Structure- and Ligand-Dynamics-Based Design of Novel Antibiotics Targeting Lipid A Enzymes LpxC and LpxH in Gram-Negative Bacteria

CONSPECTUS:

Bacterial infections caused by multi-drug-resistant Gram-negative pathogens pose a serious threat to public health. Gram-negative bacteria are characterized by the enrichment of lipid A-anchored lipopolysaccharide (LPS) or lipooligosaccharide (LOS) in the outer leaflet of their outer membrane. Constitutive biosynthesis of lipid A via the Raetz pathway is essential for bacterial viability and fitness in the human host. The inhibition of early-stage lipid A enzymes such as LpxC not only suppresses the growth of Pseudomonas aeruginosa, Klebsiella pneumoniae, Enterobacter spp., and other clinically important Gram-negative pathogens but also sensitizes these bacteria to other antibiotics. The inhibition of late-stage lipid A enzymes such as LpxH is uniquely advantageous because it has an extra mechanism of bacterial killing through the accumulation of toxic lipid A intermediates, rendering LpxH inhibition additionally lethal to Acinetobacter baumannii. Because essential enzymes of the Raetz pathway have never been exploited by commercial antibiotics, they are excellent targets for the development of novel antibiotics against multi-drug-resistant Gram-negative infections.

This Account describes the ongoing research on characterizing the structure and inhibition of LpxC and LpxH, the second and fourth enzymes of the Raetz pathway of lipid A biosynthesis, in the laboratories of Dr. Pei Zhou and Dr. Jiyong Hong at Duke University. Our studies have elucidated the molecular basis of LpxC inhibition by the first broad-spectrum inhibitor, CHIR-090, as well as the mechanism underlying its spectrum of activity. Such an analysis has provided a molecular explanation for the broad-spectrum antibiotic activity of diacetylene-based LpxC inhibitors. Through the structural and biochemical investigation of LpxC inhibition by diacetylene LpxC inhibitors and the first nanomolar LpxC inhibitor, L-161,240, we have elucidated the intrinsic conformational and dynamics difference in individual LpxC enzymes near the active site. A similar approach has been taken to investigate LpxH inhibition, leading to the establishment of the pharmacophore model of LpxH inhibitors and subsequent structural elucidation of LpxH in complex with its first reported small-molecule inhibitor based on a sulfonyl piperazine scaffold.

Intriguingly, although our crystallographic analysis of LpxC- and LpxH-inhibitor complexes detected only a single inhibitor conformation in the crystal lattice, solution NMR studies revealed the existence of multiple ligand conformations that together delineate a cryptic ligand envelope expanding the ligand-binding footprint beyond that observed in the crystal structure. By harnessing the ligand dynamics information and structural insights, we demonstrate the feasibility to design potent LpxC and LpxH inhibitors by merging multiple ligand conformations. Such an approach has enabled us to rationally design compounds with significantly enhanced potency in enzymatic assays and outstanding antibiotic activities in vitro and in animal models of bacterial infection. We anticipate that continued efforts with structure and ligand dynamics-based lead optimization will ultimately lead to the discovery of LpxC- and LpxH-targeting clinical antibiotics against a broad range of Gram-negative pathogens.

Graphical Abstract

INTRODUCTION

The global burden of multi-drug-resistant bacterial infections, particularly in hospitals with large numbers of vulnerable patients, is a major threat to public health.^4,5 Our diminished ability to treat multi-drug-resistant bacterial infections⁶ also weakens our response to the current COVID-19 pandemic, where COVID-19 patients with secondary bacterial infections have been associated with worse treatment outcomes and mortality despite antimicrobial therapies.⁷ Because of resistance, fluoroquinolones, aminoglycosides, tetracyclines, trimethoprim-sulfamethoxazole, and aztreonam are no longer recommended for empiric therapy.^6,8–10 Many β-lactam antibiotics are no longer effective due to resistance from bacteria harboring extended-spectrum β-lactamases, New Delhi metallo-β-lactamase (NDM-1), and Klebsiella pneumoniae carbapenemase (KPC).⁶ Resistance to colistin, a last-line-of-defense antibiotic, has also emerged due to the spread of the mcr-1 and mcr-like genes.^11–13 In light of these issues, the Emerging Infections Network cited the lack of treatments for Gram-negative bacteria to be the greatest area of unmet need,⁴ a view echoed by a 2017 WHO report.⁵ As a result, fundamentally new classes of agents for which no compound-specific resistance currently exists are ideally positioned to overcome this emerging public health crisis.

