Proteolysis of Latent Transforming Growth Factor-β (TGF-β)-binding Protein-1 by Osteoclasts A CELLULAR MECHANISM FOR RELEASE OF TGF-β FROM BONE MATRIX

The binding of growth factors to the extracellular matrix (ECM) may be a key pathway for regulation of their activity. We have shown that a major mechanism for storage of transforming growth factor-β (TGF-β) in bone ECM is via its association with latent TGF-β-binding protein-1 (LTBP1). Although proteolytic cleavage of LTBP1 has been reported, it remains unclear whether this represents a physiological mechanism for release of matrix-bound TGF-β. Here we examined the role of LTBP1 in cell-mediated release of TGF-β from bone ECM. We first characterized the soluble and ECM-bound forms of latent TGF-β produced by primary osteoblasts. Next, we examined release of ECM-bound TGF-β by bone resorbing cells. Isolated avian osteoclasts and rabbit bone marrow-derived osteoclasts released bone matrix-bound TGF-β via LTBP1 cleavage. 1,25-Dihydroxyvitamin D3 enhanced LTBP1 cleavage, resulting in release of 90% of the ECM-bound LTBP1. In contrast, osteoblasts failed to cleave LTBP1 or release TGF-β from bone ECM. Cleavage of LTBP1 by avian osteoclasts was inhibited by serine protease and metalloproteinase (MMP) inhibitors. Studies using purified proteases showed that plasmin, elastase, MMP2, and MMP9 were able to cleave LTBP1 to produce 125–165-kDa fragments. These studies identify LTBP1 as a novel substrate for MMPs and provide the first demonstration that LTBP1 proteolysis may be a physiological mechanism for release of TGF-β from ECM-bound stores, potentially the first step in the pathway by which matrix-bound TGF-β is rendered active.

It has long been thought that growth factors released from bone matrix play important roles in the coupling of bone resorption to bone formation and in repair processes such as fracture healing. However, at present little is known about the mechanisms by which growth factors are stored in the ECM, the ECM proteins with which they interact, or the molecular mechanisms by which they are released.
Bone ECM is the major storage site in the body for TGF-␤ (2,3). This ECM-bound TGF-␤, which is predominantly the TGF-␤1 isoform, is stored in a latent form and can be released and activated by resorbing osteoclasts (4). Once released from the matrix and activated, TGF-␤ can influence many of the steps in the remodeling pathway, including inhibition of osteoclast formation and activity, stimulation of recruitment and proliferation of osteoblast precursors, and stimulation of mature osteoblasts to produce bone matrix proteins (reviewed in Ref. 5). TGF-␤ has therefore been implicated as a coupling factor that coordinates the processes of bone resorption and subsequent bone formation. Although it is known that bone matrix-bound TGF-␤ can be released by osteoclasts (4) and that osteoclasts are capable of activating both matrix-derived and exogenously added latent TGF-␤ (6 -8), the molecular mechanism(s) by which TGF-␤ is released from ECM-bound stores remain unclear.
We have shown previously (9, 10) that a major mechanism for storage of latent TGF-␤ in bone matrix is via its association with the latent transforming growth factor-␤-binding protein-1 (LTBP1). LTBP1 is a member of an emerging superfamily of ECM proteins, which includes fibrillins 1 and 2 and LTBPs 1-4 (reviewed in Refs. 11 and 12). LTBP1 was originally identified as a component of the large latent TGF-␤1 complex (13,14). This complex consists of the 25-kDa mature TGF-␤ homodimer, which is cleaved from but remains non-covalently associated with a 75-kDa portion of the propeptide homodimer (also known as latency-associated peptide or LAP). The propeptide is then disulfide-linked to the 190-kDa LTBP1 (13,15). Although most cells produce latent TGF-␤ in the large latent complex, a small latent complex that lacks LTBP1 and consists of TGF-␤ non-covalently associated with its propeptide has also been reported as a naturally occurring form in bone and kidney (5,16).
LTBP1 facilitates secretion of latent TGF-␤ (17) and may also modulate activation of latent TGF-␤ (18). More recently it has been described as a stable component of the extracellular matrix that is important in storage of latent TGF-␤ in the matrix and may be a structural component of connective tissue microfibrils (9,10,19,20). Several studies have suggested that in addition to its role in storage of latent TGF-␤ in the ECM, LTBP1 may function as a vehicle for release of latent TGF-␤ from matrix-bound stores. Thus, purified proteases, such as plasmin and elastase, have been shown to release TGF-␤ from the matrix of fibroblasts and osteoblasts via cleavage of LTBP1 (9,21). However, it is unclear whether this is a phenomenon associated with pathologically high concentrations of proteases or whether proteolytic cleavage of LTBP1 represents a true physiologic mechanism for release of ECM-bound TGF-␤.
In the present study, we characterized the forms of latent TGF-␤ produced by primary bone cells and stored in the ECM. Next, by using osteoclast cell culture systems we investigated the molecular mechanism(s) by which TGF-␤ is released from bone ECM-bound stores during bone resorption. In particular, we examined whether proteolytic cleavage of LTBP1 may be a potential cellular mechanism for the release of TGF-␤ bound to bone ECM, and we determined whether proteases known to be important in osteoclast function may be involved in this process.
