Structural Insight into the Transglycosylation Step of Bacterial Cell-Wall Biosynthesis

See allHide authors and affiliations

Science  09 Mar 2007:
Vol. 315, Issue 5817, pp. 1402-1405
DOI: 10.1126/science.1136611


Peptidoglycan glycosyltransferases (GTs) catalyze the polymerization step of cell-wall biosynthesis, are membrane-bound, and are highly conserved across all bacteria. Long considered the “holy grail” of antibiotic research, they represent an essential and easily accessible drug target for antibiotic-resistant bacteria, including methicillin-resistant Staphylococcus aureus. We have determined the 2.8 angstrom structure of a bifunctional cell-wall cross-linking enzyme, including its transpeptidase and GT domains, both unliganded and complexed with the substrate analog moenomycin. The peptidoglycan GTs adopt a fold distinct from those of other GT classes. The structures give insight into critical features of the catalytic mechanism and key interactions required for enzyme inhibition.

The bacterial cell wall is built by glycosyltransferase (GT) and transpeptidase (TP) enzymes, which together produce a cross-linked peptidoglycan mesh that gives the bacterium its strength and shape. These enzymes are excellent drug targets because they are essential, are accessible from the periplasm, and have no equivalent in mammalian cells. TP enzymes [also known as penicillin-binding proteins (PBPs)] are the target of the β-lactam group of antibacterials, to which a disturbingly high degree of resistance has developed over the past two decades of clinical use. The peptidoglycan GT protein family, GT51 (, has no structural representatives to date and is an attractive candidate for the structure-based drug design of new antibacterialcompounds.

GT51 and TP enzymes may be present as monofunctional or bifunctional proteins (1), with the more prevalent bifunctional enzymes possessing an N-terminal GT51 domain and a C-terminal TP domain, separated by a small linker region. The GT51 domains are always membrane-bound, and they polymerize a lipid II pentapeptide substrate into β1,4-linked N-acetylmuramic acid, N-acetylglucosamine polymers [(NAM-NAG)n]. The TP domains are solvent-exposed and catalyze the cross-linking of the peptide substituents on the 3′ acetyl groups of NAM. Inhibitors of the GT51 reaction may be less subject to the development of resistance than TP inhibitors, which can be rendered ineffective by alteration of the peptide substituents—notably in vancomycin-resistant strains (2, 3). There is only one well-characterized inhibitor of the GT51 reaction, the Streptomyces natural product moenomycin, which is not effective in humans because of poor absorption properties (4). Despite its extensive use as a growth promoter in animal feed, no plasmid-borne resistance to moenomycin has been detected so far (5). Understanding of the specific moenomycin:GT51 interactions would be useful in designing a similar compound to treat human bacterial infections. Here, we report two crystal structures of a soluble truncated version of PBP2 (the only bifunctional enzyme identified in the pathogen Staphylococcus aureus), an apoenzyme structure (2.9 Å resolution), and a complex of the enzyme with the antibiotic moenomycin (2.8 Å). The growth of both crystal forms required the detergent lauryldimethylamineoxide (6). Both structures have favorable stereochemistry (table S1), with Rwork/Rfree values of 19.1/24.6% for the apoenzyme and 23.1/28.0% for the moenomycin-bound form.

The PBP2 structure reveals a bilobal fold (dimensions 55 by 60 by 110 Å), with the GT51 and TP domains separated by a short β-rich linker region (Fig. 1). The construct used (amino acids 59 to 716) omitted the N-terminal transmembrane sequence and some of the low-complexity C-terminal tail. Electron density is visible for residues 67 to 692, except for a few short loops in the GT51 domain (127 to 139 and 160 to 163 in the apoenzyme and 137 to 144 in the moenomycin complex). The TP domain structure has high structural homology to Streptococcus pneumoniae PBP1a (7) [29% sequence identity for the 293-to-679 region, with a root mean square deviation (RMSD) of 2.0 Å for Cα]. Either, but not both, of the GT51/TP catalytic domains are superimposable when comparing the two crystal forms, indicating that there is some flexibility around the linker region.

Fig. 1.

