This Article Abstract Full Text (PDF) All Versions of this Article: 26/6/743    most recent Final Manuscript Author Manuscript Purchase Article View Shopping Cart Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Request Copyright Permission Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Janssens, K. Articles by Van Hul, W. Search for Related Content PubMed PubMed Citation Articles by Janssens, K. Articles by Van Hul, W.

Transforming Growth Factor-ß1 to the Bone

Department of Medical Genetics (K.J., W.V.H.), University of Antwerp, 2610 Antwerp, Belgium; Department of Molecular Cell Biology (P.t.D.), Leids Universitair Medisch Centrum, 2333 AL Leiden, The Netherlands; and Department of Biochemistry (S.J.), University of Lausanne, 1066 Epalinges-Lausanne, Switzerland

Correspondence: Address all correspondence and requests for reprints to: Professor Dr. Wim Van Hul, Department of Medical Genetics, University of Antwerp, Campus Drie Eiken, Building T6, Universiteitsplein 1, 2610 Antwerp, Belgium. E-mail: wim.vanhul{at}ua.ac.be

	Abstract

Top Abstract I. Introduction II. The TGF-ß... III. TGF-ß1 in Bone IV. Bone Phenotypes Associated... V. Therapeutic Use of... VI. Concluding Remarks References

 
TGF-ß1 is a ubiquitous growth factor that is implicated in the control of proliferation, migration, differentiation, and survival of many different cell types. It influences such diverse processes as embryogenesis, angiogenesis, inflammation, and wound healing. In skeletal tissue, TGF-ß1 plays a major role in development and maintenance, affecting both cartilage and bone metabolism, the latter being the subject of this review. Because it affects both cells of the osteoblast and osteoclast lineage, TGF-ß1 is one of the most important factors in the bone environment, helping to retain the balance between the dynamic processes of bone resorption and bone formation. Many seemingly contradictory reports have been published on the exact functioning of TGF-ß1 in the bone milieu. This review provides an overall picture of the bone-specific actions of TGF-ß1 and reconciles experimental discrepancies that have been reported for this multifunctional cytokine.

I. Introduction

A. Functions of TGF-ß1

B. Processing and storage of TGF-ß1

C. Activation mechanisms of TGF-ß1

II. The TGF-ß Signaling Pathway

A. General aspects

B. Smad-dependent signaling

C. Smad-independent signaling

III. TGF-ß1 in Bone

A. Introduction

B. TGF-ß isoforms in bone

C. Role of TGF-ß1 in osteoblastogenesis and bone formation in vitro

D. TGF-ß1 in osteoclast formation and bone resorption in vitro

E. Interaction of TGF-ß1 with other growth factors and hormones

IV. Bone Phenotypes Associated with Abnormal TGF-ß1 Signaling

A. Bone phenotypes of knockout and transgenic mouse models of the TGF-ß signaling pathway

B. TGFB1 mutations in the pathogenesis of Camurati-Engelmann disease (CED)

C. Osteolytic metastases: a role for TGF-ß1 in malignancy

D. Bone-related association studies

V. Therapeutic Use of TGF-ß1 as Bone-Forming Agent

VI. Concluding Remarks

	I. Introduction

Top Abstract I. Introduction II. The TGF-ß... III. TGF-ß1 in Bone IV. Bone Phenotypes Associated... V. Therapeutic Use of... VI. Concluding Remarks References

 
TGF-ß1 IS THE prototype of the TGF-ß superfamily, an evolutionary conserved family of structurally related dimeric cytokines with representatives in organisms as diverse as mammals and invertebrates. Its members share a cluster of conserved cysteine residues that form a cysteine knot structure held together by intramolecular disulfide bonds. Moreover, they all have the same precursor structure with a hydrophobic signal sequence, a prodomain, and a mature C-terminal domain (1). The superfamily includes TGF-ßs, bone morphogenetic proteins (BMPs), growth and differentiation factors, activins, inhibin, and anti-Mullerian hormone. All its members play important roles in the regulation of cell proliferation and differentiation and have pivotal functions during embryogenesis.

The TGF-ß family contains three closely related mammalian isoforms—TGF-ß1, -ß2, and -ß3—that arose by duplication of a common ancestor. Similarity is most striking in the C-terminal domain (64–82%), with nine conserved cysteine residues forming four intrachain and one interchain disulfide bond. Despite this high sequence homology, analysis of the in vivo functions of the three isoforms by gene knockouts revealed striking differences, illustrating their nonredundancy (see Section IV.A.). Overall, TGF-ß1 is the most abundant isoform with the largest sources of TGF-ß1 being platelets (20 mg/kg) (2) and bone (200 µg/kg) (3).

TGF-ß1 is a ubiquitous, multifunctional growth factor. TGF-ß was first identified as a factor that synergizes with TGF- to induce colony formation of normal rat kidney fibroblasts in soft agar, hence their name (4). Since its discovery, numerous other functions have been attributed to this cytokine, and several alternative names have been proposed, such as cartilage-inducing factor, differentiation-inhibiting factor, and tissue-derived growth inhibitor. However, none of these express the multitude of functions in which TGF-ß1 is involved.

A. Functions of TGF-ß1
TGF-ß1 regulates a broad range of biological processes, including cell proliferation, cell survival, cell differentiation, cell migration, and production of extracellular matrix (ECM) (for review, see Refs.5, 6, 7, 8). The combined actions of these cellular responses mediate the global effects of TGF-ß1 on immune responses, angiogenesis, wound healing, development, and bone formation (9, 10, 11, 12). Bone formation by TGF-ß1 is promoted through chemotactic attraction of osteoblasts, enhancement of osteoblast proliferation and the early stages of differentiation with production of ECM proteins, stimulation of type II collagen expression and proteoglycan synthesis by chondrocyte precursor cells, and suppression of hematopoietic precursor cell proliferation (see Section III). Regarding the diversity of processes in which TGF-ß1 is involved, it is not surprising that this cytokine is of major importance both during embryogenesis and in maintaining tissue homeostasis during life.

B. Processing and storage of TGF-ß1
TGF-ß1 is synthesized as a 390-amino acid protein (pre-pro-TGF-ß1) consisting of three distinct parts: the signal peptide (SP; 29 amino acids), the latency-associated peptide (LAP; 249 amino acids), and the mature peptide (112 amino acids). The pre-pro-TGF-ß1 monomer is extensively processed before its secretion (outlined in Fig. 1).

View larger version (50K): [in this window] [in a new window]   FIG. 1. TGF-ß1 processing. Pre-pro-TGF-ß1 undergoes extensive posttranslational processing. 1, The signal peptide, targeting the protein to the secretory pathway, is cleaved off during transit through the rough endoplasmatic reticulum (RER). 2, Two monomers dimerize by way of disulfide bridges between cysteine residues at positions 223 and 225 in the LAP and cysteine residue 356 in the mature peptide. 3, The protein is cleaved by furin convertase at the dibasic arginine residue at position 278. This yields the LAP and the mature peptide. Noncovalent bonds between them prevent the premature activation of the mature peptide, forming the SLC. 4, The SLC can become covalently attached to a LTBP to form the LLC. Binding occurs between the cysteine residue at position 33 of the LAP and the third 8 Cys-repeat of the LTBP. 5, After its secretion, the LLC is directed to the ECM and stored through binding of the LTBP with the ECM. The SLC is more readily available for activation. The SLC to LLC ratio depends on the cell type.

