Structural requirements of heparan sulfate for the binding to the tumor-derived adhesion factor/angiomodulin that induces cord-like structures to ECV-304 human carcinoma cells.

Tumor-derived adhesion factor/angiomodulin (AGM) is accumulated in tumor blood vessels and on the endothelial cell surface (Akaogi, K., Okabe, Y., Sato, J., Nagashima, Y., Yasumitsu, H., Sugahara, K., and Miyazaki, K. (1996) Proc. Natl. Acad. Sci. U. S. A. 93, 8384-8389). In cell culture, it promotes cell adhesion and morphological changes to form cord-like structures of the human bladder carcinoma cell line ECV-304. The cord formation is prevented by heparin, which inhibits the binding of AGM to ECV-304 cells. This observation suggests that AGM interacts with cell surface heparan sulfate (HS) proteoglycans. In this study, HS glycosaminoglycans and core proteins of integral transmembrane proteoglycans, syndecan-1 and -4, were identified by immunocytochemistry on ECV-304 cells, and the structural requirements for the interaction of HS with AGM were characterized. Inhibition experiments with sulfated polysaccharides and chemically modified heparin derivatives indicated that sulfate groups were essential for both AGM-HS binding and cord-like structure formation and that the rank order of the different sulfate groups in terms of their contribution was N-sulfate > 6-O-sulfate > 2-O-sulfate. The minimum size of heparin, a chemical analog of HS, required for the binding to AGM was a dodecasaccharide as determined by competition experiments using size-defined heparin oligosaccharides. Thus, a specific sulfation pattern in the HS of cell surface syndecans of ECV-304 cells is required for AGM binding and the morphological changes.

carcinoma cell line EJ-1. This protein stimulates adhesion and spreading of vascular endothelial cells and other types of cells to plastic substrates without interacting with integrins. The cDNA of TAF has been cloned as mac25 from human brain leptomeningeal cells (2). Essentially the same cDNA as mac25 has been cloned from human fibroblasts as prostacyclin-stimulating factor, which stimulates prostacyclin production in vascular endothelial cells (3). Prostacyclin is a potent vasodilator and an inhibitor of platelet adhesion and aggregation, thus it contributes to the maintenance of vascular homeostasis. TAF is accumulated in tumor, but not in normal, blood vessels, which often have a tortuous architecture, and has been implicated in the high permeability of the tumor blood vessels and in the aberrant cell adhesion of tumor cells (1,4). TAF preferentially binds to type IV collagen among various extracellular components, and promotes adhesion of the human tumor cell line ECV-304 to type IV collagen substrate and their morphological change (4). The ECV-304 cells, which were reported to be of a spontaneously transformed human vascular endothelial cell line (5), secrete a large amount of TAF and form a cord-like structure when their confluent culture is stimulated by type I collagen (4), which is reminiscent of capillary formation. In view of its vascular modulating function, TAF has been renamed angiomodulin (AGM) (4).
Recently, however, the German Culture Collection, the American Type Culture Collection, and the Health Science Research Resources Bank (Osaka, Japan) have announced that the ECV-304 cell line is a variant of the human bladder carcinoma line T-24 based on the short tandem repeat microsatellite fingerprint and marker chromosomes. Hence, we used here the ECV-304 cell line for studying AGM-inducible morphological changes to tumor cells. In fact, AGM has been demonstrated to be accumulated in various types of carcinoma cells (4) and to be located in the invasive front of tumor nests possibly because of much production and/or binding of this protein. 2 In cell culture, AGM promotes the adhesion of cancer cells to various extracellular matrix components such as type IV collagen, laminin, vitronectin, etc. (4). These observations indicate that AGM may have some effects on cell adhesion and morphology especially during tumor development. It should be noted, however, that ECV-304 cells show a complex genotype combining markers not only of the human bladder carcinoma line T-24 but also human endothelial cells. The endothelial markers include in particular vascular endothelial growth factor receptor Flt-1 and the von Willebrand factor (6). Furthermore, these cells respond to en-dostatin with dose-dependent inhibition of migration (7). Hence, the morphological changes to ECV-304 cells may also represent some aspects of angiogenesis, and this cell line has been widely used for studying angiogenesis.
