A Novel Consensus Motif in Fibronectin Mediates Dipeptidyl Peptidase IV Adhesion and Metastasis*

Lung endothelial dipeptidyl peptidase IV (DPPIV/ CD26) is a vascular address for cancer cells decorated with cell-surface polymeric fibronectin (poly-FN). Here, we identified the DPPIV-binding sites in FN and examined the effect of binding site peptides on DPPIV/ poly-FN adhesion and metastasis. Using proteolytic fragments and maltose-binding protein fusion proteins that together span full-length FN, we found DPPIV-bind-ing sites in type III repeats 13, 14, and 15 (FNIII13, -14, and -15, respectively). DPPIV binding was mediated by the consensus motif T(I/L)TGL X (P/R)G(T/V) X and was confirmed by swapping this motif in FNIII13, -14, and -15 with the corresponding region in FNIII12, which did not bind DPPIV. DPPIV binding was lost in swapped FNIII13, -14, and -15 and gained in swapped FNIII12 (FNIII12(14)). Peptides containing the DPPIV-binding domain of FNIII14 blocked DPPIV/poly-FN adhesion and impeded pulmonary metastasis. This study adds to the classes of cell-surface adhesion receptors for FN and will help in the further characterization of the functional implications of the DPPIV/poly-FN adhesion in metastasis and possibly in cell-mediated immunity involving DPPIV-expressing lymphocytes. Dipeptidyl peptidase

Dipeptidyl peptidase IV (DPPIV 1 /CD26) is a 110-kDa type II transmembrane sialoglycoprotein. It is anchored to the plasma membrane via a single transmembrane domain that, together with a six-amino acid cytoplasmic tail, is part of the putative uncleaved signal sequence (1)(2)(3)(4). DPPIV is expressed in various epithelial tissues (e.g. liver bile canaliculi, kidney proximal tubules, prostate gland, and intestine), endothelia (e.g. lung capillaries, kidney vasa recta, and spleen sinusoids), and Tcells (4 -7). The DPPIV functions are diverse and include three major activities. The best known and most widely researched function is that of its serine exopeptidase activity. DPPIV cleaves dipeptides from the N terminus of polypeptides that have either proline or alanine in their penultimate positions (3,4). The DPPIV exopeptidase activity accounts for a variety of regulatory processes, including chemokine regulation (8 -10) and glucose homeostasis (11,12). The catalytic sequence is GWSYG and is located at amino acids 627-631 of rat DPPIV. The second major function of DPPIV is its ability to bind to the extracellular matrix, presumably involving the predicted seven-bladed ␤-propeller domain (13). Binding affinities for both collagen type 1 and fibronectin (FN) have been reported (5,14), and a major function of the DPPIV/FN adhesion has been disclosed in the colonization of the lungs by blood-borne cancer cells (6,(15)(16)(17)(18). The third major function is that of a T-cell co-stimulatory protein (reviewed in Refs. 3, 4, 9, 13, and 19). Although the exact molecular mechanisms by which DPPIV mediates T-cell activation are still unresolved, the ability of DPPIV to interact with CD45, a protein-tyrosine phosphatase (20, 21), and adenosine deaminase (22)(23)(24), each of which are capable of functioning in distinct signal transduction pathways, combined with its enzyme-and FN-binding qualities, may hold the keys to its mode of action in T-cell activation.
