Heparan Sulfate Undergoes Specific Structural Changes during the Progression from Human Colon Adenoma to Carcinoma in Vitro *

We report a detailed analysis of heparan sulfate (HS) structure using a model of human colon carcinogenesis. Metabolically radiolabeled HS was isolated from adenoma and carcinoma cells. The chain length of HS was the same in both cell populations (M r 20,000; 45–50 disaccharides), and the chains contained on average of two sulfated domains (S domains), identified by heparinase I scission. This enzyme produced fragments of approximate size 7 kDa, suggesting that the S domains were evenly spaced in the intact HS chain. The degree of polymer sulfation and the patterns of sulfation were strikingly different between the two HS species. When compared with adenoma HS, the iduronic acid 2-O-sulfate content of the carcinoma-derived material was reduced by 33%, and the overall level of N-sulfation was reduced by 20%. However, the level of 6-O-sulfation was increased by 24%, and this was almost entirely attributable to an enhanced level of N-sulfated glucosamine 6-O-sulfate, a species whose data implied was mainly located in the mixed sequences of alternating N-sulfated and N-acetylated disaccharides. The results indicate that in the transition to malignancy in human colon adenoma cells, the overall molecular organization of HS is preserved, but there are distinct modifications in both the S domains and their flanking mixed domains that may contribute to the aberrant behavior of the cancer cell.

Heparan sulfate is a widespread complex linear polysaccharide that consists of alternate hexuronic acid and N-substituted glucosamine residues. The ability to bind protein effectors such as growth factors or protease inhibitors (e.g. antithrombin III; Refs. 1 and 2) is strongly influenced by the position and density of sulfate residues that occur most frequently as N-sulfates but are also present as sulfate esters at the 6-O-(or less commonly the 3-O-) position of glucosamine or the 2-O-position of iduronic acid (3). The complexity of HS is achieved through the location of sulfate groups in domains of high and low sulfation (4), the conformational flexibility of iduronate residues (5,6), and the degree of polymorphism manifest in different tissues (7)(8)(9). The strongly anionic zones of HS, 1 the S domains, consist of contig-uous glucosamine N-sulfate-containing disaccharides that bear a variable number of O-sulfate moieties. Domains of less sulfated sequences called mixed sequences are believed to flank the S domains, separating them from the unsulfated domains, and these are liberated as tetrasaccharides by low pH nitrous acid, a reagent that cleaves HS at N-sulfated glucosaminecontaining disaccharides (i.e. GlcNSO 3 -␣1-4-hexuronic acid).
The ability of HS to act as a growth factor activator (10 -13) and as a component of focal adhesions (14) has focused attention on this molecule as a potential therapeutic target in diseases of aberrant cellular growth or migration such as cancer, diabetic retinopathy, or coronary arterial restenosis. A number of studies have investigated the structural changes in HS in animal models of malignancy with the earliest detailed analyses being performed on SV-40-transformed murine embryo cell lines (15)(16)(17). These studies showed that transformation was accompanied by a reduction in charge density that was largely due to a decrease in 6-O-sulfation in nitrous acid-resistant tetrasaccharides. Further analyses of normal and transformed mouse mammary basement membrane proteoglycans confirmed that carcinogenesis was associated with reductions in 6-O-sulfation and also of 3-O-sulfation and a reduction in the size of intact HS (18), changes that were associated with a reduced affinity for antithrombin III.
There have been no detailed studies of HS structure in human malignant tissues, but a crude comparison of normal liver and hepatoma tissue showed that there was a reduction in the charge density of HS in the neoplastic tissue (19).
The first in vitro model of the progression of human colon cancer from adenoma to carcinoma was developed recently by Paraskeva and co-workers (20). A cell line with the phenotype of an adenoma was derived from a polyp taken from a patient with familial adenomatous polyposis. Through chemical transformation, a second cell line was derived that was tumorigenic in nude mice and anchorage-independent in soft agar, in contrast to the adenoma cell line. We have used these cell lines to identify the structural changes in HS during the progression from human colon adenoma to carcinoma. Cell Culture-The adenoma and carcinoma cells were grown in conditions described by Williams et al. (20) and were used within 10 passages.