In our previous work, we demonstrated that the binding of AGM to the ECV-304 cell surface and the cord-like structure formation of these cells were completely inhibited by exogenously added heparin (Hep), a chemical analog of HS (4). AGM interacts with the cell surface heparan sulfate glycosaminoglycan (HS-GAG) of ECV-304 cells, as demonstrated by inhibition studies using Hep, HS, and an anti-AGM antibody or by HS lyase digestion experiments (4,8). The cell binding site in AGM that interacts with HS has been identified as the peptide sequence Lys 89 -Ser 90 -Arg 91 -Lys 92 -Arg 93 -Arg 94 -Lys 95 -Gly 96 -Lys 97 . This is involved in the cell adhesion and the cord-like structure formation of ECV-304 cells (8).
In this study, HS proteoglycans (HS-PGs) were demonstrated by immunocytochemistry to be present on the surface of ECV-304 cells. HS-GAG was isolated from the cultured cells, structurally characterized, and compared before and after the morphological changes. Furthermore, the structural requirements within the HS chains for the specific binding to AGM were defined.  (10,11) were gifts from Dr. Masayuki Ishihara (National Defense Medical College, Tokorozawa, Japan). Phosphatidylinositol phospholipase C from Bacillus thuringiensis was purchased from Oxford Gly-coSciences (Abingdon, United Kingdom). Anti-syndecan-1 antibody was from Immuno Quality Products (Groningen, the Netherlands). Antiperlecan antibody toward domain III (12) was a gift from Dr. Renato V. Iozzo (Thomas Jefferson University, Philadelphia, PA). Anti-syndecan-2, -3, and -4 antibodies named 10H4, 1C7, and 8G3, respectively (13,14) were gifts from Dr. Guido David (University of Leuven, Leuven, Belgium). Bovine liver HS was prepared by exhaustive chondroitinase ABC digestion of the sulfated GAG fraction that was purified by chromatography on a DEAE-cellulose column through elution with 1.0 M LiCl as described previously (15). Even numbered Hep oligosaccharides were prepared by enzymatic degradation of porcine intestinal Hep as described previously (15). [  Type I collagen was purchased from Koken Co. (Tokyo, Japan). AGM was purified to homogeneity from the serum-free conditioned medium of human bladder carcinoma cell line EJ-1 (1). The synthetic peptide, designated as AGM-peptide, which represents the HS binding site of AGM (8), was custom-synthesized by Biologica (Aichi, Japan).

Materials-Hep
Cell Culture-ECV-304 cells were obtained from the Health Science Research Resources Bank (Osaka, Japan). These cells were grown in plastic dishes using medium 199 (Life Technologies, Inc.) supplemented with 10% (v/v) fetal calf serum and kanamycin sulfate (60 g/ml) at 37°C in 5% CO 2 .
Immunofluorescent Staining-Cells were fixed with 95% methanol at Ϫ20°C for 15 min and then washed with phosphate-buffered saline. After being blocked with 1% bovine serum albumin/phosphate-buffered saline, the cells were incubated with HepSS-1 (diluted 1:100 in phosphate-buffered saline) at room temperature for 2 h, washed with 0.01% Tween/phosphate-buffered saline, and stained with fluorescein isothiocyanate-conjugated anti-mouse IgM antibody (Seikagaku Corp.) at room temperature for 2 h. Fluorescent images were obtained using a laser-scanning confocal microscope FLUOVIEW (Olympus, Tokyo, Japan).
