Archives of Biochemistry and Biophysics 505 (2011) 112

Contents lists available at ScienceDirect

Archives of Biochemistry and Biophysics

journal homepage: www.elsevier.com/locate/yabbi

Review

Dynamics of glucosamine-6-phosphate synthase catalysis

Stphane Mouilleron a,b,1, Marie-Ange Badet-Denisot b, Bernard Badet b, Batrice Golinelli-Pimpaneau a,
a b

Laboratoire dEnzymologie et Biochimie Structurales, CNRS, 1 avenue de la Terrasse, 91198 Gif-sur-Yvette, France Institut de Chimie des Substances Naturelles, CNRS, 1 avenue de la Terrasse, 91198 Gif-sur-Yvette, France

a r t i c l e

i n f o

a b s t r a c t
Glucosamine-6P synthase, which catalyzes glucosamine-6P synthesis from fructose-6P and glutamine, channels ammonia over 18 between its glutaminase and synthase active sites. The crystal structures of the full-length Escherichia coli enzyme have been determined alone, in complex with the rst bound substrate, fructose-6P, in the presence of fructose-6P and a glutamine analog, and in the presence of the glucosamine-6P product. These structures represent snapshots of reaction intermediates, and their comparison sheds light on the dynamics of catalysis. Upon fructose-6P binding, the C-terminal loop and the glutaminase domains get ordered, leading to the closure of the synthase site, the opening of the sugar ring and the formation of a closed ammonia channel. Then, glutamine binding leads to the closure of the Q-loop to protect the glutaminase site, the activation of the catalytic residues involved in glutamine hydrolysis, the swing of the side chain of Trp74, which allows the communication between the two active sites through an open channel, and the rotation of the glutaminase domains relative to the synthase domains dimer. Therefore, binding of the substrates at the appropriate reaction time is responsible for the formation and opening of the ammonia channel and for the activation of the enzyme glutaminase function. 2010 Elsevier Inc. All rights reserved.

Article history: Available online 13 August 2010 Keywords: Glucosamine-6-phosphate synthase Glutamine amidotransferase Conformational changes Ammonia channel X-ray structure

Introduction Glucosamine-6P synthase (GlmS)2 catalyzes the rst and ratelimiting step of hexosamine biosynthesis [1]. The latter product of this pathway, UDP-N-acetylglucosamine, is a precursor of several structural macromolecules such as the peptidoglycan and lipopolysaccharides in bacteria, chitin in insects and fungi, or glycoproteins in mammals. Since the deletion of the GlmS gene is lethal in fungi and bacteria in the absence of glucosamine-6P or UDP-N-acetylglucosamine exogenous import [2,3], enzyme inhibition could lead to the development of specic antibiotic or antifungal drugs [4]. Targeting the human enzyme could also limit complications associated with diabetes. Indeed, hyperglycemia associated with this pathology leads to an increased ux of intracellular glucose in excess into the hexosamine pathway. Increased formation of O-glycosylated proteins results in changes in both gene expression and protein function, which contribute to the development of insulin resistance [5] and vascular complications in type-2 diabetes [6,7].
Corresponding author.
E-mail addresses: stephane.mouilleron@cancer.org.uk (S. Mouilleron), marieange.badet@icsn.cnrs-gif.fr (M.-A. Badet-Denisot), bernard.badet@icsn.cnrs-gif.fr (B. Badet), beatrice.golinelli@lebs.cnrs-gif.fr (B. Golinelli-Pimpaneau). 1 Present address: Structural Biology Laboratory, Cancer Research UK, London Research Institute, 44 Lincolns Inn Fields, London WC2A 3PX, UK. 2 Abbreviations used: GlmS, glucosamine-6P synthase; Fru6P, D-fructose-6P; Gln, Lglutamine; Glu, L-glutamate; GlcN6P, D-glucosamine-6P; Glc6P, D-glucose-6P; DON, 6-diazo-5-oxo-L-norleucine. 0003-9861/$ - see front matter 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.abb.2010.08.008

GlmS is a member of the glutamine-dependent amidotransferase family that catalyzes the conversion of D-fructose-6P (Fru6P) to L-glutamine (Gln) using the amide nitrogen of glutamine as the unique source of ammonia, without the need of any cofactor [8]. The enzyme from Escherichia coli was shown to obey a bibi ordered mechanism, in which Fru6P binds rst, followed by Gln, whereas L-glutamate (Glu) is released before D-glucosamine-6P (GlcN6P) (Scheme 1) [9]. The reaction catalyzed by GlmS can be decomposed into two half-reactions that take place in two separate domains [10]. At the 27 kDa N-terminal domain (glutaminase domain, residues 1240), Gln hydrolysis leads to Glu and ammonia (Scheme 2A). At the 40 kDa C-terminal domain (synthase domain, residues 241608), conversion of Fru6P to GlcN6P involves sugar isomerization followed by its amination using ammonia produced at the glutaminase site (Scheme 3) [11]. The purpose of this review is to describe the dynamics of the GlmS-catalyzed reaction by comparing the various crystal structures of the full-length E. coli enzyme, which were obtained recently. In fact, obtaining well-ordered crystals of the full-length enzyme remains tricky because of the mobility of the domains and the existence of exible loops, which participate to the dynamics of catalysis. These difculties explain why the crystal structure of GlmS has been rst determined for the separately expressed domains. Structures of the glutaminase domain have been solved in the presence of the Glu product and L-glutamyl hydroxamate, a competitive inhibitor of the glutaminase site [12], and structures of the synthase domain in the presence of the GlcN6P product,

S. Mouilleron et al. / Archives of Biochemistry and Biophysics 505 (2011) 112

Scheme 1. Ordered bibi kinetic mechanism of GlmS. The crystal structures, which mimic reaction intermediates, are indicated as boxes.

