Canonical Wnt Signaling in Osteoblasts Is Required for Osteoclast Differentiation

Transcription

1 Canonical Wnt Signaling in Osteoblasts Is Required for Osteoclast Differentiation DONALD A. GLASS, II AND GERARD KARSENTY Department of Molecular and Human Genetics, Bone Disease Program of Texas, Baylor College of Medicine, Houston, Texas 77030, USA ABSTRACT: Inactivation of Lrp5, a gene encoding a likely Wnt coreceptor, results in low bone mass (osteopenia) by decreasing bone formation, suggesting that Wnt signaling in osteoblasts regulates bone formation. Here we show that Tcf1 and Tcf4 are expressed in osteoblasts during development and after birth; stabilization of -catenin, an essential component of canonical Wnt signaling, in differentiated osteoblasts results in high bone mass while its deletion from differentiated osteoblasts leads to osteopenia. Histological analysis showed that these mutations affect bone resorption. Cellular and molecular studies showed that -catenin together with TCF proteins regulates in osteoblasts the expression of Osteoprotegerin, a major inhibitor of osteoclast differentiation. These findings demonstrate that, in differentiated osteoblasts, -catenin and presumably Wnt signaling are negative regulators of osteoclast differentiation; thus they broaden our knowledge about functions that Wnt proteins may have at various stages of skeletogenesis. KEYWORDS: osteoblast; osteoclast; Wnt; -catenin; osteoprotegerin; skeletogenesis INTRODUCTION The osteoblast is defined by three major functions. First, it is responsible for the secretion of the bone extracellular matrix (ECM); second, it also expresses genes that are necessary and sufficient for the mineralization of this ECM; and third, the osteoblast is required for osteoclast differentiation and thereby for bone resorption. The two main genes required for osteoclast differentiation, M-Csf and Rankl, are expressed in osteoblasts. Osteoblasts also express a third gene that negatively regulates osteoclast differentiation. Osteoprotegerin (Opg) is a soluble tumor necrosis factor (TNF- ) receptor that functions as a decoy receptor for RANKL. However, the molecular mechanisms regulating the expression of Opg remain largely unidentified. Address for correspondence: Gerard Karsenty, Department of Molecular and Human Genetics, Bone Disease Program of Texas, Baylor College of Medicine, One Baylor Plaza, Houston, TX 77030, Voice: ; fax: karsenty@bcm.tmc.edu Ann. N.Y. Acad. Sci. 1068: (2006). C 2006 New York Academy of Sciences. doi: /annals

2 118 ANNALS NEW YORK ACADEMY OF SCIENCES The LDL receptor-related protein 5 (Lrp5) gene encodes the vertebrate homologue of arrow, which is a co-receptor for wingless, the Drosophila homologue of Wnt proteins. Loss-of-function mutations in Lrp5 in humans and in mice result in a low bone mass, associated with a decrease in bone formation, while gain-of-function mutations lead to high bone mass in humans. Forced expression of Lrp5 in NIH3T3 fibroblasts enhances canonical Wnt signaling; however, Wnt signaling can still occur in the absence of LRP5, suggesting that Lrp5 loss-of-function mutations may not represent a complete loss of canonical Wnt signaling. Collectively these data suggested, but did not establish, that canonical Wnt signaling regulates bone formation. The role of canonical Wnt signaling in a given cell type can be achieved by studying the function of intracellular proteins required for signaling. Unlike the multiple Wnts, Frizzled receptors, and LRP co-receptors, there is only one catenin molecule, which is the molecular node of the canonical Wnt signaling pathway. Thus studying its function is a powerful approach to discover the role of canonical Wnt signaling in a cell type of interest. To this end, we engineered mice harboring either gain-of-function or loss-offunction mutations in -catenin in differentiated osteoblasts. These mice had a high bone mass and low bone mass phenotype, respectively. These phenotypes, however, were primarily caused by altering bone resorption, and molecular analyses revealed that the canonical Wnt signaling pathway controls this process by regulating the expression of Opg in differentiated osteoblasts. These results establish the canonical Wnt signaling pathway in osteoblasts as a major inhibitor of bone resorption. OSTEOPETROSIS IN MICE EXPRESSING A STABILIZED FORM OF -CATENIN IN OSTEOBLASTS Mutations of serine and threonine residues within -catenin prevent its phosphorylation by glycogen synthase kinase 3 (GSK-3 ) and its subsequent degradation via ubiquitination, which in effect creates a constitutively active molecule. 6 Mice expressing a mutant allele of -catenin in which exon 3, the exon encoding the aforementioned serine and threonine residues, is flanked by loxp sequences had been generated previously (Catnb +/lox(ex3) ) 3 and were crossed with 1 (I) Collagen-Cre transgenic mice that express Cre in osteoblasts starting at E to generate 1 (I) Collagen-Cre; Catnb +/lox(ex3) mice [called hereafter cat(ex3) ob mice]. cat(ex3) ob mice were born at the expected Mendelian ratio, and appeared normal at birth and during the first 2 weeks of life; however, these mutant mice failed to thrive and died a few days after weaning. Physical examination of the cat(ex3) ob mice revealed that they lacked their lower incisors (FIG. 1A); this failure of tooth eruption was suggestive of a defect in osteoclast differentiation. 10 Histological analysis of 3-week-old cat(ex3) ob mice revealed in their vertebrae a massive increase in bone mass (FIG. 1B). The

