Crystal structure of LpxK, the 4′-kinase of lipid A biosynthesis and atypical P-loop kinase functioning at the membrane interface

Abstract

In Gram-negative bacteria, the hydrophobic anchor of the outer membrane lipopolysaccharide is lipid A, a saccharolipid that plays key roles in both viability and pathogenicity of these organisms. The tetraacyldisaccharide 4′-kinase (LpxK) of the diverse P-loop–containing nucleoside triphosphate hydrolase superfamily catalyzes the sixth step in the biosynthetic pathway of lipid A, and is the only known P-loop kinase to act upon a lipid substrate at the membrane. Here, we report the crystal structures of apo- and ADP/Mg²⁺-bound forms of Aquifex aeolicus LpxK to a resolution of 1.9 Å and 2.2 Å, respectively. LpxK consists of two α/β/α sandwich domains connected by a two-stranded β-sheet linker. The N-terminal domain, which has most structural homology to other family members, is responsible for catalysis at the P-loop and positioning of the disaccharide-1-phosphate substrate for phosphoryl transfer on the inner membrane. The smaller C-terminal domain, a substructure unique to LpxK, helps bind the nucleotide substrate and Mg²⁺ cation using a 25° hinge motion about its base. Activity was severely reduced in alanine point mutants of conserved residues D138 and D139, which are not directly involved in ADP or Mg²⁺ binding in our structures, indicating possible roles in phosphoryl acceptor positioning or catalysis. Combined structural and kinetic studies have led to an increased understanding of the enzymatic mechanism of LpxK and provided the framework for structure-based antimicrobial design.

The unique cell envelope of Gram-negative bacteria consists of a phospholipid-derived inner membrane, a thin peptidoglycan layer, and an asymmetric outer membrane, the outer leaflet of which is composed mainly of LPS. LPS protects the cell from environmental stresses and contributes significantly to toxicity during infection of a host (1–3). Anchoring LPS to the outer membrane is the hexa-acylated disaccharide of glucosamine, known as lipid A, which when decorated with one or two 3-deoxy-d-manno-oct-2-ulosonic acid (Kdo) sugars is canonically the minimum structure required for robust growth of these microbes (1, 2). Additionally, lipid A is able to elicit a mammalian host’s inflammatory response through activation of the macrophage Toll-like receptor 4 and myeloid differentiation protein 2 complex (4–6). Thus, the lipid A moiety is of particular interest to those wishing to understand viability and pathogenicity of Gram-negative bacteria, and its biosynthetic pathway is an attractive target for the development of novel antibiotics (7–9).

In Escherichia coli, nine enzymatic steps make up the constitutive pathway of Kdo₂-lipid A biosynthesis, the sixth being the phosphorylation of disaccharide-1-phosphate (DSMP) at the 4′-position catalyzed by the membrane-bound kinase LpxK to form lipid IV_A (Fig. 1) (1, 2, 10, 11). Although mutants lacking the next enzyme of the pathway, the Kdo-transferase KdtA, are still able to survive upon overexpression of downstream modification and transport proteins (12–14), no suppressor mutation has been found that can confer viability to a knockout of the LpxK gene (15, 16). Therefore, LpxK has the distinction of being the last absolutely essential enzyme of the pathway. Because of the presence of Walker A and B motifs, LpxK has been classified as a member of the P-loop containing nucleoside triphosphate (NTP) hydrolase superfamily (Pfam02606, CL0023), which is named for the Walker A motif’s glycine-rich loop that binds the β- and γ-phosphates of NTPs (Fig. S1) (17–20). Although crystal structures of a handful of other P-loop kinases have been determined previously (19, 21, 22), LpxK is at least 100 aa longer than the majority of these family members and is the only known membrane-bound lipid kinase in the group, thus making its study important for expanding our knowledge of both lipid kinases and the diverse P-loop kinase family. Purification and characterization of E. coli LpxK has been hampered because of instability of the enzyme (11). Thus, our focus has shifted to the LpxK ortholog from the hyperthermophile Aquifex aeolicus, which shares 27.2% identity and 44.9% similarity with its E. coli counterpart and is predicted to lack the putative N-terminal integral membrane helix of E. coli LpxK by hydropathy analysis (23).

