Characterizing the Non-reducing End Structure of Heparan Sulfate*

The reducing end of heparan sulfate has been known for a long time, but information on the non-reducing end has been lacking. Recent studies indicate that the non-reducing end of heparan sulfate might be the place where fibroblast growth factor signaling complex forms. The non-reducing end also changes with heparanase digestion and, thus, might serve as a marker for tumor pathology. Using high performance liquid chromatography-coupled mass spectrometry, we have identified and characterized the non-reducing end of bovine kidney heparan sulfate. We find that the non-reducing end region is highly sulfated and starts with a glucuronic acid (GlcA) residue. The likely sequence of the non-reducing end hexasaccharides is GlcA-GlcNS6S-UA±2S-GlcNS±6S-Ido2S-GlcNS±6S (where GlcNS is N-sulfate-d-glucosamine, S is sulfate, UA is uronic acid, and Ido is iduronic acid). Our data suggests that the non-reducing end of bovine kidney heparan sulfate is not trimmed by heparanase and is capable of supporting fibroblast growth factor signaling complex formation.

Heparan sulfate (HS) 2 is a long linear polysaccharide attached to core proteins in the extracellular matrix and the surface of nearly all mammalian cells. The nascent HS chain consists of disaccharide repeats of glucuronic acid (GlcA) and N-acetyl-D-glucosamine (GlcNAc) (1)(2)(3). During the process of maturation, several modifications can occur on the HS chain, which include N-deacetylation and N-sulfation of GlcNAc, epimerization of GlcA to L-iduronic acid (IdoA), 2-O sulfation of IdoA, and 6-O and 3-O sulfation of the glucosamine (GlcN). These modifications usually focus on separate regions and result in domain structures (4 -6). It is the modified domains to which various extracellular proteins bind (2,3). The region at the reducing end of HS is usually unmodified (5,(7)(8)(9). One major function of HS is to interact with both fibroblast growth factor (FGF) and its cognate receptor and promote the FGF signaling complex formation (10 -13). Defects in HS can cause complete losses of FGF as well as Hedgehog and Wingless signaling pathways and lead to severe abnormality in embryonic development (14,15). HS also plays important roles in cell migration and cancer cell metastasis (16).
Despite these known structural information and biological functions of HS, the non-reducing end (NRE) of HS has not been studied. Assuming that each HS chain has a similar content of sulfation, the fact that the reducing end of HS is devoid of sulfation (7)(8)(9)17) suggests that the NRE of HS might be sulfated and assume biological functions. Indeed, there are indications that HS uses its NRE to direct the FGF signaling complex formation (11,13,18). Consequently, an FGF signaling complex model where two molecules each of FGF and FGF receptor bind to the NREs of two approaching HS chains has been proposed (13). Whether the NRE of HS has the ability to support the complex formation relies on its structure.
It is known that heparanase activity is closely related to the metastatic potential of tumor-derived cells (19). Because heparanase can trim off the original NRE of cell surface HS, this correlation might be caused by changing the NRE of cell surface HS. Heparanase is an endo-glucuronidase, and its action on HS will release oligosaccharides from the NRE and expose a glucosamine residue at the newly formed NRE (20). This change on the NRE of HS may possibly be detected by mass spectrometry, because different end structures could have different molecular masses.
The identification of the NRE of HS may also allow us to establish a new strategy to sequence HS. The positions of internal residues on a HS chain can be assigned based on their distances to the NRE. Previously, HS sequencing strategies were established on purified heparin/HS oligosaccharides (21)(22)(23)(24). One shortcoming of these strategies is that the localization of these oligosaccharides on the whole chain of HS could not be pinpointed. Here, we describe the identification and characterization of the NRE of bovine kidney HS.