Phospholipase C Digestion-Phosphatidylinositol-specific phospho-lipase C digestion was performed according to the method of Low et al. (16). Cells were treated with 0.1 IU of phospholipase C in supplemented medium 199 at 37°C in 5% CO 2 for 1 h. Cells were then thoroughly washed prior to immunofluorescence staining. Preparation of GAGs from the ECV-304 Cell Culture-To determine their composition, the ECV-304 cell GAGs were metabolically labeled with [ 3 H]GlcN. After the cells had been grown to confluence, the culture medium was replaced with fresh medium supplemented with 135 nM [ 3 H]GlcN (50 kBq/ml) with or without 50 g/ml type I collagen, and incubation continued for 24 h. The cells cultured in the presence of type I collagen formed a network of cord-like structures (17,18). The conditioned medium was removed from each cell culture, and the cells were digested with trypsin for 30 min. Following incubation, the digest was centrifuged, and the supernatant fluid was recovered as a trypsinate fraction. The radioactivity of each fraction was quantified by liquid scintillation counting.
Analysis of 3 H-labeled GAGs-ECV-304 cell [ 3 H]GAGs were isolated from each trypsinate fraction by actinase E digestion and subsequent trichloroacetic acid treatment as described previously (19) and purified by chromatography on a DEAE-cellulose column through stepwise elution with 0.5, 1.0, and 2.0 M LiCl. Because 2.0 M LiCl-eluted fractions contained little radioactivity, 0.5 and 1.0 M LiCl-eluted fractions were used for the composition analysis of GAGs. These fractions were desalted using a Sephadex G-50 column, and the GAGs were identified by enzymatic digestion using Streptomyces hyaluronidase, chondroitinase ABC, or a mixture of heparinase and heparitinases I and II. Streptomyces hyaluronidase digestion was performed using 10 mIU of the enzyme in a total volume of 50 l of 20 mM acetate buffer (pH 5.0) at 60°C for 30 min. The enzyme digests were fractionated on a Sephadex G-50 column, and the radioactivity was measured in a liquid scintillation counter.
The disaccharide composition of CS and HS of the 1.0 M LiCl-eluted fractions was determined as follows. The samples were digested with chondroitinase ABC or the mixture of heparinase and heparitinases I and II. The resultant disaccharides were separated from the resistant materials by gel filtration and subjected to disaccharide composition analysis by HPLC on an amine-bound silica PA-03 column (YMC Co., Kyoto, Japan) (19,20). Samples were co-chromatographed with standard unsaturated disaccharides on HPLC, and eluates were collected at 0.5-min intervals for radioactivity measurement in an Aloka LSC-700 liquid scintillation counter. Individual radioactive peaks were identified by comparison with standard disaccharide peaks detected by absorbance at 232 nm.
Filter Binding Assay-Various amounts of AGM or AGM-peptide were incubated with [ 3 H]HS from ECV-304 cells (50,000 dpm) or bovine liver [ 3 H]HS (26,000 dpm), and AGM with any associated GAG was collected on nitrocellulose filters (15). Filters were monitored by liquid scintillation counting.
Formation of Cord-like Structures by Cultured ECV-304 Cells-ECV-304 cells were grown to confluence in 24-well dishes containing 500 l of medium 199, 10% fetal calf serum. The culture medium was replaced with 500 l of the fresh medium supplemented with 50 g/ml type I collagen, and the cells were incubated further. On the following day, the formation of the network of cord-like structures was assessed by phasecontrast microscopy. To examine the effects of various GAGs on the formation of cord-like structures, various amounts of GAGs, in addition to type I collagen, were added to the culture of ECV-304. On the following day, the cell morphology was examined under a phase-contrast microscope.

RESULTS
ECV-304 cells produce AGM, which promotes the adhesion of these cells to type IV collagen and morphological changes to form cord-like structures when they are stimulated using exogenous type I collagen (4). The cord formation is prevented by Hep, which inhibits the binding of AGM to the tumor cells, suggesting that AGM interacts with cell surface HS-PGs and that such interactions are indispensable to the morphological changes in ECV-304 cells. In this study, HS-GAG was isolated from ECV-304 cells and characterized for its structure as well as for the ability to interact with AGM and to regulate the cord-like structure formation.
