Pathophysiological Mechanism of Bone Loss in Type 2 Diabetes Involves Inverse Regulation of Osteoblast Function by PGC-1α and Skeletal Muscle Atrogenes: AdipoR1 as a Potential Target for Reversing Diabetes-Induced Osteopenia

Type 2 diabetes typically is a middle age–onset (>40 years) metabolic disease that affects multiple organ systems. Increasing evidence indicates that type 2 diabetes is associated with increased fracture risk, especially vertebral and hip fractures in older patients (1–4). Because chronic inflammation participates in diabetes pathogenesis and is a prerequisite for osteoclast activation, increased bone resorption is considered the likely cause of increased fracture risk in this disease. However, some studies intriguingly suggest that patients with type 2 diabetes have an increased risk of hip fractures at a higher bone mineral density (BMD) value than patients without diabetes (5) as well as an increased incidence of vertebral fractures at BMD values comparable to that of patients without diabetes (6). Thus, the reasons for increased skeletal fragility in patients with type 2 diabetes remain largely unexplained.

Patients with type 2 diabetes also have delayed fracture healing, resulting in poorer outcomes after hip fracture (7). Reduced fracture healing in type 2 diabetes is attributable to decreased collagen content, defective cross-linking, alterations in collagen subtype ratios, and collagen defects due to accumulation of advanced glycation end products (8–10), which could impair osteoblast function. These could also lead to compromised bone material strength and increased cortical porosity, affecting bone quality as shown in postmenopausal women with type 2 diabetes (11,12).

Several mouse models of type 2 diabetes are available; however, they poorly represent the human skeletal fragility observed in patients with typical type 2 diabetes (13). Unlike typical middle-aged type 2 diabetes onset in humans, when skeletal maturity has already been attained in genetically engineered diabetic mice, the onset of diabetes occurs at 4–8 weeks, an age comparable to adolescence in humans at which skeletal maturity has not been attained (14). However, the average onset age is falling in humans and is becoming increasingly common among those aged <30 years, including children and adolescents in various ethnic groups (15–18). This early-onset diabetes is characterized by increased disease severity and pancreatic β-cell failure than is typical of type 2 diabetes (19–21). Data on the impact of early-onset type 2 diabetes on human skeletal health are limited, although one report indicated that children with prediabetes with impaired glucose tolerance have low mineral content and low bone mass (22). We thus believe that the distinction in skeletal phenotype between early- and middle age–onset type 2 diabetes might be important for the following reason: Although skeletons of children/adolescents predominantly undergo modeling-directed growth (resulting in a net increase in bone mass due to enhanced osteoblastic activity), adult skeletons predominantly experience remodeling (no net bone gain), and thus, type 2 diabetes in these two cases may affect skeletal health differently.

Although monogenic obese and diabetic mouse models, such as leptin receptor–deficient diabetic mice, essentially differ from the polygenic disease origin in human type 2 diabetes, these may serve as a suitable model for deciphering the skeletal outcomes in this disease. We investigated the skeletal phenotype in leptin receptor–deficient genetically obese diabetic mice (in C57BLKS background) (db) that manifest severe diabetes, including pancreatic β-cell failure. The study also involved the identification of factors crucial for type 2 diabetes–induced skeletal effects. Furthermore, modulation of such factors by therapeutic intervention on diabetic skeleton was assessed.

Cell culture reagents were from Life Technologies. Fine reagents were from Sigma-Aldrich unless indicated otherwise. Globular adiponectin (gAd) was from ATGen Global. ¹²⁵I (20 MBq) was from Bhabha Atomic Research Center (Mumbai, India). Kits for plasma biochemical parameters were glucose (Pointe Scientific), lipids and creatinine (Randox Laboratories Ltd., Mumbai, India), insulin (Millipore), adiponectin (B-Bridge International Inc.), and osteocalcin (OCN) (Uscn Life Science Inc.). ELISA kits were plasminogen activator inhibitor 1, MCP-1, leptin, and resistin (Millipore). The TUNEL assay kit was from Roche Applied Science. 6-C-β-d-glucopyranosyl-(2S,3S)-(+)-5,7,3′,4′-tetrahydroxydihydroflavonol (GTDF) (purity >98%) was purified as previously reported (23), and metformin (Met) (purity 97%) and pioglitazone (Pio) (purity ≥98%) were from Sigma-Aldrich.

