Transforming Growth Factor-β-induced Osteoblast Elongation Regulates Osteoclastic Bone Resorption through a p38 Mitogen-activated Protein Kinase- and Matrix Metalloproteinase-dependent Pathway*

Transforming growth factor-β (TGF-β) is a powerful modulator of bone metabolism, and both its anabolic and catabolic effects on bone have been described. Here we have tested the hypothesis that TGF-β-induced changes in osteoblast shape promote bone resorption by increasing the surface area of bone that is accessible to osteoclasts. The addition of TGF-β1 to MC3T3-E1 cells resulted in cytoskeletal reorganization, augmented expression of focal adhesion kinase, and cell elongation, accompanied by an increase in the area of cell-free substratum. TGF-β1 also triggered activation of Erk1/2 and p38 mitogen-activated protein (MAP) kinase. The p38 MAP kinase inhibitor PD169316, but not an inhibitor of the Erk1/2 pathway, abrogated the effect of TGF-β1 on cell shape. The matrix metalloproteinase inhibitor GM6001 also interfered with osteoblast elongation. Treatment of MC3T3-E1 cells seeded at confluence onto bone slices to mimic a bone lining cell layer with TGF-β1 also induced cell elongation and increased pit formation by subsequently added osteoclasts. These effects were again blocked by PD169316 and GM6001. We propose that this novel pathway regulating osteoblast morphology plays an important role in the catabolic effects of TGF-β on bone metabolism.

The skeletons of developing and adult mammals undergo constant remodeling, i.e. old bone is regularly removed and new bone is regularly laid down (1,2). The major players in the remodeling of bone are two specialized and functionally coupled cell types: osteoblasts, which deposit organic and inorganic matrix, and osteoclasts, which remove bone matrix. Osteoclast function is controlled by both systemic and local factors, most of which act through or are produced by osteoblasts. Similarly, the ability of osteoblasts to deposit bone matrix is stimulated by factors that are produced by osteoclasts or released from bone in the wake of matrix dissolution. Thus, in living bone, osteoblasts play a paramount role in determining the functional state of osteoclasts, and osteoclasts are of supreme importance for osteoblast function.
Most of the arguments concerned with the quest for identi-fication of key mechanisms that produce bone resorbing osteoclasts, i.e. those mechanisms involved in osteoclast precursor proliferation and commitment, and osteoclast migration, fusion, and resorption focus on the direct effect of molecules found in the extracellular space or on the cell surface of osteoblasts or other stromal cells on osteoclast precursors and maturing osteoclasts (3). However, based on the observation that cells of the osteoblast lineage, the bone lining cells, display a cobblestone morphology and cover the bone surface in an epithelium-like manner, an additional mode of how osteoblasts may modulate the ability of osteoclasts to resorb bone has been proposed (4,5). In this model, bone lining cells mechanically hinder the access of osteoclasts to the bone surface. Consequently, in order to allow osteoclasts to degrade bone matrix, bone lining cells must retreat from part of the bone surface.
One of the few examples in the literature that illustrates how this may take place is parathyroid hormone (PTH)1-induced osteoblast retraction (6,7). In response to PTH, osteoblasts in culture adopt a stellate morphology and expose the underlying substratum. Osteoblast retraction occurs within 1 h, is reversible, and involves cyclic AMP signaling and activation of intracellular proteases. However, whether or not the retraction of osteoblasts triggered by PTH is indeed able to increase recruitment of osteoclasts to the bone surface and, hence, to promote bone resorption has never been investigated. Transforming growth factor-␤ (TGF-␤) is another growth factor that profoundly alters osteoblast shape (8). It is one of the most abundant growth factors in skeletal tissues (9) and, in mammals, comprises three isoforms, TGF-␤1, TGF-␤2, and TGF-␤3, all of which are expressed by bone cells (10) and interact with the known TGF-␤ receptors types I, II, and III (betaglycan) (11). All TGF-␤ isoforms display similar biological activities (12), although one isoform may be more potent than another in a given assay (13).
Whereas the role of TGF-␤-induced changes in osteoblast morphology is largely unknown, numerous studies suggest that TGF-␤ has multiple functions in bone metabolism. In vivo studies involving the application of exogenously administered recombinant TGF-␤ show that TGF-␤ can increase bone formation and promote fracture healing (12,14). TGF-␤1 knock-out mice display an about 30% decrease in tibial length and a reduction in bone mineral content (15), consistent with the idea that TGF-␤ functions as a bone-forming agent. On the other hand, overexpression of TGF-␤2 controlled by the osteoblastspecific osteocalcin promoter leads to bone loss (16), and mice expressing a dominant-negative TGF-␤ type II receptor have no obvious skeletal defects apart from joint abnormalities that are probably due do chondrocyte malfunction (17). These apparently conflicting results are also reflected in numerous studies in cell and tissue cultures, where TGF-␤, in dependence on particular experimental parameters, modulates various bone cell activities in opposite ways. For instance, both increases and decreases in osteoclast formation, bone resorption, osteoblast proliferation, and osteoblast differentiation have been reported (12,18).
