Crystal Structure of the N-Acetylmuramic Acid α-1-Phosphate (MurNAc-α1-P) Uridylyltransferase MurU, a Minimal Sugar Nucleotidyltransferase and Potential Drug Target Enzyme in Gram-negative Pathogens

Background: N-Acetylmuramic acid α-1-phosphate (MurNAc-α1-P) uridylyltransferase MurU contributes to a shortcut in Pseudomonas putida cell wall biosynthesis accounting for intrinsic fosfomycin resistance. Results: This work provides the first structural analysis of a MurNAc-α1-P uridylyltransferase. Conclusion: MurNAc-α1-P and UTP are coordinated at distinct sites in the active center. Significance: The presented structure represents a “minimal nucleotidyltransferase domain” and may provide structural guidelines for inhibitor design. The N-acetylmuramic acid α-1-phosphate (MurNAc-α1-P) uridylyltransferase MurU catalyzes the synthesis of uridine diphosphate (UDP)-MurNAc, a crucial precursor of the bacterial peptidoglycan cell wall. MurU is part of a recently identified cell wall recycling pathway in Gram-negative bacteria that bypasses the general de novo biosynthesis of UDP-MurNAc and contributes to high intrinsic resistance to the antibiotic fosfomycin, which targets UDP-MurNAc de novo biosynthesis. To provide insights into substrate binding and specificity, we solved crystal structures of MurU of Pseudomonas putida in native and ligand-bound states at high resolution. With the help of these structures, critical enzyme-substrate interactions were identified that enable tight binding of MurNAc-α1-P to the active site of MurU. The MurU structures define a “minimal domain” required for general nucleotidyltransferase activity. They furthermore provide a structural basis for the chemical design of inhibitors of MurU that could serve as novel drugs in combination therapy against multidrug-resistant Gram-negative pathogens.

The bacterial peptidoglycan is a large, meshlike macromolecule consisting of polysaccharide chains of alternating N-acetylglucosamine (GlcNAc) and N-acetylmuramic acid (MurNAc) 4 links that are connected via short oligopeptide bridges (1). The biosynthesis of peptidoglycan is highly con-served and essential for bacteria, and enzymes within this synthetic pathway therefore represent attractive targets for antibacterial drug design. The first committed precursor molecule in this pathway is UDP-MurNAc, which is synthesized in the cytosol from phosphoenolpyruvate and UDP-GlcNAc in two steps that are catalyzed by the enzymes MurA and MurB. The naturally produced phosphoenolpyruvate analog fosfomycin ((1R,2S)-1,2-epoxypropylphosphonic acid) irreversibly inhibits MurA, resulting in a reduced pool of UDP-MurNAc, decelerated biosynthesis of peptidoglycan, and eventually cell lysis (2). Many bacteria, particularly Gram-negative organisms such as Pseudomonas aeruginosa, are characterized by a high intrinsic resistance to fosfomycin, which limits its use for the treatment of infections (3). In the bacterial life cycle, the peptidoglycan polymer is constantly degraded and resynthesized, and bacteria are able to reuse a large proportion of their peptidoglycan in a process called cell wall recycling (4 -6). We have recently discovered a connection between cell wall recycling and intrinsic fosfomycin resistance (7,8).
In Escherichia coli, peptidoglycan catabolites are imported into the cytoplasm (6,9) where they are further degraded by the etherase MurQ, a distinctive enzyme of the E. coli peptidoglycan recycling pathway with orthologs in many other organisms (10,11). MurQ cleaves the D-lactic acid substituent of 6-OH phosphorylated MurNAc, thereby merging the catabolism of the two cell wall sugars at the stage of the common metabolite GlcNAc-6-P that is subsequently converted to glucosamine-6-P by the deacetylase NagA (6,10,12). Recent studies revealed that the occurrence of cell wall recycling enzymes differs among bacteria. Specifically, all pseudomonades and many other Gram-negative bacteria, including important pathogens, lack an ortholog of MurQ. These bacteria instead possess a shortcut from MurNAc-6-P to UDP-MurNAc that has been described recently for Pseudomonas putida KT2440 and P. aeruginosa PA01/PA14 (7,8). This pathway bypasses the de novo biosynthesis of UDP-MurNAc and thus endows the bacterium with an intrinsic resistance against the antibiotic fosfomycin. Three novel, sequentially acting enzymes were identified to account for this shortcut: a MurNAc-6-P phosphatase, the kinase AmgK that rephosphorylates the resulting MurNAc at its C 1 hydroxyl group, and the nucleotidyltransferase MurU that generates UDP-MurNAc from UTP and MurNAc-␣1-P.