Wild type (wt) (C57BLKS/J) or db (BKS.Cg-Dock7^m +/+ Lepr^db/J and B6.db; B6.BKS(D)-Lepr^db/J) mice were housed at 22 ± 3°C on a 12-h light/dark cycle. All animals had access to a standard chow diet and water ad libitum. Time course studies were conducted at the American Association for Laboratory Animal Care–accredited facility of Zydus Research Centre (ZRC) (Ahmedabad, India) following approval from the Institutional Animal Ethics Committee of ZRC. The db or wt mice used were originally from The Jackson Laboratory, and the colonies were maintained at ZRC. Drug treatment studies using 10-week-old db mice from Harlan Laboratories (the Netherlands) (BKS.Cg- + Lepr^db/+Lepr^db/OlaHsd) were conducted at Syngene International Ltd. (Bangalore, India) in its American Association for Laboratory Animal Care–accredited facility following ethical approval from the Institutional Animal Ethics Committee. In both studies, all the animals were randomized into groups based on blood glucose levels and body weight. Vehicle groups received 0.5% carboxymethylcellulose, and treatment groups received GTDF (10 mg/kg), Met (350 mg/kg), and Pio (10 mg/kg) once a day by oral gavage for 30 days. Doses of Met in adult humans range from 850 to 2,550 mg/day, which comes to 14.6–42.5 mg/kg (considering the average human weight to be 60 kg). The adult human dose of Pio is 15–45 mg/day, which corresponds to 0.25–0.75 mg/kg. The equation for dose translation from human to mouse was based on body surface area: human equivalent dose (mg/kg) = animal dose (mg/kg) × (animal K_m/human K_m) (24). Feed intake and body weight were measured every day, and on day 31, the animals were killed. Plasma and tissues were collected and stored at −80°C until further analysis.

Microcomputed tomography (μCT) of excised bones was carried out using a SkyScan 1076 CT scanner (Aartselaar, Belgium) as described earlier for mouse bone (25,26) and following the general guidelines for the assessment of bone microarchitecture in rodents using μCT (27). For scanning, source voltage was set to 50 kV and current to 200 μA. The X-ray source rotation step size was 0.84° over a trajectory of 180°. Reconstructions were made using NRecon version 1.6.9.4 software (SkyScan) to create two-dimensional (2D) 2,000 × 2,000–pixel images with a beam hardening correction set to 10% with dynamic range of −1,000 to 11,000 Hounsfield units. By drawing ellipsoid contours, trabecular bone was extracted using the CT analyzer software. In the femur epiphysis region, 200 slices were selected, leaving 50 slices from the start of the growth plate as a reference point. Cortical parameters were determined at femur mid-diaphysis by 2D analysis. From the start of the growth plate as a reference point, 200 slices were selected in the cortical region, leaving 500 slices as offset (to exclude the trabecular region). For BMD calibration, 2-mm-diameter hydroxyapatite phantom rods with known BMD (0.25 and 0.75 g/cm³) were used. For each analysis, the estimated BMD was determined based on linear correlation between the μCT attenuation coefficient and BMD (28).

Three-point bending test on femur was performed using a bone strength tester (TK-252C; Muromachi) as previously reported (29).

Deparaffinized and hydrated femoral epiphysis sections (5 μm) of various groups were stained with hematoxylin-eosin (H-E), and bone lining cells were visualized by light microscopy. Ten sections per mouse (n = 6) were used for counting by two independent researchers blinded to the experimental design.

Mouse calvarial osteoblasts (MCOs) were obtained from 1–2-day-old mouse pups as described earlier (26,30). Bone marrow cells from 10–12-week-old male wt and db mice were isolated, cultured, and differentiated into osteoblasts as described earlier (31).

Ten micrograms gAd was radioiodinated by iodogen method using precoated iodination tubes (Pierce) according to the manufacturer’s instructions. Excess ¹²⁵I was removed by a PD-10 desalting column (GE Healthcare). For binding assays, osteoblasts and myocytes in 24-well plates were incubated with increasing concentrations of ¹²⁵I-gAd in PBS supplemented with 0.1% BSA for 2 h (at which time binding equilibrium was achieved), after which the cells were washed and lysed. Nonspecific binding for each concentration was determined using a 200-fold excess of cold gAd. Specific binding was calculated by subtracting nonspecific binding from total binding.

Small interfering (si) RNAs were from Dharmacon. MCOs were transfected with 0.1 μmol/L of each siRNA per well using DharmaFECT 1 transfection reagent (Dharmacon) in six-well plates. Seventy-two hours after transfection, cells were treated as indicated and analyzed as required.