Antibodies-Two different rabbit polyclonal antibodies against LTBP1 were used: Ab39 that recognizes mouse, rat, and human LTBP1 (kindly supplied by K. Miyazono, Japanese Foundation for Cancer Research, Tokyo, Japan), and a second antibody, termed the "LTBP1 hinge antibody," raised against a synthetic peptide corresponding to residues 721-744 of the rat LTBP1 sequence (GenBank TM accession number M55431) that recognizes mouse and rat LTBP1. The specificity of these antibodies has been described elsewhere (10,13,20). A mouse monoclonal antibody against human LTBP1, as well as neutralizing antibodies to TGF-␤1 and -␤2, a pan-specific TGF-␤-neutralizing antibody, and a goat polyclonal antibody against latent TGF-␤1 propeptide were purchased from R & D systems (Minneapolis, MN). The detection antibody for Western blotting was a peroxidase-conjugated donkey anti-rabbit (Amersham Biosciences).
Cell Culture-Tissue culture reagents were purchased from Invitrogen or JRH Biosciences (Lenexa, KS). Fetal rat calvarial osteoblasts (FRC) were isolated as described previously (9,10). The cells were stored for up to 1 year using standard cryopreservation procedures and used for experiments after one further passage. UMR-106 cells were a gift from T. J. Martin (St. Vincent Institute of Medical Research, Fitzroy, Victoria, Australia). TMLC-C32 cells were a gift from D. B. Rifkin (New York University, New York).
Rabbit marrow cells were isolated using a modification of the method of Tezuka et al. (22). Femora, humeri, ulnae, and radii of 11-day-old New Zealand White rabbits were removed. Connective tissues were dissected away, and the bones were minced in ␣-modified minimal essential medium (␣MEM). The bone pieces were vortexed to dissociate the cells, and bone particles were removed by sedimentation under gravity. The supernatant was centrifuged, and the cells were resuspended in ␣MEM supplemented with 10% fetal bovine serum (FBS), 2 mM L-glutamine (LG), and 100 units/ml penicillin/streptomycin (P/S). The cells were then seeded into 12-well plates coated with 35 S-labeled bone ECM (see below).
Avian osteoclast precursors were isolated as described elsewhere from the medullary bone of egg-laying White Leghorn hens fed on a low calcium diet (23). The cells were plated in 150-mm Petri dishes as described elsewhere (23), and after overnight incubation, adherent cells were harvested by treatment with 2 mM EDTA in PBS for 5 min at 37°C. The cells were then replated into 12-well plates coated with 35 S-labeled bone ECM (see below) in ␣MEM containing 2.5% FBS, 2.5% chicken serum, 2 mM LG, 100 units/ml P/S, and 6 g/ml arabinose-␤-D-cytosine furanoside. These cells have been shown to fuse and become multinucleated in the presence of 1,25-D 3 (23) (see also Fig. 5, e and f).
The cells were cultured for up to 6 days, with or without 1,25-D 3 and/or protease inhibitors as described below. Culture media were changed every 2 days.
Fast Pressure Liquid Chromatography (FPLC) Analysis-FPLC fractionation was performed as described previously (24). FRC cells were grown in 150-cm 2 flasks in ␣MEM supplemented with 10% FBS, 2 mM LG, 100 units/ml P/S, and 30 g/ml gentamycin. At confluence the media were changed to ␣MEM, supplemented as above but with the addition of 50 g/ml ascorbic acid and 3 mM ␤-glycerophosphate (␤-GP) and the reduction of the serum to 5%. Thereafter the media were changed every 3 days. Conditioned medium was harvested from four time points representing sequential stages in the in vitro differentiation of FRC cells to form mineralized bone-like nodules. The "pre-confluent" stage was from proliferating cultures at 90% confluence; the "early post-confluent" stage was from 2-day post-confluent cultures; the "nodule-forming" stage was from 6-day post-confluent cultures in which multilayered cellular nodules had formed, but were not yet mineralized; and the "mineralization" stage was from 14-day post-confluent cultures in which mineralized bone nodules were present (see Fig. 1, inset micrographs). For collection of conditioned media, phenol red-free Dulbecco's modified Eagle's medium was used containing 2 mM LG, 100 units/ml P/S, 0.1% bovine serum albumin, 50 g/ml ascorbic acid. 25 ml of conditioned medium was collected per flask over a 48-h culture period. A total of 150 ml of conditioned medium per time point was concentrated 10-fold over a 50-kDa cut-off membrane using a minisette concentrator (Millipore, Bedford, MA), and the samples were lyophilized and reconstituted with 1-2 ml of distilled water, pH 7.2. They were then dialyzed against 20 mM tris buffer, pH 7.2, and applied to an analytical Mono-Q anion exchange column (Amersham Biosciences). The column was eluted with a linear gradient of 0 -0.5 M NaCl/20 mM tris buffer. Fractions were tested for TGF-␤ activity using the alkaline phosphatase microassay as described below.
TGF-␤ Measurement-TGF-␤ was measured as described previously by using either the ROS 17/2.8 microassay (24), which measures stimulation of alkaline phosphatase activity by TGF-␤ in ROS 17/2.8 osteosarcoma cells, or by using the mink lung epithelial cell luciferase bioassay (25), which measures stimulation of activity of the plasminogen activator inhibitor-1 promoter by TGF-␤. To determine the total (active ϩ latent) TGF-␤ levels, the samples were acidified to pH 2 using 1 M HCl and then reneutralized using 1 M NaOH immediately before addition to the assay plate. Latent TGF-␤ values were determined by subtracting active TGF-␤ measurements from total TGF-␤.
Pulse-Chase Metabolic Labeling and Immunoprecipitation-To examine further the secreted and matrix-bound forms of latent TGF-␤ produced by FRC cells, pulse-chase metabolic labeling and immunoprecipitation were performed. Cells were plated into 12-well multiwell plates at 10,000 cells/cm 2 growth area. At 90% confluence, the cells were washed twice in PBS and incubated for 1 h in cysteine-free ␣MEM supplemented with 5% dialyzed FBS, 2 mM LG, 100 units/ml P/S, 50 g/ml ascorbic acid. The cells were then labeled for 30 min using 100 Ci/well [ 35 S]cysteine (Amersham Biosciences) in cysteine-free ␣MEM supplemented with 5% dialyzed FBS and additives as above. This was followed by a "cold chase" in complete ␣MEM (containing 0.1 mg/ml L-cysteine), supplemented with 5% FBS and additives as above for a time course of 15 and 30 min, and 2, 6, 24, and 48 h.
