Cellular Fibronectin Binds to Lysyl Oxidase with High Affinity and Is Critical for Its Proteolytic Activation*

Lysyl oxidase (LOX) is a copper-containing amine oxidase known to catalyze the covalent cross-linking of fibrillar collagens and elastin at peptidyl lysine residues. In addition, its involvement in cancer, wound healing, cell motility, chemotaxis, and differentiation reflect a remarkable functional diversity of LOX. To investigate novel mechanisms of LOX regulation and function, we performed a yeast two-hybrid screen to identify LOX-interacting proteins. Three overlapping positive clones were identified as C-terminal fragments of fibronectin (FN). Glutathione S -transferase pull-downs and solid phase binding assays confirmed this interaction. LOX binds to the cellular form of FN (cFN) with a dissociation constant ( K d ) of 2.5 n M . This was comparable with our measured K d of LOX binding to tropoelastin (1.9 n M ) and type I collagen (5.2 n M ), but LOX demonstrated a much lower binding affinity for the plasma form of FN (pFN). Immunofluorescent microscopy revealed co-localization of FN and LOX in normal human tissues, where these proteins may interact in vivo . LOX enzymatic activity assays showed that cFN does not seem to be a substrate of LOX. How-ever, cFN can act as

Lysyl oxidase (LOX), 1 a copper-dependent amine oxidase, catalyzes the oxidative deamination of peptidyl lysine residues in collagen and elastin molecules to ␦-aminoadipic-␤-semialde-hyde or allysine, which can then spontaneously condense with neighboring amino groups or other peptidyl aldehydes to form covalent cross-links in fibrillar collagens and elastin (1,2). The covalent cross-linking of these extracellular matrix (ECM) proteins by LOX is essential to the formation of insoluble collagen and elastic fibers and for normal mammalian development (3,4). In many human pathologies, including cardiovascular disease, fibrosis, and cancer, LOX expression and activity have been demonstrated to be altered as compared with normal conditions (1,2).
The human LOX protein is synthesized as a prepro-enzyme of 417 amino acids with an N-terminal signal peptide sequence for secretion. During protein trafficking in the Golgi, a copper atom is incorporated at the copper-binding site (residues 286 -296) (5)(6)(7), which allows the formation of a lysyl-tyrosylquinone cofactor derived from Tyr-355 and Lys-320 residues (8). After secretion into the extracellular space, proteolytic cleavage between residues Gly-168 and Asp-169 removes the N-terminal propeptide, yielding an active enzyme of 30 kDa (9 -11). This processing is mainly accomplished by the procollagen C-proteinase bone morphogenic protein-1 (BMP-1), and to a lesser degree, by mammalian Tolloid-like 1 protein (12,13). However, it is not known where and how the proteolytic activation and amine oxidase activity of LOX are regulated in physiological and pathological conditions in the ECM.
Fibronectin (FN) has been well characterized as an extracellular matrix glycoprotein that can regulate many cellular functions such as proliferation, differentiation, migration, adhesion, and apoptosis (14 -18). FN accomplishes this diverse array of functions through interactions with a large variety of proteins including collagens, fibrin, thrombospondin, cell surface integrins, heparin sulfate proteoglycans, and tenascin-C. FN is secreted as a dimer bound together by C-terminal disulfide bonds, and it has been traditionally categorized as either soluble plasma FN (pFN) or insoluble cellular FN (cFN) (14,19,20). It has been shown that activation of soluble FN induces a conformational change resulting in an insoluble molecule that is highly adhesive through revealed binding domains (21)(22)(23)(24)(25).
In tissues, activated FN is deposited and organized into a polymeric matrix, which has long been observed to be associated with collagen fibers. Classic studies of wound healing have revealed a temporal sequence of FN deposition followed by cell invasion, then type III collagen deposition followed by type I collagen deposition, and finally, loss of FN (17,18). Recently, the formation of the FN matrix has been shown to be critical for the subsequent assembly of types I and III collagen fibrils (26,27). It has also been demonstrated that cFN is required for collagen gel contraction in wound healing models (28 -30). Lysyl oxidase enzyme activity has also been shown to be a critical factor in similar collagen gel contraction models (31,32). The * These studies were supported by Grants AR47713 and G12RR03961 from the National Institutes of Health (to K. C.) and by a Fellowship 0315258Z from the American Heart Association (to B. F.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U. S FN matrix was proposed to act as a scaffold to regulate the deposition and assembly of type I and III collagens, but the molecular mechanisms by which the FN matrix may accomplish this have not been established.
In this study, we report the identification of FN as a binding partner for LOX detected by a yeast two-hybrid screen using the mature 30-kDa LOX protein as bait. We demonstrate that LOX has a high binding affinity for cFN, but not pFN, and that the cytokine receptor-like (CRL) domain of LOX seems to play a role in stabilizing this binding. We also show that FN does not seem to be a substrate of LOX but that FN-bound LOX retains its amine oxidase activity. Immunofluorescent confocal microscopy revealed LOX and FN co-localization in cultured fibroblasts and various human tissues. In FN-null mouse embryonic fibroblast cultures, we detected dramatically decreased proteolytic processing of LOX with a corresponding decrease in catalytic activity.

