Construction of a Chondroitin Sulfate Library with Defined Structures and Analysis of Molecular Interactions*

Background: Chondroitin sulfate (CS) is a linear polysaccharide, composed of repeating disaccharide units and modified with sulfate groups at various positions. Results: A CS library was constructed with defined structures using CS polymerase and sulfotransferases. Conclusion: The CS library provided details of interactions with CS-binding molecules. Significance: Chemo-enzymatic synthesis provides a useful tool for studying the biological functions of CS. Chondroitin sulfate (CS) is a linear acidic polysaccharide, composed of repeating disaccharide units of glucuronic acid and N-acetyl-d-galactosamine and modified with sulfate residues at different positions, which plays various roles in development and disease. Here, we chemo-enzymatically synthesized various CS species with defined lengths and defined sulfate compositions, from chondroitin hexasaccharide conjugated with hexamethylenediamine at the reducing ends, using bacterial chondroitin polymerase and recombinant CS sulfotransferases, including chondroitin-4-sulfotransferase 1 (C4ST-1), chondroitin-6-sulfotransferase 1 (C6ST-1), N-acetylgalactosamine 4-sulfate 6-sulfotransferase (GalNAc4S-6ST), and uronosyl 2-sulfotransferase (UA2ST). Sequential modifications of CS with a series of CS sulfotransferases revealed their distinct features, including their substrate specificities. Reactions with chondroitin polymerase generated non-sulfated chondroitin, and those with C4ST-1 and C6ST-1 generated uniformly sulfated CS containing >95% 4S and 6S units, respectively. GalNAc4S-6ST and UA2ST generated highly sulfated CS possessing ∼90% corresponding disulfated disaccharide units. Sequential reactions with UA2ST and GalNAc4S-6ST generated further highly sulfated CS containing a mixed structure of disulfated units. Surprisingly, sequential reactions with GalNAc4S-6ST and UA2ST generated a novel CS molecule containing ∼29% trisulfated disaccharide units. Enzyme-linked immunosorbent assay and surface plasmon resonance analysis using the CS library and natural CS products modified with biotin at the reducing ends, revealed details of the interactions of CS species with anti-CS antibodies, and with CS-binding molecules such as midkine and pleiotrophin. Chemo-enzymatic synthesis enables the generation of CS chains of the desired lengths, compositions, and distinct structures, and the resulting library will be a useful tool for studies of CS functions.

CS chains are synthesized onto a linkage tetrasaccharide, GlcUA-Gal-Gal-Xyl, covalently bound to the serine residues of a core protein, by the alternating addition of monosaccharide units of GalNAc and GlcUA by CS synthases (2). During polymerization, the chain undergoes sulfation at various positions, mediated by a variety of sulfotransferases (3). The chondroitin 4-sulfotransferase (C4ST) family transfers sulfate to the 0S unit of chondroitin and generates CS containing the 4S unit (4). Chondroitin 6-sulfotransferase (C6ST) transfers sulfate to the 0S unit and generates the 6S unit (5). N-Acetylgalactosamine 4-sulfate 6-sulfotransferase (GalNAc4S-6ST) catalyzes further sulfation from the 4S unit to generate the diSE unit (6). Uronosyl 2-sulfotransferase (UA2ST) transfers sulfate to an uronic acid residue at the 2-O-position and generates diSD and diSB units from 6S and IdoUA-GalNAc(4S) units, respectively (6). Enzymes that generate a triS unit have not been identified yet. These diverse and complex structures of CS chains are synthesized with these enzyme complexes in the Golgi apparatus.
In animals, CS chains are present as a part of a proteoglycan molecule and play various biological roles in development, organ morphogenesis, inflammation, and infection by interacting with cytokines and growth factors, and regulating their signal transduction (1). For example, both midkine (MK) and pleiotrophin (PT) bind CS and promote neurite outgrowth (7). These functions are mainly ascribed to the structural diversity of oligosaccharide regions containing sulfate group modifications. Anti-CS monoclonal antibodies have been used in studies of CS chains. They are as follows: MO225, thought to strongly recognize diSD unit-rich CS chains (8); CS-56, thought to recognize native CS chains containing 6S units (9); LY111, thought to recognize intact CS chains containing 4S units (10); and 2H6, thought to recognize the native CSA chain (11). These antibodies are useful for CS investigations, but their detailed epitope structures have not been determined yet, except for WF6, which shows high affinity to two different sequences of sulfation, but shares some common structural envelope. Minimum binding sequences were octasaccharides, suggesting four adjacent sulfated disacchrides were involved in defining binding sites (12).