Materials-D-[6-
Preparation Of Heparan Sulfate-Glycosaminoglycans were prepared from cells contained in four 175-cm 2 tissue culture flasks. The adenoma and carcinoma cells were labeled at 80% confluence by incubation in normal medium containing 5 mCi/ml D-[6-3 H]glucosamine and 5 mCi/ml Na 2 35 SO 4 . After 48 h, the cell culture media from the four flasks were pooled and treated with 100 g/ml Pronase at 37°C for 24 h. To prepare extracts of the cell layer/extracellular matrix, each flask was treated with 20 ml of phosphate-buffered saline (PBS), pH 7.4, containing 1% (v/v) of Triton X-100 detergent and 100 g/ml Pronase for 24 h at 37°C. Both the cell/matrix extracts and the cell culture media were pooled and frozen at Ϫ20°C until they were analyzed.
The pooled extracts were thawed, centrifuged (1000 ϫ g for 20 min. at 4°C), and the supernatant was loaded under gravity onto a DEAE-Sephacel column (1.5 ϫ 25 cm) that had been previously equilibrated with PBS. The column was washed with PBS until there was no further elution of radioactivity and then eluted with a linear gradient from 0.15 to 0.825 M NaCl in phosphate buffer, pH 7.4, at 10 ml/h. Double-radiolabeled peaks corresponding to sulfated GAGs (HS and chondroitin sulfate/dermatan sulfate) were collected, diluted with distilled water ϫ 3 (v/v), and then loaded under gravity onto a 1 ϫ 4 cm DEAE-Sephacel column pre-equilibrated with PBS. The column was then washed with 30 ml of PBS, step-eluted with 5 ml of 0.3 M NaCl in phosphate buffer, pH 7.4, to remove any residual hyaluronan, then step-eluted again with 20 ml of 1.5 M NaCl, phosphate buffer, pH 7.4, to desorb the sulfated GAGs. The latter were volume-reduced to approximately 1 ml by centrifugal evaporation. They were then loaded onto a 1 ϫ 30-cm Sephadex G-50 (superfine) column, and pre-equilibrated and eluted with 0.2 M NH 4 HCO 3 . The GAGs were mainly present in the void volume and were pooled and exhaustively lyophilized.
The GAG preparation was dissolved in 1 ml of chondroitinase buffer (50 mM Tris-HCl, 50 mM NaCl, pH 7.8 -8) and treated with 20 mIU of chondroitinase ABC (E.C. 4.2.2.4) for 24 h at 37°C. The volume was increased to 10 ml with distilled water, and the solution was reloaded onto a small DEAE-Sephacel column (1 ϫ 4 cm) under gravity. This was washed with 20 ml of phosphate-buffered saline (0.3 M), pH 7.2, to remove degraded chondroitin sulfate fragments and then step-eluted with 1.5 M NaCl to elute the HS. This was desalted on Sephadex G-50 and lyophilized as above.
Estimation of Molecular Weight-Core protein remnants were removed from the HS by alkaline borohydride elimination in 1 ml of 50 mM sodium hydroxide/1 M sodium borohydride incubated at 37°C for 24 h. The pH was neutralized by careful addition of glacial acetic acid using phenol red as an indicator. The HS was then fractionated on a Sepharose CL-6B column (Pharmacia: 1.5 ϫ 90 cm) in 0.2 M NH 4 HCO 3 at 10 ml/h. The V o marker was blue dextran 2000 (M r ϳ 2 ϫ 10 6 ), and the V t marker was sodium dichromate (M r 294).
The relationship between the K av and the molecular weight of the GAG chain has been previously determined (21), and the nomogram in the latter paper was used to calculate the average molecular weight of an HS chain. Alkaline borohydride was only used in the preparation of HS for the estimation of chain length and not for the other analytical techniques, to avoid the loss of any alkali-sensitive sulfate groups on the HS chain.
Heparinase I and Heparinase III (Heparitinase) Scission-HS was dissolved in heparinase buffer (0.1 M sodium acetate, 0.1 mM calcium acetate, 0.1 mg/ml bovine serum albumin, pH 7, at 37°C) containing 50 g of unlabeled HS. Heparinase I or heparinase III (20 mIU/ml) were added three times over 18 h at 37°C, and the progress of the digestion was followed by monitoring A 232 .
