Polymerization of Type I and III Collagens Is Dependent On Fibronectin and Enhanced By Integrins α11β1and α2β1 *

Polymerization of the ECM proteins fibronectin and laminin has been shown to take place in close vicinity to the cell surface and be facilitated by β1integrins (Lohikangas, L., Gullberg, D., and Johansson, S. (2001) Exp. Cell Res. 265, 135–144 and Wennerberg, K., Lohikangas, L., Gullberg, D., Pfaff, M., Johansson, S., and Fassler, R. (1996) J. Cell Biol. 132, 227–238). We have studied the role of collagen receptors, integrins α11β1 and α2β1, and fibronectin in collagen polymerization using fibronectin-deficient mouse embryonic fibroblast cell lines. In contrast to the earlier belief that collagen polymerization occurs via self-assembly of collagen molecules we show that a preformed fibronectin matrix is essential for collagen network formation and that collagen-binding integrins strongly enhance this process. Thus, collagen deposition is regulated by the cells, both indirectly through integrin α5β1-dependent polymerization of fibronectin and directly through collagen-binding integrins.

Collagens form a large family of proteins with more than 20 different members described to date (3). The organization of different collagens into various types of fibrils and networks in extracellular matrices (ECM) 1 is of crucial importance for the physical properties of tissues. Type I and III collagens, the predominant proteins in the body, are prototypes for the fibrillar collagen subfamily. Although these collagens are known to serve as a scaffold for numerous associated proteins, proteoglycans, and cells in the ECM, the mechanisms that regulate their own polymerization are poorly understood.
As all collagens, types I and III are composed of three ␣ chains that form triple-helical domains. The collagen triplehelices are assembled intracellularly in a process dependent on ascorbic acid as a cofactor for hydroxylation of selected prolines and lysines (4). In addition to the triple-helical region, procol-lagen molecules of type I and III collagens contain a propeptide in both the N-and C-terminal ends as well as the so-called telopeptides. In connection with secretion of the monomers into the extracellular space, the propeptides are usually removed by proteolytic cleavage. Removal of C-terminal propeptides is a prerequisite for the fibrillogenesis. For instance the ␣1(I) collagen chain, where the C-propeptide cleavage site was mutated, could not be incorporated into the fibrillar cross-linked collagen matrix (5). The fate of the N-terminal propeptides varies between the different collagens. Although it is efficiently removed from type I collagen, the N-terminal propeptide of type III collagen appears to be a normal constituent of the interstitial ECM (6 -8). The exact way by which type III procollagen is processed, the size of the pool of the N-propeptide that is not removed in different tissues, and its function(s) are currently not known. In contrast to the propeptides, the C-and N-terminal telopeptides of secreted mature collagen molecules are fully retained. The telopeptides, as well as the N-terminal propeptide of type III procollagen, serve as good antigenic markers due to their unique collagen type-specific sequences.
Type I and III collagens isolated from tissues can polymerize in vitro, and therefore the fibrillogenesis in tissues has been thought to occur in a similar way via self-assembly guided by precise interactions between collagen molecules followed by organization of fibrils into fibers (5, 9 -11). However, it is now clear that the polymerization process is influenced by several factors, in particular by other ECM components, although little is known about the mechanisms involved. For example, members of the leucine-rich repeat protein (LRRP), thrombospondin, and tenascin protein families have been found to interact with fibrillar collagens and to affect the fiber number and thickness (12)(13)(14).
One of the best-studied interaction partners of type I and III collagens is fibronectin (FN). FN can bind directly to several collagens (15)(16)(17)(18), and a collagen-binding region in FN has been characterized (19 -21). Extensive co-distribution of FN with both type I and type III collagen has been demonstrated in tissues as well as in cell cultures. Therefore, it is likely that the polymerization reactions of these proteins are somehow coordinated. Early on it was suggested that formation of the FN network precedes and regulates the deposition of type I and III collagens (22), but solid data were missing. Since then, the polymerization of FN has been intensely studied and found to be a highly regulated event that occurs on cell surfaces, being dependent on the intracellular cytoskeleton-related tension and on binding to integrins (2,23,24). Furthermore, the concept of receptor-mediated matrix deposition has been extended to the polymerization of laminins (1,(25)(26)(27) and the formation of basement membranes. This raises the question whether collagen deposition is also regulated by the cells through specific cell surface receptors.
