High-resolution crystal structure reveals molecular details of target recognition by bacitracin

Abstract

Bacitracin is a metalloantibiotic agent that is widely used as a medicine and feed additive. It interferes with bacterial cell-wall biosynthesis by binding undecaprenyl-pyrophosphate, a lipid carrier that serves as a critical intermediate in cell wall production. Despite bacitracin’s broad use, the molecular details of its target recognition have not been elucidated. Here we report a crystal structure for the ternary complex of bacitracin A, zinc, and a geranyl-pyrophosphate ligand at a resolution of 1.1 Å. The antibiotic forms a compact structure that completely envelopes the ligand’s pyrophosphate group, together with flanking zinc and sodium ions. The complex adopts a highly amphipathic conformation that offers clues to antibiotic function in the context of bacterial membranes. Bacitracin’s efficient sequestration of its target represents a previously unseen mode for the recognition of lipid pyrophosphates, and suggests new directions for the design of next-generation antimicrobial agents.

Antibiotic resistance has emerged in most bacterial pathogens, rendering many previously powerful antibiotics useless. At the same time, the number of new drugs in the antibiotic pipeline has dwindled (1), raising the specter of a return to the preantibiotic era (2) and highlighting the importance of the development of new antimicrobial drugs. When considering templates for new drug development, it is prudent to choose compounds that differ mechanistically from the most commonly used antibiotic agents, given the speed and ingenuity with which bacteria can evade new variants of such drugs (3). One such compound is bacitracin, which recognizes a lipid pyrophosphate target that is not exploited by other commonly used therapeutic agents. Further, although bacitracin has been used for 70 y as a human and veterinary medicine and an animal-feed additive (4, 5), resistance has not spread widely through different microbial species, most likely because the drug’s clinical use is largely limited to topical application. Hence, the bacitracin scaffold is an appealing candidate for structure-based design of new therapeutic agents.

Bacitracin is produced by Bacillus subtilis and Bacillus licheniformis as a mixture of closely related dodecapeptides (6). These peptides are synthesized by nonribosomal peptide synthases (7–9), and contain both d- and l-amino acids. The peptides are cyclized into lariat structures via condensation of the ε-amino group of a lysine side chain with the peptide C terminus (10). The most potent of the bacitracin congeners is bacitracin A, which contains a thiazoline ring at its N terminus, formed by the condensation of Ile-1 and Cys-2 (11, 12) (Fig. 1).

Fig. 1.

Structure of bacitracin A2. (Upper) Sequence of linear peptide precursor. (Lower) Chemical structure of mature bacitracin. The mature antibiotic is formed by the cyclodehydration of Cys-2 with the backbone to form the thiazoline ring, together with the formation of a lariat structure via an isopeptide bond between the ε-amino group of Lys-6 and the peptide’s C terminus.

As is true for many natural product antibiotic agents (13), bacitracin possesses multiple activities, only some of which are explicitly antimicrobial. For example, bacitracin binds divalent metal ions, allowing it to act as a redox agent, a property that has been exploited to produce reagents for the oxidative cleavage of DNA (14). In its metal-free form, bacitracin binds and inhibits bacterial subtilisin-type proteases (15), probably to maintain these enzymes in an inactive form until they have been secreted from the bacterium. Bacitracin also inhibits protein disulfide isomerases, albeit with low specificity (16, 17).

Bacitracin’s antimicrobial properties derive from yet another activity, namely its ability to compromise the integrity of the bacterial envelope. Bacitracin binds undecaprenyl pyrophosphate, a lipid carrier that shuttles cell-wall biosynthetic intermediates from the cell’s cytoplasm to its exterior (18). By sequestering the lipid carrier, bacitracin interrupts the flow of peptidoglycan precursors to the site of cell-wall synthesis, weakening the cell wall and ultimately leading to bacterial death (19, 20). To bind the lipid pyrophosphate molecule, bacitracin requires a divalent metal ion, forming a ternary 1:1:1 antibiotic–metal–lipid complex (21). Bacitracin can bind a variety of different metals, but zinc supports lipid binding most potently, and the zinc form of bacitracin is most commonly used in antibiotic formulations.

