The relationship between bone and fat formation within the bone marrow microenvironment is complex and remains an area of active investigation. Clinical and experimental findings suggest that acceleration of adipoge- nesis in bone marrow, known as fatty marrow, is associated with the progression of osteoporosis and aging [1] [2] . It is now clear that adipose tissue is a complex, essential and highly active metabolic and endocrine organ [3] [4] . Adipocytes express and secrete various endocrine hormones such as leptin, adiponectin (ApN), TNF-α and IL-6 [4] . Among these hormones, leptin is the first adipokine found to have function in both energy and bone metabolism [5] [6] . These observations suggest that energy metabolism and bone mass are regulated by the same hormones. Over the past decades, ApN and its receptors (AdipoR1 and AdipoR2) have been found to regulate energy homeostasis and have protective functions for metabolic and cardiovascular diseases [7] . In addition, accumulating evidences indicate that ApN measurement may serve as a useful screening tool for predicting osteoporosis [8] . ApN and its receptors are expressed in cells of osteoblastic and osteoclastic lineages, suggesting that they play roles in regulating bone metabolism [9] [10] . Here, we summarize the effects of ApN on bone metabolism and propose future research directions.

ApN was originally identified by four independent groups and was named Acrp30 [11] and AdipoQ [12] in mice, and apM1 [13] or GBP28 [14] in humans. It belongs to the complement 1q family [15] and forms multimer complexes through the collagen-like domain in the circulatory system. Multi-mer complexes include trimers, hexamers and high-molecular-weight (HMW) forms [16] . ApN-deficient mice exhibit features of insulin resistance, dyslipidemia and hypertension [17] . ApN exerts its insulin-sensitizing effect by regulating glucose utilization and fatty acid metabolism [7] . Administration of ApN to mice decreases the plasma concentration of glucose, free fatty acids and triglycerides, increases muscular fatty acid oxidation, induces weight loss [18] and reverses obesity-associated insulin resistance [19] .

Yamauchi et al. first cloned the cDNA encoding adiponectin receptors 1 (AdipoR1) and 2 (AdipoR2) from humans and mice [20] . Our group, then cloned the porcine counterparts [21] . AdipoR1 and AdipoR2 contain seven transmembrane domains with structure, topology and function distinct from those of the G-protein-coup- led receptors [20] . AdipoR1 and AdipoR2 serve as the major AdipoRs in vivo, with the former activating the AMP kinase (AMPK) pathway and the latter activating the peroxisome proliferator-activated receptor alpha (PPARα) pathway in liver to enhance insulin sensitivity and decrease inflammation [7] . Disruption of both AdipoR1 and R2 exterminate ApN binding and downstream actions including abolishment of ApN-induced AMPK activation and decreased activity of the PPAR-α signaling pathway; The result of disruption is to increase tissue triglyceride content, inflammation, oxidative stress, insulin resistance and glucose intolerance [22] . Recent research indicates ApN has anti-proliferative effects on cancer cells and a cardio-protective role increasing longevity [23] . Figure 1 represents the effects of adiponectin on peripheral tissues and hypoadiponectinemia related diseases.

Bone remodeling is a continuous process throughout adult life consisting of the resorption of senescent bone

