Modulations of glypican-1 heparan sulfate structure by inhibition of endogenous polyamine synthesis. Mapping of spermine-binding sites and heparanase, heparin lyase, and nitric oxide/nitrite cleavage sites.

Cell surface heparan sulfate proteoglycans facilitate uptake of growth-promoting polyamines (Belting, M., Persson, S., and Fransson, L.-A. (1999) Biochem. J. 338, 317-323; Belting, M., Borsig, L., Fuster, M. M., Brown, J. R., Persson, L., Fransson, L.-A., and Esko, J. D. (2001) Proc. Natl. Acad. Sci. U. S. A., in press). Here, we have analyzed the effect of polyamine deprivation on the structure and polyamine affinity of the heparan sulfate chains in various glypican-1 glycoforms synthesized by a transformed cell line (ECV 304). Heparan sulfate chains of glypican-1 were either cleaved with heparanase at sites embracing the highly modified regions or with nitrite at N-unsubstituted glucosamine residues. The products were separated and further degraded by heparin lyase to identify sulfated iduronic acid. Polyamine affinity was assessed by chromatography on agarose substituted with the polyamine spermine. In heparan sulfate made by cells with undisturbed endogenous polyamine synthesis, free amino groups were restricted to the unmodified, unsulfated segments, especially near the core protein. Spermine high affinity binding sites were located to the modified and highly sulfated segments that were released by heparanase. In cells with up-regulated polyamine uptake, heparan sulfate contained an increased number of clustered N-unsubstituted glucosamines and sulfated iduronic acid residues. This resulted in a greater number of NO/nitrite-sensitive cleavage sites near the potential spermine-binding sites. Endogenous degradation by heparanase and NO-derived nitrite in polyamine-deprived cells generated a separate pool of heparan sulfate oligosaccharides with an exceptionally high affinity for spermine. Spermine uptake in polyamine-deprived cells was reduced when NO/nitrite-generated degradation of heparan sulfate was inhibited. The results suggest a functional interplay between glypican recycling, NO/nitrite-generated heparan sulfate degradation, and polyamine uptake.

Proteoglycans (PGs) 1 are glycosaminoglycan-substituted proteins that can be found in the extracellular matrix or at the cell surface. Glypican constitutes a family of cell surface-bound PGs where the protein is covalently connected at the C terminus to membrane lipids via a so-called glycosylphosphatidylinositol anchor. The central part of the protein consists of a cysteine-rich globular domain that contains information that ensures a high level of heparan sulfate (HS) glycosaminoglycan substitution at three sites located close to the C terminus (1). So far six different human glypicans with the same overall design have been molecularly cloned (for review, see Ref. 2).
Biosynthesis of the HS glycan chains proceeds in a stepwise manner. Serine residues in sequences like DDGSGSGSD (glypican-1) are first substituted with xylose and then the common glycosaminoglycan-to-protein linkage region GlcUA-Gal-Gal-xylose is formed. HS assembly is initiated by a unique ␣-GlcNAc-transferase that adds the first GlcNAc (3). By the alternating addition of GlcUA and GlcNAc, catalyzed by HScopolymerases (4), the extended, linear heparan backbone is formed. A unique step in HS biosynthesis is the regional exchange of N-acetyl for sulfate on glucosamine (4) catalyzed by various isoforms of N-deacetylase/sulfotransferase (NDST). NDST-1, -2, and -4 have an ND/ST ratio well below 1, whereas NDST-3 has a 10-fold higher deacetylase activity (5). Therefore, expression and participation of this isoform during HS biosynthesis would yield a significant proportion of N-unsubstituted glucosamine (GlcNH 3 ϩ ). Free amino groups could also be generated by N-desulfation catalyzed by sulfamidases.
After the formation of N-sulfate (-NSO 3 ) groups, further modifications of the HS chain take place, including epimerization of GlcUA to iduronic acid (IdoUA) and O-sulfations at various positions, yielding a characteristic pattern of alternating unmodified and highly modified segments separated by transition regions. A hypothetical HS sequence is shown in Scheme 1, and the overall structural pattern is depicted in Scheme 2A, top.
We have previously studied the nature of recycling glypican glycoforms in fibroblasts and ECV 304 cells (for review, see Ref. 6). In the latter cells, a brefeldin A (BFA)-arrested, large size glypican-1 glycoform with long HS chains containing multiple GlcNH 3 ϩ residues is degraded by heparanase during a chase, generating HS oligosaccharides and a glypican-1 glycoform with truncated HS chains (7). This glycoform can serve as a precursor for the reformation of a PG with full size HS chains. Re-synthesis of HS on the stubs is prevented by nitrite deprivation and restored when an NO donor is supplied (7). We have proposed that heparanase degradation proceeds until the GlcNH 3 ϩ residues are near the nonreducing end of the stubs, then NO-derived nitrite cleaves at these residues, providing fresh acceptor sites. More recent studies have shown that the GlcNH 3 ϩ residues are indeed concentrated to sites near the reducing side of heparanase cleavage sites in the transition region between unmodified and modified chain segments near the core protein (Ref. 8; see also Scheme 2A, top).
