Structural insights into the mechanism and inhibition of eukaryotic O-GlcNAc hydrolysis

Abstract

O-linked N-acetylglucosamine (O-GlcNAc) modification of specific serines/threonines on intracellular proteins in higher eukaryotes has been shown to directly regulate important processes such as the cell cycle, insulin sensitivity and transcription. The structure, molecular mechanisms of catalysis, protein substrate recognition/specificity of the eukaryotic O-GlcNAc transferase and hydrolase are largely unknown. Here we describe the crystal structure, enzymology and in vitro activity on human substrates of Clostridium perfringens NagJ, a close homologue of human O-GlcNAcase (OGA), representing the first family 84 glycoside hydrolase structure. The structure reveals a deep active site pocket highly conserved with the human enzyme, compatible with binding of O-GlcNAcylated peptides. Together with mutagenesis data, the structure supports a variant of the substrate-assisted catalytic mechanism, involving two aspartic acids and an unusually positioned tyrosine. Insights into recognition of substrate come from a complex with the transition state mimic O-(2-acetamido-2-deoxy-D-glucopyranosylidene)amino-N-phenylcarbamate (K_i=5.4 nM). Strikingly, the enzyme is inhibited by the pseudosubstrate peptide Ala-Cys(-S-GlcNAc)-Ala, and has OGA activity against O-GlcNAcylated human proteins, suggesting that the enzyme is a suitable model for further studies into the function of human OGA.

Introduction

Since its discovery over two decades ago (Torres and Hart, 1984), it has become clear that modification of serines/threonines by an O-linked N-acetylglucosamine (O-GlcNAc) is an essential, abundant and dynamic post-translational process (recently reviewed in Zachara and Hart, 2004; Love and Hanover, 2005). O-GlcNAcylated proteins have been detected in both the nucleus and cytoplasm (Torres and Hart, 1984; Holt et al, 1986) and the levels of O-GlcNAcylation respond to nutrient levels and cellular stress (Slawson et al, 2005). O-GlcNAc has now been detected on hundreds of proteins (Zachara and Hart, 2004), many of which play key roles in cellular processes. For instance, precise levels of O-GlcNAcylation of specific sites on insulin receptor substrate 1 (IRS-1) (Park et al, 2005), protein kinase Bβ (Park et al, 2005), glycogen synthase kinase 3β (Parker et al, 2003) and glycogen synthase (Parker et al, 2003) are required for proper insulin sensitivity and response. O-GlcNAcylation of transcription factors such as c-Myc (Chou et al, 1995) and mSin3A (Yang et al, 2002) directly affects their activity, and it is therefore not surprising that proper O-GlcNAc levels are required for a normal cell cycle (Slawson et al, 2005).

Dynamic protein O-GlcNAcylation is achieved by interplay of two essential enzymes—O-GlcNAc transferase (OGT) and O-GlcNAcase (OGA). Both these enzymes are required for life of the metazoan cell, and are highly conserved from Caenorhabditis elegans to man (Zachara and Hart, 2004). Human OGT was recently cloned and characterised in terms of its enzyme activity (Lubas and Hanover, 2000) and domain structure (Lazarus et al, 2005). The enzyme is thought to possess up to 13 N-terminal tetratricopeptide repeats (TPR), which have been shown to be essential for recognition of large protein substrates (Lubas and Hanover, 2000), although the precise mechanisms of recognition and specificity are not known. The C-terminal domain contains the OGT activity and is part of the family 41 glycosyltransferases (CAZY GT 41, http://afmb.cnrs-mrs.fr/cazy), for which no structural information is currently available.

OGA is a 92 kDa protein, originally purified from rat spleen, followed by cloning (Dong and Hart, 1994; Gao et al, 2001) of human OGA (hOGA), also identified as an antigen expressed by meningiomas (MGEA5) (Heckel et al, 1998). Subsequent characterisation showed that this enzyme most likely represents the human hexosaminidase C (HexC) activity, which was identified in addition to the lysozomal HexA/HexB activities (Wells et al, 2002). The enzyme directly hydrolyses O-GlcNAcylated peptides/proteins and the pseudosubstrate p-nitrophenyl-GlcNAc (Wells et al, 2002). Bioinformatic (Schultz and Pils, 2002), genetic (Heckel et al, 1998; Gao et al, 2001) and biochemical (Wells et al, 2002; Toleman et al, 2004) data have shown that the enzyme contains two catalytic activities, the OGA activity and an additional (histone) acetyl transferase (HAT) activity, both of which have been predicted to reside in the C-terminal half of the protein (Schultz and Pils, 2002). Strikingly, the OGA and HAT activities are thought to act synergistically, opening up the chromatin structure and directly activating transcription factors (Toleman et al, 2004).

In several studies, the pharmacological agents streptozotocin (STZ) and O-(2-acetamido-2-deoxy-D-glucopyranosylidene)amino-N-phenylcarbamate (PUGNAc) were used to inhibit OGA activity, thus raising levels of O-GlcNAcylated proteins in the cell (reviewed in Zachara and Hart, 2004). PUGNAc is a synthetic competitive inhibitor of hexosaminidases that inhibits all three known human hexosaminidases (HexA/B and OGA) (Horsch et al, 1991; Haltiwanger et al, 1998). In adipocytes, this compound dramatically reduces insulin sensitivity (Park et al, 2005). STZ is selectively toxic to pancreatic β-cells and has been used for decades to generate mouse models of diabetes (Konrad et al, 2001).

