Endocan Is a Novel Chondroitin Sulfate/Dermatan Sulfate Proteoglycan That Promotes Hepatocyte Growth Factor/Scatter Factor Mitogenic Activity*

Proteoglycans that modulate the activities of growth factors, chemokines, and coagulation factors regulate in turn the vascular endothelium with respect to processes such as inflammation, hemostasis, and angiogenesis. Endothelial cell-specific molecule-1 is mainly expressed by endothelial cells and regulated by pro-inflammatory cytokines (Lassalle, P., Molet, S., Janin, A., Heyden, J. V., Tavernier, J., Fiers, W., Devos, R., and Tonnel, A. B. (1996) J. Biol. Chem. 271, 20458–20464). We demonstrate that this molecule is secreted as a soluble dermatan sulfate (DS) proteoglycan. This proteoglycan represents the major form either secreted by cell lines or circulating in the human bloodstream. Because this proteoglycan is specifically secreted by endothelial cells, we propose to name it endocan. The glycosaminoglycan component of endocan consists of a single DS chain covalently attached to serine 137. Endocan dose-dependently increased the hepatocyte growth factor/scatter factor (HGF/SF)-mediated proliferation of human embryonic kidney cells, whereas the nonglycanated form of endocan did not. Moreover, DS chains purified from endocan mimicked the endocan-mediated increase of cell proliferation in the presence of HGF/SF. Overall, our results demonstrate that endocan is a novel soluble dermatan sulfate proteoglycan produced by endothelial cells. Endocan regulates HGF/SF-mediated mitogenic activity and may support the function of HGF/SF not only in embryogenesis and tissue repair after injury but also in tumor progression.

In the last few years, the vascular endothelium has been shown to play a crucial role in inflammation, coagulation, angiogenesis, and tumor invasion, primarily through the fine regulation of receptor-ligand interactions and secretion of different mediators. Endothelial cells also express several proteo-glycans such as decorin, biglycan, PG-100, glypican, and members of the syndecan family that regulate intercellular interactions and activation processes.
Proteoglycans are complex macromolecules that consist of a polypeptide with one or more glycosaminoglycan chains covalently bound to a serine (or rarely a threonine) residue. Different families of proteoglycans have been described (1-3) as follows: heparan sulfate, chondroitin sulfate, dermatan sulfate, and keratan sulfate proteoglycans. Heparan sulfate proteoglycans have come to particular prominence recently because of their multiple regulatory interactions with growth factors, enzymes, enzyme inhibitors, and components of the extracellular matrix (4,5). Dermatan sulfate proteoglycans (DSPGs) 1 from human fluids become of particular importance during inflammation and response to injury (6 -11), contributing, for example, to the majority of the FGF-2-dependent cell proliferation involved in the regulation of wound repair (6). DSPGs are also known to bind to many HS-binding proteins, including FGF-2, interleukin-7 (12), platelet factor 4 (13), fibronectin (14), and heparin cofactor II (15). Hepatocyte growth factor/scatter factor (HGF/SF) also binds both HS and DS (16,17), and its interaction with DS occurs through a minimal octasaccharide sequence and requires nonsulfated iduronate residues in combination with 4-O-sulfated N-acetylgalactosamine (17).
HGF/SF is a mesenchyme-or stroma-derived multipotent factor that regulates the growth, motility, and morphogenesis of various cell types, especially endothelial and epithelial cells (18 -21), and mediates epithelial-mesenchymal interactions (22,23). During early embryonic development, HGF/SF supports organogenesis and morphogenesis of various tissues and organs, including the liver, kidney, lung, and gut (24 -26). In adult tissues, HGF/SF elicits a potent organotrophic function that supports regeneration following injuries of organs like the liver, lung, and kidney (27). Moreover, aberrant expression of HGF/SF or the Met receptor leads to tumor development (28). In neoplastic tissues, HGF/SF is involved in tumor progression and metastasis (29) through tumor-stroma interactions (30). HGF/SF is synthesized as an inactive monomer, which undergoes internal proteolysis to yield a biologically active, disulfidelinked heterodimer (31,32). HGF/SF is a ligand for the Met receptor tyrosine kinase that is encoded by the c-met protooncogene (33,34). However, like many other HS/DS-binding growth factors, HGF/SF must interact with both Met and HS/DS to elicit a cellular response. Thus HS (35)(36)(37)(38) and DS (35) can potentiate HGF/SF signaling.
We previously described an endothelial cell-specific molecule 1 (ESM-1) that is interestingly restricted to endothelial cells and to the lung and kidney (39). The synthesis and secretion of ESM-1 are up-regulated by tumor necrosis factor, interleukin-1, and lipopolysaccharide and down-regulated by interferon-␥ (40). ESM-1 can regulate CD11a/CD18 integrin (LFA-1)mediated activation of leukocytes through binding to the LFA-1 and its ability to inhibit ICAM-1 binding (41). We now demonstrate that the major form of ESM-1 consists of a heavily glycosylated 165-amino acid mature polypeptides. The carbohydrate moiety of ESM-1 is a single chain of the glycosaminoglycan DS covalently linked to serine 137. ESM-1 promotes the HGF/SF-mediated proliferation of human embryonic kidney cells, in a similar way to heparin. This activity is strictly dependent on the presence of the DS chain. These findings indicate that ESM-1 may be directly and specifically involved in the regulation of HGF/SF activities. Because ESM-1 is specifically secreted by endothelial cells as a proteoglycan we propose to rename ESM-1 as "endocan."