Although most cell types secrete TGF-ß as part of the LLC, bone cells form an exception as they efficiently secrete the SLC (see Section III.B). Consequently, the SLC is the predominant form in the bone environment. The remaining TGF-ß1 is bound to LTBP-1 or -3 (19, 20).

C. Activation mechanisms of TGF-ß1
Because members of the TGF-ß family are secreted as latent complexes, they need to be activated to exhibit their biological activity. In view of the ubiquitous expression of TGF-ß1 and its receptors and the multitude of processes in which TGF-ß1 is involved, it comes as no surprise that activation is tightly regulated. The goal of the activation process is the release of the epitope(s) on the mature peptide responsible for interaction with the receptor. Activation of the LLC initiates with its release from the ECM a process mediated by proteases (plasmin, thrombin, leukocyte elastase, mast cell chymase) that cleave the LTBP at a protease-sensitive hinge region and target the cleaved complex to the cell surface (21, 22). The truncated LLC and the SLC can be subjected to three different mechanisms of in vivo activation: 1) degradation of the LAP by proteases; 2) induction of a conformational change in the LAP by interaction with integrins and thrombospondin, for example; and 3) rupture of the noncovalent bonds between LAP and mature TGF-ß1. An overview of in vivo TGF-ß activation mechanisms has been presented by Annes et al. (23). The recent development of a genetic screen to discover new TGF-ß activators might enable the identification of other relevant activating mechanisms in vivo (24).

A unique activation mechanism has been proposed in bone, in which resorbing osteoclasts may activate TGF-ß1 in their acidic microenvironment (25). Acidification might break the noncovalent bonds between LAP and mature TGF-ß1, thus releasing the active peptide. However, others doubt that the mild acidic microenvironment created by osteoclasts is able to activate TGF-ß1 and suggest the activation to be a consequence of the release of proteases by these bone-resorbing cells (26).

	II. The TGF-ß Signaling Pathway

Top Abstract I. Introduction II. The TGF-ß... III. TGF-ß1 in Bone IV. Bone Phenotypes Associated... V. Therapeutic Use of... VI. Concluding Remarks References

 
A. General aspects
Once activated, TGF-ß can interact with its receptor to induce signaling. All members of the TGF-ß superfamily signal through a dual receptor system of type I and type II transmembrane serine/threonine kinases. These receptors belong to a family of glycoproteins characterized by a cysteine-rich extracellular region, a single transmembrane -helix, and a cytoplasmic domain with a kinase domain. In addition, the type I receptors, also termed activin receptor-like kinases (ALKs), share a highly conserved glycine- and serine-rich (GS) domain adjacent to the kinase domain, the GS domain. The type II receptors are characterized by their constitutively active kinase domain. Seven type I and five type II receptors transmitting signals from TGF-ß superfamily members are present in mammals (27). Despite the fact that more than 35 members form part of the TGF-ß superfamily, the combinations between the type I and type II receptors occurring under normal conditions are limited.

For members of the TGF-ß family, the TGF-ß type II receptor (TßRII) is the sole type II receptor shown to mediate signaling. This is reflected by the phenotypic identity of the tgfbr2 and those tgfb1 knockout mice that die in utero (see Section IV.A). Of the type I receptors, ALK5, ALK1, and possibly ALK2 can transmit TGF-ß signals. ALK5 (TßRI) is the most important type I receptor for TGF-ß, which is underscored by the comparable (although not identical) phenotypes of tgfb1 and alk5 knockout mice: histological examination of the yolk sacs of tgfbr1–/– embryos shows an image very similar to that of tgfb1–/– embryos that die during embryogenesis (312). In bone cells, TßRI seems to be the only type I receptor involved in signaling.

Betaglycan and endoglin are so-called type III or accessory receptors, which are indirectly involved in signaling through the modulation of ligand-binding specificity. Betaglycan (TßRIII) can bind all three TGF-ß isoforms and is implicated in the presentation of TGF-ß to TßRII (28). For TGF-ß2, which has a low intrinsic affinity for TßRII, both in vitro and in vivo data have demonstrated signaling to be dependent on presentation of this isoform by betaglycan (29, 30). However, it must be noted that a soluble form of betaglycan, shed by cells upon proteolysis in the juxtamembrane region, has a role in sequestering mature TGF-ß, thus inhibiting signaling (31). Moreover, upon TßRII-mediated phosphorylation of betaglycan, ß-arrestin-dependent internalization of the TßRII/betaglycan complex serves to down-regulate TGF-ß signaling (32). Endoglin can bind TGF-ß1 and -ß3 in the presence of TßRII (33). Mutations in ENG, the gene encoding endoglin, lie at the basis of the human disease hereditary hemorrhagic telangiectasia, an autosomal dominant disorder characterized by multisystem vascular dysplasia (34). A murine model of this disorder presents with a phenotype that is remarkably similar to that of tgfb1 and tgfbr2 knockout mice, suggesting an in vivo requirement for endoglin in TGF-ß1 signaling (35). Bone marrow stromal cells (BMSCs) and mature osteoblasts express the two types of type III receptors (36), whereas osteoclasts seem to lack betaglycan.

In the absence of ligand, both type I and type II receptors are present as homodimers. Upon TGF-ß1 binding to TßRII, TßRI can be recruited into a heterotetrameric TßRII/TßRI complex. Ligand-induced multimerization of the receptor complex is followed by transphosphorylation of the GS domain of TßRI by the constitutively phosphorylated TßRII kinase, resulting in activation of TßRI (37). This transphosphorylation is the first step in the intracellular transmission of the signal.

B. Smad-dependent signaling
Genetic studies in Drosophila melanogaster provided a breakthrough in our understanding of intracellular TGF-ß signaling through the identification of mothers against dpp (Mad). Its protein product plays a role in mediating the function of decapentaplegic (dpp), the D. melanogaster ortholog of BMP-2 or BMP-4 (38). This discovery was followed by the genetic identification of the homologous Sma genes in Caenorhabditis elegans (39) and subsequently the Smad genes (for Sma and Mad related) in vertebrates (reviewed in Ref.40). The Smads turned out to play a central role in the transmission of signals from all receptors activated by TGF-ß superfamily members to target genes in the nucleus.