The distribution of heparinase I-and heparinase III-sensitive sites were individually determined by gel filtration chromatography through a 1 ϫ 130-cm Bio-Gel P-10 (Bio-Rad) column pre-equilibrated with 0.2 M NH 4 HCO 3 at 4 ml/h. The susceptibility of a sample of HS to an enzyme was calculated by the formula, percentage of disaccharides cleaved ϭ ⌺[([ 3 H]GlcNR n )/n] ϫ 100/total cpm, where [ 3 H]GlcNR n is the number of counts in the area under a peak for a particular oligosaccharide of length n disaccharides. The average molecular weight of heparinase I-resistant HS was determined by gel filtration chromatography of the digested HS on Sepharose CL-6B as described above.
Nitrous Acid Scission-Freshly prepared low pH nitrous acid was made according to Shively and Conrad (22). Nitrous acid (100 l) was added to 10 l of sample in water. The reaction was allowed to proceed for 60 min at room temperature and then terminated by the addition of 2 M Na 2 CO 3 with phenol red as the indicator. The digested material was mixed with 3 l of saturated sodium dichromate solution and 100 l of saturated bovine hemoglobin (V o and V t markers, respectively) and loaded onto a 1 ϫ 130-cm Bio-Gel P-10 column pre-equilibrated with 0.2 M NH 4 HCO 3 at a flow rate of 4 ml/h, and 1 ml fractions were collected (22).
The radioactivity elution profile for [ 3 H]glucosamine label was used to calculate the degree of N-sulfation of HS. To analyze the nitrous acid-cleaved material further, those fractions corresponding to the disaccharides and tetrasaccharides were separately pooled and exhaustively lyophilized. The disaccharides were reduced by incubation in 2 M NaBH 4 , 50 mM NaOH at 37°C for 45 min so that they could be identified by comparison with known reduced standards. Disaccharides and tetrasaccharides from nitrous acid scission were separately applied to two Propac PA1 strong anion-exchange (SAX) HPLC columns connected in series to a Dionex HPLC system. The tetrasaccharides were eluted by a continuous gradient from 0 -1 M NaCl, pH 3.5, whereas the disaccharides were eluted with a biphasic gradient (0 -120 mM, 0 -20 min; 120 -1,000 mM, 21-60 min). The latter method allowed the separation of IdceA(2S)-aMan R and GlcUA(2S)-aMan R disaccharides.
Enzymic Depolymerization and Analysis of Composition by SAX-HPLC-HS was completely depolymerized by incubation with heparinases I, II, and III (20 mIU/ml each) for 24 h at 37°C. The activity of each enzyme against HS (0.5 mg/ml) was measured separately by monitoring the increase in absorbance at A 232 .
The digested material was loaded onto a 1 ϫ 100-cm Bio-Gel P-2 column (Bio-Rad) and eluted with 0.2 M NH 4 HCO 3 at a flow rate of 5 ml/h to confirm the completion of chain scission (always 95% digestion). Fractions corresponding to the disaccharides were pooled and extensively lyophilized to remove the NH 4 HCO 3 .
The disaccharides were dissolved in 1 ml of distilled water (pH 3.5), loaded onto a 5-m Spherisorb SAX (Technicol, Stockport, UK) column linked to a Dionex HPLC apparatus, and eluted at 1 ml/min with a linear gradient between 0 and 0.75 M NaCl, pH 3.5. Fractions of 0.5 ml were collected, and their radioactive content was determined. The column was calibrated with HS disaccharides of known composition that were detected by their A 232 using the in-line UV detector in the HPLC apparatus.