There is currently no detailed information on the structural mechanism of O-GlcNAc hydrolysis, and how OGA recognises its protein substrates and is inhibited by the known inhibitors. Indeed, the precise location of the domain responsible for the OGA activity and the residues involved in catalysis remain unknown. Unfortunately, hOGA, which is a family 84 glycoside hydrolase (GH 84) as defined by the CAZY database, has so far resisted overexpression and purification to quantities and purity required for structural studies.

Here we describe the first crystal structure of a GH 84 (from Clostridium perfringens) with significant sequence homology to the hOGA N-terminus, showing distant structural homology to the family 20 GHs, and also report the structure in complex with PUGNAc. The active site contains a tight pocket for a single GlcNAc residue, in agreement with the N-acetyl-β-D-glucosaminidase activity and the mutagenesis data we report here. Similar to hOGA, PUGNAc competitively inhibits this enzyme in the low nanomolar range and the complex and mutagenesis data are compatible with a substrate-assisted reaction mechanism as proposed recently (Macauley et al, 2005b), involving two conserved aspartic acids. Surprisingly, residues in and surrounding the active site are conserved with human OGA, explaining the inhibition of the enzyme by the pseudosubstrate peptide Ala-Cys(-S-GlcNAc)-Ala, and its activity on mammalian protein substrates.

Results and discussion

Definition of the family 84 GH fold

Despite attempts by several groups, it has so far been impossible to generate significant quantities of the human OGA (hOGA) with crystallography-grade purity. However, the analysis of the CAZY database reveals that hOGA is part of the GH 84, along with several bacterial enzymes. To gain insight into the structure and function of hOGA, we selected one of the bacterial enzymes with the highest sequence homology, that of C. perfringens NagJ (CpNagJ, 34% sequence identity, 51% sequence similarity with hOGA), for structural and biochemical studies (Figure 1C).

Figure 1.

The GH 84 fold. (A) Representative section of the 2.25 Å experimental electron density map (green mesh, contoured at 1σ), with the final model for residues 438–450 shown as sticks. (B) Comparison of the CpNagJ structure with a representative member (SpHeX, PDB entry 1H15 (Mark et al, 2001)) of the family 20 GHs. The structures are shown in a ribbon representation, with red helices and blue strands for the TIM barrel (labelled with α1–8, β1–8), magenta helices and green strands of the N-terminal domain, and transparent yellow helices/strands for the CpNagJ C-terminal domain. The catalytic residues Asp313/Glu314 are shown for the SpHeX structures, the equivalent Asp297/Asp298 are shown for CpNagJ. The thiazoline reaction intermediate mimic, in complex with SpHeX, is shown as a sticks model with orange carbons. (C) Sequence alignment of CpNagJ with the hOGA, made with ALINE (CS Bond, personal communication). The CpNagJ secondary structure is indicated, following the colour scheme and labelling as in panel B. Residues in the active site are indicated by arrows.

Residues 31–624 of CpNagJ (corresponding to the GH 84 domain plus an additional C-terminal 170 residues; Figure 1C) were cloned, overexpressed in Escherichia coli, purified using glutathione affinity chromatography and crystallised from PEG solutions. A preliminary crystallisation study of a similar fragment has been reported recently elsewhere (Ficko-Blean and Boraston, 2005). As no significant sequence homology to known protein structures could be detected, the structure was determined with experimental phases from a zinc single-wavelength anomalous dispersion experiment to 2.25 Å resolution (Table I and Figure 1A), and refined to an R-factor of 0.177 (R_free=0.221) with good stereochemistry (Table I).

Table 1.

Details of data collection and structure refinement

	CpNagJ–Zn	CpNagJ–PUGNAc
Wavelength (Å)	1.28202	1.5418
Unit cell (Å)	a=119.939	a=129.613
	b=147.380	b=145.745
	c=157.687	c=152.800
Resolution range (Å)	20.00–2.25 (2.33–2.25)	20.00–2.35 (2.43–2.35)
No. of observed reflections	1 672 845 (120 883)	221 445 (20 790)
No. of unique reflections	65 956 (6502)	56 462 (5727)
Redundancy	25.4 (18.6)	3.9 (3.6)
I/σI	7.7 (7.6)	15.8 (2.6)
Completeness (%)	100.0 (100.0)	93.7 (96.0)
Rmerge	0.114 (0.477)	0.075 (0.609)
No. of protein residues	1170	1170
No. of water molecules	472	587
R, Rfree	0.177, 0.221	0.177, 0.232
RMSD from ideal geometry
Bonds (Å)	0.017	0.013
Angles (deg)	1.5	1.3
B-factor RMSD (Å2) (backbone bonds)	1.33	1.43
〈B〉 (Å2)
Protein	23.2	49.4
Inhibitor	—	54.0
Water	23.2	49.7
Values between brackets are for the highest resolution shell. All measured data were included in structure refinement. The space group for both crystals was I212121.