MurU belongs to the enzyme family of sugar nucleotidyltransferases (SNTs; also referred to as sugar nucleotide pyrophosphorylases, EC 2.7.7.x) whose members share a similar domain organization but do not necessarily display high sequence identity (13). MurU orthologs from Pseudomonas species, however, display sequence identities of over 70%, arguing for a general relevance of the insights gained from MurU of P. putida for these organisms. Moreover, previous studies revealed functional orthologs of P. putida MurU in P. aeruginosa PAO1 (70.8% amino acid sequence identity), Neisseria meningitidis MC58 (56.2% identity), and in Caulobacter crescentus CB15 (34.3% identity) (8).
Mutations or deletions of amgK and/or murU rendered the Pseudomonas strains more sensitive to fosfomycin (7,8), which brings the shortcut from MurNAc-6-P to UDP-MurNAc into a pharmacological focus of interest. Accounting for the recent rise of multidrug-resistant bacteria, fosfomycin, because of its low toxicity and low cross-resistance with other antibiotics, has regained significant therapeutic relevance especially in combination therapy with other antibiotics (14). The structural analysis of MurU is of particular interest in this context because this enzyme is highly selective for MurNAc-␣1-P (8), and its crystal structure might thus help to elucidate some key features for this selectivity and possibly provide a rationale for designing inhibitory drugs specifically targeting MurU and the above described shortcut in cell wall recycling.
Although several SNTs have been structurally characterized, there is no structure available of an SNT that specifically accepts MurNAc-␣1-P as its sugar substrate (13). To advance the understanding of how SNTs use MurNAc-␣1-P, we have solved crystal structures of native MurU and MurU enzymesubstrate complexes. Comparisons of these structures with the structures of related SNTs provide insights into the narrow substrate specificity of the enzyme, reveal mechanistic details, and establish a platform for the design of inhibitors.

EXPERIMENTAL PROCEDURES
Cloning of murU, Protein Expression, and Purification-murU of P. putida was cloned in vector pET29b (Novagen), heterologously overexpressed in E. coli as a C-terminally His 6tagged MurU fusion protein, and purified by nickel affinity and gel filtration chromatography as described previously (8). The purified protein (yield ϳ20 mg/liter of culture) was concentrated to 5-10 mg/ml and could be stored for several weeks at Ϫ20°C in gel filtration buffer (30 mM Tris-HCl, pH 7.6, 50 mM NaCl) without significant loss of enzyme activity.
HPLC Analysis of Purified MurU-High performance liquid chromatography (HPLC; Dionex BioLC) with a size exclusion column (BioSep-SEC-S3000, 600 ϫ 7, 800 mm, 5-m particle size; Phenomenex, Aschaffenburg) was used at a flow rate of 1 ml/min (buffer, 20 mM Na 2 HPO 4 , 500 mM NaCl, pH 7.4) to assess the association state of native MurU. Standard proteins used for gel filtration separation by HPLC were albumin (molecular mass, 66 kDa), chymotrypsinogen (25.6 kDa), and ribonuclease (13.7 kDa). The theoretical molecular mass of the His 6 -tagged version of MurU is 24,879 Da. Analytical gel filtration of MurU by HPLC revealed an apparent molecular weight of a monomer, indicating that the physiologically active moiety of MurU is the monomer (see Fig. 1A).
