Active site and oligosaccharide recognition residues of peptide-N4-(N-acetyl-beta-D-glucosaminyl)asparagine amidase F.

Crystallographic analysis and site-directed mutagenesis have been used to identify the catalytic and oligosaccharide recognition residues of peptide-N4-(N-acetyl-beta-D-glucosaminyl)asparagine amidase F (PNGase F), an amidohydrolase that removes intact asparagine-linked oligosaccharide chains from glycoproteins and glycopeptides. Mutagenesis has shown that three acidic residues, Asp-60, Glu-206, and Glu-118, that are located in a cleft at the interface between the two domains of the protein are essential for activity. The D60N mutant has no detectable activity, while E206Q and E118Q have less than 0.01 and 0.1% of the wild-type activity, respectively. Crystallographic analysis, at 2.0-A resolution, of the complex of the wild-type enzyme with the product, N,N'-diacetylchitobiose, shows that Asp-60 is in direct contact with the substrate at the cleavage site, while Glu-206 makes contact through a bridging water molecule. This indicates that Asp-60 is the primary catalytic residue, while Glu-206 probably is important for stabilization of reaction intermediates. Glu-118 forms a hydrogen bond with O6 of the second N-acetylglucosamine residue of the substrate and the low activity of the E118Q mutant results from its reduced ability to bind the oligosaccharide. This analysis also suggests that the mechanism of action of PNGase F differs from those of L-asparaginase and glycosylasparaginase, which involve a threonine residue as the nucleophile.

Peptide-N 4 -(N-acetyl-␤-D-glucosaminyl)asparagine amidase F (PNGase F, 1 peptide N-glycanase) is a 34.8-kDa amidohydrolase secreted by Flavobacterium meningosepticum (1). The enzyme cleaves the ␤-aspartylglucosamine bond of asparaginelinked oligosaccharides, converting the asparagine residue to an aspartic acid (2). The 1-amino-oligosaccharide product spontaneously hydrolyzes to ammonia and the intact oligosaccharide chain (3). The enzyme has a broad substrate specificity, but both the amino and carboxyl groups of the asparagine residue have to be in peptide linkage, while the oligosaccharide must consist at least of the N,NЈ-diacetylchitobiose core, GlcNAc␤134GlcNAc. The enzyme is highly sensitive to modifications of this core; an ␣-133-fucose substituent on the asparagine-proximal, or reducing end, GlcNAc completely blocks PNGase F activity (4), but an ␣-136-fucose substituent has no effect. The enzyme is extensively used as a biochemical tool for the study and analysis of glycoproteins (5), including the analysis of oligosaccharides as well as the deglycosylation of glycoproteins for structural analysis.
Two crystal structures of PNGase F, in different crystal forms, were recently published (6,7). The molecular structure, shown in Fig. 1, consists of two eight-stranded antiparallel ␤-sandwiches that lie side-by-side so that the interface runs the full length of the ␤-sheets. Very little information regarding the mechanism of action of the enzyme was available when the structures were determined, and the assignment of a possible active site was based primarily on information obtained from the studies of other asparaginases: L-asparaginase (8) and glycosylasparaginases (9,10). The active site of L-asparaginase includes a triad of hydrogen-bonded Asp-Lys-Thr residues, where the threonine is thought to be the nucleophile in the reaction mechanism. Glycosylasparaginases also have an essential threonine residue. This information, combined with observation of surface features of the molecule, led to the identification of three possible substrate binding sites (6): a bowl on one face of the molecule that contains a group of residues that somewhat resemble the active site of L-asparaginase, a shallow S-shaped cleft on the opposite face of the molecule that contains a number of acidic residues and threonine residues, and a deep cleft at the interface between the two domains at one end of the molecule. This cleft, formed by the loops connecting ␤-strands between the ␤-sheets contains several acidic residues and serines, as well as many aromatic residues. Most remarkable is the presence of five tryptophan residues.
In this paper we describe the results of site-directed mutagenesis studies, resulting in the identification of the active site area, and of the crystallographic analysis of the complex of the enzyme with the product disaccharide N,NЈ-diacetylchitobiose. These studies lead to the unambiguous identification of the catalytic residues and the detailed analysis of the oligosaccharide recognition site.