The structure consists of three domains (Figure 1B). The N-terminal domain shows an α/β fold, consisting of a seven-stranded mixed β-sheet, with α-helices on both sides. The middle domain is a classic (β/α)₈ barrel, with the seventh helix replaced by a loop. The C-terminal domain reveals a five-helical bundle. The DALI server (Holm and Sander, 1993) was used to detect structural homology with known protein structures. Strikingly, despite the absence of any significant sequence conservation, the N-terminal and middle domains of CpNagJ adopt a fold similar to that seen in the family 20 GHs (RMSD=3.3 Å on 365 equivalenced Cα atoms for the Streptomyces plicatus N-acetyl-β-hexosaminidase (SpHeX)). The structures of CpNagJ and SpHeX are compared in Figure 1B, showing that not only most of the secondary structure elements in the N-terminal and middle domains are conserved, but in addition two CpNagJ carboxylate side chains (Asp297/Asp298) occupy equivalent positions to those of the catalytic residues Asp313/Glu314 of SpHeX (Williams et al, 2002). Notably, DALI also detected structural similarities with families GH 18 and GH 56, which, like SpHeX, are thought to use the unusual ‘substrate-assisted' reaction mechanism (Tews et al, 1997; Markovic-Housley et al, 2000; van Aalten et al, 2001; Williams et al, 2002).

A DALI search for a structural homologue of the C-terminal domain, which does not fall within the GH 84 signature, gave no significant hits. Similarly, sequence searches in the NCBI protein database with CpNagJ residues 455–624 revealed no significant hits. GHs frequently contain carbohydrate-binding modules that assist in binding of polymeric substrates (Boraston et al, 2004), and it is possible that the C-terminal domain could fulfill such a role, although further experiments are required to verify this.

CpNagJ possesses N-acetyl-β-hexosaminidase activity and is inhibited by PUGNAc and STZ

Although CpNagJ falls within the GH 84 family, its kinetics and substrate specificity are not known. Initial steady-state kinetics experiments with p-nitrophenyl-GlcNAc (pNPGlcNAc) (not shown) gave a K_m=121 μM/k_cat=0.2 s⁻¹, similar to the values reported for hOGA (K_m=1.1 mM/k_cat=1 s⁻¹) (Wells et al, 2002)). A more robust assay was obtained with 4-methylumbelliferyl-GlcNAc (4MU-GlcNAc), which gave K_m=2.9 μM/k_cat=10.5 s⁻¹ (Figure 2A and Table II), compared to the reported values for hOGA (K_m=880 μM, (Macauley et al, 2005b)) and typical GH 20 enzymes (e.g. K_m=48 μM/k_cat=222 s⁻¹ for SpHeX (Williams et al, 2002)).

Figure 2.

Enzymology. (A) Steady-state kinetics for wild-type and mutant CpNagJ, with Michaelis Menten parameters shown in Table II. (B). Chemical structure of PUGNAc and Lineweaver–Burk analysis showing competitive inhibition, with K_i=5.4 nM (K_m=2.6 μM, k_cat=8.0 s⁻¹). (C) Chemical structures of streptozotocin and the inhibitory peptide together with dose–response curves, giving IC₅₀'s of 64±10 and 4.9±0.8 μM, respectively.

Table 2.

Steady-state kinetics and PUGNAc inhibition of wild-type and mutant CpNagJ

	Km (μM)	kcat (s−1)	kcat/Km (μM−1 s−1)	PUGNAc IC50 (nM)
WT	2.9±0.2	10.5±0.2	3.6	8.6±0.8
D297Na	3.0±0.7	0.0198±0.0006	0.001	ND
D298Na	4.0±1.4	0.0013±0.0001	0.0003	ND
Y335Fa	41±4	0.0028±0.0001	0.0001	ND
N390A	3.1±0.2	11.0±0.2	3.6	10.0±1.4
N396Aa	13±8	0.0025±0.0005	0.0002	ND
D401Aa	15±9	0.0044±0.0007	0.0003	ND
W490A	28±5	4.0±0.4	0.1	24.0±6.5
aApproximate Km and kcat values for these rather inactive mutants were determined from a kinetics run over a 9-h period with 5–250 μM substrate concentrations.
Steady-state kinetics data (see also Figure 2A) were fitted to the standard Michaelis Menten equation, with the resulting Km and kcat values shown below. IC50 values for inhibition with PUGNAc are also shown. ND=not determined.

The N-acetyl-β-hexosaminidase inhibitor PUGNAc (Figure 2B)) (Horsch et al, 1991) is a potent hOGA inhibitor that has been extensively used to raise levels of O-GlcNAc in cells, by selectively inhibiting hOGA, but not hOGT (Haltiwanger et al, 1998; Zachara and Hart, 2004). The compound shows diabetogenic properties, inducing insulin resistance by modulating O-GlcNAc/O-phosphate levels on various proteins in the insulin signalling pathway (Vosseller et al, 2002; Zachara and Hart, 2004). It has been shown to competitively inhibit hOGA with a K_i of 70 nM (Macauley et al, 2005b). PUGNAc also competitively inhibits CpNagJ with a K_i of 5.4 nM (Figure 2B and Table II), suggesting that the active sites of CpNagJ and hOGA are similar.

