Modulation of the Heparanase-inhibiting Activity of Heparin through Selective Desulfation, Graded N-Acetylation, and Glycol Splitting*

Heparanase is an endo-β-glucuronidase that cleaves heparan sulfate (HS) chains of heparan sulfate proteoglycans on cell surfaces and in the extracellular matrix (ECM). Heparanase, overexpressed by most cancer cells, facilitates extravasation of blood-borne tumor cells and causes release of growth factors sequestered by HS chains, thus accelerating tumor growth and metastasis. Inhibition of heparanase with HS mimics is a promising target for a novel strategy in cancer therapy. In this study, in vitro inhibition of recombinant heparanase was determined for heparin derivatives differing in degrees of 2-O- and 6-O-sulfation, N-acetylation, and glycol splitting of nonsulfated uronic acid residues. The contemporaneous presence of sulfate groups at O-2 of IdoA and at O-6 of GlcN was found to be non-essential for effective inhibition of heparanase activity provided that one of the two positions retains a high degree of sulfation. N-Desulfation/ N-acetylation involved a marked decrease in the inhibitory activity for degrees of N-acetylation higher than 50%, suggesting that at least one NSO3 group per disaccharide unit is involved in interaction with the enzyme. On the other hand, glycol splitting of preexisting or of both preexisting and chemically generated nonsulfated uronic acids dramatically increased the heparanase-inhibiting activity irrespective of the degree of N-acetylation. Indeed N-acetylated heparins in their glycol-split forms inhibited heparanase as effectively as the corresponding N-sulfated derivatives. Whereas heparin and N-acetylheparins containing unmodified d-glucuronic acid residues inhibited heparanase by acting, at least in part, as substrates, their glycol-split derivatives were no more susceptible to cleavage by heparanase. Glycol-split N-acetylheparins did not release basic fibroblast growth factor from ECM and failed to stimulate its mitogenic activity. The combination of high inhibition of heparanase and low release/potentiation of ECM-bound growth factor indicates that N-acetylated, glycol-split heparins are potential antiangiogenic and antimetastatic agents that are more effective than their counterparts with unmodified backbones.

Heparanase is a mammalian endo-␤-D-glucuronidase that cleaves heparan sulfate (HS) 1 chains at a limited number of sites (1)(2)(3). Cloning of the heparanase cDNA by several groups (1)(2)(3)(4)(5)(6) suggests that a single functional HS-degrading endoglycosidase is expressed in mammalian cells. The enzyme is synthesized as a latent 65-kDa precursor that undergoes proteolytic cleavage, yielding 8-and 50-kDa subunits that heterodimerize to form a highly active enzyme (7,8). Heparanase enzymatic activity participates in degradation and remodeling of the extracellular matrix (ECM), facilitating, among other activities, cell invasion associated with cancer metastasis, angiogenesis, and inflammation (1)(2)(3)9). Heparanase upregulation has been documented in a variety of human tumors correlating, in some cases, with increased vascular density and poor postoperative survival (10 -13). Heparanase overexpression has also been noted in several other pathologies such as cirrhosis (14), nephrosis (15), and diabetes (16). In addition to its intimate involvement in the egress of cells from the blood stream, heparanase activity releases from the ECM and tumor microenvironment a multitude of HS-bound growth factors, cytokines, chemokines, and enzymes that affect cell and tissue function, most notably angiogenesis (17,18). These observations, the anticancerous effect of heparanase gene silencing (ribozyme and small interfering RNA) (19) and of heparanaseinhibiting molecules (non-anticoagulant species of heparin and other sulfated polysaccharides) (20,21), and the unexpected identification of a predominant functional heparanase (1)(2)(3) suggest that the enzyme is a promising target for development of new anticancer drugs.
HS and the structurally related heparin are present in most animal species. They are glycosaminoglycans constituted by repeating disaccharide units of a uronic acid (either D-glucuronic acid (GlcA) or L-iduronic acid (IdoA)) and D-glucosamine (either GlcNAc or D-glucosamine N-sulfate (GlcNSO 3 )) and bear sulfate substituents in various positions (22)(23)(24)(25). Although derived from the common biosynthetic precursor Nacetylheparosan (-GlcA-GlcNAc) n , HS and heparin have differ-ent structures: HS is less sulfated and more heterogeneous than heparin. The two glycosaminoglycans also have different locations in tissues: whereas HS is a component of the ECM and of the surface of most cells, heparin is stored in granules of mast cells and co-released with histamine into the circulation upon cellular degranulation mainly in cases of allergic and inflammatory reactions and anaphylactic stress. On the other hand, exogenous heparin is widely used as an anticoagulant and antithrombotic drug and is of increasing interest for novel therapeutical applications (24 -27).
As an analog of the natural substrate of heparanase, heparin is commonly considered to be a potent inhibitor of heparanase (20, 21, 28 -31). This activity is attributed, in part, to its high affinity interaction with the enzyme and limited degradation, serving as an alternative substrate. Early reports (20,21,30,31) showed that heparin and some chemically modified species of heparin as well as other sulfated polysaccharides (22,32) that inhibit tumor cell heparanase also inhibit experimental metastasis in animal models, while other related compounds that lack heparanase-inhibiting activity fail to exert an antimetastatic effect (20 -22, 30 -32). Regardless of the mode of action, heparin and low molecular weight heparin (LMWH) were reported to exert a beneficial effect in cancer patients (33), stimulating research on the potential use of modified, nonanticoagulant species of heparin and HS in cancer therapy.
Screening of heparin derivatives has permitted the identification of some of the structural features of heparin associated with inhibition of the enzyme. As a general trend, the heparanase-inhibiting activity increases with increasing degrees of O-sulfation. However, N-sulfates seems to exert little effect since they can be replaced by N-acyl (N-acetyl, N-succinyl, or N-hexanoyl) groups without substantial loss of inhibitory activity (20,34). No significant differences were found between the currently used unfractionated heparins and low molecular weight heparins and a tetradecasaccharidic fragment (34). 2-O-Desulfated derivatives were shown to retain the inhibitory activity, whereas N-desulfated, N-acetylated derivatives displayed a reduced activity (35). In the present study, relationships between structure and heparanase-inhibiting activity of heparin were studied using a larger number of heparins and heparin derivatives, including some with various degrees of 6-O-sulfation of GlcN and 2-O-sulfation of IdoA residues as well as "glycol-split" derivatives obtained by controlled periodate oxidation/borohydride reduction of natural (36) or partially 2-O-desulfated heparins (37,38). Glycol splitting of C-2-C-3 bonds of nonsulfated uronic acid residues was suggested to interfere with the biological interactions of heparin by providing flexible joints between protein binding sequences (37)(38)(39). When framing heparin sequences that bind FGF-2, glycol-split residues were shown not to impair the binding to FGF-2. However, they prevented activation of FGF-2 and FGF-2-induced angiogenic activity (37,38). The present study shows that glycol splitting enhances the heparanase-inhibiting activity of heparin. Based on the observation that N-acetyl groups do not prevent and may even assist recognition by heparanase (40, 41) and taking into account that N-acetylheparin, as opposed to heparin, does not release angiogenic factors from ECM (34), we prepared and tested heparins with various degrees of N-acetylation/N-sulfation together with some of their glycol-split derivatives. N-Acetylated, glycol-split heparins were shown to inhibit heparanase more efficiently than the corresponding non-glycol-split N-acetylated heparins.
NMR spectra were recorded at 500 MHz for 1 H and 125 MHz for 13 C with a Bruker AMX spectrometer equipped with a 5-mm 1 H/X inverse probe. The spectra were obtained at 45°C from D 2 O solutions (15 mg/0.5 ml D 2 O, 99.99% D). Chemical shifts, given in parts per million down field from sodium-3-(trimethylsilyl)propionate, were measured indirectly with reference to acetone in D 2 O (␦ 2.235 for 1 H and ␦ 30.20 for 13 C). The 13 C NMR spectra were recorded at 300 or 400 MHz with a Bruker AC-300 or AMX-400 spectrometer.

