Detailed Studies on Substrate Structure Requirements of Glycoamidases A and F*

Glycoamidases (peptide-N 4-(N-acetyl-β-glucosaminyl)asparagine amidase, EC 3.5.1.52; also known as peptide: N-glycanases (PNGases) release N-linked oligosaccharides from glycopeptides and/or glycoproteins by hydrolyzing the glycosylated β-amide bond of the asparagine side chain. The most widely used glycoamidases are those from Flavobacterium meningosepticum (glycoamidase F or PNGase F) and almond emulsin (glycoamidase A or PNGase A). To study the substrate structure requirement of these enzymes systematically, we synthesized >30 glycopeptides containing cellobiose, lactose, GlcNAc, and di-N,N′-acetylchitobiose (CTB). The length of the peptide was varied from one to five amino acids, and glycosylamines were linked to either Asn or Gln located at different positions in the peptide, including NH2 and COOH termini. Neither enzyme could cleave cellobiose and lactose glycopeptides, indicating that the 2-acetamido group on the Asn-linked GlcNAc is important in the recognition by both glycoamidases A and F. GlcNAc peptides could be cleaved by both enzymes, albeit not as effectively as CTB glycopeptides. Neither enzyme requires the Asn-Xaa-(Ser/Thr) sequence (required for N-glycosylation) for activity. Glycoamidase A could even hydrolyze a Gln-bound CTB glycopeptide, whereas the action of glycoamidase F on this substrate is minimal. While glycoamidase A could act on CTB dipeptides, glycoamidase F preferred a tripeptide or longer. The K m andV max values of glycoamidase A fort-butoxycarbonyl-(CTB)-Asn-Ala-Ser-OMe were 2.1 mm and 0.66 μmol/min/mg, respectively. A natural glycodipeptide, Man9-GlcNAc2-Asn-Phe, was also completely hydrolyzed by glycoamidase A.

