Functional differences between AMPK α1 and α2 subunits in osteogenesis, osteoblast-associated induction of osteoclastogenesis, and adipogenesis

The endocrine role of the skeleton-which is impaired in human diseases including osteoporosis, obesity and diabetes-has been highlighted previously. In these diseases, the role of AMPK, a sensor and regulator of energy metabolism, is of biological and clinical importance. Since AMPK’s main catalytic subunit α has two isoforms, it is unclear whether functional differences between them exist in the skeletal system. The current study overexpressed AMPKα1 and α2 in MC3T3-E1 cells, primary osteoblasts and mouse BMSCs by lentiviral transduction. Cells overexpressing AMPKα2 showed higher osteogenesis potential than AMPKα1, wherein androgen receptor (AR) and osteoactivin played important roles. RANKL and M-CSF were secreted at lower levels from cells overexpressing α2 than α1, resulting in decreased osteoblast-associated osteoclastogenesis. Adipogenesis was inhibited to a greater degree in 3T3-L1 cells overexpressing α2 than α1, which was modulated by AR. An abnormal downregulation of AMPKα2 was observed in human BMSCs exhibiting the fibrous dysplasia (FD) phenotype. Overexpression of AMPKα2 in these cells rescued the defect in osteogenesis, suggesting that AMPKα2 plays a role in FD pathogenesis. These findings highlight functional differences between AMPKα1 and α2, and provide a basis for investigating the molecular mechanisms of diseases associated with impaired functioning of the skeletal system.


Several important hormones secreted by bone cells regulate energy balance and mineral ion homeostasis.
This endocrine function of the skeleton is impaired in various diseases including osteoporosis, obesity, and diabetes-associated bone diseases 1 . Elucidating the molecular basis for the regulation of energy metabolism and hormone production in the skeleton is therefore of biological and clinical importance, and can provide insight into the pathogenesis of these diseases.
Adenosine triphosphate (ATP) is an immediate source of energy in living cells and must therefore be maintained at a relatively high level. In eukaryotic cells, the adenosine monophosphate (AMP)-activated protein kinase (AMPK) signaling cascade detects and initiates a response to decreases in cellular ATP concentration 2 by coupling changes in the intracellular level of ATP to the phosphorylation of downstream substrates, resulting in increases or decreases in the rates of ATP production and consumption, respectively 3 .
Bone is a dynamic organ that is continuously remodeled throughout the lifetime of an organism and is susceptible to alterations in metabolic status and physiological state. Recent studies have revealed that bone metabolism is regulated by the brain and is closely linked to whole body energy homeostasis 1,4,5 . There are two main neuronal populations within the arcuate nucleus of the hypothalamus regulating energy homeostasis: The orexigenic,

Results
AMPK α1 and α2 mRNA expression is upregulated during osteogenesis.  in the current study are preosteoblasts derived from mouse calvaria and have been used extensively as an in vitro model system to examine the osteogenic differentiation. On days 0, 2, 4, and 7 after induction, the expression of Runt-related transcription factor 2/core-binding factor α 1 (Runx2), alkaline phosphatase (Alp), Phospho1, osteocalcin (Ocn), and Ampk α1 and α2 subunits was examined by qRT-PCR. An increase in the transcript levels of Runx2 (Fig. 1A), Alp (Fig. 1B), and Phospho1 (Fig. 1C) between days 2 and 7 and Ocn (Fig. 1D) on day 7 relative to day 0 was observed, indicating that MC3T3-E1 cells differentiated into osteoblasts. Ampk α1 (Fig. 1E) and α2 (Fig. 1F) mRNA expression was also upregulated on days 4 and 7 after induction.
Also, BMSCs were induced to osteogenesis. On weeks 0, 1, 2, and 3 after induction, the expression of Runx2, Alp, Phospho1, Ocn, and Ampk α1 and α2 subunits was examined by qRT-PCR. As an early-stage marker gene of osteogenic differentiation and one of the most important transcriptional factors to initiate osteogenesis, Runx2 mRNA levels showed a biphasic response during osteogenic differentiation, that is, increased from weeks 0 to 1 and reached its maximal levels at weeks 1 and thereupon fell progressively till weeks 3. Of these, there were statistically higher mRNA levels from weeks 1 to 3, relative to weeks 0 (Fig. 1M). An increase in the transcript levels of Alp (Fig. 1N) between weeks 1 and 3 and Phospho1 (Fig. 1O) and Ocn (Fig. 1P) between weeks 2 and 3 relative to day 0 was observed, indicating that BMSCs differentiated into osteoblasts. Ampk α1 (Fig. 1Q) and α2 (Fig. 1R) mRNA expression was also upregulated between weeks 1 and 3 compared with week 0.