Recombinant Human Heparanase
Recombinant enzymatically active heparanase was purified from heparanase-transfected Chinese hamster ovary cells (4). Briefly Chinese hamster ovary cells were harvested with trypsin and centrifuged, and the cell pellet was suspended in 20 mM citrate-phosphate buffer pH 5.4. The suspension was subjected to four cycles of freeze/thaw (Ϫ70/ 37°C, 5 min each), the cell extract was centrifuged (18,000 rpm, 15 min, 2-8°C), and the supernatant was collected and filtered through a 0.45-m filter. The filtrate was applied onto a Source 15 S column (Amersham Biosciences) equilibrated with 20 mM phosphate buffer, pH 6.8. The column was washed (20 mM phosphate buffer, pH 6.8, followed by 20 mM phosphate buffer, pH 8.0), and heparanase was eluted with a linear gradient (0 -35%) of 8 column volumes of 1.5 M NaCl in 20 mM phosphate buffer, pH 8.0. Active fractions were pooled and applied onto a Fractogel EMD SO 3 Ϫ (Merck) column equilibrated with 20 mM citratephosphate buffer, pH 5.4. Heparanase was eluted with a linear gradient (0 -22%) of 1 column volume followed by 10 column volumes (22-25%) of 1.5 M NaCl in 20 mM phosphate buffer, pH 8.0. Finally heparanase eluted from the Fractogel column was applied onto a HiTrap heparin column (Amersham Biosciences) equilibrated with 20 mM phosphate buffer, pH 8.0, and eluted with a linear gradient of 1 column volume (0 -20%) and 15 column volumes (20 -28%) of 1.5 M NaCl in 20 mM phosphate buffer, pH 8.0. Eluted fractions were analyzed by gradient SDS-PAGE, stained with Gelcode ® (Pierce), and pooled according to their purity. An at least 90% pure, highly active heparanase preparation was obtained, containing the active 50-and 8-kDa heparanase subunits and, to a lower extent, the 65-kDa proheparanase (8). Active recombinant human heparanase was also produced in insect cells as described previously (7). The construct encoding the 8-and 50-kDa heparanase subunits was kindly provided by Dr. E. McKenzie (Oxford GlycoSciences Ltd., Abingdon, Oxon, UK) (7). Similar results were obtained with both preparations.

6-O-Desulfated Heparins
Procedure A-An extensively 6-O-desulfated heparin also partially (ϳ15%) 2-O-desulfated ( 71 6OdeS-H(A) where the superscript denotes the degree of 6-O-desulfation), M w 16,000, was prepared according to Nagasawa et al. (44), starting from the pyridinium salt of heparin H-1, under solvolytic conditions (10 mg/ml in Me 2 SO:water 9:1) at 100°C for 2.5 h followed by resulfation of free amino groups with sulfur trioxidetrimethylamine complex in alkaline aqueous medium (45). was converted into its pyridinium salt and soaked in pyridine (20 ml). After addition of 4 ml of N-methyl-N-(trimethylsilyl)trifluoroacetamide, the solution was heated for 4 h at 80°C to yield 73 6OdeS-H or for 8 h at 60°C to yield 77 6OdeS-H. Heparin (H-1) was converted into its pyridinium salt and soaked in pyridine (30 ml). After addition of 6 ml of N,O-bis(trimethylsilyl)acetamide, the solution was heated for 2 h at 60°C to yield 46 6OdeS-H.