Polyamines (putrescine, spermidine, and spermine) are essential for growth and differentiation of all cells, and they bind electrostatically to polyanions like nucleic acids. The intracellular polyamine levels are tightly regulated by synthesis, degradation, and transport (for review, see Ref. 9). Inhibition of endogenous synthesis, e.g. by inhibition of ornithine decarboxylase with ␣-difluoromethylornithine (DFMO) results in increased polyamine uptake from the environment. The partly disappointing results with DFMO in anti-cancer trials may be explained by compensatory retrieval of extracellular polyamines by tumor cells (9). Inhibition of both polyamine synthesis and uptake could therefore be a useful anticancer strategy. However, the nature of a polyamine transport system in mammalian cells still remains elusive. Studies from this laboratory show that HS binds the polyamine spermine with an affinity that is 10 times greater than that of DNA (10). We have also obtained direct evidence for an involvement of HSPG in the uptake of polyamines by cultured fibroblasts (11). Removal of cell surface HS or inhibition of PG synthesis or sulfation reduces spermine uptake. Upon depletion of the intracellular polyamine pool, cells respond by synthesizing increased amounts of HSPG forms that have higher spermine affinity. Mutant CHO cells deficient in PG synthesis have a reduced polyamine uptake and fail to proliferate and form colonies in the presence of the ornithine decarboxylase inhibitor despite the presence of exogenous spermine. Transfection with cDNA for the missing enzyme restores polyamine uptake. 2 It is thus conceivable that the HS chains of recycling glypi-an could carry polyamines into cells when endogenous synthesis is inhibited by DFMO. After degradation of the HS chains, polyamines bound to HS oligosaccharides would be separated from the recycling truncated PG. To explore whether polyamine deprivation induces changes in HS structure and function that would facilitate polyamine uptake, we have made a compara-tive study of glypican HS from ECV 304 cells with undisturbed polyamine synthesis and from cells treated with DFMO. The PG and its HS chains were examined for spermine affinity as well as for content and location of sulfated hexuronic acids (HexUA) and heparanase and nitrite cleavage sites. We provide evidence that polyamine deprivation induces structural changes that favor increased formation of HS oligosaccharides with higher spermine affinity and that inhibition of HS degradation reduces spermine uptake.
Cell Treatments and Radiolabeling-Cells (ECV 304 and CHO-K1) were maintained as described (7,8,11) and preincubated with the appropriate medium before radiolabeling. Pretreatments with 5 mm DFMO and 1 M spermine were carried out as described (11). Radiolabeling was carried out with 20 Ci/ml D-[6-3 H]glucosamine and 50 Ci/ml [ 35 S]sulfate in sulfate-poor medium (7,8). Radiolabeling in the presence of BFA was generally carried out with cells that had been preincubated with D-[6-3 H]glucosamine alone to achieve labeling of the HS stubs on the small, resident precursor PG (7). Trypsin digestions were performed as described (8).
Extraction and Isolation of PG and PG Products-Cells were extracted with radioimmune precipitation buffer followed by immunoisolation of glypican-1 glycoforms using anti-glypican antiserum as described previously (7). Cells were also extracted with Triton X-100 and PG, and PG-derived material were recovered either by passage over DEAE-cellulose or by desalting on PD-10 (7,8). Separation into PG and the various degradation products was performed by gel-permeation chromatography on Superose 6 or Superdex peptide, and further purification of PG material was achieved by ion exchange chromatography on MonoQ (gradient elution) and sometimes by hydrophobic interaction chromatography on octyl-Sepharose (7,8). Affinity chromatography on spermine-substituted agarose has been described elsewhere (11).
Degradation Procedures-Purified radiolabeled PG was incubated with 75-cm 2 cultures of CHO-K1 cells in 6 ml of medium. Degradation products were recovered from the cell layer after extraction with Triton X-100. HS chains and chain stubs were released from the core protein by treatment with alkaline borohydride (7,8). Enzymatic digestions of HS were performed with HS or heparin lyase and deaminative cleavage at GlcNH 3 ϩ with HNO 2 at pH 3.9 as previously described (7,8). Unsaturated nonreducing terminal hexuronic acid was removed with mercury (II) acetate (12). Radioactivity measurements, buffer changes, concentrations, and recovery procedures as well as carriers were the same as described previously (7,8).
Spermine Uptake Measurements-ECV cells were grown in regular medium with or without 5 mM DFMO for 24 h. The rate of uptake of different concentrations of [ 14 C]spermine was measured as described (11).

RESULTS
General Strategy-A flow chart for the structural analysis of glypican-1 HS is shown in Scheme 2. We first examined HS chains of large (A) and small (B) glypican-1 glycoforms as well as HS oligosaccharides from ECV cells with undisturbed polyamine synthesis and then corresponding HS material from ECV cells that were treated with DFMO to up-regulate polyamine import. HS chains and oligosaccharides were analyzed for the content and location of GlcNH 3 ϩ residues, sulfated IdoUA, and heparanase cleavage sites in relation to the unmodified and modified regions and for high affinity sperminebinding sites (C). Finally, we explored whether inhibition of HS degradation could affect polyamine uptake.
