Transcriptional program of mouse osteoclast differentiation governed by the macrophage colony-stimulating factor and the ligand for the receptor activator of NF κ B

Cytokines macrophage colony stimulating factor (M-CSF) and the receptor activator of NFkappaB ligand (RANKL) induce differentiation of bone marrow hematopoietic precursor cells into bone-resorbing osteoclasts without the requirement for stromal cells of mesenchymal origin. We used this recently described mouse cell system and oligonucleotide microarrays representing about 9,400 different genes to analyze gene expression in hematopoietic cells undergoing differentiation to osteoclasts. The ability of microarrays to detect the genes of interest was validated by showing expression and expected regulation of several osteoclast marker genes. In total 750 known transcripts were up-regulated by > or =2-fold, and 91% of them at an early time in culture, suggesting that almost the whole differentiation program is defined already in pre-osteoclasts. As expected, M-CSF alone induced the receptor for RANKL (RANK), but also, unexpectedly, other RANK/NFkappaB pathway components (TRAF2A, PI3-kinase, MEKK3, RIPK1), providing a molecular explanation for the synergy of M-CSF and RANKL. Furthermore, interleukins, interferons, and their receptors (IL-1alpha, IL-18, IFN-beta, IL-11Ralpha2, IL-6/11R gp130, IFNgammaR) were induced by M-CSF. Although interleukins are thought to regulate osteoclasts via modulation of M-CSF and RANKL expression in stromal cells, we showed that a mix of IL-1, IL-6, and IL-11 directly increased the activity of osteoclasts by 8.5-fold. RANKL induced about 70 novel target genes, including chemokines and growth factors (RANTES (regulated on activation, normal T cell expressed and secreted), PDGFalpha, IGF1), histamine, and alpha1A-adrenergic receptors, and three waves of distinct receptors, transcription factors, and signaling molecules. In conclusion, M-CSF induced genes necessary for a direct response to RANKL and interleukins, while RANKL directed a three-stage differentiation program and induced genes for interaction with osteoblasts and immune and nerve cells. Thus, global gene expression suggests a more dynamic role of osteoclasts in bone physiology than previously anticipated.


Introduction
Bone is a dynamic tissue that is constantly remodeled, i.e. degraded and renewed.
These two processes are accomplished by two main types of bone cells: bone-forming osteoblasts, of mesenchymal origin, and bone-resorbing osteoclasts, of hematopoietic origin (1). Understanding the generation and activation of these two cell types will help to unravel many processes involved in bone metabolism and remodeling. This knowledge may be used to develop novel, better treatments for osteoporosis, a disease prevalent in old age and characterized by bone loss and high risk of fractures.
Osteoclast generation (osteoclastogenesis) is a multi-step process that can be reproduced in vitro. The in vitro osteoclastogenesis systems used to comprise mixtures of stromal or osteoblastic cells together with osteoclast precursors from bone marrow (2,3). In such systems, stromal / osteoblastic cells are usually stimulated by 1α,25-dihydroxyvitamin D 3 to produce factors that support osteoclast formation.
More recently, autonomous mouse osteoclastogenesis systems have been developed using bone marrow cells cultured with soluble forms of the cytokines M-CSF (macrophage-colony stimulating factor) and a soluble form of RANKL (receptor activator of nuclear factor κB ligand) (4,5). These two cytokines are now recognized as the major factors contributed by stromal cells / osteoblasts for support of osteoclastogenesis (reviewed in 6). Therefore, their addition to the culture medium overcomes the need for stromal cells.
M-CSF, a homodimeric cytokine of the colony-stimulating factor family, is well known for its ability to stimulate proliferation and subsequent differentiation of cells of the macrophage / osteoclast lineage (7, reviewed in 8). The recently identified cytokine RANKL, also named ODF, OPGL or TRANCE (4,9), is a member of the family of tumor necrosis factor (TNF)-like transmembrane or soluble cytokines, which usually act as trimers. RANKL has been identified and characterized as a longsought factor stimulating osteoclast development (10). Expression of the receptor for M-CSF (c-Fms) is a crucial feature of an osteoclast precursor. The addition of M-CSF to osteoclast precursors induces the expression of the receptor for RANKL (RANK), which together with cell adherence allows differentiation to proceed (11,12). The main features of intracellular signaling by c-Fms and RANK have been elucidated: c-Fms is a transmembrane tyrosine-specific protein kinase receptor, whose intracellular signaling involves Src and MAP kinase Erk activation (13; 14). RANK is a member of the TNF receptor superfamily (15), and, like other family members, signals to the activation of the transcription factors NFκB and, through the kinase Jnk, c-Jun (16, reviewed in 6). Thus, the present knowledge on signaling by M-CSF and RANKL does not distinguish them from the other family members. The question remained open on how these two cytokines can drive such a complex and a specific process, such as the formation of strongly adherent, large, multinucleated bone-resorbing cells from a non-adherent, heterogeneous population of small bone marrow cells.
Therefore, to better understand the process of osteoclastogenesis at the molecular level, we have analyzed gene expression patterns induced by M-CSF and RANKL during mouse osteoclast differentiation, using oligonucleotide microarrays representing about 9,400 mouse genes. The patterns of gene induction and their identity provide a basis for further molecular examination of osteoclast formation.