The involvement of DPPIV in the colonization of the lungs by blood-borne cancer cells was first discovered when we screened cancer cells with different organ colonization preferences for binding to endothelial cells from various tissue sources. Rather than relying on tissue culture-isolated endothelial cells, screening was performed with outside-out membrane vesicles freshly harvested from the luminal membrane of endothelial cells by in situ vascular perfusion of rat lungs with a formaldehyde/dithiothreitol solution to induce endothelial cell-surface vesiculation (6,15). Consistently, lung endothelial vesicles adhered in large numbers to lung-metastatic cancer cells only, but not to liver-metastatic or non-metastatic cancer cells, whereas control vesicles prepared from a rarely metastasized organ (e.g. leg musculature) adhered at background levels to lung-and livermetastatic and non-metastatic tumor cells. Using a passive/ active immunization schedule involving emulsified leg endothelial vesicles for passive immunization and lung endothelial vesicles for active immunization, we identified a monoclonal antibody that totally blocked the adhesion of lung endothelial vesicles to lung-metastatic cancer cells. The molecule blocked by the antibody was identified as DPPIV (6). Subsequently, we showed that vascular arrest of lung-metastatic cancer cells was mediated by DPPIV adhesion to cancer cell surface-associated FN (16 -18). In contrast to the poor ability of DPPIV to bind to soluble plasma FN, the binding to cell surface-associated FN was of high avidity, suggesting that cell surface-associated FN presents itself in a different configuration than soluble plasma FN (25)(26)(27). Indeed, detailed biochemical analyses of cancer cell surface-associated FN revealed that FN is assembled into large fibrillar polymers (poly-FN) that are dispersed over the cancer cell surface in multiple globular aggregates, apparently exposing multiple DPPIV-binding sites (16,17). Participation of the DPPIV/poly-FN adhesion in lung metastasis was substantiated by the findings that (a) a soluble DPPIV polypeptide representing the entire extracellular domain totally abolishes adhesion of lung-metastatic breast cancer cells to DPPIV and, accordingly, prevents lung colonization (17); (b) the Fischer 344/CRJ rat substrain, which harbors a G633R substitution in DPPIV that leads to retention and degradation of much of the mutant protein in the endoplasmic reticulum (18,28), as well as DPPIV Ϫ/Ϫ mice (obtained from Dr. Marquet, Centre d'Immunologie de Marseille Luminy, INSERM-CNRS) (11) allow lung metastasis at dramatically reduced levels relative to wild-type Fischer 344 rats (18) and C57BL/6 mice, 2 respectively; (c) lung-metastatic tumor cell lines derived from various human, mouse, and rat cancers invariably carry a strong message for endogenous FN and are able to assemble poly-FN on their surfaces; (d) FN surface expression in clones derived from a rhabdomyosarcoma correlates with lung metastasis (29); and (e) human and mouse melanoma cell lines selected for enhanced lung colonization overexpress FN (30,31). Together, these data suggest that the DPPIV/poly-FN adhesion might be a quite common mechanism in the colonization of the lungs by blood-borne cancer cells, where it may operate alone or collaborate with other adhesion molecules, including FN ligands other than DPPIV (e.g. integrins, heparan sulfate proteoglycans, and CD44), to mediate vascular arrest (32)(33)(34)(35).
In this study, we authenticate and refine the role played by the DPPIV/poly-FN adhesion in pulmonary metastasis, identifying DPPIV-binding sites within FN and generating synthetic peptides of these sites that, by their anti-metastatic effect, clearly distinguish DPPIV/poly-FN adhesion from other FN adhesions. The binding sites were found by generating and screening a series of successively shorter FN sequences for DPPIV binding. The sites are located in type III repeats 13, 14, and 15 (FNIII13, -14, and -15, respectively) and represent a common consensus motif. Their DPPIV-binding specificity is confirmed in biochemical, mutational, adhesion, and lung colony assays.
PROTOMAT was used to search for conserved motifs in FNIII13, -14, and -15. The MOTIF program (H. O. Smith, Institute for Genomic Research, Rockville, MD) was first run to provide candidate blocks in the FNIII repeats. It was followed by an automated implementation of the GIBBS sampler program (C. E. Lawrence, Department of Statistics, Harvard University, Cambridge, MA) for reality checking (39). A data base of blocks was constructed by successive application of the automated PROTOMAT system to individual entries in the PROSITE catalog (40) of FNIII repeats keyed to the Swiss Protein Sequence Database (41). 3 If sequences truly have motifs in common, then both runs typically yield similar or, more significantly, identical sets of blocks (39).
