Cooperative interactions between bone forming osteoblasts and bone-resorbing osteoclasts have been a central focus of orthopaedic research for almost two decades, following the landmark hypothesis advanced by G. Rodan and T. J. Martin in 1981 (1) . A perplexing problem and major impediment to studying osteoclasts was the experimental observation that physical contact between mononuclear osteoclast precursor cells and stromal "fibroblasts" was required for the differentiation of functional multinucleated osteoclasts in bone marrow cultures. The solution to this problem, reviewed below, represents an exciting breakthrough with important applications to orthopaedic research and clinical practice.

The New Pathway for Osteoclast Formation: Discovery of the RANK / RANKL Complex

Osteoclast differentiation and activation was recently proven to be dependent on formation of the RANK/RANKL receptor/ligand complex (Table 1, Fig. 1) (2) . RANK (Receptor Activator of NF-KappaB) was discovered in 1997 in dendritic cells (antigen-presenting cells of the immune system), and was shown to be critical for their survival during interaction with T-helper cells (3) . The novel ligand that binds to RANK was cloned and named RANKL ("RANK Ligand") by the Immunex research group (3) and TRANCE by another lab (4) . Independently, osteoclast research teams lead by T. Suda and H. Yasuda cloned osteoclast differentiation factor (ODF) and its receptor (ODFR), which proved to be identical to RANKL and RANK, respective-ly (Table 1) (2) . RANK is a type II membrane protein in the tumor necrosis factor receptor (TNFR) family, and is related to CD40, Fas, TNFR1 (p55), TNFR2 (p75), OX40, and to OPG/OCIF, the novel bone-protecting decoy receptor for RANKL discussed below. The TNFR family operates on a highly specific one receptor-one ligand principle, and RANKL apparently serves as the obligate ligand for RANK (3) . RANKL is homologous to other TNFR family ligands (TNF , FasL, CD40L, TRAIL), some trig-gering and others protecting against apoptosis, depending on the recruitment of death domain proteins and TNFR-associat-ed factors (TRAF) in the target cell.

	Linkage of RANK-RANKL signaling to osteoclast formation was established by co-culture studies of spleen-derived osteoclast precursors with ST2 stromal cells. To activate the osteoclast program of differentiation, maturation, and bone resorption, the membrane-bound receptor RANK must be expressed, and the receptor must be occupied by its ligand RANKL throughout the pathway from mononuclear precursor to mature osteoclast (2) . Without RANK-RANKL signaling, osteoclasts may cease their bone remodeling and undergo apopto-sis. RANKL, the essential ligand for osteoclastogenesis, is anchored in the surface membranes of stromal cells (and osteoblasts), and is expressed strongly in response to osteoly-sis- stimulating agents including 1,25(OH)2 vitamin D 3 , PTH (and PTH-rp), IL-6, and IL-11 acting through their cognate receptors (2) . Thus, the mysterious requirement for cell-cell contact in osteoclast formation is now explained mechanistically by the requirement for engagement of two membrane-anchored proteins RANK and RANKL (see Fig. 1). Osteoprotegerin (OPG): Decoy Receptor for RANKL which Blocks Osteoclast Formation OPG is a fortuitous discovery from Amgen Corporation's expressed sequence tag research program (5) . OPG is a non-signaling member of the TNFR family, related to Fas and TNFR1. OPG is a secreted glycoprotein with very high expression levels in developing bone. OPG lacks a trans-membrane domain but has two homologous death domain sequences. Its only known function is to act as a decoy receptor for RANKL, binding to membrane-anchored RANKL with a K d of 289pM (6) . Thus, when OPG is overexpressed in transgenic mice, it protects against osteoclastic resorption, and protects against bone loss in ovariectomized (estrogen-deficient) animals (5) . Presumably, OPG ties up RANKL and prevents binding to its cognate receptor RANK on osteoclasts (7) . OPG specifically inhibits osteoclas-togenesis in vitro elicited by 3 distinct effectors: 1,25(OH)2 vit-amin D 3 , PTH, and IL-11) (5, 6) . All 3 agents are now known to signal to a common pathway which upregulates RANKL expression on the stromal cell component of the stromal cell/preosteoclast co-culture model (2) . PTH has been shown to reciprocally induce RANKL expression while inhibiting OPG expression (8) . Clinical and animal trials of OPG are beginning to show striking inhibitory effects on osteoclast activity.
Fig.1 Osteoclast Development and the RANK / RANKL / OPG Signaling Pathway
Pre-osteoclasts arise from stem cells of the monocyte/macrophage lineage, requiring macrophage colony stimulating factor (M-CSF) for proliferation. In order to mature, survive, and actively resorb bone as a mature osteoclast, the pre-osteoclast must express the receptor RANK, and this receptor must engage its ligand RANKL. This critical ligand is expressed on the surface membranes of stromal cells and osteoblasts, under the influence of bone resorption-stimulating agents including 1,25(OH)2 vitamin D3 , PTH (and PTH-rp), IL-6, and IL-11 acting through their cognate receptors on target cells in the stromal/osteoblastic lineage. Cell-cell contact allowing RANK-RANKL signaling is therefore normally required for osteolysis. A soluble decoy receptor for RANKL, named osteoprotegerin (OPG), is produced by various cell types. OPG inhibits the osteoclast-mediated osteolytic pathway (see text for details).

