Fibronectin Regulates Latent Transforming Growth Factor-β (TGFβ) by Controlling Matrix Assembly of Latent TGFβ-binding Protein-1*

Latent transforming growth factor-β-binding proteins (LTBPs) are extracellular matrix (ECM) glycoproteins that play a major role in the storage of latent TGFβ in the ECM and regulate its availability. Here we show that fibronectin is critical for the incorporation of LTBP1 and transforming growth factor-β (TGFβ) into the ECM of osteoblasts and fibroblasts. Immunolocalization studies suggested that fibronectin provides an initial scaffold that precedes and patterns LTBP1 deposition but that LTBP1 and fibronectin are later localized in separate fibrillar networks, suggesting that the initial template is lost. Treatment of fetal rat calvarial osteoblasts with a 70-kDa N-terminal fibronectin fragment that inhibits fibronectin assembly impaired incorporation of LTBP1 and TGFβ into the ECM. Consistent with this, LTBP1 failed to assemble in embryonic fibroblasts that lack the gene for fibronectin. LTBP1 assembly was rescued by full-length fibronectin and superfibronectin, which are capable of assembly into fibronectin fibrils, but not by other fibronectin fragments, including a 160-kDa RGD-containing fragment that activates α5β1 integrins. This suggests that the critical event for LTBP1 assembly is the formation of a fibronectin fibrillar network and that integrin ligation by fibronectin molecules alone is not sufficient. Not only was fibronectin essential for the initial incorporation of LTBP1 into the ECM, but the continued presence of fibronectin was required for the continued assembly of LTBP1. These studies highlight a nonredundant role for fibronectin in LTBP1 assembly into the ECM and suggest a novel role for fibronectin in regulation of TGFβ via LTBP1 interactions.

Recent evidence suggests that the binding of growth factors to the extracellular matrix (ECM) 1 is a major mechanism for regulation of growth factor activity and plays a fundamental role in tissue morphogenesis and repair (1). The latent transforming growth factor ␤-binding proteins (LTBPs) are members of a family of ECM proteins that are key regulators of transforming growth factor-␤s (TGF␤s) (2)(3)(4). LTBP1, the prototype member of this family, regulates TGF␤ at multiple levels. First, LTBP1 associates with latent TGF␤ inside the cell and facilitates secretion of the latent complex (5). LTBP1 then targets latent TGF␤ to the ECM for storage (6,7). LTBP1 may also provide a vehicle for release of the latent growth factor, following proteolytic cleavage of LTBP1 and release of C-terminal LTBP1 fragments, still bound to the latent TGF␤ (7)(8)(9). Finally, there is also evidence that LTBP1 plays a role in activation of the latent TGF␤ complex (10,11).
LTBPs are members of a larger superfamily of matrix proteins that include fibrillins 1 and 2, the recently reported fibrillin 3 (12), and the latent TGF␤-binding proteins 1-4 (reviewed in Refs. 2, 3, 13, and 14). At least three of the LTBPs (LTBPs 1, 3, and 4) can form complexes with latent TGF␤1, whereas LTBP2 appears not to bind to latent TGF␤ (15). LTBP1 is also secreted by many cell types in a free form that is not bound to TGF␤. The percentage of free LTBP1 varies from 10 to 90%, depending on the cell type and the differentiation stage examined (6,9,16). These observations raise the possibility that LTBP1 may also have important functions that are independent of TGF␤ and may be related to its properties as an ECM protein.
At present, little is known about how LTBPs function as matrix proteins. Like the fibrillins, their primary structure consists predominantly of 6 cysteine (epidermal growth factorlike) repeats similar to the motifs found in the epidermal growth factor precursor and 8 cysteine repeats (termed "TB repeats") unique to the LTBPs and fibrillins. Recent studies have demonstrated that the third TB repeats in LTBPs 1, 3, and 4 are the sites for covalent binding of latent TGF␤ (15,17).
Immunolocalization studies have shown that LTBP1 co-localizes with fibrillin-1 in 10-nm microfibrils in the ECM of osteoblasts (18). LTBP1 has also been shown to co-localize with fibronectin (18 -20), and a binding interaction between LTBP1 and fibronectin has been suggested by ligand blotting studies (19). Fibronectin is required for the assembly of several ECM proteins, including type I collagen (21), fibulin I (22), thrombospondin (23), and fibrinogen (24). We therefore hypothesized that fibronectin may play a critical role in LTBP1 assembly into the ECM and may thereby play a role in regulation of TGF␤. Multiple approaches were used to disrupt fibronectin assembly, and the effect on LTBP1 and TGF␤ incorporation was determined. These studies highlight a novel role for fibronectin in the regulation of TGF␤ via LTBP1 interactions.

MATERIALS AND METHODS
Reagents-Human plasma fibronectin was purified as described previously (25) or purchased from Invitrogen. The 70-kDa N-terminal and 160-kDa fibronectin fragments were prepared as described previously (26). 30-and 40-kDa fibronectin fragments as well as superfibronectin were purchased from Sigma. The H120 recombinant fibronectin fragment was prepared as described previously (27). Fibronectin-stripped serum was prepared by passing the serum over a gelatin-Sepharose column as described elsewhere (28).