Purification of Fusion Proteins and Rat DPPIV-MBP-FN fusion proteins (New England Biolabs Inc.) were purified according to the manufacturer's instructions. Briefly, 2 liters of Escherichia coli culture were spun after a 2-h isopropyl-␤-D-thiogalactopyranoside (0.3 mM) induction. Cell pellets were sonicated in 100 ml of column buffer (20 mM Tris-HCl (pH 7.4), 200 mM NaCl, and 1 mM EDTA) and then spun at 19,000 rpm for 20 min at 4°C. Supernatants were diluted 1:3 with column buffer and passed through an amylose resin column. Columns were washed with 10 volumes of column buffer and eluted with 10 mM maltose in column buffer. The purity of the elutes were evaluated by Coomassie Blue staining of SDS-polyacrylamide gels and Western blotting with anti-MBP pAb. Protein concentrations were measured by the Bradford method (Bio-Rad). DPPIV was immunopurified from rat lungs using mAb 6A3 (16 -18). HA-tagged GST fusion proteins were purified with anti-HA mAb-conjugated agarose beads.
Gel Overlay Assay (Far-Western)-Equal amounts of N-terminal FN proteolytic fragments (29, 45, (36,42). Gel-separated proteins were transferred to nitrocellulose membranes and denatured with 4 M guanidine HCl in Tris-buffered saline containing 0.1% Tween 20 (TTBS) for 2 h. Proteins were then slowly renatured in TTBS. After blocking with 5% skim milk in TTBS for 2 h at room temperature, membranes were incubated overnight with 2 g/ml DPPIV in 5% skim milk at 4°C, followed by extensive washing with TTBS and incubation with anti-DPPIV pAb CU31. Bound anti-DPPIV antibodies were detected with horseradish peroxidase-conjugated goat anti-rabbit secondary antibody and visualized by ECL TM (Amersham Biosciences).
Other Biochemical Methods-A modified enzyme-linked immunosorbent assay (ELISA) was used to measure the binding of DPPIV to FN fusion proteins. FN fusion proteins were coated at the indicated concentrations onto 96-well microtitration plates overnight at 4°C. The plates were washed three times with phosphate-buffered saline (PBS) and then incubated overnight with 5 g/ml immunopurified DPPIV at 4°C. DPPIV binding was detected with anti-DPPIV pAb CU31, horseradish peroxidase-conjugated goat anti-rabbit antibody, and ECL. DPPIV affinity precipitation (pull-down assays) was carried out with DPPIV-conjugated Affi-Gel 10 beads (16). FN fusion proteins (5 g/ml in 300 l of PBS containing 0.1% ␤-octyl glucoside) were incubated overnight with DPPIV beads at 4°C. The beads were washed with 0.1% ␤-octyl glucoside-containing PBS and then boiled for 10 min in SDS sample buffer. Bound SDS-PAGE-separated proteins were Westernprobed with anti-MBP pAb. For the co-immunoprecipitation of DPPIV and MBP-FNIII repeats, DPPIV (5 g/ml in 300 l of 0.5% ␤-octyl glucoside-containing PBS) was incubated with the same concentrations of wild-type or mutant MBP-FNIII single repeats for 6 h at 4°C. Complexes were then incubated overnight with anti-MBP pAb-conjugated protein A-agarose beads at 4°C. After washing with 0.5% ␤-octyl glucoside-containing PBS, immunoprecipitates were subjected to SDS-PAGE and immunoblotting with anti-DPPIV pAb CU31 or anti-MBP pAb. To examine the heparin-binding abilities of wild-type and mutant MBP-FNIII fragments, fusion proteins were incubated overnight with heparin-agarose beads at 4°C. After washing with 0.5% ␤-octyl glucoside-containing PBS, the beads were boiled for 10 min in SDS sample buffer and subjected to SDS-PAGE, and the resolved proteins were immunoblotted with anti-MBP pAb.