Like the functional aspects of TGF-␤ action on bone cells, the intracellular signals triggered by binding of TGF-␤ to its receptor are numerous and complex. They include phosphorylation and activation of Smad transcription factors, regulation of Ras, Rho, and other small G-proteins, activation of protein kinases such as TGF-␤-activated kinase-1 (TAK1), mitogen-activated protein (MAP) kinases, and Src, recruitment of adaptor proteins and regulation of ion channels (19). Not surprisingly then, TGF-␤ regulates expression of a spectrum of genes with potential significance for bone cell function, such as those coding for various extracellular matrix molecules and their receptors, proteinases and proteinase inhibitors, cell-cell adhesion molecules, and growth factors. The fact that some of the signaling pathways downstream from the TGF-␤ receptor are also common to and/or modulated by a variety of other growth factors and extracellular matrix molecules would also explain the seemingly contradictory results obtained with TGF-␤ in various functional assays as outlined above, i.e. the precise nature of the cellular microenvironment may determine the outcome of TGF-␤ action.
Promotion of osteoclast recruitment through morphological transformation of osteoblasts as a result of local release and/or activation of TGF-␤ may represent an attractive model to explain spatiotemporal regulation of bone resorption and the functional coupling of osteoblast and osteoclasts. In this study, we set out to better define the effect of TGF-␤ on cell shape and to investigate the mechanism by which TGF-␤ induces osteoblast elongation. We provide evidence that two classes of molecules that are regulated by TGF-␤1, MAP kinases and matrix metalloproteinases (MMPs), play a pivotal role in TGF-␤1induced alterations in osteoblast shape and that osteoblast elongation promotes osteoclastic bone resorption.

MATERIALS AND METHODS
Cell Culture-The mouse osteoblast cell lines MC3T3-E1 (MC) (20) was routinely maintained and passaged in growth medium consisting of ␣MEM (Life Technologies, Inc.) containing 5% fetal bovine serum (Life Technologies), 100 units/ml penicillin and 100 g/ml streptomycin.
For experiments, MC cells were plated at a density of 30,000 cells/ cm 2 into 96-well tissue culture plates (Costar) or onto glass coverslips and maintained under serum-free conditions in ␣-MEM (Life Technologies, Inc.) for 24 h unless otherwise indicated. Substrata were either left untreated or coated with 30 g/ml type I collagen (Nitta Collagen). Unless otherwise indicated, growth factors and inhibitors were added at the time of plating at the following concentrations: TGF-␤1 2.5 ng/ml, bFGF 2 ng/ml, and PTH 10 nM (R&D Systems); GM6001  To analyze osteoclastic bone resorption in the presence of a bone lining cell layer, MC cells were seeded at a density of 80,000 cells/cm 2 onto slices of bovine femur cortical bone (6 mm in diameter, 0.2 mm thick) placed in 96-well plates, and cultured under serum-free conditions for 24 h, after which time they had formed a confluent layer of bone lining cells. MC cells were then washed once with serum-free medium, and 200 l of cell suspension containing ϳ500 osteoclasts in ␣-MEM with 2% fetal bovine serum were added on top of the bone lining cell layer. After a settling period of 90 min, nonadherent cells were removed by replacing the medium with ␣MEM containing 0.5% fetal bovine serum. The culture was then continued for a further 48 h. To analyze osteoclastic bone resorption in the absence of a bone lining cell layer, MC cells were mixed with osteoclasts at the time of plating. In one set of experiments, bone lining cells were fixed with 96% ethanol for 20 min at Ϫ20°C and washed three times with medium before osteoclasts were seeded on top of the bone lining cells.
Osteoclasts were obtained as unfractionated bone cells according to the method described by Tezuka et al. (21) with minor modifications. Briefly, long bones from 10-day-old rabbits were minced in ␣MEM and gently agitated for 30 s with a vortex mixer. After sedimentation of bone fragments for 1.5 min, the supernatant was harvested and washed twice by centrifugation at 45 ϫ g for 2 min. Cells were then resuspended and plated as described above.
To quantify cell number, the Alamar Blue assay (Trek Diagnostics) was used according to the manufacturer's instructions.
