Role of K22 and R120 in the covalent binding of the antibiotic fosfomycin and the substrate-induced conformational change in UDP- N -acetylglucosamine enol pyruvyl transferase Alison M. Thomas 1, *, Cristian Ginj 1, *, Ilian Jelesarov 2 , Nikolaus Amrhein 1 and Peter Macheroux 1 1 Eidgeno ¨ ssische Technische Hochschule Zu ¨ rich, Institute of Plant Sciences, Department of Agricultural and Food Sciences and Department of Biology, Zu ¨ rich, Switzerland; 2 Universita ¨ tZu ¨ rich, Institute of Biochemistry, Zu ¨ rich, Switzerland UDP-N-acetylglucosamine enolpyruvyl transferase (MurA), catalyzes the first step in the biosynthesis of peptidoglycan, involving the transfer of the intact enolpyruvyl moiety from phosphoenolpyruvate to the 3¢-hydroxyl group of UDP-N- acetylglucosamine (UDPNAG). The enzyme is irreversibly inhibited by the antibiotic fosfomycin. The inactivation is caused by alkylation of a highly conserved cysteine residue (C115) that participates in the binding of phos- phoenolpyruvate. The three-dimensional structure of the enzyme suggests that two residues may play a decisive role in fosfomycin binding: K22 and R120. To investigate the role of these residues, we have generated the K22V, K22E, K22R and R120K single mutant proteins as well as the K22V/ R120K and K22V/R120V double mutant proteins. We demonstrated that the K22R mutant protein behaves simi- larly to wild-type enzyme, whereas the K22E mutant protein failed to form the covalent adduct. On the other hand, the K22V mutant protein requires the presence of UDPNAG for the formation of the adduct indicating that UDPNAG plays a crucial role in the organization of productive inter- actions in the active site. This model receives strong support from heat capacity changes observed for the K22V/R120K and R120K mutant proteins: in both mutant proteins, the heat capacity changes are markedly reduced indicating that their ability to form a closed protein conformation is impeded due to the R120K exchange. Keywords: transferase; fosfomycin; antibiotic; mutagenesis; protein conformation. A rigid cell wall is essential for the survival of most bacteria. Compounds that interfere with cell wall biosynthesis or function, such as b-lactams, are powerful antibiotics and the bacterial enzymes involved in cell wall biosynthesis are attractive targets for the development of new drugs [1]. The biosynthesis of the cell wall component peptidoglycan (or murein) commences with the transfer of the intact enolpyr- uvyl moiety of phosphoenolpyruvate to the 3¢-hydroxyl group of UDP-N-acetylglucosamine (UDPNAG) [2]. This reaction, catalysed by UDP-N-acetylglucosamine enol- pyruvyl transferase (MurA), leads to the generation of UDP-N-acetylenolpyruvylglucosamine (Scheme 1A). The naturally occurring antibiotic fosfomycin, produced by some Streptomyces and Pseudomonas species [3–5], irreversibly inhibits MurA activity by alkylating the thiol group of a catalytically important cysteine residue, C115 (Scheme 1B) [6]. The rate of MurA inactivation by fosfomycin is increased considerably in the presence of UDPNAG [7]. This accel- erating effect is not due to a change in the reactivity of the thiol group, as the pK a of the thiol group is not affected by UDPNAG binding [8]. Crystallographic studies have shown that MurA is subject to a large conformational change upon binding of UDPNAG and fosfomycin or UDPNAG and (Z)-3-fluorophosphoenolpyruvate, respectively, to the free, unliganded enzyme [9–11] (Fig. 1). In the unliganded form, the active site of MurA is readily accessible (ÔopenÕ confor- mation) whereas in the liganded form (ÔclosedÕ conforma- tion) a loop in the upper domain forms a lid on the active site, thereby shielding the ligands from solvent and gener- ating a compact structure. This loop movement places the reactive C115 closer to fosfomycin or (Z)-3-fluorophos- phoenolpyruvate in the active site (Fig. 1). Hence it can be assumed that fosfomycin and the thiol group of C115 are optimally positioned in the closed conformation, so that the nucleophilic attack of the thiol group is facilitated. In a recently initiated site-directed mutagenesis program, we have discovered that replacement of K22 leads to a more than 300-fold decrease in enzymatic activity [12]. Using isothermal titration calorimetry (ITC), fosfomycin binding was detected for the conservative mutation K22R in the presence of UDPNAG while the K22V and K22E mutant proteins appeared to have lost this ability completely [12]. According to the three-dimensional structure of MurA [11], Correspondence to P. Macheroux, Graz University of Technology, Institute of Biochemistry, Petersgasse 12/II, A-8010 Graz, Austria. Fax: + 43 316 873 6952, Tel.: + 43 316 873 6450, E-mail: peter.macheroux@tugraz.at Abbreviations: fosfomycin, (1R,2S)-1,2-epoxypropylphosphonic acid; glyphosate, N-(phosphonomethyl)-glycine; ITC, isothermal titration calorimetry; MurA, UDP-N-acetylglucosamine enolpyruvyl trans- ferase; TPCK, L -(tosylamido-2-phenyl) ethyl chloromethyl ketone; UDPNAG, UDP-N-acetylglucosamine; DC p , heat capacity change; DG, free energy change; DH, enthalpy change; DS, entropy change. *Note: The first two authors contributed equally to this work. (Received 16 February 2004, revised 26 April 2004, accepted 30 April 2004) Eur. J. Biochem. 