STZ (Figure 2C) is a widely used diabetogenic compound, selectively killing β-cells in the pancreas, generating an insulin-deficient mouse model (Konrad et al, 2001). There is some evidence that this compounds acts as a suicide inhibitor of OGA (Konrad et al, 2001), although this has been controversial (Gao et al, 2001; Macauley et al, 2005b). Although STZ does inhibit CpNagJ (IC₅₀=64 μM; Figure 2C), we were unable to detect a time-dependent change in inhibition or a covalent enzyme–inhibitor complex by mass spectrometry (data not shown). This is similar to what has been observed in a recent enzymological study of hOGA (Macauley et al, 2005b).

PUGNAc inhibits CpNagJ/hOGA by mimicking the transition state

Although the synthesis of PUGNAc (Horsch et al, 1991) and its activity against N-acetyl-β-hexosaminidases (e.g. Horsch et al, 1991; Haltiwanger et al, 1998; Macauley et al, 2005b) has been known for more than a decade, its structural mode of inhibition has not been defined. To gain insight into the mechanism of PUGNAc inhibition and to study the details of the CpNagJ/hOGA active site, a CpNagJ–PUGNAc complex was determined by soaking the inhibitor into native crystals, followed by refinement against 2.35 Å X-ray diffraction data (R, R_free=0.177, 0.232; Table I). Well-defined ∣F_o∣−∣F_c∣, φ_calc electron density allowed building and refinement of the complete PUGNAc molecule. The inhibitor binds with the GlcNAc sugar deep in a pocket on the enzyme surface, whereas the phenylcarbamate moiety projects towards the solvent. The oxime is in the Z configuration, consistent with a recent study showing that inhibition of hOGA with PUGNAc is dependent on the Z oxime stereochemistry (Perreira et al, 2006). The E oxime stereoisomer would not fit the shape of the CpNagJ active site. The sp² hybridisation of the C1 carbon helps the pyranose ring assume a ⁴E envelope conformation. Strikingly, the oxygen of the 2-acetamido group approaches the C1 carbon to within 3.0 Å. This has also been observed in the GH 18, 20 and 56 families, which have been shown to employ substrate-assisted catalysis, where the oxygen of the 2-acetamido group, rather than a protein side chain, is the catalytic nucleophile, leading to formation of an oxazolinium intermediate (Terwisscha van Scheltinga et al, 1994; Tews et al, 1997; Markovic-Housley et al, 2000; Mark et al, 2001; van Aalten et al, 2001). Indeed, an elegant mechanistic enzymology study of hOGA has recently shown that it uses the same substrate-assisted double-displacement mechanism with overall retention of stereochemistry (Macauley et al, 2005b), compatible with the structural data described here. The GlcNAc moiety of PUGNAc appears to be tightly tethered in the active site, through 10 hydrogen bonds to nine residues, which are all conserved in the hOGA sequence (Figures 1C and 3). This tight hydrogen-bonding network explains why, unlike the human GH 20 hexosaminidases, hOGA is not inhibited by GalNAc (Gao et al, 2001). Asp298 and Tyr335 are within hydrogen bonding distance of the oxime nitrogen (approximately equivalent to the position of the substrate glycosidic oxygen), while Asp297 and Asn396 influence the conformation of the acetamido group. The 3, 4 and 6 sugar hydroxyls point deep into the pocket, where they are hydrogen bonded to Lys218, Gly187, Asn429 and, in particular, Asp401. Notably, there are only two residues that contact the inhibitor and that are not conserved between CpNagJ and hOGA. Trp490 projects from the CpNagJ C-terminal domain towards the catalytic domain, and stacks with the PUGNAc phenyl ring, similar to what is observed for the interaction of the PUGNAc-related compound, HM508, with a GH 18 chitinase (Vaaje-Kolstad et al, 2004) (Figure 3). The other nonconserved residue is Val331, at the very bottom of the active site (Figure 3). Interestingly, this residue is a cysteine in hOGA, compatible with the early observation that the enzyme can be inactivated by a thiol-reactive compound (Dong and Hart, 1994). Notably, a cysteine close to the active site of the protein phosphatase PTP1B has recently been shown to play a key role in redox-dependent regulation of phosphatase activity (Salmeen et al, 2003). Further experiments are currently in progress to study a potential similar mechanism in hOGA.

Figure 3.

Details of the active site. The CpNagJ–PUGNAc complex is compared to the SmChiB–HM508 complex (PDB entry 1UR9 (Vaaje-Kolstad et al, 2004)), the SpHeX–thiazoline complex (PDB entry 1H15 (Mark et al, 2001)) and the AmHya–substrate complex (PDB entry 1FCV (Markovic-Housley et al, 2000)). The structures were superimposed using the C2, C3, C5 and O5 atoms of the ligands in the −1 subsite. The ligands are shown as sticks with green carbons. For PUGNAc, the unbiased 2.35 Å ∣F_o∣−∣F_c∣, φ_calc electron density map is shown in cyan, contoured at 2.5σ. The CpNagJ–PUGNAc complex is shown in stereo, and residues contacting the inhibitor are labelled (see also Figure 1C). For the other complexes the two key carboxylate side chains are identified. The CAZY GH family numbers are also shown.