Rationale for Studying the RANK / RANKL / OPG System in Skeletal Disease

This pathway for osteoclast development provides a con-vergent, unified mechanism for understanding the regulatory actions of many disparate agents (1,25(OH)2 -vitaminD (3) , PTH/PTHrp, IL-6, IL-11, M-CSF, and OPG) in skeletal homeostasis, and pathology (e.g., metabolic bone disease, inflammatory bone destruction, and osteolytic metastasis). Novel interpretations of earlier studies on skeletal metastasis are now possible in light of the hypothesis that we and others are testing: namely that the RANK / RANKL / OPG pathway is utilized by breast adenocarcinoma to establish osteolytic sites. A recent study suggests that breast cancer-derived PTHrp, acting on osteoblasts to upregulate RANKL expression, is a major stimulus to osteoclast formation in situ (9) . Future therapies for bone metastases will undoubtedly target this new class of tumor-bone interactions and the regulatory mechanisms involving RANK, RANKL, and OPG.

Model Systems for Osteoclast Study

With the availability of recombinant soluble forms of RANKL that can act on RANK without being tethered to a membrane (3) , osteoclasts can now be formed efficiently in vitro. Our attention has recently focused on the mouse macrophage cell line RAW264.7 that expresses RANK and responds strongly to mRANKL, forming TRAP+ multinucleated osteoclasts within 4 days (7) . A newly developed resorbable hydroxyapatite film culture system allows us to study resorptive activity of individual osteoclasts. Importantly, the RAW osteoclasts can be transfected in pure culture for gene promoter studies and the possible discovery of novel osteoclast-specific transcription factors (see below).

Transcriptional Regulation in Osteoclasts

Maturation of osteoclast precursors involves the expression of several osteoclast enriched and osteoclast specific proteins: structural proteins, receptors (e.g. calcitonin receptor, RANK), signaling molecules (e.g. Src), enzymes (e.g. cathep-sin- K, tartrate-resistant acid phosphatase, osteoclast proton-pump components, and carbonic anhydrase), membrane channels (e.g. Cl - channel and Cl - /HCO 3 - anion exchanger), and adhesion molecules (e.g. á V â 3 integrin) (10) .

The integrin á V â 3 is a member of a large family of heterodimeric, trans-membrane, cell-matrix attachment proteins which recognize specific sequences in a variety of extracellular matrix proteins. Integrins not only mediate cell-matrix attachment but also transmit bi-directional signals across the plasma membrane (11) . Integrin adhesion and signal transduction con-tribute to a wide range of cellular activities including motility, spreading, differentiation, secretion of matrix proteins, matrix assembly, matrix mediated cell cycle progression, and apoptosis (11) .

Mature osteoclasts predominantly express the á V â 3 form of integrin, with levels estimated at 10 to the 7th molecules per osteoclast (12) . Pharmacological blockade of ligand binding, by the highly expressed á V â 3 osteoclast integrin, inhibits osteoclast adhesion to bone and to specific bone matrix proteins, while inhibiting the spreading of osteoclasts. Osteoclast functional activity, as measured in vitro by pit formation and resorption of radioactive bone fragments, is also inhibited by blockade of á V â 3 binding (13) . In addition, ovariectomy-induced bone loss is inhibited by blockade of á V â 3 in animal models of post-menopausal osteoporosis (14, 15) . The requirement for â 3 integrin expression in osteoclast formation and function in vivo was determined by investigating mice in whom the â 3 gene was genetically deleted. Mice that lack the â 3 gene, and therefore do not express the á V â 3 integrin, produce defective osteoclasts. Osteoclasts from â 3 null mice do not spread normally in vitro, do not form the characteristic actin ring structure, and have disorganized ruffled membranes in vivo. These defects, while countered by a 3-fold increase in osteoclast numbers, lead to increased bone mass and osteosclerosis, confirming the importance of á V â 3 for normal osteoclast function (16) .