LIPID A ENZYMES AS NOVEL ANTIBIOTIC TARGETS IN GRAM-NEGATIVE PATHOGENS

Gram-negative bacteria are characterized by the enrichment of lipopolysaccharide (LPS) or lipooligosaccharide (LOS) in the outer leaflet of their outer membrane (Figure 1A). The membrane anchor of LPS and LOS is a unique saccharolipid known as lipid A. Lipid A is the active component of endotoxin that causes Gram-negative septic shock during bacterial infection, and its constitutive biosynthesis is required for bacterial viability and fitness in the human host.^14–16

Figure 1.

Lipid A biosynthesis in E. coli. (A) Schematic illustration of Gram-negative bacterial membranes. (B) Raetz pathway of lipid A biosynthesis. Small-molecule inhibitors have been reported for LpxA, LpxC, LpxD, and LpxH (colored in pink). LpxH is found in nearly all of the clinically important Gram-negative pathogens, whereas its functional orthologs, LpxI and LpxG (colored in green), are narrowly distributed in α-proteobacteria and Chlamydiae, respectively.

Lipid A biosynthesis is most thoroughly characterized in E. coli, involving nine enzymes in the Raetz pathway (Figure 1B).^14,15 It starts with the acylation of UDP-GlcNAc catalyzed by LpxA.¹⁷ The UDP-GlcNAc acylation reaction in E. coli is reversible and unfavorable;¹⁸ therefore, the subsequent deacetylation of UDP-(3-O-acyl)-GlcNAc catalyzed by LpxC is considered the committed step of lipid A biosynthesis.¹⁹ After deacetylation, a second (R)-3-hydroxymyristate chain is added by LpxD to make UDP-2,3-diacylglucosamine (UDP-DAGn).^20,21

The fourth and fifth enzymes in the pathway, LpxH and LpxB,^22,23 are membrane-associating enzymes. LpxH cleaves the pyrophosphate moiety of UDP-DAGn to make 2,3-diacyl-GlcN-1-P (lipid X). LpxB next generates β,1′-6-linked disaccharide 1-phosphate (DSMP) by condensing a second molecule of UDP-DAGn with lipid X. Although the chemical transformation of the LpxH reaction is conserved among Gram-negative bacteria, enzyme LpxH itself is replaced with a functional ortholog, LpxI in α-proteobacteria, which has no structural or mechanistic relationship to LpxH,^24,25 or LpxG in Chlamydiae, which falls within the same protein family but shares very little sequence identity with LpxH (Figure 1B).²⁶ These functional orthologs never coexist in the same bacterium.

The rest of the lipid A biosynthetic enzymes, LpxK, KdtA, LpxL, and LpxM, are inner membrane proteins.²⁷ LpxK phosphorylates the 4′-position of DSMP to generate intermediate lipid IV_A.²⁸ Next, two Kdo residues are incorporated by bifunctional enzyme KdtA.²⁹ The last steps of E. coli lipid A biosynthesis involve the addition of secondary laurate and myristate chains by LpxL and LpxM,^30–32 which generate Kdo₂-lipid A, a moiety sufficient to maintain bacterial viability in most Gram-negative organisms. Because disrupting lipid A biosynthesis is bactericidal for most Gram-negative bacteria, developing novel antibiotics targeting essential lipid A enzymes, such as LpxC, has become a major research focus for industrial groups and academic laboratories for the last two decades.^33,34 Here we summarize the ongoing research on developing novel antibiotics targeting LpxC and LpxH in the Zhou and Hong laboratories at Duke University.

A BRIEF HISTORY OF LPXC

In 1969, Normark, Boman, and Matsson isolated a mutant E. coli strain (D22) that changed the morphology of the parental strain after chemically induced mutagenesis.³⁵ This strain has the tendency to form chains that can be transformed into filaments (Figure 2). Because such a phenotype reflected the change in the bacterial envelope, the mutated gene was denoted as envA.³⁵ The envA mutation profoundly sensitized cells to a variety of antibiotics,³⁵ and its genuine function as a deacetylase in lipid A biosynthesis was established in 1995.¹⁹ envA was subsequently named lpxC.¹⁹ The validation of lpxC as an essential gene³⁶ and the knowledge that the lpxC/envA mutant compromised bacterial membrane integrity and sensitized E. coli to a variety of antibiotics highlighted the many benefits of inhibitors disrupting lipid A biosynthesis: these compounds not only would serve as standalone bactericidal antibiotics but also can be used to enhance the effectiveness of other therapeutics.

Figure 2.

Mutation in envA/lpxC that caused cell morphology changes in E. coli. Cells of the D21 parental and D22 envA/lpxC mutant strains are shown in (A) and (B), respectively. Adapted with permission from ref 35. Copyright (1969) American Society for Microbiology.

