Extensive circumstantial evidence suggests that cells of the osteoblast lineage are an important source of the RANKL that stimulates osteoclast formation. This evidence consists mainly of localization of RANKL-expressing cells in bone sections [16
] and detection of RANKL expression in cells isolated from rodent calvaria, which are rich in osteoblast progenitors [4
]. However, these localization studies have provided inconsistent results. For example, some studies detected RANKL in osteocytes while others did not. Moreover, calvarial preparations are heterogeneous and contain other cell types in addition to osteoblast precursors [30
], some of which may be sources of RANKL. More importantly, conditional ablation of osteocalcin-expressing cells in mice did not alter osteoclast number or function [32
], demonstrating that mature osteoblasts and their immediate precursors cannot be an essential source of RANKL in bone. Thus while osteoblast precursors may be an important source of RANKL, this has yet to be demonstrated experimentally. This uncertain relationship between osteoclast-support cells, defined by the expression of RANKL, and cells of the osteoblast lineage, defined here as cells that express relatively osteoblast-specific genes such as osteocalcin, has led to the use of vague terms to describe such support cells, such as stromal or stromal/osteoblastic, and it is the latter term that will be used here.
Parathyroid hormone (PTH) is one of the most potent stimulators of RANKL expression in stromal/osteoblastic cells [8
]. The concomitant reduction in RANKL levels and osteoclast number in rodents lacking PTH suggests that this regulation is biologically relevant [33
]. PTH stimulation of RANKL expression occurs primarily through activation of the protein kinase A (PKA) - cAMP pathway [35
] and requires the well-known target of this pathway, cAMP response element binding protein (CREB) [35
] (). CREB is also required for stimulation of RANKL by 1,25(OH)2
and the gp130 cytokine oncostatin M (OSM), suggesting that this transcription factor may play a central role in coordinating the actions of multiple signaling pathways [35
]. A potential CREB binding site has been identified 962 bp upstream from the transcription start site of the murine RANKL gene [38
]. However, additional studies have not detected significant stimulation of transcriptional reporter constructs containing up to 2 kb of the murine 5′-flanking region by PTH, 1,25(OH)2
or OSM [39
]. This latter set of results suggested that important transcriptional regulatory regions might reside outside the proximal 5′-flanking region of this gene.
Signals and transcription factors that control RANKL expression in stromal/osteoblastic cells
To identify potential distant regulatory regions, my laboratory developed an approach in which bacterial artificial chromosomes (BACs) are used to create transcriptional reporter constructs for the RANKL gene. BAC clones can harbor DNA fragments up to 200 kb in length and can be propagated in the same manner as plasmids. Because they contain such large segments of DNA, BAC-based reporter constructs allow regulatory studies to be performed with genes in a more native structure compared to traditional promoter-reporter constructs [42
]. In many cases, DNA fragments containing an entire gene, spanning tens of kb, can be obtained. Quite often these large fragments also contain the naturally occurring insulators that define the borders between genes [45
]. Thus these fragments frequently exhibit copy number-dependent, position-independent expression when used to generate stable cell lines or transgenic animals [42
]. In addition, they often contain the necessary regulatory elements to confer appropriate cell type-specificity and responsiveness to extracellular signals [43
The initial BAC-based RANKL reporter construct consisted of the full-length murine RANKL gene, as well as extensive 5′- and 3′-flanking region, in which the 3′-untranslated region was replaced with the luciferase coding sequence using recombineering techniques [44
]. In contrast to reporter constructs containing only the proximal RANKL 5′-flanking region, the resulting BAC-based construct was robustly stimulated by PTH, as well as by 1,25(OH)2
or OSM [41
]. The recombineering approach was then used to delete various regions within the BAC-based reporter construct and thereby localize the PTH-responsive region to a 2 kb fragment located 76 kb upstream from the transcription start-site. Sequences within this 2 kb fragment are highly conserved among mammals and contain a motif harboring two highly conserved cAMP-response elements (CREs) as well as a binding site for the osteoblast-specific transcription factor Runx2 [51
]. CREB and Runx2 were shown to bind these sites using gel mobility shift and chromatin immunoprecipitation (ChIP) assays [41
], suggesting that these sites are functional.