3 GLASS & KARSENTY: CANONICAL WNT SIGNALING IN OSTEOBLASTS 119 FIGURE 1. Activation of -catenin in osteoblasts leads to osteopetrosis. (A) Lower incisors are absent in a 3-week-old cat(ex3)ob mouse. (B) Histomorphometric analysis of 3-week-old wild type (wt) and cat(ex3)ob vertebrae shows increased bone volume (BV/TV) in cat(ex3)ob mice (25 ). (C) TRAP staining in 3-week-old wt and cat(ex3)ob vertebrae. cat(ex3)ob osteoclasts are smaller and fewer in number (400 ). (D) Urinary deoxypyridinoline crosslinks (Dpd) elimination is decreased in 3-week-old cat(ex3)ob mice compared to wt. (E) Number of osteoblasts in 3-week-old wt and cat(ex3)ob vertebrae is similar. (F and G) Calvarial RNA microarray results identify many genes dysregulated greater than twofold in cat(ex3)ob mice and in Lrp5 / mice compared to wt. The asterisk ( ) indicates statistical significance. t-test value <0.05.

4 120 ANNALS NEW YORK ACADEMY OF SCIENCES

5 GLASS & KARSENTY: CANONICAL WNT SIGNALING IN OSTEOBLASTS 121 number of multinucleated osteoclasts (identified by positive Tartrate Resistant Acid Phosphatase [TRAP] staining) was decreased in cat(ex3) ob mice (FIG. 1C), as was the urinary excretion of the collagen degradation byproduct deoxypyridinoline (Dpd) (FIG. 1D). In contrast, the number of osteoblasts was normal (FIG. 1E). Thus, the analysis of the cat(ex3) ob mice revealed, unexpectedly, a defect in osteoclast differentiation leading to osteopetrosis. Opg AS A WNT SIGNALING TARGET GENE The unexpected differences between the bone phenotype of Lrp5 / mice (a low bone mass due to a defect in bone formation) and that observed in cat(ex3) ob mice (a high bone mass due to a failure of bone resorption) raised the hypothesis that the genes regulated by -catenin-dependent signaling may differ from those regulated by LRP5-dependent signaling. Microarray analysis of wt, Lrp5 /, and cat(ex3) ob osteoblast RNA was used to test this hypothesis. A total of 12,000 genes were analyzed between the mutant and wt osteoblasts. A number of genes were upregulated (more than twofold) in Lrp5 / compared to wt osteoblasts, and a number of genes were downregulated in cat(ex3) ob compared to wt osteoblasts (FIG. 1F). However, very few of these genes were common to both groups, that is, upregulated in Lrp5 / and downregulated in cat(ex3) ob osteoblasts. Similar results were obtained when looking for genes downregulated in Lrp5 / osteoblasts and/or upregulated in cat(ex3) ob osteoblasts (FIG. 1G). Furthermore, none of the genes common to both groups had an obvious function in osteoblasts. These data suggest that the -catenin mutation has a more drastic effect on bone mass regulation; for this reason we focused our analysis on the molecular bases of the osteopetrosis of the cat(ex3) ob mice. A review of the microarray data identified only one gene regulating osteoclast differentiation whose expression was noticeably increased (3.2 fold) in cat(ex3) ob mice. This gene, Opg, encodes a soluble TNF- receptor that inhibits osteoclast differentiation by functioning as a decoy receptor for RANKL, an osteoclast differentiation factor. 8,13 Opg expression was markedly FIGURE 2. Opg expression is increased in cat(ex3) ob osteoblasts. (A) In situ hybridization for Opg (a f) at E14.5 (a, b), E16.5 (c, d), and E18.5 (e, f) on wt (a, c, e) and cat(ex3) ob (b, d, f) forelimbs. Opg expression is increased in cat(ex3) ob forelimbs (50 ). (B) Northern blot analysis of wt and cat(ex3) ob long bone and osteoblast RNA. Blots were sequentially probed for Opg, Rankl, and Gapdh as a control. Note the increased Opg expression in cat(ex3) ob cells. (C) ELISA analysis for OPG in serum from wt and cat(ex3) ob mice. Note the increased levels of OPG in cat(ex3) ob mice. (D) Northern blot analysis of wt and cat(ex3) ob long bone RNA. Blots were sequentially probed for Runx2, Osterix, ATF4, Fra1, and FosB. No detectable change in expression levels of these genes is observed. The asterisk ( ) indicates statistical significance. t-test value <0.05.