One year after the elucidation of the functional role of LpxC,¹⁹ Onishi and co-workers at Merck reported the first nanomolar inhibitor of LpxC, L-161,240.³⁷ L-161,240 displayed a minimum inhibitory concentration (MIC) of ~1 μg/mL against E. coli in vitro and rescued mice from a lethal dose of E. coli infection in vivo, establishing LpxC as a viable antibiotic target.³⁷ However, L-161,240 was completely inactive against wild-type (WT) P. aeruginosa (MIC > 100 μg/mL), limiting the potential of its clinical application.³⁷ The discovery of CHIR-090,³⁸ a potent LpxC inhibitor capable of killing WT P. aeruginosa and many other Gram-negative pathogens,³⁹ profoundly changed the narrow-spectrum perception of LpxC inhibitors, and LpxC has since become a mainstream novel antibiotic target pursued by academic laboratories as well as pharmaceutical and biotechnology companies (comprehensively reviewed in refs 33 and 34).

STRUCTURE OF LPXC

Despite the discovery of several nanomolar inhibitors of LpxC in the early 2000s, the lack of structural information has severely hindered the effort to optimize LpxC inhibitors. As a first step toward successful drug design, our laboratory determined the first structure of Aquifex aeolicus LpxC in complex with substrate analog inhibitor TU-514 by solution NMR (Figure 3A,B).^40,41 This coincided with the elucidation of the crystal structure of the same enzyme bound with a fatty acid by David Christianson’s laboratory.⁴² These structures revealed a unique sandwich fold of LpxC unlike any human amidases: it consists of two domains each with a layer of α-helices packing against a layer of β-sheets. The two domains come together by packing the helices inside and the β-sheets outside, resembling a hotdog sandwich (Figure 3B).

Figure 3.

Ligand binding features revealed by the structure of the A. aeolicus LpxC/TU-514 complex. (A) Chemical structure of TU-514. (B) Structure of the A. aeolicus LpxC/TU-514 complex. LpxC is shown in the cartoon model, with domains I and II colored in dark green and pale green, respectively. Inserts I and II are colored in coral and blue, respectively. (C) Structural features of the substrate-binding site. A. aeolicus LpxC residues forming the hydrophobic passage, basic path, and hydrophobic patch are listed. Adapted with permission from ref 41. Copyright (2005) American Chemical Society.

LpxC is a zinc-dependent metalloamidase, though Fe²⁺ also supports enzymatic activity.⁴³ The active site is located between two domains, with the catalytic metal ion coordinated by two histidine residues and an aspartate residue. The fourth position is occupied by the hydroxamate group of substrate analog inhibitor TU-514.^40,41 An analysis of the LpxC/TU-514 complex revealed distinct features for inhibitor interaction (Figure 3C): the catalytic zinc ion is coordinated by a metal chelating group that is often described as the warhead of the LpxC inhibitors; conserved residues form two patches in the active site—one hydrophobic and one basic—that constitute the binding surfaces for the headgroup of LpxC inhibitors. Finally, LpxC contains two unique insertion regions (inserts I and II) that surround the active site, with insert II forming an L-shaped hydrophobic passage accommodating the tail group of LpxC inhibitors.

STRUCTURE OF THE LPXC-CHIR-090 COMPLEX AND THE EFFORT TO EXPAND THE SPECTRUM OF LPXC INHIBITION

In 2007, we reported the structure of the A. aeolicus LpxC/CHIR-090 complex, revealing the molecular details of the first antibiotic bound to LpxC (Figure 4A).¹ The structure provided a satisfactory explanation of the outstanding potency of CHIR-090: its tail group with a kink at the methylene group between the terminal morpholine group and the distal phenyl ring naturally conforms to the L-shaped hydrophobic passage, and its threonyl hydroxamate headgroup efficiently picks up three key interactions: zinc chelation via the hydroxamate group, van der Waals (vdW) interaction with the hydrophobic patch via the threonyl methyl group, and polar interaction with the basic patch via the threonyl hydroxyl group.

Figure 4.

Molecular basis of LpxC inhibitor specificity. (A) Structure of the A. aeolicus LpxC/CHIR-090 complex. (B) The presence of a critical glycine residue in insert II is associated with the LpxC sensitivity to CHIR-090 inhibition. (C) Substituting the glycine residue with serine or alanine creates steric clashes with the distal phenyl ring of CHIR-090. (D) Replacing the distal phenyl ring of CHIR-090 with the acetylene group enhances the antibiotic activity against E. coli (W3110), E. coli with its genomic lpxC replaced with that of R. leguminosarum (W3110RL), and P. aeruginosa (PAO1). (E) Structures of LpxC enzymes from E. coli (coral), P. aeruginosa (blue), and A. aeolicus (green) in complex with LPC-009 (2) reveal distinct structural features of individual LpxC orthologs. The relative movements of the insert II helix and the insert I loop are labeled. (F) Binding of L-161,240, but not LPC-009 (2), induces a backbone flipping of the insert I loop of E. coli LpxC, with the side chain of C63 swinging out of the active site. Complexes of E. coli LpxC/L-161,240 and E. coli LpxC/LPC-009 (2) are shown in green and orange, respectively. The side chain of residue C63, the backbone of residues in the insert I loop, and inhibitors are shown as sticks. Zinc ions are shown as spheres. Panels (A) and (B) were adapted with permission from ref 1. Copyright (2007) National Academy of Sciences. Panels (C–E) were adapted with permission from ref 44. Copyright (2011) Elsevier. Panel (F) was adapted with permission from ref 45. Copyright (2013) American Chemical Society.