Because several comprehensive reviews on Smad signaling by members of the TGF-ß superfamily have been published recently (37, 41, 42, 43), this signaling pathway will be discussed only briefly. Based on their structural and functional properties, Smads can be classified into three groups: receptor mediated (R)-Smads, common mediator (Co)-Smads and inhibitory (I)-Smads. R-Smads 2 and 3 are responsible for transmitting most signals from TGF-ß. The only Co-Smad identified so far in mammals is Smad4, which is commonly used by all TGF-ß superfamily members. The R- and Co-Smads share a similar structure with conserved amino- and carboxy-terminal domains, the Mad homology (MH)-1 and MH2 domains, connected by a more divergent linker region. In addition, the R-Smads contain a carboxy-terminal phosphorylation site, the SSXS motif. Lacking any recognizable enzyme activity, Smads achieve their signaling capacity mainly through protein-protein or DNA-protein interactions, exerted by the different domains. The MH1 domain can mediate direct DNA binding, whereas the MH2 domain is implicated in receptor interaction, Smad oligomerization, and transcriptional activation. Both domains further drive nuclear import and allow binding to various transcription factors and cofactors (see below). The divergent linker region contains multiple phosphorylation sites, allowing fine tuning of Smad functioning by many different signaling pathways in the cell, which converge on phosphorylation of this region. Furthermore, the linker region of R-Smads (with the exception of Smad8) contains the PY motif, which directs interaction of the Smad proteins with the E3 ubiquitin ligases of the Smurf or SCF families, targeting the protein for degradation (reviewed in Refs.44 and 45). The class of the I-Smads comprises Smad6 and -7. Smad6 is an inhibitor of BMP signaling, whereas Smad7 inhibits both TGF-ß/activin and BMP signaling. I-Smads share the MH2 domain with the R-Smads but show only weak similarity to the MH1 domain. Via its MH2 domain, Smad7 is recruited to the receptor complex, thereby mechanically blocking the access of R-Smads. Moreover, it can direct TßRI for ubiquitination and degradation through binding of Smurfs to its PY motif (44, 45). Finally, it acts as an adaptor protein in the formation of a protein phosphatase holoenzyme that targets TßRI for dephosphorylation (46).

In Fig. 2, the canonical Smad-dependent signaling pathway has been outlined. Upon binding of TGF-ß to its type II receptor and formation of the heterotetrameric type II/type I receptor complex, TßRII transphosphorylates and activates TßRI. Through their MH2 domain, R-Smads can bind the GS domain of TßRI, an interaction promoted by adaptor proteins such as Smad anchor for receptor activation (SARA). SARA interacts specifically with TßRI and functions to recruit Smad2 and Smad3 to the activated receptor complex, presumably in the endocytotic compartment (47, 48, 49). In the basal, unphosphorylated state, the MH1 and MH2 domains of the R-Smads inhibit each other reciprocally. Binding of the R-Smad to TßRI is followed by phosphorylation of the former at its carboxy-terminal SSXS motif by the TßRI kinase domain, which causes the R-Smad to dissociate from the receptor complex. This induces a conformational change that relieves reciprocal MH1-MH2 domain inhibition and promotes the formation of heteromeric complexes with variable stoichiometry with the Co-Smad4 (50, 51, 52, 53). The Smad complex translocates to the nucleus, where it can bind directly to DNA or recruit other DNA binding partners (transcription factors). The MH1 domains of phosphorylated Smad3 and Smad4 can bind to Smad-binding elements (SBEs) or GC-rich regions in the promoter of TGF-ß target genes. However, the interaction of Smads with DNA is of both low specificity and low affinity and is not sufficient to induce transcriptional activity. DNA binding partners that bind to recognition sequences in close proximity of the SBE site and to the MH1 or MH2 domains of the Smads are required. In addition, Smads recruit general cofactors with activating or repressive capacity (mostly obtained through interaction with histone acetylases or deacetylases, respectively), which further determine the outcome of the Smad-mediated gene transcriptional response. An extensive list of Smad binding partners and cofactors can be found in a recent review by Miyazawa et al. (54).

View larger version (54K): [in this window] [in a new window]   FIG. 2. Signaling by the TGF-ß family members through the Smad-dependent and MAPK-dependent pathways. A, After activation, TGF-ß can bind to its type II receptor, TßRII. B, Binding does not alter the phosphorylation state of TßRII, which is constitutively active, but induces the formation of a heterotetrameric receptor complex of TßRII and a type I receptor, in most cases TßRI (ALK5). C, TßRII activates TßRI by transphosphorylation of the GS domain. 1, Smad-dependent signaling. 1D, Activated TßRI can, in turn, phosphorylate one of the R-Smads at the C-terminal SSXS domain. Presentation of R-Smads to TßRI can be promoted through binding with SARA. 1E, Phosphorylation of the R-Smad relieves the reciprocal inhibition of its MH1 and MH2 domains and allows its interaction with a Co-Smad, forming a heteromeric complex. 1F, The R-Smad/Co-Smad complex translocates to the nucleus. 1G, Smads 3 and 4 are able to bind DNA through their MH1 domain but are unable to induce transcription independently. Instead, they modulate the transcription of diverse genes through their interaction with a set of corepressors, coactivators, and transcription factors. 2, MAPK-dependent signaling. 2D, The factors between the receptor complex and the MAPKKKs remain largely unknown. 2E,Through sequential phosphorylation of different MAPKKKs, MAPKKs, and MAPKs (ERK, JNK, and p38 MAPK), transcription factors of the ATF, Jun, and Fos families are activated. 2F, Transcription factors homo- and heterodimerize into AP-1 complexes, which can bind to AP-1-binding sites in the DNA sequence. KD, Kinase domain; P, phosphorylation; ATF, activating transcription factor; MAPKK, MAPK kinase; MAPKKK, MAPK kinase kinase.

Crosstalk between Smad and MAPK pathways adds to the complexity of TGF-ß signaling. Crosstalk can be obtained through physical interaction between Smad2, -3, and -4, and members of the Jun, Fos and ATF families bound to their AP-1 site in the promoter of target genes, possibly stabilized by Smad-DNA binding at an adjacent SBE site (76, 77). In addition, JNK (activated by TGF-ß) can phosphorylate Smad3, thus facilitating activation and nuclear translocation of the latter in response to TGF-ß (66). On the other hand, TGF-ß-activated c-Jun was shown to antagonize Smad signaling by enhancing interaction of Smad2 with a corepressor (78, 79). Recently, a hierarchical model of gene regulation by TGF-ß was proposed (80). Upon investigating the expression profile of hundreds of TGF-ß-controlled genes in fibroblasts deficient in Smad2, Smad3, or ERK signaling, respectively, Smad3 was demonstrated to be the critical mediator for expression of immediate-early target genes. Smad2 and the ERK pathways were found to function predominantly in the transmodulation of immediate-early and intermediate gene regulation. It would be also interesting to investigate this expression profile in p38 MAPK- and JNK-deficient cell lines.

Despite ample in vitro evidence in the literature for the involvement of MAPKs in the TGF-ß signaling cascade, data that unequivocally demonstrate the need for MAPK pathways in in vivo TGF-ß-mediated responses are lacking. Although knockout and transgenic mouse models of numerous MAPK signaling intermediates are available (81) (for mouse models with a bone phenotype, see Refs.82 and 83) none of them are scored for defects in TGF-ß signaling. However, keratinocytes derived from MEKK1-deficient mice show no migration in response to TGF-ß1 (84), and MKK3(–/–) mesangial cells are defective in TGF-ß1-induced vascular endothelial growth factor expression (85). These observations clearly show the requirement for MAPK-dependent signaling in transmitting TGF-ß signals.

Although it has been established that these Smad-independent pathways, like the Smad-dependent pathways, are initiated by the ligand-induced activation of the TßRI/TßRII receptor complex, the mediators acting between the receptor complex and the MAPKKKs have not been fully elucidated. In vitro studies have established a role for hematopoietic progenitor kinase-1 and TAK1 binding protein (TAB)-1 as TGF-ß-activated MAPKKKKs (70, 86) and for the X-linked inhibitor of apoptosis protein as a cofactor in TGF-ß signaling, possibly linking TßRI and TAB1 (87, 88). Upon disruption of tab1, delayed ossification and decreased responsiveness to TGF-ß stimulation are observed, suggesting a role for this MAPKKKK in TGF-ß-mediated bone formation (89).