To determine the importance of this 2 kb enhancer for control of the endogenous RANKL gene, we have generated mice lacking this region. Deletion of the enhancer, designated the RANKL distal control region (DCR), increased bone mass in both the axial and appendicular skeleton [52
]. Loss of the enhancer also reduced PTH stimulation of RANKL mRNA in primary bone marrow cultures, as well as stimulation of RANKL mRNA in bone [52
]. DCR-deficient mice also had reduced basal RANKL mRNA levels in bone, thymus, and spleen. The increase in bone mass was due to reduced osteoclast and osteoblast formation leading to a low rate of bone remodeling, similar to that observed in humans and mice with hypoparathyroidism. These findings demonstrate that control of RANKL expression via the DCR is a critical determinant of the rate of bone remodeling. Subsequent studies indicate that mice lacking the DCR are also completely protected from the loss of cancellous bone that occurs with secondary hyperparathyroidism (C. Galli and C.A. O'Brien, unpublished results). Interestingly, cortical bone loss in this model is unaffected by deletion of the DCR, suggesting that different transcriptional enhancers of the RANKL gene mediate the response to secondary hyperparathyroidism in different skeletal compartments.
is another potent inducer of RANKL expression in stromal/osteoblastic cells [8
]. In addition to its effect on PTH action, deletion of the DCR also reduced the response to 1,25(OH)2
in vitro and in vivo [41
]. The DCR was independently identified as a potential 1,25(OH)2
-responsive region in studies by Pike and colleagues that utilized a combined ChIP-DNA microarray approach [53
]. These studies identified functional vitamin D response elements (VDREs) within the DCR as well as several additional VDREs within other conserved regions that lie 16, 22, 60, and 69 kb upstream from the transcription start site of the murine gene. Despite the high level of sequence conservation between species, conventional promoter-reporter studies suggest that in the human RANKL gene, one of these more proximal enhancers may play a dominant role in mediating the response to 1,25(OH)2
]. Several earlier studies had identified a potential VDRE located approximately 1 kb upstream from the transcription start site [55
]. However, as neither the BAC-based mapping approach nor the ChIP-DNA microarray approach identified this region as mediating effects of 1,25(OH)2
, the significance of this proximal site to the overall control of the RANKL gene remains to be determined.
Cytokines that utilize the gp130 signal transducer, also known as IL-6-type cytokines, stimulate RANKL in stromal/osteoblastic cells [8
]. Members of this family play roles in both normal and pathological bone resorption [59
]. Cytokines that utilize gp130 activate the signal transducers and activators of transcription (STAT) and mitogen-activated protein kinase (MAPK) pathways. Studies using dominant-negative proteins demonstrated that STAT3 is essential for stimulation of RANKL by gp130 cytokines and that blockade of gp130 or STAT3 blunted the response to IL-1, suggesting that IL-1 stimulates RANKL indirectly by stimulating the expression of IL-6-type cytokines [58
]. This latter finding may also be relevant to the actions of TNFα, as IL-1 has been shown to be required for the actions of TNFα on RANKL expression in stromal/osteoblastic cells [61
]. BAC-based reporter assays indicated that an important response element mediating the effects of gp130 cytokines lies within a 10 kb fragment beginning 82 kb upstream from the transcription start site [41
]; however, the sequences within this region that bind STAT3 have yet to be identified.
Compelling evidence indicates that canonical Wnt signaling stimulates OPG expression in bone and thereby suppresses bone resorption [62
]. A growing number of studies suggest that the same pathway may also be an important regulator of RANKL in stromal/osteoblastic cells. In vitro, canonical Wnt signaling or over-expression of β-catenin, a transcriptional co-factor that mediates the effects of the canonical Wnt pathway, suppressed RANKL expression, either under basal conditions or after stimulation by PTH or 1,25(OH)2
]. Conversely, deletion of β-catenin from calvarial cultures stimulated RANKL expression [63
]. The T cell factor (TCF) family of transcription factors function together with β-catenin to mediate the transcriptional effects of canonical Wnt signaling [67
]. Spencer et al. identified several potential TCF binding sites within the human RANKL 5′-flanking region that may mediate the effects of Wnt signaling and over-expression of β-catenin was able to suppress the activity of a murine RANKL promoter-reporter construct containing many of these sites [68
]. Several in vivo studies also suggest that canonical Wnt signaling suppresses RANKL expression. Administration of DKK1, which blunts canonical Wnt signaling, stimulated RANKL expression and osteoclast formation in bone [69
] and suppression of DKK1 in ovariectomized rats had the opposite effect [70
]. Moreover, in mice with a hypomorphic LRP6 allele, canonical Wnt signaling in bone is decreased and this is associated with increased bone resorption and RANKL expression [71
]. Taken together, these studies suggest that canonical Wnt signaling in bone exerts a tonic negative control of RANKL expression. The signaling pathways, transcription factors, and regulatory regions that control RANKL expression in stromal/osteoblastic cells are summarized in .
