal cells were seeded in six-well plate under different conditions. BODIPY^TM staining kit (Thermo Fisher Scientific) was applied to trace the endocytosis of lipid substance in BMSCs (DiDonato and Brasaemle, 2003), and 4,6-diamidino-2-phenylindole staining solution (C00065, Solarbio, Beijing, China) was used to locate the nucleus. After BODIPY^TM application, the plate was scanned and photographed under × 40 magnification using EVOS Cell Imaging System (EVOS M7000, Thermo Fisher Scientific). Cells with positive signals in cytoplasm were counted, and the fluorescence intensity was calculated by BODIPY^TM-positive cell counts over total cell in the region. Statistical comparisons were conducted with GraphPad Prism software using the Holm–Sidak test. For each analysis, four different views were adopted to carry out the quantification.

Adenoviruses expressing CMV-Cre-recombinase and CMV-Null vectors were purchased from Beyotime (Beijing, China). BMSCs were infected with viruses at 100 multiplicity of infection (MOI) in the presence of 0.1% poly-L-lysine (Sigma-Aldrich) for 1 h as previously described (Buo et al., 2016).

Total protein was extracted from BMSC cell lysates, murine bone tissue, and other organs using TRIzol (Thermo Fisher Scientific) following the standard protocols. cDNA was obtained using PrimerScript RT (TaKaRa, Japan). Real-time PCR was performed using TB Green Master Mix kit (TaKaRa, Japan) on Applied Biosystems StepOnePlus Real-Time PCR System (Thermo Fisher Scientific). The primers used for qPCR analysis were listed in the Supplementary Table. mRNA expression levels of Alpl, Sp7, Col1a1, Tcf1, Lef1, and Axin2 were normalized to glyceraldehyde 3-phosphate dehydrogenase (Gapdh). Statistical comparisons were conducted with GraphPad Prism software using the Holm–Sidak test. Each experiment was repeated three times.

Total protein was extracted from BMSC cell lysates, murine bone tissue, and other organs using radioimmunoprecipitation assay (RIPA) buffer (Thermo Fisher Scientific). The protein concentration was determined by bicinchoninic acid (BCA) protein assay kit (Thermo Fisher Scientific). Western blot was performed using a standard protocol. Anti-LRP5 (rabbit monoclonal, ab223203, Abcam, 1:1,000), anti-Type I Collagen (COL1; rabbit monoclonal, ab270993, Abcam, 1:1,000), anti-RUNX Family Transcription Factor 2 (RUNX2; rabbit monoclonal, ab236639, Abcam, 1:1,000), anti-beta catenin (rabbit monoclonal, ab32572, Abcam, 1:5,000), and anti-GAPDH (rabbit monoclonal, #2118, Cell Signaling Technology, 1:1,000) were applied as primary antibodies, and horseradish peroxidase (HRP) goat anti-rabbit immunoglobulin G (IgG) (H + L) (ZB-2301, ZSGB Biotech, Beijing, China, 1:2,000) was applied as secondary antibody to carry out immunoblotting. Protein expression levels were visualized by automatic chemiluminescence imaging system (C300, Azure Biosystems). The quantification of band intensity was performed by ImageJ Software. Overall expression levels of LRP5, COL1, RUNX2, and beta-catenin were normalized to GAPDH levels. Statistical comparisons were conducted with GraphPad Prism software using the Holm–Sidak test. Three repetitions were carried out for each of the above experiments.

Control and CKO mice at P21 were euthanized. Femora and humeri were obtained and fixed in 4% paraformaldehyde (P1110, Solarbio, Beijing, China) for 48 h, washed in cold PBS three times, and then preserved in 75% ethanol at 4°C. Samples were scanned and analyzed using SCANCO micro-computed tomography (micro-CT) μ100 system (SCANCO Medical). The three-dimensional structural parameters including humerus length, mean bone density, trabecular bone volume fraction (Tb.BV/TV), average cortical thickness (Ct.Th), trabecular number (Tb.N), and trabecular separation (Tb.Sp) were analyzed to determine the bone quality as previously described (Bouxsein et al., 2010). Each analysis was performed in samples collected from 12 individual animals (six control mice and six CKO mice). Statistical comparisons were conducted with GraphPad Prism software using unpaired t-test.

Here, 2-month-old control or CKO mice were euthanized. Intact adipose tissues in gonadal, mesenteric, and subcutaneous regions were carefully dissected using forceps and scissors. The percentage of fat pad mass (g) over body weight (g) was calculated. Each analysis was repeated with 12 individual animals (six control mice and six CKO mice). Statistical comparisons were conducted with GraphPad Prism software using the Holm–Sidak test.