Although CHIR-090 displayed potent inhibition against a variety of Gram-negative bacteria and LpxC enzymes, its inhibition was 600-fold weaker against the Rhizobium family of LpxC enzymes.¹ It also displayed weaker activity against Acinetobacter and Francisella LpxC enzymes. A knowledge of the CHIR-090 binding mode enabled us to divide LpxC enzymes into two different categories (Figure 4B). We found that the presence of a glycine residue in the insert II passage (i.e., G198 in A. aeolicus LpxC or G210 in E. coli LpxC) accommodated the distal phenyl ring of CHIR-090 and rendered LpxC enzymes sensitive to CHIR-090 inhibition, whereas substituting the glycine with a serine or alanine residue as seen in LpxC enzymes from R. leguminosarum, A. baumannii, or F. tularensis reduced the CHIR-090 inhibition due to the clash of the serine or alanine side chain with the distal phenyl ring of CHIR-090 (Figure 4B,C). Indeed, replacing S214 with glycine (S214G) rendered the Rhizobium enzyme 100-fold more susceptible to CHIR-090 inhibition.¹

We reasoned that if resistance to CHIR-090 was caused by vdW repulsions with its distal phenyl ring (Figure 4C), then replacing it with a small functional group should mitigate such an effect. Indeed, although LPC-004 (1), lacking the morpholine group from CHIR-090, had activity similar to that of CHIR-090, diacetylene-based LpxC inhibitor LPC-009 (2, Figure 4D) not only effectively reduced the MIC values against WT and R. leguminosarum lpxC knock-in E. coli strains but also displayed enhanced potency against P. aeruginosa (Figure 4D),⁴⁴ thus improving and expanding the antibiotic activity profile of CHIR-090.

The discovery of broad-spectrum inhibitor LPC-009 (2) has made it possible to investigate the species-specific structural features of LpxC enzymes. By cocrystallizing LpxC enzymes from E. coli, P. aeruginosa, and A. aeolicus with the same inhibitor, LPC-009 (2), we showed that different LpxC enzymes have distinct shapes of the hydrophobic substrate binding passage reflected by dissimilar orientations of the insert II helix forming the dome of the passage (Figure 4E).⁴⁴ These features likely reflect the difference in the acyl chain length of natural LpxC substrates and are important factors to consider for designing broad-spectrum LpxC inhibitors. Likewise, our structural comparison of E. coli LpxC bound to L-161,240 and LPC-009 (2) revealed that the insert I loop of E. coli LpxC undergoes backbone flipping to accommodate the altered position of the carbon atom next to the hydroxamate headgroup of L-161,240 due to its opposite stereospecific configuration in comparison with LPC-009 (2) (Figure 4F), suggesting that the intrinsic conformational dynamics in E. coli LpxC may underline its ability to bind distinct classes of LpxC inhibitors promiscuously.⁴⁵

DRUG DESIGN FROM THE CRYPTIC LIGAND ENVELOPE

Because LpxC inhibitors containing the diacetylene group display a broad spectrum of enzymatic inhibition, our effort has focused on the optimization of the headgroup of LpxC inhibitors.^45–47 Among these, the most successful design has arisen from a combined structural and dynamics analysis of the cryptic inhibitor envelope of LpxC.²