In this study, we chemoenzymatically synthesized CS saccharides with defined lengths and structures, using the chondroitin polymerase K4CP and various recombinant sulfotransferases, and constructed a CS library. We then analyzed interactions between the immobilized glycosaminoglycan conjugates and glycosaminoglycan-binding molecules using an enzyme-linked immunosorbent assay (ELISA) and surface plasmon resonance (SPR) analysis. Our study clearly demonstrates the specificity of CS structure and molecular interactions of CS and provides insight into its potential applications.

Expression and Purification of Recombinant Chondroitin
Sulfotransferases-HEK293T cells were transfected with the pIRESpuro expression plasmids using Lipofectamine 2000 (Invitrogen). Stable clones were selected with 5ϳ10 g/ml of puromycin in Dulbecco's modified Eagle's medium (Invitrogen) containing 10% (v/v) fetal bovine serum (HyClone Laboratories, Logan, UT) and tested for the synthesis and secretion of the recombinant proteins by measuring enzyme activities and Western blotting using an anti-FLAG M2 antibody (Sigma). The enzymes (C4ST-1, C6ST-1, GalNAc4S-6ST, and UA2ST) secreted into the conditioned media were purified by affinity chromatography using an anti-FLAG-agarose gel (Sigma), and eluted with 10 mM Tris-HCl, pH 7.4, containing 150 mM NaCl, 20% glycerol, and 200 g/ml FLAG peptide (Sigma). The recombinant enzyme solutions obtained were stored at Ϫ80°C.
Sulfotransferase Assay-Sulfotransferase activities were assayed by the radioisotope labeling method using [ 35 S]PAPS (PerkinElmer Life Sciences) as described previously (17). The standard reaction mixture in a final volume of 50 l consisted of 50 mM sodium potassium phosphate, pH 6.8, containing 2 mM DTT, [ 35 S]PAPS (1 nmol, 0.1 Ci), CH (20 g, for C4ST-1 and C6ST-1), or CSA (20 g, for GalNAc4S-6ST and UA2ST), and the enzyme. The enzymatic reaction was carried out at 37°C for 60 min and stopped by heating in a boiling water bath for 1 min. The radiolabeled products were isolated by gel filtration using a Superdex Peptide HR10/300 column and quantified by liquid scintillation counting. One unit of enzyme is defined as the amount required for catalyzing the transfer of 1 pmol of sulfate per min.
Preparation of CS-hexamethylenediamine (HMDA) Conjugates-CS oligo-and polysaccharides (CH6, CH, CSA, CSC, CSD, CSE, and DS) were conjugated with HMDA at the reducing end by the reductive amination method. CS (1 mol) solution in 1.8 M HCl containing 100 mol of HMDA (100 l) was heated at 65°C for 2 h, followed by the addition of sodium cyanoborohydrate (150 mol), and further heating at 65°C for 16 h. The HMDA-conjugated saccharides were purified by gel filtration chromatography on a Superdex 30 HR16/600 column (1.6 ϫ 60 cm) using 0.2 M ammonium acetate as an eluent and obtained with freeze-drying repeated three times. The structure of purified CH6-HMDA was confirmed using a MALDI-TOF MS spectrometer (Autoflex; Bruker, Bremen, Germany) (18). The disaccharide compositions of CS-HMDA were measured using fluorometric post-column HPLC after digestion with chondroitinase ABC or AC II, as described below.
Preparation of sCS-HMDA Library-A mixture of CH6-HMDA (0.2 mol), UDP-GlcUA (6 mol), UDP-GalNAc (6 mol), and K4CP (3.0 units) in 50 mM Tris-HCl, pH 7.2, containing 0.15 M NaCl, and 0.2 mM MnCl 2 was incubated at 30°C for 24 h, and then chromatographed on the Superdex 30 column using 0.2 M ammonium acetate as the eluent. The elution was monitored by UV absorbance at 225 nm. The average molecular weight (M r ) of the synthesized chondroitin polysaccharide HMDA conjugate (sCH-HMDA) was estimated using analytical gel filtration chromatography on a Superose 12 HR10/300 column using short chains of hyaluronan as molecular weight standards (11).