Occurrence of HS-PGs on the Surface of ECV-304 Cells-By indirect immunofluorescent staining using the HepSS-1 antibody and subsequent confocal microscopic examinations, a dense staining for HS-GAG was observed on the surface of the ECV-304 cells at three different culture stages. In the logarithmic growing phase, the cell surface staining was evident as shown in Fig. 1A. Although the cells shown in Fig. 1A were permeabilized with methanol prior to staining, a similar cell surface staining pattern was observed using unpermeabilized cells (data not shown). A dense staining of the basal surface, with dark nuclei, was seen in cells undergoing contact inhibition (Fig. 1C) and in cells forming the cord-like structures (Fig.   1E). The staining might be attributable to shed HS (21) and intracellular HS, because the cells were fixed with methanol. The uniform staining of the latter cell culture was in sharp contrast to the intense accumulation of AGM in the cord-like structure (4). Apparently, the undersulfation (see below) of the HS observed concomitantly with the cord formation did not alter the HepSS-1 epitope. Treatment of the cells with a mixture of heparinase and heparitinases I and II, which eliminate the epitope for HepSS-1, completely abolished the immunoreactivity, confirming the specific binding of HepSS-1 to cell surface HS-GAG (data not shown).
The core proteins of the HS-PGs were also examined by immunocytochemical analysis. The binding of the anticore protein antibodies toward a basement membrane PG, perlecan, as well as integral transmembrane PGs, syndecan-1, -2, -3, and -4, to the ECV-304 cells was tested. Anti-syndecan-1 and -4 antibodies bound to the cells, whereas the others did not (Fig. 2). The failure to detect perlecan is surprising inasmuch as it is widely distributed in human tissues and is present not only in the basement membrane zone of virtually all vascularized tissues but also in the tumor stroma of several human cancers (22). The present results may indicate that perlecan is not expressed in ECV-304 cells as in the case of hepatocytes and keratinocytes (22). Alternatively, it may be released into the medium without being retained in the matrix of the cultured ECV-304 cells. The immunostaining of HepSS-1 antibody was not diminished by the treatment with phosphatidylinositol phospholipase C, which releases glypicans from cell surface, suggesting that glypican family members are not appreciably expressed by ECV-304 cells, being consistent with a previous report (6). These results indicate that the major cell surface HS-PGs of ECV-304 cells are syndecan-1 and -4.
To demonstrate directly the HS-GAG production by ECV-304 cells, we labeled cells using [ 3 H]GlcN and analyzed the sulfated GAGs extracted from the cell culture by means of specific GAG lyases. ECV-304 cells form the cord-like structure network when the confluent monolayer is incubated for 24 h in medium containing type I collagen (17). To investigate if the sulfation pattern of the ECV-304 cell surface HS changes upon induction of the cord formation, PGs were labeled with [ 3 H]GlcN for 24 h in the presence and absence of type I collagen, and polyanionic materials were isolated by ion-exchange chromatography. Addition of the type I collagen did not markedly affect the number of ECV-304 cells after 24 h: type I collagen-treated and -untreated control cultures contained 1.4 ϫ 10 3 and 1.7 ϫ 10 3 cells/mm 2 , respectively, whereas the total radiolabel incorporation into the GAG fraction was increased ϳ1.5-fold by induction of the cord formation (Fig. 3). The isolated polyanionic materials were examined for their susceptibility to Streptomyces hyaluronidase, chondroitinase ABC, or a mixture of heparinase and heparitinases I and II. Each digest was subjected to gel filtration on Sephadex G-50 (results not shown), and the amounts of hyaluronic acid, CS, and HS were determined by digestion products from each enzyme treatment. The 0.5 M LiCl-eluted fractions obtained from both cultures contained predominantly hyaluronic acid and a small proportion of CS but no HS (Fig. 3). The mixture of heparinase and heparitinases I and II fragmented ϳ52 and 37% of the [ 3 H]GlcN-labeled GAGs in the 1.0 M LiCl-eluted fraction obtained from cultures in the presence and absence of type I collagen, respectively (Fig. 3). These results indicated that ECV-304 cells synthesized an appreciable amount of HS. CS represented 40 and 32% of the 1.0 M LiCl-eluted fractions isolated from the control and the cord-forming ECV-304 cells, respectively (Fig. 3). The unidentified materials may include keratan sulfate and/or sialylated or sulfated O-linked oligosaccharides derived from glycoproteins.