Antibodies-Antibodies against LTBP1 included a rabbit polyclonal antibody (Ab39; kindly supplied by K. Miyazono, Japanese Foundation for Cancer Research, Tokyo) (18,29). A rabbit polyclonal antibody against a peptide in rat LTBP1 was also used, which recognizes mouse and rat but not human LTBP1 (18). Fibronectin antibodies included a mouse monoclonal antibody (IgM) against cellular fibronectin (Sigma), a mouse monoclonal antibody (IgG) against the human fibronectin ED-A domain (Harlan), a mouse monoclonal antibody IgG (39B6) directed against type II repeats 12-14 of human fibronectin (30), and a rabbit polyclonal antibody against fibronectin purified from human platelets (Neomarkers, Freemont, CA).
A peroxidase-conjugated donkey anti-rabbit antibody (Amersham Biosciences) was used for Western blotting and ELISA. All secondary antibodies for immunofluorescent staining were purchased from Jackson ImmunoResearch (West Grove, PA). Various combinations of fluorochrome-conjugated secondary antibodies were used as appropriate for each combination of primary antibodies, as stated in the figure legends.
In some experiments, biotinylated secondary antibodies were used in conjunction with fluorescein isothiocyanate-conjugated streptavidin (Vector Laboratories, Burlingame, CA).
Cell Culture-Tissue culture reagents were purchased from Invitrogen. Primary fetal rat calvarial (FRC) osteoblasts were prepared as described previously (7,18). UMR-106 cells were a gift from T. J. Martin (St. Vincent Institute of Medical Research, Fitzroy, Victoria, Australia) and were maintained in Dulbecco's modified Eagle's medium supplemented with 2 mM L-glutamine, 100 units/ml penicillin/streptomycin, and 10% fetal bovine serum.
FN-null (FNϪ/Ϫ) and heterozygous (FNϩ/Ϫ) embryonic fibroblasts were derived as described elsewhere (31). For experiments examining LTBP1 assembly in FN-null fibroblasts, the cells were cultured in the presence of FN-stripped serum with or without the addition of 10 g/ml plasma fibronectin. For growth curve experiments, cells were plated in 96-well plates at 20,000 cells/cm 2 . Cells were trypsinized at the specified time points, and cell number was determined by using a Coulter counter (Beckman Coulter Inc.).
Immunocytochemistry-For immunocytochemistry, cells were plated in Lab-Tek chamber slides at 20,000 cells/cm 2 in Dulbecco's modified Eagle's medium supplemented with 10% FN-stripped fetal bovine serum, 2 mM L-glutamine, and 100 units/ml penicillin/streptomycin with or without 10 g/ml plasma fibronectin. At confluence, the media were changed to Dulbecco's modified Eagle's medium supplemented with 5% FN-stripped fetal bovine serum, 50 g/ml ascorbic acid, and other additives as above, with or without the addition of fibronectin or fibronectin fragments. The cells were cultured for the specified periods, with media changed every 3 days. Co-localization of LTBP1, fibronectin, and type I collagen was performed using double staining indirect immunofluorescence techniques as described previously (18). The specimens were viewed and photographed digitally using a Leica DMRXA microscope with epifluorescence illumination and a Photometric Coolsnap FX CCD camera. For double labeling experiments, multiple controls were performed, including using nonimmune serum or control IgG in place of primary antibody, testing each primary antibody individually using both secondary antibodies to exclude cross-species reactivity of the secondary antibodies, and preincubating the primary antibody with the immunizing peptide (where available).
TGF␤ ELISA and Modified ELISA for Detection of Relative Amounts of LTBP1-A commercial ELISA (TGF␤1 Emax; Promega Corp., Madison, WI) was used for measurement of TGF␤1 in conditioned media samples collected over 48 h or in guanidine HCl extracts of the ECM, prepared as described elsewhere (32). Values were normalized to cell number. For measurement of total (active ϩ latent) TGF␤, the sam-ples were acidified using HCl and then reneutralized prior to measurement using NaOH according to the ELISA manufacturer's instructions.
To quantify the relative amounts of LTBP1 in the ECM, a modified ELISA was used. Cells were grown in 96-well plates in media with or without fibronectin. The plates were fixed in 95% ethanol, then blocked with 5% bovine serum albumin ϩ 1% milk, followed by incubation in primary antibodies against LTBP1. After washing, the amount of bound LTBP1 antibody was determined using the Vectorstain-Elite ABC immunodetection kit (Vector Laboratories, Burlingame, CA) in conjunction with O-phenylenediamine as a reaction substrate. Alternatively, the vector VIP substrate kit was used, followed by solubilization of the reaction product by incubating for 10 min at room temperature in 50 l of 2 M KCl, then adding 50 l of Me 2 SO and incubating for a further 10 min. The plates were read on an EL X 800 plate reader (Bio-Tek Instruments, Winooski, VT) at 450 nm, and background subtraction was performed using a blank control that had been incubated with nonspecific IgG in place of primary antibody. Values were normalized to total protein content of the cell lysate or to cell number as determined using a Coulter counter.
