Efficient synthesis of O-linked glycopeptide by a transglycosylation using endo alpha-N-acetylgalactosaminidase from Streptomyces sp.

Gal beta-(1-->3)-GalNAc-linked hexapeptide was synthesized by a transglycosylation using Gal beta-(1-->3)-GalNAc beta-pNP as a donor and a serine-containing hexapeptide as an acceptor using endo GalNAc-ase from Streptomyces sp.. The Gal beta-(1-->3)-GalNAc residue was transferred to the hydroxyl group of the serine residue of the peptide. The total yield of the glycopeptide via this process was better than that of the chemoenzymatic method. This process was confirmed to be a versatile method for the synthesis of O-linked glycopeptides.

Analysis of the microheterogeneity of the IgA1 hinge glycopeptide having multiple O-linked oligosaccharides by capillary electrophoresis.

It was found that the self-aggregation of IgA1 was closely connected with the glycoform of a mucin-type sugar chain on its hinge portion. In this report, normal human serum IgA1 was separated into two subfractions by a jacalin column. The elution condition, 25 mM galactose, used here was similar to that reported for the glycoprotein with a single mucin-type sugar chain per molecule. The IgA1 eluted under this condition was substantially the monomeric form. In contrast, the remaining IgA1 eluted from the column with 0. 8 M galactose was substantially the aggregated form. An analytical method for the microheterogeneity of the IgA1 hinge glycopeptide (HGP33) was developed to determine the difference between these IgA1 fractions by capillary electrophoresis (CE). Native HGP33 from both IgA1 fractions was separated into peaks 1-11, depending on their glycoforms. Because the sialic acid-rich component migrated slowly on CE, the 25 mM fraction was abundant in the sialic acid-rich components (peaks 7-11), but the 0.8 M fraction was abundant in the sialic acid-poor components (peaks 1-4). Comparison of the number of sugar chains per hinge peptide indicated that the 25 mM fraction was relatively well glycosylated. Thus, application of CE analysis to the HGP33 indicated that the monomeric IgA1 was composed of a relatively complete molecule with respect to the glycoform rather than the aggregated IgA1.

Human serum immunoglobulin G3 subclass bound preferentially to asialo-, agalactoimmunoglobulin A1/Sepharose.

As we reported before, exoglycosidase treatment of human serum IgA1 changed it to a sticky molecule. In order to examine the presence of the specific binding protein to the sticky IgA1 in human serum, IgA1, asialo-IgA1 (IgA1-S) and asialo-, agalacto-IgA1 (IgA1-SG)/Sepharose column chromatography of normal human serum was carried out. Purified hinge glycopeptide (HGP33) prepared from IgA1 was used for the preparation of HGP/Sepharose. A portion of the serum protein was bound to those columns and eluted with the buffer containing 1.0 M NaCl. About four times the amount of protein was eluted from the IgA1-SG/Sepharose column than that from IgA1/Sepharose. Most of the eluted protein was IgG, and the IgG1 and IgG3 subclasses but neither IgG2 nor IgG4 was dominant. Under the lower salt concentration, a portion of IgG was also bound to the HGP-SG/Sepharose column. The obtained results coincide well with the previous report of the co-present of IgG1 and IgG3 with deposited IgA1 in IgA nephropathy patients (Aucouturier, P., et al. (1989) Clin. Immunol. Immunopathol. 51, 338-347). Thus, the results solved the question of why the IgG3 was co-present with deposited IgA1 in IgA nephropathy patients.

Study of the relationship between sticky human serum IgA1 and its O-glycan glycoform.