2-O-Desulfated Heparins
Procedure A-2-O-Desulfated heparin in the IdoA form (H, IdoA(A), M w 17,700) was prepared according to Jaseja et al. (47). Heparin (500 mg) was simply dissolved in 500 ml of 0.1 M NaOH, and the solution was frozen and lyophilized. The residue dissolved in 500 ml of distilled water was dialyzed, and the product was isolated by evaporation under reduced pressure. Its 13 C NMR spectrum closely corresponded to the one reported in the literature (48), indicating an essentially complete conversion of the original IdoA2SO 3 residues into IdoA residues.
Procedure B-2-O-Desulfated heparin in the GalA form (H, GalA(B), M w 12,600) was prepared by a modification of methods used by Jaseja et al. (47) and Rej and Perlin (49) essentially as described previously (48). Heparin (500 mg) was dissolved in 10 ml of 1 M NaOH and then heated at 85°C for 1 h. After cooling below 30°C, the solution was brought to pH 7 with 0.1 M HCl and heated at 70°C for 48 h to give (after cooling, dialysis, and freeze-drying) the GalA derivative with a typical 13 C NMR spectrum (48).

N-Acetylated Heparins
N-Acetylated heparins ( x NAH, where the superscript x denotes the degree of N-acetylation as referred to total GlcN) were prepared by time-controlled N-desulfation under solvolytic conditions (44). Briefly the pyridinium salt of heparin was stirred at 20 -25°C in Me 2 SO:water (9:1) for different times (30,60,90, 100, and 120 min and 8 h) to obtain intermediates with different degrees of N-desulfation, which upon Nacetylation with acetic anhydride in alkaline aqueous medium (NaHCO 3 , 4°C, 2 h) (50) gave 29

Glycol-split Heparins and Glycol-split N-Acetylated Heparins
Glycol-split heparins and glycol-split N-acetylated heparins were prepared by exhaustive periodate oxidation and borohydride reduction of heparin and N-acetylheparins, respectively, without (36) or with (37,38) prior partial 2-O-desulfation. For the first series of glycol-split N-acetylheparins, 250-mg samples of H-1, 29 NAH, 39 NAH, 50 39 NAH, 58 NAH, and 70 NAH were dissolved in 5 ml of 1 M NaOH and then heated at 60°C for 30 min. After cooling below 30°C, the solutions were brought to pH 7 with 0.1 M HCl and heated at 70°C for 48 h to give (after cooling, dialysis, and freeze-drying) partial conversion of IdoA2SO 3 to GalA. Products were treated as described above to yield the corresponding glycol-split derivatives H, 52 gs (M w 11,000), 29 NAH, 60 gs (M w 6,000), 43 NAH, 60 gs (M w 8,500), 57 NAH, 64 gs (M w 9,500), and 70 NAH, 59 gs (M w 9,300). The glycol-splitting (gs) percentages were evaluated by integration of the anomeric 13 C NMR signals at 106.5 ppm (A) and at 102 ppm (B), corresponding to the split uronic acid residues and 2-O-sulfated iduronic acid residues, respectively; gs ϭ (A/(A ϩ B)) ϫ 100. Products obtained without generation of additional nonsulfated uronic acid residues had a content of glycol-split residues (mainly arising from GlcA) of 24 Ϯ 1% and are designated as "reduced oxyheparins" (RO-H) (36). Products obtained by glycol splitting of both the preexisting and the newly generated nonsulfated uronic acids (IdoA or GalA) were designated as H, x gs (or NAH, x gs if derived from N-acetylheparins) where the superscript x indicates the percentage of glycolsplit uronic acid.

Low Molecular Weight Derivatives
Low molecular weight derivatives of H-1, H, 44 gs, and 50 NAH, 25 gs ( 50 NA,RO-H) were prepared by nitrous acid depolymerization of the corresponding polysaccharides (51). A solution of polysaccharide (4 g) was dissolved in 65 ml of H 2 O and cooled at 4°C. After the addition of 75 mg of NaNO 2 the pH was adjusted to 2 with 0.1 M HCl. The solution was stirred at 4°C for 20 min, and then the pH was brought to 7. Solid NaBH 4 (1 g) was added in several portions under stirring. After 2-3 h, the pH was adjusted to 4 with 0.1 M HCl, and the solution was neutralized with 0.1 M NaOH. The products (low molecular weight H-1, 6,500; LMWH, 49 gs, 6,300; LMWH, 49 gs, 3,000; and 50 NA,RO-H, 5,400) obtained by precipitation with 3 volumes of ethanol were dissolved in water and recovered by freeze-drying. The depolymerization degrees and the corresponding molecular weight values were determined by integration of the 13 C NMR signals at 98 -107 and 82, 85, and 87 ppm, corresponding to total C-1 and C-2, C-3, and C-5 of the anhydromannitol unit, respectively. The percentage of glycol splitting, expressed as glycol-split residues referred to total uronic acids, was evaluated by integration of the 13 C NMR signals at 106.5 and 102 ppm, corresponding to C-1 of the split uronic residues and 2-O-sulfated iduronic residues, respectively.

