The High Affinity Heparin-binding Domain and the V Region of Fibronectin Mediate Invasion of Human Oral Squamous Cell Carcinoma Cells in Vitro *

Fibronectin is an extracellular matrix molecule composed of repeating subunits that create functional domains. These domains contain multiple binding sites for heparin and for various cell-surface receptors that modulate cell function. To examine the role that the high affinity heparin-binding region and the alternatively spliced V region of fibronectin play in tumor invasion, we expressed and purified four complementary recombinant fibronectin proteins. These proteins either included or excluded the alternatively spliced V region and contained either a mutated, non-functional high affinity heparin-binding domain (Hep−) or an unmutated heparin-binding domain (Hep+). Cultured oral squamous cell carcinoma cells were assayed for invasion into a Matrigel/collagen matrix supplemented with these four purified recombinant proteins, and for spreading and motility on plastic. Increased invasion was observed in gels supplemented with the V−Hep+ protein when compared with the V−Hep− protein. Inclusion of the V region in the proteins enhanced the invasion and migration associated with both Hep+ and Hep−proteins, whereas cell spreading was enhanced with the Hep+recombinant proteins. These data demonstrate that both the high affinity heparin-binding domain and the V region of fibronectin play important roles in invasion, motility, and spreading of oral squamous cell carcinoma cells.

Squamous cell carcinomas are the most common type of malignant oral neoplasm and account for a major portion of deaths related to oral cancer. These tumors represent 4% of all cancers of males in the United States, but in some Asian countries they are the most common malignant tumor. At present the survival rate approximates 50% and has not improved significantly in patients treated over the past several decades (1). Much of the morbidity and mortality associated with these tumors is related to their invasive characteristics. Studies relating to mechanisms used by squamous cell carcinomas for adhesion, motility, and invasion are, therefore, an important aspect of cancer therapy.
Previous studies have implicated cell-surface receptors belonging to the integrin and the proteoglycan families in the invasion and metastasis of tumors (2)(3)(4). Furthermore, coordinate interactions among these receptors have been implicated in tumor cell adhesion to extracellular matrix (ECM) 1 components (5,6). Since some of these adhesive interactions can be blocked by small synthetic peptides derived from ECM molecules, peptide blocking experiments have also identified specific molecules that are potentially useful for therapy (7).
One such molecule which has been studied for its potential therapeutic use is fibronectin (FN), an ECM adhesion molecule composed of multiple functional domains that interact with multiple cell-surface receptors. The domains are composed of repeating structural units, as evidenced by sequence homology, and are designated as type I, type II, and type III repeats. Repeats are numbered from the amino terminus of the molecule. The functional regions include a cell-binding domain, which contains an arginine-glycine-aspartic acid (RGD) adhesive sequence, and a carboxyl-terminal heparin-binding domain (8,9). The translated protein also includes additional regions that arise through alternative splicing. In rat FN these domains are designated EIIIB, EIIIA, and the V region (see Fig. 1).
Many cell-surface receptors interact with specific portions of the FN molecule. The RGD site in the 10th type III repeat (III-10) has been shown to interact with both the ␣ 5 ␤ 1 (10) and ␣ v ␤ 1 integrins (11). The ␣ 4 ␤ 1 integrin binds to the V region as well as to sequences in the adjacent 14th type III repeat (III-14) (12). FN also contains numerous heparin-binding sites within the carboxyl-terminal portion of the molecule, including a high affinity binding site in the 13th type III (III-13) repeat and multiple low affinity binding sites within III-14 (5,6). Many of these FN domains retain biological activity when isolated as purified proteolytic or recombinant protein fragments (13). Fragments from the RGD cell-binding, heparin-binding, and alternatively spliced V regions of FN have been used extensively to examine interactions between cells and FN in general (14) and to better understand tumor cell adhesion, motility, and invasion (5,6,(15)(16)(17)(18)(19)(20).