Although the biological function of the enzyme is increasingly attracting attention, the most popular use of the enzyme currently remains to be de-N-glycosylation of glycoproteins/ glycopeptides, most often for structural elucidation of glycoconjugates (15,16). The commercially available glycoamidases are from sweet almond (glycoamidase A) (2) and Flavobacterium meningosepticum (glycoamidase F) (4). Both enzymes act on all three types of N-linked oligosaccharide chain, i.e. high mannose, hybrid, and complex types (17,18). However, the Nglycoside chain with a Fuc(␣1,3)GlcNAc-Asn segment, found in glycoproteins from plants and insects (19 -22), could be liberated only by glycoamidase A, but not by glycoamidase F (23). It is also acknowledged that a certain length of peptide is needed for the enzyme activities of both enzymes. Recently, Altmann et al. (24) showed, by stepwise degradation of a biantennary glycopeptide with exoglycosidases, that the size of the carbohydrate moiety in the substrate has little influence on the enzyme activities of both glycoamidases and that the hydrolysis rates of both enzymes may be primarily determined by the length of the peptide. However, the glycopeptides used in this study contained at least three amino acids, and the glycopeptides with one or two amino acids were not studied.
To study the influence of the length and sequence of the peptide part on the glycoamidase activity, we synthesized four series of glycopeptides, with each series containing cellobiose, 1 lactose, GlcNAc, or di-N,NЈ-acetylchitobiose and containing one to five amino acids with the sugar attached to either Asn or Gln. The results obtained with these substrates are presented below.
grade or higher purity.
General Methods-Unless otherwise specified, reactions were carried out at 22-24°C. Organic layers from various extractions were dried over anhydrous sodium sulfate. Evaporation was performed with a rotary evaporator at 20 -45°C under diminished pressure. For thinlayer chromatography, Silica Gel 60 F 254 (coated on an aluminum sheet, layer thickness of 0.25 mm; E. Merck AG, Darmstadt, Germany) was used. Spray reagents for TLC were 15% (v/v) sulfuric acid in 50% (v/v) ethanol for carbohydrates and 0.5% (w/v) ninhydrin in 95% ethanol for amino groups. Preparative column chromatography of synthetic compounds was performed with Silica Gel 60 (15-40 m; E. Merck AG). 1 H NMR spectra were recorded with a Brucker WH 360 spectrometer using solutions in CDCl 3 (for acetylated sugars) or Me 2 SO-d 6 (for protected glycopeptides).
Enzymatic Reaction with Glycoamidases A and F-A mixture of 10 nmol of substrate and 0.2 milliunit of glycoamidase A or 0.5 milliunit of glycoamidase F in 15 l of 0.1 mM ammonium acetate buffer (pH 5.0 for glycoamidase A and pH 8.0 for glycoamidase F) was incubated at 37°C for 18 h. The mixture was evaporated with a SpeedVac using a vacuum pump and redissolved in 200 l of H 2 O. One-fourth of the sample was analyzed using the HPAEC-PAD system (see below), and the enzymatic reaction product was quantified by comparison of peak areas with standard compounds analyzed in a separate run. One enzyme unit is defined as the quantity of enzyme that releases 1 mol of oligosaccharide from the ovalbumin glycopeptide Gln-Glu-Lys-Tyr-(Man 5 -GlcNAc 3 )-Asn-Leu-Thr-Ser-Val for glycoamidase A or from ribonuclease B for glycoamidase F per min, as suggested by the manufacturers.
High Performance Anion-exchange Chromatography-The HPAEC system consisted of a Bio-LC (Dionex Corp., Sunnyvale, CA) equipped with a pulsed amperometric detector (PAD-II), and a Dionex CarboPac PA-1 column (4 ϫ 250 mm) was used for analysis of the enzymatic reaction products. The column was eluted at a rate of 1.0 ml/min with 100 mM sodium hydroxide for 5 min, followed by a linear gradient of sodium acetate (0 -250 mM) developed in 25 min. A cycle of a 5-min washing with 250 mM sodium acetate and a 15-min equilibration with the starting eluent was inserted between runs. The PAD sensitivity was set at 1 A.
High Performance Liquid Chromatography-The HPLC purification of the synthetic product was performed with a Gilson HPLC system equipped with a Rheodyne 7125 injector, a Fiatron CH-30 column heater, and a UV detector (Model V 4 , Isco Inc., Lincoln, NE). The flow rate was 1.0 ml/min, with the effluent being monitored at 215 nm. ODS columns (4.6 ϫ 250 mm; Rainin Instrument Co. Inc. (Woburn, MA) and Shiseido (Tokyo, Japan)) and a Hypercarb graphitized carbon column (4.6 ϫ 100 mm; Shandon Scientific, Cheshire, United Kingdom) with a Direct-Connect guard cartridge column (Alltech Associates Inc., Deerfield, IL) were used for separation depending on the hydrophobicity of the compounds. The eluent was 25 mM ammonium acetate buffer (pH 6.0) with a linear gradient of CH 3 CN (10 -40% for ODS columns and 0 -20% for the graphitized carbon column) developed in 40 min at 40°C. The purified glycopeptides were freeze-dried and stored at Ϫ20°C.
Amino Acid Analysis-The synthetic glycopeptides purified by HPLC were hydrolyzed by 4 M HCl at 100°C for 6 h, and the acid was removed by evaporation with a vacuum pump. The hydrolysates were then reacted with phenyl isothiocyanate for amino acid determination according to a published method (28). The phenylisothiocarbamyl-derivatives were separated and quantified by HPLC with an ODS column (4.6 ϫ 250 mm; PhaseSeparation, Norwalk, CT).

Synthesis of Cellobiose-and Lactose-containing Glycopeptides
The synthetic strategy for cellobiose glycopeptides is summarized in Scheme 1. The synthesized glycopeptides are listed in Table I.
Per-O-acetylcellobiosyl bromide (33) was prepared from cellobiose (32) by a one-pot reaction (29) and then converted to per-O-acetylcellobiosyl azide (34) In Method A, a glycosylamine derivative was condensed with Boc-Asp-O-Bzl, followed by stepwise elongation of the peptide portion. In Method B, a glycosylamine was coupled to the preformed peptide.
b All carbohydrates are attached to the ␤-amide group of Asn or to the ␥-amide group of Gln. c Cel, cellobiose; Lac, lactose.
pentapeptides were prepared by Method B described below. Synthesis of lactose glycopeptides was based on the same strategy as shown in Scheme 1.