271, 2682–2690 (2004) Ó FEBS 2004 doi:10.1111/j.1432-1033.2004.04196.x the positively charged side chain of K22 participates in fosfomycin binding, thus providing a rationale for the loss of fosfomycin binding to the K22V and K22E mutant proteins. However, two other positively charged amino acid side chains, R397 and R120, are also involved in fosfomycin binding, and in view of the multitude of interactions, it was assumed that deletion of a single interaction would not lead to a complete loss of fosfomycin binding. Therefore it was argued that K22 may play a role in the transition of the open and closed conformations, i.e. it is part of a molecular switch mechanism. To shed more light on the energetics of the conform- ational change, we have recently completed a thermody- namic study of UDPNAG binding to MurA [13]. Based on the analysis of the measured heat capacity changes (DC p ), and on surface accessibility calculations, we have proposed that binding of UDPNAG alone is accompanied by a significant structural shift toward the closed conformation, in agreement with evidence from studies of small angle X-ray scattering [14] and the protective effect of UDPNAG on proteolysis of MurA [15]. Here we report the thermo- dynamic profile of UDPNAG binding to two MurA single mutants, K22V and R120K, and the double mutant K22V/ R120K. The emphasis is put on analysis of the heat capacity decrement as this parameter is a sensitive indicator of both the changes in hydration and the conformational changes involved in protein–ligand interactions [16,17], and thus may provide further information on the molecular mechanism of fosfomycin and UDPNAG binding to MurA. Fig. 1. Structural representation of MurA in the open (A, PDB entry 1NAW [9]), and closed (B, PDB entry 1UAE [11]), conformation. The thiol group of C115 is highlighted in green and the nitrogen atoms of R120 and K22 in blue. The open form of MurA is unliganded (A) and the closed conformation (B) contains fosfo- mycin (orange) and UDPNAG (purple) in the active site. The loop that carries C115 and R120 is highlighted in yellow (A) and red (B), respectively. The figure was prepared using the program MOLMOL [24]. Scheme 1. Reaction catalyzed by MurA (pathway for biosynthesis of UDP-N-acetylmuramic acid) (A) and inactivation mechanism of MurA by fosfomycin as a result of the covalent linkage between Cys115 of MurA and fosfomycin (B). Ó FEBS 2004 Lysine 22 and arginine 120 in MurA (Eur. J. Biochem. 271) 2683 In a previous study, we have combined proteolysis with MALDI-TOF mass spectrometry analysis to demonstrate that fosfomycin forms the covalent adduct with C115 of wild-type MurA even in the absence of UDPNAG [13]. Unlike ITC, which relies on measurable heat changes, this method is independent of the thermodynamic and kinetic properties of the binding process and is solely based on the detectability of the peptides of interest. We have used this method to show that binding of fosfomycin to the K22 mutant proteins strongly depends on the charge of the amino acid side chain occupying this position. Taken together, our combined calorimetric and protein chemical approach leads to a more detailed understanding of the role of K22 and R120 with respect to fosfomycin binding as well as the ligand-induced conformational switch. Experimental procedures Chemicals and enzymes Fosfomycin (disodium salt) and UDP-N-acetylglucosamine (sodium salt) were from Sigma, Buchs, Switzerland. 1,4- dithio- D , L -threitol, EDTA, isopropyl thio-b- D -galactopyr- anoside and Hepes were from Fluka, Buchs, Switzerland. Tris was from BDH Laboratory Supplies, Poole, England. Trypsin (EC 3.4.21.4) from bovine pancreas (12 200 UÆmg )1 ; TPCK-treated to inactivate any remaining chymotryptic activity) was from Sigma, Buchs, Switzerland. Site-directed mutagenesis Mutagenesis was carried out using the Quik-change site- directed mutagenesis kit from Stratagene as previously described [18]. The following oligonucleotides were used to change the R120 to a lysine and a valine, respectively (the codon changes are underlined): 5¢-primer (R to K): 5¢-GGT TGCGCCATTGGCGCG AAACCTGTTGACCTGC ATATC-3¢;3¢-primer (R to K): 5¢-GATATGCAGGTCA ACAGG TTTCGCGCCAATGGCGCAACC-3¢;5¢-primer (R to V): 5¢-GGTTGCGCCATTGGCGCG GTTCCTGT TGACCTGCATATC-3¢;3¢-primer (R to V): 5¢-GATATG CAGGTC AACAGGAACCGCGCCAATGGCGCA ACC-3¢. The template used for amplification was the pKK233-2 plasmid containing Enterobacter cloacae MurA wild-type (for generation of the R120K single mutant) or the plasmid encoding the K22V mutant (for generation of the K22V/ R120K double mutant) [12]. The whole gene was resequ- enced after site-directed mutagenesis in order to confirm the introduction of the desired