Quantification of Cell Shape Changes and Pit Formation-To study cell shape, cells were fixed with 5% glutaraldehyde in phosphate-buffered saline (PBS), washed extensively with water, and stained with 0.5% toludine blue (Sigma) in 2.5% Na 2 CO 3 . Cell shape changes were quantified with ImagePro software by measuring the areas of the substratum that were covered and not covered by the cell monolayer.
To quantify pit formation by osteoclasts at the end of the culture period, cells were scraped gently off of the bone slices with a cotton stick. The bone slices were then washed with water and stained for resorption pits with Mayer's hematoxylin (Sigma). After brief sonication in a water bath and subsequent washing, the resorbed area was measured using CAST-GRID software (Microsoft, Olympus).
Indirect Immunofluorescence and Actin Staining-For detection of FAK by indirect immunofluorescence, cells were maintained on glass coverslips, fixed with formalin, and processed as described (23). FAK expression was visualized using mouse monoclonal antibodies from Transduction Laboratories and rhodamine-conjugated donkey antimouse secondary antibodies from Jackson Laboratories.
For actin staining, cells were fixed with 4% paraformaldehyde for 30 min, washed with PBS for 15 min, and permeabilized with acetone for 5 min at Ϫ20°C. Cells were rinsed several times with PBS and incubated with fluorescein isothiocyanate-conjugated phalloidin (Sigma) for 1 h in PBS. Cells were mounted in Vectashield containing 4Ј,6-diamidino-2-phenylindole (DAPI, Vector Laboratories) for visualization of nuclei.

Osteoblast Elongation Induced by TGF-␤1 Is Accompanied by an Increase in Cell-free Substratum and Reorganization of Focal Contacts and the Actin Cytoskeleton-Osteoblasts
maintained in the presence of serum have repeatedly been shown to transform from a cuboidal or cobblestone to an elongated morphology in response to TGF-␤1 (8, 24 -26). When MC cells that were plated at high cell density on tissue culture plastic in chemically defined medium were exposed to TGF-␤1, cell elongation was apparent after 16 -24 h in culture but not at earlier time points (Fig. 1A and not shown). These shape changes were accompanied by a decrease in cell spreading and, hence, by an increase in the area of cell-free substratum (Fig. 1, A and B). Furthermore, they occurred regardless of whether TGF-␤1 was added at the time of cell plating ( Fig. 1A) or when cells were allowed to establish a cobblestone morphology prior to TGF-␤1 treatment, i.e. when TGF-␤1 was added 1 day after cell plating (not shown). The effect of TGF-␤1 on cell shape was dose-dependent and maximal at 2.5 ng/ml (Fig. 1B). When cells were plated on a type I collagen substratum, TGF-␤1 also induced cell elongation, albeit to a lesser extent than on tissue culture plastic (Fig. 1A), an observation consistent with the reported decrease in expression of the TGF-␤1 receptors in cells exposed to type I collagen (27). The effect of TGF-␤1 on cell shape was reversible when cells were treated for 1 day with TGF-␤1 and subsequently maintained for another 4 days without TGF-␤1 (Fig. 1C). Similar results were also obtained with cells maintained in serum-containing medium (Fig. 1C). Under serumfree conditions, cell numbers increased only slightly during a 5-day culture period and were unaffected by TGF-␤1 (Fig. 1D). In contrast, TGF-␤1 reduced the time-dependent increase in cell numbers observed in serum-containing medium (Fig. 1D).
These results indicate that the effect of TGF-␤1 on cell shape is unrelated to and independent of its effect on osteoblast proliferation.
The morphological transformation in response to TGF-␤1 suggests that TGF-␤1 alters focal adhesion contacts and cytoskeletal organization in MC cells. To address this question, cells were labeled with phalloidin, and antibodies against FAK. As shown in Fig. 2, A and B, stress fibers were much more elaborate in osteoblasts treated with TGF-␤1, and localization of FAK at focal contacts was increased. Furthermore, expression of FAK protein was increased in response to an 8-and 24-hour treatment with TGF-␤1 (Fig. 2C).
p38 MAP Kinase Mediates the Effect of TGF-␤1 on Cell Shape-MAP kinases control cell function and shape through integration of signals from the extracellular matrix and growth factors (28). Furthermore, the activation of MAP kinases by TGF-␤1 has been described in several cell types (19). We therefore analyzed the expression and activation of MAP kinases in MC cells stimulated with TGF-␤1. Expression levels of the three MAP kinase members, Erk1/2, p38 MAP kinase, and JNK, were not altered by TGF-␤1 (Fig. 3). However, Erk1/2 and p38 MAP kinase, but not JNK, were activated as determined by phosphorylation state-specific antibodies (Fig. 3). Phosphorylation/activation of Erk1/2 and p38 MAP kinase was obvious between 1 and 9 h after the addition of TGF-␤1. At 24 h, Erk1/2 activation was no longer detectable, but p38 MAP kinase still displayed increased phosphorylation (Fig. 3). This suggests that the effect of TGF-␤1 on cell shape, which is apparent after 24 h but not after 9 h of TGF-␤1 treatment, is mediated by sustained activation of p38 MAP kinase but not by Erk1/2.