Overall structure of PBP2. The modular nature of the GT51 and TP folds is shown, with the predicted position of the transmembrane region represented by a vertical blue rectangle. The active-site residues for the GT51 (lower, E171; upper, E114) and TP (S398) functionalities have their Cα atoms shown as red spheres. The GT51 domain is proposed to be at least partly submerged in the bilayer in order to access the lipid II substrate, which is consistent with observations that catalytic activity and crystallization were dependent on the presence of detergents.

The GT51 domain has two putative catalytic glutamic acid residues [E114 and E171 (8)], and their importance has been demonstrated in mutational studies on the Escherichia coli PBP1b enzyme (9). The overall fold of the lipid II-dependent GT51 domain shows no similarity to either of the GT-A or GT-B folds that have been observed in all preceding GT enzyme–class structures. The structure is primarily α-helical and is composed of two segments: one globular “head” region situated next to the linker domain and a smaller “jaw” region beneath this, closer to the membrane (Fig. 1). The globular segment has seven α helices and a small β region, and it has low structural similarity to bacteriophage λ lysozyme (λL) [DALI (10) gives a Z score of 3.5, 12% sequence identity for the 98-to-266 region, with an RMSD of 4.2 Å for Cα]. Given that the two enzymes act on similar substrates, it is possible that the GT51 module was acquired by the bacteriophage at a point long ago in its evolution. Despite low sequence identity, there is good overlap of the secondary structure between the two enzymes, and the first putative glutamate catalytic residue of the GT51 domain overlays that of λL (E114 with E19, respectively; Fig. 2A). The helical topology is preserved between the lysozyme and GT51 folds, with the soluble β-rich domain I of λL replaced by the membrane-associated GT51 jaw subdomain (residues 121 to 182). The jaw is composed of three α helices, with the outermost helix containing the second putative catalytic residue, E171. A pocket formed by residues close to the linker region shows sequence conservation in other GT51 enzymes and may be important for contacting the end of the transmembrane segment. A cleft between the globular and jaw subdomains connects this pocket with a larger pocket on the opposite side of the molecule. The cleft superimposes with the substrate binding site in λL and related lysozyme folds (Fig. 2A).

Fig. 2.

GT51 domain comparison with lysozyme and conformational change between the apoenzyme and liganded forms. (A) Structure-based alignment of the GT51 domain (blue) with that of λL(green). Helices H1 and H3 of λL have counterparts in all related G-, C-, and V-type lysozyme folds (20). The putative E114 catalytic residue of the GT51 domain aligns well with that of λL (both in stick representation) at the end of λL helix H1. (Inset) Orthogonal view with loops omitted. The lysozyme structure from deposition 1D9U (15) is shown, with the poly-NAG ligand in stick form. The figure highlights both the striking similarity in secondary structure (localized only to the GT51 head subdomain) and the conservation of an active-site cleft between the two enzyme types. The regions below the cleft have no structural similarity and are presumably different because of the need for the GT51 domain to be localized at the membrane and to bind a lipidic substrate. (B) Conformational change in the jaw subdomain upon binding moenomycin (represented in stick form). The moenomycin-bound form is blue, and the apoenzyme is yellow.

The active site of the TP domain contains the characteristic SXXK, (S/Y)X(N/C), and (K/H)(T/S)G motifs (1). The TP catalytic S398 residue lies ∼70 Å from the putative catalytic residue E114 of the GT51 domain. The GT51 fold possesses five signature motifs, conserved in both the monofunctional and bifunctional GT51 enzymes (1) (fig. S1). Motifs I and III contain the GT51 catalytic glutamic acids (E114 and E171, respectively), with their Cδ atoms ∼14 Å apart in the apoenzyme. Of the five signature motifs, motifs IV and V appear to largely play a role in maintaining the structure of the GT51 fold, whereas motif II forms the innermost helix of the jaw subdomain that divides the two pockets and is probably involved in substrate recognition. Constraints imposed by the location of N-terminal residues 68 to 73, the hydrophobicity of the lower part of the jaw subdomain, and the hydrophilic nature of the TP domain lead to a model for membrane orientation (Fig. 1).