H
-OOC

H NH3+
-OOC

NH3+

H NH3
-OOC

NH3+

H
-OOC

NH3+

H
-OOC

NH3+ L-Glutamate

Gly99
L-Glutamine H2N H HO H S H H2N O H O H3N HO H2N S O O HO S O

Gly99
HO

Asn98 Cys1

H O H HO

Asn98
H S

O H

Thr606

Cys1

H2N

Cys1

H3N HO

Cys1
H2N HO

Cys1

Thr606

Thr606
NH3+

Thr606

Thr606

H
-OOC

NH3+

H
-OOC

NH3+
-OOC

DON O S H2N HO H N+ N-S

+ N2 O N+ N S O

Cys1

H2N HO

Cys1
HO

H2N

Cys1

Thr606

Thr606

Thr606

Scheme 2. Reaction mechanism at the glutaminase site. (A) Glutamine hydrolysis. (B) Inactivation of GlmS by the glutamine analog DON.
D-glucose-6P (Glc6P, the product of the reaction in the absence of glutamine), and 2-amino-deoxyglucitol-6P, an analog of a reaction intermediate, which is a competitive inhibitor of the synthase site [13,14]. The structures of the individual domains, combined with biochemical experiments, have allowed to identify the catalytic residues involved in Gln hydrolysis and Fru6P isomerization and to propose a mechanism at each active site (Schemes 2A and 3). GlmS is a member of the N-terminal nucleophile hydrolase family, which includes one class of glutamine amidotransferases as well as penicillin acylase, aspartyl glucosaminidase and the 20S proteasome [15]. The main catalytic residues responsible for glutamine hydrolysis in GlmS are the amino-terminal cysteine, whose free a-amino group is proposed to act as a base and thiol group as a nucleophile (see discussion below), and Asn98/Gly99, which form the oxyanion hole that stabilizes the tetrahedral intermediates during hydrolysis (Scheme 2A). Lys603, which was shown to form a Schiff base with

the carbonyl group of Fru6P [16], Glu488 the likely base that deprotonates C1 of Fru6P, Lys485 proposed as the base that deprotonates the O5 hydroxyl group and His504* potentially involved in the opening/closure of the sugar ring are the key catalytic elements for the synthesis of the GlcN6P product using ammonia formed at the glutaminase site [13] (Scheme 3). Catalytic His504* belongs to the so-called His-loop (composed of the tripeptide 503*505*) from the second synthase monomer that forms the synthase dimer, which indicates that the enzyme is active only in the dimeric form. The activity decrease of the proteins, in which the corresponding residues have been mutated, is in agreement with their critical function in catalysis (Table 1). Five crystal structures of full-length E. coli GlmS are known (Scheme 1, Fig. 1 and Table 2). We recently determined the crystal structure of the unliganded form (GlmS crystal) [17]. We improved the resolution of the crystals obtained in the presence of the rst substrate Fru6P (GlmSFru6P crystal), in which the sugar is present

S. Mouilleron et al. / Archives of Biochemistry and Biophysics 505 (2011) 112

O3PO

H OH O O H

O3PO OH

HO HO

His504*

HO HO H2N

HR

= Hs OH

O3PO OH

HO HO

HR

"NH3"

Glu488
O3PO OH HR

OH D-fructose-6P

Lys603

NH Lys603

Hs OH

HO HO

Hs OH NH2

Lys485

OPO3= HO O HO

O3PO OH H NH3

His504*

O3PO OH

Glu488
Hs O H

HO HO

HO HO

NH3 OH D-glucosamine-6P

NH3

Lys485

Scheme 3. Reaction mechanism at the synthase site.

Table 1 Kinetic parameters for the glutaminase (Gln hydrolysis in the absence of Fru6P), hemi-synthase (Gln hydrolysis in the presence of Fru6P), and synthase activities (GlcN6P synthesis) of the wild-type and mutant GlmS proteins. Glutaminase activity (in the absence of Fru6P) K Gln (mM) m Wild-typea G0b H504Q W74Aa C1Ac K603Rc 0.05 1.6 0.009 0.2 nd 0.17 kcat (min1) 10 1.4 0.012 41 0 10 kcat/K Gln (M1 s1) m 3300 14 22 3400 0 980 Hemi-synthase activity (glutaminase activity in the presence of Fru6P) K Gln (mM) m 0.27 0.9 0.02 0.38 nd 0.03 kcat (min1) 1030 36 3 181 <1 23 kcat/K Gln (min1) m 63,600 670 2500 7900 nd 12778 Synthase activity (formation of GlcN6P) K Fru6P (mM) m 0.36 nd 3.6 0.93 nd 0.6 kcat (min1) 865 nd 1 7 <1 7 kcat/KmFru6P (min1) 40,000 nd 4.6 125 nd 175

nd, not determined. a Ref. [23]. b Addition of Gly at the N-terminus [12]. c Ref. [45].

Fig. 1. (A) Crystals of full-length E. coli GlmS. Using 78% PEG4000 as precipitant, both GlmS and GlmSGlcN6P crystallize in the same space group, with similar cell parameters, and present the same aspect (size up to 0.15 mm 0.15 mm 0.15 mm). (B) Crystals of the GlmSFru6P complex (0.1 mm 0.1 mm 0.8 mm) were obtained using 0.8 M LiCl as precipitant. (C) Crystals of the GlmSGlc6PDON complex (0.05 mm 0.1 mm 0.4 mm). The DON covalently modied protein was rst puried and then crystallized in the presence of Fru6P using 12% PEG8000 as precipitant. (D) Crystals of the C1A-GlmSGlc6PGlu complex (0.5 mm 0.5 mm 0.5 mm) were obtained using 22% PEG3350 as precipitant.

in the linear form [18,19]. We crystallized GlmS, covalently modied by the Gln analog 6-diazo-5-oxo-L-norleucine (DON) (Scheme 2B), in the presence of Fru6P [19,20]. In these crystals, the majority of Fru6P was isomerized into linear Glc6P (GlmSGlc6PDON crystal), which is consistent with the isomerase activity exhibited by the isolated synthase domain [11]. Finally, the structure of GlmS in complex with cyclic GlcN6P (GlmSGlcN6P crystal) was solved [17]. These structures represent snapshots of several reaction intermediates along the catalytic cycle (Scheme 1)

and their comparison with each other therefore sheds light on the dynamics of catalysis. Unfortunately, there is no structure available to date corresponding to the covalent iminium intermediate formed by the nucleophilic attack of Lys603 on the Fru6P carbonyl group (Scheme 3), which could be stabilized upon reduction [16]. Interestingly, such an unstable iminium intermediate could be trapped in crystals of fructose-1,6-bisphosphate aldolase using a mutant with highly decreased activity [21]. Such crystallization strategy would be worth trying in the case of GlmS in the future.