To study the possible role of FN and collagen-binding integrins in collagen matrix assembly, we have used a FN knockout mouse fibroblast cell line that also lacks collagen-binding integrins. Using these cells, and subclones obtained after transfection with integrin ␣ 11 and ␣ 2 subunits, we have found that FN has a crucial role in the polymerization of fibrillar type I and III collagens and that integrins ␣ 11 ␤ 1 and ␣ 2 ␤ 1 strongly enhance this process.

EXPERIMENTAL PROCEDURES
Transfections-The FN-deficient mouse embryonic fibroblasts (clone 4D) were grown to 70 -80% confluence on a 96-well plate, and a total of 15 g of ␣ 2 -pBJ-1 (28) and ␣ 11 -pBJ-1 (29) constructs were used in separate transfections with GenePORTER reagent (Gene Therapy Systems, Inc.) according to the manufacturer's instructions. The constructs were co-transfected with the PGK/puro vector (30) at a ratio of 5:1, and resistant clones were selected in the presence of 5 g/ml of puromycin. Individual clones were isolated and further assayed for ␣ 11 and ␣ 2 expression by cell attachment and FACS analysis (␣ 2 transfectants) followed by metabolic labeling and immunoprecipitation with specific antibodies for ␤ 1 , ␣ 11 , and ␣ 2 . The clones chosen for further studies are referred to as 4D␣ 11 and 4D␣ 2 in this article.
Cell Culture-Generation of the FN-deficient mouse embryonic fibroblasts has been described previously (31). The untransfected FN Ϫ/Ϫ cells (clone 4D) were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, L-glutamine (2 mM), Fungizone (2.5 g/ml), streptomycin (100 g/ml), and penicillin (100 units/ml) at 37°C in humidified atmosphere supplemented with 5% CO 2 . For the transfected cells, puromycin was added to 5 g/ml. 50 g/ml ascorbic acid was used with all cells, and 10 g/ml human plasma FN was added to the culture medium where indicated. Cell cultures were used at about 70 -80% confluence after 3-5 days in culture.
Antibodies-Anti-integrin antibodies used were the following: mouse anti-human ␣ 2 (MAB 1950, Chemicon International, Inc.), hamster anti-rat ␤ 1 monoclonal antibody (BD PharMingen), and the polyclonal anti-␣ 11 antibody, produced against a synthetic peptide comprising the cytoplasmic domain of the ␣ 11 chain, which is described elsewhere (32). The polyclonal anti-ICTP (carboxyterminal telopeptide of type I collagen) and anti-PIIINP (aminoterminal propeptide of type III procollagen) antibodies were obtained by immunizing rabbits with human antigens isolated from human bone (33) and ascitic fluid (34), respectively. The monospecific antibodies were purified from the antisera by immunoadsorption on immobilized ICTP or PIIINP. The chicken-antihuman FN antibody has been described previously (35). Secondary antibodies were goat-anti-rabbit Cy3 and donkey-anti-chicken fluorescein isothiocyanate (Jackson ImmunoResearch).
Cell Attachment Assay-Cell attachment assay was performed essentially as described before (2). Briefly, wells in a 96-well microtiter plate were coated with type I collagen (Vitrogen) in a range of concentrations from 1 to 50 g/ml human plasma FN (30 g/ml) and 1% BSA at 37°C for 1 h and blocked with 1% BSA at 37°C for 1 h. Cells in a serum-free medium were seeded (10 5 cells/well) and incubated for 1 h at 37°C. After washing the wells once with PBS, the attached cells were fixed for 10 min in 96% ethanol at RT, stained with crystal violet for 30 min, washed, and lysed in 1% Triton X-100 for 20 min on a shaker. Bound dye was quantified by measuring the absorbance in the enzyme-linked immunosorbent assay reader at 595 nm. Each experiment was performed in triplicate. Adhesion to FN was used as 100% reference, and the background found on BSA-coated wells was subtracted from all measurements.