Individual sCS-HMDA conjugates were prepared as follows. For sCSA-HMDA, a solution of 50 nmol sCH-HMDA in 3 ml of 50 mM Tris-HCl, pH 6.8, containing 2 mM DTT, 11.5 mol PAPS, and 2.0 kilounits C4ST-1 was incubated at 37°C for 24 h. For sCSC-HMDA, a solution of 10 nmol sCH-HMDA in 2 ml of the same buffer containing 2.3 mol of PAPS and 400 units of C6ST-1 was incubated at 37°C for 24 h. For sCSAC-HMDA, a solution of 30 nmol sCH-HMDA in 2 ml of the same buffer containing 6.9 mol of PAPS, 720 units of C4ST-1, and 520 units of C6ST-1 was incubated at 37°C for 24 h. Each sulfated chondroitin product was purified using Superdex 30 column chromatography. Using these sulfated CS species, we further prepared sCS-HMDA as shown in Fig Biotinylation of CS-HMDA Conjugates and ELISA-CS-HMDA species were modified using a sulfo-NHS-activated biotin reagent (sulfo-NHS-LC-biotin; Pierce) at the amine group of the HMDA residue at the reducing end of the saccharide chains, and resultant CS-biotin derivatives were immobilized onto streptavidin-coated microplates for ELISA system and on sensor chips for SPR analysis. Solutions of the CS-HMDA conjugates (5 nmol) in 50 mM phosphate buffer, pH 7.4, containing 50 nmol of sulfo-NHS-LC-Biotin were incubated at room temperature for 1.5 h and then further incubated at room temperature for 10 min after 100 nmol of ethanolamine was added to the solutions. The CS-biotin products were purified using Superdex Peptide column chromatography.
SPR Analysis-Interactions between CS-biotin conjugates and the CS-binding molecules MK and PT were analyzed using an SPR biosensor (BIAcore 1000; GE Healthcare). A streptavidin-conjugated sensor chip (SA chip) was used to immobilize the biotin-conjugated CS chains (CH-, CSA-, CSC-, CSD-, CSE-, sCH-, sCSA-, sCSC-, sCSAC-, sCSAD-, sCSE-, sCSDE-, and sCtriS-biotin). The CS-biotin solutions (10 g/ml) in 10 mM phosphate buffer, pH 7.2, containing 0.005% (w/v) Tween 20 (70 l) were injected at a flow rate of 5 l/min to immobilize the CS chains, and then the sensor chip was washed with PBST containing 2 M NaCl to remove unbound material. Resonance unit values for immobilization of CS-biotin species as shown in supplemental Table 2. Binding assays were performed at 25°C at a constant flow rate of 20 l/min during both the association and dissociation phases. MK and PT solutions (50, 80, 150, and 250 nM) in 10 mM HEPES buffer, pH 7.4, containing 0.15 M NaCl, 3 mM EDTA, and 0.005% Tween 20 (HBS-EP, 60 l) were injected into the flow cells at a flow rate of 20 l/min (association), and then HBS-EP without the cytokines was flowed at the same flow rate for 180 s (dissociation). After each run, regeneration of the sensor chip surface was accomplished by an injection of 20 l of HBS-EP containing 2 M NaCl. Changes in resonance units were recorded, and the values for rate constants were determined by nonlinear regression analysis using BIAevaluation software (version 4.1, GE Healthcare). Association rate constants (k a ) were calculated from the linear portions of the sensorgrams during the early association phase. Dissociation rate constants (k d ) were calculated from the early portion of the dissociation phase, which occurs after the completion of sample injection during the washout period. The apparent equilibrium dissociation constants (K d ) were calculated as the ratios of k d /k a .
Compositional Analysis of CS Derivatives-The synthetic CS derivatives were digested with chondroitinase ABC or ACII (10 milliunits) at 37°C for 1 h. Unsaturated disaccharide products were analyzed using fluorometric post-column HPLC system as reported previously (20).
