The mice were purchased from Dashuo Company (Chengdu, China). Mice were housed in box cages, maintained on a 12 h light/12 h dark cycle, and fed a chow diet ad libitum (Sakai et al., 2019). The diabetes mellitus model of mice was constructed as described previously (Zhou X. et al., 2019). In brief, male C57BL/6J mice aged 8 weeks were fasted for 12 h and then injected intraperitoneally of 1% streptozotocin dissolved in saline solution with a dose of 50 mg/kg for 4 consecutive days. Tail vein blood was taken to test the fasting blood glucose level on the 1^st and 2^nd week post-injection. The mice with a fasting blood glucose higher than 16.7 mmol/L were regarded as successful diabetes model (Jiang et al., 2021). The mice were randomly divided into following three groups of eight mice each: 1) normal mice (Control); 2) normal mice treated with PPARβ/δ agonist GW501516 (Control + GW); 3) diabetes mellitus group (DM) and 4) diabetes mellitus group treated with PPARβ/δ agonist GW501516 (DM + GW).

Control + GW and DM + GW groups were injected with GW501516 (5 mg/kg/d, Sigma-Aldrich) dissolved in 0.1 ml dimethyl sulfoxide (DMSO) every other day (Mosti et al., 2014). Control group and DM group were injected with 0.1 ml DMSO every other day. The mice were sacrificed after 4 and 8 weeks of injection. The femurs were collected for subsequent histomorphological analysis.

The femurs were dissected and fixed in 4% paraformaldehyde for 2 days and then stored in 70% ethanol at 4°C. Micro-computed tomography (μCT) analysis was performed (μCT50, SCANCO Medical) with a spatial resolution of 10 μm (55 kV, 114 mA, 500 ms integration time). The regions of interest (ROI) were defined as the trabecular bone and cortical bone at the distal femur (Bouxsein et al., 2010). Bone mineral density (BMD), bone volume/total volume (BV/TV), trabecular number (Tb.N), trabecular separation (Tb.Sp), trabecular thickness (Tb.Th), cortical bone thickness (Ct.Th), and cortical bone porosity (Ct.Po) were evaluated within the delimited ROI (Wang et al., 2020).

The mouse femurs were fixed and decalcified, and then embedded in paraffin. Sections of 4.5 μm were prepared. H&E staining (Solarbio, Beijing, China) was performed as the manufacturer’s instruction. The stained sections were observed using an inverted microscope (IX81, Olympus).

TRAP staining was performed according to the procedure described previously (Liu et al., 2016). TRAP staining solution (Sigma) was added to cover the tissue sections. After incubating at 37°C for 40 min, the sections were stained with 0.1% methyl green solution (Sigma) for 10 s. The stained sections were observed using an inverted microscope (IX81, Olympus).

Eight-week-old C57BL/6J mice were sacrificed to separate the femur and tibia. The bone marrow was flashed into a petri dish and centrifuged. The supernatant was discarded, and the red blood cell lysate was added. The cells were seeded in a 10 cm diameter petri dish, cultured in DMEM medium with 10% fetal bovine serum (FBS), and supplied with 50 ng/ml m-CSF (R&D). After 4 days, the suspension cells were discarded and bone marrow-derived macrophages (BMDMs) were obtained for subsequent experiment.

DEME medium with 5.6 or 30 mM glucose were used for normal glucose (NG) or high glucose (HG) culture, and 5 μM GW501516 was added to HG + GW group. After 3 days of culture, different macrophage culture supernatants were obtained. To avoid the inhibition of osteoclast differentiation by high glucose concentration, DMEM medium without glucose was used to mix with the culture supernatant. Osteoclast-induced BMDMs are divided into three groups: NG Supernatant group, HG Supernatant group, and HG + GW Supernatant group.

For osteoclast induction, BMDMs (4×10⁴ cells/well) were seeded in 24-well plates. 50 ng/ml RANKL (R&D) was added to the BMDM culture medium, and the medium was changed every 2 days. After 3 days of culture, mature osteoclasts could be observed and subsequent experiments were performed.

TRAP staining (Sigma-Aldrich) was performed according to the previously described procedure (Liu et al., 2016). The cells were fixed with 4% paraformaldehyde for 10 min, added with TRAP staining solution, and incubated at 37°C for 40 min. TRAP-positive cells containing three or more nuclei were considered osteoclasts.

The total RNA of BMDMs and osteoclasts was extracted using Trizol reagent (Invitrogen, Carlsbad, CA, United States), and then reversely transcribed to obtain stable cDNA using PrimeScript™ RT reagent Kit with gDNA Eraser (TaKaRa Bio, Otsu, Japan). The RT-PCR was performed using SYBR Premix Ex Taq II (TaKaRa Bio) in Quant Studio™ three real-time fluorescent PCR instrument (ThermoFisher Scientific, China). Glyceraldehyde 3-phosphate dehydrogenase (Gapdh) was used as an internal reference to normalize the gene expression (Wang et al., 2020). The result was calculated using the 2^−ΔΔCt method and expressed as a multiple change relative to Gapdh. The primer sequences are summarized in Supplementary Table S1.