Conditioned media were harvested and protease inhibitors added (1 mM phenylmethylsulfonyl fluoride, 50 g/ml aprotinin, 1 M pepstatin A, 10 M leupeptin). The cells ϩ matrix were washed twice in ice-cold PBS and then lysed in ice-cold radioimmunoprecipitation buffer (RIPA) (50 mM Tris, 150 mM NaCl, 1% Nonidet P-40, 0.5% sodium deoxycholate). The deoxycholate-insoluble material (essentially matrix) was washed two more times in ice-cold PBS, transferred to a 1.5-ml tube, and centrifuged for 5 min at 14,000 rpm. The pellet was then digested for 2 h at 37°C on an end-over rotator in 500 l of a solution consisting of 0.2 units/ml plasmin in plasmin digestion buffer (see under "Protease Digestions"). This digestion releases any LTBP1 that may still be bound to the ECM (9,21).
LTBP1 and TGF-␤ in the supernatant and in the plasmin digest of the matrix were determined by immunoprecipitation followed by SDS-PAGE and autoradiography as described previously (9, 24) using a rabbit polyclonal antibody specific for LTBP1 or using a goat polyclonal antibody specific for LAP.
Preparation of Bone ECM-To prepare bone ECM for culturing with osteoclasts, FRC cells were plated into 12-well plates as described above. At confluence, the medium was changed to ␣MEM supplemented with 5% FBS, 2 mM LG, 100 units/ml P/S, 30 g/ml gentamycin, 50 g/ml ascorbic acid, and 3 mM ␤-GP. Thereafter, the culture media were changed every 3 days. Cultures were maintained for 10 -14 days, by which time mineralized bone nodules could be seen, and a thick ECM layer had formed. For preparation of ECM-coated plates, a modification of the method of Jones et al. (26) was used. The cells were washed twice in ice-cold PBS and lysed in PBS containing 0.5% Triton X-100 for 10 min at 4°C. This leaves a membrane-like layer of bone ECM adhered to the bottom of the culture well. The matrix layer was then washed twice in ice-cold PBS and treated with ice-cold 50 mM ammonium acetate, pH 7.5, for 10 min at 4°C (the pH was reduced from 9.0, as described in Jones et al. (26), to prevent activation of TGF-␤ by the alkaline pH). The matrix layers were then washed 4 times in PBS, air-dried, and stored at Ϫ70°C.
Measurement of Release of LTBP1 and TGF-␤ from [ 35 S]Cysteinelabeled Bone ECM-An immunoprecipitation-based assay was developed for examining release of bone matrix-bound LTBP1 and latent TGF-␤ by cells cultured on bone ECM (27). FRC cells were cultured in 12-well plates as described above until 10 -14 days of culture. The cells were then labeled for 48 h using 100 Ci/well [ 35 S]cysteine in ␣MEM containing one-fifth the normal cysteine content and supplemented with 5% dialyzed FBS, 50 g/ml ascorbic acid, 3 mM ␤-GP, 100 units/ml P/S, 2 mM LG. The labeled bone ECM was prepared as described above.
To examine release of LTBP1/TGF-␤ by osteoclasts, two sources of osteoclast precursors were used, rabbit marrow cells and avian osteoclasts. The cells were plated onto the radiolabeled ECM at 2.5 ϫ 10 6 cells/well (rabbit marrow cells) or 5 ϫ 10 6 cells/well (avian osteoclast precursors) in 2 ml of ␣MEM supplemented as described above. The cells were incubated with or without addition of 1,25-D 3 , PTHrP, or protease inhibitors as specified under "Results." Release of LTBP1 and latent TGF-␤ was monitored by immunoprecipitating the labeled protein fragments released into the culture media using anti-LTBP1 antibodies (the LTBP1 antibodies will precipitate both free LTBP1 and LTBP1 complexed to latent TGF-␤). Because the osteoclasts themselves were never exposed to [ 35 S]cysteine, any radiolabeled proteins present in the culture supernatants were assumed to have been released from the pre-labeled matrix. The supernatant was transferred to a fresh tube and protease inhibitors added as described above. The deoxycholateinsoluble pellet was digested in 0.2 units/ml plasmin as described above. LTBP1 in the supernatant and in the plasmin digest of the matrix was then determined by immunoprecipitation with anti-LTBP1 antibodies, as described above.
Western Blotting-Samples were separated on SDS-PAGE using 4 -20% gradient gels, transferred onto nitrocellulose, and immunoblotted as described previously (9,10). The immunostained bands were visualized using the ECL detection system according to manufacturer's instructions (Amersham Biosciences). The immunostained Western blots were then exposed to X-Omat AR film (Eastman Kodak).
Immunofluorescent Staining-For immunofluorescent staining, FRC cells were cultured on lab-tek 8-chamber slides at 10,000 cells/cm 2 growth area and maintained for 10 -14 days as described above. ECM was prepared as described above and incubated with or without proteases as described above. After incubation in proteases, the ECM layers were fixed in 95% ethanol and stained by immunofluorescence as described previously (10).