Laser Light Scattering-The synthetic CH polymers were separated on a TSK-gel G4000 pwXL column (Tosoh, Tokyo, Japan) eluted with 0.2 M NaCl at 0.5 ml/min and analyzed using a laser light scattering photometer ((DAWN DSP; Wyatt Technology, Santa Barbara, CA) to determine their absolute average molecular weights (21,22).

Preparation and Characterization of CS Derivatives-Ini-
tially, we conjugated the CH6 oligosaccharide and various CS polysaccharides (CH, CSA, CSC, CSD, CSE, and DS) with HMDA at the reducing ends of the saccharides, as described under "Materials and Methods." The CS-HMDA conjugates have a free amino group at the reducing end, which is useful to conjugate another active group such as biotin and to immobilize the CS chains on activated solid materials. The average molecular weights (M r ), disaccharide compositions, and sulfation degrees (SD, number of sulfate groups per disaccharide unit) of the CS-HMDA conjugates are summarized in Table 1.
The conjugation reactions did not alter the molecular weights or sulfation patterns of the CS chains. The chain size of chemically desulfated CH was smaller than the original CSC, probably due to the partial cleavage of the glycosyl bonds (23 Following this, we enzymatically elongated the carbohydrate chain of CH6-HMDA, using K4CP and two donor substrates   DECEMBER 21, 2012 • VOLUME 287 • NUMBER 52 described under "Materials and Methods." The chain size could be elongated up to 350 kDa by manipulating the conditions of the enzymatic reactions, such as the concentrations of the donor and acceptor substrates, the amount of enzyme, and the reaction time (supplemental Fig. 2).

Construction of a CS Library with Defined Compositions-
We prepared various recombinant chondroitin sulfotransferases from stable clones of HEK293T cells as described under "Materials and Methods." The activities of the recombinant sulfotransferases purified from the culture media from the stable cells were 2,000ϳ3,000 units/ml (C4ST-1), 3,000ϳ6,000 units/ml (C6ST-1), 900ϳ1,600 units/ml (GalNAc4S-6ST), and 300ϳ600 units/ml (UA2ST), respectively. Using these sulfotransferases and sCH-HMDA, we generated various sCS-HMDA species by the sequential addition of sulfate residues at different positions, as shown in Fig. 1B. Disaccharide compositions and the SD values of the sCS-HMDA derivatives are summarized in Table 2.
Upon incubation with 50 nmol of sCH-HMDA, 11.5 mol of PAPS, and 2 kilounits of C4ST-1 at 37°C for 24 h, the product designated as sCSA-HMDA contained up to 95ϳ98% C4S disaccharide units. When a lower amount of C4ST-1 or PAPS was used, the product showed a lower SD (supplemental Fig. 3). Upon incubation with 10 nmol of sCH-HMDA, 2.3 mol of PAPS, and 400 units of C6ST-1 at 37°C for 24 h, the product designated as sCSC-HMDA contained up to 92ϳ99% C6S units. When a lower amount of C6ST-1 or PAPS was used, the product showed a lower SD (supplemental Fig. 3). The simultaneous reaction of C4ST-1 and C6ST-1, upon incubation with 30 nmol of sCH-HMDA, 6.9 mol of PAPS, 720 units of C4ST-1, and 520 units of C6ST-1 at 37°C for 24 h, yielded a product (sCSAC-HMDA) that contained almost the same amounts of 4S (47.9%) and 6S (50.6%) units. By changing the ratio of C4ST-1 to C6ST-1 in the reaction, the ratio of 4S and 6S units in the product could be altered accordingly (supplemental Fig. 4). These results indicate the establishment of bioengineering techniques for the generation of CS chains with defined ratios of monosulfated and non-sulfated disaccharide units.
CS-biotin derivatives were prepared from the CS-HMDA species using the sulfo-NHS-LC-biotin reagent and were immobilized on streptavidin-coated microplates for ELISA and on sensor chips for SPR analysis. The chain sizes and disaccharide compositions of the CS-biotin conjugates did not change from their corresponding CS-HMDA origins.
ELISA-For the ELISA system, the native and synthetic CSbiotin conjugates were immobilized on streptavidin-coated microplates in a dose-dependent manner (0.001ϳ10 g/ml). The amounts of anti-CS monoclonal antibodies (MO225, CS56, LY111, and 2H6) bound were measured using the ELISA system as described under "Materials and Methods" (Figs. 2 and  3). The half-maximal effects (ED 50 ) of the binding activities of the antibodies to the CS derivatives were estimated from the dose response profiles against the concentrations (g/ml) of the immobilized CS derivatives (Table 3).