EXPERIMENTAL PROCEDURES
Reagents-Monoclonal anti-c-Myc antibody (9E10) was purchased from Covance, Inc. Monoclonal anti-FN antibody (FN8 -12), which recognizes all isoforms of FN, was purchased from Takara Bio Inc. Monoclonal anti-glutathione S-transferase (GST) antibody was purchased from Upstate Biotechnology. Polyclonal rabbit anti-LOX antibody was previously characterized and recognizes both the full-length and the processed LOX from human and mouse (33). Polyclonal anti-BMP-1 antibody, which recognizes the CUB-2 domain, was purchased from Affinity Bioreagents, Inc. Purified rabbit IgG, used as an isotype-negative control for immunofluorescent staining, was purchased from Zymed Laboratories Inc. Secondary antibodies for immunofluorescent staining, which included anti-mouse antibody conjugated to Alexa Fluor 546 (red) to visualize the anti-FN antibody and anti-rabbit antibody conjugated to Alexa Fluor 488 (green) to visualize the anti-LOX antibody and the purified rabbit IgG-negative control, were purchased from Molecular Probes, Inc.
Purified active bovine LOX (bLOX) was generously provided by Dr. Hebert Kagan (34). Purified pFN, cFN, bovine serum albumin (BSA), and chick tropoelastin proteins were purchased from Sigma. Purified rat tail collagen type I was purchased from BD Biosciences.
Yeast Two-hybrid System-Primers were designed to introduce restriction sites 5Ј and 3Ј to sequence verified LOX cDNA to clone the inserts into Clontech's pGBKT7 vector downstream and in-frame with the GAL4 DNA-binding domain. The primers used to amplify the fulllength LOX cDNA were: 5Ј-ACA TGC CAT GGT GAT GCG CTT CGC CTG GAC CG-3Ј (forward) and 5Ј-CGC GGA TCC CTA ATA CGG TGA AAT TGT GCA GC-3Ј (reverse), with the introduced NcoI and BamHI restriction sites underlined. To express only the mature form of LOX (amino acids 169 -417), a forward primer was designed that started at the known BMP-1 cleavage site (10): 5Ј-CCG GAA TTC GAC GAC CCT TAC AAC CCC TA-3Ј (forward), with the introduced EcoRI site underlined. After PCR amplification, the inserts were digested and ligated into the pGBKT7 vector. The finished constructs, pGBKT7-proLOX and pGBKT7-LOX, were each verified by DNA sequencing.
Two-hybrid screening was conducted according to Clontech's System 3 protocol using a human placental cDNA library (Clontech), which contained over 3.5 ϫ 10 6 independent clones generated from human placental mRNA with oligo(dT) primers. The library was transformed into AH109 yeast that contained pGBKT7-LOX with an efficiency measured by dilutions of transformants on non-selective plates. Clones containing interacting proteins were selected by growth on a drop-out minimal medium lacking leucine, tryptophan, histidine, and adenine and by activation of the lacZ gene. The positive clones were streaked multiple times on these selective plates to get single clones and to verify their growth under selective conditions. The library plasmids were isolated from the positive yeast clones, and the cDNA inserts were sequenced using an ABI 310 capillary sequencer. The interactions of the identified FN clones were further characterized by performing direct interaction assays in AH109 yeast. pGBKT7, pGBKT7-proLOX, pG-BKT7-LOX, or pGBKT7-lamin C (Clontech) was co-transformed with pGADT7 or pGADT7-FN. The selection for positive interactions was carried out with the same stringent conditions as the original twohybrid screen.
Cell Culture-Human neonatal foreskin fibroblasts were obtained from Clonetics, Inc. and grown using Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS) (Se-rologicals Corp.), penicillin, streptomycin, and amphotericin. Mouse embryonic fibroblasts (MEFs) from FN-null mice (FNϪ/Ϫ) and heterozygous siblings (FNϩ/Ϫ) were generously provided by Dr. Deane Mosher (35). MEFs were kept in culture under the same conditions as human neonatal skin fibroblasts as described above, except 10% FNdepleted FBS was used to supplement the DMEM. The removal of FN from the FBS was accomplished by mixing 4 ml of gelatin-Sepharose (Amersham Biosciences) with 25 ml of FBS overnight and then centrifuging to remove the resin. Only human and mouse primary fibroblasts of less than passage 10 were used for protein analysis. For experiments analyzing cell medium proteins, cells were grown to 2 days post-confluency, washed with PBS, and incubated with serum-free, phenol red-free DMEM. After an additional 48 h in culture, the conditioned cell medium (CCM) was collected. To concentrate total proteins from the CCM for Western blotting, we added 10 l of Strataclean Resin (Stratagene) to the appropriate volume of CCM, and then the sample was vortexed and mixed for 30 min at 4°C (36). After centrifugation, the supernatant was removed, and we added 2ϫ Laemmli buffer to the resin, boiled it for 5 min, and then froze the sample.
GST Protein Expression and Pull-downs-The plasmid pGEX-4T-1 (Amersham Biosciences) was used to generate expression constructs for various LOX fragments containing an N-terminal GST tag. The pG-BKT7-proLOX plasmid was used as a cDNA template for cloning the LOX fragments. Recombinant GST⅐LOX fusion proteins, corresponding to amino acids 1-417 (proLOX), 169 -417 (mature LOX), 1-168 (propeptide), 169 -348 (mature LOX minus the CRL domain), and 349 -417 (CRL domain), were cloned and transformed into the BL21 strain of Escherichia coli (Stratagene). GST fusion protein expression was induced by adding 0.1 mM isopropyl-1-thio-␤-D-galactopyranoside to growing cultures and shaking for an additional 2 h at 37°C. Recombinant proteins were extracted from inclusion bodies using a solubilization buffer (8 M urea, 10 mM K 2 HPO 4 , 5 mM dithiothreitol, pH 8.2), filtered through a 0.45-m syringe filter, and refolded with a rapid dilution into 10 mM K 2 HPO 4 buffer (pH 8.2). GST⅐LOX fusion proteins were captured and purified using glutathione-Sepharose 4B (Amersham Biosciences) and then eluted using the inclusion body solubilization buffer and refolded. Protein concentrations were measured using Bradford reagent (Bio-Rad).
For GST pull-down experiments, 5 g of GST⅐LOX or GST alone was attached to glutathione-Sepharose 4B resin and incubated with 10 g of purified pFN, 10 g of purified cFN, or 1 ml of human neonatal skin fibroblast CCM for 1 h at 4°C. The resin was washed three times with PBST (0.05% Tween 20) and then resuspended in 30 l of 2ϫ Laemmli buffer and boiled for 5 min, and equal amounts were electrophoresed by SDS-PAGE. Bound FN was detected by Western blot analysis using an anti-FN monoclonal antibody.
Solid Phase Binding Assays-To determine the binding affinity of LOX to FN as compared with various ECM and control proteins, solid phase binding assays were performed (37). Wells of high protein-binding EIA/RIA microplates (catalog number 3590; Corning) were coated with purified proteins at 200 nmol/well in PBS or 10 mM K 2 HPO 4 (pH 8.2) overnight at 4°C. Wells were blocked with 1% BSA in PBS for 3 h at 37°C. After removing the blocking solution, wells were washed three times with 0.1% BSA in PBS. Purified soluble "ligand" protein was added to the wells at various concentrations (0 -100 nM) in PBS or 10 mM K 2 HPO 4 (pH 8.2) and incubated overnight at 4°C. The wells were then washed three times with PBST, and bound ligand was reacted with anti-GST (Upstate Biotechnology) or anti-FN primary antibody and then detected with peroxidase-labeled anti-mouse secondary antibody (Amersham Biosciences). Peroxidase activity was quantitated using the Quant-aBlu fluorogenic peroxidase substrate kit (Pierce), with absorbance measured at 405 nm. All samples were performed in triplicate. The dissociation constants were calculated using non-linear regression analysis performed by Prism3 statistical software (Graphpad, Inc.).
Immunofluorescent Microscopy-Cultured human fibroblasts were grown, as described above, on glass coverslips until 6 days post-confluency and then washed with PBS three times and fixed with 10% formalin for 10 min at room temperature. After fixation, the cells were treated with 0.1% Triton X-100, washed again in PBS, and blocked with 0.1% BSA in PBS for 30 min at room temperature. The fixed cells were reacted with primary antibodies diluted in 1% BSA in PBS for 1 h at room temperature and then washed in PBS with 0.1% BSA three times for 5 min each. The bound primary antibodies were reacted with fluorescently labeled secondary antibodies diluted in 1% BSA in PBS for 1 h at room temperature, and then the coverslips were washed three times in PBS with 0.1% BSA for 5 min each. After a final wash in PBS for 5 min, the coverslips were mounted using Vectashield mounting medium (Vector Laboratories). Immunofluorescent staining on fixed human tis-sues was performed using human tissue macroarrays (catalog number 70312-3 and 70313-3, Novagen). The tissue sections were deparaffinized by heating at 60°C for 30 min and washing three times in xylene for 15 min. Tissue sections were then rehydrated in 100, 95, 80, and 70% ethanol followed by immersion in PBS. When needed, antigen retrieval was performed by microwaving the sections in 0.01 M citric acid on high, medium, and then low power for 5 min each. The sections were washed in PBS three times for 5 min and then blocked with 5% normal goat serum (Pierce) at room temperature for 30 min. The sections were reacted with primary antibodies for 2 h at room temperature, washed three times in PBST, and then incubated with fluorescently labeled secondary antibodies for 45 min at room temperature. After three final washes in PBS, Vectashield was used to mount coverslips. All slides were analyzed with a Zeiss LSM Pascal confocal microscope.
LOX Activity Assays-The LOX enzyme activity was measured using the Amplex Red fluorescence assay (38), which we have adapted for use in microplate format. The assay reaction mixture consisted of 50 mM sodium borate (pH 8.2), 1.2 M urea, 50 M Amplex Red, 0.1 units/ml horseradish peroxidase, and 10 mM 1,5-diaminopentane (cadaverine) substrate. The protein samples were added to the reaction mix, in the presence or absence of 500 M BAPN, which was then incubated at 37°C. The fluorescent product was excited at 560 nm, and the emission was read at 590 nm using a BMG Labtechnologies Inc. Polarstar Optima. To measure whether LOX activity is affected by binding to FN, 10 pmol of purified bLOX was incubated with 10 pmol of purified cFN or BSA at 4°C for 1 h. To determine whether LOX retains amine oxidase activity when bound to a solid phase FN matrix, we performed an experiment similar to the solid phase binding assays. We coated microplate wells with cFN, and after blocking with BSA, we incubated bLOX in the wells at a concentration range of 0 -100 nM overnight at 4°C. We washed away the unbound bLOX three times with PBST and once with PBS and then added 200 l of the above Amplex Red reaction mix (with 1,5-diaminopentane substrate) without urea. Reactions were incubated at 37°C for 60 min, and the fluorescence was measured as above. To measure LOX activity in MEF culture medium, an equal microgram amount of CCM proteins from FNϩ/Ϫ and FNϪ/Ϫ cultures were added to each microplate well and incubated with the Amplex Red reaction mix listed above, using 1,5-diaminopentane as a substrate. Parallel samples were incubated with 500 M BAPN as a negative control. The reactions were incubated at 37°C, and the fluorescence was measured every 5 min. The LOX activity was calculated as the increase in fluorescent units over time (120) above the BAPN controls. Replicate samples were used to decrease experimental error. LOX activity assays performed from MEF CCM required the concentration of medium proteins using 10,000 molecular weight cut-off Amicon Ultra centrifugal filter units (Millipore).

RESULTS
Yeast Two-hybrid Screening Identified LOX Interactions with FN-To investigate novel mechanisms of LOX regulation and function, we performed a yeast two-hybrid screen to identify LOX-interacting proteins. The expression of the lysyl oxidase gene is high in placental tissues, and we anticipated that using a placental library would be optimal for discovering proteins with LOX-binding ability. It was unknown whether the N-terminal propeptide region of LOX (amino acids 1-168) would inhibit protein interactions, as it has been theorized to electrostatically interact with the active site region of LOX to inhibit enzyme activity (39). Additionally, the LOX protein in the ECM has been shown to be predominantly the processed active 30-kDa form. Therefore, a bait designed to encode the 30-kDa mature form of LOX (amino acids 169 -417) was used to screen a human placental yeast two-hybrid library (Fig. 1A).
After transformation of pGBKT7-LOX into yeast strain AH109, expression of the fusion protein GAL4-BD-LOX was confirmed by Western blot (data not shown). Assays for expression of the reporter genes ade2, his3, and lacZ showed no autonomous transcriptional activity by GAL4-BD-LOX alone. Subsequently, the placental cDNA library was transformed into the yeast strain containing pGBKT7-LOX, and the cells were plated to screen for expression of the three reporter genes. From plated dilution controls, the library transformation efficiency was over 5 ϫ 10 6 colony-forming units/g. Yeast colonies that expressed all three reporter genes were isolated, and the library insert cDNA was sequenced.
From this yeast two-hybrid screen, three of the positive clones we isolated were overlapping FN cDNA fragments. Because the placental cDNA library was constructed using oligo(dT) primers, all the FN fragments were on the C-terminal end of the protein. The shortest FN cDNA encoded amino acids 2057-2446, which contained a heparin-binding domain, a REDV cell adhesion site, the IIICS alternative splicing domain, one type III repeat, and the fibrin-binding domain (Fig. 1). All three FN cDNAs included part of the alternatively spliced IIICS domain (Fig. 1), which is more commonly expressed in cellular forms of FN.
To verify that the binding of FN was specific to LOX and to determine whether FN could also bind to the full-length form of LOX (proLOX), several combinations of baits and target plasmids were co-transformed into yeast and streaked on ϪAde/ϪHis/ ϪLeu/ϪTrp-selective plates. Yeast were also co-transformed with combinations of LOX and FN proteins with empty vectors as negative controls. A non-interacting bait protein, nuclear lamin C, was also used as a negative control to show specificity of the FN interaction. Only the yeast cells that contained LOX or pro-LOX and FN were able to grow on the nutritional dropout plates (Fig. 2). These data demonstrated that the detected interaction between LOX and FN was not the result of nonspecific binding and that the propeptide region does not inhibit the interaction of LOX with the FN clones isolated from the screen.

FIG. 1. Schematic of LOX constructs used in this study and identified LOX-interacting FN fragments.
A, proLOX is secreted and proteolytically processed in the extracellular environment to the active LOX enzyme, which was used as bait for a yeast two-hybrid screen of a human placental library. The numbers designate the included amino acids. SS, signal sequence; arrowhead, BMP-1 proteolytic activation site; Cu, copper-binding site; LTQ, lysyl-tyrosyl quinone cofactor. B, the fibronectin protein has three types of repeated motifs and can be alternatively spliced at the ED-A, ED-B, or the IIICS domain. The three C-terminal FN fragments shown were clones isolated from the yeast two-hybrid screen as LOX-interacting proteins. The numbers designate the included amino acids.

LOX-FN Interaction in Vitro-
To verify and to biochemically characterize the interaction of LOX and FN, further experiments required adequate amounts of purified LOX protein.
Since LOX has been successfully expressed in bacteria (40,41), we cloned LOX cDNA into the pGEX-4T-1 vector to express LOX as a GST fusion protein in E. coli. The N-terminal GST tag allowed for rapid purification of large amounts of the recombinant LOX protein through the high affinity binding to glutathione. We were able to express and purify sufficient amounts of GST⅐LOX from the bacteria inclusion bodies using a slightly modified method of protein refolding based on that of Jung et al. (41). We also cloned the full-length proLOX as a GST fusion protein, but despite attempts to optimize the growing conditions, GST-proLOX never expressed at the correct molecular weight, and it was not used for further experiments. For analysis of individual LOX domains, we cloned three other cDNA fragments that span the entire length of the LOX protein: GST⅐LOX 1-168 , GST⅐LOX 169 -348 , and GST⅐LOX 349 -417 , where the numbers refer to the included amino acids. We successfully expressed the fusion proteins and were able to purify them in sufficient quantities using the same method as for GST⅐LOX (Fig. 3A).
To verify the protein interactions observed in yeast cells, purified recombinant GST⅐LOX was used for pull-down experiments. Equal molar amounts of GST⅐LOX or GST alone were bound to glutathione resin and incubated with FN inputs, and after washing, bound FN was detected by Western blotting. Using purified FN as input, we were able to pull down both plasma and cellular forms of FN with GST⅐LOX but not with GST alone (Fig. 3B). In addition, using CCM from human skin fibroblasts as input, we were able to pull down secreted FN with GST⅐LOX but not with GST alone (Fig. 3B).
To further characterize the binding of LOX to FN and to estimate an equilibrium disassociation constant (K d ) for this interaction, we performed solid phase binding assays. These assays also allowed us to compare the binding affinity of LOX to FN versus type I collagen and tropoelastin. In the first experiment, microplate wells were coated with purified pFN, cFN, type I collagen, tropoelastin, or BSA as a negative control. Purified GST⅐LOX, in a concentration range of 0 -100 nM, was incubated with the immobilized proteins overnight at 4°C. After washing, the quantity of bound GST⅐LOX was measured in each well. The GST⅐LOX protein bound to cFN, type I collagen, and tropoelastin with high affinity, as K d was calculated at 2.5, 5.2, and 1.9 nM, respectively (Fig. 4A). No binding to BSA was detected. The binding of GST⅐LOX to pFN was lower than that to cFN in these assays (Fig. 4A), but above that of the BSA-negative control, showing that there may be a weak interaction with the plasma form of FN.
To test the binding of cFN to solid phase GST⅐LOX and individual LOX domains, we coated microplate wells with purified GST⅐LOX, GST⅐LOX 1-168 , GST⅐LOX 169 -348 , GST⅐LOX 349 -417 , or GST alone. Purified cFN, in a concentration range of 0 -100 nM, was incubated with the immobilized proteins as described. We observed that cFN bound to solid phase GST⅐LOX with a calculated K d of 3.8 nM (Fig. 4B), consistent with the high affinity detected in the previous solid phase binding assay. We also detected cFN binding to the GST⅐LOX 169 -348 protein, although at a lower affinity than the 30-kDa form of LOX, which includes the CRL domain. cFN demonstrated little binding to GST⅐LOX  and the CRL domain alone in GST⅐LOX 349 -417 . No binding to GST alone was detected.
FN and LOX Co-localize in Normal Human Tissues-To determine whether LOX and FN co-localize in vivo, immunofluorescent staining was performed on cultured human neonatal skin fibroblasts and commercially available fixed human tissue sections and analyzed by confocal microscopy. These tissues were co-stained with FN and LOX antibodies, with serial sections stained with the negative control for the LOX antibody, purified rabbit IgG, to demonstrate the specificity of the LOX signal. In the ECM of cultured human skin fibroblasts, LOX was localized in a very punctate pattern as compared with the continuous matrix of FN (Fig. 5A). Co-staining in normal human tissue arrays demonstrated many areas of co-localization, supporting the possibility of an in vivo biological interaction between FN and LOX. The punctate staining pattern of LOX, observed in cultured fibroblasts, was consistent with that seen in distinct microenvironments of fixed tissues. In placental tissues, intense FN and punctate LOX staining surrounded the epithelial cells of the placental villi. LOX was also localized to the stroma and along the basement membrane beneath the epithelium (Fig. 5, B and C). In the esophagus, LOX co-stained strongly with FN along the basal layer beneath the stratified squamous epithelium and above the lamina propria (Fig. 5D). Adipocytes have individual basement membranes surrounding each cell, and adipose tissues showed LOX and FN co-localization closely bordering each cell (Fig.  5E). In vascular tissues where ECM deposition is critical for proper function, FN and LOX were co-localized in the walls of FIG. 3. GST⅐LOX fusion proteins were expressed and purified for in vitro binding assays. A, Coomassie Blue-SDS-PAGE analysis of GST⅐LOX fusion proteins that were purified from E. coli using glutathione resin. The fusion proteins were all expressed at the predicted molecular mass. B, immunoblots of GST pull-down assays using purified pFN, purified cFN, and CCM from human cultured skin fibroblasts were used as inputs. Equal amounts of GST or GST⅐LOX were bound to glutathione resin and incubated with the input. After washing, bound FN protein was detected by immunoblotting using an anti-FN antibody.
veins and arteries (Fig. 5, F and G). In the kidney, both proximal and distal tubules showed strong LOX staining that overlapped with the FN staining around the tubule epithelial cells, particularly the basolateral side of the epithelium (Fig. 5H). Intense FN staining was also observed in the connective tissue between the kidney tubules.
cFN Is Not a Substrate for LOX, but cFN-bound LOX Remains Enzymatically Active-Since cFN bound to LOX with approximately the same affinity as two classic LOX substrates, tropoelastin and type I collagen, we tested whether LOX could oxidize lysyl residues of cFN in vitro. Purified bLOX was incubated with equal molar amounts of purified cFN, tropoelastin, type I collagen, or 1,5-diaminopentane (cadaverine), and the quantity of H 2 O 2 generated during the oxidative reaction was measured using Amplex Red and horseradish peroxidase (38). As a negative control, BAPN was added to parallel reactions to inhibit bLOX activity. The LOX activity was calculated as the increase in fluorescence above BAPN controls. The results show that cFN was not oxidized at statistically significant levels as compared with type I collagen, tropoelastin, or 1,5-diaminopentane (Fig. 6A). Tropoelastin had the highest rate of oxidation, which was 7.9-fold higher than 1,5-diaminopentane and 2.2-fold over type I collagen. Although the data showed that 1,5diaminopentane is less efficiently oxidized by LOX than tropoelastin or type I collagen, in subsequent in vitro LOX activity assays, 1,5-diaminopentane was used because it was possible to use a higher molar excess of substrate in the reaction for more sensitive detection of LOX activity.
To determine whether binding to cFN would inhibit the amine oxidase activity of LOX, we performed an activity assay with bLOX that had been preincubated with an equal amount of cFN. No decrease in LOX activity was seen, as compared with non-bound bLOX and bLOX preincubated with BSA control (Fig. 6B). In addition, we used a combination of solid phase binding assay and amine oxidase activity assay to determine whether LOX bound to solid phase cFN would remain in an active conformation. First we incubated 0, 10, or 100 nM of purified bLOX with immobilized cFN as described above, allowing the proteins to bind overnight. Then we washed away the unbound bLOX protein and performed an amine oxidase activity assay in the same microplate wells by adding the Amplex Red reaction mixture and 1,5-diaminopentane substrate, with BAPN controls in parallel. The results show that LOX, when bound to immobilized cFN, remains enzymatically active toward 1,5-diaminopentane in vitro (Fig. 6C).
LOX Activity Is Decreased in FN-null Cells Due to Decreased Proteolytic Activation-To investigate the role of the cFN-LOX interaction on the LOX protein and its proteolytic activation in vivo, we analyzed LOX produced by embryonic fibroblasts derived from FN-null mice and their heterozygous siblings (35). These cells were cultured in DMEM with FN-depleted FBS. Like many ECM proteins, LOX levels dramatically increased when the MEFs were grown past confluency. Therefore, further LOX protein analysis was performed with cells grown to 2 days post-confluency and then incubated with serum-free medium for another 2 days.
Conditioned cell medium was collected from the FNϩ/Ϫ and FNϪ/Ϫ cells to analyze the LOX protein. Western blot analysis of protein fractions from these cells using a LOX-specific antibody showed dramatically decreased levels of the 30-kDa LOX in the medium of FNϪ/Ϫ cells. In addition, increased levels of 46-kDa proLOX was clearly observed in the FNϪ/Ϫ medium as compared with the FNϩ/Ϫ medium (Fig. 7A). By analyzing three replicates and quantifying the relative amounts of LOX in the medium, the amount of 30-kDa LOX was decreased by ϳ90% in the FNϪ/Ϫ medium as compared with the FNϩ/Ϫ medium. However, the total sum of detected LOX protein, both processed and unprocessed, was approximately equal in both FNϪ/Ϫ and FNϩ/Ϫ lanes. No difference was detected in the cell lysate protein fraction, which contained only low levels of proLOX in both cell lines (data not shown). Western blot analysis of the same medium samples also demonstrated that there was no decrease in BMP-1 protease levels in the FNϪ/Ϫ medium (Fig. 7A).
To determine whether the changes in LOX proteolytic processing observed by immunoblotting corresponded to changes in enzyme activity, the LOX activity was measured from equal microgram amounts of CCM proteins collected from FNϩ/Ϫ and FNϪ/Ϫ cells. Activity was calculated as the increase in fluorescent units over time (120 min) minus the increase of the BAPN controls. The results showed that LOX amine oxidase activity was decreased by 89.2% in the FNϪ/Ϫ cell medium as compared with the FNϩ/Ϫ cell medium, with p Ͻ 0.0001 (Fig.  7B). The dramatic decrease in 30-kDa LOX protein and corresponding decrease in LOX activity in FN-null conditions were verified in three independent experiments. DISCUSSION The results of this study demonstrate that FN binds the LOX protein and provide evidence suggesting this as a novel mechanism by which FN may regulate the proteolytic activation of LOX and contribute to the assembly of the collagen matrix. Co-distribution of FN and collagen matrices have long been noted (17,18), and recent studies have shown that the assembly of types I and III collagen into fibers is dependent on the preceding polymerization of FN (26,27). FN has been also shown to directly bind to several collagen molecules (42)(43)(44), but this binding does not fully explain the role of FN in the regulation of the collagen matrix assembly. LOX initiates the covalent cross-linking of fibrillar collagen molecules, which is the final enzymatic step in the formation of mature collagen fibers. Therefore, the discovery that FN can directly bind LOX and regulates the activation of LOX provides an attractive explanation about how FN can guide the formation of mature collagen fibrils.
From the solid phase binding assays, we demonstrated that the binding affinity of LOX for cFN was much greater than its affinity for pFN. Although the categorization of plasma versus cellular FN may have been originally based on solubility, another discovered difference is that pFN circulates in a compact, inactive conformation with concealed binding sites. In contrast, cFN is usually present in an activated, open conformation (14,16). The higher affinity of LOX for cFN may be due to this open conformation through a revealed LOX-binding site. The other main difference between pFN and cFN, which arise from the same gene, is alternative splicing. It is also possible that the alternative spliced domain IIICS, which is found more prevalently in cFN molecules (14,19), contains the LOX-binding site. Each FN cDNA fragment isolated in the yeast two-hybrid screen contained the middle 192-bp IIICS segment, and two also contained the 75-bp IIICS segment.
The FN-binding domain within the LOX protein appears to be located between residues 169 and 348, but the C-terminal CRL domain may play a role in stabilizing the interaction. In the solid phase binding assays, we detected binding of cFN to GST⅐LOX 169 -348 but at a significantly lower affinity than GST⅐LOX, which contains the CRL domain. By sequence, the CRL domain is homologous to the N-terminal extracellular domain of the Class 1 cytokine receptor superfamily (45,46). However, no function for the CRL domain of LOX has previously been shown. Our results suggest that this C-terminal domain may add stability to the protein interactions of LOX. In a separate experiment, we tested whether our polyclonal anti-LOX antibody, which was designed against amino acid residues FIG. 6. LOX activity assays revealed that cFN is not an efficient substrate for LOX, but cFN binding does not inhibit the amine oxidase activity of LOX . A, LOX enzyme activity was measured using equal molar amounts of cFN, tropoelastin, type I collagen, or 1,5-diaminopentane as substrate. With cFN, the measured LOX activity was not significantly above zero. B, purified bLOX was incubated with soluble cFN or BSA to determine whether binding to cFN inhibited the enzyme activity of LOX. No significant decrease in activity was measured. C, to test whether cFN-bound LOX showed activity toward other substrates, 0, 10, or 100 nM of bLOX was incubated with solid phase cFN overnight, the unbound bLOX removed, and an activity assay performed with 1,5-diaminopentane substrate. For each activity assay, the measured increase in fluorescence over that of BAPN controls is shown.
176 -197, could be used to block the interaction with cFN. No significant decrease in cFN binding was measured when GST⅐LOX was preincubated with excess amounts of antibody (data not shown), which is consistent with the site of interaction being closer to the CRL domain of LOX.
Confocal microscopy of FN and LOX immunofluorescent costaining in fibroblast cultures showed a pattern of continuous FN matrix with overlapping punctated LOX. At this high resolution, the distribution of LOX seen in the fibroblast extracellular matrix appeared non-random and co-localized with the FN meshwork at many points. Co-expression of LOX and FN was also observed in a wide range of normal human tissues. Primarily, we observed high resolution co-localization around areas of heavy ECM deposition, such as in blood vessels and connective tissue, but we also noted co-localization at the basal surface of epithelial cells in the esophagus and kidney. In the placenta and adipose tissues, LOX staining was observed in a pattern suggesting localization to basement membranes. There have been no previous reports of LOX being associated with basement membranes, although FN localization to basement membranes has been well characterized (17,18,47). Since type I, II, and III collagens and elastin are not typical components of the basal lamina, this localization of LOX suggests that LOX may have yet unidentified functions or substrates.
The results of in vitro activity assays suggest that cFN is probably not a substrate for the LOX enzyme. These results do not exclude the possibility that lysyl residues within FN are oxidized by LOX under specific conditions in vivo. Covalent crosslinks in the FN matrix have not been commonly reported, except in conditions of wound healing and clotting when FN is crosslinked to fibrin by factor VIIIa transglutaminase (48 -50). The data do show that cFN-bound LOX retains its enzyme activity toward 1,5-diaminopentane, indicating that the active site of LOX is open. A detailed characterization of how these complex molecular interactions influence collagen assembly remains to be determined. However, this ability to bind an active LOX enzyme at specific domains may allow the FN matrix to provide specific microenvironments to guide the cross-linking of ECM molecules.
To analyze LOX in the absence of FN in vivo, we obtained embryonic fibroblasts generated from FN-null mice and their heterozygous siblings. By Western analysis, the quantity of 30-kDa LOX in the cell culture medium was decreased by ϳ90% in FN-null conditions. We also observed a corresponding increase of the 45-kDa unprocessed proLOX in the cell medium of the FNϪ/Ϫ cells and an increase of a 40-kDa band, which does appear in the FNϩ/Ϫ lanes in longer exposures. This 40-kDa protein remains unidentified, but it may represent an intermediate processed form of LOX or perhaps a novel embryonic-specific LOX variant. Consistent with the dramatic decrease in 30-kDa LOX seen by Western analysis, LOX amine oxidase activity assays demonstrated that LOX activity in FNϪ/Ϫ cell medium was ϳ10% of what was measured in FNϩ/Ϫ medium. The relatively equal amounts of active BMP-1 protease in FNϩ/Ϫ and FNϪ/Ϫ culture medium indicated that the decrease in LOX activation was not due to an absence of BMP-1 protease levels. These data suggest that the presence of FN is critical for LOX activation in vivo.
Results showing FN regulation of LOX activation support the hypothesis that the extracellular microenvironment is a determinant of the rate of LOX enzyme activation. This regulation may be a key factor in explaining the reported temporal differences in LOX mRNA expression and LOX activity. Many studies have described up-or down-regulation of LOX in response to biological stimuli, but in many of these studies, the level of alteration between LOX mRNA and LOX activity was not consistent (51)(52)(53)(54)(55)(56)(57). In addition, the production of LOX mRNA in response to dermal injury has been demonstrated to be rapid (peaking after 3 days) (58). However, LOX enzyme activity has been shown to peak later (8 -10 days post-wound) (59), which suggests that the LOX pro-enzyme may be held in the ECM for a significant period of time before activation.
Proteolytic activation of LOX has been shown to be primarily performed by BMP-1/procollagen C-proteinase, which is also responsible for cleaving the C-terminal end of type I, II, and III procollagens (11,12,60). No evidence for a relationship between FN and BMP-1 has previously been shown; however, our results suggest that there may be some connection. Perhaps FN facilitates LOX processing by bringing BMP-1 into close proximity or by altering the conformation of proLOX to make the cleavage site more accessible. FN may even act as a guiding scaffold to bring LOX, BMP-1, and procollagen molecules together in areas of collagen fiber assembly. Further studies may elucidate the molecular mechanism by which FN regulates LOX processing and how the FN-LOX interaction affects the assembly of the extracellular matrix in various tissues.
The FN-LOX interaction may also have implications for pathological processes that have been associated with misregulation of either of these individual proteins, such as carcinogenesis, cardiovascular disease, and fibrosis. Some of the consequences of increased or reduced FN in pathological conditions may be attributed to the resultant change in LOX activity. In pathological conditions in which alternative FN splicing may play a role, such as the oncofetal FN form present in certain cancers and fibrotic diseases, it will be important to investigate the effects on LOX activation.