Mapping of GlcNH 3 ϩ Residues in HS of the Large Glypican Glycoform from Cells with Undisturbed Polyamine Synthesis-HS chains from the [ 3 H]glucosamine-and [ 35 S]sulfate-labeled large size glypican PG of BFA-treated ECV cells were cleaved with nitrite at pH 3.9 at GlcNH 3 ϩ residues (Scheme 1, site 4), and the various products (Scheme 2) were separated by chromatography on Superdex peptide (Fig. 1A). As expected, most SCHEME 1. Cleavage sites in a hypothetical heparan sulfate chain. The indicated cleavage sites are: 1, HS lyase; 2, heparanase; 3, heparin lyase (when GlcNR is GlcNS); 4, NO/nitrite. R is either acetyl or sulfate (S); HexUA (HexA) is either GlcUA (GlcA) or IdoUA (IdoA). Under forcing conditions HS lyase may also cleave adjacent to IdoUA (13,14), especially in sequences close to the linkage region (14). Similarly, heparin lyase may also cleave at the rare sulfated GlcUA residues (15). Moreover, heparin lyase cleaves preferentially exolytically (16), which means that cleavage of internal linkages and short fragments takes place at a slower rate (16 -18). SCHEME 2. Flow chart of the strategy for analyzing HS structure and spermine affinity. ECV cells not treated or treated with DFMO to deplete endogenous polyamine pools were used in all cases (A-C). In A, large size glypican-1 precursor was isolated from cells that were also exposed to BFA, and in B, smaller-size glypican-1 and HS oligosaccharides obtained from cells not treated with BFA were separately isolated. In C, selected fractions were examined for spermine affinity. ϩ to regions near the core protein attachment (Ser) and clustering of sulfated IdoUA to the highly modified domains (8). Heparanase cleavage sites are located at the reducing (filled arrow) or nonreducing side (open arrow) of highly modified regions (20). The glypican preparations were either treated with alkali to release HS chains or directly with heparanase (see II and III) to generate HS fragments and core protein with remaining HS stubs. Intact, alkali-released HS chains were exposed to HNO2 at pH 3.9 to identify the location of GlcNH 3 ϩ residues. Fragments obtained were separated and digested with heparin (Hep) lyase to identify sulfated HexUA and by HS lyase to cleave the unmodified backbone. Alternatively, HS chains were directly cleaved by HS lyase to liberate the modified sections and those containing GlcNH 3 ϩ . These products were also separated and subsequently treated with either HNO 2 at pH 3.9 or heparin lyase. Cleavage of HS in intact glypican by heparanase yields both long and short fragments (II and III, respectively). The former pool includes stubs attached to the core protein via Ser. The core protein has a C-terminal, hydrophobic glycosylphosphatidylinositol-anchor (filled circle with two rods). The latter property was used to separate the degradation products, and HS stubs were released from the core protein by alkali treatment. Heparanase-released HS fragments (from II and III) were further degraded by heparin lyase or HNO 2 at pH 3.9.
Mapping of Heparanase Cleavage Sites in HS of the Large Glypican Glycoform-Heparanase is expected to cleave HS at certain glucuronidic linkages (Scheme 1, site 2; see also Ref. 19) located on either side of the highly modified regions (Scheme 2A, for review, see Ref. 20). However, CHO cell-derived heparanase may not require the 2-O-sulfate group on the HexUA (20). Moreover, the site closest to the core protein (Scheme 2A) may not always be cleaved (8). Because the unmodified stretches generally are longer than the more modified ones, two types of fragments should be obtained (Scheme 2A, bottom half), i.e. longer ones (pool II), comprising the unmodified and less modified stretches and including the stubs still attached to the core protein as well as shorter ones (pool III) carrying the highly modified regions. Previous studies show (8) that ECV cells express a surface-located, suramin-inhibited heparanase that degrades exogenously supplied, mature glypican PG into the above-mentioned products. Because we subsequently found that many CHO cell lines express more potent heparanase activity, these cells were used in most of the following experiments. To obtain sufficient amounts of material for further analysis, several batches of glypican PG were repeatedly incubated with the CHO cells, and the products were recovered and separated on Superose 6. One such example is shown in Fig. 2.
[ 35 S]Sulfate-labeled glypican PG obtained from BFA-treated ECV cells and purified via both gel-permeation and ion exchange chromatography (8) was incubated with heparanasecontaining CHO-K1 cells, and the products were separated by gel permeation chromatography ( Fig. 2A). Pool I consisted of undegraded and partially degraded PG. Pool II, which should consist of long HS-chain fragments and glypican core protein with truncated HS chains (see Scheme 2A, pool II), was passed over octyl-Sepharose. The unbound, alkali-resistant (data not shown) free HS-chain fragments were rechromatographed both before (Fig. 2B) and after treatment with nitrite at pH 3.9 (Fig.  2C). The results indicated the presence of internally located nitrite-sensitive sites in these fragments. Hence, the less mod-ified stretches contain solitary GlcNH 3 ϩ residues (see Scheme 2A, pool II). The position of the GlcNH 3 ϩ residues relative to heparin-type repeats in this and other heparanase-released fragments was assessed in the experiments shown below.