When we performed ELISA using sCS species, MO225 showed high binding affinity to sCSDE, sCSAD, and sCtriS, with ED 50 values of 0.0058, 0.012, and 0.031 g/ml, respectively, indicating their stronger binding affinity than CSD showing the strongest affinity among native CSs. MO225 showed an affinity for sCSAC that was similar to its affinity for CSD and low binding affinity for cCSE. It did not bind to sCH or sCSA. CS56 showed similar binding affinities to sCSAD, sCSAC, and sCSDE as to CSC. It did not bind to sCH, sCSA, sCSC, sCSE, and sCtriS. LY111 bound to sCSAC, sCSDE, and sCSAD with higher affinity than to CSA. It did not bind to sCH, sCSC, sCSE, and sCtriS. 2H6 had high binding affinity to sCSAC (ED 50 ϭ 0.0036 g/ml), and low affinity to sCSAD (ED 50 , 0.014 g/ml) and sCSDE (ED 50 , 0.023 g/ml) plates. Again, it did not bind to sCH, sCSA, sCSC, sCSE, and sCtriS.
SPR Assay-We then analyzed the interactions of the CSbinding cytokines MK and PT with immobilized CS-biotin derivatives. The sensorgrams of various concentrations of MK and PT against the immobilized CS derivatives are shown in Figs. 4 and 5, respectively. The k a , k d , and K D values of the cytokines for the CS derivatives were calculated using a 1:1 (Langmuir) binding model using the BIAevaluation 4.1 software, and these results are summarized in Table 4.
The K D values of MK for CSD, CSE, and sCSAD were 78, 80, and 99 nM, respectively. These were 1.3-to 1.7-fold lower than that of CSC, and 5-to 12-fold lower than those of CSA, sCSA, sCSC, sCSAC, and sCSDE. The K D values of MK for CH, sCSC, sCSE, and sCtriS were over 1,000 nM, and the sCH chip showed the lowest affinity to MK (K D of 11,200 nM). Although the CSC

DISCUSSION
Recent studies have revealed that CS exhibits various functions via specific binding to physiologically active molecules.
Although native CS sources designated as CSA, CSC, CSD, and CSE are available, their CS disaccharide compositions are not uniform, preventing our understanding of the precise relation-  ship between CS function and structure. In this study, we chemoenzymatically synthesized various CS species with defined compositions, some of which are of uniform structure, and therefore with defined structure. This is the first report of the construction of a CS library with a specified length (M r Ϸ 10,000) and defined structures. Using this library, we performed the characterization of individual CS sulfotransferases especially with regard to their substrate specificity, epitope analysis of different anti-CS antibodies, and determination of the CS structure required for binding to MK and PT. These results clearly demonstrate that our CS library is indeed applicable to various studies of CS functions. For the construction of the CS library, the selection and preparation of enzymes were critical. At present, six CS synthases have been identified in mammals (2,24). However, any one or a mixture of two or more of these enzymes could not elongate CH chain in vitro, and co-expression of at least two of these enzymes in the cell is required for CS polymerization (25). In contrast, the single enzyme K4CP, a capsular enzyme of bacteria (14) can polymerize the CH chain with a desired length presumably over 500 kDa (supplemental Fig. 2). In addition, CH prepared using K4CP is totally non-sulfated, whereas chemically desulfated CH species still contain ϳ10% sulfated disaccharide units.
For the transfer of sulfate groups, we used animal sulfotransferases, which have been well characterized. Regarding their expression, bacterial expression systems have failed to produce sufficiently active enzymes except in a few cases (26 -28), and baculoviral infection systems were unsuitable because of chondroitinase activity in the virus-infected medium (29). CS sulfotransferases obtained from mammalian expression systems efficiently modified the CH and CS chains. C4ST-1 and C6ST-1 generated sCSA and sCSC possessing Ͼ95% 4S and 6S units, respectively, indicating that 22 disaccharide units of 23 were uniformly sulfated. In addition, we could generate CS molecules with the desired compositions of 4S and 6S units by changing the reaction conditions (supplemental Fig. 3), allowing simultaneous reactions with both C4ST-1 and C6ST-1 (supplemental Fig. 4), and all of the chains were sulfated rather equally (supplemental Fig. 5).