Immunoelectron Microscopy-For immunoelectron microscopy, FNnull and heterozygous fibroblasts were cultured on Thermanox coverslips (Nalge Nunc International) in media with or without added fibronectin as described above. Media were changed every 3 days, and coverslips were harvested on days 6, 14, and 21. Immunogold staining was performed on unfixed specimens as described previously (18) using anti-LTBP1 and anti-fibronectin antibodies each diluted 1:10. For detection, a donkey anti-rabbit secondary antibody-6-nm gold conjugate and a donkey anti-mouse secondary antibody-18-nm gold conjugate were used (Jackson ImmunoResearch). The samples were fixed after immunogold staining and processed for transmission electron microscopy as described previously (18). They were then embedded in Taab epoxy resin (Taab Laboratories Equipment Ltd., Aldermaston, UK). After polymerization, the coverslips were removed to leave the cell layer exposed on the outer surface of the block. Ultrathin sections were cut at a slight angle to the growth substrate by using a diamond knife. Sections were examined with or without contrast in a Philips EM 301 transmission electron microscope at an accelerating voltage of 60 kV.
LTBP1 Stably Expressing Cell Lines-To generate UMR-106 cell lines stably overexpressing LTBP1, a full-length human LTBP1 construct in the PSV7d vector (kindly provided by K. Miyazono, Japanese Foundation for Cancer Research, Tokyo) was co-transfected at a 10:1 ratio with an RSVneo selection vector, using calcium phosphate precipitation as described elsewhere (33). The transfected cells were selected with 400 g/ml G418, and resistant single cell clones were screened for LTBP1 expression using an ELISA, as described previously (7). Three high expressing clones (0.2-1.5 g/ml) were selected for further study. Control clones transfected with RSVneo alone were used for comparison.
FPLC Analysis-FPLC fractionation was performed as described previously (16) using 48-h serum-free, phenol red-free conditioned media harvested from 90 -95% confluent cultures. 150 ml of conditioned medium was concentrated 10-fold over a 50-kDa cut-off membrane using a minisette concentrator (Millipore, Bedford, MA). The samples were then lyophilized, reconstituted, and dialyzed against 20 mM Tris buffer, pH 7.2, prior to application over 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 previously (16).
Metabolic Labeling and Immunoprecipitation-For metabolic labeling and immunoprecipitation, cells were plated into 12-well multiwell plates at 10,000 cells/cm 2 growth area. At 90% confluence, the cells were metabolically labeled as described previously (8) by using 100 Ci/well [ 35 S]cysteine for 6 h. LTBP1 was measured in the culture supernatants as well as the ECM by immunoprecipitation followed by SDS-PAGE and autoradiography as described previously (8,16). A plasmin digestion was used to release the ECM-bound LTBP1 (7,19).
Western Blotting-Proteins in concentrated conditioned media samples or plasmin matrix digests were separated by SDS-PAGE using 4 -20% gradient polyacrylamide mini gels, and immunoblotting was performed as described previously (7). The immunostained bands were visualized using the Renaissance ECL detection system according to manufacturer's instructions (PerkinElmer Life Sciences). Samples were normalized to total protein content or to cell number prior to loading on the gels.

Time Course of LTBP1 and Fibronectin Assembly in Primary
Osteoblast Cultures-To determine the time course of LTBP1 and fibronectin assembly, immunolocalization studies were performed over a time course of 1-21 days (Fig. 1). Fibronectin initially appeared as short fibrillar structures on the surface of fetal rat calvarial cells after 24 h in culture ( Fig. 1, d1). At this stage, no immunoreactive LTBP1 was present on the cell surface or in the extracellular matrix. By 2 days in culture ( Fig. 1,  d2), the fibronectin became organized into a fibrillar network in the ECM, and a small amount of LTBP1 was observed that co-localized with fibronectin. By 3 and 5 days of culture ( Fig. 1, d3 and d5), LTBP1 was localized in a fibrillar network that co-distributed with fibronectin. The incorporation of LTBP1 therefore lagged 1-2 days behind that of fibronectin, and there was always a proportion of fibronectin-positive fibrils that did not stain for LTBP1. Between days 5 and 21, there was an increase in formation of LTBP1 fibrils, which became organized into long parallel fibrillar arrays. At the same time, there was a progression from distinct fibronectin fibrils to fibrils with a more diffuse appearance. The co-localization of LTBP1 and fibronectin started to diverge until day 21, when LTBP1 became localized in fibrils that were clearly distinct from fibronectin. Identical results were obtained by using the human MG63 osteosarcoma cell line as well as with human skin fibroblasts and using two independent antibodies for LTBP1 and fibronectin (data not shown).
Double-labeled immunogold localization for LTBP1 and fibronectin demonstrated co-localization of these two molecules in fibrillar structures in 6-day fetal rat calvarial cell cultures (Fig. 2). In later (14-day) cultures, some fibrils could be found that were labeled with both antibodies; however, LTBP1 was also frequently found in bundles of 10-nm microfibrils that were not labeled with fibronectin antibodies (Fig. 2).
Inhibition of LTBP1 and TGF␤1 Incorporation by a 70-kDa N-terminal Fibronectin Fragment-To determine whether fi-bronectin was required for assembly of LTBP1 into bone ECM, fetal rat calvarial cell cultures were treated for 6 days with a 70-kDa N-terminal fibronectin fragment (Fig. 3a). This fragment blocks assembly of both endogenous and exogenous fibronectin in fibroblasts by occupying binding sites that are required for fibronectin self-association (28). Treatment with 70 g/ml of the fibronectin fragment (Fig. 3a, 70K) resulted in a dramatic reduction in fibronectin incorporation compared with untreated controls (control) or control cultures treated with plasma fibronectin (pFN). This was paralleled by a reduction in LTBP1 incorporation into the ECM. Dose-response experiments showed a significant reduction in LTBP1 incorporation at doses between 10 and 70 g/ml, as determined by immunofluorescence (data not shown) and also by quantitation of the relative amounts of incorporated LTBP1 using a modified ELISA (Fig. 3b).