Although many domains of FN have been implicated in mediating tumor cell functions, the contributions of each domain and the relative importance of each family of receptors to these processes are difficult to assess. However, these evaluations are important since they may suggest potential therapeutic interventions by targeting the portions of the FN molecule that play the greatest role in the invasion process. These inquiries are often best made using recombinant proteins that exhibit altered function because of specific point mutations rather than deletions of large protein segments, which may alter protein function nonspecifically. In the present study, we used four different purified recombinant FN proteins, with or without function-perturbing point mutations in the high affinity heparin-binding region and with or without the alternatively spliced V region, to evaluate the relative contributions of these domains to the process of invasion by oral squamous cell carcinoma cells. Our results showed that the high affinity heparinbinding domain and the alternatively spliced V region of FN both contribute to the invasive behavior of these cells.

EXPERIMENTAL PROCEDURES
Materials-Growth factor reduced Matrigel basement membrane matrix was purchased from Collaborative Biochemical Products, Bedford, MA. Vitrogen 100-purified collagen was purchased from the Collagen Corp., Palo Alto, CA. Nutridoma-HU, a serum-free medium supplement, was purchased from Boehringer Mannheim. Transwell porous cell culture inserts, polycarbonate membranes in 24-well plates (6.5 mm diameter, 8.0 mm pore size), were purchased from Costar, Cambridge, MA.
Cells and Cell Culture-Human oral squamous cell carcinoma cells (HSC-3 and HOC313) were a gift from Dr. Randall Kramer (University of California, San Francisco) and have been described previously (21). These cells were maintained in ␣-minimum essential medium containing 10% fetal bovine serum, 1% penicillin, and 1% streptomycin.
Recombinant Fibronectin Proteins-Rat fibronectin cDNA was engineered between type III repeats 10 and 15 ( Fig. 1A) to form four constructs, which were expressed as recombinant FN proteins and subsequently purified. The proteins spanned repeats III-10 to III-15, and all contained the RGD cell-binding region in III-10, the alternatively spliced EIIIA type III repeat, and the low affinity heparin-binding sequences in the carboxyl-terminal heparin-binding domain. However, the proteins either included (V ϩ ) or excluded (V Ϫ ) the alternatively spliced V region and differed in the high affinity heparin-binding region of III-13 (22). In the latter region, two point mutations were introduced into the cDNA to replace adjacent arginine residues with threonines ( Fig. 1B). Since the threonine mutations have subsequently been shown to virtually abrogate heparin-binding function, these recombinant proteins have been designated as either Hep ϩ for the unmutated protein or Hep Ϫ for the mutated protein. The four proteins are therefore identified as V Ϫ Hep ϩ , V Ϫ Hep Ϫ , V ϩ Hep ϩ , and V ϩ Hep Ϫ .
The cDNAs used to generate the four recombinant proteins were engineered in two parts as follows. The cDNAs corresponding to repeats III-13, -14, and -15 between the 5Ј ApaI and 3Ј HincII restriction sites in the FN cDNA were cloned into the pECE expression plasmid (23). Two cDNAs that either included or excluded the V region were used. The 3Ј HincII portion was ligated into the blunted XbaI site of the pECE polylinker to attach an engineered three-frame stop within the vector to the 3Ј end of the FN cDNA fragment. The ApaI restriction site was used to insert synthetic oligonucleotides into the 5Ј end of III-13. These nucleotides introduced an XbaI restriction site into the cDNA corresponding to the hinge region between III-13 and III-12 in the FN protein (Fig. 1B, shown in bold type). The new restriction site was engineered by altering a single nucleotide in the FN cDNA at the third position of a codon. This mutation preserved the protein sequence of the cDNA but inserted a unique restriction site, the XbaI recognition sequence, into the gene. The FN sequence between XbaI and ApaI could then be altered by inserting synthetic oligonucleotides between these two restriction sites. As indicated in Fig. 1B, one such pair of oligonucleotides was inserted between these restriction sites to mutate adjacent arginine residues to threonines. This mutation was subsequently shown to virtually abrogate heparin-binding function in the expressed recombinant protein.