Solution NMR studies have established that protein receptors can adopt a variety of conformations in solution. We reasoned that the protein-bound ligand might similarly adopt different conformations that are not visualized by crystallography. LpxC inhibitors such as CHIR-090 and LPC-011 (3) contain an amino acid derivative in their headgroup (Figure 5A), allowing us to conduct rotamer analysis of the headgroup based on the nomenclature for amino acids. When we crystallized the LpxC/LPC-011 (3) complex, we observed a single conformation of the threonyl headgroup (Figure 5B), yet solution NMR measurements based on scalar couplings showed that the threonyl headgroup readily samples different χ¹ rotamers (Figure 5C). Merging these ligand conformations mapped out an inhibitor envelope with three binding pockets around the Cβ position of the threonyl headgroup (Figure 5D), allowing the design of 3-hydroxy valine hydroxamate (LPC-037, 4) and 3-amino valine hydroxamate (LPC-040, 5) compounds that occupy all three binding pockets (Figure 5A,E). Taking a step further, we examined isoleucine-hydroxamate derivative LPC-023 (6, Figure 5A) to determine if we could expand the dynamically accessible LpxC inhibitor envelope at the Cγ position. Despite the observation of a single gauche+ χ² conformation of the isoleucine headgroup in the crystal structure (Figure 5F), solution NMR measurement revealed the presence of the trans χ² rotamer at ~25% population (Figure 5G), suggesting that there exist two methyl-sized pockets near the Cγ position. Because these additional pockets are near the conserved histidine and lysine residues on the basic patch, two fluorine atoms were chosen because fluorine is both strongly electro-negative and lipophilic, rendering it well-suited for forming hydrophobic interactions with the deprotonated histidine side chain and the stem of the lysine group or for forming electrostatic interactions with a protonated histidine imidazolium and a positively charged lysine terminal ammonium group (Figure 5H). Such a design yielded LPC-058 (7) (Figure 5A,I), which is a very slow, tight-binding inhibitor with a K_I* value of 3.5 pM. The improvement of the LpxC binding affinity by optimizing the tail and head groups is accompanied by a dramatic reduction of MICs of LPC-058 (7) over parent compound CHIR-090 for a wide range of Gram-negative pathogens, sometime enhancing the antibiotic activity by more than 100-fold.²

Figure 5.

Optimizing LpxC-targeting antibiotics by exploring the cryptic inhibitor envelope. (A) Chemical structures of LpxC inhibitors. MIC, minimum inhibitory concentration; Ec, E. coli W3110; Kp, K. pneumoniae 10031; Pa, P. aeruginosa PAO1; and Ct, Chlamydia trachomatis LGV-L2. (B) Crystal structure of LPC-011 (3) bound to LpxC showing a single ligand conformation. (C) Solution-NMR-based scalar coupling measurements reveal the presence of alternative rotamers of the threonyl group of LPC-011 (3) when bound to LpxC. (D) Merging multiple rotamers reveals three binding pockets near the Cβ position of the threonyl group. (E) Design and structural validation of Cβ-triple-substituted compound LPC-040 (5). (F) Crystal structure of LPC-023 (6) bound to LpxC showing a single ligand conformation. (G) Solution NMR analysis reveals the presence of two χ² rotamers of LPC-023 (6) when bound to LpxC. (H) Design of difluoro-methyl-substituted compound LPC-058 (7). (I) Structural validation of the binding model of LPC-058 (7). Adapted with permission from ref 2. Copyright (2016) Lee et al.

LPC-058 (7) displays a synergistic effect with conventional antibiotics such as β-lactams, amikacin, and ciprofloxacin against multi-drug-resistant Gram-negative pathogens.⁴⁸ The comprehensive evaluation of LPC-058 (7) against a large panel of clinical strains showed that LPC-058 (7) is highly effective against both susceptible and multi-drug-resistant Gram-negative bacteria from 369 clinical isolates with no major MIC difference between the two groups.⁴⁸ These observations highlight the impact of LpxC inhibitors as a novel class of antibiotics to overcome the threat of multi-drug-resistant Gram-negative infections.

An evaluation of LPC-058 (7) in a murine plague infection model showed that LPC-058 (7) was able to rescue mice with lethal Y. pestis infections at a dose of 40 mg/kg Q8H via IV delivery, though it appeared to elevate the blood alanine aminotransferase level.⁴⁹ This adverse effect is likely associated with the aniline group in LPC-058 (7) because related compound LPC-069 (8) was well tolerated at 200 mg/kg Q8H with no adverse clinical signs.⁴⁹ At this dosing level, a 5-day consecutive treatment was able to fully rescue mice with a lethal dose of Y. pestis infection.⁴⁹

FUTURE OF LPXC INHIBITORS

Potent LpxC inhibitors display outstanding antibiotic activity against P. aeruginosa, K. pneumoniae, Enterobacter spp. of the ESKAPE pathogens, and other Enterobacteriaceae and non-fermentative Gram-negative bacilli.^48,50 Despite the outstanding antibiotic activity of hydroxamate-containing LpxC inhibitors, there have always been concerns due to the poor performance of hydroxamate-containing matrix metalloproteinase (MMP) inhibitors.⁵¹ The failure of Achaogen’s ACHN-975, a hydroxamate-containing LpxC inhibitor, in phase I clinical trials due to adverse cardiovascular effects has cemented such fear and motivated the development of nonhydroxamate LpxC inhibitors, such as the ones developed by Forge Therapeutics^52,53 and Taisho Pharmaceutical Co.^54,55 Intriguingly, recent work from Achaogen showed that eliminating the hydroxamate group did not mitigate the cardiovascular side effect. Instead, modifications in other parts of the molecule improved the cardiovascular safety window.⁵⁶ Conversely, some nonhydroxamate-based LpxC inhibitors similarly displayed an adverse cardiovascular effect (e.g., compounds 17 and 18 in ref 55). Taken together, these observations raise an intriguing possibility that compounds with an enhanced potency of LpxC inhibition such as those offered by the difluoromethyl-allo-threonyl-hydroxamate compounds might offer a sufficient window of safety.

