Structure of the E. coli bifunctional GlmU acetyltransferase active site with substrates and products

Abstract

The biosynthesis of UDP-GlcNAc in bacteria is carried out by GlmU, an essential bifunctional uridyltransferase that catalyzes the CoA-dependent acetylation of GlcN-1-PO₄ to form GlcNAc-1-PO₄ and its subsequent condensation with UTP to form pyrophosphate and UDP-GlcNAc. As a metabolite, UDP-GlcNAc is situated at a branch point leading to the biosynthesis of lipopolysaccharide and peptidoglycan. Consequently, GlmU is regarded as an important target for potential antibacterial agents. The crystal structure of the Escherichia coli GlmU acetyltransferase active site has been determined in complexes with acetyl-CoA, CoA/GlcN-1-PO₄, and desulpho-CoA/GlcNAc-1-PO₄. These structures reveal the enzyme groups responsible for binding the substrates. A superposition of these complex structures suggests that the 2-amino group of GlcN-1-PO₄ is positioned in proximity to the acetyl-CoA to facilitate direct attack on its thioester by a ternary complex mechanism.

The bifunctional enzyme GlmU catalyzes the CoA-dependent acetylation of GlcN-1-PO₄ and the metal-dependent condensation of the resulting GlcNAc-1-PO₄ with UTP to form the activated nucleotide sugar UDP-GlcNAc (Mengin-Lecreulx and van Heijenoort 1994). UDP-GlcNAc occupies a position as a fork metabolite leading to the formation of peptidoglycan, lipopolysaccharide, and teichoic acid components of the bacterial cell wall (Anderson and Raetz 1987). GlmU is an essential enzyme in both Gram-positive and Gram-negative bacteria, and is viewed as an attractive target for the development of antimicrobial compounds.

The bacterial GlmU has been studied from Escherichia coli and Streptococcus pneumoniae (Mengin-Lecreulx and van Heijenoort 1993, 1994; Gehring et al. 1996; Pompeo et al. 1998). The acetyltransfer step is catalyzed by the C-domain of the protein, the N-terminal domain being the site of the subsequent uridyltransfer reaction. There is apparently no substrate channeling of GlcNAc-1-PO₄ between active sites (Gehring et al. 1996). The X-ray structures of GlmU in complex with substrates have been reported for the E. coli enzyme in complex with CoA and UDP-GlcNAc (Olsen and Roderick 2001) and in a truncated form as apoenzyme in complex with UDP-GlcNAc (Brown et al. 1999), and for Streptococcus pneumoniae GlmU as the full-length apoenzyme in complex with UDP-GlcNAc (Kostrewa et al. 2001), and as the full-length apoenzyme in complexes with acetyl-CoA and acetyl-CoA/UDP-GlcNAc (Sulzenbacher et al. 2001).

The presence of tandem-repeated copies of an imperfect six-residue theme of [LIV]-[GAED]-X-X-[STAV]-X termed a hexapeptide repeat or an isoleucine patch (Dicker and Seetharam 1992; Vaara 1992) encoding a left-handed parallel β-helix (LßH) domain (Raetz and Roderick 1995) places GlmU into the hexapeptide repeat superfamily of acyltransferases. These enzymes catalyze a variety of CoA- or acyl carrier protein-dependent acyltransfer reactions whereby acetyl, succinyl, or long-chain fatty acyl groups are transferred to primary hydroxyl or amines of amino acids, sugars, metabolic intermediates or xenobiotic compounds. The crystal structures of several hexapeptide acyltransferases have been determined, and all have been found to adopt the characteristic coiled LßH structural domain and either a trimeric or hexameric (dimer of trimers) oligomeric structure. In all cases, the acyltransferase active site is formed at the junction of two adjacent LßH domains, leading to an enzyme with three apparently independent acyltransferase sites per trimeric enzyme. GlmU is unusual, in that all three polypeptide chains are used to form each of its three acetyltransferase active sites (Olsen and Roderick 2001; Sulzenbacher et al. 2001), with the third subunit forming a limited set of interactions with the 3′–5′ ADP group of CoA.