Fig. 1.

The role of LpxK in Kdo₂-lipid A biosynthesis. LpxK, the sixth enzyme of the pathway in E. coli, catalyzes the transfer of a phosphate group (purple) from ATP to the 4′-position of disaccharide-1-phosphate to form lipid IV_A. Positions on the glucosamine sugars (blue) are indicated (red). LpxK is the only kinase of the pathway and the phosphoryl transfer is an essential step in lipid A biosynthesis.

LpxK is one of six orthologs of the E. coli Kdo₂-lipid A biosynthetic pathway identified by sequence homology in A. aeolicus, the missing enzymes being LpxH, LpxL, and LpxM (24). Based on analysis of Aquifex Lipid A (25, 26), the LpxK substrate in A. aeolicus is predicted to be analogous to E. coli DSMP with a few exceptions. The 3- and 3′-hydroxymyristoyl groups on the disaccharide backbone are N-linked and the 2- and 2′-positions are occupied by hydroxypalmitoyl moieties instead of hydroxymyristate. Galacturonic acid moieties decorate the 1- and 4′-positions of Aquifex Lipid A, but are likely added after transport to the outer membrane has occurred, as in various Rhizobium species (1, 2, 27).

We here report the crystal structures of both apo- and ADP/Mg²⁺-bound A. aeolicus LpxK solved to 1.9 Å and 2.2 Å, respectively. Along with analysis of Walker motif point mutants, the structures have provided insight into the specificity of this unique kinase, have helped illuminate how the ubiquitous phosphoryl transfer function adapts to interfacial catalysis at the membrane, and increase LpxK's chance of becoming an inhibitor target, potentially improving our arsenal against the devastating infectious diseases caused by Gram-negative bacteria (28).

Results

Crystal Structure and Overall Topology of Apo LpxK.

The initial crystal structure of A. aeolicus LpxK was solved using a single anomalous dispersion dataset of selenomethionine-derivatized protein in a primitive tetragonal space group (SI Materials and Methods). Primitive orthorhombic crystals of the same apo form of LpxK were later obtained, which improved the resolution to 1.9 Å. The model includes the entire polypeptide chain and was refined to R_work and R_free values of 17.2% and 20.0%, respectively (Table S1).

LpxK is composed of two domains, both adopting a Rossmann-like α/β/α sandwich fold, that are connected by two twisted, antiparallel β-strands (Fig. 2A). The larger N-terminal domain (residues 1–206, 298–315) consists of an internal nine-stranded parallel β-sheet (with the exception of β3 and β13) flanked by seven α-helices (α1–α6, α11), a 3₁₀-helix (α7), and intervening loops. The smaller C-terminal domain (residues 210–294) consists of a six-stranded parallel β-sheet (with the exception of β8) surrounded by two α-helices (α9 and α10), a 3₁₀-helix (α8), and loop regions. The strand order of the β-sheets is β13, β8, β7, β6, β1, β5, β2, β4, β3, and β9, β8, β13, β12, β10, β11 for the N- and C-terminal domains, respectively, with β8 and β13 linking the two domains. These features distinguish LpxK from other P-loop kinases, which usually consist of single Rossmann-fold domain with an internal five-stranded β-sheet containing the strand order β2, β3, β1, β4, β5 (19–22). The active site Walker A/P-loop and Walker B motifs are located on the N-terminal domain in a pocket formed between the two domains (Fig. 2A). As found in the majority of P-loop kinases, the Walker A motif is located between the first β-strand (β1) and the following α-helix (α2) of the polypeptide chain (19–22). Even though all residues of LpxK were successfully modeled, loops encompassing residues 74–79 (L1), 234–236 (L2), and 261–262 (L3) displayed weaker density, with an increased average B-factor (128.0 Ų compared with 43.8 Ų for the entire polypeptide), indicating a higher degree of flexibility in these regions (Fig. 2A).

