Novel Carbohydrate-binding Activity of Pancreatic Trypsins to N-Linked Glycans of Glycoproteins*

How glycosylation affects the reactivity of proteins to trypsin is not well understood. Bovine and porcine pancreatic trypsins were discovered to bind to α-Man, Neu5Acα2,6Galβ1,4Glc, and α-galactose sequences by binding studies with biotinylated sugar-polymers. Quantitative kinetic studies supported that phenylmethylsulfonyl fluoride (PMSF)-treated trypsin binds to glycolipid analogues possessing α-Man or α-NeuAc but not to those possessing β-galactose or β-GlcNAc residue. Enzyme-linked immunosorbent assay (ELISA) showed that trypsin binds to six kinds of biotinylated glycoproteins possessing high mannose-type and complex-type N-glycans but not to bovine submaxillary mucin, which possesses only O-glycans. Further, the binding of trypsin to glycoproteins was differentially changed by treatments with sequential exoglycosidases, endoglycosidase H, or N-glycosidase F. Quantitative kinetic studies indicated that PMSF-treated trypsin binds with bovine thyroglobulin with the affinity constant of 1010 m–1, which was the highest among the glycoproteins examined, and that α-galactosidase treatment decreased it to 105 m–1. PMSF-treated trypsin bound to other glycoproteins, including ovomucoid, a trypsin inhibitor, with the affinity constants of 108-105 mol–1 and were markedly changed by glycosidase treatments in manners consistent with the sugar-binding specificities suggested by ELISA. Thus, the binding site for glycans was shown to be distinct from the catalytic site, allowing trypsin to function as an uncompetitive activator in the hydrolysis of a synthetic peptide substrate. Correspondingly the carbohydrate-binding activities of trypsin were unaffected by treatment with PMSF or soybean trypsin inhibitor. The results indicate the presence of an allosteric regulatory site on trypsin that sugar-specifically interacts with glycoproteins in addition to the proteolytic catalytic site.

Numerous biological phenomena are mediated by recognition of specific oligosaccharide signals. This recognition implies quality control in polypeptide folding, cellular interactions, and protein targeting (1)(2)(3). In contrast, some functions of protein glycosylation seem to be widely applicable to various types of glycosylation, for example, protecting against proteolysis, stabilizing active conformations, and affording solubility to proteins (3). These functions have been attributed to ambiguous steric effects of glycosylation in the absence of clear structural specificity, but the involvement of glycan recognition in achieving these functions has not yet been elucidated. Clarification of the molecular mechanism by which glycosylation plays a role in protecting or stabilizing the active conformation of proteins would enable the use of glycosylation in molecular engineering of recombinant products for therapeutic purposes.
Trypsin is a principal pancreatic serine protease that plays a key role in digestion in the duodenum by activating zymogens and degrading dietary proteins. Trypsin acts specifically on peptide bonds of the carboxyl side of positively charged lysine and arginine and catalyzes the activation of many pancreatic proenzymes, such as trypsinogen, chymotrypsinogen, proelastase, and carboxypeptidase, and protease-activated receptors to control digestive efficiency in the intestines (4,5). When, however, trypsin is activated in the pancreas, the activated proteinases induce the destruction of pancreatic cells. The modulation of trypsin activity is therefore important for controlling digestive efficiency and preventing pancreatitis.
Porcine pancreatic ␣-amylase (PPA) 2 is activated by interaction with glycoproteins. Previously we reported that PPA exhibits carbohydratebinding activity toward N-glycans of glycoproteins (6). To further elucidate the biological functions of the carbohydrate-binding activity found in PPA, we investigated whether other pancreatic digestive enzymes possess similar carbohydrate-binding activity. In this study, we found that trypsin exhibits remarkable carbohydrate-binding activities to the sequences present in the N-glycans of glycoproteins with a specificity distinct from PPA. This finding provides new insights into the interaction between the proteases and glycoproteins related to protease resistance and the biological functions of carbohydrate-specific interactions in the digestive organs.