Although GlmU has been crystallized in a variety of forms, no set of crystal structures is available for the acetyltransferase active site bound to its GlcN-1-PO₄ substrate or its GlcNAc-1-PO₄ product, leading to uncertainty concerning the enzyme groups that contact the substrates and the chemical mechanism of the acetyltransferase reaction. We report here the structures of E. coli GlmU in complex with acetyl-CoA, CoA/GlcN-1-PO₄, and desulpho-CoA/GlcNAc-1-PO₄ at the acetyltransferase active site. These structures identify the active site groups and their relationships to the substrates and suggest a ternary complex mechanism of action for the GlmU acetyltransferase reaction.

Results and Discussion

Structure of trimeric GlmU

The asymmetric unit of the crystals studied here contains two subunits. Each subunit is operated on by a crystallographic threefold axis to produce two distinct copies of the trimeric enzyme. One of these trimers binds UDP-GlcNAc and Co²⁺ at the pyrophosphorylase active site, while the other binds UDP-GlcNAc without a metal ion. The acetyltransferase active sites of these trimers is contained within the C-terminal β-helical domain (residues 260–437 of the 456 residue polypeptide) (Fig. 1A). In contrast to the pyrophosphorylase active sites, the two crystallographically distinct acetyltransferase active sites are structurally similar.

Figure 1.

E. coli GlmU in complex with substrates. (A) Native trimeric structure of GlmU. The three acetyltransferase active sites associated with the left-handed β-helical domain (shown in complex with CoA and GlcN-1-PO₄) are depicted at the bottom. The three pyrophosphorylase active sites (in complex with UDP-GlcNAc and Co²⁺) are at the top. For the proximal acetyltransferase active site, the A and B subunits are colored blue and green, respectively, with the insertion loop of subunit B in red. (B) Acetyltransferase active site viewed from the same direction as in A. The inserted loop of subunit B is shown (red). (C) Acetyltransferase active-site interactions with the pantetheinyl group of acetyl-CoA. Close hydrophilic interactions are indicated by dotted segments. The interactions formed to CoA are similar. (D) Acetyltransferase active-site interactions with GlcNAc-1-PO₄. The interactions formed to GlcN-1-PO₄ are similar. (E) Stereoview superposition of GlmU acetyltransferase active-site complexes; acetyl-CoA/SO₄ (blue bonds), CoA/GlcN-1-PO₄ (yellow), and desulpho-CoA/GlcNAc-1-PO₄ (green). The residue labeling refers to the A or B subunits of a single acetyltransferase active site. The superimposed coenzyme and phosphosugar molecules are labeled for acetyl-CoA and GlcNAc-1-PO₄, respectively. Residues 386B and 387B from the insertion loop are visible. Figure prepared with PyMOL (DeLano 2002).

Structure of GlmU substrate and product complexes

The structures of three GlmU complexes were solved: as bound to acetyl-CoA, GlcN-1-PO₄/CoA, and GlcNAc-1-PO₄/desulpho-CoA (Fig. 1A,B). All of these structures bind UDP-GlcNAc to one of the two pyrophosphorylase domains present in the asymmetric unit. Desulpho-CoA rather than CoA was included with GlcNAc-1-PO₄ in order to avoid the use of a catalytically active substrate pair. The acetyltransferase active site is formed at the junction of two adjacent LßH domains, but also makes limited use of the C-terminal tail of the third subunit of the trimeric enzyme to interact with the 3′–5′ ADP group of the cofactor. The two adjacent subunits that donate the majority of interactions to the substrates are termed A and B (Fig. 1A). In addition to residues from the coiled LßH domains, an insertion loop (residues 386B–393B) also donates residues to the acetyltransferase active site.