The moenomycin-bound structure (Fig. 3) was obtained by cocrystallization. The electron density of this complex inhibitor is well defined for all groups other than the C25 lipid unit (Fig. 3C). Moenomycin binds in the cleft between the head and jaw regions of the GT51 fold, in an extended conformation that probably mimics that of the growing sugar-chain substrate or product. Rings B to F (Fig. 3B), modeled in a chair conformation, form a twisted plane, with ring F nearest the putative E114 catalytic residue, ring B located toward the linker region, and ring D projecting out into the bulk solvent. Consistent with this model, moenomycin derivatives lacking ring D retain antibiotic activity (11). Ring A exits the channel on the side furthest from the membrane and may be positioned similarly to a peptide substituent from a NAM substrate. The phosphoric acid diester group of moenomycin is under the plane of rings B to F, in an orientation that potentially directs the lipid group toward the membrane. The interactions between the protein and inhibitor are extensive (Fig. 3A), consistent with moenomycin and related compounds possessing some of the lowest median inhibitory concentration values known for antibiotics (12). The head subdomain is largely identical between the apoenzyme and moenomycin-bound GT51 structures, whereas large conformational changes occur in the jaw subdomain (Fig. 2B). The strictly conserved D115, H121, K184, and I187 residues and the largely conserved A123, T150, and E179 residues are located in the hinge region between the two subdomains and may be important in the ability of the GT51 domain to undergo such a conformational change. Based on chemical similarity, it has been proposed that rings C and E of moenomycin would bind in the same way as a NAM-NAG disaccharide of the substrate. In the complex of GT51 with moenomycin, the ring oxygens and N-acetyl groups of rings C and E and all three β1,4 linkages are spaced in agreement with a linear glycan chain. Arranging the β1,4 linkages of the growing chain substrate to mimic those of the inhibitor places the reducing 4-OH end of the chain furthest from the catalytic machinery, identical to that of the lysozyme family of enzymes.

Fig. 3.

Detail of moenomycin binding. (A) Interactions of moenomycin with the donor substrate site of GT51, with selected side chains shown. The phosphoric acid diester group is positioned in a region of high positive charge, and the sugar rings form an extended plane in the active-site cleft. Ring D projects out into solution, and ring A is located closest to the linker region. (B) Chemical structure of moenomycin. For clarity, an R1 group is used in place of the C25 moenocinol lipid unit. (C) Electron density of the moenomycin molecule. The map is of a 2fo-fc format, contoured at 1σ, from the final model at 2.8 Å resolution. For reference, the putative catalytic residues E114 and E171 are shown in stick form.

The differences between the moenomycin ring F/phosphoric acid diester/C25 moieties and the substrate NAM/pyrophosphate/C55 moieties must be responsible for GT51 inhibition by moenomycin. To investigate this hypothesis, we modeled a productive complex of both the lipid II and growing chain substrates on the moenomycin-bound structure. Whether lipid II is the donor or acceptor in the polymerization reaction has been extensively debated (13). From constraints imposed by the structures of the apoenzyme- and moenomycin-bound forms, we can now almost definitely conclude that lipid II is the acceptor, and the growing chain is the donor, with support from numerous experimental sources [summarized by Welzel (13)]. Modeling the growing chain (donor) substrate onto the moenomycin structure, a NAG group is placed over ring E, and the adjoining NAM sugar super-imposes with ring F. The moenomycin-based model also situates the NAM peptide substituents so that no steric clashes occur with the active-site cleft. Replacing the β1,4 linkage between a NAM-NAG disaccharide with a β1,2 linkage between rings E and F in moenomycin results in each ring F substituent of moenomycin correlating with the C+2 substituent of NAM (e.g., the C3 OCONH2 group of ring F with the C5 CH2OH group of NAM). Because of this staggering, the C1 phosphoryl groups of both the inhibitor and the substrate point in similar (axial) orientations, and the C4 OH of moenomycin is roughly equivalenttothe modeledringoxygenof NAM [as suggested earlier in (14)]. This positions the axial methyl on C4 of the inhibitor above the plane of NAM, over the ring oxygen. The methyl group lies between the donor C1 site of nucleophilic attack and the region occupied by E114 and the incoming 4-OH end of the acceptor. In our structures, there is ample space to model the lipid II acceptor adjacent to the proposed donor site, with the 4-OH end available for interaction with both E114 and C1 of the donor sugar. This position of lipid II is comparable to that of subsites +1 and +2 of the substrate in λL (15), with the prereaction GT51 substrates similar to the postreaction lysozyme products.