Osteoblast-associated osteoclastogenesis of bone marrow monocytes is attenuated by AMPK α2 compared with α1. Osteoclasts differentiation and maturation depends on RANKL and M-CSF secreted by osteoblasts, with the two cell types interacting through direct contact as well as paracrine signaling 14 .
Further, we investigated the osteoclast function induced by LV-AMPKα 1 and LV-AMPKα 2 MC3T3-E1 cells by evaluating resorption pit formation. The results revealed lower osteoclast function induced by LV-AMPKα 1 and LV-AMPKα 2 MC3T3-E1 cells compared with LV-ctr cells (Fig. 5D,E). Moreover, lower osteoclast function induced by LV-AMPKα 2 cells was observed compared with LV-AMPKα 1 cells (Fig. 5D,E). Bone metabolism markers expression in MC3T3-E1 cells overexpressing AMPK α1 and α2 subunits. Given that overexpressing the α 2 as compared to the α 1 subunit of AMPK conferred a greater osteogenic potential and diminished osteoblast-associated induction of osteoclastogenesis, the potential mechanisms underlying this difference were investigated in cells treated with BMP2 for 7 days by evaluating the protein expression levels of 36 markers of bone metabolism using an antibody array (Fig. 6A). The relative expression level of specific proteins was represented by the fluorescent signal intensity. Among the 36 markers, five markers that were differentially expressed between LV-AMPKα 1 and LV-AMPKα 2 MC3T3-E1 cells were of particular interest (Fig. 6A,B). AR and osteoactivin were upregulated by 180% and 110%, respectively, and M-CSF, RANKL, and matrix metalloproteinase (MMP)-24 were downregulated by > 90%, 80%, and 70%, respectively, in LV-AMPKα 2 as compared to LV-AMPKα 1 cells (Fig. 6C). There was no significant difference in OPG expression level between LV-AMPKα 1 and LV-AMPKα 2 MC3T3-E1 cells, so that higher OPG/RANKL ratio was observed in LV-AMPKα 2 cells as compared to LV-AMPKα 1 cells (Fig. 6C). The expression profiles of the other markers are shown in Supplementary Figure 1. In LV-AMPKα 1 and LV-AMPKα 2 cells treated with BMP2 for 7 days, qRT-PCR analysis revealed an upregulation of Ar and Osteoactivin and downregulation of M-csf, Rankl, and Mmp-24 mRNA levels in LV-AMPKα 2 relative to LV-AMPKα 1 cells (Fig. 6D), consistent with the protein expression data from the antibody array.
Osteoactivin and AR are involved in osteogenesis of AMPK α subunit-overexpressing cells. Osteoactivin and AR play essential roles in osteogenic differentiation [15][16][17][18] . Since the expression of both proteins was higher in MC3T3-E1 cells overexpressing the α 2 compared with the α 1 subunit of AMPK, we investigated whether this was responsible for a greater osteogenic potential in LV-AMPKα 2 as compared to LV-AMPKα 1 cells. To answer this question, LV-AMPKα 1 and LV-AMPKα 2 MC3T3-E1 cells were infected, respectively, with LV particles expressing AR and osteoactivin (Fig. 7A) and siRNA against the two proteins (Fig. 7A). The cells were then induced to osteogenesis. A qRT-PCR analysis performed on day 7 showed that the mRNA expression of Runx2, Alp, Bsp, and Ocn was upregulated in response to AR and osteoactivin overexpression in LV-AMPKα 1 cells (Fig. 7B-E). In contrast, AR knockdown resulted in the downregulation of Runx2, Alp, Ibsp, and Ocn mRNA expression in LV-AMPKα 2 cells (Fig. 7B-E). Osteoactivin knockdown also decreased the mRNA expression of Runx2, Alp, and Ocn, but had no effect on the level of Ibsp. On day 21 after induction, Alizarin Red staining and quantification revealed greater mineralization in LV-AMPKα 1 cells overexpressing AR and osteoactivin than in controls, whereas AR and osteoactivin knockdown reduced mineralization in LV-AMPKα 2 cells (Fig. 7F).