Stable Isotope Labeling of Heparan Sulfate-34 S incorporation by 6-OST-1 was done as described previously (26). The labeling buffer (2ϫ) contained 50 mM MES (pH 7.0), 1% (w/v) Triton X-100, 5 mM MgCl 2 , 5 mM MnCl 2 , 2.5 mM CaCl 2 , 0.075 mg/ml protamine chloride, and 1.5 mg/ml bovine serum albumin. To label HS with 34 S, 10 g of HS, 10 l of labeling buffer, 1 l of 6-OST-1, 2 l of 34 S-labeled 3Ј-phosphoadenosine 5Ј-phosphosulfate (3 mM), and an appropriate amount of water were mixed to the total volume of 20 l. The reaction was incubated for 2 h at 37°C. The modified HS was purified on a DEAE column. * This work was supported in part by National Institutes of Health Grants 1PO1 HL66105 and 5RO1 HL59479. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. □ S The on-line version of this article (available at http://www.jbc.org) contains supplemental Fig. 1 (on the inability of the non-reducing end disaccharide to be labeled by 6-OSTs) and supplemental Fig. 2  Heparan Sulfate Lyase Digestion-For analyzing HS end structures, HS samples with or without 34 S labeling were digested thoroughly with heparinase, heparitinase I, and heparitinase II individually or in combination at 37°C for 2 h. For each case, 10 g of an HS sample was digested in 20 l of a buffer containing 40 mM NH 4 Ac (pH 7.0), 1 mM CaCl 2 , and 1 milliunit each of the enzymes. One g of the digestion product was then used for LC/MS analysis.
Heparan Sulfate Oligosaccharide ␤-Glucuronidase Digestion-Briefly, 10 g of HS lyase digest was treated with 1 l of ␤-glucuronidase (50 units) in 15 l of digestion buffer (0.05 mM sodium acetate, pH 5.0, and 0.1 mg/ml bovine serum albumin). The reaction was incubated overnight at 37°C.
Capillary Liquid Chromatography and Mass Spectrometry-High performance liquid chromatography and mass spectrometry analysis was described previously (26,27). Separations were performed on an Ultimate capillary high performance liquid chromatography work station (Dionex, Sunnyvale, CA) employing dibutylamine as an ion-pairing agent. A gradient elution was performed using a binary solvent system composed of water (eluent A) and 70% aqueous methanol (eluent B), both containing 8 mM acetic acid and 5 mM ion-pairing agent. High performance liquid chromatography separations were performed on a 0.3 ϫ 250-mm C18 column at a rate of 5 l/min and a sample volume of 6.3 l. The elution profile was 0% eluent B for 5 min, 6% eluent B for 19 min, 18% eluent B for 17 min, 34% eluent B for 13 min, and 55% eluent B for 16 min. After each run, the column was washed with 90% eluent B for 15 min and equilibrated with 100% eluent A for 28 min.
Mass spectra were acquired on a Mariner BioSpectrometry work station electrospray ionization time-of-flight mass spectrometer (PerSeptive Biosystems, Framingham, MA). Nitrogen was used as a desolvation gas as well as a nebulizer. Conditions for electrospray ionization-mass spectrometry were as follows: nebulizer gas (N 2 ) flow rate, 1 liter/min; nozzle temperature, 140°C; drying gas (N2) flow rate, 0.6 liters/min; spray tip potential, 2.8 kV; and nozzle potential, 70 V. Negative ion spectra were acquired every 4 s by scanning m/z from 40 to 4000. Total ion chromatograms and mass spectra were processed with the Data Explorer software version 3.0.