Mapping of GlcNH 3 ϩ Residues and Heparin-type Repeats in Heparanase-generated Fragments Comprising Unmodified Regions-Because heparanase cleaves a glucuronidic bond, subsequent cleavage of a glucosaminidic bond by nitrous acid or heparin lyase would yield odd-numbered saccharide fragments (see Scheme 1 if sites 2 and 4 were cleaved). To assess whether we could distinguish between odd-and even-numbered saccharides, HS-lyase digests of [ 3 H]glucosamine and [ 35 S]sulfatelabeled HS chains, comprising a series of even-numbered saccharides, were chromatographed on Superdex peptide either  (Fig. 3A) or after subsequent removal of nonreducing terminal ⌬HexUA residues to obtain odd-numbered saccharides (Fig. 3B). The results showed that it might be possible to distinguish between these saccharide series, at least when they appear separately.
The long HS-chain fragments released by heparanase treatment of [ 3 H]glucosamine and [ 35 S]sulfate-labeled large size glypican PG (see Scheme 2A, pool II, and obtained as in Fig. 2, A-C) were treated with nitrite at pH 3.9 and chromatographed on Superdex peptide (Fig. 3C) to demonstrate that no small odd-numbered terminal saccharide fragments had been released. Hence GlcNH 3 ϩ residues and heparanase cleavage sites were not located close to one another in the peripheral parts of HS chains. To check whether IdoUA(2-OSO 3 )/GlcUA(2-OSO 3 ) residues were located close to the nitrite cleavage sites, the nitrite-treated material was subsequently digested with heparin lyase and rechromatographed (Fig. 3D). Again, there was no release of small saccharides. We can thus conclude that IdoUA(2-OSO 3 )/GlcUA(2-OSO 3 ) residues were rare in these segments of the HS chain (Scheme 2A).
Mapping of GlcNH 3 ϩ Residues in HS Segments near the Core Protein-The glypican core protein with HS stubs remaining after heparanase treatment of the large size PG (see Scheme 2A) was recovered from pool II (as in Fig. 2A) by adsorption to octyl-Sepharose. This material, which was alkali-sensitive (data not shown), was excluded from Superdex peptide (Fig.  3E). To identify GlcNH 3 ϩ residues, it was treated with nitrite and rechromatographed (Fig. 3F). A series of saccharide fragments, probably mostly small odd-numbered ones but also larger even-numbered ones, were obtained. A monosaccharide that might be present could only be derived from a sequence (GlcUA)-GlcNH 3 ϩ -HexUA(-OSO 3 )-provided that the heparanase can cleave adjacent to a GlcNH 3 ϩ residue (19). A trisaccharide could be derived from a sequence (GlcUA)-GlcNR-HexUA(-OSO 3 )-GlcNH 3 ϩ -. However, few of the small fragments obtained (1-5 in Fig. 3F) appeared to be sulfated. Therefore, this arrangement was probably very rare. Consecutive GlcNH 3 ϩ residues that were not detected in preceding experiments (Fig.  1A) should have yielded disaccharides. The smallest fragment obtained from the core protein-attached HS stubs may thus be a monosaccharide (1 in Fig. 3F). Tetra-, hexa-, and octasaccharides that were obtained (4, 6, and 8 in Fig. 3F) should be derived from sequences containing clustered GlcNH 3 ϩ residues that are known (8) to be concentrated to the region near the core protein (see Scheme 2A). The radiosulfate-labeled material eluting in the excluded volume (v o in Fig. 3F) may correspond to long stubs still attached to the core protein (see Scheme 2A, pool II) but devoid of GlcNH 3 ϩ residues and connected to a highly modified region (from pool III in Scheme 2A) because a heparanase site was lacking or not cleaved. Alternatively or in addition there could be fragments obtained by cleavage of widely spaced heparanase and nitrite cleavage sites (Scheme 1, sites 2 and 4) with the latter located close to the core protein.
Mapping the Distance between the Core Protein and the First GlcNH 3 ϩ Residue-To determine the distance between the core protein and the first GlcNH 3 ϩ residue, [ 3 H]glucosamine-and [ 35 S]sulfate-labeled glypican PG from BFA-treated ECV cells was directly subjected to cleavage by nitrite, passed over octyl-Sepharose to recover core protein with HS stubs (see Scheme 2A, bottom right), and treated with alkali to release these stubs, which were then chromatographed on Superdex peptide (Fig. 3G). Most of the material was still eluted in the excluded volume, indicating that there were at least 12-14 GlcUA-Glc-NAc repeats between the core protein and the first GlcNH 3 ϩ residue. The minimum size of a 3 H-labeled deaminative cleavage product would be the hexasaccharide HexUA-GlcNR-GlcUA-Gal-Gal-xylose generated by cleavage at the second glucosamine. Small amounts of material was detected in the retarded fractions (Fig. 3G), indicating that a free amino group in the second GlcN position was rare. The longest HS stubs released by alkali from nitrite-cleaved glypican PG and which eluted in the excluded volume of Superdex peptide (Fig. 3G) were also chromatographed on Superose 6 and compared with the elution position on the same column of HS stubs released by alkali from the glypican core protein after heparanase treatment (see Scheme 2A, pool II, and bar in Fig. 3E). Their elution profiles coincided, indicating that in segments near the core protein, heparanase cleavage sites and GlcNH 3 ϩ residues could be closely spaced (data not shown).  Fig.  2A. To check whether this pool included some small saccharides, it was also chromatographed on Superdex peptide (Fig.  3H). Some trisaccharide was obtained (3 in Fig. 3H), indicating that cleavage of a nitrite-sensitive site near the heparanase site could take place when exogenous PG is incubated with cells. However, these sites appear to be rare.