Immunofluorescence-Cells cultured on coverslips were washed in PBS and fixed for 10 min in 2% paraformaldehyde at RT. Nonspecific binding sites were blocked with 10% goat serum (Statens veterinä rmedicinska anstalt) diluted in PBS. The cells were then incubated with primary antibodies for 1 h at 37°C, washed 3ϫ5 min in PBS/Tween followed by incubation with Cy3-conjugated goat-anti-rabbit and fluorescein isothiocyanate-conjugated donkey-anti-chicken IgG secondary antibodies for 40 min at 37°C. Stained cells were washed 3ϫ5 min in PBS/Tween, mounted in ProLong Antifade mounting medium (Molecular Probes), and visualized and photographed under a Zeiss Axiophot microscope equipped with optics for observing fluorescence.
Metabolic Labeling and Immunoprecipitation-Cells in 25 cm 2 culture flasks were washed three times with Dulbecco's modified Eagle's medium devoid of cysteine and methionine and metabolically labeled in the presence of 25 Ci/ml of [ 35 S]methionine/cysteine (pro-Mix [ 35 S] cell labeling mix, Amersham Biosciences). Integrins were immunoprecipi-tated from cells after labeling for 16 h, and collagen ␣ chains after pulse-labeling for 30 min, followed by chase periods of 1, 5, and 3 h. Proteins were extracted from the cells with 1 ml of solubilization buffer (1% Triton X-100, 0.15 M NaCl, 20 mM Tris-HCl, pH 7.4, 1 mM MgCl 2 , 1 mM CaCl 2 ) containing protease inhibitors (2 mM phenylmethylsulfonyl fluoride, 1 g/ml pepstatin, 1 g/ml leupeptin). Solubilized proteins were centrifuged for 3 min at 13,000 ϫ g, and the supernatant was precleared with protein A-Sepharose CL 4B (Amersham Biosciences) for 2 h. Following centrifugation, the cell extract was incubated with the immune IgG for 2 h. Antibody-antigen complexes were recovered with protein A-Sepharose, washed three times with lysis buffer, and boiled for 3 min in SDS-sample buffer. Proteins were separated on 7.5% SDS-polyacrylamide gels and processed for fluorography.
Western Blotting-Equal numbers of transfected and control 4D cells were seeded into 25 cm 2 cell culture flasks. The culture medium was supplemented with 50 g/ml ascorbic acid for all cells and 10 g/ml of human plasma FN where indicated. After 3-5 days, the cells were washed twice in PBS, and cellular fraction was extracted by incubating the cultures twice with 500 l of 1% Triton X-100 solubilization buffer for 10 min at 4°C with gentle rocking. Lysates from both incubations were pooled, and the total protein content was measured using the BCA Protein Assay Reagent (Pierce). The Triton-insoluble matrix fraction remaining attached to the plastic was solubilized directly in SDS sample buffer without reducing agents. The samples were loaded on a 7.5% SDS-polyacrylamide gel in proportion to the amount of the total protein in cultures, and after electrophoretic separation the proteins were transferred to nitrocellulose membranes (Schleicher & Schuell). The membranes were blocked in 1% BSA/TBS ϩ 0.1% Tween, sequentially incubated with primary antibodies and HRP-conjugated anti-rabbit secondary antibody at RT, and bound antibodies were detected by ECL (Amersham Biosciences).