EXPERIMENTAL PROCEDURES
Materials-Porcine pancreatic trypsin (PPT), N-␣-benzoyl-L-arginine ethyl ester (BAEE), N-␣-benzoyl-DL-arginine-p-nitroanilide hydrochloride (BAPA), soybean trypsin inhibitor, bovine serum albumin (BSA), 3,3Ј-diaminobenzidine tetrahydrochloride, methyl-␣-Dmannoside, and mannitol were purchased from Wako Pure Chemical Industries, Ltd., Osaka, Japan. Bovine pancreatic trypsin (BPT), bovine submaxillary gland mucin (BSM), human holotransferrin, fetuin from fetal calf serum, hen ovomucoid, human orosomucoid, bovine thyroglobulin, streptavidin-biotinylated horseradish peroxidase complex (ABC complex), and 4-nitrophenyl phosphate magnesium salt were purchased from Sigma. Sugar-biotinylated polyacrylamide probes (sugar-BP probes) were purchased from Lectinity Holdings, Inc. Moscow, * This work was supported in part by Grants-in-aid for Scientific Research (C) 14580622 and 17570109 (HO) from the Japan Society for the Promotion of Science and Grantsin-aid for Scientific Research on Priority Areas 15040209 and 17046004 (HO) from the Ministry of Education, Culture, Sports, Science, and Technology. 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. 1 To whom correspondence should be addressed.  (7). ␤-Galactosidase and ␤-N-acetylhexosaminidase from jack bean and ␣-galactosidase from Mortierella vinacea were purchased from Seikagaku Corp. EZ-Link sulfo-N-hydroxysuccinimide-biotin and Sambucus nigra bark lectin were purchased from Funakoshi Co. Ltd., Tokyo, Japan. Phenylmethylsulfonyl fluoride (PMSF) and methyl-␣-D-galactoside were purchased from Nacalai Tesque, Inc., Kyoto, Japan. N-Heptyl-␤-D-thioglucoside was purchased from Dojindo Laboratories, Kumamoto, Japan. Peanut lectin and Ricinus communis agglutinin I were purchased from Seikagaku Corp. Lactose was purchased from Kanto Kagaku, Tokyo, Japan. SDS-PAGE molecular weight standards were purchased from Bio-Rad. Preparation of Glycoprotein Probes-All biotinylated glycoprotein probes, their deglycosylated derivatives, and biotinyl lectins were prepared in our laboratory. Biotinylation was performed using EZ-link TM sulfo-N-hydroxysuccinimide-biotin according to the instruction manual. Briefly, 2 mg of each glycoprotein was dissolved in 1 ml of 50 mM sodium bicarbonate buffer (pH 8.5), and 74 l of sulfo-NHS-biotin (1 mg/ml) was added. After incubation for 30 min at room temperature, the reactant was dialyzed against water to remove excess biotin. Asialoglycoproteins, asialoagalactoglycoproteins, and asialoagalactoahexosaminoglycoproteins were prepared from biotinylated glycoproteins by sequential glycosidase treatments with neuraminidase (0.1 units/mg glycoprotein) in 20 mM acetate-buffered saline (pH 5.5), ␤-galactosidase (0.14 units/mg glycoprotein) in 50 mM sodium-citrate buffer (pH 3.5) overnight, and then ␤-N-acetylhexosaminidase (1.43 units/mg of glycoprotein) in 50 mM sodium citrate buffer (pH 5.0) at 37°C overnight. The glycan structures and carbohydrate concentrations of the glycoproteins used in this study are summarized in Scheme 1. Besides the sequential treatments described above, biotinylated bovine thyroglobulin was treated with ␣-galactosidase (0.14 unit/mg of glycoprotein) in 20 mM acetate-buffered saline (pH 5.5) for agalactosylation of the major oligosaccharide, Gal␣1-3Gal␤1-4GlcNAc. Fetuin was treated with N-glycosidase F (600 units/mg of glycoprotein) in 10 mM Tris-buffered saline (TBS) at pH 7, or de-O-glycosylated with a mixture of neuraminidase SCHEME 1. Major oligosaccharide structures of glycoproteins used in this study.
Biotinylated porcine thyroglobulin (20 g) was denatured in glycoprotein denaturing buffer (5% SDS, 10% ␤-mercaptoethanol) at 100°C for 10 min, a 10% volume of 0.5 M sodium citrate buffer (pH 5.5) was added, and then it was incubated with 1500 units of endoglycosidase H at 37°C overnight. Deglycosylation of all biotinylated glycoprotein probes was checked by ELISA for a change in reactivity with Ricinus communis agglutinin I for agalactosylation, Psathyrella velutina lectin for desialylation and ahexosaminylation, Galanthus nivalis lectin for endoglycosidase H treatment, and peanut lectin for de-O-glycosylation to recognize each carbohydrate structure, and by mobility on SDS-PAGE to demonstrate the decrease in molecular weight (data not shown).
SDS-PAGE-To check the purity of trypsin, SDS-PAGE was performed according to the method of Laemmli (8) using a 14% gel in the presence of 2-mercaptoethanol. The bovine and porcine trypsins (10, 20, or 40 g of protein per lane) were loaded onto the gel together with a set of markers and run at 20 mA for 1.5 h. After electrophoresis, protein bands were visualized by Coomassie Brilliant Blue R-250.