We confirmed these results in LV-AMPKα 1 and LV-AMPKα 2 primary osteoblasts (Fig. 7G). A qRT-PCR analysis performed on day 9 showed that the mRNA expression of Osterix, Alp, Ibsp, and Ocn was upregulated in response to AR and osteoactivin overexpression in LV-AMPKα 1 cells (Fig. 7H-K). In contrast, AR and osteoactivin knockdown resulted in the downregulation of Osterix, Alp, Ibsp and Ocn mRNA expression in LV-AMPKα 2 cells (Fig. 7H-K).
We also confirmed the results described above in LV-AMPKα 1 and LV-AMPKα 2 BMSCs (Fig. 7L). A qRT-PCR analysis performed on day 7 showed that the mRNA expression of Runx2, Alp, Ibsp, and Ocn was upregulated in response to AR and osteoactivin overexpression in LV-AMPKα 1 cells ( Fig. 7M-P). In contrast, AR knockdown resulted in the downregulation of Runx2, Alp, and Ocn mRNA expression, but had no effect on the level of Ibsp in LV-AMPKα 2 cells (Fig. 7M-P). Osteoactivin knockdown also decreased the mRNA expression Runx2, Ibsp and Ocn, but had no effect on the level of Alp ( Fig. 7M-P).
Adipogenic potential is suppressed by overexpression of the AMPK α2 subunit. The balance between osteogenesis and adipogenesis in BMSCs is impaired in several human diseases 19 , therefore its regulation is of medical importance 20 . The role of AMPK in adipogenesis was investigated in preadipocyte 3T3-L1 cells stably expressing AMPK α 1 or α 2 subunits. Exogenous AMPK α 1 and α 2 subunits were expressed at high levels in LV-AMPKα 1 and LV-AMPKα 2 3T3-L1 cells (Fig. 8A) respectively, as determined by RT-PCR. The expression of endogenous AMPK α 1 or α 2 was unaffected by the overexpression of the exogenous proteins (Fig. 8A).
AR level is negatively correlated with adipogenesis in BMSCs 18 . Given the upregulation in AR expression and inhibition of adipogenesis induced by AMPK α 2 as compared to AMPK α 1 overexpression, a role for AR in adipogenesis was examined by infecting LV-AMPKα 1 and LV-AMPKα 2 3T3-L1 cells with a virus expressing AR and a siRNA against AR respectively, and inducing adipogenic differentiation. On day 15, Oil Red O staining and quantification were used to evaluate adipogenesis and cells were analyzed for Pparγ, aP2, and Glut4 expression by qRT-PCR. AR overexpression suppressed adipogenesis in LV-AMPKα 1 3T3-L1 cells. The inhibition of adipogenesis was rescued in LV-AMPKα 2 3T3-L1 cells by AR knockdown (Fig. 8C-G).
Further, we confirmed the results described above in LV-AMPKα 1 and LV-AMPKα 2 BMSCs. LV-AMPKα1 and LV-AMPKα 2 BMSCs were infected with a virus expressing AR and a siRNA against AR respectively, and inducing adipogenic differentiation. On day 25, Oil Red O staining and quantification were used to evaluate adipogenesis. AR overexpression suppressed adipogenesis in LV-AMPKα 1 BMSCs. The inhibition of adipogenesis was rescued in LV-AMPKα 2 BMSCs by AR knockdown (Fig. 8H-I).