Time course experiments indicated that at early time points (days 1-3) there was a virtual absence of both fibronectin and LTBP1 in the ECM of cells treated with the fibronectin 70-kDa fragment. However, by day 5, fibronectin and LTBP1 incorporation began to partially recover. This recovery of LTBP1 incorporation appeared to be due to the fact that even the highest dose of the 70-kDa fibronectin fragment failed to completely block fibronectin incorporation. The above results were replicated by using an LTBP1 antibody that recognizes the prolinerich hinge region and using MG63 human osteosarcoma cells (data not shown). TGF␤ 1 is the major TGF␤ isoform produced in bone (7,34). To determine whether inhibition of fibronectin assembly also affected incorporation of TGF␤ into the ECM, the amount of TGF␤1 in the ECM was measured using a commercial ELISA (Fig. 3c). Treatment of fetal rat calvarial cells with the 70-kDa fibronectin fragment reduced the amount of latent TGF␤1 stored in the ECM by 80%. This was not because of reduced synthesis and secretion of TGF␤ as increased amounts of latent

FIG. 1. Immunofluorescence time course showing co-localization of LTBP1 (rabbit anti-LTBP1 detected with Cy3 anti-rabbit) and fibronectin (anti-fibronectin IgM monoclonal detected with biotinylated anti-mouse and fluorescein isothiocyanate streptavidin) in FRC cells over a time course of 1-21 days.
Note that fibronectin appears first on the cell surface after 24 h (d1) and that LTBP1 appears later (d2) and is associated with the underlying fibronectin network. From days 5 to 21, the fibronectin network becomes more diffuse, and the localization of LTBP1 and fibronectin begins to diverge until by day 21 LTBP1 is organized into fibrils that are distinct from fibronectin. Bar ϭ 50 m.
TGF␤ were detected in the conditioned media of the treated cultures (Fig. 3c).
LTBP1 Assembly in Fibronectin-null Embryonic Fibroblasts-To define further the role of fibronectin in the assembly of LTBP1 into the ECM, we examined LTBP1 assembly in fibroblasts differentiated from mouse embryonic stem cells that were null for the fibronectin gene (FN-null) (Fig. 4). Both heterozygous (ϩ/Ϫ) and FN-null (Ϫ/Ϫ) embryonic fibroblasts synthesized and secreted LTBP1 in amounts that were grossly equivalent to the amounts produced by fetal rat calvarial cells. 2 Fig. 4a shows double-stained immunolocalization of LTBP1 and fibronectin in the ECM of 6-day cultures of FN-null fibroblasts (Ϫ/Ϫ), as compared with heterozygous controls (ϩ/Ϫ). The cells were cultured in fibronectin-stripped serum to ensure that no exogenous fibronectin was present. In control cells, a network of fibronectin fibrils was present in the ECM, and LTBP1 was found to co-distribute with this network. In contrast, no fibronectin was detected in the ECM of FN-null cells, and this was associated with a failure in LTBP1 incorporation. Western analysis (Fig. 4b) indicated that this effect was not because of reduced expression of LTBP1 in the FN-null cells, as they secreted a large amount of LTBP1 into the conditioned media (Fig. 4b, black arrow) but failed to incorporate it into the ECM (Fig. 4b, gray arrow). In contrast, the heterozygous cells had lower amounts of LTBP1 in the conditioned media, with the major proportion in the ECM.
Time course studies demonstrated a virtual absence of LTBP1 incorporation into the ECM of FN-null cells between days 3 and 6 of culture (Fig. 5a). However, with longer times in culture (14 -21 days), a few sparse LTBP1-positive fibrils were formed, which occurred in areas where the cells were multilayered. At no time was any fibronectin detected in the ECM. Identical results were obtained using two independent LTBP1 antibodies (data not shown).
Immuno-EM analysis of 14-day cultures confirmed that the LTBP1 localized to 10-nm microfibrils in heterozygous cells and that collagen fibrils were unlabeled (Fig. 5b, ϩ/Ϫ). The few small LTBP-1 positive fibrils that formed in the 14-day FN-null cell cultures also appeared to be 10-nm microfibrils (Fig. 5b, Ϫ/Ϫ). (23,31) have shown that an exogenous source of fibronectin can restore the ability of FN-null cells to assemble fibronectin. We therefore determined whether addition of plasma fibronectin could rescue incorporation of LTBP1 into the ECM of FN-null cells (Fig. 6a). Treatment with 10 g/ml plasma fibronectin (Fig. 6a, Ϫ/Ϫ ϩ FN) rescued incorporation of LTBP1 into the ECM of FN-null cells. When the cultures were double-stained using an antibody against the fibronectin ED-A domain, which only recognizes the endogenous cellular form of fibronectin, no fibronectin staining was seen in the FN-null cells, confirming that they did not produce any endogenous fibronectin (Fig. 6a). However, when an antibody against fibronectin type III repeats 12-14 was used, which recognizes the exogenously added human plasma fibronectin, it was clear that the added fibronectin was assembled into the ECM of the rescued FN-null cells and that LTBP1 co-distributed with the added fibronectin (Fig. 6b). Time course studies showed that the rescue effect was evident as early as 3 days and that the FN-null cells went on to produce a well organized parallel fibrillar LTBP1 network by 21 days (data not shown).