RESULTS
Adenoma and carcinoma cells were taken at 80% confluence and metabolically radiolabeled for 48 h with [ 3 H]glucosamine and 35 SO 4 . GAGs were extracted from the cell layer by treatment with Pronase in 1% (v/v) Triton X-100; the culture medium was also digested with Pronase, and the medium and cell layer extracts were then pooled and clarified by centrifugation (see "Experimental Procedures"). The GAGs were fractionated by anion-exchange chromatography, and the HS (identified by its degradation by nitrous acid) eluted in a double-labeled peak at ϳ0.55 M NaCl (not shown). There was no difference in the NaCl concentration required to desorb the HS produced by adenoma and carcinoma cells. The 3 H/ 35 S-labeled HS was cleared of any contaminating chondroitin sulfate and hyaluronan by treatment with chondroitinase ABC. The Sepharose CL-6B elution profiles of intact adenoma and carcinoma HS species were symmetrical, and the position of the peak of elution of each sample corresponded to a chain of average molecular mass 20 kDa, equivalent to 45-50 disaccharides (Fig. 1).
Low pH Nitrous Acid Scission of HS-The adenoma and carcinoma HS chains were cleaved with nitrous acid, which breaks HS at the glycosidic bond between N-sulfated glucosamine and uronic acids. The products of scission were eluted by Bio-Gel P-10 gel filtration, and the chromatographs (Fig. 2, a  and b) show that the major low molecular mass products were either disaccharides derived from the S domains (contiguous N-sulfated disaccharides) or tetrasaccharides of structure hexuronic acid-GlcNAc-GlcUA-aMan R , where the aMan R (anhydromannitol) is derived from GlcNS in the HS chains. The 35 S in these peaks represents O-sulfated sugars in the tetrasaccharides, and O-sulfated sugars plus inorganic 35 S, released from the N-sulfate groups, in the disaccharide peak.
The nitrous acid profiles (Fig. 2, a and b) can be used to calculate the degree of N-sulfation of HS (3,25). The adenoma HS was more susceptible to nitrous acid scission (37% disaccharides) than the carcinoma HS (32% scission), and 16% of the adenoma HS chain and 13% of the carcinoma HS were present as contiguous N-sulfated sequences (Table I), as revealed by the level of 3 H radioactivity in the disaccharide peaks in Fig. 2,  a and b). In general the data also suggest that the adenoma and carcinoma HS chains maintain a close clustering of O-and N-sulfate groups and that the general distribution of N-sulfated residues is similar in both chains.
The samples of digested adenoma and carcinoma HS gave broadly similar elution profiles, and it was calculated that 69% of the adenoma HS was sensitive to heparinase III scission, whereas 74% of the carcinoma HS was cleaved by the enzyme (Fig. 2, c and d; Table I). One notable difference, however, was that the adenoma HS yielded more hexa-than octasaccharide, but the converse was true in the carcinoma (Fig. 2, c and d).
The heparinase III profiles demonstrate that in both HS species, the majority of the GlcUA-bearing disaccharides occur in contiguous sequences, as revealed by the large disaccharide peaks (Fig. 2, c and d). The fragments that are resistant to heparinase III digestion are the sulfated domains (S domains).
Heparinase I Scission-To gain further information on the structure and location of the S domains, the enzyme heparinase I was used. This enzyme cleaves HS essentially where GlcNS(Ϯ6S)-IdceA(2S) residues occur (23,24), although the enzyme is active against glucuronate-2-O-sulfate (28), a rare constituent in HS (29,30). Scission of the intact adenoma and carcinoma chains with heparinase I and elution of the products by Sepharose CL-6B gel filtration (Fig. 1) showed that the K av of the main peak of eluted material was 0.7, suggesting that the average molecular weight of heparinase I-resistant fragments was 7,000 (21).
Smaller fragments were also produced by heparinase I, and these were examined by chromatography on Bio-Gel P-10. The results showed that the adenoma HS gave a significantly higher yield of low molecular weight products (disaccharides and tetrasaccharides) than the carcinoma HS (Fig. 2, e and f). Although 13% of the hexosaminidic linkages in the adenoma HS were susceptible to heparinase I, only 7% of such linkages were cleaved in the carcinoma HS. In addition, the chromatographs suggest that approximately 4% of the adenoma HS disaccharides were present as contiguous heparinase I-sensitive disaccharides, whereas the corresponding figure for the carcinoma HS was about 2% (Fig. 2, e and f; Table I). This was a consistently observed difference and was noted in three separate experiments.