Peptide Binding to Purified DPPIV and DPPIV-expressing Cells-Immunopurified DPPIV (10 g/ml) was coated onto 96-well microtitration plates overnight at 4°C. After blocking with 1% bovine serum albumin (BSA), the coated plates were incubated with biotinylated PEP14 or PEPcon (or GST-PEP14 or GST-PEPcon) for 2 h at room temperature. Bound peptide was detected by ELISA using horseradish peroxidase-streptavidin or anti-GST pAb. HEK293 cells were transfected with either rat DPPIV cDNA or vector (pRc-CMV) alone (18). Surface expression of DPPIV was evaluated by fluorescence-activated cell sorting using anti-DPPIV mAb 6A3. To determine the ability of biotinylated PEP14 and PEPcon to bind to cell surface-expressed DPPIV, transfected cells were incubated with 0.1 mg/ml biotinylated peptide for 1 h at room temperature. The cells were then washed with PBS, and bound peptide was detected by staining with Texas Redconjugated streptavidin for 30 min at 4°C. Staining patterns were observed by fluorescence microscopy.
Cell Adhesion and Lung Colony Inhibition Assays-Cell adhesion and lung colony assays were described previously (16,42). For adhesion inhibition experiments, DPPIV-binding MBP-FNIII fusion proteins (160 nM) were coated onto the wells of microtitration plates overnight at 4°C (control, 30 g/ml poly-L-lysine) and then blocked with 1% BSA in PBS and conjugated with DPPIV at a concentration selected to equate the amount of DPPIV bound to plastic as determined by ELISA with anti-DPPIV pAb CU31. The wells were then seeded with tumor cells (3-5 ϫ 10 4 cells/well) and incubated for 30 min at 37°C. The percent tumor cell adhesion was determined as described (16). Alternatively, DPPIV (10 g/ml) was coated onto the microtitration plate directly overnight at 4°C, blocked with 1% BSA in PBS, and incubated with GST-PEP14 (control, GSP-PEPcon) for 2 h at room temperature. After washing with PBS, tumor cells were added, and adhesion was determined as described above. For lung colony inhibition assays, GST-PEP14 or GST-PEPcon (1 mg/mouse) was injected intravenously 15 min prior to tumor cells (1 ϫ 10 5 cells/0.2 ml of Dulbecco's modified Eagle's medium/mouse). To exclude adverse reactions of GST-PEP14 with DP-PIV-expressing T-cells, lung colony assays were performed in T-celldeficient female 4-week-old Scid/beige mice (Charles River Laboratories, Wilmington, MA).
The functional significance of the DPPIV/FNIII12-15 adhesion was substantiated in adhesion inhibition assays with four cancer cell lines. Three of these cell lines, MTF7, B16-F10, and MDA-MB-231, consistently colonized the lungs of rats and mice upon tail vein injection, expressed strong endogenous messages for FN, and accumulated significant amounts of aggregated FN on their surfaces (Fig. 3A). In contrast, MCF7 tumor cells did not colonize the lungs and were negative for endogenous message and surface accumulation of FN. Accordingly, MTF7, B16-F10, and MDA-MB-231 cells strongly adhered to DPPIV-coated dishes as well as to DPPIV bound to dishes coated with poly-L-lysine, but failed to adhere to DPPIV immobilized on dishes coated with FNIII12-15 (Fig. 3B). MCF7 cells did not adhere to DPPIV (Fig. 3B). In these experiments, the DPPIV coating concentration was selected such that equal binding to the three substrates (plastic, MBP-FNIII12-15, and poly-L-lysine) was achieved as determined by ELISA.