These studies were performed using standard procedures as previously described (32,33). Primer sequences for quantitative PCR (qPCR) are listed in Supplementary Table 3. Antibodies and dilutions for immunoblotting were as follows: anti-peroxisome proliferator–activated receptor γ (PPARγ) coactivator 1-α (PGC-1α) (ST1202; Calbiochem; 1:2,000); anti-AMPK, phospho-AMPK (pAMPK) (Thr 172), and cleaved caspase-3 (Cell Signaling Technology; 1:1,000); and adiponectin receptor (AdipoR) 1, AdipoR2, muscle ring finger protein-1 (MuRF1), and β-actin antibodies (Santa Cruz Biotechnology; 1:1,000 except for β-actin, which was 1:3,000). TUNEL assay was performed as described earlier (33). For immunohistochemistry, femur epiphysis transverse sections (5 μm) were deparaffinized, hydrated, and after antigen retrieval, incubated with mouse anti-Runx2 (Abcam; 1:1,600) along with anti-PGC-1α (1:2,000), pAMPK (1:500), or MuRF1(1:500) at 4°C overnight. The sections were then washed and incubated with fluorescent Alexa Fluor goat anti-mouse and goat anti-rabbit IgG (H + L) (1:1,500) (Life Technologies) for 1 h at room temperature. Sections were also stained with DAPI and visualized by fluorescent microscopy. Image-Pro Plus 6.1 software (Media Cybernetics) was used for quantification of microscopic data where five randomly selected fields from six bone sections per group were analyzed.

MCOs at 70–80% confluence were trypsinized, and 2 × 10³ cells/well were seeded in 96-well plates. After 24 h, cells were given various treatments for 48 h in osteoblast differentiation medium. Alkaline phosphatase (ALP) activity was measured using a fluorometric kit (BioVision) according to the manufacturer’s protocol.

MCOs in 96-well plates (2 × 10³ cells/well) were treated with increasing concentrations of glucose, palmitic acid, and dexamethasone (Dex) with or without GTDF and Met for 48 h. Osteoblast viability was assessed by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay as previously reported (33).

Annexin V-FITC Apoptosis Detection kit (Sigma) was used to determine apoptosis. Briefly, the treated cells were trypsinized and washed with PBS, and 1 × 10⁶ cells/mL were resuspended in binding buffer and labeled with 5 μL annexin V-FITC and 10 μL propidium iodide (PI) for 10 min in the dark. Cell fluorescence was measured on a FACSCalibur flow cytometer (Becton Dickinson) and analyzed using CellQuest Pro software.

Results are expressed as mean ± SE. All data were analyzed using GraphPad Prism 5.0 software. Statistical analyses were performed using one- or two-way ANOVA as appropriate followed by Bonferroni posttest analysis.

Three-dimensional μCT evaluation of trabecular microarchitecture in 8-, 12-, and 16-week-old C57BLKS/J (wt) and db mice revealed that db femur epiphysis at all ages displayed a loosely connected trabecular network compared with wt mice (Fig. 1A). The wt mice displayed bone gain at 12 weeks characterized by significantly higher BMD, trabecular bone volume (BV/TV), trabecular number (Tb.N), and connectivity density (Conn.D), followed by trabecular loss at 16 weeks manifested by decreases in these parameters (Fig. 1A). The db mice did not gain bone at any age and were osteopenic throughout, characterized by significantly lower BMD, BV/TV, Tb.N, trabecular thickness (Tb.Th), and Conn.D and higher trabecular separation (Tb.Sp) than wt mice (Fig. 1A). Age-based comparison in db mice also revealed progressive osteopenia characterized by significantly higher Tb.Sp at 16 weeks, and other parameters compared with 8-week-old db mice showed a decreasing trend with age.

Femur mid-diaphysis of db mice representing cortical bone showed a thinner cortex than wt mice at all ages (Fig. 1B). At 8 weeks, BMD, average cortical thickness (Ct.Th), and cortical area (Ct.Ar) were comparable between wt and db, the latter group displaying significantly lower periosteal perimeter (Ps.Pm) and endocortical perimeter (Ec.Pm). At 12 and 16 weeks, all cortical parameters in db were significantly lower than corresponding wt groups (Fig. 1B). Like trabecular parameters, cortical parameters, including BMD, Ct.Th, and Ct.Ar, in 12-week-old wt mice were significantly higher than 8-week-old wt mice, and at 16 weeks, these parameters displayed a decreasing trend compared with 8-week-old wt mice (Fig. 1B). The db mice showed no cortical gain at any age (Fig. 1B).