To determine the consequences of MAP kinase activation for osteoblast morphology, cells were treated with the MEK1/2 FIG. 1. Effect of TGF-␤1 on osteoblast shape and cell number. A, micrographs of toluidine blue-stained MC cells plated on untreated tissue culture plastic or collagen-coated substrata and maintained for 24 h in serum-free medium in the presence (TGF-␤) or absence (w/o) of TGF-␤1. B, MC cells were maintained for 24 h in serum-free medium in the presence of increasing concentrations of TGF-␤1. Cells were then fixed and stained, and the area of the substratum that was not covered by cells was quantified and expressed as % of the total area analyzed.
Error bars indicate standard deviations from one representative experiment. C, micrographs of toluidine blue-stained MC cells maintained for 5 days on tissue culture plastic in serum-free (Ϫ serum) or serum containing (ϩ serum) medium. Cells were kept in the absence (w/o) or presence (TGF-␤ continuous) of TGF-␤1 for the total culture period or in the presence of TGF-␤1 only for the first day of culture (TGF-␤ first day only). D, the number of cells maintained under the same conditions as in C was quantified using the Alamar Blue assay. The results are normalized, with values obtained at 1 day of culture in the absence of TGF-␤1 set to 100. Error bars indicate standard deviations from one representative experiment.
inhibitor PD98059, which specifically blocks the Erk1/2 pathway and not the p38 MAP kinase pathway (29,30,36), and with the kinase inhibitor PD169316 (31,36,73,74), which specifically blocks the p38 MAP kinase and not the Erk1/2 pathway. Cells treated with TGF-␤1 and PD98059 had a morphology similar to cells maintained with TGF-␤1 alone (Fig.  4A). In contrast, cells treated with TGF-␤1 and PD169316 had morphologies similar to cells not treated with TGF-␤1 (Fig. 4A). Furthermore, PD169316 but not PD98059 strongly reduced the increase in cell-free substratum area observed in response to TGF-␤1 (Fig. 4B). Results similar to those obtained with PD169316 were also obtained with the related p38 MAP kinase inhibitor SB203580 (not shown). None of the MAP kinase inhibitors tested affected cell shape in the absence of TGF-␤1 (not shown). To verify the specificity of the inhibitors used, we tested the effect of SB203580, PD169316, and PD98059 on both the p38 and p44/42 MAP kinase activity. PD98059 did not inhibit the p38 MAP kinase activity but totally abrogated the p44/42 MAP kinase activity as previously reported (30). SB203580 and PD169316 did not inhibit the p44/42 MAP kinase but blocked p38 MAP kinase activity as previously reported (31,36,73,74) (data not shown).
The broad-spectrum tyrosine kinase inhibitor genistein and the phosphatidylinositol-3 kinase inhibitor LY294002 had no significant effect on cell shape in the presence (Fig. 4, C and D) or absence (not shown) of TGF-␤1. The protein kinase C inhibitor chelerythrine had a slight but insignificant effect on cell shape and only in the presence of TGF-␤1 ( Fig. 4D and not shown). Taken together, these data indicate that p38 MAP kinase plays a central role in the TGF-␤1-induced conversion of osteoblasts from cuboidal to elongated.