In vitro mouse osteoclastogenesis
Cytokine-driven stromal cell-free mouse osteoclastogenesis was done as described by Shevde et al. (5). Briefly, bone marrow was collected from about twenty 5-weeks old male mice (C57BL, MA612, TIF/SPF). The cells were plated on 10-cm tissue culture dishes at 2x10 8  Aliquots of PCR products, supplemented with a loading buffer (final concentrations: 5% glycerol, 10 mM EDTA, 0.01% SDS, 0.025% xylene cyanol and bromophenol blue dyes), were fractionated on 8% polyacrylamide gels. The gels were vacuumdried, exposed to phosphor-storage screens (Molecular Dynamics) and imaged by

Analysis of gene expression on microarrays
The hybridization data were analyzed using the software provided by Affymetrix  To assess the sensitivity and accuracy of DNA microarrays for quantitative gene expression analysis, we first analyzed the microarray hybridization data obtained for the osteoclast markers and compared them to those obtained by RT-PCR (Fig. 1c).

Cellular and molecular characterization of primary mouse osteoclasts
The microarray analysis detected three out of seven osteoclast markers (c-fms, RANK and TRAP), which, consistent with the RT-PCR, were found to be up regulated.
Judging from threshold cycle numbers for detection by RT-PCR, these three markers were expressed more strongly than the others. The housekeeping genes were detected and found not to be up regulated, in agreement with the RT-PCR data. These results indicated that microarrays can detect the expression and regulation of osteoclast marker genes, albeit at a lower sensitivity than RT-PCR.
The expression of other genes characteristic for the osteoclast phenotype was further

Up regulated expression of 750 known genes is associated with osteoclastogenesis
In the rest of the present study, we analyzed the expression of known genes and not of uncharacterized "expressed sequence tags" (ESTs). In particular, we focused on up regulated genes, expecting that the contributors to the osteoclast differentiation process would be among them. As a threshold for expression regulation, we selected a 2-fold difference from the control, a criterion with statistical and biological meaning (details in Experimental Procedures).
We detected 750 known genes that were induced at least 2-fold by M-CSF and RANKL at days 1, 3, or 6, relative to day 0. This subset of genes was further analyzed by a nonhierarchical clustering method, defining 10 classes of genes, grouped according to their temporal patterns of regulation (Fig. 2). There was a good concordance between the general gene expression patterns observed in independent primary cultures (Fig. 2, compare A, B and C). Already one day after stimulation, 39% of the genes were induced (clusters I-IV, representing a total of 296 genes), either transiently (clusters I-III) or continuously (cluster IV). The remaining genes were induced either at day 3 (clusters V-IX, 391 genes, 52 %), or at day 6 (cluster X, 63 genes, 9%). These data indicate that the major transcriptional changes take place before day 3 (91% of genes), which is the time when the osteoclast phenotype becomes detectable in the cultures (TRAP staining, bone-resorbing activity). Further progression of osteoclast differentiation after day 3 involved up regulation of few new genes and already expressed genes were either preserved or down regulated (Fig. 2).
Thus, it appears that early commitment to the phenotype (attachment, cell

RANK and NFκB, interleukins, interferons and chemokines
We focused on genes up regulated with time in adherent cultures containing M-CSF, a survival and differentiation factor for the myeloid and osteoclast lineages. The induction of these genes was calculated relative to day 0, which comprised undifferentiated non-adherent bone marrow mononuclear cells (Fig. 3a, Fig. 3a), the other comprising molecules related to signaling by interleukins, interferons, and chemokines (Fig. 3b).
The importance of NFκB for osteoclast formation was previously best documented by the osteopetrotic phenotype (high bone mass due to dysfunctional osteoclasts) observed in knockout mice (24,25). RANK is the receptor for RANKL and its role in osteoclast differentiation and activation is well established in vitro, as well as in mice and in humans (15,(26)(27)(28). We have detected the induction of RANK, as previously reported (Fig. 1b,c; 11 (Fig. 4). Thus, in contrast to M-CSF-induced genes, RANKL-induced genes stem mainly from mRNA induction / stabilization in a given population of myeloid / osteoclast precursors. We identified 255 such RANKL-induced genes, among which 11 encoding cytokines, chemokines or growth factors, 13 receptors, 27 signaling molecules and 27 transcription factors and modulators. Genes in these four functional groups were further analyzed by a clustering algorithm and sorted according to the chronology of induction (Fig. 4a-d).
With the exception of the prostaglandin receptor EP2 (PTGER2), which is involved in bone resorption (45), for most of these receptors a link to osteoclast biology is not obvious at present. The most intriguing observation is the increased expression of two receptors for biogenic amines and neurotransmitters (H2R and ADRA1A, the receptors for histamine and epinephrine, respectively).
Among the RANKL-induced subset of signaling molecules (Fig. 4c) Within the group of transcription factors and their modulators (Fig. 4d), some genes can be connected to bone biology (ETL1, RELB and NFκB2). ETL1, a member of the SWI2/SNF2 family of helicases and nucleic-acid-dependent ATPases, has been shown, through gene inactivation in the mouse, to be important for normal bone growth (47). RELB encodes a transcription activator that participates to the NFκB transactivator complex (48). Since NFκB, a target of RANK signaling (49)