Role of the AMPK α2 subunit in the impaired osteogenesis of BMSCs with FD. FD is characterized by an impairment in osteogenic differentiation potential in BMSCs due to activating missense mutations in the guanine nucleotide binding protein α -stimulating (Gnas) gene. In normal BMSCs, AMPK α 2 mRNA expression was progressively upregulated during osteogenic differentiation (Fig. 9A). Two in vitro models were established that mimicked the pathological features of FD. In the first, BMSCs were generated that expressed a Gnas mutant harboring an R201H mutation (Gsα R201H ); in the second model, BMSCs were treated with an excess of cell membrane-permeable cAMP 21 . Osteogenic differentiation was induced in the cells, and the mRNA expression of AMPK α 2 was assessed on days 0, 2, 4, and 7. LV-Gsα R201H -BMSCs and excess cAMP-treated BMSCs showed the different pattern of AMPK α 2 subunit mRNA expression with normal BMSCs: no statistical difference was observed at day 2 compared with day 0 and there was significant downregulation at day 4 (by 28%) and 7 (by 67%) compared with day 0 in LV-Gsα R201H -BMSCs (Fig. 9B). Accordingly, in excess cAMP-treated BMSC, there was gradual downregulation in AMPK α 2 subunit mRNA expression at day 2 (by 46%), 4 (by 62%) and 7 (by 76%) compared with day 0 (Fig. 9C). To determine whether the downregulation of the AMPK α 2 subunit is responsible for the reduced osteogenic differentiation potential in BMSCs treated with excess cAMP, the AMPK α 2 subunit was overexpressed by LV transduction in these cells, which were then induced to undergo osteogenic differentiation. Overexpression of the AMPK α 2 subunit rescued the impaired osteogenic differentiation in cAMP-treated BMSCs, as evidenced by the increased ALP, IBSP, and OCN transcript levels ( Fig. 9D-F), implicating the role of α 2 subunit of AMPK in FD pathogenesis.

Discussion
The results of the present study revealed three key functional differences between the AMPK α 1 and α 2 subunits. The α 2 subunit increased osteogenic potential in MC3T3-E1 cells, primary osteoblasts and mouse BMSCs, which involved AR and osteoactivin. Compared to the α 1 subunit of AMPK, the α 2 subunit inhibited osteoblast-associated induction of osteoclastogenesis in MC3T3-E1 cells via downregulation of M-CSF and RANKL. Finally, compared to α 1, overexpression of the α 2 subunit suppressed adipogenesis in 3T3-L1 cells, which also involved AR.
Mammalian AMPK has two isoforms of the α and β and three of the γ subunit, and the various isoforms of each subunit have distinct biological functions 22,23 . A769622 activates AMPK both allosterically and by inhibiting the dephosphorylation of AMPK α Thr172 by specific phosphatases 24 . Its mode of action does not involve binding to the γ subunit, unlike other AMPK activators such as AMP and AICAR, but instead depends on the presence of the β subunit 25 . Interestingly, it was shown that A769622 is selective for the β 1 isoform and does not activate AMPK heterotrimers containing β 2 subunits 26 . In contrast, only AMPK β 2 can be modified posttranslationally by PIASy-dependent SUMOylation, leading to activation of the AMPK complex 25 . One study screened 12,000 genes in the H1299 human lung carcinoma line and found 133 genes that were either induced or repressed in response to p53-dependent cell growth arrest and apoptotic conditions, including β 1 but no other AMPK subunits 27 . Another study demonstrated that α 1 and γ 1 are almost exclusively localized in the cytoskeleton, while α2 and γ2 are present in all subcellular fractions, including the nucleus 28 . These data suggest that pharmacological interventions targeted to specific AMPK subunit isoforms can selectively modify particular AMPK functions.
The α subunit of AMPK is the main catalytic domain of the AMPK complex. A dominant-negative α 2 subunit attenuated the mutant AMPK γ 2 phenotype, and AMPK complexes containing α 2 rather than α 1 subunit mediate the effects of AMPK γ 2 mutations 22 . AMPK α 2 is the main effector of basal and AICAR-stimulated AMPK activity, including AICAR-induced glucose uptake 29 . Clot retraction was impaired in platelets from AMPKα 2 −/− but not AMPKα 1 −/− mice 23 , and AMPK α 2 knockout mice showed increased sensitivity to diet-induced obesity and insulin resistance, whereas no metabolic defects were observed in α 1 knockout mice.