Theoretical Calculation of the Monoisotopic m/z Value of a Terminal Oligosaccharide Released from the Non-reducing End of an HS chain by
Bacterial Lyases-An oligosaccharide released from internal sections of an HS chain with bacterial lyases begins with a ⌬ 4,5 -uronic acid (⌬UA); however, the terminal oligosaccharide released from the NRE of an HS chain should begin with a saturated residue (Fig. 1). If the NRE of an HS chain is a uronic acid (UA), the molecular mass of the terminal oligosaccharide will be 18.01 Da (addition of H 2 O 1 ) more than that of its ⌬UA-containing internal counterpart; if the NRE of an HS chain is a GlcN, the molecular mass of the terminal oligosaccharide will be at least 181.09 Da (addition of C 6 H 15 O 5 N 1 ) more than that of the ⌬UA-containing internal counterpart (Fig. 1). In negative mode mass spectrometry, the molecular formula of a ⌬UA-containing oligosaccharide can be identified with Equation 1 (26), where n is the number of disaccharides, p is the number of acetyl groups, and q is the number of sulfates. The molecular formula of a terminal oligosaccharide with a UA at its NRE can be identified with Equation 2, and the molecular formula of a terminal oligosaccharide with a GlcN at its NRE can be identified with Equation 3.
Identification of the Non-reducing End of Bovine Kidney HS-A bovine kidney HS sample was first subject to complete digestion with HS lyase mixture and then analyzed with LC/MS (Fig. 2). Most of the m/z values in the integrated mass spectrum had solutions with Equation 1 (TABLE ONE) and therefore should belong to internal oligosaccharides. Only m/z 514.05 had a solution with Equation 2, suggesting that it belongs to a terminal oligosaccharide with a UA at the NRE. This terminal oligosaccharide was then designated as a dp2-2S* to distinguish it from internal disaccharides dp2-2S at m/z 496.04 (Fig. 2B). Because the NRE UA had no preceding glucosamine residue that was required for its modification (1), this UA was likely a GlcA. The following GlcN residue could not be labeled with 6-OSTs on the whole chain of HS (supplemental Fig. 1, available in the on-line version of this article), suggesting that it might have been sulfated at its 6-O position. Also, considering the rareness and positioning of 3-O-sulfate and free amino group (28,29), this dp2-2S* was likely to be GlcA-GlcNS6S. No m/z value was found to have solution with Equation 3, suggesting no HS chain terminated on GlcN. Because the dp2-2S and dp2-2S* had similar chemical structures, the ratio of their m/z signals would quantitatively reflect their molar ratio. This ratio was found to be ϳ15 (Fig. 2B), suggesting each bovine kidney HS chain contains ϳ15 dp2-2S.
Confirmation of the Non-reducing End by Exo-glycosidase Treatment -To confirm that the dp2-2S* did begin with a GlcA residue, the above HS lysate was further digested with ␤-D-glucuronidase and analyzed with LC/MS. An extracted ion current chromatogram was obtained at m/z 514.05 and 576.00 from the total ion chromatogram (Fig. 3). m/z 576.00 belonged to internal trisulfated disaccharides dp2-3S and caused two peaks in the extracted ion current chromatogram. m/z 514.05 caused a single peak at elution time of 37.6 min. Because the dp2-3S began with ⌬UA and was resistant to ␤-D-glucuronidase digestion, it could serve as a control. The ␤-D-glucuronidase treatment eliminated the peak of dp2-2S* but not the peaks of the dp2-3S (Fig. 3B), indicating that the dp2-2S* did start with a GlcA residue.