mutation and to exclude that unwanted mutations occurred elsewhere in the gene (Microsynth, Balgach, Switzerland). Expression and purification of proteins Wild-type MurA (from Enterobacter cloacae)andtheK22V, K22E, K22R, R120K, K22V/R120V and K22V/R120K mutant proteins were expressed and purified as previously described [12,19]. The yield of purified K22V/R120V protein was very low (< 1% as compared to wild-type) shown by the lack of binding to the ÔReactive YellowÕ material used in affinity chromatography. Therefore microcalorimetric measurements (see below) could only be performed with the K22V, R120K and K22V/R120K mutant proteins. The protein concentration was determined using an extinction coefficient of 24 020 M )1 Æcm )1 at 280 nm or with Bradford reagent (Pierce) using bovine serum albumin (BSA) for calibration. MALDI-TOF-MS analysis of fosfomycin binding Wild-type MurA, the K22V, K22R and K22E single mutant proteins and the K22V/R120K and K22V/R120V double mutant proteins (100 l M ) were incubated for 50 min at 25 °Cin50m M Tris/HCl, pH 7.4 under each of the following conditions: (a) no substrates; (b) 10 m M fosfo- mycin; (c) 10 m M fosfomycin and 1 m M UDPNAG; and (d) 1 m M fosfomycin and 10 m M UDPNAG. Trypsin (0.5 mgÆmL )1 ) was added to the reaction and incubated for 16 h at 25 °C. Removal of excess substrates was achieved by desalting the reactions using pre-equilibrated Sep-Pak C 18 columns (Waters) as previously described [13,18]. MALDI-TOF-MS spectra were recorded with a Voyager Elite mass spectrometer using the reflectron mode for increased mass accuracy; interpretation of the spectra was based on the analysis reported earlier [15]. Isothermal titration calorimetry All measurements were carried out in 50 m M Hepes/NaOH, pH 7.4, containing 2 m M dithiothreitol and 0.5 m M EDTA. Sample preparation and titrations were performed as described previously [12] using a MCS isothermal titration microcalorimeter from Microcal Inc. The K22V, R120K and K22V/R120K MurA variants (200–420 l M )were titrated with UDPNAG from a 5 m M stock solution at temperatures between 10 and 30 °C. The double mutant proteins were not stable at 30 °C with stirring in the ITC cell; the data obtained at this temperature was not included in the analysis. The raw data were integrated and normal- ized for molar concentration. The dissociation constants, K d values, enthalpies of binding and stoichiometries were determined from the binding isotherm by fitting a 1 : 1 binding model to the data using the software provided by the manufacturer. Results Reaction of fosfomycin with wild-type and the K22V, K22E and K22R mutant proteins We have shown previously by ITC that fosfomycin binding to wild-type MurA and the K22 mutant proteins in the presence of UDPNAG is accompanied by heat release in the case of wild-type enzyme and the K22R mutant protein, but is calorimetrically silent with the K22V and K22E mutant proteins. This was interpreted as a lack of fosfomycin binding to these mutant proteins [12]. Binding of fosfomycin to either wild-type MurA or the K22 mutant proteins in the absence of UDPNAG is not associated with a heat change in any case. To gain further information on fosfomycin binding to MurA, we have developed a rapid and reliable method to detect the covalent adduct formed between the thiol group 2684 A. M. Thomas et al. (Eur. J. Biochem. 271) Ó FEBS 2004 of C115 and fosfomycin (Scheme 1B). Trypsinolysis of wild- type MurA and the K22 mutant proteins (Fig. 2A and B) produces a peak at 1616 Da (m/z) comprising amino acids 104–120 [15]. Incubation of wild-type MurA with fosfo- mycin prior to trypsinolysis in the absence or presence of UDPNAG results in the appearance of a new peak at 1754 Da (m/z) which is assigned to the Cys115–fosfomycin covalent adduct (observed mass difference ¼ 138 Da, as expected) (Fig. 2C, E and G). At the same time the peak of the unlabeled peptide fragment disappears or becomes Fig. 2. Comparison of MALDI-TOF spectra between the masses of 1600 and 1800 for wild-type Enterobacter cloacae MurA (left) and the K22V mutant protein (right). (A and B) Tryptic digest of wild-type MurA and the K22V mutant protein (no additions). (C and D) Wild-type E. cloacae MurA and the K22V mutant protein incubated with 10 m M fosfomycin prior to digestion. (E and F) Wild-type MurA and the K22V mutant protein incubated with 1 m M UDPNAG and 10 m M fosfomycin prior to digestion. (G and H) Wild-type MurA and the K22V mutant protein incubated with 10 m M UDPNAG and 1 m M fosfomycin prior to digestion. The mass peak of 1616 (m/z) is due to the peptide of amino acids 104– 120 and on binding of fosfomycin this mass shifts to 1754 (m/z). The peak at 1657 (m/z) is attributed to the peptide fragment comprising amino acids 295–310 and is attributed to tryptic and chymotryptic cleavage. Ó FEBS 2004 Lysine 22 and arginine 120 in MurA (Eur. J. Biochem. 