Rescue of LTBP1 Incorporation by Addition of Exogenous Fibronectin and Fibronectin Fragments-Previous studies
Immuno-EM studies on day-14 cultures demonstrated that LTBP1 in the rescued cultures was localized to abundant 10-nm microfibrils, similar to the amounts seen in the heterozygous controls (data not shown). In contrast, the FN-null cells cultured in the absence of fibronectin produced only a few sparse microfibrils. Western analysis (Fig. 7a) and quantitation of ECM-bound LTBP1 using a modified ELISA (Fig. 7b) confirmed that LTBP1 incorporation into the ECM was dramatically reduced in FN-null cells and could be returned to control levels by addition of exogenous fibronectin. Growth curves demonstrated that the lack of incorporation of LTBP1 in the FN-null cells was not because of reduced cell growth compared with heterozygous control cells (Fig. 7c). In fact, the FN-null cells proliferated more rapidly than the heterozygous control cells. Addition of fibronectin to the cultures had no significant effect on cell growth in either FN-null or heterozygous cells between days 1 and 6; however, a small but significant increase in proliferation was seen in FN-null cells treated with fibronectin at days 10 and 14.
We next performed rescue experiments using a panel of fibronectin fragments to determine whether specific fragments were able to rescue LTBP1 incorporation and whether assembly of fibronectin fibrils was required or whether RGD-mediated activation of integrins was sufficient (see Fig. 8a for a schematic diagram of the fibronectin fragments used). Of the fragments tested, only full-length plasma fibronectin or superfibronectin, which are both capable of assembly into ECM fibrils, was able to rescue incorporation of LTBP1 into the ECM (Fig. 8b). Other fragments, including N-terminal 30-, 40-, and 70-kDa fragments as well as the 160-kDa and H120 fragments, were not able to rescue LTBP1 incorporation (Fig. 8b).
The Continual Presence of Fibronectin Is Required for Continued Assembly of LTBP1-A recent paper by Sottile and 2 S. L. Dallas, unpublished observations.

FIG. 2. Double-labeled immunogold EM localization of LTBP1 and fibronectin in 6-and 14-day cultures of fetal rat calvarial cells.
Fibronectin was localized using 18-nm gold-labeled secondary antibodies and LTBP1 was localized using 6-nm gold particles. Note that at 6 days, fibrils are seen that are positive for both LTBP1 and fibronectin. By day 14, many microfibrillar bundles (MF) can be found that contain LTBP1 but not fibronectin. A control in which a nonimmune IgG was used in place of the primary antibody is shown on the right.
Hocking (23) has shown that the continual assembly of fibronectin is required to maintain the stability of several ECM components. To determine whether the continual assembly of fibronectin is required to maintain the presence of LTBP1 in the ECM, we performed rescue experiments in FN-null cells followed by withdrawal of fibronectin (Fig. 9). FN-null cells were rescued with fibronectin treatment (10 g/ml) for 3 days, after which the cultures were either fixed immediately or continued for a further 3 days in the presence or absence of fibronectin. Immunostaining confirmed that FN-null cells rescued with fibronectin treatment for 3 days showed a well organized fibronectin fibrillar network and incorporated LTBP1 into the ECM. If the cultures were continued for a further 3 days in the presence of fibronectin, additional fibronectin and LTBP1 assembly occurred. However, withdrawal of fibronectin for days 4 -6 resulted in loss of fibronectin immunoreactivity from the ECM, suggesting that the continual presence of fibronectin is required to maintain stability of the fibronectin fibrillar network. In contrast, withdrawal of fibronectin did not result in significant loss of the LTBP1 that had already been incorporated, but it did prevent further LTBP1 assembly (i.e. LTBP1 staining in cultures following withdrawal of fibronectin for days 4 -6 resembled cultures that had been fixed on day 3).
Type I Collagen and Fibrillin-1 Are Not Required for Association of LTBP1 with Fibronectin-Similar to the report of Velling et al. (21), we also found that type I collagen incorporation was compromised in FN-null fibroblasts compared with heterozygous controls (data not shown). To exclude the possibility that the lack of incorporation of LTBP1 into the ECM in FN-null fibroblasts was secondary to an effect on collagen assembly, experiments were performed in the absence of ascorbic acid. Under these conditions, no fibrillar collagen was detected, and all the type I collagen was localized intracellularly. However, the relationship between LTBP1 and fibronectin was unaffected by the absence of ascorbic acid, i.e. LTBP1 incorporation in FN-null cells could be rescued by addition of fibronectin in the absence of fibrillar collagen (data not shown).
In addition to reduced incorporation of LTBP1 and type I collagen, we also found that fibrillin-1 assembly was impaired in FN-null fibroblasts. 2 Because LTBP1 and fibrillin-1 both appear to be dependent on fibronectin for assembly into the ECM and because we have previously reported that LTBP1 and fibrillin-1 co-localize in 10-nm microfibrils, we next examined whether fibrillin-1 was required for incorporation of LTBP1 into the ECM in association with fibronectin. LTBP1 was stably transfected into UMR-106 rat osteosarcoma cells, which do not produce endogenous LTBP1 or fibrillin-1. We have shown previously (16) that these cells produce exclusively the 100-kDa small latent TGF␤ complex that lacks LTBP1. However, these cells do produce an organized fibronectin matrix, making them a useful host cell for LTBP1 overexpression to examine its association with fibronectin in the absence of fibrillin-1.