The debate about hydroxamate versus nonhydroxamate LpxC inhibitors will likely linger. Regardless of which form of LpxC inhibitors will ultimately advance to human clinical trials, we anticipate that the valuable lessons learned from the structural and dynamics analysis of LpxC enzymes and their bound inhibitors will ultimately contribute to the successful development of LpxC-targeting antibiotics.

OVERCOMING MULTIDRUG RESISTANCE OF GRAM-NEGATIVE PATHOGENS BY TARGETING LPXH

Besides the extensive development of LpxC inhibitors, small-molecule inhibitors have also been discovered for LpxA,^57,58 LpxD,^57,59 and LpxH.^3,60–62 Among these targets, LpxH is particularly interesting because it is the only late-stage lipid A enzyme. The division of lipid A enzymes into early stage and late stage arises from the surprising discovery that A. baumannii strains deficient in lpxA, lpxC, and lpxD are viable⁶³ while those lacking lpxH, lpxB, or lpxK are not.^64–66 As the substrates of these enzymes become membrane-embedded, deletion or inhibition of these enzymes results in the toxic accumulation of lipid A intermediates, the distortion of the bacterial inner membrane (Figure 6), and ultimately cell death even when lipid A biosynthesis is dispensable to the viability of the bacteria. In this regard, the discovery of a sulfonyl piperazine compound (denoted as AZ1 below) as the first LpxH inhibitor was significant.⁶⁰ AZ1 displayed modest antibiotic activity against WT K. pneumoniae and E. coli strains with a compromised outer membrane (yhjD*ΔkdtA)³ or with an efflux pump deletion (ΔtolC).⁶⁰

Figure 6.

LpxH depletion results in gross morphology changes in A. baumannii that can survive without lipid A biosynthesis. Transmission EM images of A. baumannii with IPTG-inducible lpxH expression in the presence and absence of IPTG are shown in (A) and (B), respectively. Distortion of the inner membrane is indicated by the arrow. Adapted with permission from ref 64. Copyright (2016) Richie et al.

ESTABLISHING THE PHARMACOPHORE OF SULFONYL PIPERAZINE LPXH INHIBITORS

AZ1 consists of three regions: a phenyl group, a sulfonyl piperazine linker, and an N-acetyl indoline group (Figure 7A). We analyzed a suite of AZ1 derivatives in order to establish the pharmacophore.⁶¹ First, to elucidate the role of the CF₃ (R¹) group in the phenyl ring, we substituted various functional groups (such as H, Br, alkyl, Ph, OH, CO₂Me, and CO₂H) in place of CF₃. Although no substitution showed better activity than AZ1, analogs with a hydrophobic substituent on the phenyl ring (e.g., Br and Ph) were active. Analogs with a polar functional group, such as OH and CO₂H, were inactive. These results indicated the importance of the hydrophobicity and size of the substituent on the phenyl ring in LpxH inhibitory activity. In the case of the sulfonyl piperazine linker, compounds with expanded ring systems, a flexible acyclic linker, or an extended piperazine linker showed reduced activity. These data suggested that both the rigidity of the piperazine linker and the orientation of the trifluoromethyl-substituted phenyl ring are crucial to AZ1 activity. To investigate the role of the indoline region, various long-side-chain linkers and sulfanilide rings were introduced. Interestingly, extended N-acyl groups were mostly comparable to AZ1 and demonstrated only slightly reduced activity. On the basis of SAR analysis, we generated a pharmacophore model represented by AHHRR (Figure 7B), indicating that it has one hydrogen-bond acceptor (A), two hydrophobic groups (H), and two aromatic rings (R). The two hydrophobic groups and one hydrogen-bond acceptor are mapped to the piperazine ring, the indoline ring, and the carbonyl group, respectively (Figure 7B).

Figure 7.

LpxH inhibition. (A) Structures of AZ1 and derivatives. (B) Pharmacophore generation results (green, hydrophobic groups; orange, aromatic rings; rose, hydrogen-bond acceptor). Adapted with permission from ref 61. Copyright (2019) American Chemical Society.