Few reports have compared the functions of AMPK α 1 and α 2 subunits. The AMPK agonists AICAR and metformin induce osteogenesis in MC3T3-E1 cells [30][31][32][33][34] , and metformin causes increases in ALP activity, collagen synthesis, OC production, and extracellular Ca 2+ deposition in vitro, possibly by increasing Runx2 expression. When primary osteoblasts were co-treated with AICAR and the AMPK antagonist compound C, the latter suppressed the stimulatory effect of the agonist on bone nodule formation. The present study analyzed differences in the osteogenic potential of MC3T3-E1 cells expressing the α 1 or α 2 subunits of AMPK. The findings that the α 2 subunit conferred the cells with greater osteogenic potential as compared to α 1 are at variance with those of another study, which found no differences in tibial bone mass between AMPK α 2 knockout and wild-type mice, although both cortical and trabecular bone compartments were smaller in the mutants 35 . In addition to possible differences attributable to the model systems that were used (in vivo vs. in vitro in this study), one possible reason for the discrepancy between the findings is the low expression of α 2 relative to the α 1 isoform in the skeletal system. α 1 is the predominant AMPK isoform expressed (albeit at low levels) in BMSCs, preosteoblasts, and preadipocytes 8,35,36 . The antibody array data presented here showed that the levels of many markers of bone metabolism differed between MC3T3-E1 cells expressing α 1 or α 2 subunits, including AR, basic fibroblast growth factor, interleuking-6 and -17, MCP-1, M-CSF, macrophage inflammatory protein 1a, MMP9, osteoactivin, tumor necrosis factor α , RANKL, and MMP24 ( Supplementary Fig. 1). Of these, osteoactivin and especially AR were implicated in the differences in osteogenesis potential between α 1 and α 2. This is consistent with studies showing that AR deficiency leads to tissue-nonspecific ALP downregulation followed by decreased phosphate production, ultimately reducing bone mineralization. Similar results were observed in osteoblast-specific AR knockout mice, in which AR was found to stimulate osteoblast differentiation and suppress bone resorption 37,38 ; AR mutants developed osteoporosis and showed decreased BMSC osteogenesis resulting from the downregulation of Runx2 39 . In addition, an enhancement of osteogenesis, AR also suppresses adipogenesis. Androgen treatment inhibits adipocyte differentiation and body fat formation in vitro and in rodent and nonhuman primate models, as well as rosiglitazone-induced adipogenesis in human BMSCs; it also promotes interactions between β -catenin, AR, and T-cell factor 4 to suppress adipogenic differentiation of 3T3-L1 cells 40 , which were induced to undergo adipogenesis by overexpression of the α 1 and α 2 subunits of AMPK in this study. The weaker induction in α 2-AMPK 3T3-L1 corresponded to a greater upregulation in AR expression as compared to the adipogenic induction and AR level in α 1-AMPK 3T3-L1 cells. Adipogenesis was rescued in α 2-AMPK 3T3-L1 and inhibited in α 1-AMPK 3T3-L1 cells upon AR knockdown and overexpression, respectively, suggesting that the functional differences between α 1 and α 2 involve AR signaling.
Osteoclast differentiation and maturation depend on cues from the microenvironment, especially from osteoblasts through cell-cell contact and paracrine signaling. This study provides evidence that osteoblast-associated osteoclastogenesis was reduced in bone marrow monocytes co-cultured with MC3T3-E1 cells expressing α 2 as compared to α 1 and likely involves RANKL and M-CSF, as the expression of these two proteins was downregulated in the former. AR, which was upregulated in AMPK α 2-relative to α 1-expressing MC3T3-E1, may also be involved, since it inhibits bone resorption 41 and AR-deficient mice exhibit a calvarial and femoral bone loss phenotype.
The most intriguing finding of the present study as that AMPK α 2 expression was impaired during osteogenesis in an FD model. In normal BMSCs, α 2 subunit transcript level gradually increased during osteogenic differentiation; however, this was abrogated and even reversed in BMSCs exhibiting the FD phenotype, and the defect in osteogenesis in these cells was rescued by overexpression of AMPK α 2, suggesting that the α 2 subunit may be a critical factor in the pathogenesis of FD. For instance, α 2 may be involved in the overactivation of bone resorption as our findings demonstrate a negative relationship between α 2 expression and osteoclastogenesis.