The structure of the E. coli GlmU acetyl-CoA complex is similar to that reported from S. pneumoniae (Sulzenbacher et al. 2001), with the acetyl oxygen of the cofactor accepting a hydrogen bond from the main-chain amide group of Ala 380A (Fig. 1C). The pantetheinyl arm of the cofactor forms hydrogen bonds to the main-chain amide groups of Ser 405A, Tyr 387B, and Ala 423A. The re face of the planar acetyl group faces toward a pocket that binds sulfate from the medium. This pocket had been previously identified as the likely binding location for GlcN-1-PO₄ on the basis of its proximity to the cofactor thioester group and to the positively charged side chains of S. pneumoniae Arg 332B, Lys 350B, and Lys 391B (Sulzenbacher et al. 2001). Electron density corresponding to the 3′–5′ ADP moiety of the cofactor is generally good, and its conformation and interactions with the enzyme are the same as those observed in the CoA complex (Olsen and Roderick 2001). The structure of E. coli GlmU in complex with CoA and GlcN-1-PO₄ indeed demonstrates that the 1-PO₄ group of the substrate binds in this pocket. The sugar 1-PO₄ of the substrate interacts with three cationic side chains (E. coli Arg 333B, Lys 351B, and Lys 392B) as well as the side chains of Asn 386B and Tyr 366B. The 3- and 4-hydroxyl groups form hydrogen bonds to Asn 377A, while the 2-amino group donates to His 363A NE2 (distance 3.4 Å) and to Asn 386B OD1 (3.3 Å). The hexose ring stacks with the phenolic ring of Tyr 387B. This residue and Asn 386B are included in a highly conserved Asn-Tyr-Asp-Gly sequence motif in the residue range 386–389 and project from LßH coil 8 of the B subunit as an insertion loop that contacts the sugar substrate. The structure of GlmU in complex with GlcNAc-1-PO₄ and desulpho-CoA is similar to the GlcN-1-PO₄/CoA structure (Fig. 1D). The acetyl group re face of the GlcNAc-1-PO₄ product faces toward the cofactor. The carbonyl oxygen of the GlcNAc-1-PO₄ acetyl group is not oriented to form a hydrogen bond to any enzyme group.

Superposition of the GlmU substrate complexes

A superposition of the acetyl-CoA/SO₄, CoA/GlcN-1-PO₄, and desulpho-CoA/GlcNAc-1-PO₄ structures using the C_α coordinates of the LßH domain reveals little conformational change of GlmU between these forms, in contrast to the conformational change that occurs on binding of the cofactor to the apoenzyme (Sulzenbacher et al. 2001; Fig. 1E). This superposition reveals the proximity of the 2-amino group of the GlcN-1-PO₄ substrate to the acetyl-CoA thioester carbonyl carbon (3.3 Å). For the reverse reaction direction (cofactor acetylation), the CoA thiol is positioned 2.4 Å from the acetyl carbonyl carbon of the GlcNAc-1-PO₄ product. The acetyl groups of acetyl-CoA and GlcNAc-1-PO₄ bind to similar locations on the enzyme—their superimposed carbonyl carbons are separated by only 1.6 Å, and only minimal movement would be required to overlay these acetyl groups from their pre- and post-catalytic binding positions (Fig. 1E).

The proximity of the 2-amino group of GlcN-1-PO₄ to the acetyl group of the acetyl-CoA suggests that a direct SN₂ attack on the acetyl-CoA thioester is possible. Presumably, His 363A would activate the 2-amino group by removing a proton either prior to or concomitant with the attack. This histidine residue pairs with Glu 349B and additionally stacks against the phenolic ring of Tyr 366B. These interactions would serve to promote the proper tautomeric form of His 363 lacking a proton on NE2 and to increase its basicity. Both His 363A and Tyr 366B adopt energetically unfavorable eclipsed side-chain conformations (χ₁ ≈ 0°) in order to achieve this ring stacking interaction. Whether the 1-PO₄ group of the substrate is involved in acid-base chemistry is unknown. After nucleophilic attack, the resulting tetrahedral oxyanion intermediate could be stabilized by the main-chain amide of Ala 380A, although the hydroxyl group of Ser 405A could also be positioned to interact with the oxyanion.

An alternative chemical mechanism for the acetyltransferase reaction involving an acetyl-enzyme intermediate appears less likely than the ternary complex mechanism proposed here, due to the existence of GlmU/acetyl-CoA complex crystals prepared from both E. coli and S. pneumoniae GlmU. These crystals clearly depict the location of the acetyl group bound to the coenzyme and not as an acetyl-enzyme intermediate. It may also be relevant to note that all enzymes of the hexapeptide acyltransferase superfamily for which chemical mechanisms have been proposed are thought to operate by a similar ternary complex mechanism, despite dissimilarities in the environment of their active-site histidine residues, the subunit from which this residue is donated, and the chemical identity of the substrate acceptor group as hydroxyl or amine.