Cell Cultures and Materials
CHO cells were cultured in ␣-minimum Eagle's medium (Life Technologies, Inc.) supplemented with 10% fetal calf serum. SV40-transfected human endothelial cells (SV1 cells) (42) were cultured in RPMI 1640, containing 2 mM L-glutamine and 10% fetal calf serum. Human embryonic kidney cells (293 cell line) were cultured in DMEM with 10% fetal calf serum. 293 cells used for the test of proliferation were cultured in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) supplemented with 10 g/ml insulin and 10 g/ml transferrin. Proteinase K and chondroitinase ABC were purchased from Roche Molecular Biochemicals; chondroitinases B, ACI, and C were from Sigma; and heparinase III was from Grampian Enzymes (Orkney, UK). Human HGF/SF was from R & D Systems (Abingdon, UK), and heparin was from Sigma. Anti-endocan monoclonal antibodies were produced and purified as described previously (40).

Development of Cell Lines Expressing Endocan
The full-length endocan cDNA was excised from the pCDM8 vector, purified, and inserted into XhoI-HindIII-digested pcDNA3 expression vector (Invitrogen, Groningen, The Netherlands). The construct was transfected into CHO and 293 cells using LipofectAMINE (Life Technologies, Inc.), followed by selection with G418 (1000 or 300 g/ml, respectively). Stably transfected cell lines were obtained by limit dilution, and the resulting cell lines were named CHO-Endocan and 293-Endocan.

Determination of the O-Glycosylation Site of Endocan by Mutation Analysis
Two potential O-glycosylation sites in the endocan polypeptide, serine 137 and threonine 120, were predicted by the NetOglyc 2.0 Prediction Server. The codons for serine 137 and threonine 120 in the endocan cDNA were each mutated to an alanine codon by PCR with the Quick-Change site-directed mutagenesis kit, according to the manufacturer's recommendations (Stratagene, Cambridge, UK). The resulting alanine mutations in the cDNA were confirmed by sequencing on an ABI Prism 377 automated DNA sequencer (PE Biosystems, Courtaboeuf, France). The mutated cDNAs were then transfected into 293 cells to obtain transient and stable transfectants (293-S137A/endocan and 293-T120A/ endocan cell lines).

Purification of Endocan
Purification of Endocan from Established Cell Lines-The cell lines were cultured in suspension in medium without fetal calf serum (CHO-SFM II and 293-SFM, Life Technologies, Inc.). After 3-4 days in culture, the medium was collected and stored at Ϫ20°C until use. The pH of the medium was adjusted to pH 8, before application to a column (2.5 ϫ 10 cm) of DEAE-Sepharose (Amersham Pharmacia Biotech). The column was washed with 0.2 M NaCl, 50 mM Tris, pH 8, and then eluted with a gradient of 0.8 -1 M NaCl in the same buffer. Collected fractions were adjusted to 0.2 M NaCl, 50 mM Tris, pH 8.0, and applied to an immunoaffinity column made by immobilizing an anti-endocan mAb (MEC4) on Affi-Gel Hz Hydrazide gel, following the recommendations of the manufacturer (Bio-Rad). After washing with 0.2 M NaCl, 50 mM Tris, pH 8.0, endocan was eluted with 3 M MgCl 2 , concentrated, and dialyzed against the same solution using an Ultrafree 10-kDa molecular mass cut-off membrane (Millipore, Bedford, MA). The final material was then quantified by immunoassay for endocan, and its purity was verified by SDS-PAGE followed by Coomassie Blue or Alcian Blue staining. The nonglycosylated form of endocan (S137A/endocan) was purified in only one step by immunoaffinity chromatography. The degree of purity of WT/endocan and S137A/endocan was checked by gel filtration chromatography on Superdex 75/200 HR 10/30 columns (0.13 ϫ 30 cm; Amersham Pharmacia Biotech) and yielded a single peak. The preparations were shown to be endotoxin-free by the limulus amebocyte lysate test (BioWhittaker, Verviers, Belgium).
Purification of Endocan from Human Plasma-Eight hundred ml of spent plasma, kindly provided by the Etablissement de Transfusion Sanguine (Lille, France), was first precipitated with 60% ammonium sulfate. The precipitate was dissolved and dialyzed against 0.2 M NaCl, 50 mM Tris, pH 8.0. The dialysate was applied to a 50-ml Affi-Gel pre-column (Bio-Rad) prior to immunoaffinity chromatography, which was then performed as described above.