The high-affinity IgA1 toward jacalin was mostly composed of aggregated IgA1 and abundantly contained the asialo disaccharide, Galbeta1,3GalNAc, in the O-linked oligosaccharide in the hinge region [Journal of Biochemistry 120, 92-97 (1996)]. Meanwhile, the removal of sialic acid from IgA1 accelerated the aggregation of the IgA1 molecule [J. Am. Soc. Nephrol. 9, 2048-2054 (1998)]. In order to examine the nature of such a sticky IgA1, affinity chromatography using asialo-IgA1 (deSIgA1)-Sepharose was carried out. Seventeen percent of normal human serum IgA1, 27% of asialo-IgA1 (IgA1-S), and 48% of asialo-, agalacto-IgA1 (IgA1-SG) were bound to the column. Removal of the N-acetylgalactosamine residue from IgA1-SG resulted in a decreasing affinity toward deSIgA1-Sepharose. Thus, the binding ability toward the column was the highest for the IgA1-SG among the deglycosylated IgA1s. On the other hand, heat treatment of IgA1 accelerated the aggregation but decreased its binding ability toward the column. Such heat denaturation probably destroys the structure of the binding site. Since the enzymatic removal of the N-glycan sugar chains did not induce the aggregation and exhibited no effect on the binding, the incomplete O-linked sugar chain on the hinge portion should be directly related to the sticky characteristics of the IgA1 molecule. The binding was non-covalent and not strong because the asialo-, agalacto-hinge glycopeptide was eluted slightly slower than the native one from the column and the bound IgA1 was dissociated in the presence of 1 M NaCl.

Mutual separation of hinge-glycopeptide isomers bearing five N-acetylgalactosamine residues from normal human serum immunoglobulin A1 by capillary electrophoresis.

Immunoglobulin A1 (IgA1) from normal human serum is known to have O-linked sugar chains, sialylated Galbeta1,3GalNAc, in the hinge portion. In order to reduce the microheterogenity of the sugar chain, the hinge glycopeptide prepared from IgA1 was sequentially treated with neuraminidase and beta-galactosidase. The asialo-, agalacto-hinge glycopeptide (HGP-SG) composed of a 33-mer peptide (HP33) and N-acetylgalactosamine (GalNAc) residues was obtained. The HGP-SG was separated into three major peaks, A, B and C, by high-performance liquid chromatography (HPLC). Each glycopeptide fraction was further separated by capillary electrophoresis (CE). Peaks A, B and C with HPLC abundantly contained HP33 bearing five and six N-acetylgalactosamine residues (HGP33-5,6GN), HGP33-4,5GN and HGP33-3,4GN, respectively. Among these glycopeptide peaks, only the HGP33-5GN peak was partly split into two peaks based on the CE analysis - HGP33-5GN-alpha and -beta. The glycopeptide, HGP25-5GN shortened by the thermolysin digest of HGP33-SG was also well separated into the alpha and beta forms by CE analysis. No differences in their mass and peptide portion were observed between HGP25-5GN-alpha and -beta. Therefore, the obtained result might indicate that HGP25-5GN-alpha was an isomer of HGP25-5GN-beta differing in its stereospecific structure of the peptide portion and/or the attachment site of the GalNAc residue.

Aggregated human serum immunoglobulin A1 induced by neuraminidase treatment had a lower number of O-linked sugar chains on the hinge portion.

A part of human serum immunoglobulin A1(IgA1) was aggregated by treatment with neuraminidase. Aggregated IgA1 was separated from non-aggregated IgA1 by gel permeation chromatography. The prepared asialo-hinge glycopeptide (asialo-HGP) from both IgA1 subfractions was treated with beta-galactosidase to determine the number of beta-linked sugar chains attached on the hinge region. Removal of the galactose residue from asialo-HGP resulted in the HPLC separation of three major peaks. MALDI-TOFMS analysis of the glycopeptides also indicated the presence of three HGP components with three, four and five N-acetylgalactosamine (GalNAc) residues, respectively. Comparison of their relative content among the glycopeptide components showed a higher content of the HGP component with a lower number of GalNAc residues on aggregated IgA1. Thus, asialo-HGP prepared from aggregated IgA1 induced by neuraminidase treatment had an incomplete core structure of O-linked oligosaccharides. Especially, the result suggested that the reduced number of the attached O-linked oligosaccharides on IgA1 take part in phenomena such as self-aggregation of asialo-IgA1.

Application of matrix-assisted laser desorption ionization time-of-flight mass spectrometry to the analysis of glycopeptide-containing multiple O-linked oligosaccharides.