Osteoblast apoptosis is associated with both primary and secondary osteoporosis (34) and compared with wt femur epiphysis in db mice across all age-groups, displayed remarkably increased osteoblast apoptosis (Fig. 1C and Supplementary Fig. 1A). Periosteal or bone lining cells are a source of osteogenic precursors (35). At 12 weeks, wt but not db bones displayed significantly higher periosteal cell numbers than the 8-week groups (Fig. 1D and Supplementary Fig. 1B), and db mice at all ages had significantly lower periosteal cell numbers than wt mice (Fig. 1D). The osteogenic surrogate serum OCN level dropped significantly with age in both mice; however, as reported earlier (36), db mice had significantly lower OCN levels than wt mice (Fig. 1J).

Compared with wt, nonfasting and fasting glucose was significantly higher in 8-week-old db mice (>280 and >150 mg/dL, respectively), which further increased to >450 and >200 mg/dL at 12 weeks and remained steady thereafter (Fig. 1E and F). The db mice also displayed significant nonfasting and fasting hyperinsulinemia (Fig. 1H and I); however, insulin levels in 16-week-old db mice was significantly less than in 8-week-old db mice, probably due to β-cell apoptosis, which is typical of the db strain. Despite the decrease in insulin, 16-week-old db mice had intriguingly comparable blood glucose levels to that of 12-week-old db mice, although we cannot explain it. Presumably, this finding was due to elimination of glucose through urine, which again, is a trait in db mice. Nonesterified fatty acid (NEFA) level in db mice was significantly higher than wt at all ages, and no age-dependent change was observed (Fig. 1G). Consistent with earlier reports (37,38), db mice at 8 weeks displayed a significantly lower adiponectin level than wt, which decreased further with age (Fig. 1K).

We next assessed whether NEFA and glucose, the two major mediators of diabetic pathology, could directly affect osteoblast viability. Both palmitate and glucose induced loss of MCO viability and apoptosis in vitro (Fig. 2A and B). Apoptosis-related cysteine protease caspase-3 activation assay revealed that palmitate and Dex but not glucose enhanced cleaved (active) caspase-3 levels (Fig. 2C).

Consistent with enhanced osteoblast apoptosis, db femur epiphysis displayed significantly higher p53 expression than wt at all ages (Fig. 3A). Among the db group, 16-week-old mice had significantly higher p53 than 8-week-old mice (Fig. 3A). Consistent with peak bone gain (Fig. 1A and B), wt mice displayed significantly higher Runx2 (key osteogenic factor) expression at 12 weeks followed by a decline at 16 weeks, whereas Runx2 mRNA in db femur epiphysis was significantly lower than wt at all ages (Fig. 3A). Furthermore, among the db group, Runx2 expression significantly declined with age (Fig. 3A).

Muscular PGC-1α expression is suppressed in diabetes (39,40). Because PGC-1α is parathyroid hormone responsive (41) in osteoblasts and its expression increases during osteoblast differentiation (42), we assessed its skeletal expression. Like Runx2, PGC-1α expression in wt but not db mice peaked at 12 weeks and then declined, and compared with wt, PGC-1α expression in db bones was significantly lower and showed significant decline with age (Fig. 3A).

Diabetes and obesity negatively influence muscular health by increasing atrogenes (43) that are involved in protein catabolism. Increasing PGC-1α expression and activity downregulates these atrogenes and prevents muscle atrophy under diverse stresses (44). Because some of these atrogenes are reported in bone and genetic ablations of the E3 ubiquitin ligase MuRF1 and lysosomal protease (cathepsin L) prevent unloading (45) and ovariectomy-induced bone loss (46), we assessed their skeletal expression. Compared with wt, db femur had significantly higher MuRF1, atrogin-1, and cathepsin L transcripts, and their levels in db but not wt mice increased significantly with age (Fig. 3A). Consistent with mRNA expression, PGC-1α protein level was drastically lower in db than wt mice of corresponding age-groups (Fig. 3B). Conversely, MuRF1 protein level was modestly higher in 8-week-old db mice than in wt mice but dramatically increased at 12 and 16 weeks (Fig. 3B).

Consistent with their ability to directly induce osteoblast apoptosis, palmitate or glucose alone robustly enhanced MuRF1 and atrogin-1 and decreased PGC-1α mRNA and protein in MCOs (Dex was used as positive control) (Fig. 3C and D).

We next asked whether modulating PGC-1α expression and activity could ameliorate diabetes-induced osteopenia. Adiponectin signaling through AdipoR1 in particular modulates PGC-1α (47,48); therefore, we first assessed its expression in wt and diabetic skeleton at various ages.