271) 2685 much less intense. When the same experiment is carried out with the K22 mutant proteins, marked differences are observed: in the case of the K22V mutant protein, covalent binding of fosfomycin requires the presence of UDPNAG (Fig. 2D, F and H) while the K22E mutant protein lacks the ability to form the covalent adduct completely (Table 1). On the other hand, the K22R mutant protein behaves very similar to wild-type enzyme (data not shown). While the results obtained with the K22R and K22E mutant proteins are in agreement with the ITC measurements, the K22V mutant protein clearly binds fosfomycin covalently, albeit only in the presence of UDPNAG. This finding is in contrast to the lack of a heat signal in ITC measurements. Hence it can be concluded that the absence of an ITC signal with this mutant protein is not due to a lack of adduct formation, but rather indicates that the binding process is not associated with a measurable net heat change. The finding that binding of UDPNAG to the K22V mutant protein restores the ability to form the covalent adduct with fosfomycin raises a question about the molecular mechanism of this salvage process. The three- dimensional structure of the ternary complex [11] indicates that UDPNAG interacts with the phosphonate group of fosfomycin and also with the guanidinium group of R120, which in turn forms a salt-bridge to the phosphonate group (see below). This amino acid residue is invariant in all known MurA sequences and is part of the loop region that forms a lid on the active site upon the formation of the closed protein conformation (Fig. 1). Hence, it is plausible that UDPNAG plays a direct role and/or an indirect role to engage residues in the loop for binding interactions with the phosphonate group of fosfomycin (or with the phosphate group of phosphoenolpyruvate during normal catalysis). In order to investigate the importance of R120 for the formation of the covalent fosfomycin adduct, we construc- ted the K22V/R120V and K22V/R120K double mutants (see Materials and methods) and repeated the analysis of adduct formation in the absence and presence of UDP- NAG. As shown in Fig. 3, the K22V/R120K mutant Table 1. Covalent adduct formation with fosfomycin and enzymatic activity of wild-type MurA and the protein mutants investigated in this study. Proteins were incubated with fosfomycin alone (10 m M final concentration) or with fosfomycin and UDPNAG. The latter experi- ment was performed using either 10 m M fosfomycin and 1 m M UDPNAG or 1 m M fosfomycin and 10 m M UDPNAG, respectively. The results obtained under these two different conditions were essen- tially the same. +, indicates detection of the fosfomycin-labelled tryptic fragment comprising amino acids 104–120; –, no detection. The effect of the mutation(s) on catalytic activity is also listed. *, Residual activities too small to be measured reliably. Protein Fosfomycin Fosfomycin + UDPNAG Activity (%) Wild-type + + 100 K22R + + 0.3 K22V – + 0.03 K22E – – 0.05 R120K – – <0.05 K22V/R120K – – * K22V/R120V – – * Fig. 3. Incubation of the K22V/R120K double mutant protein with fosfomycin followed by MALDI-MS analysis. The experimental con- ditions are as described in the legend to Fig. 2. (A) The double mutant proteinwithnoadditions;(B)incubationwith10m M fosfomycin; (C) incubation with 10 m M fosfomycin and 1 m M UDPNAG and (D) incubation with 1 m M fosfomycin and 10 m M UDPNAG. The peak of the unlabeled tryptic fragment (amino acids 295–310) is at 1588 (m/z) and when labelled with fosfomycin a peak at 1726 (m/z)is expected. The position of the expected mass of the fosfomycin-labelled peak is marked by an arrow in each panel. 2686 A. M. Thomas et al. (Eur. J. Biochem. 271) Ó FEBS 2004 protein did not form the covalent fosfomycin adduct, even in the presence of UDPNAG (Fig. 3C,D). Note that the mass shift of the unlabeled fragment (amino acids 104–120) from 1616 Da (MH + ) to 1588 Da (MH + ) is due to the arginine to lysine exchange in position 120 (expected mass shift ¼ 28 Da). The data collected for wild-type MurA, the three single and two double mutant proteins are summarized in Table 1. From this analysis, it is clear that even a conservative amino acid exchange from arginine to lysine at position 120 disables the UDPNAG-dependent rescue mechanism for the formation of a covalent adduct. This in turn suggests that UDPNAG assists in the assembly of the proper interactions in the active site by stabilizing the closed conformation of the protein. In the closed conformation, R120 moves from an ÔoutsideÕ position into an ÔinsideÕ position close to the negatively charged phosphonate (in the case of fosfomycin) or phosphate (in the case of phos- phoenolpyruvate) group (Fig. 1, compare panel A and B). From this it would follow that R120 is an important residue in the stabilization of the closed conformation of the protein. This hypothesis can be tested by assessing the extent of the conformational change occurring upon addition of UDPNAG to a mutant protein carrying a different amino acid residue in position 120. The experimental strategy chosen to test our