Several stable single cell clones were obtained, which expressed LTBP1 in the range 0.2-1.5 g/ml, as determined by ELISA. FPLC analysis was used to examine the forms of latent TGF␤ produced by the transfected UMR-106 cell lines (Fig.  10a). In empty vector-transfected cells, a single peak of TGF␤ activity was observed, eluting at 0.2 M NaCl (Fig. 10a, gray bars), consistent with the known elution position of the 100-kDa small latent TGF␤ complex (16,35). LTBP1 was undetectable by ELISA in empty vector-transfected cells (Fig. 10a, solid  lines). Transfection of UMR-106 cells with LTBP1 altered the elution profile so that ϳ30% of the latent TGF␤ activity eluted at a later position (0.3 M NaCl), corresponding to the known elution position of the 290-kDa large latent TGF␤ complex, containing LTBP1 (16,35). Consistent with this, the LTBP1 peak, as measured by ELISA, overlapped with the large latent TGF␤ complex peak. The LTBP1 peak also extended beyond the TGF␤ peak, which presumably represents free, uncomplexed LTBP1. Fig. 10b shows immunoprecipitation analysis on two LTBP1-overexpressing UMR-106 clones. As expected, LTBP1 was undetectable in empty vector-transfected cells in either the media or matrix. In contrast, when LTBP1 was overexpressed in UMR-106 cells, bands were observed in the conditioned medium and in plasmin digests of the ECM at the expected size for free LTBP1 (Fig. 10b, black arrowhead) and LTBP1 complexed to TGF␤ (Fig. 10b, white arrowhead), suggesting that the overexpressed protein was incorporated into the ECM. Double-stained immunofluorescence confirmed that LTBP1 was undetectable in the ECM of empty vectortransfected cells (Fig. 10c). In contrast, in LTBP1-overexpressing cells, abundant LTBP1 staining was observed in the matrix, which co-localized with fibronectin. No fibrillin-1 immunostaining was detected in either control or LTBP1transfected cells. Identical results were obtained with three independent LTBP1-overexpressing clones. Thus, in the absence of fibrillin-1, LTBP1 still co-localized with fibronectin. FIG. 4. a, immunofluorescent staining of LTBP1 (Ab39, detected with Cy3 anti-rabbit antibodies) and fibronectin (monoclonal antibody against ED-A domain, detected with biotinylated anti-mouse IgM followed by streptavidin-fluorescein isothiocyanate) in embryonic fibroblasts from FN-null (Ϫ/Ϫ) and heterozygous (ϩ/Ϫ) mice (6-day cultures). Note that LTBP1 and fibronectin are co-localized in the heterozygous cells, but there is a complete absence of LTBP1 and fibronectin in the ECM of the FN-null cells. Bar ϭ 50 m. b, Western blot showing LTBP1 in the conditioned medium (black arrow) and in the matrix (gray arrow) of FN-null (Ϫ/Ϫ) and heterozygous (ϩ/Ϫ) embryonic fibroblasts. In order to detect the matrix-bound LTBP1, a plasmin digestion was performed, hence fragments of 110-130 kDa are observed. Note that in the heterozygous cells, the major proportion of LTBP1 is incorporated into the ECM. In contrast, although FN-null fibroblasts secrete high levels of LTBP1 into the conditioned medium, they fail to incorporate it into the ECM. Samples were normalized to cell number prior to loading on the gel. In FN-null cells, no fibronectin is detected at any time point. LTBP1 initially fails to incorporate at the earlier time points; however, by days 14 -21 a few sparse fibrils are able to form, mainly in areas where the cells are multilayered. Bar ϭ 50 m. d, day(s). b, electron micrographs showing localization of LTBP1 in FN-null (Ϫ/Ϫ) and heterozygous (ϩ/Ϫ) embryonic fibroblasts (14-day cultures) as demonstrated by immunogold labeling using antibody to LTBP1. Note that the 6-nm gold particles are localized on 10-nm microfibrils (MF) and that the banded collagen fibrils (COL) are unlabeled.

DISCUSSION
In this study we provide evidence that fibronectin is a key regulator of LTBP1 and TGF␤ incorporation into the ECM of osteoblasts and fibroblasts. Immunolocalization studies showed that fibronectin is assembled into the ECM prior to LTBP1 and that LTBP1 is then deposited onto and co-localizes with fibronectin. However, in long term cultures of both primary osteoblasts and fibroblasts, fibronectin and LTBP1 are localized in separate fibrillar networks. These data suggest that fibronectin provides an initial template for deposition of LTBP1 (i.e. formation of a fibronectin matrix precedes and patterns LTBP1 deposition) but that this template only acts as a temporary scaffold (i.e. the initial fibronectin template is lost).