STRUCTURE OF LPXH IN COMPLEX WITH AZ1

The lpxH gene in E. coli was cloned in 2002.⁶⁷ Subsequent sequence analysis and biochemical studies established LpxH as a member of the calcineurin-like phosphoesterases (CLPs), which requires detergent and a dimanganese cluster for full activity.⁶⁸ These biochemical properties have been nicely corroborated by the structural analysis of H. influenzae LpxH by our laboratory⁶⁹ and P. aeruginosa LpxH reported by Yao’s laboratory,⁷⁰ E. coli LpxH by Bohl and colleagues in Aihara’s laboratory 2 years later,⁷¹ and most recently K. pneumoniae LpxH by our laboratory.³ Collectively, these structures have revealed a unique insertion lid above the conserved core architecture of calcineurin-like phosphoesterases (CLPs) (Figure 8A). The active site, situated between the lid domain and core CLP architecture, contains a dimanganese cluster that is chelated by residues of the signature metal chelating motifs of CLP enzymes. The glucosamine-1-phosphate headgroup shared by lipid X and LpxH substrate UDP-DAGn sits in the active site with its phosphate group located above the dimanganese cluster, whereas the 2-N-linked acyl chain is buried inside a hydrophobic chamber between the core domain and the insertion lid, and the 3-O-linked acyl chain rises through an open area above the active site (Figure 8A,C).

Figure 8.

Structure of LpxH complexes. (A) Structure of the K. pneumoniae LpxH/lipid X complex. (B) Structure of the K. pneumoniae LpxH/AZ1 complex. LpxH is shown in the cartoon model. The CLP fold and the lid domain are shown in green and coral, respectively. The dimanganese cluster in the active site is shown in the sphere model. Lipid X and AZ1 are shown in the stick model. The purple meshes represent omit (2mFo-DFc) maps of lipid X and AZ1 contoured at 1σ. (C) Overlay of AZ1 with the 2-N-acyl chain of lipid X in the K. pneumoniae LpxH complex structures. (D) Interactions of AZ1 with K. pneumoniae LpxH residues. Adapted with permission from ref 3. Copyright (2020) National Academy of Sciences.

The structural elucidation of LpxH enzymes has made it possible to model AZ1. Such effort generated divergent docking poses with opposite orientations,⁷¹ though the preferred docking orientation⁷¹ is unsupported by our pharmacophore model of AZ1.⁶¹ The structural elucidation of the K. pneumoniae LpxH/AZ1 complex by our laboratory has ultimately resolved this confusion.³ The structure, determined at 2.26 Å resolution, allowed a clear visualization of AZ1 and its orientation in the hydrophobic substrate chamber (Figure 8B): AZ1 binds with the N-acetyl indoline ring extending toward the active site and with the piperazine group reaching out to the far side of the chamber. Such a structural observation of the AZ1 orientation in K. pneumoniae is further supported by solution-based photo-cross-linking studies of E. coli LpxH using a diazirine derivate of AZ1.³ Interestingly, the hydrophobic chamber also has an “L” shape similar to that of the acyl chain binding passage of LpxC,¹ and the sharp kink of the sulfonyl group of AZ1 coincides with the L turn observed in the 2-N-acyl chain of LpxH product lipid X (Figure 8C).

A significant number of vdW interactions were observed between AZ1 and the surrounding hydrophobic chamber residues from K. pneumoniae LpxH (Figure 8D). A classic parallel cation–π stacking interaction was also observed between the AZ1 indoline ring and the guanidinium side chain of R80. In addition, hydrogen bonds were observed between the AZ1 N-acetyl group and the side chain of N79 of LpxH and between the AZ1 sulfonamide group and the side chain of R157 and the backbone amide group of W46 of LpxH. These structural observations correlated well with our previously reported pharmacophore model of AHHRR.⁶¹

LIGAND DYNAMICS-BASED DESIGN OF LPXH INHIBITORS

Early-generation inhibitors may not occupy all of the space within the inhibitor envelope. As discussed previously for LpxC inhibitors,² solution NMR offers an excellent tool to probe the cryptic inhibitor envelope sampled by multiple conformations of the bound compound in solution even when the crystal structure shows a rigid ligand conformation. This is also the case for the LpxH inhibitor, AZ1. In our crystal structure of the LpxH/AZ1 complex determined at 2.26 Å resolution, only one conformation of AZ1 was observed, with the trifluoromethyl group pointing upward to the exit of the substrate chamber. Intriguingly, two ¹⁹F signals were observed for the LpxH-bound AZ1 whereas only one ¹⁹F signal was detected for free AZ1, and these signals do not overlap (Figure 9A). Fitting of the two signals suggested that the LpxH-bound AZ1 adopted two conformations with a major state population of ~85% and a minor state population of ~15% (Figure 9A). We reasoned that these two populations could indicate a flipping of the trifluoromethyl-substituted phenyl ring, revealing the existence of another binding pocket near the trifluoromethyl-substituted phenyl ring (Figure 9A). Furthermore, the smaller population indicated that such a state is unfavorable, presumably because the trifluoromethyl group is too big for this position. On the basis of this analysis, we designed JH-LPH-33 (9) containing a chlorine atom at the meta position of the trifluoromethyl group (Figure 9B). The crystal structure of JH-LPH-33 (9) in complex with K. pneumoniae LpxH solved at 2.25 Å confirmed our prediction, revealing a conformation with the trifluoromethyl pointing up toward the exit of the hydrophobic chamber and the chlorine atom buried within the hydrophobic chamber (Figure 9B). The overlay of JH-LPH-33 (9) with LpxH product lipid X showed that the chlorine atom partially occupied the binding pocket of the terminal methyl group of the 2-N-acyl chain (Figure 9B). Remarkably, our further investigation of the possible substitutions of the chlorine atom showed that JH-LPH-33 (9) hit the sweet spot: replacing the chlorine atom with either smaller functional groups (H, F, CH₃) or larger functional groups (Br or I) reduced the compound activity (Figure 9B).⁶²