Fig. 2.

Overall structure and surface properties of apo A. aeolicus LpxK. (A) Cartoon diagram of A. aeolicus LpxK tracing the polypeptide chain from the N (blue) to C (red) terminus. The Walker A/P-loop and Walker B motifs residing between the two domains are labeled. Secondary structure and select loop regions denoted by “L” are labeled. (B) A surface representation of LpxK in which each residue was assigned a color based on its hydrophobic (green) or hydrophilic (white) nature (34). The orientation is preserved from A. A bound Hepes molecule is shown as gray sticks along with corresponding simulated annealing omit electron density (blue mesh) calculated with coefficients F_o − F_c, contoured at 2.5 σ. Potential hydrogen bonds are indicated with dashed lines. The Walker B residues D138 and D139 are also included for reference. (C) The electrostatic surface of LpxK as calculated by the Adaptive Poisson-Boltzmann Solver (35). The electrostatic potential is scaled from −3.0 (red) to +3.0 (blue) kT/e. Figures were rendered using PyMOL (41).

The two-domain architecture of LpxK resembles known structures of Gram-negative diacylglycerol kinases YegS and DgkB; however, the C-terminal domains of these enzymes primarily consist of β-sheets and the interdomain linkers are loop segments (29, 30). The closest known structural homolog of LpxK is HypB (15.9% sequence identity and 27.5% similarity), a GTPase involved in nickel processing that resembles LpxK’s N-terminal domain (Fig. S2A) (31, 32). Strikingly, the backbone rmsd between HypB and the N-terminal domain of LpxK is 2.54 Å (33) since the strand order and placement of some helices surrounding the sheets is relatively conserved between the two proteins. However, HypB is one strand shorter at each end of its long central β-sheet and lacks the N-terminal helix of LpxK.

During purification of A. aeolicus LpxK, it was noted that the protein was unable to be separated from membranes without detergent treatment (SI Materials and Methods). Although we cannot completely rule out the possibility that the N-terminal helix of A. aeolicus LpxK (residues 6–26) is transmembrane in vivo, we hypothesize, based on topology prediction (23) and multiple hydrogen bonding interactions between various side chains of the helix (R3, S10, Y13, R20, N21, D25) and the core of the N-terminal domain that the position of this helix in the structure is not merely a consequence of crystal packing (Fig. S3). To illuminate how LpxK may associate with the inner membrane, the surface hydrophobicity of the enzyme was mapped, revealing the outward face of the N-terminal helix to be hydrophobic in nature mainly because of contributions from the side chains of residues L2, L6, F9, L12, I16, F19, L23, and F28 (Fig. 2B and Fig. S3) (34). This surface could have an analogous function to the putative N-terminal transmembrane domain of E. coli LpxK, anchoring A. aeolicus LpxK to the membrane. To shed further light on this matter, the electrostatic surface of A. aeolicus LpxK was calculated using the Adaptive Poisson-Boltzmann Software (Fig. 2C) (35). Interestingly the N-terminal domain surface, especially adjacent to the hydrophobic face of the N-terminal helix, is significantly more cationic than the C-terminal domain surface. Combined, these observations suggest that the N-terminal domain is responsible for LpxK’s membrane affinity via the hydrophobic lower surface of the N-terminal helix and cationic residues surrounding this region, enabling the enzyme to interact strongly with the anionic phospholipids of the inner membrane. Positive difference electron density was observed in a basic patch of the N-terminal domain. Because the crystallant contained 100 mM Hepes and the contour of the density resembled its chemical structure, a molecule of Hepes was modeled and successfully refined. The sulfate group extension off the piperazine core is bound to the side chains R72, R119, and H143 (Fig. 2B). Although this binding event is probably not physiological, these residues are highly conserved among orthologs of LpxK (Fig. S1) and may be involved in binding the DSMP substrate's 1-phosphate.