Characterization of Secreted Latent TGF-␤ Complexes Produced by Primary Bone Cell
Cultures-As matrix laid down by primary osteoblast cultures was to be used to examine release of bone matrix-bound TGF-␤ by osteoclasts, it was first necessary to characterize the latent TGF-␤ complexes produced by the osteoblast cultures and determine what proportion of the latent TGF-␤ was complexed to LTBP1. Fig. 1 shows results from Mono-Q FPLC analysis of conditioned media samples from FRC cultures at the pre-confluent, early post-confluent, nodule-forming, and mineralizing stages (see "Materials and Methods"). The insets in Fig. 1 show the appearance of the cultures at each of these stages. Two major peaks of latent TGF-␤ activity were observed, one eluting at 0.22 M NaCl (peak II), which is the expected elution position for the 100-kDa small latent TGF-␤ complex (16,24), and the other eluting at 0.3 M NaCl (peak III), which corresponds to the 290-kDa large latent TGF-␤ complex, containing LTBP1 (16,24). The pre-confluent and early post-confluent cultures produced predominantly the small latent TGF-␤ complex (peak II). However, with maturation in culture, the cells switched to producing larger amounts of the 290-kDa (LTBP1-containing) complex (peak III), which made up ϳ40% of the total latent TGF-␤ secreted. In addition, a minor peak eluting at 0.05 M NaCl (peak I) was also observed in nodule-forming and mineralizing cultures. At present the nature of this peak is unknown. Antibody neutralization studies indicated that the latent TGF-␤ activity was predominantly TGF-␤1 in all three peaks with smaller amounts of TGF-␤2 (data not shown). In the pre-confluent and early post-confluent cultures, peak III showed ϳ20 -30% TGF-␤2 activity, in contrast to the nodule-forming and mineralizing cultures, which showed no detectable TGF-␤2 activity in this peak (data not shown).
Analysis of Matrix-bound Latent TGF-␤ in Primary Bone Cell Cultures-LTBP1 has been shown to be cross-linked in the ECM via the action of transglutaminase (28) and is highly insoluble. Previous studies have shown that proteases such as plasmin and elastase can release proteolytic fragments of LTBP1 from the ECM (9,21) and that plasmin can activate latent TGF-␤ (29,30). These serine proteases were therefore used to examine the ECM-bound forms of latent TGF-␤ in primary bone cell cultures (see Fig. 2). Treatment of 35 S-labeled bone ECM with plasmin or elastase resulted in release of radiolabeled products, which could be immunoprecipitated with antibodies against LTBP1 (see Fig. 2a). Plasmin released a doublet at ϳ110 -130 kDa (black arrowhead), corresponding to cleaved fragments of free LTBP1, as well as a minor band at ϳ230 kDa (white arrowhead), which is the expected size of the large latent TGF-␤ complex containing cleaved LTBP1. Elastase released similar fragments but appeared to cleave LTBP1 at a different site, as evidenced by a faster migration. As expected, in both cases the upper band, corresponding to the large latent TGF-␤ complex, disappeared under reducing conditions (data not shown). To confirm that the upper band at 230-kDa contained TGF-␤, supernatants from plasmin-digested bone ECM were immunoprecipitated with antibodies to LTBP1 and to the latent TGF-␤1 propeptide (also known as latency-associated peptide or LAP) to demonstrate co-migration (see Fig. 2b). Immunoprecipitation with anti-LAP antibodies produced a band at ϳ230 kDa (white arrowhead), which exactly co-migrated with the upper band precipitated with anti-LTBP1. An additional band was observed at ϳ200 kDa, presumably representing latent TGF-␤ bound to an as yet unidentified protein.
To determine whether plasmin and elastase released both latent and active TGF-␤ from bone ECM, the TGF-␤ concentrations in the plasmin and elastase digests were measured by bioassay (see Fig. 2, c and d). Both plasmin and elastase released latent TGF-␤ from bone ECM (i.e. detectable after acid activation). Plasmin also released detectable levels of active TGF-␤. However, this active TGF-␤ represented less than 10% of the total TGF-␤ released. In contrast to plasmin, elastase did not release any detectable active TGF-␤ from bone ECM (Fig.  2d). The doses of plasmin and elastase used are sufficient to abolish completely the LTBP1 fibrillar staining in bone matrix preparations from fetal rat calvarial cells (9). 2 To analyze further the production of LTBP1-bound forms of latent TGF-␤ in primary bone cell cultures, pulse-chase immunoprecipitation experiments were performed (see Fig. 3). Immunoprecipitation analysis, using a 30-min pulse of [ 35 S]cysteine labeling, indicated that significant amounts of LTBP1 were secreted into the culture media of FRC cells after only 15 min of chase (Fig. 3a). A major band at 170 -190 kDa was observed (black arrowhead), corresponding to free LTBP1. This band represented about 90% of the immunoprecipitable LTBP1. In addition a minor band at ϳ290 kDa was observed (white arrowhead), representing LTBP1 complexed to TGF-␤. These bands increased in intensity up to 6 h. However, by 24 and 48 h an additional lower molecular weight band was evident at ϳ100 kDa (gray arrow), which presumably represents a proteolytically processed form of LTBP1.
To detect LTBP1 that is cross-linked in the ECM, the matrix was digested with plasmin (see above). This resulted in release of proteolytic fragments of free LTBP1 of 110 -130 kDa (Fig. 3c,  black arrowheads), and a band representing cleaved LTBP1 complexed to TGF-␤ at 230 kDa (Fig. 3c, white arrowhead). These experiments indicated that not only was LTBP1 rapidly secreted but it was also rapidly incorporated into the matrix, with significant amounts of free LTBP1 and large latent complex detected in the ECM after only 15 min of chase. Again the intensity of these bands increased up to 6 h in culture, after which no significant further increase was observed.