Binding Studies with Sugar-BP Probes or Biotinylated Glycoprotein Probes-PPT and BPT were preincubated in the presence or absence of 0.5 mM PMSF, 0.5 mM soybean trypsin inhibitor, or 5 mM EDTA in 10 mM TBS (pH 7.5) for 1 h and then immobilized at concentrations of 0.01-0.5 g/100 l in wells of a microtiter plate (Immulon 1, Dynatech Laboratories) at 4°C overnight. All other procedures were performed at room temperature using 10 mM TBS (pH 7.5) as the dilution buffer. After immobilization, the wells were blocked with 3% BSA for 2 h. Aliquots (100 l) of various sugar-BP probes (shown in Scheme 1A) or biotinylated glycoprotein probes at concentrations of 10 g/ml were added to each well, followed by incubation for 1 h. After incubation, the wells were washed three times, and 100 l of ABC complex (1 g/ml) was added, and the mixture was incubated for 1 h. After washing three times, color was developed by adding 200 l of o-phenylenediamine/H 2 O 2 , and then 50 l of 2.5 M H 2 SO 4 was added to stop the reaction. Absorbance was measured with a microplate reader (Bio-Rad MPR-80) at 490 nm.
Quantification of Interactions between Trypsins and Glycolipid Analogues by Surface Plasmon Resonance-For binding studies between trypsins and glycolipid analogues or various glycoproteins, a BIAcore 2000 SPR apparatus (BIAcore AB, Uppsala, Sweden) was used. Structure of glycolipid analogues used in this study was illustrated in Scheme 2B, and their synthesis will be described elsewhere. Glycolipid analogues were immobilized on a HPA sensor chip (BIAcore AB) by preparing liposomes containing each glycolipid analogue/phosphatidylcholine at a molar ratio of 40/60 as described previously (9, 10). PMSF-treated PPT was injected onto the sensor chip at various concentrations in 10 mM TBS buffer (pH 7.5) at a flow rate of 20 l/min at 25°C using a BIAcore biosensor. The reference cell was prepared by immobilizing phosphatidylcholine and used to correct for bulk effect. The chip was regenerated each time by injection of 20 l of 0.1 M phosphoric acid.
Quantification of Interactions between Trypsins and Various Glycoproteins by SPR-After equilibration of a CM5 sensor chip (BIAcore AB) with HEPES-buffered saline, the surface of the sensor chip was activated with an amine coupling kit. BPT or PPT (each 1.8 M) in 10 mM sodium acetate buffer (pH 6) containing 0.1 mM PMSF and 0.2 M methyl ␣-Dmannoside was injected onto the activated surface, and then the remaining N-hydroxysuccinimide esters were blocked with 1.0 M ethanolamine hydrochloride (pH 8). Each step was performed for 14 min at a constant flow rate of 10 l/min at 25°C. The reference flow cell was prepared with BSA as a ligand.

Carbohydrate Binding Activity of Pancreatic Trypsins
mM TBS (pH 7.5) were separately injected onto the trypsin-immobilized sensor chip at concentrations of 1, 0.5, 0.25, 0.125, and 0.0625 M for 150 s at a flow rate of 20 l/min at 25°C. The chip was regenerated each time by injection of 20 l of 10 mM HCl. To assay inhibition, a trypsin-immobilized sensor chip was equilibrated with TBS containing 50 mM or 0.2 M methyl-␣-mannoside, methyl-␣-galactoside, or lactose for 15 min, and then the glycoprotein dissolved in the same buffer was injected. Kinetic parameters were calculated mainly by global analysis, or affinity analysis when necessary, using the BIAevaluation software version 3.1.
Measurement of Enzyme Activity of Trypsin-Enzyme activity was measured in a test tube according to the method previously described using BAEE (11) or BAPA (12) as the substrate. To estimate the effect of sugar on the BAEE-hydrolytic activity, it was measured after preincubation of PPT in 100 l of 1 mM HCl (0.02 mg/ml) with 100 l of 0.2 M methyl-␣-D-mannoside, 0.2 M methyl-␣-D-galactoside, or 0.2 M lactose in 20 mM TBS (pH 7.6) for 10 min at 25°C. A control experiment was done by preincubating PPT without sugar. After preincubation, the PPT solution was added to 500 l of 0.025 mM BAEE in 10 mM TBS (pH 7.6) and incubated at 25°C for 1 min. Absorbance was immediately measured at 253 nm, and trypsin activity that increased absorbance by 0.003 at 25°C for 1 min was regarded as 1 USP unit of trypsin.