MMPs Regulate TGF-␤1-induced Osteoblast Elongation-It has recently been shown that MMPs can trigger the morphological transformation of epithelial cells from a cobblestone to an elongated and fusiform morphology, both in culture and in vivo (23,32). Thus, to further delineate the mechanism underlying TGF-␤1-induced changes in osteoblast morphology, we next examined whether TGF-␤1 alters expression of major MMPs produced by osteoblasts, namely MMP-2 (gelatinase A), MMP-13 (collagenase-3), and MMP-14 (membrane-type 1 MMP). As judged by gelatin zymography, MC cells expressed proteolytic activity corresponding to the molecular mass of latent (72 kDa) but not active MMP-2 (67 kDa) (Fig. 5). Expression of latent MMP-2, which was found in both cell culture medium and cell

p38 MAP Kinase and MMPs Mediate TGF-␤ Effects on Bone Cells
lysate, was not altered by TGF-␤1 (Fig. 5). Likewise, MMP-14 expression, as visualized by immunoblotting, was unaffected by TGF-␤1 (Fig. 5). In contrast, using immunoblot analysis, we found that MMP-13, which was barely detectable in untreated cultures, was strongly up-regulated when cells were incubated with TGF-␤1 (Fig. 5). Most of the MMP-13 induced by TGF-␤1 was found in the cell culture medium at both 6 and 24 h after the addition of TGF-␤1, whereas only a small amount of MMP-13 was found in cell lysates and only 6 h after initiation of TGF-␤1 treatment (Fig. 5). Neither expression nor activation of MMP-13 induced by TGF-␤1 was influenced by PD169316 (not shown), indicating that p38 MAP kinase-independent signals are involved in the up-regulation of MMP-13 expression by TGF-␤1.
To investigate whether MMPs contribute to the regulation of morphology by TGF-␤1, cells were maintained in the presence of the broad-spectrum hydroxamate MMP inhibitor GM6001 (33). GM6001 reduced cell elongation and inhibited the increase in cell-free substratum area in response to TGF-␤1 by about half (Fig. 6, A and B). Similar results as those with GM6001 were also obtained with the hydroxamate-type MMP inhibitor BB-94 (34) and with the phosphinate-type MMP inhibitor 11A (35) (not shown). This effect of MMP inhibitors was specific, because the serine proteinase inhibitor aprotinin, the cysteine proteinase inhibitor E64, and the aspartyl proteinase inhibitor pepstatin were without effect on the TGF-␤1-induced changes in cell shape (Fig. 6, A and B). None of the proteinase inhibitors used affected cell number or cell shape in the absence of TGF-␤1 (not shown). GM6001 did not affect the induction by TGF-␤ of either the p38 or p44/42 MAP kinase (not shown). Thus, MMPs appear to play an important role in the TGF-␤1-induced shape changes.

p38 MAP Kinase and MMPs Do Not Impair the Effect of TGF-␤1 on Osteoblast Proliferation and Do Not Interfere with Changes in Osteoblast Morphology Induced by PTH and
bFGF-To examine whether p38 MAP kinases and MMPs are selectively involved in the regulation of cell shape by TGF-␤1, the ability of PD169316 and GM6001 to interfere with the inhibition of proliferation by TGF-␤1 was studied in serumcontaining medium. Under these conditions, Neither PD169316 nor GM6001 affected cell number in the absence or presence of TGF-␤1 (Fig. 7A). The same observation was also made when cells were maintained under conditions that induce osteoblast differentiation, i.e. in the presence of serum, ␤-glycerol phosphate, and ascorbic acid (not shown). Likewise, the decrease in alkaline phosphatase expression by TGF-␤1 that was observed in the latter condition was not modulated by the two inhibitors (not shown).
To determine whether p38 MAP kinase and MMPs are regulators of osteoblast morphology in general or are specifically required for TGF-␤1-induced shape changes, the effect of PD169316 and GM6001 on MC cells maintained in the pres-ence of PTH and basic fibroblast growth factor (bFGF, FGF-2) was examined. When osteoblasts were treated with PTH for 1 h, a subpopulation of cells adapted a stellate, retracted morphology (7) (Fig. 7B). Neither PD169316 nor GM6001 affected osteoblast retraction induced by PTH (Fig. 7C). Treatment of MC cells with bFGF resulted in elongated and retracted cells (Fig. 7D), a phenotypic transition that was still observed, albeit to a slightly reduced extent, in the presence of PD169316 and GM6001 (Fig. 7D).
TGF-␤1-induced Shape Changes Facilitate Osteoclastic Bone Resorption-To examine whether TGF-␤1-induced changes in osteoblast shape would increase bone resorption by osteoclasts, we developed an assay in which the bone lining cell layer shielding the bone surface from osteoclasts was mimicked by plating MC cells at confluence onto bone slices. Under these conditions, TGF-␤1-promoted cell elongation was even more pronounced than on tissue culture plastic (Fig. 8A). As expected, both PD16 and GM6001 inhibited this morphological transformation (Fig. 8A).
