Temporally controlled transgene expression was used to trace cells of the osteoblast lineage using Dmp1-CreERt2 and mTmG mice. We used the mouse 10-kb Dmp1 promoter to drive the expression of the inducible CreERt2 in transgenic mice because it is expressed not only in osteocytes but also in mature osteoblast cell populations. The mutated ERt domain responds only to the synthetic estrogen receptor ligand tamoxifen. Administration of tamoxifen induces transient nuclear translocation and CreERt-mediated gene recombination. Dmp1-CreERt2 mice were crossed with mTmG mice, a double fluorescent reporter mouse strain. A reporter gene, such as the Rosa26R mTmG reporter transgene, can visualize the recombination since the expression of cre-recombinase induces the permanent excision of the upstream cassette encoding the membrane-targeted dTomato (mT) reporter protein to allow expression of a downstream cassette encoding a membrane targeted eGFP (mG) reporter protein (21, 22). In this manner, cells expressing cre-recombinase are indicative of mG expression. Then, mG is expressed continuously in the target cells and all possible progeny. Therefore, in this system, we can chase the fate of GFP-expressing cells such as from osteoblast lineage cells to BMAds (12). The 10-kb Dmp1-CreERt2 mice the Rip-Cre mice were kindly provided by Dr. Henry Kronenberg (Endocrine unit, Massachusetts General Hospital, USA) and Dr. Hye Seung Jung (Seoul National University Hospital). The 10-kb Dmp1-CreERt2 mice were generated by nuclear injections of a transgene encoding CreERt2 under the control of a 10-kb fragment of Dmp1 in B6C3F1 hybrid mice (Taconic, Hudson, NY, USA). Reporter mice (Gt(ROSA)26Sortm4(ACTB-tdTomato,-EGFP)Luo/J, stock no. 007576) were purchased from The Jackson Laboratory (Bar Harbor, Maine, USA).

These studies were approved by the Institutional Animal Care and Use Committee of Seoul National University.

For experiments with Dmp1-CreERt2:mTmG mice, 4-hydroxytamoxifen (4-OHTam, Takeda, Osaka, Japan) was used. For 4-OHTam injections, 2.5 mg of 4-OHTam were dissolved in 100 mL of dimethylformamide (Thermo Fisher Scientific, Waltham, MA, USA) and then diluted to 2.5 mg/mL in corn oil (Sigma-Aldrich, St. Louis, MO, USA). Dmp1-CreERt2 were injected with 0.5 mg of 4-OHTam once weekly in postnatal weeks 4–8.

In postnatal week 12, the mice were fed a 20 mg/kg/day rosiglitazone-supplemented diet (containing 140 mg rosiglitazone/kg diet; R0106, Tokyo Chemical Industry, Japan) or chow diet ad libitum for 8 weeks (randomly assorted into cages and group-housed), weighed, humanely euthanized, and analyzed. In addition, mice were injected subcutaneously with vehicle (phosphate-buffered saline [PBS]) or 50 mg/kg Scl-Ab (ratized antibody; Amgen Inc., Thousand Oaks, CA, USA; UCB, Brussels, Belgium) twice weekly for 4 weeks starting in postnatal week 16 (Figure 1). Mice were sacrificed in postnatal weeks 8 (+2 days), 12, 16, and 20.

Blood was collected via orbital sinus puncture before animals were euthanized. The serum levels of N-terminal propeptide of type I procollagen (P1NP) were measured using an ELISA kit (Immunodiagnostic Systems, Fountain Hills, AZ, USA).

Whole femurs were harvested from the mice after euthanasia, fixed in formaldehyde solution for 48 h, placed in 70% ethanol, and stored at 4°C until imaging. The distal femur from each mouse was scanned using high-resolution micro-CT (SkyScan 1173, Bruker microCT, Kontich, Belgium) at 90 kV and 88 μA with an isotropic voxel size of 7.1 μm using a 1.0-mm aluminum filter. For the metaphyseal tibia, a 1.5-mm section (starting 500 μm below the growth plate) was analyzed. Scanned images were reconstructed using NRecon v.1.6 software (Bruker microCT) by correcting for beam hardening and ring artifacts. Data were analyzed using CT Analyzer version1.6 (Bruker microCT). For trabecular bone regions, bone volume density (BV/TV), trabecular thickness (Tb.Th), trabecular separation (Tb.Sp), and trabecular number (Tb.N) were assessed.