Among these terminal tetrasaccharides, dp4-4S* accounted for the majority and had 34 S incorporation (Fig. 4B). To see if dp4-4S* contained a single compound, extracted ion current at m/z 505.05 Ϯ 0.10 was obtained. Two peaks at 63.9 and 64.3 min were found, indicating FIGURE 2. Identifying the NRE terminal disaccharide of bovine kidney HS. A, total ion current chromatogram (TIC) of a bovine kidney HS digest separated on a reverse phase C-18 column. Bovine kidney HS was first digested with a mixture of heparinase, heparitinase I, and heparitinase II and then analyzed with LC/MS. The NRE terminal disaccharide of dp2-2S* caused a peak at elution time 37.6 min. B, integrated mass spectrum of the entire separation. Each m/z peak could be assigned to certain oligosaccharides (TABLE ONE). The monoisotopic peak at m/z 514.05 was assigned to a NRE terminal disaccharide of dp2-2S*. The monoisotopic peak at m/z 496.04 was assigned to internal disaccharides of dp2-2S. The isotopic cluster of dp4-1Ac-4S (at m/z 517.05) happened to be in close proximity to that of dp2-2S*.   OCTOBER 7, 2005 • VOLUME 280 • NUMBER 40 that dp4-4S* had two structural isomers (Fig. 5A). Only the isomer of the second peak had 34 S incorporation. Because the first disaccharide from the NRE of bovine kidney HS could not be labeled by 6-OST-1 (supplemental Fig. 1), the 34 S labeling should occur at the 6-O position of the second disaccharide. Considering the rareness and unique positioning of free amino group and 3-O-sulfate on HS (28,29), the 34 Slabeled dp4-4S* isomer was likely to be GlcA-GlcNS6S-UA-GlcNS6S, and the unlabeled dp4-4S* isomer was likely to be GlcA-GlcNS6S-UA2S-GlcNS. The dp4-3S* had a single peak in its extracted ion chromatogram (data not shown) and was likely to have a single structure of GlcA-GlcNS6S-UA-GlcNS. Only trace amounts of dp4-1Ac-3S* and dp4-5S* were detected. Because heparitinase II was known to have overlapping substrate specificity with heparinase, the heparitinase II-digested HS sample was also examined in parallel. Both enzymes were able to release dp2-3S, but otherwise they were different. The major NRE oligosaccharides released by heparitinase II and heparinase were dp2-2S* and dp4-4S*, respectively (Fig. 6). These data suggest that the two enzymes have overlapping but distinct substrate specificities and thus are able to release different NRE oligosaccharides. .03 corresponded to dp4-3S*, dp4-1Ac-3S*, dp4-4S*, and dp4-5S* respectively (TABLE  TWO). dp4-4S* accounted for Ͼ80% of the total end structures and showed significant 34 S incorporation.

TABLE TWO
The oligosaccharides in Fig. 4 The monoisotopic m/z values were related to the numbers of charges (z), disaccharides (n), acetyl groups (p), and sulfates (q) with Equation1, Equation

DISCUSSION
We have identified and characterized the NRE structures of bovine kidney HS. The first disaccharide of bovine kidney HS was likely to be GlcA-GlcNS6S. The second disaccharide was variable and could be UA-GlcNS6S, IdoA2S-GlcNS, or UA-GlcNS. Because the heparinase that released tetrasaccharides from the NRE has a preference to cleave the glycosidic bond between GlcNS and IdoA2S-GlcNSϮ6S (30), the third disaccharide was likely to be IdoA2S-GlcNSϮ6S (Fig. 7). When Chinese hamster ovary cell HS and porcine intestine HS were examined, the first disaccharides were found to be dp2-2S* too (supplemental Fig.  2, available in the on-line version of this article). Overall, limited sequence variation was observed at the NRE of bovine kidney HS, which suggests that the sequences at the NRE of bovine kidney HS are probably strictly controlled.
It is not known if this restricted sequence variation is an overall theme of the whole chain of bovine kidney HS. The sequences of sulfated FIGURE 5. The major end structure dp4-4S* contained two isomers. A, ion current chromatogram (XIC) extracted at m/z 505.05 Ϯ 0.10 from the total ion current chromatogram of Fig. 4 and the corresponding mass spectra. Two overlapping peaks at 63.9 and 64.3 min were observed. The second peak exhibited 34 S incorporation. B, simulated m/z cluster of dp4-4S* according to natural isotope abundance. FIGURE 6. Heparitinase II and heparinase released different NRE structures. These two enzymes both released trisulfated disaccharide but were very different in the composition of the oligosaccharides that they released, reflecting similar but distinct substrate specificities. A, total ion current (TIC) chromatogram of heparitinase II-digested, 34 S/6-OST-1-labeled bovine kidney HS sample and corresponding integrated mass spectrum. The major NRE oligosaccharide was dp2-2S*. B, total ion current chromatogram of heparinase-digested, 34 S/6-OST-1-labeled bovine kidney HS sample and corresponding integrated mass spectrum. The major NRE oligosaccharides were dp4-4S*. domains of fibroblast HS were shown to be strictly limited (31). The reducing end regions of HS samples from various sources were shown to be unmodified (5,(7)(8)(9). These findings together suggest that HS could have limited sequences.