by osteoclasts (Oc) and the formation of new bone from osteoblasts (Ob). Differentiated Oc are derived from hematopoietic stem cells [24] , whereas Ob are derived from mesenchymal stem cells [25] . In the process of bone remodeling, Oc adhere to bone to remove the exterior rigid structure by acidification and proteolytic digestion. Once the resorption site is created, Ob intrude and begin new bone formation along with the secretion of osteoid, a mineralizing factor. The process is completed by coverage of the bone surface by lining cells, a type of terminally differentiated Ob [26] . Bone disease is caused by an imbalance in bone remodeling. At the physiological level, mechanisms of bone remodeling are regulated locally by cytokines and systemically by hormones [27] . PTH and 1,5-dihydroxy vitamin D have anabolic actions, which are opposed by calcitonin [27] -[29] . Several systemic regulators, such as, insulin-like growth factor (also acting through local regulators) [30] , glucocorticoids [31] , thyroid hormone [32] and estrogen [33] contribute to bone formation and bone resorption. In addition, local factors derived from bone cells, like TGF-β, have inhibitory functions on osteoclastogenesis [34] regulate cell growth and differentiation and induce anabolic activity in Ob [35] . IL-1 and TNF-α serve to quiesce Oc and participate in bone resorption pathophysiology [36] . The receptor activator of NF-κB ligand (RANKL), RANK and osteoprotegrin (OPG) are essential for Oc differentiation. RANKL is expressed by Ob as a soluble factor that binds with RANK on Oc with recruitment of tumor necrosis factor receptor-associated factor 6 (TRAF6) and activation of the transcription factor, nuclear factor of activated T-cells cytoplasmic 1 (NFATc1) to induce of Ob differentiation related genes via the NF-κB pathway [37] [38] . OPG is a decoy receptor for RANKL which plays an inhibitory role in osteoclastogenesis [39] . OPG-deficient mice exhibit a decrease in total bone density and develope osteoporosis [40] . The regulation of bone remodeling not only involves osteoblastic and osteoclastic cell lineages but also other marrow cells. Adipocytes are derived from the same progenitor cells as Ob, and the equilibrium between the two cell types is important in bone remodeling [41] [42] . Several regulatory pathways of adipocyte differentiation from bone marrow mesenchymal stem cells interact with osteoblast differentiation pathway [43] . The major adipogenic transcriptional factors, C/EBPα and PPARγ, are regulated by extracellular signaling involved in osteoblastgenesis [43] . Mesenchymal stem cell differentiation into Ob requires Wnt signaling activation, which stimulates osteoblastogenesis and suppresses adipogenesis by blocking PPARγ and C/EBPα [44] [45] . Cell lineage allocation in the bone marrow microenvironment is shown in Figure 2.

Bone remodeling is an energy intensive process and bones need to balance energy constantly in response to nutrient availability during growth and bone turnover. During the last decade adenosine 5’-monophosphate-acti- vated protein kinase (AMPK) emerged as a key in the regulation of energy homeostasis and as the mediator of

central and peripheral effects of numerous hormones [53] . In addition, AMPK signaling activates ApN receptor signaling pathways in liver, muscle and adipose tissues [54] -[57] . In vitro studies demonstrate that AMPK modulates bone cell differentiation and function. AMPK is expressed in mouse tibia and ROS 17/2.8 osteoblastic cells [58] . Murine Ob and human mesenchymal stem cells treated with AMPK activators (AICAR or metformin) stimulate Thr-172 phosphorylation of AMPK in a dose/time-dependent manner. In contrast, treatment with the AMPK inhibitor (compound C) or knockdown of AMPK by shRNA-lentivirus infection inhibites AMPK phosphorylation and induces osteogenesis [58] [59] . Shah et al. (2010) analyzed the bone phenotype of 4 month-old male WT and AMPKα1−/−knockout mice by micro-CT. The AMPKα1−/−KO mice have less trabecular bone [58] . Furthermore, AMPK acts as a negative regulator in differentiation of Oc [60] . Treatment of pre-Oc with compound C (AMPK inhibitor) potentiates bone resorption and formation of TRAP-positive multinucleated cells in a dose-dependent manner [60] . Treatment with the globular form of ApN inhibits TNF-α/ RANKL-induced Oc formation from a RAW264 clone via AMPK signaling [61] . The elucidation of the importance of AMPK signaling in bone is still in early stage, but already reveals that AMPK activation affects bone formation and bone mass; therefore, AMPK signaling might act as a significant pathway in skeletal physiology.