Preparation of Dishes Coated with ECM
Bovine corneal endothelial cells were plated into 35-mm tissue culture dishes at an initial density of 2 ϫ 10 5 cells/ml and cultured as described above except that 4% dextran T-40 was included in the growth medium (4, 52). On day 12, the subendothelial ECM was exposed by dissolving the cell layer with PBS containing 0.5% Triton X-100 and 20 mM NH 4 OH followed by four washes with PBS (52). The ECM remained intact, free of cellular debris, and firmly attached to the entire area of the tissue culture dish. To produce sulfate-labeled ECM, Na 2 35 SO 4 (Amersham Biosciences) was added (25 Ci/ml) on days 2 and 5 after seeding, and the cultures were incubated with the label without a medium change and processed as described previously (4,52). Nearly 80% of the ECM radioactivity was incorporated into HSPG.

Heparanase Inhibition Activity
Heparin species were tested for their ability to inhibit heparanase using metabolically sulfate-labeled ECM as a substrate (28,29). Briefly sulfate-labeled ECM coating the surface of 35-mm culture dishes was incubated (4 h, 37°C, pH 6.0) with recombinant human heparanase (40 ng/ml) in the absence and presence of three concentrations (0.2, 1.0, 5.0 g/ml) of each heparin species. The reaction mixture contained 50 mM NaCl, 1 mM dithiothreitol, 1 mM CaCl 2 , and 10 mM phosphate-citrate buffer, pH 6.0. To evaluate the occurrence of proteoglycan degradation, the incubation medium was collected and applied for gel filtration on Sepharose 6B columns (0.9 ϫ 30 cm). Fractions (0.2 ml) were eluted with PBS at a flow rate of 5 ml/h and counted for radioactivity. The excluded volume (V o ) was marked by blue dextran, and the total included volume (V t ) was marked by phenol red. Nearly intact HSPGs are eluted from Sepharose 6B just after the void volume (K av Ͻ 0.2, fractions 1-10), while HS degradation fragments are eluted toward the V t of the column (peak II, 0.5 Ͻ K av Ͻ 0.8, fractions 15-35) (4, 28, 29). Labeled fragments eluted in peak II were shown to be degradation products of HS as they were 5-6-fold smaller than intact HS chains of HSPGs, resistant to further digestion with papain and chondroitinase ABC, and susceptible to deamination by nitrous acid (29). Heparanase activity ϭ K av ϫ total cpm in peak II. Recovery of labeled material applied on the column ranged from 85 to 95% in different experiments. Each experiment was performed at least three times, and the variation in elution positions (K av values) did not exceed Ϯ15%.

Release of ECM-bound FGF-2
ECM-coated wells (4-well plates) were incubated with iodinated FGF-2 (1-2 ϫ 10 5 cpm/ng, 1.5-2.5 ϫ 10 4 cpm/0.25 ml/well, 3 h, 24°C), and the unbound FGF-2 was removed by four washes with PBS containing 0.02% gelatin (34). The ECM was then incubated (3 h, 24°C) with the various heparins and modified heparins, and aliquots (0.25 ml) of the incubation medium were counted in a ␥-counter to determine the amount of released material. The remaining ECM was washed twice with PBS and solubilized with 1 N NaOH, and the radioactivity was counted in a ␥-counter (34). The percentage of released 125 I-FGF-2 was calculated from the total ECM-associated radioactivity. "Spontaneous" release of 125 I-FGF-2 in the presence of incubation medium alone was 7-12% of the total ECM-bound FGF-2 (34). Each experiment was performed three to five times, yielding similar results.

Stimulation of FGF-2 Mitogenic Activity
A cytokine-dependent, heparan sulfate-deficient lymphoid cell line (BaF3) engineered to express fibroblast growth factor receptor-1 (53,54) was applied to investigate the effect of heparin derivatives on FGF-2-mediated cell proliferation. These cells (clone F32) respond to FGF-2 only in the presence of exogenously added heparin, HS, or some modified species of heparin. Briefly F32 cells (2 ϫ 10 4 /well) were plated into 96-well microtiter plates in the presence of 2.5 or 5.0 ng/ml FGF-2 and increasing concentrations of the test compound in a total volume of 250 l. Forty-eight hours later, 1 Ci of [ 3 H]thymidine was added per well, and the cells were incubated for an additional 6 h and collected with a cell harvester. Incorporated thymidine was determined by liquid scintillation counting using a TopCount microplate counter (53,54).

Gel Permeation Analysis of NAH and ϳ50% Glycol-split Heparin
(H, 52 gs) before and after Digestion with Heparanase 2 mg of each compound were incubated for 48 h at 37°C in 40 mM phosphate-citrate buffer, pH 5.8, with or without 4 g of recombinant heparanase in a total volume of 50 l. The samples were lyophilized, then redissolved in 0.5 ml of water, and analyzed by GPC-HPLC using 300 ϫ 7.8-mm TSK PW 2000 and PW 3000 columns.