The region engineered between III-13 and III-15 in the pECE expression vector was removed from the plasmid by digesting with XbaI and EcoRI. EcoRI digests within the plasmid vector downstream from the three-frame stop and polyadenylation sequence. The cDNA fragment removed from pECE therefore contained repeats III-13 to III-15, an engineered translation termination site and a polyadenylation sequence. This cDNA fragment was bounded by restriction sites for XbaI and EcoRI. Thus, fragments containing the adjacent arginines as well as fragments with adjacent arginines mutated to threonines were isolated. Each of these fragments also contained cDNA in which the V region was either present or removed by alternative splicing.
The upstream portion of the FN cDNA between type III repeats 10 and 13 was engineered by adding restriction sites, using primers in a polymerase chain reaction. These primers introduced the unique restriction site for XbaI at the 3Ј end (in the hinge region between III-12 and III-13) and a second site for XhoI at the 5Ј end (in the cDNA corresponding to the hinge region between III-9 and III-10 in the FN protein). Sense primers (5Ј-AA TCT CGA GTT TCC GAT GTC GGC TCT-3Ј) provided the recognition sequence (CTCGAG) for XhoI. Antisense primers (5Ј-TT CTC TAG AGT CGT GAC GAC TCC CTG AGC-3Ј) provided the recognition sequence (TCTAGA) for XbaI. The product of the polymerase chain reaction was therefore bounded by restriction sites for XhoI and XbaI. Products were subcloned into TA subcloning vectors (Invitrogen, San Diego, CA), and inserts were removed by digestion with these two restriction enzymes. Two cDNA fragments (III-10 to III-13 and III-13 to III-15) were then ligated into the pTrcHis-C bacterial expression plasmid (Invitrogen) in a three-part ligation as follows. The cDNA inserts encoding repeats III-10 to III-13 were bounded by XbaI and XhoI. The cDNA inserts encoding repeats III-13 to III-15 plus terminator were bounded by XbaI and EcoRI. These two cDNA fragments were ligated into the pTrcHis-C expression plasmid vector (in the appropriate reading frame) following digestion with XhoI and EcoRI. The two resulting sets of expression plasmids contained type III repeats 10 -15, with the FN sequences stopping and starting in the hinge regions of the FN molecule adjacent to these repeats. The V region was either included or excluded in each set, as was the Hep ϩ or Hep Ϫ sequence. Initiation and termination signals were provided by the pTrcHis-C expression plasmid and the three-frame stop from the pECE plasmid.
All expressed proteins also contained the six-histidine (His 6 ) purification sequence on the amino terminus, which was used for purification with metal-binding columns (Invitrogen). Proteins were eluted from the columns under denaturing conditions (pH 5.0) and were separated from bacterial proteins, as described by the manufacturer. After purification, elution buffer was removed by passage over G25 Sepharose columns.
All of the recombinant FN proteins contained the 10th type III repeat (with the RGD cell-binding region), the alternatively spliced EIIIA domain, and the 11th to 15th type III repeats, which include the low affinity heparin-binding sequences in the carboxyl-terminal heparinbinding domain. These four proteins also either included or excluded the V region, and they differed in III-13 such that it either contained the adjacent arginines or the substituted threonines. The predicted molecular masses of the four recombinant proteins are 69.3 kDa for the V Ϫ proteins and 82.5 kDa for the V ϩ proteins.