In bone, examples of TGF-ß-mediated processes involving MAPK signaling are chemotaxis of osteoclasts (90), osteoclastogenesis (91), changes in osteoblast shape (92), osteoblast-to-osteocyte transdifferentiation (93), Runx2 expression in mesenchymal precursor cells (94), 1 and 2 collagen I expression in osteosarcoma cells (95, 96), collagenase 3 expression in osteoblastic cells (97), suppression of osteoblastic osteocalcin expression (98), and inhibition of alkaline phosphatase (ALP) activity and mineralization by osteoblasts (99).

In conclusion, we can state that the JNK, ERK, and p38 MAPK pathways contribute considerably to the whole of TGF-ß-induced responses, but further characterization is needed to assess their importance in relation to the Smad-dependent and other TGF-ß-induced signaling pathways (reviewed in Ref.100).

	III. TGF-ß1 in Bone

Top Abstract I. Introduction II. The TGF-ß... III. TGF-ß1 in Bone IV. Bone Phenotypes Associated... V. Therapeutic Use of... VI. Concluding Remarks References

 
A. Introduction
Bone is a mineralized tissue that serves many functions: providing mechanical support to joints, tendons, and ligaments; protecting soft tissues; supporting hematopoiesis; regulating blood calcium levels, etc. It consists largely of an organic matrix of type I collagen and noncollagenous proteins mineralized with hydroxyapatite crystals. Formation, deposition, and mineralization of bone tissue are executed by the osteoblasts that differentiate from mesenchymal precursor cells. The key transcription factor that drives the mesenchymal precursor cell toward the osteoblast lineage and controls bone formation is Runx2 (Cbfa1), which regulates the expression of all known marker genes expressed by the osteoblast (101). In addition, several other transcription factors and homeobox proteins, such as Dlx5, Msx2, Bapx1, Hoxa-2, Osx, and AP-1, affect osteoblast differentiation (102, 104). Local growth factors and cytokines regulating osteoblast differentiation include BMPs, fibroblast growth factors, platelet-derived growth factor (PDGF), IGFs, and Indian Hedgehog (103). Moreover, bone formation is regulated by endocrine factors such as sex steroid hormones, PTH , 1,25-dihydroxyvitamin D₃ [1,25-(OH)₂D₃, the active metabolite of vitamin D], and leptin (104, 105).

Throughout life, bone tissue is continuously remodeled by the balanced processes of bone resorption and consecutive bone formation. Bone resorption by the osteoclasts involves demineralization of the inorganic matrix by acidification followed by enzymatic degradation of the organic matrix by cathepsin K and matrix metalloproteinases (106). Osteoclasts are large, multinucleated cells (MNCs) of hematopoietic origin that differentiate from monocyte/macrophage precursor cells within the bone environment. The recognition that osteoclast differentiation requires the presence of marrow stromal cells or osteoblasts led to the discovery of the two osteoblast-derived factors essential and sufficient to promote osteoclastogenesis: macrophage-colony stimulating factor (M-CSF) and receptor activator of nuclear factor (NF)-B ligand (RANKL). Upon binding to their respective receptors on the osteoclast precursor cell surface (c-fms and RANK), two prominent transcription factor complexes, the NF-B and AP-1 proteins, are activated, and signaling cascades essential for proper osteoclast differentiation, fusion, function, motility, and survival are initiated. Downstream factors that are indispensable during osteoclast differentiation and which control expression of genes that typify the osteoclast lineage include PU.1 and MITF (107, 108), NFAT-c (109), and the transmembrane adaptor protein DAP12 (110). Molecules that identify the mature, functional osteoclast and are essential for cell survival, cell motility, and bone resorption, comprise c-Src, TRAP, carbonic anhydrase II, and cathepsin K (111). Osteoclastogenesis and bone resorption are modulated by a series of growth factors, cytokines, and hormones that can act directly on the osteoclast or indirectly through the osteoblast/stromal cell. Examples of proresorptive factors are TNF-, IL-1, 1,25-(OH)₂D₃, PTH, and PTHrP; among the antiresorptive factors, we find estrogens, calcitonin, BMP-2 and -4, PDGF, calcium, interferon-, and IL-4, -10, -17, and -18 (112).

A pivotal role in the bone-remodeling process has been assigned to TGF-ß1 because it was proven to affect both bone resorption and formation. TGF-ß1 is secreted in a latent form by bone cells and is stored in the ECM. Active, resorbing osteoclasts are capable of activating TGF-ß1, which in turn attenuates further bone resorption by impairing osteoclastogenesis and promotes bone formation through chemotactic attraction and stimulation of proliferation and differentiation of osteoblast precursors. Although this seems straightforward, the story is much more complicated because it turned out that the in vitro effects of TGF-ß1 on cells of the osteoblast and osteoclast lineage depend greatly on factors such as cell differentiation stage, cell density, TGF-ß1 concentration, the presence of serum, and other culture conditions. In vivo, the presence of other growth factors in the bone environment and the environment as such determine the exact outcome of TGF-ß1 functioning.

B. TGF-ß isoforms in bone
All three TGF-ß isoforms are detected in bone, but the TGF-ß1 isoform is the most abundant at the protein level (113). In cartilage of mouse embryos, TGF-ß1 is highly expressed in perichondrial cells, TGF-ß2 is expressed in chondrocytes, and TGF-ß3 is expressed in both (114). In embryonic bone, TGF-ß1 levels are high in periosteum and osteocytes. Little TGF-ß2 or -ß3 is detected in the periosteum, but they are readily detected in osteocytes (114). In neonatal and adult mice, TGF-ß1 protein is detected in bone marrow cells, chondrocytes, and cartilaginous matrix (115). To our knowledge, no detailed information on protein expression of TGF-ß2 or -ß3 in the adult mouse is currently available. In neonatal human bone, all isoforms can be found at sites of endochondral and intramembranous ossification but, again, the patterns of expression differ. At sites of endochondral bone formation, TGF-ß1 and TGF-ß3 are detected in proliferative and hypertrophic zone chondrocytes, and TGF-ß2 is detected in all zones of the cartilage. During intramembranous bone formation, TGF-ß1 and -ß2 colocalize with sites of mineralization, whereas TGF-ß3 is more widely distributed (116). Osteoclasts also express TGF-ß, mostly TGF-ß1, in high amounts (117). It is also important to note that expression of all three TGF-ß isoforms is up-regulated during fracture healing, suggesting that their roles are not restricted to embryonic bone development, but extend to adult bone remodeling (118).

Almost all nonmalignant cells secrete TGF-ß1 as LLC in conjunction with a LTBP (see Section I.B). Bone cells, however, form an exception: the SLC is the predominant form and is secreted with great efficiency (19, 119, 120). The SLC has not been observed in such large amounts in other cell systems (121), suggesting an important function for this form in bone. The SLC probably represents a pool of TGF-ß1 that is readily available, while another part is deposited in the bone matrix for storage through covalent binding of the LAP with LTBP-1 (19) and possibly LTBP-3 (20). Moreover, TGF-ß1 is present at a physiologically significant level in plasma (122), and this source may contribute to the reservoir stored in the bone matrix.