The CpNagJ/hOGA active sites are similar to GH 18, 20 and 56

Comparison of CpNagJ with GH 18, 20 and 56 structures reveals a number of significant differences and similarities in the active sites, that allow the putative identification of residues involved in catalysis. CpNagJ Asp298 occupies a position equivalent to a conserved glutamic acid in the GH 18, 20 and 56, which has been shown to be the catalytic acid, protonating the glycosidic bond in the first step of the reaction (Tews et al, 1997; Markovic-Housley et al, 2000; Mark et al, 2001; van Aalten et al, 2001). Unusually, Tyr335, also observed in GH 56, but not in GH 18/20, is also within hydrogen-bonding distance of the glycosidic oxygen (Figure 3). CpNagJ Asp297 appears to be structurally conserved in the GH 18, 20 and 56 enzymes, where it is thought to stabilise the conformation of the acetamido group and the developing positive charge on the oxazolinium intermediate (Tews et al, 1997; Markovic-Housley et al, 2000; van Aalten et al, 2001; Williams et al, 2002). CpNagJ Asn396 hydrogen bonds the acetamido oxygen, a role fulfilled by a tyrosine in GH 18, 20 and 56 enzymes (Figure 3).

To understand the contributions of active site residues to substrate/inhibitor binding and catalysis, seven mutants were investigated (Figure 2A and Table II). Asn390, a surface-exposed residue away from the active site, was mutated to alanine as a control, and showed kinetic parameters indistinguishable from the wild-type enzyme. The remaining mutants approximately fall into two distinct groups, mainly affecting either binding of the substrate (effects on K_m) or catalysis (effects on k_cat). Residues Asp401 and Trp490 make interactions with the PUGNAc inhibitor on the verges of the active site (Figure 3) and, in agreement with this, mutations of these residues (D401A, W490A) have significant effects on the ability of the enzyme to bind the substrate/inhibitor (5–10-fold increase in K_m) (Figure 2A and Table II).

Mutations of several residues showed significant effects on catalysis. Asp297 and Asn396 are involved in stabilising the conformation of the acetamido group and the oxazolinium intermediate, compatible with their 530- and 4200-fold reduction in k_cat, respectively. Similar selective effects on k_cat have also been demonstrated for mutation of the equivalent residues (an Asp and a Tyr, respectively, see Figure 3) in the GH 18 (e.g. van Aalten et al, 2001; Bokma et al, 2002) and GH 20 (e.g. Williams et al, 2002) families. This has been taken as evidence for the involvement of these side chains in a substrate-assisted catalysis mechanism.

In addition to the increase in K_m for the D401A mutation noted above, there is also a significant drop in k_cat (2400-fold; Figure 2A and Table II). Asp401 interacts with the O4/O6 hydroxyls, well away from the glycosidic bond (Figure 3). It is possible that the tight hydrogen bonds to these hydroxyls aid in formation of the ⁴E envelope conformation of the pyranose ring in the transition state, providing a possible explanation for the large mutational effect on k_cat.

Asp298, proposed to be the catalytic acid, and Tyr335, of hitherto unknown function, are both in a position to interact with the glycosidic oxygen (Figure 3). Compatible with its proposed role as protonator of the glycosidic bond, mutation of Asp298 to Asn reduces k_cat 8100-fold (Figures 2 and 3 and Table II). Strikingly, mutation of Tyr335 to Phe dramatically also reduces activity (3800-fold). GH 18 and GH 20 enzymes do not possess an equivalent of this residue, whereas it is observed in the only GH 56 structure available (Markovic-Housley et al, 2000) (Figure 3). Apart from the glycosidic oxygen, there are no hydrogen-bonding donors or acceptors nearby the tyrosine hydroxyl (it is 4.8 Å away from the pyranose ring oxygen), suggesting that it is protonated in the PUGNAc/Michaelis complex. Interestingly, a very recent study has suggested that hydrolysis of O-glycosides by hOGA proceeds via a late transition state in which cleavage of the glycosidic linkage is well advanced (Macauley et al, 2005a). Tyr335 would then be well positioned to stabilise any partial negative charge that develops at the glycosidic oxygen in such a transition state, consistent with the requirement of this residue for catalysis.

CpNagJ possesses OGA activity against human proteins and is inhibited by an S-GlcNAc peptide