Almost all (ϳ90%) 4S units in sCSA-HMDA were sulfated by GalNAc4S-6ST at the C-6 position of the GalNAc(4S) residue. Although the enzyme was reported to preferentially catalyze GalNAc(4S) residues at the nonreducing end of CSA (6,30), it also catalyzed inside 4S units. Although UA2ST sulfated a GlcUA residue with only 4% of the 6S units in sCSC-HMDA, it transferred ϳ90% of the 6S units in sCSAC-HMDA, consistent with a previous report (31) showing that UA2ST preferentially sulfates the GlcUA residue of GalNAc(4S)-GlcUA-GalNAc(6S). UA2ST sulfated GlcUA, not IdoUA, residue with 9% of the 4S units in sCSAC-HMDA to form the diSB units, which also confirms a previous report (32), demonstrating the preparation of the diSB unit from CSA with UA2ST. When reacted with sCSE-HMDA as substrate, UA2ST sulfated the GlcUA residue of the diSE unit forming the triS unit, which has been rarely found in animals. However, no triS unit was generated from the diSB unit in sCSAD-HMDA with GalNAc4S-6ST. These results suggest that UA2ST requires C4-and/or C6-sulfation in the GalNAc residue adjacent to GlcUA and that GalNAc4S-6ST does not catalyze GalNAc C6-sulfation after C2 sulfation in the GlcUA residue with UA2ST. The sequential catalytic reaction by GalNAc4S-6ST and UA2ST followed by C4ST and C6ST is strictly defined in vivo.
Epitopes of anti-CS antibodies have been determined by inhibition of interactions with defined oligosaccharides. In the ELISA system, we analyzed the direct interactions of antibodies with CS with defined composition and structure. The most distinct feature was that neither of the CS antibodies tested bound to uniform CS species such as sCH (100% 0S), sCSA (Ͼ95% 4S), sCSC (Ͼ95% 6S), and sCSE (ϳ88% diSE), except for LY111, which bound to sCSA very weakly, indicating that anti-CS antibodies, in general, require a block of complex CS structures. The interactions with native CS species suggest that these antibodies require minor CS disaccharide residues for binding, in addition to the major CS unit. Another feature is that all the antibodies bound to sCSAC at substantial levels, indicating that the alternate repeating structure of 4S and 6S units is necessary for binding. MO225 bound to sCSDE, sCSAD, and sCtriS, suggesting that it preferentially recognizes a block containing a diSD unit, which confirms previous reports (8). CS56, LY111, and 2H6 bound to sCSAC, sCSAD, and sCSDE, suggesting the preferential requirement of a complex structure composed of various sulfated saccharides. These indicate that these antibodies require combinations of non-uniform mixed sequences in their epitopes, and actually, this is in agreement with the epitope sequences for mAb WF6 (12).
In SPR analysis, native CSD and CSE showed high affinity to MK and PT, consistent with previous reports (33)(34)(35), and a comparison between the native and synthetic CS reveals distinct features. MK bound to sCS containing diSD units, but with a lower affinity than to one containing diSE units. Its binding to native CSE may be due to the long chain size (M r ϭ 100,000) and/or the heterogeneous binding motif. MK also bound to sCSA, sCSC, and sCSAC, at certain levels, indicating that it binds to a uniform CS structure and does not always require a diSD unit. PT binds to sCSC, sCSAC, and sCSAD, similar to MK. As the dissociation constants of these to sCSE were similar, the CSD unit may not be important for stable binding.
The binding motifs of many glycosaminoglycans to various molecules and receptors such as that of heparan sulfate for heparin-binding growth factors and growth factor receptors are 8ϳ12-mer oligosaccharides (36). Similarly, CS may contain a binding motif of certain saccharide residues. DS structure, containing IdoUA instead of GlcUA, is important for binding to cytokines, for example heparin cofactor II (37). Chemoenzymatic synthesis enables the generation of CS chains with specific lengths, distinct compositions, and defined structures. Therefore, it will provide a useful tool for studies of physiological functions of CS in development and disease.