Immunoprecipitation, Immunoblotting, and Amino Acid Sequencing
Cells in 80-cm 2 flasks were lysed by incubation in PBS containing 0.5% Nonidet P-40 and an anti-protease mixture (Roche Molecular Biochemicals) for 30 min at 4°C with agitation. Cell lysates were cleared by centrifugation at 10,000 ϫ g for 15 min. One g of antiendocan mAb (MEP19) or ICAM-1 mAb (clone 164B) was added to the cleared lysates and cell supernatants and incubated overnight at 4°C with agitation. Fifty microliters of anti-mouse immunoglobulin conjugated to agarose beads (Sigma) was added to the samples at 4°C for 90 min, and then the agarose beads were collected by centrifugation, washed twice with lysis buffer, and then twice with PBS. The beads were resuspended in 20 -40 l of SDS-PAGE sample buffer, boiled for 2 min, centrifuged, and the polypeptides separated by SDS-PAGE. Following SDS-PAGE, polypeptides were transferred to nitrocellulose membranes according to standard procedures. After a blocking step, the membranes were incubated for 1 h with endocan mAb (MEP14) at 1 g/ml, followed by a 1-h incubation with an anti-mouse Fc horseradish peroxidase-conjugated secondary antibody (Sigma). Immunoreactivity was detected by ECL (Amersham Pharmacia Biotech). For amino acid sequence analysis, purified endocan was transferred to a polyvinylidene difluoride membrane (Millipore, Bedford, MA) after SDS-PAGE and stained with 0.1% Coomassie Blue. The protein band of 50 kDa was excised from the membrane, and the N-terminal sequence was determined by automated Edman degradation on an ABI 473A protein sequencer.

Digestion by Proteinase K and Purification of the GAG from Endocan
To determine the size of the glycosaminoglycan chains, purified endocan was digested with proteinase K with an enzyme:endocan ratio of 1:50 (w/w) in 0.2 M NaCl, 10 mM Tris, pH 8.0, at 56°C for 3 h. The sample was then incubated overnight with 0.5 ml of DEAE-Sepharose at 4°C. The DEAE-Sepharose was poured into a Bio-Spin chromatography column (0.5 ϫ 3 cm; Bio-Rad) and washed with 10 column volumes of buffer. GAG chains were eluted with 1 ml of 1.2 M NaCl, 50 mM Tris, pH 8, and then aliquoted and stored at Ϫ70°C. Losses of GAG chains during purification were estimated by applying untreated endocan to the same volume of DEAE-Sepharose; from an input of 120 g, 100 g of endocan was recovered in 1 ml of elution buffer. Thus, total endocan input to the digest was adjusted to give a final estimated GAG concentration of 200 g/ml. Samples were analyzed on 12% SDS-PAGE, followed by Coomassie Blue or Alcian Blue staining.
Structural Analysis of the GAG Chain 3 H Labeling of the GAG Chains of Endocan-293-Endocan cells were cultured in suspension in 250 ml of 293 SFM medium until the end of the log phase of cell division. The cells were then centrifuged and resuspended in 250 ml of fresh medium containing 5 Ci/ml D-[6-3 H]glucosamine hydrochloride (Amersham Pharmacia Biotech). After 48 h of incubation at 37°C, the cell supernatants were processed for the purification of endocan as described above, in two steps including DEAE-Sepharose ion-exchange chromatography followed by immunoaffinity chromatography. This yielded a specific activity of 24,000 cpm/g of endocan.
The 3 H-labeled GAG chains were released from endocan by ␤-elimination with 50 mM NaOH, 1 M NaBH 4 at 45°C for 24 h. After neutralization with 1 M acetic acid, the GAG was precipitated by addition of 3 volumes of ethanol containing 1.3% (w/v) potassium acetate at Ϫ20°C. The GAG precipitate was collected by centrifugation, washed with 75% (v/v) ethanol, re-centrifuged, air-dried, and then re-dissolved in water. To remove some residual 3 H-labeled small oligosaccharides (probably O-linked glycosylation), the sample was applied to a 0.5-ml column of DEAE-Sepharose and washed with PBS. The GAG chains were step eluted with 1 ml of 1.5 M NaCl. The sample was then desalted on a PD-10 column equilibrated in distilled water and concentrated by centrifugal evaporation.
Disaccharide Analysis-[ 3 H]GAG chains (40,000 cpm) were exhaustively digested to disaccharides with two additions of 10 mIU of chondroitinase ABC in 50 mM NaCl, 50 mM Tris-HCl, pH 8.0, at 37°C for 24 h. Digests were boiled and centrifuged to remove protein and then diluted to 1 ml with distilled water adjusted to pH 3.5 with HCl. Samples were applied to a 5-m Hypersil (0.46 ϫ 25 cm; ThermoQuest, Runcorn UK) strong anion-exchange high pressure liquid chromatography column equilibrated in water pH 3.5. After a wash with pH 3.5 water, the disaccharides were resolved over a linear gradient of 0 -0.4 M NaCl, pH 3.5, at a flow rate of 1 ml/min. Fractions of 0.5 ml were collected and counted for radioactivity. The 3 H-labeled disaccharides were identified by comparison with the elution positions of known CS/DS and heparan disaccharide standards monitored by on-line UV detection at 232 nm. Analyses were performed on duplicate enzyme digests.
Glucuronate:Iduronate Ratio-[ 3 H]GAG chains (50,000 cpm) were exhaustively digested with either chondroitinase ACI (40 mIU in 50 mM NaCl, 50 mM Tris-HCl, pH 7.3) or chondroitinase B (8 mIU in 50 mM NaCl, 50 mM Tris-HCl, pH 8.0). Digests were initiated with half the enzyme and incubated at 37°C for 16 h, followed by the addition of the second enzyme aliquot and incubation for a further 2 h. Digests were then boiled and centrifuged to remove the protein, and the cleared supernatants were applied to a Bio-Gel P10 gel filtration column (1 ϫ 170 cm) equlibrated with 0.1 M NaCl and run at a flow rate of 4 ml/h. Fractions of 1 ml were collected and counted for radioactivity. The V o and V t values of the column were determined with undegraded GAG and sodium dichromate, respectively. The percentage of galactosaminyl bonds cleaved by each of the specific enzymes was calculated from the distribution of 3 H radiolabel across peaks corresponding to known oligosaccharide size fractions, relative to the total eluted 3 H radiolabel, by the use of a standard algorithm.