Enzymatic Preparation of MurNAc-␣1-P-The enzymatic preparation of the MurU substrate MurNAc-␣1-P was conducted as described (8), and the concentration of MurNAc-␣1-P in preparations was determined by assaying the phosphate release with malachite green upon treatment with alkaline phosphatase according to previously established protocols (15).
Enzyme Activity Assay-The malachite green assay also served to evaluate the enzymatic activity of purified MurU. No MgCl 2 had to be added to yield active enzyme; however, addition of EDTA (1 mM) led to complete loss of MurU activity, indicating the requirement of divalent cations, e.g. Mg 2ϩ , which has been shown previously for other nucleotidyltransferases (13).
Substrate Specificity Assays-The nucleotidyl triphosphate specificity of MurU was analyzed using 5 l of radiolabeled AmgK reaction product (MurNAc ␣-1-[ 32 P]phosphate) in Tris-HCl buffer, pH 7.6 containing 0.25 unit of bakers' yeast inorganic pyrophosphate (Sigma-Aldrich) to drive the transferase reaction toward completion and a 50 mM concentration of either UTP, ATP, CTP, GTP, or TTP nucleotidyl triphosphate (Sigma-Aldrich) in 25-l reaction volumes at 37°C. The nucleotidyl transfer reactions were started in parallel by the addition of 1 g of purified MurU, and 5 l of the reaction mixtures were spotted immediately (0 min) and after 180 min of incubation on a TLC plate. Reaction products were separated in basic solvent with n-butyl alcohol, methanol, 25% (w/v) ammonium hydroxide, water (5:4:2:1). The radioactive products were detected using a Typhoon Trioϩ biomolecular imager (GE Healthcare). The 32 P-radiolabled substrate MurNAc-␣1-P was prepared by adding 50 mM MurNAc to a reaction mixture containing 100 mM Tris-HCl, pH 7.6, 100 mM ATP, pH 7.6, 10 mM MgCl 2 , 140 kBq of [␥-32 P]ATP in a total volume of 100 l. The reaction was started by the addition of 25 g of purified AmgK and incubated for 3 h at 25°C. MurNAc-␣1-P but not GlcNAc-␣1-P or Glc-␣1-P was the only accepted sugar phosphate substrate that yielded UDP-MurNAc along with UTP (8). The presence of 0, 0.5, 1, 5, and 10 mM GlcNAc-␣1-P in the same setup did not have an effect on product formation. MurU also displays narrow specificity for UTP, and no other tested nucleotide triphosphate (ATP, CTP, TTP, or GTP) served as substrate (Fig. 1B).
Protein Crystallization-MurU crystallized in space group P 6 1 22 with one monomer in the asymmetric unit and a solvent content of the crystals of around 50% (16). Crystals were grown using the sitting drop vapor diffusion method at 20°C. In 96-well crystallization plates, 300 nl of protein solution (7.5-10.5 mg/ml) were mixed with 300 nl of a well solution that contained 0.05-0.1 M buffer (MES, pH 6 or 6.5; HEPES, pH 7; or Tris, pH 8.5) and 1.0 -1.5 M (NH 4 ) 2 SO 4 .
X-ray Diffraction-All data were collected at 100 K on PILATUS detectors at the synchrotron radiation beamline ID14-4 of the European Synchrotron Radiation Facility in Grenoble, France as well as at beamlines X06SA and X06DA of the Swiss Light Source in Villigen, Switzerland. Diffraction patterns of the non-liganded enzyme as well as of the enzymeligand complexes were generated using an x-ray wavelength of 1.0 Å, whereas diffraction data of the NH 4 I-soaked crystals for experimental phasing were obtained at a wavelength of 1.7 Å.