Establishment of Collagen-binding FN-deficient Cells-The
FN-deficient cell line 4D was found to attach avidly to FN but to be unable to adhere to type I collagen (Fig. 1). To establish cells able to bind collagens, 4D cells were stably transfected with cDNAs encoding the integrin subunits ␣ 11 or ␣ 2 . The cells expressing either ␣ 11 or ␣ 2 integrin chains exhibited collagenbinding ability, with saturating attachment at ϳ10 g/ml of collagen coating concentration (Fig. 1). The ␣ 11 and ␣ 2 integrin chains could be immunoprecipitated with the specific antibodies, as well as co-immunoprecipitated with the ␤ 1 subunit, from the metabolically labeled transfected cells. A mobility shift characteristic to the ␣ 11 and ␣ 2 chains analyzed in SDS-PAGE under reducing conditions was seen (data not shown).
Changes in Type I and Type III Collagen Distribution Caused by FN and/or Integrins ␣ 11 ␤ 1 and ␣ 2 ␤ 1 -Antibodies specific for FN, the ICTP and the PIIINP, respectively, did not stain the ECM of 4D cells cultured in FN-free medium (Fig. 2, a). Addition of soluble FN to these cells resulted in formation of FN networks as expected, and both type I and type III collagen were detected as fibrils co-localizing with FN (Fig. 2). The exogenously added FN was organized into a dense network much faster compared with collagens, and by the time the cells had built a continuous collagen network, the underlying FN matrix was so dense that the staining became very unclear. Therefore, capturing both FN (green) and collagen (red) stainings on the same image was complicated, and the focus was put on collagen. When the cells were cultured without ascorbic acid, there was essentially no collagen in the matrix, although the FN network appeared to be normal (not shown). To analyze the possible role of collagen-binding integrins in the formation of the ECM, the 4D␣ 11 and 4D␣ 2 cells were cultured in the presence of soluble FN and stained with antibodies to FN, ICTP, and PIIINP. The strongest indication for the role of ␣ 11 ␤ 1 and ␣ 2 ␤ 1 integrins in collagen deposition was obtained from the 4D␣ 11 and 4D␣ 2 cells cultured without soluble FN where short collagen deposits were found in close apposition to the cell surfaces (Fig. 2, c and e). These fibrils had an abnormal morphology, were disorganized, and were never arranged into a network. The type I collagen network assembled by the 4D␣ 11 and 4D␣ 2 cells in the presence of FN was not remarkably different from that in cultures of 4D cells (Fig. 2, ICTP, d and  f). However, the type III collagen detected in the cultures of transfected cells contained prominent thick fibers in addition to the thin network observed in the case of type I collagen (Fig. 2,  PIIINP, d and f). In contrast to the thin collagen strands, the thick fibers did not co-localize with the FN network, and they never occurred in the 4D cells (Fig. 2, PIIINP). The network of type III collagen fibrils was more dense, and the individual fibrils were thicker in cultures of 4D␣ 11 cells than in cultures of 4D␣ 2 .
Detection of Type I and Type III Collagens in the Insoluble Matrix Deposited by the 4D, 4D␣ 11 , and 4D␣ 2 Cells-Isolated ECM from cells grown under identical conditions as for immuonocytochemistry was analyzed by Western blotting. Antibodies against both ICTP and PIIINP detected prominent bands corresponding to type I and III collagens, respectively, in the matrix fractions of the 4D␣ 11 and 4D␣ 2 cells grown in the presence of FN (Fig. 3). Weak signals were also detected in the samples from 4D␣ 11 and 4D␣ 2 cells grown without FN, probably corresponding to the secreted collagens seen in Fig. 2, c and e. Essentially no signal was obtained from the untransfected   11 , and 4D␣ 2 cells. Cells were grown, and the insoluble ECM fraction was extracted as described under "Experimental Procedures." Equal amounts of proteins from the insoluble matrix fraction of different cell lines were separated by SDS-PAGE, transferred to the nitrocellulose membranes, and the membranes were incubated with the antibodies against ICTP and PIIINP followed by the HRP-conjugated secondary antibody and detection by the ECL. A filter containing identical samples was incubated with purified rabbit IgG instead of a primary antibody. cells. No bands were detected on an identical filter incubated with non-immune rabbit IgG (Fig. 3, PI).