Total Disaccharide Composition-HS from adenoma and carcinoma cells was completely depolymerized with heparinases I, II, and III. The disaccharides were separated by strong anionexchange HPLC, and the results are shown graphically (Fig. 3,  a and b) and numerically (Table II). The major disaccharide in both HS species was ⌬UA-GlcNAc, which comprised just over half the total disaccharide units; the ⌬UA-GlcNSO 3 was also a prominent constituent (Table II). However, a number of reproducible differences were present in the O-sulfated disaccharides. In particular, progression to carcinoma was associated with an increase in ⌬UA-GlcNAc(6S) but a reduction in ⌬UA(2S)-GlcNS. The latter disaccharide is most likely to contain IdceA(2S) in the HS chain and to be located in the S domains, where it will be cleaved by heparinase I. The findings on the content of IdceA(2S) in the two HS species are compatible with the reduced heparinase I sensitivity in the carcinomaderived material noted earlier (Fig. 2, e and f).
The data also show that the average number of sulfate groups per 100 disaccharides in the adenoma and carcinoma HS was 73 and 61, respectively. The lower sulfation in the carcinoma HS was due to a 20% reduction in N-sulfation and a 33% reduction in 2-O-sulfation, the observed 24% increase  2. Bio-Gel P-10 gel filtration chromatographic analysis of adenoma HS (a, c, and e) and carcinoma HS (b, d, and f) after  scission with nitrous acid (a and b), heparinase III (c and d), and heparinase I (e and f). The insets in e and f represent enlargements of the 3 H radiolabel profiles in each graph. Solid line, 3 H label; dotted line, 35 S label; dp, degree of polymerization; dp2, disaccharides, dp4, tetrasaccharides; dp6, hexasaccharides. (Fig. 3, c and d). This component corresponds to the ⌬hexuronic acid(2S)-GlcNS unit in the heparinase-released disaccharides ( Table II). Disaccharides of structure GlcUA(2S)-aMan R (6S) that elute after IdceA(2S)-aMan R (6S) (peak 5) on SAX-HPLC were not detected. The data indicate that the reduction in 2-O-sulfation on transformation occurs chiefly in the S domains. The major 6-O-sulfated unit in the nitrous acid-released disaccharides was IdceA(2S)-aMan R (6S), and only very small amounts of the mono-6-O-sulfated units GlcUA/IdceA-aMan R (6S), were present (Fig. 3, c and d). This indicates that the majority of the ⌬UA-GlcNS (6S) detected in the heparinase digests (Table II) must be present in the tetrasaccharides released by nitrous acid, which correspond to the mixed sequences. Because of the close coupling of N-and O-sulfation in HS, the ⌬UA-GlcNAc (6S), which represents 12% of adenoma and 16% of carcinoma HS disaccharides, will also be present mainly in the mixed sequences, which thus contain significantly more 6-O-sulfates than the contiguous N-sulfated regions. Combining the data in Table II and Fig. 3, c and d, we can calculate that approximately 70 and 80% of the total 6-Osulfates are present in the mixed sequences in the adenoma and carcinoma HS, respectively.
Analysis of Nitrous Acid-derived Tetrasaccharides-The structure of the tetrasaccharide products of nitrous acid scission were analyzed by SAX-HPLC and shown to contain non-, mono-, and disulfated species (Fig. 3, e and f). 35 S radiolabel in these peaks is O-sulfate that is mainly present at C-6 of Glc-NAc and C-6 of aMan R (see "Discussion"). The sulfation of each peak was based on the ratio of 3 H/ 35 S. The results showed that the adenoma-derived nitrous acid-resistant tetrasaccharides consisted mainly of non-and mono-O-sulfated structures (Fig.  3e), whereas the carcinoma contained a significantly higher proportion of monosulfated structures, with some di-O-sulfated material also present (Fig. 3f). These differences were reproduced in three experiments. The results confirm that variation in O-sulfation of the mixed sequences is an important distinguishing feature between adenoma and carcinoma HS. DISCUSSION This paper describes a detailed analysis of the changes in structure of HS in a model of human colon carcinogenesis in vitro (20). The study shows that HS from cultured adenoma and carcinoma cells share a common molecular organization in the form of similar chain lengths and spacing of S domains (Fig.  1). When HS was degraded with either low pH nitrous acid or heparinase III (Fig. 2), the results showed that many of the linkages susceptible to these reagents were contiguous especially in the case of heparinase III (Fig. 2). The data indicate that the HS from both samples conform to a domain structure that is characteristic of the HS family (4,31). The adenoma and carcinoma HS were approximately 45-50 disaccharides in length (20 kDa), and the fragmentation by heparinase I to yield a major peak of resistant material with a molecular mass of 7 kDa (Fig. 1) implied that the HS chains contained two evenly spaced S domains.