hypothesis was recently established with wild-type enzyme demonstrating that the conformational changes occurring upon ligand binding are accompanied by significant heat capacity changes. Determination of the heat capacity changes for the K22V, R120K and K22V/R120K mutant proteins In a recent study, we measured the thermodynamic parameters for UDPNAG binding to wild-type enzyme as a function of temperature [13]. The heat capacity change (DC p ) was obtained from the slope of Kirchoff’s plots (DH vs. T). Analysis of the experimental DC p with DC p calculated from the change of solvent accessible surface upon transition from the open to the closed MurA conformation, indicates that UDPNAG binding alone induces the formation of the closed conformation to a large extent [13]. We have carried out the same analysis with the K22V and R120K single and the K22V/R120K double mutant proteins in order to obtain information on how the replacement of these two important side chains affect the protein conformational change. The binding of UDPNAG is only slightly weaker (threefold) for the single and double mutant proteins. Therefore, the experiments were conducted under similar experimental conditions as for wild-type enzyme [12]. All thermodynamic parameters derived from our experiments are summarized in Table 2. As shown in Fig. 4B, binding of UDPNAG to the K22V mutant protein is exothermic, and DH, DS and DC p are the same (within error) as the values reported previously for wild-type enzyme [13]. On the other hand, UDPNAG binding to the K22V/R120K double mutant protein (Fig. 4A) is less exothermic in the entire temperature interval and DC p ¼ 1.38 kJÆmol )1 ÆK )1 is sig- nificantly smaller than DC p measured for both the wild-type MurA and the K22V variant ()1.9 kJÆmol )1 ÆK )1 and )2.0 kJÆmol )1 ÆK )1 , respectively). The unfavorable entropic contribution is also reduced. As the strength of UDPNAG binding is not much affected by either mutation, the thermodynamic profiles indicate that replacement of K22 by valine does not interfere with the conformational change induced by UDPNAG binding, whereas the additional replacement of R120 by lysine reduces the probability of forming the fully closed form of the protein. The results obtained with the K22V/R120K double mutant protein point toward a central role of arginine 120 in the conformational process occurring during catalysis. Therefore, we have generated the R120K single mutant protein to define its importance in catalysis and the conformational change. Similarly, the heat capacity change observed for UDPNAG binding is the same as for the K22V/R120K double mutant protein, supporting a model in which R120 plays the crucial role in the open-closed transformation in MurA. Discussion MurA undergoes a pronounced, conformational change upon ligand binding. The loop region of the upper domain moves towards the active site of the enzyme, thus shielding the substrates (and ligands) from bulk solvent. A detailed thermodynamic study of this process has indicated that the shift of the equilibrium towards the closed conformation is induced to a large extent by UDPNAG [13]. Here we have shown that the thermodynamic characteristics of this process are identical in the K22V mutant protein, indicating that the mode and extent of the conformational change is very similar to wild-type MurA. This finding is in clear contrast to an earlier hypothesis that K22 plays a key role in Table 2. Thermodynamic parameters for the binding of UDPNAG to the K22V, R120K and K22V/R120K mutant proteins in comparison to wild-type MurA. Values at 20 °C were chosen as example. All measurements were performed in 50 mm Hepes, pH 7.4. The calculated values for DH, DG, TDS and K d are from duplicate experiments. DC p was calculated from the slope of the regression of DH vs. T. All values except K d and DC p are in kJÆmol )1 . K d is in l M and DC p is in kJÆmol )1 ÆK )1 . DG was calculated from DG ¼ RT ln K d . The errors are estimated to be ± 1% for DG,±3% for DH,±9%forDS and 15% for DC p . The calculated and experimental DC p for wild-type MurA was taken from a previous report [13]. wt MurA K22V MurA R120K MurA K22V/R120K MurA DH )46.8 )44.0 )45.9 )33.8 DG )25.5 )23.0 )23.0 )22.8 TDS )21.4 )21.0 )22.9 )11.0 K d (l M ) 28.4 80.5 80.0 87.6 DC p )1.9 )2.0 )1.4 )1.38 Ó FEBS 2004 Lysine 22 and arginine 120 in MurA (Eur. J. Biochem. 