C. Role of TGF-ß1 in osteoblastogenesis and bone formation in vitro
The cellular events involved in bone formation are chemotaxis and proliferation of osteoblast precursors, differentiation to the mature osteoblast phenotype with synthesis of ECM proteins (e.g., type I collagen, osteopontin), and mineralization of the resulting matrix. Finally, osteoblasts either undergo apoptosis or transdifferentiate to osteocytes or bone-lining cells. All these events are under the control of both systemic hormones and local growth factors.

Data from numerous in vitro experiments have demonstrated the role of TGF-ß1 in every stage of bone formation. Despite conflicting results (see below), most data support the following model. TGF-ß1 increases bone formation in vitro mainly by recruiting osteoblast progenitors and stimulating their proliferation, thus expanding the pool of committed osteoblasts, as well as by promoting the early stages of differentiation (bone matrix production). On the other hand, it blocks later phases of differentiation and mineralization (123, 124). These later stages are regulated by other growth factors such as the BMPs (125). Apoptosis of osteoblasts is blocked by TGF-ß1 through maintenance of survival during transdifferentiation into osteocytes (93, 126, 127).

In contrast to the BMPs, TGF-ß1 is unable to induce osteogenesis in mesenchymal pluripotent cells, although it can inhibit differentiation to myogenic cells (128). However, once committed to osteogenesis, TGF-ß1 increases the pool of osteoprogenitors both by inducing chemotaxis (129, 130, 131) and proliferation. Indeed, most studies illustrate the mitogenic effect of TGF-ß1 on osteoprogenitors and osteoblast-enriched cell cultures (132, 133, 134, 135, 136, 137), although some have reported growth inhibition of osteoblast-like cells by this cytokine (138, 139). The biphasic, concentration-dependent effect of TGF-ß1 on osteoblast proliferation, with inhibition of DNA synthesis at high concentrations, lies at the basis of this discrepancy (140). Moreover, variables such as cell density, serum concentration, and differentiation stage were found to affect the outcome of TGF-ß1 treatment (138, 139, 140). In sparse cultures, TGF-ß1 is inhibitory at concentrations above 0.15 ng/ml, but this dose shifts to higher levels as the cell density increases, with a peak response at 15 ng/ml in confluent cultures. The underlying mechanism is unknown, but it seems plausible to assume that the in vivo situation, in which osteoblasts are found in tight clusters, corresponds best to a confluent culture, implying that at physiological (low range) concentrations, TGF-ß1 stimulates osteoblast proliferation.

Differentiation of osteoblast precursors can be monitored by the expression of proteins that compose the bone matrix, e.g., type I collagen, osteopontin, and osteonectin, as well as by the expression of the osteoblast differentiation markers, ALP and, in a later stage, osteocalcin. Conflicting data concerning the effect of TGF-ß1 exist for most of these markers. This is the case for the expression of type I collagen (132, 133, 135, 137, 138, 140, 141, 142, 143), other organic matrix components such as fibronectin, plasminogen activator inhibitor-1, osteonectin, osteopontin, and decorin (135, 136, 138, 141, 142, 143, 144), and ALP activity (136, 137, 138, 140, 141, 142, 145). Osteocalcin expression has been shown to be inhibited (142, 146, 147). We believe that these ambiguous observations can again be attributed to differences in the osteoblastic cell model system (tumorigenic vs. nontumorigenic), culture conditions (e.g., serum concentration), cell density, TGF-ß1 concentration, and the presence of other growth factors. However, the most important variable is the differentiation stage of the target cell population with a stimulatory effect of TGF-ß1 on differentiation of bone-forming cells in the early stage but an inhibitory effect in later stages.

In recent years, some of the molecular mechanisms underlying TGF-ß1 actions in bone formation have been elucidated. AP-1 binding sites and/or SBEs were identified in the promoters of many bone matrix proteins, e.g., osteopontin, fibronectin, type I collagen, osteocalcin, and ALP, pointing to a role for both MAPK- and Smad-dependent signaling (71, 73). Crosstalk between the Smad and MAPK pathways is also relevant in osteogenesis. Thus, despite the fact that TGF-ß1 inhibits ALP activity and mineralization in vitro, Smad3 was shown to stimulate both processes in osteoblasts (148). Inhibition of the ERK and JNK pathways antagonizes the inhibitory effect of TGF-ß1 on ALP activity, showing that the MAPK pathways negatively regulate the Smad pathway (99). The observation that an anti-Smad4 antibody or expression of dominant-negative Smad3 or Smad4 can down-regulate TGF-ß1-induced AP-1 DNA binding in osteoblasts also points to a role for Smads in modulating AP-1 activity (98).

In addition to acting directly, TGF-ß signaling can also affect bone formation indirectly. A master factor in bone formation is Runx2, also known as Cbfa1, a DNA-binding transcription factor specific for cells of the osteogenic lineage (149). Runx2 binding sites are found in the promoters of several bone formation markers including collagen 1, ALP, osteopontin, RANKL, and osteocalcin (149). Runx2 is a common target of TGF-ß1 and BMP2, mediating the inhibitory effect of these factors on myogenic differentiation of C2C12 pluripotent mesenchymal precursor cells (150). Induction of osteoblast-specific gene expression in these cells requires coordinated action between Runx2 and BMP2-induced Smad5 (151). In the early differentiation stage, TGF-ß1 induces the expression of Runx2 in combination with BMPs. However, in later stages of differentiation and maturation of osteoblasts, TGF-ß1 opposes BMP2 actions (152). Smad3, activated by TGF-ß1, physically interacts with Runx2 at Runx2-responsive elements, thus suppressing the expression of Runx2 and other osteogenic genes (collagen 1, ALP, osteocalcin) by an autoregulatory feedback mechanism (123). Furthermore, Smad2 overexpression decreases Runx2 mRNA levels (153). Recently, menin was identified as an additional regulatory factor essential in promoting the BMP-induced commitment of mesenchymal stem cells to osteoprogenitors through physical and functional interaction with Smads1/5 and Runx2 (154). It was shown that after osteoblast commitment, menin turns into a repressor of osteoblast maturation: by binding to TGF-ß1-activated Smad3, it mediates the inhibitory effect of the latter on Runx2 activity (154). The inhibitory role of TGF-ß1 in late phase osteogenesis in vitro was further confirmed through the use of a TßRI kinase inhibitor to suppress TGF-ß signaling. Whereas BMP-induced osteoblast commitment was unaltered, osteoblast differentiation and matrix mineralization were stimulated (124).

We conclude that TGF-ß1 generally inhibits mineralization of the matrix it helps to produce. However, the question is how responsive osteoblasts normally are to TGF-ß1 in their late differentiation stage. The answer might lie in the flux of TGF-ß receptors on the osteoblast membrane. A decrease in TßRI and TßRII expression is observed as human BMSCs progress from osteoprogenitor cells to maturing osteoblasts (36), confirming earlier findings in murine and rat osteoblastic cells that TGF-ß/receptor interactions decrease during osteoblast differentiation (155, 156). This would imply that osteoblasts in later stages are less sensitive to TGF-ß1. Moreover, TGF-ß1 itself has been shown to transiently or persistently (at low and high concentrations, respectively) down-regulate the levels of all receptor types on the osteoblast surface, primarily in late differentiation stage cells (157, 158). We propose the hypothesis that receptor down-regulation provides a way to decrease the responsiveness of the osteoblast toward TGF-ß1 to circumvent late-phase inhibition by this cytokine.