Confirmation of the gene expression profiles by quantitative PCR
As an independent confirmation of the gene expression patterns determined by microarray hybridization, we measured the expression levels of selected genes using SyBr Green-based fluorigenic real-time PCR or radioactive quantitative PCR. As exemplified for PDGF-α, Scya5/RANTES, H2R, CLOCK and CSRP, the expression profiles obtained by microarray analysis and quantitative PCR were very similar (Fig.   6).

Discussion
Our study provides a first comprehensive view of coordinated regulation of the its receptor CCR5, whose expression provides a basis for an autonomous regulation of osteoclast motility. Together, from these data a picture of osteoclast is emerging, in which this cell type is more independent, versatile and dynamic than currently appreciated.
With the aim of getting more insights into the mechanisms of action of RANKL, we searched for genes whose expression was more specifically regulated by this cytokine. Currently, there are only two known RANK target genes: c-Jun (5) and FosL1 (57). In this study we identified about 70 novel RANK-responsive genes. We observed that, among the genes induced by RANKL, a significant proportion functions in controlling transcription. Among these transcription factors, several have or may have a role in cell differentiation and would, therefore, deserve further investigation to determine their contribution to osteoclast differentiation. Another prominent group of RANKL-induced genes encodes signaling molecules, whose connection to osteoclast differentiation or activity is less obvious. However, transcriptional modulators and signaling molecules share a remarkable feature, their coordinated patterns of regulation over the time. Indeed, in both groups, most of the genes are not up regulated by RANKL at all time points studied (day 1, 3 and 6), but rather in waves, peaking at day 1, 3, or 6 ( Fig. 5). This observation does not only suggest that these genes may be involved in the control of different stages of osteoclastogenesis, but also allows the definition of at least three previously unidentified steps in the osteoclast differentiation process, based on a transcriptional fingerprint.
A series of chemokines, cytokines and growth factors genes are also induced by RANKL. Some of them may play a role in osteoclast differentiation or activation (e. implies a novel direct coupling between osteoclasts activation and osteoblasts recruitment. These factors were already known to be present in bone, but so far osteoclasts were not implicated as the cells producing them. A significant number of genes encoding receptors are also induced by RANKL. One of them, encoding the prostaglandin receptor EP2 (PTGER2), has a proven role in osteoclastogenesis (45). Again, as for pro-resorptive cytokines, prostaglandin E2 is thought to act mainly via stromal cells / osteoblasts, but not directly on osteoclasts.
Our identification of the up regulated prostaglandin receptor EP2 during osteoclast differentiation supports the unexpected notion that osteoclast can directly respond to prostaglandin E2.
Surprisingly, two receptors for biogenic monoamines were up regulated: the receptors for histamine (H2R) and the α1-adrenergic catecholamine receptor (ADRA1A).
Biogenic monoamines are neurotransmitters and vasoactive substances released by the neurons in the central and peripheral nervous system. Bone is a well-innervated tissue where the nerve fibers come in a direct contact with osteoclasts (58). In addition, human and mouse osteoclasts have been recently reported to express mRNAs for axon guidance molecules, such as semaphorin 3B, suggesting that they can guide growing nerve fibers (59). We have also detected the expression of semaphorin 3B (data not shown). The up regulation of biogenic monoamine receptors during osteoclastogenesis indicates that osteoclasts, in addition to modulating neurite outgrowth, also have the potential to respond to stimuli released by the neurons.
Histamine is also produced by the tissue mast cells, and the expression of its receptor in osteoclasts suggests that osteoclasts have a potential to respond to stimuli from the immune system as well. This regulation may have a pathophysiological and pharmacological relevance, since mast cell numbers strongly increase after ovariectomy, an in vivo situation of increased osteoclastogenesis and bone loss (60), and because histamine receptors antagonists can modulate osteoclast resorption in vivo (61).
In conclusion, by genome-wide identification of genes regulated during osteoclastogenesis, our study provides a basis for further molecular studies aiming at a better understanding of the osteoclast differentiation process. The identity of some of the genes described in this report sheds light on unsuspected potential of osteoclasts to regulate their own activity without mediation of stromal cells and to communicate with some other osseous and non-osseous cell types present in bone.
Together, our data suggest that the cells of the osteoclast lineage have a role in bone physiology that is more dynamic and versatile than it is currently viewed.   For the GeneChip microarray analysis, detection is indicated by symbols on a scale ranging from "-" (no significant signal) to "+++" (signal > 1000).