4 Table 2 Crystal structures of full-length E. coli GlmS.

S. Mouilleron et al. / Archives of Biochemistry and Biophysics 505 (2011) 112

GlmS PDB code Space group Unit cell a, b, c () a, b, c Resolution () No. of monomers per asymmetric unit pH of crystallization 2VF4 H32 144.7, 144.7, 171.7 90, 90, 120 2.95 1 7.2

GlmSFru6P 2BPL C2 132.2, 109.7, 176.3 90, 97.1, 90 2.05 3 7.5

GlmSGlc6PDON 2J6H P212121 83.3, 91.2, 185.0 90, 90, 90 2.35 2 7.2

GlmSGlcN6P 2VF5 H32 144.7, 144.7, 171.7 90, 90, 120 2.9 1 7.2

C1A-GlmSGlc6PGlu 3OOJ H32 247, 247, 630 90, 90, 120 2.6 8 7

In addition, the C1A mutant of GlmS (Table 2) was crystallized in the presence of both substrates (S. Mouilleron, M.-A. Badet-Denisot, B. Badet and B. Golinelli-Pimpaneau, in preparation). In these crystals, Fru6P was mainly converted to Glc6P and Gln to Glu (C1AGlmSGlc6PGlu crystal). Since the C1A mutant does not catalyze glutamine hydrolysis (Table 1), the presence of glutamate in the crystals likely comes from spontaneous hydrolysis of glutamine during crystallization. On the other hand, anisotropic renement was carried out for each crystal structure [17,20], which gave informations about the movement of the GlmS domains inside the crystals. Collective motions of the protein were also studied by a normal mode analysis [22] and the dynamics of the ammonia channel by steered molecular dynamic simulations [23].

Conformational change of Lys503* and translation of helix CF Upon Fru6P binding, the glutaminase domains get a xed position relative to the synthase domains, which brings in particular Trp74 near the His-loop (Fig. 2E). This movement causes steric hindrance and is accompanied by a conformational change of the Hisloop. On the one hand, catalytic His504* comes nearer the synthase site, which supports its role in sugar ring opening (Scheme 3). On the other hand, the swing of the Lys503* side chain maintains the ionic interaction with Glu535* and creates a new salt bridge with Glu608 of the C-loop. This latter interaction between the His-loop of one monomer and the C-loop of the second monomer participates to the closure of the synthase site. The concomitant translation of helix CF by one helix turn allows a bidentate ionic interaction between Arg539* and Asp29 from the incoming glutaminase domain (Fig. 2E). Formation of a closed channel When Fru6P is bound at the synthase site, a channel, which is constituted mostly by hydrophobic residues, starts to get formed. It is made up of the C-loop, the His-loop of the second monomer (including the side chain of Lys503* whose conformation has changed), as well as Arg26 and Trp74 from one glutaminase domain (Fig. 3A and B). Since the C-loop is a major building block of the channel and since it is not ordered in the absence of Fru6P, the channel is not formed when no sugar is bound at the synthase site. In the presence of Fru6P but in the absence of Gln, the channel is almost completely formed, although the two active sites cannot communicate through the channel because of its obstruction by the indole group of Trp74 (Fig. 3B). This closure of the channel in the absence of Gln hinders GlmS to use free ammonia instead of glutamine for sugar amination. Mobility of the domains in the crystals After isotropic renement of the GlmS structure, the two subdomains that form one synthase domain were shown to have very different temperature factors [17]. This led us to carry out TLS renement of the crystallographic data to model the rigid-body motion of the subdomains in the crystals in terms of Translation, Libration and Screw motions. In this way, X-ray crystallography can provide experimental information about the atomic displacements in the crystals and therefore about the dynamic structure of the protein [25,26]. Analysis of the anisotropic displacements inside the GlmS crystals indicates a relatively large rigid-body translational motion of the subdomain that is linked to the mobile glutaminase domain (0.376 2 compared to an estimated value of 0.165 2 for a domain of this size [27]). The GlmSFru6P structure contains three copies of the monomer in the asymmetric unit (Table 2), which presented large anisotropic atomic displacements inside the crystals and could not be properly modeled using isotropic renement programs [20]. Anisotropic

Conformational changes upon Fru6P binding The conformational changes occurring upon Fru6P binding can be addressed by comparing the GlmS structure, solved in the absence of ligand, and the GlmSFru6P structure (Scheme 1). Closure of the synthase site by the C-terminal loop The GlmS structure displays no electron density corresponding to the C-terminal residues, as shown by the superposition of the GlmSFru6P model on the electron density corresponding to the GlmS structure (Fig. 2A). Therefore, the C-terminal loop (C-loop, residues 600608) is exible in the GlmS structure and therefore, the synthase site is accessible to solvent. Upon Fru6P binding, the C-loop gets ordered and covers the synthase site, protecting the sugar from solvent (Fig. 2B). Ordering of the glutaminase domains The GlmS structure displays no electron density corresponding to the glutaminase domains (Fig. 2C) although SDSPAGE analysis of the dissolved crystals indicates that full-length GlmS is present (Fig. 2D). This result demonstrates that the glutaminase domains are not ordered in the absence of ligand at the synthase site in this crystal form, indicating their intrinsic mobility. In contrast, electron density corresponding to the glutaminase domains is observed in the GlmSFru6P structure, which indicates that the glutaminase domains become ordered once Fru6P is bound. In the GlmSFru6P structure, two synthase domains form the dimer interface, whereas the two glutaminase domains are located on each side (Fig. 2C). Upon Fru6P binding, the aromatic groups of Tyr28 and Trp74 belonging to the glutaminase domain get anchored onto the C-loop (Fig. 2E), which indicates that the closure of the C-loop on the synthase site precedes the functional positioning of the glutaminase domains relative to the synthase domains. These large conformational changes occurring upon Fru6P binding are expected to be slow, in agreement with the formation of the fructosimine-6P intermediate being rate-limiting (Scheme 3) [24].