Preparation (and Schematic Presentation) of Chemically
Modified Species of Heparin-The relationship between sulfation patterns and heparanase-inhibiting activity of heparin species with unmodified backbones was studied using O-desulfated heparin derivatives and N-desulfated, N-acetylated heparins of various degrees of substitution prepared starting from a well characterized pig mucosal heparin (H-1) using established procedures with slight modifications. 6-O-Desulfation was accomplished using two different procedures, the first (procedure A) involving solvolytic desulfation (44) and the second (procedure B) through activated silyl acetamides (46). As reported (46), attempts to obtain extensively 6-O-desulfated heparins using procedure A also involved partial 2-O-desulfation (10 -15%, by NMR analysis). Therefore, the extensively 6-O-desulfated heparin H, 77 6OdeS was obtained only via procedure B. 2-O-Desulfation of heparin was also performed using two different procedures, leading to different products. The first (procedure A), involving lyophilization of alkaline solutions, quantitatively removes the 2-OSO 3 groups retaining the IdoA residues in their original configuration (47). The second one (procedure B), which involves heating of alkaline solutions, is more easily controlled to generate also partially 2-O-desulfated heparins (37) and to convert the 2-O-sulfated IdoA residues into GalA residues (47)(48)(49). Glycol-split derivatives were prepared by periodate oxidation/borohydride reduction of both unmodified heparin and partially 2-O-desulfated heparins as described previously (37). The same procedure was applied to obtain glycol-split derivatives of N-acetylheparins of various degrees of N-acetylation and 2-O-desulfation. Details are given under "Experimental Procedures." Fig. 1 shows the general scheme for the preparation of derivatives of heparin, N-acetylheparins, and the corresponding glycol-split derivatives. Scheme 1 shows the formulas of heparin and heparin derivatives. The structure of a heparin chain is schematized as composed of N-acetylated (NA), N-sulfated (NS), and mixed NA/NS domains; the prevalent sequences are those of the trisulfated disaccharide (formula 1) (24). For simplicity, only chains containing the antithrombin binding region are presented, although this region is contained in only about one-third of the chains of heparin. The prevalent structure of 6-O-desulfated heparins is shown in formula 2, and that of 2-O-desulfated heparins retaining the IdoA configuration is shown in formula 3. Formula 4 shows the prevalent structure of heparins that underwent inversion of configuration to GalA during 2-O-desulfation of IdoA. Fully N-acetylated heparins are represented by the general formula 5. Partially N-acetylated heparins have different percentages of N-acetyl groups; the NSO 3 groups are the complement to 100% N-substitution. Glycol splitting is depicted in Fig. 1 in two different ways. Fig.  1A refers to heparin and fully N-acetylated heparin (NAH) glycol-split without any previous modification of their structures. The corresponding glycol-split products are referred to as RO-H (36) and "N-acetyl-reduced oxyheparins" (NA,RO-H), respectively. Fig. 1B refers to heparin and NAH (in the example, 50% N-acetylated heparin 50 NAH) that were glycol-split after removal of 2-OSO 3 groups to reach a total content of 50% nonsulfated uronic acid residues; the nonmodified residues were IdoA2SO 3 (37). De-O-sulfation of one of two IdoA residues followed by glycol splitting gave heparins with prevalent structure 6, corresponding to sequences of polypentasulfated trisaccharides separated by glycol-split uronic acid residues (37). Low molecular weight species of heparin and representative heparin derivatives were obtained by controlled nitrous acid depolymerization (51) of heparin, 50% glycol-split heparin, and the RO derivative of 50% N-acetylated heparin.
All compounds were analyzed by one-and two-dimensional 1 H and 13 C NMR spectroscopy (37). Analytical data, expressed as relative molar content of 6-OSO 3 , 2-OSO 3 , and NSO 3 groups are summarized in Table I Heparanase Inhibition by Heparin Derivatives-Typical heparanase inhibition curves showing the gel filtration profiles of sulfate-labeled degradation fragments released by heparanase from metabolically labeled ECM in the absence (control) and presence of 1 g/ml unmodified heparin and fully N-acetylated heparin are presented in Fig. 2. Inhibition is reflected by the decreased amounts and K av values of HS fragments released from ECM and eluted as peak II (fractions 20 -35) in comparison with control incubation of the ECM with recombinant heparanase in the absence of inhibitors (28,29). Heparanase activity is calculated as the total amount of cpm eluted in peak II multiplied by the K av (i.e. elution position) of these fragments. The heparanase-inhibitory activity (expressed as percent inhibition of heparanase) of almost all heparins at concentrations of 5, 1, and 0.2 g/ml is shown in Table I. Most of the data represent the average of several separate experiments (numbers indicated). Standard deviations, indicated for each heparin, were usually lower than 5 for the most active compounds and did not exceed 20 as a mean for the less effective ones.
Data in Table I confirm that heparin is a strong inhibitor of heparanase (ϳ70% inhibition at 1 g/ml). No significant differences in inhibitory activity were found among H-1 and other heparin preparations from pig mucosa, beef mucosa, and beef lung (data not shown) despite significant differences in their sulfation patterns (detailed under "Experimental Procedures"). Also activity differences found between the parent heparin and its low molecular weight species as well as between glycol-split 50 NAH and its low molecular weight species were not significant. On the other hand, well defined significant differences in heparanase-inhibiting activity were associated with specific chemical modifications of heparin. As illustrated in Fig. 3, whereas either 6-O-desulfation or 2-O-desulfation with reten-tion of IdoA configuration had little or no effect on the heparanase-inhibitory activity of heparin, 2-O-desulfation with change of configuration of the IdoA residues to GalA markedly decreased the inhibitory activity of heparin.
Also complete removal of N-sulfate groups followed by Nacetylation resulted in a substantial decrease of the inhibitory activity (Fig. 3). However, as illustrated in Fig. 4, this effect was only noted for N-acetylation degrees higher than ϳ50%. On the other hand, glycol splitting markedly increased the heparanase-inhibiting activity of both heparins and N-acetylated heparins and restored the inhibitory effect lost upon N-acetylation of heparin ( Fig. 4 and Table I). This effect is illustrated in Fig. 4 and Table I for N-acetylated heparins of the RO type (i.e. 25% glycol-split), which almost completely inhibited the heparanase activity (to less than 10% of the control at 1 g/ml and to 20 -30% at 0.2 g/ml) irrespective of their degree of N-acetylation. Glycol splitting extended to newly generated nonsulfated IdoA/GalA residues in heparin and N-acetylated heparins gave products showing high heparanase-inhibitory activity. The dose dependence of the heparanase-inhibitory activity is illustrated in Fig. 5 for heparin, fully N-acetylated heparin, and its RO derivative. IC 50 values calculated from the corresponding curves are Ͼ5 g/ml for NAH, ϳ0.4 g/ml for H-1, and ϳ0.2 g/ml for 100 NA,RO-H.
Gel permeation chromatographic analysis of some products of heparanase digestion, performed under conditions of the  (24). For simplicity, only chains containing the antithrombin binding region are represented, and the GlcA residue of this region is shown as the only nonsulfated uronic acid residue in the NS region. B, schematic representation of partial 2-O-desulfation and glycol splitting of both the preexisting and the newly generated nonsulfated uronic acid residues of heparin (H) and 50% N-acetylated heparin ( 50 NA-H). The represented schematic structure of glycol-split heparin corresponds to derivative H, 50 gs with splitting of about 50% of the total uronic acid residues, prevalently represented by repeating sequences of polypentasulfated trisaccharide separated by split uronic acid and actually consisting of about 25% preexisting and 25% newly generated uronic acid residues (38). The example for glycol-split N-acetylated heparin corresponds to the model derivative 50 NAH, 50 gs. enzyme inhibition assay, indicated that whereas heparin (not shown) and N-acetylheparin are cleaved by heparanase (as previously shown for heparin) (40), their glycol-split derivatives are not susceptible to cleavage as illustrated for fully N-acetylated, RO-heparin in Fig. 6A and 52% glycol-split heparin (H, 52 gs) in Fig. 6B.