Expression of the á V â 3 integrin heterodimer is regulated at the level of the â 3 integrin mRNA, with no significant changes in á V mRNA levels during hormone and cytokine induced á V â 3 expression in macrophages and osteoclast precursors (17-19) . The increase in â 3 integrin mRNA steady state level occurs through transcriptional up-regulation and not via increased message half-life. Importantly, the huge increase of á V â 3 expression during osteoclast formation results from a transcriptional increase in â 3 integrin mRNA levels (19) .

Little is known about the regulation of gene transcription during osteoclastogenesis, primarily because a tractable cell model system for osteoclast formation has not been available. Recent advances in our understanding of the signals and factors directing osteoclastogenesis led to the identification of such a model cell line, RAW264.7 (7 ). To study the regulation of â 3 gene expression in osteoclastogenesis, we generated â 3 gene promoter/luciferase (luc) gene reporter constructs. In tran-sient transfection with these constructs, expression of the light producing luc gene reporter is a direct measure of the â 3 gene promoter activity. By transient transfection of osteoclasto-genic cells with the â 3 promoter/luc reporter construct, and subsequent induction of osteoclast formation with RANKL, we have shown that a 1.1 kilobase fragment of the â 3 promoter directs an 7 to 11 fold increase in reporter expression [McHugh, unpublished]. Deletion analysis of the 1.1 kilobase promoter have further identified an approximately 65 base pair region containing sequences which are responsible for the bulk of promoter induction during osteoclast formation. Further deletion and mutation of the â 3 promoter, in luc reporter con-structs, will allow us to identify the DNA sequences and their cognate transcription factors that direct â 3 gene transcription in osteoclastogenesis.

The osteoclast transcriptional machinery of the â 3 gene is likely shared by several other osteoclast-specific genes. Therefore, in addition to significantly increasing our under-standing of the molecular and cellular events directing osteo-clast differentiation, the identification of osteoclast-specific transcriptional machinery may identify novel targets for anti-osteoclastogenic therapies.
		Fig. 2. Caveolae and coated pit structures in the osteoblast plas ma membrane. Transmission electron micrograph of a murine MC3T3-E1 osteoblast [50,000 X].

Caveolae and Pre-assembled Signal Transduction Complexes in Osteoblasts

Signal transduction generally begins with the binding of a ligand to a membrane-anchored receptor, initiating a cascade of intracellular events leading to protein phosphorylation, protein- protein interaction, gene regulation, and a variety of cellular changes or responses. Signaling events in pre-osteoclasts that lead to osteoclast maturation (see above) often originate from osteoblasts that are actively responding to hormones or cytokines with their own specific set of signal transduction pathways. Thus it is essential to understand the signaling events in osteoblasts.

Recently, we demonstrated that osteoblasts have specialized pre-assembled signal transduction complexes and caveolae (Fig. 2), tiny membrane structures enriched with signaling molecules. Before describing the form and function of caveolae and pre-assembled signal transduction complexes in osteoblasts, we will address the origin of the membranes and nature of the lipids and proteins that form caveolae.

Small Detergent-resistant Membrane Domains (Detergent-resistant Microdomains)

Sphingolipids and cholesterol form detergent-resistant membrane microdomains (DRM) by self-aggregation during transport from the Golgi to the cell surface. The detergent insolubility of these aggregates stems from the properties of the lipids and their ability to form fluid vs. highly ordered membranes. DRM can be thought of as discrete lipid rafts in the general lipid milieu of the plasma membrane. These membrane patches, although rich in proteins, likely represent only 10-15% of the plasma membrane area. A good way to conceptualize these membrane microdomains is to visualize them as islands in the great lipid sea of the plasma membrane.

At least two varieties of DRM exist on many cell surfaces (20) : One variety of cell surface DRM appear by electron microscopy (EM) as "flat" areas of the cell membrane and are referred to as G domains. The second variety of DRM are called caveolae and contain caveolin as well as other components described below (20) . By EM caveolae appear as striated 50-100 nm membrane "invaginations" coated with oligmerized caveolin. While many cell types have both caveolae and G domains, some cell varieties do not express caveolin (i.e. haemopoietic cells) and may only have G domains (21) .