Structure Determination and Model Building-Indexing, integrating, and scaling of diffraction data were done with the XDS software package (17). Experimental phases for the nonliganded enzyme data were determined in a single anomalous dispersion experiment (18,19), exploiting the anomalous signal of iodine in NH 4 I-soaked crystals at a wavelength of 1.7 Å. Initial phases were derived by determining the substructure of four iodine atoms per asymmetric unit with the help of the program suite SHELXC/D/E (20) and further extended and refined with SHARP/autoSHARP (21) and PHENIX (22). Molecular replacement for all ligand-bound structures was done with Phaser (23) using the structure of the native enzyme as a search model. Model building, refinement, and validation was performed with Coot (24), PHENIX, the CCP4 suite (25), the Protein Data Bank data validation and deposition portal, and the MolProbity web page (26). To remove model bias, simulated annealing was performed for all structures, and simulated annealing omit density maps were calculated for the ligand-bound structures using phenix.refine. The TLSMD (Translation Libration Screw Motion Determination) web server (27) was used for TLS parameterization of the protein.
Coordinate and parameter files for MurNAc-␣1-P, UpNHpp, and UppNHp were generated using phenix.elbow. Ligands were built manually according to the difference and simulated annealing omit density maps using Coot, and their coordinates were further refined with PHENIX.
The final structures contain all residues with the exception of some poorly ordered segments between Gly 134 and His 136 and around Ala 154 in the native and complex 1 structures (Table 1) as well as the C-terminal His 6 tag (in all structures). Structure figures were generated with PyMOL (28), and electrostatic potentials were calculated with PyMOL-implementations (29). For structural comparison, structural models were superimposed using the alignment algorithm implemented in PyMOL with complex 1 as the reference structure. For the native structure 177 atoms and for complex 2 175 atoms were aligned with a root mean square deviation of 0.179 and 0.144 Å, respectively.

Overall Structure and Substrate-binding Region of MurU-
The structure of MurU was determined at 1.8-Å resolution using single anomalous dispersion ( Table 1). The polypeptide chain folds into a three-dimensional structure similar to those observed in other SNTs with a large N-terminal "core" domain that contains the catalytically active center, a dinucleotidebinding region showing a Rossmann fold, and a C-terminal auxiliary domain (13). The core domain of MurU adopts an ␣/␤sandwich fold with a central, mixed, seven-stranded ␤-sheet flanked on both sides by a total of eight ␣-helices (Fig. 1). The C-terminal auxiliary domain essentially consists of a single, long ␣-helix (␣9) that projects from the core domain.
MurNAc-␣1-P, but not the related GlcNAc-␣1-P or Glc-␣1-P compound, can function as sugar phosphate substrate for MurU, yielding UDP-MurNAc along with UTP (8). Moreover, the presence of GlcNAc-␣1-P did not have an inhibitory effect on product formation in an inhibition assay (data not shown). MurU also displays narrow specificity for UTP as no other tested nucleotide triphosphate (ATP, CTP, TTP, or GTP) is accepted as a substrate according to a 32 P radioassay (Fig. 1B). With the aim of understanding how MurU binds and interacts with its substrates and catalyzes the uridyl transfer reaction, we solved structures of MurU in complex with (i) MurNAc-␣1-P and the ␤,␥ non-hydrolyzable UTP analog UppNHp (complex 1 in Table 1) and with (ii) MurNAc-␣1-P, the ␣,␤ non-hydrolyzable UTP analog UpNHpp and 20 mM MgCl 2 (complex 2 in Table 1). The catalytically active site is located at a solventexposed cleft whose bottom is formed by one edge of the central ␤-sheet and at two of its sides by helices ␣5 and ␣8 as well as the N-terminal amino acids 1-13. This N-terminal section, which also comprises strand ␤1, is widely conserved among SNTs and carries the canonical signature motif GXGXR (Gly 9 -Arg 13 in MurU) that has been referred to previously (13). The third side of the substrate-binding cleft is flanked by a subdomain that is inserted between strands ␤5 and ␤6 of the central ␤-sheet of MurU. This subdomain consists of a three-stranded antiparallel ␤-sheet (strands ␤8 -␤10) that is connected to ␤5 by a highly flexible loop that was not fully visible in any of our crystal structures (Fig. 1). Functionally, this subdomain is involved in the coordination of MurNAc-␣1-P (or generally the sugar moiety in structurally related enzymes) and can be referred to as the sugar-binding domain (see also Ref. 30).