RANKL was originally identified as a TNF family member highly expressed in T lymphocytes [9
]. Shortly thereafter, several studies demonstrated that activated T and B cells can support osteoclast formation in vitro via expression of RANKL [77
], suggesting that these cell types may be important sources of RANKL during inflammation. Subsequently, studies in humans demonstrated that RANKL protein levels on B and T lymphocytes isolated from the bone marrow of post-menopausal women were elevated compared with pre-menopausal women or post-menopausal women treated with estrogen [80
], suggesting that lymphocyte expression of RANKL may also play an important role in the elevated bone resorption associated with loss of sex steroids. It should be noted, however, that there are as yet no studies demonstrating direct control of RANKL gene transcription by sex steroids in any cell type.
The first evidence that lymphocyte activation can play a role in stimulating osteoclast formation in vivo was obtained by examining mice lacking the gene cytotoxic T-lymphocyte-associated protein 4 (ctla4). Ctla4 is expressed on the surface of T cells and acts as a homeostatic suppressor of their activation and mice lacking this gene exhibit constitutive T cell activation [81
]. This increase in T cell activation was found to be associated with increased osteoclast number and reduced bone mass, which were reversed by administration of OPG [78
]. Based on this result, the authors of the study concluded that elevated expression of RANKL on activated T cells is sufficient to increase bone resorption. However, a more recent study has shown that ctla4 directly inhibits osteoclast differentiation in the presence of constant amounts of RANKL [82
], raising the possibility that the increased osteoclastogenesis in ctla4-deficient mice may not involve RANKL expression by activated T cells.
It is important to note that administration of OPG inhibits osteoclast formation and bone resorption in all in vivo models because osteoclast differentiation in vivo is absolutely dependent on RANKL expression [1
]. Therefore, blockade of bone resorption in a particular model by addition of exogenous OPG cannot reveal whether changes in RANKL expression are responsible for an increase in bone resorption. Moreover, elevation of many other cytokines such as M-CSF [83
], IL-1 [61
], and TNFα [84
] can synergize with constant levels of RANKL to increase osteoclast differentiation. Thus, the ability of OPG to reduce osteoclast numbers in ctla4-deficient mice does not reveal whether the increase in osteoclasts was due to elevated RANKL expression on T cells or any other cell type.
Although it is clear that activated T cells can stimulate osteoclast formation in vitro, a seminal study by Hiroshi Takayanagi demonstrated that isolated live T cells, activated by anti-CD3 antibody, completely inhibited osteoclast formation in vitro [85
]. The mechanism for this inhibition involved production of IFNγ by the T cells which promoted degradation of TRAF6, a scaffolding protein essential for osteoclast differentiation, in osteoclast precursors. Thus the ability of activated T cells to stimulate or inhibit osteoclast differentiation in vitro depends on the method of activation and type of culture system used. Perhaps more importantly, inflammation due to lipopolysaccharide injection, collagen-induced arthritis, or serum transfer-induced arthritis, can induce bone loss in T cell-deficient mice [86
]. Thus, T cells cannot be the sole source of RANKL in inflammatory bone loss. Nonetheless, depletion or genetic deletion of T cells significantly blunts bone loss in some murine models of inflammation [89
] or estrogen deficiency [90
]. However, in inflammatory models, lack of T cells also blunts the inflammatory response in general. Therefore, it remains unclear whether the role of T cells in inflammation-associated bone loss is to express RANKL, to promote the inflammatory response, or both.
At the transcriptional level, little is known about the control of RANKL expression in lymphocytes. Only a single study has examined the activity of RANKL promoter-reporter constructs in a T cell model [91
]. This study demonstrated that RANKL promoter-luciferase constructs, containing approximately 2 kb of the proximal 5′-flanking region, were stimulated by phorbol-12-myristate-13-acetate (PMA) and ionomycin in the human Jurkat T cell line. The combination of PMA and ionomycin treatment, which mimics activation of the T cell receptor, increased binding of NFκB and Egr family transcription factors to gel shift probes corresponding to putative sites in the RANKL proximal promoter region [91
]. However, the requirement of these binding sites for promoter activity was not determined.
The decrease in RANKL expression observed in the thymus and spleen of mice lacking the DCR enhancer may reflect reduced expression in T and B cells, respectively. If this is found to be the case, it will be important to determine whether the same transcription factor binding sites that mediate DCR activity in stromal/osteoblastic cells, such as the CREs, also play a role in lymphocytes.