To analyze the effect of sugar by a double reciprocal Lineweaver-Burk plot, the initial rates of the enzyme-catalyzed reaction were measured using BAPA as the substrate. The substrate stock solution was prepared by dissolving 0.0217 g of BAPA in 1.5 ml of Me 2 SO and then diluted to final concentrations of 0.1-0.5 mM with 100 mM TBS (pH 7.5) in the presence or absence of 0.2 M methyl-␣-mannoside, methyl-␣-galacto-FIGURE 2. Quantification of interaction between PPT and glycolipid analogues by SPR. Glycolipid analogues were immobilized on an HPA sensor chip as described in the text. PPT was pretreated with 0.1 mM PMSF and injected onto the sensor chip at various concentrations in 10 mM TBS buffer (pH 7.5) at a flow rate of 20 l/min at 25°C using BIAcore. Binding curves of PPT on the sensor chip immobilized with glycolipid analogues containing ␣-Man (A), ␣-NeuAc (B), and ␤-Lac (C) are shown. The response is expressed as the change in the number of resonance units induced by the binding of PPT to the glycolipid analogue-immobilized flow cell, which was corrected for bulk effect by subtracting the change on the phosphatidylcholine-immobilized reference cell. D, kinetic parameters for the interaction between PPT and glycolipid analogues. Kinetic parameters were calculated by global analysis for ␣-Man and affinity analysis for ␣-NeuAc and ␤-Lac. k a , association rate constant; k d , dissociation rate constant; K A , association constant.  side, or lactose. The substrate solutions were heated for 5 min at 37°C. The enzyme stock solution was prepared by dissolving 2.5 mg of PPT in 3 ml of 1 mM HCl containing 13 mg of CaCl 2 -2H 2 O. A 10-l aliquot of the enzyme solution was added to 3 ml of the substrate solution, and absorbance was measured at 410 nm every 30 s.

Interaction between Trypsins and Sugar-BP Probes
The carbohydrate-binding activities of BPT and PPT were analyzed using synthetic sugar-BP probes (Scheme 2A), and specificities toward sugar residues and oligosaccharides were determined. As shown in Fig. 1, both BPT and PPT exhibited high binding activity toward ␣-Man-, ␣-Man-6-phosphate-, NeuAc␣2,6Gal␤1,4Glc-, ␣-Gal-, and ␤-Glc-BP probes among the 17 kinds of sugar-BP probes tested. Both trypsins bound to NeuAc␣2,3Gal␤1,4Glcand NeuAc-BP to a lesser extent than to NeuAc␣2,6Gal␤1,4Glc-BP, indicating a preference for sialyl linkages. On the contrary, trypsin did not bind with ␣-GalNAcor mucin core 2-type BP probes or with ␤-Gal, LacNAc, Lac, ␤-GlcNAc, or ␤-Gal3-sulfate-BP probes. None of the carbohydrate-binding activities of the trypsins was affected by preincubation with PMSF and EDTA or soybean trypsin inhibitor, suggesting that binding is independent of the catalytic site. The bound sugar residues other than ␤-D-Glc are components of N-glycans, demonstrating basic specificities toward monosaccharides or short sequences that include linkages for trypsin binding.

Interaction between Trypsin and Glycolipid Analogues Analyzed by SPR
To verify the carbohydrate-binding specificity of trypsin by quantitative measurement, interaction analyses were performed by SPR using five kinds of synthetic high sensitivity glycolipid analogues (Scheme 2B). The total amounts of immobilized glycolipid analogues containing ␣-Man, ␣-NeuAc, and ␤-Lac were 1593, 1560, and 1680 BIAcore resonance units (RU, 1000 RU ϭ 1 ng/mm 2 ), respectively. As shown in Fig.  2 (A-C), PPT concentration-dependently bound to immobilized glycolipid analogues. The differential binding of PPT to the analogues clearly indicates its relative binding affinity toward sugar residues: PPT binds best with the analogues containing ␣-Man, then ␣-NeuAc, and to a lesser extent ␤-Lac. The binding and dissociation occurred rapidly at the start and end of injection of the glycolipid analogues except the ␣-Man derivative, demonstrating the specific binding of PPT to those glycolipid analogues with quick association and dissociation rates. PPT did not bind to other analogues containing ␤-galactose or ␤-GlcNAc even at 2 M. The association constants (K A ) were calculated to be 10 6 -10 5 (M Ϫ1 ) for glycolipid analogues containing ␣-Man and ␣-NeuAc (Fig. 2D), which are comparable to the K A obtained for the interaction between ricin and the glycolipid analogues containing ␤-galactose 3 and higher than that obtained between concanavalin A and ␣-Man-derivatized glycolipid containing phosphatidylethanolamine aglycon (13). The carbohydrate-binding specificity indicated by SPR corresponded with that obtained by using sugar-BP probes (Fig. 1) demonstrating that the affinity of trypsin for specific carbohydrates is comparable to those of known plant lectins. Therefore, the binding activity of trypsin for glycoproteins was examined.