6 122 ANNALS NEW YORK ACADEMY OF SCIENCES

7 GLASS & KARSENTY: CANONICAL WNT SIGNALING IN OSTEOBLASTS 123 increased in cat(ex3) ob developing long bones (FIG. 2Aa Af) and Northern blot analysis using 1-week-old bones or primary osteoblast RNA showed that Opg overexpression in cat(ex3) ob bones persisted beyond birth (FIG. 2B). Consequently, higher levels of OPG were detected in cat(ex3) ob serum than in wt serum (FIG. 2C). In contrast to the expression of Opg, the expression of transcription factors implicated in osteoblast differentiation was unchanged between wt and cat(ex3) ob osteoblasts (FIG. 2D). LOW BONE MASS AND HIGH BONE RESORPTION IN MICE LACKING -CATENIN IN OSTEOBLASTS To more accurately determine the biological function(s) of -catenin in osteoblasts we utilized a loss-of-function approach. Mice in which loxp sites had been introduced 5 of exon 2 and 3 of exon 6 of -catenin had been generated previously to achieve cell-specific inactivation of -catenin. 1 These mutant mice were crossed with 1 (I) Collagen-Cre mice to generate 1 (I) Collagen-Cre; -catenin floxed/floxed mice (hereafter called cat ob mice). cat ob mice were born at the expected Mendelian ratio, had a normal lifespan, and showed no overt phenotypic abnormalities. Histological analysis of 1-month-old mutant mice revealed a low bone mass phenotype (FIG. 3A). A marked increase in the number of TRAP-positive multinucleated osteoclasts along with an increase in urinary Dpd elimination demonstrated an increase in bone resorption (FIG. 3B, C), while the number of osteoblasts was similar between wt and cat ob mice (FIG. 3D). Opg expression was decreased in cat ob bones and osteoblasts while Rankl expression was not (FIG. 3E). In situ hybridization analysis showed that the expression of molecular markers of osteoblast differentiation and function was similar in wt and cat ob developing long bones (FIG. 3Fa Fj). Thus, the molecular and histological analyses concur to establish that an important function of -catenin occurring in osteoblasts once they are differentiated is to control osteoclast differentiation by regulating Opg expression. FIGURE 3. Osteoporosis and decreased Opg expression in mice lacking -catenin in osteoblasts. (A) Histomorphometric analysis of 1-month-old wt and cat ob vertebrae shows decreased bone volume (BV/TV) in cat ob mice (25 ). (B) TRAP staining of 1-mo-old wt and cat ob vertebrae. cat ob osteoclasts are larger and greater in number (400 ). (C) Dpd elimination is increased in cat ob mice. (D) Number of osteoblasts in 1-mo-old wt and cat ob vertebrae is similar. (E) Northern blot analysis of wt and cat ob long bone and osteoblast RNA. Blots were sequentially probed for Opg, Rankl, and Gapdh as a control. Opg expression is decreased in cat ob cells. (F) In situ hybridization for Runx2 (a, b), Osterix (c, d), 1 (I) Collagen (e, f), 2 (I) Collagen (g, h), and Osteocalcin (i, j) at E16.5 (a j) on forelimbs of wt (a, c, e, g, i) and cat ob (b, d, f, h, j) embryos. The asterisk ( ) indicates statistical significance. t-test value <0.05.