Crystal Structure of ADP/Mg²⁺-Bound LpxK.

To elucidate nucleotide-binding residues of A. aeolicus LpxK, the structure of ADP/Mg²⁺-bound enzyme was solved by molecular replacement to a resolution of 2.2 Å, resulting in R_work and R_free values of 17.8% and 21.3%, respectively (Table S1). Electron density was not well-defined for the first 8 aa of the N-terminus, and they were not included in the model.

ADP and Mg²⁺ reside in the pocket formed between the N- and C-terminal domains (Fig. 3A). The P-loop surrounds the β-phosphate of ADP, and the adenosine moiety is bound between the second helix of the smaller C-terminal domain and the interdomain linker (Fig. 3B and Fig. S4). A global alignment of the apo- and ADP/Mg²⁺-bound forms of LpxK using the DynDom server (36) reveals a 25° hinge motion in which the C-terminal domain closes around the nucleotide, highlighting the significant conformational change that accompanies ligand binding (Fig. 3A). Residues 206–212 and 294–297 make up the “hinge” regions, which lie on the β-sheet linker at the base of the C-terminal domain. As a consequence of binding, loops L2 and L3 of the C-terminal domain are better ordered, with decreased B-factors (48.4 Ų and 54.6 Ų, respectively) compared with the apo LpxK structure (120.3 Ų and 145.3 Ų) (Fig. 3 A and B). Loops L2 and L3 are involved in adenosine/α-phosphate interactions and Mg²⁺ coordination, respectively. The stabilization of these loops in conjunction with ligand binding indicates they are flexible and upon domain closure are locked into place.

Fig. 3.

Domain closure induced by binding of ADP/Mg²⁺. (A) Shown is the globally aligned cartoon illustrations of apo- (gray) and ADP/Mg²⁺-bound (green) A. aeolicus LpxK with ADP (cyan sticks), Mg²⁺ (magenta sphere), and Cl⁻ (yellow sphere). Corresponding simulated annealing omit electron density (blue mesh) was calculated with coefficients F_o − F_c, contoured at 4 σ. Relative positions of helix α9 and loops L2 and L3 are indicated. When rotated 90°, the overlay more clearly illustrates the 25° hinge motion of the C-terminal domain as it closes around the ligands. (B) A stereoview close-up of the ADP/Mg²⁺-binding pocket containing ADP (cyan sticks) and Mg²⁺ (magenta sphere). Hydrogen bonds are indicated with dashed lines. Select waters are shown (red spheres). (C) Schematic overview of the nucleotide/magnesium binding site. Hydrogen bonds are indicated with green dashes and their distances indicated. Select waters are shown in blue and the orange arrow indicates the relative distance between K51 and D139.

C-terminal domain movement positions residue Q240 to hydrogen bond to the N6 and N7 nitrogen atoms of the adenine ring (Fig. 3C and Fig. S4). Residues L235, F241, T278, and P279 of the C-terminal wall of the binding pocket provide van der Waals contacts with adenosine. Residues R206, F208, and F296 on the domain linker β-sheet contribute additional interactions with the opposite face of the nucleoside and the back wall of the binding pocket is formed by V244 and L294. The ribose of ADP is secured by hydrogen bonds to the backbone amide nitrogen atom of K280 and the S53 hydroxyl group, and is within van der Waals distance of M56 and L103. Spherical electron density was present between the ribose of ADP and S53, which refined well as a chloride ion, likely a consequence of including 130 mM NaCl in the crystallization drop solution.