Mapping of GlcNH
Most of the heparanase-generated fragments that contained the highly modified regions (Scheme 2A, pool III) were excluded from Superdex peptide (Fig. 3H). This material was pooled (see bar), digested with heparin lyase, and rechromatographed (Fig. 3I). Small sulfated saccharides were formed, indicating the presence of IdoUA(2-OSO 3 )/GlcUA(2-OSO 3 ) residues. Some of them appeared to be odd-numbered (1 and 3 in Fig. 3I). A monosaccharide could only be derived from the nonreducing end of an heparanase-generated oligosaccharide because cleavage of the glucuronidic (bold face) linkage in -GlcNR-HexUA-GlcNR-GlcUA-GlcNR-HexUA(-OSO 3 )-GlcNRwould yield GlcNR-HexUA(-OSO 3 )-GlcNR-terminating fragments. If the HexUA was IdoUA, R in the nonreducing terminal GlcNR would have to be N-sulfated (5), and then heparin lyase would release the monosaccharide GlcNSO 3 . The results were inconclusive (Fig. 3I), and it appears more likely that most of the heparin lyase-released saccharides were derived from the highly modified regions in the center of the heparanase-released fragments (Scheme 2A, pool III). Treatment with nitrite did not cleave the oligosaccharide backbone (Fig. 3J), confirming that GlcNH 3 ϩ residues are absent from the highly modified regions.
Localization of High Affinity Spermine-binding Sites in HS of the Large Glypican Glycoform-Spermine binding was assessed by affinity chromatography on spermine-substituted agarose eluted with a linear guanidinium chloride gradient (Fig. 4). BFA-arrested, [ 3 H]glucosamine-and [ 35 S]sulfate-labeled glypican PG precursor with long, GlcNH 3 ϩ -containing HS chains (Scheme 2A and Ref. 8), eluted as a uniform peak centered at an ionic strength of ϳ0.8 m (Fig. 4A). The degradation products obtained after heparanase treatment of this PG (Scheme 2A), i.e. long HS-chain fragments (from pool II), core protein with HS stubs (from pool II) and smaller HS fragments (pool III) were chromatographed separately on spermine-agarose (Fig. 4, B-D). Most of the long and free HS-chain fragments comprising the unmodified regions appeared at a somewhat lower ionic strength of ϳ0.7 m (Fig. 4B). The core protein with HS stubs showed relatively low affinity and eluted as a broad distribution between 0.2 and 0.7 M (Fig.  4C). A major population of the smaller HS fragments comprising the modified regions eluted at the same ionic strength as  Fig.  2A, pools II and III, respectively). Free HS-chain fragments and core protein with HS stubs were separated by passage of pool II over octyl-Sepharose (see Scheme 2A). Free chain fragments were treated with nitrite at pH 3.9 (as in Fig. 2C) and rechromatographed on the Superdex peptide column (C). The same material was both treated with nitrite at pH 3.9 and digested with heparin lyase and rechromatographed (D). Core protein with HS stubs (see Scheme 2 ) was chromatographed both before (E) and after treatment with nitrite at pH 3.9 (F). [ 3 H]Glucosamine-and [ 35 S]sulfate-labeled glypican PG from BFA-treated ECV 304 cells was directly subjected to cleavage by nitrite, passed over octyl-Sepharose to recover core protein with HS stubs, and treated with alkali to release these stubs, which were then chromatographed (G). HS oligosaccharides (as in pool III in Fig. 2A) were chromatographed directly (H), after digestion with heparin lyase (I), and after treatment with nitrite at pH 3.9 (J). In the latter case, only 3 H was recorded. Numbers 2, 4, 6 etc. refer to di-, tetra-, hexasaccharide and so on; 1, 3, 5 etc. refer to mono-, tri-, pentasaccharide and so on. E, 3 H; f, 35 S; V o , void volume; V t , total volume. the long fragments, indicating that the smaller fragments have the highest relative spermine affinity (Fig. 4D). As shown previously (10), there is also a nonspecific correlation between HS chain size and spermine affinity. The high affinity HS fragments displayed reduced affinity upon digestion with heparin lyase (Fig. 4E), indicating that IdoUA(2-OSO 3 )/GlcUA(2-OSO 3 ) residues contribute to the high affinity as shown previously (10).
Localization of High Affinity Spermine-binding Sites in HS of the Small Glypican Glycoform-The small size, immunoisolated glypican-1 glycoform present in untreated ECV cells appeared to contain three populations of spermine binding material (Fig. 4F). The most retarded major component eluted at an ionic strength of ϳ0.9 M, i.e. somewhat later than the large glypican PG from BFA-treated cells (Fig. 4A). Although the large glycoform carries HS chains with high affinity sperminebinding sites (see Fig. 4D), there is also a greater number of GlcNH 3 ϩ residues. The presence of positive charges is expected to reduce overall spermine affinity.