All data analysis in this study was carried out on GraphPad Prism 8 (GraphPad Software). At least three independent experimental groups were performed to generate a qualified data set. Statistical strategies adopted in each experiment were described above. P-values < 0.05 (^∗), 0.01 (^∗∗), and 0.001 (^∗∗∗) were considered significant. Error bars on all graphs are presented as the standard deviation of the mean unless otherwise indicated.

To investigate the effects of lipid availability on BMSC osteogenesis, we began by performing osteogenic assays using murine BMSCs isolated from P14 wild-type C57BL/6J mice. Osteogenic medium was prepared using normal FBS (10%), lipid-reduced serum (LRS; 10%), serum deprivation (SD), and LRS with additional oleate (OL, 75 μM), an unsaturated fatty acid, and treated with isolated murine BMSCs. ALP staining was performed on days 5, 7, and 11 after osteogenic induction, and ALP activity was measured to determine the osteogenesis in different mediums (Figures 1A,B). Moreover, an MTT assay was carried out to evaluate the cell viability 7 days after osteogenic induction under different lipid supplements (Figure 1C). Here, we observed significantly decreased ALP activities of BMSCs in LRS or SD medium, and the suppression could be partially rescued by additional supplementation with oleate (Figure 1B). Notably, the cell survival was significantly reduced in SD medium (Figure 1C; p < 0.001), indicating that the viability of BMSCs was impaired due to complete deprivation of serum supplement. Further in vitro analysis revealed a significant decrease in protein expression of osteogenic markers such as COL1, RUNX2, and beta-catenin, serving as an intracellular Wnt signaling transducer (Figures 1D,E) 7 days post-induction. We then asked whether the suppressed osteogenesis was also related to the downregulation of Wnt/beta-catenin signaling. Relative mRNA expression of several Wnt target genes (Tcf1, Lef1, Axin2) and osteogenic genes (Alpl, Sp7, Col1a1) was examined and turned out to be reduced significantly in qRT-PCR assays 7 days after induction (Figure 1F). Notably, the suppressed expressions of protein and mRNA of both osteogenesis and Wnt/beta-catenin signaling were rescued by oleate supplementation (Figures 1A–F).

To further validate the rescue effect of oleate supplementation in lipid-deprived BMSCs, we performed an osteogenic assay of BMSCs using increasing dosage of oleate (0, 25, 50, 75, and 100 μM, respectively) in LRS medium for 5, 7, and 11 days (Figure 2A). Quantification of ALP activity revealed that the enhancement of osteogenesis was dosage-dependent on oleate supplementation (Figure 2B). Moreover, the protein and mRNA expressions of osteogenic markers and Wnt target genes were also increased dependent on oleate treatment 7 days after induction (Figures 2C–E). These results suggested that lipid availability could modulate osteogenic differentiation via Wnt/beta-catenin signaling in BMSCs.

As an important co-receptor for Wnt/beta-catenin signaling and potential endocytic mediator of multiple substances (Li et al., 2001; Schneider and Nimpf, 2003; Chung and Wasan, 2004), LRP5 was previously reported essential in bone development (Clement-Lacroix et al., 2005; Yadav and Ducy, 2010; Chin et al., 2015; Frey et al., 2015; He et al., 2020) and lipid metabolism (Magoori et al., 2003; Loh et al., 2015). We generated CKO mice by crossing Lrp5-floxed ready mice with Prrx1-cre mouse line to inactivate Lrp5 gene in limb mesenchymal cell lineage, including BMSCs. The loxP sequence was inserted into the murine genome to conditionally remove the exon 2 of Lrp5 gene in Cre-expressing lineages. We collected the genome DNA from P10 mouse tail, and a PCR program and the following electrophoresis were performed to determine the floxed Lrp5 allele (Figure 3A). To further validate the efficiency of Cre-mediated knockout in vivo, we harvested multiple organs from control (Lrp5^fl/fl) and CKO (Lrp5^fl/fl; Prrx1-cre) mice at P0, including bone and skeletal muscles in forelimbs, skin, brain, heart, kidney, and subcutaneous adipose tissue. The following Western blot and qRT-PCR analysis confirmed that the LRP5 protein expression and Lrp5 mRNA expression were lost in CKO mice in bone and skeletal muscles of forelimbs compared to control mice (Figures 3B,G; all p < 0.001). It turned out that the mRNA expression of Lrp5 was significantly increased in liver tissue (p < 0.001), indicating a compensatory enhancement of gene expression in liver. In addition, in P21, the CKO mice presented a reduced body size and shortened limbs in gross appearance (Figure 3C). Besides, the body weight was decreased in CKO mice significantly (Figure 3F; p < 0.001).