Purity of BPT and PPT on SDS-PAGE
The carbohydrate-binding activities shown in Figs. 1 and 2 have never been reported for trypsin. To eliminate the suspicion that some contaminant in the trypsin preparation might exhibit such an activity, we analyzed the purity of the trypsin preparations used in this study by SDS-PAGE. As shown in Fig. 3, both BPT and PPT showed only a single band without any detectable contamination even if 40 g of trypsin preparation was applied per lane to the polyacrylamide gel. Therefore, the carbohydrate-binding activities observed in this study were attributed to the trypsins.   PPT and BPT were immobilized on a CM5 sensor chip, and each glycoprotein was injected onto the sensor chip at various pH, as described in the text. Fetuin (f) or porcine thyroglobulin (E) was dissolved at concentrations of 1 or 0.5 M, respectively, in 10 mM acetate buffer (pH 4.5-6.5), 10 mM TBS (pH 7-8), or 10 mM bicarbonate buffer (pH 9 -10) and injected onto the sensor chip. The bound amounts of glycoprotein are expressed as relative response (%) by taking the response at pH 7.5 as 100%. A, relative response on immobilized PPT; B, relative response on immobilized BPT.

Interaction between Trypsins and Biotinylated Glycoprotein Probes
Interactions between the glycoprotein probes and PPT or BPT were studied by ELISA at pH 7.5, which is the physiological pH in the duodenum. Because the amount of biotin incorporated into each glycoprotein probe was almost equal, as judged by the color intensity of each probe developed with the ABC complex being within a 10% error, the value of A 490 corresponds to the amount of probe bound. As shown in Fig. 4(A and B), BPT and PPT were found to bind to various glycoproteins with very similar binding patterns. The trypsins bound best to bovine thyroglobulin and to a lesser extent to fetuin, porcine thyroglobulin, ovomucoid, orosomucoid, and transferrin, in that order, but not to BSM. All the bound glycoproteins contain 5-30% (w/w) N-linked oligosaccharides, whereas BSM possesses up to 60% (w/w) O-linked glycans, which are mainly sialyl-Tn and core 3-type (Scheme 1). Combined with the finding that trypsins did not bind with ␣-GalNAcand ␤-GlcNAc (Fig. 1), this indicates that trypsins do not interact with O-linked glycans.
The involvement of the N-glycan structure in the binding with PPT and BPT was shown by deglycosylating the glycoprotein probes with endo-type glycosidases. As shown in Fig. 5C, de-N-glycosylation of fetuin by N-glycosidase F treatment markedly decreased the reactivity toward both trypsins (data not shown for PPT), showing that their binding was mostly due to the affinity for the sialylated complex-type N-glycans of fetuin. The reactivity of trypsins for porcine thyroglobulin (Fig.  5B), which contains almost equal amounts of high Man-type and complex-type glycans (Scheme 1), was decreased to about half that of intact porcine thyroglobulin by endoglycosidase H treatment as well as asialoagalactoahexosaminylation, indicating that trypsins bind with high Man-type N-glycans as well as the sialylated complex-type.
As shown in Fig. 5, the effects of exo-type glycosidase treatments of the glycoprotein probes illustrate the contribution of each sugar residue to the interaction with BPT. Remarkably, binding of BPT with bovine thyroglobulin was found to be diminished by ␣-galactosidase treatment (Fig. 5A), clearly indicating that the ␣-galactosyl residue at the nonreducing terminal, which is unique to the N-glycan of bovine thyroglobulin (Scheme 1), is an epitope for BPT binding. Other exoglycosidase treatments of bovine thyroglobulin did not affect the binding. On the other hand, neuraminidase treatment of other glycoproteins that possess sialylated complex-type N-glycans, such as fetuin, ovomucoid, orosomucoid, and porcine thyroglobulin, considerably decreased the binding, as shown in Fig. 5 (B-E). For these glycoprotein probes ␤-galactosidase treatment subsequent to desialylation did not significantly change the binding, but exposure of ␣-mannosyl residues of the trimannosyl core of N-glycans by ␤-hexosaminidase treatment subsequent to degalactosylation restored the binding to trypsin. These results strongly indicate that ␣-NeuAc and ␣-mannosyl residues of complextype multiantennary N-glycans contribute to the binding of trypsins, but ␤-galactose and ␤-GlcNAc residues of the lactosamine sequence do not. The biantennary complex type of transferrin did not show significant affinity for trypsins (Fig. 5F). As a whole, the sugar-binding specificities of BPT and PPT indicated in Fig. 1 coincide with and account for the binding specificities toward the glycoprotein probes.