Effects of Polyamine Deprivation on the Structure of HS in the Small Glypican Glycoform-ECV cells were made dependent on an exogenous supply of spermine by treatment with DFMO during growth (11). When the cultures had reached confluency after growth in spermine-containing medium, cells were incubated with [ 3 H]glucosamine and [ 35 S]sulfate for 24 h in the continued presence of DFMO with or without exogenous spermine. The small glypican-1 glycoform was immunoisolated from both cultures, and the HS chains were released by alkali and analyzed for size and GlcNH 3 ϩ content (Scheme 2B). As shown in Fig. 5A, the HS chains of glypican-1 from cells radiolabeled during spermine uptake comprised both large and smaller size chains, whereas those obtained in the absence of spermine were mainly of smaller size (Fig. 5E). Furthermore, HS chains of glypican-1 from cells radiolabeled during spermine uptake were partly degraded by nitrite to a series of fragments (Fig. 5B) including sulfated oligosaccharides (V t ), indicating the presence of multiple sites containing GlcNH 3 ϩ residues close to the sulfated, highly modified regions. In contrast, HS chains derived from glypican radiolabeled in the absence of concomitant spermine uptake contained few if any free amino groups (Fig. 5F).
The ratio of [ 35 S]sulfate-to-[ 3 H]glucosamine was higher in HS chains radiolabeled during spermine uptake (Fig. 5A) than in HS from cells not supplied with spermine (Fig. 5E). Although this may reflect greater sulfation, it is also possible that the

FIG. 4. Affinity chromatography on spermine-agarose of various glypican PG glycoforms and its degradation products produced in untreated or DFMO-treated cells. [ 3 H]Glucosamine and [ 35 S]sulfate-labeled large size glypican PG obtained by gel and ion exchange chromatography from BFA-treated ECV 304 cells was subjected to cleavage by heparanase of CHO-K1 cells followed by separation on Superose 6
into large and small cleavage products (as in Fig. 2A, pools II and III, respectively). Free HS-chain fragments and core protein with HS stubs were separated by passage of pool II over octyl-Sepharose (see Scheme  presence of cationic spermine affects uptake of sulfate and glucosamine. Effects of Polyamine Deprivation on the Structure of HS in the Large Glypican Glycoform-The HS chains of the BFAarrested, large glypican PG precursor produced in cells with up-regulated polyamine uptake were also examined for free amino groups. When the cultures had reached confluency under the same regimen as above, cells were incubated with [ 3 H]glucosamine and [ 35 S]sulfate in the presence of both DFMO and BFA with or without exogenous spermine. Glypican PG was isolated from both cultures, and the HS chains were released by alkali and analyzed for size and GlcNH 3 ϩ . As shown in Fig. 5, C-D and G-H, respectively, the HS chains were similar in size and extensively degraded by nitrite in both cases. The number of GlcNH 3 ϩ residues clearly exceeded that seen in HS chains of BFA-arrested glypican from ECV cells with undisturbed endogenous polyamine synthesis (see Ref. 8 and Fig. 4).
To investigate possible effects of intracellular polyamine deprivation on the sulfation of HS, we measured the content of sulfated IdoUA residues by degradations with heparin lyase. Mature HS chains derived from the large BFA-arrested PG precursor isolated from ECV cells with undisturbed endogenous polyamine synthesis were digested with heparin lyase and chromatographed on Superose 6 (Fig. 6A). One pool of larger, lower sulfated fragments and another pool of smaller, higher sulfated fragments were obtained. The latter were examined for GlcNH 3 ϩ residues by treatment with nitrite at pH 3.9 followed by gel permeation chromatography on Superose 6 ( Fig. 6B). No degradation could be observed. Furthermore, there was no release of small saccharides from the termini of the fragments, as indicated by chromatography on Superdex peptide (data not shown). Hence, there was little clustering of IdoUA(2-OSO 3 )/GlcUA(2-OSO 3 ) and GlcNH 3 ϩ residues in accordance with the results presented above (Fig. 1H and 3J).
The results obtained with HS chains of corresponding PG from DFMO-treated cells obtained after radiolabeling in the presence or absence of spermine are shown in Fig. 6, C and D, respectively (for comparisons with undigested chains, see Fig.  5, C and G, respectively). Both HS preparations were sensitive to degradation by heparin lyase and afforded, as in the preceding case, one pool of larger fragments and one pool of smaller ones. The yields of the latter fragments (see the bars in Fig. 6, A, C, and D) were 1.5-fold greater in the case of chains derived from cells treated with DFMO (Fig. 6, C and D) than from untreated cells (Fig. 6A). The heparin lyase-generated HS fragments were examined for GlcNH 3 ϩ residues by treatment with nitrite at pH 3.9 followed by gel permeation chromatography on Superdex peptide (Fig. 6, F and H). The small heparin lyasereleased fragments were sensitive to deaminative cleavage yielding small saccharides, mostly tetra-, hexa-, and octasac-charides (2)(3)(4)(5)(6)(7)(8) in Fig. 6, F and H; cf. Fig. 6, E and G). Hence, polyamine deprivation causes structural changes in HS that increases the number of both IdoUA(2-OSO 3 )/GlcUA(2-OSO 3 ) and GlcNH 3 ϩ residues and brings them closer together.