S. Mouilleron et al. / Archives of Biochemistry and Biophysics 505 (2011) 112

Fig. 2. The comparison of the GlmS and GlmSFru6P structures highlights the conformational changes occurring upon Fru6P binding. (A) Superimposition of the GlmSFru6P and GlmS structures on the synthase site. The 2Fo Fcalc electron density map displayed for the GlmS structure (contoured at the level of 1r, in gray), shows no density for the C-loop. Therefore, the empty synthase site is accessible to solvent, whereas the C-loop (in coil) covers the synthase site when Fru6P is bound. (B) View of a molecular surface section of the synthase site of the GlmSFru6P structure. The C-loop covers the synthase site and shields the sugar from solvent. (C) The superposition of the models for the GlmS and GlmSFru6P structures with the 2Fo Fcalc electron density map of the GlmS structure contoured at the level of 1r (in red), shows no density corresponding to the glutaminase domains in the absence of ligands. (D) SDSPAGE analysis of dissolved crystals of unliganded GlmS (lane 2) compared with the puried protein before crystallization (lane 1) indicates that the GlmS crystals contain the full-length protein. (E) Ordering of the glutaminase domains upon Fru6P binding. The anchoring of Trp74 from the glutaminase domain onto the C-loop is accompanied by a conformational change of the Lys503* side chain and the translation of helix CF. When Fru6P is bound, His504* is positioned close to the sugar, supporting its role in sugar ring opening. The two synthase domains are indicated in light green and forest green for GlmS, and cyan and slate for GlmSFru6P. Both glutaminase domains of GlmSFru6P are colored marine Fru6P is shown in cyan sticks.

S. Mouilleron et al. / Archives of Biochemistry and Biophysics 505 (2011) 112

Fig. 3. Formation and opening of the ammonia channel upon substrates binding. (A) Overview of the intramolecular cavity linking the two active sites of one GlmS monomer. (B) Formation of a closed channel upon Fru6P binding. The channel walls are built by the C-loop, the His-loop from the synthase domains as well as Arg26 and Trp74 from the glutaminase domains. (C) Opening of the channel upon DON binding thanks to the swing of the indole group of Trp74. (D) Trp74 acts as the gate that opens the ammonia the channel upon Gln binding, isolating it from solvent. One synthase domain is indicated in cyan for GlmSFru6P, and yellow for GlmSGlc6PDON. One glutaminase domain is colored marine for GlmSFru6P and orange for GlmSGlc6PDON. Fru6P is shown at the bottom in yellow sticks and DON at the top in wheat sticks.

renement indicated individual displacements of the glutaminase domains with different directions of the libration axes. In addition, a particularly large individual libration motion of one of the glutaminase domains, with mean-square displacement up to 8.12 was observed [20]. Therefore, the glutaminase domains, which display extreme mobility in the GlmS structure (no xed position relative to the synthase domains), become more constrained when Fru6P is bound but still remain highly exible. In contrast, the synthase do-

mains forming one dimer move together, indicating the rigidity of the synthase dimer core. Conformational changes upon glutamine binding The conformational changes occurring upon Gln binding can be visualized by comparing the GlmSFru6P and GlmSGlc6PDON structures (Scheme 1) [19]. DON forms a covalent adduct with

S. Mouilleron et al. / Archives of Biochemistry and Biophysics 505 (2011) 112

Fig. 4. The comparison of the GlmSFru6P and GlmSGlc6PDON active sites structures highlights the conformational changes occurring upon glutamine binding. (A) Synthase site. The main changes are the ipping of the peptide bond of Lys603 and the conformational changes of the side chains of Ser401 and Ser604. (B) Glutaminase site. Huge rearrangements occur including the closure of the Q-loop to protect the glutaminase site from solvent, the swing of Trp74 to close the channel, the anchoring of the amino and carboxylate functions of the ligand by Arg73 and Asp123, the formation of the oxyanion hole with the repositioning of Asn98 and Gly99, the conformational change of the Arg26 side chain. (C) Environment of the C-loop (left: GlmSFru6P structure; right: GlmSGlc6PDON structure). The major rearrangement of Asn98 and Cys1 mediated by Arg26 activates the glutaminase function. The two synthase domains are indicated in cyan and slate for GlmSFru6P, and wheat and yellow for GlmSGlc6PDON. Both glutaminase domains are colored marine for GlmSFru6P and orange for GlmSGlc6PDON. Fru6P is shown in cyan sticks for GlmSFru6P and in yellow sticks for GlmSGlc6PDON. DON is shown in wheat sticks.

the amino-terminal cysteine, which mimics the glutamyl-thioester formed during glutamine hydrolysis (Scheme 2B).

the ipping of the peptide bond of Lys603. These conformational changes are linked to a rearrangement of Arg26 from the glutaminase domain, which is involved in the activation of the glutaminase function, as discussed below.