In our previous report [Iwase et al., J. Biochem., 120 (1996) 393], the number of O-linked oligosaccharide chains on the hinge region of IgA1 was estimated by matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOFMS). In this experiment, the number of non-substituted N-acetylgalactosamines and Galbeta1,3GalNAc residues, as the core O-linked oligosaccharide structure per heavy chain of normal human serum IgA1, was estimated by digestion of the asialo-hinge glycopeptide with alpha-N-acetylgalactosaminidase (GalNAc-ase) or endo-alpha-N-acetylgalactosaminidase (endo-GalNAc-ase). GalNAc-ase treatment of the asialoglycopeptide produced two major peaks, one being a glycopeptide containing four GalNAc and four Gal residues, and the other contained three GalNAc and three Gal residues. Treatment with endo-GalNAc-ase also produced a nearly equal amount of the two peaks, with the naked hinge peptide and the peptide having one GalNAc residue. From those results, we concluded that the asialo-hinge glycopeptide was composed of three components bearing four Galbeta1,3GalNAc and one GalNAc, only four Galbeta1,3GalNAc, and three Galbeta1,3GalNAc and one GalNAc, respectively. This method was useful for determining the glycoforms on the IgA1 molecule with respect to the core O-linked oligosaccharide structure.

Evidence for a site-specific fucosylation of N-linked oligosaccharide of immunoglobulin A1 from normal human serum.

Glycopeptides containing the N-linked oligosaccharide from human serum IgA1 were analyzed by matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOFMS). Two glycopeptides, GP1 and GP2, prepared from the endoproteinase Asp-N digest of the IgA1 heavy chain, were derived from the CH2 domain (N-glycan site at Asn263) and the tailpiece portion (N-glycan site at Asn459), respectively. The structure of the attached sugar chain was deduced from the mass number of the glycopeptide and confirmed by a two-dimensional mapping technique for a pyridylaminated oligosaccharide. GP1 was composed of two major components having a fully galactosylated bianntena sugar chain with or without a bisecting N-acetylglucosamine (GlcNAc) residue. On the other hand, the GP2 fraction corresponded to the glycopeptides having a fully galactosylated and fucosylated bianntena sugar chain partly bearing a bisecting GlcNAc residue. Thus, the site-specific fucosylation of the N-linked oligosaccharide on the tailpiece of the alpha1 chain became evident for normal human serum IgA1.

Structural determination of the O-linked sialyl oligosaccharides liberated from fetuin with endo-alpha-N-acetylgalactosaminidase-S by HPLC analysis and 600-MHz 1H-NMR spectroscopy.

The endo-alpha-N-acetylgalactosaminidase from the culture medium of Streptomyces sp. OH-11242 (endo-GalNAc-ase-S) hydrolyzed the O-glycosidic linkage between GalNAc and Ser (Thr) in fetuin, liberating oligosaccharides. The O-linked oligosaccharides liberated from the fetuin with endo-GalNAc-ase-S were pyridylaminated following fractionation on a Bio-Gel P-4 column. The structure of the pyridylaminated O-linked oligosaccharides from fetuin has been determined by reverse-phase HPLC and 600-MHz 1H-NMR spectroscopy. The chemical shifts and the coupling constants of pyridylaminated (PA) NeuAc alpha2-3Gal beta1-3GalNAc were refined by computer simulation of the spectrum. The structures of NeuAc alpha2-3Gal beta1-3(NeuAc alpha2-6)GalNAc-PA and NeuAc alpha2-3Gal beta1-3(NeuAc alpha2-3Gal beta1-4GlcNAc beta1-6)GalNAc-PA were determined by their structural reporter groups.

Estimation of the number of O-linked oligosaccharides per heavy chain of human serum IgA1 by matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOFMS) analysis of the hinge glycopeptide.

In our previous study, gas-phase hydrazinolysis was used to analyze the glycoform of the O-linked oligosaccharide of human serum IgA1. All O-linked oligosaccharide chains are known to be present in the hinge portion. However, the number of O-linked oligosaccharide chains on IgA1 remained unclear. In order to determine the number of linked sugar chains, we applied matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOFMS) to the hinge glycopeptide prepared from human serum IgA1. MALDI-TOFMS did not show clear peaks, probably due to the microheterogeneity of the structure of each sugar chain. However, elimination of peripheral sialic acid and galactose residues by sequential treatment with neuraminidase and beta-galactosidase gave clear mass spectra with several sharp peaks. On the basis of these spectra, we conclude that IgA1 prepared from normal human serum carries different numbers of sugar chains. There are two major populations, one contains five GalNAc residues and the other four GalNAc residues. On the other hand, the hinge glycopeptide prepared from myeloma IgA1 was composed mainly of one population containing four GalNAc residues. Earlier, we reported incomplete glycosylation of IgA1 isolated from the serum of an IgA1 myeloma patient. In this experiment, the presence of four O-linked oligosaccharides per heavy chain of IgA1 from a myeloma patient was found. The reason why only four out of five sites on the hinge glycopeptide were fully glycosylated in the IgA1 from the IgA1 myeloma patient is not clear.