Effect of Modified Heparins on Release of ECM-bound FGF-2 and Stimulation of FGF-2 Mitogenic
Activity-Some of the heparin derivatives were tested for their capacity to release FGF-2 from ECM (18,34). As demonstrated in Fig. 7, doseresponse curves of the FGF-2-releasing activity of glycol-split heparin (H, 52 gs) and its corresponding low molecular weight derivative (LMWH, 49 gs) were almost superimposable to those reported for heparin (34), indicating that glycol splitting does not substantially modify the FGF-2-releasing properties of hep-arin. Also the curves of the RO derivative and of heparin are superimposable (data not shown). Fig. 7 also shows that glycolsplit, N-acetylated heparins behave similarly to non-glycolsplit NAH (34) in that they release ECM-bound FGF-2 consistently less than unmodified heparin. 100 NAH (not shown) and 100 NA,RO-H exhibited the lowest FGF-2-releasing activity among the tested compounds, yielding only about twice the spontaneous release observed in presence of the buffer (PBS) alone.
The ability of heparin, 100 NAH, and 100 NA,RO-H to promote the mitogenic activity of recombinant FGF-2 was investigated using cytokine-dependent, heparan sulfate-deficient lymphoid cells (BaF3) engineered to express fibroblast growth factor receptor-1 (53,54). Unlike heparin, both fully N-acetylated heparin ( 100 NAH) and its glycol-split counterpart molecule SCHEME 1. Formulas of heparin and heparin derivatives.
( 100 NA,RO-H) failed to stimulate the mitogenic activity of FGF-2 beyond the basal level obtained in the absence of added heparin (Fig. 8). Thus, while glycol splitting of NAH fully restored its heparanase-inhibiting activity, it failed to induce a similar restoration of the ability to displace ECM-bound FGF-2 and to stimulate the mitogenic activity of recombinant FGF-2.