The detection of ROS levels in BMDMs was carried out in accordance with the recommended protocol (Pang et al., 2019). In brief, BMDMs (2×10⁵ cells/well) were seeded in 6-well plates under NG, HG, or HG + GW culture. Then, the ROS fluorescent probe DCFH-DA (Beyotime Biotechnology) was diluted with serum-free DMEM medium at a ratio of 1:1,000. After culturing for 72 h, the culture medium was removed and then DMEM medium containing DCFH-DA was added. After incubation at 37°C for 20 min, the cells were washed three times with serum-free DMEM medium without phenol red and digested by trypsin. The fluorescence intensity was analyzed on a flow cytometer (ThermoFisher).

For flow cytometry analysis, the mouse femur was cut into pieces, digested in 1 mg/ml collagenase I, 1 mg/ml dispase II at 37°C for 30 min, centrifuged. After lysis of red blood cells on ice, the samples were passed through a 70-μm filter, centrifuged, and ready for staining. Anti-mouse CD45, anti-mouse CD11b, anti-mouse CD86, and anti-mouse CD206 were purchased from BD Biosciences, and the permeabilization/fixation kit was purchased from eBioscience. All staining processes were performed in 100 μl PBS. The cells were stained on ice for 30 min and then analyzed by flow cytometry (BD Biosciences).

The expression levels of TNF-α and IL-1β were measured by mouse ELISA kits (Cusabio) according to the recommended protocol. Briefly, 100 μl of culture supernatant was added to a 96-well plate with high binding capacity and incubated for 2 h, and the absorbance was measured with a microplate reader. Serum concentration of PINP, CTX and ANGPTL4 were measured using ELISA kits (CUSABIO). Mice were fasted for 4 h and then we collected the blood samples by puncturing the cheek pouch and allowed the blood to coagulate on ice for 1 h before centrifugation to obtain the serum.

siRNA sequence for Angptl4 and Ppard were designed and synthesized by Sangon Biotech (Shanghai, China). BMDMs were transfected with Lipofectamine® RNAimax (Invitrogen) in serum-free DMEM medium followed by the manufacturer’s instructions. In all experiments using siRNA, control siRNA and Lipofectamine® RNAimax were added to other group to eliminate other potential influence.

For immunofluorescence staining, sodium citrate solution was used for antigen retrieval. After blocking by 5% BSA at 37°C for 1 h, primary antibodies (anti-CD86, anti-CD206, Abcam) were incubated overnight at 4°C. Then the fluorescent secondary antibodies (Abcam) were incubated for 1 h at room temperature. DAPI was used to mark the nucleus with 15 min of staining. The stained sections were observed under a fluorescent microscope (Olympus BX53).

Chromatin immunoprecipitation (ChIP) assay was performed according to the manufacturer’s instructions with EZ-Zyme Chromatin Preparation Kit (Millipore) and Magna Chip HiSens (Millipore). Rabbit IgG (Sigma) was used as the control, and antibody against PPARβ/δ (Santa Cruz) was set as the experimental group. The DNA-protein complex was dissociated, PCR primers were designed to detect the target area, and then the pull-down DNA and input DNA were tested by PCR analysis with the primers flanking the PPRE region of Angptl4.

For ChIP-seq analysis, we downloaded the ChIP-seq data of PPARβ/δ in GEO database, with the accession number GSE50144 (Inoue et al., 2014). The binding peak of PPARβ/δ in ANGPTL4 gene segment was visualized via integrative genomics viewer (IGV) software.

All quantified data were expressed as mean ± standard deviation (SD). Statistical differences were performed via unpaired two-tailed Student’s t test for comparison between two groups and by one-way or two-way analysis of variance (ANOVA) followed by the Tukey’s post hoc test for multiple comparisons. p values < 0.05 were considered to be statistically significant.

First, we explored the effect of PPARβ/δ agonist on the osteoporotic phenotypes of diabetic mice (Figure 1A). Through μCT analysis of the distal femur, we found that both trabecular bone and cortical bone were affected by the diabetic condition and manifested as bone loss. PPARβ/δ agonist could partially relieve the diabetic osteoporosis. In addition, PPARβ/δ agonist treatment had no significant influence on bone phenotype in healthy control mice (Figures 1B,C). H&E staining results showed the number and thickness of trabecular bone and the thickness of cortical bone decreased in DM group, while in the DM + GW group, the osteoporotic phenotype was restored (Figure 2A). Through TRAP staining, we found that osteoclast activity in DM group was significantly increased, and PPARβ/δ agonist treatment could reduce the activation of osteoclasts under diabetic condition (Figures 2B,C). Also, bone resorption marker CTX decreased after PPARβ/δ agonist treatment (Figure 1D). Overall, we found that at the 4^th week and 8^th week, the bone mass of the DM group was significantly lower than that of the control group, and PPARβ/δ agonist alleviated the diabetic bone loss.