Abundance of Gal beta 1,3GalNAc in O-linked oligosaccharide on hinge region of polymerized IgA1 and heat-aggregated IgA1 from normal human serum.

Gas-phase hydrazinolysis was used to analyze the glycoform of the O-linked oligosaccharide of human serum IgA1. In our previous report, only one glycoform was obtained from the IgA1 of healthy individuals. However, it was found to be composed of heterogeneous IgA1 components having mutually different glycoforms. First, the IgA1 was separated into two subfractions having different affinities toward jacalin. Among them, the high-affinity subfraction was mainly composed of polymerized IgA1. Comparative study of the carbohydrate chain showed a relative abundance of Gal beta 1,3GalNAc in the polymerized form. A simultaneous analysis of the N-glycan of these subfractions was also carried out. Three major components, two biantennary and one triantennary oligosaccharides, were obtained from both subfractions and the relative contents of these components were almost the same. On the other hand, IgA1 was artificially polymerized by heating at 63 degrees C for 2 h. The heat-stable IgA1 was separated from the heat-aggregated material on a Sephacryl S-300 column. The obtained heat-stable IgA1 (approximately 20%) was not further aggregated by more heating under the same conditions. The heat-stable IgA1 contained a much higher amount of the sialylated Gal beta 1,3GalNAc. Thus, it was shown that the degree of completeness of the hinge O-linked oligosaccharide might be correlated with the stability and polymerization process of the IgA1 molecule.

The presence of 2-keto-3-deoxygluconic acid and oxoaldehyde dehydrogenase activity in human erythrocytes.

2-Keto-3-deoxygluconic acid (3-DGA) is produced from 3-deoxyglucosone (3-DG:a highly reactive glycation intermediate) through oxidation by the enzyme oxoaldehyde dehydrogenase (OAD) in animals. We developed a specific assay method for 3-DGA using high-performance liquid chromatography [Fujii, E. et al. (1994) J. Chromatogr. B 660, 265-270] and measured it in the hemolysate and plasma of diabetic patients and healthy subjects. Both human erythrocytes and plasma contained considerable amounts of 3-DGA. However, human erythrocyte contained about 30-50 times higher 3-DGA than human plasma did and also had the same ability to convert 3-DG to 3-DGA as OAD had. Erythrocyte 3-DGA levels of diabetic patients were 990 +/- 370 nmol/gHb (n = 57, Mean +/- SD) and were significantly higher compared with healthy subjects (527 +/- 194 nmol/gHb, n = 7, p < 0.01). In all diabetic patients and healthy subjects (n = 64), there was only one patient who had a very low level of erythrocyte 3-DGA and lacked the ability to convert 3-DG to 3-DGA. When erythrocytes were incubated at 37 degrees C for 8 hours in phosphate buffer containing 0.35 mM 3-DG, 3-DG was easily taken into the erythrocytes and was converted to 3-DGA. Our results suggest the contribution of OAD not only to the prevention of glycation of hemoglobin but also to that of blood vessels by scavenging plasma 3-DG into erythrocytes.

Quantitation of the glycation intermediate 3-deoxyglucosone by oxidation with rabbit liver oxoaldehyde dehydrogenase to 2-keto-3-deoxygluconic acid followed by high-performance liquid chromatography.

A simple and sensitive method for the detection of 3-deoxyglucosone was developed using oxidation with crude oxoaldehyde dehydrogenase to 2-keto-3-deoxygluconic acid followed by high-performance liquid chromatography (HPLC). Oxoaldehyde dehydrogenase was prepared from rabbit liver and partially characterized. 2-Keto-3-deoxygluconic acid produced from 3-deoxyglucosone by oxoaldehyde dehydrogenase was derivatized with 1,2-diamino-4,5-methylenedioxybenzene, and the fluorescent products were detected and quantitated by HPLC using a solvent containing borate. In the presence of borate, 2-keto-3-deoxygluconic acid was completely separated from N-acetylneuraminic acid. The detection limit of 3-deoxyglucosone was 2.5 pmol/injection (10 microliters) at a signal-to-noise ratio of 3. This method was used to confirm the inhibitory effect of aminoguanidine on glycation.