8 124 ANNALS NEW YORK ACADEMY OF SCIENCES OSTEOBLAST-AUTONOMOUS NATURE OF THE OSTEOCLAST DIFFERENTIATION DEFECT IN -CATENIN MUTANT MICE To determine if the bone resorption phenotype in these mutant mice was osteoblast autonomous we performed co-culture experiments using wt, cat ob, or cat(ex3) ob osteoblasts to trigger osteoclast differentiation of bone marrow macrophages (BMMs) of various genotypes. 9 Co-culture of wt osteoblasts with wt or mutant BMMs generated a similar number of TRAP-positive multinucleated osteoclasts (FIG.4A,B), indicating that the defect in osteoclast differentiation in the cat(ex3) ob and cat ob mice did not originate in cells of the osteoclast lineage. Co-culture of cat ob osteoblasts with wt BMMs generated a greater number of TRAP-positive multinucleated osteoclasts (FIG. 4A) while co-culture of cat(ex3) ob osteoblasts with wt BMMs failed to generate any osteoclasts (FIG. 4B); this latter defect could be rescued by adding a neutralizing antibody against OPG to the co-culture (FIG. 4C). These data establish that the origin of the osteoclast differentiation defect observed in both -catenin mutant mice resides in osteoblasts. EXPRESSION OF CANONICAL WNT SIGNALING PATHWAY COMPONENTS IN OSTEOBLASTS If -catenin controls Opg expression by signaling through the canonical Wnt signaling pathway, then one or more of the Lef/Tcf transcription factor family members must be present in osteoblasts to interact with -catenin. To identify which Lef/Tcf transcription factors were expressed in cells of the osteoblast lineage we performed in situ hybridization on wt embryos at various stages of development and compared their expression to that of Runx2 and 1 (I) Collagen, two osteoblast-specific markers. 4 Of the four family members, only Tcf1 and Tcf4 were detected in Runx2-expressing and 1 (I) Collagenexpressing cells: Tcf1 expression was detected as early as E14.5 while Tcf4 expression became detectable at E16.5. Tcf1 has various isoforms; some contain a -catenin interacting domain (BID) and act as transcriptional activators while others lack this domain and serve as repressors. 11 To assess whether the Tcf1 transcripts present in osteoblasts could belong to the former category we performed Real-Time PCR (polymerase chain reaction) using osteoblast and thymus RNA and appropriate primers and probes. The percentage of BID-containing Tcf1 transcripts is 40-fold higher in osteoblasts than in thymus (FIG. 5A), suggesting that TCF-1 proteins in osteoblasts can activate transcription. Thus, at least two members of the Lef/Tcf family of transcription factors are expressed in osteoblasts during development and after birth, and can serve as transcriptional activators of the canonical Wnt signaling pathway.

9 GLASS & KARSENTY: CANONICAL WNT SIGNALING IN OSTEOBLASTS 125 FIGURE 4. The bone resorption defect in -catenin mutant mice is osteoblast autonomous. (A) TRAP (+) cells were counted following coculture of wt or catob BMMs with either wt or catob osteoblasts. catob osteoblasts increased osteoclast differentiation regardless of the BMM genotype. (B) TRAP (+) cell number following co-culture of wt or cat(ex3)ob BMMs with either wt or cat(ex3)ob osteoblasts. cat(ex3)ob osteoblasts decreased osteoclast differentiation regardless of the BMM genotype. (C) TRAP (+) cell number following co-culture of wt BMMs with either wt or cat(ex3)ob osteoblasts, with or without an anti-opg antibody. cat(ex3)ob osteoblasts decreased osteoclast differentiation of wt BMMs while the anti-opg antibody prevented this effect. The ( ) asterisk indicates statistical significance. t-test value <0.05.