ApN and its receptors are expressed in bone and bone stromal cells which suggest that they play critical roles in bone metabolism. The present review summarizes effects of ApN on bone metabolism, both in physiological states and in Ob or Oc culture in vitro. There are still several controversial results on the correlation of ApN and bone mineral density or other bone parameters. The critical role of ApN in the bone remodeling process and bone metabolism is an indisputable fact. Herein, we propose some possible future research directions.

1) In the circulatory system, ApN exists multi-mer complexes, such as trimers, hexamers and HMW forms [16] . Several researches suggest that different forms of ApN may give rise to different activities and have different roles in insulin sensitivity [62] [63] . In both mice and human diabetic patients, an increased ratio of HMW ApN to total ApN in the plasma correlates with improvement in insulin sensitivity during treatment with an insulin-sensitizing drug, TZD; there is no corelation with total amount of ApN [63] . Only one research group indicates that HMW ApN affects bone metabolism in both male and female hemodialysis patients [64] . We spe- culate that different multi-mer forms of ApN cause different influences on bone formation, but more molecular level observations are needed to confirm this speculation.

2) Energy equilibrium in mammals is controlled by the actions of circulatory hormones that coordinate fuel production and utilization in metabolically active tissues. Bone remodeling can be affected by metabolic related hormones, implying that bones are involved in the control of energy homeostasis [65] [66] . Osteocalcin, the most abundant non-collagenous protein of the bone extracellular matrix regulates β-cell proliferation, insulin secretion and glucose homeostasis [66] [67] . Furthermore, there are several consistent results showing the insulin-osteocalcin endocrine loop in humans [68] [69] were decreased in adipocyte and pancreatic islets respectively co-culture with osteocalcin knockout osteoblasts [65] . Whether osteocalcin has any interaction with ApN in bone metabolism or as a feed-back loop via ApN needs further investigation.

3) Mao et al. [70] used the cytoplasmic domain of AdipoR1 as bait to screen a yeast two-hybrid cDNA library derived from human brain. They found that APPL1 (adaptor protein containing pleckstrin homology domain, phosphotryosine binding domain (PTB) and leucine zipper motif) interacted with ApN receptors in mammalian cells and the interaction was stimulated by ApN. It has been suggested that APPL1 mediates the effect of ApN on inhibition of osteoclastogenesis and bone resorption [52] . Although this observation is an indirect evidence, we speculate that APPL1 may have functions to mediate ApN effects on bone development. APPL1 not only has an AdipoR1 binding domain in the PTB domain, but also has a FSH receptor binding domain [71] . The existence of FSH receptor domain in APPL1 increases the possibility of APPL1 mediating the regulation of ApN on bone metabolism [71] [72] . In order to investigate this possible regulator, we have to further investigate the role of APPL1 by culture system or loss-of-function and gain-of-function model.

4) Epigenic regulation—MicroRNAs (miRNAs) are small endogenous RNA fragments (19 - 25 nt) that regulate gene expression by targeting mRNA in post-transcriptional stage [73] . Controling the differentiation of the mesenchymal stem cells into osteogenic or adipogenic lineage is important for maintaining healthy bone and is necessary for prevention of bone related disease. Recently, there are some evidences showing that the differentiation between Ob and adipocytes is tightly controlled by miRNAs [74] [75] . We speculate that miRNA may have the effects on regulating expression of ApN and therefore affect bone formation or bone resorption.

5) AdipoR2 is expressed in bone-forming cells [10] . So far, there is no evidence showing the effect of ApN through AdipoR2 in bone formation or bone resorption. The role of AdipoR2 in bone metabolism must be investigated.

6) Bone marrow microenvironment and its endocrine or paracrine regulatory systems are actively studied. Osteoporosis models such as ovariectomy and senescence have been used extensively in rodents or other mammals [76] [77] . Ovariectomy mimics postmenopausal women whose bone loss is caused from acute ovary hormonal deficiency. Ovariectomy destroys the overall hormonal balance, but is a difficult model to investigate changes in bone microenvironments. More appropriate models are needed to investigate regulatory mechanisms.

The study was supported by grants from the National Science Counsel in Taiwan.