Anti-coagulant Activity
Control platelet-poor plasma (PPP) was prepared from blood anticoagulated with sodium citrate (30 healthy donors) by centrifugation at 2500 ϫ g for 15 min. All the reagents were purchased from Stago Diagnostica (France). Three parameters were evaluated, by adding endocan or buffer or heparin to PPP.
APTT-Activated partial thromboplastin time explores the intrinsic pathway of blood coagulation (FI, FII, FV, FVIII, FIX, FX, FXI, and FXII). Deficit or inhibition of one of these factors enhances coagulation time of the reactive mixture (PPP, cephalin, activator, and CaCl 2 ).
TCT-Thrombin clotting time is performed on PPP ϩ thrombin. With a standard concentration of thrombin, the clotting time of plasma is constant. Abnormalities of fibrin formation induce an increase of the coagulation time.
Anti-Xa Activity-Anti-Xa activity of heparin, or of other inhibitors acting on FXa, is detected by this competitive assay. The studied sample (PPP ϩ endocan, buffer, or heparin) is mixed with FXa and a FXaspecific chromogenic substrate. The final coloration is inversely proportional to the inhibitor concentration.

Thrombin Generation Test
This global sensitive assay can detect platelet or plasmatic abnormality inducing a lag time or a decrease of thrombin generation. In five healthy donors, platelet-rich plasma (PRP) was prepared from blood anticoagulated with sodium citrate by centrifugation at 150 ϫ g for 10 min. Thrombin generation test was performed for each subject, in samples without endocan, with unfractionated calcium heparinate (0.5 IU anti-Xa/ml) or with 0.2, 0.5, or 1 g/ml endocan (final concentration). Endocan was added 10 min before the assay. At 37°C, 1 ml of plasma was mixed with 1 ml of CaCl 2 . Aliquots of 0.1 ml were removed from the reaction each minute for 15 min. Clots formed in the reactive mixture were regularly removed. Aliquots were mixed with 0.2 ml of fibrinogen (Sigma, 4/1000 in Owren buffer) at 37°C, and clotting time was measured for each aliquot. Thrombin formed in the reactive mixture acts on fibrinogen, inducing fibrin formation. The clotting activity was maximal between 4 and 8 min and then decreased on account of neutralization of thrombin by anti-thrombin.

Gel Filtration of Endocan and Its GAG Chain
For native endocan, 50 g of purified WT/endocan or S137A/endocan in 0.5 M NaCl, 50 mM Tris, pH 8, were chromatographed on Superdex 200 or Superdex 75 HR10/30 columns (0.13 ϫ 30 cm; Amersham Pharmacia Biotech), respectively, using a Bio-Rad Biologic Chromatography System (Bio-Rad) at a flow rate of 1 ml/min in the same buffer. Fractions of 1 ml were collected, and endocan was detected by a specific ELISA. The following proteins in the low and high molecular mass calibration kits (Amersham Pharmacia Biotech) were used as standards: ribonuclease A (bovine pancreas, 13.7 kDa), ovalbumin (hen egg, 43 kDa), albumin (bovine serum, 67 kDa), aldolase (rabbit muscle, 158 kDa), ferritin (horse spleen, 440 kDa), and thyroglobulin (bovine thyroid, 669 kDa). Molecular weight standards were run immediately after WT/endocan and S137A/endocan. The elution times of the standards were used to construct a standard linear curve, K av ϭ f(log MR), to determine the apparent molecular mass of WT/endocan and S137A/endocan.