Figure 9.

Design of sulfonyl piperazine LpxH inhibitors based on ligand dynamics analysis of the cryptic inhibitor envelope. (A) Solution ¹⁹F NMR reveals the presence of two conformations of AZ1 bound to LpxH caused by ring flipping of the trifluoromethyl-substituted phenyl ring. (B) Design and structural validation of JH-LPH-33 (9). (C) Antibiotic activities of AZ1 and JH-LPH-33 (9) against K. pneumoniae. Adapted with permission from ref 3. Copyright (2020) National Academy of Sciences.

Our enzymatic characterization established that JH-LPH-33 (9) is a competitive inhibitor with a K_I value of ~10 nM against K. pneumoniae LpxH.³ JH-LPH-33 (9) showed an ~14-fold enhancement in potency over AZ1, which has a K_I value of ~145 nM. It also improved the antibiotic activity of AZ1 by more than 40-fold against K. pneumoniae, reducing the MIC value from >64 to 1.6 μg/mL (Figure 9C). The main challenge for the sulfonyl piperazine LpxH inhibitors appears to be limited bacterial outer membrane permeability. These compounds currently lack activity against WT E. coli strains. However, the inclusion of 10 μg/mL PMBN, which increases the outer membrane permeability but does not have antibiotic activity by itself, yielded MIC values in the sub-μg/mL range for JH-LPH-33 (9) against E. coli.³

FUTURE OF LPXH INHIBITORS

A surprise from the structures of LpxH-inhibitor complexes is that neither AZ1 nor JH-LPH-33 (9) occupies the active site.³ This sets the stage for a fragment-based hit-to-lead optimization campaign for sulfonyl piperazine compounds. As a first step in this direction, we recently synthesized derivatives of JH-LPH-33 (9) that have an N-acyl chain extension containing a hydroxamate group.⁶² Although our structural analysis of one of these compounds, JH-LPH-41, showed that the extended N-acyl chain indeed reached into the active site, it unexpectedly found its way to interact with the protein backbone instead of the dimanganese cluster.⁶² Nevertheless, with further optimization of the N-acyl chain we anticipated the discovery of potent LpxH inhibitors that chelate the dimanganese metal cluster. Considering the conservation of the lpxH gene in K. pneumoniae, A. baumannii, P. aeruginosa, and Enterobacter spp., a broadly effective LpxH inhibitor may provide the much needed therapeutic solution for treating multi-drug-resistant Gram-negative infections.

CONCLUSIONS

The rapid progress in structural biology has now elucidated all six essential lipid A enzymes, rendering the inhibition of the lipid A pathway an attractive strategy for developing novel antibiotics for treating multi-drug-resistant Gram-negative infections. Should these inhibitors display desirable clinical properties, effectiveness, and safety, they would become a new class of antibiotics with a novel mode of action against Gram-negative pathogens over half a century.

Biographies

Pei Zhou received his B.S. degree from the University of Science and Technology of China (USTC). Following his Ph.D. study with Gregory L. Verdine in the Department of Chemistry and Chemical Biology at Harvard University, he pursued postdoctoral research in structural biology with Gerhard Wagner at Harvard Medical School. In 2001, he established his own research laboratory at Duke University School of Medicine and is currently a professor of biochemistry and chemistry. His main research interest centers on the role of structure and dynamics in macromolecular complex assembly and function, enzyme inhibition, and structure- and dynamics-based drug development.

Jiyong Hong received B.S. and M.S. degrees from Seoul National University (Korea). Following his Ph.D. study under the guidance of Dale Boger at The Scripps Research Institute (TSRI), he was a postdoctoral research associate in the laboratory of Peter Schultz at TSRI. In 2005, he began his independent career at Duke University and is currently a professor of chemistry. His research interest includes the synthesis and study of mechanisms of action of biologically important natural products, the development of synthetic methods for the rapid construction of molecular complexity and structural diversity, and the identification of small-molecule modulators for biological processes.