271) 2687 the conformational change, possibly as part of a molecular switch mechanism [12]. On the other hand, analysis of tryptic fragments obtained after incubation with fosfomycin alone and in combination with UDPNAG have provided further insight into the molecular mechanism driving the formation of the covalent C115–fosfomycin adduct. A conservative exchange of K22 to arginine maintains fosfo- mycin binding, while a charge reversal in the K22E mutant enzyme completely abolishes the ability to form the covalent adduct (Table 1). Inspection of the three-dimensional structure of MurA complexed with fosfomycin and UDPNAG provides a rationale for these findings: the side chain amino group of K22 engages in a salt-bridge interaction with the phospho- nate group of fosfomycin (Fig. 5). Clearly, the guanidinium group of arginine can, at least in part, fulfil this function while the negatively charged glutamate side chain weakens or even prohibits fosfomycin binding due to charge–charge repulsions. Most interestingly, the uncharged valine side chain completely abolishes fosfomycin binding in the absence of UDPNAG. Clearly, UDPNAG binding creates additional favorable interactions that promote fosfomycin binding to the enzyme. As shown in Fig. 5, UDPNAG may exhibit a direct and/or an indirect effect on the binding environment of fosfomycin. The direct influence comes from two hydrogen bonds formed by the nitrogen of the N-acetyl group and the 3¢-oxygen to the phosphonate and hydroxyl group of fosfomycin, respectively (Fig. 5). An indirect effect of UDPNAG may be exerted via a salt-bridge of its diphosphate group to the distal guanidinium nitrogen atom of the invariant R120 (Fig. 5). R120 is located in the flexible loop and hence does not interact with fosfomycin in the open form of the protein. However, as the closed conformation of the protein is stabilized by UDPNAG, R120 is recruited as a binding partner for fosfomycin. Thus the bidentate character of the H-bonding network involving R120 appears to be critical for formation of the fosfomycin– MurA covalent adduct. The critical role of R120 in generating productive interactions in the active site, is emphasized by the 2000-fold decrease in catalytic activity (Table 1) that is greater than the loss of activity found with all other single mutant proteins characterized so far [12,18]. We envisage the following putative mechanism: K22 provides an essential binding site by positioning and/or stabilizing the fosfomycin phosphonate group via an Fig. 4. Temperature dependence of DH (circles, solid lines), DG (squares, dotted line) and –TDS (triangles, dashed lines) for UDPNAG binding to the K22VR120K (A), K22V (B) and R120K mutant proteins (C). Thermodynamic data for wild-type MurA is included for com- parison and is represented by filled symbols in (A)–(C). Open symbols are parameters measured for the corresponding mutant. In a typical experiment 200–420 l M protein was titrated with 5 m M UDPNAG. DC p was determined from the slope of the regression line describing the temperature dependence of DH. Fig. 5. Schematic respresentation of the active site of MurA with UDPNAG and fosfomycin bound. The dashed lines indicate hydrogen bond interactions and the numbers give the distances in A ˚ (based on the structure reported in [11]). 2688 A. M. Thomas et al. (Eur. J. Biochem. 271) Ó FEBS 2004 H-bond. If the N e atom of K22 is missing (as in the K22V mutant), the fosfomycin binding/reaction with C115 occurs only in the presence of UDPNAG as the substrate promotes the interaction of R120 and fosfomycin in order to form a complex optimal for the nucleophilic attack of the thiol group. This complex greatly resembles the closed confor- mation. The observed lack of fosfomycin binding and covalent attachment to the K22V/R120K double mutant protein supports this scenario. Lysine in position 120 cannot form simultaneously the H-bonds depicted in Fig. 5. If fosfomycin is not initially oriented/bound by K22, a lysine in position 120 possibly makes (if at all) only a single H-bond to the N-acetyl group of UDPNAG. As the lysine side chain is somewhat shorter than the arginine side chain, it is no longer positioned in an optimal way to form potential H-bonds with fosfomycin. As a consequence, the required H-bonding partners of fosfomycin are not present in the K22V/R120K MurA variant and the covalent adduct is no longer formed. At the same time, formation of the intricate H-bond pattern involving two H-bond donors for bidentate interaction (Fig. 5) may be a stringent prerequisite for the complete closure of the lid. The thermodynamic profile of UDPNAG binding to the MurA variants studied here supports this model. UDPNAG binds with very similar affinities to wild-type MurA and its K22V, R120K and K22V/R120K variants. The free energy difference, DDG ¼ DG mut – DG wt , is only 2–3 kJÆmol )1 ,and DG R120K and DG K22V equal DG K22V/R120K within error. Moreover, we have demonstrated that the presence of fosfomycin is also not critical to UDPNAG binding [13]. It follows that UDPNAG binds in a preformed pocket and does not require significant interactions with either K22 or with fosfomycin, nor with the guanidinium group of R120. Also, the heat capacity decrement associated with UDP- NAG binding is not related to interactions involving K22. However, the heat capacity decrement is dependent on the presence of a short H-bond with R120, as it is significantly lower in the K22V/R120K and R120K mutant proteins. Replacement of R120 by lysine causes a 25% reduction of DC p for both the single R120K and the K22V/R120K double mutant protein indicating that