The MurNAc-␣1-P-binding Site of MurU-The data sets for complex 1 and complex 2 revealed very clear, unbiased difference electron density (mF o Ϫ DF c ) for MurNAc-␣1-P, and thus this ligand could be modeled with high confidence (Fig. 2). The MurNAc-␣1-P sugar ring assumes a 4 C 1 chair conformation in all our MurU structures. MurNAc-␣1-P forms hydrogen bonds to the side chain carboxylates of Asp 205 and Asp 140 as well as polar contacts to the side chain of Asn 105 ( Fig. 2A). Furthermore, there is a partial polar interaction to the backbone carbonyl of Phe 160 (for the sake of clarity, it is not explicitly shown in Fig. 2). The side chains of Phe 141 and Phe 160 provide a hydrophobic patch that accommodates the hydrophobic portion of the MurNAc-␣1-P lactyl moiety and the sugar ring. The side chains of Leu 181 and Leu 185 , which are also facing the lactyl-CH 3 , also contribute to this interaction. The carboxyl group of the lactyl moiety is accommodated in the vicinity of Lys 180 in a solvent-filled groove displaying a predominantly positive surface potential (Fig. 2C). The structures of the two complexes also contain a bound sulfate ion that was likely incorporated from the crystallization solution. The sulfate binds in the vicinity of the MurNAc-␣1-P lactyl group (Fig. 2). The sulfate ions have elevated temperature factors (B-factors) in all structures, indicating low occupancy and/or high mobility.
The UTP-binding Region of MurU-In contrast to MurNAc-␣1-P, the densities obtained for the uridyl substrates were much weaker in both complexes and did not allow for a confident placing of the uridyl ring. Nevertheless the "pNHp" portion of UppNHp ( Fig. 2A) or the pyrophosphate of UpNHpp (Fig. 2B), respectively, could be clearly localized. They are being coordinated by backbone nitrogens of residues 10 -12 of the nucleotidyltransferase signature motif as well as the side chain of Lys 23 (Fig. 2, A and B). The positive surface potential of this motif is significantly enhanced by the side chain of Arg 13 and helps compensate the negative charge of the pyrophosphate (Fig. 2C). The slightly different orientation that is adopted by the pNHp in comparison with the ␣-phosphate of UpNHpp can be attributed to the different geometrical environment of nitrogen compared with oxygen in the respective ligands.
As the uridyl moieties were not visible in either structure, we modeled their approximate location and orientation based on the visible portion of the nucleotides and geometrical restraints (Fig. 2, B and C, black tracings). Independently performed molecular dynamics simulations (data not shown) arrived at a conformation of the uridyl moiety that almost coincided with the orientations modeled, suggesting that the depicted conformation might be a good approximation of the structure that UTP actually adopts when bound to MurU. However, it would not be valid to deduce any specific interactions with the protein from these models.  APRIL 24, 2015 • VOLUME 290 • NUMBER 17

JOURNAL OF BIOLOGICAL CHEMISTRY 10807
In the non-liganded structure of MurU, a SO 4 2Ϫ is coordinated instead of the pyrophosphate at the respective site. However, in contrast to the native data set, the densities in the complexes are larger, and their elongated shape cannot be explained by a spherical sulfate ion, thus clearly arguing for the presence of the pyrophosphate moiety of the non-hydrolyzable substrate UpNHpp or the pyrophosphate mimic pNHp, respectively, being retained at the active site.