Release of LTBP1 and LTBP1 Complexed with Latent TGF-␤ by Rabbit Marrow Cells-To determine whether proteolytic cleavage of LTBP1 could be a potential cellular mechanism for release of matrix-bound latent TGF-␤, experiments were per-2 S. L. Dallas, unpublished observations. formed using cultures of neonatal rabbit marrow cells. These cultures consist of a mixed population of cells that is highly enriched for mature osteoclasts but also contain some stromal cells and cells in the osteoblastic lineage, which can support and activate the osteoclasts. When neonatal rabbit marrow cells were cultured on [ 35 S]cysteine-labeled bone ECM, specific radiolabeled products were released into the culture medium, which could be immunoprecipitated with antibodies against LTBP1. Fig. 4, a and b, shows radiolabeled fragments released into the culture medium during 0 -24 h and 24 -48 h of culture. In control cultures (NC) in which no cells were seeded onto the labeled matrix, a very small amount of LTBP1 was detected in the medium, presumably representing passive diffusion of LTBP1 from the ECM and/or low level degradation by serum components in the medium. In contrast, when rabbit marrow cells were cultured on the labeled ECM (C), a major band of ϳ130 kDa, corresponding to a cleaved fragment of LTBP1 (black arrowhead), and a minor band of ϳ230 kDa, corresponding to cleaved LTBP1 complexed to small latent TGF-␤ (white arrowhead), were released into the culture medium. Release was approximately equivalent during the 0 -24-and 24 -48-h culture periods. Treatment with PTHrP, a stimulator of osteoclast activity through its action on osteoblasts and stromal cells, produced a slight stimulation in release of LTBP1 and LTBP1 complexed to latent TGF-␤, which was most evident in the first 24-h culture period. The serine protease inhibitor, aprotinin (AP) completely blocked release of LTBP1 under both control and PTHrP-stimulated conditions. As expected, release of LTBP1 and LTBP1/TGF-␤ by rabbit marrow cells into the culture medium was concomitant with a reduction in LTBP1 bound in the matrix (see Fig. 4c).
Osteoclasts but Not Osteoblasts Release LTBP1 from Bone ECM-To rule out that osteoblasts and/or stromal cells in the rabbit marrow cultures were releasing LTBP1 from bone ECM and to verify that osteoclasts were responsible for release of LTBP1 and large latent TGF-␤ complex, experiments were performed using pure populations of osteoblast-like cells followed by the use of relatively pure populations of osteoclasts derived from avian osteoclast precursors. Primary cultures of fetal rat calvarial osteoblasts as well as the rat osteosarcoma cell line, UMR-106, failed to release LTBP1 and large latent TGF-␤ complex from bone matrix, either under control or PTHrP-stimulated conditions (data not shown).
In contrast, when avian osteoclast precursors were cultured on radiolabeled bone ECM, cleaved fragments of LTBP1 were released into the culture medium (see Fig. 5, a-d). These avian osteoclast precursors fuse in culture over a 6-day period to form highly purified populations of multinucleated cells (Ͼ95% pure), which are positive for tartrate-resistant acid phosphatase (TRAP) and are capable of resorbing bone (8,23) . The avian osteoclast cultures released LTBP1 under control, unstimulated conditions (C) (see Fig. 5a, black arrowheads). This release occurred at a low level throughout the entire 6-day culture period and was most evident as a dramatic decrease in the amount of ECM-bound LTBP1 and LTBP1/TGF-␤ at the end of the culture period (Fig. 5b). Treatment of the avian osteoclast cultures with 1,25-D 3 , which enhances fusion of precursors and stimulates osteoclastic activity (23), resulted in a marked stimulation in release of LTBP1 and LTBP1/TGF-␤ from bone ECM. Release was maximal between days 3 and 4 of culture and resulted in the release of essentially all the ECMbound LTBP1 and LTBP1/TGF-␤ into the culture medium by the end of the 6-day culture period (Fig. 5b). For subsequent experiments, the media samples were therefore collected over days 1-4 of culture. Fig. 5, e and f, shows photomicrographs of the avian osteoclasts formed after 6 days of culture on bone ECM under control culture conditions (Fig. 5e) and in the presence of 1,25-D 3 (Fig. 5f). Note that in the presence of 1,25-D 3 , fusion of the precursors to form multinucleated osteoclasts is enhanced, and TRAP staining is much more intense.
To confirm that the avian osteoclasts were releasing TGF-␤ from the bone matrix, the TGF-␤ content of the bone ECM was measured after culturing with avian osteoclasts for 6 days (see Fig. 5g). Bone ECM that had been cultured with avian osteoclasts showed an ϳ60% reduction in TGF-␤ content compared with controls cultured without cells. Stimulation of the avian osteoclasts with 10 Ϫ8 M 1,25-D 3 further enhanced this release, resulting in the loss of ϳ90% of the matrix-bound TGF-␤.
Dose-response experiments indicated a dramatic stimulation in release of ECM-bound LTBP1 by avian osteoclast precursors treated with doses of 1,25-D 3 as low as 10 Ϫ10 M compared with unstimulated controls (see Fig. 5, h and i). Maximal release was seen between 10 Ϫ8 and 10 Ϫ7 M and subsequent experiments were performed using the 10 Ϫ8 M dose of 1,25-D 3 .

Effects of Protease Inhibitors on Release of Bone Matrixbound LTBP1 and LTBP1 Complexed to Latent TGF-␤ by
Avian Osteoclasts-The osteoclast culture systems we have used produce complex mixtures of proteases including cathepsins, plasminogen activators, and matrix metalloproteinases (MMPs) (reviewed in Ref. 31). In particular, osteoclasts express high levels of cathepsin K (32) and MMP9 (33,34). To investigate which protease(s) were responsible for cleavage of LTBP1, a number of protease inhibitors were examined (see Fig. 6). 35

S-labeled bone ECM by rabbit marrow cells as demonstrated by immunoprecipitation using anti-LTBP1 antibodies.