The four recombinant proteins were analyzed by standard sodium dodecyl sulfate-polyacrylamide gel electrophoresis using 10% polyacrylamide gels stained with 0.5% Coomassie Brilliant Blue R250 and destained in a solution of 50% methanol and 5% glacial acetic acid. The proteins were further positively identified by standard Western immunoblots probed with a mouse monoclonal antibody to rat FN (24), used at a dilution of 1:200. Primary antibody was then exposed to an alkaline phosphatase-conjugated goat anti-mouse secondary antibody (Life Technologies, Inc.) used at a dilution of 1:5000. Bound antibody was detected using nitro blue tetrazolium chloride and 5-bromo-4-chloro-3indolyl phosphate, toluidine (Promega, Madison, WI). The concentration of the four recombinant proteins was determined by comparison to a standard curve of sequential dilutions of bovine serum albumin. Molecular mass standards (Life Technologies, Inc.) for electrophoresis included the following prestained proteins with apparent molecular mass as follows: myosin (214 kDa), phosphorylase b (111 kDa), bovine serum albumin (74 kDa), ovalbumin (45 kDa), carbonic anhydrase (29 kDa), ␤-lactoglobulin (18 kDa), and lysozyme (15 kDa).
Heparin-binding Function-To determine the presence or absence of heparin-binding function, the V Ϫ recombinant proteins were analyzed for binding to heparin-Sepharose. Heparin-Sepharose columns were prepared by packing Poly-Prep chromatography columns (Bio-Rad) with heparin-Sepharose CL-6B (Pharmacia Biotech Inc.) according to the manufacturer's instructions. The column was then equilibrated with 10 mM Tris, pH 8.0, and the sample was applied to the column and allowed to adhere to the heparin-Sepharose matrix for 6 h at 4°C. Then, the sample was eluted from the column, using a continuous salt gradient of 0 -0.5 M NaCl in 10 mM Tris, pH 8.0. Fractions of 3 ml were collected for each of the assays and analyzed using standard Western immunoblots as described above.
Invasion Assays-In vitro assays for invasion were conducted as described previously (25). The gel solution was prepared by mixing 300 l of a Matrigel solution and 300 l of a collagen solution (Vitrogen, final concentration of collagen 0.8 mg/ml). The Matrigel solution was prepared by mixing 200 l of Matrigel with 100 l of serum-free medium, which contained ␣-minimum essential medium supplemented with 0.66 ϫ Nutridoma solution, 0.036 mg/ml sodium pyruvate, and 3.3 mM Hepes, pH 7.4. The collagen solution contained 1.6 mg/ml collagen, 25 l of a 10 ϫ phosphate-buffered saline that was calcium-and magnesium-free (PBS-CMF), 125 l of a 1 ϫ PBS-CMF solution, and 6.6 mM NaOH. The test proteins were subsequently added to this combined gel solution to a final concentration of 0.1 mM. The control gels contained the gel solution without the test protein. The gel solution was coated onto polycarbonate filters inside the Costar inserts in 24-well plates and allowed to set for 1 h at 37°C before cells were added to the wells. The transwell membranes were coated with 10 l of the gel solution.
Following preparation and setting of the gels, human squamous cell carcinoma cells were trypsinized, pelleted, rinsed twice with serum-free medium, and counted. A 200-l aliquot from a suspension of 1 ϫ 10 6 cells/ml was added to each well. Cells were allowed to invade the gel on the membrane filter for approximately 18 h. At the end of the incubation time, transwell filters were washed three times with PBS, fixed in a 1:3 glacial acetic acid:methanol solution for 30 min, washed three times with double distilled water, and stained for 15 min with 0.05 mg/ml of Hoechst stain (bis-benzimide 2Ј-[hydroxyphenyl]-5-[4-methyl 1-1-piperazinyl]-2,5Ј-bis-1H-benzimidazole; Sigma). Filters were washed another three times with double distilled water, placed on a glass slide, covered, and sealed to a glass coverslip using mounting medium (Vectashield, Vector, Burlingame, CA). The cells on the underside of the membrane that invaded the gel and passed through the pores of the membrane appeared bright against a dark background and were visualized for counting with an Axiophot Photomicroscope equipped with a filter for Hoechst stain detection (Zeiss, Oberkochen, Germany). The number of cells from five high power fields (400 ϫ) from each membrane was counted. Each experimental condition was tested in triplicate chambers, with cell counts averaged between the triplicate samples for each experiment.