Despite bacitracin’s importance as a therapeutic agent, the structural basis for how the drug recognizes its lipid pyrophosphate target has remained elusive, thwarting efforts to use bacitracin as a template for the structure-based design of new antibiotics. Structures have been determined for metal-free bacitracin alone in solution (22, 23) and in complex with several different proteases (24, 25). Notably, the antibiotic adopts a broad range of conformations in these structures, indicating that the molecule possesses considerable intrinsic flexibility. The solution structure of the binary antibiotic–cobalt complex has also been determined, and shows that metal binding induces a more compact configuration of the antibiotic (26); however, the antibiotic still retains substantial residual flexibility in this binary complex. The binary antibiotic–cobalt complex structure has been used as the basis for modeling the structure of the ternary complex of antibiotic, metal, and lipid pyrophosphate (14), but, until now, no experimental structure has been available for the ternary complex.

We now report the 1.1-Å resolution crystal structure of the ternary complex of bacitracin A, zinc, and a lipid pyrophosphate ligand. This structure shows a substantially different antibiotic conformation than any described to date, and makes clear the details of target binding. Our results demonstrate that bacitracin employs a degree of substrate sequestration previously unseen with peptide antibiotics, and also reveal an unsuspected role for sodium ions in target recognition.

Results

Crystal Structure of the Bacitracin A–Target Complex.

Because bacitracin is flexible, we expected it to be difficult or impossible to crystallize the antibiotic by itself. However, we reasoned that ligand binding might induce bacitracin to assume a rigid and compact conformation that could be crystallized. We chose geranyl (diprenyl) pyrophosphate as the lipid ligand because it is more soluble and has lower entropy than the native undecaprenyl pyrophosphate target, but is still expected to bind bacitracin with high affinity (18). We chose zinc as the metal for the ternary complex because of its clinical relevance and because its anomalous scattering properties would lend themselves to phase determination. After a rigorous screening process, we obtained crystals of the ternary complex of bacitracin A with zinc and geranyl pyrophosphate that diffracted to 1.1 Å. Phase determination by the single-wavelength anomalous dispersion method yielded a high-quality electron density map in which the entire ternary complex could clearly be discerned (Fig. 2).

Fig. 2.

(A) Stereo view of bacitracin in complex with its geranyl-pyrophosphate ligand. Bacitracin is shown in a cyan ribbon representation, with atoms in side chains being colored gray (carbon atoms), red (oxygen), and blue (nitrogen); the sulfur of the thiazoline ring is colored gold. The pyrophosphate group of the ligand is colored red (oxygen) and orange (phosphorus); the geranyl chain is colored yellow. The sodium and zinc ions are shown as light green and light blue spheres, respectively. (B) Experimental electron density maps from SAD phasing. In light blue is the SAD-phased, solvent-flattened map used to fit the model (shown at a contour level of 1σ). In magenta is the anomalous difference Fourier map, contoured at 6σ. The pyrophosphate group of the lipid is shown, together with the two metal ions. (C) Sequestration of the pyrophosphate group by bacitracin. A semitransparent surface representation for bacitracin is shown in cyan. (D) An opaque surface representation is shown to highlight the almost complete burial of the pyrophosphate group. (Left) Same view as seen in C; two additional rotated views are shown (Center, Right) to emphasize that the pyrophosphate group is inaccessible to solvent, regardless of the angle of approach.

In our crystal structure, bacitracin A adopts a compact configuration, with the antibiotic wrapping tightly around the lipid pyrophosphate and zinc ion (Fig. 2). For the first eight residues of the antibiotic, the backbone curves around the pyrophosphate group; at residue 9, the backbone forms a reverse turn, so that residues 10 to 12 lie above residues 6 to 8, in an antiparallel orientation. The antibiotic forms a highly curved C-shaped “wall” that encloses the ligand. The two ends of this curved structure—the antibiotic’s N terminus on one side, and the reverse turn containing residues 8 to 10 on the other—do not interact directly with each other, but rather are bridged by the lipid pyrophosphate ligand, which acts as a clasp to maintain the antibiotic in a closed form. Side chains from either end of the antibiotic (Ile-1, Phe-9, and His-10) wrap around the opening, completing the sequestration of the target from its environment. This results in the almost complete burial of the target’s pyrophosphate group—less than 2% of the pyrophosphate’s surface area remains solvent-accessible in the complex (Fig. 2).