MATERIALS AND METHODS
Site-directed Mutagenesis-The coding sequence of PNGase F was cloned into pLITMUS 29 (New England Biolabs). Site-directed mutagenesis was performed following the Kunkel protocol (11), using 39base pair oligonucleotides bearing the site of mutation at positions 22-24. All substitutions were confirmed by DNA sequencing at the mutation site in the pLITMUS 29 vector and again by sequencing of the full PNGase F gene in the pMAL p2 construct. Single colonies were picked until at least two independent clones could be derived that * This work was supported in part by Grants GM-50431 (to P. V. R.) and GM-30471 (to T. H. P., and A. L. T.) from the National Institutes of Health, United States Public Health Service. 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.
harbored the same mutation. The New England Biolabs pMAL p2 fusion kit and protocol were used for expression and purification. Cells were grown at 37°C and expression was induced after 3 h by addition of 1 mM isopropyl-1-thio-␤-D-galactopyranoside. Incubation was continued for 8 h at 30°C followed by 12 h at room temperature. After purification over an amylose column, the fusion protein was cleaved using a 12-h room temperature incubation with 3% Factor Xa, resulting in an 85% yield of the cleavage product. A Q-Sepharose (Pharmacia Biotech Inc.) column was used for the final purification to homogeneity. Enzymatic activity of the fusion proteins was measured by enzyme concentration-dependent gel-shift assays of the deglycosylation of denatured RNase B. An HPLC assay (12), using a didansylfetuin glycopeptide, Leu-Ala-Asn(CHO)-AeCys-Ser as the substrate, was used to compare the activities of the most interesting mutants with that of the wild-type enzyme and for the competitive inhibition experiments.
X-ray Crystallography-Solutions containing a 30:1 molar ratio of N,NЈ-diacetylchitobiose (Fine grade, Seikagaku America) to PNGase F were prepared by adding lyophilized disaccharide to 100-l volumes of protein at concentrations of about 10 mg/ml. These solutions were incubated at 37°C for 1 h prior to set-up of crystallization experiments. X-ray diffraction data for the N,NЈ-diacetylchitobiose complex of the wild-type enzyme were measured using a Rigaku R-Axis image plate area detector equipped with focussing mirrors. The crystal was flashcooled to approximately Ϫ140°C immediately after addition of an equal volume of 50% glycerol to the crystallization drop. A total of 104,022 reflections were measured to 2.0-Å resolution. These were averaged to unique 26,902 data with an R sym (⌺͉I Ϫ ϽIϾ͉/⌺ϽIϾ) of 0.044. A F Ͼ 2(F) cut-off was applied to the data, which are 94.1% complete. The structure was refined with the program X-PLOR (13), using the protein coordinates of the room temperature wild-type structure as the starting point. After the initial simulated annealing refinement of the protein molecule, electron density for the N,NЈ-diacetylchitobiose was observed in (2F o Ϫ F c ) and (F o Ϫ F c ) maps displayed with the program CHAIN (14). Further refinement using alternating X-PLOR molecular dynamics refinement and model building led to an R value of 0.197 for the data between 10.0 and 2.0 Å. The validity of the N,NЈ-diacetylchitobiose modeling was checked intermittently with omit maps that were calculated after simulated annealing refinement in which atoms within an 8-Å sphere of the binding cleft area were omitted. The final model consists of 2458 protein atoms, 29 N,NЈ-diacetylchitobiose atoms, one sulfate anion, and 266 water molecules. The root mean square deviations from ideality of the final model are: bonds 0.010 Å, angles 1.67°, dihedrals 27°. Only Trp-86 falls outside the allowed regions of the Ramachandran plot, while 90.4% of the non-glycine residues are in the most favored region as defined in the program PROCHECK (15). The average thermal parameters are 18.4 Å 2 for the main chain atoms, 19.7 Å 2 for the side chain atoms, and 16.6 Å 2 for the chitobiose. Details of the structure analysis of the mutant enzymes and their complexes will be published elsewhere. Table I lists the mutants prepared as part of the search for catalytic residues. Initially, selected residues in the three possible substrate binding sites were mutated: acidic residues to the corresponding amides and serine or threonine residues to alanine. Two of the mutants of residues in the bowl area, T101A and D102N, have somewhat reduced activity, while the third one, D99N, is fully active. The latter residue was considered to be the most likely candidate for the acidic residue in the L-asparaginase-like active site. Likewise, the two mutants of residues in the S-shaped cleft, T71A and T72A, are fully active. Therefore, these two areas were eliminated from further consideration. The cleft at the interface between the two domains at the top end of the molecule has two areas of interest that are about 12 Å apart at opposite ends of the cleft. Both areas include acidic residues as well as other polar residues that could be involved in catalysis or substrate recognition. One group includes Glu-118, Ser-154, Ser-155, and Asp-157. The E118Q mutant has greatly reduced activity, but the S154A, S155A, and D157N mutants are fully active, indicating that only the glutamic acid directly interacts with the substrate. At the other end of the cleft, Asp-60 and Glu-206 are in hydrogen bonding contact with each other via a very tightly bound water molecule. This water molecule also makes additional contacts with Tyr-85 and Arg-248. When the acidic residues were mutated, the D60N mutant was found to have absolutely no activity, while the activity of E206Q was extremely low. To further evaluate this site, three additional mutants were prepared: D60E, E206D, and Y85F. The very low activity levels of the acidic residue mutants confirm their role in catalysis while suggesting strict geometric requirements. The hydroxyl group of Tyr-85, however, is not of great importance in view of the relatively high activity of the Y85F mutant. No other hydroxyl groups are present in the active site cleft. Therefore, the mechanism of action of PNGase F, unlike those of L-asparaginase and glycosylasparaginase, does not appear to involve an hydroxyl group as the nucleophile.