Cell Adhesion and Spreading-HSC-3 cells were trypsinized, pelleted under centrifugation, washed twice with PBS, and suspended in control medium (serum-free ␣-minimum essential medium supplemented with 0.2% lactalbumin hydrolysate (Life Technologies, Inc.) and 1% penicillin/streptomycin) at a density of 3 ϫ 10 5 cells/ml. 100 l of this cell suspension was aliquoted per well into 96-well plates and subsequently incubated at 37°C in a humidified 5% CO 2 incubator. The four recombinant FN proteins were added to the wells immediately before plating cells, yielding a final protein concentration of 100 M/ well. Cells were photographed after 2 h of incubation with the recombinant proteins at 200 ϫ magnification (26).
Cell Migration-Cells were prepared as described for the cell adhesion assay, except that they were suspended at a density of 1 ϫ 10 6 cells/ml; 100 l of this cell suspension was aliquoted per well into a 96-well plate. After 2 h of incubation, a scratch was made down the center of the well with a sterile needle, removing cells from this area and creating a clear zone into which cells could migrate (27). Photographs of the cell-free scratched area were taken, and then the 96-well plates were incubated for another 7 h. After this time, photographs were again taken to determine if cells had migrated into the scratched area in the presence of the recombinant test proteins. FIG. 1. Recombinant FN proteins. A, experiments were performed with four different recombinant FN protein fragments engineered to examine functions of the high affinity heparin-binding domain and the alternatively spliced V region of FN. All four proteins span type III repeats 10 -15. Two proteins contain the alternatively spliced V region (designated with a V ϩ symbol), while the other two do not (designated with a V Ϫ symbol). Two proteins also contain the high affinity heparinbinding domain (designated with a Hep ϩ symbol), while the other two proteins have a mutated, non-functional heparin-binding domain (designated with a Hep Ϫ symbol). The RGD cell-binding sequence (ϩ) and high affinity heparin-binding sequence (*) are indicated within the clear boxes. The alternatively spliced repeats (EIIIB, EIIIA, and V region) are shaded. The type I, type II, and type III repeating structural domains within the FN molecule are indicated as shown in the key. B, the high affinity heparin-binding domain was rendered non-functional by insertion of synthetic oligonucleotides between restriction sites for ApaI and XbaI. These oligonucleotides preserved the native FN protein sequence but changed two adjacent arginines to threonines in the mutated protein. Positions of restriction sites are underlined. Altered nucleotides are indicated in bold type. The heparin-binding consensus sequence is indicated at the top of the figure. The sequences depict the wild-type FN sequence, the engineered Hep ϩ sequence (nucleotide sequence contains a new restriction site for XbaI but preserves the native FN protein sequence and heparin binding of the expressed protein), and the engineered Hep Ϫ sequence (the two adjacent arginines are mutated to threonines, and heparin binding of the expressed protein is virtually lost).

RESULTS
The diagrams in Fig. 1A depict the III-10 to III-15 repeats of FN, which were expressed as fusion proteins in bacteria. These complementary proteins either included (V ϩ ) or excluded (V Ϫ ) the V region and contained either the wild-type high affinity heparin-binding region (Hep ϩ ) or the mutated sequence (Hep Ϫ ). The high affinity heparin-binding consensus sequence (22) was mutated in III-13 to create the Hep Ϫ proteins (Fig.  1B). Point mutations were incorporated into the FN cDNA such that in the resultant proteins, adjacent arginines were replaced with threonines. These mutations effectively abrogated heparin-binding function.