Within the curved structure formed by bacitracin, the pyrophosphate group of the ligand is flanked by the zinc ion and an additional metal, which we have modeled as a sodium ion (Fig. 3). The sodium appears to serve electrostatic and structural roles: it helps to neutralize the negative charge of the pyrophosphate group, and also bridges two oxygen atoms of the pyrophosphate with side-chain and main-chain oxygen atoms of the antibiotic.

Fig. 3.

(A) Recognition of the lipid’s pyrophosphate group by bacitracin. Distances (in Å) are shown between oxygen atoms of the ligand, hydrogen bond donors on the antibiotic, and metal ions. Red numbers in circles denote the residue number of a particular backbone or side chain group of the antibiotic. (B) Coordination of the zinc ion. (C) Coordination of the sodium ion. Color scheme is the same as in Fig. 1A.

In the ternary complex, bacitracin adopts a highly amphipathic configuration (Fig. 4). All the polar side chains of the antibiotic segregate on one side of the molecule, whereas all the nonpolar side chains fall on the opposite side. Most of the polar side chains are directly involved in recognizing the metals or the pyrophosphate group, whereas the nonpolar side chains form a hydrophobic collar that surrounds the exit channel through which the lipid chain extends out of the binding site. Within the crystal lattice, the hydrophobic faces of multiple bacitracin molecules associate with one another (Fig. S1). In the presence of a bacterial membrane, however, the hydrophobic face of the antibiotic will presumably associate with the lipid membrane, with the lipid’s undecaprenyl tail passing through the hydrophobic collar and entering the membrane interior.

Fig. 4.

(A) Amphipathic structure of bacitracin. The antibiotic molecule is shown in the same orientation as in Fig. 1A. The peptide backbone is colored yellow, hydrophilic side chains are colored blue, and hydrophobic side chains are colored red. (B) Bacitracin binding to tethered lipid bilayers as measured by surface plasmon resonance. The four panels show antibiotic binding to bilayers in the presence (Upper) or absence (Lower) of Zn²⁺, and using membranes that contain (Right) or do not contain (Left) undecaprenyl-pyrophosphate (C₅₅-PP). The color coding corresponds to an antibiotic concentration series of 0, 0.75, 1.5, 3, 6, and 12 µM. (C) Bactericidal doses of bacitracin fail to disrupt bacterial membranes. K⁺ efflux was measured from M. luteus cells using a potassium-sensitive fluorescent dye; the indicated concentrations of antibiotic were added to the cuvette at the time indicated by the arrow. Gramicidin S was included as a positive control.

Target Recognition by Bacitracin A.

Bacitracin recognizes the pyrophosphate group of the lipid target by direct antibiotic–lipid interactions and indirect interactions mediated by the two metal ions (Fig. 3). Direct interactions take the form of hydrogen bonds between backbone and side-chain amide groups of the antibiotic and oxygen atoms of the lipid. Five of the antibiotic’s 12 residues contribute backbone hydrogen bonds to the ligand, with a sixth hydrogen bond being contributed by the side chain of Asn-12. These interactions are augmented by interactions between the pyrophosphate and the two metal ions, which sit on either side of the pyrophosphate group. Each metal interacts with two oxygen atoms of the lipid, one each from the terminal and bridging phosphate groups. The metals are in turn chelated by groups on the antibiotic.