The samples are as follows: NC, control with no cells; C, rabbit marrow cells; PTHrP, rabbit marrow cells treated with 100 ng/ml PTHrP; AP, rabbit marrow cells treated with 50 g/ml aprotinin; PTHrP ϩ AP, rabbit marrow cells treated with 100 ng/ml PTHrP and 50 g/ml aprotinin. Gels a-c show samples immunoprecipitated with antiserum specific for LTBP1, and gels d-f show corresponding control samples immunoprecipitated with non-immune rabbit serum. a shows cleaved LTBP1 (black arrowhead) and large latent TGF-␤ complex (white arrowhead) released into the culture medium by rabbit marrow cells during 0 -24 h of culture. b shows cleaved LTBP1 and large latent TGF-␤ complex released into the culture medium during 24 -48 h of culture. c shows LTBP1 and large latent TGF-␤ complex bound in the matrix at the end of the 48-h culture period (the matrix was treated with plasmin to release the bound LTBP1 and LTBP1/TGF-␤, hence a doublet is seen for free LTBP1, see black arrowheads). Note the release of LTBP1 into the media by rabbit marrow cells with a corresponding decrease in matrixbound LTBP1. This release is blocked by aprotinin.
Under basal, unstimulated conditions (C), only a small amount of LTBP1 was released from radiolabeled bone ECM by avian osteoclasts when compared with matrix cultured without cells (NC). This low level release was blocked by the serine protease inhibitor, AP. The other protease inhibitors tested did not have any notable effect on basal release of LTBP1/TGF-␤ from bone ECM. These included the serine protease inhibitor leupeptin, the cysteine protease inhibitor pepstatin A, and the tissue inhibitor of matrix metalloproteinases-1 (TIMP1). In contrast, under 1,25-D 3 -stimulated conditions, essentially all the matrix-bound LTBP1 and large latent TGF-␤ complex were released into the culture media. Both the serine protease inhibitor, aprotinin, and the matrix metalloproteinase inhibitor, TIMP1, blocked release of LTBP1 and LTBP1/TGF-␤ almost to the level seen in unstimulated control cells . The serine protease inhibitor leupeptin also partially blocked release of LTBP1 and LTBP1/TGF-␤. In contrast, the aspartic (acid) protease inhibi-tor, pepstatin A, was without effect. In separate experiments we have also found that the cysteine protease inhibitor E64 is without effect in this assay (data not shown).
Cleavage of LTBP1 by Purified MMPs-The above experiments suggested the involvement of serine proteases and/or matrix metalloproteinases in release of LTBP1 and large latent TGF-␤ complex from bone ECM by avian osteoclasts. To confirm that MMPs were able to cleave LTBP1, experiments were performed using the purified proteases. LTBP1 was immunoprecipitated from the conditioned media of CHO-L76 cells which stably overexpressed human LTBP1 using a mouse monoclonal antibody against LTBP1. The immunoprecipitates were then digested with plasmin, MMP2, or MMP9 and analyzed by immunoblotting using a rabbit polyclonal antibody against LTBP1 (see Fig. 7a). LTBP1 immunoprecipitates incubated with digestion buffer alone (C) gave a single LTBP1 immunoreactive band at ϳ190 kDa (black arrow). Digestion with plasmin (PL), MMP2, or MMP9 produced degradation products in the range 125-165 kDa (gray arrows), confirming cleavage of LTBP1. As expected, no immunoreactive LTBP1 bands were detected in samples precipitated with control IgG.
We also examined the ability of MMPs to cleave LTBP1 in its ECM-bound form, which is localized to 10 nm matrix microfibrils in primary osteoblast cultures (10). Treatment of osteoblast-derived ECM with either plasmin or MMP2 resulted in the release of cleaved fragments of LTBP1 into the digest supernatant in the range 125-150 kDa (see Fig. 7b). In contrast, MMP9 failed to release significant amounts of LTBP1 from the ECM. Activity of the MMP2 and MMP9 preparations was confirmed by gelatin zymogram analysis (data not shown). Western blot data were confirmed by parallel immunofluorescent staining for LTBP1 in matrix preparations that were digested with plasmin, MMP2, or MMP9 (see Fig. 7c). Fibrillar LTBP1 staining was observed in control cultures (C). Treatment with MMP2 dramatically reduced fibrillar LTBP1 immunostaining, and treatment with PL almost completely abolished LTBP1 staining. In contrast, treatment with MMP9 produced only a slight reduction in staining intensity.