Electrophoresis of the four purified recombinant FN proteins produced two bands at approximately 70 kDa, which correspond to the V Ϫ Hep ϩ and V Ϫ Hep Ϫ proteins ( Fig. 2A, lanes 1  and 2, respectively), and two bands at approximately 80 kDa, which correspond to the V ϩ Hep ϩ and V ϩ Hep Ϫ proteins ( Fig.  2A, lanes 3 and 4, respectively). Proteins were detected by staining with Coomassie Blue to demonstrate efficacy of purification. The four recombinant proteins all reacted with a mouse monoclonal antibody to rat FN (24) (Fig. 2B).
We examined the functional significance of the heparin-binding mutation by determining the ability of the V Ϫ proteins to bind to columns of heparin-Sepharose. The recombinant pro-teins were applied to the column and eluted in 3-ml fractions with continuous salt gradients from 0 to 0.5 M NaCl. Recombinant protein was detected in the 3-ml fractions by Western immunoblots. The mutated V Ϫ Hep Ϫ protein bound to heparin-Sepharose but eluted from the column in 0.07-0.08 M NaCl (Fig. 3A). In contrast, the native protein, V Ϫ Hep ϩ (Fig. 3B), required approximately 0.5 M NaCl to elute. Since physiologic salt concentration is 0.14 M NaCl, the mutated protein would not bind heparin under physiologic conditions.
Having purified and functionally characterized the recombinant FN proteins, we next examined the invasive properties of HSC-3 squamous carcinoma cells in response to these FN proteins in a collagen/Matrigel matrix. Initial assays performed with only the V Ϫ Hep ϩ and V Ϫ Hep Ϫ proteins demonstrated that both stimulated cell invasion, but that the Hep ϩ protein was more effective than the Hep Ϫ protein in facilitating this process (Fig. 4, B and C). There was at least a 3-fold increase in invasion in gels supplemented with the Hep ϩ protein as compared with gels supplemented with the Hep Ϫ protein (Table I). Additional experiments were performed with all four recombinant proteins (Table II). Both V ϩ proteins induced significantly elevated levels of invasion when compared with gels containing either the V Ϫ Hep Ϫ protein or Matrigel alone. The V Ϫ Hep ϩ protein was also more effective than the V Ϫ Hep Ϫ protein in facilitating invasion. The V Ϫ Hep Ϫ protein was marginally better at inducing invasion than was Matrigel alone.
These results were confirmed in assays with HOC313, a second oral squamous cell carcinoma cell isolate. These cells invaded the gels with the same order of efficacy as the HSC-3 cells (Fig. 5, Table III). The relative magnitude of invasion induced by the FN proteins was as follows: Since adhesion and migration are components of invasion, these processes were evaluated separately for HSC-3 cells in the presence of the recombinant FN proteins. Migration assays  Tables I  and II. were conducted by observing cell movement into a denuded "scratched" area in the culture dishes. As was seen with the invasion assays, the V region played the dominant role in this process (Fig. 6). At 7 h of migration, differences in migration between HSC-3 cells exposed to V ϩ and V Ϫ recombinant proteins were easily apparent, with substantially increased motility in cells exposed to the V ϩ proteins.
The importance of the heparin binding function was more readily demonstrated in assays measuring cell spreading. Cell spreading was maximized in wells supplemented with the Hep ϩ FN proteins but not the Hep Ϫ proteins (Fig. 7). The V ϩ Hep Ϫ protein, which was highly effective in inducing invasion, was not effective in mediating cell spreading. These observations demonstrate that properties of the V region and the heparin-binding domain are both important factors influencing the behavior of the squamous cell carcinoma cells in culture, but that these domains may modulate functions that are separable, depending on the assay. DISCUSSION In this study, invading HSC-3 squamous cell carcinoma cells showed a 3-fold increase in invasion in gels supplemented with the V ϩ Hep Ϫ recombinant FN protein when compared with gels containing the V Ϫ Hep Ϫ protein. Inclusion of the V region of FN enhanced the invasion associated with both Hep ϩ and Hep Ϫ proteins. Similar results were obtained with HOC313, another oral squamous cell carcinoma isolate. Assays for cell migration demonstrated that inclusion of the V region substantially increased motility of HSC-3 cells. However, cell spreading was most enhanced when HSC-3 cells were incubated with recombinant FN proteins in which the high affinity heparin-binding domain was functional. These data demonstrate that both the high affinity heparin-binding domain and the V region of FN mediate invasion by human oral squamous cell carcinoma cells. However, cell spreading is more associated with the heparinbinding function.