The zinc ion adopts an octahedral coordination geometry. Two of the metal’s ligands are oxygen atoms of the lipid pyrophosphate. Three additional ligands are provided by the antibiotic: the side chain of d-Glu-4, the N-terminal amine, and the nitrogen of the thiazoline ring. The sixth zinc ligand is a water molecule. Metal-ligand distances vary between 2.0 and 2.2 Å (Table S1), and fall within the normal ranges for Zn-O and Zn-N coordination distances (27, 28). The N-terminal amine is presumably unprotonated, even though the crystals were grown at neutral pH, evidently because of a metal-induced pK_a shift (29). The N terminus does not contribute a hydrogen bond to the nearby pyrophosphate group, as has been suggested (14). The sodium ion adopts a slightly distorted square pyramidal coordination geometry; the metal’s five ligands are two oxygen atoms from the pyrophosphate group, the side chain of d-Asp-11, and the backbone carbonyl oxygens of residues 5 and 8. Metal-oxygen distances for the sodium ion range from 2.2 to 2.4 Å (Table S1), in good agreement with the predicted ideal value of 2.36 Å for a pentacoordinate Na⁺ ion (30).

The diprenyl chain of the geranyl pyrophosphate emerges from the enveloping grasp of the antibiotic through a channel lined by the side chains of Ile-1, Leu-3, Ile-5, Ile-8, and d-Phe-10. With the exception of Ile-1, these hydrophobic side chains all pack directly around the geranyl group, approaching the lipid to within 3.5 to 4 Å. Most of these close contacts are with the first prenyl group (i.e., the one closer to the pyrophosphate), or with the proximal atoms of the second prenyl group. The antibiotic makes no interactions with the more distant atoms of the second prenyl group. The electron density for the second prenyl group is extremely weak, reflecting a high degree of disorder stemming from the failure of the antibiotic to bind and stabilize the distal end of the lipid chain. This mode of interaction is consistent with the observation that bacitracin binds short-chain and long-chain lipid pyrophosphate species equally well (18). Thus, the primary determinants for ligand recognition are the pyrophosphate group and the first prenyl group, whereas the remainder of the undecaprenyl lipid chain contributes only indirectly to target recognition by tethering the pyrophosphate group in the membrane environment.

Membrane Binding and Disruption by Bacitracin.

The amphipathic structure of bacitracin suggests it might possess an innate affinity for membranes, even in the absence of the undecaprenyl pyrophosphate target. To test this, we used surface plasmon resonance to measure bacitracin binding to immobilized lipid bilayers in the presence and absence of zinc and the undecaprenyl pyrophosphate ligand. Although the sensorgrams obtained were refractory to simple kinetic analysis, they showed clear differences in the amount of antibiotic bound under the different conditions (Fig. 4). The binary bacitracin–zinc complex showed modest binding to membranes in the absence of the ligand, but binding was enhanced dramatically in the presence of undecaprenyl pyrophosphate. In the absence of zinc, bacitracin bound only weakly to membranes containing the pyrophosphate ligand, highlighting the importance of the zinc ion as an organizing center around which the antibiotic and the pyrophosphate group can assemble. In the absence of metal or the pyrophosphate ligand, very little membrane binding was observed.

Having established that specific target recognition is required for bacitracin to bind membranes with high affinity, we next asked whether the antibiotic permeabilizes or otherwise disrupts the membrane subsequent to binding, as is seen with nisin (31). Using bacitracin-zinc, we measured antibiotic-mediated potassium efflux from Micrococcus luteus cells with the potassium-sensitive dye PBFI. Bacitracin failed to elicit leakage of potassium ions at concentrations as high as 40 mg/L (Fig. 4), although some slow leakage was observed at higher concentrations (Fig. S2). In contrast, we found that bacitracin’s minimum inhibitory concentration (MIC) for this same strain is 0.5 mg/L, in good agreement with a previously reported value (11). Zinc alone showed no effect on membrane permeability, and the positive control, the known membrane-permeabilizing agent gramicidin S, caused immediate and rapid potassium efflux upon addition. Hence, although bacitracin does appear to be able to disrupt membrane integrity, the concentrations required to do so are approximately two orders of magnitude higher than those necessary for antibiotic activity.