To examine whether the shape changes induced by TGF-␤1 in MC cells maintained on bone slices would facilitate bone resorption, cells were kept on bone slices for 1 day in the presence of TGF-␤1. Subsequently, osteoclasts from rabbit long bones were added to the bone slices, and incubation was continued for another 2 days in the absence of TGF-␤1. Under these conditions, pit formation was strongly enhanced by TGF-␤1 (Fig. 8B). Similar data were also obtained when TGF-␤1 was present during the whole 3-day culture period (Fig. 8C). Thus, when MC cells were presented to osteoclasts as a layer of bone lining cells, TGF-␤1 increased pit formation. The fact that TGF-␤1 increased pit formation not only when present during the whole culture period but also when it was present in the culture only during the time prior to addition of osteoclasts suggests that shape changes induced by TGF-␤1 in the bone lining MC cells prior to arrival of osteoclasts at the bone surface are sufficient to promote bone resorption. When MC cells were mixed with osteoclasts and seeded at the same time, TGF-␤1 did not affect pit formation (Fig. 8D). This suggests that altered gene expression of MC cells in response to TGF-␤1 does not affect pit formation, i.e. the effect of TGF-␤1 on pit formation appears to be due primarily to its effect on osteoblast morphology. GM6001 was without effect on pit formation when osteoclasts were mixed with MC cells at the time of plating, irrespective of the presence of TGF-␤1 (Fig. 8D). In contrast, GM6001 reduced pit formation to control levels when osteoclasts were added to MC cells that covered the bone surface and were exposed to TGF-␤1 (Fig. 8, B and C). No effect of GM6001 on pit formation was observed in the bone lining cell set-up when TGF-␤1 was not present. These data further support the idea that MMPs control osteoclastic bone resorption by inhibiting TGF-␤1-induced shape changes. PD169316, in contrast to GM6001, inhibited pit formation under all culture conditions (Fig. 8, B-D), presumably because of an effect both on bone lining cells and osteoclasts. This seems reasonable taking into account the ubiquitous distribution of the MAP kinases in question.
To further investigate whether TGF-␤1 increases pit formation in the presence of a monolayer of MC cells by altering cell shape rather than by stimulating the production of soluble factors by MC cells that may promote osteoclast activity, MC cells were seeded onto bone slices, incubated for 24 h with TGF-␤ and inhibitors, and then fixed with ethanol prior to the addition of osteoclasts. Also, this procedure was designed to sperate the effects of TGF-␤, PD169316, and GM6001 on osteoblasts from their possible direct effects on osteoclasts. Therefore, this assay is a study of the bone lining cells mediating osteoclast function. This treatment resulted in a 50% reduction of pit formation compared with bone slices that did not harbor MC cells (Fig. 8E). In cultures where MC cells plated onto bone slices were treated with TGF-␤1 prior to fixation and addition of osteoclasts, pit formation was restored to levels observed in the absence of MC cells (Fig. 8E). This effect was abolished when TGF-␤1 pretreatment was carried out in the additional presence of PD169316 or GM6001 (Fig. 8E). PD169316 inhibited pit formation to an extent that was below that observed in the absence of TGF-␤1. This may be because the MC cell monolayer on bone slices appeared even more compact in the presence of both TGF-␤1 and PD169316 than in the absence of both molecules. In conclusion, in the presence of a bone lining cell-like monolayer of MC cells, TGF-␤1 appears to promote bone resorption by increasing the area of the bone surface that is accessible to osteoclasts. DISCUSSION Regulation of cell shape is an important process that controls tissue morphogenesis, cell migration, differentiation, proliferation, and survival, presumably by regulating expression of genes central to these processes (28,37,38). Thus, it is possible that TGF-␤1-induced osteoblast elongation contributes to several aspects of TGF-␤1 function in bone, for instance the regulation of osteoblast differentiation and expression of proteins affecting osteoclast maturation. In this study, we have focused on the direct consequence of TGF-␤1-induced osteoblast elongation for the ability of osteoclasts to access a substratum covered by MC cells mimicking a layer of bone lining cells. Osteoblast elongation in response to TGF-␤1 was observed on both tissue culture plastic and bone slices and was accompanied by a severalfold increase in the cell-free substratum area. Furthermore, treatment of osteoblasts with TGF-␤1-promoted bone resorption by osteoclasts added to bone slices covered by a confluent layer of MC cells.