Synthesis of GlcNAc-and CTB-containing Glycopeptides
GlcNAc and CTB glycopeptides were synthesized by either of the following two procedures, as exemplified by the CTB glycopeptides.
Method A (Stepwise Elongation)-This synthetic strategy for CTB glycopeptides is summarized in Scheme 2. Di-N,NЈ-acetylchitobiose (39) was per-O-acetylated in pyridine, and then the isolated crystalline peracetate (40) was converted to glycosyl azide (42) via its oxazoline derivative (41) according to a published method (30). Compound 42 was reduced to chitobiosylamine (43) by hydrogenation and subsequently condensed with Boc-Asp-O-Bzl to provide 44. Compound 44 served as the building block for synthesis of various CTB glycopeptides. A stepwise elongation strategy by cycles of COOH-or NH 2 -terminal deprotection and condensation with the desired amino acids followed by de-O-acetylation with methanolic ammonia led to a series of CTB glycopeptides.
Method B (En Bloc Synthesis)-A simpler strategy was also used to synthesize diverse glycopeptides as described in Scheme 3. Cellobiose (32), lactose (49), GlcNAc (50), and CTB (39) were directly converted to the corresponding glycosylamines (31-33) by dissolving in a saturated NH 4 HCO 3 solution and keeping at room temperature for 10 days, followed by repeated lyophilization to remove excess NH 4 HCO 3 . The peptide portion was synthesized by stepwise elongation as shown in Scheme 3. The ␤-O-Bzl of Asp was removed by hydrogenation to expose the ␤-carboxyl group for subsequent coupling with the glycosylamine. The condensation of peptide with glycosylamine using 2-ethoxy-1-ethoxycarbonyl-1,2-dihydroquinoline as a condensing agent gave a Ͼ60% yield. The glycopeptides thus obtained were purified by gel filtration on a Sephadex G-10 column (1 ϫ 60 cm) eluted with 20 mM ammonium acetate buffer (pH 6.0), followed by lyophilization to yield white powder.

Identification of Synthetic Glycopeptides
Glycopeptides with O-acetylated sugars were analyzed by 1 H NMR analysis in Me 2 SO-d 6 . The signals were assigned by decoupling. Table II shows the chemical shifts of some of the glycopeptides. All H-1 signals were doublet (J 1,2 ϳ 8 Hz), sup-

Action of Glycoamidases A and F on Synthetic Glycopeptides
All the glycopeptides synthesized were tested as substrates under the conditions described under "Experimental Procedures" and also in Footnote a to Table III. Results are summarized in Table III. The following conclusions can be made from  Table III. 1) Neither glycoamidase A nor F can hydrolyze any of the lactose and cellobiose glycopeptides. 2) When the carbohydrate is linked to a single amino acid (Asn or Gln), neither enzyme can hydrolyze such glycopeptides, even when both ␣-NH 2 and ␣-COOH groups are masked. A natural glycopeptide containing only Asn (i.e. Man 9 -GlcNAc 2 -Asn) could not be hydrolyzed either. 3) Glycopeptides containing only a single GlcNAc can be cleaved by both enzymes, albeit not as effectively as CTB peptides. Fig. 1 (A and B) shows examples of GlcNAc release from Cbz-(GlcNAc)-Asn-Ala-Thr-OMe (11) by glycoamidases A and F, respectively. Hydrolysis of the GlcNAc peptide (13) by glycoamidase F was considerably slower (3.3%) than that of the corresponding CTB peptide (27) (98%). 4) The length and the nature of the peptide affect the hydrolysis by both enzymes. The CTB pentapeptide (27) was totally hydrolyzed by both enzymes under the conditions described for Table III. However, the rate of hydrolysis of CTB dipeptides by glycoamidase A varied considerably, depending on the position and the type of amino acids. CTB dipeptides with Asn at the COOH terminus (17 and 18) were much poorer substrates than the dipeptides with Asn at the NH 2 terminus (16 and 19). The contrast between 18 and 19 is especially striking; the reversal in the order of Asn and Phe caused the extent of hydrolysis to drop from 100% (Asn-Phe) to 9.2% (Phe-Asn). Among the glycotripeptides, the presence of Gly at the COOH terminus seems to lower the hydrolysis rate. This is true for both enzymes and for both the GlcNAc peptide (12) and the CTB tripeptide (22). Both enzymes could hydrolyze Gln glycopeptides (Fig. 1C), but at diminished rates (compare 27 versus 31 and 20 versus 30). 5) The consensus sequence for N-glycosylation is not required by either enzyme (Fig. 1D). 6) Glycoamidase F tends to work more sluggishly than glycoamidase A (Fig. 2, compare A and B).

Kinetic Constants of Glycoamidase A Digestion of Boc-(CTB)-Asn-Ala-Ser-OMe (20) as Substrate
Boc-(CTB)-Asn-Ala-Ser-OMe (20) was used as substrate for the kinetic study of glycoamidase A. The initial rate (v) was determined by HPAEC analysis. The K m and V max values of the enzyme were 2.1 mM and 0.66 mol/min/mg, respectively (from the Lineweaver-Burk plot shown in Fig. 3).