Proliferation Assay
Growth-promoting activity was determined by measuring [methyl-3 H]thymidine incorporation into 293 cells. Cells were seeded at a density of 1 ϫ 10 4 /well in Techno Plastic Products (TPP) 96-well microplates and maintained for 24 h in DMEM supplemented with transferrin and insulin. Recombinant human HGF/SF was diluted in PBS containing 0.1% bovine serum albumin and added to triplicate wells to obtain a final concentration of 50 ng/ml. WT/endocan, S137A/ endocan, and the purified GAG derived from endocan or heparin were added alone or together with HGF/SF, as indicated in the figure legends. After 96 h 0.5 Ci of [methyl-3 H]thymidine/well was added for 16 h and [methyl-3 H]thymidine incorporation into DNA was determined on a TopCount Microplate Scintillation Counter (Packard Instrument Co., Rungis, France). Assays were performed in triplicate. Cell viability was measured by the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide reduction test.

Binding Assays
Binding reactions were carried out at 20°C in an IAsys Autoϩ dual channel resonant mirror biosensor (Affinity Sensors, Saxon Hill, Cambridge, UK) using streptavidin-derivatized aminosilane surfaces as described (43, 44) with minor modifications. Endocan was biotinylated on the amino groups of the core protein (43,44) and then immobilized on the streptavidin-derivatized surface. HGF/SF did not bind to the streptavidin-derivatized planar aminosilane surfaces (43). A single binding assay consisted of adding the HGF/SF at a known concentration in 5 l of PBST to a cuvette containing 45 l of PBST. The association reaction was followed over 210 s. The cuvette was then washed three times with 50 l of PBST, and the dissociation of bound ligate into the bulk PBST was followed over time. The surface was regenerated by washing twice with 50 l of 2 M NaCl, 10 mM Na 2 HPO 4 , pH 7.2. Binding parameters were calculated from the association and dissociation phases of the binding reactions using the non-linear curve fitting FastFit (Affinity Sensors).
A single binding assay yielded four binding parameters: the slope of initial rate of association, the on-rate constant (k on ), and the extent of binding, all calculated from the association phase, and the off-rate constant (k off , equivalent to the dissociation rate constant, k d ), calculated from the dissociation phase.