DISCUSSION
Although LTBP1 was originally identified as a component of the large latent TGF-␤ complex (13,14), it is now clear that the LTBPs 1-4 are members of the fibrillin superfamily that have FIG. 7. a, Western blot showing cleavage of LTBP1 by plasmin (PL), MMP2, and MMP9. LTBP1 was immunoprecipitated from the conditioned media of CHO-L76 cells stably transfected with human LTBP1, using a monoclonal antibody specific for human LTBP1. The immunoprecipitate was digested with plasmin, MMP2, or MMP9 (25 g/ml) for 6 h at 37°C and then electrophoresed in 4 -20% gradient SDS-PAGE gels under non-reducing conditions. The samples were immunoblotted using a polyclonal antibody to LTBP1 (Ab39). The samples are as follows: IgG, sample immunoprecipitated with non-immune IgG; C, sample immunoprecipitated with LTBP1 monoclonal antibody and incubated in digestion buffer alone; PL, MMP2, and MMP9, sample immunoprecipitated with anti-LTBP1 monoclonal antibody and then digested with the indicated protease. Note cleavage of immunoprecipitated LTBP1 by plasmin, MMP2, and MMP9 to give fragments in the range 125-165 kDa. b, Western blot showing LTBP1 fragments released by digestion of ECM from fetal rat calvarial cells with plasmin (PL), MMP2, and MMP9. ECM preparations were digested with plasmin, MMP2, or MMP9 (25 g/ml) for 6 h at 37°C and then electrophoresed in 4 -20% gradient SDS-PAGE gels under non-reducing conditions. The samples were immunoblotted using a polyclonal antibody to LTBP1 (Ab39). Note that plasmin and MMP2 released cleaved LTBP1 fragments; however, MMP9 failed to release LTBP1 fragments from the ECM. c, immunofluorescent images of ECM from fetal rat calvarial cells following digestion with PL, MMP2, and MMP9. Note the well organized fibrillar network in control cultures (C), the reduced staining intensity following MMP2 digestion, and virtual absence of fibrillar staining following plasmin treatment. MMP9 treatment failed to reduce the fibrillar staining significantly compared with undigested controls (bar, 25 m). multiple functions as both structural ECM proteins and as regulators of TGF-␤ availability (11,12). Our data, showing release of ECM-bound LTBP1 and LTBP1 complexed to latent TGF-␤ by osteoclasts, represent the first demonstration that proteolytic cleavage of LTBP1 may be a physiological mechanism for the release of latent TGF-␤ from ECM-bound stores. Although previous studies (9,20,21) using purified proteases have implied a proteolytic mechanism for release of latent TGF-␤ from ECM through cleavage of LTBP1, the amounts of proteases used in these studies are high relative to physiological concentrations. Our studies using osteoclast populations show that these cells, which are actively involved in matrix turnover, are able to release cleaved fragments of LTBP1 and LTBP1 complexed to latent TGF-␤ from matrix that is laid down by osteoblasts.
In the present studies we have also characterized the forms of latent TGF-␤ produced by primary osteoblasts and incorporated into the ECM. The major secreted forms were the small latent complex containing TGF-␤1 and large latent complexes containing predominantly TGF-␤1, with smaller amounts of TGF-␤2. Previously, we reported characterization of the forms of latent TGF-␤ produced by osteosarcoma cell lines (24). Whereas one cell line (UMR-106) produced exclusively the small latent TGF-␤ complex, and one (MG63) produced exclusively the large latent TGF-␤ complex, a third cell line (ROS 17/2.8) produced both complexes and therefore resembled most closely the primary bone cells examined in the present studies.
Pulse-chase immunoprecipitation studies demonstrated that the large latent complex containing LTBP1 is rapidly secreted and rapidly incorporated into the ECM of primary osteoblasts, similar to data reported using a human fibrosarcoma cell line (19). However, in contrast to the fibrosarcoma cells, the major proportion of the LTBP1 produced by primary osteoblasts remained in the culture medium. This suggests that either the osteoblast cells produce an excess of LTBP1 over that which is required for matrix incorporation or that the mechanisms controlling matrix incorporation may be different in these two cell types. Interestingly, only a minor fraction (ϳ10%) of the LTBP1 produced by primary osteoblasts was complexed to TGF-␤, with the remaining 90% occurring in a free, uncomplexed form. In various cell types examined, including fibroblasts (19,24), osteoblasts, (9,24) and breast cancer cells, 2 the amount of free LTBP1 can vary between 50 and 95%. This suggests functions for LTBP1 that are independent of its role in TGF-␤ regulation. In previous studies (9,10) we have demonstrated that bone matrix-bound LTBP1 is present as part of an organized microfibrillar network, which also contains fibrillin-1. These observations, together with the high degree of homology of LTBP1 with the fibrillins, support an important independent role for LTBP1 as an extracellular matrix protein.
Release of bone matrix-bound TGF-␤ by resorbing osteoclasts would be expected to have important consequences for bone cells, as TGF-␤ is capable of regulating all the steps in the bone remodeling cascade (reviewed in Ref. 5). TGF-␤ has been shown to inhibit osteoclast activity, both by stimulating osteoclasts to undergo apoptosis and by inhibiting formation of osteoclasts from their precursors. TGF-␤ is also a powerful chemoattractant and mitogen for osteoblast precursors (5). Therefore, TGF-␤ released from bone matrix could attract osteoblasts to the site of bone resorption and stimulate them to proliferate. The effect of TGF-␤ on mature osteoblasts is then to inhibit proliferation and stimulate production of bone ECM proteins, including type I collagen, fibronectin, and osteocalcin (reviewed in Ref. 35). In addition to stimulating these ECM proteins, TGF-␤1 also induces expression of its own message as well as expression of LTBP1 (24,35).
Interestingly, 1,25-dihydroxyvitamin D 3 stimulated release of LTBP1 from bone ECM by avian osteoclast cells. This effect is most likely due to 1,25-D 3 enhancing resorptive activity through stimulating fusion of osteoclast precursors to form mature resorbing cells. However, a direct effect of 1,25-D 3 on protease activity in osteoclasts is also possible.
Our studies using protease inhibitors suggested the involvement of serine proteases and/or MMPs in release of ECMbound LTBP1 and LTBP1 complexed to latent TGF-␤ by avian osteoclasts and rabbit bone marrow cells. These findings are consistent with our biochemical observations that plasmin, MMP2, and MMP9 can cleave LTBP1. Our data using inhibitors of cysteine and aspartic proteases (e.g. cathepsins B, L, D, and K) suggest that this group of proteases are not involved in cleavage of LTBP1 by avian osteoclasts. This is in agreement with the biochemical data of Taipale and co-workers (21) showing that cathepsins B and D were negative or showed a very limited ability to cleave LTBP1 from the ECM of epithelial cells.