We recently reported that compared with B6.db, db mice at 12 weeks displayed severely depleted muscular AdipoR1 protein and impaired response to gAd (32). Of note, AdipoR1 was readily detectable in the wt, B6.db, and db bones, whereas consistent with our previous report (32), muscular AdipoR1 was markedly reduced in db but not B6.db mice (Fig. 4A). In agreement with differential AdipoR1 expression, acute gAd exposure caused AMPK phosphorylation in both muscle and bones of wt mice but only in bones of db mice (Fig. 4B).

To decipher intact AdipoR1 expression in bones from its depletion in db skeletal muscle, we assessed microRNA-221 (miR-221) and RNA binding protein polypyrimidine tract binding protein (PTB) levels because they negatively regulate AdipoR1 expression (49). miR-221 level in bone and muscle across all ages was significantly higher in both diabetic strains than wt, although compared with bone, differences in muscle was higher (two- to threefold in bone vs. three- to sixfold in muscle) (Fig. 4C). The PTB expression pattern, however, was different from miR-221. Although 8-week-old bones had similar PTB expression across groups, 12- and 16-week-old db bones displayed modest, but significantly higher PTB expression than both wt and B6.db (Fig. 4C). Skeletal muscle, however, showed a dramatically higher PTB expression in 12- and 16-week-old db mice (5- to 20-fold) than both wt and B6.db mice. Taken together, the PTB and miR-221 expression pattern appeared to correlate with the loss of AdipoR1 in skeletal muscle of db mice.

We previously have shown that the osteoanabolic agent GTDF (23), acting as an AdipoR1 agonist, ameliorates diabetes in B6.db but not in db mice (32). Because AdipoR1 action in bone induces an osteogenic effect (50) and db bones were AdipoR1 intact (Fig. 4A), we reasoned that despite its inability to ameliorate diabetes, GTDF may still show an osteoanabolic effect in db mice.

Testing the effect of GTDF along with standard antidiabetic drugs necessitated a sufficient number of age- and sex-matched db mice that was not available in the present colony; thus, we procured fresh db mice with identical genetic background. This also allowed us to confirm that the skeletal phenotype observed in db mice (Fig. 1) were not due to breeding and maintenance-associated local factors. Trabecular and cortical parameters of both db mice were comparable (Supplementary Fig. 2A and B), and both displayed AdipoR1 expression in bone but not skeletal muscle (Supplementary Fig. 2C). These newly acquired db mice were then used for further studies.

We treated db mice with GTDF for 4 weeks at a dose (10 mg/kg) that failed to rescue diabetes in them (32), so any osteogenic outcome would not be a consequence of improved diabetic phenotype. We compared the skeletal effects of GTDF with clinically used antidiabetic drugs Met (AMPK/PGC-1α activator) and Pio (PPARγ agonist).

Assessment of metabolic parameters revealed that although final body weight significantly increased in Pio-treated mice as expected, GTDF or Met did not alter it (Supplementary Table 1). EchoMRI data showed that Pio but not GTDF or Met significantly increased fat mass, whereas none of the treatments altered lean mass or water content (Supplementary Table 1). Pio but not GTDF or Met significantly decreased fasting and nonfasting blood glucose levels, and Pio alone significantly reduced plasma triglyceride and VLDL levels (Supplementary Table 2).

In gross observation by μCT, deterioration of femoral and tibial trabecular architecture was readily observed in vehicle-treated db mice, whereas both GTDF and Met caused substantial improvement (Fig. 5A). Compared with vehicle-treated db mice, GTDF-treated db femur had significantly higher BV/TV, Tb.N, Tb.Th, and Conn.D and lower Tb.Sp, and all the parameters were comparable to wt mice (Fig. 5A), suggesting complete trabecular restoration. Although BV/TV, Tb.N, and Tb.Th in the Met-treated group were restored to wt levels, these mice still had significantly lower Conn.D and higher Tb.Sp than wt mice (Fig. 5A), indicating a partial restoration. Pio modestly but significantly improved BV/TV, Tb.N, Tb.Th, and Conn.D in db femur but failed to reduce Tb.Sp, and except for Tb.Th and Tb.N, could not restore other parameters to wt levels (Fig. 5A). Tibial trabecular data show that except for Tb.Sp, all parameters in db mice were significantly lower than in wt mice (Fig. 5A). GTDF-treated db mice displayed significant improvement in all tibial parameters, and except for BV/TV and Tb.Sp, the rest were restored to wt levels (Fig. 5A), suggesting substantial tibial restoration. Met caused partial tibial restoration because only Tb.Th and Conn.D were comparable to wt (Fig. 5A). Pio was ineffective in restoring tibial cancellous bone because all parameters were lower than wt (Fig. 5A).