DPPIV Binds to FNIII13, -14, and -15-Further dissection of the DPPIV-binding polypeptide MBP-FNIII12-15 was done to yield dual (MBP-FNIII12-13, MBP-FNIII13-14, and MBP-FNIII14 -15) and single (FNIII12, -13, -14, and -15) type III repeats to narrow the DPPIV-binding site to a single repeat (Fig. 1). Among these constructs, only the single repeat FNIII12 lost DPPIV-binding ability; all other fragments strongly bound DPPIV as observed by Far-Western and pulldown analyses (Fig. 4A, single repeats shown only). These binding interactions were specific because DPPIV bound neither to FNIII12 nor to MBP, and control beads failed to pull down any of the FN repeats (Fig. 4A). Parallel ELISA analyses also demonstrated that DPPIV bound to FNIII13, -14, and -15 (Fig. 4B). However, these data did not reveal whether DPPIV binding to FNIII13, -14, and -15 was mediated by three distinct binding domains or a common binding site present in each of the three repeats. To address this issue, we performed MTF7/ DPPIV adhesion inhibition experiments in which DPPIV was bound to dishes coated with any one of the three repeats. The data show that all three fragments totally abolished the binding interaction between MTF7 cells and DPPIV (Fig. 4C). Similarly, FN fragments containing one or more of the three DPPIV-binding repeats also inhibited MTF7 cell adhesion to DPPIV (data not shown), suggesting that FNIII13, -14, and -15 might harbor a common DPPIV-binding motif (provided DPPIV contains a single FN-binding site).  (43,44) is the most conserved among the three DPPIVbinding repeats FNIII13, -14, and -15, but is most divergent in FNIII12, which does not bind DPPIV (Fig. 5A). The underlined motif includes the last four amino acids of the E ␤-strand and extends over the entire length of the solution-accessible EF loop region (44).
To test this putative DPPIV-binding motif, we swapped the T(I/L)TGLX(P/R)G(T/V)X motif sequence in FNIII13, -14, and -15 with the corresponding loop in FNIII12 rather than deleting the entire loop and causing structural alterations in the repeat that might interfere with function (Fig. 5B). In DPPIV pull-down assays, we showed that wild-type FNIII13, -14, and -15 repeats were successfully pulled down by DPPIV, whereas the three FNIII13(12), FNIII14 (12), and FNIII15(12) mutants were not, although equal amounts of wild-type and mutant proteins were subjected to pull down (Fig. 5C, upper and middle panels). To ensure that our swapping strategy did not perturb the inherent repeat structure and inadvertently affect DPPIV binding, we examined the FNIII13(12) and FNIII14 (12) mutants, which also contain the second heparin-binding domain (45) located remotely from the swapped sequence, for  (14)). C, binding of DPPIV and heparin to wild-type or mutant FNIII single repeats is shown by DPPIV and heparin pull-down assays (middle and lower panels, respectively) (see "Experimental Procedures"). Bound and loading amounts of the MBP fusion proteins were revealed by Western blotting with anti-MBP pAb. D, anti-MBP co-immunoprecipitates (IP) of wild-type or mutant FNIII single repeats and DPPIV were Western-probed with anti-DPPIV pAb CU31 (upper panels) and anti-MBP pAb (lower panels). Note that DPPIV failed to pull down and co-immunoprecipitate the mutant constructs, whereas heparin binding was unaffected by the mutant, indicating that the DPPIV-and heparinbinding sites are distinct. E and F, FNIII12 gained DPPIV binding by placing the consensus sequence of FNIII14 in lieu of the corresponding site of FNIII12 (FNIII12(14)). Binding of DPPIV to wild-type FNIII12 or FNIII12 (14) was examined by DPPIV pull-down assay (E, lower panel) and anti-MBP co-immunoprecipitation and Western blotting with anti-DPPIV pAb CU31 (F, upper panel) and anti-MBP pAb (F, lower panel). The presence of equal amounts of wild-type and mutant FNIII12 used in the assays is shown in the upper panel of E. IB, immunoblot. their abilities to mediate heparin binding. Both wild-type and mutant FNIII13 and FNIII14 were equally competent to mediate heparin binding (Fig. 5C, lower panels). The same DP-PIV-binding behavior for wild-type and mutant FNIII13, -14, and -15 was also observed in co-immunoprecipitation experiments. DPPIV co-immunoprecipitated with wild-type (but not mutant) FNIII13, -14, and -15 using anti-MBP antibodies (Fig.  5D). Although the data presented so far clearly show that the identified consensus motif is essential for DPPIV binding to FNIII13, -14, and -15, the ultimate test for its serving as the DPPIV-binding domain was obtained when we introduced the consensus motif of FNIII14 into FNIII12 in lieu of the corresponding sequence. This domain exchange resulted in gain of DPPIV binding in the FNIII12(14) mutant, as observed by DPPIV pull-down assays (Fig. 5E) and co-immunoprecipitation (Fig. 5F).