Discussion

Here we describe the crystal structure of bacitracin A in complex with zinc and a lipid pyrophosphate ligand. The structure reveals that the antibiotic wraps its ligand in an intimate embrace, forming an amphipathic shell that completely shields the pyrophosphate group from solvent and efficiently sequesters the ligand. Within this shell, the pyrophosphate group of the lipid is flanked by two metal ions; one is a zinc ion that has long been known to promote antibiotic function, whereas the other is a sodium ion, the requirement for which has not been previously appreciated. The zinc ion organizes the N-terminal region of the antibiotic around the ligand, whereas the sodium ion plays a similar role for the C-terminal region of the antibiotic.

This structure is only the second to describe the interaction of an antibiotic with a lipid pyrophosphate component of the cell-wall biosynthesis pathway; the first was that of the lantibiotic nisin in complex with lipid II (32). The bacitracin structure reveals a mode of recognition for its lipid pyrophosphate target that is distinct from that used by nisin. Nisin forms a concave “pyrophosphate cage” along one side of lipid II’s pyrophosphate group, wherein hydrogen bonds link antibiotic amide groups with pyrophosphate oxygen atoms. Unlike bacitracin, nisin uses no metals to bind lipid II, nor does it completely enclose its ligand. These structural differences allow us to rationalize the antibiotics’ ligand specificity. Nisin recognizes bulky ligands such as lipids II, III, and IV, in which sugars are attached to the pyrophosphate group (31, 33); the lantibiotic’s open architecture allows it to accommodate these sugar substituents, which are much too large to fit within the compact enclosure formed by bacitracin. Thus, by forming a closed structure that tightly encircles the pyrophosphate group, bacitracin becomes highly specific for undecaprenyl pyrophosphate over lipid II and related molecules. It is worth noting that the convergent evolution of bacitracin and nisin to recognize similar lipid pyrophosphate targets (albeit by dissimilar mechanisms) highlights the physiological importance of these targets.

Bacitracin’s amphipathic dome structure implies the topology with which the antibiotic interacts with its ligand in the context of a bacterial membrane. The dome likely assembles around the lipid head group with its bottom binding to the membrane, the lipid’s prenyl tail threading through the opening at the bottom of the dome, and the dome’s polar top facing away from the membrane. Consistent with this model, structural changes that reduce amphipathicity reduce potency. For example, substituting any of the isoleucines in bacitracin A with less hydrophobic valine residues gives rise to less effective drugs, as is seen in bacitracins B, D, and E (11). Also consistent with this model, the basic residues ornithine-7 and histidine-10 are located at the sides of the dome, where they are well positioned to interact with the negatively charged membrane-lipid head groups.

Bacitracin’s binding of undecaprenyl pyrophosphate and the subsequent disruption of cell-wall biosynthesis offer a convincing rationale for the molecule’s antibacterial activity, which is wholly consistent with the structure presented here. However, the highly amphipathic nature of the structure also prompts the question of whether bacitracin functions solely as an inhibitor of cell-wall biosynthesis, or whether it might also permeabilize cell membranes. Indeed, early work revealed that at sufficiently high concentrations, bacitracin could induce leakage of low molecular weight species from liposomes (34) and cause ultrastructural changes in membranes (35, 36); these observations have given rise to lingering questions about bacitracin’s membrane activity (19). Accordingly, we directly tested bacitracin’s membrane-binding ability, and found that both zinc and the lipid pyrophosphate target are required for efficient membrane localization. Although consistent with bacitracin acting as an inhibitor of cell-wall biosynthesis, this does not rule out the possibility of membrane perturbation; for example, nisin binds specifically to lipid II and subsequently inserts into the membrane to form pores (31). However, our finding that minimum inhibitory antibiotic concentrations are two orders of magnitude lower than concentrations required to permeabilize bacteria strongly argue that bacitracin’s antimicrobial activity stems purely from its ability to sequester undecaprenyl pyrophosphate, and that membrane disruption is an artifact occurring at high concentrations of the drug, and is not required for antibacterial action.