Disaccharide Composition Analysis of the ECV-304 Cell HS-The HS preparations derived from the ECV-304 cell culture, which had been labeled with [ 3 H]GlcN in the presence or absence of type I collagen, were subjected to a disaccharide composition analysis after digestion with a mixture of heparinase and heparitinases I and II. The resulting disaccharides were analyzed by anion-exchange HPLC. Overall, the disaccharide compositions of both HS samples were similar, and the major disaccharide in both HS fractions was nonsulfated disaccharide ⌬HexA-GlcNAc (Table I). However, there was an increase in the proportion of the nonsulfated disaccharide and decreases in the sulfated disaccharides in the type I collagentreated HS preparation. Based on the disaccharide composition data, sulfation degrees at specific positions of the individual saccharide units were calculated (Table II). The data showed that the average number of sulfate groups/100 disaccharides in the HS preparations from the type I collagen-treated and untreated cell cultures was 93 and 113, respectively. The lower sulfation (82.1%) in the HS preparation from the collagentreated cultures as compared with that from the untreated cultures was attributable to the 16% reduction in N-sulfation, 19% reduction in 6-O-sulfation, and 21% reduction in 2-O-sulfation.
Binding of the ECV-304 Cell HS to AGM-The [ 3 H]GlcNlabeled GAG preparations isolated from the confluent ECV-304 cell culture were incubated with AGM, and their binding ability was evaluated using the nitrocellulose filter binding assay (see "Experimental Procedures"). As shown in Fig. 4, GAG fractions eluted with 1.0 M LiCl on DEAE-cellulose chromatography bound to AGM. The retained counts on the nitrocellulose filters were comparable between the 1.0 M LiCl-eluted fractions prepared from type I collagen-treated and untreated cultures. The binding of the 1.0 M LiCl-eluted fractions to AGM was abolished by treatment with a mixture of heparinase and heparitinases I and II (Fig. 4). A chondroitinase ABC treatment did not affect the bound radioactivity (data not shown). These results indicated that AGM bound to HS rather than CS. In contrast, very little binding to AGM was observed for the GAG fractions eluted with 0.5 M LiCl on DEAE-cellulose chromatography (Fig. 4), being consistent with the observation that the 0.5 M LiCl-eluted fractions contained no HS (Fig. 3).
Binding of Bovine Liver HS to AGM-Because the availability of the 3 H-labeled ECV-304 cell HS was limited, bovine liver HS (15) was used to investigate the structural feature of HS responsible for its binding to AGM by competition experiments.  Bovine liver HS strongly inhibited the binding of the 3 H-labeled ECV-304 cell HS to AGM (data not shown), suggesting that it also contained the binding sequence for AGM. Therefore, bovine liver HS was N-deacetylated with hydrazine and radiolabeled with [ 3 H](CH 3 CO) 2 O. The resultant 3 H-labeled HS (specific radioactivity, 2.6 ϫ 10 5 dpm/g) was evaluated for its ability to bind AGM using the nitrocellulose filter binding assay. Direct binding was observed in a concentration-dependent manner as shown in Fig. 5. Saturation of the binding was obtained at the AGM concentration of 40 g/ml, and a maximal binding of 15% of the added bovine liver [ 3 H]HS was achieved. The binding specificity of the bovine liver [ 3 H]HS for AGM was then investigated by inhibition studies described below using various kinds of oligo-and polysaccharides including size-defined sulfated oligosaccharides and chemically modified GAGs.