Consistent with a critical role for fibronectin in LTBP1 assembly, disruption of fibronectin fibrillogenesis using a 70-kDa N-terminal fibronectin fragment impaired incorporation of LTBP1 and latent TGF␤1 into the ECM. Furthermore, embryonic fibroblasts that lacked the fibronectin gene failed to incorporate LTBP1 into the ECM even though they secreted a large amount of LTBP1 into the culture medium. Addition of exogenous fibronectin was able to completely rescue fibrillar assembly of fibronectin and to rescue LTBP1 incorporation into the ECM as well as its co-localization with fibronectin. Together, these data suggest a crucial non-redundant role for fibronectin in the assembly of LTBP1 into the ECM. The lack of LTBP1 assembly in the absence of fibronectin suggests that LTBP1 does not use the same integrin receptor system as fibronectin (i.e. ␣5␤1 or ␣4␤1) directly for its assembly, since these integrins are expressed in the FN-null cells (31) and their presence is not sufficient for LTBP1 assembly in the absence of fibronectin. The assembly of LTBP1 therefore appears to be fundamentally different from that of fibronectin.
Although LTBP1 assembly was virtually absent in 1-6-day cultures of FN-null fibroblasts, with extended culture times (14 -21 days) a small amount of LTBP1 incorporation was observed, which was localized in areas where the cells were multilayered and had the appearance of short, disorganized fibrils. One possibility is that this small amount of LTBP1 incorporation may be due to low levels of residual fibronectin remaining in the stripped fetal calf serum used for culturing. However, by using two independent antibodies that recognize plasma fibronectin, we have not detected any incorporation of However, no fibronectin is detected in these rescued cultures using the ED-A antibody that only recognizes the endogenously produced cellular form of fibronectin and not the added plasma fibronectin. b, FN-null embryonic fibroblasts rescued by addition of 10 g/ml plasma fibronectin and double-stained using anti-LTBP1 and anti-fibronectin monoclonal antibody against type III repeats 12-14. In this case the fibronectin antibody recognizes the added plasma fibronectin. Note that the added fibronectin has been assembled into the ECM and that the rescued LTBP1 is co-localized with it. Bar, 50 m.
plasma fibronectin into the ECM of these cells by immunofluorescence. This suggests that a very limited amount of LTBP1 assembly can occur via an alternative pathway that does not require fibronectin.
In rescue experiments using a panel of fibronectin fragments, only fragments that were capable of assembly into fibronectin fibrils were able to rescue LTBP1 incorporation in FN-null cells. This suggests either that the critical event for LTBP1 assembly is the formation of a fibrillar fibronectin network and/or that multiple interacting sites on the fibronectin molecule are required for LTBP1 incorporation and that the individual fragments tested were lacking one or more of these sites. Future studies will therefore require the use of fibronectin constructs that contain all the necessary domains for assembly into fibrils but contain deletions in other internal domains (36). The ability of the 70-kDa fibronectin fragment to inhibit incorporation of LTBP1 in primary osteoblasts favors the hypothesis that fibronectin fibril assembly is the critical event, since this fragment works by blocking fibronectin selfinteracting domains but does not inhibit binding of fibronectin to cell surface integrins. Furthermore, the 160-kDa fibronectin fragment that contains the RGD sequence and is known to stimulate integrin-mediated signaling was unable to rescue LTBP1 incorporation in FN-null cells, suggesting that integrin ligation alone is not the critical event for LTBP1 assembly.
Similar to LTBP1, we have observed a time-dependent colocalization between fibrillin-1 and fibronectin, with initial deposition in association with fibronectin, followed by localization in separate fibrillar networks. 2 We have also found that fibrillin-1 assembly was disrupted by inhibition of fibronectin assembly. However, fibrillin-1 showed a somewhat greater capacity for assembly in the absence of fibronectin compared with LTBP1. Together, these data show that fibronectin is a critical regulator of assembly for at least two members of the fibrillin superfamily and that these microfibrillar proteins may have a limited capacity for self-assembly in living cell systems in the absence of fibronectin.
A number of studies have shown that fibronectin acts as an orchestrator for the assembly of other ECM proteins. For example, assembly of type I and III collagen is impaired in FNnull fibroblasts (21,23). Although collagens are clearly capable of self-assembly in various cell-free systems, these studies suggest that additional levels of control may be exerted in living cell systems. In addition to reduced LTBP1 and fibrillin-1 incorporation, we also observed that assembly of type I collagen was compromised in FN-null cells. 2 However, the fact that LTBP1 assembly could be rescued in FN-null cells in the absence of ascorbic acid (i.e. in the absence of fibrillar collagen) suggests that the impairment of LTBP1 incorporation was not a secondary consequence of reduced collagen assembly.
At present, the mechanism by which fibronectin and LTBP1 interact is unclear. Solid phase binding assays performed in our laboratory by using purified LTBP1 and fibronectin fragments suggest that there is no direct binding interaction between LTBP1 and fibronectin and that the binding is indirect. 3 Although Taipale et al. (6) reported a binding interaction between LTBP1 and fibronectin by ligand blotting, these studies were performed with conditioned media rather than purified LTBP1 and are therefore consistent with an indirect binding mechanism. Because it has been reported that LTBP1 can interact directly with fibrillin-1 (37) and because fibrillin-1 co-localizes with fibronectin in a manner similar to LTBP1, we examined whether fibrillin-1 may mediate LTBP1 binding to fibronectin. By overexpressing LTBP1 in UMR-106 cells that lack endogenous LTBP1 or fibrillin-1, we showed that fibrillin-1 was not required for LTBP1 incorporation in association with fibronectin. Our current data suggest that heparan sulfate proteoglycans may be responsible. 3 In primary osteoblasts, human fibroblasts, and osteoblastic cell lines, we have consistently observed that LTBP1 initially co-localizes with fibronectin but later is localized in a separate fibrillar network. In accordance with this, we have also reported that LTBP1 and fibronectin show a partial overlap in distribution in periosteal osteoblasts in vivo (18). At present, the mechanism for the change in co-distribution patterns is unknown. One possibility is that LTBP1 is laid down in association with fibronectin as part of an initial phase of ECM assembly designed to rapidly assemble a supportive "temporary ECM." However, this ECM is later remodeled to produce a more highly organized ECM containing specific fibrillar structures. This could be achieved either by breakdown and reassembly of the original ECM or by an active cell-mediated reorganization of the original ECM. Recent studies (23) have shown that the continual assembly of fibronectin into the ECM is required to maintain stability of ECM components, such as type I collagen and thrombospondin, suggesting that the ECM is a highly dynamic structure, which may be constantly undergoing remodeling.