Seizing the evidence that PPARβ/δ agonist alleviated the diabetic osteoporosis phenotypes and reduced osteoclast activity in vivo, we speculated that PPARβ/δ might directly inhibit osteoclast differentiation. Considering that high glucose environment inhibited osteoclast differentiation in vitro (Wittrant et al., 2008), we explored the regulatory effect of PPARβ/δ agonist on osteoclast differentiation under normal glucose culture. However, PPARβ/δ agonist exhibited no effect on osteoclast differentiation in vitro (Figures 3A,B). RT-PCR results showed there is no significant difference in osteoclastogenesis-related genes before and after PPARβ/δ agonist treatment (Figure 3C).

Given that immune homeostasis imbalance was an important factor leading to diabetic complications, we explored the effect of PPARβ/δ agonist one the ROS level and pro-inflammatory polarization of macrophages induced by high glucose. The results showed that PPARβ/δ agonist effectively reduced the high glucose-induced ROS production (Figures 4A,B) and the expression of M1 signature genes (iNOS, IL-1β, and TNF-α), while improved the expression of M2 signature genes (CD206, Arg-1, and IL-10) (Figure 4C).

Our previous studies proved that the macrophage polarization could affect osteoclast differentiation (Zhang et al., 2021). Therefore, we supposed that the reduced inflammatory phenotypes of macrophage might be the key mechanism for PPARβ/δ agonist to alleviate diabetic osteoclast activity. After adding macrophage culture supernatant of NG, HG, and HG + GW groups, the differentiation levels of osteoclasts were different (Figure 4E). The macrophage culture supernatant from HG group could significantly increase both the number and size of osteoclasts, while supernatant from HG + GW group reversed the stimulating effect on osteoclast differentiation (Figures 4F,G). The results of RT-PCR showed that the macrophage culture supernatant from HG group could significantly increase the expression of osteoclastogenesis-related genes, which were partially reduced by HG + GW Supernatant group (Figure 4H). To clarify which factor in the culture supernatant affected the osteoclast differentiation, we extracted the macrophage culture supernatant and measured the protein levels of IL-1β and TNF-α by ELISA. We found that high-glucose culture increased the expression level of IL-1β and TNF-α, and PPARβ/δ agonist treatment significantly decreased IL-1β and TNF-α level, and the reduction of TNF-α was the most significant (Figure 4D). In addition, after neutralizing antibodies treatment, we found that TNF-α blockade had the most obvious effect on inhibiting osteoclast differentiation, and IL-1β blockade also inhibited the osteoclast differentiation (Supplementary Figure S1). We speculate that TNF-α is the most important inflammatory factor that stimulates osteoclast differentiation under high glucose, and PPARβ/δ agonist treatment can reduce osteoclast differentiation by reducing the level of TNF-α.

To validate that PPARβ/δ agonist could reduce the inflammatory phenotypes of macrophage in vivo, immunofluorescence staining (Figures 5A–C) and flow cytometry analysis (Figures 5D,E) of bone marrow tissue were performed to detect the expression of CD86 (M1 marker) and CD206 (M2 marker). The CD86⁺ cells in the DM group significantly increased compared to control group, while the number of CD206⁺ cells decreased, indicating that M1 polarization of macrophages increased while the M2 polarization decreased in DM group. PPARβ/δ agonist reduced the proportion of CD86⁺ cells and enhanced the proportion of CD206⁺ cells, indicating that PPARβ/δ agonist treatment in vivo could effectively rectify the macrophage dysfunction in skeletal system under diabetic condition.

After determining the regulatory effect of PPARβ/δ agonists on macrophage polarization in vivo and in vitro, we hope to explore the regulatory mechanism and downstream signaling pathways of PPARβ/δ agonist. As an important transcription factor, PPARβ/δ had been reported to bind to the PPRE region of genes and activate gene transcription. By analyzing ChIP-seq data in GEO database, we found that there is a binding peak at the intronic region of Angptl4, which is reported to coincide with the PPRE region (Inoue et al., 2014) (Figure 6A). Based on the above evidence, we speculated that Angptl4 could serve as a potential downstream gene of PPARβ/δ.

In mice BMDM, PPARβ/δ agonist treatment also up-regulated the expression level of Angptl4 (Figure 6B). Through ChIP assay, we proved that PPARβ/δ had a binding site in the gene segment of Angptl4 in BMDM (Figure 6C). Knockdown of Angptl4 by siRNA in BMDM eliminated the protective effect of PPARβ/δ agonist on the inflammatory phenotypes of macrophages under high-glucose culture (Figure 6D). We evaluated the expression level of ANGPTL4 in serum of diabetic mice by ELISA test and verified ANGPTL4 expression in bone marrow tissue through immunohistochemical staining. We found that after PPARβ/δ agonist treatment, ANGPTL4 was significantly up-regulated (Supplementary Figure S2). However, as for the effect of secreted ANGPTL4, blockade of ANGPTL4 by neutralizing antibody in culture supernatant did not significantly affect osteoclast differentiation, which suggested that the regulation of osteoclast differentiation by PPARβ/δ agonist seemed to be independent of secreted ANGPTL4 (Supplementary Figure S3). Therefore, we believed that Angptl4 could serve as an important downstream gene of PPARβ/δ to regulate macrophage polarization (Figure 6E).