Rearrangement of the synthase active site Superposition of the synthase sites of the GlmSFru6P and GlmSGlc6PDON structures indicates that binding of DON induces only few changes (Fig. 4A). The major difference concerns two residues from the C-loop: the rotation of the side chain of Ser604 and Closure of the glutaminase site and anchoring of glutamine Comparison of the glutaminase sites of the GlmSFru6P and GlmSGlc6PDON structures shows that, upon DON binding, the

S. Mouilleron et al. / Archives of Biochemistry and Biophysics 505 (2011) 112

Q-loop (residues 7381) closes the glutaminase site, shielding it from solvent and allowing Arg73 to anchor the carboxylate group of DON via bidentate direct ionic interactions (Fig. 4B). Binding of DON is also ensured by the rearrangement of several other residues from the glutaminase site: the carboxylate of Asp123 and the main-chain carbonyl group of Gly99 bind to the DON amino group via a salt bridge and an H-bond, respectively. Activation of the catalytic residues and formation of the oxyanion hole Before hydrolysis of glutamine takes place, the amino-terminal function of Cys1 is likely deprotonated and could act as a catalytic base to deprotonate the nucleophile, as proposed for other members of the N-terminal nucleophile enzyme family [28,29] (Scheme 2A). Actually, the H-bond network observed in the GlmSFru6P structure is not in contradiction with such a protonation state since the terminal amine is bound to the main-chain carbonyl group of Arg26 and to the hydroxyl side chain of Thr606 via a water molecule (Fig. 4C, left). Although the pKa of a free a-amino group in a protein is expected to lie in the range 6.87.9 [30], a pKa of 55.3 for the a-amino-terminal group of GlmS complexed with Fru6P was calculated with the program propka (http://propka.ki.ku.dk/pka/) [31], which suggests it is deprotonated at the reaction pH of 7.5. Although the uncharged a-amino group could directly deprotonate the cysteine thiol group, it is hydrogen-bonded to a water molecule in the GlmSFru6P structure (Fig. 4C, left). After abstraction of a proton by the a-amino group from this water molecule, the latter could act as a base and abstract the proton from the thiol group of Cys1. The activation of the nucleophilic serine by its own amino group via a bridging water molecule has previously been proposed for penicillin acylase [29]. The critical function of the free a-amino group of Cys1 in glutamine hydrolysis is illustrated by the activity of the protein in which a Gly residue was added at the N-terminus of GlmS (Gly0 mutant, Table 1) [12], which displays a 40-fold decrease in afnity for Gln, as well as 8-fold and 26-fold lower kcat for glutaminase activity in the absence and presence of Fru6P, respectively. After deprotonation, the thiolate group of Cys1 attacks the carbonyl carbon of glutamine to form an oxyanion intermediate, which moves towards the glutamyl-thioester intermediate by releasing ammonia (Scheme 2). In the GlmSGlc6PDON structure, which mimics the latter intermediate, the carbonyl group is bound by the oxyanion hole formed by the main-chain nitrogen of Gly99 and the side chain amino group of Asn98, which has undergone an important rearrangement upon DON binding, as emphasized by the 100 rotation of this side chain (Fig. 4B). The functional orientation of Asn98 is maintained by an H-bond to the side chain of Arg26, whose conformation has also changed upon DON binding. Arg26 also locks the position of Cys1 by forming H-bonds with its carbonyl and a-amino groups. The C-loop lies in the region located between the two active sites and participates in interdomain signaling by relaying the conformational change of Arg26. In the GlmSFru6P and GlmSGlc6PDON structures, on the sugar side, the C-loop takes part in the binding of the sugar via water-mediated H-bonds, which allows the sugar exibility necessary for the reaction. In the GlmSFru6P structure, on the glutaminase site side, the C-loop contributes to maintain the conformation of Arg26 through two direct H-bonds involving the carbonyl group of Lys603 and the hydroxyl group of Thr606. Upon DON binding, the C-loop does not form direct H-bonds with Arg26 any longer. The hydroxyl group of Thr606 (belonging to the C-loop) is H-bonded to the a-amino group of Cys1 via a water molecule in the GlmSFru6P structure but directly in the GlmSGlc6PDON structure. It also forms a direct H-bond to the oxyanion hole residue Asn98 in the latter structure. Thr606 has therefore a crucial role in the activation of the glutaminase function of GlmS. The contribution

of the C-loop to the enzyme function is highlighted by the 38-fold lower activity of the protein that contains a His6-tag at the C-terminus, compared to the wild-type enzyme [32]. The Gln hydrolysis rate is increased 100-fold when Fru6P is bound, as calculated by the ratio of the catalytic constants for the hemi-synthase and glutaminase activities of GlmS (Table 1). This result suggests that sugar binding has profound effects on the catalytic residues involved in Gln hydrolysis. Indeed, the oxyanion hole is not formed and the thiol group of Cys1 points away from the glutaminase active site in the structure of the isolated glutaminase domain in complex with Glu [12] (see Fig. 3D in [19]). In contrast, the oxyanion hole is fully formed in the GlmSGlc6PDON structure. Therefore, it is the presence of Fru6P and Gln at both active sites that activates the glutaminase function of GlmS. This mechanism of glutaminase function activation couples the glutaminase and synthase activities, ensuring that only a low amount of ammonia is lost into the medium in the absence of the sugar. Opening of the channel Along the closure of the Q-loop, the v1 torsion angle of Trp74, which belongs to this loop, changes by 75 (Fig. 3C and D). This motion opens the channel (i.e. the two active sites are now linked by a solvent-inaccessible channel), indicating that the indole group of Trp74 serves as the gate of the channel, while allowing the sequestration of ammonia from bulk solvent. Whereas ammonia transfer was inefcient in the W74A mutant (Table 1), it was partially restored by increasing the size of the side chain [23]. Indeed, the efciency of ammonia transfer, calculated as the ratio between kcat for the synthase activity (formation of GlcN6P) and kcat for the hemisynthase activity (Gln hydrolysis in the presence of Fru6P) decreased from 84% for the wild-type enzyme to 12.8% for the W74F mutant and to 4.7% for the W74A and W74L mutants. Molecular dynamic simulations indicate that local dynamical motions of a few side chains (Ala602 and Val605) allow ammonia to move through the channel [23]. In addition, it was shown that, in the Trp74 mutants, ammonia may use an alternative channel formed in the second GlmS monomer to escape to bulk solvent. However, since exogenous ammonia still cannot be used by the mutant enzymes for the synthesis of GlcN6P, it appears that it cannot clear a path to the glutaminase site via this second channel. Rotation of the glutaminase domains relative to the synthase domains The effect of Gln binding on the global conformation of the enzyme can be appraised by superposing individually the synthase or the glutaminase domains of the GlmSFru6P and GlmSGlu6PDON structures (Fig. 5A and B). The superposition of the synthase domains of the two structures indicates a rotation of the glutaminase domains relative to the synthase domains upon DON binding, whereas the superposition of the glutaminase domains shows that this motion induces the closure of the glutaminase site by the Qloop while maintaining the dimeric architecture of the enzyme. The domain movements between the GlmSFru6P and GlmSGlu6PDON structures were more precisely analyzed using the program DynDom [20]. DynDom determines domains, hinge axes and hinge bending residues in proteins, where two conformations are available, by comparing their rotational properties [33,34]. Two dynamic domains, which correspond closely to the synthase and glutaminase structural domains, were identied. The movement of the glutaminase domains relative to the synthase domains upon glutamine binding was described as a 22.8 rotation around a hinge axis running approximately parallel to helix 300 317 of the synthase domain. The hinge regions were characterized as the Q-loop and the peptide linking the two domains (residues