The FN constructs used in this study contain all known low affinity heparin-binding sequences within the carboxyl-terminal type III repeats, yet these sequences did not impart significant functional heparin binding or in vitro invasion. Inclusion of functional high affinity heparin-binding sites did, however, increase invasion. These experiments therefore reveal a structure-function correlation between high affinity heparin binding and important cellular responses. We demonstrated that the V Ϫ Hep ϩ protein induced an increase in cell invasion of HSC-3 cells when compared with the V Ϫ Hep Ϫ protein. The invasion induced by the V Ϫ Hep ϩ protein may be mediated in part through heparan-sulfate and chondroitin-sulfate proteoglycan receptors (5,6). However, since the V region, which contains the principal binding site for the ␣ 4 ␤ 1 integrin, is not present, these responses are probably not mediated by this integrin (28). Since the constructs do contain the III-10 repeat and the RGD cell-binding sequence of FN, any interactions with known members of the integrin family probably involve the ␣ 5 ␤ 1 integrin (13).
The V region of FN enhanced the invasive phenotype of both HSC-3 and HOC313 cells. In fact, its presence may predict a The total number of invading cells in 5 high power (400 ϫ) fields was counted. Data represent the mean Ϯ the standard error of the mean derived from triplicate invasion chambers. Data were analyzed using an ANOVA followed by the Scheffé F post hoc test.
*, significantly greater than cells in unsupplemented chambers and in chambers supplemented with the V Ϫ Hep Ϫ and V Ϫ Hep ϩ proteins (p Ͻ 0.0001).  Table III. The total number of invading cells in 5 high power (400 ϫ) fields was counted. Data represent the mean Ϯ the standard error of the mean derived from triplicate invasion chambers. Data were analyzed using an ANOVA followed by the Scheffé F post hoc test.
*, significantly greater than cells in unsupplemented chambers (p Ͻ 0.0002) or in chambers supplemented with the V Ϫ Hep Ϫ protein (p Ͻ 0.003). The total number of invading cells in 5 high power (400 ϫ) fields was counted. Data represent the mean Ϯ the standard error of the mean derived from triplicate invasion chambers. Data were analyzed using an ANOVA followed by the Fisher's post hoc test.
*, significantly greater than cells in unsupplemented chambers and in chambers supplemented with the V Ϫ Hep Ϫ protein (p Ͻ 0.01).
change toward a more invasive phenotype. The increase in invasion may be assisted by the high affinity heparin-binding region, but the V region appears to play the dominant role in this process. Other portions of the FN molecule such as low affinity heparin-binding sequences may also be important factors in the malignant process but were not examined in this study.
The high affinity heparin-binding sequence likely mediates interactions with other FN domains and receptors over large spans of FN. This sequence is separated from the binding sites for the ␣ 4 ␤ 1 integrin in the V region by approximately two type III repeats, and from the RGD binding site for the ␣ 5 ␤ 1 integrin by either two or three type III repeats (depending on whether the alternatively spliced EIIIA domain is excluded or included). Analysis of HSC-3 cells using fluorescence-activated cell sorting has in fact shown that these cells express ␤ 1 , ␣ 4 , and ␣ 5 integrin subunits. 2 Interactions between cell-surface receptors such as proteoglycans, which may bind the high affinity heparin-binding domain, and either of these integrins must, therefore, occur between separated portions of the FN molecule. In contrast, virtually all other known heparin-binding sequences in the type III repeats of FN are clustered in repeat III-14 and are immediately adjacent to the leucine-aspartic acid-valine (LDV) binding site for the ␣ 4 ␤ 1 integrin in the alternatively spliced V region (14). In fact, all three of these low affinity heparin-binding sites and the LDV sequence are located within a stretch of 91 amino acids that encompasses the carboxylterminal portion of III-14, the amino portion of the V region, and the intervening hinge region between these repeats. This entire span is the approximate size of one type III repeat. Thus, it seems likely that the low affinity heparin-binding sites me-diate cellular interactions coordinately with the ␣ 4 ␤ 1 integrin over a short distance.