As discussed above, the CpNagJ active site is highly conserved with hOGA, and the enzymes show similar kinetic parameters for pseudosubstrates. Strikingly, the residues surrounding the active site are also highly conserved between CpNagJ and hOGA (Figure 4A). Whereas the GlcNAc binds deep in a conserved pocket, there also appears to be sequence conservation in the bottom and the walls of a groove running over the surface of the protein, perhaps representing the peptide-binding site in case of hOGA. This leads to the hypothesis that, although the natural substrate of CpNagJ is probably a GlcNAc-containing carbohydrate polymer (e.g. hyaluronan/chitin/peptidoglycan), the enzyme may also be suitable as a model system to study hydrolysis of O-GlcNAcylated peptides/proteins in vitro. Indeed, the enzyme is able to hydrolyse O-GlcNAcylated proteins in Swiss 3T3 cell lysate (Figure 4B), although some proteins appear not to be a substrate. It is possible that on these proteins the O-GlcNAc sites are masked, or that the unconserved Trp490 residue in CpNagJ (Figure 4A) prevents binding of some substrate proteins. The inactive D401A mutant of CpNagJ does not show activity against O-GlcNAcylated proteins. In agreement with the competitive inhibition of PUGNAc, this compound inhibits the OGA activity of CpNagJ in the cell lysates completely at 10 μM (Figure 4B). As a final verification of the ability to bind O-GlcNAcylated peptides, we synthesised the thioglycosidic peptide Ala-Cys-(S-GlcNAc)-Ala. Thioglycosides have previously been successfully used to trap glycosidase–substrate complexes (e.g. Varrot et al, 2003), and S-linked glycopeptides have attracted attention as glycopeptide mimetics due to their enhanced chemical stability and resistance to glycosidases (e.g. Zhu et al, 2004). Strikingly, CpNagJ is potently inhibited by this peptide (IC₅₀=4.9 μM; Figure 2C), providing further evidence for the ability of CpNagJ, like hOGA, to bind to and hydrolyse O-GlcNAcylated peptide substrates. Interestingly, data from a recent study suggest that inhibition with this peptide may be transient, as it was shown that hOGA is able to hydrolyse model thioglycoside substrates with aromatic leaving groups (Macauley et al, 2005a). However, such (good) leaving groups are significantly different from those in the true peptide substrates and further work will be required to study this.

Figure 4.

Surface sequence conservation and OGA activity. (A) The protein surface of the CpNagJ–PUGNAc complex is shown in three different colours to reflect sequence conservation with the hOGA: grey (not conserved), red (identical) and dark red (similar), following sequence conservation in the alignment shown in Figure 1C. (B) Western blot analysis of O-GlcNAc levels in Swiss 3T3 cell lysates, after incubation with different amounts of the wild-type enzyme (0.5, 3 and 5 μM), the inactive D401A mutant (5 μM) and the PUGNAc inhibitor (10, 50 and 100 μM). O-GlcNAc was detected with a commercially available antibody.

Concluding remarks

The data presented here give the first detailed structural and mechanistic insights into a bacterial GH 84 with significant sequence homology to the hOGA. The data show that the N-acetyl-β-hexosaminidase activity must be, by sequence homology, located to the N-terminal domain of hOGA, in contrast to what has been proposed previously by a bioinformatics study (Schultz and Pils, 2002). The active sites of the human and bacterial enzyme are, with the exception of two residues, highly conserved, and the sequence conservation extends to the groove around the active site, where peptide binding and recognition take place. The GlcNAc portion of the substrate is tightly bound in a deep pocket, by hydrogen bonds and stacking with conserved aromatic residues. This is somewhat similar to the protein phosphatases, where the O-linked phosphate binds deep in a conserved pocket, while the peptide binds in a groove on the surface (e.g. Jia et al, 1995; Salmeen et al, 2000). Interestingly, hOGA possesses a cysteine in the active site, but away from the catalytic center—similar to the tyrosine phosphatases, where this residue has been shown to be involved in redox-dependent regulation of the enzyme (Salmeen et al, 2003).

The structure and enzymology suggest an unusual variant of the substrate-assisted catalysis mechanism previously identified for the GH 18, 20 and 56 enzymes (Tews et al, 1997; Markovic-Housley et al, 2000; van Aalten et al, 2001; Williams et al, 2002). In these families the catalytic acid is a glutamic acid residue on the protein, whereas in the GH 84 enzymes it is an aspartic acid (Asp298). Furthermore, a tyrosine appears to be involved in stabilisation of the transition state. An equivalent tyrosine has been observed in the GH 56 bee venom hyaluronidase (Markovic-Housley et al, 2000), but its function was hitherto unknown. After more than a decade of use as a modulator of cellular O-GlcNAc levels, the PUGNAc complex reveals that this compound inhibits OGA by mimicry of the transition state through the geometry at the C1 carbon. Tight stacking interactions are observed between the PUGNAc acetamido group and a conserved trytophan/tyrosine pair. Interestingly, a recent mechanistic enzymology study has shown that hOGA possesses a cavity near the acetamido methyl group, whereas such cavity is absent in the GH 20 hexosaminidases (Macauley et al, 2005b). This is supported by the structure described here, where Val331 forms the bottom of a pocket, whereas a tryptophan tightly packs against the acetamido methyl in the GH 20 enzymes.

In agreement with the sequence homology between the hOGA and the bacterial enzyme, CpNagJ shows OGA activity on several proteins in human cell lysate, and the enzyme is also inhibited by a thioglycosidic inhibitory peptide. This suggests that CpNagJ may be a suitable model for further studies into the mechanisms of recognition/specificity of O-GlcNAcylated peptides. Furthermore, the structural data presented here could aid in rational design of more potent/drug-like OGA inhibitors, which could be used to probe or modulate the role of O-GlcNAc in diseases such as Alzheimer's, diabetes and cancer.