Inhibition of the Binding of Bovine Liver HS to AGM by Various GAGs-Effects of various kinds of sulfated polysaccharides were examined on the binding of the bovine liver [ 3 H]HS
to AGM. The [ 3 H]HS (1.0 g/ml) was incubated with AGM (0.76 g) in the presence of increasing amounts (0.5-6.0 g/ml) of various GAGs (Fig. 6). Unlabeled bovine liver HS prevented the binding of the 3 H-labeled bovine liver HS to AGM, and 50% inhibition was observed at 2.0 g/ml (Fig. 6A). Bovine lung Hep and porcine intestinal Hep inhibited the binding also, whereas bovine kidney HS exhibited no inhibition even at 3.0 g/ml. Most of the CS isoforms exhibited no inhibition even at 6.0 g/ml, whereas CS-E unexpectedly had a strong inhibitory effect (50%) at 2.0 g/ml (Fig. 6B). The IC 50 value was approximately equivalent to that of bovine liver HS. Hep preparations were more potent than the HS and CS preparations in these inhibition assays.
The effects of various kinds of sulfated polysaccharides on the binding of bovine liver [ 3 H]HS to AGM were examined to investigate the binding specificity (Fig. 7). [ 3 H]HS was achieved. These results confirmed our previous proposal (8) that the AGM peptide was the HS binding site and that AGM bound to the cell surface HS through interaction with this particular peptide sequence.

Effects of Various GAGs on the Formation of the Collageninduced Cord-like Structures of ECV-304
Cells-We have previously shown that ECV-304 cells produce AGM, which is intensely accumulated in the cord-like structures induced by type I collagen (4). The synthetic AGM-peptide and Hep inhibit the formation of the cord-like structures (4,8). Thus, the cord-like structure formation involves interactions between the cell surface HS and AGM, as well as between the cell surface and collagen (17). To characterize the structural specificities of GAGs responsible for the inhibition of the cord-like formation, we evaluated the effects of various polysaccharides including bovine kidney HS, CS-A, CS-E, chemically desulfated Hep derivatives, and dextran sulfate on the formation of the cord-like structures of ECV-304 cells ( Fig. 10 and Table III). 2ODS-Hep (Fig. 10F) and dextran sulfate (not shown) prevented the cord formation at a concentration similar to that of Hep, whereas bovine kidney HS, CS-A, CDSNAc-Hep, and CDSNS-Hep showed no inhibition. Indeed, bovine kidney HS stimulated to some extent the formation of cord-like structures (Fig. 10E).
6ODS-Hep, NDSNAc-Hep, and CS-E had a modest effect even at high concentrations (100 -150 g/ml). These observations indicate that the type of sulfation is critical to the inhibition of the cord-like structure formation. Experiments using chemically desulfated Hep derivatives suggested that the contribution to inhibition of the cord formation increases with N-sulfate Ͼ 6-O-sulfate Ͼ 2-O-sulfate groups in Hep/HS. This rank order of the sulfate groups is consistent with that for the AGM binding activity described above. These results reinforce the notion that the cord formation of ECV-304 cells depends not only on the interaction between the cell surface and collagen (17) but also on the interaction between the cell surface HS-PGs and AGM.
Interestingly, squid cartilage CS-E inhibited the cord formation as well as the AGM-HS binding (Fig. 6B). CS-E is expressed by ECV-304 cells as demonstrated by the disaccharide analysis after chondroitinase ABC digestion of the GAG preparation. A little yet appreciable amount of the disaccharide unit ⌬HexA-GalNAc(4,6-O-disulfate) typical for CS-E was found, representing 2-3% of the total CS disaccharides as shown in Table IV. Recently it has been demonstrated that CS-E binds to various types of collagen to different degrees (24) and also to Hep-binding growth/differentiation factors midkine (25, 26) and platelet factor 4 (27). CS-E is expressed in a cell-type-and tissue-type-specific manner and in a developmentally regulated fashion (reviewed in Ref. 28). It remains to be determined whether CS-E is a physiological ligand of AGM and is involved in the expression of biological activities of AGM in vivo.