Our studies identify LTBP1 as a new substrate for MMPs. Interestingly, both MMP2 and MMP9 were able to cleave the soluble form of LTBP1; however, only MMP2 cleaved the ECMbound form of LTBP1. This suggests that the cleavage sites used by MMP9 may be unavailable when LTBP1 is incorporated into the ECM. Thus, efficient cleavage of soluble forms of LTBP1 by a protease may not necessarily imply that the same protease will cleave LTBP1 once it is assembled into ECM microfibrils. Another possibility is that degradation of other collagenous or non-collagenous ECM components may be required prior to LTBP1 cleavage to expose cryptic proteolytic cleavage sites on LTBP1. Thus, a cascade of proteases that are involved in the degradation of bone ECM by osteoclasts may be required for release of LTBP1 and TGF-␤ from bone ECM.
Recent studies (36,37) have shown that other members of the fibrillin superfamily, such as fibrillins 1 and 2, are also substrates for MMPs. Breakdown of microfibrillar proteins by MMPs may be important in pathological conditions and may contribute to the disease phenotype in inherited disorders such as Marfan's and related syndromes. Interestingly, fibrillin-1 mutations associated with Marfan's syndrome and ectopia lentis have been shown to result in the production of mutant protein, which is more susceptible to degradation by proteases including MMPs (36,37). This effect appears to be through disruption of calcium binding and apparent exposure of cryptic protease cleavage sites. Breakdown of fibrillin-containing microfibrils that also contain LTBP1 would be expected to result in the release and activation of matrix-bound TGF-␤, which has been strongly implicated in a number of fibrotic diseases and could play a role in scleroderma, another disease for which the fibrillin-1 gene has been implicated (reviewed in Ref. 38).
Our studies showing release of TGF-␤ from bone matrix by plasmin and elastase suggest that these proteases are much more efficient at cleaving LTBP1 and releasing latent TGF-␤ from bone matrix than they are at activating the latent TGF-␤ released. Thus, concentrations of plasmin that can cleave essentially all the LTBP1 and abolish LTBP1 immunoreactivity in the ECM are capable of activating only a small proportion of the latent TGF-␤ released. Although several studies (29,30) have implicated plasmin as an activator of TGF-␤, these studies have been performed using conditioned medium containing predominantly the 100-kDa small latent TGF-␤ complex, which may be more readily activated than the large latent complex, or samples in which the latent TGF-␤ complexes are not defined. Other studies (reviewed in Ref. 39) have been performed using co-culture systems, where again the form of latent TGF-␤ is not defined and may involve release from ECM-bound stores. Our data support the notion that there may be several sequential steps in the process by which ECM-bound TGF-␤ is rendered active, as suggested by Rifkin and co-workers (40). Proteases such as plasmin and MMPs may be important initially in release of latent TGF-␤ from ECM-bound stores. Activation may then take place by cell surface-localized mechanisms, perhaps involving interactions with mannose 6-phosphate receptors (40), integrins (41), or thrombospondins (42). A recent report (43) has also implicated surface-bound forms of MMP2 and MMP9 in activation of the small latent TGF-␤ complex. Further studies are therefore clearly warranted to clarify the steps in the activation pathway.
This study has focused on one potential pathway for release of latent TGF-␤ from bone ECM. However, other mechanisms for TGF-␤ storage and release in bone ECM may also play an important role. For instance, many growth factors can bind to matrix via heparan sulfate-containing proteoglycans and other members of the proteoglycan family. Mature TGF-␤ can bind to heparin (44) and our own unpublished studies indicate that LTBP1 itself binds to heparin. 3 Studies by Tiedemann et al. (45) have demonstrated the importance of heparan sulfatecontaining proteoglycans in assembly of fibrillin-1 into the ECM. Thus heparan sulfate-containing proteoglycans may play a similar role in LTBP1 (and by implication TGF-␤) incorporation. At present the potential role of proteoglycan-degrading enzymes, such as heparanase, in release of TGF-␤ from bone ECM remains to be determined. However, Taipale et al. (21) reported that various glycosidases, including heparinases I and III and chondroitinase ABC as well as a combination of all three, were unable to cleave LTBP1 from the ECM of epithelial cells.
Mature TGF-␤ is also known to bind to proteoglycans such as decorin and biglycan, which may provide an alternative pathway for storage and release of TGF-␤ in bone ECM (46). These latent TGF-␤ complexes could be viewed as "secondary" complexes because it is the mature TGF-␤ homodimer (i.e. after activation) that is bound. In contrast, the LTBP1-bound TGF-␤ can be thought of as a "primary" latent TGF-␤ complex, because this latent complex is assembled inside the cell prior to secretion (17). Although our studies have demonstrated a likely role for LTBP1 in both the storage and release of latent TGF-␤ in bone ECM, other secondary latent TGF-␤ complexes, such as the decorin and biglycan-bound forms, may also be released and activated during bone resorption.
In summary, the data presented here, using osteoclast culture systems, provide the first demonstration that cell-mediated release of ECM-bound TGF-␤ can occur via proteolytic cleavage of LTBP1. LTBP1 is known to regulate TGF-␤ activity at multiple levels including enhancing secretion of the latent TGF-␤ complex, facilitating storage of the latent TGF-␤ complex in the ECM and modulating activation of latent TGF-␤. The involvement of LTBP1 in the cell-mediated release of latent TGF-␤ from ECM-bound stores confirms an additional regulatory role for this ECM protein and may provide an important pathway for communication between cells and the extracellular matrix.