Biomechanical strength of femur diaphysis assessed by 3-point bending showed that db mice had significantly lower strength parameters than wt mice (data not shown) and that GTDF- and Met- but not Pio-treated db mice displayed significantly higher resistance to bending than vehicle-treated db mice, as evidenced from higher energy, failure load, and stiffness (Fig. 5B).

Furthermore, GTDF- and Met- but not Pio-treated db mice had higher periosteal cell numbers (Fig. 5C and Supplementary Fig. 3). Osteogenic effect of GTDF and Met was also evident from significantly higher Runx2 and OCN mRNAs in femur epiphysis of db mice treated with them, whereas Pio had no effect (Fig. 5D). Additionally, GTDF and Met but not Pio significantly lowered expression of adipogenic markers PPARγ and CCAAT-enhancer-binding protein α in db mice (Fig. 5D). In support of direct effects of GTDF and Met on osteoblasts, the cytotoxic and apoptosis-inducing effects of glucose and palmitate on osteoblasts were mitigated by GTDF and Met but not Pio (Fig. 5E and F).

AMPK is a key downstream mediator of AdipoR1 signaling, which phosphorylates and thereby activates PGC-1α (47). Osteoblasts in vehicle-treated db femur epiphysis had severely depleted pAMPK compared with wt (Fig. 6A). Consistent with the AdipoR1-intact status of diabetic bone, GTDF-treated db epiphysis had significantly higher pAMPK than vehicle-treated db (Fig. 6A). Met activates AMPK and, thereby, stimulates osteoblast differentiation (51,52), and consistent with this finding, Met-treated db mice displayed significantly higher pAMPK than vehicle controls (Fig. 6A). Pio-treated db also had modest but significantly higher pAMPK than vehicle-treated db (Fig. 6A). Similar to AMPK, PGC-1α expression in osteoblasts was severely lower in vehicle-treated db mice, and GTDF and Met but not Pio restored it (Fig. 6A and Supplementary Fig. 4). MuRF1 displays an inverse correlation with PGC-1α (Fig. 3) (44). In the current study, MuRF1 was also strongly expressed in vehicle-treated db osteoblasts, and in GTDF- and Met- but not Pio-treated db mice, its expression was significantly lowered (Fig. 6A and Supplementary Fig. 4). These results were confirmed by immunoblotting using whole marrow–free femur and tibia, where although PGC-1α was undetectable in vehicle- and Pio-treated db mice, it was strongly expressed in GTDF- or Met-treated db mice (Fig. 6B). Conversely, MuRF-1 level was high in vehicle- and Pio-treated db bones, whereas in GTDF- and Met-treated bones, it was drastically lower (Fig. 6B).

We next determined the roles of AdipoR1, AdipoR2, PGC-1α, and MuRF1 in osteoblasts by individually silencing them. Figure 7A shows confirmation of silencing of these proteins in osteoblasts.

MuRF1 depletion reduced both palmitate- and glucose-induced apoptosis in MCOs by >50% (Fig. 7B), suggesting the mediatory role of this atrogene in osteoblast apoptosis. We asked whether the protection conferred by GTDF against palmitate- and glucose-mediated MCO apoptosis (Fig. 5G) was AdipoR1 dependent. Indeed, siAdipoR1 but not siAdipoR2 or nonsilencing control siRNA (siC) abolished antiapoptotic effects of GTDF (Fig. 7C and Supplementary Fig. 5). siPGC-1α also abrogated the GTDF effect (Fig. 7C). As expected, the antiapoptotic effect of Met was attenuated by siPGC-1α but not siAdipoR1 (Fig. 7C).

Because both GTDF and Met restored trabecular and cortical parameters in db mice, which essentially indicated osteoanabolic actions of these compounds, we assessed GTDF- and Met-mediated osteoblast differentiation. In control MCOs (siC transfected), ALP activity was significantly stimulated by GTDF and Met (gAd and bone morphogenetic protein 2 [BMP-2] were used as positive controls for AdipoR1 and differentiation, respectively) (Fig. 7D). Silencing PGC-1α abolished the induction of ALP activity stimulated by all agents (Fig. 7D). Upon AdipoR1 silencing, the ALP stimulatory effect of gAd or GTDF but not BMP-2 was abolished (Fig. 7D). Silencing AdipoR2 failed to have an impact on the stimulatory effect of gAd or GTDF on ALP activity (Fig. 7D). These data suggest a specific role of AdipoR1 and PGC-1α in GTDF-mediated osteoblast differentiation. Of note, basal ALP activity in MCOs was also significantly depleted in the presence of siPGC-1α and siAdipoR1 but not siAdipoR2 (Fig. 7D), indicating that the autonomous activities of the two proteins might also be required for osteoblast differentiation.