Materials and methods

Cloning, expression and crystallisation

A fragment corresponding to the N-terminal GH 84 catalytic domain of CpNagJ (GenBank accession no. BAB80940) was PCR amplified (forward primer, 5′-GGATCCGTAGGACCTAAAACTGGG-3′; reverse primer, 5′-CTCGAGTTATCATATTAATGTTAAATCAAAAC TTAAAGC-3′) from genomic DNA from C. perfringens (Sigma D5139). The PCR product was ligated into pCR 2.1—TOPO (Invitrogen) and subcloned into the pGEX-6P-1 vector (Amersham Biosciences) using the BamHI and the XhoI restriction sites. Single amino-acid residue changes were made using the Quick Change Site Directed Mutagenesis Kit (Stratagene), following the manufacturer's protocol. All plasmids were verified by DNA sequencing (The Sequencing Service, School of Life Sciences, University of Dundee, Scotland, UK). Upon sequencing of several independent PCR reactions, it became apparent that there were several nucleotide changes on the template DNA, compared to the GenBank entry. These thus represent strain-specific changes, which result in six conservative amino-acid substitutions that are located in surface loops of the protein.

The CpNagJ–pGEX-6P-1 construct was transformed into E. coli BL21 (DE3) pLysS cells. Cells were grown overnight in Luria-Bertani medium (LB)+ampicillin (100 μg/ml). From this culture, 10 ml of cells were used to inoculate 1 l of LB media. The cells were grown to OD₆₀₀=0.5 before expression was induced by the addition of 250 μM of isopropyl-β-D-thiogalactopyranoside, and then the cells were cultured overnight at room temperature. The cells were harvested by centrifugation at 2500 g for 30 min, flash frozen in liquid nitrogen, thawed at 37°C, and resuspended in 25 ml of buffer A (50 mM HEPES, 250 mM NaCl, pH 7.5) with half a protease inhibitor tablet (Roche). Cells were lysed by the addition of 10 mg of DNAse-1 and the use of a cell disrupter at 30 K psi. The lysate was centrifuged at 18 900 g for 30 min at 4°C and passed through a 0.45 μm filter. The filtrate was then incubated at 4°C for 2.5 h on a rotating platform with 2 ml of glutathione-Sepharose beads (Amersham Biosciences), prewashed with buffer A. The N-terminal GST tag was removed from the GST–CpNagJ fusion protein by incubating the beads with PreScission Protease (80 μg) at 4°C overnight. The supernatant of the beads and a subsequent wash were passed over a Bio-Rad 20 ml disposable column to remove the beads. The resulting filtrate was concentrated to 4 ml, and loaded onto a Superdex 75, 26/60 gel filtration column pre-equilibrated in buffer A. The pure fractions were verified by SDS–PAGE, pooled and then dialysed overnight in 50 mM Tris (pH 8.0) at 4°C. The CpNagJ mutants were expressed and purified using an identical protocol, exluding the final gel filtration step.

Pure CpNagJ protein was spin concentrated to approximately 28 mg/ml. Vapour diffusion crystallisation experiments were set up by mixing 1 μl of protein, 1 μl of mother liquor (0.2 M ammonium sulfate, 0.1 M sodium cacodylate, pH 6.5, and 30% PEG 8000) and 0.25 μl of 40% v/v γ-butyrolactone. Rod-shaped crystals appeared after 4 days, growing to a maximum size of approximately 0.3 × 0.1 × 0.1 mm.

Data collection, structure solution and refinement

The crystals were cryoprotected by a 5 s immersion in a solution containing 0.17 M ammonium sulfate, 0.085 M sodium cacodylate, pH 6.5, 25.5% PEG 8000 and 15% v/v glycerol, and then frozen in a nitrogen cryostream. Crystals used for phasing were soaked in 1.5 M ZnSO₄ for approximately 5 s and cryoprotected as above. An inhibitor complex of PUGNAc (Toronto Research Chemicals Inc., Canada) was obtained by adding 0.25 μl of 10 mM PUGNAc to the crystals for 10 min, followed by cryoprotection.

A 25-fold redundant 2.25 Å single-wavelength anomalous dispersion data set was collected from the ZnSO₄-soaked crystals at beamline BM14 at the European Synchrotron Radiation Facility (ESRF). Initial phases were calculated with SOLVE (Terwilliger and Berendzen, 1999) to 3.25 Å (figure of merit=0.35), from the 11 Zn sites located. With the help of PROFESS (Collaborative Computational Project, 1994), two-fold noncrystallographic symmetry was detected, compatible with eight of the 11 sites. Solvent flattening, phase extension to 2.25 Å and twofold averaging was then performed with DM (Cowtan, 1994), yielding a good-quality electron density map. This was used as input for warpNtrace (Perrakis et al, 1999), which was able to build 976 out of 1186 possible residues, covering two molecules in the asymmetric unit. Iterative model building (with O (Jones et al, 1991) and COOT (Emsley and Cowtan, 2004)) and refinement (with CNS (Brunger et al, 1998) and REFMAC (Murshudov et al, 1997)) then yielded the final model with statistics shown in Table I.