Peptides Harboring the DPPIV-binding Site of FNIII14 Impede Tumor Cell Adhesion to DPPIV and Pulmonary Metastasis-To examine whether the consensus motif by itself is sufficient for DPPIV binding, we synthesized biotinylated peptides based on the consensus motif of FNIII14 (PEP14). Because the corresponding peptide from FNIII12 was insoluble, we used a sequence from the same loop in FNIII EDA (the second alternatively spliced repeat) as the control peptide, which displayed similar charge and hydrophobicity characteristics as PEP14, i.e. EDTAELQGLRPGSEYTVS (PEPcon). First, we measured the binding abilities of these biotinylated peptides to immobilized DPPIV by ELISA using streptavidin detection. PEP14 bound to DPPIV in a dose-dependent manner, whereas PEPcon failed to bind DPPIV at all test concentrations (Fig. 6A). Second, we performed peptide affinity precipitation assays in which biotinylated peptides were bound to streptavidin-conjugated beads before being exposed to immunopurified soluble DPPIV (Fig. 6B, upper panel). Only PEP14 was able to precipitate DPPIV (Fig. 6B, middle panel), whereas beads alone or PEPcon-conjugated beads left DPPIV in the flow-through (Fig.   6B, lower panel). To test whether PEP14 also binds to DPPIV protein in its native transmembrane state, we transfected HEK293 cells with rat DPPIV and measured the ability of the transfected cells to bind the biotinylated peptides by Texas Red-conjugated streptavidin staining. Again, PEP14 bound selectively to the DPPIV-expressing HEK293 cells, but not to the mock-transfected HEK293 cells, whereas PEPcon bound to neither cell type (Fig. 6C).
Finally, we examined PEP14 in tumor cell adhesion and lung colony assays. For economical reasons and to guarantee longer survival in serum, we conducted our experiments with a GST-PEP14 fusion protein rather than with synthetic PEP14. GST-PEP14 blocked adhesion of MTF7 cancer cells to DPPIV in a dose-dependent manner, causing a 50% inhibition of adhesion at a concentration of 3.0 M (Fig. 7A). The control fusion protein GST-PEPcon had no adhesion inhibitory effect at any of the test concentrations. In lung colony assays, GST-PEP14 consistently reduced the colonization of the lungs by MTF7 cells, allowing the generation of fewer and smaller tumor colonies than GST-PEPcon (Fig. 7B). DISCUSSION Fibronectin is a "pro-metastatic" gene that is overexpressed in cancer cells selected for enhanced lung colonization (16 -18, 28 -30). This pro-metastatic property of FN has been associated with the ability of many lung-metastatic cancer cells to assemble FN on their surfaces into multiple, randomly dispersed aggregates generated by FN self-association (16,17). FN polymers facilitate adhesion to DPPIV, a distinct pulmonary vascular address that promotes homing of blood-borne cancer cells to the lungs (6, 16 -18). Binding to FN does not engage DPPIV enzymatic activity (16) and is therefore distinct from the binding of a non-FN lung-targeting peptide to membrane dipeptidase, a second pulmonary dipeptidase-type vascular address (46). To learn more about the DPPIV/poly-FN "lung homing" mechanism and to identify "tumor homing" peptides for future FIG. 6. Binding between DPPIV and PEP14 (representing the consensus motif of FNIII14). A, 96-well microtitration plates coated with DPPIV (10 g/ml) overnight at 4°C were incubated for 2 h with various concentrations of the synthetic biotinylated peptide PEP14 or the corresponding control peptide from FNIII EDA (PEPcon) at room temperature. Peptide binding was detected as described under "Experimental Procedures." B, biotinylated peptides (PEP14 and PEPcon) bound to streptavidin-conjugated agarose beads were used to precipitate DPPIV (bead-bound peptides (upper panel)). Both bead-bound and -unbound (flow-through (F.T.)) materials were analyzed by Western blotting with anti-DPPIV pAb CU31 (middle and lower panels, respectively). C, biotinylated PEP14 (arrowheads), but not PEPcon, bound to DPPIV-transfected HEK293 cells as shown by staining with Texas Red-conjugated streptavidin (left and middle panels, respectively). Mock-transfected HEK cells served as a negative control (right paned). IB, immunoblot.