Two Magnesium Ions Are Likely Part of the Active Site-An EDTA-induced enzymatic inactivation indicated that MurU requires Mg 2ϩ or equivalent divalent cations as a cofactor to achieve its full enzymatic activity (8). We therefore also included 20 mM MgCl 2 in the soaking solution when solving the structure of complex 2. An unbiased difference electron density map (mF o Ϫ DF c ) revealed two spherical features near the ␣-phosphate of MurNAc-␣1-P and the pyrophosphate moiety of UpNHpp that could result either from water molecules or from an ion of similar size (Fig. 2, A and B). However, in MurU, MurNAc-␣1-P binding was obviously independent of the presence or absence of Mg 2ϩ in the soaking solution, indicating that the Mg 2ϩ is not required for substrate binding per se. To justify the modeling of Mg 2ϩ in our structure, we analyzed the coordination environment at the putative cation positions (which will be termed position A/ A Mg 2ϩ and position B/ B Mg 2ϩ hereafter). At position A, the side chains of Asp 205 and Asp 107 (Fig.  2B) as well as the backbone nitrogen of Gly 207 (not explicitly shown in Fig. 2B) would stabilize a putative A Mg 2ϩ . A Mg 2ϩ in position B in contrast would not be coordinated by any amino acids of MurU but to the phosphates of the uridyl substrate. A similar coordination of divalent cations in an SNT has been described previously (31). Moreover, we performed molecular dynamics simulations that also predicted the presence of Mg 2ϩ at the MurU active site. With the crystal structure as a starting model, the positions and radii of putative metal ions could be calculated and optimized by keeping only the protein atoms close to their crystallographic positions through the application of a high force field constant on the protein coordinates (data not shown). Although it was not suitable to distinguish among Mg 2ϩ , NH 4 ϩ , or Na ϩ , the ion positions predicted in this molecular dynamics approach were quite robust in different simulation scenarios and coincided nicely with the ion positions deduced from the x-ray data. Taken together, it seems reasonable and plausible to model Mg 2ϩ ions at the two catalytically relevant positions A and B.
Structural Changes upon Ligand Binding-The comparison of the complex 1 and 2 structures with the non-liganded enzyme reveals no large domain rearrangements upon ligand binding. In particular, there is no significant movement of the C-terminal helix relative to the N-terminal core domain, which is in contrast to what has been observed for the larger C-terminal domains of other SNTs (32). However, there are some distinct side-chain reorientations of residues involved in Mur-NAc-␣1-P coordination upon substrate binding (Fig. 3). The side chain of Asp 140 is rotated by almost 90°and reoriented toward the sugar substrate, thereby facilitating hydrogen bonding to C 4 -OH of MurNAc-␣1-P (Fig. 3). Interestingly, this side chain has rather low B-factors in both its substrate-bound and its non-liganded state, indicating that the described movement is a switch between two distinct states. Another, more subtle reorientation takes place at the conserved Arg 13 of the signature motif. Here, we observe a slight reorientation of the Arg 13 guanidinium group toward the pyrophosphate moiety of UpNHpp (Fig. 3) or UppNHp, respectively. Other residues involved in substrate binding, such as Asp 205 , Phe 106 , and  Table 1). Substrates are depicted as sticks and colored according to the atom type (oxygens in red, nitrogens in blue, carbons in yellow, and phosphorus in orange). The protein is shown in schematic representation with the colors highlighting the subdomains referred to in the text (blue, C-terminal helix; orange, sugar-binding domain; marine blue, nucleotidyltransferase signature motif). Secondary structure elements are numbered from the N to C terminus. Bottom, schematic representation of the reaction catalyzed by MurU. Ligands and side chains are colored by atom type (oxygens in red, nitrogens in blue, phosphorus in orange, sulfur in gold, magnesium in green, and carbons in the colors used for the domains in Fig. 1. C, surface representation of the substrate-binding site colored according to the surface potential (gradient from blue (positive) to red (negative)). Note that the parts of UpNHpp for which no reliable density could be obtained were modeled in the most likely possible conformation by relying on geometrical restraints and the clear electron density for the pyrophosphate part as an anchor point.
Asn 105 , remain essentially unchanged upon substrate binding (Fig. 3). It appears that MurU provides a preformed binding pocket for MurNAc-␣1-P that requires no large domain rearrangements for substrate binding.