The pyrophosphate moiety of ADP is held in place by an intricate network of hydrogen bonds with backbone amide nitrogen atoms, various side chains, and water molecules (Fig. 3 B and C). Interestingly, the phosphate groups are found in a strained conformation with two dihedral angle outliers between O5′-Pα and O3A-Pβ (37), indicating that this geometry may reflect the catalytically competent ATP conformer necessary for phosphate transfer. The α-phosphate group is secured by interactions with the side chain of K280 and the backbone nitrogen atom of G236. Compared with the apo structure, the P-loop is found in a nearly identical overall conformation except for G47, which is flipped, bringing the backbone nitrogen atoms of G48 and S49 closer to the β-phosphate of ADP, thereby stabilizing its negative charge. The backbone amides and side chains of P-loop residues S49, K51, and T52 hydrogen bond to the β-phosphate, and the carboxylate of E100 and the β-phosphate of ADP are monodentate ligands of Mg²⁺ (Fig. 3 B and C). The octahedral coordination shell of the ion is completed by water molecules, which are held in place by interactions with the side chains of D260, K51, H261, and E100.

Point Mutagenesis of Walker Motif Residues.

To investigate the importance of various conserved Walker motif residues of LpxK, a number of alanine point mutants were created and partially purified through solubilization from membranes. A standard thin-layer chromatography (TLC) radioassay, slightly modified from previous conditions (11) (SI Materials and Methods), was performed in triplicate for each mutant and activity assessed (Table 1). To determine how critical P-loop flexibility is for LpxK binding and catalysis, alanine point mutants of residues G47, G48, and G50 were generated and demonstrated 2-fold, 50-fold, and near complete reduction of specific activity to background levels, respectively. P-loop alanines, especially at positions closest to the proposed catalytic residues at the base of the P-loop, likely displace the backbone amides from their favored orientation and interfere with the optimal hydrogen bonding network necessary for catalysis.

Table 1.

Activity of Walker motif point mutants

LpxK mutant	Specific activity compared with WT*
G47A	43.0 ± 5.1%
G48A	1.8 ± 0.4%
G50A	< 0.1%
K51A	< 0.1%
T52A	< 0.1%
S53A	9.9 ± 4.1%
D138A	< 0.1%
D139A	< 0.1%

*Activity of vector control subtracted before calculation; specific activity of WT = 1.54 μmol⋅min^–1⋅mg^–1.

Conserved Walker A motif residues K51, T52, and S53 were also mutagenized to help determine their roles in LpxK activity. K51, an absolutely conserved residue among kinases of the P-loop containing NTP hydrolase superfamily, is thought to contribute to γ-phosphate stabilization during transfer (19, 20). The crystal structures presented herein suggest T52 may be crucial for positioning the β-phosphate through hydrogen bonding. Both K51A and T52A mutants lowered specific activity to near background levels. The S53A mutant resulted in a 10-fold activity reduction, indicating that its interaction with the ribose hydroxyl group is very important, but not entirely critical for the overall activity of LpxK. Alanine substitutions of D138 and D139 were also generated to analyze the importance of these conserved Walker B motif moieties. Both mutations reduced activity to near background levels. These residues could have a number of functions including lipid substrate binding, coordinating the catalytic Mg²⁺ in a chemically active conformation, or acting as a catalytic base by deprotonating the 4′-hydroxyl group of DSMP (20).

Discussion

The crystal structures of both the apo- and ADP/Mg²⁺-bound forms of LpxK have allowed us to gain valuable insight into the specificity and mechanism of this kinase within the lipid A biosynthetic pathway and novel antibiotic target. Among other bacterial lipid kinases, only YegS, DgkB, and DgkA have known structures, but these enzymes belong to the DAGK superfamily, lacking the Walker motifs that distinguish the P-loop NTPases (29, 30, 38). Of these three kinases, only DgkA, a functional trimer with nine transmembrane helices, is membrane-bound like LpxK.