Hydrolysis of a Natural Glycodipeptide, Man 9 -GlcNAc 2 -Asn-Phe, by Glycoamidase A
Glycoamidase A could totally hydrolyze a glycodipeptide, Man 9 -GlcNAc 2 -Asn-Phe, derived from soybean agglutinin. As shown in Fig. 4, the substrate disappeared completely after incubation of the enzyme with substrate at 37°C for 18 h, and a new peak eluted at the position corresponding to Man 9 -GlcNAc 2 , clearly indicating that the substrate was digested by glycoamidase A. However, glycoamidase F failed to hydrolyze the same substrate (data not shown).

Synthesis of Glycopeptides-
We have chemically synthesized a number of glycopeptides to systematically study the substrate requirement of glycoamidases A and F. The chemical construction of the glycosylamine-␤-L-aspartyl bond has been achieved with protected sugars and protected Asp by various methods (34 -37), and the Asp was further extended to glycopeptides (38,39). When we, as well as others (40,41), followed a published method (34) for glycopeptide synthesis, the target compounds could be obtained only after extensive purification, and the yields were low. However, direct condensation of unprotected glycosylamines with L-aspartic acid and/or L-aspartic acid-containing peptide (42,43) provided much higher overall yields and produced fewer by-products.
Unnatural Glycopeptides-We tested the unnatural cellobiose and lactose glycopeptides to determine whether the 2-   acetamido group is required by the enzymes. Neither glycoamidase A nor F could act on any cellobiose and lactose glycopeptides. Since both enzymes released GlcNAc from GlcNAc-containing glycopeptides (e.g. 13), the resistance of cellobiose and lactose glycopeptides to the glycoamidases clearly indicates that at least the 2-acetamido group of the GlcNAc linked to Asn is involved in the recognition by the enzymes. Our data agree well with the results of a recent crystallographic analysis (44), which showed that the N-acetyl group of the Asn-linked GlcNAc is wrapped into the glycoamidase F molecule and makes contact with Asp-60, the purported primary catalytic residue of glycoamidase F. The glycopeptides with carbohydrate linked to the ␥-amide nitrogen of Gln instead of the ␤-amide nitrogen of Asn could also be hydrolyzed by glycoamidase A, with a slower hydrolytic rate. This indicates that glycoamidase A is not sensitive to the distance between the carbohydrate and the peptide backbone.
Oligosaccharide Chain Length-It has been reported that glycoamidase A is capable of releasing GlcNAc from peptides containing a single GlcNAc (17,24), although at 15-3000 times slower rates than the corresponding glycopeptides with larger glycans, but glycoamidase F requires at least the di-N,NЈacetylchitobiosyl core unit because the enzyme is unable to cleave the GlcNAc-Asn bond in ribonuclease B or external invertase treated with endo-␤-N-acetylglucosaminidase H (45). In agreement with the published results, we found that glycoamidase A could hydrolyze all of the synthetic peptides containing only GlcNAc if the peptide was larger than a tripeptide, although the hydrolytic rates were slower than those of the corresponding CTB-bearing peptides. However, contrary to the previous results (45), we found that glycoamidase F was also capable of acting on a GlcNAc tripeptide, Cbz-(GlcNAc)-Asn-Ala-Thr-OMe (11), and a GlcNAc pentapeptide, Boc-Tyr-Ile-(GlcNAc)-Asn-Ala-Ser-OMe (13), although at rates slower than those of glycoamidase A. Interestingly, glycoamidase F failed to act on Cbz-(GlcNAc)-Asn-Ala-Ser-OMe (10), which was a fair FIG. 1. Determination of glycoamidase digestion of glycopeptides. The enzymatic reactions were carried out as described under "Experimental Procedures," except that the glycoamidase F used was from Boehringer Mannheim. The hydrolysates were analyzed by the HPAEC-PAD system. The eluents were 16 mM isocratic NaOH for A and B and 100 mM NaOH with a gradient of NaOAc (0 -400 mM) developed in 30 min for C and D. The peaks at ϳ12 min in A and B and at ϳ6 min in C and D were coincidental with GlcNAc and CTB, respectively. The chromatograms shown in C and D were obtained by subtraction of the chromatogram of substrate alone from those obtained from the reaction mixture. Therefore, the hydrolytic product appeared as a peak, and the digested substrate formed a negative peak. A, glycoamidase A with 11; B, glycoamidase F with 11; C, glycoamidase A with 31; D, glycoamidase A with 28. Length of the Peptide Chain-The structural requirement on the oligosaccharide chains for the hydrolysis by glycoamidases was studied previously by trimming the oligosaccharide chain with various exoglycosidases (24,45). Only small differences were noted between the di-N,NЈ-acetylchitobiose and a larger sugar chain, and it was concluded that the hydrolytic rates are controlled more by the peptide portion than by the carbohydrate portion. In this study, we tried to determine the minimum peptide requirement of the two enzymes by using a series of synthetic glycopeptides with di-N,NЈ-acetylchitobiose.