10 126 ANNALS NEW YORK ACADEMY OF SCIENCES

11 GLASS & KARSENTY: CANONICAL WNT SIGNALING IN OSTEOBLASTS 127 To confirm that the bone phenotype observed in the -catenin mutant mice was due to alterations in the canonical Wnt signaling pathway we analyzed mice lacking Tcf1, one of the two Tcf genes expressed in osteoblasts. Tcf1 / mice have a normal life span and have no apparent abnormalities, 12,7 yet bone histological analysis revealed a low bone mass in 1-month-old Tcf1 / mice (FIG. 5B). As with the cat ob mice there was an increase in the number of TRAP-positive multinucleated osteoclasts and in urinary elimination of Dpd (FIG. 5C, D) while the number of osteoblasts was normal. Northern blot analysis confirmed that, similar to cat ob osteoblasts, the expression of Opg was decreased in Tcf1 / osteoblasts (FIG. 5E). REGULATION OF Opg EXPRESSION BY -CATENIN AND TCF PROTEINS The phenotypic abnormalities present in cat(ex3) ob, cat ob, and Tcf1 / mice and the changes in Opg expression in these mice suggest that Opg might be a direct target of TCF-1 and/or TCF-4 in osteoblasts. Consistent with this hypothesis, there are three potential TCF-binding sites within the Opg promotor (FIG. 6A). The respective importance of these sites was first tested by electric mobility shift assay (EMSA) using osteoblast nuclear extracts and labeled double stranded oligonucleotides, containing the sequences of each of these three sites, as probes. Of the three possible sites, a protein-dna complex only formed upon incubation of osteoblast nuclear extracts with site 1 oligonucleotides (FIG. 6B). To verify this finding in vivo, chromatin immunoprecipitation analysis using wt primary osteoblasts was performed. Antibodies against TCF-1, TCF-4, or -catenin could immunoprecipitate site 1 of the Opg promotor but could not immunoprecipitate the first exon of Opg or the Osteocalcin promotor (FIG. 6C). To assess the relative importance of sites 1, 2 and 3 in determining Opg promotor activity DNA co-transfection experiments in COS cells, which do FIGURE 5. Regulation of osteoclast differentiation by -catenin requires TCF proteins. (A) (left) Schematic representation of activating (top) and repressive (bottom) isoforms of Tcf1. The location of the BID and Real-Time PCR primers for detection of Tcf1 isoforms (Tcf1 ACTIV and Tcf1 ALL, respectively) are indicated. (right) Real-Time PCR analysis for Tcf1 ACTIV and Tcf1 ALL expression in osteoblasts and thymus, normalized to 18S. Activating isoforms of Tcf1 are more abundant in osteoblasts than in thymus. (B) Histomorphometric analysis of 1-month-old wt and Tcf1 / vertebrae shows decreased bone volume (BV/TV) in Tcf1 / mice (25 ). (C) TRAP staining of 1-month-old wt and Tcf1 / vertebrae. Osteoclast number is increased in Tcf1 / mice (400 ). (D) Dpd elimination is increased in Tcf1 / mice. (E) Northern blot analysis of wt and Tcf1 / osteoblast RNA. Blots were sequentially probed for Opg, Rankl, and Gapdh. Note the decreased Opg expression in Tcf1 / cells. The asterisk ( ) indicates statistical significance. t-test value <0.05.

12 128 ANNALS NEW YORK ACADEMY OF SCIENCES FIGURE 6. Opg expression is regulated by -catenin and TCF proteins. (A) Schematic representation of the proximal 3.6 kb of the Opg promotor with the three putative canonical TCF-binding sites indicated by black boxes. (B) EMSA. Osteoblast nuclear extracts (NE) were incubated with labeled Opg site 3 (lane 1), site 2 (lane 2), or site 1 oligonucleotides (lane 3). A protein-dna complex formed only upon incubation of NE with site 1. (C) Chromatin immunoprecipitation for Runx2, TCF-1, TCF-4, and -catenin from primary osteoblast cultures. Antibodies against the TCF proteins and -catenin can immunoprecipitate site 1 of the Opg promotor, but not the Osteocalcin promotor (OSE) or the Opg coding sequence (CDS). (D) COS7 cells were transiently transfected with popg 3.6-luc,pOpgm1-luc,pOpgm3-luc,orpOpgm2+3-luc reporter constructs with or without HA-Lef1 and/or -catenin activ. LEF-1 increased transactivational activity of popg 3.6-luc,pOpgm3-luc, and popgm2+3-luc, which was further increased with -catenin. In contrast, LEF-1 failed to activate popgm1-luc.