Preincubation of PPT and BPT with PMSF and EDTA or soybean trypsin inhibitor before the binding studies did not affect the binding activities toward glycoproteins (data not shown). Moreover, the binding of trypsin to ovomucoid, a natural inhibitor that blocks the catalytic site of trypsin, was found to be significantly affected by glycan trimming (Fig. 5D). The observations indicate that the trypsin binding to glycoproteins is independent of their catalytic activity.

Interaction between Trypsins and Various Glycoproteins Analyzed by BIAcore
The total amounts of immobilized PPT, BPT, and BSA were 4,012, 4,028, and 4,031 RU, respectively. The responses are expressed as the change of resonance units induced by the binding of analytes to each  Carbohydrate Binding Activity of Pancreatic Trypsins MARCH 31, 2006 • VOLUME 281 • NUMBER 13 flow cell, which was corrected for bulk effect by subtracting the change on the BSA-immobilized reference cell.
pH Dependence of Interaction between Trypsin and Glycoproteins-Fetuin or porcine thyroglobulin was injected onto a trypsin-immobilized chip at pH 4.5-10. As shown in Fig. 6, the amounts of the glycoproteins bound to immobilized PPT and BPT changed within 30% in the pH range examined, showing a maximum at around pH 7.5-8.0. The weakly alkaline pH coincides with the enteric pH indicating that the carbohydrate-binding activity of trypsin is optimal in the milieu of the intestine. Based on this observation, 10 mM TBS at pH 7.5 was thereafter used for the binding studies.
Effect of Sugars on Binding of Trypsin to Glycoproteins- Fig. 7 shows the effect of monosaccharide trypsin binding to glycoproteins. Me-␣-Man and Me-␣-galactoside, which showed the highest binding to trypsin ( Fig. 1) were used as inhibitors in comparison with Lac, which showed relatively lower binding in the binding study with BP-sugars. The binding of both trypsins to fetuin and porcine thyroglobulins was decreased by 30 -44% in 50 mM Me-␣-Man and 56 -64% in 0.2 M Me-␣-Man, and binding to bovine thyroglobulin was decreased by 44 -50% and 66 -70% in 50 mM and 0.2 M Me-␣-Gal, respectively. The binding of trypsins was decreased by only 10 -18% even at 0.2 M Lac, indicating the weak inhibitory activity of Lac compared with those of Me-␣-Man and Me-␣-galactoside coinciding with the relative affinity of sugars shown by BP-sugar binding studies. The results strongly support the hypothesis that trypsins bind glycoproteins by sugar-specific interaction.
Kinetic Parameters for Binding between Trypsins and Glycoproteins-The sensorgrams are shown in Fig. 8 (A and B), and the binding parameters were calculated for each glycoprotein. The binding of all glycoproteins except BSM fit best a 1:1 binding model among the fitting models in global analysis. The interaction between bovine thyroglobulin and trypsins was analyzed at concentrations lower than 0.25 M (Fig. 8, A  (panel e) and B (panel o)), because bovine thyroglobulin gave fluctuating, irregular-shaped binding curves at concentrations higher than 0.5 M. The interaction between trypsins and BSM showed box-shaped binding curves suggesting high association and dissociation rates (Fig. 8,  A (panel j) and B (panel t)) and was analyzed using affinity analysis. The kinetic data of the binding are summarized in Table 1.
Trypsin bound to the glycoproteins possessing N-glycans with significantly high affinity, K A ranging from 10 10 -10 6 M Ϫ1 . In contrast, the K A for BSM was as low as 1-8 ϫ 10 4 M Ϫ1 , indicating very weak interaction with trypsins. The K A for bovine thyroglobulin binding to both PPT and BPT was 10 10 M Ϫ1 , which is the strongest among glycoproteins and equal to that of high affinity antibodies, followed by the K A of porcine thyroglobulin, 4-5 ϫ 10 7 M Ϫ1 . The high K A of trypsin for bovine thyroglobulin is attributable to the extremely low dissociation rate constants (k d ) (10 Ϫ7 -10 Ϫ6 s Ϫ1 ) compared with those of other glycoproteins (10 Ϫ4 -10 Ϫ3 s Ϫ1 ), suggesting that thyroglobulins hardly dissociate from trypsin. The K A of trypsin for bovine thyroglobulin was markedly decreased from 10 10 to 10 5 (M Ϫ1 ) by ␣-galactosidase treatment, indicating that ␣-galactose residues are essential for high affinity binding to trypsin. On the other hand, the K A of trypsin for porcine thyroglobulins (4 -5 ϫ 10 7 M Ϫ1 ) was decreased to 15-20% by treatment with endoglycosidase H or neuraminidase, showing that high-Man types as well as sialyl residues of complex types contribute to the binding. The K A for fetuin was decreased by N-glycosidase F treatment by ϳ10 Ϫ3fold but not by O-glycosidase treatment, indicating that the sialylated complex-type N-glycans of fetuin contributed absolutely to the binding, but O-glycans did not, even if they are sialylated. The results correlated well with the reactivities of trypsin toward intact and glycosidase-modified glycoprotein probes by ELISA.