DISCUSSION
The HS chains of HSPG in the ECM and on the surface of endothelial cells are the natural substrates for heparanase. HSPGs, expressed by virtually all cells, are thought to play key roles in numerous biological settings including embryogenesis, cytoskeleton organization, cell migration, wound healing, inflammation, cancer metastasis, and angiogenesis (17,26). These multiple functions, exerted via distinct mechanisms, are modulated by heparanase through endoglycosidic cleavage of HS (1-3). The site of cleavage is the ␤-glycosidic linkage of a GlcA residue, which must be flanked by N-sulfated or N-acetylated ␣-linked GlcN residues. At least one OSO 3 group is essential for efficient recognition by the enzyme (40, 41). The three-dimensional structure of heparanase is not yet known in detail. Translation of the primary structure of heparanase over another endo-␤-glycosidase (␤-xylanase) shows clusters of basic amino acid residues at least one of which is conceivably implicated in binding to sulfate groups of the substrate (55). Our preliminary studies, applying point mutations and deletions as well as synthetic peptides, identified amino acid residues 158 -171 as the predominant HS binding domain of the heparanase molecule. 2 HS/heparin and derived oligosaccharides must have a minimal octasaccharidic size to be good substrates for heparanase (40,41). However, the enzyme also can be efficiently inhibited by shorter but more extensively sulfated oligosaccharides such as maltohexaose polysulfate and phosphomannopentaose polysulfate (PI-88) (22). Heparin, although less sulfated than these persulfated oligosaccharides, is a good inhibitor of heparanase activity (28) and is active in experimental metastasis models as well (20 -22). The inhibitory activity of heparin is lost upon extensive O-desulfation and/or a decrease in chain length below a tetradecasaccharidic size (20,34).
Previous reports on the effects of selective O-desulfation on the heparanase-inhibiting activity of heparin (35) are essentially confirmed by the present study. In fact, removal of either the 6-O-sulfate group on the glucosamine residue or the 2-Osulfate group on the iduronic acid residue only slightly reduced the inhibitory activity of heparin (Fig. 3).
The decrease in inhibitory activity observed in the present study for extensively 6-O-desulfated heparin prepared by the solvolytic method appears to be associated with a concomitant partial 2-O-desulfation. On the other hand, the consistently lower heparanase-inhibiting activity of partially 2-O-desulfated derivative obtained with procedure B (involving a change of configuration from IdoA to GalA) as compared with that of the 2-O-desulfated derivative obtained by procedure A (with  retention of the IdoA configuration) is likely associated with different conformational properties of IdoA and GalA. IdoA residues have been demonstrated to be endowed with a unique conformational flexibility ("plasticity"). Such a plasticity, associated with different equienergetic conformations of IdoA residues, all coexisting in a rapid dynamic equilibrium, can currently explain the better protein binding capacity and associated biological properties of IdoA-containing sequences as compared with the more rigid GlcA-containing ones (56). Based on simple conformational criteria, GalA is expected to have very much the same conformational rigidity as GlcA.
Replacement of N-sulfate groups with N-acetyl groups does not completely suppress the heparanase-inhibitory activity of heparin (34); its activity is reduced to about one-third (35) as confirmed by the present study. However, it is noteworthy that a substantial decrease in heparanase-inhibiting activity was only observed for degrees of N-acetylation higher than about 50%. As shown in Table I and illustrated in Fig. 4, for degrees of N-acetylation lower than 40%, the inhibitory activity remained essentially the same as that of heparin. This is taken as an indication that only one-half of the NSO 3 groups of heparin are essential for complete inhibition of heparanase. The accepted conformation of the NS domains of heparin is represented by helices where sets of three sulfate groups (NSO 3 , 2SO 3 , and 6SO 3 ) alternate on each side of the chain (57). The observation that only one of two N-sulfate groups is required to inhibit heparanase and the assumption that the N-desulfation/ N-acetylation reaction occurs randomly along the NS domains would accordingly suggest that heparin and its derivatives with unmodified backbones bind the enzyme only from one side of the chain. . Sulfate-labeled degradation fragments released into the incubation medium were analyzed by gel filtration on Sepharose 6B. K av of peak II (see Fig. 2), calculated for each compound, was multiplied by the total cpm eluted in peak II. Results are presented as percentage of control. Residual heparanase activity ϭ K av ϫ total cpm in peak II (percentage of control).
FIG. 4. Inhibition of heparanase by N-acetylheparins and the corresponding 25% glycol-split (RO) derivatives. Sulfate-labeled ECM was incubated (4 h, 37°C, pH 6.0) with heparanase (40 ng/ml) in the presence of 1 g/ml NAH with an increased percentage of Nacetylation (% NAc) or with the corresponding 25% glycol-split derivatives (NA,RO-H). Sulfate-labeled material released into the incubation medium was analyzed by gel filtration, and heparanase enzymatic activity (K av ϫ total cpm in peak II) is presented as percentage of the 100% activity obtained in the absence of inhibitor.
The effect of glycol-splitting of heparin on heparanase inhibition is largely new. Lapierre et al. (35) reported that periodate oxidation/borohydride reduction of nonsulfated uronic acids of heparin, leading to a product corresponding to H, 25 gs of the present study, did not impair the inhibitory activity of heparin, a finding taken by these authors as an indication that nonsulfated IdoA is not essential for the activity. The present results on heparins subjected to glycol splitting only at the level of preexisting nonsulfated uronic acids (RO-H) (36) and of both the preexisting and the newly generated ones (such as H, 52 gs) (37,38) indicate that in fact glycol splitting resulted in a marked general increase in the inhibition of the heparanase activity by heparin species. A reasonable explanation of such an effect is that formation of glycol-split residues, involving elimination of conformational constraints with formation of three additional degrees of rotational freedom per each split residue, generates flexible joints that separate from each other heparin sequences containing IdoA2SO 3 residues, thus facilitating the docking of these sequences to sites essential for heparanase activity. The proposed "extra flexibility" (39) FIG. 6. Glycol splitting inhibits cleavage by heparanase. Shown are gel filtration profiles of fully N-acetylated heparin (A) and 52% glycol split heparin (B) before (a) and after (b) incubation with heparanase. 2 mg of each compound were incubated for 48 h at 37°C in 40 mM phosphate-citrate buffer, pH 5.8, with or without 4 g of recombinant heparanase in a total volume of 50 l. The samples were lyophilized, then redissolved in 0.5 ml of water, and analyzed by GPC-HPLC using 300 ϫ 7.8-mm TSK PW 2000 and PW 3000 columns and a refraction index detector. The sharp peak is from salts. induced by these joints reinforces the binding driving influence of the already existing intrinsic conformational plasticity of iduronate residues. More notably, glycol splitting also increased the heparanase-inhibiting activity of N-acetylated heparins, even that of fully N-acetylated heparin, whose heparanase binding capacity is very weak when their backbone is unmodified.