Although there are numerous rodent models of type 2 diabetes, few have undergone thorough assessment of bone structure in relation to peak bone mass achievement and age-related bone loss. Among various mouse models that represent human obesity associated with type 2 diabetes, leptin receptor–deficient mice in a C57BLKS background represent severe phenotypes (53). In the current study, wt mice in C57BLKS background displayed peak bone mass achievement at both trabecular and cortical sites at 12 weeks, whereas db mice did not. Several reports have shown osteopenia in leptin receptor–deficient mice (6,13) with the exception of one (54), which showed high bone mass compared with wt littermates. The current data corroborate the finding of Williams et al. (36), showing trabecular osteopenia in ∼11-week-old male db mice. In addition, db mice showed severe age-related bone loss. From these observations, it appears that db mice had an early termination of peak bone gain and/or early onset of age-related bone loss. Bone accrual critically depends on osteoblast number and function, and because db mice displayed robust apoptosis in osteoblasts at all ages, it appears that the lack of viable osteoblasts and resultant deficiency in osteoblast function in the db bones was responsible for the failure in peak bone mass achievement and accelerated the development of age-related osteopenia. Bone formation is consistently lower in patients with type 2 diabetes compared with those without diabetes as evidenced by lower serum OCN (55), and as reported before (36), we observed that db mice also have markedly lower OCN levels at all ages than wt mice.

Glucose and palmitate induced mouse primary osteoblast apoptosis at pathologically relevant concentrations. The extent of cytotoxicity induced by palmitate and glucose was comparable to Dex, a potent inducer of osteopenia (33). It thus appears that the combined effect of glucotoxicity and lipotoxicity causes osteopenia equivalent in severity to that caused by Dex, and diabetic bones indeed shared common features of Dex-induced osteoporosis because bone was lost at both trabecular and cortical sites.

Tumor suppressor p53 is a negative regulator of osteoblast differentiation because it suppresses Runx2 expression, and hypermorphic p53 mutation in mice causes osteopenia (56). In db bones, p53 was increased and Runx2 suppressed, suggesting reduced osteoblast number and differentiation. Skeletal muscle atrogenes were also elevated in the db bones and showed an age-related increase, which could also negatively affect osteoblast survival and differentiation. In fact, silencing MuRF1 conferred robust protection against palmitate- and glucose-mediated MCO apoptosis. It is possible that other atrogenes (atrogin-1 and cathepsin L) also play key roles in osteopenia induced by various stresses because these factors were also induced in db bones and glucose- and palmitate-treated MCOs. Other reports support such a notion, as mice lacking these atrogenes are protected from osteopenia under diverse stresses (45,46). In contrast, the muscle anabolic factor PGC-1α exhibited an age-related decline in db mice. Therefore, it appears that like skeletal muscle, a reciprocal relationship between MuRF1 and PGC-1α exists in osteoblasts. Thus, loss of osteoblast population and function in db bones may occur due to suppression of PGC-1α and induction of atrogenes. The regulation and possible interaction of p53 and Runx2 with MuRF1 and PGC-1α would be an interesting future topic of investigation.

AdipoR1 increases insulin sensitivity and promotes cellular energy expenditure by activating AMPK and PGC-1α, and AdipoR1 activation represents an attractive therapeutic approach for the treatment of obesity and type 2 diabetes (48). Previously, we showed that db skeletal muscle is deficient in AdipoR1 expression (32). In contrast, we show here that diabetic bones are AdipoR1 intact. The explanation behind this appears to involve two negative AdipoR1 regulators, PTB and miR-221 (49), because both, especially PTB, were expressed at much higher levels in skeletal muscle than in bones of diabetic mice.

Adiponectin and its receptors are expressed in bone marrow stromal cells, which suggests their potential role in bone metabolism (57). C57BL6/J mice treated with adenoviral-derived adiponectin had increased trabecular bone mass and enhanced mineralization activity of osteoblasts (58). Mice harboring porcine AdipoR1 transgene had higher bone volume and trabecular numbers than age- and sex-matched controls (50). Cultures of bone marrow stromal cells from adiponectin knockout mice showed significantly less osteogenesis than cultures from adiponectin-intact mice (57). Together, these reports suggest that adiponectin may regulate bone formation in an autocrine/paracrine manner in addition to endocrine mode. Although AdipoR1 expression in db bones was comparable to wt, osteopenia in db mice could be attributed to their observed hypoadiponectinemia.