S. Mouilleron et al. / Archives of Biochemistry and Biophysics 505 (2011) 112

Fig. 5. Large amplitude movements upon DON binding. (A) Rotation of the glutaminase domains. (B) Closure of the Q-loop over the glutaminase site. The two synthase domains are indicated in cyan and slate for GlmSFru6P, and wheat and yellow for GlmSGlc6PDON. Both glutaminase domains are colored marine for GlmSFru6P and orange for GlmSGlc6PDON. Fru6P is shown in cyan sticks for GlmSFru6P and in yellow sticks for GlmSGlc6PDON. DON is shown in wheat sticks.

235248). Actually, the linker between the two domains is expected to provide the exibility necessary for the enzyme function. Interestingly, insertion of a His6-tag at position 225 corresponding to the beginning of the linker leads to a protein as active as the wild-type enzyme, whereas insertion at position 240, located close to the interface between the glutaminase and synthase domains and not far away from the C-loop, inactivates the protein [32]. Additionally, the low-frequency normal modes were analyzed using the GlmSGlu6PDON structure to examine the collective large amplitude movements of the protein that are likely functionally relevant [22]. One of the ve lowest frequency normal modes successfully reproduced the rotation of the glutaminase domain occurring upon glutamine binding, which was previously identied with DynDom using the crystal structures. Compared to the GlmSFru6P crystals, the GlmSGlc6PDON crystals displayed a much less pronounced anisotropic behavior. Anisotropic renement showed, in particular, smaller individual libration motions of the glutaminase domains, with mean-square displacement only up to 2.92 [20]. This result indicates a less pronounced intrinsic exibility of the GlmS glutaminase domains when ligands are bound at both active sites. This relative rigidity permits the maintenance of an open ammonia channel connecting the two active sites.

Conformational changes upon glucosamine-6P release The comparison of the GlmSGlcN6P and GlmS structures sheds light on the conformational changes occurring upon glucosamine6P release (Scheme 1) [17]. In the GlmSGlcN6P structure, the sugar

adopts the cyclic form and the C-loop covers the synthase site (Fig. 6A). Apart from the ordered conformation of the C-loop, the GlmSGlcN6P structure is very similar to the unliganded structure, just as the anisotropic displacements inside the crystals. In particular, the glutaminase domains, which are not observed in the electron density, are not ordered. The conformation of the His-loop is different in the GlmSGlcN6P and GlmS structures, with an important movement of Lys503*, which partially closes the synthase site in the presence of GlcN6P and the repositioning of His504*, which is positioned near the sugar in the GlmSGlcN6P structure. The C-loop displays slightly different conformations in the structures in complex with the linear Fru6P substrate and the cyclic GlcN6P product (Fig. 6B). In the GlmSGlcN6P complex, the Cloop displays a more relaxed conformation, making direct interactions with the sugar but few interactions with the rest of the protein, and the sugar is less buried in the active site pocket. The catalytic residues His504*, Lys485 and Glu488 are further away form the sugar in the GlmSGlcN6P structure compared to the GlmSFru6P structure. This different positioning of His504* is again in agreement with its putative role in sugar ring opening/closure, that of Lys485 with its involvement in the deprotonation of the sugar O5 hydroxyl group and that of Glu488 with its potential function as a base that deprotonates the C1/reprotonates the C2 sugar atom (Scheme 3). The comparison of the GlmSGlcN6P and GlmSFru6P complexes therefore indicates that the GlmSGlcN6P structure corresponds to the conformation when GlcN6P is just about to leave the enzyme. GlcN6P release involves the opening of the glutaminase domains and of the C-tail, a motion again expected to be slow. Indeed, this step was found to be partially rate-limiting [24].