Characterization of Endocan Produced by 293 and CHO
Cells-In a previous work, we showed that endocan could be immunoprecipitated from HUVEC supernatants and human plasma and that it migrated as a diffuse band of about 50 kDa on SDS-PAGE. To obtain large quantities of endocan in its native form, the 293-endocan and CHO-endocan cell lines were established. Immunoprecipitation of endocan from the culture medium of these cells followed by Western blotting showed a band quite similar in size to that observed for endocan from the culture medium of HUVEC (Fig. 1A). To determine whether endocan was processed as a secretory molecule, as suggested by the predicted N-terminal amino acid sequence, endocan was purified from the 293-endocan cell line. The experimentally determined N-terminal sequence of the 50-kDa form indicated that the signal peptide of 19 amino acids was cleaved at the predicted site, leading to a mature endocan polypeptide of 165 amino acids beginning at W (WSNNYAVD-polypeptide). The predicted size of the mature polypeptide (20 kDa) is inconsistent with the apparent molecular mass of 50 kDa observed on immunoblots, suggesting that the secreted form of endocan was post-translationally modified. Oligomerization through disulfide bonds was not involved because reducing conditions did not reduce its molecular weight (Fig. 1C). On the other hand, purified endocan stained much more strongly on SDS-PAGE with Alcian blue (Fig. 1C) than with Coomassie Blue (Fig. 1B), indicating that endocan was heavily glycosylated.
Serine 137 Supports the O-Glycosylation of Endocan-A computer search for putative sites of glycosylation identified three potential sites of O-glycosylation, i.e. serine 16, threonine 120, and serine 137, but no N-glycosylation sites. Threonine 120 and serine 137 were mutated to generate an alanine (T120A/endocan and S137A/endocan, respectively). These mutant proteins were then transiently expressed in 293 cells, immunoprecipitated from cell lysates and supernatants, and analyzed by Western blotting. The T120A/endocan mutant migrated at 50 kDa, similar to the apparent molecular mass of the wild-type form of endocan ( Fig. 2A). In contrast, S137A/endocan migrated at 22 kDa, a size close to that of the intracellular form of wild-type endocan and consistent with the predicted molecular mass of the polypeptide (Fig. 2A). Immunoprecipitations, performed from transiently transfected COS and CHO cells, gave similar results, strongly suggesting that only serine 137 supported glycosylation in these different cell types (data not shown). To examine the size of the GAG chain, WT/endocan was treated with proteinase K. Alcian blue staining after SDS-PAGE showed a shift from 50 kDa to a mean of 30 kDa after proteolysis, in clear agreement with the loss of the 20-kDa polypeptide core from endocan (Fig. 2B).
Molecular Size of Endocan Determined by Gel Filtration-To determine the hydrodynamic size of endocan under native conditions, purified WT/endocan and S137A/endocan were analyzed on Superdex 200 and 75 gel filtration columns, respec-tively. By using an endocan-specific ELISA, a single peak of S137A/endocan was observed (K av ϭ 0.235), corresponding to a molecular mass of 21.8 kDa (Fig. 3B), in clear agreement with the predicted molecular mass, and also the molecular size observed after SDS-PAGE and Western blot ( Fig. 2A). Interestingly, WT/endocan chromatographed as a single peak (K av ϭ 0.133) corresponding to an average molecular mass of 400 kDa (Fig. 3A). The deduced molecular mass of WT/endocan was in apparent discrepancy with the molecular size observed after SDS-PAGE and Western blot (Fig. 1). This result indicates that the carbohydrate moiety has a profound effect on the physicochemical properties of endocan.
The Carbohydrate Content of Endocan Is Sensitive to Chondroitinases but Not to Heparinase III-To characterize the type of glycan attached to Ser-137, endocan was treated with a variety of chondroitinases and heparinase III. Incubation of endocan with heparinase III failed to alter the apparent molecular mass of endocan on SDS-PAGE (Fig. 4). In marked contrast, chondroitinase ABC treatment decreased the molecular mass of the secreted endocan to 22 kDa (Fig. 5A). Chondroitinase ABC similarly reduced the molecular mass of se- Purified endocan (5 g) from SV1 cells was run on a 15% SDS-PAGE gel and stained with Coomassie Blue to detect the peptidic core of the molecule. Arrow indicates the expected position of endocan, which failed to stain. C, detection of purified endocan with Alcian blue. Purified endocan (5 g) from SV1 cells was run on a 15% SDS-PAGE gel and revealed by Alcian blue staining to detect the glycan component of the molecule. Arrow indicates endocan. creted endocan derived from 293-endocan cells and from the human endothelial cell line SV1. Because endocan circulates in the bloodstream, we next studied the behavior of purified endocan from human plasma. The major band of 50 kDa also shifted to 22 kDa after treatment with chondroitinase ABC. (Fig. 5A). Thus, the nature of the carbohydrate moiety of endocan on Ser-137 seemed to be a chondroitin sulfate (CS)/DS glycosaminoglycan (GAG). To determine the specific type of CS/DS chondroitinases B, ACI and C were employed. Both chondroitinases ACI and C decreased the apparent molecular mass of the complete endocan population from 50 to 22 kDa (Fig. 5, C and D). By comparison, chondroitinase B generated a similar size reduction for the majority of the endocan population. The remainder of the population was affected to lesser degrees, giving rise to a smear on SDS-PAGE extending between the intact 50-kDa endocan and the 22-kDa core protein (Fig. 5B). This susceptibility to chondroitinase B indicates that endocan specifically contains a GAG chain of the DS type.