From the above, it is clear that the effect of TGF-ß1 on in vitro osteogenesis is highly dependent on a broad range of experimental conditions and is the final outcome of many interacting factors. Interactions are expected to be fully elucidated in the coming years and will be helpful to further explain the paradoxical findings reported in the past.

D. TGF-ß1 in osteoclast formation and bone resorption in vitro
Bone resorption involves the dissolution of bone mineral and the enzymatic degradation of the organic bone matrix by osteoclasts, giant MNCs. In vitro, spleen cells, bone marrow cells, peripheral blood mononuclear cells (PBMCs), and alveolar macrophages can act as a source of osteoclast precursors. The events of recruitment of osteoclast precursors to the bone environment, differentiation to the mature osteoclast, bone resorption, and osteoclast apoptosis are all modulated by TGF-ß1. Like many of the other cytokines influencing osteoclastogenesis and/or bone resorption, TGF-ß1 does not solely modulate these processes by direct action on osteoclasts and their precursors, but also acts via osteogenic cells. Upon binding to its receptor on the osteoblast membrane, expression of proteins involved in formation and activation of osteoclasts is induced.

The role of TGF-ß1 in osteoclastogenesis and bone resorption is very complex, and many seemingly contradictory reports have been published. Several important parameters must be taken into account when evaluating the studies performed in the past: 1) Is the study looking at osteoclastogenesis (MNC formation) or at bone resorption by mature osteoclasts? 2) Is the study utilizing isolated osteoclast cultures or a system where supporting cells (lymphocytes, stromal cells) are present? 3) Which TGF-ß1 concentration has been used? 4) How long and during which differentiation stage has the growth factor been applied?

Let us first turn our attention to MNC formation. In the last few years, a general model for the action of TGF-ß1 on osteoclastogenesis has emerged. According to this model (illustrated in Fig. 3), TGF-ß1 inhibits osteoclast formation in cocultures at high concentrations, while stimulating it in isolated cultures.

View larger version (37K): [in this window] [in a new window]   FIG. 3. Role of TGF-ß1 in osteoclastogenesis. 1, Effect of TGF-ß1 on osteoclast precursors in coculture. In the presence of osteoblasts/stromal cells, TGF-ß1 binds to its receptor on the osteoblast membrane (a) and modulates the expression of OPG and RANKL through Smad (b)- and MAPK (c)- dependent pathways. OPG expression and secretion are enhanced (d), whereas RANKL expression is down-regulated (e). This balance shift between OPG and RANKL results in a maximal occupation of RANKL by OPG and interferes with RANKL/RANK interaction (f). Consequently, osteoclastogenesis is impaired at high TGF-ß1 concentrations (g). 2, When added to isolated hematopoietic precursor cultures together with RANKL and M-CSF, TGF-ß1 induces osteoclastogenesis. After binding to its receptor on the osteoclast membrane (a), TGF-ß1 induces expression of RANK (b–d). The resulting increase in RANK/RANKL interactions leads to NF-B activation (e) and expression of osteoclastogenesis-inducing genes (f).

Duration of TGF-ß1 application affects the outcome of the experiment as well. Thus, a switch from inhibition to stimulation of osteoclast formation was detected in a mixed cell population (fetal long bone) subjected to high TGF-ß1 levels in the initial part of the culture period (d 1–3) or for longer time periods (d 1–7) (167). Furthermore, contamination of osteoclast cultures by lymphocytes (when using PBMCs as starting material) was shown to influence osteoclastogenesis. Thus, high levels of TGF-ß1 present in the initial part of the culture period vastly increase RANKL-/M-CSF-induced MNC formation and bone resorption in a human lymphocyte-rich population (91, 168). When TGF-ß1 is maintained during the entire culture period (28 d) or applied at a later stage of differentiation, this effect levels out, implying that the stimulatory effects of TGF-ß1 are restricted to the monocyte stage of the culture and shift to counteracting osteoclastogenesis in pre- and mature osteoclasts (91). In a lymphocyte-poor or pure monocyte population, osteoclast formation evoked by M-CSF and RANKL was only modestly enhanced by TGF-ß1 (91, 168).

Others consistently observe a costimulation of RANKL/M-CSF-induced MNC formation by TGF-ß1, both at low and high concentrations, in cultures of isolated M-CSF-dependent bone marrow cells or other osteoclast precursors for which extreme care was taken to remove all contaminating cells (169, 170, 171, 172, 173). Direct effects of TGF-ß1 on osteoclast precursors, such as up-regulation of RANK expression (172, 174) and induction of NF-B activation (169) and suppressor of cytokine signaling expression (175), are responsible for this positive effect. Moreover, because no OPG-expressing cells are present, TGF-ß1 is unable to induce OPG to counteract RANK/RANKL interaction.

What about the effect of TGF-ß1 on bone resorption? Likewise, this completely depends on the cellular context. When the system is dependent on osteoclast recruitment from hematopoietic precursors for bone resorption, a process that is inhibited by TGF-ß1, resorption will be impaired. This is the case, for example, in fetal long bones where the marrow cavity is still developing and osteoclasts have not yet invaded. On the contrary, in calvaria or older long bones with an established marrow cavity, where mature osteoclasts are already present, TGF-ß1 stimulates them to resorb bone at all concentrations (162, 167, 176, 177, 178).

E. Interaction of TGF-ß1 with other growth factors and hormones
Upon elucidation of the Smad signaling pathway, it appeared remarkably simple for such a complex group of cytokines as the TGF-ß superfamily: the ligand assembles a membrane receptor complex that activates the Smads, and the Smads assemble complexes that regulate transcription. However, since this discovery, an intricate web of crosstalk has been revealed. On the one hand, TGF-ß-induced pathways (such as the Smad and MAPK signaling pathways) interact with each other (see Section II.C). On the other hand, crosstalk occurs with pathways initiated by other local and systemic factors. The most important hormones and cytokines are discussed in Table 1. It must be noted that some of the interactions have been reported in a specific culture model and do not necessarily apply to other models as well. For a few of these factors the molecular mechanisms of synergy and antagonism have been discussed below.

TNF-, a proinflammatory cytokine, has profound effects in the bone environment, as it induces both osteoblast apoptosis and osteoclastogenesis (290). TGF-ß has been shown to modulate both processes (239, 291), but the underlying mechanisms are thus far unknown. In fibroblasts, experimental evidence points to a role for the NF-B and AP-1 transcription factor families activated by TNF- in TNF-/TGF-ß crosstalk. Upon TNF- treatment, JNK-mediated c-Jun and JunB phosphorylation decreases Smad/DNA interactions, either through formation of off-DNA Smad/Jun complexes or through competition of Smad and Jun for binding to the coactivator p300 (241, 292). Moreover, a TNF-- and NF-B-mediated up-regulation of Smad7 synthesis antagonizes TGF-ß signaling in mouse embryonic fibroblasts (293). In a similar setup in human dermal fibroblasts, however, Smad7 expression was not observed (292). In addition, Smad7 expression was shown to be decreased by TNF- and NF-B in human embryonic kidney 293 cells through competition of NF-B with Smad3 (a stimulator of Smad7 expression) for binding to p300 (294). These examples show that the effect of TNF- on Smad signaling is influenced by the cell system used. Further research needs to be performed in bone model systems to elucidate the molecular basis of the crosstalk between these two growth factors in bone.