10

S. Mouilleron et al. / Archives of Biochemistry and Biophysics 505 (2011) 112

an intermediate state of the enzyme along the catalytic pathway. Indeed, the structure of the mutant is organized as a trimer of dimers, forming a hexamer (Fig. 7). In this architectural organization, the glutaminase domains are ordered but not oriented relative to the synthase domains in a catalytically competent way (article in preparation). Interestingly, the hexameric form of the enzyme was previously observed in the structures of the isolated synthase domain [13,14], as well as in the GlmS and GlmSGlcN6P structures [17] (Fig. 7). In all these cases, the glutaminase domains are either not present, not ordered or adopt a catalytically inactive conformation. The observation that different conformations of the enzyme accommodate different quaternary structures suggests a regulation of the catalytic activity of GlmS by a shift of the equilibrium between its quaternary forms, the hexamer being the inactive form and the dimer the active one. This behavior termed morphein has been reported for a few enzymes [35,36]. Whereas the existence of GlmS oligomers higher than a dimer has been detected in solution by sedimentation velocity experiments [37], it remains to demonstrate that GlmS exhibits kinetic phenomena characteristic of a morphein equilibrium, such as protein-concentration dependence of the activity, non-Michaelis behavior, kinetic hysteresis, dependence of activity upon the order of addition of reaction components. A rst hint that the protein activity is concentration-dependent comes from abnormal substrate binding stoichiometries [37]. Ammonia channeling in other glutamine amidotransferases Insights into the molecular mechanisms, which explain how other glutamine amidotransferases coordinate their catalytic activity in each active site to sequester ammonia from bulk solvent, have previously been reviewed [3840]. Substrate binding to the ammonia-accepting active site generally leads to reorganization of a synthase domain loop (C-loop for GlmS), which is a major building block of the channel walls. Thus, most ammonia channels appear to be transiently formed only after the substrates are bound. Acceptor binding also stimulates glutaminase activity as a result of conformational changes in the glutaminase catalytic pocket.

Fig. 6. Conformational changes upon GlcN6P binding. (A) Comparison of synthase sites of the GlmS and GlmSGlcN6P structures. When GlcN6P is bound, the C-loop is ordered. (B) Comparison of the synthase sites of the GlmSFru6P and GlmSGlcN6P structures. In the GlmSGlcN6P structure, the C-loop adopts a more relaxed conformation and the sugar is less buried in the active site pocket. Catalytic Lys485, Glu488 and His504* are positioned farther away from the cyclic sugar compared to the liner sugar. The two synthase domains are indicated in light green and forest green for GlmS, cyan and slate for GlmSFru6P, and magenta and pink for GlmSGlcN6P. Fru6P is shown in cyan sticks and GlcN6P in pink sticks.

Quaternary structure of GlmS Another strategy to obtain the structure of GlmS in complex with both substrates, besides the use of a Gln analog (GlmSGlc6PDON structure), was to crystallize an inactive mutant of GlmS in complex with Fru6P and Gln. Since Cys1 is crucial for the catalytic activity of GlmS (Scheme 2), the C1A mutant of GlmS was used to test this approach although the C1A-GlmSGlc6PGlu structure does not mimic

Fig. 7. Superposition of two different hexameric structures of GlmS. One complete hexamer present in the asymmetric unit of the C1A-GlmSGlcN6PGlu structure is shown in red and grayish-blue for the synthase and glutaminase domains, respectively. The hexamer of the GlmS structure, which contains only the synthase domains, was reconstituted using the crystal symmetry and is colored green.

S. Mouilleron et al. / Archives of Biochemistry and Biophysics 505 (2011) 112

11

Fig. 8. Structural catalytic scheme of GlmS (only one monomer is shown). In the absence of ligand, the glutaminase domains do not adopt a xed position relative to the synthase domains. The C-loop is exible and both active sites are accessible to solvent. Binding of Fru6P triggers the closure of the C-loop on the synthase site and the functional positioning of the glutaminase domains relative to the synthase domains. The ammonia channel is formed but remains closed by the Trp74 indole ring. After the formation of a Schiff base with Lys603, binding of Gln triggers the closure of the Q-loop on the glutaminase site, the opening of the channel and the activation of the glutaminase function. After sugar amination, release of Glu followed by that of GlcN6P regenerates free enzyme.

The structurefunction relationship of ammonia channels in glutamine amidotransferases was recently further elucidated by computational studies [41,42]. Interestingly, ammonium ions, when modeled instead of ammonia in glutamine 50 -phosphoribosylpyrophosphate amidotransferase, could not access the interior of the channel but leaked into bulk solvent [42]. Similarly, molecular dynamic simulations indicate that ammonium is not transferred through the channel in the small subunit of carbamoyl phosphate synthetase but remain immobilized in a water pocket [41]. These results indicate that at least these two enzymes, together with imidazole glycerol phosphate synthase [43,44], discriminate against ammonium ion. In these cases, the driving force for the translocation of ammonia is not the consequence of an electrostatic potential between the two active sites but comes instead from the solvation energy of ammonia [41] and local dynamical movements of the protein [42]. Ammonia moves through the channel by exchanging H-bonds with the surrounding residues and several water molecules, whereas the sites vacated by ammonia are relled by water molecules.

(Fig. 8). X-ray crystallography thus revealed that ammonia is transferred from the glutaminase site to the synthase site through an 18- long channel with Trp74 serving as the gate of the channel. In the absence of Gln, the channel is closed, which prevents the enzyme to use ammonia for sugar amination. Upon glutamine hydrolysis, the channel opens and is sequestered from solvent so that the ammonia intermediate product is transferred to the second active site. Taken together, the step-by-step formation of a usable channel together with the activation of the glutaminase function of the enzyme upon binding of the substrates allows efcient transfer of ammonia from glutamine only when the sugar is bound at the synthase site and ready to accept ammonia. Among the various glutamine amidotransferases that have been structurally characterized to date, the complete structural mechanism of ammonia transfer, which involves large domain movements triggered by the binding of the substrates, has been deciphered only for GlmS [40]. Further elucidation of the structure and mechanism of action of this class of enzymes will probably have signicant implications for the discovery of new therapeutic agents.

Conclusion The knowledge of the catalytic mechanism of GlmS, a two substrates-handling enzyme, has allowed to crystallize various conformations of the enzymatic complexes along the catalytic pathway. Comparison of the unliganded and liganded GlmS structures discloses the conformational changes that occur during catalysis

References
[1] S. Ghosh, H.J. Blumenthal, E. Davidson, S. Roseman, J. Biol. Chem. 235 (1960) 12651273. [2] W.L. Whelan, C.E. Ballou, J. Bacteriol. 124 (1975) 15451557. [3] M. Sarvas, J. Bacteriol. 105 (1971) 467471. [4] E. Borowski, Farmaco 55 (2000) 206208. [5] P. Bhonagiri, G.R. Pattar, E.M. Horvath, K.M. Habegger, A.M. McCarthy, J.S. Elmendorf, Endocrinology 150 (2009) 16361645.