DISCUSSION
The specific cell-binding sequence of AGM, which promotes the adhesion and morphological changes of ECV-304 cells, has been identified as a peptide sequence containing seven basic amino acids that interact with the cell surface HS (8). In the present study, detailed structural requirements of HS for the AGM-HS binding were investigated for the first time. HS-GAG was first identified immunocytochemically on the surface of cultured ECV-304 cells. In addition, core proteins of syndecan-1 and -4 were revealed, indicating that major HS chains are synthesized most likely on these PGs. The HS chains were then isolated from cultured ECV-304 cells and shown to exhibit binding to AGM in vitro. Chemical analyses revealed undersulfated properties of the HS, the sulfation degree of which was comparable to that of bovine kidney HS (29), which showed no inhibitory activity against the AGM-HS binding (Fig. 6A). In

Effects of GAGs on the collagen-induced cord formation
A confluent monolayer culture of ECV-304 cells was incubated with test agents (25ϳ150 g/ml) in addition to 50 g/ml of type I collagen to induce the cord formation. After 24 h, the formation of the network of cord-like structures was assessed by phase-contrast microscopy.

Test agents
Concentrations of test agents 25 50 100 150 The symbols ϩ, Ϯ, or Ϫ represent the degree of inhibition of the cord-like structure formation: ϩ, strong inhibition; Ϯ, weak inhibition; Ϫ, no inhibition. b ND, not determined.
contrast, sulfate groups of the HS are essential for the specific interaction with AGM as shown by inhibition assays using various chemically modified Hep preparations. The rank order of the different sulfate groups for the binding activity was N-sulfate Ͼ 6-O-sulfate Ͼ 2-O-sulfate (Fig. 7). Inhibition assays with Hep-derived oligosaccharides indicated that the minimum size required for the AGM binding was a dodecasaccharide (Fig. 8). These results enforce our previous proposal of the involvement of HS-PGs in the regulation of the morphological changes to ECV-304 cells through interaction with AGM (4). The biological significance of the cord formation of ECV-304 cells, which had been taken as a characteristic of endothelial cells forming capillaries, has to be reconsidered because the cell line is a bladder carcinoma. The observed morphological change surely involves cell adhesion and elongation. Involvement of HS-PGs, in particular syndecans, in cell adhesion and cellular morphology is being increasingly studied (reviewed in Refs. 30 and 31). Although primary receptor classes such as integrins, cadherins, and selectins play major roles in cellmatrix adhesion (32,33), a second class of cell surface molecules including syndecans may modify the type of adhesion mediated by primary receptors. Syndecans transduce extracellular stimuli directly into cytoplasmic signals, and AGM may act as a soluble effector molecule. Syndecan-1 and -4, which were produced by ECV-304 cells, are expressed in a variety of cell types including epithelial cells, endothelial cells, and smooth muscle cells and are type I integral membrane PGs with homologous transmembrane and cytoplasmic domains, the latter of which has two conserved regions and a variable region. The cytoplasmic tail of syndecan-1 interacts with intracellular microfilaments (34 -36) and that of syndecan-4 with focal adhesion molecules (37) being involved in protein kinase C-␣ activation (38). The AGM binding to the cell surface HS chains of syndecans on ECV-304 cells induces morphological changes most likely through such molecular interactions. It is also of note that the invasiveness of B-lymphoid cells derived from human plasma cell leukemia has recently been demonstrated to be regulated strictly by the HS-bearing ectodomain of syndecan-1 that interacts with type I collagen (39).