the lysine plays a pivotal role in formation of the closed conformation, i.e. the closure of the lid. This is supported by the decrease in enthalpy and the reduction of unfavorable entropy in the double mutant protein that is compatible with a less ordered structure of the lid making fewer contacts with the body of the protein. Alternatively, the observed changes in the thermodynamic parameters for the K22V/R120K and R120K mutant proteins might be caused by differences in the thermal/vibrational content of the complex. As the structure of the binary complex of MurA with bound UDPNAG is not available, it is not possible to calculate how much of the heat capacity change is due to hydration effects. Also, the lack of a binary structure does not allow us to evaluate the extent to which the conformational change is impeded by the mutations. However, our results clearly demonstrate that (a) the interaction between UDPNAG and R120 is the critical driving force triggering the conformational transition from the open to the closed state, and (b) the lid closes completely only if all interactions between properly positioned UDPNAG, fosfomycin, K22 and R120 take place. The structurally and functionally related enzyme 5-enolpyruvylshikimate 3-phosphate synthase has an invari- ant arginine residue in a corresponding position (R124). The three-dimensional structure of this enzyme in complex with the reversible inhibitor glyphosate also demonstrates that R124 forms a salt-bridge to the phosphonate group of glyphosate [20]. In contrast to MurA, however, 5-enolpyr- uvylshikimate 3-phosphate synthase lacks the flexible loop including the cysteine residue and, moreover, the enolpyr- uvyl-accepting substrate, shikimate-3-phosphate, does not interact with R124 in the ternary complex [20]. Hence, the mechanism of binding of fosfomycin by reinforcement of an initial binding site (mainly K22 and R397) through recruit- ment of secondary binding partners located in a flexible loop (R120) cannot be envisaged for 5-enolpyruvylshiki- mate 3-phosphate synthase. Despite phylogenetic, structural and mechanistic similarities between 5-enolpyruvylshiki- mate 3-phosphate synthase and MurA, it appears that these two enzymes have developed different strategies to bind the substrate and to shield the active site against bulk solvent during the catalytic process. As UDPNAG binding to the K22V mutant protein overcomes the constraints on fosfomycin adduct formation, then why is this mutant protein catalytically inactive [12]? Although phosphoenolpyruvate was shown to be capable of reacting with the thiol group of C115 in a fashion similar to fosfomycin [19,21], this adduct appears to be an off- pathway species that releases phosphoenolpyruvate in the active site of MurA, which then reacts with the 3¢-hydroxyl group of bound UDPNAG to form a O-phosphothioketal [22,23]. This species then eliminates phosphate to yield the product UDP-N-acetylenolpyruvylglucosamine. The stereo- chemical course of the reaction (anti-addition, syn-elimin- ation [10]) also dictates that the reacting molecules are properly positioned by neighboring amino acid residues. Hence, productive catalysis is subject to various chemical and steric constraints in contrast to the comparably simple adduct formation with fosfomycin. It follows that the role of K22 in catalysis is to provide a binding partner for the phosphate group of phosphoenolpyruvate as well as to achieve proper alignment of the reactants during the critical addition–elimination steps. Acknowledgements This work was supported by the ETH through an internal research grant to P. M. and N. A. (0-20-515-98). We would also like to thank A. K. Samland for many stimulating discussions and for providing the K22V, K22E and K22R mutant plasmids. We are also grateful to T. Etezady-Esfarjani for his help in preparing Fig. 1. References 1. El Zoeiby, A., Sanschagrin, F. & Levesque, R.C. (2003) Structure and function of the Mur enzymes: development of novel inhibitors. Mol. Microbiol. 47, 1–12. 2. Bugg, T.D.H. & Walsh, C.T. (1992) Intracellular steps of bacterial cell wall peptidoglycan biosynthesis: Enzymology, antibiotics, and antibiotic resistance. Natural Product Reports. 9, 199–215. 3. Hendlin, D., Stapley, E.O., Jackson, M., Wallick, H., Miller, A.K., Wolf, F.J., Miller, T.W., Chaiet, L., Kahan, F.M., Foltz, E.L., Woodruff, H.B., Mata, J.M., Hernandez, S. & Mochales, S. Ó FEBS 2004 Lysine 22 and arginine 120 in MurA (Eur. J. Biochem. 271) 2689 (1969) Phosphonomycin, a new antibiotic produced by strains of streptomyces. Science 166, 122–123. 4. Katayama, N., Tsubotani, S., Nozaki, Y., Harada, S. & Ono, H. (1990) Fosfadecin and fosfocytocin, new nucleotide antibiotics produced by bacteria. J. Antibiotics. 43, 238–246. 5. Shoji,J.,Kato,T.,Hinoo,H.,Hattori,T.,Hirooka,K.,Matsu- moto, K., Tanimoto, T. & Kondo, E. (1986) Production of fosfomycin (phosphonomycin) by Pseudomonas syringae. J. Antibiotics. 39, 1011–1012. 