MurU Defines a "Minimal Functional Domain" for Sugar Nucleotidyltransferase Activity-A Dali (33) search revealed significant structural similarities among MurU and several other SNTs. Of those, the glucose-1-phosphate thymidylyltransferase RmlA from P. aeruginosa (Protein Data Bank code 1G1L (30)), the mannose-1-phosphate guanylyltransferase GMPase of Thermotoga maritima (Protein Data Bank code 2X5Z (32)), and the N-terminal SNT domain of bifunctional GlcNAc-1-P uridylyltransferase GlmU from E. coli (Protein Data Bank code 1FWY (34)) were chosen for structural comparison with MurU. This selection was predominantly made to cover a range of substrates that, taken together, feature the specificity requirements of MurU as closely and all-encompassing as possible and thus provide a helpful guideline to integrate the structural analysis of MurU into a broader context of SNTs. The comparison of the overall structures of P. putida MurU (MurU P.p. ), P. aeruginosa RmlA (RmlA P.a. ), E. coli GlmU (Glm-U E.c. ), and T. maritima GMPase (GMPase T.m. ) clearly visualizes the common domain organization of SNTs (Fig. 4). All four enzyme folds can be subdivided into an N-terminal and a C-terminal domain with the N-terminal domain housing the sugar nucleotidyltransferase activity. The N-terminal domains of the enzymes are similar in structure as they all feature a Rossmann fold containing nucleotide binding motifs and a central sevenstranded ␤-sheet with an additional "sugar-binding" domain inserted between strands ␤5 and ␤6. The C-terminal domain of MurU, however, is strikingly different from that of RmlA P.a. , GlmU E.c. , and GMPase T.m. . Although in the latter three enzymes the C-terminal domain is rather large and projects quite prominently from the N-terminal nucleotidyltransferase domain, the corresponding region of MurU is small, essentially consisting of a single ␣-helix. The C-terminal domain is involved in the regulation of enzymatic activity in several SNTs, for example by mediating oligomerization and/or by binding allosteric inhibitors (13), and in some cases, it may even influence substrate specificity (32). By contrast, MurU is a monofunctional enzyme that, according to gel filtration profiles, is monomeric in solution (Fig. 1A).
The role of the truncated C-terminal domain in MurU is unclear. No significant reorientation of the C-terminal helix ␣9 is observed upon substrate binding, and no residues from ␣9 are involved in substrate binding or the coordination of the catalytically relevant Mg 2ϩ ions (Figs. 2 and 3). Although the C-terminal domain has some regulatory functions in related enzymes, MurU does not seem to require an "elaborate" C terminus for enzymatic activity. This defines the N-terminal core domain as a minimal functional domain required for MurNAc-␣1-P uridylyltransferase activity.

DISCUSSION
We have solved structures of MurU of P. putida in native and two ligand-bound states at high resolution. Taken together, these structures allow for the identification of critical structural differences to other, related SNTs and may help explaining the remarkably narrow specificity of MurU for MurNAc-␣1-P. They furthermore provide a structural basis for suggesting a catalytic mechanism.  MurU binds its substrate MurNAc-␣1-P through numerous polar and hydrophobic interactions, and specificity for the MurNAc group over related compounds such as GlcNAc or GalNAc is in part achieved through contacts that specifically target the unique lactyl group of the sugar. The observation that GlcNAc-␣1-P does not act as a competitive inhibitor for MurU moreover indicates that this compound does not bind to the enzyme in accordance with the specific contacts made by the substrate MurNAc-␣1-P that cannot be made by GlcNAc-␣1-P.
Although the location and orientation of the uridyl ring of the second substrate remain somewhat speculative as we do not observe density for it, the pyrophosphate portions of UpNHpp and the pNHp of UppNHp are clearly visible and allow placement of at least this part of the nucleotide with confidence. The pyrophosphate mimics one of the products of the uridyl transfer reaction and is tightly coordinated by residues of the SNT signature motif. Magnesium ions or other divalent cations have been found to be essential for or to enhance the enzymatic activities of other SNTs (5,13,31,35), and it is likely that MurU likewise relies on bound magnesium ions to perform the nucleotidyl transfer reaction. The narrow specificity for UTP does not necessarily imply a high binding affinity for the nucleotide. However, on the basis of our structural data, it would not be valid to draw conclusions on the UTP binding mode of MurU.