Materials and Methods

Overexpression, purification, and crystallization

Native E. coli GlmU was overexpressed and purified by anion exchange and gel-filtration chromatography as previously described (Olsen et al. 2001). Three GlmU complex crystals were prepared by the hanging-drop vapor-diffusion method by mixing 4 mL of a protein solution containing ligands with a similar volume of reservoir solution on a circular coverslip and inverting over a sealed 1-mL reservoir.

Acetyl-CoA/UDP-GlcNAc

The protein solution was 8.3 mg/mL GlmU, 6.6 mM Tris HCl (pH 7.5), 33 mM NaCl, 0.66 mM DTT, 0.013% sodium azide, 20 mM MgCl₂, 20 mM acetyl-CoA, 100 mM Glc-1-PO₄, and 20 mM UDP-GlcNAc. The reservoir solution was 50 mM Bis-Tris (pH 6.1), 1.8 M (NH₃)₂SO₄, and 4 mM CoCl₂. Crystals were harvested and transferred to a soak solution containing 25 mM MES (pH 6.4), 1.8 M (NH₃)₂SO₄, 28 mM MgCl₂, 10 mM acetyl-CoA, 100 mM Glc-1-PO₄, and 10 mM UDP-GlcNAc for 4 d prior to X-ray data measurement.

CoA/GlcN-1-PO₄/UDP-GlcNAc

The protein solution was 8.3 mg/mL GlmU, 6.6 mM Tris HCl (pH 7.5), 33 mM NaCl, 0.66 mM DTT, 0.013% sodium azide, 20 mM MgCl₂, 16 mM CoA, 24 mM GlcN-1-PO₄, and 20 mM UDP-GlcNAc. The reservoir solution was 56 mM Bis-Tris (pH 6.4), 1.83 M (NH₃)₂SO₄, and 4.4 mM CoCl₂. Crystals were harvested and transferred to a soak solution containing 25 mM MES (pH 6.4), 1.8 M (NH₃)₂SO₄, 28 mM MgCl₂, 10 mM CoA, 98 mM GlcN-1-PO₄, and 10 mM UDP-GlcNAc for 1 d prior to X-ray data measurement.

Desulpho-CoA/GlcNAc-1-PO₄/UDP-GlcNAc

The protein solution was 8.3 mg/mL, 6.6 mM Tris HCl (pH 7.5), 33 mM NaCl, 0.66 mM DTT, 0.013% sodium azide, 20 mM MgCl₂, 16 mM desulpho-CoA, 10 mM GlcNAc-1-PO₄, and 20 mM UDP-GlcNAc. The reservoir solution was 56 mM MES (pH 5.6), 1.67 M (NH₃)₂SO₄, and 11.1 mM CoCl₂. Crystals were harvested and transferred to a soak solution containing 25 mM MES (pH 6.4), 1.8 M (NH₃)₂SO₄, 28 mM MgCl₂, 100 mM GlcNAc-1-PO₄, and 10 mM UDP-GlcNAc for 3 d prior to X-ray data measurement.

X-Ray data measurement and structure determination

X-ray diffraction data were measured either in-house using a Rigaku R-Axis IV⁺⁺ image plate detector and RU-H3R rotating anode X-ray generator equipped with Osmic Blue optics and operating at 50 kV and 100 mA or at beamline X9A of the National Synchrotron Light Source, Brookhaven National Laboratory (Table 1). The crystals were briefly exposed to crystal-storage liquor solutions containing 30% glycerol prior to cooling in a nitrogen gas stream at 125 K. The X-ray diffraction data were reduced with HKL (Otwinoski and Minor 1997). The crystals described here were isomorphous to the rhombohedral form of GlmU crystallized in complex with CoA and UDP-GlcNAc (Olsen and Roderick 2001). The atomic coordinates of this structure (PDB accession code 1hv9) were used to solve the structure of the newly prepared GlmU complex crystals by molecular replacement. Model building and atomic parameter refinements were carried out with O (Jones et al. 1991) and CNS, utilizing thermal factor restraints (Brunger et al. 1998).

Table 1.

Data measurement and refinement statistics

Data deposition

Atomic coordinates have been deposited in the RCSB Protein Data Bank (accession codes 2OI5, 2OI6, and 2OI7).