Effect of Sugars on Enzyme Activity of PPT
As shown in Fig. 9A, the sugars that bound to trypsins enhanced the enzyme activity to various degrees, as detected using BAEE and BAPA as substrates. The hydrolytic activity of PPT for BAEE was enhanced by 1.4-fold in 0.2 M methyl-␣-D-mannoside and 1.2-fold in methyl-␣-Dgalactoside but not enhanced in 0.2 M lactose. When we used a slowly hydrolyzable BAPA as the substrate, PPT was activated with 0.2 M methyl-␣-D-mannoside and methyl-␣-D-galactoside to ϳ1.2to 1.4-fold at 300 -600 s (Fig. 9B). As shown in Fig. 9 (C-E) and Table 1, the Lineweaver-Burk plots indicate that the binding of Me-␣-Man, and Me-␣galactoside uncompetitively activates PPT with increasing V max by 2.5fold and K m by 2-to 2.5-fold, while binding of lactose slightly inhibited PPT noncompetitively and uncompetitively, indicating that the binding of carbohydrates activates the hydrolytic activity to various degrees.

TABLE 1 Binding parameters for interaction between trypsins and glycoproteins
Interactions between trypsin and glycoproteins were measured in 10 mM TBS (pH 7.5) using BIAcore. Kinetics parameters were calculated by global analysis for most glycoproteins and affinity analysis for BSM. k a , association rate constant; k d , dissociation rate constant; K A , association constant (K A ϭ k a /k a ).

DISCUSSION
This study demonstrates that mammalian pancreatic trypsin commonly binds to glycoproteins possessing N-linked glycans by carbohydrate-specific interaction. The sugar-binding specificity of trypsin was shown by the binding with sugar-BP probes and glycolipid analogues to be ␣-galactosyl, oligomannosyl, and nonreducing terminal ␣2,6-NeuAc residues (Fig. 1). Trypsin bound to glycoproteins possessing N-glycans with very high affinity, reaching 10 10 -10 6 M Ϫ1 , whereas it did not bind to BSM (Fig. 4 and Table 1). The binding of glycoprotein probes with trypsin was changed by glycosidase treatments on ELISA and SPR analyses, which coincided well with the sugar-binding specificity indicated by sugar-BP probes. The specificity of the interaction between trypsin and the glycoproteins was proven by inhibition studies with monosac-charides using SPR (Fig. 7) and conclusively demonstrated to be due to the affinity of trypsin for component saccharide residues of the N-linked glycans but not by protein-protein interaction.
Treatment of trypsin with soybean trypsin inhibitor and PMSF did not affect the binding to sugar-BP, glycolipid analogues, and glycoprotein probes, and trypsin was noncompetitively and uncompetitively activated toward synthetic substrates, BAEE and BAPA, by the binding of specific sugars ( Fig. 9 and Table 2). Therefore, the N-glycan recognition of trypsin must be exhibited at a site different from its catalytic site, and activation would be caused by an allosteric effect to make the substrate-binding site more accessible to the substrate and/or by a conformational effect that stabilizes the trypsin molecule against autodegradation, like the stabilizing effect of Ca 2ϩ binding (14).