therapeutic application (47), we screened a series of FN fragments that together span the length of the FN molecule for their abilities to bind DPPIV using various molecular, biochemical, and functional assays. We have shown that FN contains multiple DPPIV-binding sites, one in each of FNIII13, -14, and -15. The binding sites are located in the N-terminal part of a common consensus motif identified by the PROTOMAT algorithm (39). Comparison of the identified DPPIV-binding motif with sequences of the corresponding EF regions in the other FNIII repeats showed that the sequence Thr-(Leu/Ile)-Thr located at the end of the E-strand and prior to the hydrophobic core residue Leu is the most conserved sequence in FNIII13, -14, and -15 (44). Preliminary data show that replacement of Thr-Ile-Thr in FNIII14 with Ile-Val-Ser (the corresponding sequence in FNIII12) results in loss of DPPIV binding (data not shown). However, replacement of Ile-Val-Ser in FNIII12 with Thr-Ile-Thr from FNIII14 did not restore DPPIV binding in FNIII12, suggesting that the Thr-Ile-Thr tripeptide acts as a DPPIV-binding site only in the context of the identified consensus motif. This notion is supported by the observation that FNIII9, the only FNIII repeat other than FNIII13, -14, and -15 harboring Thr-Leu-Thr in the same region, also was unable to confer DPPIV binding to FNIII8 -11.
The three DPPIV-binding domains are located within an FN region (FNIII13-15) that is involved in multiple biological functions (34,48). For example, FNIII13-15 is important in FN fibrillogenesis (49); binds human immunodeficiency virus type 1 gp120/160 to reduce virus infectivity (50); and harbors adhesion sites for integrins ␣ 4 ␤ 1 and ␣ IIb ␤ 3 (33,48), the chondroitin sulfate proteoglycan CD44 (35), and heparin (second heparinbinding domain) (34,44,48,(51)(52)(53). The integrin-and heparin-binding domains are located in areas distinct from the DPPIV-binding sites, and each is characterized by a unique sequence (45,48), suggesting that they may operate in an independent manner. However, studies by Sharma et al. (44) showed that the integrin ␣ 4 ␤ 1 -and heparin-binding domains located on opposite faces of FNIII13-14 may affect binding of one binding domain by loading or unloading the other. This type of interaction between distinct binding domains has also been observed for the heparin-and DPPIV-binding domains, where occupation of the heparin-binding site with its ligand totally inhibited DPPIV binding to MTF7 cells in vitro and abolished lung colonization of these tumor cells in vivo. 4 Although we have no precise explanation for this preliminary result, it is likely that heparin binding acting in an allosteric manner perturbs FNIII13, -14, and -15 binding to DPPIV. Full disclosure of the dynamic interplay between ligand-bound and -unbound cell adhesion domains in FNIII13-15 may therefore provide new insights into the conditions that regulate FN binding engagement with integrins, heparan sulfate proteoglycans, and DPPIV and may explain the different binding behavior of FN under different microenvironmental conditions (33-35, 44, 48 -51).
Functional analysis with a peptide representing the DPPIVbinding domain of FNIII14 showed competitive blocking of the DPPIV/poly-FN adhesion and a powerful anti-metastatic effect. The extent of these anti-adhesive and anti-metastatic effects was similar to that reported for adhesion-blocking antibodies and the soluble extracellular domain of DPPIV (17). A similar anti-metastatic effect by a 22-mer synthetic peptide harboring the complete DPPIV-binding motif of FNIII14 has also been reported for T lymphoma cells colonizing liver and spleen (52). Although various integrins (for which the 22-mer peptide had no binding affinity) were implicated in the anti-metastatic effect, it is likely that the peptide blocked the binding interaction between hepatic and splenic sinusoidal FN and DPPIV, which has been shown to be prominently expressed on the T lymphoma cells used (53).