Diffraction data of PUGNAc-soaked crystals were collected on a rotating anode to 2.35 Å (Table I). Refinement was initiated from the native structure, immediately revealing well-defined ∣F_o∣−∣F_c∣, φ_calc electron density for the inhibitor, which was built in with the help of PRODRG (Schuettelkopf and van Aalten, 2004)-generated inhibitor structure and topology. Further refinement and model building then yielded the final model with statistics shown in Table I.

Enzymology

Steady-state kinetics of wild-type and mutant CpNagJ were determined using the fluorogenic substrate 4-methylumbelliferyl-N-acetyl-β-D-glucosaminide (4MU-NAG; Sigma). Standard reaction mixtures (50 μl) contained 0.2 nM enzyme in 50 mM citric acid 125 mM NaH₂PO₄, pH 5.5, 0.1 mg/ml BSA and 1.5–25 μM of substrate in water. The reaction mixture was incubated for 7 min at 20°C (RT). The reaction was stopped by the addition of a twofold excess (100 μl) of 3 M glycine-NaOH, pH 10.3. The fluorescence of the released 4-methylumbelliferone was quantified using an FLX 800 Microplate Fluorescence Reader (Bio-Tek), with excitation and emission wavelengths of 360 and 460 nm, respectively. The production of 4-methylumbelliferone was linear with time for the incubation period used, and less than 10% of the available substrate was hydrolysed. Experiments were performed in triplicate; all of the spectra were corrected for the background emission from the buffer and the protein. Michaelis–Menten parameters were obtained by fitting the fluorescence intensity with GraFit (Leatherbarrow, 2001).

The IC₅₀'s of Ala-Cys(S-GlcNAc)Ala, STZ and PUGNAc against wild type and mutant CpNagJ were determined using the same protocol, but with the following changes: the enzyme was incubated for 1 min with the corresponding inhibitor (1 nM–1 mM) prior to starting the reaction by the addition of a constant substrate concentration equivalent to the K_m determined by steady-state kinetics (e.g. 2.9 μM for wild type). All parameters were obtained by fitting the fluorescence intensity data to the standard IC₅₀ equation in the software GraFit (Leatherbarrow, 2001).

Determination of the PUGNAc K_i was performed by steady-state kinetics in the presence of different concentrations (0, 15, 25 and 45 nM) of the inhibitor. The mode of inhibition was visually verified by a Lineweaver–Burk plot, and the K_i determined by fitting all fluorescence intensity data to the standard equation for competitive inhibition in GraFit (Leatherbarrow, 2001).

Western blotting

Swiss 3T3 cells were cultured using standard techniques and lysed in a buffer containing 50 mM Tris–HCl, pH 7.5, 1 mM EDTA, 1 mM EGTA, 270 mM sucrose, 0.1% β-mercaptoethanol, 1 mM Na₃VO₄, 1% (v/v) Triton X-100, 50 mM NaF, 5 mM sodium pyrophosphate and proteinase inhibitor cocktail (one tablet/25 ml). For the OGA assay, 100 μg of cell lysate was incubated with various amounts of enzyme and inhibitor for 30 min at 20°C in a final volume of 100 μl. For Western-blot purposes, 50 μl was separated on a 10% SDS–PAGE gel and transferred onto a nitrocellulose membrane. The membrane was blocked in 50 mM Tris–HCl, pH 7.5, 0.15 M NaCl, 0.1% (v/v) Tween (TBS-Tween) and 10% (w/v) BSA for 1 h. The membrane was then incubated with the same buffer for 16 h at 4°C in the presence of 1 μg/ml of O-GlcNAc primary antibody (Pierce). Detection was performed using horseradish peroxidase-conjugated secondary antibody and the enhanced chemiluminescence (ECL Amersham Pharmacia Biotech) reagent.

Synthesis of Ala-Cys-(S-GlcNAc)-Ala

N-(tert-butoxycarbonyl)-O-β-(2-acetamido-2-deoxy-3,4,6-tri-O-acetyl-β-D-glucopyranosyl)-L-cysteine benzyl ester (Boc-L-Cys(S-GlcNAc(Ac)₃-β-D)-OBn) was synthesised from N-(tert-butoxycarbonyl)-L-iodoalanine (Aldrich) using the method of Ohnishi et al (2000). The sugar amino acid was then elaborated to the fully protected glycopeptide derivative, Z-L-Ala-L-Cys(S-GlcNAc(Ac)₃-β-D)-L-Ala-OBn, by standard peptide synthesis methodology, followed by removal of the N-benzyloxycarbonyl (Z) and benzyl ester (Bn) protection by catalytic transfer hydrogenolysis (Anwer and Spatola, 1980), and O-deacetylation by treatment with sodium methoxide in anhydrous methanol (Paulsen and Holck, 1982). The final glycopeptide was purified by reversed-phase HPLC on a Dionex HPLC system equipped with a Phenomenex Gemini 5 μM C-18 (250 × 10 mm) column (mobile phase A=0.1% trifluoroacetic acid in H₂O; mobile phase B=0.1% trifluoroacetic acid in acetonitrile; linear gradient 5–95% B in 10 min; flow rate 2.5 ml/min), and was characterised by ¹H NMR, low-resolution and high-resolution electrospray mass spectrometry.