The Rate of Collagen Synthesis in 4D Cells Is Not Affected by the Presence of Collagen-binding Integrins-
To determine whether the differences in collagen deposition between the 4D, 4D␣ 11 , and 4D␣ 2 cells were solely dependent on their capability of incorporating the collagens into matrices, or if the presence of FN and/or collagen receptors would affect collagen synthesis, the cells were cultured as for immunochemistry and Western blotting, and the cell lysates were subjected to immunoprecipitation with the anti-ICTP and PIIINP antibodies. We found that the rate of synthesis of type I and type III procollagen ␣ chains was largely the same in all three cell lines (data not shown).

DISCUSSION
In this paper we have addressed questions regarding the assembly of type I and type III collagens into the extracellular matrix. We have analyzed the role of a pre-existing FN matrix as well as collagen-binding integrins in collagen matrix assembly and in collagen synthesis by the cells in a monolayer culture. A FN-deficient mouse embryonic fibroblast cell line (clone 4D), which also lacks the integrin-type collagen receptors, was used. The cells were transfected with the full-length cDNAs for human integrin ␣ 11 (29) and ␣ 2 (28) subunits. Obtained cell lines (4D␣ 11 and 4D␣ 2 ), together with the untransfected 4D cells as controls were cultured in the presence or absence of ascorbic acid and human plasma FN and subjected to immunohistochemistry, immunoblotting, and metabolic labeling and immunoprecipitation.
The interdependence of FN and collagen networks has previously been studied by several approaches (37)(38)(39). Mov13 fibroblasts, where the gene for the collagen ␣1(I) chain had been inactivated and no type I collagen was synthesized, produce a sparse FN matrix that contains only short FN fibrils. The ability of the cells to deposit a normal FN matrix could be restored by transfection of the wild type collagen ␣1(I) chain, but not by expression of the collagen ␣1(I) chain with mutated FN-binding sites or by adding type III or V collagens to the cell cultures, suggesting that FN polymerization is specifically dependent on the presence of type I collagen (39). In Schwann cells, the assembly of FN, as well as type I and IV collagens and perlecan, into the ECM has been reported to be strictly dependent on ascorbic acid or exogenously added type IV collagen (40). Soluble FN has been shown to have a globular configuration, which has been proposed to unfold into an elongated form during fibrillogenesis (41). Interaction of soluble FN with collagens and heparan sulfate chains has been reported to induce a similar conformational change (42), suggesting that these compounds may promote FN polymerization in vivo, possibly by exposing integrin-binding sites in FN (43). The reports concerning the dependence of FN polymerization on collagens would support such a mechanism (39,40).
However, although the above works argue that FN polymerization is regulated by the presence of collagens in the ECM, the opposite has been reported by others. This may suggest that the role of collagens in FN matrix deposition is cell typedependent (40). For example, McDonald et al. (38) have shown that blocking the FN-binding sites on collagen completely abrogated the assembly of the collagen network leaving the FN matrix intact. The same authors could observe a normal deposition of FN matrix by ascorbate-deficient primary chick embryo fibroblasts and the absence of detectable collagen in these cultures. The formation of a FN matrix in cultures of tenascin-X-deficient fibroblasts was recently shown to occur normally, although the deposition of collagen into the ECM was strongly reduced (14). In our cell system, FN matrix deposition was unaffected regardless of the culture conditions applied.