Samples were dissected to remove soft tissues, fixed in 4% paraformaldehyde, incubated overnight at 4°C, and then decalcified in 15% EDTA for 7–14 days. Decalcified samples were cryoprotected in 30% sucrose/PBS followed by a 30% sucrose/PBS : OCT (1:1) solution, each performed overnight at 4°C. Samples were then embedded in OCT and cryosectioned at 8 μm (Leica, CM3050S). Images were captured using an epifluorescence microscope (Leica DMI 6000 [Inverted]) with prefigured triple-band filter settings for DAPI/FITC/TRITC and an automated fluorescent microscope with a whole-slide scanning platform. Confocal images were acquired using LEICA TCS SP8 STED CW and LAS X (Leica) software with lasers and corresponding band-pass filters for DAPI (Ex. 405 nm, BP 420–480), GFP (Ex. 488 nm, BP 505–530), and tomato (Ex. 543 nm, BP 565–595). Representative images of at least three independent biological samples are presented in the figures. Routine H&E staining was performed using previously published protocols.

Paraffin sections were postfixed in 4% paraformaldehyde for 10 min, blocked with 2% BSA for 30 min, and incubated with mouse anti-GFP monoclonal antibody (1:250, ab6673, Abcam), and anti-perilipin antibody (1:250, ab3526, Abcam) overnight at 4°C followed by Alexa Fluor 488-conjugated donkey anti-goat IgG (1:250, ab150129, Abcam) and Alexa Fluor 594-conjugated donkey anti-rabbit IgG H&L (1:250, ab150076, Abcam) for 1 h at room temperature. Sections were further incubated with DAPI to stain nuclei and observed under a microscope. Immunohistochemical staining of PPARγ (1:100, MBS8241690, Mybiosource) was performed as previously described (23). Adipocyte sizes were determined using Image J, as described previously (24).

The results were presented as the mean ± standard deviation. Statistical evaluation was performed via a nonparametric two-tailed Student’s t-test using GraphPad Prism version 6 (www.graphpad.com). Bonferroni’s correction for multiple testing was used for Figures 4 and 5. P < 0.05 was considered significant.

Figure 1 outlines the protocol used for this experiment. As found in previous studies, large numbers of GFP+ (direct fluorescence from mG) plump osteoblasts (yellow arrow) were identified on the endosteal surfaces of the femur 2 days after the last 4-OHTam injection in postnatal week 8 (Figure 2). Four weeks after the last 4-OHTam injection (12-week-old mice), the GFP+ cells, presumably BLCs (white arrow), became flat and less numerous on the endosteal surface of the femur. A few GFP+ osteocytes are shown in Figure 2 (asterisk). The flat GFP+ cells persisted on the endosteal surface in mice at week 16. Therefore, we could label mature osteoblasts and BLCs in the 10-kb Dmp1-CreERt2:mTmG mice. Additionally, to characterize the targeting specificity of 10-kb Dmp1-CreERt2:mTmG, we analyzed GFP expression in sections of the tibia at 2 weeks of age (Figure 3). The Dmp1-CreERt2:mTmG mice showed strong GFP signals in the tibia but not in the BM (Figure 3A). As expected, the tibia sections from CreERt2(-):mTmG mice did not exhibit any GFP activity (Figure 3B). Furthermore, GFP+ cells were not detected on the bone sections of Dmp1-CreERT2:mTmG mice without 4-OHTam(Figure 3C). Within the tibia, Dmp1-CreERt2 marked not only endocortical osteoblasts (Figure 3D) but also trabecular osteoblasts (Figure 3E) and the adjacent skeletal muscle (Figure 3F).

We first analyzed the skeletal effects of 8 weeks of exposure to rosiglitazone by measuring microCT parameters of the distal femur. Significant decreases in BV/TV (Figure 4A), bone surface density (BS/TV) (Figure 4B), and Tb.N (Figure 4C), and increases in Tb.Sp (Figure 4D) were noted after 8 weeks of rosiglitazone administration compared to the effects of the vehicle (P < 0.05). We then examined the effects of Scl-Ab in rosiglitazone-treated and control mice. Scl-Ab administration significantly improved trabecular parameters induced by rosiglitazone (BV/TV and Tb.Th: P < 0.01; BS/TV, Tb.N, Tb.Sp, and serum P1NP levels: P < 0.05) (Figure 4). The combined use of rosiglitazone and Scl-Ab did not lead to significant differences in these parameters compared to the effects of Scl-Ab alone. Serum P1NP levels significantly decreased from 8 to 12 weeks (P < 0.01). Scl-Ab significantly increased serum P1NP levels versus the vehicle (P < 0.05). In addition, 4 weeks of overlapping Scl-Ab treatment significantly increased serum P1NP levels compared to the effects of rosiglitazone alone (P < 0.05, Figure 4F).