Great technical progress has been made in HS sequencing and structural determination in recent years (21,24,26,(31)(32)(33), but HS whole chain sequencing has not been attempted yet. The identification of the NRE of HS may open a door to this avenue. The NRE of an HS is a logical starting point for reading sequences on an HS chain. The NRE oligosaccharides can be easily distinguished from various internal oligosaccharides because of their unique molecular masses. LC/MS also allows us to monitor HS oligosaccharides without radioisotope or chemical labeling. With these advantages, it might be possible for us to sequence the HS whole chain in the future. Current study represents an initial effort on sequencing HS whole chain from its NRE.
Interestingly, only GlcA residue was observed at the NRE of bovine kidney HS. A GlcN residue was expected at the NRE of an HS chain (Fig.  1C) if the chain had gone through heparanase digestion in vivo, because heparanase is an endo-glucuronidase that cleaves glucuronidic linkages (20,34). The fact that no GlcN residue was observed the NRE of bovine kidney HS suggests that heparanase activity in bovine kidney was negligible or had different subcellular localizations. Consistently, heparanase was shown to be localized mainly within the lysosome and to be responsible for HS turnover under normal condition (35,36). Recently, secreted heparanase protein was detected in tumor tissues and was related to the metastatic potential of tumor-derived cells (19). It will be interesting to see if HS from tumor tissues or tumor-derived cells has GlcN at its NRE. In this sense, the NRE of HS could serve as a convenient marker for secreted heparanase activity and, thus, tumor pathology.
Based on crystal structure (11) and biochemical data (13,18), an end model for FGF signaling complex formation at the NREs of two approaching HS chains had been proposed previously (13). Our current data prove that a hexasaccharide at the NRE of bovine kidney HS contains 5-7 sulfates. Among them, three are N-sulfates, one or two are 2-O-sulfates, and at least one is a 6-O-sulfate. These sulfates are pivotal and may be sufficient (11,13,37) for FGF signaling complex formation; therefore, our structural information on the NRE of HS provides strong evidence to support the end model of FGF signaling complex (13). The fact that bovine kidney HS contains a highly sulfated NRE and a nonsulfated reducing end (9) also suggests a sulfation gradient on HS chain from the reducing end to the NRE. This gradient may facilitate the movement of the binding protein toward the NRE on HS and explains the mechanism of FGF signaling complex formation (13). In this way, HS can be considered as a regulator of molecular encounter for FGF and its receptor (38).
One concern about the current study is whether we have covered all possible NRE structures. If some NRE structures were resistant to lyase digestion and could not be released as short oligosaccharides, they could have been missed in our detection. Our current mass spectrometrycoupled high performance liquid chromatography cannot separate large and highly sulfated oligosaccharides such as dp6-8S to dp6-12S, dp8-8S to dp8-16S, dp10-7S to dp10-20S et al. Because these large and highly sulfated oligosaccharides are either unlikely to occur on HS or unlikely to resist to lyase digestion, we think that we have covered most if not all possible NRE structures. FIGURE 7. Major possible end structures of bovine kidney HS. The first disaccharide was conserved. The second disaccharide was variable. The third disaccharide was proposed based on the substrate specificity of heparinase, which cleaved the linkage between the second and third disaccharide. The size of the reducing end NA domain was estimated based on previous study on HS prepared from bovine kidney glomerular basement membrane (9).