Glycol splitting involves substantial loss of the anticoagulant activity of heparin (36). It is now clear that the main reason for such an effect is the cleavage of C-2-C-3 bonds of the GlcA residue of the pentasaccharidic sequence (Fig. 1, formula 7). This residue in its unmodified form is essential for binding to antithrombin, and whenever it is glycol-split (as in sequence 8), the heparin affinity for antithrombin is completely lost (Ref. 24 and references therein).
The heparanase inhibiting properties of glycol-split, extensively N-acetylated heparins cannot be explained by the model discussed previously where only sulfate groups (perhaps in addition to the uronate carboxyl groups) on the same side of the heparin helix are involved in binding to the enzyme. Such a model is compatible with heparanase binding of heparins that contain no more than 50% N-acetyl groups. Retention of strong inhibitory activity upon removal all NSO 3 groups and their substitution with nonpolar N-acetyl groups followed by glycol splitting of nonsulfated uronic acids implies that efficient docking to heparanase, favored by the flexible joints generated by glycol splitting, occurs with a conformation different from that adopted by heparin with a degree of N-acetylation lower than 50%.
Conformational polymorphism is not uncommon in heparin sequences bound to proteins. Thus, x-ray studies showed that heparin sequences may bind FGF-1 in more than one, equally favored conformation (58). Increasing evidence is being accumulated that the plasticity of iduronate residues combined with some rotational freedom of both uronic acid and amino sugar residues around the glycosidic linkages favors several possibilities of binding to basic domains of proteins (59). As expected (37,39), glycol splitting appears to further increase the molecular flexibility of heparin chains as determined by small angle x-ray scattering (60) and NMR spectroscopy supported by molecular modeling studies (61).
It appears that heparin contains both recognition/cleavage and inhibition sites for heparanase and that its inhibition of the enzyme also involves competition (40). Size profiling of HS degradation products by heparanase in the presence of Nacetylheparin (Fig. 6A) showed that the fully N-acetylated derivative NAH is also cleaved by the enzyme. On the contrary, the gel filtration profile obtained in the presence of glycol-split heparin is practically superimposable to that obtained in the absence of the enzyme (Fig. 6B), indicating that modification of GlcA residues totally abolishes cleavage by the enzyme.
HS assembles ligands and receptors into ternary signaling complexes, best exemplified by the FGF-fibroblast growth factor receptor-heparin complex (62,63). Following cleavage by heparanase, the multitude of polypeptides sequestered and regulated by HS (17) become bioavailable (1,18), and this requires a tight regulation of their activity, applying, among other approaches, modified species of heparin and HS. The present results on the effect of glycol splitting on the ability of heparins to release FGF-2 from ECM extend previous observations on the effect of O-sulfation and N-acetylation on this property (34). As illustrated in Fig. 7, glycol-split derivatives of both heparin (including LMWH) and NAH did not substantially modify the ability to displace FGF-2 from ECM. In other words, while both native and glycol-split heparins efficiently released FGF-2 from ECM, their N-acetylated counterparts exhibited a markedly reduced ability to displace FGF-2, reflecting the essential involvement of N-sulfate groups in this interaction. This observation supports the finding that glycol-split heparins bind FGF-2 (37) and vascular endothelial growth factor (64) with very much the same affinity as unmodified heparins. However, since H, 50 gs inhibits dimerization of FGF-2 (37), it is conceivable that the ECM-bound growth factor is released by glycol-split heparins in an inactive form. Moreover we demonstrated that unlike native heparin and LMWH, NAH and even more so glycol-split NAH ( 100 NAH, 25 gs, i.e. NA,RO-H) failed to stimulate FGF-2-mediated proliferation of HS-deficient lymphoid cells (Fig. 8). The remarkable heparanase-inhibitory activity of N-acetylated, glycol-split heparins together with the low levels of FGF-2 that they release from ECM and their inability to stimulate the mitogenic activity of FGF-2 indicates this class of chemically modified heparins as potential antiangiogenic and antimetastatic agents. These compounds also markedly inhibit wound angiogenesis in transgenic mice overexpressing the heparanase gene (65). Furthermore our preliminary experiments show that some of these heparin derivatives effectively abolish experimental lung colonization of intravenously administered B16-BL6 mouse melanoma cells (Ref. 66).
Retrospective analyses suggest that treatment of venous thromboembolism in cancer patients with LMWHs is associated with additional benefits in terms of their survival (33). The experiments presented in this study were undertaken to develop heparin-based molecules for efficient inhibition of heparanase activity. Polysulfated chains such as those of heparin are expected to envelop the basic clusters of heparanase and compete with its binding to HS. Confirming previous findings (34), we showed that such an activity is retained by a low molecular weight heparin. Moreover we also demonstrated such a retention for three representative low molecular weight glycol-split derivatives. Further studies are planned to determine, for each type of derivative, the shortest chains as well as the shortest IdoA2SO 3 -containing sequences still showing significant inhibition of the enzyme and to elucidate whether the carboxylate groups of glycol-split residues participate in binding and inactivation of heparanase.
In conclusion, we applied desulfation strategies and controlled glycol splitting to remove sulfate groups not necessarily involved in heparanase recognition and inhibition and to improve the molecular flexibility and biological interactions of heparin. Generation of specific heparanase-inhibiting compounds such as those described in this study is important not only as a proof of concept but also as a promising approach to develop heparin-based anticancer lead compounds devoid of side effects.