Comparing the structures of the binary bacitracin–metal complex and ternary bacitracin–zinc–ligand complex reveals significant differences in metal coordination. A variety of techniques has revealed that in the binary complex, zinc has at least three ligands: the thiazoline nitrogen and the side chains of Glu-4 and His-10 (26, 27, 37–39). The antibiotic’s N terminus is a fourth potential ligand, which would produce a square planar coordination geometry (29), but this remains controversial (6). In the ternary complex, three of these four groups (the N terminus, Glu-4, and the thiazoline) coordinate the zinc, and occupy the same relative positions with respect to the metal as in the binary complex. However, His-10 is not a metal ligand in the ternary complex; in addition, the zinc ion shifts to an octahedral coordination scheme in the ternary complex, acquiring two new oxygen ligands from the pyrophosphate group and a third from a water molecule.

Another significant structural difference between the binary and ternary complexes involves the antibiotic’s backbone conformation. In both cases, the backbone for residues 1 through 8 curves around the metal, and then reverses direction with a tight turn. However, these turns have opposite orientations in the two structures; in the ternary complex, the backbone turns “up,” away from the membrane-binding surface (Fig. 2), whereas, in the binary complex, the backbone turns downward (26). This causes the backbone to undergo a twist in the ternary complex that is not seen in the binary complex. As a result, the two complexes display very different surface distributions of amino acid side chains. The side-chain distribution is strictly amphipathic in the ternary complex, but, in the binary complex, polar and nonpolar side chains are more or less evenly distributed around the surface, giving rise to a structure with no obvious hydrophobic moment.

Certain differences between the binary and ternary complexes are readily explained. For example, His-10 may toggle on or off the metal in response to substrate binding. In the ligand’s absence, His-10 would coordinate the zinc, keeping the metal bound to the antibiotic and maintaining the antibiotic’s N-terminal region in a ligand-ready configuration. When ligand is encountered, His-10 would be released from the metal, allowing the antibiotic’s N and C termini to swing apart and admit the lipid’s pyrophosphate group (the sodium ion would presumably also enter at this point).

Other differences are less easily rationalized, particularly the difference in backbone configuration between the binary and tertiary complexes. The backbone topology reported for the binary complex implies that ligand binding would require a complete flip of the antibiotic’s C-terminal region; further, the proposed “turn-down” backbone structure in the binary complex gives rise to an amino acid surface distribution that is not obviously amphipathic, unlike the “turn-up” conformation of the crystal structure. It is possible that such a gross conformational change does indeed occur between the binary and tertiary complexes, possibly to allow the antibiotic to avoid forming an aggregation-prone hydrophobic surface until the target is encountered. However, invoking such a major conformational rearrangement seems unnecessarily complex. It would therefore be interesting to know whether the turn-up configuration is also consistent with the Co²⁺-derived distance restraints used to determine the binary complex structure (26). Alternatively, it possible that the zinc and cobalt binary complexes are not isostructural.

A model has previously been proposed for the ternary complex formed between bacitracin, metal, and farnesyl pyrophosphate, based on a combination of NMR measurements and molecular mechanics calculations (14). This model, however, bears little resemblance to the X-ray crystal structure reported in this paper, differing in the details of metal coordination, backbone conformation, and substrate recognition. In the model, His-10 remains bound to the metal, whereas Glu-4 is released; the crystal structure shows the opposite. In the model, the antibiotic’s backbone adopts the topology seen in the binary complex, rather than that seen in the crystal structure. As a consequence, the model structure places the side chains of Ile-5, Ile-8, and Phe-9 on the opposite side of the molecule from Ile-1 and Leu-3; in the crystal structure, all these residues lie on one side of the antibiotic, where they form the hydrophobic ring that surrounds the lipid tail at the point where it exits the binding site. Also, in the model structure, the third prenyl group of the farnesyl-pyrophosphate ligand interacts with hydrophobic side chains of bacitracin, whereas the crystal structure shows that the first prenyl group makes the main interactions with the antibiotic.