Site-directed Mutagenesis-
The D60N, E206Q, and E118Q mutants were overexpressed  in E. coli, purified in large quantities, and crystallized. All three crystallized in the same space group under nearly identical conditions as the wild-type enzyme (16). The crystallographic analysis showed that their structures are essentially identical to that of the wild-type enzyme, confirming that the reduced activities are caused by diminished catalytic functionality or impaired substrate binding, not by altered protein conformation. 2 X-ray Crystallographic Analysis of the N,NЈ-Diacetylchitobiose Complex-PNGase F hydrolyzes glycopeptides containing the minimal oligosaccharide moiety N,NЈ-diacetylchitobiose, GlcNAc␤134GlcNAc. The initial oligosaccharide product of the PNGase F hydrolysis contains the 1-amino-N,NЈ-diacetylchitobiose core, which spontaneously degrades to N,NЈ-diacetylchitobiose. This compound was tested as a potential inhibitor against the standard PNGase F substrate, dansyl-Leu-Ala-Asn(CHO)-dansyl-AeCys-Ser using an HPLC assay (12). N,NЈ-Diacetylchitobiose inhibited substrate hydrolysis by 36% at a 12.5-fold molar excess over substrate but by only 50% at a 100-fold excess. Even though the inhibitory activity of the disaccharide was relatively poor, co-crystallization experiments were conducted, using a 30:1 molar ratio of N,NЈ-diacetylchitobiose to PNGase F.
The crystals of the complex were grown under identical conditions and belong to the same space group as those of the uncomplexed protein. The structure of the complex of the wildtype enzyme has been fully refined to an R value of 0.197 at 2.0-Å resolution. Fig. 2 shows the electron density of the disaccharide in the refined structure. The protein conformation is essentially unaffected by the binding of the ligand. The leastsquares fit of the main chain atoms of all but the four Nterminal residues has a root mean square deviation of 0.33 Å and a maximum deviation of 1. Aromatic residues also make important contacts with the substrate. Trp-191 is positioned nearly perpendicular to the disaccharide and forms a hydrogen bonding contact rather than a hydrophobic interaction. Trp-120 forms a hydrogen bond with O 6Ј but also appears to be positioned correctly to be able to make a hydrophobic contact with the next residue, the first mannose, in the intact substrate. The figure was prepared with the program SECTOR (22). position of N ⑀1 by 0.95 Å. The average thermal parameters for the protein molecule are significantly lower in the complex: 18.4 Å 2 and 19.5 Å 2 for the main chain and side chain atoms in the complex, compared with 23.2 and 25.4 Å 2 in the uncomplexed structure. Interestingly, the change in the thermal parameters is uniformly distributed over the molecule and does not result from large changes in the binding cleft area. The N,NЈ-diacetylchitobiose molecule has an average thermal parameter of 16.6 Å 2 with a range from 11.0 to 26.0 Å 2 . The N-acetyl group of the first GlcNAc is most deeply buried into the protein molecule and correspondingly has the lowest thermal parameters.