Several lines of evidence suggest that the increased bone resorption in response to TGF-␤1 is due to the shape alterations of osteoblasts and is not mediated through the direct action of TGF-␤1 on osteoclasts or by the production of osteoclast-activating factors produced by MC cells in response to TGF-␤1. First, TGF-␤1 promoted bone resorption only when osteoclasts were added to a preplated bone lining cell layer but not when an equivalent number of osteoblasts was mixed with the osteoclast preparation at the time of plating. Second, GM6001 only inhibited pit formation in the presence of TGF-␤1 when added to cultures where bone lining cells were present but not when osteoclasts and MC cells were mixed at the time of plating. If GM6001 had had a direct effect on osteoclasts, or had influenced MC cells to produce factors affecting osteoclasts, then GM6001 should also have inhibited pit formation when added to cultures where MC cells and osteoclasts were seeded at the same time. Thus, GM6001 is likely to have affected pit formation by reversing the effect of TGF-␤1 on cell shape. Third, when bone lining cells were fixed after TGF-␤1 treatment and osteoclasts were subsequently plated and maintained in the absence of TGF-␤1, bone resorption was increased to the same extent as observed with unfixed bone lining cells. Taken together, these results strongly suggest that a major mechanism by which TGF-␤1 promotes osteoclastic bone re-sorption in the presence of bone lining cells is the alteration of osteoblast shape, i.e. the increase in cell-free substratum, which promotes attachment of osteoclasts to bone. Thus, the bone-lining cell layer orchestrates osteoclastic bone resorption by altering osteoblast shape. This conclusion is also strengthened by observations that TGF-␤ increases expression of osteoprotegerin and decreases expression of osteoclast differentiation factor in osteoblasts (39), which would be expected to inhibit rather than promote bone resorption. Furthermore, TGF-␤1 may have a negative impact on osteoclast survival (40), which likewise would inhibit bone resorption.
We found that TGF-␤1 in MC cells activates both Erk1/2 and p38 MAP kinase but not JNK. It has been demonstrated in epithelial cells that TGF-␤ can stimulate all three MAP kinase pathways and that they mediate some of the effects of TGF-␤ on gene expression (11). Activation of Erk1/2 by TGF-␤ occurs through Ras, which activates Raf-1, which then phosphorylates the Erk1/2 kinase MEK1/2 (41)(42)(43). Stimulation of the p38 MAP kinase pathway in yeast occurs through activation of the MAP kinase kinase kinase family member TAK1 (44). The exact mechanism of TAK1 activation is unknown but presumably involves TAB1, a protein that binds to TAK1 and enhances its kinase activity (45). TAK1 potentially displays two-pronged action by its ability to phosphorylate two different sets of MAP kinase kinases, MKK3 and MKK6, which activate p38 MAP kinase (46), and MKK4, which can activate both p38 MAP kinase and JNK The asterisk indicates significant differences (p Ͻ 0.05, Student's t test) in the presence of TGF-␤1 as compared with the same culture condition but in the absence of TGF-␤1. C, same as in B, except that TGF-␤1 was present for the whole 3-day culture period. D, same as in B, except that MC cells were mixed with osteoclast preparations at the time of plating. E, MC cells were maintained for 1 day on bone slices in the absence of TGF-␤1 (Ϫ, ϩBLC), or in the presence of TGF-␤1 (Ϫ, ϩ TGF-␤), TGF-␤1 and PD169316 (P), or TGF-␤1 and GM6001 (G). Bone slices were then fixed and washed before the addition of osteoclast preparations and were maintained for 2 days in medium without additives. Osteoclast preparations were also added to fixed bone slices that had been preincubated with medium in the absence of MC cells (Ϫ, left bar). The results of pit counting are normalized with values obtained for the area of pits formed in the presence of MC cells and with medium without additives set to 100.
p38 MAP Kinase and MMPs Mediate TGF-␤ Effects on Bone Cells (47,48). Because we did not observe activation of JNK between 5 min and 24 h after TGF-␤1 stimulation, it is unlikely that MKK4 is a downstream target of TGF-␤1 signaling in MC cells.
TGF-␤1-induced osteoblast elongation was dependent on p38 MAP kinase activity, as assessed by using the p38 MAP kinase inhibitor PD169316. To our knowledge, this is the first report functionally linking the p38 MAP kinase pathway to TGF-␤1 in osteoblasts. The p38 MAP kinase in osteoblasts is also activated by bFGF, where it mediates the up-regulation of alkaline phosphatase activity in response to epinephrine (49) and may, together with JNK, play a role in the regulation of MMP-1 promoter activity (50,51). Interestingly, osteoblast shape changes induced by bFGF were only slightly inhibited by PD169316, indicating that TGF-␤1 and bFGF alter osteoblast morphology by using different intracellular signaling cascades. This finding is also consistent with the fact that the morphology of osteoblasts in response to TGF-␤1 is obviously different from the morphology of osteoblasts treated with bFGF, i.e. cells treated with bFGF are more elongated than cells treated with TGF-␤1 and often display a rather stellate morphology.