Composition of the DS Chain from Endocan-The DS chain from endocan was exhaustively digested to disaccharides using the nonspecific chondroitinase ABC. A disaccharide yield of 98% was obtained (not shown). The disaccharides were resolved and identified by strong anion-exchange high pressure liquid chromatography (Fig. 6A). This allows an analysis of the content and distribution of specific sulfate groups but not of specific uronate residues. The DS chain was found to consist of the following: 2.8% nonsulfated, 19.4% mono-6-O-sulfated, 66.5% mono-4-O-sulfated, 5.1% di-2,6-O-sulfated, 4.6% di-4,6-O-sulfated, and 1.5% di-2,4-O-sulfated disaccharides. The average sulfate density is thus 1.08 sulfates/disaccharide (with 0.73, 0.29, and 0.07 attached to C-4, C-6, and C2). The preponderance of GalNAc(4-OSO 3 ) relative to GalNAc(6-OSO 3 ) is typical for a DS species, as epimerization of GlcA to IdoUA takes place adjacent to GalNAc(4-OSO 3 ) residues. To probe the degree of epimerization of GlcA to IdoUA within the DS chain, the specific chondroitinases ACI and B were utilized. Chondroitinase ACI cleaves all galactosaminyl linkages adjacent to GlcA, except in the relatively rare occasions when GlcA is 2-O-sulfated (45). In contrast, chondroitinase B cleaves nearly all galactosaminyl linkages adjacent to IdoUA (with or without 2-O-sulfation), except on the non-reducing side of the rare nonsulfated IdoUA-GalNAc disaccharide (45). Thus the linkages individually susceptible to these two enzymes are mutually exclusive and quantitatively should approximate to the contents of GlcA and IdoUA, respectively. In practice, the values will tend to be slight underestimates, reflecting the occasional presence of rare resistant linkages, coupled with the inability to accurately quantify breakdown within the larger oligosaccharides eluting in and close to the void volume of the gel filtration column. Chondroitinase ACI extensively digested the endocan DS chain. The proportion of galactosaminyl linkages cleaved was 91.9% with the great majority of products being disaccharides (Fig. 6B, i). However a small proportion of sequences up to at least decasaccharide in length were resistant, which will correspond to contiguous sequences of IdoUAcontaining disaccharides. As expected from the above extensive breakdown, chondroitinase B elicited only 5.8% bond cleavage. Only a small proportion of disaccharides was produced, together with a large range of intermediate sized oligosaccha-rides, suggesting that the relatively infrequent IdoUA residues only occurred singly or in relatively widely spaced small clusters (Fig. 6B, ii). The additive bond cleavage obtained with both enzymes was 97.7%, indicating relatively little scope for underestimations in the individual digests.
Effect of Endocan on Coagulation-Endocan was secreted as a dermatan sulfate proteoglycan by endothelial cells. Because DS is known to affect both the generation of thrombin in vitro (46) and coagulation, the potential anticoagulant activity of endocan was tested on APTT, TCT, anti-Xa activity, and on thrombin generation test. As shown in Fig. 7A, endocan, even at high doses from 0.2 to 1 g/ml, did not modify these different tests. APTT, TCT, and anti-Xa activities were similar for PPP with buffer or with endocan. In positive controls, APTT, TCT, and anti-Xa activities were higher for PPP with heparin. In addition, endocan had no inhibitory effect on the thrombin generation test; there were no differences between 0.2, 0.5, or 1 g/ml endocan and buffer, whereas heparin induced a marked lag phase in thrombin formation (Fig. 7B).
HGF/SF Binds to Endocan and Exhibits Increased Mitogenic Activity-A recent report (17) has demonstrated that DS binds to HGF/SF. Because endocan contains a DS chain, we asked if endocan could also affect HGF/SF activity. To evaluate the consequences of such an interaction, [methyl-3 H]thymidine incorporation into 293 cells was measured in the presence of 50 ng/ml HGF/SF alone or together with different amounts of WT/endocan. Initially we observed that HGF/SF alone induced 293 cell proliferation at a level of 45% relative to serum alone, whereas WT/endocan alone did not trigger 293 cell proliferation (data not shown). However, when added together with HGF/SF, WT/endocan at 2.5 g/ml enhanced considerably the HGF/SF-induced proliferation by 160% (Fig. 8A). This enhancing effect of endocan on HGF/SF activity was similar to that of heparin (Fig. 8A) and was shown to be dose-dependent, becoming significant at a concentration of 100 ng/ml (Fig. 8B). To examine the respective involvement of the polypeptide and the DS chain of endocan, the mutant S137A/endocan (corresponding to the nonglycanated form) and the proteinase K-treated endocan (corresponding to the isolated DS peptidoglycan) were compared with WT/endocan in the proliferation assay. Neither the nonglycanated form of endocan nor the free DS chain had any effect alone on the proliferation of 293 cells. Similarly, the nonglycanated form of endocan was unable to enhance the HGF/SF-induced stimulation of 293 cell proliferation (Fig. 8B) even at high concentrations. In marked contrast, the purified DS peptidoglycan chain consistently enhanced HGF/SF-induced proliferation, with an 88% increase compared with HGF/SF alone (Fig. 8B). Although less pronounced than the effect of WT/endocan, the enhancement of HGF/SF proliferative activity by the peptidoglycan chains was dose-dependent. Thus, these results clearly demonstrated that endocan increased the HGF/SF-induced proliferation of 293 cells and that this activity was solely due to the DS chain of endocan.
Binding of HGF/SF to Endocan-The kinetics of the interaction between HGF/SF and endocan were determined in an optical biosensor. The binding of HGF/SF to immobilized endocan was monophasic at all concentrations of the growth factor. Thus a one-site binding model was used to fit all the data. HGF/SF bound endocan with fast kinetics and a high affinity (Table I). Importantly, the K d values of this interaction determined from the kinetic parameters (6.6 Ϯ 2.6 nM) and from the extent of binding at equilibrium (3.4 Ϯ 1.6 nM) were compara- ble, demonstrating the internal consistency of the binding assays.