Caveolae

Caveolae are distinctive DRM that appear as small non-coated plasmalemmal vesicles (Fig 2). These organelles are present in adipocytes, myocytes, endothelial cells, chondrocytes, osteoblasts as well as other cell types, and are formed by the oligmerization of the protein caveolin (a 22 kD phosphoprotein that is the major structural component of caveolae) in membranes with the lipid composition of DRM. As outlined below caveolae are implicated in molecular transport processes and signal transduction (22-26) .

Caveolae in Molecular Transport: Caveolae are linked in a grape-like branching structure that penetrates deep within a cell, thus, conventional EM cannot be used to determine if caveolae are dynamic structures or remain attached to the cell membrane. Despite these problems, accumulating evidence makes it clear that caveolae are transport structures. This evi-dence includes the discovery that proteins involved in vesicular transport are found in caveolae (e.g. annexin, SNAP, NSF and VAMP-2) (26) .

Caveolae in Signal Transduction: Many receptors and signal transduction molecules have been localized to caveolae these include: PDGF receptors (23-25) , GTP-binding proteins (26, 27) , Src family tyrosine kinases, PI3-kinase, PLC , PKC á and â (23) . Caveolin itself has been shown to associate with a variety of signal transduction molecules including GTP-binding proteins (27) , and Src family protein kinases (27) . It has been suggested that caveolin is a negative regulator of signaling molecule activity and may act as a ‘scaffold' to which disparate signaling elements can attach in their inactive state (27) .

Given the abundance of signaling molecules and receptors found in caveolae and DRM it has been suggested that they are pre-assembled signal transduction complexes. Caveolae and other DRM are loci for signal molecule communication because they are loci for signal molecule accumulation. To emphasize the role of caveolae in cell signaling we will use an instructive example of a growth factor receptor (platelet derived growth factor receptor: PDGFR) and explore the evidence that caveolae regulate its signal transduction. The PDGFR is a particularly interesting example as it is expressed on bone cells and bone cells are responsive to PDGF.

PDGF receptors (PDGFR) in endothelial cells are associated with caveolae (23-25) , and require intact caveolae for maximum function (23) . In osteoblasts, PDGF treatment causes the phosphorylation of PDGFR in caveolin-enriched membranes and in the membrane at large (Solomon et al., 2000, submitted). It has also been shown that components of the PDGFR signaling cascade (e.g. Raf, Erk, Src) are present in caveolae of endothelial cells (23-25) , cardiomyocytes (28) , and osteoblasts (Solomon et al., 2000, submitted). PDGF treatment of endothelial cells leads to the activation of Erk in caveolae, and caveolae in isolation respond to PDGF by increasing tyrosine phosphorylation and activating Erk (25) . In osteoblasts, treatment of cells with PDGF causes translocation of Raf and Erk2 from the caveolin-enriched membranes to other areas of the cell (Solomon et al., 2000, submitted). Thus, caveolae are important regulators of PDGFR signaling in various cell types including osteoblasts.

What Value is this Information in Understanding Osteoblast Function?

How will the characterization of pre-assembled signaling complexes and caveolae provide fundamental new knowledge pertinent to the function of osteoblasts? How may this knowledge enable us to manipulate the function of osteoblasts for the development of new treatment protocols and improvement of current therapeutic approaches to bone disease? PDGF is an important growth factor for osteoblasts that regulates replication and gene expression, furthermore PDGF already is being used to stimulate bone repair in human clinical studies (29) . PDGFR is found in the pre-assembled signaling complexes of osteoblasts, endothelial cells, and myocytes, and receptor localization to these complexes is critical for function. Our recent observations regarding caveolae in osteoblasts reveal new targets for potentially enhancing the positive actions of PDGF while diminishing the negative ones. We may now address questions that previously we did not even know to ask: Are some of the multiple actions of PDGF on osteoblasts delivered by receptors regulated by their distribution into the plasma membrane at large vs. pre-assembled signaling complexes? Can we use this knowledge to promote desired clinical effects of PDGF while discouraging deleterious side-effects? Do the caveolae of osteoblasts regulate signaling from other growth factor receptors? Are the osteogenic effects of cholesterol-low-ering statin drugs (lovastatin, simvastatin) (30) explained by modulated signaling in cholesterol-rich caveolae? Answering these queries is part of our on-going and long-term research goals.

Summary

The discovery of new signaling pathways in osteoblasts and osteoclasts provide a window for understanding the complex interactions of these cells in bone. This new knowledge establishes an exciting foundation for future research into the pharmacological control of bone cells in health and disease.