FIG. 2. Time courses of glycoamidase A (A) and glycoamidase F (B) digestion of Boc-Tyr-Ile-(CTB)-Asn-Ala
Neither enzyme could release sugars from GlcNAc (9), CTB (15 and 29), or Man 9 -GlcNAc 2 bound to Asn or Gln, even when both ends were protected, thus agreeing with published results (1). The minimum peptide length for glycoamidase A seems to be a dipeptide; Man 9 -GlcNAc 2 -Asn-Phe as well as all end-protected CTB dipeptides tested (16, 17, 18, and 19) were hydrolyzed by the enzyme, although the rate of hydrolysis varied among the peptides of different sequences. Glycoamidase A seems to prefer those substrates with glycosides at the NH 2 terminus (16 and 19) to those with them at the COOH terminus (17 and 18). The fact that Man 9 -GlcNAc 2 -Asn-Phe could be totally cleaved by glycoamidase A (Fig. 4) suggests that the protection of NH 2 and COOH termini was not necessary. This differs from the early claim that glycoamidase A can only act very slowly on the glycopeptide when the glycosylated Asn is at the NH 2 -or COOH-terminal position (6,17). In contrast, among the glycodipeptides tested, only Boc-(CTB)-Asn-Ala-OMe (16) and Boc-Ile-(CTB)-Asn-OMe (17) could be hydrolyzed by glycoamidase F, but at extremely slow rates (1.8 and 0.9% with 16 and 17, respectively). The hydrolytic rates with glycotripeptides were significantly improved over those with glycodipeptides, suggesting that glycoamidase F requires at least a tripeptide for the activity.
A Thr at the C-2ϩ position of Asn (21) improved the glycoamidase F activity compared with that of the Ser-containing glycotripeptide (20). Likewise, Cbz-(GlcNAc)-Asn-Ala-Thr-OMe (11) was cleaved by glycoamidase F (Fig. 1B), but Cbz-(GlcNAc)-Asn-Ala-Ser-OMe (10) was not. Interestingly, this difference was not observed with glycoamidase A. The glycotripeptide with a glycosyl-Asn at the COOH terminus (e.g. 23) is a poorer substrate for both enzymes than those in which the glycoside chain is located at the NH 2 terminus (e. g. 20 and 21). Extending one amino acid at the COOH terminus (25) did not yield significant improvement in glycoamidase F activity, but additional sequence resulted in a significant increase in the hydrolytic rate. Thus, having a dipeptide sequence at the COOH-terminal side of Asn seems to be important for glycoamidase F activity.
Neither enzyme requires a strict Asn-Xaa-(Ser/Thr) sequence needed for natural glycosylation because both enzymes could release CTB from Boc-Tyr-Ile-(CTB)-Asn-Ala-Gly-OMe (28). The weak activities of both enzymes for Boc-(CTB)-Asn-Ala-Gly-OMe (22), on the other hand, may suggest that Gly plays an inhibitory role because the glycoamidase A reactivity was reduced from 48% for Boc-(CTB)-Asn-Ala-OMe (16) to 3.8% by adding a Gly at the COOH terminus.
Kinetic Constants-The kinetic constants of glycoamidases A and F for various glycopeptides have been investigated by several groups (1,3,24,46,47). The K m value of glycoamidase A was reported to be 4 mM for a glycopentapeptide with a heptaoligosaccharide from pineapple stem bromelain (3) and 1.1 mM for the fetuin tryptic glycopeptide derived from the Asn-138 site (1). We found that the K m value of glycoamidase A for Boc-(CTB)-Asn-Ala-Ser-OMe is 2.1 mM, which is comparable with the literature values for naturally occurring glycopeptides, again indicating that CTB is as effective as larger oligosaccharides and that a glycotripeptide is as effective as larger glycopeptides for glycoamidase A.
Conclusion-Our results revealed finer structural requirements for glycoamidases A and F, especially for di-or tripeptides. A tripeptide or a tetrapeptide is required for full expression of the activity of glycoamidase or F. The Asn-Xaa-(Ser/Thr) sequence is not mandatory, and Asn can be replaced with Gln. Recently, Kuhn et al. determined the three-dimensional structure of glycoamidase F by x-ray crystallography (48) and the carbohydrate-binding site of the enzyme by site-directed mutagenesis (44). However, the precise interaction of the enzyme with the peptide portion of the substrate is still unclear. Our results will be useful for the further elucidation of the catalytic mechanism of glycoamidase F.