The details of ligand recognition revealed in this crystal structure demonstrate how many rounds of natural selection can produce a relatively small molecule that nonetheless functions with exquisite potency and specificity. Bacitracin inhibits bacterial cell-wall biosynthesis by targeting extracellular, membrane-associated pyrophosphate groups; such moieties are unique to the bacterial kingdom and their functionality cannot easily be modified by mutation (40). Furthermore, bacitracin sequesters its target in an impressively efficient way, by using a scaffold that completely envelopes the target’s pyrophosphate group. Such sequestration is a very different modality of binding than that used by nisin, which binds only one side of its lipid II target. Complete sequestration of the target affords protection against enzymatic and nonenzymatic hydrolysis of the target, either one of which would allow for recycling of the lipid carrier and continuation of cell-wall biosynthesis. However, fully sequestering the pyrophosphate group within a small cavity presents a practical challenge, namely how to neutralize the ligand’s strong negative charge. Bacitracin surmounts this challenge in an unanticipated way, by sandwiching the pyrophosphate between two different metals that serve to immobilize and neutralize the target.

Materials and Methods

Materials.

Bacitracin A (VETRANAL-grade) was purchased from Sigma-Aldrich. Geranyl pyrophosphate and undecaprenyl pyrophosphate were purchased from Isoprenoids; the lipid pyrophosphates were stored at −20 °C at 10 mg/mL in 10 mM ammonium bicarbonate and at −80 °C at 1 mg/mL in chloroform/methanol/1% NH₄OH (65/35/4), respectively. All other lipids were purchased from Avanti Polar Lipids and stored as 10 mg/mL solutions in chloroform.

Bacitracin A Purification.

Bacitracin A was separated from its congeners by HPLC using a semipreparative Ultrasphere 5 ODS (C18) column (Hichrom), as described previously (41). Peak fractions were diluted fivefold in water, desalted on a Sep-Pak Vac 12 mL C18 cartridge (Waters), lyophilized, and redissolved in 10 mM Hepes, pH 7.5.

Crystallization, Structure Determination, and Refinement.

Full details are provided in SI Materials and Methods. The ternary antibiotic–zinc–ligand complex was prepared in 5 mM Hepes, pH 7.5, 7 mM zinc acetate, and 10 mM sodium citrate at an antibiotic concentration of 5 mg/mL. A total of 0.5 μL of this solution was mixed with 0.5 μL of 20% (wt/vol) PEG-2000 in 0.1 M sodium acetate, pH 6.5. Crystals were immersed in cryoprotectant containing 3 volumes of glycerol plus 7 volumes of the PEG precipitant and flash-cooled by plunging into liquid nitrogen. Single-wavelength anomalous dispersion (SAD) data were collected at beam line X6A of the National Synchrotron Light Source (Table S2). Phasing and refinement were carried out by using PHENIX version 1.7.2–869 (42). An example of the final 2Fo-Fc map is shown in Fig. S3. Structure figures were generated by using the PyMOL Molecular Graphics System (version 1.5.0.4; Schrödinger). Coordinates and structure factors have been deposited in the Protein Data Bank with ID code 4K7T.

Measuring Membrane Binding Using Surface Plasmon Resonance.

All surface plasmon resonance experiments were performed on a BioRad ProteOn XPR36 using the ProteOn Liposome Capturing Kit (BioRad); full details are provided in SI Materials and Methods. Large unilamellar lipid vesicles were prepared with or without undecaprenyl pyrophosphate and captured on the same BioRad LCP chip. Bacitracin antibiotic dissolved in Hepes-buffered saline solution with or without 1 mM zinc acetate was flowed over the chip at 25 °C at a flow rate of 30 μL/min, using a blank flow strip as a reference. The sensorgrams displayed clear concentration dependence, but showed binding kinetics too complex to be modeled as simple 1:1 or 1:2 binding interactions. Therefore, we limited our assessment of bacitracin’s membrane binding to a semiquantitative analysis, focusing solely on mass bound (total response).

K⁺ Efflux and MIC Experiments.

Leakage of potassium ions from M. luteus cells (ATCC 49732; American Type Culture Collection) was measured as described previously (43), using the PBFI (Molecular Probes). The broth-dilution MIC for bacitracin was measured as described previously (44) for the same M. luteus strain, using brain/heart broth and overnight growth at 30 °C. Further details are provided in SI Materials and Methods.