Sustained release of phenytoin following the oral administration of phenytoin sodium/ethylcellulose microcapsules in human subjects and rabbits.

Phenytoin sodium was microencapsulated with ethylcellulose (EC) by a coacervation-phase separation method from ethyl acetate solution to develop a prolonged release dosage form of phenytoin. Release of phenytoin from the microcapsules (phenytoin sodium/EC) was evaluated by the JP dissolution test in JP disintegration media No. 1 and No. 2. The release rates of phenytoin from phenytoin sodium powders were extremely rapid in both media, however, the release rates from the microcapsules were much more retarded. Following the oral administration of microcapsules to rabbits, prolonged plasma concentrations of phenytoin were obtained, while microcapsules orally administered to human subjects showed prolonged urinary excretion of phenytoin metabolites.

Partial purification and characterization of an endo-alpha-N-acetylgalactosaminidase from the culture medium of Streptomyces sp. OH-11242.

For the purification of a new type of endo-alpha-N-acetylgalactosaminidase from the culture medium of Streptomyces sp. OH-11242 (endo-GalNAc-ase-S) [Iwase, Ishii, Ishihara, Tanaka, Omura & Hotta (1988) Biochem. Biophys. Res. Commun. 151, 422-428], a method for assaying enzyme activity was established. Using purified pig gastric mucus glycoprotein (PGM) as the substrate, oligosaccharides liberated from PGM were pyridylaminated, and the reducing terminal sugars of oligosaccharides larger than Gal beta 1-3GalNAc were analysed by h.p.1.c. The crude enzyme of endo-GalNAc-ase-S was prepared as an 80% (w/v) ammonium sulphate precipitate from the concentrated culture medium. The enzyme was partially purified by gel chromatofocusing and subsequent DEAE-Toyopearl chromatography. Endo-enzyme activity eluted around pI 4.8 on a gel chromatofocusing column and eluted with 0.19-0.25 M-NaCl on a DEAE-Toyopearl column. In the enzyme fraction obtained, no exo-glycosidases or proteases could be detected. The molecular mass of the enzyme was estimated as 105 kDa by gel filtration, and the optimum pH was 5.5. Endo-GalNAc-ase-S hydrolysed the O-glycosidic linkage between GalNAc and Ser (Thr) in 3H-labelled and unlabelled asialofetuin, liberating both the disaccharide (Gal beta 1-3GalNAc) and the tetrasaccharide [Gal beta 1-3 (Gal beta 1-4GlcNAc beta 1-6)GalNAc]. When endo-alpha-N-acetylgalactosaminidase from Alcaligenes sp. (endo-GalNac-ase-A) was incubated with 3H-labelled and unlabelled asialofetuin, only the disaccharide (Gal beta 1-3GalNAc) was liberated.

Analysis of glycoform of O-glycan from human myeloma immunoglobulin A1 by gas-phase hydrazinolysis following pyridylamination of oligosaccharides.

A comparative study was made on the glycoform of O-glycan from human myeloma immunoglobulin A1. By gas-phase hydrazinolysis, O-glycan was released from its hinge portion. The released oligosaccharide was pyridylaminated and separated by a two-dimensional analytical method of gel filtration and reverse-phase HPLC. Four major pyridylamino derivatives (P1-P4) were obtained. The neutral component (P4) among them was identified as Gal beta 1,3GalNAc-PA by cochromatography with an authentic standard pyridylamino sugar. The desialylation of the other components indicated the largest P1 and middle size P2 components possibly corresponded to a disialylated structure, NeuAc alpha 2,3Gal beta 1,3(NeuAc alpha 2,6)GalNAc-PA, and a monosialylated component, NeuAc alpha 2,3Gal beta 1,3GalNAc-PA, respectively. The structural assignment of P3 is still incomplete. Four similar components were also detected in bovine fetuin whose relative content (P1:P2: P3:P:4) was 16:43:19:22. The relative content (%) of P1-P4 (glycoform) in IgA1 from the healthy control was 10.1 +/- 3.3, 48.2 +/- 4.6, 7.0 +/- 2.6, and 34.7 +/- 4.5. The glycoform of O-glycan on IgA1 thus appears the same for any individual. Analysis of IgA1 myeloma protein indicated glycoforms distinct from those of the healthy controls. The relative content of these component could be classed as 2:8:0:90 (Type I, only one case designated as Kita), 5:24:3:68 (Type II, seven cases), and 9:41:5:45 (Type III, four cases). Thus, the results for IgA1 myeloma protein indicate that at least three glycoforms of O-glycan are possible for the IgA1 hinge structure. However, only one glycoform was found in the healthy controls.