12

S. Mouilleron et al. / Archives of Biochemistry and Biophysics 505 (2011) 112 [25] P.J. Artymiuk, C.C. Blake, D.E. Grace, S.J. Oatley, D.C. Phillips, M.J. Sternberg, Nature 280 (1979) 563568. [26] J. Painter, E.A. Merritt, Acta Crystallogr. D: Biol. Crystallogr. 62 (2006) 439450. [27] C. Scheringer, Acta Crystallogr. A 28 (1972) 516522. [28] C. Oinonen, R. Tikkanen, J. Rouvinen, L. Peltonen, Nat. Struct. Biol. 2 (1995) 11021108. [29] H.J. Duggleby, S.P. Tolley, C.P. Hill, E.J. Dodson, G. Dodson, P.C. Moody, Nature 373 (1995) 264268. [30] C. Tanford, Adv. Protein Chem. 17 (1962) 69165. [31] D.C. Bas, D.M. Rogers, J.H. Jensen, Proteins 73 (2008) 765783. [32] C. Richez, J. Boetzel, N. Floquet, K. Koteshwar, J. Stevens, B. Badet, M.A. BadetDenisot, Protein Expr. Purif. 54 (2007) 4553. [33] S. Hayward, R.A. Lee, J. Mol. Graph. Model. 21 (2002) 181183. [34] G.P. Poornam, A. Matsumoto, H. Ishida, S. Hayward, Proteins 76 (2009) 201 212. [35] S. Breinig, J. Kervinen, L. Stith, A.S. Wasson, R. Fairman, A. Wlodawer, A. Zdanov, E.K. Jaffe, Nat. Struct. Biol. 10 (2003) 757763. [36] E.K. Jaffe, Trends Biochem. Sci. 30 (2005) 490497. [37] M. Valerio-Lepiniec, M. Aumont-Nicaise, C. Roux, B. Raynal, P. England, B. Badet, M.A. Badet-Denisot, M. Desmadril, Arch. Biochem. Biophys. 498 (2010) 95104. [38] F.M. Raushel, J.B. Thoden, H.M. Holden, Acc. Chem. Res. 36 (2003) 539548. [39] A. Weeks, L. Lund, F.M. Raushel, Curr. Opin. Chem. Biol. 10 (2006) 465472. [40] S. Mouilleron, B. Golinelli-Pimpaneau, Curr. Opin. Struct. Biol. 17 (2007) 653 664. [41] Y. Fan, L. Lund, L. Yang, F.M. Raushel, Y.Q. Gao, Biochemistry 47 (2008) 2935 2944. [42] X.S. Wang, A.E. Roitberg, N.G. Richards, Biochemistry 48 (2009) 1227212282. [43] R. Amaro, Z.A. Luthey-Schulten, Chem. Phys. 307 (2004) 147155. [44] R.E. Amaro, R.S. Myers, V.J. Davisson, Z.A. Luthey-Schulten, Biophys. J. 89 (2005) 475487. [45] M.A. Badet-Denisot, L. Ren, B. Badet, Bull. Soc. Chim. Fr. 130 (1993) 249255.

[6] M. Brownlee, Nature 414 (2001) 813820. [7] M.G. Buse, Am. J. Physiol. Endocrinol. Metab. 290 (2006) E1E8. [8] P. Durand, B. Golinelli-Pimpaneau, S. Mouilleron, B. Badet, M.A. Badet-Denisot, Arch. Biochem. Biophys. 474 (2008) 302317. [9] B. Badet, P. Vermoote, F. Le Gofc, Biochemistry 27 (1988) 22822287. [10] M.A. Denisot, F. Le Gofc, B. Badet, Arch. Biochem. Biophys. 288 (1991) 225 230. [11] C. Leriche, M.A. Badet-Denisot, B. Badet, J. Am. Chem. Soc. 118 (1996) 1797 1798. [12] M.N. Isupov, G. Obmolova, S. Butterworth, M.A. Badet-Denisot, B. Badet, I. Polikarpov, J.A. Littlechild, A. Teplyakov, Structure 4 (1996) 801810. [13] A. Teplyakov, G. Obmolova, M.A. Badet-Denisot, B. Badet, Protein Sci. 8 (1999) 596602. [14] A. Teplyakov, G. Obmolova, M.A. Badet-Denisot, B. Badet, I. Polikarpov, Structure 6 (1998) 10471055. [15] J.A. Brannigan, G. Dodson, H.J. Duggleby, P.C. Moody, J.L. Smith, D.R. Tomchick, A.G. Murzin, Nature 378 (1995) 416419. [16] B. Golinelli-Pimpaneau, B. Badet, Eur. J. Biochem. 201 (1991) 175182. [17] S. Mouilleron, M.A. Badet-Denisot, B. Golinelli-Pimpaneau, J. Mol. Biol. 377 (2008) 11741185. [18] A. Teplyakov, G. Obmolova, B. Badet, M.A. Badet-Denisot, J. Mol. Biol. 313 (2001) 10931102. [19] S. Mouilleron, M.-A. Badet-Denisot, B. Golinelli-Pimpaneau, J. Biol. Chem. 281 (2006) 44044412. [20] S. Mouilleron, B. Golinelli-Pimpaneau, Protein Sci. 16 (2007) 485493. [21] M. St-Jean, J. Sygusch, J. Biol. Chem. 282 (2007) 3102831037. [22] N. Floquet, P. Durand, B. Maigret, B. Badet, M.A. Badet-Denisot, D. Perahia, J. Mol. Biol. 385 (2009) 653664. [23] N. Floquet, S. Mouilleron, R. Daher, B. Maigret, B. Badet, M.A. Badet-Denisot, FEBS Lett. 581 (2007) 29812987. [24] B. Golinelli-Pimpaneau, F. Le Gofc, B. Badet, J. Am. Chem. Soc. 111 (1989) 30293034.