Progression to malignancy was associated with specific structural changes, including a reduction in 2-O-sulfation in the S domains and an increase in 6-O-sulfation in the mixed sequences, which consist of alternate N-sulfated and N-acetylated disaccharides. Previous studies have suggested that Osulfates in these sequences are mainly present on C-6 of the amino sugars (32), and we assume that this is the case in the HS studied here. Structural changes in the mixed sequences have been reported before in murine models of transformation. These studies reported that tumor cells synthesized HS with a reduced content of 6-O-sulfate (15-17) and 3-O-sulfate (18) in the mixed sequences. Our results confirm that these regions are a major target of structural change in carcinogenesis. However, the data show that malignant transformation of human adenomas is accompanied by an increase in mixed sequence 6-O-sulfation mainly associated with GlcNAc residues (Table  II). This increase is also revealed by the high content of mono-O-sulfated and di-O-sulfated tetrasaccharides in the nitrous acid scission products (Fig. 3, e and f).
The biosynthesis of HS is assumed to follow the heparin pathway of polymer modification in which the addition of 2-Osulfates to iduronic acid residues occurs before 6-O-sulfation of GlcNS/GlcNAc (1,33,34). The composition of the S domains is compatible with this sequence of events, in that most of the 6-O-sulfate in these domains occur in disaccharides bearing 2-O-sulfate moieties (Fig. 3, c and d). However, in the adenoma HS and particularly in the carcinoma HS, there are more 6-O-sulfate residues in the flanking sequences than in the S domains. Taken in conjunction with the reduction in 2-O-sulfation in the carcinoma HS, the data suggest that the 6-Osulfotransferases that act on the mixed sequences operate independent of the presence of 2-O-sulfate groups, in accord with findings on the structure of HS produced by a Chinese hamster ovary cell mutant defective in 2-O-sulfotransferase (35). It is interesting that in other species of HS the content of 6-Osulfates is higher in the mixed sequences than in the S domains (28).
The data also indicate that there is a reduction in 2-Osulfation after malignant transformation (Table II; Fig. 2, e and f). Since the 2-O-sulfotransferase has a similar amount of iduronate (Table I, heparinase III-resistant fraction) to act on in both HS species, the implication is that the enzyme is acting more efficiently in the adenoma than the carcinoma.
The data presented above allow the construction of a simplified composite model of the structure of HS produced by the adenoma and carcinoma cells (Fig. 4). The chains are shown to contain two evenly spaced S domains with internal sites of cleavage for heparinase I. The variations in 2-O-sulfation and 6-O-sulfation in the S domains and mixed sequences of the two HS species are also illustrated. The proportions of material contained in the di-and tetrasaccharide peaks of heparinase I-released fragments (Fig. 2, e and f) were used to calculate the number of heparinase I scission sites in the S domains. These data suggested that, on average, heparinase I cleavage of an S domain in the adenoma HS would liberate one disaccharide and one tetrasaccharide, whereas the same treatment of a carcinoma HS S domain would release either a disaccharide (as shown in Fig. 4) or a tetrasaccharide.
In conclusion, the findings from the present study indicate that the overall molecular architecture of HS is preserved in colon adenoma and carcinoma cells, although distinct changes are imposed on this structure in the malignant cell through changes in the content and pattern of sulfation, particularly the O-sulfates. The data also imply that 2-O-sulfation and 6-O-sulfation may be differentially regulated. The significance of these changes for the malignant phenotype is unclear, but in view of the role of HS in controlling cell growth and adhesion, it seems reasonable to assume that the structural modifications in the carcinoma HS may promote a more proliferative or invasive type of behavior.