Taken overall, the N-terminal α/β/α domain most resembles other members of the P-loop containing NTP hydrolase superfamily, especially HypB (19, 20, 31) (Fig. S2A). LpxK is larger than the majority of these other family members and its possession of a second α/β/α domain involved in adenosine binding is a significant and unique deviation from typical topologies. Bringing ATP close to the anionic phospholipids of the inner membrane may require additional interactions provided by the LpxK C-terminal domain to lower the free energy of nucleotide binding enough to overcome charge repulsion at the membrane surface. In most P-loop kinases, additions to the core α/β/α domain tend to be more subtle, often consisting of only a few secondary-structure elements, and these proteins implement surface clefts and dimerization to bind ATP and exclude water from the active site (21, 22, 31). Our crystal structures, along with gel-filtration analysis of the purified enzyme (Fig. S5B), indicate that LpxK is a functional monomer. The additional domain of LpxK likely bestows specificity for ATP over other NTPs, as previously observed for E. coli LpxK (10), and may perform the water-exclusion function as well. The fact that LpxK’s closest structural homolog is the GTPase HypB and not other P-loop kinases indicates that the ancestral LpxK likely differentiated itself from other P-loop NTPases early on in the evolutionary history of this superfamily (20, 31).

During purification, we noted that A. aeolicus LpxK cannot be separated from membranes by salt treatment alone and required detergent, an unusual property for a protein without any obvious transmembrane segments, as determined by hydropathy (23) and structural analysis. Although the N-terminal helix has strong homology to its E. coli counterpart, which is predicted via hydropathy analysis to span the membrane, we believe that our crystal structures reflect the helix’s physiological position on the core of the N-terminal domain provided the number of contacts observed between the amphipathic face of the helix and adjacent domain elements L4, L5, and α7 (Figs. S1 and S3). It is possible that the predicted transmembrane domain of E. coli LpxK is in reality analogous to A. aeolicus LpxK’s N-terminal helix, being only partially buried in the membrane. Future studies could provide more insight into this issue, including determining if and how A. aeolicus LpxK functions without the N-terminal helix, or probing whether charge interactions are the main driving force for membrane affinity through studies of alkali-resistance to membrane integration.

The crystal structure’s lack of a large hydrophobic pocket as found in other structures with bound lipid A (6), combined with the knowledge that DSMP is tetra-acylated, supports the hypothesis that LpxK may only partially extract its substrate from the inner membrane to perform catalysis. Based on LpxK’s calculated surface charge, surface hydrophobicity, and the directionality of the ADP phosphates in the structure, the DSMP binding surface is likely on the underside of the N terminus (Fig. 4), where residues of loops L1 (residues 71–74), L4 (138–143), and L5 (165–172) are highly conserved among LpxK orthologs (Fig. 2 and Fig. S1). LpxK may use charged residues on these loops to bind the disaccharide portion of DSMP and then secure the acyl chains using the N-terminal α-helix as a clamp. The tightly bound Hepes molecule in the 1.9 Å resolution apo-enzyme structure may provide insight into the location of part of the DSMP binding site because the conserved basic residues binding the Hepes sulfate could be correctly positioned to coordinate the 1-phosphate of DSMP (Fig. 2B).

Fig. 4.

Hypothetical membrane association of LpxK. Based on surface charge and hydrophobicity analysis, we predict that LpxK attaches to the membrane through partial burial of its N-terminal helix, allowing the enzyme to partially extract DSMP from the membrane to perform catalysis. Illustrative positions of the ATP and partially extracted DSMP substrates before catalysis are indicated. Binding and catalysis occurs in the pocket formed between the two domains near the inner face of the N-terminal helix.

The superposition of the apo- and ADP/Mg²⁺-bound LpxK structures reveal a closure of the C-terminal domain around the ADP product in the active site (Fig. 3A). LpxK may remain in the “open” conformation and mold itself around its substrates upon binding, or alternatively bind its substrates followed by domain closure to adopt the catalytically active conformer, bringing the necessary residues within proximity of the ATP, DSMP, and the Mg²⁺ ion (Fig. 3 B and C). Outside the P-loop, residues S53, R206, G236, Q240, and K280 appear to have a significant role in ATP binding because they are hydrogen bonded to the adenosine, ribose hydroxyl groups, or the α-phosphate of ADP.