Here, we have identified and characterized a novel cellbinding domain on FN and have presented data showing that binding of endothelial DPPIV to this FN domain plays an important role in pulmonary metastasis. However, at this time, it is unclear whether the DPPIV/poly-FN adhesion acts alone in mediating vascular arrest and metastasis or whether it requires the cooperation of a secondary adhesion event, as reported for the adhesion interaction between immune/inflammatory cells and endothelial cells (reviewed in Refs. 54 and 55). Because all lung-metastatic cancer cells tested in our laboratory, including MDA-MB-231, B16-F10, 4T1, MTF7, LLC1, and K7M2, exhibit high message and protein levels for both FN and integrin ␤ 4 and adhere to both of the respective pulmonary addressins (DPPIV and CLCA), it is conceivable that pulmonary vascular arrest of blood-borne cancer cells is initiated by the DPPIV/poly-FN adhesion and is stabilized by the CLCA/ integrin ␤ 4 adhesion discovered in this laboratory (42,56). Circumstantial support for this notion came from studies of the CLCA/integrin ␤ 4 adhesion under flow conditions (57). These experiments showed that integrin ␤ 4 -expressing tumor cells adhere to CLCA-coated dishes only at low shear stresses. However, once adhesion is established, high shear stress levels are FIG. 7. PEP14 inhibits DPPIV adhesion and metastasis of MTF7 tumor cells. A, dishes coated with DPPIV (10 g/ml) overnight at 4°C were treated with GST-PEP14 or control GST-PEPcon as described in the legend to Fig. 6A, seeded with MTF7 (3 ϫ 10 4 cell/wells), and incubated for 30 min at 37°C, and the percent specific adhesion was determined (18). Data represent means Ϯ S.D. of triplicate measurements in three independent experiments. *, p Ͻ 0.01, relative to GST-PEPcon binding to DPPIV. B, GST-PEP14 (1 mg/mouse) injected intravenously 15 min prior into MTF7 tumor cells (1 ϫ 10 5 cells/0.2 ml of PBS/mouse) caused a dramatic reduction in the number and size of MTF7 lung colonies relative to GST-PEPcon. #, means Ϯ S.D. are from two independent experiments, each involving eight female 6-week-old Scid/beige mice; *, p Ͻ 0.01, relative to MTF7 binding inhibition (Inh.) by GST-PEPcon. unable to break it, suggesting that, for engagement of the CLCA/integrin ␤ 4 adhesion, a "slowdown" of tumor cells may be beneficial. We predict that such a slowdown is mediated by the DPPIV/poly-FN adhesion, which is facilitated by the prominence of tumor cell-surface FN polymers that easily engage in binding interactions with endothelial cells. The necessity for a dual tumor cell/endothelial cell adhesion principle in lung metastasis is also supported by our finding that lung-metastatic, integrin ␤ 4 -expressing tumor cells colonize the lungs of DPPIV Ϫ/Ϫ mice at reduced levels relative to DPPIV ϩ/ϩ mice. 2 Finally, the expression of DPPIV on cells other than lung endothelial cells could extend the importance of our findings beyond metastasis. A principal DPPIV expresser cell is the T lymphocyte (3,4,13,19), in which DPPIV serves the role of a coactivator (19). Although DPPIV enzymatic activity and the collaboration of DPPIV with other molecules such as CD45 and adenosine deaminase have been implicated in this activating role (20 -24), future studies will be needed to explore whether the ability of T lymphocytes to bind to fibrillar FN via surfaceexpressed DPPIV, operating independently or in concert with integrins (e.g. ␣ 4 ␤ 1 ) (58 -61), is the missing link in the activation mechanism of T lymphocytes. If this is indeed the case, then the DPPIV/poly-FN adhesion could also be involved in T-cell-mediated cytolysis of tumor cells.