Consistent with the results of Sottile and Hocking (23), we have observed that in FN-null cells that were first rescued with fibronectin supplementation followed by withdrawal of fibronectin, there is a loss of fibronectin from the ECM. This suggests that the continual presence of fibronectin is required to maintain stability of the fibronectin fibrillar network. Although the LTBP1 fibrillar network appeared to be more stable and was not degraded upon withdrawal of fibronectin, withdrawal of fibronectin halted further LTBP1 assembly so that cultures treated with fibronectin on days 1-3 followed by fibronectin withdrawal for days 4 -6 resembled cultures that had been fixed on day 3. In contrast, cultures in which fibronectin treatment was continued for days 4 -6 continued to assemble more LTBP1 into the ECM. These data suggest that not only is fibronectin required for the initial assembly of LTBP1 into the ECM but that the continual presence of fibronectin is essential for LTBP1 assembly to continue. 3 S. L. Dallas, manuscript in preparation.
FIG. 9. Immunofluorescent staining showing FN-null embryonic fibroblasts treated for days 1-3 with 10 g/ml plasma fibronectin and then either fixed immediately or continued for days 4 -6 in the presence or absence of 10 g/ml fibronectin. Note that cells cultured with fibronectin for days 4 -6 continue to assemble more LTBP1 and fibronectin into the ECM compared with cultures fixed at day 3. However, in cultures in which fibronectin is withdrawn for days 4 -6, there is a loss of fibronectin immunoreactivity in the ECM. In contrast to fibronectin, LTBP1 is not lost from the ECM, but further assembly of LTBP1 is halted so that the 6-day cultures resemble those fixed at 3 days. Bar, 50 m.
LTBP1 is a multifunctional regulator of TGF␤ activity by facilitating secretion of latent TGF␤ from the cell (5), targeting latent TGF␤ to the ECM for storage (6,7), providing a vehicle for release of the growth factor upon proteolytic cleavage (8,9), and facilitating activation of latent TGF␤ (10, 11). Our data showing a critical role for fibronectin in regulating LTBP1 incorporation therefore suggest a previously unknown function for fibronectin as a regulator of TGF␤ via regulation of LTBP1 assembly. In support of this, we have shown reduced latent TGF␤1 incorporation into the ECM of FN-null fibroblasts together with a corresponding increase in latent TGF␤1 in the conditioned medium. Interestingly, Globus et al. (38) showed that antibodies to fibronectin induced apoptosis of osteoblasts and that this could be reversed by addition of TGF␤, suggesting FIG. 10. Overexpression of LTBP1 in UMR-106 cells that lack endogenous LTBP1 and fibrillin-1. a, FPLC profile of LTBP1-transfected UMR-106 cells and empty vector-transfected controls. Note a single peak of TGF␤ activity in the control cells (gray bars) and undetectable LTBP1 (solid line). In contrast in the transfected cells, there are two peaks of TGF␤ activity (gray bars). The second peak overlaps with the elution position of LTBP1 as determined by ELISA (solid black line). LLC, large latent TGF␤ complex; SLC, small latent TGF␤ complex. b, immunoprecipitation showing the lack of detectable LTBP1 in empty vector-transfected UMR-106 cells (EV) compared with two independent cell clones transfected with LTBP1. In the transfected cells, LTBP1 is detectable in the conditioned media and in the ECM. Samples were normalized to cell number prior to loading on the gel. Ab, antibody; C, non-immune control. c, immunostaining for LTBP1 and fibronectin in empty vector (EV) compared with LTBP1-transfected cells (LTBP1). Note the absence of fibrillin-1 in either cell line. Also note that LTBP1 is incorporated into the ECM in association with fibronectin even in the absence of fibrillin-1. Bar, 50 m. an interaction between fibronectin and TGF␤ in regulation of osteoblast function. The present findings showing that fibronectin is critical for LTBP1 and TGF␤ assembly into the ECM of osteoblasts may provide a molecular explanation for such interactions between fibronectin and TGF␤ in bone cells.
In summary, we have demonstrated a novel function for fibronectin in regulating the incorporation of LTBP1 and latent TGF␤ into the ECM. Because LTBPs are major regulators of transforming growth factor-␤s, these studies may have important implications for diseases in which TGF␤ plays a pathogenic role, such as fibrotic diseases, cancers, osteoporosis, and arthritis. These studies will also have implications for current models of assembly of microfibrillar proteins.