Although in many experimental models the main function of p38 MAP kinase, similar to JNK, appears to be activation of stress-regulated genes, it is becoming increasingly evident that p38 MAP kinase also controls other important events such as differentiation, apoptosis, and cell migration. The latter point, which has been shown for smooth muscle cells, endothelial cells, and neutrophils, (52)(53)(54), is of particular importance because cell migration and shape changes are likely to use an overlapping cellular machinery. Interestingly, as in the case of TGF-␤1-induced osteoblast elongation, migration of endothelial cells and neutrophils was not affected by MEK1/2 inhibition (53,54). We found that PD169316 completely inhibited migration of osteoblasts cells regardless of whether TGF-␤1 was present (not shown). Thus, p38 MAP kinase can regulate both cell motility and shape in osteoblasts.
In contrast to p38 MAP kinase, the Erk1/2 pathway was not involved in regulating osteoblast morphology in response to TGF-␤1. The transient nature of Erk1/2 stimulation in MC cells in response to TGF-␤1 suggests that it mediates transient changes in gene expression that may be relevant for other functions of TGF-␤1 in bone metabolism. Erk1/2 activity in osteoblasts was reported to be increased by a variety of growth factors and hormones, e.g. interleukin-6, bFGF platelet-derived growth factor, and estrogen (55)(56)(57)(58), and activation of Erk1/2 has been linked to the ability of epidermal growth factor-1 and insulin-like growth factor-1 to stimulate osteoblast proliferation (59,60). We observed no effect of PD98059 on the growth of MC cells in the absence of TGF-␤1 or in its presence when growth is inhibited (not shown).
Elongation of MC cells in response to TGF-␤1 was reduced on tissue culture plastic and virtually abolished on bone slices when cells were cultured in the presence of the MMP inhibitor GM6001. Furthermore, TGF-␤1 up-regulated expression of MMP-13, further indicating a role of MMPs in osteoblast cell shape changes. However, PD169316 did not inhibit the increase in MMP-13 expression in response to TGF-␤1 (data not shown), which suggests that MMP-13 is not in a direct linear pathway downstream of p38 MAP kinase.
The ability of TGF-␤1 to alter cell shape is not confined to osteoblasts, but can also be observed in epithelial-mesenchymal transitions during normal development and progression of tumor cells toward an invasive phenotype (61,62). As we have shown for the morphological transformation of osteoblasts in response to TGF-␤1, epithelial-mesenchymal transitions involve changes in cell shape from cobblestone to mesenchymal/ elongated and more abundant actin filaments. Similarly, both processes are accompanied by increased expression of focal adhesion kinase, which regulates events such as adhesion and motility at the cell-substratum interphase (63). Furthermore, as in the case of osteoblast elongation in response to TGF-␤1, matrix metalloproteinases are able to regulate the morphological transformation during epithelial-mesenchymal transition (23). Thus, the mechanisms leading to epithelial-mesenchymal transition and TGF-␤1-induced osteoblast elongation appear to share critical molecular features.
At present, it is not clear in which situations TGF-␤ could facilitate bone resorption in vivo by altering osteoblast shape. One possibility is that active TGF-␤ is produced during bone remodeling within the basic multicellular unit (BMU) of the osteon. Both osteoblasts and osteoclasts are able to synthesize TGF-␤ (64 -66). Although osteoblasts are presumably responsible for production of most of the TGF-␤ entrapped in bone matrix (67), the majority of active TGF-␤ appears to be generated by bone resorbing osteoclasts either through release of TGF-␤ stored in bone matrix or by biosynthesis of new TGF-␤ (68). Active TGF-␤ may then be made accessible to osteoblasts through a process involving transcytosis (69). Interestingly, most of the immunodetectable TGF-␤ in bone appears to be located in the vicinity of osteoclasts (70). The selective activation of TGF-␤ by osteoclasts within the BMU may fulfill two functions. First, by inducing elongation of bone lining cells, it may propel the directional movement of osteoclasts within the BMU as well as the recruitment of more osteoclasts at the front of the BMU. Second, the chemotactic action of TGF-␤ on osteoblasts (71,72) would ensure that osteoblasts/osteoblast precursors would accumulate at the bone-forming rear of the BMU. In addition to modulating the activity of existing BMUs, the action of TGF-␤1 to increase cell-free bone surface may lead to initiation points for new sites of bone resorption or may promote bone resorption in general in response to a variety of physiological signals. Thus, although the precise role and mechanism of TGF-␤ with regard to osteoblast elongation remains to be established, our results provide the first paradigm for the physiological consequences of alterations in the morphology of osteoblasts lining the bone surface which thereby are orchestrating osteoclast function.