13 GLASS & KARSENTY: CANONICAL WNT SIGNALING IN OSTEOBLASTS 129 not express Tcf genes, were performed. We used as a reporter construct a vector containing 3.6 kb of the Opg promotor fused to the luciferase gene (popg 3.6- luc). An expression vector for LEF-1, used here as a prototype of LEF/TCF proteins, strongly transactivated popg 3.6-luc and this transactivation by Lef1 was further enhanced by co-transfection with a -catenin expression vector. In control experiments, Lef1 alone, or in combination with -catenin, could not transactivate popgm 1 -luc, which has a mutation at site 1 of the Opg promotor. Moreover, the transactivational function of Lef1 was conserved when mutations were introduced in sites 2 and 3, further establishing the importance of site 1 in the regulation of Opg expression by TCF proteins (FIG. 6D). DISCUSSION Here we present evidence demonstrating that -catenin and presumably canonical Wnt signaling have a novel function in differentiated osteoblasts: the control of osteoclast differentiation. The evidence supporting this role includes: (1) the expression of two TCF proteins in osteoblasts that arise from intramembranous or endochondral ossification; (2) the decrease or increase in bone resorption caused by stabilizing or deleting -catenin in differentiated osteoblasts, respectively; (3) the increase in bone resorption detected in Tcf1 / mice; and (4) the identification of Opg, an inhibitor of osteoclast differentiation, as a target gene for -catenin and TCF proteins. This function of -catenin was revealed by stabilizing or deleting it from osteoblasts once differentiation has occurred through the use of 1 (I) Collagen-Cre mice, which do not express Cre in osteoblasts before E Why does the loss of -catenin in differentiated osteoblasts not result in the same phenotypic and molecular abnormalities in bone cells as the loss of Lrp5? One possibility is that the lack of -catenin has a more dramatic effect on osteoblast function than the lack of LRP5, which is only one of two likely co-receptors for Wnt proteins present in osteoblasts. LRP6, another protein with homology to Arrow, is also expressed in osteoblasts, and hypomorphic mutations in Lrp6 lead to a low bone mass in mice. 5 Thus it is possible that mice lacking both Lrp5 and Lrp6 in osteoblasts would display the bone resorption abnormalities characterizing the cat ob mice. Another possibility is that the ligand for LRP5 controlling bone formation may not be a Wnt protein. To address this possibility the identification of one or more Wnts whose deletion in differentiated osteoblasts reproduces the phenotype caused by the deletion of -catenin could be used to determine if they also act through LRP5. REFERENCES 1. BRAULT, V., R. MOORE, S. KUTSCH, et al Inactivation of the beta-catenin gene by Wnt1-Cre-mediated deletion results in dramatic brain malformation and failure of craniofacial development. Development 128(8):

14 130 ANNALS NEW YORK ACADEMY OF SCIENCES 2. DACQUIN, R., M. STARBUCK, T. SCHINKE, et al Mouse alpha1(i)-collagen promoter is the best known promoter to drive efficient Cre recombinase expression in osteoblast. Dev. Dyn. 224(2): HARADA, N., Y. TAMAI, T. ISHIKAWA, et al Intestinal polyposis in mice with a dominant stable mutation of the beta-catenin gene. EMBO J. 18(21): KARSENTY, G.& E.F. WAGNER Reaching a genetic and molecular understanding of skeletal development. Dev. Cell 2(4): KOKUBU, C., U. HEINZMANN, T. KOKUBU, et al Skeletal defects in ringelschwanz mutant mice reveal that Lrp6 is required for proper somitogenesis and osteogenesis. Development 131(21): LOGAN, C.Y.& R. NUSSE The Wnt signaling pathway in development and disease. Annu. Rev. Cell. Dev. Biol. 20: ROOSE, J., G. HULS, M. VAN BEEST, et al Synergy between tumor suppressor APC and the beta-catenin-tcf4 target Tcf1. Science 285(5435): SIMONET, S., D.L. LACEY, C.R. DUNSTAN, et al Osteoprotegerin: a novel secreted protein involved in the regulation of bone density. Cell 89: TAKAHASHI, N., T. AKATSU, N. UDAGAWA, et al Osteoblastic cells are involved in osteoclast formation. Endocrinology 123(5): TEITELBAUM, S.L. & F.P. ROSS Genetic regulation of osteoclast development and function. Nat. Rev. Genet. 4(8): VAN DE WETERING, M., J. CASTROP,V.KORINEK, et al Extensive alternative splicing and dual promoter usage generate Tcf-1 protein isoforms with differential transcription control properties. Mol. Cell Biol. 16(3): VERBEEK, S., D. IZON, F. HOFHUIS, et al An HMG-box-containing T-cell factor required for thymocyte differentiation. Nature 374(6517): YASUDA, H., N. SHIMA, N.NAKAGAWA, et al Identity of osteoclastogenesis inhibitory factor (OCIF) and osteoprotegerin (OPG): a mechanism by which OPG/OCIF inhibits osteoclastogenesis in vitro. Endocrinology 139: