Fibroblast growth factor-binding protein is a novel partner for perlecan protein core.

Perlecan, a widespread heparan sulfate proteoglycan, functions as a bioactive reservoir for growth factors by stabilizing them against misfolding or proteolysis. These factors, chiefly members of the fibroblast growth factor (FGF) gene family, are coupled to the N-terminal heparan sulfate chains, which augment high affinity binding and receptor activation. However, rather little is known about biological partners of the protein core. The major goal of this study was to identify novel proteins that interact with the protein core of perlecan. Using the yeast two-hybrid system and domain III of perlecan as bait, we screened approximately 0.5 10(6) cDNA clones from a keratinocyte library and identified a strongly interactive clone. This cDNA corresponded to FGF-binding protein (FGF-BP), a secreted protein previously shown to bind acidic and basic FGF and to modulate their activities. Using a panel of deletion mutants, FGF-BP binding was localized to the second EGF repeat of domain III, a region very close to the binding site for FGF7. FGF-BP could be coimmunoprecipitated with an antibody against perlecan and bound in solution to recombinant domain III-alkaline phosphatase fusion protein. Immunohistochemical analyses revealed colocalization of FGF-BP and perlecan in the pericellular stroma of various squamous cell carcinomas suggesting a potential in vivo interaction. Thus, FGF-BP should be considered a novel biological ligand for perlecan, an interaction that could influence cancer growth and tissue remodeling.

During mammalian development, its expression is detected quite early in tissues of vasculogenesis, being deposited along nearly all the endothelial-lined vascular beds (26,27). Perlecan is present not only in the basement membranes but also within extracellular matrices (13,28) and in close proximity to cell surfaces (29), where its binding is likely mediated by members of the integrin family (30 -32). Targeted disruption of the perlecan gene causes embryonic lethality at day 10.5 with widespread cephalic and skeletal abnormalities (33,34). Notably, the basement membranes are normally formed in the homozygous null animals, but vascular and cephalic abnormalities are generated in areas of increased pressure, suggesting that perlecan is required for maintaining basement membrane integrity (35). The abnormal cartilage structure and the disregulated endochondral ossification of the perlecan null animals are suggestive of a phenotype encountered with activating mutations of the fibroblast growth factor (FGF) 1 receptor 3, thereby positioning perlecan as a negative regulator of this signaling pathway (33).
Perlecan affects cell proliferation, tumor invasion, angiogenesis, and thrombosis (13, 36 -38), but the pathways by which it is able to influence such important events are not completely understood. Increased perlecan deposition is observed in breast and colon (13) carcinomas as well as in metastatic melanomas (39), and these changes correlate with enhanced metastatic potential (40). Perlecan functions as a ligand reservoir for angiogenic growth factors that become stabilized against misfolding or proteolysis (41,42). For example, perlecan binds FGF2 (43,44) and promotes receptor activation and mitogenesis (45). In a rabbit ear model of angiogenesis, perlecan-FGF2 complexes induce blood vessel formation at levels higher than those induced by heparin-FGF2 complexes (45). FGF2 binds to the heparan sulfate chains of perlecan, and its displacement by various proteolytic enzymes offers a plausible physiological mechanism whereby a powerful angiogenic stimulus becomes operational at the site of active tumor invasion (42). Thus, we hypothesized that perlecan deposition in the newly formed tumor stroma may act as a scaffold on which capillaries proliferate to generate new vascular anastomoses (14). Suppression of perlecan expression blocks autocrine and paracrine activities of FGF2 in human melanoma cells (46) and halts melanoma cell proliferation and invasion (47). In fibrosarcoma cells, however, perlecan appears to act as a negative regulator of growth and invasion (48), suggesting that the specific cellular context may play a cardinal role in perlecan's biological function. We have recently discovered (49) that antisense targeting of the perlecan gene correlates with a reduced colon carcinoma cell growth and a markedly attenuated responsiveness to mitogenic FGF7. FGF7 binds specifically to domains III and V of perlecan (50), and exogenous perlecan efficiently reconstitutes FGF7 mitogenic activity in perlecan-deficient cells (49).
The major goal of this study was to identify novel proteins that interact with the protein core of perlecan. Using the yeast two-hybrid system and domain III of perlecan as bait, we identified a strongly interactive clone that corresponded to HBp17 (51), also known as FGF-BP (52), a protein previously shown to modulate the activity of FGF1 and FGF2 to enhance the tumorigenicity of A431 squamous carcinoma cells (51) and to act as a potent angiogenic stimulus (52). FGF-BP bound specifically to domain III within the second EGF repeat, a region very close to the major binding site for FGF7 (50). FGF-BP could be coimmunoprecipitated with an antibody against perlecan and bound in solution to recombinant domain III-alkaline phosphatase (AP) fusion protein. Immunohistochemical studies revealed a significant up-regulation of FGF-BP in various squamous cell carcinomas with a distribution similar to that of interacting with perlecan domain III by the yeast two-hybrid system. Lanes 1 and 2 correspond to the M r markers and a negative control, respectively. B, alignment of human (NP005121), bovine (AAF75792), rat (AAF23079), and mouse (NP032035) sequences encoding FGF-BP (the respective GenBank TM accession numbers are given in parentheses). Fully conserved residues are shown in red, partially conserved residues are in green, and weakly conserved residues are in blue. The 10 conserved Cys are represented by an asterisk, while the 2 heparinbinding partial consensus sequences are represented by a horizontal line. C, in vitro transcription/translation of empty plasmid (lane 1), FGF-BP-containing plasmid (lane 2), and luciferase-containing plasmid (lane 3). Samples were incubated with [ 35 S]methionine in a rabbit reticulocyte lysate, separated in a 6 -12% SDS-PAGE, and subjected to autoradiography.
FIG. 2. Specific interaction between FGF-BP and perlecan protein core. A, growth in triple minus media of FGF-BP cDNA cloned into plasmids carrying both the binding (pGB) and activating (pGAD) domains, used as either prey or bait with perlecan domain III and employing the yeast host strain SFY526. As positive and negative controls, pTD1/pVA3 and pTD1/pLAM5Ј plasmids were used, respectively. B, ␤-galactosidase assays of the cotransfectant clones. C, ␤-galactosidase assays of the positive and negative controls.
perlecan, suggesting a potential in vivo interaction. Therefore, FGF-BP should be considered a novel biological ligand for perlecan, an interaction that could influence cancer growth and tissue remodeling.

EXPERIMENTAL PROCEDURES
Materials and Cell Cultures-Media and fetal bovine serum were obtained from Hyclone Laboratories (Logan, UT). 125 I and Hybond ECL membranes were purchased from Amersham. Monoclonal antibodies 7B5 against perlecan domain III (15) or monoclonal C9 against human FGF-BP (51) have been previously described.
Yeast Two-hybrid Library Screening-The Matchmaker two-hybrid system (CLONTECH Laboratories, Inc., Palo Alto, CA) was used to screen a human keratinocyte cDNA library constructed in pGAD10 (complexity ϳ5 ϫ 10 6 recombinants) with perlecan domain III. The full-length domain III and the deletion constructs were cloned into pGBT9 and pGAD424 (50), and further deletion fragments were generated by endonuclease digestion of the constructs. Competent HF7c cells were prepared and combined with 50 g each of the library DNA and the domain III construct. The mixture was incubated at 30°C for 30 min with shaking (200 -250 rpm) and then heat-shocked for 15 min at 42°C and incubated at 30°C for 30 min in the presence of 1 ml of Trp Ϫ /Leu Ϫ medium with constant shaking. The transfected cells were plated in medium lacking Trp and Leu to obtain an indication of the primary transformation efficiency and on plates lacking Trp, Leu, and His to select for colonies expressing interacting hybrid proteins. The plates were incubated at 30°C for 8 days. The His ϩ colonies were isolated and screened by PCR for the inserts in the activation domain vector using primers specific for the GAL4 activation domain plasmid.
Yeast Two-hybrid One-on-one Interactions-FGF-BP was cloned in the yeast two-hybrid plasmids by PCR amplification, digestion of the PCR product with EcoRI/SalI, and ligation into EcoRI/SalI-digested pGBT9 and pGAD424 plasmids. Both constructs were analyzed by DNA sequencing. Transformation of the yeast reporter strain SFY526 with all the combinations of hybrid constructs, using FGF-BP and perlecan domain III as both prey and bait and utilizing the various deletion constructs of domain III to specifically map the site of interaction, was performed as described previously (50). The transfected cells were plated in Trp Ϫ /Leu Ϫ and Trp Ϫ /Leu Ϫ /His Ϫ agar plates to check for in-teractions, analyzed initially by growth in the triple minus media compared visually with that of the positive control. For qualitative ␤-galactosidase assays, cells grown in Trp Ϫ /Leu Ϫ plates were transferred onto Whatman No. 3MM paper filters and soaked in Z buffer/Xgal solution (0.1 M Na 2 HPO 4 , 45 mM NaH 2 PO 4 , 10 mM KCl, 2 mM MgSO 4 , 0.3% ␤-mercaptoethanol, and 3.3 mg/ml 5-bromo-4-chloro-3indolyl-␤-D-galactopyranoside). Yeast cells were frozen in liquid nitrogen for 10 s, thawed, transferred onto a filter, and incubated at 30°C for ϳ8 h to visualize the appearance of blue colonies. In addition, the twohybrid interactions were verified and quantified by liquid assays of ␤-galactosidase activity using the Luminescent ␤-galactosidase detection kit II (CLONTECH). This system uses a chemiluminescent substrate (Galacton-Star TM ), and the light emission can be used as a quantitative measure of ␤-galactosidase activity. Five-ml overnight cultures in Trp Ϫ /Leu Ϫ medium were transferred to 8 ml of YPD medium and incubated at 30°C for 3 h with shaking (250 rpm). The cells were harvested by centrifugation at 10,000 ϫ g for 30 s, and the pellets were resuspended in 300 l of buffer and subjected to two consecutive freeze/ thaw cycles. 25-l aliquots of each cell lysate were incubated with 200 l of Galacton-Star TM reaction mixture at room temperature for 60 min and centrifuged at 10,000 ϫ g for 1 min, and the supernatants were transferred to luminometer tubes. The light emission (measured in relative light units) was recorded at 5-s intervals, and the ␤-galactosidase activity was normalized on the relative OD, and expressed as relative light units /A 600 unit of cell culture.
In Vitro Transcription/Translation-The TNT T7 Quick-coupled Transcription/Translation System (Promega) was used for in vitro transcription and translation of FGF-BP and the positive luciferase control. FGF-BP was subcloned from pGAD424 into pcDNA3.1 downstream from the T7 promoter by EcoRI/SalI digestion followed by ligation into an EcoRI/XhoI-digested pCDNA3.1. One g of the plasmid DNA, 40 l of the TNT Quick Master Mix, and 20 Ci of [ 35 S]methionine were combined and incubated at 30°C for 90 min. A 5-l aliquot was resuspended in SDS-sample buffer and loaded on a 6 -12% SDS-PAGE. The gel was incubated with fixing solution (40% methanol, 10% glacial acetic acid) for 1 h, dried, and exposed on Kodak X-Omat AR film for 6 -15 h at room temperature.
Generation of A431 Cells Expressing Domain III-AP Fusion Protein and Co-immunoprecipitation Studies-Perlecan domain III was sub-

FIG. 3. FGF-BP interacts specifically with the second EGF-like repeat of domain III-1 of perlecan.
A, schematic representation of domain III and various deletion mutants. Domain III has been subdivided as indicated at the top according to a recently proposed nomenclature (67). Green ovals indicate globular domains (L4 modules), and red rectangles indicate the laminin-type EGFlike repeats (LE modules). The numbers within parentheses designate the amino acid position based on the mature protein core. Growth is indicated by semiquantitative assessment, with maximal growth at ϩϩϩ. B, representative ␤-galactosidase assays of all the clones carrying domain III or its deletion mutants. C, liquid assays quantization of ␤-galactosidase activity using a sensitive luminescent detection system (CLONTECH). The values are the mean Ϯ S.D. (n ϭ 5) and are expressed as relative light units (RLU) normalized on cell content (OD 600 ).
cloned by PCR (Expand TM Template PCR kit, Roche Molecular Biochemicals) using the oligonucleotides 5Ј-CGCAATTGCCCTGCCCT-GACGGCC-3Ј and 5Ј-CGGGATCCAATTGTGGGGCTTGGTTTGTCTC-3Ј, which introduced an MfeI site. The purified digested fragment was ligated into an engineered pcDNA3.1 vector linearized with EcoRI containing the sequence of the human placental AP preceded by the BM-40 signal peptide. The construct was sequenced, and 10 g of the purified plasmid were used to transfect A431 cells by electroporation (375 V, 490 microfarads, 1 ms). Cells were cultured in G418 (400 g/ml), and clones were isolated by ring cloning and expanded. Conditioned media from confluent cells were collected after 2 days and assayed for AP enzymatic activity using a SEAP (secreted alkaline phospatase) Chemiluminescence Detection Kit (CLONTECH). The expression of the recombinant RNA was analyzed by Northern blotting and the chimeric protein detected in Western immunoblotting employing a specific antibody directed against domain III. For immunoprecipitation, aliquots of the media containing domain III-AP fusion proteins were brought to 50 mM HEPES, pH 7.4, 150 mM NaCl, 5 mM EGTA, 10% glycerol, 1% Triton X-100, 100 mM NaF, 200 M Na 3 VO 4 , 10 mM sodium pyrophosphate, and a mixture of protease inhibitors (Complete TM , Roche Molecular Biochemicals). Following a 4-h incubation at 4°C, 1 g of monoclonal antibodies against domain III or FGF-BP was added and incubated overnight at 4°C. Sepharose-A/G beads (1:1 v/v) were added and incubated for 1 h with shaking, centrifuged at 10,000 ϫ g for 10 min, and washed three times with 20 mM HEPES, pH 7.4, 150 mM NaCl, 0.1% Triton X-100, 10% glycerol, 200 M Na 3 VO 4 , 10 mM NaF, and protease inhibitors. The final pellet was boiled in Laemli buffer and analyzed by SDS-PAGE. Additional details are provided in the text and and the figure legends.
In Vivo Interaction of Human Perlecan Domain III and FGF-BP in Transiently Transfected COS-1 Cells-Subconfluent cultures of COS-1 cells were transfected using LipofectAMINE Plus reagent (Life Technologies, Inc.) with 4 g of perlecan domain III/AP and FGF-BP, both cloned into pcDNA3.1 (Invitrogen) containing the sequence of the BM-40 signal peptide. After 72 h of incubation, the serum-free conditioned medium was collected and filtered, and various protease inhibitors were added (see above). Two g of either anti-domain III or anti-FGF-BP antibody were added to 1 ml of medium of transfected and untransfected COS-1 cells, and the proteins were immunoprecipitated at 4°C for 18 h with continuous rocking. The immunocomplexes were captured with protein A/G-agarose beads (Pierce). Samples were analyzed in a 3-15% SDS-PAGE gradient gel under nonreducing conditions and transferred into nitrocellulose. The membranes were incubated in 5% nonfat dry milk at room temperature for 1 h. The lower portion of the membrane (proteins Ͻ40 kDa) was then incubated with the anti-FGF-BP antibody and the upper portion with the anti-domain III antibody for 1 h. Membranes were washed three times in Tris-buffered saline containing 0.1% Tween 20 and incubated with horseradish peroxidase-conjugated secondary antibody for 1 h. After three washes, the immunocomplexes were visualized by enhanced chemiluminescence (Pierce).
Immunohistochemical Studies-A survey of normal and neoplastic frozen human tissues, obtained from the Tumor Bank of Thomas Jef-  5 and 6), in contrast to clone 4 or the wild type cells (lanes 3 and 4 and 7 and 8, respectively). The bottom panel, reacted with ␣FGF-BP, shows a significant amount of FGF-BP migrating with an average mass of ϳ23 kDa. B, alkaline phosphatase assays of various clones as indicated. C, immunoprecipitation (ip) followed by Western immunoblotting (wb) as indicated. Notice the presence of domain III-AP fusion protein and FGF-BP in both immunoprecipitates. The reactive ϳ85 kDa band (lanes 1 and 2, top panel) corresponds to IgG homo-or heterodimers since the gel was run under nonreducing conditions because the C9 antibody does not work well following reduction. Lane 3 shows immunoblotting of the total medium alone without any immunoprecipitation. Notice the presence of the predicted ϳ190and ϳ23-kDa proteins corresponding to perlecan's domain III-AP and -FGF-BP, respectively.
ferson University, was investigated using monoclonal antibodies C9 against FGF-BP (51) or 7B5 against perlecan domain III (15). A total of 38 samples was investigated including normal samples of skin, colon, kidneys, brain, tongue, pharynx, and lung. In addition we stained several samples of squamous cell carcinomas from skin, esophagus, tongue, lung and cervix, as well adenocarcinomas of the urinary bladder, lung, and kidney. Immunohistochemistry was done essentially as previously described (15) with minor variations. Routinely, we used primary antibodies at 1:1000 dilution and secondary antibodies conjugated with either AP or horseradish peroxidase at 1:2000 dilution. In all cases, the frozen sections were blocked with either lovamisole or H 2 O 2 before the immunohistochemical reaction. Images were captured with a Pixera digital camera and assembled using Adobe Photoshop version 5.0.

RESULTS AND DISCUSSION
Discovery of FGF-BP as a Binding Partner for Perlecan Protein Core-The yeast two-hybrid system (53) was used to identify candidate proteins that interact with perlecan domain III in vivo. Before proceeding with the screening, the bait plasmid harboring domain III was tested for its inability to activate the prototrophic reporter gene HIS3. As a control for possible interactions, all constructs were also tested as either bait or prey. In addition, all of the constructs were assayed for growth in double minus (Trp Ϫ /Leu Ϫ ) media as a control for transfection efficiency. The screening of ϳ0.5 ϫ 10 6 cDNAs from a human keratinocyte library resulted in the isolation of 50 independent clones identified as large colonies growing in triple minus (Trp Ϫ /Leu Ϫ /His Ϫ ) plates. Of these, 40 clones grew back and 38 clones showed inserts ranging between 0.4 and 1.6 kilobase pair. To minimize the number of false positives and to increase the likelihood of obtaining pure clones, the initial isolates were regrown in triple minus media and subjected to direct PCR screening using flanking primers specific for the pGAD plasmid. All of the bands were isolated and sequenced, and DNA homology searches using the NCBI BLAST program of the Genetics Computing Group package led to the elimination of many false positives, the majority of which were nuclear proteins that could themselves activate the transcription of the HIS3 reporter gene. A highly interacting clone contained a 1.2-kilobase pair insert (Fig. 1A) that encoded a small protein, named HBp17 (Fig. 1B) because it binds heparin-binding growth factors 1 and 2 (51) (also called FGF-BP, for FGFbinding protein (52)). The cDNA was in frame with the activating domain of the plasmid and coded for the full-length FGF-BP, further supporting the concept of a real protein-protein interaction.
The full-length FGF-BP cDNA is 1163 base pairs, and the primary structure of the encoded protein consists of 234 amino acids, including a 21-residue signal peptide sequence and a single putative N-linked glycosylation site. Computer searches revealed that homologues of human FGF-BP have been cloned previously from bovine, rat, and mouse tissues (54 -56) and that FGF-BP is well conserved among species (Fig. 1B). Human FGF-BP showed 84, 57, and 49% identical amino acid sequence to the bovine, rat, and mouse species, respectively. Ten Cys residues are fully conserved among species, and there are two partial heparin-binding consensus sequences (57) (Fig. 1B) in addition to a highly basic region (residues 110 -143) proposed to be the principal heparin binding domain in FGF-BP (58).
To establish the nature of FGF-BP cDNA, we cloned FGF-BP into the expression vector pcDNA3.1 and used in vitro transcription/translation to determine its molecular mass. The results showed a single band of the predicted size of ϳ23 kDa (Fig. 1C, lane 2) in contrast to the empty vector (Fig. 1C, lane  1), which showed no detectable bands. The positive control was provided by the luciferase cDNA, which produced the expected ϳ66 kDa protein (Fig. 1C, lane 3).
Nature of FGF-BP-FGF-BP is a secreted, heparin-binding protein originally purified from media conditioned by A431 human squamous carcinoma cells (51). Notably, FGF-BP is localized to squamous epithelia (51) and to squamous cell carcinomas (59). FGF-BP binds FGF1 and FGF2 in a noncovalent reversible manner (51) and potentiates the activity of FGF7 (59), similar to the action of perlecan (49). In addition, A431 squamous carcinoma cells transfected with FGF-BP become more tumorigenic than wild type cells, and nontumorigenic A431-4 subclones, which do not express FGF-BP, become tumorigenic upon de novo expression of FGF-BP (51). Ectopic expression of FGF-BP in adrenal adenocarcinoma cells causes a release of FGF2 and formation of highly vascularized tumors in nude mice (52). FGF-BP expression is down-regulated by retinoic acid (56,60) and induced by EGF (61), and it is often enhanced in squamous cell carcinomas and in a few colon carcinoma cell lines (54). Notably, FGF-BP can serve as an angiogenic switch in vivo because reduction of FGF-BP expression by targeting FGF-BP with specific ribozymes causes a reduction in the growth and angiogenesis of tumor xenografts (62).
FGF-BP Binds Specifically to Perlecan's Domain III-To further verify that FGF-BP was a true interactive protein, we cloned the full-length FGF-BP cDNA into plasmids carrying both the binding (pGB) and activating (pGAD) domains and adopted them as either prey or bait with perlecan domain III using a different yeast host strain (SFY526). As positive and negative controls, we utilized the pTD1/pVA3 and pTD1/ pLAM5Ј plasmids, respectively (CLONTECH). The results showed a robust growth in triple minus media of clones coexpressing FGF-BP/perlecan domain III proteins, expressed as either bait or prey, with growth rates comparable with the positive control ( Fig. 2A). To further prove this interaction, we performed ␤-galactosidase assays on the cotransfectants. In addition to growth in triple minus media, transcription of lacZ containing the upstream binding sites of GAL4 and the subsequent ability of cotransformant yeast strains to express functional ␤-galactosidase are additional strong proofs for a true protein-protein interaction (63,64). Domain III, present as either bait or prey, showed a robust production of blue colonies (Fig. 2B), comparable in intensity with the positive control pTD1/pVA3 harboring the p53 and the SV40 T antigen (Fig.   2C). This suggests that the affinity between domain III and FGF-BP is relatively high. Preliminary solid phase binding experiments using purified FGF-BP and soluble domain III-AP demonstrated a saturable and high affinity (K d ϳ 18 nM) binding (not shown).
FGF-BP Binds to the Second EGF Motif of Domain III-Next, we sought to establish the precise location of the FGF-BP-binding site within domain III of perlecan, a1172-residue domain that is encoded by 27 exons (65) and shares homology with the short arm of laminin ␣1-chain (8 -10). Domain III consists of alternating globular domains (L4 modules, characteristically devoid of cysteine residues) and short connecting rod-like segments of laminin-type EGF-like repeats (LE modules). We generated five deletion fragments of domain III, ⌬1-⌬-5 (Fig. 3A). Growth was observed in cells cotransformed with full-length domain III and the first two deletion constructs, but it was markedly attenuated using ⌬3-5 constructs (Fig. 3A). These results were corroborated by qualitative (Fig.  3B) and quantitative (Fig. 3C) ␤-galactosidase assays. In the latter, the two-hybrid interactions were quantified by liquid assays of ␤-galactosidase activity using a very sensitive luminescent detection system. For testing each interaction, five separate transformant colonies were assayed and each assay was performed in triplicate. Notably, ␤-galactosidase activity of domain III-FGF-BP cotransformants was about 50% of that observed by the positive control, further suggesting that the affinity of FGF-BP for perlecan protein core is relatively strong.
The results indicate that the second EGF-like repeat (LE2) of domain III-1 interacts specifically with FGF-BP. Coimmunoprecipitation of FGF-BP and Domain III-To verify that the interaction detected using the yeast two-hybrid system could occur outside the yeast, and to show its relevance within the context of secreted proteins from mammalian cells, we studied the interaction of secreted FGF-BP and recombinant domain III. As mentioned above, FGF-BP was originally isolated from media conditioned by A431 squamous carcinoma cells (51). To investigate the potential interaction in solution between perlecan and FGF-BP, we generated stable transfectant clones of A431 secreting the full-length domain III fused to the human placental AP to serve as a marker. The generation of domain III-AP fusion protein would avoid the potential interference of the heparan sulfate chains, because this module has been shown to be synthesized without glycosaminoglycan substitution (66), and would facilitate a direct protein-protein interaction. Several clones were isolated, and two (clones 1 and 11) were synthesized and released into the medium of the A431 cells, the fusion protein of the correctly predicted mass of ϳ190 kDa. This protein was recognized by anti-domain III monoclonal antibodies (Fig. 4A, top panel) and by an antibody against alkaline phosphatase (not shown). Interestingly, FGF-BP was detected at high levels in these cells (Fig. 4A, bottom panel) and migrated as a ϳ23 kDa protein under reducing conditions, to a position identical to that obtained with in vitro transcription/translation (cf. Fig. 1C). Proper folding of domain III-AP fusion protein was verified by detection of strong AP activity in clones 1 and 11 (Fig. 4B). This assumption is based on the fact that the placental AP is positioned C-terminally, and thus we presume that domain III is also properly folded because the fusion protein expressed high enzymatic activity. In addition, domain III folds into an individual entity, as determined by rotary shadowed electron microscopy and biophysical studies (66,67). Immunoprecipitation studies using anti-FGF-BP or anti-domain III monoclonal antibodies followed by Western immunoblotting showed the coimmunoprecipitation of domain III and FGF-BP (Fig. 4C).
We further confirmed the interaction identified in the twohybrid screen system by coimmunoprecipitation of domain III and FGF-BP in transiently transfected kidney COS-1 (African green monkey) cells, which do not express FGF-BP. The media conditioned by COS-1 cells cotransfected with domain III and FGF-BP cDNAs showed that both proteins interacted in solution and could be identified by their respective antibodies (Fig.  5). Collectively, these data indicate that FGF-BP is a binding partner for perlecan domain III and that FGF-BP can bind perlecan protein core in solution.
FGF-BP Is Increased in Squamous Cell Carcinomas and Codistributes with Perlecan-To elucidate FGF-BP distribution in human tissues, we tested a number of normal and neoplastic frozen human tissues with the monoclonal antibody directed toward FGF-BP. In adult normal skin, FGF-BP was expressed at relatively low levels in the suprabasal region of the epidermis and focally in hair follicles (Fig. 6B). Interestingly, in a few cases of squamous epithelia, FGF-BP epitopes were clearly present along the basement membrane at the dermo-epidermal junction (Fig. 6C), occasionally extending into the basement membrane of dermal blood vessels (Fig. 6D). This is similar to the distribution of human perlecan using anti-domain III antibodies (15). All of the other normal tissues, including lung, brain cortex, kidney, uterus, breast, colon, and various fibro-adipose tissues, were essentially negative. In contrast, invasive squamous cell carcinoma showed a marked induction of FGF-BP, especially around the most undifferentiated cells (Fig. 6E). Notably, even at relatively high concentrations of the primary antibody (1:200) there was no detectable staining in the normal tissues and the nonsquamous cell carcinomas tested (see below), in contrast to skin and squamous cell carcinomas where the antibody had to be diluted significantly (1:1000) to achieve optimal staining. Overall, these data are in good agreement with in situ hybridization studies in the mouse, which have shown that FGF-BP expression starts at embryonic day 9 reaches its peak perinatally and is subsequently down-regulated during adult life (54). In concert with our findings, FGF-BP mRNA expression is dramatically increased upon induction of mouse skin papillomas and carcinomas (54).