To determine whether rosiglitazone could stimulate the differentiation of osteoblast lineage cells into BMAds, we chased GFP+ cells on the endosteal surface of the femur after 8 weeks of rosiglitazone administration. To better characterize and quantify the number of BMAds, we performed immunohistochemical staining using anti-perilipin antibody and anti-GFP antibody. Therefore, BMAds were visualized with anti-perilipin antibody, and GFP/perilipin double+ BMAds should be derived from osteoblast lineage cells. Representative histological images of BMAds in the femur in each group are shown in Figure 5A. After a 4-week chase period (16 weeks), some perilipin+ BMAds were detected in the BM but no GFP/perilipin double+ BMAds were detected. After an 8-week chase period, GFP/perilipin dual expression was detected from a few BMAds (yellow arrow) in the rosiglitazone group, and very few GFP/perilipin double+ BMAds were detected in the rosiglitazone/Scl-Ab group (N = 4–6/group). As shown in Figure 5B, rosiglitazone significantly increased the overall number of adipocytes (perilipin + BMAds) in the femur BM (rosiglitazone 543.5 ± 205.9/mm² vs. vehicle 3.1 ± 2.1/mm²; P < 0.001, Figure 5A). Moreover, some GFP/perilipin double+ BMAds were identified near the endosteal surface (rosiglitazone 20.5 ± 11.2/mm² vs. vehicle 0/mm²; P < 0.001, Figure 5B). Collectively, these data suggest that a small subpopulation of endosteal osteoblast lineage cells could differentiate into BMAds during rosiglitazone treatment compared to the vehicle findings (rosiglitazone 4.3 ± 2.8% vs. vehicle 0%; P < 0.001, Figure 5B).

We then tested whether Scl-Ab could modulate this effect of rosiglitazone on BMAds and thereby alter the total number of BMAds and the number of GFP/perilipin double+ BMAds in rosiglitazone-treated mice. When 8 weeks of rosiglitazone was overlapped by 4 weeks of Scl-Ab treatment, the total number of BMAds was not significantly decreased compared to the effects of 4 weeks of vehicle treatment (rosiglitazone 543.5 ± 205.9/mm² vs. Scl-Ab/rosiglitazone 487.3 ± 175.2/mm², Figure 5B). Additionally, Scl-Ab did not decrease the adipocyte size significantly (rosiglitazone 1612.2 ± 490.1 µm² vs. Scl-Ab/rosiglitazone 1498.5 ± 286.6 µm², Figure 5C). Interestingly, Scl-Ab significantly decreased the number of GFP/perilipin double+ BMAds in the femur BM relative to the vehicle (rosiglitazone 20.5 ± 11.2/mm² vs. Scl-Ab/rosiglitazone 2.8 ± 3.7/mm²; P < 0.001, Figure 5B). Collectively, when we quantified the contribution of the differentiation of osteoblast lineage cells into adipocytes to the entire BM adipogenesis process by counting the number of GFP/perilipin double+ BMAds and total perilipin+ BMAds, Scl-Ab preferentially suppressed the differentiation of osteoblast lineage cells into adipocytes (rosiglitazone 4.3 ± 2.8% vs. Scl-Ab/rosiglitazone 0.6 ± 0.9%; P < 0.001, Figure 5B). In addition, as found in previous studies, more GFP+ endosteal cells appeared plumper in the Scl-Ab or Rosi+Scl-Ab treatment groups than in the vehicle or Rosi alone groups (Figure 5D).

To exclude the possibility that rosiglitazone affected endogenous GFP expression, Rip-Cre:mTmG mice were also treated with rosiglitazone. In Rip-Cre transgenic mice, the rat insulin II promoter drives cre expression in both β cells in the pancreas and some neurons in the brain (25). Co-immunostaining experiments showed that rosiglitazone also increased the number of perilipin+BMAds in Rip-Cre:mTmG mice, but GFP/perilipin double+ cells were not detected on the endosteal surface or in the BM of the femur (Figure 6A). As expected, Rip-Cre mice targeted islet cells in the pancreas (Figure 6B), and no GFP signals were detected in the pancreas sections obtained from Rip-Cre(-):mTmG mice (Figure 6C). Thus, rosiglitazone did not increase endogenous GFP expression.

To explore the adipogenic characteristics of osteoblast lineage cells in response to rosiglitazone treatment, we examined PPARγ expression. Immunohistochemical staining revealed numerous PPARγ−expressing cells in the trabecular and cortical bones, which were likely to be osteoblasts and BLCs in the femur at 8 and 12 weeks, respectively (Figure 7).