In solution, N,NЈ-diacetylchitobiose exists in an equilibrium state containing a mixture of ␤and ␣-configurations of the O 1 -hydroxyl group at the reducing end. In the wild-type PN-Gase F complex structure, only the ␣-conformation is observed, despite the fact that the oligosaccharide moiety attached to the N ␦ of the asparagine is in the ␤-configuration. The glycosidic link between the GlcNAc residues is in an extended conforma-tion with torsion angles, (O 5 -C 1 -O 4Ј -C 4Ј ) Ϫ85°and (C 1 -O 4Ј -C 4Ј -C 5Ј ) Ϫ123°. These angles are well within the range of those observed for other ␤-134-linked disaccharides in complexes with lectins (17,18) but are most similar to those seen in complexes of lysozyme with GlcNAc multimers (19,20). Fig. 3 shows the location and orientation of the N,NЈ-diacetylchitobiose molecule in the cleft of the enzyme. The molecule is inserted edgewise into the cleft and spans the distance between the two groups of acidic residues in the cleft, with O 1 of the reducing end GlcNAc residue forming hydrogen bonds with Asp-60 and the water molecule, Wat 346 , that connects Asp-60 and Glu-206. O 6 of the second GlcNAc forms a hydrogen bond with Glu-118. Fig. 4 shows a schematic diagram of all important intermolecular contacts. The first GlcNAc residue makes extensive contacts with the protein, although several are mediated by tightly bound water molecules. One interesting feature, shown in The solvent-accessible surface of the protein is reduced by less than 200 Å 2 in the complex: 12,395 Å 2 for the protein in the complex versus 12,587 Å 2 in the uncomplexed structure. This is explained in part by the location of the molecule toward one side of the cleft, making close contacts with many residues of the N-terminal domain but allowing for a layer of water molecules to form bridging contacts with the residues of C-terminal domain. Many of these water molecules, including Wat 146 , Wat 75 , and Wat 346 , are in nearly identical positions in the uncomplexed and complex structures. Of all the water molecules involved in contacts with the protein and the disaccharide, only Wat 348 and Wat 349 have no water molecule within 0.5 Å in the uncomplexed structure. However, all water molecules that only make contacts with the N,NЈ-diacetylchitobiose molecule are disordered in the uncomplexed structure. The carbonyl oxygen makes hydrogen bonds with two water molecules in the cavity, while the methyl group is in close proximity to a hydrophobic area, primarily formed by the Ile-82 side chain. The protein surface is colored according to its electrostatic potential. Potentials greater than 9 T are blue, those less than Ϫ22 T are red, and neutral ones are white. DISCUSSION Site-directed mutagenesis experiments were used to determine that the active site of PNGase F is located in the cleft at the interface between the two domains of the molecule and showed that three acidic residues, Asp-60, Glu-118, and Glu-206, are essential for activity.
The combination of this information with the crystal structure analysis of the complex with N,NЈ-diacetylchitobiose clearly reveals the distinct roles of these residues in the mechanism of action of the enzyme. Mutagenesis of Asp-60 to the corresponding asparagine results in total loss of activity. This residue is in direct hydrogen bonding contact with O 1 of the reducing-end GlcNAc residue. This atom replaces the N ␦ of the asparagine in the substrate. Therefore, this residue must be of primary importance in the catalytic mechanism. The fact that the enzyme selects for the ␣-configuration of O 1 of the N,NЈdiacetylchitobiose while the N ␦ of the substrate is in the ␤-configuration may prove to be important for further studies of the mechanism of action. Glu-206, whose amide mutant has very low residual activity, is not in direct contact with the product and is more likely to play a secondary role, such as stabilization of a reaction intermediate or interaction with the O ␦ of the substrate asparagine.
The geometry of the complex structure in the area of Glu-118 clearly illustrates why this residue is important for substrate binding and why the E118Q mutation impairs substrate binding. O 6 of the second GlcNAc residue accepts a hydrogen atom in its hydrogen bond with N ⑀1 of Trp-120. Therefore, it must donate a hydrogen atom in its hydrogen bond to O ⑀2 of Glu-118. O ⑀1 of the Glu-118 side chain is also involved in a hydrogen bond with another hydrogen donor, Ser-155. The glutamine side chain of the E118Q mutant cannot accept two hydrogen atoms, thus destabilizing this interaction and decreasing the ability to bind the substrate. Co-crystallization experiments of the mutants with the N,NЈ-diacetylchitobiose confirm this interpretation. While the D60N and E206Q mutants also crystallize as complexes with the N,NЈ-diacetylchitobiose bound in the active site cleft, no electron density corresponding to the N,NЈ-diacetylchitobiose molecule could be located in the electron density map of crystals of the E118Q mutant grown under identical conditions. These studies have unambiguously identified the active site residues and the oligosaccharide binding mechanism of PN-Gase F. The exact nature of the mechanism of action of this enzyme is at this time still unclear. However, it is obvious from the lack of serine or threonine residues in the active site area that, contrary to what was initially expected, the mechanism is not related to those of L-asparaginase or glycosylasparaginase. These studies will form the basis for further analysis, which will be focused on the analysis of the interaction of the enzyme with the peptide component of the substrate.