In comparison with the functionally related SNTs RmlA P.a. , GlmU E.c. , and GMPase T.m. (Fig. 4), the substrate-binding cleft of MurU is significantly more solvent-exposed. Regarding this fact and considering the distinct binding mode for MurNAc-␣1-P, we consider it likely that MurNAc-␣1-P binds first to the enzyme in a preformed binding pocket followed by Mg 2ϩ -coordinated UTP. After uridyl transfer, the remaining pyrophosphate together with the Mg 2ϩ would then be released immediately after the reaction followed by the product UDP-MurNAc.
Extensive studies of GlmU orthologs of various organisms have shown that the phosphoryl transfer reaction involves nucleophilic activation of the phosphoryl oxygen of the sugar substrate, which in turn attacks the ␣-phosphate of the NTP substrate, resulting in the formation of the NDP sugar product and a concomitant release of pyrophosphate. This has been proposed to occur in a S N 2 reaction for GlmU with the reaction proceeding through a pentavalent phosphorane involving two catalytically relevant Mg 2ϩ ions with distinct functions (31). One Mg 2ϩ ( A Mg 2ϩ ) enables nucleophile activation and substrate coordination and is coordinated by distinct active site residues, whereas a second Mg 2ϩ ( B Mg 2ϩ ) contributes to the stabilization of the transition state of the reaction. Because of its specific requirements for coordination geometry, B Mg 2ϩ promotes product formation by changing its coordination state and thus destabilizing the scissile bond (31). According to our structural data and the likely locations of the magnesium ions, a similar mechanism seems also plausible for MurU (Fig. 5). Thereby the extensive coordination of the leaving group of the UMP transfer would result in a stabilization of the reaction product, which in turn destabilizes the scissile bond between P ␣ and P ␤ of UTP and thus renders it susceptible to a nucleophilic attack by the phosphate group of MurNAc-␣1-P. Such an attack might well be facilitated by the A Mg 2ϩ ion in position A in analogy to what has been described for GlmU Mtb . However, referring to our structure, it is possible that residue Asp 205 not only coordinates the Mg 2ϩ ion but that this interaction moreover enables the Asp 205 carboxyl group to catalytically abstract a proton of the MurNAc-␣1-P phosphate group that in turn initiates the nucleophilic attack (Fig. 5).
Altering the nucleotide selectivity of SNTs through engineering of the substrate-binding site is a field of particular interest because with the biosynthesis of NDP sugars SNTs provide precursors for a large range of bioglycosylation pathways. Chemoenzymatic approaches for in vitro glycorandomization with a focus on engineering the sugar phosphate selectivity were introduced more than 10 years ago (36,37). In 2007, Moretti and Thorson (38) were able to also enhance the latent NTP substrate flexibility of the thymidylyltransferase RmlA, and in 2011, Moretti et al. (39) were able to expand its promiscuity via directed evolution based on x-ray structural data. We think that MurU forms a promising target for such an approach.
The recent rediscovery of fosfomycin in antibiotic combination therapy against multidrug-resistant bacteria has identified intrinsic fosfomycin resistance as a remarkable clinical problem (14). MurU significantly contributes to this resistance, thus making a specific inhibitor for this enzyme a desirable objective. Because MurU is highly specific for its sugar substrate (8), the MurNAc-␣1-P binding pocket might be a promising target for structure-guided drug design. Although this pocket is shallow and solvent-accessible, it still contains features that could be exploited, such as the hydrophobic pocket that accommodates the methyl group of the lactyl moiety of MurNAc-␣1-P or the charged surface that faces the carboxylate group of MurNAc-␣1-P. The design of an initial inhibitor should therefore mimic this lactyl pattern and perhaps involve optimization of contacts that favor its binding over the actual substrate.