We propose that E114 is a Brønsted base and acts to directly abstract a proton from the 4-OH group of the lipid II acceptor. The deprotonated form of E114 may be stabilized by the adjacent R249 residue, strictly conserved as part of motif V. The proton abstraction step probably occurs concomitantly with the electrophilic migration of the donor C1 toward the acceptor 4-OH group (Fig. 4, A and B). In the moenomycin complex, the conserved E171 residue lies closer to the glyceric acid moiety than the phosphate-sugar linkage (the β phosphate in our substrate model), which in combination with pH activity profiles of the E. coli PBP1b enzyme (16) casts some doubt on whether E171 protonates the sugar-phosphate linkage to assist catalysis. Furthermore, mutants of this residue in E. coli PBP1b retain some residual activity, whereas those of our predicted Brønsted base, E114, do not (9). If E171 does not act to protonate the substrate, then we propose that it helps to coordinate the pyrophosphate group of the donor, either directly or via a divalent metal cation. The variable pH optima and divalent cation requirements of the GT51 family of enzymes (1719) may result from varying local environments of the E171 residue. The SN2-like reaction occurs between donor and acceptor, causing inversion at the donor C1 anomeric carbon and formation of the β1,4-linked product. The lipid-pyrophosphate leaving group of the donor is then free to diffuse away and be recycled in lipid II synthesis. We propose that translocation of the newly formed product to the donor site is assisted by a higher affinity for the pyrophosphate moiety in the donor site than in the acceptor site, with the conserved positively charged K155, K163, R167, and K168 residues located near the donor pyrophosphate region of the active site (Fig. 4C). This model is again reminiscent of the lysozyme active site, where the +1 and +2 subsites that match the modeled GT51 acceptor sugars possess the lowest substrate affinity of all the subsites. These two structures now provide a basis for addressing further questions about the mechanism of this important family of enzymes and for the design of new antibacterials. This work also opens the door for understanding structure and function relationships in other clinically important families of lipid-sugar GTs.

Fig. 4.

Proposed mechanism for lipid II polymerization. To simplify these diagrams, the peptide substituents on lipid II have been omitted. (A) Schematic for lipid II polymerization. For clarity, R1 and R2 groups are used in place of the OAc and NHAc groups, respectively. For comparative purposes, the respective lysozyme sugar subsites are labeled in parentheses (the traditional nomenclature using subsites a to f has been avoided to prevent confusion with the moenomycin ring labeling). In this model, lipid II is the acceptor (right side), and the growing glycan chain is the donor (left side). Residue E114 acts to deprotonate the acceptor 4-OH group, which concomitantly attacks C1 of the donor, in an SN2-like reaction that inverts the α-linked precursors into a β1,4-linked product. Residue E171 may assist this process by direct protonation of the phosphate-sugar bond or by stabilizing the pyrophosphate group through interaction with a divalent cation. (B) Spatial representation of the lipid II polymerization model. The membrane interface (horizontal black line), transmembrane region (vertical blue rectangle), and missing polypeptide (dotted blue line) are shown for effect. The protein structure is unmodified from the moenomycin-bound complex, with the growing glycan-chain donor (left side) modeled over moenomycin rings E and F and the lipid II acceptor (right side) fitted manually between the glycan-chain donor and the E114 catalytic residue (shown with E171 in stick form). After polymerization, the product would be translocated in the direction denoted by the yellow arrow. Any sugar chain larger than four sugar units (discounting the incoming two sugar units of the acceptor) would project out from the GT51 domain, and there are no steric barriers in our structure to prevent this from occurring. The stronger positive charge on the left side of the active site, relative to that on the right side [see (C)], may assist in movement of the retained acceptor lipid-pyrophosphate group into the donor position. (C) Detail of active-site pockets and cleft. Residues E114 and E171 are shown in space-filling form. The electrostatic potentials (red, negative; blue, positive) indicate a conserved region of positive charge across the middle of the pocket. This region binds the phosphoric acid diester group of moenomycin in our structure and is located in a position to bind both pyrophosphates in our substrate model. (D) Details of the hydrophobic platform of the GT51 fold. The view is approximately 90° from (C), with residues shown in stick form. Green, hydrophobic platform; gray, E114 and E171.

Supporting Online Material

Materials and Methods

SOM Text

Fig. S1

Table S1


References and Notes

View Abstract

Navigate This Article