The context in which we conducted our study allowed us to investigate the assembly of collagens independently of FN. We could show that there was no collagen matrix built in the absence of a pre-formed FN matrix. Differently from the FN matrix assembly, the assembly of collagens into the ECM was strictly dependent on the presence of ascorbic acid in the culture medium. The addition of soluble FN to the cells resulted in the formation of a FN matrix followed by assembly of a FNassociated collagen network. Furthermore, we found that in 4D cells expressing integrins ␣ 11 ␤ 1 and ␣ 2 ␤ 1 , significantly more collagen, the type III collagen in particular, was deposited into the matrix. In addition to the thin collagen fibrils co-localizing with the FN in the 4D cells, thick type III collagen fibers that did not co-localize with FN fibers were seen in cultures of the 4D␣ 11 ␤ 1 and 4D␣ 2 ␤ 1 cells. Surprisingly, we could detect small amounts of type I and type III collagen deposits in cultures of cells expressing integrins ␣ 11 ␤ 1 and ␣ 2 ␤ 1 also in absence of FN, which were always found in a close apposition to the cell surface and were never arranged into a network (Fig. 2, c and e). It is possible that the ␣ 11 ␤ 1 and ␣ 2 ␤ 1 integrins present on these cells function as nucleation centers for secreted collagens, but for formation of a collagen network a FN scaffold is required. Further studies are needed to elucidate the specific roles of these integrins in collagen polymerization.
In the tissue, the collagen fibrils associate to form fibers, which in turn build a matrix stabilized by cross bridging molecules. This process involves other collagens, such as collagen IX in cartilage (44) as well as non-collagenous molecules such as fibromodulin (45)(46)(47), lumican (48), decorin (45, 49 -51), and cartilage oligomeric matrix protein (COMP) (52), all capable of specific binding to triple helical collagen. For example, decorin has been shown to regulate the thickness of collagen fibrils and fibers by binding to the collagen monomers and preventing them from associating laterally (13). The decorin-null mice have irregular collagen fibrils with gigantic diameters resulting from the uncotrolled lateral fusion (53). Thus, appropriate organization of collagens in tissues requires both negative (e.g. decorin) and positive (e.g. FN and integrins) regulatory factors.
A direct involvement of the collagen-binding integrins in regulation of collagen synthesis and collagenase/matrix metalloproteinase (MMP) production has been reported by several groups (54 -57). Mice lacking the integrin ␣ 1 subunit were shown to have strongly elevated levels of type I collagen synthesis as well as higher levels of collagenase-3 (MMP-13) (57). Integrin ␣ 2 ␤ 1 has been shown to induce the collagenase (MMP-1) expression in human skin fibroblasts in three-dimensional collagen lattice (56), but no effect on collagen synthesis has been reported. Curiously, integrin ␣ 2 subunit-deficient mice do not have any detectable extracellular matrix-related phenotype, but instead display a subtle dysfunction of platelets (58). This unexpected finding may be due to possible compensatory effects of other collagen-binding integrins, e.g. integrin ␣ 1 ␤ 1 in vivo. In our case, the dramatic effect of the presence and engagement of the integrins ␣ 11 ␤ 1 and ␣ 2 ␤ 1 on collagen deposition and polymerization by the 4D cells suggested that the levels of collagen synthesis could be altered by these integrins. However, we could not observe a change in the levels of synthesized collagen ␣ chains in our system under the cell culture conditions used. It is possible that fibroblasts need to be embedded in a three-dimensional environment to influence collagen synthesis. This does not seem to apply for epithelial cells that have been shown to up-regulate a number of mRNAs and proteins in response to adhesion to two-dimensional surfaces coated with ECM proteins (36).
In summary, our results regarding the formation of ECM in fibroblasts demonstrate that: 1) no collagen matrix forms in the absence of FN and collagen receptors and that the presence of FN alone is sufficient for collagen polymerization; 2) even in the absence of FN, integrins ␣ 11 ␤ 1 and ␣ 2 ␤ 1 are able to promote fibrillogenesis of type I and type III collagens to some extent, although no proper collagen network is formed; 3) the presence of collagen receptors and FN together results in formation of a well organized collagen network that, in the case of type III collagen, is denser and contains thicker fibers compared with type I collagen; and 4) changes in the appearance of the collagen matrix are not caused by changes in levels of collagen synthesis, and thus collagen receptors have no such regulatory role under applied cell culture conditions. Taken together this points to mutually supportive roles of FN/integrin ␣ 5 ␤ 1 and integrins ␣ 11 ␤ 1 and ␣ 2 ␤ 1 in the assembly of type I and III collagens into the ECM.