Estrogen has powerful effects on cells both of the osteoblast and the osteoclast lineage (295). It stimulates osteoblast proliferation, differentiation, deposition of ECM, and mineralization (296). On the other hand, osteoclast maturation and function are impaired, whereas osteoclast apoptosis is promoted (261, 297). Taken together, estrogen is a potent anabolic agent in bone. In vivo, this is illustrated by the bone loss upon estrogen depletion after the menopause. Accumulating evidence points to a role for TGF-ß1 in mediating some of the effects of estrogen, e.g., promotion of murine osteoclast apoptosis (261). In vivo, TGF-ß1 mRNA and protein levels decrease in ovariectomized rats (298, 299), whereas estrogen treatment of postmenopausal women stimulates TGF-ß1 and TGF-ß2 mRNA and protein production (300). Information from the mechanism of interplay between TGF-ß and estrogen signaling pathways comes from studies in human embryonic kidney 293T and human breast cancer cells, in which crosstalk is mediated by physical interaction between Smads and estrogen receptor (ER)-. These studies identified ER- as a transcriptional corepressor for Smad activity (301). Smad3, on the other hand, can enhance ER-mediated transcriptional activity (301), although Smad4 behaves as a transcriptional corepressor, suppressing the Smad3-mediated ER- transactivation (302).

From the data presented in Table 1, it can be concluded that the effects of TGF-ß1 on bone formation and resorption, both in vitro and in vivo, must be evaluated in view of the presence of other cytokines and hormones, which modulate or are modulated by TGF-ß1 signaling in a number of ways.

	IV. Bone Phenotypes Associated with Abnormal TGF-ß1 Signaling

Top Abstract I. Introduction II. The TGF-ß... III. TGF-ß1 in Bone IV. Bone Phenotypes Associated... V. Therapeutic Use of... VI. Concluding Remarks References

 
A. Bone phenotypes of knockout and transgenic mouse models of the TGF-ß signaling pathway
Information about the function of TGF-ß and its downstream signaling mediators can be gained from the study of knockout mice, in which one intermediate of the pathway has been eliminated by gene targeting. In Table 2, an overview has been presented of the targeted deletions of the genes encoding the ligands TGF-ß1, -ß2, and -ß3, the binding proteins LTBP-3 and -4, the receptors TßRI and TßRII, and the intracellular mediators involved in Smad-dependent TGF-ß signaling, Smad2, -3, and -4. With the exception of ltbp4 null mice, all knockout mice that outlive the stage of osteogenesis develop severe bone defects. This finding stresses the importance of TGF-ß signaling in the ossification process both during embryonic development and postnatally.

Tgfb2 (307) and tgfb3 (308, 309) knockouts both show abnormalities in bone development, although far more severe in the former (see Table 2). The marked differences of the bone phenotypes of the three tgfb knockout mice reflect the differences in bone-specific expression of the isoforms. It is also important to note that in tgfb1 knockout mice, TGF-ß2 and TGF-ß3 mRNA levels are unaltered (306), which provides additional evidence for their nonredundancy.

A large percentage of TGF-ß is found in association with a LTBP, which facilitates folding and secretion and targets the complex to the ECM (see Section I.B). Therefore, absence of one of the LTBP isoforms could influence normal TGF-ß processing and storage. The ltbp2 null mouse is not included in Table 2 because LTBP-2 is unable to bind TGF-ß and therefore has no role in the formation of the LLC. Although LTBP-4 can bind TGF-ß, the absence of a bone phenotype in the ltbp4 knockout mouse (311) suggests that this LTBP isoform is not involved in the folding and storage of TGF-ß in bone. The bone phenotype observed when knocking out ltbp3 (310) (see Table 2) implies that this isoform is probably responsible for the formation of a percentage of the LLC found in bone tissue. This is in accordance with earlier observations that human LTBP-3 is expressed in osteoblasts and some osteosarcoma cell lines and secreted as a LLC in conjunction with TGF-ß1 (20). Unfortunately, no knockout mouse model has been developed for ltbp1, which is presumed to be an important TGF-ß1-binding protein in bone as well (19, 323).

TßRI or TßRII deficiency is embryonic lethal, precluding any study of the effect of their absence on bone development (312, 314, 315). The severity of the phenotypes highlights the requirement of TGF-ß signaling during embryonic development. Recently, a conditional knockout of tgfbr2, which limits absence of the type II receptor to Col2a-expressing cells, was developed (324). The majority of the mice did not survive postnatally, but bone defects were examined in 13.5-d through 17.5-d embryos. Bone abnormalities (size reductions) were confined to the parts of the skull that develop through endochondral ossification and the spine, demonstrating a crucial role for TßRII in axial skeleton development. The appendicular skeleton remained normal until 1–12 wk after birth, when a progressive reduction in the length of the proximal long bones was noted. Although tgfbr2 is redundant in embryonic long bone development, it appears to function postnatally.

As intracellular mediators of the canonical TGF-ß signaling pathway, the absence of one of the TGF-ß-induced R-Smads or the Co-Smad is anticipated to profoundly affect normal development. Because Smad4 is the common mediator Smad for all TGF-ß superfamily members, it is not unexpected that its absence causes 100% embryonic lethality (320, 321). Likewise, Smad2 deficiency inevitably leads to embryonic death (314, 315). Remarkably, the phenotype observed in smad3 knockouts depends greatly on the strategy used to knock out the gene. Thus, mice with a targeted disruption in exon 1 or 2 are viable (317, 318, 319), whereas disruption of exon 8 causes death between 1 and 8 months of age (316). These discrepancies prove the need to carefully examine expression and function of the mutant protein. Indeed, Yang et al. (316) observed that Smad3^ex8/ex8 has residual activity and is capable of suppressing TGF-ß-induced reporter activity at supraphysiological levels in in vitro assays. In our opinion it cannot be excluded that at physiological levels, the mutant protein still interacts with cofactors or DNA, thus inhibiting, for example, proper functioning of the other TGF-ß/activin-specific R-Smad. In bone, loss of Smad3 was shown to result in osteopenia or degenerative joint disease when exon 1 or 8, respectively, is disrupted. In the former, TGF-ß-mediated proliferation is intact, but TGF-ß is no longer able to inhibit osteoblast differentiation, thereby increasing the osteocyte fate and eventually leading to apoptosis. In the latter, terminal differentiation of chondrocytes is no longer inhibited by TGF-ß. The reason for this discrepancy is currently unclear but might involve differences in genetic background and the presence of modifier loci. Despite these differences, the main conclusion is that Smad3 seems to lack a crucial role during embryonic development but is indispensable in adult tissues. In bone, not bone formation as such, but rather bone remodeling and maintenance are affected by loss of Smad3. Although Smad2 and Smad3 have 95% sequence identity and have been used interchangeably in mediating TGF-ß signaling in in vitro assays in the past, it is now well established that they activate separate sets of target genes, as shown recently by Yang et al. (80) (see Section II.C), as well as display differences in their mechanism of transcriptional activation (e.g., Smad3 can bind directly to DNA, whereas Smad2 is unable to do so). In this context, it is worth noting that Felici et al. (325) recently identified a new TßRII-interacting protein, TLP1, that plays a role in regulating the balance between Smad2 and Smad3 signaling. The marked differences of the respective knockout phenotypes confirm the functional specificity of these two R-Smads in vivo.