DISCUSSION
In this study we identify a novel soluble DS proteoglycan secreted by endothelial cells, which promotes HGF/SF-induced proliferation through its glycosaminoglycan chain. Based on the restricted expression of this proteoglycan to endothelial cells, we propose to call it endocan.
CS/DS are linear polysaccharides that consist of repeating disaccharide units composed of a uronic acid and an N-acetylgalactosamine. CS and DS share the same polymer precursor. Whereas CS chains uniformly contain glucuronic acid, the presence of any iduronic acid is sufficient to define operationally the GAG as a DS rather than a CS. Thus, the sensitivity of apparently all the endocan GAG chains to chondroitinase B, which cleaves glycosidic linkages on the non-reducing side of iduronate residues, defines the GAG chain as a DS, even though the degree of breakdown, and thus the iduronate content, is relatively low. The importance of iduronic acid is that its sugar ring is flexible, allowing substantial conformational change that may be essential for interactions with proteins (47,48). Synthesis of CS/DS is initiated by the formation of a tetrasaccharide linker, xylose-galactose-galactose-glucuronic acid, with the xylose covalently linked to a serine residue on the protein (49). Serine 137 in the endocan polypeptide supports the attachment of the sole DS chain. The linearity of the GAG chain explains that in gel filtration chromatography the hydrodynamic size of endocan corresponds to an apparent molecular size of 400 kDa.
Endocan does not appear to belong to the small leucine-rich repeat (SLRP) family of CS/DS PGs, such as decorin (4,5) biglycan (50), lumican, and fibromodulin (51). This family of proteoglycans is characterized by conserved cysteines that form disulfide-bonded loops near both termini of the protein core and highly homologous internal leucine-rich repeats, which consist of about 80% of the deduced sequences and have been postulated to mediate protein-protein interactions (52). Such conserved leucine-rich repeats are absent from the endocan polypeptide. The endocan core protein is also only about half the size of the core proteins of the SLRP family. Moreover, small leucine-rich PGs are localized essentially within the extracellular matrix and occasionally peripherally associated with the cell surface, whereas endocan is mostly secreted in to the bloodstream (40). From these data, endocan would appear to represent a novel soluble DSPG, which may define a new family of PGs.
In the last few years, there has been an emerging concept reporting that many families of PGs are implicated in diverse biological processes through binding to growth factors like FGF family members (53,54), transforming growth factor-␤ (55), vascular endothelial growth factor (56), and HGF/SF (35). Several studies suggest that the interactions between HS-binding proteins and heparan sulfate proteoglycans could regulate their biological activities (57,58) and orientate them toward their target cell (59,60) at the site of the inflammatory reaction. The expression of CS/DSPGs is differentially regulated in human endothelial cells by inflammatory cytokines (7), and DS released in vivo after injury can support FGF-2-mediated cell proliferation (6). Interestingly, endocan, a DSPG devoid of any effect on coagulation, is also secreted by human vascular endothelial cells, and we have demonstrated previously (39,40) that the secretion of endocan is under the control of pro-inflammatory cytokines. This suggests a possible involvement of endocan in local repair following injury.
Our results indicated that endocan significantly increased human embryonic kidney cell proliferation in the presence of HGF/SF. By contrast, the nonglycanated form of endocan had no effect. Moreover, the DS chains isolated from endocan were able to mimic the effects of endocan. These findings demonstrate that the biological activity of endocan on HGF/SF function is mediated by its GAG moiety and not related to the polypeptide. Interaction of HGF/SF with either HS or DS occurs within the IdoUA-rich domains of these GAGs (16,17). However, the minimum number of IdoUA residues required for an HGF/SF-binding site has not yet been determined. The ability of the endocan DS chains to bind HGF/SF with an affinity comparable with that of decorin (16,17) and to potentiate the growth stimulatory activity of HGF/SF, when only about 6% of their constituent uronates are IdoUA, which tend to occur in only small isolated clusters, suggests that not all the uronates in the minimal putative hexasaccharide-binding sequence may need to be IdoUA. Also, only one or two IdoUA clusters may need to be present in a DS chain for maximal activity, which may indicate that a single DS chain may only be able to accommodate one, or possibly two, bound HGF/SF molecules at any one time due to the large size (84 kDa) of this growth factor.
The in vivo significance of the ability of endocan to potentiate HGF/SF activity is as yet unknown. However, in the kidney, endocan was shown to be expressed selectively in distal tubules (40). As a recent study of a human multicystic dysplasic kidney showed that HGF/SF protein was preferentially localized in the same portion of the nephron (33), this might suggest an involvement of endocan in these HGF/SF-dependent pathological disorders. Another field of pathology where HGF/SF has been implicated recently is the development of breast (29), renal (61), and lung (62) carcinomas as well as malignant melanoma (63). HGF/SF is thought to favor extension of hyperplasia and generate cells with an increased invasive and metastatic phenotype. HGF/SF also increases angiogenesis, mainly when combined with vascular endothelial growth factor (64,65). Endocan may be implicated in these phenomena of dysregulated HGF/SF motogenic, mitogenic, and angiogenic activities. Supporting this idea, the concentration of endocan was found to be elevated in the sera of patients with lung cancer (data not shown). Moreover, the concentration of HGF/SF has been similarly reported to be elevated in sera of patients presenting with non-small cell lung carcinoma (66).
In conclusion, we demonstrate that the major form of endocan is a soluble proteoglycan that contains a single chain of DS. a The S.E. is derived from the deviation of the data from a one-site binding model, calculated by matrix inversion using the FastFit software provided with the instrument ("Experimental Procedures"). No evidence was found for a two-site model of association, and so the HGF/SF-binding sites in each length of oligosaccharide were homogeneous in this respect. Four independent sets of k on were measured, and the four resulting values for k a and their errors were combined.
b The correlation coefficient of the linear regression through the k on values.
c The k d is the mean Ϯ S.E. of 13 values, obtained at different concentrations of HGF/SF. No evidence was found for a two-site model of dissociation, and so the HGF/SF-binding sites in each length of oligosaccharide were homogeneous in this respect. d The K d value was calculated from the ratio of k d /k a , and the S.E. is the combined S.E. of the two kinetic parameters. e The K d value was calculated from the extent of binding at or near equilibrium, and the S.E. is the combined S.E. of four independent determinations of K d .
Endocan promotes HGF/SF-induced proliferation through its DS chain. The fact that endocan is specifically secreted by endothelial cells suggests that it may represent a novel factor able to regulate the functions of HGF/SF, and perhaps other GAG-binding regulatory proteins, in areas of embryonic development, tissue regeneration, or tumor progression and that it may be particularly important in the context of emerging therapeutic applications of HGF/SF.