Effect of Polyamine Deprivation on the Affinity of Heparan
Sulfate for Spermine (Scheme 2C)-We examined the effects of polyamine deprivation on the spermine affinity of the various glypican-1 glycoforms and HS-degradation products by affinity chromatography (see Fig. 4). The small size GlcNH 3 ϩ -rich glypican-1 glycoform obtained from DFMO-treated cells radiolabeled in the presence of spermine uptake contained more components with lower affinity for spermine (Fig. 4G) compared with corresponding material from cells with ongoing endogenous polyamine synthesis (Fig. 4F). The extreme clustering of IdoUA(2-OSO 3 )/GlcUA(2-OSO 3 ) and GlcNH 3 ϩ residues in HS of this glypican glycoform from DFMO-treated cells (see Figs. 5B and 6C) may have inhibited spermine binding. Accordingly, GlcNH 3 ϩ -poor glypican-1, produced by DFMO-treated cells in the absence of spermine uptake, comprised a major high affinity population (Fig. 4H) of similar affinity (0.8 -0.9 M) to that of glypican from unperturbed cells (Fig. 4F).
We also examined the effect of DFMO treatment on the spermine affinity of the large size, BFA-arrested glypican. The PG precursor produced in cells with up-regulated polyamine transport had a greater affinity for spermine (peak elution positions 0.9 -1.0 M in Fig. 4, I and J) than the PG from cells with undisturbed endogenous polyamine synthesis (0.8 M, Fig.  4A). It should be pointed out that the two PG forms from DFMO-treated cells contained more free amino groups, i.e. more positive charges, than PG from untreated cells that would counteract the binding to spermine-agarose. Hence, they should harbor sites with increased spermine affinity to compensate for their lower net negative charge. Accordingly, upon ion exchange chromatography on MonoQ the BFA-arrested PG from cells with endogenous polyamine synthesis bound more strongly than the corresponding PG from DFMO-treated cells (data not shown). This is in agreement with their net negative charges and indicates that spermine binding can be specific and not only a nonspecific electrostatic attraction.
We finally examined the endogenously generated HS oligosaccharides that are degradation products of the large size glypican PG precursor (7) from cells with undisturbed polyamine synthesis and from cells treated with DFMO to inhibit polyamine synthesis and increase polyamine uptake. HS oligosaccharides formed in cells with endogenous polyamine synthesis showed relatively low spermine affinity and eluted in three pools at 0.4, 0.5, and 0.7 m (Fig. 4K). In the oligosaccharide pool obtained from DFMO-treated cells, a separate component with high spermine affinity (elution position, 0.7 M) was markedly increased (Fig. 4, L and M).
Effect of Nitrite Deprivation on Spermine Uptake-As shown in Fig. 5D, up-regulation of spermine uptake by DFMO treatment resulted in an increased content of GlcNH 3 ϩ in the HS chains of the large size glypican PG precursor. The HS degradation products (oligosaccharides) generated endogenously from this glycoform during glypican recycling (7) were also examined for the presence of GlcNH 3 ϩ residues by treatment with nitrite and chromatography on Superdex peptide. There was no indication of cleavage in any of the samples (data not shown), indicating that both heparanase-and nitrite-generated cleavage was taking place during their endogenous formation. It is possible that NO/nitrite-generated cleavage of HS is required for spermine uptake. We therefore examined whether spermine uptake was affected by nitrite deprivation (Fig. 7). ECV cells displayed Michaelis-Menten polyamine uptake kinetics with a K m value of ϳ0.5 M, i.e. in the same range as other cultured cell lines (11). DFMO treatment resulted in a 1.5-fold increase in V max for spermine uptake but no change in K m , as indicated by a Lineweaver-Burke plot (data not shown), suggesting an increased number of spermine-binding sites or transporters. Uptake was reduced to control level either by scavenging nitrite with sulfamate or by preventing NO release from nitrosothiols by the Cu(I)-selective chelator neocuproine. By treatment with the selective endothelial NO synthase inhibitor N-nitroarginine, uptake of spermine was reduced even further, presumably because the ultimate source of nitrite (NO released from Arg) was depleted.

DISCUSSION
Cell surface HSPG are selective regulators of ligand-receptor encounters. Thereby they can regulate the signaling of growth factors and morphogens during growth and developmental patterning (for reviews, see Refs. 2, 21, and 22). HSPG are also involved in the uptake and internalization of small basic molecules such as polyamines (11) and basic peptides such as fibroblast growth factor (23) and human immunodeficiency virus-1 Tat (24) as well as of virus particles (25,26). The glycosylphosphatidylinositol-anchored HSPG glypican may localize to caveolae, membrane domains involved in specific forms of internalization (27). However, the mechanism for transfer of HS-ligand complexes from the vesicular to the cytosolic compartment, which is required for nuclear targeting, remains a mystery.