In conclusion, the results from this study indicate that the α 2 and α 1 subunits of AMPK have several functional differences, with α 2 conferring stronger osteogenic potential and a weaker ability to induce osteoblasts-associated osteoclastogenesis in MC3T3-E1 cells as well as conferring a lower adipogenic potential to 3T3-L1 cells. These findings provide a basis for developing drugs that can differentially target the α 1 and α 2 subunits of AMPK to treat diseases such as obesity and osteoporosis that are associated with mutations in specific AMPK subunits.
MC3T3-E1 cells were treated with recombinant human bone morphogenetic protein (BMP)2 (R&D, Minneapolis, MN, USA) at a final concentration of 100 ng/ml to induce osteogenesis.
Primary calvarial osteoblasts. Four-day-old male C57BL/6J mice were killed with a lethal dose of sodium pentobarbital. Calvaria was dissected, cleaned of soft tissue, and maintained in PBS buffer. The cleaned calvaria was cut into small pieces of 1-2 mm 2 and washed with PBS. The bone pieces were incubated with 2 mg/ml collagenase II (Sigma, St. Louis, MO) solution for 2 h at 37 °C in a shaking water bath. Then, the fragments were washed with α -MEM supplemented with 10% FBS and cultured in culture medium (α -MEM containing 10% FBS, 100 U/ml penicillin, 100 μ g/ml streptomycin). The culture flasks were stored in a humidified atmosphere of 5% CO 2 in air at 37 °C. After confluence, the bone fragments were removed and the confluent layers were trypsinized and the cells were replated.
BMSCs. Mouse BMSCs were purchased from Cyagen Bioscience Inc. Human BMSCs were isolated and expanded using a modified version of a previously described method 42,43 . The donor was healthy and had no metabolic or other diseases or inherited conditions that could affect the current study. Written informed consent was obtained from the donor, and the study was approved by the Ethics Committee of Shanghai Ninth People's Hospital Affiliated to Shanghai JiaoTong University School of Medicine. Methods used in this study were carried out in accordance with the the relevant guidelines and regulations.
Mouse BMSCs were induced to osteogenesis by Osteogenesis Induction Medium (OIM) containing 10 mM β -glycerophosphate, 100 μ g/ml ascorbic acid, and 10 nM dexamethasone. To investigate the effect of excess cAMP on AMPK α 1 and α 2 expression in human BMSCs, confluent cells were exposed to medium containing BMP2 and 2 mM dibutyryl-cAMP (Sigma, St. Louis, MO, USA).
3T3-L1 cells. 3T3-L1 cells were cultured in Dulbecco's Modified Eagle's Medium (DMEM) (Gibco) supplemented with 10% FBS, 100 U/ml penicillin, and 100 mg/ml streptomycin. Adipogenic differentiation of 3T3-L1 and mouse BMSCs was induced as previously described 20 . Briefly, confluent cells were treated with a complete adipogenic hormone cocktail consisting of DMEM supplemented with 10% FBS, 10 g/ml insulin, 0.5 mM methylisobutylxanthine, and 1 μ M dexamethasone (all from Sigma). The day differentiation was induced was considered as day 0. On day 3, the culture medium was replaced with DMEM containing only insulin and 10% FBS. On day 6, the complete adipogenic hormone cocktail was again added.
Osteoblast-associated osteoclastogenesis. Bone marrow monocytes isolated from long bones of 6-week-old C57BL/6 mice were cultured with MC3T3-E1 in the presence of BMP2 (100 ng/ml) for 7 days. The cells were then fixed and stained for TRAP. TRAP-positive multinuclear cells (> 3 nuclei/cell) were considered as osteoclasts and were counted. The total area of TRAP-positive regions were quantified using Image J software (National Institutes of Health, Bethesda, MD, USA) in each sample in 5 randomly selected fields of view.
Bone absorption assay. Bone marrow monocytes were cultured with MC3T3-E1 in the presence of BMP2 (100 ng/ml) for 7 days on bovine bone slices. Cells on bone slices were then removed by mechanical agitation and sonication. Bone resorption pits were visualized under a scanning electron microscope (SEM) (FEI Quanta 250), and the percentage of bone resorption area was quantified using Image J software.