The coating of oligosaccharides on glycoproteins can serve to protect the polypeptide chain from degradation by proteases (3). The contributions of sialylation to the stabilization of glycoprotein against tryptic hydrolysis have been reported for several glycoproteins, including orosomucoid (15) and vitronectin (16). The de-N-glycosylation of ovomucoid with trifluoromethanesulfonic acid has been reported to interfere with the inhibitory activity against trypsin and make ovomucoid easily hydrolyzable with trypsin (17). Although the relationship between oligosaccharide structure and the protective function against proteases has been explored for several glycoproteins (18 -20), the protecting mechanism achieved by the oligosaccharides has remained unclear. Because the removal of oligosaccharides from a mature protein does not always drastically alter its sensitivity to proteolysis, FIGURE 9. Effect of various sugars on enzyme activity of PPT. A, aliquots (100 l) of 0.2 M methyl-␣-mannoside, methyl-␣-galactoside, or lactose, or buffer for the control were added to PPT (2 g) in the same volume, and the enzyme activity was measured against BAEE as described in the text. Relative activity was expressed as a percentage, taking the control as 100%. B, time course of PPT activity in the presence of various sugars. The enzyme activity was measured against BAPA as described in the text in the presence or absence of 0.2 M methyl-␣-mannoside (f), methyl-␣-galactoside (Ⅺ), lactose (छ), and control (ϫ). C-E, Lineweaver-Burk plots of PPT. PPT activity was measured in the presence (solid line) or absence (dashed line) of 0.2 M concentrations of various sugars as described in the text and analyzed by a double reciprocal Lineweaver-Burk plot. some specific interaction between protease and glycoproteins may be involved in regulating protease attack. We found that trypsin sugar-specifically interacts with N-linked glycoproteins. The binding of trypsin to the N-glycans of glycoprotein would protect the carrier glycoprotein from hydrolysis, at least partially, by topologically restricting the substrate-binding site of trypsin. Deglycosylation of glycoproteins, which diminishes the carbohydrate-specific binding, makes trypsin interact with the peptide moiety of the glycoprotein through the substrate-binding site to hydrolyze it. In this hypothesis, glycosylation at even one site of the polypeptide can significantly affect the proteolysis of the carrier glycoprotein. It is necessary to define the relationship between the structure and position of glycosylation that affects the susceptibility to trypsin to examine the hypothesis. The binding specificity of trypsin toward carbohydrates was different from that of PPA. Trypsin bound little to N-acetyllactosamine and ␣-GalNAc, which bound well to PPA, and bound to fetuin better than to transferrin (Fig. 4), whereas PPA bound to transferrin better than to fetuin (6). The differences may suggest that the endogenous receptors for trypsin and ␣-amylase are not identical.
The carbohydrate-binding activity of trypsin was exhibited at a broad pH range with the optimum at pH 8.0, the slightly alkaline pH similar to the pancreatic fluid in the intestinal lumen. Therefore, trypsin may interact with glycoligands in the epithelial surface of the duodenum and intestine in vivo, because it is extensively glycosylated with N-glycans (21-23) containing ␣-galactose residues (24). The N-glycan-binding activity would play a role in targeting trypsin and concentrating it on the brush-border membrane. Such immobilization of trypsin would enhance the activity and/or elongate the short life span of trypsin by stabilizing it against autodegradation. Because the ability of the duodenum to digest proteins increases rapidly in the cascade of enzyme activations following trypsin activation, the enhancement of trypsin activity would be amplified to increase digestive efficiency exponentially; therefore, the 140% enhancement of trypsin activity measured by using BAEE would be amplified in the duodenum to increase more than 5-fold after the sixth stage. In addition, the binding of trypsin to intestinal epithelium makes the product peptides spatially available as a substrate for the exo-type peptidases that are naturally anchored to the intestinal brush-border membrane. The activated proteinases cooperatively break down dietary proteins to peptides that are subsequently degraded to amino acids by other exo-type peptidases either secreted or expressed in the brush border membrane of epithelial cells in the duodenum and small intestine. Rat aminopeptidase N (EC 3.4.11.2) is one of such glycoproteins with 20% (w/w) carbohydrates that possess unsialylated triand tetraantennary complex types as major N-glycans (25), which are very similar to the glycans of ovomucoid. A major part of dipeptidylpeptidase IV (EC 3.4.14.5) (26) and peptide transporter-1, a Hϩ/peptide cotransporter responsible for the uptake of small peptides (27), are among the N-glycosylated glycoproteins in the small intestine, too. The association of trypsin with this exopeptidase or transporter would enhance the rate of degradation of substrate proteins and peptide absorption by increasing the catalytic efficiency both allosterically and with mass action after. The binding of trypsin to intestinal glycoreceptors may also stimulate exocrine secretion of digestive tract hormones or pancreatic proteins as reported for exogenously administered plant lectins (28). The carbohydrate binding may regulate the reactivity of trypsin with the glycosylated protease-activated receptor 2 depending on the glycosylation state (29) and influence intestinal inflammation, cytoprotection, and cellular motility.
Together with our previous findings on pancreatic ␣-amylase, carbohydrate-binding activities of macromolecule-degrading enzymes might play essential roles in localization, activation, and stabilization of pancreatic enzymes to achieve efficient digestion. Considering the biological significance of trypsin in the activation of other proteinases and its degradative role in various tissues, the mechanism of modulating tryptic susceptibility by glycosylation of proteins must be elucidated.