At present, no data are available on the bone-specific overexpression of tgfb1. However, Erlebacher and Derynck (326) targeted a constitutively active form of tgfb2 under the osteocalcin promoter to mature osteoblasts. Heterozygous mice with a 16-fold increase in TGF-ß2 production showed a dramatic, age-dependent loss of bone mass reminiscent of high-turnover osteoporosis. The long bones were normal in length, but their cortices were thinner; clavicles were practically absent. Histologically, meta- and epiphyseal trabeculation were dramatically reduced. Osteocyte density and osteoprogenitor cell numbers were markedly increased, and mineralization in the cortical bone was impaired. These effects could be accounted for by the increased activities of both osteoblasts and osteoclasts. By expressing a cytoplasmatically truncated TßRII from the osteocalcin promoter, Filvaroff et al. (327) inhibited TGF-ß signaling in osteoblasts. The resultant bone phenotype was in many ways the opposite of that of the tgfb2-overexpressing mice. An age-dependent increase in trabecular bone mass was associated with decreased osteoblast differentiation and osteoclastic resorption. Generation of double transgenic mice, overexpressing both TGF-ß2 and the truncated TßRII, allowed workers to discern osteoblast-dependent and -independent effects (328). These studies confirm the role of TGF-ß (in this case the TGF-ß2 isoform) in bone remodeling and in keeping the balance between bone formation and resorption. The findings are surprising in the light of the in vivo stimulation of bone formation upon administration of TGF-ß1 (see Section V) or -ß2 (329, 330). Several explanations come to mind. First, restriction of TGF-ß2 expression to the mature, nondividing osteoblast is not fully representative of the natural in vivo situation, in which TGF-ß is expressed at all stages of osteoblast differentiation. It is even plausible that forced expression in the late differentiation stage inhibits bone formation, as shown in vitro for TGF-ß1 as well (see Section III.C). In addition, the major isoform in bone is TGF-ß1, accounting for up to 90% of TGF-ß found in the bone environment. Second, osteoblast activity is most perturbed in the epiphyses and diaphyses, which are the sites of highest osteocalcin expression, suggesting that the action of TGF-ß2 is restricted to its sites of production. Again, this does not correspond to the normal in vivo situation. Third, the authors report an increased bone resorption without alterations in osteoclast number. However, it should be taken into consideration that because osteoclasts lack betaglycan, they are probably relatively unresponsive to TGF-ß2, as has been shown before for hematopoietic progenitor cells (331). High levels of TGF-ß2 might overcome the low affinity of TGF-ß2 for TßRII in this mouse model, but the direct effect of physiological levels of TGF-ß2 on osteoclasts remains to be elucidated. Therefore, we feel that observations in these mouse models need further experimental support. Construction of a transgenic mouse overexpressing tgfb1 or tgfb2 under a promoter for early osteoblast differentiation, such as collagen type I, would be helpful in unraveling the in vivo effects of these isoforms in bone formation.

B. TGFB1 mutations in the pathogenesis of Camurati-Engelmann disease (CED)
CED or progressive diaphyseal dysplasia is a rare bone disorder with an autosomal dominant mode of inheritance. Radiologically, it is characterized by hyperostosis and sclerosis of the diaphyses of the long bones and sclerosis at the skull base. Patients suffer mainly from bone pain, muscle weakness, a waddling gait, and fatigue (332). In 2000, we and others succeeded in identifying TGFB1 as the disease-causing gene underlying this bone disorder (333, 334). Ten different mutations have been reported thus far (333, 334, 335, 336, 337, 338) (Table 3). With one exception, a duplication of three Leu-residues in the signal peptide, all are missense mutations in the LAP. Recently, we were able to show that the mutations can be functionally divided into two groups (337). The first group, represented by mutations concentrated around the cysteine residues responsible for dimerization of the LAP (see Fig. 1), affects activation of the mutant protein upon overexpression: secretion is normal, but the amount of active protein is approximately doubled, due to destabilization of the dimerization process. This is reflected by a major overinduction of Smad-dependent signaling, as measured by the phosphorylation level of Smad2 and luciferase activity evoked by a TGF-ß responsive transcriptional reporter. The second group of mutants, located at the N terminus, shows severe impairment of TGF-ß1 secretion. Nevertheless, the Smad-dependent transcriptional response is likewise increased. The observation of signaling in the absence of extracellular active protein led us to formulate a hypothesis of intracrine signaling in which the TGF-ß receptor complex, which is constantly internalized and recycled back to the plasma membrane, initiates signaling when encountering active TGF-ß1 intracellularly. However, this hypothesis awaits further experimental evidence.

The phenotype of CED patients contrasts with the reported phenotype of the tgfb2 transgenic mouse (see Section IV.A). This can be attributed to different reasons. First, knocking out tgfb1 and tgfb2 has shown that the isoforms are functionally nonredundant (Section IV.A), making it difficult to compare their in vivo activities after overexpression. Second, the 16-fold overexpression of TGF-ß2 is restricted to the mature osteoblast, whereas in CED patients, the limited TGF-ß1 overactivity is present during all phases of osteoblast differentiation and in all bone cells. Construction of a knock-in mouse model carrying one of the CED mutations would therefore be a valuable tool to gain information on osteoblast and osteoclast functioning upon tgfb1 overactivity.

The relatively mild phenotype of CED patients is surprising when taking into account the versatile processes in which TGF-ß1 is implicated during embryogenesis and adult life and is in sharp contrast to the severity of the phenotypes observed after knocking out the gene or overexpressing it in a tissue-specific manner. We suggest the following hypothesis. In most tissues, TGF-ß1 appears as a high-molecular weight latent complex (LLC) in conjunction with a LTBP and is stored as such in the ECM (340). However, bone cells produce predominantly the SLC, consisting of the LAP and the mature peptide, but excluding the LTBP (19, 119, 120). This form is suggested to represent a pool of readily available TGF-ß1, necessary in an environment in which this cytokine plays such an important role throughout life (19). The effect of a particular mutation might depend on the nature of the latent complex, with a role for the LTBP in neutralizing the conformational changes brought forth by the mutations, thereby restricting full activity of the mutant protein to the bone environment. Furthermore, it is intriguing to note that all affected cell types, osteoblasts, chondrocytes, adipocytes, and myocytes, originate from the same mesenchymal stem cell population. Consequently, it would be interesting to investigate the nature of the latent complex in fat and muscle tissue.

C. Osteolytic metastases: a role for TGF-ß1 in malignancy
Of the different tumor types known to be associated with osteolytic lesions, breast carcinoma is the most common: in more than 80% of patients with advanced disease, breast cancer metastasizes to bone where it gives rise to osteolytic lesions, both by activating osteoclastic bone resorption and inhibiting osteoblastic bone formation. These lesions can give rise to pain, hypercalcemia, fractures, and nerve-compression syndromes. The avidity of breast tumors for the bone environment is due to the high concentration of growth factors present, possibly in combination with favorable interactions between specific receptors on the bone marrow endothelial cells and cell surface structures on the o