In this present study, the structural requirements of Hep/HS for inhibition of the cord formation were also investigated. ECV-304 cells cultured in the presence of chlorate, which interferes with sulfation reactions (40), did not show morphological changes (data not shown), indicating that sulfation was essential for the cord formation. The inhibition experiments of the cord formation using various chemically modified Hep preparations indicated that both N-sulfate and O-sulfate groups were necessary to attain complete inhibition. The order of the sulfate groups for the inhibition was as follows: N-sulfate Ͼ 6-O-sulfate Ͼ 2-O-sulfate ( Fig. 10 and Table III), being consistent with the order of sulfate groups essential for the AGM-HS binding. It should be emphasized that the synthetic 20-mer AGM peptide containing the major cell binding site also inhibits the cord formation of ECV-304 cells (8). These findings altogether indicate that the cord-like structure formation of ECV-304 cells induced by type I collagen is dependent on the cell surface HS interacting synergistically with both soluble AGM and type I collagen in the matrix.
Structural differences were observed when HS-GAGs produced by ECV-304 cells in the presence and absence of type I collagen were compared ( Fig. 3 and Table I). The amount of HS-GAGs was increased 2.4-fold by treatment of type I collagen. Disaccharide composition analysis revealed that the HS preparations from both collagen-treated and -untreated control cultures were similar sharing a high proportion of the nonsulfated disaccharide unit. In view of the generally accepted concept of the existence of sulfated saccharide clusters for various HS-binding proteins (41), sulfated disaccharide units are assumed to be in clusters scattered along HS-GAG polysaccharide chains of ECV-304 cells as well, forming a putative high affinity binding domain for AGM and presumably also for type I collagen.
We observed an increase in the nonsulfated disaccharide unit and a decrease in the di-and trisulfated disaccharide units during the morphological transition to cord-like structures, resulting in further undersulfation. The total sulfate groups in the HS preparations were 113 and 93/100 disaccharides before and after the collagen treatment, respectively (Table II). Both N-sulfation and O-sulfation decreased during the formation of the cord-like structures, and the AGM-binding ability of the HS-GAG isolated from the collagen-treated culture is lower than that from the untreated confluent culture (Fig. 4). Dynamic structural changes involving undersulfation associated with decreased AGM binding appear to take place during the morphological changes. There are other examples of subtle yet dynamic structural changes to HS-GAG during the occurrence of prominent biological phenomena. During murine neural development, a single species of HS-PG undergoes a rapid, tightly controlled change in growth factor-binding specificity concomitant with the temporal expression of acidic fibroblast growth factor and basic fibroblast growth factor (42). The HS-GAGs show distinct alterations in sulfation patterns, total chain length, and the number of sulfated domains. A number of earlier studies have indicated malignancy-associated undersulfation of cell surface HS (43)(44)(45). Recent detailed structural analysis has revealed the specific structural changes of HS isolated from human adenoma and carcinoma cells during the colon carcinogenesis (46). The structural alterations associated with the morphological changes to ECV-304 cells observed in this study are subtle but could be significant as in the case of the aforementioned examples and the well established antithrombin III-binding motif of Hep/HS where the 3-O-sulfation leads to activation of the precursor pentasaccharide sequence (47,48). The structural changes to HS chains of ECV-304 cells appear to be the key to switching the binding properties of AGM.
AGM modulates the adhesion and morphology of target cells including tumor cells and primary human endothelial cells. 3 ECV-304 cells targeted by AGM, exhibit endothelial markers as mentioned in the "Introduction," and it is well established that the tube formation of endothelial cells is specifically promoted by type I collagen (17). Type IV collagen, which is colocalized with AGM in tumor blood vessels (4), also promotes the tube formation of endothelial cells in vitro (49), and the 3 A. Ito, J. Sato, and K. Miyazaki, unpublished results. inhibition of type IV collagen biosynthesis prevents angiogenesis in vivo (50). However, the role of type I collagen in the ECV-304 cell culture can not be played by type IV collagen (4). Specific interactions of the cell surface HS with type I collagen in particular have recently been revealed during the capillary formation of cultured human umbilical vein endothelial cells (51). It is tempting to speculate that distinct specific functional structural domains embedded in HS chains are also involved in the interaction with type I collagen.