Reverse transcription (RT)-and real-time PCR. Total cellular RNA was isolated from cultured cells using the RNeasy Mini Kit (Qiagen, Valencia, CA, USA) according to the manufacturer's protocol. Primers used in all reactions are shown in Supplementary Table 1. For RT-PCR, single-stranded cDNA was reverse-transcribed from 1 μ g total RNA using reverse transcriptase with an oligo-dT primer. PCR was carried out with 1 μ l cDNA using the following cycling parameters: 30 cycles of 94 °C for 40 s, 60 °C for 40 s, and 72 °C for 40 s. PCR products were analyzed by agarose gel electrophoresis. Real-time PCR was carried out on a 96-well plate ABI Prism 7500 Sequence Detection system (Applied BioSystems, Foster City, CA, USA) using SYBR Green PCR Master Mix (Takara Bio Inc., Otsu, Japan). Cycling conditions were as follows: 40 cycles of 94 °C for 5 s, 60 °C for 34 s, and 72 °C for 30 s. The comparative 2 −ΔΔCt method was used to calculate the relative expression of each target gene as previously described 42 ; the expression levels of all genes were normalized to that of β -actin. Lentiviral transduction. LV vectors containing the coding sequences of human AMPK α 1 and α 2 subunits, AR, and osteoactivin and short interfering (si)RNAs against AR and osteoactivin were purchased from Genecopoeia (Rockville, MD, USA). LV particles were generated as previously described 21 . Briefly, 1.3-1.5 × 10 6 293T cells were plated in a 10-cm dish and the transfection complex was added directly to the culture medium when cells reached 70-80% confluence. After incubation in a CO 2 incubator at 37 °C, LV particle-containing medium was collected 48 h post-transfection.
Bone metabolism markers array. Semi-quantitative sandwich-based bone metabolism markers arrays (Human Cytokine Array L-Series; RayBiotech, Atlanta, GA, USA) were used to detect 36 markers of bone metabolism on a glass slide matrix. Biotin-conjugated antibodies used for detection were combined as a single cocktail for later use. Printed slides were placed in chambers to allow incubation of each array with a different sample. Arrays were incubated in blocking buffer followed by whole cell lysis samples. After extensive washing to remove Scientific RepoRts | 6:32771 | DOI: 10.1038/srep32771 non-specific binding, the antibody cocktail was added to the arrays. After additional washes, the arrays were incubated with the streptavidin-conjugated HiLyte Fluor 532 (Anaspec, Fremont, CA), and the fluorescent signal was visualized using a GenePix 4200A laser-based scanner system (Molecular Dynamics, Sunnyvale, CA) on the green channel. Two replicates were spotted for each antibody, and the average median signal intensity for both spots (with the local background subtracted) was used to calculate protein level.
Oil Red O staining and quantification. Cultured cells were stained with Oil Red O as previously described 19 to evaluate adipogenesis. Briefly, cells were washed twice with phosphate-buffered saline (PBS) and fixed with 4% paraformaldehyde for 2 h at 4 °C. After two washes in PBS, cells were stained for 2 h in freshly diluted Oil Red O solution consisting of six parts Oil Red O stock solution (0.5% in isopropanol) and four parts water at 4 °C. The stain was removed from the cells with two PBS washes and cells were examined with an inverted microscope. To measure the quantification of lipid accumulation, Oil red O was eluted by adding 100% isopropanol and optical density was detected using a spectrophotometer at 520 nm.  (4) β -TCP loading LV-AMPKα 2 MC3T3-E1 cells (LV-AMPKα 2). A modified version of a previously described method was used 44,45 . All MC3T3-E1 cells were treated with rhBMP-2 at a final concentration of 100 ng/ml to induce osteogenesis for 3 days. Then approximately 2.0 × 10 6 cells were seeded on each β -TCP disk (ϕ 6 mm × H 2 mm, Bio-lu Biomaterials Company, Shanghai, China). After 12 hours, the cell-scaffold complex was implanted into the intramuscular pocket of the femur of 8-week-old nude mice (BALB/c, nu/nu; SIPPR-BK Laboratory Animal Co. Ltd, Shanghai, China). Eight weeks after implantation, samples were harvested, fixed in 4% paraformaldehyde for micro-CT. Then these complexes were decalcified, and embedded in paraffin. Thin sections (5 μ m) were stained with hematoxylin and eosin (H&E).