We then investigated the expression of FGF-BP vis á vis that of perlecan in dysplastic skin and various tumors, including carcinomas of the lung, uterus, bladder, kidney, and squamous cell carcinomas of the esophagus, skin, tongue, pharynx, lung, and penis. A total of 38 samples were investigated. In dysplastic skin overlaying an infiltrating squamous cell carcinoma of the tongue, FGF-BP staining was markedly induced (Fig. 7B). Perlecan distribution was primarily along the basement membrane and blood vessels of the upper dermis (Fig. 7, C and F) in agreement with previous studies (15,68). Again, there was a marked expression of FGF-BP in the invasive squamous cell carcinoma cells of the esophagus (Fig. 7E), and consecutive sections showed a significant codistribution of FGF-BP (Fig.  7G) and perlecan (Fig. 7, H and I). Keratin pearls, considered to represent a sign of cellular differentiation, did not contain either FGF-BP or perlecan epitopes (Figs. 6E and 7I). No significant FGF-BP immunoreactivity was observed in all the other nonsquamous carcinoma tumors (not shown) with the exception of focal positivity in colon and lung carcinomas. The latter finding is in agreement with the reports that FGF-BP is expressed by some colon carcinoma cell lines (52) and that FGF-BP transcript can be detected in the developing mouse intestine and lung (54).
Overall, our data indicate that under normal conditions, the distributions of FGF-BP and perlecan overlap only focally. The former is located in the suprabasal layer (stratum spinosus) of the epidermis, whereas the latter is located primarily in the dermis, along the basement membrane zone at the dermalepidermal junction and along the vasculature. When the squamous epithelium becomes transformed, regardless of its site of origin, FGF-BP is deposited in the pericellular space in close proximity to the cell surface of the most aggressive squamous carcinoma cells. This distribution clearly overlaps with that of perlecan, suggesting a potential in vivo interaction between these two proteins.
Conclusions-Because of its eukaryotic nature, the yeast two-hybrid system has been widely used to study protein-protein interactions, primarily those occurring among intracellular proteins (69). Only recently, however, has it been successfully used to investigate interactions among extracellular proteins such as those involving collagen types VI and IV (70), thrombospondin (71), EMILIN (72), or matrix metalloproteinase 2 (73). Binding sites for FGF7 and platelet-derived growth factor-AA and -BB have been identified in subdomains III-1 (50) and III-2 (74), indicating that there are unique binding specificities for perlecan modules containing highly repetitive sequences. The results of this study indicate that the second EGF-like repeat (LE2) of domain III-1 interacts specifically with FGF-BP and that this interaction could modulate the ability of squamous epithelia to respond to FGF-mediated signaling. Binding of FGF-BP to subdomain III-1 is the second extracellular ligand identified so far for this particular perlecan region. Coimmunoprecipitation and codistribution in cancer tissues suggest that these interactions may take place in vivo. Perlecan may function as an extracellular sink for FGF7 and FGF-BP, acting as a reservoir for these growth factors.
FGF-BP appears to be located within the keratinocyte layer of squamous epithelia, primarily in the stratum spinosus. During cancer growth and invasion, FGF-BP could be released from the transformed epithelial cells and interact with perlecan, which is highly expressed in the stroma and perivascular microenvironment. It is plausible that FGF-BP, FGF7, and perlecan form a trimolecular complex that potentiates the activity of FGF7. It is seemingly possible that perlecan may modulate the targeting of FGF7 and FGF-BP to the epithelium and may function as an important molecular entity in the signaling of these proteins. Upon displacement by partial proteolysis of the protein core, FGF7 and FGF-BP would become available to the surrounding cellular environment and could behave as a promoter of growth and differentiation. An alternate pathway would involve an FGF-BP-mediated displacement of FGF2 and FGF7 stored in the cell surface heparan sulfate proteoglycans. This pathway does not require the need of active proteolysis or glycolytic enzymes and could be operational at the site of tumor growth. According to this working model (Fig. 8), overproduction of FGF-BP in malignant cells or induction by EGF would liberate FGF-BP in the microenvironment where it would displace FGF2 or FGF7 bound to heparan sulfate chains of syndecan, glypican, or perlecan proteoglycans. In addition, FGF-BP could interact and displace FGF7 bound to perlecan's domain III, without the need of proteolytic activity. Various FGFs could then be released into the microenvironment where they would stimulate angiogenesis and tumorigenesis. Hence, perlecan would play a central role not only as molecular storage of growth factors but also as repository for FGF-BP, and possibly related proteins, that would act as angiogenic modulators.