Our other mutational studies have shown that conserved residues of the Walker A/P-loop motif are critical for LpxK function, as has been observed for other P-loop kinases (39, 40). Mutation of three P-loop glycine residues inhibited activity significantly, especially those nearer the conserved P-loop lysine. This result indicates that conformational flexibility of the glycine-rich loop is important for proper substrate binding and catalysis, as is generally accepted for nearly all enzymes that catalyze phosphoryl transfer chemistry, including P-loop kinases (19, 20). Sizeable conformational movement of the entire loop is suggested in the 1.9 Å resolution apo structure, where weak positive difference electron density is observed sitting adjacent to the currently modeled P-loop position. We hypothesize that this density indicates the presence of a rarely populated alternate P-loop conformation. Without ligand, the P-loop may exist in a more dynamic state and is “pinned” down in the active conformation only upon nucleotide binding.

We believe that our model of ADP/Mg²⁺-bound LpxK, namely the position of Mg²⁺ bound exclusively to ADP and the E100 carboxylate, likely represents a postcatalysis conformation of the enzyme. Our reasoning for this conclusion is twofold. First, only LpxK incubated with ATP/Mg²⁺ crystallized with ADP and Mg²⁺ in the active site, indicating that ATPase activity may have been achieved in the absence of DSMP during crystallization. Second, in other P-loop kinases, the Walker A hydroxyl-containing residue (T52 in LpxK), which follows the P-loop lysine, is often found to be directly involved in Mg²⁺ binding (19–22) and is sometimes assisted by a Walker B carboxylate (19, 20, 31, 40). The precatalytic conformation of ATP/Mg²⁺-bound LpxK may be well-represented by the GTPγS/Mg²⁺-bound model of LpxK’s closest structural homolog HypB, in which the γ-phosphate occupies an analogous position to the Mg²⁺ ion of the LpxK structure (Fig. S2B) (31). In Methanococcus jannaschii HypB, the side chains of T47, D75, and E120 (structurally analogous to T52, E100, and D138 of A. aeolicus LpxK) directly coordinate the Mg²⁺ ion. LpxK complexed with ATP/Mg²⁺ in a precatalytic state may adopt a similar conformation with the Mg²⁺ bound in the W₅ position of the current ADP/Mg²⁺ LpxK model and the γ-phosphate bound to E100 (Fig. 3 B and C). In this case, the nearby H261 may act as a catalytic base because of strong hydrogen-bonding interactions with adjacent D99. Alternatively, because of its proximity the K51 and the bound phosphates of ADP, Walker B residue D139 may play this role (Fig. 3C). Whether the aspartic acid residues of the Walker B motif help coordinate Mg²⁺, position the DSMP substrate, or act in catalysis, we have demonstrated that both D138 and D139 are essential for activity.

Overall, the structural characterization of both the apo- and ADP/Mg²⁺-bound forms of LpxK represent an important first step in understanding this unique member of the P-loop–containing NTP hydrolase superfamily. The insights gained from these structures have provided a glimpse into how a common phosphorylation reaction must adapt to interfacial catalysis on the membrane. Mutagenesis studies of various conserved residues have begun to reveal the mechanism of substrate binding and catalysis. Furthermore, the crystal structure of LpxK in complex with ADP/Mg²⁺ should aid the development of novel antibiotics that target this essential kinase of Gram-negative bacteria. The number of available kinase inhibitor libraries makes the selective inhibition of LpxK a real possibility for the development of potent antimicrobials.

Materials and Methods

Full details of the methods used can be found in SI Materials and Methods. Briefly, both GST-tagged and nontagged A. aeolicus LpxK could be purified and crystallized in apo- and ADP/Mg²⁺-bound forms by sitting-drop vapor diffusion at 20 °C. The structure was solved using the phases generated from selenomethionine-derivatized LpxK extended with higher-resolution data. Alanine point mutants were generated and their activities characterized based using a TLC-based radioassay slightly modified from previous conditions (11). See Tables S2–S4 for primers used in this study, strains and plamids, and a representative purification table for A. aeolicus LpxK purified via an N-terminal GST tag, respectively.