To determine the effect of Lrp5 knockout in vivo, we performed micro-CT scan on the humerus and femur samples harvested from P21 mice. The humerus length was remarkably decreased in CKO mice (Figures 3E,F; p < 0.001). The reconstruction imaging of distal femur recorded decreased density in trabecular architecture and thinner cortical bone in CKO compared to control mice (Figure 3D). Further analysis revealed a significant decrease in mean bone density (p < 0.001), average Ct.Th. (p < 0.001), Tb.BV/TV (p < 0.001), and Tb.N (p < 0.01) and increased Tb.Sp (p < 0.001) in the distal femur of CKO mice. These data indicated that the bone formation in limb was impaired as a result of suppressed osteogenesis differentiation of skeletal progenitors, including BMSCs by Lrp5 knockout.

Previous report showed the lipid metabolism and adipose tissue were changed after Lrp5 gene was knockout in osteoblast lineage (Frey et al., 2015), indicating the role of LRP5 in modulating lipid availability in skeletal progenitors. In order to test the similar regulatory effect in BMSC-CKO mice, we dissected the adipose tissue from control and CKO mice at 2 months old. Interestingly, there was increased fat pad mass in gonadal (p < 0.001) and mesenteric areas (p < 0.05; Figures 3I,J), but the subcutaneous fat in limb was not changed (data not shown). In addition, the expression of lipid metabolic genes (Acadl, Acat2, Acsl1, Cpt1b, and Hadha) in limb BMSCs was significantly decreased (Figure 3K; all p < 0.001). There data suggested that the ability of using lipid substances was impaired in BMSCs after Lrp5 knockout and the accumulation of adipose tissue was increased in mesenteric and gonadal regions, strongly indicating the role of LRP5 in lipid uptake and metabolism in BMSCs.

To further understand the role of LRP5 in lipid uptake of BMSCs, we used adenovirus carrying CMV-Cre-recombinase (Ad-Cre) or CMV-null (Ad-CMV) expressing vector to infect BMSCs isolated from P14 Lrp5^fl/fl mice. The expression of LRP5 at protein and mRNA levels was examined to validate the efficiency of Cre-recombination (Figures 4C,D; p < 0.001). To trace the endocytosis of lipids in vitro, we applied BODIPY^TM staining (DiDonato and Brasaemle, 2003) in FBS and LRS + OL culture condition 48 h after viral infection (Figure 4A). The fluorescence intensity of lipid droplet signals in cytoplasm was found significantly decreased after Lrp5 ablation both in FBS and LRS + OL conditions (Figure 4B; both p < 0.001), suggesting impaired endocytosis of lipid due to Lrp5 ablation. We next performed osteogenic assays on adenoviral-treated BMSCs in FBS and LRS mediums (Figure 4E). Consistent with in vivo results, the osteogenesis of BMSCs was significantly inhibited by Lrp5 ablation (Figures 4E,F; all p < 0.001) after 7-day induction in both assays, indicating the modulatory role of LRP5 in BMSC osteogenesis.

In light of the fact that Wnt-LRP5 signaling was found to participate in the lipid metabolism and bone formation in skeletal progenitors, we then examined whether this modulation is regulated by lipid availability or not. After 7-day osteogenic induction, Lrp5 knockout significantly reduced the expression of Wnt target genes (Tcf1, Lef1, Axin2; Figure 4G; all p < 0.001), as well as osteogenic genes (Alpl, Sp7, Col1a1; Figure 4G; all p < 0.001) in FBS-treated BMSCs, which was consistent with previous reports (Ayturk et al., 2013; Sebastian et al., 2017). However, the changes in Wnt target gene expression were not significant after Lrp5 ablation in lipid deprivation medium (Figure 4G). These data suggested that the modulation of LRP5 in BMSC osteogenesis was mediated by Wnt/beta-catenin signaling and was dependent on lipid availability. We further used recombinant Wnt3a protein, a classic Wnt ligand, in the lipid-deprived BMSCs with/without Lrp5 ablation (Figure 4E). The osteogenic activity was significantly enhanced even compared with Lrp5-intact BMSCs (Lrp5^fl/fl treated with Ad-CMV) in FBS medium (Figures 4E,F; p < 0.001). These results indicated that the Wnt-activating osteogenesis was independent of lipid scarcity.