6. Kahan, F.M., Kahan, J.S., Cassidy, P.J. & Kropp, H. (1974) The mechanism of action of fosfomycin (phosphonomycin). Ann. NY Acad. Sci. 235, 364–385. 7. Marquardt, J.L., Brown, E.D., Lane, W.S., Haley, T.M., Ich- ikawa, Y., Wong, C H. & Walsh, C.T. (1994) Kinetics, stoichio- metry, and identification of the reactive thiolate in the inactivation of UDP-GlcNAc enolpyruvoyl transferase by the antibiotic fosfomycin. Biochemistry 33, 10646–10651. 8. Krekel, F., Samland, A.K., Macheroux, P., Amrhein, N. & Evans, J.N.S. (2000) Determination of the pK a value of C115 in MurA (UDP-N-acetylglucosamine enolpyruvyltransferase) from Enterobacter cloacae. Biochemistry 39, 12671–12677. 9. Scho ¨ nbrunn, E., Sack, S., Eschenburg, S., Perrakis, A., Krekel, F., Amrhein, N. & Mandelkow, E. (1996) Crystal structure of UDP- N-acetylglucosamine enolpyruvyltransferase, the target of the antibiotic fosfomycin. Structure 4, 1065–1075. 10. Skarzynski, T., Kim, D.H., Lees, W.J., Walsh, C.T. & Duncan, K. (1998) Stereochemical course of enzymatic enolpyruvyl transfer and catalytic conformation of the active site revealed by the crystal structure of the fluorinated analogue of the reaction tetrahedral intermediate bound to the active site of the C115A mutant of MurA. Biochemistry. 37, 2572–2577. 11. Skarzynski, T., Mistry, A., Wonacott, A., Hutchinson, S.E., Kelly, V.A. & Duncan, K. (1996) Structure of UDP-N-acetylgluco- samine enolpyruvyl transferase, an enzyme essential for the synthesis of bacterial peptidoglycan, complexed with substrate UDP-N-acetylgucosamine and the drug fosfomycin. Structure 4, 1465–1474. 12. Samland, A.K., Amrhein, N. & Macheroux, P. (1999) Lysine 22 in UDP-N-acetylglucosamine enolpyruvyl transferase from Enterobacter cloacae is crucial for enzymatic activity and the formation of covalent adducts with the substrate phosphoenol- pyruvate and the antibiotic fosfomycin. Biochemistry 38, 13162–13169. 13. Samland, A.K., Jelesarov, I., Kuhn, R., Amrhein, N. & Macher- oux, P. (2001) Thermodynamic characterization of ligand- induced conformational changes in UDP-N-acetylglucosamine enolpyruvyl transferase. Biochemistry 40, 9950–9956. 14. Scho ¨ nbrunn, E., Svergun, D.I., Amrhein, N. & Koch, M.H.J. (1998) Studies on the conformational changes in the bacter- ialcellwallbiosyntheticenzymeUDP-N-acetylglucosamine enolpyruvyltransferase (MurA). Eur. J. Biochem. 253, 406–412. 15. Krekel, F., Oecking, C., Amrhein, N. & Macheroux, P. (1999) Substrate and inhibitor-induced conformational changes in the structurally related enzymes UDP-N-acetylglucosamine enolpyruvyl transferase (MurA) and 5-enolpyruvylshikimate 3-phosphate synthase (EPSPS). Biochemistry 38, 8864–8878. 16. Spolar, R.S. & Record, M.T. Jr (1994) Coupling of local folding to site-specific binding of proteins to DNA. Science 263, 777– 784. 17. Baker, B.M. & Murphy, K.P. (1998) Prediction of binding energetics from structure using empirical parameterization. Methods Enzymol. 295, 294–315. 18. Samland, A.K., Etezady-Esfarjani, T., Amrhein, N. & Macher- oux, P. (2001) Asparagine 23 and aspartate 305 are essential residues in the active site of UDP-N-acetylglucosamine enolpyruvyl transferase from Enterobacter cloacae. Biochemistry 40, 1550–1559. 19. Wanke, C. & Amrhein, N. (1993) Evidence that the reaction of the UDP-N-acetylglucosamine 1-carboxyvinyltransferase proceeds through the O-phosphothioketal of pyruvic acid bound to Cys115 of the enzyme. Eur. J. Biochem. 218, 861–870. 20. Scho ¨ nbrunn, E., Eschenburg, S., Shuttleworth, W.A., Schloss, J.V., Amrhein, N., Evans, J.N.S. & Kabsch, W. (2001) Interaction of the herbicide glyphosate with its target enzyme 5-en- olpyruvylshikimate 3-phosphate synthase in atomic detail. Proc. NatlAcad.Sci.USA98, 1376–1380. 21. Brown, E.D., Marquardt, J.L., Lee, J.P., Walsh, C.T. & Anderson, K.S. (1994) Detection and characterization of a phospholactoyl-enzyme adduct in the reaction catalyzed by UDP- N-acetylglucosamine enolpyruvoyl transferase, MurZ. Biochem- istry. 33, 10638–10645. 22. Marquardt, J.L., Brown, E.D., Walsh, C.T. & Anderson, K.S. (1993) Isolation and structural elucidation of a tetrahedral intermediate in the UDP-N-acetylglucosamine enolpyruvoyl transferase enzymatic pathway. J. Am. Chem. Soc. 115, 10398– 10399. 23. Ramilo,C.,Appleyard,R.J.,Wanke,C.,Krekel,F.,Amrhein,N. & Evans, J.N.S. (1994) Detection of the covalent intermediate of UDP-N-acetylglucosamine enolpyruvyl transferase by solution- state and time-resolved solid-state NMR spectroscopy. Biochem- istry 33, 15071–15079. 24. Koradi, R., Billeter, M. & Wu ¨ thrich, K. (1996) MOLMOL: a program for display and analysis of macromolecular structures. J. Mol. Graph. 14, 29–32. 2690 A. M. Thomas et al. (Eur. J. Biochem. 271) Ó FEBS 2004 . Role of K22 and R120 in the covalent binding of the antibiotic fosfomycin and the substrate-induced conformational change in UDP- N -acetylglucosamine enol pyruvyl. with the K22V and K22E mutant proteins. This was interpreted as a lack of fosfomycin binding to these mutant proteins [12]. Binding of fosfomycin to either