To investigate whether AdipoR1 in bone, and more specifically osteoblasts, of diabetic mice could be pharmacologically targeted to mitigate osteopenia through an osteoanabolic mechanism, we used GTDF. We have shown that GTDF improves diabetic phenotypes in B6.db mice with intact AdipoR1 in skeletal muscle and liver but not in db mice that lacked functional AdipoR1 in these tissues (32). Thus, db mice with skeletal (but not muscular) expression of functional AdipoR1 allowed us to selectively activate AdipoR1 in bone without correcting the diabetic phenotype. The most common clinically used drugs, Met and Pio, were used at their pharmacologically relevant doses for comparison of skeletal effects between the drugs.

Neither GTDF nor Met improved diabetic phenotypes in db, but Pio was modestly effective, although it failed to reduce NEFA levels but did reduce fasting and nonfasting serum glucose levels, which were still significantly higher than in wt. GTDF and Met improved trabecular microarchitecture and increased bone strength and bone lining cells. Pio was mostly ineffective because some improvement in femur epiphysis was seen, but tibia metaphysis, where bone is actually accrued less than in femur, showed no improvement. Additionally, strength parameters were not improved by Pio, and Pio did not increase periosteal cell number. The modest skeletal improvement by Pio could be due to modest mitigation of diabetic phenotypes because Pio failed to confer protection against gluco- and lipotoxicity on isolated osteoblasts. The skeletal improvement by GTDF and Met without altering diabetic phenotype as well as the protection imparted by them on isolated osteoblasts against glucose and palmitate assault clearly indicate that GTDF and Met had direct osteoblastic effects. The current data are also supported by previous reports where Met was shown to directly induce osteoblast differentiation (51,59) and protect against high-glucose–mediated inhibition of osteoblast growth (60).

Although both GTDF and Met restored bones of db mice by likely preserving osteoblasts against gluco- and lipotoxicity, the molecular mechanism of action of the two were different. Silencing AdipoR1 but not AdipoR2 abrogated the prosurvival and differentiation-promoting effects of GTDF, thus suggesting its AdipoR1-specific effect in osteoblasts as observed earlier in myocytes (32). Skeletal effects of Met, however, were blocked by silencing PGC-1α but not AdipoR1, suggesting that the osteoanabolic effect of the drug was mediated by PGC-1α and not AdipoR1, which was supported by Met failing to replace ¹²⁵I-gAd in a radioligand competition assay (data not shown). Salient findings of the current study are summarized in Fig. 8.

In conclusion, this study showed that diabetic mice in a BLKS background have poor achievement of peak bone mass and age-related development of severe osteopenia. Inducing and activating PGC-1α in osteoblasts by GTDF or Met completely reverse osteopenia in these mice without correcting diabetic phenotypes.

Acknowledgments. The authors acknowledge the sophisticated analytical instrument facility of the Central Drug Research Institute for help with confocal microscopy and flow cytometry. The authors also acknowledge technical support provided by Kavita Singh of the Sophisticated Analytical Instrument Facility, CSIR-Central Drug Research Institute for confocal microscopy.

Funding. M.P.K. and M.Y. were supported by the Indian Council of Medical Research. A.K.S. and A.G. were supported by fellowships from CSIR. S.Sh., H.K., J.S.M., and S.K. were supported by the University Grants Commission. S.Sa. and N.C. acknowledge funding from the CSIR Network project ASTHI (Anabolic Skeletal Targets in Health and Illness).

Duality of Interest. A.A.J. and M.R.J. are employees of Zydus Research Centre, the research and development arm of Cadila Healthcare Ltd. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. M.P.K. contributed to the experimental design and performance, data analysis, and discussion. A.K.S. contributed to the experimental design, performance, and supervision; data analysis; and discussion. A.A.J., M.Y., S.Sh., H.K., A.G., J.S.M., M.C.T., G.K.N., and S.K. contributed to performing the experiments and to the data analysis and discussion. R.R., M.R.J., A.K.T., and R.M. contributed to supervising the experiments and to the data analysis and discussion. A.S. contributed to the data analysis and discussion. M.M.G. contributed to the experimental performance and supervision, data analysis, and discussion. J.R.G. contributed to the experimental design and supervision, data analysis, and discussion. S.Sa. contributed to the study concept, experimental design and supervision, data analysis, discussion, and writing of the manuscript. N.C. contributed to the experimental design and supervision, data analysis, discussion, and writing of the manuscript. N.C. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.