Isolation and partial characterization of serine- and threonine-rich porcine gastric mucus glycopeptides.

1. Two subfractions from purified porcine gastric mucus glycopeptide were found to separate from each other by cesium chloride equilibrium centrifugation. The highest density fraction and two lower density fractions separated were designated VHD, HD and LD, respectively. A comparative study of these components was made. 2. The high and low density fractions, HD and LD, appeared almost the same or identical, while VHD differed completely from either of them in the following respects: (1) VHD exhibited strong alcian blue binding activity. (2) 57% of VHD bound to the DEAE-Toyopearl column equilibrated with 0.2 M NaCl. (3) VHD eluted from the Sephacryl S-400 column as a lower molecular subunit. (4) One third of the sialic acid as a minor component in VHD was constituted by N-glycolylneuraminic acid. (5) Carbohydrate composition showed typical mucus glycoprotein with slightly higher fucose content. (6) Amino acid compositions of the anionic components prepared from VHD showed the highest Ser/Thr ratio, 1.92 compared to 0.46 for LD and 0.62 for HD. (7) Oligosaccharide released from VHD by alkaline-sodium borohydride treatment was larger than that from HD or LD. 3. The above results indicate the minor component, VHD, separated from the major components, to be a quite similar but not identical component to the so-called sulfated mucus glycoprotein reported previously [Slomiany et al. (1972) J. biol. Chem. 247, 5062-5070].

Analysis of porcine gastric mucus glycoprotein added to a culture medium of Streptomyces sp. OH-11242 as the only source of carbon.

1. In the process of obtaining the degradation enzymes of mucus glycoprotein, porcine gastric mucus glycoprotein (PGM), added as the only source of carbon, was removed from the culture medium of Streptomyces sp. OH-11242 [Iwase et al. (1988) Biochem. biophys. Res. Commun. 151, 422-428] and analysed. 2. The amino acid and carbohydrate compositions of porcine gastric mucus glycoprotein (PGM-m) recovered from a culture medium were similar to those of original PGM. 3. However, the elution profile of PGM-m on Sephacryl S-400 differed from that of PGM and closely resembled that of performic acid-treated PGM or protease-treated PGM. 4. Either of these corresponded to the so-called subunit of approximately 550,000 in mol. wt, as reported by Scawen and Allen [(1977) Biochem. J. 163, 363-368]. 5. Performic acid treatment of PGM-m led to the production of a smaller unit (unit m) having a mol. wt of about 72,000. Separate treatment of different sized components prepared from PGM-m showed the above unit m to be produced from each molecule. 6. Thus, PGM-m is a molecule partly modified by various glycosidases including endo-alpha-N-acetylgalactosaminidase and exposure of the modified part to performic acid results in oxidation. 7. Production of unit m from both larger and smaller molecules indicates the part susceptible to performic acid to exist at regular intervals on the mucus glycoprotein molecule.(ABSTRACT TRUNCATED AT 250 WORDS)

Extraction method for preparing pyridylamino sugar derivatives and application to porcine gastric mucus glycoprotein analysis.

For the sensitive detection of free sugars and oligosaccharides, the use of their pyridylamino derivatives has now found general acceptance. To remove excess 2-aminopyridine from this derivative in a reaction mixture, gel filtration and ion-exchange chromatography were conducted. It was found in the present study that contaminated 2-aminopyridine could be selectively removed from the reaction mixture by adjusting the pH with saturated sodium bicarbonate at above 8.5 followed by extraction with benzene. By using this method, fewer purification steps and less time are required, with minimum loss of pyridylamino sugar derivatives.