Polyamines are natural constituents of all cells and essential for cell growth. They are generally believed to support nuclear processes such as transcription and export of mRNA (9). Glypican core proteins contain nuclear localization signals and have been localized to nuclei of neurons and glioma cells (28). Earlier studies by Conrad and co-workers (29) demonstrate transport of HS into the nuclei of hepatocytes. Studies on polyamine uptake may therefore offer novel insights into the mechanism of HS-ligand internalization and nuclear targeting. Specific inhibitors of polyamine biosynthesis, e.g. DFMO, can be used to up-regulate polyamine uptake.
Recycling glypicans with their polyamine binding HS side chains are thus potential vehicles for the transport of polyamines into and out of cells (6 -8, 10, 11). Here we show that the mature HS chains of the glypican-1 precursor produced by ECV cells with undisturbed polyamine synthesis have the general design depicted in Scheme 3A. The non-sulfated backbone structure with its occasional and clustered GlcNH 3 ϩ residues displays relatively low affinity for the polyamine spermine. As expected, the high affinity spermine-binding sites are localized to the highly modified and sulfated regions that are scattered along the chain. These spermine binding regions are released as HS oligosaccharides upon degradation by heparanase. Degradation probably begins at the non-reducing end and progresses toward the core protein as glypican is transported to an intracellular compartment (Scheme 4). In unperturbed ECV cells at steady state there is a major pool of potential spermine binding HS oligosaccharides and a minor pool of glypican PG with truncated HS chains that are on the average 30 -50% the size of the mature ones (7,8). A substantial portion of the polyamines taken up by cells may thus return to the environment bound to the HS chains of recycling glypican. It is known that cells continuously leak polyamines into the environment (9).
When cells are treated with the ornithine decarboxylase inhibitor DFMO, endogenous polyamine synthesis diminishes, and net uptake from the environment increases. It appears that cells somehow adapt to this situation by appropriate alterations of the glypican-HS structure (Scheme 3B). The struc- FIG. 7. Effect of nitrite deprivation on spermine uptake. The curves show [ 14 C]spermine uptake by untreated (E) or DFMO-treated cells (q) and uptake by DFMO-treated cells that were simultaneously exposed to either 10 mM N-nitroarginine (OE), 0.01 mM neocuproine (f), or 10 mM sulfamate (ࡗ).
tural and functional changes include (a) increased sulfation (see S in Scheme 3) and, thereby, increased spermine affinity and (b) an increased number of NO/nitrite cleavage sites (ϩ in Scheme 3), resulting in increased potential to form spermine binding oligosaccharides. The modulation of HS structure can take place in BFA-treated cells. Hence de novo synthesis of glypican is not required. Polyamine deficiency may modulate HS on recycling glypican either by signaling to the biosynthetic machinery (up-regulation of NDST-3 and 2-OST?) or to the degradative pathway (induction/activation of sulfamidase?). Because IdoUA(2-OSO 3 )/GlcUA(2-OSO 3 ) and GlcNH 3 ϩ residues also become more adjacent to one another, even a sequence containing a heparanase recognition site (bold face) like GlcNR-HexUA-GlcNR-GlcUA-GlcNR-HexUA(-OSO 3 )-GlcNR-may contain N-unsubstituted glucosamines on either side. Thus, fragments generated by the combined action of heparanase and nitrite could both be small (trisaccharide) and highly sulfated.
Although the mature HS chains on the BFA-arrested PG precursor produced in DFMO-treated cells contained a large number of GlcNH 3 ϩ residues, the HS oligosaccharides derived from it contained very few if any GlcNH 3 ϩ residues. The HS stubs on the core protein retained such residues only when spermine uptake and radiolabeling took place at the same time (Fig. 5). However, degradation of HS by nitrite appears to be required for spermine uptake, because nitrite deprivation (a) increases the GlcNH 3 ϩ content of HS (8), (b) suppresses HS degradation, (c) prevents glypican recycling (7), and (d) inhibits spermine uptake (Fig. 7). The reason why some GlcNH 3 ϩ residues remained in some, presumably undegraded HS chains of the small glypican glycoform when spermine uptake occurred could be because of a negative feedback mechanism, whereby increased spermine uptake inhibits nitrite formation and, thus, also HS degradation (Scheme 4). It is known that polyamines in general and spermine in particular inhibit NO synthase (30,31), thereby reducing the formation of NO, the precursor of nitrite (32). The present results suggest a functional interplay between glypican recycling, NO/nitrite generation, HS degradation, and polyamine uptake (Scheme 4). Studies on possible synergistic anti-tumor effects of DFMO and inhibitors of glypican recycling/HS biosynthesis are in progress. SCHEME 4. Postulated functional interplay between glypican recycling, HS degradation by heparanase and nitrite, and polyamine uptake. Polyamines are bound electrostatically to the HS chains of glypican and are carried into the cell when glypican is internalized, possibly via caveolae. As heparanase degrades HS, polyamines are released, probably still bound to the HS fragments. Glypican with HS stubs return to the Golgi where HS chains are resynthesized provided that the GlcNH 3 ϩ units are cleaved-off via deaminative cleavage catalyzed by endogenously formed NO-derived nitrite. Polyamines exit from caveolae, endosomes, or possibly caveosomes (27) by some unknown mechanism. Small HS fragments generated by the combined action of heparanase and nitrite on the structurally altered HS in polyamine-deprived cells could support membrane penetration of polyamines.