Other studies have investigated cellular responses to heparin-binding regions of FN by using either of two approaches. In one series of multiple studies (7,26,29,30), synthetic peptides were constructed from short heparin-binding sequences, and cellular responses to these peptides were evaluated in vitro and in vivo. The results demonstrated that some of the peptides were highly effective in modulating functions such as migration and invasion in many different cell types. Most of these peptides were clustered in the 91-amino acid sequence within III-14 and the V region. Interestingly, the synthetic peptide corresponding to sequences within the high affinity heparinbinding domain was often ineffective in modulating cellular functions (7).
The second approach was to express and purify recombinant FN type III repeats (12,15). The intervening sequences between the repeats of interest were often deleted in an attempt to eliminate amino acids believed to be nonessential to the biologic responses being measured. These studies have shown that repeats containing either the RGD cell-binding sequence or the high affinity heparin-binding sequence were marginally effective at promoting invasion of HT1080 fibrosarcoma cells (14). Mixtures of these two individual repeats were not appreciably better. The recombinant proteins were, however, highly effective in promoting invasion when repeats containing the RGD cell-binding site were linked in tandem with repeats containing the high affinity heparin-binding segment. Taken together, these two sets of approaches suggest that the high affinity heparin-binding domain needs to be presented to cells as a contiguous unit, in tandem with other functional domains, such as the RGD site. In addition, these data suggest that for maximal cell function, the high affinity heparin-binding region may indeed mediate interactions over large stretches of the FN molecule.  On the basis of these studies, the carboxyl-terminal heparinbinding sites of FN can be classified into two groups. One group is the low affinity heparin-binding sequences located on repeat III-14 immediately adjacent to the V region, which contains the LDV binding site for the ␣ 4 ␤ 1 integrin. Because of their proximity, the low affinity heparin-binding sequences and the V region may therefore be primarily responsible for modulating interactions between heparin sulfate proteoglycans and the ␣ 4 ␤ 1 integrin. Since these interactions are constrained to a small portion of FN, any agent such as an antibody or a synthetic peptide would be highly effective in altering functions associated with this complex and may interfere with binding of this integrin. Also, since the heparin-binding synthetic peptides with blocking functions are contained within the 91amino acid segment that also contains the LDV binding site, it is possible that the main effect of these peptides is in blocking receptor complexes that are formed with the ␣ 4 ␤ 1 integrin.
The second group is the high affinity heparin-binding site located on repeat III-13, which may require other, nonadjacent cell adhesion sequences to modulate cell function. Previous studies have indicated that small synthetic peptides corresponding to sequences in the high affinity heparin-binding domain are often ineffective in blocking biological activity in vivo (7). This region, in isolation, may also be ineffective in inhibiting tumor metastasis (15). However, our results would indicate that the high affinity heparin-binding domain plays an important role in tumor invasion. Other studies in which type III repeats containing cell-binding sequences were joined to and expressed with repeat III-13 have also indicated that these larger fragments are effective in inhibiting tumor metastasis (16). Interactions of other sequences with the high affinity heparin-binding domain might also be sensitive to differential splicing of either the EIIIA or the V region, since these splicing events would change the spatial relationship between the binding sites for integrins and heparin.